LAZAR DRAGON MATHEMATICAL METHODS IN AERODYNAMICS ll KLUWER ACADEMIC PUBLISHERS EDITURA ACADEMIEI ROMANE MATHEMATICAL METHODS IN AERODYNAMICS Mathematical Methods in Aerodynamics by LAZAR DRAGO$, Roqsonime Academy. H charm Romania L7 KLUWER ACADEMIC PUBLISHERS EDITURA ACADEMIEI ROMANE DORDRECHT/BOSTON FLONDON BUCURE$11 A C.I.P. Catalogue record for this book is available from the Library al' Congress. ISBN 1-4020-1663-8 ISBN 973-27-0986-3 Published by Kluwer Academic Publishers and Editura Academrei Romans. Kluwer Academic Publishers. P.O. Box 17. 3300 AA Dordrecht. The Netherlands. Editura Academici Romans. P.O. Box 5-42.050711 Ducure ti. Romania. Sold and distributed in North. Central and South America by Kluwer Academic Publishers, 101 Philip Drive. Nor well. MA 02061. U.S.A. In all other countries. except for Romania and the Republic of Moldavia, sold and distributed by Kluwer Academic Publishers, P.O. Box 322. 3300 AH Dordrecht. The Netherlands. In Romania and Republic of Moldavia sold and distributed by Editura Academici Rom5nc. P.O. Box 5-42.050711 tiucuresti. Rominia. 1'rintrrl an "id -free paper All Rights Reserved 0 2003 Kluwer Academic Publishers and Editura Academiei Rorn inc No part of this work may be reproduced. stored in a retrieval system. or transmitted in any form or by any means, electronic, mechanical, photocopying. microfilming. recording or otherwise, without written permission from the Publisher, with the exception of any material supplied specifically for the purpose of being entered and executed an a computer system. for exclusive use by the purchaser of the work. Printed in Romania Table of Contents xiii Preface I The Equations of Ideal Fluids .. . ...................... .... ................. . The Equations of Motion 1.1.1 Elements of Kinematics 1.1.2 The Equations of Motion ... 1.2 The 1.2.1 Helmholtz's equation. Bernoulli's integral 1.1 . . ... .... ......... Potential Flow ............................. 1.2.2 1.2.3 The Linear Theory ......................... I 1 3 5 5 6 8 ... .......... .. ..... The Shock Waves Theory ............ ............. 1.3.1 The Jump Equations ........................ 11 11 ilugoniot's Equation 13 15 16 17 19 1.2.4 1.3 ............ The Equation of the Potential ................... 1 1.3.2 1.3.3 1.3.4 1.3.5 1.3.6 The Acceleration Potential . .. . . . . . . .. . . . . . . . . . . . . . The Solution of the .lump Equations .............. . Prandtl's Formula ....................... .. The Shock Polar .......................... The Compression Shock past a Concave Bend .......... 9 2 The Equations of Linear Aerodynamics and its Fundamental Solutions 21 2.1 2.2 21 21 2.1.2 2.1.3 2.1.4 2.1.5 2.1.6 22 24 26 28 29 30 The Steady Solutions ....................... 30 .... ... ................. 34 ............ ... The Fundamental Solutions for the Fluid at Rest ........ On the Interpretation of the Fundamental Solution ....... The Fundamental Solutions of the Steady System ............ 2.2.2 2.2.3 2.2.4 2.2.5 2.2.6 2.2.7 2.4 . ......... The Equation of the Potential ..... ... ........... .. ............ . The Linear System ...... The Uniform Motion in the Fluid at Rest ..... ... ... . The Equations of Motion ........ ... The Equations of Linear Aerodynamics ............. The Fundamental Solutions of the Equation of the Potential ...... 2.2.1 2.3 ................. . ......... The Equations of Linear Aerodynamics 2.1.1 The Fundamental Problem of Aerodynamics Oscillatory Solutions ...... ........... Oscillatory Solutions for Al = 1 The Unsteady Solutions . . . . . The Unsteady Solutions for At = I . . . . . . . . . 32 . . . . . . . . 41 42 43 2.3.1 The Significance of the Fundamental Solution 2.3.2 2.3.3 2.3.4 2.3.5 2.3.6 The General Form of the Fundeunental Solution ......... 2.4.2 2.4.3 . . . . . . . . . . . ........... ........... 44 44 ..... .... 45 46 47 48 48 50 The Determination of the Velocity Field ............. 51 Other Fortes of the Components V and W 53 The Subsonic Plane Solution .... .... The Three-Dimensional Subsonic Solution The TwoDimensional Supersonic Solution . . . . . . . . . . . The Three-Dimensional Supersonic Solution ........... The Fundamental Solutions of the Oscillatory System 2.4.1 36 The Determination of Pressure ... ... ............ 50 .. . ... .... vi 2.4.4 2.4.5 2.5 The Fundamental Solutions in the Case Af = I ......... Fundamental Solutions of the Unsteady System I .......... .. 55 57 57 58 .. Cauchy's Problem ........ ............ . .. .. The Perturbation Produced by a Mobile Source ... . ..... Fundamental Solutions of the Unsteady System If .. ... ..... 2.5.1 2.5.2 2.3.3 2.3.4 2.6 The Incompressible Fluid ..................... 55 Fundamental Solutions . Fundamental Matrices . . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . 61 62 64 64 66 . 2.6.1 2.6.2 The Fundamental Matrices .. ............. ..... The Method of the Minimal Polynomial .... .... ... .. 3 The Infinite Span Airfoil In Subsonic Flow 3.1 3.1.2 3.1.3 3.1.4 3.1.5 3.1.6 3.1.7 3.1.8 3.1.9 3.2 3.3 69 . ...... . .... ... ... .. ... .... .... .. A Classical Method ........................ The Fundamental Solutions Method . ....... ... .. .. The Airfoil in the Unlimited Fluid 3.1.1 The Statement of the Problem . . 69 69 70 . . The Function f (z). The Complex Velocity in the Fluid The Calculation of the Aerodynamic Action . . . . . . . . . . 72 75 76 . . . .. .................... . 77 Examples ........ . The General Case ......................... 80 Numerical Integrations ... ...... .............. 81 . The Integration of the Thin Airfoil Equation with the Aid of Gauss-type Quadrature Formulas . ... . ....... ..... 81 The Airfoil in Ground Effects ....................... 82 3.2.1 The Integral Equation .. ..... ....... .... ..... 82 3.2.2 A Numerical Method .................. . ... .. 85 3.2.3 The Flat Plate ........................... 85 3.2.4 The Symmetric Airfoil .. ..................... 86 The Airfoil in Tunnel Effects .. ... . .... ... . ......... 88 . . .. . 88 3.3.1 The Integral Equation . .. 90 3.3.2 The Integration of the Equation (3.3.9) . . . . . . . . . . . . . . . . . 91 . . .. .... ... . ...... The Numerical Integration .............. ...... 92 92 95 The Integral Equation ................. ...... 97 . . . . . . . . . . . . . . . Numerical Results . . . . . . . . . . 3.4 Airfoils Parallel to the Undisturbed Stream 3.4.1 The Integral Equations . . 3.3.3 .. .. . . . . . . ... . . . . . . . . . . . . . . . .............................. 97 3.4.2 3.5 Grids of Profiles 3.5.1 3.5.2 The Numerical Integration 3.6 Airfoils in Tandem . . . . . . . ... .. .................. 100 . . .... . ................ . 3.6.1 3.6.2 3.6.3 3.6.4 . . . . . . . . The Integral Equations The Determination of the Functions f, and ff . . . . . . . . 101 . 101 .. .... .. 103 . ....... ..... 104 ................. ......... The Lift and Moment Coefficients Numerical Values . . . 105 4 The Application of the Boundary Element Method to the Theory of the Infinite Span Airfoil In Subsonic Flow 4.1 109 ... . .... ... .............. 109 .. ... ... . .......... ....... 109 .. . ............ .. 110 The Equations of Motion 4.1.1 Introduction 4.1.2 The Statement of the Problem . 4.1.3 . . The Fundamental Solutions .. .. . ............... 112 vii 4.2 Indirect Methods for the t idimited Fluid Case ............. 113 4.2.1 The integral equation for the Distribution of Sources ...... 113 4.2.2 The Integral Equation for the Distribution of Vortices 4.2.3 4.2.4 4.2.5 4.2.6 4.3 4.4 The Determination of the Unknowns .............. . The Circular Obstacle . . . . . . . . . . . . . . . . . . . . . . . 115 115 117 120 The Elliptical Obstacle ...................... 121 4.3.1 4.3.2 4.3.3 4.3.4 4.3.5 The representation of the solution ............... . 122 . .................... .... 125 4.3.6 4.3.7 4.3.8 Appendix ...... ................... ... .. 129 The Integral Equation ............. ... .. .... 123 . The Circulation .. The Discretization of the Equations ............... 126 The Lifting Profile ......................... 126 The Local Pressure Coefficient .................. 128 Numerical Determinations ................. .. . .................... .. The Representation of the Solution . ............. .. 131 131 131 . . The Airfoil in Ground Effects 4.4.2 4.4.3 4.4.4 1.4.5 The Integral Equation .................. .. .. 134 The Computer Implementation ... .......... ..... 135 . 136 .. The Treatment of the Method . .. . . . . . . . . . . . . . . The Circular Obstacle in a Compressible Fluid ......... 137 .. .... . ... ................ . .. 138 .1.4.6 Appendix 4.5.1 The Representation of the Solution ............. .. . The Airfoil in Tunnel Effects .... ................ ... 140 4.5.2 4.5.3 4.5.4 4.5.5 4.6 .. ....................... .... . The Direct Method for the Unlimited Fluid Case ............ 122 4.4.1 4.5 . The Boundary Elements Method ................ . The Integral Equation ................ ....... 144 The Verification of the Method ...... ... ......... 146 Appendix ...... ................... ..... 149 Other Methods. The Intrinsic Integral Equation .......... . 4.6.1 140 . Green Functions ...... ......... ........... 142 150 . The Method of Regularization .................. 150 5 The Theory of Finite Span Airfoil in Subsonic Flow. The Lifting Surface Theory 5.1 The Lifting Surface Equation 5.1.1 5.1.2 5.1.3 5.1.4 155 . ... ... ............. .. 155 . The Statement of the Problem .................. 155 Bibliographical Comments ..................... 158 The General Solution ................ ... .. .. 159 The Boundary Values of the Pressure ... ........... 161 .. 163 .The U.S Boundary Values of the Component w . . . . . . ... .. .. . . . . . . . x.1.6 The Integral Equation =..1.7 .5.1.8 The Plane Problem ....... ..... .. ... ... ... . . . . . . . . . . . . 164 Other Forms of the Integral Equation .............. 166 5.1.9 The Aerodynamic Action in the First Approximation . . . . . . . 168 169 5.1.10 A More Accurate Calculation ................... 171 5.2 5.1.11 Another Deduction of the Representation of the General Solution 173 Methods for the Numerical Integration of the Lifting Surface Equation 175 5.2.1 5.2.2 The General Theory ........................ 175 Multhopp's Method ............. ........... 178 Viii 5.3 5.4 .. 179 180 The Third Method ......................... 181 5.2.3 The Quadratum Formulas Method 5.2.4 5.2.5 The Aerodynamic Action ... .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .............. 184 . ...................... 184 Ground Effects in the Lifting Surface Theory 5.3.1 The General Solution 5.3.2 The Integral Equation 5.3.3 The Two-Dimensional Problem .... .... ....... ... 188 . . The Wing of Low Aspect Ratio . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .. . .. . . . . . 189 . . 193 . 5.4.2 The Integral Equation ......... ........ ... .. 189 The Case h = h(x) .......... . .. ... .... ... 192 5.4.3 The General Case 5.4.1 . . . . . . . . . . .. . . . . . . . . . . . . . 6 The Lifting Line Theory 6.1 186 19T Prandtl's Theory .............................. 197 6.1.1 The Lifting Line Hypotheses. The Velocity Field . ...... 197 6.1.2 Prandtl's Equation . .. ..... .............. 200 6.1.3 The Aerodynamic Action ...... ........ . ...... 202 . . . . ...................... 203 6.1.4 The Elliptical Flat Plate 6.2 The Theory of Integration of Prandtl's Equation. The Reduction to Fredholm-Type Integral Equations ....... ..... ........ 205 6.2.2 The Equation of Trefftz and Schmidt ............... 205 Existence and Uniqueness Theorems ............... 209 6.2.3 Foundation of Glauert's Method 6.2.4 6.2.5 The Minimal Drag Airfoil 6.2.1 . . . . . . . . . . . . . . . . . 210 Glauert's Approximation ..................... 212 ..... .. .............. 212 6.3 The Symmetrical Wing. Vekuas Equation. A Larger Class of Exact Solutions ... ....... .... . ... .. . ............. 214 The Integral Equation ....... .. . ... .......... 215 6.3.1 Symmetry Properties ........................ 214 6.3.2 6.3.3 Vekua's Equation ......................... . 217 The Elliptical Wing ........................ 220 6.3.5 The Rectangular Wing . . . 221 6.3.6 Extensions .................. . ...... ... 222 Numerical Methods . .. . 223 6.4.1 Multhopp's Method ........................ 223 6.4.2 The Quadrature Formulas Method .. ... .. .. ...... 228 6.3.4 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6.4 . . . . . . . . . . . . . . . .. ... ................. 231 Various Extensions of the Lifting Line Theory .............. 234 6.5.1 The Equation of Weissinger and Reissner .. .......... 234 6.5.2 Weissinger's Equation. The Rectangular Wing ......... 236 The Lifting Line Theory in Ground Effects .. ... .. 2.18 . 6.5 6.6 6.4.3 The Collocation Method 6.6.1 The Integral Equation .......... ...... ....... 238 The Elliptical Flat Plate ...................... 240 6.6.2 6.6.3 6.7 Numerical Solutions in the General Case ............. 241 The Curved Lifting Line ................ 6.7.1 6.7.2 6.7.3 The Pressure and Velocity Fields The Integral Equation . . . . .... ...... 242 ................. 242 . .. .. . .. 246 The Numerical Method .... . ................. 247 . . . . . . . . . . . ix 7 The Application of the Boundary Integral Equations Method to the 251 Theory of the Three-Dimensional Airfoil In Subsonic Flow 7.1 The First Indirect Method (Sources Distributions) ........... 251 7.1.1 7.1.2 7.1.3 7.1.4 7.1.5 7.1.6 7.1.7 7.1.8 7.2 The Integral Equation The Integral Equation . ................. . ... . . .... .............. 253 253 . . . . The Discretixation of the Integral Equation ........... 255 The Singular Integrals . ... ................. .. 258 The Velocity Field. The Validation of the Method ....... 258 The Incompressible Fluid. An Exact Solution .......... 259 .. . . . . . . . . . 263 The Expression of the Potential . . The Second Indirect Method (Doublet Distributions). The Incompress. . . . . ible Fluid .................................. 265 7.2.1 7.2.2 7.2.3 7.2.4 7.3 The General Equations ... . .................. 251 The Integral Equation ..... ... ............. .. 265 The Flow past the Sphere. The Exact Solution ......... 267 The Velocity Field ......................... 268 The Velocity Field on the Body. N. Marcov's Formula ..... 268 The Direct Method. The Incompressible Fluid ............. 271 7.3.1 The Integral Representation Formula ........ .. ..... 271 7.3.2 7.3.3 7.3.4 7.3.5 ..... . ................ .. ..................... ... 275 The Integral Equation Kutta's Condition 274 The Lifting Flow .......................... 276 The Discretization of the Integral Equation ........... 279 8 The Supersonic Steady Flow 8.1 8.1.1 8.1.2 8.1.3 8.1.4 8.1.5 8.1.6 8.2 283 The Thin Airfoil of Infinite Span .. ................ ... 283 The Analytical Solution ...................... 283 The Fundamental Solutions Method ....... .... .... 286 The Aerodynamic. Action 287 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . The Graphical Method ....................... 289 The Theory of Polygonal Profiles ................. 290 .. 294 Validity Conditions . Ground and Tunnel Effects .. .. .................. .. 295 8.2.1 The General Solution ............... .... .... 295 298 8.2.2 The Aerodynamic Coefficients ................. . The Three-Dimensional Wing .............. ... ... .. 300 . . . . . 8.3 . 8.3.1 Subsonic and Supersonic Edges ..... ............ . 8.3.2 8.3.3 8.3.4 8.3.5 8.3.6 8.3.7 8.3.8 8.3.9 The Representation of the General Solution ........... 302 The Influence Zones. The Domain Di ....... .. ..... 304 The Boundary Values of the Pressure ........... . .. 305 The First Form of the Integral Equation . . . . . . . . . . . . . 300 306 The Equation D in Coordinates on Characteristics .... .. 308 The Plane Problem ......................... 310 The Equation of Heaslet and Lomax (the 11L Equation) . . . 311 The Deduction of HL Equation from D Equation ...... 313 8.3.10 The Equation of Homentcovschi (II Equation) . .... .... 318 8.4 ... .... .... 320 Abel's Equation .......................... 320 The Theory of Integration of the H Equation .. 8.4.1 8.4.2 . The Solution of the H Equation in the Domain of Influence of the Supersonic Trailing Edge ...... ........ ... . 321 X 8.4.3 The Solution in the Domains of Influence of the Subsonic Lead- ing Edge .......................... .. ... 323 The Wing with Dependent Subsonic Leading Edges and Independent Subsonic Trailing Edges . . . . . . . . . 324 8.4.5 The Wing with Dependent Subsonic Trailing Edges . .. ... 326 8.4.6 The Solution in the Zone of Influence of the Subsonic Edges under the Hypothesis that the Subsonic Leading Edges are letdependent . . . . . . . . . . . . . . . . . . . . . . . . . . 327 8.4.7 The Wing with Dependent Subsonic Trailing Edges . . . . . . 337 8.4.4 . . 8.5 . . . . . The Theory of Conical Nloticaas ...... ...... .......... Introduction ............... .. ... .. ... . 8.5.1 339 8.5.2 8.5.3 340 . 8.5.4 8.6 . The Wing with Supersonic Leading Edges ..... ....... The Wing With a Supersonic Leading FAlge and with Another Subsonic Leading or Trailing Edge .. .... . ...... ... The Wing with Subsonic Leading Edges .... ......... Flat Wings ... . ... .. .. .... .... ... ...... .. ... The Trapezoidal Wing with Subsonic Lateral Edges ...... 352 The Trapezoidal Wing with Lateral Supersonic Edges ..... 355 The Triangular Wing. The Calculation of the Aerodynamic Action . . . . . . . . . . . . . . . . . . . . . . . . . . . 349 . . . . The Equations of the Transonic Flow .......... .... ... .. 359 :3.59 .. . .... .. .. 359 9.1.1 The Presence of the Transonic Flow .... 9.1.2 The Equation of the Potential .. ... .... .... . ..... 361 The System of 'transonic Flow . ... . ....... . ... .. 364 The Shock Equations . . . .. . . .. ......... ... .. 368 The Plane Flow ..... ............... . .......... 369 . . . . . . . . . . . . . . . . . . .. .369 . . . . . . . . . . . . . . . . . . . . . . . . .. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9.2.2 The Fundamental Solution The General Solution . . 9.2.3 9.2.4 9.2.5 The Lift Coefficient The Symmetric Wing The Solution in Real . 9.2.6 The Symmetric Wing 9.:3.1 The Fundamental Solution .......... . .. .... ... .M1 The Study of the Singular Integrals ........ ....... 386 9.2.1 . . . . . . .. 371 376 376 377 . .. .. ......... ......... .. .. 380 The Thre"Dimensional Flow ... ..... . ......... ..... 383 9.3.2 9.3.3 9.3.4 9.4 347 8.6.3 8.6.4 9.1.3 9.1.4 9.3 :343 The Angular Wing with Supersonic Leading Edges .... ... 347 9 The Steady Transonic Flow 9.2 342 8.6.1 8.6.2 . 9.1 :339 . . . ... ... ... . ....... ..... 387 .......... :389 The Lifting Line Theory .. ... ........ .. .......... 392 The General Solution . Flows with Shock Waves .. ...... .. .. . 9.4.1 9.4.2 The Velocity Field ......................... :392 The Integral Equations ......... . .. .. 394 . . . . . . . . 10 The Unsteady Flow 397 ... ... . .... . .. 397 10.1.1 The Statement of the Problem ..... . .. .... .... .. 397 10.1.2 The Fundamental Solution ..... . ............ .. 397 10.1 The Oscillatory Profile in a Subsonic Stream 10.1.3 The Integral Equation . . . . . . . . . . . . . . . . . . . . . . . .ie.19 10.1.4 Considerations on the Kernel ..... ....... .. ..... 402 xi ...... 404 10.2 The Oscillatory Surface in a Subsonic Strewn ..... 10.2.1 The General Solution ............ ......... . 404 10.2.2 The Integral Equation ....................... 405 . 10.2.3 Other Expressions of the Kernel Function ......... . . 10.2.4 The Structure of the Kernel . . 10.2.5 The Sonic Flow ........ . . . . . . . . . . 409 412 . .................. 413 . . . . . . . . . 10.2.6 The Plane Flow ........ . .................. 414 10.3 The Theory of the Oscillatory Profile in a Supersonic Stream ..... 415 10.3.1 The General Solution 10.3.2 . . . . . . . . ..... . . . . . . . . . . The Integral Equation and its Solution ....... ....... 10.3,3 Formulas for the Lift and Moment Coefficients 10.3.4 The Flat Plate .............. . . . . . . .. . .......... . . 415 418 421 423 424 .. .......... 10.4 The Theory of the Oscillatory Wing in a Supersonic Stream ...... 426 . 10.3.5 The Oscillatory Profile in the Sonic Flow 10.4.1 The General Solution ............ .... ... . ... 426 10.4.2 The Boundary Values of the Pressure .............. 428 10.4.3 The Boundary Values of the Velocity, The Integral Equation . 430 10.4.4 Other Expressions of the Kernel ................. 433 . ....... ............ . ....... 435 10.4.5 A New Form 10.4.6 The Plane Problem . . . . . . 10.5 The Oscillatory Profile in a Sonic Stream .. .. .... . . . . . . . . . 436 .. .......... .... 438 . . 10.5.1 The General Solution. The Integral Equation .. ........ 438 10.5.2 Some Formulas for the Lift and Moment Coefficients . . . . . . 441 10.6 The Three-Dimensional Sonic Flow .................... 442 .. .. .. 442 .. .................. 443 10.6.3 The Plane Problem ......................... 446 10.6.4 Other Forms of the Kernel . .. . ... . 447 10.6.1 The General Solution . . . . . . . . . . . . . . . 11 The Theory of Slender Bodies 11.1 The Linear Equations and Their Fundamental Solutions ..... .. . . . . . . . . . . 10.6.2 The Integral Equation ... . . . . . . 11.1.1 The Boundary Condition. The Linear Equations ... 11.1.2 Fundamental Solutions . . 449 449 . .... 449 ... .... ........ . .... . 11.2 The Slender Body in a Subsonic Stream ... ............. . 452 454 . ... . ...... ........ 454 11.2.1 The Solution of the Problem 11.2.2 The Calculus of Lift and Moment Coefficients ..... ..... 456 11.3 The Thin Body in a Supersonic Stream ................. 458 11.3.1 The General Solution ..... .................. 458 11.3.2 The Pressure on the Body. The Lift and Moment . ............... 461 . ... .. ........ . ........... .. 463 Coefficients ............. 11.3.3 The wing at zero angle of attack .... ............. 463 11.3.4 Applications A Fourier Transform and Notions of the Theory of Distributions 465 .. ............... .. 465 A,1 The Fourier Transform of Functions A.2 The Spaces V and S .. ................... .... .. 466 A.3 Distributions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 467 A.4 The Convolution. Fundamental Solutions ................ 470 A.5 The Fourier Transform of the Functions from S ............ 472 A.6 The Fourier Transform of the Temperate Distributions ......... 473 xii ............ 475 .............. 478 A.7 The Calculus of Some Inverse Fourier Transforms A.8 The Fourier Transform in Bounded Domains B Cauchy-type Integrals. Dirichlet's Problem for the Half-Plane. The Calculus of Some Integrals 481 Cauchy-type Integrals ........................... 481 .................. 482 B.3 Plemelj's Formulas ............................. 483 B.l B.2 The Principal Value in Cauchy's Sense ................ 483 ........ 485 B.4 The Dirichlet's Problem for the Half-Plane B.5 The Calculus of Certain Integrals in the Complex Plane B.6 Glauert's Integral. Its Generalization and Some Applications ................................. 489 B.7 Other Integrals ............................... 491 C Singular Integral Equations C. i 493 The Thin Profile Equation ......................... 493 C.2 The Generalized Equation of Thin Profiles ................ 496 C.3 The Third Equation ............................ 498 C.4 The Forth Equation ............................ 502 C.5 The Fifth Equation ............................. 504 D The Finite Part 509 D.1 Introductory Notions ............................ 509 ............................. 509 ................ 510 D.2 The First Integral D.3 Integrals with Singularities in an interval D.4 Hadamard-Type Integrals ......................... 513 D.5 Generalization ............................... 515 E Singular Multiple Integrals 517 F Gauss-Type Quadrature Formulas 521 ............................. 521 F.2 Formulas of Interest in Aerodynamics .................. 525 F.3 The Modified Monegato's Formula .................... 529 FA A Useful Formula .............................. 530 F.1 General Theorems Bibliography 533 Index 571 Preface The researchers in Aerodynamics know that there is not a unitary method of investigation in this field. The first mathematical model of the airplane wing, the model meaning the integral equation governing the phenomenon, was proposed by L. Prandtl in 1918. The integral equation deduced by Prandtl, on the basis of some assumptions which will be specified in the sequel, furnishes the circulation C(y) (see Chapter 6). Using the circulation, one calculates the lift and moment coefficients, which are very important in Aerodynamics. The first hypothesis made by Prandtl consists in replacing the wing by a distribution of vortices on the plan-form D of the wing (i.e. the projection of the wing on the plane determined by the direction of the uniform stream at infinity and the direction of the span of the wing). Since such a distribution leads to a potential flow in the exterior of D and the experiences show that downstream the flow has not this character, Prandtl introduces as a supplementary hypothesis another vortices distribution on the trace of the domain D in the uniform stream. The first kind of vortices are called tied vortices and the second kind of vortices are called free vortices. On the basis of this model one developed later the main theories of Aerodynamics namely the lifting surface theory (after 1936, more precisely in 1950, when Multhopp gave the equation of this theory), the lifting surface theory for the supersonic flow (after 1946) and the lifting theory for oscillatory wings and surfaces for the subsonic, sonic and supersonic flow (after 1950). In the framework of the last theory the wing is replaced by doublets distributions. From a physical point of view, there is no reason for replacing the wing with vortices or doublets distributions. It is true that the vortices are detaching from the wing, but these are effects , not causes of the presence of the wing. The fact that these replacements lead to correct results shows how subtle was Prandtl's in- tuition. We specify that the distributions on D and its trace do not result from the equations of motion (they have been introduced outside the mathematical model). Taking into account this inconvenient, we have shown in (5.7) how it can be removed. We have to consider that the wing and the fluid constitute an interacting material system. If we want to study the fluid flow, then according to Cauchy's stress principle xiv (the principle of the internal forces; see for example [1.11), p.35), we have to assume that there exists it forces distribution on the boundary, which has against the fluid the same action like the wing itself. We shall replace therefore the wing with it forces distribution instead of a vortices, sources or doublets distribution and we shall find out the density of this distribution such that it should have the same action against the fluid like the wing itself. We shall proceed by imposing to the fluid flow determined by the forces distribution to satisfy the slipping condition on the wing, condition which is also satisfied by the flow determined by the wing. In this way it follows an integral equation for determining the forces density. This equation constitutes the mathematical model for the wing we have in view. This method is an unitary one and it is based only on the classical principles of mechanics (in fact. Cauchy's stress principle). It may be applied to all configurations: see [5.7) for the wing in a subsonic stream, [8.4] for the wing in a supersonic stream, [10.15], [10.16], [10.17) for the oscillatory wings in subsonic, sonic or supersonic stream etc. All these results are given in this book (see chapters 5, 8, 10, 11). We called this method (in[5.7)): the fundamental solutions method. It may be utilized to all cases in which one can calculate the fundamental solutions of the equations of motion. We have to notice that in the framework of this method, the existence of the vortices downstream the wing follows from the model (i.e. from the equations of motion) and it must not be introduced artificially. In the sequel we shall present some of the models of aerodynamics. For two-dimensional configurations, in a subsonic stream, the models are one-dimensional singular integral equations considered in the sense of Cauchy's principal value. One may integrate analytically only the equation of thin profiles in a free stream. For other geometries one determines numerical solutions with the aid of Gauss-type quadrature formulas (see Chapter 3). For three-dimensional wings in a subsonic stream, the models are two-dimensional integral equations with strong singularities, which are defined in the sense of Finite Part (see Chapter 5). For other geometry (for example the wing in ground effects) the models are generalized equations. All these models are solved only numerically. For the wing in it free stream, Multhopp's method is available. In this book we introdace a more general method - the quadrature formulas method. In the last part of Chapter 5 one presents the theory of low aspect wings which was extended by the author to the general case of asymmetrical wings. The lifting line theory may be deduced from the lifting surface theory with the aid of Prandtl's assumptions (6). This theory is developed xv by presenting analytical and numerical methods for solving Prandtl's equation; one considers also extensions of this theory, all the methods representing one-dimensional integral-differential equations. The author shows how these equations may be reduced to integral equations with strong singularities and for this type of singularities he gives a Gausstype quadrature formula, which allows the equation to be reduced to a linear algebraic system which is solved numerically. This method, which is very general. allows to obtain numerical solutions both in the case of the lifting line (Chapter 6) and the case of the lifting surface (Chapter 5). In the case of supersonic flow, the integral equations are solved analytically. For the three-dimensional wing (the lifting surface) we present in Chapter 8 a nice solution given by D. Homentcovschi in 18.16]. The integral equations describing the flow past oscillatory wings and profiles (chapter 10) have the same nature like the equations utilized in the case of steady flow but the kernels are more complicated. However for the sonic and supersonic flows these equations may be solved exactly by means of the Laplace transform, as it is shown in [10.17]. Chapter 9, devoted to the transonic motions. begins with a new asymptotic deduction of the equations of motion. The two and three-dimensional integral equations are obtained following the papers of the author and D. Homentcovschi. The theory of subsonic and supersonic flow past slender bodies (in Chapter 11) relies also on the fundamental solutions theory. In Chapter 2 one deduces the equations of the linear aerodynamics, on the basis of an asymptotic analysis. assuming that the small parameter depends on the thickness of the profile. In the classical aerodynamics this deduction is performed under the assumption that the unknowns and their derivatives have the same order of magnitude, but this fact cannot he a priori assumed. Then one calculates the fundamental solutions for the equation of the potential (paper [2.11]) and the fundamental solutions for the systems of equations of aerodynamics : the steady system[2.8], the oscillatory system [10.17], the unsteady system [2.6], (2.7]. On these solutions will rely the theories from the forthcoming chapters. The models we have already presented are the so called classical or linear models. They are suitable for the thin wings and thin profiles because they rely on the following assumptions: 1) one uses a linear boundary condition, 2) the boundary condition is imposed on the support of the wing (the segment (-1.1] for the profile, the plan-form D for the three-dimensional wing), :3) the equations of motion are linearized. The development of the scientific computing allows us to develop more exact methods. Indeed we can give up to the first two assumptions using xvi the boundary integral equations method (BIEM). also called the boundary element method (BEN), which was employed for the first time by Hess and Smith [7.9], [7.10]. The integral equations on the boundary are obtained imposing the exact boundary condition on the boundary of the wing. The integral equation is discretize d using, for example, the collocation method. One obtains an algebraic system which is solved numerically. The linearization of the equations of motion is necessary only in the case of compressible fluids. The theory that we have developed is thus valid for every body in an incompressible fluid and for at thin body in a compressible fluid. Two chapters from this book, Chapter 4 for the 2d airfoil and Chapter 7 for the 3d airfoil are baLsed on our papers (L. Drago.1 and A. Dinu). The comparison between the known analytical results and the numerical results shows a very good agreement. In the Appendices we give some results concerning The Distributions Theory, The Singular Integral Equations Theory, The Principal value and The Finite Part, Gauss-type Quadrature Formulas, etc. In every work one finds, in a certain measure, both the achievements of the predecessors and of the researchers contemporaneous with the author. Among the people which have directly collaborated with tae, I have to mention at first my professors Victor Va lcovici and Caius lacob, who introduced me in the field of aerodynamics. I also mention my younger colleagues Nicolae Marcov, Liviu Dint, Dorel Homentcovschi, Adrian Carabineanu, Victor Tigoiu, Vladimir Cardoi, Gabriela Marinoschi, Stelian Ion and Adrian Dinu. They were my students at the University of Bucharest, but I learned a lot from their papers. Some of them were my fellow - workers in the aerodynamics research, many of them stimulated me with their youth and their way of thinking in our seminars from the Faculty of Mathematics of the University of Bucharest. I am very grateful to all of them. My special gratitude goes to Adrian Carabineanu for his work in performing the English translation of the book, to Adrian Carabineanu and Stelian Ion for typesetting the monograph in Latex and to Victor 'igoiu for his activity in finalizing the 195D Grant with the World Bank. I acknowledge that the hook was sponsored by MEC-CNCSIS Contract 49113/2000, Grant 195D with World Bank. LAZAR DRACOS Chapter 1 The Equations of Ideal Fluids 1.1 1.1.1 The Equations of Motion Elements of Kinematics In this Chapter we present the equations governing the flow of ideal fluids. On the basis of these equations we shall develop the theory in the forthcoming chapters of the book. It is well known (see, for exannple, [1.11J), that the fluid flow is defined by the di eomorphism x = X(t, X), (1.1.1) where X is the vector of position of a particle P in the reference configuration (for fluids this is the initial configuration), and x is the vector of position of the same particle at the moment t. For a fixed X and a variable t , the equation (1.1.1) furnishes the motion law for the particle having the vector of position X. Hence the velocity and the acceleration of the particle will be given by the formulas v (t, X = It X (t, X ), a (t, X = cit. v (t, X). (1.1.2) The fluid is a continuum medium. It means that the support of the initial configuration is it domain Do. The image of this domain by the diffeomorphism (1.1.1) will be denoted by D and one demonstrates [1.11J that it is a domain. The functions X;(t, Xi, X2i X3) appearing in (1.1.1) belong to the class C2(D0) and the Jacobian is 0(X1, X2, Xa) V(XI, X2, X3) # (} (1.1.3) The velocity field defined in (1.1.2) may be discontinuous in isolated points, on abstract. curves or across abstract surfaces. This kind of surfaces will be named shock waves. THE EQUATIONS OF IDEAL FLUIDS 2 The fluid flow may be described by functions defined on Do, i.e. functions having the form 46(t, X) , or by functions defined on D, i.e. functions having the form F(t, x). The first presentation is called the material description, because it utilizes quantities attached to the material particles, the second is called the spatial description, because it furnishes information about the particles which are located at the moment t in the points of a domain D. The material derivative of the quantity 0 attached to the fixed and it is given by the formula particle X, is denoted by 46(t, X). (1.1.4) In order to obtain the derivative of F when X is fixed, we have to take into account that F depends on the coordinates X, through the functions Y{. Using the derivation rule for the composite functions, we obtain for the material derivative OF OF d X; OF OF OF +(v V)F. (1.1.5) F(t' x) dt + as d t +V'r7x; t For studying the motion of fluids we employ the spatial description. The main quantities are: the density (or specific mass) p(t, x ), the pres- , T. sure p(t. x ), the velocity field v (t, x ), the temperature T(t, x ), the entropy s(t, x) etc. The acceleration which is the material derivative of the velocity is obtained by means of the formula: a = at + (v 0)v . (1.1.6) Utilizing the derivation rule for the determinants, from (1.1.3) one obtains Eider's theorem J = J div v (1.1.7) and then (see for example [1.11]) Reynolds's formulas for continuous inotions (i. e. motions characterized by fields belonging to the class) C1((to,tII X DWff dtJDF(t,x)dv= r - 1 ID (F'+Fdivv)dv= r ( 1.1.8) BtF dv+ / DFv nda fD [OF + div(Fv)J dv = JD and Reynolds's formula for motions with shock waves dtI F(t,x)dv= J_a_dv.f-J E++E_ F v neda-J fFlld da, s ( 1.1.9) 3 THE EQUATIONS OF MOTION - F_, and d is the displacement velocity of the surface of discontinuity S (see formula (2.5.1) and figure 2.5.1 from where OFJ = F.,. (1.110. The Equations of Motion 1.1.2 The principle of conservation of mass is d dt,Dpdv=O (V)DCD. (1.1.10) For continuous motions one utilizes the formula (1.1.8) and one obtains the equation of continuity p+pdivv =0. (1.1.11) The general expression for the principle of variation of the momenturn is t L"= dPwdv -jDpnda+ JDpf dv (V)DCD, (1.1.12) f representing the force per unity of mass. For continuous motions, from (1.1.11) and (1.1.6), one obtains Euler's equation p[ + (v V) v] =pf -grade- (1.1.13) The balance equation of the energy (the first principle of the thermodynamics) is j ( aD JOD + /Dpf - vda- /D q nda (V)DCD, (1.1.14) e representing the specific internal energy, and q, the flux of heat vector. For continuous motions, taking into account (1.1.11) and (1.1.13), we deduce Pe = -pdiv v - div q . (1.1.15) In ideal fluids the processes are reversible. Eliminating div v by the aid of equation (2.1.11) and employing the second principle of thermodynamics (1.11], p.54, one obtains the fundamental equation of thermodynamics de=Tds - pdv, (1.1.16) THE EQUATIONS OF IDEAL FLUIDS 4 where s is the specific entropy and v is a notation for 1/p. Eliminating div v from (1.1.15) and taking into account (1.1.16), we deduce the following remarkable form of the equation of energy pTs= - divq. (1.1.17) This shows that if it is possible to neglect the change of heat (it is the case of aerodynamics where the velocities are great), then s = 0. (1.1.18) The equation (1.1.18) indicates that s is constant on trajectories, the constant varying from one trajectory to the other. One calls such a motion isentropic motion.. If there exists a configuration where the entropy constant is the same everywhere, then in every configuration arising from the first one, the constant will be the same everywhere. Such a motion is called homentropic or isentinpic everywhere. The perfect gas is characterized by the following equations of state p=pRT, (1.1.19) The first one (which is obtained from the laws of Boyle-Mariotte and Gay-Lussac) is the thcrmic equation (or Clapeyron's equation) and the second is the caloric equation. It is easy to prove (see for example [1.11) p. 57-58) that for this gas we have p }+Co, e= P7 //J 1 -+C1. -Y-1p (1.1.20) c, being the so called specific heat at constant volume and y = cp jc where cP is the specific heat at constant pressure. For the air 7 = 1.405. From the expression of a it follows for the homentropic motion: p = kp'', (1.1.21) k representing a constant. The quantity c, defined by the formula = (dp/ ' (1.1.22) has the dimension of a velocity. One shows (see for example [1.10)) that it gives just the speed of propagation of the surfaces of discontinuity of the pressure (sound waves). For the ideal gas in homentropic motion it follows c2 = 7plp. (1.1.23) 5 THE POTENTIAL FLOW The Potential Flow 1.2 1.2.1 Helmholtz's equation. Bernoulli's Integral If the tnassic forces possess a potential f = grad II (in aerodynamics these ford (representing the weight of the air) are neglected) and if the fluid is characterized by a thermodynamic law having the form p = p(p), where p is a derivable function, defined for p > 0 such that ;7(p) > 0, then we deduce f- Igrad p= grad II -1 pp) Utilizing the identity (v V) v = curl v x v + grad (v 2/2), Euler's equation (1.1.13) becomes / f +curlvxv=grad(n-1 dp-2v2). jt P \\ (1.2.1) This is Hetnahoftss equation. The flow of a fluid is irrotational in a domain D if everywhere in D. we have curl v = 0. This equation constitutes the necessary and sufficient condition for the existence of a differentiable function jp(x, x), such that v=grad 44. (1.2.2) Such a flow is called potential. Applying the operator curl in (1.2.1), we eliminate the term in the right hand side of the equation. The resulting equation is integrated [1.11] as follows 0 P (.VX) = X(t,X), (1.2.3) WO and m) representing the vortex (2w =- curly) and respectively the density in a reference configuration. The formula (1.2.3) shows that if the motion is potential in a configuration, it remains potential in every theorem). configuration arising from the first one We can therefore put curl v = 0 in (1.2.1). It results Bernoulli's integral Ot+2vz+ J LP -n=C(t) 4t=8W Ot. (1.2.4) One may give to the spatial constant C(t) the value zero ([1.11], p.90). Neglecting the messic forces, Bernoulli's integral is tt+2vz+ J Pp=0. (1.2.5) 6 THE EQUATIONS OF IDEAL FLUIDS The motion of a fluid is stationary or steady if the velocity field does not depend explicitly on t. Nowr (1.2.4) becomes 2v2+ J ap-n=C. (1.2.6) P In this case the constant cannot be zero. For the perfect gas in homentropic motion we have the formula (1.1.21). With a convenient notation of the constant, from (1.2.6) we deduce v- + 'Y P = 7 PO = _ , (1.2.7) 2 T::_1 -P 7-lpo 7-1 pa, pn and co representing the pressure, the density and the square of the sound velocity for the fluid at rest. From (1.2.7) one deduces that in the compressible fluid there is a superior limit v of the velocity which is obtained for p = 0 and a critical value v, which is obtained for v = c. These values are VCr = Co Vmax = Co V7 +2 _ (1.2.8) For v < v ., the flow is subsonic and for vrr < v < vmm, the flow is supersonic. 1.2.2 The Equation of the Potential The equation of the potential is obtained from the equation of continuity (1.1.11), Bernoulli's integral (1.2.5) and the equation (1.1.21), assuming that v has the form (1.2.2). Calculating the material derivative of (1.2.5), we obtain: dt(5t+2tr2)+P=0. (1.2.9) Taking into account that p depends on t through the agency of p and utilizing the notation (1.1.22), we deduce _± d dpP P (Oe + 2v2) (1.2.10) Replacing in the equation of continuity, we obtain the equation of the potential c2A -(v .V)(v -V)o -2(v V).6e-#u=0, (1.2.11) *r11F: POTENTIAL FLOW where c' depends on 40. We find this dependence in the case of the perfect gas in homentropic flow. Taking into account (1.1.21), Bernoulli's integral (1.2.5) becomes v2 4+ + )-i {n7-t -1710-t} = o, (1.2.12) representing the density in a reference state. From (1.1.21) and (1.1.22) it follows cc = kyp0-1, and from (1.2.12) we deduce: po 1 (+v2)]7_1 =p 11 P=poll-7 1 (fit+2v2)Jy-1 (1.2.13) , (1.2.14) CS +2v2), (1.2.15) representing the pressure and the sound velocity in the CO reference state of density po. For the steady flow one obtains the equation Po and c2A4'-(v .V)(v cI)4'=0 (1.2.16) and the formulas // 1 P=po1--y21 } Jy-1 y P=poll-y 21V1\-V-1 , (1.2.17) c2 = c9 - y-1 v2 2 (1.2.18) where po, po and co are quantities corresponding to the fluid at rest (v = 0). One demonstrates [1.10], p. 207, that the equation (1.2.11) is hyperbolic and the equation (1.2.16) is elliptic for the subsonic flow (v2 < c'-) and hyperbolic for the supersonic flow (c2 < v2). The equality v2 = c2 occurs only on curves in the two-dimensional flow or surfaces in the case of three-dimensional flow. These varieties are separating the domains where the flow is subsonic from the domains where the flow is supersonic. This kind of motion is called transonic. We denoted tit = I v 12. 8 THE EQUATIONS OF IDEAL FLUIDS 1.2.3 The Linear Theory When the equations (1.2.11) and (1.2.16) are utilized in order to determine the perturbation produced by a thin body in a fluid having a known flow, they may be linearized Let us consider for example, that the uniform flow of a fluid having the velocity U,,, i, the pressure p,, and the density p,,,, is slightly perturbed by the presence of a body. We shall denote by v = UU i + v' (1.2.19) the velocity field for the perturbed flow ( v' is the perturbation of the velocity field). We assume that all the coordinates of v' have the same order of magnitude a (e representing a small parameter which characterizes the body). Hence we assume that. (1.2.20) where I- VI is bounded and e << 1. According to Lagrant o-Cauchy's theorem, the perturbed flow is potential, since it arises from a potential flow. Hence, (3)¢, such that v = grad O and from (1.2.19), it follows v' = grad p and -0 =U,,,r+V. (1.2.21) The condition (1.2.20) is equivalent to Sit) = (Yr . ;Py, s5z) . (1.2.22) Replacing in the formulas above and neglecting the terms O(E2), we deduce v = (U,o + E ;5;r, E v, 63z), 2J2 = U200 + 21J. pa + 0022) (1.2.23) and then (1.2.24) U p7= - 2U.v*1 - pc1 = 0, (1.2.25) (1 - Af`)Vls +Wv + Vz_ = 0, where Af(= U,,,/c,,,) is Mach's number in the perturbed flow. One easily find that the equation (1.2.24) is hyperbolic, and the equation (1.2.25) is elliptic if Al < 1 and hyperbolic if Al > 1. The motions characterized by Al < I are named subsonic, and the motions with Af > 1 arc named supersonic. The pressure is deduced from Bernoulli's integral, (1.2.5), which may be written as follows 1 2 actual v2) 0o + p f clp -;p = 0. (1.2.26) THE POTENTIAL FLOW 9 if we neglect the massic forces. It obviously results de+2 (v2-U2)+f dr=0, (1.2.27) Let us assume that the relation p=Poo+0(e). (1.2.28) is also valid. Then we deduce p(p) = poo + J r dp rx 1 = I loo (p `poo) dr + 1 -- J pW JJ j 0 I P 0(e2) = p + P c2px pay + O( r2) 0(6 c?C p 2) . poll (1.2.29) Replacing (1.2.21), (1.2.23) and (1.2.29) in (1.2.27), we find in the linear approximation the formula for the determination of the pressure P - P,x = -f).G%e + U,otp,.), (1.2.30) Since the previous formulas will be deduced in a different manner in Chapter 2, in dimensionless variables, we are going to introduce here these variable (x', y'. z', t'), by means of the formulas (x, y, z) = Lo(r`, y', z'), t = (Lo/UU)t', y: = Uaip'p - P = PoUp', (1.2.31) Lo representing a reference length which is not specified yet. So, the equation (1.2.22) becomes j112Vt.e. = 0. (1.2.32) and the formula (1.2.30). l), = - (vi. + co=.) . 1.2.4 (1.2.33) The Acceleration Potential Another function, often utilized in aerodynamics, is the acceleration potential ;', introduced by Prandtl. Its existence is ensured for the unlimited fluid characterized by the lawn p = p(p), by the equation a = -brad / .11 dam' P (1.2.34) TILE EQUATIONS OF IDEAI. FLUIDS 10 resulting from (1.1.6). (1.1.13) and (1.2.1). Denoting a= grad !, , (1.2.35) we get v" = - J clp' + C(t) . (1.2.36) where C(t) is an arbitrary function depending on t, which can be determined Lw imposing the value of 0 for a certain state. For example, if the fluid is incompressible and one consider, that t/' vanishes at infinity (where p = p,), it follows C = p,/p, whence (1.2.37) V, = (P. - P)//). Since t' represents the perturbation of the pressure, it is usually called the prrssum function. The relation between 4) and 0 may be deduced from the equations (1.2.5) and (1.2.26). We have r = grad grad (wt + (1/2)v2) _ -grad J (1.2.38) If the flow is uniform at infinity, with the velocity U., it results tl =Vet + (1/2){tom - U ) (1.2.39) In the framework of the linearized theory this relation becomes V'= P +U.oP.- (1.2.40) Taking into account the permutahility of the operator a/Ot + U,,(9/0x with the operators appearing in (1.2.24), it follows that '0 satisfies the sane equations (1.2.24) or (1.2.25). From (1.2.29) and (1.2.39), it results that ,, has the same expression (1.2.37) in the case of the linear approximation for compressible fluids. In order to write the boundary conditions, we have to indicate the relation between 0 and the coordinates of the velocity. For steady motions, we obtain from (1.2.40), (taking into account that rv must vanish at infinity) : Uc = jW4)dx, Ucw as-dX. (1.2.41) For unsteady flow, deriving (1.2.40), we get .2.42) (1.2.42) THE SHOCK WAVES THEORY 11 Performing the change of variables t, x --+ r, £ : r=t, t;=x-U".t (1.2.43) and denoting w(t, x, y, -) = w (T, 0(t, x, y) z) _ V"'(r, U, , y, z) , (1.2.44) we deduce ay//Or = N//&r, whence, integrating Ta (1.2.45) 00 Returning to the variables t and x, we get w(t, x, y, z) = f +U.r',y,z)dr' _ (1.2.46) fe 1.3 The Shock Waves Theory 1.3.1 The Jump Equations The principles of motion for the continuous media are expressed in terms of balance equations. The general form of a balance equation is dt,npQdv - Jen R nda+JnpSdv, (1.3.1) where D is a material domain, Q and S are tensors having the same order and R is a tensor having the order greater with a unity. For the ideal fluid, the principle of conservation of mass (1.1.10) has the form (1.3.1), with Q=I, R=0,S=0. (1.3.2) The principle of variation of the momentum (1.1.12) is obtained for Q=v, R=-pI, S= f, (1.3.3) I representing the unity tensor, and the balance equation for the energy (1.1.14) for Q=a+(1/2)v2,R=-pv-q,S= f v. (1.3.4) THE EQUATIONS OF IDEAL. FLUIDS 12 Let us see now the balance equation (1.3.1) in case that D is crossed by a surface of discontinuity with the displacement velocity d ([1.11), p.28). Applying the derivation formula (1.1.9), the equation (1.3.1) gives .I [ UL 51 pQ)-AS1 l dv + i;:++E.. pQ(v n)-Rnda- llpQQdda=O. We pass to the limit superposing F,+ and F_ on S (fig. 1.3.1). In this case, n+ becomes n, and n_ becomes -n. Fig. 1.3.1. If we assume that the integrand from the first integral is bounded, the first term tends to zero, because vol D± ---. 0. Hence one obtains: jDPQ(v.n_d)_Rnoda=07 (1.3.5) where we denoted (1.3.6) The limit values encountered here are continuous on S (they are not continuous across S), and S is arbitrary, because the balance equations are formulated for every D. By virtue of the fundamental lemma ([1.11) p.31), from (1.3.5), we deduce OpQ(v n - d) - Rn = 0. (1.3.7) 13 THE SHOCK WAVES THEORY This is the equation which leads to the balance equation. It establishes the connection between the limit values from the two parts of the shock waves. If across a surface the fields have discontinuities, this surface is a shock wave. Utilizing (1.3.7) for (1.3.2), one obtains; =0, (1.3.8) for (1.3.3) Opv(v n - d) +pn, = 0, (1.3.9) and for (1.3.4), (1.3.10) The conduction laws utilized in practice (particularly Fourier's law) make impossible the discontinuity for q. Hence we shall utilize the equation (1.3.10) as follows 1 p(e+2v2 =0. (1.3.11) To the previous equations we shall add the inequation: Ops(v . n - d)Q > 0, (1.3.12) coming from the Second Principle of Thermodynamics ([1.111, p.50). For the ideal fluid we have an equality. 1.3.2 Hugoniot's Equation The jump Equation (1.3.11) is quite complicated. It can be replaced by a simple equation (Hugoniot's equation), which establishes a connection only between the thermodynamic quantities from the two sides of the shock wave. In order to deduce these equations, for the sake of simplicity we introduce the propagation speed P = d - v n. In this way, the jump equations become r OPPI =0, (1.3.13) OpPv-png =0, (1.3.14) =0, (1.3.15) THE F.QUATIOSS OF IDEAL FLUIDS 14 [pPsO <0. (1.3.16) From these equations we have to find the unknowns behind the shod; wave and its propagation speed if we know the state of the fluid in front of the shock wave. For the salve of simplicity too, we mark by the index 1 the limit values (on S) in front of the shock wave and by the index 2 the limit values behind the shock wave. The normal to S is positively orientated from the region 2 toward region 1. We exclude the situation Pi = P-2 = 0, because in this case the surface of discontinuity does not cross the fluid (it is a material surface moving together with the fluid). For deducing Hogoniot's equation, we notice that from (1.3.13) it follows that the quantity p P is continuous across the shock wave. N e denote by rn=pjP1=P2P2 (1.3.17) this quantity. Since in the jump equations we employ only the velocity on the shock wave in a current point of the wave, we write v=v,,n+vet. (1.3.18) v'2 = vn representing the normal component and vi the tangential component (situated in the plane determined by n and v). We denoted by t the versos of the tangent. Projecting (1.3.14) on the tangent and on the normal to the shock wave and taking into account that by hypothesis, in T 0, we deduce UVt =0, rnOv,, 11 - H =0. (1.3.19) The first equation shows that the tangential component of the velocity remains constant across the shock wave. From the definition of the propagation speed P and from the conservation relation (1.3.17), we deduce Vn =d-P=d-nzT, r- 1/p, (1.3.20) the second equation from (1.3.19) becoming rn2OTO + 0. (1.3.21) From (1.3.15) taking into account the equations (1.3.19), (1.3.20) and (1.3.21), we notice that: pPvt 9 =III vie) = 0, OpP'el1 = -21n2dOrl + rn3[r20 = 2dOpE - Innllpll (rl + r'2) THE SHOCK WAVES THEORY 15 From these equations and from (1.3.15), we get: 2Ohl = OpO(T1 +r2), (1.3.22) where h = e + pr is the enthalpy ([1.11], p.55). This is the first form of Hugoniot's equation. Performing the calculations in (1.3.22), we get also the form (1.3.23) 2Oej + rlJ(PI + P2) = 0, and. for the perfect gas, when e is given by (1.1.20), one finds the third expression: P2 P1 = b+ 1)p2 - (7 - 1)PI (1.3.24) (7 + 1)pt - Of - 1)P2 which is useful in applications. The inequality (1.3.2) shows that if m > 0 (the shock wave surpasses the fluid), then 81 < 82; if in < 0 (the fluid surpasses the shock wave), then 82 < s1. In all cases behind the shock wave the entropy is not decoeasing. It increases or remains constant. 1.3.3 The Solution of the Jump Equations In the sequel we are going to demonstrate, for the perfect gas, that giving the state of the medium in front of the shock wave and the propagation speed Pt, one may determine completely the state behind the shock wave. To this aim, we introduce the numbers: M1 = Pt/cl , M2 = P2/c2, (1.3.25) where c1 and c2 represent the sound velocity in front of the shock wave respectively behind the shock wave (CI = 7Pi/Pl, 4 = 7P2/p2). By hypothesis, M1 is known and M2 is unknown. From (1.3.21), we deduce: P2 - P1 = in2(rl _,r2) = p1P1 (1 - r2/rl) = 7P1Mi (1 - 72/r1) . The function (pa/pl) -1 may also be obtained from (1.3.24). Comparing the expressions we get:\ T2-1= 71 2 (M1-11,E-1= 27 7+1 / Pt 7+ (M2 -1). (1.3.26) From the relation P1r2 = P2r1i it follows: f -1)I(1-Jt't2)=11th-1 (1.3.27) THE EQUATIONS OF IDEAL FLUIDS 16 From the perfect gas law p = pRT and from (1.1.20), we obtain: - 81 82 C11 fi-C )(p1)' = in J (i) r 111 r2/rl and p2/pt being replaced from (1.3.26). Prandtl's Formula 1.3.4 For the stationary wave (d = 0), we have v = -rnr, such that the equation (1.3.11), for the perfect gas, becomes Q1v2+ ry PO y-1P 2 =0. (1.3.29) The expression written here intervenes in Bernoulli's integral (1.2.7). Taking into account the meaning given there to v,,,,,x , we have the relation P1=1 2 2+ iv max=l 2vt.7-1pt 1 P1=12v2msx 2 11 (1.3.30) which shows that vmax is invariant across the shock wave. Taking into account that 0, from (1.3.30) we deduce: 1 y P1 PI 2 t[ 2 0 _ 'r' - l pj P11}2 1l (1.3.31) / We eliminate p2/pt by the aid of equation (1.3.24) and pre/pl, by the aid of relation PIV1.. = P2v2n which results from (1.3.17). Employing once again (1.3.30), we deduce: ('y + 1)v1nv2n - (/ - 1)v? = 2yP1/Pl 1)(v 1 - vl ). Writing vu = v2i = vi, we obtain Prandtl's fonnula v1nt'2n= -llvmax-Vi y+1 ry+1 )=t2.r(1.3.32) the significance of vt. being that of (1.2.8). In the expressions vma and va. utilized here, cp is the sound velocity for the fluid in front of the shock wave (at rest). The relation (1.3.32) gives c2,, when the state of the fluid in front of the shock wave is known. For the normal shock According to this relation, if (:;( = 0), the relation becomes vttt2 = vt is supersonic, then v2 is subsonic. THE SHOCK WAVES THEORY 17 The Shock Polar 1.3.5 Buselnann (Vortrage arcs dem Gebiete des Aerod amik, Aachen, 1929), gave a graphic method for constructing v2 if the angle with v1 is known. To this aim, we consider in the current point Al from the shock wave (fig. 1.3.2) the AIX axis, pointing in the direction and sense of v1 and the AIY axis in the plane determined by v1 and the normal n to the shock wave. Fig. 1.3.2. We use the normal pointing towards the state (2); the change of the sense of the normal does not affect the jump relations (excepting the relation (1.3.16), whose signification is known), because in these relations the normal intervenes linearly. Because of the continuity of the tangential component of the velocity, v, lies also in the plane determined by v1 and n and it will have on the tangent versor t the same projection AIQ like vl. We denoted by PI and P2 the extremities of the vectors v 1 and v2 and by 0, the angle between v2 and v1. Our aim is to find the geometric locus of the point P2 when 0 is varying. We shall use, obviously, Prandtl's formula (1.3.32). Denoting by (X, Y) the coordinates of the vector v2, from the figure, it results X = v2 cos 0, Y = 1v2 sin 0, vl = v1 sin o, vt = v1 cos B , v2., = v-± sin(v - 0) = X sin Or - Y cos a . (1.3.33) So, Prandtl's formula becomes: Xsin2a-YsinorCosa= vi - 7+1 VICOS2a (1.3.34) THE EQUATIONS OF IDEAL FLUIDS 18 From the triangle P1P2R, where the angle P1P2R = a, it follows tana = (vi - X)JY, such that (1.3.34) becomes (1.3.35) Y2(b- X) = (vl X)2(X - a), - %%'here we denoted 2 a= Vt , b=v`,+ vl 2 y+1 vi. (1.3.36) From the assumption that vt is supersonic, it follows that a < vi, we deduce vj < b < v.. and from vi < The curve representing eq. (1.3.35) is symmetric with respect to MX, it intersects the MX axis in the points A(a) and Pt (v1) (which are inverse with respect to the sonic circle x2 + y'' = v,2,,), the last point being double and has a vertical asymptote in B(b). The curve is real for a < X < b and it is represented in figure 1.3.3. It is called foliurn of Descartes. It is the shock polar. Fig. 1.3.3_ Let us show now how we construct v2 when we know the angle 0 made with vl. First of all one constructs the shock polar. This is determined only by the state (1). The same state determines the angle ©p made by the tangent MT with the MX axis. If 0 < Bo, there exist three intersection points 1,1_,13, of the radius vector with the polar, but only one is P2. The points on AT are eliminated because they correspond to instable shocks. One eliminates the points THE SItOCI' WAVES THEORY 19 1:1 because. generally, the shocks are compressive. (v2 < vt ). P2 will coincide therefore with 12. If C is the intersection of the polar with the sonic circle and 9° is the corresponding polar angle, then v2 will be supersonic if 0 < B, and subsonic, if 0 > O. Once 11 determined, it results the direction of the versor t (the orthogwua from M on P1P2) and then the angle a. The density, the pressure. the temperature and the entropy behind the shock wave is obtaincNi front (1.3.26) and (1.3.28), setting M1 = -vln/et = -MI sin or, where .111 = v1/c1 is Mach's number in front of the shock wave. 1.3.6 The Compression Shock past a Concave Bend We consider a supersonic flow having the velocity v1 i, the density p) and the pressure P1 in the presence of a concave bend having the opening b (fig. 1.3.4a)). The wall ME produces a compressive shock, the discontinuity line MM' being characterized by the unknown angle a. Behind the shock, the velocity which has to be tangent to ME, will make the angle 6 with MX . If 6 < Bo, the position of P2 will be given by 11 from the polar corresponding to this motion. It follows like above, a and the flow behind the shock. If 6 > B(), one cannot satisfy for v2 the condition to be parallel to AI E. As the experience confirms, the assumption of a rectilinear shock wave Al Al' cannot be taken into consideration. In this case one admits the existence of a detached curvilinear shock wave which is formed in front. of Al (fig. 1.3.4b)). The experience confirms this assumption. Flence the detached shock waves are formed in front of the dihedron (S > A() (fig. 1.3.5a)), or in front of the bodies with rounded leading edge (fig. 1.3.5b)). As an information we give the following values : (10 = 10° for M1 = 1.42 and Bo = 22.55°, for M1 = 2. Depending on Fig. 1.3.4. the shape of the body, i.e. on the values of 0, behind the shock wave, THE EQUATIONS OF IDEAL FLUIDS 20 C 0<00 Pig. t.3.5. we meet regions where V2 is subsonic (11M2 < 1), if 9 > 9, and regions where u2 is supersonic, if 9 < O. When 9 is big, the compression is big. When 9 is passing to small values, it appears a detente and the velocity becomes again supersonic. On the direction MV the shock is normal (vi v2 = v,2 and v2 is subsonic. The subsonic regions are separated from the supersonic ones through sonic lines CD and C'D'. Behind the shock wave, the flow is transonic. The shock waves theory will be present in the transonic flow and the hypersonic flow. Chapter 2 The Equations of Linear Aerodynamics and its Fundamental Solutions 2.1 2.1.1 The Equations of Linear Aerodynamics The Fundamental Problem of Aerodynamics The fundamental problem of aeroclynamica consists in determining the perturbation produced in a given state of a fluid by a certain motion of a body. The given state of the fluid is called basic state or unperturbed state. The unperturbed state of the fluid may be the rest state, the uniform flow state, or more generally, the state given by the flow with an imposed non-uniform velocity field. In this book the unperturbed state will be either the rest or the uniform flow. In his turn, the body may he fixed, moving uniformly or may have a general imposed motion. Obviously, a fixed body in at rest state of a fluid does not produce any perturbation. The most common are the case when the unperturbed fluid moves uniformly and the body is fixed and the case when the fluid is at rest and the body has a given uniform motion. As we shall see in 2.1.6, these cases are equivalent, from the mathematical point of view. In both cases, the resulting perturbation will he stationary. If the perturbing ixxly has a non-uniform motion, the perturbation will be non-stationary, whatever should be the state of the fluid. In the sequel we shall consider only the cases when the unperturbed fluid is at rest or has an uniform flow. The problem of determining the perturbation may be practically solved only in the case of small perturbations, when we can neglect their products (we keel) only the principal parts of the equations). In these cases, the linear systems of equations obtained for the perturbations may be investigated either with the methods of the classical analysis or with the methods of the theory of distributions. The systems will be linear 22 LASEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS with constant coefficients in case that the basic state is uniform and they will be linear with variable coefficients in case that the basic flow is not uniform. Since, as we have already mentioned, the last case will not be treated in this book we give here some references, [2.4], [2.14], [2.15], [2.20], (2.22), for the reader interested in this subject. The order of magnitude of the perturbation is determined by the basic flow and by the shape of the body. For some basic flows, a slender body with it small incidence changes slightly the flow, i.e. produces small perturbations (governed by linear systems of equations). We cannot establish in advance the conditions of validity of the linear theory. This will be done after determining the solution of the linearized equations, imposing not to obtain results which cannot be accepted from a physical point of view. In this way, from Chapters 3 and 8, it will follow that in the case of steady flow, the linear theory is not valid when the basic flow has approximately the sonic velocity (M = 1), or has an hypersonic velocity (M > 3) even if the body is slender with a small incidence. In these cases one has to employ the non-linear equations for determining the perturbation. A special feature has the unsteady flow because the system of equations of motion is hyperbolic and it is well known (see, for example, [1.6]), that for this sort of problems, Cauchy's problem is correct, i.e. the solution depends continuously on the initial data. If these data are small, the perturbations will remain small at every instant. 2.1.2 The Equations of Motion First of all we assume that the basic motion (the unperturbed motion) (the Ox axis is of a fluid is an uniform motion with the velocity taken to be on the direction and in the sense of the unperturbed stream and we denote by i the versor of this axis), the pressure po,, and the density p,,. We denote by xi, yi, zi the generic spatial coordinates, by t i the time and we introduce the dimensionless coordinates x, y, z, t by mean., of the relations (2.1.1) (xi, yi, zi) = Lo(x, y, z), Uwtl = Lot. where LO is a characteristic length which has to be specified in every problem we have in view. We assume that the uniform motion defined above is perturbed by the presence of a body which has a prescribed motion. Let F(ti, xi, yi, zt) = 0. (2.1.2) 23 TILE EQUATIONS OF LINEAR AERODYNAMICS be the equation of the surface of the moving body. Denoting by V 1, p1 and pt the velocity, the pressure and the density for the perturbed flow, we shall write: V 1 = UU(i + v), Pt = P.- + P"Uoop, pi = Poa(1 + P), (2.1.3) the first terms defining the basic motion, and the last ones, the perturbation. Obviously we have: lien (v, p, p) = 0. _-.-OD (2.1.4) We neglect the heat changes (in aerodynamics this assumption is plausible, because the variation of the phenomena are very rapid and there is not enough time for the heat change) and we assume that the fluid obeys to the perfect gas law. In these conditions, the perturbed flow will be determined by the following equations: 01+p1divlV1 =0, sl =0, P1V1+grad, pi =0, (2.1.5) s1 representing the specific entropy and "the point", a notation for the material derivative: (2.1.6) s 1 = c,.1n(Pl /Pi) + C, f = of /0t1 + (V 1 V 1)f . The index 1 attached to the differential operators indicates that the derivatives are calculated with respect to x1, yl, zl Taking into account the expression of the entropy, from the equation (2.1.6), we get: P1p1 = 'YPt11, (2.1.7) so that we can eliminate pl from the first equation (2.1.5). Hence, we have to take into consideration the system (2.1.8) 01 + ypldiv3 V1 = 0, p1Vl +grad, p1 = 0 which, taking into consideration (2.1.1) and (2.1.3) becomes: A!2[pt+(1+u)p=+VVlpy+wp;]+(1+-tM2p)(u?+vy+w=) +p)[ut+(1+u)U=,+ ruy+wu,]+ps=0, +p)[Vt++u)vs+vuy+wvZ]+Py=0, (1+p)[wt+(1+u)w2+trwy+wWz]+p: = 0, = 0, (2.1.9) (2.1.10) (2.1.11) (2.1.12) 24 LINEAR AERODYNAMICS. F'UNDA'MENTAL SOLUTIONS where (u, v, w) = v, pt = dp/at, Ps = Op/tax,... and M = U00/c., coo = 'YPoo/Poo , (2.1.13) Al representing Mach's number in the basic motion and 'y, the ratio of the specific heats The condition for the perturbation surface (2.1.2) to be a material surface is F = 0 (the Euler-Lagrange criterion)and it may be written as follows Ft +(l+u)F,+vFy+wF =0. (2.1.14) This condition must be satisfied for F(t, x, y, z) = 0. The system of equations (2.1.9) - (2.1.12), together with the boundary condition (2.1.14), will determine the perturbation (p,v). A. The Linearization around the Uniform Motion 2.1.3 The Equations of Linear Aerodynamics We assume now that the equation of the perturbing surface is z = h(t, x, y) = ei (t, x, y), (2.1.15) where e is a small parameter and h(t, x, y) is a known function with continuous first order derivatives (fig. 2.1.1a)). If the perturbation surface is cylindrical with generators parallel to Oz (fig. 2.1.1b)), then the equation of the profile determined in the xOy plane is assumed to have the form y = h(t, x) = J (t, x) . (2.1.16) In this case, the perturbation will be plane. For the surface (2.1.15), we write F = eh(t, x, y) - z in (2.1.14). One obtains Eht + E(1 + u)h= + evhy = w (2.1.17) which has to be satisfied for z = Eh(t, x, y). The principal part from the left. hand side of the equality (2.1.17) has the order of e. We deduce that the right hand part must have the same order of magnitude. Hence it follows w(t,z) y,eh) = e'w(t)x,y,eh). We consider that this relation is valid all over the fluid, whence w(t, x, y, z) = e1 F(t, x, y, Z). (2.1.19) 25 TIIE EQUATIONS OF LINEAR AERODYNAMICS --s 0 a) Fig. 2.1.1. Taking into account (2.1.18). the principal part of the boundary condition (2.1.17) is h, (t, x, y) + h..(t, x, y) = w(t, x, y, 0) . (2.1.20) Taking into account (2.1.19), we deduce that the principal part of the product from (2.1.12) has the order e. It follows therefore that p has the same order. Hence, p(t, x, y, -) = Ep(t. X, r, =) , (2.1.21) the residual equation from (2.1.12) being Dw+p: =0, (2.1.22) where D is the material derivative operator for the unperturbed motion: D = a/at + 9/ax . (2.1.23) Taking into account (2.1.21), from (2.1.10) and (2.1.11), we deduce u(t, x, y, z) = F-u(t, x, y, z) , v(t, x, y, :) = e'u(t, x, y, z) (2.1.24) and the residual equations Du+p,,=0, Dv+py=0. (2.1.25) Now, the equation (2.1.9) becomes ?J2Dp+ dive = 0. (2.1.26) This equation together with the equations (2.1.22) and (2.1.25) which have the vectorial form Dv + grad p = 0, (2.1.27) LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS 26 constitute the fundamental system of the linear aerodynamics. Obviously, this system has to be integrated with the boundary condition (2.1.20) and the condition at infinity upstream lim (p, v) = 0. (2.1.28) If the perturbation is plane, the condition (2.1.20) will be replaced by ht(t, x) + h=(t, x) = v(t, x, 0) . (2.1.29) If the perturbation surface is fixed (i.e. h does not depend explicitly on t), the perturbation will be stationary, determined by the system 1v12p= + div v = 0 , vx + grad p = 0 limo(p, v) = 0 2 30) . (.1 and by the boundary conditions w(x, y, 0) = hr (x, y) , (2.1.31) v(x, 0) = hr (x) . (2.1.32) Obviously, the equations (2.1.26), (2.1.27) and (2.1.29) could be easily obtained from (2.1.9)-(2.1.12) and (2.1.14), supposing that all the perturbations have the same order of magnitude. But this thing has to be demonstrated. In the case of transonic flow, for example, the perturbations have different orders of magnitude (see Chapter 9). 2.1.4 The Equation of the Potential Applying the operator curl in (2.1.27), we obtain: Dcurlv = 0. (2.1.33) Taking into account the significance of the operator D, we deduce that curl v is constant on the parallels to the Ox axis, the constant generally varying from one parallel to the other. Since at infinity upstream (x = = -oo) the constant vanishes on every parallel to Ox, it follows that curl v = 0 on every trajectory coming from --cc. This property is not true for trajectories detaching downstream from the body. This fact will be better put into evidence by the fundamental solutions. In the irrotational zone we have v = grad cp(t, x) . (2.1.34) THE EQUATIONS OF LINEAR AERODYNAMICS 27 Taking into account this representation of the field v, from (2.1.27) we deduce p = -D<p = -(vt + ip:) . (2.1.35) The function f (t) which should be added in the right hand side of the equation may be considered equal to zero, since W is determined with the approximation of an arbitrary additive function of t. From (2.1.26) and (2.1.35) it follows Dip = M2D2co or, explicitly (1 - M2)fpxs + Wy, +'pz:. - 2M2WfZ - M2Vu = 0. (2.1.36) This is the equation of the potential. In case of the stationary flow, from (2.1.33) it results curly = f(y,z), f representing an arbitrary field which must be considered zero because for x -- -oo we have curl v = 0. Again it is true the observation that curl v is not zero on the parallels at the Ox axis detaching downstream from the body. In the irrotational zone we have therefore v = grad W(x) . (2.1.37) From the first equation (2.1.25), taking into account (2.1.28), it follows p(x) = -u(x) . (2.1.38) Replacing (2.1.37) and (2.1.38) in the first equation (2.1.30), it follows (1 - M2)9u + ,pyy + ,P:: = 0. (2.1.39) This equation may be obviously deduced from (2.1.36). For the incompressible fluid (M = 0) these equations become Al P = 0. (2.1.40) Applying the operator grad in (2.1.39), we get: (1 - M2)v22 + vyy + USX = 0. (2.1.41) Hence, the coordinates of the velocity (and the pressure) satisfy the same equation (2.1.39). The equations (2.1.36) and (2.1.39) have been obtained in another way in Chapter 1. There we utilized the Lagrange-Cauchy theorem in order to prove that the perturbed flow is potential. Here, without 28 LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS utilizing this theorem, we demonstrated that in the first approximation the perturbation is potential. In addition, we see here that downstream the perturbing body, the perturbation does not possess this property any longer. B. The Linearization around the Rest State 2.1.5 The Linear System Let us assume now that the basic state of the fluid is the rest state, which is perturbed, as above, by a prescribed arbitrary motion of a body. We denote by po and po the pressure and the density of the fluid at rest. We denote also by ca = -ypo/po the square of the sound velocity for the same state. Since, in this case, there is no characteristic velocity, it is not recommended to use the variables (2.1.1) and (2.1.3). It is natural to use dimensional coordinates and to put V1 = V, P1 = PO + POP, Pt = Po(1 + p), (2.1.42) litn (p, v, p) = 0. (x}-00 (2.1.43) The system of equations of motion has the shape p+(1+p)divv=0, (1+p)P=(c2 +7p)p, (1+p)v+grad p=0. (2.1.44) If the equations of the perturbing surface have the form (2.1.15), we shall have the boundary condition Eli, + EU) X + Evii, = w (2.1.45) which has to be satisfied for z = Eh(t, x, y). We deduce w(t, x, y, d) = eii (t, x, y, ch) (2.1.46) and the boundary condition w(t, x, y, 0) = ht (t, x, y) . (2.1.47) Assuming like in (2.1.3), that (2.1.46) is valid all over the fluid, w(t, x, y, z) = ew(t, x, y, z), 29 THE EQUATIONS OF LINEAR AERODYNAMICS from the projection of the second equation from (2.1.44) onto the Oz axis, we deduce P(t,x,v,z) = C PO, z, v, z) and the residual equation wt+pz=0. Acting in the sequel like in subsection (2.1.3), we obtain at last p t +divv = 0 , p t = c2pt , vt + grad p = 0 (2.1.48) We deduce the system pt+c2divv=0, vt+gradp=0 lim (p, v) = 0 III--= (2.1.49) and the solution p = c2p. The system (2.1.49) has to be integrated with the boundary condition (2.1.47). From (2.1.49) one obtains the fundamental equation of acoustics pa - cRAp = 0. 2.1.6 (2.1.50) The Uniform Motion in the Fluid at Rest Let us consider the particular case of the uniform motion of the per- turbing body in the fluid at rest. We assume that the motion is performed with the velocity U0 in the negative sense of the Ox axis. Putting V, = Uov, pi = poll + A, Pt = po + poop (2.1.51) and utilizing the variables (2.1.1), we get the system: Mop+(1+yAf,p)divv=0, (1+p)Mop=(1+yM p)P, (1+p)v+gradp=0, (2.1.52) where MQ is Mach's number for the fluid at rest. As above, one deduces the boundary condition (2.1.47) and the system (2.1.52) reduces to the residual form Af2pt+divv=0, vt+gradp=0 and M p = p. (2.1.53) 30 LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS In a frame of reference R' = Cx'y'z' (fig. 2.1.2) solidary with the body, having the axes parallel and having the same sense with the axes of the frame R = Oxyz (the frame R' is inertial), the system (2.1.53) has the form of the steady system (2.1.30). Indeed, we pass from the frame R to the frame R' by means of the Galilean transformation t'=t, =x+t, Y ,=Y, z'=z. (2.1.54) With this transformation, the system (2.1.53) gains the form of the system (2.1.26). (2.1.27) and the condition (2.1.47), the form (2.1.20). But, since we deal with a translation of the body, h in (2.1.47) has the form h(x + t, y) which is transformed in h(x', y'). The boundary condition in the variables x', y', z' will have therefore the form (2.1.31), so that it will determine a steady motion. The system of equations of motion in z', y', z' will have the form (2.1.30). Hence we have demonstrated that the problem of determining the perturbation produced by a body moving uniformly with the velocity -Uoi, in a fluid at rest, is equivalent to the problem of determining the perturbation produced by the same fixed body in an uniform stream with the velocity Uoi. 2.2 The Fundamental Solutions of the Equation of the Potential 2.2.1 The Steady Solutions This subsection is written on the basis of the paper [2.111. The fundamental solutions of the equation (2.1.39) are the solutions of the equation (1 - e ,+ (2.2.1) a representing Dirac's distribution. These solutions are, obviously, dis-; tributions. Utilizing the Fourier transform method, we shall obtain ternperate solutions. Taking into account (A.5.1), valid also for temperate solutions, we obtain using the notation a2 = IaI2, (a2 - M202).? = -1, (2.2.2) whence E_ 1 a2-M2ni --V-' 1 [(1 -A12)a +a2+a (2-2.3) FUNDAMENTAL SOLUTIONS OF THE EQUATION OF THE POTENTIAL 31 In the subsonic case M < 1, we shall denote fl= 1-M2, (2.2.4) and in the supersonic case (M > 1), k= M2-1. (2.2.5) In the subsonic case one utilizes the formulas (A.7.10) and (A.7.11). One obtains the following fundamental solutions: x+ 4w E 1 2Aft In 1(y +z) , n = 3, x f2 +y2, n = 2, (2.2.6) (2.2.7) For the two-dimensional case we have not written the additive constamt C - in j3 appearing in (A.7.11), because the fundamental solution is determined with the approximation of a solution of the homogeneous equation. In the supersonic case, one utilizes the formulas (A.7.14) and (A.7.15). One obtains the following fundamental solutions: 1 H(x - ky2+z2)} 27r .6 = - x-k(y+z) n=3, H(x - klyl), n = 2. (2.2.8) (2.2.9) H repn-senting, as we have considered in Appendix A, Heaviside's func. tion. From the definition of this function (A.3.13), it follows that the three-dimensional solution is different from zero only for x > kk/,/2 +z2. (2.2.10) This inequality implies x > 0, xs > k2(y2 + zs). The set of points from the space verifying these inequalities constitutes the interior of the cone with the vertex in the origin of the system of coordinates (the perturbation point) and the symmetry axis along the Ox axis (fig. 2.2.1a)). This cone is called Mach's cone. In fact, it is the characteristic cone associated to the partial differential equation (2.1.39) LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS 32 Y b) a) Fig. 2.2.1. in the hyperbolic case. The radius of the cone is x/k, and the angle p (the semi-opening) is determined by the formula tan y = xxk =k = Ml -11 (sin µ = M J . (2.2.11) In the two-dimensional case, the solution is different from zero only inside the dihedrou made by the plane; x = ±ky where x > 0 (fig. 2.2.1b)). This is Mach's dihedrun. 2.2.2 Oscillatory Solutions The fundamental oscillatory solutions are defined by the equation (1 - M2)En + Cm + e - 2M26 - M2Ea = 6(x)exp(iwt) , (2.2.12) associated to the equation (2.1.38). They will have, obviously, the shape E = E(x)exp(iwt), (2.2.13) where (1- M2)E;= + E,,,, + E.: - 2iwAf2E= + w2M2E = 6(x). (2.2.14) Performing the change of functions E - e: E = exp(Ax)e (2.2.15) and nullifying the coefficient of the derivative e=, one obtains the equar tion: (1 - M2)e", + eyy + e1t + ae = exp(-Ax)6(x), (2.2.16) FUNDAMENTAL SOLUTIONS OF THE EQUATION OF THE POTENTIAL 33 where A ikM, w2M2 a= 1-M2 (2.2.17) In the subsonic case (M < 1) one performs the change of variable (x, y, 0 (X, Y, Z): X =x, Y = fly, Z=Az (2.2.18) and one takes into account (A.3.11). One obtains the equations: cxx + cyy + e z + k2e = exp(-AX)8(X ), n = 3, e x x + eyy + k2e = ftuexp(-)LX)b(X), n = 2, (2.2.19) where k = WA1/02 = wM. In (2.2.19) we have Helmholtz's non-homogeneous equations. It is well known (see, for example, [A.12], §9), that Helmholtz's equation has two fundamental solutions in the three-dimensional case and the same number of solutions in the two-dimensional case, the choice of the solution depending on the type of oscillation defined by the equation. Taking into account that we obtained Helmholtz's equation looking for oscillations having the form (2.2.13), it follows (see, for example, [1.41) Chapter 7, §2) that the following solutions have a physical meaning: ea = exp(-ikIXI) 4i(XI 1 fr(2) e2 = 4 (kIXI) , 63 representing the solution for the three-dimensional case and 'e2 the solution for the two-dimensional case. Hoe) is Hankel's function. Performing the convolutions of these solutions with the right hand member from (2.2.19) (A.4.6 formula), we get the following fundamental solutions e2 = 4'-Ho2)(kjXI), e3 = so that, taking into account the changes already made, we find: E3 ik(Mx - R) , (2.2.20) E2 = -'blIt HQ(kR)exp (kMx) , with the notations R= a +f2(y2+z2), 17 = Vx2 ++Q2y2. 34 LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS In the supersonic case ; (Al > 1) one performs the change of variable X =x, Y=ky, Z=kz. (2.2.21) The equation (2.2.16) becomes exx - eyy - ezz + v2e = -exp(-AX) 6(X), (2.2.22) where v = wM/k2. In (2.2.22) we have the non-homogeneous KleinCordon-Fock equation. The fundamental solutions of this equation are also known (JA.8J [A.111). Performing the convolutions of these equations with the right hand member of (2.2.22), we get: Y Y X27tH ( e3 C2 y- -2+-Z Z2) cos X X. Z. V Z2 --y2) 2kH(X - IYI)Jo (vvX2 such that finally it follows - E E3 1 H `x - k Vy2 +zs) cosv x -k y +z) x_k ( kf.M2 .i.i)exp, 1-M=x) (2.2.23) (v 2L, H(x - IyI)Jo x2 - k2y2) exp (12X) Due to the presence of Heaviside's function in (2.2.23), it is obvious that these solutions will be different from zero only in the interior of Mach's cone with the vertex in the origin for x > 0, respectively in the interior of Mach's dihedron (with the edge on Oz) for x > 0. 2.2.3 Oscillatory Solutions for M = 1 If M = 1, the equation (2.2.14) becomes Ev + E..s - 2iwE= + w2E = 6(--). (2.2.24) Introducing the Fourier transform 2 with respect to the variables y and z one obtains A 2iwE=-(w2-az-a3)E_-6(x). FUNDAMENTAL SOLUTIONS OF THE EQUATION OF THE POTENTIAL 35 The solution of this equation (see (A.21)) has the shape E = H(x)E, where 2iwE. - (w2 - az It follows 0,3 ) E(o) E = O , = -1/2iw. E - - H(x) w2 2iwP - a3 - a3. [ (2.2.25) 2iw For obtaining E we take into account that L 0 exp (-au2)d u = (2.2.26) we notice that +00 exp (-iAa - Ba2)da = FOG exp -B (a + 2B l)a - da /// 42 I. One obtains: E 4irx) p (- i'd r2) , (2.2.27) with the notation r2 = z2 + y2 + z2. In the two-dimensional case we have: [__(x x 2+y2)], 11 (2.2.28) where 2o- 2Riw = 1. The solutions (2.2.27) and (2.2.28) may also be obtained as limits of the subsonic solutions (2.2.20) for M - 1. For obtaining (2.2.27), we notice that M2 2 jj(Mx - R) _ - lim M+ R = -U , limy (2.2.29) x being positive, as it follows from the supersonic solution (2.2.23). For obtaining (2.2.28) one takes into account the asymptotic behaviour of Ho2) for great values of the argument ((1.42)). HO(2(k1) t/2 (-h) exp (-ik R + i) (2.2.30) and one performs a calculcalculus similar to (2.2.29). In the two-dimensional case, the limit of the supersonic solution for M --+ 1 is obtained in (10.171. 36 LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS The Unsteady Solutions 2.2.4 In the sequel we are going to determine the fundamental solutions of the equation (2.2.14), i.e. the solutions of the equation (I - M2)e11 + £yy + E, - 2Af2£u - Af 2£tt = 6(t, x) . (2.2.31) This equation determines the perturbation produced in the uniform stream, defined in 2.1, by a source of potential acting at the moment t = 0 in the origin. The problem is plane (two-dimensional) if the source is uniformly distributed on the Oz axis. Applying the Fourier transform, we deduce M2d2£/d t2 - 2M2ial d E/d t + (a2 - A12c )£ (2.2.32) We know from Appendix A that the solution of this equation has the form £ = H(t)E, where H(t) is Heairiside's function and k, is a solution of the problem MV P/dt2 - 2M2ia1 dE/d t + (a2 - Al2a2 )E = 0, k(0) = 0 (dE/d t)(0) = -AI--2. (2.2.33) We get: A'IE = _sin J a jf- ti t e t°'t . (2 . 2 .34) Utilizing the formulas (A.7.14) and (A.7.15), we deduce £(t , x) = -- Irt 6(t-MR) , 1 2r where R= H(t - MR) n=3 , n = 2, (2.2.35) (2 . 2 . 36) t2 - M2R` (x-t)2+1,2+z2, R= (x-t)2+12. We shall not write any longer the factor H(t) from the right hand member, because obviously this member is different from zero only for t>0. Further we shall perform a detailed investigation of the solutions (2.2.35), (2.2.36). We denote h(t)=t-AIR. ;'.2.37) FUNDAMENTAL SOLUTIONS OF THE EQUATION OF THE POTENTIAL 37 We are interested to find the zeros t; of this function, as we are going to utilize the formula: (2.2.38) from [A.10], page 20. We have also to know the sign of the function h, because the two-dimensional solution differs from zero only for h > 0. One notices in (2.2.37) that the zeros t; are positive. Obviously. h(0) < 0 and h(oo) = (1 - M2)oo. For the graphic representation of the function h : (0, oo) - R, we have to separate the cases M < I and M > 1. The zeros of the function h(t) are tt- M2x+M x +(1(2.2.39) 1 112 In the subsonic case (Al < 1), h(t) has a single positive root namely h(t) t, t Fig. 2.2.2. t+. The graphic of the function h is represented in figure 2.2.2. Utilizing the formula (2.2.38) and taking into account that t+=Ai (x-t+)2+y2+z2, we deduce -- 1 t+ 4r t x b(t - t+) + (y` + x ) (2.2.40) , in the three-dimensional case and H(t - t°.) 1 27r t -A (2 , . 2 41) . 38 LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS in the two-dimensional case. Here we denoted to = t f(z = 0). The solution (2.2.40) was given for the first tine' in [2.11], and the solution (2.2.41) is given in [1.2]. A given point P (fig. 2.2.3) perceives in different manners the perturbation produced at the moment t = 0 in the origin of the system of coordinates in the three-dimensional case and the perturbation produced at the moment t = 0 uniformly on the O: axis, in the two-dimensional case. In the three-dimensional case the perturbation is perceived at the moment t+ and only at this moment (see the solution (2.2.40)). In the two-dimensional case, the perturbation is perceived by P continuously, beginning from the moment t° (see the solution (2.2.41)). The explanation of this difference is that in the case of the plane problem, one admits that, at the moment t = 0, the entire Oz axis emits perturbations, to representing the moment when the perturbation emitted by the origin reaches the point B(x, y, 0) and t > to , the period when the perturbations emitted at the moment t = 0 by the other points of the Oz axis reach P. In figure 2.2.3, Q is the position of the source (moving with the stream) at the moment to, the distance QP being given by the formula J(x - t.)2 + y2 = M'lt+ Fig. 2.2.3. In the supersonic case, we have to determine the zeros of the function h'(t). They are given by the equation (x-te)2+y2+z2 = M(to-x). (2.2.42) Noticing that x < to, it follows that there exists a single zero, namely to = x + y2 -+z2/k. (2.2.43) A simple calculation gives h(to) = x - k y2 -+z 2 . (2.2.44) FUNDAMENTAL SOLUTIONS OF TILE EQUATION OF THE POTENTIAL 39 We shall distinguish three cases (fig. 2.2.4): h(to) = 0, h(to) > 0, h(to) < 0. (2.2.45) Ni) t, 2 x>k(y+x) b) a) c) Fig. 2.2.4. In the first case the fundamental solution is £3 1 t+b(t - t+) + t_8(t - t_) t x- ky+ 41r 0, E2 _ _ 1 2;r 1 t -M (-x t) t<t+,t_<t t+ <t<t_, Tf P ' (2.2.46) , (2.2.47) Ea was given for the first time in (2.11]. One interprets this solution as follows: given a point P(x, y, z) in the interior of Mach's cone, there exist two and only two moments t+ and t_ when the perturbation produced in the origin at the moment t = 0 affects this point (fig. 2.2.5). The moments t+ and t_ are calculated as functions of the coordinates of P by means of the formula (2.2.39). In the two-dimensional case, P(x, y, 0) is affected during the interval It, to_) the explanation being similar to that given in the subsonic case. In this interval, the expression under the square root is positive. 40 LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS In the three-dimensional case, figure 2.2.5 presents the intersection of the z = 0 plane with Mach's cone and with the spheres having the centers Q1 and Q2. In the two-dimensional case the figure indicates the intersection of the same plane with Mach's dihedron and with the cylinders having the radii M- t9+ and Mtoi . Since from the relation sing = Q,P;/OQ; we deduce sinp = M-1 it follows that p is Mach's angle, i.e. the two spheres (cylinders) are inscribed into Mach's cone (dihedron). Fig. 2.2.5. If P belongs to Mach's cone, there exists only one moment t+ = = t_ = to = xM2/k-2, when the point P is affected by the perturbation. At that moment, the source is located in the intersection of the Ox axis with the normal at the cone in the point R. If P is in the exterior of the cone, it is not affected by the perturbation. Hence, for the unsteady flow, the perturbed zone is also in the interior of Mach's cone (respectively Mach's dihedron), but unlike the case of steady or oscillatory flow (when an arbitrary point P from the zone was affected all the time) in the unsteady motion it is affected only at two moments t+ and t_ (in the interval between these moments in the two-dimensional flow) or in a single moment if P is on the cone (dihedron). This happens because in the first two cases the fundamental solutions correspond to sources acting all the time in the origin (on the Oz axis), while in the last case the solutions correspond to sources acting only in the moment t = 0. FUNDAMENTAL SOLUTIONS OF THE EQUATION OF THE POTENTIAL 41 2.2.5 The Unsteady Solutions for M = 1 If M = 1, the solutions of the equation (2.2.31) will be (2.2.35) and (2.2.36) with M = 1. The function h(t) = t - R has an unique zero (it' > 0) namely the moment tj = r2/2x. It follows that the fundamental solutions E3 -_ 5 ( t8;rt x2 ) 2x ) 4rrx a (t - 2x (2.2.48) 1 H(2xt-x2-y2) 2n 2zt. _ x2 _ y2 vanish only for x > 0. Every point P from the half-space (half'-plane) x > 0 is affected by the perturbation. In the three-dimensional case, the perturbation emitted by the origin at the moment t = 0 is received in the point P(x > 0, y, z) at the moment ti, and only at this moment. If the perturbation is emitted by the Oz axis (the two-dimensional flow), an arbitrary point P(x > 0. y, 0) receives this perturbation continuously, beginning with the moment tj = (x2 + y2)/2x. Fig. 2.2.6. The moment of the first reception tj is given by the equation (x - ti), + y2 + z2 = ti. (2.2.49) The graphic of this equation is a sphere (a cylinder in the two-dimensional case) with the center on the Ox axis in the point tj and the radius 42 LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS t1 (figure 2.2.6). The point t1 is obtained graphically, taking the intersection of the Ox axis with the mediator plane of the segment OP (the mediator line of OP in the two-dimensional case). 2.2.6 The Fundamental Solutions for the Fluid at Rest As we have seen in subsection (2.1.5) , the propagation of a small perturbation in the fluid at rest, is described by the equation (2.1.50). The fundamental solutions of this equation will be defined by Eu-c2(E +Eyy+t )=6(t,z). (2.2.50) They will give the pressure produced by an unitary perturbation in the origin of the system of coordinates at the moment t = 0. Applying the Fourier transform in (2.2.50), we get: d2 ?/d t2 + c2a2E = d(t) . (2.2.51) Utilizing (A.3.17), we deduce that the solution of the equation (2.2.51) is E = H(t)E, where d2E/d t2 + c2a2E = 0, E(0) = 0, k(0) = 1. (2.2.52) Solving this problem, we find: sinclcxIt cla By virtue of formulas (A.7.14) and (A.7.15), we obtain: 6(ct - r) C3 = £2 = - 6(ct - r) 4zrc2t 1 2rc H(ct - r) 41rrr (2.2.53) c t- -r E1 representing the solution in the three-dimensional care and t2 the solution in the two-dimensional case. The solution E shows that the perturbation produced in the origin at the moment t = 0 is concentrated, for t > 0, on a spherical surface having the radius ct and the center in origin. The perturbation propagates as a spherical wave having the velocity c. On the sphere of radius r the perturbation is only at the moment t = r/c. We notice the validity of Huygens' principle. The FUNDAMENTAL SOLUTIONS OF THE EQUATION OF THE POTENTIAL 43 solution E2 shows that the perturbation emitted on the Oz axis at the moment t = 0 lies at a moment t > 0 in the interior of the cylinder of radius ct. There is a foregoing front of the wave propagating with the velocity c, but there is no posterior front. Unlike the three-dimensional case, behind the foregoing front, the perturbation differs from zero at every moment t. In this case Huygens' principle is not valid any longer. We know from 2.2.4 the explanation of this fact. 2.2.7 On the Interpretation of the Fundamental Solution The equation p+pdivv = pgb(x) (2.2.54) may be interpreted as the equation of continuity when there is a source with the intensity pq in the origin. Indeed, integrating on every domain D containing the origin and taking into account (A.7.3), we get: (p+pdivv)dv=pq. (2.2.55) ID The integral gives the variation of the mass from D in the unity of time. This is given by the intensity of the source. In every domain D which does not contain the origin, the mass is preserved. The presence of the term b(x) in (2.2.54) represents the cause of the motion. Since the term has a spherical symmetry, it follows that the flow will have this property too. Hence v = F(r)x/r and v = grad V with YP = f F(r)dr. An uniform flow, with the velocity U,,i is also potential, hence the flow resulting by overlapping the uniform flow over the flow due to a source is also potential. According to the calculus from subsection 1.2.2, it follows c20O - (V - V)(V V)¢ = 95(x) Setting q = eq, the equation may be linearized and for q = 1 one obtains (2.2.1). The solution of the equation (2.2.1) could therefore represent the perturbation produced into the uniform stream defined by M, by a mass source of intensity p, placed in the origin of the system of coordinates. In the same way, the equation (2.2.31) could be obtained from the equation A+ pdiv v = pgb(t, x), (2.2.56) which would represent the equation of continuity in case that a source of intensity pq is acting in the origin at the moment t = 0. 44 LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS 2.3 The Fundamental Solutions of the Steady System 2.3.1 The Significance of the Fundamental Solution In order to put into evidence the physical significance of the fundamental solutions of the steady linear system, we consider the equations: Af 2px + div v = 0, vx + grad p = F, (2.3.1) lim (p, v) = 0. As it is already known, these equations determine, in the first approximation. the perturbation produced into the uniform stream defined in 2.1 by a force density F. For being able to utilize these equations in the case of a force of intensity f = (fl, f2, f3), applied in a point C, we have to define a density F whose action against the fluid must have the same torsor (resultant and resultant moment) like the force f. We state that this density is F = f b(x - t), (2.3.2) where 6 is Dirac's distribution. Indeed, taking into account (A.7.3), we deduce that the torsor of this density is f fb(x - lr)dx = f, f x x fb(x - F)dx = E x f , (2.3.3) i.e. just the torsor of the force f. Hence the system M2p,, + div v = 0, v= + grad p = f b(x) lim (p, v) = 0, (2.3.4) x--.-oo determines the perturbation produced in the uniform stream, defined in 2.1, by the force of intensity f = E 7f, applied in the origin of the system of coordinates. By definition, this system determines the fundamental solutions of the steady linearized system of aerodynamics. One obtains the plane solutions if one considers that the force having the constant intensity f is uniformly distributed along the Oz axis, being parallel to the xOy plane, i.e. f = (fl, f2i 0). In this case we have the same conditions for every z; and the perturbation is plane. THE FUNDAMENTAL SOLUTIONS OF THE STEADY SYSTEM 45 The General Form of the Fundamental Solution 2.3.2 We are interested in those solutions of the system (2.3.4) which can be obtained by means of Fourier transform. They are, obviously, distributions. Utilizing the formulas (A.6.4), from (2.3.4) we deduce: A12alp+et v=0, ialy+iap=-f. (2.3.5) From (2.3.5) we deduce: P" = is f (2.3.6) a2 - M2a'I Then, from the second equation (2.3.5) it follows f ink (-ia)(-ia f) (2.3.7) iai(02-M2a1) Utilizing (A.6.9) and (2.2.3) from (2.3.6) we deduce: P(x, y, z) = -(f . V)F_i [a2 - A12ai = (f V)E. I (2.3.8) Taking into account (A.7.7), and (2.3.7) we obtain: v = fH(x)5(y,z) - V(f 0)S-' [ini(cr2 1 Af2c 2)J (2.3.9) But X [iaia a2 - A1201] = _ AI2a2)J E, (2.3.10) whence, integrating with respect to x, by virtue of Lebesgue's theorem [A.9}, it follows J.--`1 ial(a"- ' .1(2 " 1fI = f Ed x. (2.3.11) The integration limits have been appropriately imposed in order to satisfy the last condition from (2.3.4). It results therefore v(x, y, z) = f H(x)6(y, z) + VV, where r f o0 Ed x. (2.3.12) (2.3.13) 46 LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS The formula (2.3.12), which is valid both in the subsonic and supersonic cases, shows that the perturbation is potential, excepting the Ox axis for x > 0 (in the two-dimensional case, one excepts the xOz plane for x > 0), where the first term does not vanish. Hence the perturbation is not potential downstream the point (or the fine) where the perturbing force is acting. From (2.3.8) and (2.3.12) it also results: u(x, y, z) = ft H(x)ft, z) - p(x, y, z) . (2.3.14) With the exception mentioned above one obtains (2.1.38). In the sequel we shall utilize also some expressions of the components of the velocity which do not result from (2.3.12). In the two-dimensional case, performing the change az = a2 - M2a1 - (1 - M2)ai in the component v resulting from (2.3.7), we deduce: V=- a ia2f + (1 - M2) iaif2 a2-1b12a (2.3.15) whence (2.3.16) v(x, y) _ - ftEy + (1 - M2)f2El . Analogously, in the three-dimensional case, replacing a3 = Maa1 - (1 - M2)a1 a2, in the component w resulting from (2.3.7), - we obtain: w _ ioaf1 -a2-M2 1 ia1f3 +(1-M2 )a2-M2a1 (2.3.17) 1a2f3 ia2a3f2 crl(a2 M2a1) a1(a2 - M2at) ' - whence ,,2 w(x, y, z) = -f1Es + (1- M2)f3e2 - f2 ys Ed x + f382yy f Ed x , 'f-C00 00 (2.3.18) with the notations = 82/8y8z, .. . 2.3.3 The Subsonic Plane Solution If the perturbation is plane and the free flow is subsonic, E has the expression (2.2.7). From (2.3.8) one obtains P(x, y) - 1 xf1 + 132yf2 27x/3 x2 + /32y2 (2.3.19) 47 THE FUNDAMENTAL SOLUTIONS OF THE STEADY SYSTEM and from (2.3.16), 0x (2.3.20) - yf +g2y 2 If the force f is distributed on a straight line parallel to the Oz q), then the axis intersecting the xOy plane in the point 2nx2 v(x, y) = . perturbation will be determined by the system 12px+divv=0, v=+gradp= f6(x-t,y-q), (2.3.21) litn (p, v) = 0. After performing the change of variables (x, y) -, (xo, yo): xo=x-e,yo=y-17 , (2.3.22) the system (2.3.21) is transformed into the system (2.3.4). Hence, for (2.3.21) one obtains I P(x, y) = 2T'8 xofl + X02 02yof2 + p2ya , v(x, J) = Q x0f2 - yofl 27r XT o 2y02 - (2.3.23) These solutions have been obtained in [2.10]. 2.3.4 The Three-Dimensional Subsonic Solution In this case, E is determined by (2.2.6). From (2.3.8) one obtains 4ap(x, y, z) = --(fia: + flay + f30z)(1/R) , (2.3.24) where R= x2+f2(y2+z2). (2.3.25) Taking into account that _ dx J (x2 + a2)3/2 x a2(x2 + a2)1/2 (2.3.26) we deduce j'a()dX=_22 (2.3.27) 1 R dx -y+z2 48 LINEAR AERODYNAMICS. EUNDAMEN"TAL SOLUTIONS Employing these results, we deduce from (2.3.13) the expression of the potential cP(x'y,z) 1 fi _ = 4zr R (1+j)12-22 (1+)f3] y y2+z (2.3.28) which gives the possibility to calculate the components of the velocity. From (2.3.18) we also obtain: / 4-1r8,(I?)-4 fsa=[ w(x,y,z)= (2.3.29) - -,(f2a: - f3ay)y2 + z2 Cl +R . Considering A = 1, we find the solutions for the incompressible fluid. These results have been obtained in [2.8). 2.3.5 The Two-Dimensional Supersonic Solution In this case, E is given by (2.2.9). Taking into account (A.7.17), from (2.3.8) we deduce 2kp(x, y) _ (-f, + k f2sign y)a(x - klyl) (2.3.30) , and from (2.3.16), 2v(x, y) = (-f1sign y + kf2)6(x - klyl). (2.3.31) Obviously, the perturbation differs from zero only in Mach's dihedron. The solution was given and utilized in [2.10]. 2.3.6 The Three-Dimensional Supersonic Solution In this case, E is given by (2.2.8) and (2.3.32) P(x, y, z) = (flame + fear + fsa.)E . Denoting s = k y2 + z2 and taking into account (A.3.9) and (A.3.14), we deduce ay T Ed x = a H(x - s) o H(x - 8)(9y J rl dx Js I X7 9 2 z2E - = 11+ x 7X=F== R (2.3.33) f 49 THE FUNDAMENTAL SOLUTIONS OF THE STEADY SYSTEM such that, By means of formula (2.3.12) we determine the velocity field. The similarity between this potential and the subsonic potential is striking (2.3.28). Let us show now that both p given by (2.3.32) and the velocity field resulting from the potential (2.3.33), are different from zero only in the interior of Mach's cone. The assertion follows immediately if we use the formula d [H(x) dx xa )j = x(x .A) = -aH(+i , A # 0,1,2,... (2.3.35) demonstrated in the theory of distributions (see, for example, [A.5], §2.2). Indeed, we can write: 1 H(x - s) 1 (2.3.36) 27.- (x - s)1/2 (x + s)1/2 such that the derivatives of £ will have the factor H(x - s). From (2.3.18), it follows for w : (2.3.37) w(x, y, z) _ -(fl(9t + k2f3a=)£ + (f2as - f3ay)./2 +/z'£ In the sequel, we shall utilize also another expression of the component w. This follows writing W ia3f1 a2 - M2c, _ ia2a3f2 al (a2 - J%f 2a1) k2iai - ia2 oil (a2 - Af 2ai) f3 instead of (2.3.17). Utilizing (2.3.11), one obtains: Fx w(x, y, z) = f2t.1 £dx - f3(k28L - 01y) J z £d x. (2.3.38) 50 LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS The Fundamental Solutions of the Oscillatory System 2.4 2.4.1 The Determination of Pressure The fundamental oscillatory solutions are defused by the system: M2(pt + p=) + div v = 0 vt + yr + grad p = f d(x)exp (i wt) (2.4.1) lim (p, v) = 0. z-+-oo They will be complex. Since the system is linear, the real part of the solutions will correspond to the case when exp (i wt) will be replaced by coswt, and the imaginary part will correspond to the case when exp (i wt) will be replaced by sin wt. The solutions of the system (2.4.1) determine the perturbations produced in the uniform stream defined in 2.1, by a force having the periodic intensity f exp (i wt), applied into origin in the three-dimensional case and uniformly on the Oz axis in the two-dimensional case. Obviously, the solution of the system (2.4.1) has the form p = P(x)exp (iwt), v = V(x)exp(iwt) , (2.4.2) P and V satisfying the system AI2{iwP-}-Ps)+divV =0 iwV + V,, + grad P = f 8(x) (2.4.3) Urn (P, V) = 0. s-+-oo Applying the Fourier transform in (2.4.3), we get A12(w-as)P=a'V, (w--al)V-aP=-if, whence P _ ia- f M2(as - w)2 - a2' V _ if a ia- f as - w + as - w M2(as - w)2 -- a2 (2.4.4) Applying the Fourier transform, from equation (2.2.14) we deduce: E = -F-s [a2 - M2(as - w)2j , P = (f V)E. (2.4.5) THE FUNDAMENTAL SOLUTIONS OF THE OSCILLATORY SYSTEM 51 The pressure P will be expressed by means of the solutions (2.2.9) and (2.2.23). In the two-dimensional subsonic case, we shall have: P= -0(f V)[Ho2)(k1)exp(ikdlx)] , (2.4.6) and in the throe-dimensional subsonic case, P Q) {expJux - R)] (2.4.7) 4r(f k being given in (2.2.19) and R, R in (2.2.20). In the two-diinensional supersonic case, we have: P = -(f' V) [H(x - klyl)Jo (v x2 - k2y2)exp (-ivMx)] , (2.4.8) and in the three-dimensional supersonic case, 1 P (f 0)[H(x-Z: 21r y2 +z2) 1 cosy x--k2( +z7) exp(-ivMx)] \/X- 2 - k (y +:) (2.4.9) v being given in (2.2.22). Taking into account the formula (2.3), we deduce that the solution (2.4.9) differs from zero only in the interior of Mach's cone. The Determination of the Velocity Field 2.4.2 Since the velocity field V vanishes at -oo, we deduce like in (2.3.12): f _ i _ iai(a2-M'-a? +2alw+w2) - f 1 " Gdx, where G = -f- 1a2 1 - M2a1 + 2alw + w2 Using the change ai - ca -+ aI, we get: ( ] r Gdx. -i a 1 w NIz a lW)2 - a2] = -ie'""d 00 (2.4.10) (2.4.11) If we replace al by al - w in (A.7.9) we find ai - w ] = e-"`H(x)b(y, z) i (2.4.12) 52 LINEAR AERODYNAMICS. FUNDAMENTAL. SOLUTIONS Utilizing (A.6.9), on the basis of formulas (2.4.11) and (2.4.12), from (2.4.4) we deduce: V = fe-k"H(x)b(y,z) + Vcp, (2.4.13) where _ -(f 0)e-k'1X / G dx. (2.4.14) f oc We notice that the perturbation is potential, excepting the Ox axis for x > 0, i.e. excepting the trace of the source in the uniform stream. Let us determine now the distribution G. We notice, applying the Fourier transform, that it is the solution of the equation (1 - M2)Gu + G - 2iwGx - w2G = 6(x). (2.4.15) The solution of this equation may be obtained like in 2.2. So, in the two-dimensional subsonic case we find G = _ ZI-0 Hoe) (k1) exp (iox) , (2.4.16) and in the three-dimensional case G 41rR exp [ik(_R)] (2.4.17) . In the two-dimensional supersonic case one obtains- G = -- 2k H(x - klyl)Jo (ui/2 - k2y2) exp (lJx) = H(x - kEyl)g, not (2.4.18) and in the three-dimensional case G = - 2 H (x - k y2 + --2) x= - k (y' + z-) cos v X- r - k-(y- + -) exp (izux) - not H(x-k. Vy2 -+z2) g, (2.4.19) w being defined in (2.2.19). Taking into account the behaviour of Ho2) for great values of the argument (2.2.30), we deduce that in all cases we have: lim (G, G=) = 0. s----oo (2.4.20) THE FUNDAMENTAL SOLUTIONS OF THE OSCILLATORY SYSTEM 53 2.4.3 Other Forms of the Components V and 14 In the two-dimensional case from the component V given by (2.4.4) one eliminates a2 by means of the identity a2=a2-M112(al -w)2+M2(a1-w)1-ai. Thus one obtains if, ai -W ft, a2 h12(a1-W)2-a2 + if2 (M2 -1)a, - 2M2wa1- M2w2 a, -w M2(a1 - w)2 - a2 whence, utilizing the inversion formulas (A.6.9) and (2.4.11), V (X, y) {f, , + f2 [(M2 -1)O + 2M2iwt7, - M2 W2] } (e-iWX X J Gdx = -f,e';v,x Gy - iw 00 J m -fee-i4" llI2iwG - (I - M2)Gx + w2 J X G d xl . J oo (2.4.21) We also obtain this form using the expression of V given in (2.4.13) and (2.4.14), V= `H(x)6(y, z) - f14 f/ X x a-wx J \ G d x) - 00 Gyy, dx, ;aid eliminating Gy, with the aid of the equation (2.4.15). Indeed, in the two-dimensional case, utilizing (2.4.20), we deduce G., d x = H(x)d(y) - (1- it f2)Gx + 2iwG + w2 J Gdx . From the last taro formulas we find again (2.4.21). For the supersonic solution, taking into account (2.4.33), with the definition of g given in (2.4.18) we have s r - 8y Gdx=H(x-kIyI) r gdx, f Ivl r r (x Gdx=H(x-kjyj)OOJ x gdx, J oo lyi (2.4.22) 54 LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS such that V given by (2.4.21) is zero outside Mach's dihedron. In a similar manner, in the three-dimensional case we deduce for the component W W (x, y, z) = - f ie " (G c - iw rx G$ d x) J oo - f2e'11x foo G,, d xx -he-i- 2"-(1-M2)Gx+(w2-822y) x Gdx, 00 (2.4.23) We notice that using the notation: s=k y 2 ++z2 (2.4.24) we may write in the supersonic case W (x, y, z) = H(x - s)w(x, y, z) , (jX) (2.4.25) where taking into account the definition of g from (2.4.19), w = -fie ir'" k.ix (. - f3e-u.,x [2iwg+k2g+(,2_ay)jXgdxj whence we deduce that W is different from zero only in the interior of Mach's cone with the vertex in the origin. Indeed, utilizing the formulas (A.3.9), (2.3.35) and noticing that s8/8y = k2y8/8s, we may write : 82yy /x Gdx=822 [H(x-s) JT gdxJ = 00 = [H(z - s)8(k2y/s)8,j gd= H(x - s)8J g d x and from GdxJ =H(x-S) {g=-iw f g2dx) 8s\\`G-iwJ 00 it follows the formula (2.4.25). .J. 55 THE FUNDAMENTAL SOLUTIONS OF THE OSCILLATORY SYSTEM 2.4.4 The Incompressible Fluid The (oscillatory) solutions for the incompressible fluid may be obtained considering Al = 0 in (2.4.4), or directly in the subsonic solution. So, in the two-dimensional case we deduce from (2.4.5): P(x, Y) = I xf! + 02 (2 . 4 . 26 ) 2 The general form of the equation (2.4.15) for M = 0 is G = g exp (iwx), (2.4.27) where Og = 0. One obtains therefore: G(x, y) = exr (i x) ln(x2 + y2) . (2.4.28) For the three-dimensional problem we have: (f V) T , P(x, y, z) with the notation r = 2.4.5 G(x, y, z) ex (iwx) (2.4.29) x2 + y2 +-2. The Fundamental Solutions In the Case M = 1 If Al = 1, E is the solution of the equation (2.4.23), and G, the solution of the equation Gyy + C_. - 2iwG,. - w2G = 8(z) . (2.4.30) C is determined like E. They have the form: E = H(x)e, G = H(a)g. (2.4.31) In the two-dimensional case. e and g have the expressions: e=- a .exp (x+ T \ g=- 2 exp )J , l2 (x- (2.4.32) with the notation utilized in (2.4.26) for a. The solution given by (2.4.5) and (2.4.21) is P(x, Y) = H(x)(fies + f2eg) (2.4.33) V (x, y) = H(x)v(x, y), 56 LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS where v(x, y) = - f le-iwx [gy - iw J xgy (r, y) d r] - Q (2.4.34) [2ig + w2 f2C- g(r, y) d TJ Ix J0 In the three-dimensional case, utilizing (2.2.6) we deduce: p-( 1 C = - ----- ex 4irx iw 2x t2 _ 1 J , -1 4'rr 2 2 x- y+Z iw exp x 2 (2.4.35) Taking into account (2.4.5) and (2.4.23), it follows P(x, y, z) = H(x)(fie. + f2e,, + f3e=) (2.4.36) W(x, y, z) = H(x)w(x, y, z), where -fie-Lox w(x, y, z) = rgz - iw jx1 ''j x { - fee gy d x- // - f3e-twx [2iwg + (w2 - ayy) x g dx . J Q The integrals from (2.4.37) have a strong singularity in the origin, but they are convergent, as it results from the following calculus indicated by V.Iftiinic: ! /Oz gd r fJf 47r bin , 4T hm e\O r -exp !02 r I i (br ` fT r exp { i (6T F lL r= exp (ibr) - 1 exp 2l) l d r -2\1Jff T! f J dT 1 (-i- I d r + EimQ fJ T exp f\-- i idt. The first limit exists because exp (-ice /r) is a bounded function. In the second integral we perform the change of variable r = 1/t and we integrate by parts. We obtain: inn J T ce/z (_!)dr=j exp (-iC'-t) jt = 57 FUNDAMENTAL SOLUTIONS OF THE UNSTEADY SYSTEM 1 x iC2 = 2 exp (x ) f 0c 2 dt exp (-ic t) t2 . ~ ir2 The last integral is obviously convergent. The solution (2.4.33) was given in {10.20) and the solution (2.4.36), in [10.17]. They can be also obtained as limits of the subsonic solutions. Indeed, for P given by (2.4.6) and (2.4.7) one utilizes (2.4.27) and (2.4.28). A similar calculus may be performed for G given by (2.4.16) and (2.4.17). Passing to the limit for M \ 1, we get g given by (2.4.32) respectively g given by (2.4.35). This section was written entirely on the basis of the papers [10.17]-[10.20]. Fundamental Solutions of the Unsteady System I 2.5 2.5.1 Fundamental Solutions In this section we determine the fundamental solution of the system (2.1.51), i.e. the solution of the system pt+c2divv=0, vt+gradp= f6(t.x) (2.5.1) liim(p,v)=0. 00 We already know that this system determines the perturbation produced in the fluid at rest, having the density po and the pressure po, by a force having the intensity f, applied instantaneously (at the moment t = 0) in the origin of the system of coordinates (on the Oz axis in the case of the two-dimensional problem). Utilizing Duhamel's principle, we deduce that the solution of the system (2.5.1) has the form: (p, v) = H(t)(P, V) , (2.5.2) where H(t) is Heaviside's function and (P, V) is a solution of the system Pt+c2divV =0, P(0, x) = 0, lim(P, V) 00 Vt+gradP=0 V (0, a) = f 6(x) (2.5.3) 0. Applying the Fourier transform, we deduce the system (2.5.4) 58 LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS which has to be integrated with the conditions P(0) = 0, V (O) = f . (2.5.5) The solution of this system is f)(1-coscla(tl P=cia. f5inlcall It, V = f+ia(ia J 2 , such that, by inversion, we have: P(t,x) _ -c(f . V).F-i Csinclaltl . ) 1C O (2.5.6) (1- Cos cla It 1 V (t' x) = fo(x) + V (f V ).F`' C,,2 . Utilizing the formulas (A.7.14) and (A.7.22), with the notation r = lxl, we deduce the following fundamental solution for the three-dimensional problem: P(t,x)=-(f.V)d(ct-r) 41rt (2.5.7) V(t,x) = f6(x) + V(f V) H(ct - r) 47rr and for the two-dimensional problem, P(t, x) = - c H(ct - r) (f V) ct2-r' 21r (2.5.8) r) Inct+ VCr t -r V(t,x)=f6(x)+V(f These are the solutions of the system (2.5.3). 2.5.2 Fundamental Matrices The solution of the system (2.5.1) has been determined for the first time by means of matrices in (2.6]. As we shall see, in some applications it is preferable to use fundamental matrices. We introduce therefore the matrices: 0c200 00c20 000c2 0000 1000 ] 0000 Q= 0000 ,R- 1000 ,S= 0000 1000 0000 0000 , (2.5.9) FUNDAMENTAL SOLUTIONS OF THE UNSTEADY SYSTEM 1 VT = fir, U, V, W1, FT 59 = (0. f1, f2, f3J, u, v, w representing the components of the vector v and fl, f2, f3, the components of the vector f. The system (2.5.1) is written as follows: Vt+QV=+RV,+SVL =F6(t,x) (2.5.10) V(t, oo) = 0. Introducing the matrix L. such that V .- LF, (2.5.11) we deduce: Le + QL,, + RLy + SL,. = E6(t, x) (2.5.12) L(t, oo) = 0, E representing the unit matrix with 4 x 4 elements. Utilizing Duhamel's principle, we deduce that the solution of the equation (2.5.12) has the form L = H(t) K(t, x), where (2.5.13) Kt+QKx+RKy+SKs=0 (2.5.14) K(0, x) = EJ(x) , K(t, oo) = 0 . Applying the Fourier transform to the problem (2.5.14), we deduce: k,=.U, k(0,a) = E, (2.5.15) where 0 A=icr1Q+i02R+ia3S=i C2a1 C202 C2a3 01 0 0 0 a2 0 0 0 0 a3 0 0 The solution of the problem (2.5.15) has the form: K = Eexp(At). (2.5.16) It is well known that for determining the function exp (At) one may employ two classical methods: the method of matrix functions and the LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS 60 method of the minimal polynomial. We shall utilize herein the method of the minimal polynomial described in 2.6.2. The eigenvalues of the matrix A are Al = 0, \2,3 = ±i laic, the first one being multiple of order two. One deduces that the minimal polynomial has the form a(A - a2)(-\ - a3) = a(a2 + a2c'2) . Comparing this form with (2.6.13), we obtain (?p=a2=0, ai =-a2c2 Hence, the equations (2.6.18), (2.6.1.9) and (2.6.17) become - 9oft 2c2921 go = 0, 92(0) = 9'2(0) = 0, !f2(0) go (0) = 1, 9i = -c12c292 + 90, 91(0) = 0 . K has the form Ego + Agl + A-2g2. We get: k = E+ Asmclalt + A21 - COSCktIt C202 Clal Using the inverse Fourier transform we obtain: K(t,x)=E6(x)+AY-1 I sinclalt+A2.._t 1 -c:asclaltclal c2a2 J where, for the three-dimensional case, 0 A C2ar Clay c2as ox ay 0 0 0 0 0 0 a_ 0 0 0 Utilizing the formulas (A.7.14) and (A.7.22), we deduce that the matrix K has the form: H(ct - r) (2.5.17) K(t,x) = Eb(x) + 4tAb(ct - r) + 4ac2A2 71 In the two-dimensional case we have: K(t, x) = Eb(x) + + 1 27rc2 1 21rc A H(ct - r) + c t - r' VC-It- A2 H(ct-r)Inct+ r -r (2.5.18) FUNDAMENTAL SOLUTIONS OF THE UNSTEADY SYSTEM 1 A=- 61 0 CZC7z c28y 0 0 ax a, 0 0 The formulas (2.5.13), (2.5.17) and (2.5.18) give the fundamental matrices. 2.5.3 Cauchy's Problem We shall prove in the sequel that the solution of the problem Vt + QVz + RVt, + SV. = F(t, x) V(0, x) = 0, (2.5.19) V(t, oo) = 0 (2.5.20) is c V(t,x)=1 K*Fdt, (2.5.21) 0 where K*F= f 3K(t-r,x-4)F(r,4)d4=f K(r,4)F(t-r,x-4)d4 is a convolution. Indeed, taking into account (2.5.14), we have: Vt=K*FLt + t f K,*Fdr, 0 t QVz+RVv+SVs=J (QKx+RKv+SK=)*Fdt, 0 V +QVs+RVy+SVz F(t,x). fR3 The conditions (2.5.20) are obviously verified, because K(t, oo) = 0. Taking into account (2.5.17) and (2.5.18), with the notation p = _ fit', it follows for the solution of the problem (2.5.19), in the three- 62 LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS dimensional case, V(t,x) .fu F(-r,x)d-r + 4c A r c Jr3 F(t- -r, x - t)6(cr - p) d t + , A2 I d r fa F (t- irC" -r,x - ls) 3 H(cr-P)d P (2.5.22) and in the two-dimensional case, r1 V (t, x) = foF(r,x)dr+2xc AI u -T,x~4) H(crr _ I dTI2F(te p) d4+2 VR-T P2 A2 p2dt. -r,x-E)H(cr-P)III Cr+ P (2.5.23) 2.5.4 The Perturbation Produced by a Mobile Source We shall apply the above formulas for determining the perturbation produced in the fluid at rest (having the density po and the pressure po) by a force of intensity f . whose application point moves uniformly, with the velocity v, in the direction of the Ox axis and in the negative sense. This problem, also considered in [2.6), is important in aerodynamics because it models the leading edge of an uniformly moving airplane in the air at rest. The solution is obtained from (2.5.22) replacing F(t, x) = Fo6(x + vt)6(y, x), (2.5.24) 63 FUNDAMENTAL SOLUTIONS OF THE UNSTEADY SYSTEM 1 where Fo' = (0, fl, f2, f3). One finds V(t,x)=FDH(x,+vt)o(y,a)+4I R,) AFD.I D t l A2Fo f +zirc- H( dr+ b(cr _ RT) d (2.5.25) where R, _ x + vt - vr)2 + y2 + z2 . (2.5.26) For calculating these integrals we have to study the behaviour of the function h(r) = c-r - R,r , (2.5.27) on the interval (0, oo). We shall proceed like in 2.2.4. Because h(O) < 0 and h(oo) = (c2 - v2)oo. we have to separate the subsonic flow case (v < c) from the supersonic flow case (v > c). The zeros of the function h(r) are given by the formula (c2 where - v2)r+ = -v(x + vt) ± R, R = [(x + vt)2c2 + (c2 - v2)(y2 + z2), 1/2 (2.5.28) . (2.5.29) In the subsonic case, only r+ is positive. If the velocity of the source is supersonic (v > c), the function h(r) has two zeros (r+ < 'r-) if /to > 0 and no zero if !to < 0. We denoted ho = h(ro), ro representing the zero of the derivative of the function h, vro=x+vt+ cr0 Vv- (2.5.30) It results, vho = c(x + vt) - (v2 - c2)(y2 + z2). (2.5.31) Utilizing the relation crfh'(r ) = ±R which may be verified directly, and the formula (2.2.38), we deduce f ' 6(cr - R') T 0 dr = cR-1 H(t -,r+), if v < c, cR-1JH(t-r+)+H(t-r_)J, if v>candho>0, 10, if v > c and ho < 0. (2.5.32) 64 LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS In the same time, taking into account the definition of Heaviside's, function we have: JtH(CRr) dRr v - v In c(x + vt) c(x + vt) + R R, if(v<c)U(v>c,ho>0,t<r_), if v > c, ho > 0, t >T-, -;- if v>c,ho<0. 0, (2.5.33) Now the solution (2.5.25) is explicit. The solution of this problem may be also obtained directly (without using the formula (2.5.22)) [1.10]. The interpretation of the solution is similar to the interpretation from (2.2.4). Fundamental Solutions of the Unsteady System II 2.6 2.6.1 The Fundamental Matrices In this section we determine, using the papers [2.71 and [2.9], the fundamental solutions of the system (2.1.26), (2.5.27), i.e. the solutions of the system M2(pt+p=)+divv=0 vi + v;r + grad p = f a(t, x) (2.6.1) limp, v) = 0. 00 Introducing the matrices Q= 1 1 0 0 0 0 00 10 ' 01 000M-2 00x1'20 1 M'200 R= 00 0 10 0 00 0 0 0 0 _ ' S 000 0 000 0 100 0 and the matrices V and F defined in (2.5.9), the system (2.6.1) may be written like in (2.5.10). We shall proceed like in the previous section. Considering V = H(t)K(t, x)F, (2.6.2) 65 FUNDAMENTAL SOLUTIONS OF THE UNSTEADY SYSTEM II one obtains the following equation for k : Kt = AK, where K(0, a) = E, (2.6.3) al a1M-2 02M -2 a3M-2 al A=i a2 CM 1 0 0 0 0 1 0 0 1 The roots of the characteristic polynomial are: Al = ial, A2,3 = i (al ± lalM-1), (2.6.4) the first root being double. One verifies that the minimal polynomial has the form (A-A1)(A-A2)(A-,13) = (A-ial)3+a21tf-2(a-ia1). Comparing with (2.6.13) it results: a2 = 3i a1, al = 3a2 - 02M-2, ao = i a1(a2M-2 - a2). (2.6.5) k has the form K = Ego(t) + Ag1(t) + A2g2(t) , (2.6.6) the functions 90, 91, 92 being determined by the equations gg2 - a292 - a19o2 - aog2 = 0, 92(0) = g2(0) = 0, 92(0) = 1, (2.6.7) goo = ao92, 90(0) = 1, 901 = a1.92 + go, 91(0) = 0. Determining these functions, .%v find: go = I 1 - ioM lal t la) sin M M2 - a?a2 / I 1 - o0c1 lalt`1 M JJ / m sin lal t - 2ia?M2 a2 1 -cos lal t J 91 = 9'2 (al 1 exp(ialt). = a/ - aft _ M M exp(i alt), exP(i a1 t), 1112 1 (2.6.8) LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS 66 In the three-dimensional case one utilizes the formulas (A.7.15) and (A.7.23). From (2.6.6) it follows the fundamental matrix: K(t, x) = Eb(2- t)b(y, z) + 4 + (E t R)+ (EOY + +2A8,.+A')11(tA1R -R} (2.6.9) where R = V(2--- t)2 + YT-A- z2 and A = - AI'28y M 2(9: 8r A1'2O 0. a2 0 8= 0 0 0 0 Sr 0 ay In the two-dimensional case one utilizes the formulas (A.7.15) and (A.7.23). From (2.6.6) one obtains the following fundamental matrix: A12 K(t, x) = Eb(x - t, y) + -(EO + A) 2" + where R= ( A122 2-r H(t -- AM) jt72 + A12R2 (O_ + 2A87 + A2}H(t - AIR) In t i +_ t2 - A12 nip (2.6.10) - t)- y2 and 8? A AI'28,. M-28y 8= 0.1 0 8y 0 8t lit [1.10] and [2.7] one gives some applications which are important for aerodynamics. We mention the mobile source on the direction of the unperturbed stream and the mobile source on the direction perpendicular to the same stream. 2.6.2 The Method of the Minimal Polynomial In the sequel we present the method utilized for determining the function exp (At). Let A be a matrix with n x n elements and let At, ... , Ak(k < n) be its eigenvalues, i.e. the roots of the characteristic polynomial A(A) = det (AE - A) . FUNDAMENTAL SOLUTIONS OF TIIE UNSTEADY SYSTEM II 67 Obviously, 0(a) = (A - A1)m1 ... ()1 - A )mk , (2.6.11) where ml + ... + ink = n. The minimal polynomial associated to the matrix A is by definition the polynomial with the smallest degree which has the property P(A) = 0. (2.6.12) Let -... - a1,\ - ao P(A) = am be this polynomial. Obviously, m < n, because .(A) = 0. am-la,n-I (2.6.13) From (2.6.12) and (2.6.13) it results m-1 Am= EajA'' j=0 and step by step m-1 Am+p = E aln) A3 , p = 0, 1, 2, .. . jO Taking into account that the exponential is a power series, we deduce that there exist the functions go(t), ... , g.-1(t) . such that exp(At) = goE+g1A+...+gm_1A"-1 (2.6.14) We have: A exp (At) = goA +... + +9,n-1(ao E + al A + ... + a,,,-1 Am-1 ) , gm_2A'n-1+ ( 2. 6 .15) On the other hand, A exp(At) = (d/(It) exp(At) = g(E+g'A+...+g',,,-lAm-1. (2.6.16) Since P(A) is unique, from (2.6.15) and (2.6.16) we obtain: 90=ao9m-1791 =a19+n-1+go,...,9in_1 =a+n-19m-1+9m-2 (2.6.17) 68 LINEAR AERODYNAMICS. FUNDAMENTAL SOLUTIONS We deduce that 9m_1 is the solution of the differential equation 1rn?1 = am-1g(n1 1' -#- ... + olgVn- 1 + (109,n-1 , (2.6.18) with the following initial conditions: g,, 1(0) = girl -1(0) = ... = girl1(0) _ 0, g (in-1) 1' (0) = 1 . (2.6.19) The functions go, ... , gin-2 are determined from (2.6.17) with the cu conditions 90(0) = 1,91(0) = 0,---,9,,.-2(0) = 0- (2.6.20) imposed by (2.6.14). In fact, for determining the minimal polynomial, one takes into ac- count that it is a divisor of the characteristic polynomial, hence it has the form: P(A) = (A - A0," (A - A2)u'2 ... (A - Ak)' = = A"-on,-1Arn-1 -...-ao, (2.6.21) nj < mj and n1 + n2 + ... + nk = in. To conclude, we try with divisors having the form (2.6.21), beginning with the simplest where (ill = n2 = ... = nk = 1) in order to satisfy the equation P(A) = 0. from the University of Bucharest We mention that dr. *t. has communicated us this method. Chapter 3 The Infinite Span Airfoil in Subsonic Flow The Airfoil in the Unlimited Fluid 3.1 3.1.1 The Statement of the Problem In this subsection we determine the perturbation produced by a thin infinite cylindrical body in a subsonic uniform stream having the velocity UQ,i, the pressure pQ, and the density p,,. We assume that the generators of the cylinder are perpendicular on the velocity of the unperturbed stream which is parallel to the xOy reference plane. The cylindrical bodies are bodies with constant cross-section and the infinite cylinders are cylinders which are long enough, such that the effects of the end conditions over the flow in the rOy plane may be neglected. Under these assumptions ( the conditions which determine the perturbation are not varying in time and they are the same in every plane parallel to xOy), the perturbed flows of the fluid will be stationary and plane. Ua Y Fig. 3.1.1. % Te utilize the variables r , y, z introduced in (2.1.1) and the per- turbations p and v defined by (2.1.3). In figure 3.1.1 one presents the profile determined by the intersection of the cylindric body with the xOy plane. We denote by y = la(x) ± hl(x) (3.1.1) the equations of the curves C+ (the upper surface) and C_ (the 70 THE INFINITE SPAN AIRFOIL IN SUBSONIC FLOW lower surface). The curve having the equation y = h(z) gives the so called skeleton of the profile. For imposing the boundary conditions we use the projection of the profile onto the Ox axis (the direction of the unperturbed stream). Defining Lo from (2.1.1) as the half length of this projection and taking the origin of the axes of coordinates in the middle of the projection, it follows that the functions h(x) and h1(x) have to be defined on the segment [-1,+1]. We assume that these functions have continuous first order derivatives and that the derivative h(x) satisfies Holder's condition (see Appendix B). The cylinder is thin if h(x) = Fh(x),h1(x) = Eh1(x) e representing a small parameter. We notice that whatever would be the equations of the curves C+ and C_, they may be written like in (3.1.1). Indeed, if the equations would be y = h±(x) = ehf(x), they might take the form (3.1.1) putting h+(x) + h_ (x) = 2h(x), h+(x) - h- (x) = 2h1(x). Taking into account the equations (3.1.1) the following conditions have to be satisfied v(x, ±0) = h(x) ± 14 (x), Ix I < 1 (3.1.2) as it follows from (2.1.32). 3.1.2 A Classical Method The classical methods for solving this problem rely on the assumption that the perturbation is potential. We have therefore v = grad Sp(x, y). (3.1.3) The potential p which is defined in the exterior of the segment [-1, +11 from the xOy plane, is the solution of the equation (3.1.4) P"P:: +'p1,s, = 0 with the boundary conditions jpi,(x, f0) = h'(x) ± h4 (x) , JxJ< 1 (3.1.5) and the condition at infinity z lim00(P_, spy) = 0 (3.1.6) 71 THE AIRFOIL IN THE UNLIMITED FLUID which follows from (2.1.28). Using Clauert's transformation x, y -, X, Y: X=x, Y=fiy, (3.1.7) one obtains from (3.1.3) and (3.1.4) u=six -- U, v =A(PY=QV sOxx +'PYY = 0 (3.1.8) (3.1.9) in the exterior of the segment [-1,+11 from the XOY plane. The boundary conditions (3.1.5) and the conditions at infinity (3.1.6) become V (X, fO) = h'(X) f hi (X), 1X 1 < 1 lira (U,V)=0. X -»- cc (3.1.10) The problem was solved by many authors (see for example Birnbaum [3.51, [3.61, Sohngen [A.34), Sedov [1.38], p.47, Iacob [1.21), p.661), by means of one of the following two methods: reduction of the problem to an ntegral equation which is then solved or the reduction of the problem to a boundary value problem in the complex plane. We also mention the methods relying on the expansion of the solution in a Fburier series. (Glauert [3.211 and Atei. inger (3.491). In the sequel we give one of the most natural methods for reducing the problem to an integral equation. We represent the holomorphic function W = U - iV in the Z = X + iY plane with the cut [-1, +1] on the real xis, by means of the Cauchy integral W(Z)=-L r -f(t)+I t(t)dt, (3.1.11) where f(t) and fl(t) are unknown real Holder functions. It is easy to understand the significance of this representation. W(Z) Is the sum of the complex velocity determined by a continuous superposition of squrm having the intensity fm on the segment [-1, +1) and the complex velocity determined by a continuous superposition of vortices having the circulation f, on the same segment. In fact, the integral containing f may also be regarded as a continuous superposition of doublets on the segment [-1, +1] from the (Z) plane. Obviously; W(Z) is a holomorphic function in the exterior of the segment [-1, +11 and it vanishes for Z -+ oo. Hence, U and V THE INFINITE SPAN AIRFOIL IN SUBSONIC FLOW 72 satisfy the equation (3.1.9) and the condition (3.1.10). In order to prove this we pass to the limit considering Z - X ;k i 0, where X E (-1, + 1). Using Pleme(j's formulas (B.3.1) and separating the real part from the imaginary one we obtain: 2f(X)+2- / U(X,±0)= V(X,±0) = 2ft(X) - t 2zr, it()dt (3.1.12) tfit) dt. (3.1.13) The mark"' " indicates the principal value in Cauchy's sense (Appendix B). From (2.1.38), (3.1.8) and (3.1.12) it follows p(x, +0) - p(x, -0) = U(X, -0) - U(X, +0) = f (x). (3.1.14) This relation puts into evidence the significance of the function f (x). It will be fundamental for the determination of the lift and moment coefficients. Imposing the conditions (3.1.10) in (3.1.13), adding and subtracting the relations just obtained, we find Q T / tdt _ = -2h' ( X) JX i < 1 , Oft(X) = -2h4(X) JXi < 1. (3 . 1 . 15) (3.1.16) The relation (3.1.16) determines the function Of, and (3.1.15) represents an integral singular equation for the determination of the function Of. It is the well known equation of thin profiles which was put into evidence by Birnbaum in 1923 and solved by Sohngen in 1939. Its solution, given in Appendix C, will be written in (3.1.25). We notice that the thickness of the profile is ensured by the distribution of sources (ht = 0 ft = 0) and the discontinuity of the pressure (hence the lift and moment coefficients) is ensured by the distribution of vortices. 3.1.3 The Fundamental Solutions Method There is no physical justification for replacing the wing by a continuous distribution of sources and vortices. It is more natural to replace the wing by a continuous distribution of forces and to determine the 73 THE AIRFOIL IN THE UNLIMITED FLUID intensity of these forces, in order to obtain the same action against the fluid like in the case of the wing. Indeed, the fluid and the wing are two interacting systems. If we want to determine the motion of the first subsystem, we have to replace the action of the second subsystem by a distribution of forces. This is the basic idea of the fundamental solutions method. We shall replace the wing by a continuous distribution of forces f = (fl, f) on the segment [+1, -1], having the intensity f a priori unknown (the intensity will depend on the point). In fact, for obtaining a plane perturbation, the forces have to be uniformly applied on parallels to the Oz axis, intersecting the xOy plane in the points of the segment [-1,+11. Taking into account the formulas (2.3.23), it will result the following perturbation in the fluid: p(x,y) = (x, v) _ 1 27r#I f +1 xof1() X02 + #2y2 (3.1.17) /+t xof(t) -yfi(E)d 21r ./ 1 xo + 82y2 In view of imposing the boundary conditions (3.1.2) we have to calculate ylimo v(x, y) for jxj < 1. Denoting by v(x, ±0) these limits, we have u(x, f0) = Q lim 2ir y--±o f+1 xof W 1 (3.1.18) xo + 02y2 The first integral represents the tangential derivative of a simple layer potential and the second the normal derivative of the same potential. Although the formulas which give the limits of these derivatives are known, we prefer to calculate them directly. If we pass to the limit in the first integral making y = 0, we cannot simplify with x0 because for = x it is zero. We shall divide this integral in three parts. Since for a e small enough, we may approximate f on the segment (x - e, x + E) with f (x), we deduce performing the substitution - x = u, :+C xoM) _ #2y2 +` ud u I-C U2 +Qty 2 74 THE INFINITE SPAN AIRFOIL IN SUBSONIC FLOW the integrand being an odd function. It remains +1 lim xof W d+ r:- E f d- -lim E-.0 l J ( V-±0 J1 xo2+ 1 zo 1 r1 f(_) Ji x-4 f d J = d excepting zero, The second limit from (3.1.18) is f1 Jx.c xo the (3.1.19) interval (x - s, x + e) where the integrand is infinite for y = 0 and f = x. Approximating again f j (C) by f, (x), or applying the mean formula, since f i is integrable and multiplied by an integrable factor having a constant sign, we get: Yli m J lm Y 114 +a fi(x)vuy f, + uZ v(x, 0) °J y -Z0' f ±f1(x) - d (_) Imposing the boundary conditions (3.1.2), adding and subtracting, we get: Q f (4) d _ -2h'(x), (3.1.21) (x). (3.1.22) fi(x) _ We also have P(x, fU) = fi(e f2f(x) + 2WQ (3.1.23) such that f (x) has the same significance like in (3.1.14). The sntegral equations (3.1.15) and (3.1.21) coincide. In fact, the representation (3.1.11) may also be obtained from the fundamental solution. This representation is valid for the points which do not belong to the semi-axis Ox (x > 1). Indeed in these points we have u = -p, such that u- v= 1 1 2 i -f (S) + i P J_i -Z .11(x) df ' (3.1.24) 75 THE AIRFOIL IN THE UNLIMITED FLUID where Z = x + i fly. We see from (3.1.16) and (3.1.22) that instead of Of, from (3.1.11) we put here fl. The equivalence just established shows that the formula (3.1.11) is not valid on Ox (x > 1). On this half-line the flow is not irrotational. We mention finally that the proof of Plemelj's formulas (which are necessary for obtaining the integral equation (3.1.15)) is anyway more difficult than the proof of the formulas (3.1.19) and (3.1.20). 3.1.4 The Function f (x). The Complex Velocity in the Fluid The solution (bounded in the trailing edge) of the integral equations (3.1.15) and (3.1.21) are obtained from the formula (C.1.9). We have therefore: (x)= Of 2 V -x' 44 /iTih'(t)d t. tt x l+xf (3.1.25) 1 Obviously, for the symmetric profile (h = 0) it results f = 0. One obtains the complex velocity substituting (3.1.16) and (3.1.25) in (3.1.11). Since, taking into account the formulas (B.5.5) and (B.5.7), we deduce _ dt 11-t 1+t (t-Z)(s-t) +1 1 +1 1 t 1 a 1 Z+t(tZ+st)dt= Z 1 s-Z Z+1' it follows W(Z) h' I -- A J-11 t'(Zdt+ i +zr(3 Z 1 Z+1 f, +1 (3.1.26) 1+ t h` (t) 1-tt-Z d t For the incompressible fluid (f3 = 1) one obtains the formula (t) u-iv=-1AJf 1 hit-z dt+ i + - z_ i +1 J +l /rTih (t) 1 (3.1.27) 1-tt-zdt' where z = x+iy. The formula (3.1.27) was given for the first time by Sedov [1.38], p.51 and deduced in a different manner by lacob [1.21), p.664, 76 THE INFINL'I'E SPAIN AIRFOIL IN SUBSONIC FLOW which solved the problem (3.1.9), (3.1.10) reducing it to a boundary value problem in the complex plane. It is not simpler to solve the bound- ary value problem than solving the singular integral equation (C.1.1). Moreover, the complex velocity field is not of much interest in aerodynamics. It is utilized only for determining the jump of the pressure on the profile (which is calculated directly in the framework of the method of the integral equations). 3.1.5 The Calculation of the Aerodynamic Action In some papers the aerodynamic action is calculated by means of a curvilinear integral on the contour of the profile. This calculation is wrong for contours with angular points. A correct calculation is performed using a control contour (surface in the three-dimensional case) surrounding the profile (wing). We shall perform this calculation where it will be absolutely necessary. Here we shall give a simple calculation, observing that in the first approximation, the action of the fluid comes from the jump of the pressure pl(xl, -0) - pi(xi, +0) = Dpi 1, which gives the lifting force, parallel to the Oy axis. Taking into account the formulas (2.1.1) and (2.1.3) on the unity of length of the cylinder, it follows the lifting force L= I (IPi I dxI = P.UULo J I i [!pO d x (3.1.28) and the following momentcalculated with respect to the point x°: MI =fJ to (xi - x°) x pt ?d xi , where j represents the versor of the Oy axis. We obtain MI = Mk, where rfi M = poU020Lo J (r - x°)OpOdx. (3.1.29) We denoted Bp0 = P(x, -0) - p(x, +0) (3.1.30) Instead of the dimensional quantities L and M, it is preferable to use the dimensionless quantities CL and chl, named the lift coefficient respectively the moment coefficient. These aerodynamic coefficients are defined by the formulas: L M CL, = (l/2)PooU2 (2Lo) ' cA' = (1/2)p.U, (3.1.31) 77 ME AIRFOIL IN THE UNLIMITED FLUID It is preferable to use the dimensionless aerodynamic coefficients because the numerical calculations are performed for dimensionless quantities. From (3.1.14), (3.1.28)-(3.1.30), and (3.1.31) it results CL =- f +1 f (x)d x, cAt = i -12 J ' (x - x°)f (x)d x . (3.1.32) i Finally, utilizing the solution (3.1.25), (B.5.4) and (B.6.9) we get cL- h(3.1.33) r+i JY CAf=-i13 J-1 f V ±tth'(t)dt- 1 2x°cL. 1 (3.1.34) In the case of the profiles which are symmetric with respect to the Ox axis, (h = 0), cL and cM vanish. Obviously we cannot use the method for calculating the drag because it has the order e2 3.1.6 Examples The flat plate. For the flat plate having the angle of attack (fig. 3.1.2a)), the equation (3.1.1) is E y = -xtgE = -Ex. It follows h(x) = -.ex whence CL= 27re ire CM= -(x° + 2)A 1 (3.1.35) These formulas were given for the first time by Glauert (1928) and Prandtl (1930). For the incompressible fluid (M = 0) one obtains CL = 27rE, 1l cAf = - z° + 2) Ire (3.1.36) We notice that CL is increasing (c,%f is decreasing) for M / 1 (fig. 3.1.3) (the lift is increasing because of the compressibility). In the vicin- ity of M = 1 (starting approximately with Al = 0.8) the lift has very great values, in contradiction with the reality. We deduce that in the vicinity of M = 1 the linear theory is not valid any longer. Therefore, for the transonic flow (Chapter-9) we shall utilize other equations. From (3.1.25) one obtains the jump of the pressure THE INFINITE SPAN AIRFOIL IN SUBSONIC FLOW 78 U. b) r c) Fig. 3.1.2. CI 2ttE M 0.8 1 Fig. 3.1.3. /if = -2e 1 V + x (3.1.37) The parabolic profile (fig. 3.1.2b)). This profile is obtained for h(x) = e (I - x2), h1 = 0. Utilizing (B.6.9) we find CL = Ze 73 °f=- 7r Cm 1 - xa (3.1.38) with the same interpretation for cG. The S-profile (fig. 3.1.2c)). For h(x) = e(x - x3) and hl (x) = 0, we obtain CL - x°) A , CH = 4,6 ( j , f = V-x + Tx 1+x --X 2 (3.1.39) 1 The profile with thickness, having the shape of an elliptic sector For a profile having the shape of a sector bounded by two arcs THE AIRFOIL IN THE UNLIMITED FLUID 79 of ellipse having the small semi-axes e2 < ei (fig. 3.1.4), we have h+(x) _ -62 VI -X h_(x) , x7, (3.1.40) 2h(x) = -(i, + s2) 1 - x 2h1(x) = (El -- 62W1 -X* , . uw Fig. 3.1.4. For f, and f , taking into account the representation (3.1.16) and (3.1.25) we deduce Aft = (F1 - E2) lx x v- 3f = , - dt t 1-t 1-t t -x X e2 1 ?r +xI f' 1-x t-x where 1+t £l x dt + 1 1-x (3.1.41) +1 dt 1 1-t Per forming the change of variable 1 -- t = u and taking into account (D.2.3), we get 7* +l 1ddu 1 ,/ 1- t = `r2 u = in 2. (3.1.42) Utilizing (B.5.8) we get 1(3f = El i Ex / I \ x . In 2 + x In l + x) (3.1.43) Ale also have £1 +F2 2- dt _ 61+12(2--1n2), (3.1.44) CMS _ - £12x°(2 - In 2). 2/3 The action of the fluid is equivalent to a lifting force passing through the origin. 80 THE INFINITE SPAN AIRFOIL IN SUBSONIC FLOW 3.1.7 The General Case It is well known that the Chebyshev polynomialsof first kind T,,(t) = cos(n arccost) (3.1.45) constitute a basis on the interval (-1, +1). We shall consider the series expansion of h(t) in this basis: h'(t) = E an cos(n arccos t). (3.1.46) t=ooso (3.1.47) Putting it follows co h'(cos a) _ E an cos no n-U whence j" h'(cos8)d6, ao an = 2 0 7r 7T jh1(cos0)cosn0d0, (3.1.48) n= 1,2,... Substituting (3.1.46) in (3.1.33) and (3.1.34) and performing the change of variable (3.1.47), one obtains 2n cf, ao + al 2 a2\ 7r V cdt = - a ao + al + 2 J. (3.1.49) Theses formulas were given by Homentcovschi in (A.201. It is important to calculate the distribution of the jump of the pressure on the profile. Using the change of variable (3.1.47) and Glauert's formula (B.6.6) we deduce +l 1 + t h'(t) °Q (1 + cos a) cos no- J 1-tt-xdt=EanJ n-o coso-cos9 U = (3.1.50) = 7r cot 8 °O an sin n9, x = coo 8. -o Replacing in (3.1.25) it follows: pf(x)=2Eansinn8 00 n=0 It is Glauert who proposed this type of solution. (3.1.51) THE AIRFOIL IN THE UNLIMITED FLUID 3.1.8 81 Numerical Integrations In the case of an arbitrary profile, for calculating the integrals (3.1.33) and (3.1.34) one may employ quadrature formulas (F.2.24). Using the notations _ tQ-cos 2a-1 a=1,...,n 2n + 1' (3.1.52) , one obtains n > (1 + 41r 13(2n + 1) cL (3.1.53) CM - n 2n -T3 1 a (1 + ta)tali (tQ) - 2 CL. (2n -+I) a- 1 For calculating the integral from (3.1.25) one utilizes the formula (F.3.1). In the collocation points 2i7r xj = t = 1,...,n, Cos Fn + 1 , (3.1.54) it results f(T,) 4 p(2n +' 1) X. E 1 + t Q=1 Q - x1 h (tQ) (3.1.55) The results obtained with these formulas are much more accurate than the results obtained with other methods. 3.1.9 The Integration of the Thin Airfoil Equation with the Aid of Gauss-type Quadrature Formulas We may use the quadrature formulas from Appendix F to create and extremely efficient method for determining the solution of the integral equation (3.1.15). There were made many such attempts in various papers but nowhere one can find the good solution because there is not prescribed the behaviour of the solution in the points ±1. We know from Appendix C that the solution of the above mentioned equation depends on the behaviour imposed in ±1. The solution satisfying the Kutta-Joukovsky condition in the trailing edge is /3f (t) _ -E + t F(t) 1 (3.1.56) 82 THE INFINITE SPAN AIRFOIL IN SUBSONIC FLOW Using (F.2.19) and posing h = -Eh, we reduce the equation (3.1.15) to the algebraic system n Ai,F, =hl, j = 1,...,n, (3.1.57) a=1 where Fa = F(ta), h1 = h (xj), A,a = t- 2n + I ' 2n + 1 Qa- z1 = x,' (3.1.58) + 17r' In (3.1.57) we have a linear algebraic system with n equations for n unknowns Fl, ... , Fn. We can create a computer code for Cos solving the system. The coefficients AJa, the weight points to and the collocation points xj are the same for every profile. For a given profiles we have to change only the column with the elements h1, ... , K. After determining the unknowns Fa, the lift and moment coefficients result from the formulas (3.1.32) and (F.2.18). CL = - +1 f(t)d t = F t F(t)d t = j_1I l+t Q (3.1.59) 1 cl=-Z J +1 1 WE tf(t)dt=-2ENA1, where n ta)F0, NL a=1 (3.1.60) a=I For verifying the method we utilized the analytic solution for the flat plate (3.1.37). From (3.1.56), it follows that for hJ = 1 we must have F = 2. The results obtained with the numerical method described above, with n = 20 gave for F. values situated between 0.999 and 2.001 and for NL and NM the value 0.999. Hence the method is extremely efficient [3.12]. 3.2 3.2.1 The Airfoil in Ground Effects The Integral Equation When an airplane is landing or taking off we have to take into account the ground effects. Some of the first papers devoted to these effects belong to Tomotika and his fellow-workers (3.44], [3.45]. We have also to 83 THE AIRFOIL. IN GROUND EFFECTS mention the papers of Pancenkov [1.33], [1.34], [3.36], and especially the papers of Plotkin and his fellow-workers [3.37]-(3.39}, where one gives the integral equation of thin profiles in ground effects and one proposes approximate solutions. The is considered a small parameter. Widnall and Barows (5.37J and Tuck [3.46] used asymptotic methods for investigating the problem. In fact one encountered two small parameters: the arrow of the airfoil and the distance from the airfoil to the ground. The fluid was considered incompressible. In the sequel, following 13.131, we shall utilize the method of funda- mental solutions, in order to obtain the integral equation for the compressible fluid and for the airfoil with thickness. The small parameter is the arrow of the airfoil. In case that the distance from the airfoil to the ground is also small, we have to elaborate a new theory. We use the notations from the previous subsection. We denote by a/2 the distance from the airfoil to the ground (fig. 3.2.1). The perturbation has to satisfy the following boundary conditions: v(x, f0) = h(x) ± hI (x), v(x, -a/2) = 0 U. Ix1 < I. (3.2.1) - oo < x < oo . (3.2.2) -I +1 x T a/2 an -I --- - - +1 Fig. 3.2.L. According to the method of fundamental solutions, we have to replace the airfoil by a continuous distribution of forces (ft, on the segment [-1, +1] from the y = 0 axis. For satisfying the boundary condition (3.2.2) we shall consider a symmetric distribution of forces (fl, -f on the symmetric segment (-1, +11 of the y = -a axis and we shall determine the intensity of the distributions from the condition (3.2.1). Taking into account that the perturbation produced within the uniform stream having the velocity U,i by the force (fl, f2) is (2.3.23), it follows that the two distributions will determine in the fluid 84 THE INFINITE SPAN AIRFOIL IN SUBSONIC FLOW the perturbation 1 J-tf +t P(-T' y) = + v(x, y) = 1 xoft W + J32yf (t) x02 +,82y2 f +t xofi(Q - 132(y + a)f (C) + M(y + ;)-2 A f+t z0f(0 - bf1(0 d xa +U 1 - (3.2.3) 1 r+t xofW + (y + a)ft (_) 21r x0#2(N+a)2 1 f and f j having to be determined. One easily verifies that this representation satisfies the condition (3.2.2). We denoted xo = x - . Only the first integrals become singular. when passing to limit for y ±0, Using the formulas (3.1.19) and (3.1.20), we deduce p(x, +0) - p(r., -0) = f (x) , f(o d t- t v(x, f0) _ 2ft(x) + 27r (3.2.4) , (3.2.5) +1 io+m 11-r where in = fla. adding, we get: Imposing the conditions (3.2.1), subtracting and ft(x) _ -2hi(x) ar FI fro) 2l jxi < 1 t1 oo+ in dt = H(x), (3.2.6) (3.2.7) where H(x) = h'(x) - A ,l +t,,,2d 1 xo (3.2.8) The equation (3.2.7) is the integral equation In Appendix C it was called the generalized equation of the thin airfoil. For the incompressible fluid (/3 = 1) and for the airfoil without thickness (ht = 0) it coincides with the equation given by Pancenkov and Plotkin. 85 THE AIRFOIL IN GROUND EFFECTS 3.2.2 A Numerical Method The equation (3.2.7) is obviously a singular integral equation. As we have shown in Appendix C, it may be reduced to a Fredholm equation, but the problem remains still unsolved because there are not available general methods for solving this type of equations (excepting the method of successive approximations). As we have already shown in 13.131, the equation (3.2.7) may be solved numerically utilizing the quadrature formulas from Appendix F. Looking for the solutions of (3.2.7) having the form (3.1.56), putting H = -e and using (F.2.18) and (F.2.19), one obtains the linear algebraic system n 7 1j . (3.2.9) a=1 where FQ = F(t0), T-1 j = H(xj), B ra 1 (t° - 1)(xj - t0) 2n + I (xj - tu)2 + m2 (3.2.10) Aj,,, t,, and xj being given in (3.1.58). From the system (3.2.9) we determine the unknowns F1,. .. , F,,. Now the lift and moment coefficients (3.1.32) (for x° = 0) will be obtained from the formulas CL J_ !+1 CAI 2{3 I. (3.2.11) _ + t tF(t)d t 2 Nit , where we have utilized the notations (3.1.60). One may write computer programs for solving the system (3.2.9). The coefficients Aja and Bj0 do not depend on the shape of the airfoil, such that the program may be utilized for every airfoil. One changes only the matrix with one column Wj . 3.2.3 The Flat Plate In the case of the flat plate with the angle of attack we have 77j = 1. We solve the system (3.2.9) and then we determine NL and THE INFINITE SPAN AIRFOIL IN SUBSONIC FLOW 86 NM with the formulas (3.1.60). We write the aerodynamic coefficients (3.2.11) as follows CL = cL NL, (3.2.12) CAf = where c' and cM are the coefficients for the free stream (a = oo). 5. NL1 4 3 2. 1.0.6 Mp0 } 1. 0.0 4.0 2.0 6.0 8.0 0.0 4.0 2.0 6.0 8.0 a a a) b) Fig. 3.2.2. The variation of NL and NAf versus a is shown in figures 3.2.2a) and 3.2.2b). The compressibility effects on the lift and moment are greater than it is shown in the above mentioned figures, because A also intervenes in c -L4 cy. Obviously, the ground effect is great when a is small and it quickly decreases when a increases. 3.2.4 The Symmetric Airfoil As we have seen in (3.1.5), for the symmetric airfoil in a free stream with zero angle of attack, the lift and the moment vanish. The situation is different for the wing in ground effects. As we may observe from (3.2.7) and (3.2.8), i i the case of symmetric airfoil (h = 0, hl 0 0), H does not vanish, hence the solution of the equation. (3.2.7) is different from zero. For the Joukovsky symmetric profile considered in 11.28), [3.37], [3.13) y = +E(1 - X) VI - x2 (3.2.13) THE AIRFOIL IN GROUND EFFECTS 87 we obtain using the notation hl = -(1 - x) 1- x' and the formulas (F.2.12) and (F.2.18): H =-m ;r J-1 (xj - e)2 + m2 in P d_ 1 - (-2, '1 2m to (1 - trt ) +1F-(,)2+fn2 - 2n+I E(xj-tn)2)+m2' (3.2.14) In this case one presents NL and N11 versus a in figures 3.2.3a) and 3.2.3b) for n = p = 20. The lift coefficient is where cos y7r p+ negative i.e. the resultant is a force pointing towards the ground. Hence when the airplane is landing or taking off it becomes heavier. The pilot has to take into account this fact. - al.bo - - - Msean6 I : _ F- a'l 0.001 l 000 1.60 / -=- l 4.00 5.20 a30 410 OAO. I 000 400 I 1 a 1.60 '20 4.90 640 900 b a) Fig. 3.2.3. The fact that both in the case of the flat plate and the case of the symmetric profile, the lift and the moment become very great when a is very small, is not true in reality. Hence, for small values of the parameter a we have to elaborate a new theory based on two small parameters (see [5.37]). 88 3.3 3.3.1 THE INFINI'T'E SPAN AIRFOIL IN SUBSONIC FLOW The Airfoil in Tunnel Effects The Integral Equation The experiments for determining the aerodynamic parameters are performed in wind tunnels. We have therefore to take into account the influence of the walls of the tunnel an the aerodynamic characteristics, of the airfoil. Since many papers dedicated to this subject are secret, we cannot give the history of the research in this field. We have cited in the bibliography some authors, without consulting their papers. We shall present therefore, only the model that we gave in [3.14]. This model can be easily obtained with the method of fundamental solutions and it is in the spirit of the theory previously presented in this book. We formulate the problem as follows: an uniform stream, having the velocity U ,i, the pressure pao and the density p,,, flowing between two infinite flat plates parallel to the O.T. axis, encounters a thin airfoil of infinite span with the generatrices parallel to the Oz axis. The fluid is compressible, and the velocity of the uniform stream is subsonic. One requires to determine the perturbation and the influence of the stream on the airfoil. We utilize the variables (2.1.1) and (2.1.3). Let (3.1.1) represent the equations and a the distance between the plates (walls) (fig. 3.2.1). For determining the perturbation, we have to impose the following boundary condition: v(x, ±0) = h'(x) ± hl(x),JxJ < 1 (3.3.1) v(x, ±a/2) = 0, -oo < x < oo (3.3.2) Y +11 w2 xx I /2 Fig. 3.3.1. In order to utilize the method of fundamental solutions, we shall f)(t) replace the airfoil by a continuous distribution of forces defined on the segment [-1, +1). For satisfying the boundary conditions (3.3.2), we have also to take into account symmetric distributions on the images of the strip [-1,+1) in the planes y = ±a/2 and symmetric distributions on the images of the images in the planes y = ±3a/2 etc. 89 THE AIRFOIL IN TUNNEL EFFECTS (the method of images). In this way one obtains the following general representation of the perturbation: +oo P(x, Y) - 27rp na)f(e) d xo + 02(y - na)2 +1 1 J- , (3.3.3) ( v x, y) = p }OO f 27r 1 (-1)"Xof(t) - (y ro + (y - na)2 with xo = x - . We can easily verify that v given above satisfies (3.3.2). For imposing the conditions (3.3.1) we have to pass to the limit considering y -+ 10, -1 < x < 1. The only singular integrals correspond to n = 0 and they are calculated using the formulas (3.1.19) and (3.1.20). Taking into account the equalities (1.161: 00 _ 1 k2 + n2 n=1 zr cosh kn 2k sinh kir 1 2k2 ' A 2k sink ka "-1 k2 + n2 2k2 ' we deduce: 1 p(x, f0) = f 2 f (x) + v(x,±0) _ f xo )d + (3.3.4) +1 1 7rp -L J_1 2f1(x) + 2r , rLd+ (3.3.5) +p where m = ap Kl (xo) _ I fi K(xo)f(t)dt`, -o K(xo) = m sink-' (m xo' - , coth \\(m xo// , \ XO (3.3.6) From (3.3.4) we deduce again the significance of the function f : f (x) = p(x, +0) - p(x, -0), (3.3.7) 90 THE INFINITE SPAN AIRFOIL IN SUBSONIC FLOW and from (3.3.5) and (3.3.1), fl(x) = -2hi(x), (3.3.8) +I x 2, K(xo)f()d= h'(x) . (3.3.9) 1 The formula (3.3.8) determines fl and the equation (3.3.9), the function f. The integral equation of the problem is in fact the generalized equation of the thin profiles. The kernel K has no singularity for t = x. We notice that for a - oo (m -+ oo) the equations (3.3.4) and (3.3.9) are reduced to the equations corresponding to the airfoil in a free (unlimited) airfoil. 3.3.2 The Integration of the Equation (3.3.9) We utilize the quadrature formulas from Appendix F for integrating the equation (3.3.9). Putting h = -eh, we shall look for solutions f having the form (3.1.56). Utilizing (F.2.18) and (F.2.19), we reduce the equation (3.3.9) to the system n j=1,...,n, >2CfaFo =hj. (3.3.10) 0=1 where F. = F(ta), hi = h(xj), C'° = m 2n 1 sinh-1 I -(xj - to,)} , (3.3.11) §i xj being given in (3.1.58). After determining the unknowns F1,.. . , F, from (3.3.10), the lift and moment coefficients may be obtained by means of the formulas: tQ (3.3.12) cA, _ CL = NL and NA, having the expressions given in (3.1.60). The distribution of the pressure on the airfoil may be obtained from (3.3.4), taking into account (3.3.8). We get p(to,T0)=ff& 2J3 l+ta 1 rr/3 1 hl(x)dx1 tp - x THE AIRFOIL rN TUNNEL EFFECT'S 91 The values of the pressure in the points to, may be determined in the wind tunnel by means of pressure plugs. We have thus the possibility to verify the theory presented herein. We may use a computer for solving the system (3.3.10). Since the coefficients C1,, do not depend on the shape of the airfoil, we have to change in the program only the column containing the elements 1 . Numerical Results 3.3.3 = 1. In For the flat plate having the angle of attack e we have this case, the formulas (3.3.12) become cL = cL NL, cm = cMNM, cOLO and cc representing the coefficients corresponding to the unlimited fluid. - M. b 0 » 1 MWI"06 za -I I 1.4 1.3 a 00 02 0.4 0.6 01 1.0 '10.1 00 01 0.4 06 02 1.0 to" h) a) Fig. 3.3.2. The figures 3.3.2 present the variations of NL and Niv versus the width of the tunnel for M = 0 (incompressible fluid) and M = 0.6. We notice that the lift in the wind tunnel is greater then the lift in the unlimited fluid and it decreases when the width of the tunnel increases. For the same width, the lift is an increasing function of M. The theory presented herein allows us to determine these variations. 92 3.4 3.4.1 THE INFINITE SPAN AIRFOIL IN SUBSONIC FLOW Airfoils Parallel to the Undisturbed Stream The Integral Equations In the literature, the problem of the uniform flow past a configuration of airfoils has been solved for the incompressible fluid and for airfoils whose exterior may be mapped conforma.lly on the exterior of a circle. There were given especially solutions for the biplane [1.36], [3.38]. We shall consider in the three forthcoming subsections, following the paper [3.15], the general problem of the compressible fluid in subsonic flow and airfoils with thickness of arbitrary shape. YA UM a/2 x Fig. 3.4.1. We shall consider in this subsection, two airfoils parallel to the undisturbed stream (fig. 3.4.1), not necessarily identical. The free stream has the velocity U,,,i, the pressure p,, and the density pp,;,. One utilizes the variables (2.1.1) and (2.1.3) and one considers that the equations of the airfoils are: y = h(x) . h1(x), Y = 1(x) ± 11(x), krl < 1 (3.4.1) 1 XI < 1 (3.4.2) the first corresponding to the upper airfoil and the second to the lower airfoil. The dimensionless distance between the projections of the airfoils is denoted by a. Utilizing the method of fundamental solutions, we shall replace the upper airfoil by a distribution of forces of intensity and the lower airfoil by a distribution (gl, g)(t), both of them (fl, defined on the segment [-1, +1]. Starting from the above mentioned perturbation produced by a force in the free stream (the fundamental solution (2.3.23)), we deduce that the perturbation produced by the two AIRFOILS PARALLEL TO THE UNDISTURBED STREAM 93 distributions is: +t 1 Ax, U) - 2yrQ xof 1(e) + QZ(y - a12)f (t) 1- 1 xa +,Bs(y - a/2) s d t+ +t X091 (t) + O2(y + a/2)9(t) 1 + 27rp J-1 d (y + a/2)2 xo + (3.4.3) x, y) = ( p T,,-- 1 +1 xof (e) - (y - a/2)ft(e) d t+ xp +,02(y - a/2)2 1 + Q r+t xo9(t) - (y + a/2)gt () d t, 2tr f 1 xo + /32(y + a/2)2 where x0 = x - .. The distributions of forces will be determined from the boundary conditions When y v(x, (a/2) f 0) = h'(x) ± hi (x), fix( < 1 (3.4.4) v(x, -(a/2) f 0) = !'(x) ± li (x), jxj< 1 (3.4.5) (a/2) ± 0), the first integrals from the representation (3.4.3) become singular and they have to be calculated with the formulas (3.1.19) and (3.1.20). One obtains P(x, (a/2) f 0) 2 f (x) + +1 1 + 2aQ ' fl , 270 T-1 -;To (3.4.6) xo.91 + 7n,89 xo + jn2 _1 1 v(x, (a/2) 10) _ 1 ft (x) + 2a T-1 d+ d f (')d t+ xo (3.4.7) 1 1 Qxog - f + 2rr ,l 1 mgt xo + m2 d , where rn = ap. From (3.4.6) it follows f(x) = P(x,(a/2) + 0) - p(x,(a/2) - 0), (3.4.8) and from (3.4.4) and (3.4.7) fi(x) = -2h'1(x) (3.4.9) 94 THE INFINITE SPAN AIRFOIL ICY SUBSONIC FLOW J t f x()) d t + 2 , i t xo +(r roe d = H(x), where (3.4.10) +I H(x) = h'(x) + 2 d . (3.4.11) In the same way, passing to the limit in (3.4.3) when y -i -(a/2)±O (in this case the second integrals become singular), from the boundary condition (3.4.5), we obtain g(x) = p(x, -(a/2) + 0) - p(x, -(a/2) - 0), '(mot 9(e) 1 gi(x) = -21c(x), +t f( )2dC + I 0 I X2 +,fn (3.4.12) (3.4.13) = L(x), (3.4.14) 0 where L(x) = l'(x) - ,1 t I x('2d. (3.4.15) The distributions of forces on the two chords are determined by solving the system of integral singular equations (3.4.10) and (3.4.14). From (3.4.9) and (3.4.13) the functions H(z) and L(x) are known. The field of pressure on the profiles are given by the formula (3.4.6) and the corresponding formula for p(z, -(a/2) ± 0). Symmetric Airfoils. If the two airfoils are symmetric with respect to the Ox axis, then l(z) = -h(x), 1I (x) = hI (x) . (3.4.16) It follows L(x) = -H(x) whence xz(+tn r, I If gdt;+ o J (3.4.17) I Since the solution of the generalized equation of thin profiles is unique, it results f = -g, and the 'system of equations (3.4.10) and (3.4.14) reduces to a single equation 2 Ifro) d 1 2 f +I I +(n) 'TO f s d4= H(x) (3.4.18) o which is just the equation (3.4.7) of the airfoil in ground effects. The result is natural. The lift and moment coefficients for the upper airfoil 95 AIRFOILS PARALLEL TO THE UNDISTURBED STREAM are given by the formulas (3.2.11). For the lower airfoil they have the opposite sign. For the entire configuration we have cL = cAt = 0. Identical Airfoils. If the airfoils are identical, then 1(x) = h(x) - 2a, it(x) = hj(x). (3.4.19) Moreover, if the airfoils have no thickness, then L = H. Subtracting the equations (3.4.10) and (3.4.14), it results g = f whence A f I_ z2 +(m)2d = h'(x). 1 1 27r (3.4.20) This equation has the form of the generalized equation of thin profiles and it can be integrated numerically like in (3.2.2) and (3.3.2). 3.4.2 The Numerical Integration In order to solve numerically the system (3.4.10) and (3.4.14), we shall use the quadrature formulas from Appendix F. For thin airfoils, the functions H(x) and L(x) have the form H(x) = -eH(x), L(x) _ = -iL(x), hence we shall look for the following type of solutions Qf (t) = -E V 1 +t F(t), 3g(t) = -E 1 + tG(t) (3.4.21) which satisfy the Kutta-Joukowsky condition on the trailing edge. Utilizing (F.2.18) and (F.2.19), the system (3.4.10) and (3.4.14) is reduced to n E(Aj.F.-Bj.Go)=Hl, ,7=1,...,n, a=1 (3.4.22) n >(A1oGo-BjoFa)=1i, a=1 A,,,, tQ and x, are given in (3.1.58) and Bj, in (3.2.10). Like always, F. = F(t0), G. = G(t0), H, = 77(x,), LJ = L(x,). We have to find out the unknowns Fl,. .., Fn, Gl, ... , Gn from the system (3.4.22). The coefficients Aj0 and B 0 are the same for all the airfoils; only H and L, are varying. The lift and moment coefficients for the entire configuration are given by the formulas THE INFINITE SPAN AIRFOIL IN SUBSONIC FLOW 96 CL =-1 1+1 +1 (f+9)dt= /(F+G)dt--Ni 1+t (3.4.23) +1 CAI --2J-, +1 t(f+9)dt=2Q11 where t l+t(F+ c)dt = -ASE2Q1V2, n N, = 1 2n + 1 :(1 - ta)(F. +G(), Q=1 (3.4.24) n N2 2n 1 1 1)(F.+G.). 0=1 We notice that N, and N2 contain only the unknowns Fa + Ga which may be determined from the system Aja(Fa + Ga) - Bja(Fa + Ga) = Nj + Lj (3.4.25) a=l of n equations (j = 1, ... , n) with n unknowns. The unknowns Fa and Ca are separating only when we calculate the pressure on the two airfoils. For example, the field of the pressure on the upper airfoil is obtained from (3.4.6) with the formula p(t,., (a/2) f 0) = f 29E f 1 -to F. 0 - 4)G(ti) f3(2n + 1) (ta - t{)2 + m2 - x)li(x) h', (x) dctQx(ta-x)2+m2d x, iro f!" 1 t cos 2iIr 2n+1' (3.4.26) to being given in (3.1.58). In the same way we may determine the field of the pressure on the lower airfoil. The numerical determinations have been performed for the biplane having the angle of attack E(h' = 1' = -E) (fig. 3.4.2a)) and for the symmetric biplane (h' = -1' = -e) (fig. 3.4.2b)) taking n = 10. In GRIDS OF PROFILES 97 a) b) Fig. 3.4.2. the first case HJ = L, = 1, and in the second 77J = -Tj = 1. The coefficients A, , Bin depend only on the parameter m = a#. In the second case we obtain N1 = N2 = 0 for all the values given to m (it is natural). In the first case, the values of N1Q'1 and N2Q-1 depend on a and M. They are given in tables 1 and 2. In the first line we may find the values of 13-1N1 and 0-1N2 for the monoplane. We notice that the lift coefficient for the biplane is much greater then the lift coefficient for the monoplane and it increases with a. The lift increases also with M. The same conclusions are true for the moment coefficient. Table 2 Table I The values of N,$-1 a=0.5 a=1 a=5 3.5 3.5.1 M=0 M=0.6 1.00 1.461886 1.25 1.741425 1.70842 1.981013 2.034937 2.46087 The values of N2 a=0.5 a=1 a=5 1 M=O M=0.6 1.00 1.624288 1.25 1.941250 2.209987 2.477736 1.827670 1.991025 Grids of Profiles The Integral Equation The classical problem may be also solved in a simple manner by means of the method of fundamental solutions. One obtains again the generalized equation of thin profiles. Let us consider a grid of identical airfoils having the equations y = na + h(x) ± hl (x) , n=0,±1,±2,..., (3.5.1) 98 THE INFINITE SPAN AIRFOIL IN SUBSONIC FLOW which perturb the uniform subsonic stream defined in section 2.1.1. Since every airfoil produces the same perturbation, we shall replace every profile by the same distribution of ford (fl, f) (c). Since the perturbation produced by the force (fl, f2) in the uniform stream is (2.3.23), we deduce that the perturbation produced by the grid is 1 C` P(x, U) = +t xoft(s) + Q2(y 1-1 (3.5.2) Q 21r u(x, A = +°° -00 f +' xof W - (U - na)fl W d x0 + (32(y - na)2 L where xo=x-t;. The profile corresponding to n = 0 will be called the reference Passing to limit, when y -, ±0 the integrals corresponding to n = 0 become singular and they are calculated with the formulas profile. (3.1.19) and (3.1.20). Employing the notation k = xo/m. we get +00 n*o -00 x° A0 + m2n2 - 2x° -00 nx n - 1 2x° 1r k2 - n2 - m2 \2k m2 = +00 00 n coth -xo m 7T coth k7r - 1 2 1 K(xo), r TO 2n=0' (3.5.3) whence: '6 P(x, ±0) = f 2 f (x) + j,- f x°) d t+ f1 (3.5.4) +z 1 +2- j h (xo)fi (4)d t, P(x, +0) - P(x, -0) = f W, (3.5.5) d+ 1 u(x, f0) _ 2 Q ft (x) + 2 J t o (3 .5.6) +i +2 fl 99 GRIDS OF PROFILES Imposing the boundary condition v(x, f0) = h'(x) ± M, (x) (3.5.7) fi(x) = -2hi(x) (3.5.8) we deduce +1 2 j K(xo)f(t)dt If (o) d t + = h'(x) (3.5.9) 1 This is the integral equation of the problem. It has the same form like the equations (3.2.7), (3.3.9) and (3.4.10), (3.4.14), excepting the non - singular kernel. Obviously, from (3.5.3) we deduce that the kernel K(xo) is not singular. After determining f by means of (3.5.9), the field of pressure on the reference profile will be obtained from (3.5.4), and (3.5.5) will be utilized for the calculation of lift and moment coefficients. In order to obtain the pressure on the profile corresponding to n = I and to impose the boundary condition on this profile, we must pass to the limit in (3.5.2) y - a :E 0. In this case, the integrals corresponding to n = 1 become singular. With the change of variable y - a = Y we have for example rx, +t Ax, a t 0) = lim to J- xxo + #2Yf d + 27r,3 Y + +00 1 n#1 27 r3 2 -00 +1 xof1 +,82(1 - n)af d ,l-1 x0 +1112(1 -n)2 Putting I - n = nl in the last integral we deduce +00 n'At -00 _ +ao 1 To' + m2(1 - -a0 00 1-n +00 "#1 n)2 n,00 X.2 + m2(1 - n)2 1 m2n' -0 such that, taking into account (3.5.3), we obtain for p(x, a t 0) the same expression like for p(x, ±0). In the same way, for v(a, a f 0) one obtains the same formula like for v(x, ±0). In this way, imposing the boundary condition v(x,a±0) = h'(x) ±14(x), (3.5.10) 100 THE INFINITE SPAN AIRFOIL IN SUBSONIC FLOW one obtains the equations (3.5.8) and (3.5.9). We draw the conclusion that if f1 and f are determined by (3.5.8) and (3.5.9), the boundary conditions arc satisfied on each profile and the field of pressure coincides with the field on the reference airfoil. The lift and moment coefficients are calculated on every profile by means of the formulas cj =-J 1 1 2 1 3.5.2 r f(t)dt, cal=--- + tf(t)dt. (3.5.11) t The Numerical Integration If f haa, the form (3.1.56) and It has the form -sh, using the quadrature formulas (F.2.18) and (F.2.19), we obtain from the integral equation (3.5.9) the system n Di,,Fa = Itj, j = 1, ....:a , (3.5.12) a=1 where T 1 - t Goth in 2n + 1 71 in (xj - ta) , (3.5.13) to and xj being given in (3.1.58). Considering for CL and cAf the form (3.1.59), we obtain for NL and N,%j the expressions (3.1.60). For example, for a grid of flat plates having the angle of attack E(hj = 1), the variations of the coefficients Nf and N,tt versus a, are given in figures 3.5.1. We notice that the coefficients cL and cAt are increasing functions of a and Al. They tend asymptotically to I when a - cc, hence the lift and moment coefficients tend asymptotically to the taken in the case of a single profile. The pressure in the control points (nod(s) is obtained from the formula P(tu>f0) = 213 1 7T,9 x- tQda.- (3.5.1 l) /' ,ro -1 AIRFOILS IN TANDEM 101 .o N 12 1.02 a% IN 09$ 014 091 066 011 US 0n 00 a 000 020 010 0.60 030 040 1.00 X10-, a) 000 am 040 060 0.*0 1.00 X10' b) Fig. 3.5.1. 3.6 3.6.1 Airfoils in Tandem The Integral Equations In the sequel we shall determine the perturbation produced in the uniform stream defined in 2.1 (under the assumption that it is subsonic (M < 1)), by a configuration of airfoils in tandem (fig. 3.6.1), having the equations y=11(x)±h1(z), al <x<bl, (3.6.1) a2 < z < b2 . (3.6.2) y =12(x) f h2(z), For the incompressible fluid this problem was studied by Chaplygin and Sedov (1.38) who determined the complex velocity. If we reduce the problem to integral equations and we solve them like in 13.151, we can determine directly the quantities of interest in aerodynamics (the field and the jump of the pressure on the airfoils, the lift and moment coefficients). Moreover, the fluid may be compressible and we don't need to use elliptic integrals. In the sequel, we shall present the results from the last paper cited above. Replacing the airfoils by distributions of forces having the intensities (gl . fl) respectively (g2, f2) and taking into account the fundamental THE INFINITE SPAN AIRFOIL IN SUBSONIC FLOW 102 Y K a, b, a3 b, Fig. 3.6.1. solution (2.3.23), we get the following representation of the perturbation: jbX0g1(t)+#2yf1(C) 1 P( x, y) = 2aQ 1 x0 + /y2 r"' xo92( +2Q 1 S ) + 02Yf2(t) d xo + Q2 y2 , (3.6.3) v(x,y)= xofiW --y91( )d xo + 132y2 2,r J a Q rb' + 2a az x0 + dtt y2 We kept the same notations like above. Passing to limit for al < x < b1 and taking into account (3.1.19) and (3.1.20), one obtains: P(x,f0) = 2fi(x) + 27r# xo J a2 (3.6.4) p(x, +0) - p(x, -0) = f 1(x) (3.6.5) f,a, 0 Q, f bi fi(t) v(x, t0) = bz o1( )dC+ 91(x) + 27r J n, xo d+ fb aZ xo d ' (3.6.6} Imposing the boundary condition v(x, ±0) =1 (x) ± hi (x) it follows 91(x) = -2hi(x) Q fl(O) 27r1 axo 1 d 2a Jxo 2xo L d=11(x) (3.6.8) Q f + + (3.6.7) AIRFOILS IN TANDEM 103 In the same way, for a2 <x < b2 we deduce: P(x,±0)=f2f2(x)+2a V(x, f0) = -T 192W + 2 27r - C g 0{dt+2 ) ., f )dt+ 2 Mo 22(i)dt(3.6.9) Y J., x0 f oQd(, z whence, utilizing the boundary condition v(x, f0) = !'(x) f h2(x) we obtain p(x, +0) - p(x, -0) = MX), (3.6.10) 92(x) _ -214(x), (3.6.11) 2f a, i t 'f : _ (x). f ( 3. 6 . 12) Hence the functions fl(x) and f2(x) will be defined by the system (3.6.8) and (3.6.12). This is an interesting mathematical problem, since the first equation is defined for al < x < bl and the second is defined for a2 < z < .b2. 3.6.2 The Determination of the Functions f, and f2 For f, the equation (3.6.8) has the structure (C.1.5) which has the solution (C.1.9). Utilizing (B.6.11) and the last formula from (B.5.4) we deduce: Of, (x) = Lt (x) + x xt- t 2(f)d f V - bt (3.6.13) where L1(z) = a b-a t n, (3.6.14) i Substituting fl given by (3.6.13) in (3.6.12) and taking (B.6.11) and the last formula from (B.5.4) into account, one obtains the following integral equation for f2: 2 r.2 V/F - a' f2(t)dt=L2(x), a2<x<b2, (3.6.15) THE INFINITE SPAN AIRFOIL IN SUBSONIC FLOW 104 La(x) r x- ,(x) = a1 b1 -! - a, h' 7r Ja, Y -- x 4b1 -- clt . (3.6.16) The type of the equation (3.6.15) is (C.1.1) and its solution has the form (C.1.9). One obtains: 2 ((x - b1)(b2 - x)11/'r,ll (x - a t)( - aa) /3f2(x) = /t- a l. L2(t)dt t-x t l or, taking (3.6.16) and the first formula from (B.6.11) into account, 13f2(2) = 2 (x - bi)(b2 _ x) 7r x - '1)(x - n2) + ( ( -a l}(aa (b1 - .)(b2 - t-n 1)(t-a- I.' ) (t-b1)(b2-t) t - .r a ) - x`l + a2 < x < b2 . dt (3.6.17) Once 13f2 determined, from (3.6.13) it results f3f1: (b1 - x)(ax - fllx) ! 2 a) r (x - al)(x - J a2 2 V(() - ) li(S} xd + --a1}(a2 (t - a1)(t - a2) 12(E) (t - b1)(ba - t) t - x a1 < c < b1. dt (:3.6.18) In this way we determined the solution of the system (3.6.8) and (3.6.12). 3.6.3 The Lift and Moment Coefficients From the formulas (3.6.5) and (3.6.10), we deduce for the lift and moment. coefficients: b; 1 f;(x)dx, cAr cL = - 1 b, xfi(a)dx (3.6.19) Utilizing the solutions (3.6.17) and (3.6.18), changing the order of integration and taking (13.6.5) into account, we find c1 bi -2j + a, b= 102 E)1t(ti)dti (b1 -- /(t - al)(t - a2)12(t)cit (t - b1)(b2 - t) . (3.6.20) AIRFOILS IN TANDEM 105 Utilizing the same solutions, changing the order of integration and taking (B.G.6) into account, we get for the moment coefficient: C.%/ ( - al)(a2 - ) L -13 (bi b2 FLI- al)(t-a2) bi)(t - t) +,._ n alt (od + , tt2(t)d t + al+as-b1 -b2 4 CL C. (3.6.21) Obviously, the formulas (3.6.20) and (3.6.21) are generalizations of the formulas (3.1.33) and (3.1.34). Utilizing (B.5.14) we may prove by induction that in the case of n profiles in tandem we have: r cL b kt /i_ak1(t)d k=1 1 (3.6.22) " n cet=-1R fj11 k=1 3.6.4 k t t Q I n bktlJ(t.)dt+(ak-bk) k=1 Numerical Values In the case of two flat plates having the angle of attack fs (fig. 3.6.2a) or 3.6.2b)), the integrals (3.6.20) may be expressed with the aid of elliptic functions. The first one has the form of the integral 252.21 from (1.4), page 105, and the second has the form of the integral 256.21 from the same book, page 122. For more complex configurations, these kinds of expressions are not available. However, in all cases, the integrals intervening in the expressions of c z, and ctit (for every n), may be calculated numerically by means of the quadrature formulas (F.2.24). For example, if n = 2, performing the change of variable C -+ t: = b1-al al+bl 2 + 2 -1<t<+1 (3.6.23) THE INFINITE SPAN AIRFOIL IN SUBSONIC FLOW 106 a) b) Fig. 3.6.2. and utilizing (F.2.24) we deduce: _ (b - ) (b'i - ) d - 2n+ 1 or=l (1 Jo 1 I1 Jl = n bl - at al)(a2 - 1 1 (a2f)) {d f t)(62 al + bl 2 (3.6.24) - - it a to b - to + ta) (bi - al)2 n 1 + 2(2n E to(1 + to) + 1) :to 0=1 a = 2a2 - (al + bl) b , = bl - al 252 - (al + bi) bl - al where to are given by (3.1.52). Similarly we get: '2 1 2 1 J2 , (t - al)(t - a2) (t - bl)(b2 - t) =a2 2 C= -' 1 + t 2n+1 am1 (t - bl)(b2 - t) 7r J2 (t-al)(t-a,) = b2-a2 n to (3.6.25) c-tom 2(2n+1) 2al - (a2 + b2) b2-a2 td t c d= Vd-to' 0=1 251 - (a2 + b2) b2-a2 Hence for two flat plates in tandem having the angle of attack -- we 107 AIRFOILS IN TANDEM have 0ir2 CL = £(11 +12), cM = re`(Jt +J2) (3.6.26) 13 where 11, 12, J1, J2 may be calculated numerically using a computer. For example, for at = 1, b1 = 2, a2 = 3 and b2 = 4 one obtains: 4 11 = 0.370671 and 12 = 0.629328. For a1 = 1, b1 = 2, and b2 = 5 one obtains: I1 = 0.415458 and 12 = 0.584541. For a1 = 1, b1 = 2, a2 = 6 and b2 = 7 one obtains: I1 = 0.449746 and 12 = 0.550253. In all the cases we have 11 + 12 = 0.999999, hence the lift has practically the value of the lift for a single plate and this is true even if the distance between plates increases. Since --11 + 12 takes the values 0.258657, 0.169083, 0.100507 it results that if the first plate has the angle of attack -c and the second e, the lift decreases when the distance between plates increases. If the first plate has the angle of attack -s and the first has the angle of attack e, the lift becomes negative. The previous formulas enable us to obtain numerical values for cL and cAl in the case of n airfoils with different shapes. Chapter 4 The Application of the Boundary Element Method to the Theory of the Infinite Span Airfoil in Subsonic Flow 4.1 4.1.1 The Equations of Motion Introduction The theory exposed in the previous chapter relies on the following three assumptions: 10 The linearization of the boundary condition. 2° The linearized condition is imposed on the chord of the profile, not on the boundary (as it is natural to do). 3° The linearization of the equations of motion. The assumptions are plausible for thin airfoils, otherwise they may be the cause of great errors. The application of the boundary integral equations method (BIEM), which is also called the boundary elements method (BEM), enable its to give up the first two assumptions. Hence we shall utilize the non-linear boundary condition which will be imposed on the contour (boundary) of the airfoil. For the incompressible fluid, we shall employ the exact equations (4.1.4) i.e. the equation of continuity and the equation of irrotationality, such that, in this case, the mathematical model (the equations of motion and the boundary conditions) will be valid for every airfoil (not only for the thin ones). In the case of the compressible fluid, we shall utilize the linearized equations, hence the results will be valid only for thin profiles, but even in this case the results are better then the results obtained by adopting the first two assumptions. 110 THE BOUNDARY PLEXENTSNIFTHOD The first step in solving a problem with the boundary integral equations method (or the boundary elements method) consists in reducing the boundary value problem to a boundary integral equation (whence it follows the first name of the method); the second step consists in solving approximately the integral equation, replacing the boundary with a polygonal line (whence it follows the second name). In addition we have to mention that there exists a direct method and an indirect method. In, the direct method we utilize the equations of motion for deducing a representation of the solution in the domain occupied by the fluid with the aid of the solution on the boundary. Then we pass to the limit on the boundary. In the indirect method, we replace the wing with a distribution of fundamental solutions (sources, vortices, doublets) on the boundary and and we determine their intensity imposing the boundary conditions to be verified; in this way one obtains an integral equation for the intensity of the fundamental solutions; the method is therefore indirect. As we shall see in 4.2, it is easier to use the indirect method, but the results are less accurate. It is more difficult to utilize the direct method (4.3) but the results are more accurate. Therefore, in the forthcoming subsections we shall use mainly this method. 4.1.2 The Statement of the Problem The problem was already formulated in the previous chapter. A subsonic stream having the velocity U,,.i (the Ox, axis has the direction of the unperturbed stream), the pressure p. and the density p is perturbed by the presence of a cylindric body with the generatrices perpendicular on the x1Oyu plane, having a given cross section (see fig. 4.1.1). One requires to determine the perturbation and the action of the fluid against the body. Denoting by X, Y the dimensionless coordinates introduced by means of the relation (xi,yi) = L0(X,Y), (4.1.1) Lo being an unspecified reference length, and by U,,,,V and pc,.U"'2'P, the perturbation of the velocity and the perturbation of the pressure, we have: V1 =UU(i+V), P1 =p +pWU,P. (4.1.2) The system which determines the perturbation is (2.1.30), written with capital letters. Projecting the second equation on OX and taking the damping condition at infinity into account, we deduce P = -U. (4.1.3) 111 THE F . UNrIONS OF MOTION y U., x Fig. 4.1.1. In this way we deduce the equations which determine the perturbation: /28U/ax + aV/aY = 0, 8V/ax - au/ay = 0, (4.1.4) U and V representing the coordinates of the vector V. and ,0 having the usual significance ([f = l - Af ). On the boundary C, of the airfoil we shall impose the condition (l +U)N,y +VNy = 0, (4.1.5) N,V and Ny representing the coordinates of the inward normal to C1. Performing the change of variables x=X. y=0Y (4.1.6) u= f3U, v=V, the system (4.1.4) becomes au/ax + aV/ay = 0, av/ax - au/8y. (4.1.7) In order to transform the condition (4.1.5), Ave notice that if the boundary C1 has the parametric equations X = X (S), Y = Y(S) with S increasing when the curve C1 is traversed like in figure 4.1.1, then N = (dY/dS, -dX/dS). (4.1.8) Performing the change of variable (4.1.6), we have x = x(X (S)), y = y(Y(S)) and THE BOUNDARY ELEMENTS METHOD 112 Cl and the transformed curve C being traversed in the same sense (a increases with S). It follows NX - # dS' Ny =n a-S (4.1.9) , such that (4.1.5) becomes (,3 + u)nx + l2vny = 0 on C. (4.1.10) One also imposes the damping condition at infinity lim(u, v) = 0. (4.1.11) 00 4.1.3 The Fundamental Solutions The first step in applying the BIEM consists in determining the fundamental solution of the system of equations (in our case, the system (4.1.7) ). We call source-type solution the solution of the system au'/ax + 8v /ay a(x - t, y - 11), (4.1.12) &v'/Ox -au.la!l = 0. The fact that the perturbation represented by 5 intervenes in the equation of continuity justifies the name of the solution. We apply the Fourier transform and like in Chapter 2 we obtain: U0 1 x-t 27r(x-e)2+(y- n)2' v* _ I y q 21r(s-t)2+(y-17)2 (4 1.13) We call vortex-type solution the solution of the system au /Ox + W /ay = 0, (4.1.14) Ov'/Ox Obviously one obtains: W = -v*, v = u' . (4.1.15) 113 INDIRECT METHODS FOR THE UNLIMITED FLUID CASE Indirect Methods for the Unlimited Fluid Case 4.2 The integral equation for the Distribution of Sources 4.2.1 Taking (4.2.13) into account, we deduce that if we replace the airfoil having the contour C (fig 4.1.1) with a distribution of sources of intensity f (x) apriori unknown, the perturbation of the velocity in the point Al(t) from the fluid will be given by the formulas: v( ) _ 2-4 -1 (4.2.1) ,Cf(x)Ix-4I2ds. In order to determine the intensity f (x) we impose the boundary condition (4.1.10). To this aim, we have to calculate the boundary values in (4.2.1) when C -+ zo a current point Qo of the boundary C. We are going to prove that if f (x) satisfies the Holder condition on C, i.e. there exist two positive constants A and p (p < 1) such that, for every two points Q(=j) and Q(x2) belonging to C, we have: If (z1) - f (w2)1 < Alaet - x21', then - 2 . PC f(x) v(xo ) = lim _ - 2 f (xo)n ° -2 - 2a Pf (m) (4.2.2) d8= (4 .2 .3) X Isa X0 d =012 it, in every regular point Geo. We denoted PC = li , Pi-C (4.2.4) where c is the arc cut out from C by the circle of radius E and center Qo (the arc QI Q2 from fig. 4.2.1). We denoted by no the inward normal at C in Qo. Indeed, taking the definition (4.2.4) into account, we notice that it suffices to evaluate the integral on c. We have limo Eli ( _ 2 - f f (x) \ = Ii n { L= liX0 PC 12-4 TV JC I-T-C ds= f [f (x) - f (xo)) I d aJ - f Z o) 44.2.5) THE BOUNDARY ELEMENTS METHOD 114 Fig. 4.2.1. Taking into account that f satisfies the condition (4.2.2), we deduce that the first integral from the right hand side of the equality (4.2.5) vanishes when E -- 0. For calculating L we shall replace the arc Q1Q2 with the segment on the tangent Q'Q'2 (fig. 4.2.1) and we shall put on this segment: x = xO + STO, s E [-E, +EJ . (4.2.6) We assume, for the sake of simplicity, that M(C) tends to Qo(xo) on the direction of the normal n°. We have therefore 4 =xo--9n° cu q>0. (4.2.7) It follows L= Um 17-0 r +C E o+ r1no s2 . ,2 q.e.d. Imposing the condition (4.1.10), we obtain from (4.2.1) and (4.2.3) the following integral equation: (flo2 + N2n02) f (xo)+ + (x - xo)n° +f2(y - bo)n° Uds=2pn2. cf(x) Ix-xo12 (4.2.8) INDIRECT' METHODS FOR THE UNLIMITED FLUID CASE 115 From this integral equation we are going to determine the intensity of the sources. For the incompressible fluid (M = 0) it may be solved exactly for the circular obstacle (4.12). 4.2.2 The Integral Equation for the Distribution of Vortices If we replace the airfoil with a distribution of vortices having the intensity (circulation) g(x) , defined on C, then, according to the formulas (4.1.15), we obtain the following expressions for the components of the perturbation u and v in a point M(t) from the fluid: da, u(0 = 2 where l; = (1:, rl), x = (x, Y) Taking the formula (4.2.6) into account, we obtain: Y-YO d s W012 U(X0) = 2g(xo)ny + 9(x) Ix - , (4.2.10) f 2 g(xo)n: v(xo) - -1 P9(m) 1 -Z X0 12 d and from the boundary condition (4.1.10): -M2g(xe)niny+ +1 ir '(x)p2(x-xo)ny-(V -bb)nrd8=2f3n=, c (4.2.11) Ix - xol where M is Mach's number in the unperturbed flow. For the incompressible fluid (A! = 0) we have a first kind integral equation. 4.2.3 The Boundary Elements Method One utilizes the following collocation method in order to solve the equations (4.2.8) and (4.2.11): one approximates the boundary C by it polygonal line {!,,}(j = 1 T), with the end points on C (fig. 'TIME BOUNDARY ELEMENTS METHOD 116 Fig. 4.2.2. 4.2.2) and one approximates on every segment Li, the unknown f (respectively g) by the value fj (resp. gi) from the midpoint of the segment. Denoting by :xf the vector of position of the midpoint, we have fi = f (x°), (j = L N). We may also consider a linear variation of f on L. In this case we have to determine the two constants from the values of f in the end points of the segment L,. i.e. from the values of f on C. Now the equation (4.2.8) may be written as follows: ( 02 +/32802) f(xo) + 1 +_7 r (x - X0 710 + )`12(y - ?ko)tty 'Y fi L, ds = 28nr . Ix - xo1Imposing the equation to be satisfied in every midpoint, i.e. putting successively xo = x9 . i = 1N, we get the system: N ai fi + Ai i f) = Ai, i (4.2.12) where tai = nr(xP) + '3ly1y(x)) (4.2.13) AiJ n'r(x:))Uii + /3lnv(x°)V i + Ai = fln (xn) , 117 INDIRECT METHODS FOR THE UNLIMITED FLUID CASE with the notations Ucj = / u'(x,x°)da L (4.2.14) Y1 = jv(zz?)ds1 f u' and v' being given in (4.1.13). The linear algebraic system (4.2.12) consisting in N equations will determine the N unknowns f{. Now the perturbation of the velocity in the points Po(x°) follows from (4.2.3) N 1 vi =-2f.n(x°)-EV1jfj J-1 (4.2.15) V>, = (U+f,V ) In the same way, from the equation (4.2.11) we get the system N hi9, + 1: Bij9j = A, , i = 1,(4.2.16) j:1 where 2b; _ -1tnz(x°)n,,(x°) , (4.2.17) Bij = #'nr(x°)U,j - n(x°)Vij. From (4.2.10) we obtain the following velocity field N 1 w = 29jnr(x0)+Evj9j j..1 (4.2.18) 1 Vf 4.2.4 N t'4s(x'i')-Uijgj. - The Determination of the Unknowns The integrals (4.2.14) may be computed numerically using quadrature formulas. They may be also calculated exactly for every shape of THE BOUNDARY ELEMENTS METHOD 118 the contour C. Since the integrals UUJ, V, are singular we shall calculate their Finite Parts. For avoiding the singular integrals, we may utilize the regularization method 4.8. For obtaining a parametrization of the segment L,, we shall denote by (x1, yl) and (x2,112) the coordinates of the end points PI, respectively P2, considered in the sense of traversing C (fig. 4.2.2). Then, the coordinates of the generic point on L., will be y°+-2ylt, -1<-t<1, x°-i-x22xlt, (4.2.19) where, obviously, x° X2 2 x1 yj=112 2 111 (4 . 2 .20) . Utilizing the formula ds = 1(&-)2 -+ (dy) , we deduce gds=l,dt, lJ = (x2-x1)2+(112-111)2. (4.2.21) Denoting 4a = I,b= (x°-x°)(x2-xl)+(i4 -Y?)(112-Yi), c = (xjI - x? )2 + (yp - y1)2, - +1 k 1 tkdt ate+bt+c' k = 0, (4.2.22) 1, we get (x - x°)2 + (y - y°)2 = ate + bt + c (4.2.23) whence, U,, = Io 47f [ (xj_ x°) +x2 2 x1J l] , (4.2.24) ii4i[(y°- y° ) Io+2 yi V 119 INDIRECT METHODS FOR THE UNLIMITED FLUID CASE Taking the running sense on C into account, we shall have in (4.2.13), (4.2.15), (4.2.17) and (4.2.18): n,,(x°) _ _-V1 r+y(x 0) = x1 !J X2 1 ii V2 (4.2.25) 1 11z`) n,,(xo) _ n (x?) = x I j. where (xl'), yi')) and (X( 2" Y2()) are the coordinates of the end points of the segment Li and 1; is the length of this segment, i.e. xz') - xi'1)2 + (yz;l - y1'))2]1/2 (4.2.26) . We have to notice that in (4.2.24) the integrals Ik may be calculated exactly. Indeed, since <0, A =- b=-4ac = we have to = V'--5 Il - I 2a In arctan. (4.2.27) dc-a' a+b+c - (4.2.28) b a-b+c a/ arc t an vr---s c-a Taking (4.2.23) into account, we deduce a + b + c = (x2 - x°) + (y2 (4.2.29) a-b+c= (xl -x°)2 +(yi -y°)2. For i = j we have b = c = 0, such that the integrals Ik become singular. Utilizing the formulas (D.2.3) we obtain Io = -2/a, Il = 0. (4.2.30) So, the coefficients ai, At,, A{, b1 and B1j depend on the coordinates of the end points of the segments L, on C. Hence, we may solve the systems (4.2.12) and (4.2.16) using a computer. 120 TllE BOUNDARY ELEMENTS METFIOD 4.2.5 The Circular Obstacle For testing the two methods presented above and the direct method from the following subsection, we shall use the exact solution for the incompressible flow past a circular obstacle. This solution is already known. If 0 is the origin (center) of the obstacle and R its radius, then the complex velocity in dimensionless variables is (see (5.2.4) from 11.111) R UI -iVI =U01- Z2) \\ . (4.2.31) t Taking R as reference length and putting z = exp(iO) it results u = - cos 2O , y _ -sin 2O (4.2.32) In figures 4.2.3 and 4.2.4 we compare the numerical values obtained for u by mens of (4.2.15) and (4.2.18) with the values obtained by means of (4.2.32). We observe that the solution obtained by means of the distribution of vortices is better. It is also useful to compare the numerical values obtained on the three ways for a quantity of aerodynamic interest. This is the local pressure coefficient defined by the formula A P. Cy _ = (1/2)P..oUx , (4.2.33) with the notations from (4.1.2). Taking into account that for the incompressible fluid, the system (4.1.7), for which we have determined the approximate solutions, is the exact system of the equations of motion, we shall calculate also Cn exactly. This may be obtained from Bernoulli's integral 1 V2 Pi 1 r r2 . Poo _ (4.2.34) Poa 2 Poo It follows Cp=1-V2=1-(1+ u)2-t'2. (4.2.35) Using this formula, from (4.2.15) we get C,, in the case of the distribution of sources, from (4.2.18) we get. C,, in the case of the distribution of vortices and from (4.2.32) we get C,, exactly. We also calculate Cn with the direct method from the following subsection. The graphic rep- resentations show that the direct method gives better results than the indirect methods and among these ones, the method based on the distributions of vortices is better. INDIRECT METHODS FOR THE UNLIMITED FLUID CASE 121 t 1. I es m As .03 .1s .1s a .1 a n 6e m IJI If Im )m >r )10 1L ))a us a W b b Ip Ip 110 7q 30 M RD 30 No Fig. 4.2.3. 4.2.6 Fig. 4.2.4. The Elliptical Obstacle The profiles with smooth boundaries (for example the circle or the ellipse) are non-lifting profiles. The profiles with angular points determine a circulatory motion of the fluid and therefore it appears the lift. The circulation (and the lift) are determining from the Kutta-Joukovsky condition (see for example, (1.101 p. 179), hence these profiles are lifting profiles. In the first case we encounter d'Alembert's paradox. As we shall see in the following subsection, the indirect methods are utilized mainly for non-lifting bodies, while the direct formulation is suited for writing the k - J condition. It is well known the exact solution for the elliptic obstacle in the incompressible fluid. The complex potential from the (Z1) plane is given by the formula (see for example, (1.101 p.178). F(zi) = 2 {( zi + za'°+ zl + (zi ' at + bt e1°l (4.2.36) t at - bj where at and bl are the semiaxes, 4 = at - b1, and a is the angle of attack. Deriving (4.2.36) we deduce the complex velocity which has the same form in dimensionless variables z = Z, a, b, c . Hence it follows 1+u-iv= ]e`1°+ 2[(1+ (4.2.37) +(l ` z a+bei°l a - b 122 THE BOUNDARY ELEMENTS METHOD We determine the components u and v separating the real part from the imaginary one. The graphic representations for the exact solution and for the solution obtained with the first indirect method, are given in [4.7J. We notice that the results are almost identical. 4.3 The Direct Method for the Unlimited Fluid Case 4.3.1 The representation of the solution As we have specified in 4.1, in the framework of the direct method we represent the velocity field in the fluid with the aid of its values on the boundary of the body and then we pass to the limit. In order to obtain this representation, we shall consider a domain D exterior to the boundary C and limited by a circle CR having the radius big enough, such that C should be situated in the interior of CR. For f and g continuously differentiable function C', we have the identity [f (aulax + avlay) + 9(8v/ax - 8u/8y)Jda = 0 (4.3.1) which follows from (4.1.7). Noticing that fau/ax = a(fu)/ax - ua f /ax, .. . and applying Gauss's formula, we obtain the identity J[u(Of/ax - 89/8y) + v(8f/ey + 09/8x)Jda = (4.3.2) = j[u(fn - 9ny) + v(9nx + f ny)Jds . the integral on 47R vanishing when R - oo, because of the behaviour at infinity of f and g which will be identified with u' and v' and because of the condition (4.1.11). For f = u' and g = -v' (4.1.13), taking (4.1.12) into account, we deduce: u(f) =JC u(x) [u'(x, )n (x) (4.3.3) +v(x) [u'(x, Ony(x) - v'(x, )n=(x)] }ds , THE DIRECT METHOD FOR THE UNLIMITED FLUID CASE 123 and for f = v' and g = u' 2L(x) [v`(x, S)ns(x) C (4.3.4) +v(x)[u'(z,4)n=(x) + v'(x,C)ny(x)]}d8. This is the integral representation of the solution, valid for every point M(4) from the exterior of C. 4.3.2 The Integral Equation For obtaining the boundary integral equation, we have to pass to the limit in (4.3.3) and (4.3.4), letting the point M(r;) to tend to the generic point Qo(zo) belonging to C. To the limit, the integrals containing the fundamental solution (u', v') become singular. Hence we have to adopt the definition (4.2.4). We calculate the integrals on c, as follows: lim Ju(z) [u*(x,F)n (x) + v*(x,C)nyr(x)]ds lim J [u(x) - u(xo)] [u'(x,4)n=(x)+ (4.3.5) +v (x, 4)nr(x)]ds + u(xo)Li L 1 = lim t-M. 1c [u' (x, , where v`(x, 4)ny(x)]ds . If c = is replaced by the segment Q'1Q'2 (fig. 4.2.1), then we Z have the parametrizations (4.2.6) and (4.2.7) and n = no. Hence, L= 27r re s2 + ds = 2 . Passing to the limit, the first integral from the right hand side of the equality (4.3.5) vanishes when e -+ 0. 124 THE BOUNDARY ELEMENTS METHOD Similarly we have: Eli m J v(x)[u*(x,Lr)nv(x) - v`(x,t)nt(x)]ds _ slim J [v(x) - v(xo)] [u`(x, )ny(x)-v*(x, t)nz(x)]ds + v(xo)L2 , where L2 = lim = (4.3.6) J[u*(x)n(x) +v(x,)n(x)]ds = ny - -r,0, no.) r s2 it + ds = O. 172 Therefore, from (4.3.3) we deduce: 2u(xo) _ {u(x) [u*(x, xo)ns(x) + v*(x, xo)ny(x)]+ (4.3.7) +v(x) [u* (x, xo)nv (x) - v* (x, xo)nz(x)] }ds . In the same way, from (4.3.4) we obtain: 2V(xo) {u(x)[u*(z,xo)nz(x) - u*(x,xo)nv(x)]+ (4.3.8) +v(x) [u* (x, xo)n=(x) + v*(x, xo)ny(x)] Ids. The formulas (4.3.7) and (4.3.8) constitute a system of two singular integral equations on C for the unknowns u(xo) and v(xo). Now we introduce the function G = (13 + u)ni, - vnx . (4.3.9) We may utilize on C the condition (4.1.10). It follows Q+u= n2n2G, v= -nz v 2 2C. (4.3.10) nv Replacing these relations in (4.3.7) and (4.3.8) and taking into account the relations PC. (u* n,, + v*nv)ds = -1/2; P(u*nv - v*nr)ds = 0, (4.3.11) THE DIRECT METHOD FOR THE UNLIMITED FLUID CASE 125 which will be demonstrated in (4.3.7), we get: u ( xo ) = ,3 + 2 P C (v ' n,n V u ')Gd s - M2 n2 + #2n2 x y 4.3.12) u (xo )=2TS and then G(xo) - 2 \ - u '-A12 n2+ 2n2v ' Gds x y ) I PC + hf2 [u'nox + v'rty+ n22 nzny + g2nyh (v n o ' - u ' ny° )JGd s = 2An y , (4.3.13) with the notation no = n(xo). The equation (4.3.13) is the singular integral equation of the problem. Putting M = 0, one obtains the equation for the incompressible fluid. 4.3.3 The Circulation In the case of the profiles with angular trailing edge, we have to utilize the circulation. Taking the sense defined on C1 into account and denoting by stl) and n(') the (dimensional) versors of the tangent and inward normal, we have 41), sod = -11y1), 41) = (4.3.14) such that rl = V1 si'idwi' (U1nyll - Vinzll)ds(') = = -U0LoP[(1 + U)Ny - VN,;Jds. With he changes of variables (4.1.6) and (4.1.9), we deduce [' 1 = - U 0 ° j[(0 + u)ny - vn=Jds. (4.3.16) Denoting r1 = U,,,Lor, r representing the dimensionless circulation and utilizing (4.3.9), it results the simple formula: r = - PCds. (4.3.17) For the incompressible fluid one obt ins: r=- G ds. (4.3.18) These formulas also emphasize the significance of the function G. THE BOUNDARY ELEMENTS METHOD 126 4.3.4 The Discretization of the Equations Acting like in 4.2.3, we reduce the equation (4.3.13) to the algebraic system N i = 1, 2, ... , N , (1/2)G{ + F A;jG,i = J3;, (4.3.19) j=1 where Ci = C(x°), A = Onv(x°), A,j = -nh(x°)Utj - n,(x°)V J-- (4.3.20) nz(x°)n1r(x°) n2(x') nr(x°)V, -- nv(x°)U,j] _M2 n2(xj), Uy and being defined (4.2.14). After determining the unknowns G1 from (4.3.19), we may determine the circulation from the formula: N -/3I' = E 1jGj , (4.3.21) j=1 representing the length of the segment Lj. We deduce the components of the velocity 1j + u; = A2ny(x°)G= n2(x°) + A2n2(x°) ns (x°)Gi n2(xe) + n2(x°) ' (4.3.22) from (4.3.10). Obviously, for n(x°) and n(x°) we shall utilize the expressions (4.2.25) and for U;j and [;j, (4.2.24). 4.3.5 The Lifting Profile If the profile has an angular point (fig. 4.3.1), or a cusp, like the Joukovsky-type profiles, then we have to determine the circulation using the Kutta-Joukovsky rule. The circulation determines the lift. For determining the circulation, we impose the equality of the pressure in the points P, (on the upper side of the angle) and P; (on the lower side of the angle) when THE DIRECT METHOD FOR THE UNLIMITED FLUID CASE 127 Fig. 4.3.1. these points (Pp and Pi) tend to the angular trailing edge P1. From Bernoulli's integral it follows that the equality of the pressures implies the equality of the tangential components of the velocity, (according to the slipping condition the normal components are null). Taking into account the orientation of the versor of the tangent s(1) on the upper and lower sides of the angle (fig. 4.3.1), we deduce the condition: Vi(P,).stt}+V1(Pi),alt} =0. (4.3.23) With the notation (4.3.9), we have: V1 slit = U)sal + VsY)J = UM +u J"'J 13 +vn=1ds pJdS d5 such that (4.3.23) implies G(Pi) = 0. (4.3.24) Consequently, for such a profile, we have to add the condition (4.3.24) to the system (4.3.19). In order to solve this problem we introduce [4.18J the regularization variable A and we obtain the system: A+(1/2)Gi+AijGj =;3i, i = 1,...,N, (4.3.25) G(P,) + G(Pi) = 0, for the unknowns C1,. .. , Gtr , A. in this way the number of equations equals the number of unknowns. For big values of N we obtain a small A i.e. (A - 0). THE BOUNDARY ELEMENTS METHOD 128 4.3.6 The Local Pressure Coefficient As we have seen in 4.2.5, the local pressure coefficient for the incompressible fluid (4.2.33) is (4.2.35). Let us deduce the expression of the local pressure coefficient for the compressible fluid. For the potential flow we have the formulas 11.101 page 206 which become, using the notations from this chapter: Vi ;fir Pi = Pu 2 Pl-potl-ry-1 l C? : c2o V12\i (4.3.26) J 2 _co2t1-72 For the uniform flow that we have in view, these formulas give 7-1UW 7-l 2 cam) 7-1 U2 Pco - Po 2 T (4.3.27) co =cfltl-7-1 U" cclo 2 czo From the last relation we deduce k ,y-1 1 Uz 7-tlli2+ 2 (4.3.28) such that, denoting r=72142, (4.3.29) it follows 1 k(7 - 1) =1+1. z (4.3.30) Introducing the parameter k in the formulas (4.3.26) and (4.3.27), we THE DIRECT METHOD FOR THE UNLIMITED FLUID CASE 129 get: P1=PO I- V"k 00 \ r (4.3.31) Poo=P0(1-721k}T , 7-1 Poo = Po 1-2 k) T These formulas will be replaced in (4.2.27). Taking out the factor (7 - 1)k/2 and utilizing (4.3.30), it results _ 2 2 ( )]'-T-11. (4.3.32) Cp = j -- 1+ ry 2 1 M2 ( U2 .yM2 For the incompressible fluid (M = 0) one obtains the formula (4.2.29). For the compressible fluid, V12/U.2 from (4.3.32) will be replaced by (1 + u/#)2 + v2. We shall employ the local pressure coefficient in order to define the lift coefficient which gives a global measure of the action of the fluid against the profile CL = 'f Li Cl,nylidat , (4.3.33) where Ll is the length of the contour C1. It is also important to get the expression of the critical velocity v, because the subsonic character of the flow is maintained only if Vl < vi,, . This velocity is obtained from (4.3.26) putting vl = cl and utilizing the definition of k. We obtain: U02 4.3.7 11+721 (M2 -1)] . (4.3.34) Appendix In this subsection we shall prove the formulas (4.3.11). To this aim, we denote by C, the arc (semicircle) of the circle of radius e with the center in Q0, interior to the profile (fig. 4.3.2) and we notice that we have the parametrization x-xo=Ecos9, y-yo=EsinO (4.3.35) 00<0<00+ir THE BOUNDARY ELEMENTS METHOD 130 Fig. d.3.2. which, for u', z}' (4.1.13) implies: U* (x, xa) 1 cos 0 2;r v' (x, x) E 1 sin y 27,, (4.3.36) E We also consider the circle CR with the center in Qo, and the radius R big enough, such that the profile C is in the interior of the circle. On this circle we have the parametrization x-x0=Rcos4 , ,- such that 9l' lcoscp = in R y-yp=RsincF, V* (4.3.37) 1Sill 27r R In the domain D. exterior to the contour C - c + Cs and interior ' to the circle CR, we may integrate the equation: au'/ax + av'/(9y = a(.r - x0, y - yo) (4.3.38) Applying Green's formula we get: (u'n., + v'ny)ds + Prr (0n, + v'ny)ds+ (4.3.39) + (u'n2 + Ti'ny)ds = 1. THE AIRFOIL IN GROUND EFFECTS 131 or, PC lu*(x, xo)ns(x) + v'(x, xo)ny(x)jds + 1 + 2n f"' o 2a 1 dO +- j dip = 1. , 0 One obtains the first of the relations (4.3.11). Integrating on D the equation 8u'/8y - 8v-/8x = 0, one obtains the second of the relations (4.3.11). 4.3.8 Numerical Determinations In figure 4.3.3 we compare u calculated by means of the direct method with u calculated exactly. We observe that the values are almost the same. In figure 4.3.4 we give 4 graphic representations: u exact, u direct, u sources, u vortices. We notice that u obtained by means of distributions of sources (4.2.15) does not give very good results and u obtained by means of distributions of vortices (4.2.18) gives better results. The values of u obtained with the direct method are constantly in the vicinity of exact values of u . Is tS as a3 .1 Ti rr-FT lTF 43 TG as a a o IN laI10us 110 210 210 010 110 a to .a 10 Ib ISO to ]b M 7s Im in IM Fig. 4.3.3. 4.4 4.4.1 Fig. 4.3.4. The Airfoil in Ground Effects The Representation of the Solution When the airplane is landing or taking off we have to take into account the influence of the ground. We shall consider this problem herein '1'm; BOUNDARY ELEMENTS METHOD 132 Y; x> Fig. 4.4.1. replacing the ground with the x10:1 plane (fig. 4.4.1). We have the same problem like in the previous sections. Art uniform subsonic stream, having the velocity U,,,, the pressure p and the density p", flowing in the yt > 0 half-space, is perturbed by the presence of an infinite cylindrical body, having the generatrices parallel to :r, IOzt, the Oxt axis having the direction of the stream. The x10y1 plane determines a cross section whose boundary C1 is assumed to be a thin profile. Using the smite notations like in 4.1 we have to integrate the system (4.1.7) in the ha lf-plane y > 0, with the boundary condition (4.1.10) on C and the supplementary condition zr(x, 0) = 0, (4.4.1) (V)X. In this case, instead of the fundamental solution (u', v') from 4.1.3 we shall utilize Green's distribution (U', V') determined by the system au; 119X l aVV/ay = 6(x - y -11) (4.4.2) (9DU/ay = 0, I) representing a point front the y > 0 half-plane, and the boundary condition U'(a,0) = 0. V+(r.0) = 0. (V)x. (4.4.3) This distribution is obtained by means on the fundamental solution (4.1.13) using the method of images. Denoting 2r, (x - )2 + (y - rj)2 (4.4.4) vt (T, y, , tl) _ 1 27r (x y-rI - )2 + (y - 1)2 133 THE AIRFOIL IN GROUND EFFECTS we have Uf = u*(x,11,Cq) f u*(x,y,C-q) (4.4.5) Vf = v'(x,y, ,q) ±v'(x,y,C -q) Indeed, u' (x, y, t, q) and v'(x, y, t, q) verify the system (4.1.12) for ev- ery t in the xOy plane. Obviously, u'(x, y, t, -q) and v' (x, y, t, -q) verify the system 8u'/8x+ 8v'/811= b(x - C y +q) (4.4.6) 8v'/8x-8u'/Oy=0. When t is in the superior half-plane (y > 0) this system becomes 8u'/8x + 8v'/ay = 0 (4.4.7) 8v'/8x + 8t,'/8y = 0. With these specifications one easily finds that (4.4.5) verifies (4.4.2). It is also very easy to verify the conditions (4.4.3). Let us return now to the representation of the solution. Denoting by D the domain from the superior half-plane y > 0 which is exterior to C and delimited by a half-circle CR with the center in the origin, the diameter on Ox and the radius big enough in order to contain C in the interior (fig. 4.4.1), we deduce like in 4.3, for R - oo, the identity [u(8f 1,9x - 8g/8y) + v(8f/8y + 8g/8x)] ds = (4.4.8) IL) =: j [(u f + vg)n= + (v f - ug)ny] ds + J u(x, 0)9(x, O)dx. 0000 We took into account the fact that on the diameter of the half-circle we have n = (0, -1) and that the integral on CR vanishes for R - oo when f and g are distributions having the form U., V. Putting now f = U+ and g = -V;, from (4.4.2) and (4.4.3), we deduce u( ) = J (u(x) [U+(x, 4)n1(x) + V+(x, )ny(x)] + c +v(x) [U+(x, C)n,(x) - V+(x, 4)ns(x)] Ids. (4.4.9) THE BOUNDARY ELEMENTS METHOD 134 In the same way, for f = V. and g = U , from (4.4.9) we obtain vW = PC {u(x) [V -*(x, F)nr(x) - V* (4.4.10) +v(x)[U-w (x, )ns(x) + V`(x,Ony(x)]}ds. Formally, the solution is identical to (4.3.3) and (4.3.4). It differs only by Ut, V}. 4.4.2 The Integral Equation Acting like in 4.3, we deduce f (1/2)u(xo) = PC {u(x)[U+(x,xo)n:(x) + V+ (m, xo)ny (x)] + +v(x)[U..(z,xo)ny(x) (1/2)v(xo) _' - V+(x,xo)ny(x)]}ds, (4.4.11) {u(x) [V=` (x, xo)n: (x) - U* (x, xo)ny(x)] + +v(x) [U' (x, xo)n=(x) + V! (m, xo)nv(x)} }ds. Utilizing the identities PC (U}nt + V.ny)ds = -2, P(U_ny - V'n=)ds = 0 (4.4.12) which will be demonstrated at the end of this section, we obtain for G defined by (4.3.9), the following integral equation: (1/2)G(xo) - PC {U (x, xo)n.(xo) + VV (x, xo)ny(xo)+ +1bllni(z)+ (ny,x) [V_(x,xo)n=(x)- (4.4.13) -U, .(x, xo)n.,(x)] }Gds = Qnv(xo), where the mark "prime" has the same significance like in (4.2.4). The formulas (4.3.10) which determine u and v in function of C, remain valid. 135 THE AIRFOIL IN GROUND EFFECTS 4.4.3 The Computer Implementation The equation (4.4.13) is discretized like in 4.3.4. One obtains the system N (1/2)G; + i1 A;}Gj= B;, I = 1, ... , N , (4.4.14) where AU = -n,,(x°)U,j o o -M2ny(x')+Q2n2(xq)( (x)V (4.4.15) + with the notations U, = J Lf U; (Z' x0)ds , V j = j V(x, x9)ds, (4.4.16) s B; remaining unchanged. The system (4.4.14) will determine the unknowns C1,.. . , GN. From (4.3.22) it follows u;, v;, (i = 1, ... , N). The circulation is obtained with the formula (4.3.21). For lifting airfoils, we also consider the Kutta-Joukovsky condition (4.3.24). We solve the problem introducing the regularization variable A and acting like in 4.3.5. In order to obtain the expressions of the unknowns U, j and Vtj*, we use the notations (4.2.22) and the relations Y= (xs - x°)(x2 - x1) + (y° + y°)(y2 - y1), c1= (xi -x°)2+(y;+y°)2 Ik =!-i1 (4.4.17) ate+bit+ ' In this way one obtains (z-x°)2+(y+y?)2=ate+b't+c', 4[(xjo -x0)(Io U,3 Vi'* = -Lj 41r kla)+x22x1(Ilfli)f 12 yl (Il f li)J [Y-v?Iodv?+Y?I6+ (o (4.4.18) THE BOUNDARY ELEMENTS METHOD 136 n(x0) and n(xti) being defined in (4.2.25). For It) and It one obtains the expressions (4.2.28). Since A = b2 - 4ac' (?1j + Y9)(x2 - XI) - (xj - xe))(?M - x!1)]2 < 0, it follows 4, = a 2Qr arctan , C' -a (4.4.19) 1t=2a III +b+c'-afwhere ,arctar. -© c' - a , a+b'+c' = (>'2-x°)2+(J2+J9)2 (4.4.20) a -b<+cl = For i = j the integrals Io and It become singular and they have the values (4.2.30). 4.4.4 The Treatment of the Method The problem presented in this subsection was studied in (4.10] and we present with small modifications the solution given in that paper. For testing the method we use the exact solution which is available for the circular obstacle in ground effects (fig. 4.4.2), (4.12]. If the center of the circle is in the point (0, b) and the radius is a (a < b), then, employing the usual notations, we have the exact solution: u - iv= -2 E Z2 C" + n>1 where the sequences +4 E n>t cn b» Z2 + b2 (4.4.21) and {cn} are defined by the following {bn} formulas: 2 2 b"+1=b-b+a c,++1=(b+bC11 b,,, n)2n--0, > (4.4.22) b1=b, c1=a2. Decomposing (4.4.21) in simple fractions, we get at - iv = - E Gn [(z - bn)2 + n>1 i1 (1 (4.4.23) + ibn )2} THE AIRFOIL IN GROUND EFFECTS 137 YA U- Fig. 4.4.2. whence, separating the real part from the imaginary one, we deduce u = - L'` n>1 x22-(y-b,,)2 + x2-(y+bn)2 f x2 + (y - bn )212 [x2 + (y + b.)212 (4.4.24) v= -2xEG& rx y - b + (y - bn)212 + n>1 y +bn Jx2 + (1/ + bn)?12 Putting r=acos8, y=b+asinO, we find uIC and vJ,. Using the formula (4.2.35), we represent in figure 4.4.3 the exact solution by a continuous curve and the approximate solution (in the case of the incompressible fluid, rt = 0) by small squares. We considered r = 0, a = 1 and b = 1.1. Since the two solutions practically coincide, we conclude that the method presented herein is very good. It may be applied for every shape of the airfoil. 4.4.5 The Circular Obstacle in a Compressible Fluid In this case there is not available any exact solution, such that we have only the information provided by the BIEh1. We have to take care to use THE BOUNDARY ELEMENTS METHOD 138 4 C, 0 -4 .8 1 -12J ace METHOD EXACT -16.J -20 J -24. r 0 90 180 I 270 360 Fig. 4.4.3. for the radius of the circle values that are small with respect to the other characteristic lengths (for example, the distance to the ground) because only in these circumstances we can employ the linearized equations. In figure 4.4.4, we represent the coefficient CL defined by (4.3.33) for the circular obstacle with the radius a = 1, considering 60 nodes on the boundary, . Obviously, the compressibility determines the increase of the lift (as we have already seen in the case of the free fluid). We considered 7 = 1.405. Giving the distance to the ground, we may calculate (in the absence of the circulation) the values of M for which the flow remains subsonic (one utilizes vQ given by the formula (4.3.34). 4.4.6 Appendix In order to prove the formulas (4.4.12) we take into account that in the entire plane, the solutions (U7, VV) satisfy the systems aUU/ax+av}/ay=a(x-t,y-17)± 5(x-C y+ ii) (4.4.25) aVV/ax - aUl/ay = 0, where 7> 0. Like in 4.3.7, we denote by Cf the arc of circle interior to the profile C, having the center in Qo(zo) and the radius a and by CR the THE AIRFOIL IN GROUND EFFECTS 139 000C, -0.02 -0.01 -0.06 -0.08 -0.10 -012 -0.14 -016. -02000 10 0.5 1.5 2.0 30 2.5 3.5 H Fig. 4.4.4. circle having the center in Qo and the radius R big enough, such that the profile C belong to the interior of the circle. The domain exterior to the curve C - c + CF and interior to the circle CR will be denoted by D. We shall integrate on this domain the first equation from (4.4.25) and and we shall apply Green's formula. One obtains: 2 = P[U+(x, xo)nr(x) + V. (x, xo)nv(x)Ids+ + find P (U+r1r + V. n.,,)ds + -00 f)(U*nr +V. nv)ds . 1z (4.4.26) With the parametrizatiou (4.3.37) we deduce on CR: :c - :r.0 COs (r-xo)2+(y- ))2 - R x-x'0 _ (x - xo)2 + (y + yo)2 Cos / ' _ T. - xU R2 + 411o(y - yo) + 4Jo I 1 +4sin R z R +42) 1 = Cos +O(R `) whence, u+ = RV - + O(Rr2), V+ si + O(R--) 140 THE BOUNDARY ELEMENTS METHOD V. = O(R-2) . Utilizing the parametrization (4.3.35) on C1 we have for example U! = O(R-2), y-yo+2yo y+yo (x-xo)2(y+yo)2 (x-xo)2+(y-vo)2+4yo(y-1m)+4y 2 (,+,,sin 2yo j (1 + Esyoe + 2yo 2 bI + 0(c). Hence, Uf _ 1 =tar C0619+0(6), s _ V1 sine 1 e + 4Ayo +0(6). Taking into account that on C. we have d s = Ed 9 and on CR, d s = Rdw, we get: 6o+x lim (U+n= + V nv)d s = PC. limo PC tar , (U..n + V+ny)d s= 1 d9 2t I f0 de = 2. Replacing all these in (4.4.26) it follows the first relation (4.4.12). 4.5 4.5.1 The Airfoil in Tunnel Effects The Representation of the Solution Assuming that the airfoil is situated between two planes parallel to the direction of the unperturbed stream (fig. 4.5.1), we have to integrate the system (4.1.7) in the domain exterior to the contour C and bounded by the two planes, with the boundary conditions (4.1.10) and the supplementary condition (4.5.1) v(x, 0) = 0, v(x, a) = 0 (V)x, Denoting by D the domain of motion delimited by the sides x = fd (fig. 4.5.2), we deduce for d - oo the identity ID [u(Of I8x - 89/8y) + v(8f/8y + 89/&)I da = = JR/u + gv)n, + (f v - 9u)nv)ds+ +J +00 00 q(x, 0)u(x, 0)dx 00 g(x, a)u(x, a)d x . +00 (4.5.2) 141 THE AIRFOIL IN TUNNEL EFFECTS Fig. 4.5.2. We took into account (4.5.1) and the fact that for x = td the integrals vanish when d --+ oo if f and g behave like the solution of (4.1.13) and it and v have the property (4.1.11). For f =1.t and g = -V+, where (17+, V+) verifies in the domain occupied by the fluid, the system (4.1.12) and the relations V+(x, 0) = 0, V+(x, a) = 0, (4.5.3) we get from (4.5.2): u(4) = lc 1u(x)[U+(x,01=(x) (4.5.4) +v(x) [U+(x, 4)ny(x) -V+(=. F)n7(x)] )d s. Analogously, for f = V_ and g = U_, where (U_,V_) verify the system (4.1.12) and have the supplementary property U_(x,0) = 0, U_(x,a) = 0, (4.5.5) we get from (4.5.2): v() _ (4.5.6) +v(x) [U-(x, t)rns(x) - V-(x, Ony(x)] }d s. 142 THE BOUNDARY ELEMENTS METHOD Hence we obtain for u(x) and v(4) the same representation like in (4.4.1), with the difference that (UU, V V ) become (U_+, V ). 4.5.2 Green Functions The system (4.1.12) determines the fundamental solutions of the system (4.1.7). These are distributions. The fundamental solutions which are defined only in a portion of the plane and satisfy the boundary conditions are named Green functions, or, more properly, Green distributions. The solutions (U+,V+) and (U_,V_) are therefore Green functions. For obtaining these solutions we employ the method of images. If M(t) is a point of the strip -oo < z < +oo 0 < y < a (fig. 4.5.1), then the solution (without restrictions) of the system (4.1.12) is (4.1.13). As we already know, it represents the perturbation produced in the entire plane xOy by a source located in M. For satisfying the first condition of (4.5.3), we have to place a source having the same intensity (taken to be equal to the unity) in the symmetric point M(t, -ri). For satisfying the second condition of (4.5.3), we have to place sources having the same intensity in the points which are the symmetric points of M and M with respect to the y = a axis, i.e. in the points Mi (t, 2a - rl), Mi (t, 2a + q). These sources disturb the condition on y = 0, such that we have to place sources in the symmetrical points of Mi and Mi with respect to the y = 0 axis, i.e. in the points M+1 (t, -2a + i) and -2a - ,) etc. The perturbations produced by these sources are x-t 1 + V(")+ 27r (x-t)2+[(y-(2na±q)12 1 y-(2nat9) (4.5.7) 27r (x - t)2 + [y - (2na ±, )12 and they satisfy the system OUP }/Ox + OV n)/Oy = a(x - t, y _ (2na ± ii)) (4.5.8) OVA" i /Ox + OD'") /ay = 0. In the strip -oo < x < +oo, 0 < y < a, all the perturbations satisfy the homogeneous system, excepting the perturbation (Uo , V+) THE AIRFOIL IN TUNNEL EFFECTS 143 which satisfies the corresponding non-honwgeneous system. lI+(x, ) = 2;r (x' + [y - (2nn + 11)]' + (4.5.9) r: I 'V-4' { (r =-00 y - (2nn + rl) } (x - )2 + [y - (2nn + n)]2 2,r n=-00 y - (2na - r1) [y- (2na-17)]2 +{:r. satisfies the system (4.1.12) and the boundary conditions (4.5.3). It is just the solution utilized in (4.5.4). In the same way we deduce that the solution U-(x, ) = 2;r { (x - C)2 + [y - (2na + 11)12 - x-t (4.5.10) )-2, L I n=-1 (2na + 11) (a-ti)2+ [y-(2na+1))]2+ y-(2na-11) + (x - )2 + [y - (2na -11)]1 satisfies the system (4.1.12) and the boundary condition (4.5.5). This is just the solution utilized in (4.5.6). The series (4.5.9) and (4.5.10) are the Green functions of the problem. The sums intervening in (4.5.9) and (4.5.10) may be calculated using the formula = ircot az, (4.5.11) where z = x + iy. Separating the real part, from the imaginary one, we 144 THE BOUNDARY ELEMENTS METHOD get from (4.5.11): +oo y y2 + (x - n)2 = 7r sink 27ry cosh 2iry - cos 21rx (4.5.12) +00 0 x-n sin 2irx n=-ao y2 + (x - n)2 = A cosh 2iry - cos 27rx Using these relations we obtain the expressions of the Green functions: sinhQ(x- (x - ) - cos Q (y - q) cosh Q 71 f=4n cosh 1 sinh -(x - .) a (x - ) - cos a (y + +)) (4.5.13) 7r _ V f (x, ) = sin -(y - q) n 1 sin 7;(y cosh Q(x - e) - 4.5.3 + A `a +,q) 7r(y + n) The Integral Equation In the sequel we shall proceed like in 4.3.2, i.e. we pass to the limit -i xo E C in (4.5.4) and (4.5.6), we introduce the function G by means of the formula (4.3.9) and we utilize the relations +(x, xo)n.(x) + V+ (x, xo)ny(x)]d s = - 2 (4.5.14) lu-(x,xo)ny (x) - V-(x,xo)n.(x)1ds = 0, 145 THE AIRFOIL IN TUNNEL EFFECTS which will be demonstrated in Appendix 4.5.5. One obtains the following integral equation: (1/2)G(xo) - c {U_(x,xo)n(x) V+(x, xo)nt,(x)+ +M2ni(X) + f(X) [V-(x, xo)nx(xo)- (4.5.15) -U+(x, xo)n,(xo)]}Gds = Onv(xo) , and discretizing, we deduce the system N (1/2)Gi + E AijGj = Bi, j = 1, ... , N, (4.3.16) i.t where Aij is A!. from (4.5.15) with the notatipns U= f Dx, x°)ds, l= J V f(x, x°)d s . (4.5.17) ; .. In order to determine these coefficients we consider on Lj the parametrization (4.2.19). Further we have two possibilities: the exact calculus like in 4.2.4, or the approximate calculus, with Gauss-type quadrature formulas (A.55], for the resulting integrals +l Ut(t)d 1-' Uii = 2 t, Vj = 1_j t1 f V±(t)d t . 2 t (4.5.18) t In case that i = j, these integrals become singular for t = 0, because of the terms sinh a [x(t) - x?d t +1 t cosh a sin t ui [x(t) - x°] - cos'-` [y(t) - A a a (y(t) - y°) d t [x(t) - x°] - cos [y(t) - y°j cosh a a 146 THE BOUNDARY ELEMENTS METHOD but, taking into account that the integrands are odd functions, we deduce that they vanish. We have therefore 1. +1 8a i Uf=± ` }- `" - sink - [x(t) - x9 'd t a cosh [x(t) -;J - cos a [y(t) - y°] (4.5.19) a sin +e Co.), a [y(t) - y,)jd t n 7r [x(t) - x,)[ - cos [y(t) - yp[ a a These integrals may be also calculated numerically, using the Gauss-type quadrature formulas. 4.5.4 The Verification of the Method In order to perform this verification the shall use the solution given by the complex potential f (z) = z + Ctanh (4.5.20) z where C and a are real constants. Let us study the flow determined by this potential. Employing Euler's formulas we deduce the identities sinh iy = i sin y, cosh iy = cos y cosh 2 Ax cos2 Ay + sinh2 Ax sins Ay = 2 (cosh 2Ax + oos 2Ay), cosh2 Ax cost Ay - sinh2Ax sin Ay = 2 (1 + cosh 2Ax cos 2Ay), (4.5.21) such that, separating the imaginary parts, we get from (4.5.20): C sin 2Ay - y + cosh 2Ax + cos 2Ay ' (4.5.22) We denoted A = 7r/a. Obviously the straight lines y = 0 and y = a are streamlines ('P = 0 and 'I' = a). Another streamline is the line y = a/2(41 = a/2) but not entirely because for x = 0 the fraction from (4.5.22) is not determined. Further we shall study the velocity field. From (4.5.20) it results w(z) = u iv = I + AC cosh2 Az , (4.5.23) THE AIRFOIL IN TUNNEL EFFECTS 147 whence, utilizing (4.5.21), u. = 1 + 2AC 1 + ch2 Ax cos 2Ay (cosh 2Ax + cos 2Ay)2 (4.5.24) V = 2AC sinh2 Ax sin 2Ay (cosh 2Ax + cos 2Ay)2 Now we return to the streamline y = a/2. This line may be continued ('I' = a/2) on the set of points (x, y) where we have: a z = -y+ C sin 2Ay cosh 2Ax + cos 2Ay (4.5.25) This curve is symmetric with respect to the Oy axis, because the equation is even in the x variable. It is also symmetric with respect to the y = a/2 axis, because the points having the ordinates y = a/2 + yo and y = a/2 - yo, where _ yo C sin 2Ayo cosh 2Ax - cost Ayo (4.5.26) simultaneously belong to the curve. Hence the curve having the equations (4.5.25) is an oval which orthogonally intersects the axes y = a/2 and x = 0. In the point P(-a, a/2) we have therefore u = 0. From (4.5.24) it follows the equation sinh2An = AC, (4.5.27) for the half-diameter a as a function of C. The semi-diameter 6 on the Oy axis may be determined from (4.5.26) as follows _ C sin 2A,0 1-cos2Af3' whence it results a simpler equation $tan AQ = C. (4.5.28) We draw the conclusion that the complex potential (4.5.20) characterizes the uniform flow with the velocity (1, 0), in the channel having the sides y = 0 and y = a, in the presence of the oval having the diameters 2a and 2/3 with the center on the Oy axis (fig. 4.5.3). It is not difficult to see how this potential was obtained. One knows that the uniform flow at infinity, in the presence of a doublet, determines THE BOUNDARY LLPMENTSMEITHOD 148 y=0 L. H=8110 x=0 Fig the flow past a circular obstacle. Let us study now the circular flow in the presence of a doublet in a channel. For the sake of simplicity, we consider the doublet situated in the origin and y = a/2, y = -a/2, the equations of the lines which are the walls of the channel. These right lines become streamlines for the flow determined by the doublet, if we add, according to the method of images. symmetric doublets. The potential f0(z) describing the flow produced by these doublets is A(z)_...+ I z+ia +1+ +... 1 = _ro _ let -cot Trz Ia IT 1 Ir irz u a -coth Ave take fo(z) = CcothA>. A = 7r/a, C representing it real o constant. If the Ox axis coincides with the lower wall. as we have previously considered. then the potential becomes fo(z) = Ceoth A (:_ i a2) = Ct.unh Az . Adding the potential of the uniform flow having the velocity (1,0) we obtain (4.5.20). In the paper [4.9] that we utilized for writing this subsection, we made tests for the local pressure coefficient C,,, defined by (4.2.35). In figure 4.5.4 we present the values of Ci, calculated exactly (by means of (4.5.24) and the values calculated with the numerical method presented above with H = a/10. The maximum relative error is 0.73% in the case of a discretiration with 80 nodes. For the circular obstacle, symmetrically placed in a tunnel, we present the compressibility effect in figure 4.5.5 where one indicates the variation of the maximum velocity THE AIRFOIL IN TUNNEL EFFECTS 149 i- I -o- - na i \ 0 J Fig. 4.5.4. _ Fig. 4.5.3. against the distance to the walls for 'r = 1.405 in the cases M = 0 and M = 0.3 and in the absence of the circulation (F = 0). We considered 60 equidistant nodes on the circumference. In [4.9) one presents the graphic representations for C, in the case of the NACA-4412 profile at 0 angle of attack and Mach=0.5 in a free stream and in a tunnel. The example is instructive, because, in this case the trailing edge is angular, and we have to impose the equality of the pressures on the two sides of the edge in order to determine the circulation. 4.5.5 Appendix In order to prove the first formula (4.5.14) we shall notice that (U+, V+) (x, y, xo, yo) satisfy in the domain D, exterior to the contour C-c+C1, and bounded by the walls y = 0 and y = a and by the segments x = ±d, the equation: aU+/49x+aV+/Oy=6(x-xo,y-yo). Integrating this equation over D and applying Gauss's formula one obtains: J (U+nx + V+n1)d s + ii;t a 1c, + d im ra J0 (U+n" + s+ U+(d, y, xo, yo)d p + slim J t7+(-d, y, xo, yo)d p = 1, -oo J.a (4.5.29) THE BOUNDARY ELEMENTS METHOD 150 the integrals on y = 0 and y = a vanishing by virtue of the relations (4.5.1). The integrals on C6 are calculated utilizing the parauiietrizntions (4.3.35) which imply sin0+C, cos0+0(e), Uf= I I where C_ 1 2ryo yn 1 1 + a2 n=0 (yo/a)2 - n2 To the limit, when a -i 0, this is 1/2. One calculate the last two integrals taking into account that d t Id _ 2 (a. - b)tan (t/2) a -b a+bcost arctan a -b' To the limit, when d -- 0, every integral is 1/2. One obtains the first formula (4.5.14). Analogously, we obtain the second formula if we integrate the equation aTJ_/ay - 8TH-/ax = 0 on the same domain D. 4.6 4.6.1 Other Methods. The Intrinsic Integral Equation The Method of Regularization The methods that we have already utilized have been given by L. Drago§ [4.6] and by L. Drago§ and A. Dinu [4.7], [4.8], [4.9], [4.10]. Specific to these methods is the fact that one utilizes physical variables the velocity and pressure fields), such that from the solution one obtains directly the elements of interest in aerodynamics (the velocity and the pressure on the profile). Other methods (Biudolino a.o. [4.2], Morino and Luo [4.20]) utilize the real potential and others (Griello a.o. [4.15], Carabineanu [4.4]), the stream function. All these methods may be utilized only for incompressible fluid. The theory that we are going to present in the sequel will be also applied to the incompressible fluid, but it has the advantage to utilize the physical fields. Moreover, the singular integrals are avoided and one utilizes the intrinsic elements of the flow (the tangential component of the velocity). Therefore it is suitable to call this method the method of regularization [4.11]. OTHER METHODS. THE INTRINSIC INTEGRAL EQUATION 151 Denoting V = Un(i+v), (4.6.1) we shall determine the perturbation from the equations dive=0, rotv=0 (4.6.2) with the boundary condition (4.6.3) and the condition at infinity limn v(x) = 0. (4.6.4) We may write the equations (4.6.2) as follows div (v - c) = 0, rot (v - c) = 0, (4.6.5) c representing a constant vector. We put into evidence the normal and tangential components of the velocity: v = (v n)n + (v s)s = v,8. (4.6.6) By virtue of the condition (4.6.3) we may write v = -n?n + v,s. (4.6.7) Rewriting the scalar formulas from 4.3 as a vectorial formula we deduce that for every two continuously differentiable functions or distri- butions f and g , by virtue of the equations (4.6.5) (k - the versor of the Oz axis), we have: ID if div (v - c) + (gk) rot(v - c)]da = 0, (4.6.8) D representing, like, in 4.3.1, the domain exterior to C and interior to the circle Cn having the radius R big enough. Utilizing the formulas div f (v - c) = (v - c) grad f + f div (v - c) (4.6.9) div [gk x (v - c)] = (v - c) rot (gk) - gk rot (v - c) and applying Gauss's theorem, we obtain the identity: J(v -c) gradf-rot(gk))da= f(v - c) (fn-nxgk)ds, }- CR (4.6.10) 152 THE BOUNDARY ELEMENTS METHOD n being the normal (pointing outwards the domain D (i.e. the inward pointing normal with respect to C and the outward pointing normal with respect to CR). We shall write the distribution (4.1.13) as follows v,- 1 x - xn 2r, Ix - xo12 (4.6.11) and the system (4.1.12) that it satisfie : divv'=b(x-x()), rote'=0, (4.6.12) the equations being valid for every point xn from D + C. For (f, g) (u', -v'), one obtains from (4.6.10) the projection of the identity (v - c)divv'da = J {[n (v - c)]v + [n x (v -- c)] x v}ds, +CR JJJD (4.6.13) on the Or axis. For (f,g) (v'.u') we obtain the projection of the same identity on the Oy axis. Hence the relation (4.6.13) is valid. By virtue of (4.6.12) we obtain from (4.6.13) v(xa) - c = f {[n (v - c)]v' 4. [ri x (t+ - c)] x v}tl s. (4.6.14) TCu llere we shall put c = v(xo) - vo and we shall evaluate the integral on CJt. Utiliziug the parametrization (4.3.37), we shall deduce for the projection of this integral on the Ox axis xli {n - (v - vo)u' + (u - ua)(u'n7 + v'n,)2m -riI[(ei - 110W + (v - vo)v']}d3 = 1 2=r j (u - t!o)dV = -110, because for R , oo, u vanishes according to the condition (4.6.4). One deduces a similar formula for the projection on the Oy axis. Utilizing these results one obtains from (4.6.14) the following representation V(x0) = PC {[n (v - vo)]v' + In x (v - vo)] x v' }d s. (4.6.15) The integral is not singular because the factor v - vo tends to zero when x xo E C. This is a regularized integral. The formula (4.6.15) is valid for both xo in the fluid and on the boundary C. OTHER METHODS. THE INTRINSIC INTEGRAL EQUATION 153 We notice utilizing (4.6.7) that on C we have nxv=v,k (4.6.16) In - (v-vo)]v' = (-n=+ such that, after elementary calculations, from (4.6.15) one obtains V(xo) _ P(v,k x v' + v;((v' . s°)n-(n v')s° - (n s°)v'] - nxv'- (4.6.17) -79x((n° v`)n - (n . v')n° - (n n°)v']}d s. For obtaining the unknown v, outside the integral, we shall perform the operation no x (4.6.17) and we shall take into account that we have no_ soy, n o_ y - -soy, (4 6 18) no x v' = (so v')k (4.6.19) :- . . We deduce no x s° whence 11°= 4.6.20) ((s° . v')n' - (so v')ns]d s. This is the integral equation of the problem. It is a regularized (nonsingular) integral equation because when x - x°, the denominators and the numerators (n° v')v°) of the integrand simultaneously vanish (see (4.6.11)). The equation (4.6.20) was obtained in a different manner by V.Cardng (4.5]. It is, obviously an intrinsic equation. This equation may be also solved by means of the boundary elements method. One obtains, like above, a linear algebraic system of N equations with N unknowns. In the case of lifting airfoils the following boundary condition is added (see also (4.3.23)). -+Pf. (4.6.21) 154 THE BOUNDARY ELEMENTS METHOD Taking (4.6.1) and (4.6.18) into account, we apply this condition as follows: v. (P.) + v.(A) = ny(P.) + ny(Pi) , (4.6.22) P, and Pi being sufficiently close of Pf. From the equation (4.6.20) and the condition (4.6.22) we shall determine v, on C. 'Since we have more equations than unknowns, we shall treat the problem like in 4.3.5, introducing the auxiliary variable A. The reader may find examples in [4.11]. Chapter 5 The Theory of Finite Span Airfoil in Subsonic Flow. The Lifting Surface Theory 5.1 5.1.1 The Lifting Surface Equation The Statement of the Problem We assume that an uniform flow, having the velocity (the Ox axis has the direction and the sense of the uniform flow), the pressure p,, , the density p,, and Af(= U,./c,.) < 1, is perturbed by the presence of a finite span wing, perpendicular on the flow direction. Any body which has a characteristic dimension much larger then the two other dimensions is considered to be a wing. We call the span of the wing, the length of the wing taken along the direction of the large characteristic dimension (figure 5.1.1). Fig. 5.1.1. One requires to determine the action of the fluid against the wing. In fact. as we have already seen when we studied the two-dimensional case, for determining the action, one has to calculate the perturbation THE LIFTING SURFACE THEORY 156 fields p(x) and v(x). We employ the Cartesian variables x,y, z introduced in (2.1.1) and the fields p and v introduced by (2.1.3). The Oy-axis has the span direction and the origin 0 is situated in the middle of the wing. The Oz-axis is taken perpendicular to the Ox and Oy-axes in order to detennine a Cartesian, positive oriented frame. We assume that the projection of the wing onto the xOy-plane is a simple connected domain D, with a piecewise smooth boundary &D such that every straight line parallel to the Ox-axis (which has the direction of the unperturbed flow) should intersect the boundary in almost two points Q. and Q f. As we may see in figures 5.1.2 and 5.1.3, the lateral edges of the wing may be exceptions from this assumption if they are parallel to the Ox -axis . The intersection of 8D with a lateral edge parallel to the Ox -axis consists of a point, two confounded points or a straight segment. Assuming that the two lateral edges consist of a point (denoted respectively by B and B') we notice that they divide the boundary into two arcs The front arc BB' which is attacked by the stream is called the leading edge and the rear arc is called the trailing edge. The equation of the leading edge (consisting of the points Q,) is x = x_(y), and the equation of the trailing edge is x = x+(y). Hence Qa has the Cartesian coordinates (x- (y), y) and Qf (x+ (y), y). For wings with lateral edges consisting of simple or confounded points we obviously have: x+(±b) = x-(±b), (5.1.1) where 2b is the span (in dimensionless variables). We shall call these ones I wings, and the wings whose lateral edges consist of straight segments will be called II wings. For the II wings the condition (5.1.1) will be replaced by the condition of the continuity of the pressure along the edge (the Kutta-Joukowski condition (see (5.1.30)). In figures 5.1.2 we indicated the domain D for some I wings having lateral edges consisting of simple points (the delta wing, figure a), the gothic wing, figure b), the trapezoidal wing, figure c), the rhombic wing, figure d), the swallow tail wing, figure e) or lateral edges consisting of double confounded points (the elliptical wing, figure f)). In figure 5.1.3 we present the domain D for the arrow shaped wing (with lateral edges consisting of straight segments). The domain D is named the plane form of the wing or the planar form of the wing. 157 THE LIF"TINC SURFACE EQUATION b) a) d) f) e) Fig. 5.1.2. We notice that the wing has two surfaces: the upper surface, denoted by S+ and the lower surface denoted by S_ . The equations of these THE LIFTi G SURFACE THEORY 158 surfaces may be written (for the sake of simplicity) in the following form z = h(x, y) f hj (x, y) , (x, y) E D . (5.1.2) Indeed, considering z = f (x, y) the equation for S+ and z = g(x, y) the equation for S_, then, setting 2h(x, y) = f (x, y) + g(x, y) (5.1.3) 2h1(x,y) = f(x,y) - g(x,y), we obtain (5.1.2). B. Fig. 5.1.3. The wing is thin if and only if h and h1 have the form h(x, y) = A-(x, y), h1(x, y) = Fhl (x, y) , (5.1.4) E(<<1) being a real parameter (max h, h1 in D), and !t $i hl bounded functions. We also assume that h(x,y) and hl (x, y) are defined on D and have the first order derivatives with respect to x, denoted by h,,, hi=. The two surfaces S+ and S_ do not intersect each other if and only if h(x, y) + hl (x, y) > h(x, y) - hi (x, y). 5.1.2 (5.1.5) Bibliographical Comments The bibliography concerning this subject is extremely rids and it is not possible tp mention it entirely. In fact, this is not necessary 'ME LIFTING SURFACE EQUATION 159 because the fundamental solutions method, utilized herein is different from the previously utilized methods. As we have already mentioned in the introduction, the methods utilized so far rely on the equation of the potential (we assume that the perturbation is irrotational, the wing being assimilated to a distributions of bound vortices or doublets. Downstream, behind the wing, where the experience shows that the flow is not irrotational, one introduces in an artificial manner a free vortices distribution. The perturbation of the velocity results from the velocity field induced by the two distributions. This is Prandtl's model [6.21) for the lifting line theory, which assumes that the wing may be replaced by the segment BB'. The lifting surface theory, which does not replace the domain D with the segment BY, was developed after 1936 by Prandtl [5.26], (5.27), Weissinger [5.35], Reissner [5.29], Multhopp [5.24], Flax, Lawrence [5.12), Truckenbrodt (5.32), Mangler and Spencer (5.22), etc. We also mention the unpublished paper [5.6] of P.Cociirlan. Some synthesis of the research in this domain may be found for example in [1.1), [1.2], (1.38) and [1.41]. D. Homentcovschi [5.16] presents, for the compressible fluid, a theory relying on the equations of motion written in distributions, without assuming that the flow is potential. Another theory is given in (5.7]. Here one introduces the fundamental solutions method. It is not necessary to assume that the flow is potential and a vortices sheet is formed behind the the wing. These properties result from equations. 5.1.3 The General Solution As we have already mentioned in 3.1.3, there is no physical reason to replace the wing by a distribution of vortices. It is reasonable to replace the wing by a distribution of forces which should have against the fluid the same action as the wing. Indeed, the fluid and the wing must be considered as an interacting material system. According to Cauchy's tension principle (see [1.11), p.35), there exists a ford distribution f on the wing (or on D) which has the same action against the fluid like the wing itself. The perturbation of the fluid flow will be determined by this distribution and by the slipping condition on the surface of the wing. We try to employ a forces distribution on D, having the form f = (f1,0,f)(,t1) If we manage to satisfy the boundary conditions, on the basis of the uniqueness theorem, we deduce that we found the solution of the prob- 160 THE LIFTING SURFACE THEORY Obviously, there exist various distributions which enable us to reach the solution. In chapter 7 we shall employ distributions on the boundary of the body. From (2.3.24) we deduce that the perturbation of the pressure due 1em. to the force f applied in the point (t, q) is ft(c rt)8 + f(c n)8 u-.) , (5.1.6) xo=x-t,yo=y-v1, Ri= xo+(yo+z2), (5.1.7) P(x,IV,x) J where A continuous superposition of such forces on D, will give [fi(ii)- + f P(x, y, z) = -1 n) ez ] JJD since the problem is linear. For the potential jp, from (2.3.28) we deduce ( X y, z) = T 7r- JD [f (R1 ) -- uo+ z z2 (_) d t d n , + RI) f (t, n)1 d (5.1.8) dn (5.1.9) and for the velocity field from (2.3.12) v(x, y, z) = 6(z) Jf f(, t7)H(xo)b(yo)d d r< + Oip (5.1.10) which can be written explicitly: u(x, y, z) = 6(z) Jj h (, r1)H(xo)a(yo)d d r- p(x, y, z) v(x,y,z) = AD I fl , (f,n)' (i;) -f(t,rl)Z [y (i+9.)] }ddt 2 (5.1.11) w(x, y, z) = 6(z) IfD f (t, tt)H(xo)5(yo)df d n+ +4- fD f fi(t,n)g ()dd_ r1 -4Rffnf(F,n)az [1l0 z2 THE LIFTING SURFACE EQUATION 161 For w, from (2.3.29) we get: w(x,y,z) 4rr lID [fi(i); (RI} -Q2f(E,rl) CRI) JdC dq+ +4 AD f(C17)Zy (5.1.12) (l+ Xg Jo2+YO Ri)J z2 dtdn From formula (5.1.10) we deduce that the perturbation is potential, excepting the strip from the xOy - plane, -b < y < b, x > x_ (y), where the first term does not vanish. We notice that the existence of the vortices sheet behind the wing, which was a hypothesis in the classical theories (see Prandtl [6.21]), represents herein a consequence following from the equations of motion. As the experience confirms the existence of such a sheet, it means that the experience confirms the mathematical model based on the fundamental solutions method. 5.1.4 The Boundary Values of the Pressure In view of calculating the boundary values of the pressure when z -' ±0 and (x, y) E D, we notice that the first term from (5.1.8) represents the tangential derivative of a simple layer potential and the second, the normal derivative of the same potential. From the potential theory one knows (see for example, [1.6), [1.39]) how to obtain the boundary values of these derivatives but we prefer to calculate them directly. After deriving we have to calculate It limo f i (s, y) E D (5.1.13) =- lim z-'.±O JD (x, y) ED dd f(t,n) . 1 We notice that setting z = ±0 and (x, y) E D, Rl vanishes when the generic integration point Q(t, n) coincides with M(x, y) E D. The integral Il becomes singular. Denoting by DE the disk with the radius THE LIFTING SURFACE THEORY 162 E and the center h1, we shall write and we shall consider by definition , hm (5.1.14) U IJD 110 = JJ-D, +AD, if the limit exists. For calculating the integral on Dt, one makes the change of variable O7D, q -+ r, 0: - x = Or c os 0, tj - y = r sin 0 (5.1.15) 0<r<e, 0<0<21r and one notices that if f is a continuous function, for a small enough, we may approximate f on the disk with f (x, y). We deduce: AD, 1 and then, the formula dtj= fJ r2 cos f(xJ 1!) 12,, 0c ( f1 r2 +z 2)3/2 d9dr=0 =o1 f(e,r1)0dedt, R (5.1.16) D where R= Vxg+ (5.1.17) As it is shown in (E.0.12), the last integral exists. As regarding 12, it vanishes outside a vicinity DD of M. We have therefore I2 Zo : (x. y) E D .f(C,t]) z dCdti lID 74 2s =f li n) fo z Jo (5.1.18) rd 0d r = f2tr f (x, y) Jo (r2 + z2)3/2 (;, v) E D Taking into account these results, from (5.1.8) we deduce: p(x, y, f0) = - 0 ID f i (4, r]) ? d d of 2f (x, y), (x, y) E D. (5.1.19) This is the formula for the pressure on the upper and lower surfaces of the wing. We deduce the formula (5.1.20) p(x, y, +0) - p(2, y, -0) = f (x, y) , which gives us the significance of the function f (x, y). Like in the twodimensional case, f is the jump of the pressure on the wing. This is a fundamental result for the calculus of the aerodynamic action. THE LIFTING SURFACE EQUATION 5.1.5 163 The Boundary Values of the Component w If one utilizes (5.1.11), then taking the limit, when z -+ ±0, the first term disappears because 5(±0) = 0. It is more difficult to calculate the limit of the last term because it contains the derivative with respect to z. The calculation is performed in (5.221 and in 11.111, p.208, 209. We utilize here the expression (5.1.12) for w. In this expression, the limits of the first two terms are given by the formulas (5.1.17) and (5.1.18) and the derivative with respect to y from the last term commutes with the limit. We have therefore, lira 13 a8y yo + z I1 a z (1 + La Rl )] dC drl = (s, y) E D (5.1.21) 8 _ j22K(xty,z7ii)dt, y) ED +Cr, with the notation x+(7) K(x,y,z,r1)=1_ (5.1.22) 00 After setting z = 0 in the last expression of 13, we notice that we may not simplify with yo, because it vanishes for 77 = y. We shall divide the integral into three parts: J-bb _ 1b £+Jy+E+ JyyEa The last integral vanishes because for F small enough we may consider K(x, y, z, v7) = K(x, y, z, y) whence +E TY-C yoz K(x,y,z,rl)dn=-K(x,v,z,y) f +E +z =0. E u 2du2 For the integrals on (-b, y - s) and (y + e, +b) we may pass to the limit, setting z ± 0 and we may then simplify by yo, because it does THE LIFTING suitFACE THEORY 164 not vanish. It follows 13 = a Hen I I `J b 3y = line + r rb\ ! (r., rl, 0, 0 d rJ Jy+il 1C(x,'y, 0, y - E) l f +b\l ,,...s + t3 = Yo 1i (x, y. L' + \J-b + Jv+c/ ay \ yo / 0, y + `) (5.1.23) d tJ Expanding into a Taylor series the first two terms and taking into account the definition of the "Principal Value" (D.3.1) and the definition of the "Finite Part" (D.3.6), we deduce 3'/`' b 1 +1 tb h'(x, II _JJDf -7( o- `1+ Rddy+2 ayh(x,y,0,J1)di, _ JJ, f(e,iJ)x (5.1.24) Taking into account the expressions of 11,11 and 13 from (5.1.12) we obtain: u,(x, y, ±0) 5.1.6 = f fl (?', y) - 41, 1fn f (y, i) (l + ) d d r1 . (5.1.25) The Integral Equation F om the boundary conditions (2.1.33) and the equatiolLS (5.1.2) we deduce: ua(x, y, f0) = Jr=(3:. y) t h' (x, y) , (:r, y) E D . (5.1.26) Replacing the last relation in (5.1.25), subtracting and adding the two relations, the first corresponding to the upper sign (the condition on the upper surface) and the second corresponding to the lower sign, we get (5.1.27) 4 _ ly JJ n C1 yo \ + 0) d diJ = hr(x,y), R (x, y) E D. (5.1.28) THE LIFTING SURFACE EQUATION 165 The relation (5.1.27) determines directly f, (w, y) and the relation (5.1.28) constitutes an integral equation for determining the function f . This is the well known equation of the lifting surface. It is obviously a singular integral equation with a strong singularity. The mark * is for the "Finite Part". For the incompressible fluid, The equation has been written for the first time by Multhopp in 1950 [5.24]. A rigorous demonstration is given by Mangler in the Appendix of the above mentioned article. The demonstration is quite complicated because the author utilizes the representation w(x,y,z)=-47r (5.1.29) which follows from (5.1.11). Short deductions are presented in [1.1], [1.38]. For wings with lateral edges consisting from straight segments (figure 5.1.3) the equation (5.1.28) is integrated with the conditions f (x, ±b) = 0 (5.1.30) which, taking into account (5.1.20), means the continuity of the pressure (the Kutta-Joukowski condition). On the trailing edge x = x+(y), -b < y < +b (5.1.31) one imposes the boundedness condition for the function f (x, y). An equation equivalent to (5.1.28) was given by Lawrence and Flax in 1951 [6.11]. Formally, it can be obtained from (5.1.28) using the identity )= R (I+0 yo xo+R _ 8 rxo+Ro xpyo) 8y 1 \ xoyo 8 (5.1.32) It follows: 4v 1 A f( AC 17) (1+)ddtih'(xY)1 (x, y) E D. (5.1.33) Sometimes it is preferable to use this equation instead of (5.1.28), because it has weaker singularities (Cauchy-type singularities). A rigorous deduction of equation (5.1.33) from (5.1.28) must start from the definitions of the Finite Part and Principal Value . Using the function C(y) = - / +(n) s . (n) f (t, t1)d e , (5.1.34) TILE LIMNG SURFACE THEORY 166 whose significance will be given in 6.1, one demonstrates the identities r f(art) d z; d q - - = Iv yJD dy ff X f(C rl) d (5.1.35) d d(5.1.36) ff aUo d e which lead to the equivalence in view (see also 6.1). 5.1.7 Other Forms of the Integral Equation 1. We are going to give another form for the expression (5.1.12) of the component w . In this expression the integrals are not singular, so we may employ Green-type formulas. We have (1+.)]dd1l= fin _aly 1 70, +_ - i yo z2 {1 + a)} drl d (1+)]ddY! 1111 Lf YD_2 R1 + y02 + JJ D a yoZa YO (1 + RI ) d d l = r 1 Jo J = PODf(4,r!)yo a .a YO 1/o+z2 (5.1.37) Using the identity: ( ax `Ri) 1 l/ + 3 R z 8 (_ i xo Of oR (5.1.38) and Green's formula, we deduce I fDf(c,rl)8x (_)ddT!= -POD M,17) +132z2ffD x(oR1)dedn-JJ1 l Lf (5.1.39) YO T9Ridr dry. 167 THE LIFTING SURFACE EQUATION Replacing (5.1.37) and (5.1.39) in (5.1.12) and passing to the limit, we obtain 1 1 w(x, y, ±O) = t 2 f 1(x, y) +47rJaL - f (c n) ( yo \l + R moo) d e+ (5.1.40) +4n JJD 5; yo \1 + o} dt dn, whence it follows the integral equation: R f (C n) 1 47rJoD yo (5.1.41) of +4R JJD yo + J de do ° h=(x,y) This is the equation given by Homentcovschi equation herein, utilizing Green's formulas before passing to the limit is preferable to that of Homentcovschi which utilizes Green's formulas for singular integrals. Formally, the equation (5.1.41) is deduced from (5.1.28) by means of the identity (5.1.32) as follows: ACT)) AD (i+ xo _ 1r 8 (1+ R 1 R (1+ l o)d / ` dn(5.1.42) 2. In other papers (see for example, 11.21) we find the equation: 47r AD 1 yo I+ o) d do - h=(x, y) . (5.1.43) It can be obtained from Homentcovschi's equation, imposing the condition f (x, y) = 0 on 8D. (5.1.44) Usually this condition may be imposed either on the trailing edge or on the leading edge and on the lateral edges. Indeed, imposing the 168 THE LIFTING SURFACE THEORY condition (5.1.44) both on the leading and trailing edges means that for the plane profile obtained through it section of the wing with a plane parallel to the xOz-plane, one has to determine the jump of the pressure which vanishes at the both edges. But it is well known that the jump of the pressure for the plane profile satisfies the equation (C.1.1) which has a bounded solution at both edges only if the free term Satisfies the condition (C.1.13). Hence the condition (5.1.44) is valid only after imposing some restrictions for hx(x, y). However, for wings whose trailing (leading) edge is a straight line perpendicular on the Ox-axis, like, for example, the delta wing with the base representing the trailing (leading) edge perpendicular on Ox, or the trapezoidal wing with the big base representing the trailing (leading) edge perpendicular to Ox, the entire integral equation (5.1.41) reduces to the integral equation (5.1.43), because on the perpendicular trailing (leading) edge we have d = 0, and on the lateral edges we may impose the condition (5.1.41). For the rectangular wing the equation (5.1.43) is exact. 5.1.8 The Plane Problem In order to obtain the solution of the plane problem from the threediniensional solution we assume that the intersection of the wing with each plan parallel to the xO.-; -plane is a profile having the same shape (figure 5.1.1). This means that the equations (5.1.2) have the form z = h(x) ± hl (x) , (V)y, (5.1.45) and the domain D is the rectangle (- I < x < 1, -b < y < b) with no (half of the chord of the wing wsas taken as the reference length La). From these hypotheses it follows the general representations of the solutions (5.1.8) and (5.1.12), b fi = fi(b), f = fW Taking into account that J dy y (a2 + ff2y2)3/2 (12 f a'- + / y2 we deduce lim r b-..oo_b _ (!)(17)_$(229) R 2:co 169 THE LIFTING SURFACE EQUATION lin) b- -oo ./ b Oy 1 y02 V+ (1 + Z2 2 =- lim r+b0 ( 1, R y° )J cl vi = (5.1.47) (i+)Jdi=o. Jl[yo So, the above mentioned representation becomes I +I nQ xofi(o)+42 #2Zf dE Q r+1 xofg) - zpIWd, w(C z) = 2a JI xO+/32z2 (5.1.48) and v(x, z) = 0. Taking into account that here Oz has the position of the Oy -axis from the problem considered in 3.1,we find again the formula- (3.1.17). The integral equation is obtained from (5.1.28). For calculating the integral )r1 b b yo (1 + Lo dtl, R we use the definition of the Finite Part and the identity (5.1.32) valid on each of the intervals (-b, y - s), (y + E, b). Thus we obtain I+b 1 b yo (1 +°J d lint f x0 + , e-'0 l xoyn R +x yRl+b c - 2 (1+ Uf Ix011 /-bby ( l+ R)drl=-o. whence urn (5.1.49) The integral equation r educes to (3.1.21). 5.1.9 The Aerodynamic Action in the First Approximation We indicated in [1.11], p. 79 the way to calculate the action of the fluids against the bodies by applying the transport equations of the momentum and the moment of momentum to the fluid filling the domain THE LIFTING SURFACE THEORY 170 bounded by the body surface E and the control surface E0 surrounding E. This calculus is absolutely necessary when the body has corners, like in the case of thin bodies. However, in a first approximation the calculations may be done taking into account that the action of the fluid is given by the jump of the total pressure OPi 0 = Pt (x1, yl, -0) - PI (x, yi, +0) which determines the lift L. On the unit of area the force is pljk (fig. 5.1.4), k representing the versor of the Ozl-axis, and on the entire area of the domain D1, L = Lk, where L= f 1 pl I 1dy1 = f fD POdxdy . (5.1.50) Fig. 5.1.4. For the resulting moment one obtains the formula M = zl x Opt I kdxtdyl = PooUZLo fJ z x QpO kdxdy, (5.1.51) where according to (5.1.20), 64 = P(x, y, -0) - p(x, y, +0) _ -f (x, b) . (5.1.52) 171 THE LIFTING SURFACE EQUATION Denoting by At the area of the domain DI (which is the domain D in dimensional coordinates) we have Al = dx1dy1 = LoJ1 dxrdy = LOA. (5.1.53) U IL, For the lifting coefficient et, one obtains the formula cl def L (1/2)paoU.A1 = - A AD f(x,y)dxdy, (5.1.54) and for the moment coefficients (5.1.55) 2 Aao dot der _11f = Z Aao l,layf(x, J)dxdy, xf (x, y)dxdy,, (5.1.56) where al = a0Lo is the medium chord of the domain D1 on the direction of the unperturbed stream. As we shall see in the sequel, the drag (the projection of the resultant on the direction of the unperturbed stream ) and the gyration moment (the projection of the moment Ozlaxis) have the order of magnitude e2, for this reason they are not present in this calculation. M? is the rolling moment, and Afy, the pitching moment. 5.1.10 A More Accurate Calculation We obtain a more accurate calculation if we have in view that the resultant and the resultant moment are determined by the jump of the tension vector i.e. by the formulas R = - Jf pt nOdxtdyl , M = - fJ x1i x OplnO&rtdyl , (5.1.57) I w here n is the outward normal on E. The presence of the normal enables us to take into account the shape of the wing. If the equation of the wing is h(x, y) - z = 0(h = FIT), the normal n is +h k n=kri1+hr+hy =hri+hyj-k+ ta'' + h2 2 172 THE UFTINC SURFACE THEORY Neglecting the terms of order O(E3) and taking into account that h is h(x, y) f h1(x, y), it follows - p'nD = 2pco [h1=i + hlyj + (hxh1 + hyhlu)k] -POOU,2, IpI (h=i + by j - k) + pooU.2, < p > (hi1i + hlyj) , where we denoted < p >= p(x, y, +0) + p(x, y, -0) . (5.1.58) Hence the lift is: R. = -PpoUULo ff .f (y)dxdy + 2p.Lo ff (h,,hl + hyhiy)dxdy, (5.1.59) and the drag Rr = 2PooL2 IfD h1:dxdy + p UULo Jf f (x, y)h,,dxdy+ D (5.1.60) f < p > hudxdy. +p,oUO2OL0 ifD For the rolling moment one obtains the formula: h!= = -P.UULo LID yf (x, y)ddy +2pLy(hh+ (5. 1.61) for the pitching moment one obtains = pULx f (x, y)dxdy - 2pLx(hhl,, + hhly)dxdy, jj (5.1.62) and for the gyration moment, AI. = 2pooL03 Jj ( xhl- yhl)dxdy+ +PoU;,Lo JJ f(xhy - yhx)dxdy+ .l D (5.1.63) +PooUo,LO f f < p > (xhty - yh1m)dxdy, D where <p>=-- xo h1xR3dEdrl, D (5.1.64) THE LIFTING, SURFACE EQUATION 173 as it results from (5.1.19) and (5.1.27). If we keep only the terms of order O(E), it follows: R.- _ -Poo U.2 Lo2 Rr = 2pxL JJ JJ D f (x, y)dxdy, 1i1=(x, y)dxdy, (5.1.65) (A fr, Afy) = p A-Y'X)f (x, y)dxdy, D M, = 2pmLo Jf (xh- yhtr)dxdy. For the lifting surface (hl = 0) the formulas (5.1.55)-(5.1.59) become R. fD f (x, y)dxdy, R: = -PU Lo f f f(x,y)h.(x,y)dxdy, D (5.1.66) (M1. A.fy) =p LG ff(-y, x)f (x, y)dxdy, A9, = 2pxLo ff (xhy - yhr)dxdy.. D Rz, Air §i My being O(s), and Rr and Al;, O(E2). 5.1.11 Another Deduction of the Representation of the Gen-. eral Solution In the sequel we shall deduce again the representation of the solution (5.1.8), (5.1.12). We start from D. Homentcovschi's idea (exposed in (A.81) to utilize the Fourier transform for bounded domains (A.6]. The method synthesizes the problem of determination of the fundamental solution and the problem of replacing the wing with a forces distribution. In addition it justifies the assimilation of the wing with a distribution of forces having the form (fl, 0. f ). Indeed, employing the formulas THE LIFTING SURFACE THEORY 174 (A.8.2) to the system of linearized aerodynamics (2.1.30) and taking into account that D is a surface of discontinuity, we get 0=-ia1A (5.1.67) 0=-iali) - iap- Pk, with the notations [jpj)e,(QI x+a2Y)d x d y, (W' P) = JJ D ( where, taking into account (5.1.20), (5.1.25) and (5.1.27), awj = w(x, y, +0) - w(x, y, -0) = 2hi=(x, y) (5.1.68) W = p(x, y, +0) - p(x, y, -0) = f (X., Y) From (5.1.67) one determines first p and then w. One obtains ia3P - ia1W a2-M201 (5.1.69) ia3W W =ate +,Q2 a ia1P ia2P + a ( 2 - M22) Utilizing the expressions of P and W, we find: 1Q3 p- J fp(t, r?) a2-A?2aj e(-t4+0tin)9 dv7is l dd d11, -2 [1D hit (e, t?) a2 - M2a2 whence, taking into account (A.6.9), P(x, y, z) fJD'02 ,-I i(o,+ozq) a2 - 2ai d dy+ (5.1.70) rrrr +2JJD hlt ,rl) az a2 - M2a2 Employing (A.7.2) we obtain P(x,y,z) =-4a + 27r f fD rr JJD (_)d+ (5.1.71) hlt (, q) 8x (R1 d dr7 NUMERICAL I`TECRATION OF THE LIFI1NC SURFACE EQUATION 175 where R1 is that from (5.1.7). We obtained in this way the representation (5.1.8). Similarly, utilizing (2.3.11) and (2.3.27), we have: F`1 fly _ ial(a2 - M1ai) (5.1.72) r 0 aJ-110 1)dr 47r ya+z2 yo (i+) jai such that R, having the same expression, it follows: w(x,y,z) =-T7r - 1,! n Yj) dFdij- /32 + 4T 11 (5.1.73) (R1) di;d;+ 0 [ ?100+Z2 yo (1+)].d rl, which is just the representation (5.1.12). We have to notice that from (5.1.70) one deduces the inversion formulas 1 [Y2 at(G2 - h12a21) 1 4z y2 y 1+ X r.' + (a" +z-) (5.1.74) which could be utilized in 2.3. 5.2 5.2.1 Methods for the Numerical Integration of the Lifting Surface Equation The General Theory There are not yet known exact solutions of the equation (5.1.28). The first numerical solution was given by Multhopp in 1950 15.24]. Previously the same author had given in 1937 the approximate solution (to which one assigned his name), for the lifting line equation (Prandtl's equation (6.1.16)). Multhopp's method relies on the Gauss-type quadrature formulas for non-singular integrals. At that time there were not available quadrature formulas for singular equations. For the singularity appearing in the lifting surface equation, Multhopp utilizes a series expansion 176 THE I..IFTINC SURFACE THEORY with Chebyshev polynomials of first kind, which is truncated in order to obtain an algebraic system. The method is analogous to Glauert's method for Prandtl's equations, except that the sin functions are replaced by Chebyshev polynomials. In 1958, at a Meeting in Fort Worth, Hsu gave a quadrature formula for integrals with a strong singularity and employed this formula for the singularity from the equation (5.1.28). However, in Hsu's formula the unknowns are present even in the collocation points and this is a drawback. Starting from a formula given by Monegato [A.52), Drago§ gives [6.5] the formula (F.3.5) where there are present supplementary unknowns in the collocation points. One utilizes successfully this formula for solving Prandtl's equation in 6.5 and for solving the lifting surface equation in [5.10] and [5.11]. These solution will be presented in 5.2 and 5.3. In 5.2 we shall sketch the solution of the equation S via the collocation method. We have to solve the equation: I 4rr N(xo, lyo)dt dr1= -hr(x, y), i f( Q (5.2.1) o where N(xo,yo)=1+ To V2'or+ Y4 (5.2.2) with the following conditions: f (t, ±b) = 0, x-(fb) < < x+(±b) (5.2.3) -b < rl < +b. (5.2.4) The first two conditions mean the continuity of the pressure on the f (x+(rl), q) = 0, lateral edges in case that they are straight segments (if the lateral edges are represented by points the conditions disappear by virtue of (5.1.1)). The condition (5.2.4) ensures the boundedness of the pressure along the trailing edge. One performs the following reasoning: each intersection of the wing with a plane parallel to xOz determines a thin profile; as it is known from 3.1 for such a profile one imposes the boundedness condition on the trailing edge. In order to utilize the quadrature formulas we shall perform the change of variables (x, y) - (u, v) TI) --4 (a. defined by the equations x = a(y)u + c(y) Zr = a(ri)a + c(q) (5.2.5) y by rl = b13 177 NUMERICAL INTEGRATION OF THE LIFTING SURFACE EQUATION with a(y) = x+(y) - z-(y) 2 , c(y) = x+ (y) + x-(y) 2 (5.2.6) Writing the equation (5.2.1) as follows: r 1 T UFO +(n) f q)N(xo, yo)d do = -4xhz(x, y) and taking into account that e(t,g) = a(F)b, .9(a, /3) we deduce f a(A) 1 (v - A)2 f (a, /3)N(u, v, a, Q)da] dfl = 9(u, v), (5.2.7) where denoting by a(v), c(v), f (a, /3) etc., the functions a(y), c(y), f((, q) in the new variables, we have a(v)u + c(v) - a(/J)a - c(/3) N(u, v, a, /3) = 1 + 1/2 {[a(v)u + c(v) - a(Q)a - c(f)J' + k2b2(v - /3)21 (5.2.8) (5.2.9) g(u, v) = -4rrbhi(u, v), with the notation k The Kutta-Joukowski conditions (5.2.3) become -1 < a < +1) , f(1,/3)=0, -1</3<+1). (5.2.10) (5.2.11) The solution of the equation (5.2.7) depends on the behaviour that we impose at the extremities of the intervals (±1). For a fixed 0 , one obtains within the wing a thin profile. For such a profile we have imposed in 3.1 the behavior given by ++ a . Along the span one imposes in the behaviour 1 --02. Such a behaviour ensures that the conditions (5.2.10) are satisfied. We shall seek for solutions f (a/3) having the shape f(a,A) = f -l ++aF(a,A), (5.2.12) 178 THE LIMING SURFACE THEORY with F bounded. For F one obtain the equation JT (p1- v a(13) I- t '1 + a F(Q, 3)N (ti: vj rx $)da1 d8 _ -1<u,v<+1. =y(u.v), (5.2.13) Utilizing the quadrature formula (F.2. [8), this equation becomes ni 1 ^F t 2. (1 - a) 2 r) , a(;T):"(u.. v,(xi,O)F(ai, 1)d 8 = (5.2.14) = (tin + 1)g(u.v), where `=cos2m= (5.2.15) i= lm. a +1' From now on Nttlithopp's method and the quadrature: formulas method are separating. Multhopp's Method 5.2.2 Using the formula (F.4.2) for (f3 m v)-2 , equation (5.2.14) becomes p --4irEE(1+k)(1i=1 k-i J1 -or)Uk(u) JJJ - 3 `u(f3)N(a. v, ! = (2m + 1)g(u,v). (5.2.16) One calculates the integral from this equation utilizing (F.2.12). One Obtains r -.1-r it E(l+I)(I r=1 k-l,t=1 (5.2.17) =(2m+ 1)(n.+ 1)9(u, V), -1 < u,v <+1, NUMERICAL INTEGRATION O THE LIFTING SURFACE EQUATION with cosn+l' 7 =r . 179 (5.2.18) Introducing the notation H(ai,f3j) _ (1 - ai)(1 -QJ)a(#j)F(ai,f3j), (5.2.19) the system (5.2.17) becomes m rn p -4irrG.rL.r (1 k)Uk(v)Uk(Yj) (u, v,ai,Qj)H(ai,QJ) _ i=1 j=1 k-i (5.2.20) =(2m+1)(n+1)g(u,v), -1<u,v<+1. There are m x n unknowns, H(ai, /3j ). In order to obtain the same number of equations in (5.2.20) we shall assign m values to u and n values to v, obviously, all of them in the interval (-1, +1). As we shall see later, the aerodynamic coefficients are functions of H(ai, /3j ). It is sufficient to find out these unknowns. We may write computer programs for solving the system (5.2.20). 5.2.3 The Quadrature Formulas Method Utilizing in equation (5.2.14) the formula (F.3.5), we obtain the equation m n tar E F` [1i=1 j=1 (1 - ai)(1 ()3j - Ak)2 m -7r(2m + 1)(n + 1)2 E(1 - ai)a(.3k)N(u, Qk, ai, Qk)F(ai, Qk) _ iml = (2m + 1) (n + 1)g(u,13k) , (5.2.21) where Aj=ten+1, for k = T. (5.2.22) 180 THE LIFTING SURFACE THEORY In (5.2.21) we have a system of n linear algebraic equations with m x n unknowns F(a1, /3j ). Imposing the system to be verified in m points ul - oOS Vir 2m+1 =al, t=T, (5.2.23) the number of equations equals the number of unknowns. The system will be written as follows m n Bik, a=Tm k=I-n(5.2.24) Aekij Ha 13 i=1 where we denoted Alki; (2 [1- (-1)j+k] =t ()3j -)3k)2 -(2m + 1)(n + 1)21 j -A }N(atflkoi/3i) (5.2.25) Btk = (2m + 1) (n + 1)9(a1,Pk) 5.2.4 The Aerodynamic Action The lift and moment coefficients are calculated by the aid of formulas (5.1.54)-(5.1.56). Using (5.2.12), (F.2.18) and (F.2.12) we obtain: CL A ff D Af +1 =Af f +L +1 f(t, 1- fl2a(Q) 1 41r2b m [f 1 l f(a, d/3 = +1 1 1+ aF(a, l)da] dQ = (5.2.26) " A(2m+ 1)(n+ 1) FF H(a"Pj)' w1 j=1 where H(a;, (3j) are (5.2.19). Formula (5.2.26) is valid for both Multhopp's method and the quadrature formulas method. NUMERICAL INTEGRATION OF THE LIFTING SURFACE EQUATION 181 In a similar way we obtain: 2 = IL of 4R Aao +I J_ I ja(,3)f (a,13)da df3 = n in 2b2 f +I aoA(2m + 1)(n + 1) E E Qj H(ai,13j) i=1 j=1 , (5.2.27) cy Aao 2b - Aao 11 f r')dCdr' = +1 +I J-1 4-Ir 2 ta(,B)a + c(R)I a(A)f (a, Q)da d/.3 = b aoA(2m+1)(n+1) ao m r( ( it (5.2.28) q ti_1 j=1 representing the length of the medium chord and di = CD = - A !ID 41r2 b A(2n+1)(n+1) n (5.2.29) n i=1 j=1 For the flat plate of incidence E, we have h(x, y) = -Ex then g(u, v) _ 47rb. For the rectangular flat plate we have x_ = -1, x+ = 1, a = 1, c=0, u-a N(uvapl=1+ , , , 5.2.5 (5.2.30) The Third Method Some numerical experiments show that it is not always indicated to impose the behaviour along the direction a under the form We shall use in (5.2.7) the following quadrature formula +1 m f (a, Q)N(u, v, a, /3)da = > f(ai, i9)Ki (u, v, f3) , 1 i=1 (5.2.31) 182 THE LIFTING SURFACE THEORY 2i where a; = - m , i = 0, 1'.. . , m are equidistant nodes on the interm val (-1,+1), and K1 (u, v, /3) = joi-I N(u, v, a, fl)da = 2 m 1 a(Q) /[a(v)u + c(v) - a(A)a; -- c(/3)1+ k2b2(v + - p)2+ (a(v)u + c(v) - a(/3)a{-1 - c(,6)J2 + k2b2(v - /3)2. (5.2.32) The behaviour in the span direction remains the same (like in the previous methods), hence: 1--#2F(&,,6), f (a, A) = (5.2.33) where F(a, 3) is finite for rQ = ±1. We shall use the same formula (F.3.5) with respect to p. So, the equation (5.2.7) furnishes the following discretized form of the equation (5.2.7) ,n n HtkiiF(a (3:) = bh=(ailk) , (5.2.34) i=I j=1 where Htk;j _ = 1 4(n + 1) {i_(_i+k] 2 Ak)2a(13i)Hj(ac,,6k, Qj) (Qj - , k (5.2.35) and n +l 1 H10919 = a(/3k)Ha(at, Qk, /3k) . (5.2.36) The algebraic system (5.2.34) has m x n linear equations with m x n where #j =coo 17r unknowns n+l* NUMERICAL INTEGRATION OF THE LIFTING SURFACE EQUATION 183 The lift and moment coefficients are calculated by means of formulas CL _ - 2b +1 A 11 +1 1-,2a(fl)F(a,A)da dQ = J_1 1 2bir (n + 1)A j (1-1,)a(13j) J F(a, $$)da 1 (5.2.37) 1 n 4bnr (1 in n+ 1 )rr1A Jul i=1 C, 22 2 1 1+1 #a(13) J1 = 1- J32F(a, /3)da dfl _ (5.2.38) nm 4b2ir E I6ja(0j)(1 - AJ)F(ai, Qi) = (n + 1)mAao, Jul i=1 Aao f 1111 a(#)[a(A)a + c(Q)] 1- 02fla,,O)da d/3 = n cm -4b,r In + 1)mAan Jul iml EEa(AJ)(1- _Q1)(2i-'m-1a(pi)+c(fii),F(ai,QP)(5.2.39) In the paper 15.101, from which we presented this method, there are presented computer programs for the elliptical flat plate h(x, y) = -cx, x2 + e < 1, b = 2 and for the wing whose projection on the Oxy - plane is an rectangle -b<y<b, 0<x<1, x(y)=0, x+(y)=1, and whose normal section is an arc of parabola having the equation h(x,y) = e(1 - x2), e <<l, (5.2.40) THE LIFTING SURFACE THEORY 184 The fluid in considered incompressible. For the rectangular wing one obtains an analytic solutions in the framework of the theory dealing with the wings of low aspect ratio. In this way we have a test solution for the third method. One deduces that this method furnishes very good results. 5.3 5.3.1 Ground Effects in the Lifting Surface Theory The General Solution The ground effects in the lifting surface theory were taken into consideration in (5.311 and [5.371 where one utilizes asymptotic methods. An approach of this subject in the framework of curvilinear lifting line theory may be found in [1.32], [1.331. The present subsection is elaborated following (5.8], where one gives the general theory. One utilizes the fundamental solutions method. The geometry of the problem is presented in figure 5.3.1. The origin of the reference frame is located in the middle of the span, the Ox-axis has the direction of the unperturbed stream and the Oy-axis has the span direction. The ground is considered to be the plane 1I having the equation z = -d/2. The unperturbed flow is characterized by the velocity Ui , the pressure p(,,, and the density po and it is considered to be subsonic like in 5.1. The field of velocity v1, the pressure p1 and Fig. 5.3.1. the density p1 for the perturbed flow have the form (2.1.3). One utilizes 185 GROUND EFFECTS IN THE LIFTING SURFACE THEORY dimensionless variables. The equation of the upper and lower surfaces are given by (5.1.2). It results the following boundary conditions w(x, y, ±0) = hr(x, y, ±hiz(x, y) (5.3.1) (x, y) E D (5.3.2) w(x, y, -d/2) = 0, (V )x, y. For satisfying the last condition we shall utilize the images method. This means to replace the wing with a forces distribution f + = (fl, 0, f ) defined on D and with a symmetric distribution f - = (fl, 0, -f) defined on the domain D', which is the symmetric of D with respect to the plane II. In order to write the general form of the perturbation fields p and w we have at first to write the form determined by the concentrated forces f+ in P+(4+) and f- in P_(4-). We have therefore to determine the solution of the equations M28p/8x + divv = 0 (5.3.3) 8v/8x + grad p = f+b(x - 4+) + f -b(x - 4-) corresponding to the system (2.3.4). The system (5.3.3) is linear and its solution is the sum of the corresponding solutions from (2.3.4). Taking into account (2.3.24) and (2.3.29), and the form of the fields f + and f -, we deduce: p(x, y, z) =-4 w(x* y, z) =4 0T (f, +fa ) `R+l 8z ( R+) 4 f8 x 47r (fie -fe) (i)' R+) + +47r 5; [(y_+)2+(z_z+)2 (i+xR+ )J+ 2 +4 Oz f R_)+4 f8x (R_)y - 17 4ir 8y (y - 17- )2 where R+= (x - t:1)2+ (z - C-)2 + -- (1 + x R- 2n*)2 +(z - (±)2). l )J ' (5.3.4) (5.3.5) 186 THE LIFTING SURFACE THEORY The points P+ and P- are symmetric if £+ _ {- = f, tl+ = n- = n, t+ _ C, C- _ C - d. When P+ are in D((= 0), the symmetric points P-((- _ -d) are in V. Assimilating the wing with a continuous superposition of forces defined on on D, it results the following general solution: p(x, y, z) _ - 47r JJv [fi,-+f,q)-} ()d_ d-9(5.3.6) rr f (e, n) 8 l ((Rid) dkc di J w(xy, z) 4s IL +47r'ID [f(t, n)ex JJDf(t,rl) 4r AD A2f(t, n) ] (R,' ) de cI + (L)d+ (i+)}d_ [yo-0 If,V,17)z + 1 +47r - 02f (f, n) 8J !0- Lyo+ dC (1+ -10-d/J dt dn, (5.3.7) where xo=x-t, yo=y-Hand Rt = xo+(32(yo+z2), Rld= 90 +p2[yo+(z+d)2J. (5.3.8) This is the general representation of the perturbation, the functions f, (t, r)) and f (t, t) being determined from the boundary conditions (5.3.1). The condition (5.3.2) is obviously satisfied. 5.3.2 . The Integral Equation Acting like in 5.1, we shall pass to limit considering z -- ±0, (x, y) C- E D. To the limit, only the integrals related to R are singular. Using the formulas (5.1.16), (5.1.18) and (5.1.24), from (5.3.6) and (5.3.7) we CROU\l) EFFECTS IN THE LIFTING SURFACE THEORY 187 get P(x, y, ±0) = 4rID fi(f,t,)Rgdt dry ± 2f (x, y) + + 4a JJ f1(, q) ddi 4j r R3 dti, , (5.3.9) P(x, y, +0) - P(-T, V, -0) = f (x, y), and w(x, y, f0) = 2 f1 (x, y) 47 if - 47r ,IJD R (i+)d_ f ff f (t, t)N(xo, (Z' Y) - d'l 41r 3Nn)dt dr1, (5.3.10) where N yo) _ (d2-# ) + (d + ;'F d 2 2 (1 + xd° ) I (5.3.11) with the notations R= xo + $2y$ , Rd = xo + 02 (y02 + d2) . (5.3.12) Imposing the boundary conditions (5.3.1) we deduce fi (x, y) = -2hi:(x, y) (5.3.13) and then 4Rv f (YOT L1+'0 R) dk drl + 4,r AD f (f, rl)N(xo, yo)dC dy = (5.3.14) The equation (5.3.14) may be called the generalized lifting surface equation. It was given in [5.8]. The sign "*" is for the Finite Part. Using 188 THE LIFTING SURFACE THEORY the identity (5.1.32) we may write the equation as follows: 4n 8y jf f (,'1) 1+ o) = h' (x, y) drl - 4a If f (t, r!)N(xo Uo)df do = .JJD hlfR3 1) dt dtl . (5.3.15) The Two-Dimensional Problem 5.3.3 Like in 5.1.7, in order to obtain the representation of the solution and the integral equation for the wing of infinite span we assume that the normal sections in the wing determine profiles having the same shape. It means that the equations (5.1.2) have the form z = h(x) ± hl(x) (V)y and the domain D is a rectangle (-1 < x < 1, -b < y < b) with oo. We notice that in the representation of the general solution b (5.3.6) and (5.3.7) ff and f have the form f,(£) and f(t). Relying on formulas like (5.1.46), from (5.3.6) we deduce P(x, z) - 1 21rp 1 )324f (t) dt+ f,2 xoft Wxo ++ $2z2 r+1 xofl (t) +,32(z + d)1 (t) 0 +$2(x2+d)2 which are just (3.2.3)1. Using formulas like (5.1.47), from (5.3.7) it results: w(x,z) __ r11 x01(0+ z2(C)dc _2 J_li xOf +fl ( +d)2( )41 2a i.e. just (3.2.3)2. For obtaining the integral equation we shall use (5.1.45). We also have oo do J ydq 2 , l 00 The integral 1= +00 L -0o (y[+ d2 _ yo xo dn )2 R1 = 0. (5.3.18) 189 THE WING OF LOW ASPECT RATIO reduces to integrals having the shape Jo (u2 + r du d2) VU-T-+31 Je du (u2 + d&)Z u + which may be calculated using the change of variable u -+ v where u V (5.3.17) VU-27 71 One obtains I= 2,0 d 1- xo xo arctan xo dQ (5.3.18) Using the same change of variable (5.3.17) one obtains also +O0 xo d 1 (plc 01? 11 9 1 + ;01 o xo x arctan d (5.3.19) such that we finally obtain oo -l (5.3.20) xo + d2 02 Taking into account the previous results, the integral equation (5.4.12) becomes ,r+i 1 Jl 1 xo _ dt - r+i xof () 27r JJ 1 xo + d2#2 L" /+1 clk = (5.3.21) h'(1) which is just (3.2.7). In (5.111 one extends the method from 5.2.3 also to the equation (5.3.14). 5.4 The Wing of Low Aspect Ratio 5.4.1 The Integral Equation In the sequel we shall pay attention to the ratio A = (2b)2/A, (5.4.1) 190 THE LIFTING SURFACE THEORY called the aspect ratio. We denoted by 2b the span and by A the area of the domain D. For wings characterized by a small A (wings of low aspect ratio) one develops herein a theory which leads to the integration of the lifting surface equation (5.1.33). For wings characterized by a big A we shall develop in the following chapter the lifting line theory. These are the two asymptotic theories of the lifting surface equation. If the square of the span is small with respect to the area A, we deduce that on the greatest part of the wing we have yo << xo (see fig. 5.4.1) and we may write R Ixoi xo xo l L1 + O ( )J , (5.4.2) o where R is (5.1.17). On the contrary, when the square of the span is big with respect to A (see figure 6.1.1), on the greatest part of the wing we have xo << yo, so that we can make the hypothesis (see 6.1). Fig. 5.4.1. R = kayo) (5.4.3) Returning to the wings of low aspect ratio, we notice that under the hypothesis (5.4.2) the equation (5.1.33) may be approodmated by 4rr 8y J JD f (n) (j+LX-01 df drl =1. (z, y) (5.4.4) Noticing that the integrand of this equation is different from zero only on t < x, we shall consider only wings whose trailing edge is a straight line perpendicular on Ox like in 5.4.1. Introducing the unknown function F(x, q) JS-(n) f(s, q)d, (5.4.5) THE WING OF LOW ASPECT RATIO 191 utilized for the first time by Jones 15.171, 15.18), the equation (5.4.4) becomes 8 1 v+(=) { dry = h=(x, y) , 2w 8y Jy_ (x) (5.4.6) yo where y = y+(x) represents the equation of the leading edge OB, and y = y- (r), the equation of the leading edge OA. From (5.4.5) and from the figure one notices that F(x, y- (x)) = F(x, y+ (x)) = 0. (5.4.7) In (5.4.6), utilizing the definition of the principal value, we derive taking into account (5.4.7) and then we integrate by parts. It follows a F(x, n) ay _f dq - e F' 1 yo! d _ 'v+ SF do - Jy- an ' (5.4.8) where, from now on, y- = y-(x), y+ = y+(x) In this way equation (5.4.6) becomes + OF dr) 1 Jv- 8q - h. (x, y) (5.4.9) This equation can be also obtained starting from Homenteovsch's equation (5.1.41) which, in case that one imposes the condition (5.1.44) on the leading edge, becomes equation (5.1.43). Then we can see that from (5.4.5) it follows 8F_ 8f (5.4.10) The equation (5.4.8) is a thin profiles-type equation (C.1.1) and it has one of the solutions (C.1.9), (C.1.10), (C.1.11) or (C.1.14). The solution of the equation (5.4.8) must satisfy the condition J +OFdq=0 y_ a! (5.4.11) arising from (5.4.7). The only solution satisfying this solution is (C.1.11) with C = 0. Hence, 8F(x,TI) -2 1 (t-y-)(y+-t)h=(x,t)dt. t-77 (5.4.12) 192 THE LIFTING SURFACE THEORY Integrating this solution on the interval (y_, y) and taking into account (5.4.7) and (B.6.11), we get F(x, y) _ ' dq 2ly (n-y-)(y+-q) - + (t - V -)(y+ -' t) t y- (x, t)dt . -q (5.4.13) As from (5.4.5) it results f (x, y) = F, (x, y), we deduce f (x, y) = =2 8 8x ydq ru+ (t - (qy-)(y+-q) y- ly- t) t-q (x, t)dt . (5.4.14) Employing (B.6.11) ones easily check that this solution satisfies the vanishing condition on the leading edge f(x,y-)=f(x,y+)=0. 5.4.2 (5.4.15) The Case h = h(x) In case that h does not depend on the variable y, taking into account (B.5.6) and performing the change of variable q = y_ cos20 + y+ sin20, we get : Jv- 0 < 0 < 7r/2, (5.4.16) ' V+ dt7 At - Y-) (y+ - t)dt = t-q n - y-)(y+ - q) Iv_7r Y+-Y- r 2 -a J y+ v- dq On - y-)(y+ - r!) ydq=sV(y-y-)(y+- y) Hence, f(x, y) = 2 [h'/i_ y)] . (5.4.17) 193 THE WINO OF LOW ASPECT RATdO Utilizing (5.1.54)-(5.1.56) and Green's formula we deduce: CL - A If 19 [W(x)Vv - v-}(v+ - Y1 f (y-a (b-y)dy=-2Ah'(1)(b-a)2, A -S = - A ,,D ybx = - Ah'(1)1 cv A v= Y [h!(x)-,/(y - v-)(v+ - v)] dxdy = (v - a)(b - y)dy = - bWrh'(l)(b - a)z, I/I; [zWv'?j- v-)(v+ - v)1 dx dy_ A If h'(x) vir(v -y-)(Yu)dxdy = - _ -CL - 1 0 [v+(z) - y_(x)12 h'(x)dx, (5.4.18) the reference chord ao bung 1. For example, fnr the rectangular flat plate of incidence -e with the span 2b one obtains CL = wk, c, = - c= = 0, ! , (5.4.19) and for the triangular flat plate of incidence -£, we have, CL=xe(b - a), because y+ = bx, c, = (b2 - a') r , cy = - 3 xc(b - a), (5.4.20) y- sax. 5.4.3 The General Case We shall write (5.4.14) as follows f(x, y) = 2b x8s r_ (t - -y-)(y+ - t)h.(x, t)I(y, t)dt, (5.4.21) THE LIFTING SURFACE THEORY 194 where Y d-q I(y,t) = (5.4.22) (r1-y-)(y+-rl) t - n with y- < y < y+, y- < t < y+. This integral may be calculated l fv- explicitly. For y < t the integral is not singular. For t < y the integral is singular but noticing that we have 1(y+, t) = 0 (it results from (B.5.8)) we get IY+ 1 I(y,t) d'l y+-rl) t-17 (5.4.23) which also is not a singular integral. Since A (n-a)(b-n) t-11 (5.4.24) (t-a)(b-rt)+ (b-t)(q -a) 1 in (t - a(b-it)- -v(((r1-a) (t - a)(b - t) we deduce the fundamental formula f(x,y) = = 2 8 f+h (t-y-)(y+-y)+ (y+-t)(y-y-) dt. (x,t) a 8x v- (5.4.25) For determining the lift and moment coefficients we have to calculate the integral (t V+ G(a, t) = - y-)(y+ - y) + (y+ - t)(y - y-) dy. J(t-y-)(y+-y)- V(-y+---4(y -y-) (5.4.26) Performing the changes of variables t=y++y-+y+-y-moo, y=y++y-+y+-y-cos9, 2 2 2 2 (5.4.27) we deduce Gxt - y+ 2 y- fo X in sm2 sm B+a 2 sin 8 d9, (5.4.28) 195 THE WINC OF LOW ASPECT RATIO the integral having an integrable singularity for 8 = a. The integrand from (5.4.28) is the kernel from the equation (6.2.15). Using for 0 91 a the expansion (6.2.17), we get: G(x, t) = w y+ y- sin o = 2 (y+ - t)(t - y-) . (5.4.29) It follows (t - a(b-y)+ (b-t)(y-a) fb G(I, t) =J In Ja IWa)(b-y)-'(b-t)(y-a) (5.4.30) =x (b - t(t-a). Similarly one deduces b (t - a) (b - y) + JQ y (b - t) (y - a) dy= IV'(t-a)(b-y)-i(b-t)(y-a) (5.4.31) =att+b2a) (b-t)(t-a). Utilizing Green's formula and the expression (5.4.25) for f (z, y) taking into account (5.4.30), (5.4.31), we get: A CL - -4 , lb 41 h`(1, t)G(1, t)d t = rb a _ b cs = - AJa t (b - t)(t - a)h=(1, t)d t + !-+-b ct, Cy = -CL - r1 4 rA Ju 4 r2 - -CL - A dx y+ y_ h.(x, t)G(x, t)d t = ry+ dT o f (5.4.32) hs(z, t) (Y+ - Wt - y-)d t . When h(x, t) does not depend on t we find again the formulas (5.4.18). 196 THE LIFTING SURFACE THEORY This problem, for wings symmetric with respect to the Ox-axis, was studied in [5.171. It was also included in [1.1], [1.2]. The general problem for the arbitrary wing as it was presented herein, was solved in 15.91. Chapter 6 The Lifting Line Theory 6.1 6.1.1 Prandtl's Theory The Lifting Line Hypotheses. The Velocity Field Prandtl's theory is the first mathematical model for the three - dimensional wing (the finite span airfoil). It was elaborated in 1918 16.211 and it remained until the years '40 the only theory for this wing. The german scientist, gifted with an extraordinary engineering intuition, guessed very well the simplifications which may be performed. Prandtl's method consists in replacing the wing with a distribution of vortices on its plane - form (the domain D from 5.1). Since the experience indicates that downstream the wing the flow is not potential, Prandtl introduced a vortical distribution defined on S, the trace of the plane-form D in the uniform stream (figure 6.1.1), the velocity field in the fluid being determined by the two distributions. The vortices on D are called tied vortices and the vortices on S are called fee vortices. This idea continued to dominate the aerodynamics, the models concerning the subsonic, supersonic and transonic, steady or oscillatory flows, elaborated in the years '50, '60, being conceived on the basis of this method. In 1975 D. Homentcovschi, utilizing the theory of distributions proved that the hypothesis of the existence of free vortices is not necessary because it follows from equations. L. Dragon [5.71 obtained the same result utilizing the method of the fundamental solutions. In this subsection, we deduce the lifting line theory from the lifting surface theory, as we proceeded in [5.7], utilizing the following three hypotheses (the hypotheses of Prandtl's theory): 10 One neglects the thickness of the wing, therefore in (5.1.2) one considers hl = 0. From (5.1.27) it results fl = 0, such that the representations (5.1.8), (5.1.11) and (5.1.12) give P(x,y,z) = -4 7r- IL f(CTT)1 (-) dcdi 19Z (6.1.1) THE LIFTING LINE THEORY 198 u(x,y,z) _ -p(x,y,z) v(x,y,z) _- + jo z2(1+Rl)]dfdi7 (6.1.3) )Io qJ w(x, y, z) (6.1.2) fD f (C, 17) 8x 1)dt d t7+ JJD0Y (6.1.4) -+ z2 (1 + 11 )]d t d n; 2° One considers that the unknown is the circulation C. along the contour c, resulting from the intersection of the wing with an arbitrary plane n parallel to the xOz plane at the distance y(-b < y < b) (fig. 6.1.1). This will be, obviously, a function of y Fig. 6.1.1. Pudx, C(y)= because on this contour d y = 0 and wdz = O(E2) (it results from (2.1.17) and (2.1.21) ). Taking into account (6.1.2) and (5.1.20), we obtain r C(y) = Jx u(x, y, -0)d x + / + u(x, y, +0)d x x We have also f X f(x, y)d x . (u) (6. 1 .5) x+ (y) C(n) = - x_ (y) f (f, n)d t. (6.1.6) PRANDTL'S THEORY 199 From (5.1.1) or (5.1.30) it results C(±b) = 0; (6.1.7) 30 The domain D, i.e. the projection of the wing on the xOy plane, is replaced by the segment [-b, +bJ (fig. 6.7.1); for this reason we call this theory the lifting line theory. when For studying the behaviour of the integrals (6.1.1)-(6.1.4) x_(n) -+ 0 .- x+(il), we notice that in the vicinity of this segment, for a given n , the function f (e, n) keeps a constant sign, such that we may apply the mean formula. For a function h(z, y, z, n), continuous in f, when x- (q) -+ 04-- x+(n), we have therefore f ( n)k(x, y, z, limJ1 r+b hm ti)d d r! _ r+(n) J-b [ Js_ty) -lim , f (t, n)k(x, y, z, t, r1)d d r! +b -b k(x, y, z, (', n)C(n)d n = - fb k(-T, y, z, 0, n)C(n)d n , (6.1.8) where £ E (x_(rl), x+(n)). Applying this formula in (6.1.1)-(6.1.4) and taking into account (6.1.7), we obtain 2 P (x, y, z) = - 4 rb C(n) d n = -u(x, y, z) , f+b v(x, y, z) = 4 .! b _- a 4n w(x,y,z) =-Q2 41r ly2+z2 J b C('l) f 1 )]dii C(17)ay [ o + z2 (1 + = rT o (6.1.9) C'(n)yo+z2 ( 1 + C(n) rb -4 J b Ro }n, d drl- b 0,07) y02 YO Z2 (1 + X d 200 THE LIFTING LINE THEORY where we denoted RD = x2 +#(y2 + z2) (6.1.10); Fbr Q = 1 we obtain Prandtl's representation ((1.21], p. 708). 6.1.2 Prandtl's Equation In the sequel we shall deduce the equation for C(y). This will obviously result from the lifting surface equation by virtue of the above hypotheses. We shall employ the equation (5.1.33). Taking (6.1.6) and (D.3.6) into account it results 8 f( , n) d 8y JD y-n 1 fd n= dy -d- f' +b C(q) -b n-y - d rJ C(q) d n. O -b (n - y)2 Utilizing (D.3.7) and (6.1.7), we also deduce: . & c(_) = l1+b C'(n) J b 'n-yd (6.1.12) In the second term from (5.1.33) we take into account that when x_(n) - 0 4- z+(n), on the greatest part of the domain D, we have so << 14, before E which intervenes in the definition of the principal value of the integral with respect to n becomes zero, i.e. before yo becomes zero. In the principal value we shall perform therefore the approximation R = fl I. An exact evaluation of this approximation is, made by D. Homentoovschi in [6.9) as follows jjf(r) ryxTo + dd XDYO Taking into account that sign yo = 2H(yo) -1, H'(yo) = 6(y - q), H representing Heaviside's function and 6 Dirac's distribution we 201 PRANDTL'S THEORY have: Z_ JJ 20 f (f, n) xovo +b I J-b JJ d {d ~ Q f (txon) t LJ d C, f (on) s yod fd n = ' f=+ (v) f (t'17) _ d 5(y - n)d n= 2Q xo (y) (6.1.13) Hence the equation (5.1.33) becomes: 1 +(v) ,rte C'(n) d n + sr f b n - y f (c Y) d 'r s_(v) xo = 2h' (a, y) (6.1.14) x]-1/2 (6.1.15) Multiplying this equation by [x - x-(y)]11.2[x+(y) - and integrating it with respect to x on the interval obtain from the first formula (B.5.4): AC(y) = (x- a(2) 'f+b C'(n)dn+1(y), J bn-y (y), x+(y)), we (6.1.16) where we denoted a(y) = J(y) _ -2 W x+ (y) - x- (y) (6.1.17) 2 IT x(6.1.18) a(y) representing a half of the chord of the profile at the distance y. The equation (6.1.16) is the well known Prandtt's equation (the lifting line equation). This equation, together with the conditions (6.1.7), has to determine the unknown C(y). It is an integro-diferential singular equation, with a Cauchy-type singularity. Utilizing (6.1.12), the equation for C(y) may be written QC(y) = a 2y) t +b r b (n + (ny)2 d n 9(y) . (6.1.19) In this form, the equation is not integro-differential any longer, it is only integral, but with a stronger singularity. The mark "s" is for the Finite Part. As we have already shown in [6.5] [6.6), this form is more adequate for the numerical integrations (see also 6.4.2, 6.4.3). 202 THE LIFTING LINE THEORY z A Fig. 6.1.2. The significance of the function j (6.1.18) is given in [1.2J. In the case of the flat plates having the angle of attack c (fig. 6.1.2) we have hx = -e whence (6.1.20) j(y) = 2irea(y). 6.1.3 The Aerodynamic Action Taking into account (6.1.6), from (5.1.54) and (5.1.56) we deduce 2 CL 2 +b C(y)d y, c +b Aao 1-b yC(b)d y, cv = 0. (6.1.21) The expression of c, is natural because for the wing reduced to the segment [-b, +b] the fluid cannot create a moment which should rotate this axis. By virtue of the first hypothesis (h1 = 0), in the framework of the lifting surface theory, the quantities Rs and M, which have the order of magnitude 0(e2) are reduced to those given in (5.1.56) and (5.1.59). Utilizing (6.1.6) and (6.1.7), we obtain: R. = -p.U.Lo C(y)h=(0, y)d y, b (6.1.22) +b M: = P,,. U,2,. L. f yC(y)hi(0, y)d y b But, from the boundary condition w(x,y,0) = h,(x,y) 6.1.23) and from (6.1.9) it results h=(0, y) = w(0, y, 0) _ - 1 lim ( 41r c-.o \ f V-s b + b y+c ) C'M d Yo rl , 203 PRANDTL'S THEORY because, with the change of variable n - y = u, we have y+E Eiti y-E C(rl)' 11o +-Z 2 d r! _ -C'(y) lim E-0 f +6 z = 0. 1 y+bCl(I1)d b tl y rJ , ( 6.1.24) we deduce 2 CU f A J-b C(y)w(y)d y, (6.1.25) 2 MS cZ A (1/2)Poo `k 2 Alai Aa_ Jb yC(y)w(y)d y, representing the area of the domain D and ao, the length of the dimensionless medium chord along the direction of the unperturbed stream. In fact, for w we may also utilize the expression w(y) = QC(y) - ?(y) (6.1.26) 2wa(y) which results from (6.1.18) and (6.1.26). 6.1.4 The Elliptical Flat Plate Until the appearance of the papers of Magnaradze [6.16] and Vekua [6.28] in the years 1942, 1945 it was available only one exact solution of Prandtl's equation corresponding to the elliptical flat plate wing. Usu- ally this solution is obtained as an answer to the following minimum problem: "To determine among the wings having the same lift, the wing corresponding to the minimum drag" (see, for example, (1.201). Sometimes one utilizes Glauert's approximate method (see 6.2.4). In the sequel we shall present a simple method which does not need any special considerations on Prandtl's equation. Let us consider a flat plate with the angle of attack E, whose projection on the x10y1 plane is an ellipse having the semi-axes Lo and bLo (figure 6.1.3). In dimensionless coordinates, the equation of the ellipse is x2+ +y2/b2 = 1. For the edges and the chord it results y2 xf = f 1 - 2 52 , a(y) = 1 - 2' (6.1.27) 204 THE LIFTING LINE THEORY YI bLo L=am Fig. 6.1.3. For a flat plate j(y) has the form (6.1.20) hence it is proportional to a(y). From Prandtl's equation it results that C(y) is also proportional to a(y). We shall look therefore for solutions having the form: C(y) = k 1- a Eb2 , (6.1.28) k being a constant which will be determined by imposing (6.1.23) to verify (6.1.27). Using the change of variables q = boos 0, y = b cos o, and Glauert's formulas (B.6.6), it results: di7 11-y such that k (2TE A + 7r/2b) _ir, (6.1.29) (6.1.30 Since the area of the ellipse is w b, from the formulas (6.1.23) and (6.1.27) we obtain s cL = k, CD = , cj = cs = 0. (6.1.31) 46 Obviously, cD = 0(0). For b - oo one obtains the infinite span flat plate. From (6.1.23) and (6.1.27) it results: CL = 21r8 (6.1.32) 205 THE INTEGRATION OF PRANDTL'S EQUATION 6.2 The Theory of Integration of Prandtl's Equation. The Reduction to Fredholm-Type Integral Equations 6.2.1 The Equation of Trefftz and Schmidt The general method of solving the integro-differential equations con- sists in reducing them to Fredholm-type integral equations. As it is well known, for the last ones a general theory is available (existence and uniqueness theorems, exact and approximate methods for solving the equation). The first investigation of Prandtl's equation was performed by Trefftz in 1921. He reduced the problem of solving Prandtl's equation to the problem of determination of a harmonic function in the superior half-plane with mixed conditions on the boundary. We shall prove in the sequel that such a problem may be reduced to a Fredholm-type equation. To this aim we shall consider the harmonic function U(y, z) in the superior half-plane z > 0 from the yOz plane, with mixed boundary conditions in the Z = y + iz complex plane (fig. 6.2.1). We shall also consider the complex function F(Z) = U(y, z) + iV(y, z) = 1 2m : ZdR. Jb+1 COO (6.2.1) z (Z) -b +b y Fig. 6.2.1. Obviously this is an holomorphic function in the (Z) plane with the cut (-b, +b) and it vanishes to infinity. U(y, z) is therefore a harmonic function in the half-plane z > 0. It results obviously U(y, +0) = 0, y E (-oo, -b) U (b, oo). (6.2.2) The limit values on the segment (-b, +b) are obtained by means of Plemelj's formulas. We deduce 2U(y, +0) = C(y), y E (-b, +b). (6.2.3). 206 THE LIFTINC LINE THEORY Taking the conditions (6.1.7) into account, we get d -dn= f C(t1) ZJ bbtj b =-f C(rl)'3; rl b C(rl)aZ Z)dr)_ Cn 1 Z) d+l - f b 71- Zdrl Deriving we deduce from (6.2.1): F'(Z) = I-i 8 = 2m J-b W(Zd q (6.2.4) and with Plemelj's formulas ' +b az (y, +0) = d (y, +0) = 2a f d, (6.2.5) n representing the inward pointing normal of the superior half-plane on the segment (-b, +b) of the real axis. Taking (6.2.3) and (6.2.5) into account, Prandtl's equation is transformed into the following condition on the segment (-b, +b): dU 21U(y, +o) A(y) - J(y), (6.2.6) A(y) = ra(y), A(y)J(y) =1(y) , (6.2.7) in (y, +0) = where, from now on, we denote So, Prandtl's equation was replaced by the following mixed boundary value problem: "to determine the harmonic function U(y, z) in the z > 0 half-plane, vanishing at infinity and satisfying the boundary conditions (6.2.2) and (6.2.6)". Then, the function C(y) will result from (6.2.3). The mixed problem is reduced to a Fredholm-type integral equation. Proceeding like in [1.211, p. 713, we notice that because of the condition (6.2.2), the function F(Z) may be extended by symmetry in the lower half-plane z < 0. In this way, on the lower margin of the cut (-b, +b1 we shall have the condition dU d n (y, -0) = 2QU(y, -0) + Ay), A(y) (6.2.8) So, we reduced the above mixed problem to the problem of determination of the harmonic function U(y, z), in the yOz plane, with the cut THE INTEGRATION OF PRANDTL'S EQUATION 207 [-b, +b) and the conditions (6.2.6) and (6.2.8) on the two margins of the cut. It is known (see for example, 11.111 [1.201 [1.31J) that the Joukovsky- type conformal mapping Z - W: Z = I (W + W (6.2.9) maps the exterior of the cut [-b, +b] from the (Z) plane onto the exterior of the circle of radius b and the center in origin from the (W) plane (fig. 6.2.2), the superior margin of the cut being mapped on the superior half-circle from the superior half-plane. The points fb are double and singular. One obtains the correspondence of the boundaries putting W = be''. We have y=boost, a=0. (6.2.10) Fig. 6.2.2. For 0 < a < r, (6.2.10) gives the correspondence between the halfcircle r+ and the superior margin of the cut, and for -W < a < 0, the correspondence between r- and the inferior margin (fig. 6.2.2). Since from the extension by symmetry it results U(y, +0) = -U(y, -0), we deduce that U(a) is an odd function on F. With the same application (6.2.10), the functions A(y) and j(y) become even functions. We denote them by A(a) , respectively j(o). Now we shall see how the boundary conditions (6.2.6) and (6.2.8) are transformed. To this aim we remind to the reader that, after performing a conformal mapping, 208 THE LIFTING LINE THEORY the ratio of the lengths is given by the modulus of the derivative. More precisely, let W = f (Z) be a conformal mapping and let M(Z) and N(Z+AZ) be two neighboring points and M1(W) and N1(W+AW) their images. Obviously we have AMN' lim I 1 = lim I1Z 1 = lim I AZ = If'(Z)1 Hence, returning to our problem and denoting by N the outward normal to 1', we shall have dU_dU dN_dUIdW do dN (a) do TN- d Taking (6.2.10) into account, the boundary conditions (6.2.6) and (6.2.8) on the two half - circles r+ and r_ give dU (a) = 2/3I sin a) dN A(a) U(a) - J(a) sin a (6.2.12) Denoting W = it + i v, the harmonic function U(y, z) in the yOz plane becomes the harmonic function U(u, v) in the exterior of the circle r. This one vanishes at infinity and has the normal derivative (6.2.12) known on 1'. It is a Neumann problem. Its solution is given by Dini's formula (see for example (1.20] p. 31). We obtain U(u,v)=n J_" U(a)Inlbei°-Wlda+ko dN (6.2.13) being an unknown constant. Considering that W tends to a point be' ° from C, one obtains U(s) = b r+ir J JJJ aN (o )1n I2 sin s 2 a I da + kO . (6.2.14) : Since U(a) is an odd function it results dN (a) dN (-a) , and integrating only on the interval (0, 7r), we deduce U(s) b f0" ddN (a)S(s, a)d a, 7r (6.2.15) THE IN'I'ECRATION OF PRANDTL'S EQUATION 209 where we denoted S(s, o) = In (6.2.16) am 2 We have (see, for example, 11.16]): 00 S(s, o) _ -2 F sin ks sin ko km1 (6.2.17) k the series being absolutely and uniformly convergent (s 0 o). In (6.2.15) we did not encounter ke, because we have U(s) _ _ -U(-s) whence U(0) = 0. Taking into account the relation (6.2.12) in which 2U(o) is replaced by C(a) according to the condition (6.2.3), one obtains the following Fredholin-type integral equation C(S) = A J C(°) si) S(s, a)d or + Jo(a) , (6.2.18) where A= 2bQ z Jo(s) , 2b x J(o)S(s, o') sin o d o . (6.2.19) To Hence we reduced Prandtl's integro-differential equation to the Fredhoimtype equation (6.2.18). The kernel of this equation has an integrable singularity. Equivalent integral equations have been given by Betz and Gebelein in 1936 and Trefftz in 1938. 6.2.2 Existence and Uniqueness Theorems For proving the existence and uniqueness of the solution of Prandtl's equation , we shall use the first theorem of Fredholm. This may be enunciated as follows: the equation b i P(s) = A K(s,)p(o)ds + f(s), (6.2.20) fa has an unique solution for a given value of A and for every fee term f if and only if the corresponding homogeneous equation admits only the trivial solution V(s) = 0. Hence, we must show that the equation 210 THE LIFTING LINE THEORY (6.2.18) which is homogeneous (Jo = 0) has only the trivial solution. But the homogeneous equation corresponds to the boundary problem (6.2.12) which is homogeneous (J = 0). Applying Green's formula / JD U)2du dv = -JOD Ud S, (6.2.21) where D is an annulus, exterior to the circle r, bounded by an concentric circle of radius R > b and observing that for R -- oo the last term vanishes (see, for example, §6.1 from 1 fx A we deduce UdNda < 0. Utilizing the homogeneous condition (6.2.12), we obtain: W IAsin 0. (6.2.22) This inequality implies U = 0, because, as it results from the definition (6.1.17), we have A > 0. Hence, Prandtl's equation has an unique solution. This result is very important, because, as we have already seen in 6.1.4 and as we shall see in the sequel, we manage, on various ways, to determine a solution of this equations. The above result ensures that if we find a solution, this is the unique solution of the equation. 6.2.3 Foundation of Glauert's Method The integral equation (6.2.20) is a Fredholm-type equation of the second kind. This equation has a symmetric kernel if K(s, a) = K(a, a). (6.2.23) The equation (6.2.18) has not a symmetric kernel, but it can be symmetrized. Indeed, multiplying the equation with (on the inte- A) gration interval the quantity under the radical is positive) and, taking the function as an unknown c(8) = sins C(s) A(s) (6.2.24) 211 THE INTEGRATION OF PRANDTL'S EQUATION one obtains the following equation: c(s) _ -2a Ja c(a) VsTn,; i "in a A 00 sink sin ka a ()k_i d a + F;i(S) A(s) (6.2.25) 00 K=-2E k i1 sin s smo sin ressin rca k A(s) A(a) is obviously symmetric. The kernel is even degenerate, but not of finite rank. As it is known (see, for example, [1.22], vol.3, p.193), the integral equations with degenerate kernel may be reduced to infinite algebraic systems and the equations with degenerate kernel of finite rank may be reduced to linear algebraic systems with a finite number of equations. We are not in this situation, and according to (6.2.25)we shall take only the property of symmetry of the kernel into account. According to the theory of Hilbert and Schmidt (see, for example, (1.22] v. 3, p. 243) the solution of the integral equation may be expanded, with respect to the eigenfunctions of the kernel, into absolutely and uniformly convergent series. Hence the solution of the equation (6.2.25) has the form: 00 C(a) = A(a E At, sin ka. )k-1 Taking (6.2.24) into account, it results: 00 C(a) _ E Ak sin ka. (6.2.26) k=1 The (constant) coefficients Ak will be determined replacing (6.2.26) in the equation (6.2.18), or easier, performing this replacement in Prandtl' equation (6.1.18), which, with the change of variables y = b cos s, (6.2.27) 1 ) = b cos a and with the notation C(s) for the function C(y) composed with (6.2.27)1 etc., becomes 2b/3C(s + a(s) C'(a)d a= casa-cuss 2itbaOff( s s ) ( 6.2.28 ) Before discussing about how to determine the coefficients A, from (6.2.26) and (6.2.27), we must notice that the form (6.2.26) of the so- lution C(a) may result directly from (6.2.18), without utilizing the 212 THE LIFTING LINE THEORY theory of Hilbert and Schmidt. Indeed, taking (6.2.17) into account, the equation (6.2.18) becomes: _'inks C(s) = -2a fo" k=1 A(( )) in k sin a d a+ V +4rb sin ks Air) sin ko sin od a . I0 The integrals are constants and the solution will have the form (6.2.26). 6.2.4 Glauert'e Approximation We shall return now to the problem of determination of the coefficients Ai from (6.2.26). Replacing C from (6.2.26) in (6.2.28) and using Glauert's formula (B.6.6) (herein is the origin of this formula), we deduce: E Ak[2bfl sins + k7ra(s)) sin ks = 21rba(s) j (s) sins . (6.2.29) k=1 Glauert's approximation consists in keeping the first n terms from the expansion (6.2.26) and then imposing (6.2.29) to be satisfied for n distinct values of the variable s. The coefficients Ak are the solution of an linear algebraic system, but we cannot evaluate the error of the approximation. Many other approximations have been given in the literature (see Lotz in [1.24), Carafoli in [1.5)). 6.2.5 The Minimal Drag Airfoil The foundation of Glauert's method,which consists in establishing the formula (6.2.26) gives the possibility to give an answer to the following problem of practical interest: to determine among the wings with the same lift, that one which has the minimum drag. In view of this determination we shall calculate, utilizing the formulas (6.1.21), (6.1.25) and (6.2.26), the lift and drag coefficients cL and CD. For determining the lift and the drag, we multiply these coefficients by the same factor 1 2 p00UUA1. Since 2! sin ka sin lad v = irdkl , k,1=1,2,... 213 THE INTEGRATION OF PRANDTL'S EQUATION we deduce +b f Jb ct, = A C(y)d y = A f C(a) sin ado, = JO 'At v. (6.2.30) Utilizing Glauert's formula (B.6.6), we obtain: () C'(s)s a 1 1 4ab To Cos a - oos or such that: CD / 2 C(V)w(y)d, y = - b A °° 1 2A 1 (ksmka) ksI f 4b sin ka kAi sin a w(a)C(a) sin ad a = o0 Alsinio do = (6.2.31) !:1 oc kAj. k=t The formulas (6.2.30) and (6.2.31) indicate that among all the wings with the same lift (with the same A1), the minimum drag corresponds to the wings for which A2 = A3 = ... = 0. The solution of Prandtl's equation for these wings is C(o) =A1sin or =C(y)=At V, where Al = C(O) may be determined obviously from the equation (6.1.26). We have _C(_) A(N) 5 = 2w(y) + J(y) _ - 1 At + J(y). (6.2.33) In the case of the flat plate, j has the form (6.1.20). It results J = 2e whence we deduce that the member from the left hand side of (6.2.33) is constant. Hence, a(y) = ao 1 - , (6.2.34) the constant ao being determined by the relation 0 + 2b) Al = 2E, (6.2.35) 214 THE LIFTING LINE THEORY if one gives Al. The same relation determines Al if one gives ao. For example, when Al = k like in (6.1.28), it results ao = 1, like in (6.1.27) and vice versa. The expression (6.2.33) shows that the wings which have the above property are the elliptical flat plates. 6.3 The Symmetrical Wing. Vekua's Equation. A Larger Class of Exact Solutions 6.3.1 Symmetry Properties Very often in aerodynamics we encounter the case when the wing is symmetric with respect to the xOz plane. In this situation we have x*(y) = xt (-y) , -b < y S +b h(x, y) = h(x, -y) , (6.3.1) From (6.1.19) and (6.1.20) it results a(y) = a(-y), j(y) = j(-y). (6.3.2) Let us prove that we also have C(y) = C(-y) . (6.3.3) Indeed, changing in Prandtl's equation (6.1.18) y by -y and taking (6.3.2) into account, it results AC(-y) = a2 y) 1c (q) d q +.7(y) . Cc(q) n (6.3.4) y Putting in the integral n = -u and observing that C(q)d17 = dC = C'(u)du, we deduce r -16 C'(rl)d f b 1+31 _ '/'_b C(u)du =,+b Cl(,i)d+l J.fb -u+y b 7I-y (6.3.5) Introducing this relation in (6.3.4) and comparing with (6.1.18), we get (6.3.3). THE SYMMETIUC'AL WING. VEKt'A'S EQUATION 215 The Integral Equation 6.3.2 We shall present in the sequel the simplest method for obtaining the equation (6.2.18). The demonstration is inspired from [A.27), where, on his turn, it was taken from Magnaradze (6.16] and Vekua (6.28). With the notations (6.2.7) Prandtl's equation is I 'r+b C"(n) d 27r ,!-b rl - y n = 0C(y) - J(y) . (6.3.6) A(y) For the existence of the principal value we have to assume that C'(y) satisfies Holder's condition on the segment (-b, +b). We shall invert this equation assuming that the right hand member is known. As it is known from (C.1.1) the solution C'(y) depends on the behaviour imposed in the points ±b. We know that we cannot obtain a bounded solution in the two points without imposing a restriction to the right hand member. In the same time, because of the symmetry of the wing, we cannot consider C bounded only in an extremity. Hence C' is unbounded in the two extremities, i.e. the solution has the form (C'.1.11). Further, for inverting the equation (6.3.6), A(y) and J(y) have to satisfy Holder's condition on [-b, +b]. If A(y) and hs(x, y) (with respect to the y variable) have this property we deduce the same thing for J(y) . Moreover, a(y) must not vanish on (-b,+b). If all these conditions are satisfied, then, using the formula (C.1.11), we obtain C'([/) _ -2 'r+b 1 b2y2 b bz - I' [#'(") - J(n)] d n+ n-y A(q) (6.3.7) B representing a constant which has to be determined. It is zero because from (6.3.3) we have C'(y) = -C'(-y), whence C'(0) = 0. Imposing this in (6.3.7) and observing that the integrand is an odd function, it results the assertion. Utilizing now the identity P) --r12 d dy In -y2- 62-n2 i(y-n)+ i(y-n)+ b2-y2+ P- (6.3.8) 216 THE LIFTING LINE THEORY and integrating (6.3.7) on the interval (-b, y), from (6.3.3) one obtains: C(y) _ +b 2 W 1 110A(+l) b (6.3.9) i(y-tl)+ &2y2i(y-q)+ b -y + /bbl- -J(17)] In because the modulus is equal to the unity for y = -b. Performing the change of variable y = b cos s , tI = b cos a (6.3.10) and taking into account that in -y2- (y-tl)+ (y-17)+ b2-y2+ b2-q2 is - e is = In l e-i s _ eio I e - S($ a) we obtain obviously the equation (6.2.18). Using the notation A(tr-a) we have A(bcoe(7r-a)) = A(-boos a) _ = A(-y). Hence, taking (6.3.2) and (6.3.3) into account, it results A(tr - a) = A(a), J(tr - a) _ i(a), C(tr - a) = C(a) (6.3.11) whence: sin ad a = I"/2 [#A(a) - J(a)J In s+a --/z pc(o') - J(a)J In o cos cos 2 s - a sin ad a . 2 But, . sin In sin $-a 2 s+a 2 cos s+a 2 s-a cos 2 = In sins - sin or sin s+sin or THE SYMMETRICAL WING. VEKUA'S EQUATION 217 In this way, the equation (6.2.18) for the symmetric wing becomes 26 C(s) _ J"/2 [(;) (6.3.12) - sing -J(a) In sins sinada, sins + sin a J with s in the interval (0, n/2). Vekua's Equation 6.3.3 In 1945, I.N.Vekua [6.28] gave for the symmetric profile whose chord has the form bz a(y) = 2 -y with p(y) = p(-y) > 0, , P(y) (6.3.13) (where p(y) is an analytic function on [-b, +b]), a Fredholm-type integral equation which has the great advantage that it may be integrated exactly for a large class of profiles. Vekua's method was extended immediately by Magnaradze [6.16] to wings for which the function p(y) is not necessarily analytic on the interval f-b,+b]. Since we had not the occasion to read this papers, we present herein a a synthesis due to Muschelisvili [A.27]. To this aim, we write the equation (6.3.7), where we considered 13 = 0, as follows 20 A(y) C'(y) + +b C(11 f-b ' ' 1 (6.3.14) = A(y)J1(y) - tb a(y) b2 - y2 R(y,n)C(rl)d>j b where 2(3 R(y, rl) = Ji (y) = 1 (i-r a(n) 7r 17 - y 2 - _ a(R-7-ill y) (6.3.15) Ifb +b 62 - y2 J(q)d i1. (6.3.16) Ji (y) _ -J1(-y) . (6.3.17) rl - y Obviously we have: R(y, ) _ -R(-y, -y) THE LIFTING LINE THEORY 218 Further we shall assume the continuity of the first order derivative of the function P(y) = (6.3.18) a(y) In this case, R(y, i) will be a continuous function. Since according to (6.1.7) we have: C(q ay i-bb r1 11 = dq = o d y \Jb ` + Jy+a, v 11 tab do _ J-b 17 - dn, from (6.3.14) it results dy [A(y)C'(y)] + 2Q ' +b J-b rl (y d q B(y), (6.3.19) where B(y) = dy [A()Ji() - lr+b a(y) J R(y, n)C(r1)d n I . (6.3.20) Obviously, B(y) = B(-y) (6.3.21) Eliminating the integral from (6.3.19) by means of Prandtl's equation, we obtain the following differential equation: A(y) b [A(y)C'(y)] + 4/32C(y) = A(y) [B(y) + 4i3J(y)I . (6.3.22) Assuming that the right hand member is known, we have in (6.3.22) a differential linear equation for C(y). The homogeneous equation has the linear independent solutions cos s(y), sin s(y), where 20 8(y) = (6.3.23) Jo a(rl) Utilizing Lagrange's method of variation of constants, we deduce that the equation (6.3.22) has the following solution: C(y) = Co cos s(y) + Cl sin s(y)+ y sin [s(y) + 2Q f [B('l) + 4QJ(n)] - s(n)]d n, (6.3.24) THE SYMMETRICAL WING. VEKUA'S EQUATION 219 Co and C, being constants. Obviously, Co = C(0). Calculating C(-y), taking into account that s(y) is an odd function (its derivative is an even function) and B(q) and J(n) are odd functions, and imposing (6.3.3), it results C1 = 0. Introducing B given by (6.3.20) in (6.3.24), performing an integration by parts and observing that the integrated term is zero because JI(0) = 0, 1-b it results the following integral equation: +1 C(y) K(y, ii) = a f +b (6.3.25) K(y, q)C(t7)d n= g(y) , R(qj, q) cos [s(y) - s(rh)]d nl , (6.3.26) To +2 10yJ g(y)=Cocoss sins s d+ (6.3.27) + [iivi)ccs[a(v) - s(n)1 d q. The equation (6.3.25) for J9 = 1 is the equation given by Vekua and Magnaradze. Unlike the equation of Trefftz (6.2.18), this is regular (the kernel has no singularity). Moreover, in case that the function p(y) given by (6.3.18) is a rational function, more precisely in case that a(y) has the form: a(y)=a 0 -y21+ply2+...+pny2n 9 1+qly +...+gny2" (6.3.28) as we shall see in an example, the equation of Vekua and Magnaradze reduces to an algebraic finite system. This form for a(y) is suitable for approximating every wing of practical interest. We have to mention that, for the wings having the form (6.3.28), the case when qt = 92 = ... = q,, = 0, has been solved by H.Schmidt in 1937, 16.241, and the case when pi = pl = ... = pn = 0 belongs to a larger class, considered by the author of the present book in 1958, (6.41. For this class one obtains the exact solution. Before passing to applications we notice that if constant, then, taking (B.5.6) into account, we deduce g(y) = Co cos s(y) + 2k1, J(tl) = k is a (6.3.29) 220 THE LIFTING LINE THEORY where I (y) _ {sin[s() - s(rl)) - cos[s(y) - s(n)J q dq. (6.3.30) 6.3.4 The Elliptical Wing Denoting by a and b the semi-axes of the ellipse from the xOy plane, we deduce s 1-b V a(y)=aa b2-y2, as=a/b. Obviously, R = 0 whence K(y, r)) = 0. The equation (6.3.25) gives directly the solution C(y) = g(y), where g is calculated with the formula (6.3.29). In I one performs an integration by parts. Since from (6.3.23) it results 2fi s'(tl) = ira(y) we deduce +wao I= b2-y2-bcoos(y). Since from (6.1.20) and (6.2.7) it results J = 2e, using the notation 4wea o aao+2/9' we deduce C(y) = Co cos s(y) + k - kb cos s(y) . (6.3.31) For determining the constant Co we shall employ the condition C(b) = 0. Since from (6.3.23) it results s(y) = 20 wao arcsin b , we deduce Co = kb whence C(y)=k b2 -y2. For ao = 1/b one obtains exactly the solution (6.1.30). (6.3.32) THE SYMMETRICAL WING. VEKUA'S EQUATION 221 The Rectangular Wing 6.3.5 We shall consider now that a(y) has the form a( E/)=ao b2 - y 21+ (6 . 3 . 33) , the real numbers p and q being chosen in order to ensure only positive values of the fraction one [-b, +b]. In the sequel we shall see that one imposes pb2 > -1 whence qb2 > -1. From (6.3.15) and (6.3.33) we deduce R(nt, n) = c(q + Th) (1+pip)(1+pni)' 20(g - C p) (6.3.34) irao and from (6.3.26) K(y, r1) = ' wo(y) +'P1(y) (6.3.35) 1+pr7 where acs{s(by} -_aq(m)) 1 + MY) = c 1 1 d'ri (6.3.36) Taking into account that for pb2 > -1 we have d ,q q 1 + pb2 1 I (1+p) b2- 1+ arct&n from (6.3.23), for p # 0, we deduce s(y) 2Q f q arcein EI + _ 7rao Lp b 1 +'2 J p- q p arctan y 1+ pb2 l (6.3.37) and for p = 0, (I + s(y) _ L c) aresin b - Zy vfb2 --y2 l . (6.3.38) Replacing K(y, q) given by (6.3.35) in the integral equation (6.3.25) and observing that the first term vanishes because the integrand is an odd function, we deduce: C(y) + tPiny) r-b 1 C +(p dt = Cocoss(y) + 2k1(y). (6.3.39) THE LIFTING LINE THEORY 222 The integral is a constant Cl which may be determined by multiplying (6.3.39) with (1 + py2)-1 and integrating with respect to y on the interval (-b,+b). We obtain J III[ b 1 + P?1 1 (6.3.40) bbl+py2dy. -Col bb +a( dy=2A; Imposing (6.3.39) and the condition C(b) = 0, we deduce the relation Co cos s(b) - 91(b) CI = -2k1(b). (6.3.41) Determining the constants Co and Cl from the system (6.3.40) and (6.3.41), we find the exact solution of Prandtl's equation C(y) = Cocoss(y) - !C1SOj(y) +2kI(y). (6.3.42) Using the inverse method, i.e. considering various values for the con- stants p and q and calculating the form of the chord, we may find important wings for which the exact solution (6.3.42) is valid. So, in (A.271, considering q = 0 and pb2 = 0, 9, one obtains an almost rectangular wing (the variation of the chord versus the span is very small). Indeed, we have y/b a/boo 0. 1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1.00 1.02 1.03 1.05 1.06 1.06 1,03 0.95 0.75 Importantresults ofthis meth method are givenin [1.331. 6.3.6 Extensions Modifying Vekua's method, we managed in (6.4) to give the exact solution of Prandtl's equation for wings whose chord satisfies the relation ys a(y) = P(y) , (6.3.43) where p(Z) is a holomorphic function in the Z = y + i z complex plane, excepting the vicinity of the point at infinity where one admits 223 NUMERICAL METHODS the following series expansion: k>O p(Z) =n=-oo E PAZ" (6.3.44) . The determination of the solution of Prandtl's equation is reduced to solving a Hilbert-type problem, whose exact solution is given. The polynomials belong to the class (6.3.44), whence the great importance of this solution. According to Weierstrass's theorem (see for example, [6.13], p.61), every continuous function on the interval [-b, +b] (i.e., every possible form of the wing) may be approximated by polynomials. Practically, for this one may employ an interpolation method (for example, Newton or Lagrange's method). Even if p(y) is a polynomial, in this method it is not necessary to be symmetric, like in the theory of Vekua and Niagnaradze. Numerical Methods 6.4 6.4.1 Multhopp's Method The idea of biulthopp's method consists in approximating the func- tion C(a) by the trigonometric polynomial P"(c) obtained by the Lagrange interpolation in the basis (sinkc)kl,...,n. For determining P", we notice that after introducing the matrices ST = (sin c,...,sin nc), aT = (al,...,an), it may be written as follows = P. n (c) ak sin ka = sTa. (6.4.1) kal The points where one imposes for PP(o) to coincide with C(o) (the nodes) are given by the uniform grid aI n+1 -o, a2 2a n+1 -2a,...,c"= na n 1 =nor, (6.4.2) whidi is usual in the theory of interpolation [6.13] p.20. These are equidistant on the half-circle with the diameter on the span (fig. 6.4.1). The points xJ = cos a1 (6.4.3) THE LIFTING LINE THEORY 224 +z x -1 1 b y Fig. 6.4.1. are the zeros of the Chebyshev polynomial of order two (F.2.8) on the interval (0,7r). Denoting c; = C(a1) we have to determine the matrix a from the system Sa = c, (6.4.4) where sin ai sin 2a1 ... sin nal S = sin a2 sin 202... sin nag c= (6.4.5) sin a, sin 20,E ... sin na One obtains [6.131, p.21 S-1 = IS. n2 (6.4.6) NUMERICAL METHODS We shall present at the end of the section this calculus. Hence, n+l sin v1... sin na1 Cl ... sin na c (sin a...sin no) I sin on 1Ecinko l 2 n+ (6.4.7) k=1 1(sin a... sin na) ctsinka 2 + sin koj s i n j=1 k=1 For C(a) we have the Wowing expansion: C(a) = n +2 1 n n ck sin ka j sin jo. (6.4.8) We must notice that this expression could be also obtained from (6.2.26) approximating the Fourier coefficients according to the definition of the integral for an equidistant division a. Indeed, f om (6.2.26) we have: C(Q) sin yoda Aj TO +1 k,i C(ck) sin jai, . Since jak = jka = kof, eraplaying Glauert's approximation Al sin jo, C(Q) _ j-1 one obtains (6.4.8). Utilizing Multhopp'e expansion (6.4.8), from Prandtl's equation (6.2.28) 226 THE LIFTING LINE THEORY and from Glauert's formula, we deduce: sin s 4b,3 n+1 n n ck > sin jka sin js+ k=1 j=1 (6.4.9) 27r s sins. sin jka sin js = 2irba s j=1 k=1 Giving to s the successive values La(£ = 1, ... , n) and taking into account the formulas (6.4.17) and (6.4.19) from below, we deduce the system (£ = I,-, n), BtkCk = Bt, Oct + (6.4.10) k=1 where 2Bt = (n + 1)ira(£k)j(£a). (6.4.11) Fbr k $ t, Btk - lra(£a) 2bsin£a I - (-1)n-t 8 _ I sine (k + t)a 2 1 sine (k - £)o` ' (6.4.12) 2 and for k = e, Att = 7ra(£a) n(n + 1) 2b sin la 4 (6.4.13) Determining the unknowns c1, ... , c from the system (6.4.10), one obtains the solution of Prandtl's equation from (6.4.8). Since Atk vanishes when k - £ is an even number, the system (6.4.10) may be separated; more precisely, the unknowns with odd indices may be expressed by means of the unknowns with odd indices and vice versa. This fact was proved in [1.2]. As it is already mentioned in (1.2] and in (6.17] an iterative procedure for determining the unknowns is also established. The author had not at his disposition this paper. The procedure simplifies in the case of the symmetric wing (C(y) = C(-y)), i.e. when ck = Cn+l _k (k = 1, 2, ... , [(n+ 1)/2]) or in the case of the antisymmetric wing (ck = -c,+1_k). In the first case the system reduces to [(n + 1)/2] equations, and in the second to ((n - 1)/2]. The square brackets indicate the integer part of the number from the interior. 227 NUMERICAL METHODS In the sequel we shall calculate the sums that intervened in the above formulas. Let it be for the beginning n r j=1 i=1 11.Ecosrja, 12 - Esinrja. (6.4.14) Denoting z = e1r`r, for r 96 0, we obtain _ z - zn+i - = l- z 1 whence, separating the real part from the imaginary one, I1 = -1 + 1 - (-I}r , 2 12 = 1 - (-I)r 2 cot ra . 2 (6.4.15) One obtains the sum n r Ejcosrja- 1 j=1 4 I era sin 2 (6.4.16) 12 with respect to a. We deduce therefore, noticing that k - f and k + f are odd or even simultaneously, deriving n n 2 E sin kja sin eja = E cos(k - e) jaj=1 j=1 (6.4.17) n -Ecos(k+f)ja= (n+l)6kt. j=1 With these formulas, the relation (6.4.6) written as follows S-1S n2 S2=1 may be immediately proved, since we have: n Sz n _ E sine ja F sins ja sin 2ja ... sine ja sin nja j=1 ial j=1 (6.4.18) THE LIFTING LINE THEORY 228 Utilizing now (6.4.16), for k q& a we deduce n 2 E j sin kja sin eja = j=1 (6.4.19) 4 sin2 and for k = e: (k+2 e)a sin2 e (k 2 n 4 E j sine kja = n(n -}- 1). (6.4.20) j=1 Analogously it results: n 2 E sin jka sin ja = > cos j(ka - a) - > cos j(ka + a) _ j-1 j,1 j -Re z-Zn+1 Z-Zn+1 _Re 1- f I-Z s,,01(ka+.) (-1)k+1 sin(n + 1)a sin ak COs a - COS ak (6.4.21) If we utilize this identity, for the formula (6.4.8) which gives the solution of Prandtl's equation , we obtain the final form C(a) __ 1 n+l n k=1 k+1 ` sin(n + 1)a sinker . coo a - cos ka (6.4.22) In [A.23], p. 98-111 one gives a mathematical justification of this method. More precisely, one demonstrates that under certain circumstances, the iterative procedure that one utilizes for solving the system (6.4.10) is convergent and the solution (6.4.22) converges uniformly to the solution of Prandtl's equation. 6.4.2 The Quadrature Formulas Method In [6.5) we gave a numerical method for solving Prandtl's equation by means of Gauss-type quadrature formulas. It is well known that these 229 NUM ERICAG METHODS formulas give the best approximation. The key of this method consists in writing Prandtl's equation in the form (6.1.21). With the change of variables y = bs , q7 = bz (6.4.23) and, keeping the notations C(s), a(s) and j(s) for C(bs), a(bs) and respectively j(bs), the equation (6.1.21) becomes . /3C(s) 2b) 1 J-1 (z (s)s d x + j (s) , (6.4.24) and the conditions (6.1.7), C(±1) = 0. (6.4.25) The solution of the equation (6.4.24) has the form C(s) = 1 -_3 2 c(s) . (6.4.26) Employing the quadrature formulas (F.3.5), the equation (6.4.24) reduces to the algebraic system n EAkici =7k, k = 1nn, (6.4.27) i=1 where we denoted ak = a(zk), Ck = C(xk), era k A 2b(n + 1) (-1):+kl 1- Akk=fV1 -xk+6 jk = j(zk) (x1x2 -1 i#k, zk)2 (6.4.28) n41; zi=Cosn+1,t I'n' the unknowns being c1,... , c,t. The system has to be studied theoretically and solved numerically using a computer. After determining the unknowns, the lift, drag and moment coefficients, defined in (6.1.21) and (6.1.25), CL C= 2b win the form f +1 1 - s2 c(s)d s, cD = - A 1 = mA J 1 `s 1 - s2 c(s)d s, c; r1 ZA- j 1 - s2 w(s)c(s) ds, 11 s 1 - s2 w(s)c(s) ds, (6.4.29) 230 THE LIFTING LINE THEORY give with the formula (F.2.12) CL = (n + 1)A E(1 i=1 - x?)c" CD = - (n + 1)A isl " _ {n + 1)mA E(1- x; 2)xc:, 27rb2 i_I x?)w,c{ >, 2nb2 2 c= (n + 1)mA E(1 - xd )x:wig i_1 (6.4.30) where w; = 13 VI - x, . 2aai (6.4.31) In (6.4.30) we used, like in (6.1.21), the notations A for the area of the domain D and m for dimensionless length of the mean chord (in the direction of the unperturbed stream). For verifying the method, we applied it in [6.5) to the elliptical Sat wing, for which the exact solution is known (6.1.28), (6.1.30). With the notation (6.4.26) it results c(s) = k. Putting in (6.4.27) Aki = 1- x4Ak;, c; = k4, it results that the system Ak.;ci = Q+ it/(2b) (6.4.32) must have the solution dl = d2 = ... = c;, = I. For b = 10 one obtains numerically dl = 4 = ... 1000. We deduce therefore that the proposed method is very good. In the sequel we shall give the numeric solution for the rectangular wing (in this case there exists no exact solution). Taking the reference length Lo in the definition of the dimensionless variables (2.1.1), to coincide with the half of the chord, we deduce xt = :L1. It results a(y) = 1 and j(y) = 27rc. Putting c; = 27red', the system (6.4.27) becomes Akic! = 1, k = I- n, (6.4.33) where Ak; are given by (6.4.28) where we put ak = 1. The quantities NUMERICAL 1METHODS 231 of interest in (6.4.30) are 1 kl =_ 71+1 (1-xi)c', k2=-.1 (1-x?)3,2(c,')2, (6.4.34) V(1-x?)x,c;, k4= n+1 i_i A n+1 E(1-x?)3/2x,(cI')2. i=1 One obtains CL = 7r2ek1 , CD = 7r2e2(kl - k2), (6.4.35) 2c= = 7r2bek3 , 2c; = 7r2bC2(k4 - k3) . For b = 10 we obtain the following values: fi 1 0.8 kl k2 k3 k4 0.2347 0.2544 0.0907 0.0856 0 0 0 0 The result c= = c. = 0 is natural because of the symmetry of the wing. The lift and the drag increase because of the compressibility. This result is also natural. The value 2.307c obtained here for cL in the case of the incompressible fluid is smaller than the values 7.29 e, 5.28 £, ... , obtained with Glauert's method [1.12], but the values obtained with Glauert's method come closer to the values given here if A(= 2b/m) increases, i.e. the span is great with respect to the chord. Just in this situation the lifting line theory is valid. We may think therefore that the method we have just exposed is at the same time very simple and very efficient. 6.4.3 The Collocation Method The simplest numerical integration method is certainly the collocation method [6.6]. In case that Prandtl's equation has the form (6.1.19) and satisfies the conditions (6.1.7), the solution has the form C(y) = b2 - y2 c(y) . (6.4.36) According to the collocation method, the segment [-b, +b] is divided into iV elements L; and the function c(y) is approximated on each 232 THE LIFTING LINE THEORY element with its value c; from the mid - point y° of the segment. So, the equation (6.1.19) gives , N [ 4 c(y) = a(y) 2,8 fd ix1 (b - y)2 d rI + 2j(y), (6.4.37) where, as we have already stated, Cj = c(y?). Imposing this equality to be satisfied in every point yk, k =I,-, N, one obtains N 2 Vr2b2 - yk2ck = ak E Aloe: + 2jk k = I , , (6.4.38) i=1 where ek = c(yk), ak = a(yko), ?k = ?(yk), Ak, - w+1 f J(y (6.4.39) b-q - y y;+1 representing the extremities of the segment L;(y1 = -b, yN+1 = = b). So, (6.4.35) represents an algebraic linear system consisting of N equations with N unknowns cj. For calculating Ak{, we notice that for e V (a, ?'J we have I a ( 7 7-e )2 d , y-e a -- e + nee In [ (ry-e)2 + ( b -e + +aresin a b2_.72s)2 (a-e) - ry b --a- 7 (6.4.40) For e E (a, 7) we shall write II = o - (71e)2 d q b2-772 s =J 66 (n - e)2 v° 'I d1 (6.4.41) T (77 - e)2 d b -V2 7 (rl - e)2 d ij. We use the formula (6.4.37) for calculating the last two integrals. Since (D.3.9) r-b (q - e)l d n= -7r , (6.4.42) 233 NUMERICAL METHODS it results that I1 is also determined. Utilizing these formulas, for k 96 i we obtain: Aki V - y? b2 y - 77i+I yk0 yi+1 - yk yl (yi+1 02is b2 72 + (Vb2_YA. b2-t/i+l In y` (Vi-Yk)2+ yi +arcsin (\ b2-yk2 - yko yi+1 - yk y,2 + + R--i/i+l - -Eli F1 b2 y,2 P (6.4.43) and for k = is y2 b2 Aii =- -7r - Ii yt I - Vbi b2 - y<+1 y2 + + (y +1 I Vb2 - y2 +arcsin b2 1,2+ - yI `2 Vb2-y°2+ b2-y, ) / 2]n t? + (VFb2 2b (6.4.44) Vb2 y°2 + tJ+i where we denoted 2ti = yi+1 - yiFor testing the method one utilizes also the elliptic flat wing. For this wing the exact solution is (6.1.28) with k determined in(6.1.30). Comparing (6.4.33) with (6.1.28), it results c = k/b. For E = 0.1,,0 = 1 (the incompressible fluid) b = 10 and N = 50, the numerical values obtained for ci are situated between 0.050 and 0.053, and the value of k/b is 0.052. For E = 0.0872, A = 1, b = 20 and N = 50 the values obtained for ci are situated between 0.0243 and 0.0252 and the value of k/b is 0.0252. The results given by this method are more accurate when the angle of attack is small. The accuracy also increases when N increases. THE LIFTING LINE THEORY 234 Since the method is simple and can be easily applied one can accept the error of order 10-3 that it contains. 6.5 Various Extensions of the Lifting Line Theory The Equation of Weissinger and Reinner 6.5.1 In the sequel we shall study new integral equations in order to improve Prandtl's model. In the past the researchers did not pay too much attention to these equations because the integrands contain both the unknown C and the derivative C'. Now we may use successfully these equations because some fundamental results concerning the Finite Part are known and, as we have mentioned in 16.5], after replacing the Principal Value of an integral which contains C' and has a Cauchy type singular kernel by the Finite Part of an integral which contains C and has an hypersingular kernel, we may employ the numerical methods. First we shall study the equation of Weissinger [6.30] and Reissner [5.29]. For obtaining it, we return to the lifting surface equation (6.1.28). Taking (6.1.6), (6.1.11) and (6.1.12) into account, we deduce +b rlydn. l Cdn-7dt J-e 9 ff bod JJ dq+2 (6.5.1) Utilizing (5.1.35), from (5.1.28) it results the first form of the integral equation r+b fi(n) d n + 1 ri - y 1'rj (1 4w JJD lJo \ _ z o d td r1= rt, (x, y). R (6.5.2) Multiplying this equation by (6.1.15) and integrating it on the interval [x_(y), x+(y)] one obtains j+b a(y)C(rl)d71+ brl - Y + 1 +b 2a J-b +(v) x+(n) Tx-) x-x-(y) Mm) (i_i) dx d d E rl f(v), x+(v)-x (6.5.3) j(y) being (6.1.18). 235 VARIOUS EXTENSIONS OF THE LIFTING LINE THEORY Further we shall perform an approximation which consists in replac- ing f(f,9) by C(n) A(n) f (f, rl) x+(n) - f (6.5.4) This is just the solution of the two-dimensional problem when h' is constant. Indeed, according to the formula (C.1.9), for a given y , the equation of the two-dimensional problem 1+(v) f(f,y) T ( 6 .5.5 ) = hx (x,y), n ,-(,) f - x becomes, in case that h' does not depend on x: f (x , y) x+(y) - x = -h(y) x - x_(y) (6 5 6) . . Multiplying (6.5.5) by (6.1.15), integrating the result over (x_ (y), X+ (y)] and utilizing the formulas (B.5.4) one obtains: -+(v) C(y) __ - J:-(v) (6.5.7) f (f, y)d f = ira(y)h'(y) Eliminating h' from (6.5.6) and (6.5.7) one obtains (6.5.4). With this approximation, the equation (6.5.3) becomes: a(y) J bbC n)di? +27r YO T-b (6.5.8) TO where N (y, n) :+(v) _+(») xo rIl R) x+(ti) - F x- x-(y) - x-(+)) x+(y) -- x dxd f . (6.5.9) Using the notation r)) No ( y, n) = N(y, A(y)A(n) = 1 x(y) f A(y)A(q) J=_(vl 1 (6.5.10) _+(,) J=-(») x+(r1) - f x -x_ (y) f- x-(rI) V x+(y) - x d xd f, THE LIFTING LINE THEORY 236 we may write the equation (6.5.8) as follows: CYO - f eb )dn+ 2A fC o N°(b,n)dn = J(y). (6.5.11) and J(y) being defined in (6.2.7). This equation has been deduced by Reissner [5.29). Other simplifications are performed in the papers [5.24] [6.30] in order to obtain an approximate solution. The A(y) difficulty consists in the presence of both C(n) and C'(n) in the integrand (the equation is integro-differential). Employing, as we have already shown, the Finite Part, i.e. taking (6.1.12) into account, we may write the equation (6.5.11) as follows C(q) 2 _1 y! `Ni(y, n)d, = J(y) , (6.5.12) where we denoted NI = No - 2. This is the equation we are going to use in applications. This is an integral equation. 6.5.2 Welssinger's Equation. The Rectangular Wing In the case of the wings for which the lifting surface equation cannot have the form (5.1.41) (this class is established in 5.1.7, and it contains the rectangular wing) one obtains an equation which may be easily utilized in applications. This equation was given by Weissinger [6.31), and received his name. Starting from the definition (6.1.6), we have: -i',}(n)f (x+(n) n) + x'-(n)f (x-(n), vi)- f - f=-(") x+(+r) a f 8dt rx+(n) 8 f _ J=-0) (6.5.13) dt. We are going to show why the sum -x+(n)f (x+(n), n) + x' (rl)f (x- (n), n) can be neglected. The first term vanishes by virtue of Kutta-Joukowski condition. The last term vanishes for a straight leading edge, perpendicular on the Ox axis (for example this is the case of the rectangular wing). It also can be neglected when the leading edge is slightly curved (x'_ (q) = 0). In general we can neglect this term considering that we have prolonged the theoretic leading edge in front of the real leading 237 VARIOUS EXTENSIONS OF THE LIFTING LINE THEORY edge and in this zone f = 0. With this approximation, the equation (5.1.41) may be written as follows: '+b 1 C'(17) Ud,1 -a 2w AD 09 xo -d fd =- "(x, N) (6.5.14) We make a second hypothesis, having in view to be satisfied in the case of the rectangular wing of width, let's say, 2a, namely we substitute a' instead of zo from R. This approximation is justified by the fact that (x - jj varies from 0 to 2a, hence on the greatest part of the domain D we have IxoJ = a. So, the equation (6.5.14) becomes: '+b 1 27r 1-6 dn - we - f 1 27r f+A "+ _. aq (6.5.15) x0yo Multiplying by (a +- x a - x) and integrating with respect to z on the interval (-a, +a) one obtains a -d n+ L jr a C 'm 2Ja 2 30(y) _ -a -a C (rl)d rl rah(y) , a±xh"(x,y)dx. (6.5.16) (6.5.17) Introducing the non-singular kernel K(pa) = f, fd a+ _ a yo go (6.5.18) the equation (6.5.4) may be written as follows C' r7) Ir YO n+ Zax f C'(rj)K(ya)d tj_ jo(y) . (6.5.19) This is the definitive form of Weiseinger's equation. It was created mainly for the rectangular wing. Before presenting the method of integration of this equation, we have to notice that it is not necessary to determine C(q). It suffices to find C'(t) because from (6.1.9), integrating by parts, it results CL b yC'(y)d y, es = -aQ = -- A J b y2C`(y)d y. (6.5.20) 238 THE LIFTING LINE THEORY The solution of the equation (6.5.19) has the form Cr(y) tiP(J) (6.5.21) with sp satisfying the equation '''+6 1 d i7 b2 b 2 Yo 1 + lira f (t ) h (Ju)d q = lo(U) . (6.5.22) b2 - rf2 6 Performing the change of variables 'i = brl', y = by' one obtains 1 rrb 'r+t J 777=7 it y sp(rl) 1 d>> ;p(aT) 27ra J i i - >> h0(bJo)d n = .?o(bJ) (6.5.23) This is Weissinger's equation. It looks like the generalized equation of thin profiles (C.2.1). The numeric solution is determined by means of the formulas (F.2.5) and (F.2.6). 6.6 6.6.1 The Lifting Line Theory in Ground Effects The Integral Equation The lifting line theory in ground effects is obtained from 5.3, as well as Prandtl's theory is obtained from the lifting surface theory. We are not interested in finding the velocity and pressure fields. The integral equation is obtained from (5.3.13) utilizing the formulas (6.1.11)-(6.1.13) and the relation (6.1.8). One finds 1 T7r +b E + f.?+ (y) f y, y) ! d d U f $ b 7-y y n (y) + (6.6.1) 1 +b +2- J C(t)i(x, yo)dTl = 2/is(x, y) , where d2fl2x d2 - yo yo) _ (d2 +y()2 )R3 + (dl + y02)2 ( \1 + R2 = x + e2 , e2 = 02(y2 + d2). x RI) ' (6.6.2) 239 THE LIFTING LINE THEORY IN GROUND EFFECTS Multiplying (6.6.1) by (6.1.15) and integrating the relation just obtained with respect to x on the interval (x_(, ), x+(rj)), we obtain, taking (B.5.4) into account: C'(11) d 2#C(y) + a(y) Lb Y-17 q r + ,l C(n)No(y, yo)d n = 2j(y) , (6.6.3) b where we denoted s+(v) x - x_ (y) N(x, J:^(v) x+(y) 1 No(y, yo) (6.6.4) yo)d x, j being (6.1.18). The equation (6.6.3) is the lifting line equation in ground effects. It was given in (5.8J. It is obviously an integro-differential equation which generalizes Prandtl's equation (6.1.16) (Um N = 0). Further we shall make some considerations concerning the kernel No(y, yo)-Taking (5.3.11) into account and introducing the functions Iv(y, yo) _ - +(v) r J z+(y) (yx (xz + e2)"/2 (v) d x, v = 1 , (6.6.5) we obtain: No(y, yo) = (a + y)2 [I, (y, yo) - a(y)] + #2 13(y, yo). (6.6.6) Performing the usual substitution x = s(y) + a(y)t, (6.6.7) where s(y) = x+(y) x-(y) , a(y) = x+(y) X-(Y) (6.6.8) 2 2 the integrals I become +i 1 IN = -na(y)s(y) J l+ t 1 t dt P(t)v aZ(y) 7r r+i .!-1 -1+-t td t 1-t P(t). , (6.6.9) where P(t) = a2t2 + last + e2 + s2 . (6.6.10) It is well known that the best approximation of these integrals is given by the Gauss-type quadrature formulas (F.2.24). With this form (6.6.9) one may study the asymptotic behaviour of the kernel No(y, yo) given by (6.6.6) depending on the parameter A2 = d2/b2. 240 THE LIFTING LINE THEORY 6.6.2 The Elliptical Flat Plate Without considering the ground effects, the solution of this problem is (6.1.28), (6.1.30). In the sequel we shall determine the influence of the ground. We assume that the wing has the equation x2/e2 + y2/b2 = 1. We deduce xf(y)=fb b2-y2, a(y) =b Vb2 - y2s(y) = 0. (6.6.11) Assuming that the span is much larger than the chord (e << b), we shall neglect the terms of order (e/b)2. From (6.6.9) we deduce I = 0, whence ` Integrating by parts, we deduce: f +b +b C(n)No(y,yo)dn=-a(y)f C'(n)&?yodn, b b and the integral equation (6.6.3) becomes: 2fC(y) + a(y) J b C,n) d q - 2a(y) J bb C'(1) d2 + y02 d n = 4_-Tra(y). (6.6.12) For d - oo one obtains Prandtl's equation. (6.1.16). Taking into account the shape of a(y) from (6.6.11), we shall look for solutions of (6.6.12) having the form C(y) = k b2 - y2. (6.6.13) The first integral was calculated in (6.1.29). The second may be calcu- lated with the substitution y = by, n = biz. If b < < d we neglect the terms O(b/d)4 and we obtain f b 2ir \a)2 b (n)d2+yody=k 2 We assumed that the span is much smaller than the distance to the ground. Replacing (6.6.13) in (6.6.12), we obtain the relation which determines k: (b `2 Tr = 1Ezr6 . (6.6.14) k 2p+ -n b - -I 241 THE LIFTING LINE THEORY IN GROUND EFFECTS Denoting by k the value of the constant when there are no ground effects, i.e. (6.1.30), we deduce k=kak, where (6.6.15) 2f3 + (t/b)ir k0 = 20 + (t/b)ir - (e/b) (b/d)27r (6 .6 . 16) Obviously, k > 1. The lift, drag and moment coefficients are given by the formulas (6.1.21) and (6.1.25), where for w(y) = w(0, y, 0) we have 15.81: W(y) C ,(n) J 47r bb d rl - 4n C'(Tl) b & yo d n. (6.6.17) These formulas (6.1.21), (6.1.25) and (6.6.17) are valid for every shape of the wing. For the elliptical flat plate wing we obtain: w(y) = - 4 1 2 + C ! (6 .6 . 18) such that cL=kocL, cD=klcp, c==c.=c, with the notation (6.6.19) 2 k1=ko li(1+2 )>1, (6.6.20) lift Cr and cD representing the respectively drag coefficients in the absence of ground. Obviously, both the lift and the drag are increasing in the presence of the ground. The coefficients ko and k1 depend on M, t/b, b/d. The numerical calculations from 15.81 show that for the lift the increase is not significant but for the drag it is considerable. The ground effect is a decreasing function of d. For the same values of the ratios t/b and b/d the influence coefficient ko is an increasing function of Mach's number M. 6.6.3 Numerical Solutions in the General Case Utilizing the formula (6.1.12), one may write the equation (6.6.3) as follows: 2#C(y) - a(y) f +b b c(n) d n+ (n - y)2 f +b C(n)N(y, yo)d n = 2?(y) b (6.6.21) 242 THE LIFTING LINE THEORY As it is known, this is an integral equation, not an integro-differential one, but the singularity is stronger than in (6.6.3). In (6.6.3) we have a Cauchy - type singularity and in (6.6.21) we have to consider the Finite Part of a hypersingularity. But for this kind of equations there are available quadrature formulas. In order to apply this method, we have to perform the change of variables y = by, q = bq' for calculating the integrals on the interval (-1, +1). We obtain 2bIC(y) - a(y) f +t C(n) Y)2 d n + b21 C(i7)N(y, yo)d = +1 1 (n - j(y) 1 (6.6.22) Since the solution of this equation has the form C(y) = V171- y2 c(y), (6.6.23) we obtain 1 - yj cj - aj 2b/3 1 --17C(17) d n+ Ti (n - yj) 1 1 - i2c(yl)N(yj, yj - n)d n = 2bjj , yj = cos n +b21+ + 1, t j=T . (6.6.24) Using (F.3.5) and (F.2.12) one obtains the system: n Ajcj + E Ajkck = 2bjj , j = 3n-, (6.6.25) k=1 where A.;= MO 1-y;+aj7r Aak -a L n+1 2 b2k. 7r n + 1 - yj)2 + 1(yk - (-1)'+k n + 1 N(yj, yj - yk) (1 - yk2) J (6.6.26) In the first term from Ajk one excepts k = j. The system (6.6.25) is solved numerically. 6.7 6.7.1 The Curved Lifting Line The Pressure and Velocity Fields In this subsection, we shall pay a special attention to the aspect ratio A = (2b)2/A introduced in (5.4.1). Usually, if A is small, one applies 243 TILE CURVED LIFTING LINE the theory from 5.4 concerning the wings of low aspect ratio. If A is large one applies the lifting line theory. These are the two asymptotic theories of the lifting surface theory. As it is known, one of Prandtl's hypotheses consists in replacing the domain D by the segment [-b, +b] taken along the span (the Oy axis). This hypothesis is plausible for the wings having the shape of an ellipse, triangle, trapezium or rhombus (see fig. 6.7.1) but it can be the source of great errors in the cage of the wings having the shape of a swallow tail or the shape of an arrow. In the first case it is natural to replace the wing by the curvilinear median (see fig. 6.7.2), and in the second case one approximates the wing by the median broken line (fig. 6.7.3). For birds, the nature preferred the curvilinear median. These are enough reasons for studying in this subsection the curved lifting line. In YA b j'I i Y Y. b i b rt 1 1 0 x -4 i t aX 0 1 1 b) 1 Fig. 6.7.1. Fig. 6.7.2. Fig. 6.7.3. 244 THE LIFTING LINE THEORY this case, one starts too from the general representation (5.1.8)-(5.1.12) and Prandtl's hypotheses. We assume therefore that the wing is without fl = 0) and that the unknown is C(y) defined thickness (hl = 0 by (6.1.5), with the conditions (6.1.7). When the domain D reduces to the curved line r (fig. 6.7.2) having the equation x = x.(y), the formula (6.1.8) is replaced by r` / JD f lim / il)k(x, y, z, t, q)d d y = - f +b k(x, y, z, x. (q), n)C(i )d +1, b (6.7.1) for x-(n) -. x. (n) - x+ (n) So, with the notation R. = ix - x.(n)]2 + A2(y02 + z2), (6.7.2) the formulas (6.1.9) become z P(x,y,z) _ -4 C(ri) z dn. (6.7.3) rt v(X' Y, z) = - I :Fir =fl lim. o-)] J 1D f (C, n) gEl l yob z .1 JD f (C W 1+ RI (1 + R- ) z y2 + z2 d d )dtdn+ +XOZ C(n)11+x(n)lal\ l Jb 47r L o+z2)dn- JJ p2yo +6 [x - x.(n)]z - T7r 14 C(rl) yof-' R3 d n 1 (6.7.4) Utilizing now the identity r x - x` R. R _ 8( 8 ) &1 z z z(x-x') /2yo y x-x.`1 R* jl + z2 R3 _ (6.7.5) 32z , THE CURVED LIFTING LINE 245 we obtain, after performing an integration by parts, j f v(x, y, z) = +b CI(rl) f1+ x R'(q)] d,,R. C (6.7.6) Q2 47T / +b J b zZl= C(i) R. d q . Analogously, w(x,y,z) =wi +w2, (6.7.7) where =4a fffri)_ j+bx R, x}m.(_)ddtl= (n) d n 4nx*-x.JJDf(C v?) 2[y02 W-2 47r xlim- . If,,, f (C rT) l 1 + Rt 1 lr ,ffDf( , l) o+Z2 YO X02 lice = art l 6r+.2) d d'r (6.7.8) (_)dd= R 1 r+b 4 J b C(") ['x_x,(,i)l 09l ( T2 + Z2 l d YO Y7+ J +Q rb c(n) yo +'Z2) dn. Introducing the identity $2(x _ R. yg +Z2) 110+z2 1 T3 (6.7.9) Jyoyo ly +z2 [1+x-x.(n)1I+Q2yO2X1 Rl J Rs ) 246 THE LIFTING LINE THEORY and, integrating by parts, we obtain u2. In fact, it results: x w(x' y, z) ('l) d 4a fb C(r1) R3 +b -4 f b+6 q+ 4n bb C(17) - --d n- r (+I)] d>1. y0+2 I1 + x r b l` (6.7.10) 6.7.2 The Integral Equation We start from the lifting surface equation having the form (5.1.28). Utilizing (6.1.6) and (6.1.12), we deduce: Y=dtdn Eb (q) 0 (6.7.11) dr= C 1 =;F +b C'(rl)dq. 1 4'rJ-b 11-y 0 b Taking (6.2.1) into account, we obtain: 12 = - lim 1 47r 11D f( x0 d l;d R To = (6.7.12) C(n) x-x.(rl)drl, 1 R° where R; = [x-x.(n)12+(.32y . (6.7.13) Utilizing the formula (D.3.7) and taking into account that C(±b) = 0, we deduce Iz 1 'r+6 41r ,l -b n 8r 1 u ON {C(rl) x- x. (n)1 Ro. Jd_ (6.7.14) 1 '' +6 C11) -x- 4ir.!_b n-y x. Ro d'l+la 247 THE CURVED LIFTING LINE whore 1 13 = - Co 8 - n v 8n _- d *! R° (6.7.15) C(n)xx x(n 4(n)dn 4sr From (5.1.28), (6.7.11), (6.7.12) and (6.7.14) we obtain the equation: t bb K(x, y, n)d n + 4A ,!_b n (ny air C('i)L(x, y,,7)d n = hs(x, v) (6.7.16) where K(z,y,n)=1+ (6.7.17) x - x. (n) - ray'. (n) L(x, y, (R.) are non-singular kernels. The equation (6.7.10) was obtained in another way by Prosadorf and Tordella (6.22]. It is a singular Integro-differential equation. For the straight line (x.(q) : 0) one deduces Ci(r!) Ib n +b 47r j' C'(n) x - yd n + Oar J iJ n - y Rd n - 4 x C(q) n = h: (6.7.18) x+ . This equation is a first approximation of the lifting line equation. Fbr deducing the equation (6.1.16) we had A < < yp on the greatest part of the domain D, such that we might where R consider Re = i4jyo(. Here we cannot perform this approximation. 6.7.3 The Numerical Method Using (D.3.7), one denaozutrates the identity J-bb (n - y2K(x, y, n)dn = Jn vK(x, wn)dn+ (6.7.19) C(tl) 8 + 14 n-y b K(x,y,y)dn, 248 THE LIFTING LINE THEORY such that (6.7.16) becomes: C(rl) K(x, y, ri)d rl - 1 7 T-b (11-y)'" } C(q) 8 K(x, b, y)d q+ J b n-y q (6.7.20) This is an integral equation (not an integro-differential one) but with a strong singularity, for which the Finite Part is considered. We denote AI (x, y, y) = 8 5; K(x, y, y) (6.7.21) and we perform the substitution y = by', v = br1'. The equation (6.2.14) becomes: '( A l C(eta 2 K (x, y, r1) d 1l - I b iJ 1 C(+1) + n-y M(x, y, y) d r1+ (6.7.22) + b2 r+1 C(i) 7r _ L(2:, y, +1)d 17 = 4bh'. (x, y) , I where K(x , y , rl) = L(x, y, n) xx.(rl) 1+ 1(x - x,)2 + b2R2yo11/2 x-x,('7)-byox: _ -F 2 1(x - x.)2 + 62Q2yo13/2 (6 . 7. 23) M(x,y,y) = 6 W (T, y, Y) Utilizing the quadrature formulas method, we shall take into account that the solution of the equation (6.7.22) has the form: C(n) = --q2 c(r1) (6.7.24) and we shall utilize the formulas (F.2.12), (F.3.4) and (F.3.5). Denoting IM =cosnk+l, k=in-, (6.7.25) one obtains from (6.2.16) the algebraic system 'AJ,ec* = 4bh2(x,g3), j = ln, A,cj + L= I (6.7.26) 249 THE CURVED LIFTING LINE where AJ = +1 2 K(x, n;, nj), Ask= (1 - rlk) 1[1-(-1) +k :i+1 (nK(k x, nj n,,)nk) - I 2+ (6.7.27) +b [1- (-i)+k1 M(x,17;, nk) + b2ajkL(x, n;, nk) J rlk - ni For writing explicitly this system we have to know the shape of the wing. For example, for the flat plate having the shape of an arrow with the angle of attack e we adopt the broken line model (fig. 6.7.3). We have f = -E and utilizing the substitution y = by, 0<y<1 -y, -1 <y< 0, Y, K(x,y,n) = L(x,y,n) = K1=1+xRn, 0<n<l K2 =1+ x+r7, R -1 < n < 0, (6.7.28) L,=-#2x-"-yo 0<n<l L2=-#2X + + y0, -1<n<0, R3 where 1/2 R= [(z_y)2+b22(y_q)2] The system (6.7.26) may be solved numerically using a computer. Chapter 7 The Application of the Boundary Integral Equations Method to the Theory of the Three-Dimensional Airfoil in Subsonic Flow 7.1 7.1.1 The First Indirect Method (Sources Distributions) The General Equations The superiority of this method in comparison with the classical meth- ods has been exposed in Chapter 4. Using this method we succeed to impose the non-linear boundary condition just on the boundary of the wing. Moreover, it allows to solve numerically the integral equation of the problem, approximating the boundary by a polygonal line in the two-dimensional case, or by a polyhedral surface (consisting of panels) in the three-dimensional case. We deal with the problem considered everywhere in this book. A subsonic stream, having the velocity U,,.i, the pressure poo and the density per. is perturbed by the presence of a fixed body, having a known surface E. One requires to determine the perturbed flow and the action of the fluid against the body. Introducing the dimensionless variables X, Y, Z related by the dimensional variables xl, yl, zl as follows (XI, y1, zt) = Lo(X, Y, Z) and putting V1=Uoc(i+V), P1=p,, +p,oU,2,.P (7.1.1) we obtain the system for the perturbed fields: R12OP18X + Div V = 0, OV/OX + Grad P = 0. (7.1.2) Projecting the last equation on the OX axis, we deduce: P=-U (7.1.3) 252 HIEM. THREE-DIMENSIONAL AIRFOIL Taking into account this result, the first equation from (7.1.2) and the projections of the second on OY and OZ, give fl29U/aX + OV/aY + 8W/0Z = 0 (7.1.4) av/ax - are/ay = o, aw/ax - av/az = 0, U, V, W representing the coordinates of the vector V. Denoting by F(X, Y, Z) = 0 the equation of the boundary E, we have to impose the condition (1+U)Nx+VNy+WNZ=O, F=0, where N_ (7.1.5) Grad F (Grad F1 (7.1.6) With the change of variables x=X, y=AY,z=QZ u= j3U,v=V, w=W. (7.1.7) the system (7.1.4) becomes au/ax +8v/ay + aw/az = 0 (7.1.8) au/ax + 8u/ay = 0, raw/ax - su/az = 0. (7.1.9) Performing also a change of variables in F we have aF/8X = OF/0x, (9F/8Y = (30F/8y 8F/8Z =130F/az , such that the boundary equation (7.1.5) becomes un=+ f32(vny+wns) = -13b?,F = 0, (7.1.10) where grad F (7.1.11) Igrad F1 We agree to utilize the inward pointing normal to the body. We also impose the damping conditions at infinity lim(u, v, w) = 0. 00 (7.1.12) The first equation from (7.1.9) represents a necessary and sufficient condition for the existence of a function io(x, y, z) such that u= v = acp/ay . 253 THE FIRST INDIRECT METHOD (SOURCES DISTRIBUTIONS) From the second equation and from the damping condition it results w=8o/8z. With this representation, the equation (7.1.8) gives App= 0. It is well known that the fundamental solution of this equation, tp(x)_-47r 1, r r=Ix - tI, (7.1.13) represents the potential of the flow determined by a source of intensity f, having the position vector . The velocity field is v = grad p = 7.1.2 4 xr (7.1.14) The Integral Equation Replacing the body with a continuous distribution of sources on E, having the unknown intensity f (x), the velocity field in fluid will be U( ) _ -4; Iff(x) Ix - 3da, (7.1.15) t representing the position vector of the generic point M in the fluid. 7.1.3 The Integral Equation In order to impose the boundary condition (7.1.10) we have to pass to the limit in (7.1.15) considering that M(4) tends to the generic point Qo(xo) E E. To the limit, the integral from (7.1.15) becomes singular. Following the procedure from the two - dimensional case (see 4.2.1), we shall prove that if f (x) satisfies Holder's condition on E, then v(xc) 4o - \ 4r jf f (x) I x - 41 3 d a1 =-2f(xo)r+o-,- 11 (X)Ix-xolda, (7.1.16) 254 BIEM. THREE-DIMENSIONAL AIRFOIL where (7.1.17) 6-+O1 JE-O J JE a representing the surface cut from E by a sphere E, having the center in Qo and the radius E. Indeed, writing IL = J (7.1.18) f,0+0 we have to calculate L= lim JJf(z)x x3 da, i.e. the last term from (7.1.18). Writing this term as follows L= lim t-.z0 J where Lo d a + f (zo)Lo J (f (z) - f (xo)l x li n f f f IX , (z)1013da, (7.1.19) we notice that the first integral from the expression of L tends to zero when a - 0 because f satisfies Holder's condition. For calculating Lo we shall replace o by A, the projection of the surface a on the plane 11 which is tangent to E in Qo (fig. 7.1.1). Fig. 7.1.1. 255 THE FIRST INDIRECT METHOD (SOURCES DISTRIBUTIONS) On this projection we shall use the parametrization x = xo + r(cos 9io + sin 9 jo), 0 < r < e, 0 < 0 < 27r, (7.1.20) io and jo being versors orthogonal to the plane n. Also, taking into account that the limit is the same on every path on which 4 --+ xo, we consider the limit on the direction of the inward normal no = n(xo) to E in Qo. We have therefore f =xo-qno, q>0. (7.1.21) Hence r 12" r(cos 9io + sin 9jo) + rlno rd rd 8 = 2zrn (r2 + 712)3/2 'i-.0 0 L0 = lim . (7.1.22) Now the formula (7.1.16) is demonstrated. Imposing the condition (7.1.10), we deduce the following integral equation {n2(xo) + li2En2(xo) + n2(xo)]}f (xo)+ +T7r A Zf ff (x) ( x-xo)no+fit[(y-yo)noo+(z-zo)noj da= 2 Ix - xo13 . (7.1.23} For the incompressible fluid it becomes Lf = f (-To) + 27r A f (x) ( Ix no xxo13 d a = 2n0, . (7.1.24) The integrals (7.1.23) and (7.1.24) are singular. 7.1.4 The Discretization of the Integral Equation We shall approximate the surface E of the body by a set of triangular panels TJ (j = 1, ..., N) and we shall approximate on every panel Ti , the function f by the value f; of the function in the center of mass G2 (x°) of the triangle. The equation (7.1.23) reduces to {(no)2 + f 2[('nV)2 1 //r + (n°)2]}f(xo)+ (x - xo)n? +,02[(y - yo)ny + (z - zo)n?} Ix-xo13 0 da-21nx. 256 BIEM. THREE-DIMENSIONAL AIRFOIL Imposing this equation to be satisfied in the centers of mass GG, i.e. putting xa = x°(i = 1, ... , N), we obtain the linear algebraic system N a,f,+EA,jfj=b;, i=1,...,N, (7.1.25) i-1 where we have no summation with respect to i. We denoted a = n2 (mg) + p2n?(x°) + #2n2 tx°), bi = 2 n (x°) A=i = X,in.(x9) + O2Yjnv(x°) + Q2Z,jn:(x1 (7.1.26) (7.1.27) 0 X`, = 2.-r, 1 M W1 13da. (7.1.28) For determining the quantities n(x°) and X,j we shall denote by mil, xt2 and x,3 the vectors of position of the vertices of the triangle T{ ; we choose the sense on the sides of the triangle such that the normal n(x9) is positively oriented towards the interior of the body. Taking into account the definition of the vector product, we obviously have n(xo) = (x12 - Wit) X (X,3 - Oil) 2S; ' (7.1.29) Si being the area of the triangle Ti expressed by means of the coordinates of the vectors x,l, xi2, x13. The integrals X;j are singular when i = j. We consider at first the case i 0 j. Denoting by xjl, xj2i xj3 the vectors of position of the vertices of the triangle T1 j, we shall consider the parametrization of the triangle x=xjl+(xj2-xj1)Ai+(xj3-xjl)A2. (7.1.30) F\xrther we shall proceed like in 5.3. Introducing the polar coordinates by means of the formulas Al = rcos0, A2 = rsin8 0 < 0 < r/2, 0 < r < p, (7.1.31) where p is defined (fig. 7.1.2) by the relation (cos 0 + sin 0)p = 1 (7.1.32) and denoting e(O) = (xj2 - xj1) cos 0 + (xj3 - 0, (7.1.33) 257 THE FIRST INDIRECT METHOD (SOURCES DISTRIBUTIONS) Fig. 7.1.2. we deduce x-x9=xj1-x°+re(O) (7.1.34) I0 -x,012=ar2+br+c, where a = lel2, b = 2(xjl - x°) e, c = jxj1 - x912 (7.1.35) 6 = b1- 4ac = -4(xj; - x9)e - (vjl <0. For the element of area of the triangle Tj one obtains (7.1.36) d a = 2Sjd Al d A2 = 2rSjd rd 8, S; being the area of the triangle. Hence, X;; = S. f"21(xii - x)I1(e) + e(e)I2(8)]de where _ rdr P Jo (ar2 + br + c)3/2 = 2bp + 4c 6(ap2 + by + c)1 2 _ (7.1.37) 4f 6 (7.1.38) P alt (O) - f (art + br + C)3/2 with the notations J - IP dr are+br+c - 2 %r-7 J1 - b11 - C/C , ( arctan Zap + b vr-7 17) 258 13IEM. THREE-DIMENSIONAL AIRFOIL K _ dr )r° J0 (art + (n + c)3/2 4ap + 2b 2b 6(ap2 + by + c)1/2 + W c Ones employ the formulas (7.1.29) and (7.1.37) for calculating the coefficients A1. 7.1.5 The Singular Integrals The integrals (7.1.28) are singular when i = j. Following the model from 5.3 we shall write: T,j = T(12) + T(23) + 7 31) (7.1.39) where T!kl) is the triangle GjPkP, and we shall consider the parametriza- tion of the triangle T,(12) putting x - x° = (x,1 - x° )J11 + (xj2 - x°)a2 . (7.1.40) Passing to polar coordinates and denoting E12=(xj1-x°)cos6+(xj2-x°)sin8, (7.1.41) we deduce x - x,, = rE12, Ix - x°I = rIE121 (7.1.42) Utilizing also (D.2.3) it results X(12) » = 0 1 2n T.02) I x X0 13 da= (7.1.43) = 1 Ir ' J JO 3S X?j = s/2 E12 (0)1np(8) UU (E121' (7.1.44) + These expressions are utilized for determining the coefficients A.,,.. X123) 7.1.6 The Velocity Field. The Validation of the Method The numerical values of the velocity field are obtained from (7.1.16) with the formula N 2v(x°) _ - f X y f3 j=1 . (7.1.45) THE FIRST INDIRECT METHOD (SOURCES DISTRIBUTIONS) 259 For testing the method we shall use the exact solution in the case of the sphere placed in an uniform incompressible stream. We know (see, for exaunple, [I.11 J, p.163) that if the sphere has the radius a and the center in the origin of the coordinate axes, and the uniform stream has the velocity Uk , then the potential of the perturbed flow is a3) (7.1.46) and\\\V1 Calculating V1 = grado = U,,(i+ v) which results from (7.1.1) under the hypothesis that the fluid is incompressible (A = 1) and using for tv the values (7.1.45), it results the comparison from figure 7.1.3 for IVI I /U,,. We notice that the approximate method and 1. V 1.0 Exact ...._. I talho d + a ao Fig. 7.1.3. the exact one give very closed results. In the paper [7.2], that we have utilized for writing this subsection, we may find the approximate results for the ellipsoid for various angles of attack and for various Mach numbers. We may also find approximate results for the wings whose cross section is a NACA - 64 - A - 008 profile. 7.1.7 The Incompressible Fluid. An Exact Solution We have seen that, for the incompressible fluid the integral equation is (7.1.24). In this subsection we shall determine the exact solution of DIEM. THREE-DIMENSIONAL AIRFOIL 260 the equation in case that the perturbing body is a sphere (7.4). INg. 7.1.4. Considering that the points on the sphere (x and Qo(xo) have the coordinates (fig. 7.1.4): X = R(sin q, cos 92, sin ql sin 42, cos ql ) 0<gl<zr, 0<qq<27r, (7.1.47) xo = R(sin qi cos 92, sin qi sin gZ, cos qi) we deduce for no (the inward pointing normal to the sphere in Qo) no = -(sin g° cos q2, sin q, sin qz, coe g°) . (7.1.48) We have therefore: f (x) - f (ql, q2), f (xo) - f (q°, q2) (7.1.49) da = R2 sin gldgidq2, no -- no = - cos qo and Ix - xo12 = 1X12 + Ixo12 -- 2x0 xo = 21122(1 - cos9), (7.1.50) cos 6 = sin q1 sin q, - cos(Q2 - q2) + cos ql cos go. (7.1.51) where Hence it results (x - xo) (no) - (x-xo) (x - xo) no Ix-x013 da= = R(1-cosB), (1 - cos 8) sin gldgl dq2 -sing1dgldq2 2(1-cos0) 2(1-cos9) 2 1-(oO) (7.1.52) THE FIRST INDIRECT METHOD (SOURCES DISTRIBUTIONS) 261 With these formulas the equation (7.1.24) becomes 2w I f(gi,gz)+47 rJo Jo f(gi,g2)sin2(igld)q2 =-2cpsq0j. (7.1.53) the sign "' indicating that one eliminates the vicinity of the point Qo, i.e. 0 = 0 according to the definition (7.1.17). Denoting 1 K(qi, qq, q1, 42) _ 4ir 2(1 - cos 9) ' (7.1.54) D = (0, 7r) x (0, 2r), we deduce that the integral operator of the equation which determines the density f (q,, 92), is a bounded operator with respect to the uniform convergence norm for real functions, continuous on the closed rectangle b. This boundedness follows from the property KsingldQ1dg2 = 1. (7.1.55) We sludl also prove that the integral operator has the invariance property J j(cosqi)Ksintdqidq2 = cosq°. (7.1.56) 3 the change of variables For proving these properties we shall perform (qj, q2) -- (0, A) defined by (7.1.51) and by the following relation, which is a direct consequence of the sine rule from the spheric trigonometry sin A sin 8 = sin q1 sin(g2 - q2), (e, A) E D . (7.1.57) Hence we choose the spherical coordinates relative to the point x for which, instead of the Oz axis we take the direction -no, and instead of the xOz plane we take the plane of the versors k and nO (fig. 7.1.5). In this way, for the element of area in the generic point Q one obtains d a = R2 sin Q1dQ1dQ2 = R2 sin 6dOdA. (7.1.58) Using the cosine theorem from the spherical trigonometry, we obtain cos q1 = cos q° cos 0 + sin q0 sin 9 cos A. Hence, r 1 ID KsingtdQldg2 = lID Ksin8d9dA (7.1.59) BIEti1. THREE-DLMMENSIONAL AIRFOIL 262 Fig. 7.1.5. (2" d A /"` Jo o =I sin 8d 9 2(1 - cos9) J f(cosqi)Ksinqidqidq2 = 1 (cos Jo n cos) d A) / (sin 8)K sin dd 9 = 0 /R 2-( J(cosqi)KsinodldA = J ID cos 6K sin Bd9d.1+ r2x +(sin q°) ( -_ 1 ' sin8cbs9 de_ 1 qo qo) 0 3 (1 - CM e) Utilizing the identity (7.1.56), we may solve the equation (7.1.53) by means of the successive approximations method. Indeed, putting f1 _-2cosg0,, we deduce f2 = - f fnf1(41,92)K(4i,42,91,92)sing1dg1d42 = _ -3(-2cosgi) fs = -If U (,)2 (2(41,42)h'sin41d41d92 = 2cosgi) 263 THE FIRST INDIRECT METHOD (SOURCES DISTRIBUTIONS) I fk+I = `J f Jofk(q ,gs)Ksingtdqidqs = (3 \k J (-2cos401) whence 00 f(gl,g2) =Ffk+1(gi,92) _ k=o 00 = (-2oosgt) (7.1.60) (1)k = -i cogt. k=O This is the exact solution for the spherical obstacle. 7.1.8 The Expression of the Potential For testing the integral equation, we must prove that the potential calculated with the density (7.1.60) coincides with the exact potential given by (7.1.46). The potential in the generic point M(F) determined by a source having the intensity f placed in Q(x) is according to (7.1.13), WW _ 4r Ix i EI ' whence it results that the potential determined by a continuous distribution, of intensity f (c), on the surface E, will have the expression w(f) = --I Jf IE IM-41 da. (x) (7.1.61) In the case of the sphere E = S(O, R), with the density (7.1.60), the potential is Arr) = 3 87r cos fJ Ix - qlEI d a . (7.1.62) Considering the point M(L) exterior to the sphere, having the coordinates f = r(sin gicosgs,sinq°singscoo q),r > R, we get Ix - Cl' = R2 - 2rReoe6 + r2 (7.1.63) BIEM. THREE-DIMENSIONAL AIRFOIL 264 where cos O is defined by (7.1.51). Taking (7.1.58) into account, it results 3 8n 3 R2 cosglsing1dqldQ2 - '?r R cos 8+ r ID f c 2 8? R- l Jj, qi sin Od Od A R - 2rR cos 0 + r- On the basis of the formula (7.1.59) and the periodicity of cos A, we obtain 3 V 3 2m R2 _ $R cosrqo1 ff d A fo 8n(coc; gi) cosOsinOd0 J r = [(R2 + r2)(R2 - 2rRcvsO + 1.2)1/2_ o -(R2 - 2rRcosO + r2)1/2 I sinOd0 = 3 16 lr- RI)- -(r+R)3- Ir-R131 and finally, since r > R, 2 T6 This is also the perturbation potential given by (7.1.46). (7.1.64) THE SECOND INDIRECT' METHOD (DOUBLET DISTRIBUTIONS) 7.2 265 The Second Indirect Method (Doublet Distributions). The Incompressible Fluid 7.2.1 The Integral Equation The theory in this subsection follow the paper [1.11]. We denote by po(x) the potential of a known flow in the entire space. We assume that this flow is perturbed by the presence of a body whose support is the simple connected and bounded domain D. We assume that the boundary of this domain, E is smooth, such that we may apply Poisson's formula. We denote by n the outward pointing normal to E. The potential po is a harmonic function, such that we may write (1.11) (D.3.2) J JE [fi(x) n !x 1 F{ 1doi 2(pe(4), IX-41 tin 4ED to (7.2.1) EE EE, E representing the exterior of D. When we write F E D we understand that the point Al whose vector of position is t belongs to D. Now we assume that the perturbation is produced by the body having the support D and let So(x) be the potential of the perturbation in E. According to the equation of continuity (7.1.8), cp will be a harmonic function which has to satisfy to infinity the damping condition line S'(F} = 0. 141-Under these conditions, V will have the representation (D.3.10) from [1.11) 2T J JE - (x x7- x' tEE 1 1 t) 0 T+p(x)1 d a= So(t), 0 the normal being the same like in (7.2.1). F= xo E LED, (7.2.2) 266 BIEM. THREE - DIMENSIONAL AIRFOIL Adding (7.2.1)3 to (7.2.2)1 we obtain W(E) = f 4a J { [,Po(x) + v(x)] 1 41 Ix (7.2.3) (s o+9)(x)}da, t E E. 1 I Determining V in order to have A A (7.2.4) W E E, co(x) = the representation (7.2.3) becomes 00 = f [PO(x) + W(x)] Ix 1f du. (7.2.5) Subtracting (7.2.1)2 from (7.2.2)2 one obtains ,P(X0) - Po(xo) = .1 g (Ix 1 f [,Po(x) + P(x)J n Introducing the function j u: x01 ) E E. d a, xo (7.2.6) E -+ R by means of the formula p(x) = fi(x) + (V)x E E, (7.2.7) the equation (7.2.6) reduces to .U(X0) _ 2zr J Jl U(x) , \ Ix 1 xol d a = JPo(xo), (V)--o E E, (7.2.8) and (7.2.5) to da. (7.2.9) WW = 4r ,/ / The equation (7.2.8) is the integral equation which determines the function µ, and the equation (7.2.9) shows that this function is just h (); (Ix 1 the doublets density E which replaces the body. We may write the equation (7.2.8) as follows lt(xo) + 2a J Js µ(x)n(x) IxX x013 d a = 2cp(xo), (d)xo E E, (7.2.10) n representing the normal pointing towards the fluid (the outward normal to E). This equation which determines the doublets density, is similar to the equation (7.2.21) which determines the sources density. Only the free terms and the normals from the kernels are different. In (7.1.24) it appears n(xo) while in (7.2.10) this is replaced by n(x). THE SECOND INDIRECT METHOD (DOUBLET DISTRIBUTIONS) 267 The Flow past the Sphere. The Exact Solution 7.2.2 Considering the problem from (7.1.7), i.e. the integral equation in the case of the uniform flow with the velocity U,,,k past the sphere S(O, R) and utilizing the formulas (7.1.47) we have: oo= -- =Rcosq° (7.2.11) such that the equation (7.2.10) becomes ir i II ( g , q0 ) ± 47r 2w j1 0 II (gl, 92, ) s1 - -la- gc V 2(l d. B ) = 2R cos q a . (7 . 2 . 12) This equation differs from (7.1.53) only by the right hand side term. The equation (7.2.12) has therefore the exact solution it = 3 R cos q1 (7.2.13) . For testing the integral equation (7.2.10), we shall prove that the potential ip, obtained from (7.2.9) with the density (7.2.13), coincides with the exact perturbation potential (7.1.64). Indeed, (7.2.9) becomes - 4-r1 I Jz u(x) ' 4 d a. ix 413 Using the notations (7.1.47) and (7.1.63) we have cp(l;) = (7.2.14) (x - la) n = R - r cos 9 x F = Rr cos 6, da = R2 sin glclgldq2, such that, utilizing (7.2.13), we deduce 1 3k 1P 4ir 2 0 2n' (R-rcos0)cosgjsingtd (R2 - 2Rr cos 9 + r2)3/2 J0 91 d q2. Passing to the variables 0 and A and taking the formulas (7.1.58), (7.1.59) and the periodicity of the function cos A into account, it results 3 'P = - 4 R 3cos ql0 fW sin O cos 0(R - r cos 9) (R2 J - 2Rr cos 0 + r2)3/2 Performing the change of variable 0 -+ u : it = (R2 - 2Rr cos 0 + r2)1/2 we deduce f x sin0cos0(R - Jo (R2 - 2Rr cos 0 + r2)3/2 d and finally, (7.1.64). 6 2 - - 3r2 d0, 268 I3IEM. THREE - DIMENSIONAL AIRFOIL 7.2.3 The Velocity Field Since the potential has the expression (7.2.14) we deduce: v(t) = grade = - 4 Jj;(x)gracI£ 47r n(x)] d a = L lx - 41 l I µ(x) (n grad) 1 Ix - X12 Ida and finally vW = 4zr ! j1`(x) [n(x) - 3(x - ,)(IX - I2nl Ix - 413 (7.2.15) This formula gives the velocity field in the fluid. We deduce that far away the kernel has the order I4I-3, as it is natural to be (see for example (6.1.16) from 11.11]). On the boundary of the body, the integrals from (7.2.15) have strong singularities. We have therefore to transform these integrals. 7.2.4 The Velocity Field on the Body. N. Marcov's Formula In the beginning we have to notice that the double layer potential of constant density equal to 1 defines a piecewise constant function whose gradient is obviously zero in the continuity points. Indeed, from (7.2.1) it results: 1 1, fr x 1 t) TT(1)da= 1/2, 0 4ED E (7.2.16) EE, and for the gradient we have 1 4n ff n(x) - 3(x -) da Ix - tp2 ] Ix - CI3 (7.2.17) in E (and D). Multiplying this identity by the constant jt(xo) and subtracting it from (7.2.15), it results J P( X) !t(xo) Ix-EI - n n-3(:c-F) (xIx-412 da Ix-EI2 (7.2.18) TILE SECOND INDIRECT METHOD (DOUBLET DISTRIBUTIONS) 269 for AI(D) in E. For calculating the velocity when AI(F) Qo(xo), a point on the boundary E, we shall denote again by or the portion cut from E by the sphere with the center in Qo and the radius s and we shall use the definition of Cauchy's principal value lim JJa- JJE (7.2.19) if it exists. With no the outward normal in Qo, we put t = xo +qno (q > 0). For the limit value of the velocity in this point, we have: =bin tje(xo) = 1-0 v(xo + qno) = It + 12, (7.2.20) for every F > 0. We denoted 11 = lira 1 if ti--o 47r -3(x I2 no4r 12(x) -11(xa) _o x - 41 da j Ix - 412 Ix _ X12 ' (7.2.21) 11 !l( Ix - 4Ixa) [n- -3(x-4) (x - t) - n da Ix-XI2, Ix-£12 where = xo + q no. The first limit and the integral may be interchanged because the integrand has no singularities on E - a. Hence one obtains 1t = /e(x) _ 14(x0) 1 47r E-o Ix - xol [n (7.2.22) (x - xo) n Ix-xo12 da ix-x012 For every e > 0 this integral exists if µ(x) satisfies Holder's condition. So, we deduce v(xo) = Jim (1) + 12) = w(xo) + link 12 , (7.2.23) 270 RIEM. THREE - DIMENSIONAL AIRFOIL where w(xo) = 1 Jf_ /L(x) - JL(xo) Ix-xol -3(x - xo) L (7.2.24) J Ix - xo12 da IX - X012 For calculating the last limit from (7.2.23), when s is small enough, we may replace a with its projection A on the tangent plane in Qo and we shall utilize (7.2.17). It results that we have x - t = r(cos Oio + sin Ojo) - i'no, Ix - t12 = r2 +.q2 , (7.2.25) µ(x) - /&(xo) = (V /)(x0) ' (x - xo) + ... _ = rvp(xo) (CDs NO +Sill 8jo, where V1z(xo) = (V/c)(xo), according to the usual notation in Analysis. Considering the scalar product Vy - (cos Bin + sin Ojo) , in the basis io,)o we obtain the identity [V -(cos Oio + sin 8jo)](cos Oio + sin 8jo) = = [(Viz io)cos8+ (V/z jo)sinO](cos8io + sinOjo) = (7.2.26) = (Vp - io)(cos2 Oio + cos8sin 6jo)+ +(Vp . jo)(sin Ocos 8io + sine` 0jo) . Noticing that some terns vanish after integrating with respect to 0, from 12 it remains 3 12 t 2-x r- 0 47 r Jo JO r3)? (r2 + i,2 )5/2 X (Qµ io) x (cost 8io + cos O sin 8jo)+ drdO, x +(QJJt jo) x (sill 8 cos 8io + sine 8jo) I2 = lim 4 J' (r2 r3'1 [(VJt - io)io + (ViA - ?o)7o)Id r . 't'ilt: l)IRECT ME IIOI). THE INCOMPRESSIBLE FLUID 271 One obtains (Vi)(xo) in the square bracket and we take it off from the integral. So, t I., = r3 17 (r2+ 2)b/2dr= Op(io) Hence. taking (7.2.23) into account, we obtain the velocity on E by ineans of the formula v(xo) = 2(V1)(xo) + w(xo), (7.2.27) w(xp) being given in (7.2.24). This formula was obtained by N.Markov in a unpublished paper. Since P(x) and Qo(zo) from (7.2.25) belong to the plane which is tangent to r in Qo, we deduce that Vp(zo) is in the tangent plane. Hence, the boundary condition (i+v).n=O pe E determined the following integral equation: 1 -1s(xo) f Fi(x) Ix - xol 4A E da =-n=, VxoEE -3(x-xc) no Ix-moll Ix-xo12 which is an alternative to (7.2.8). 7.3 7.3.1 The Direct Method. The Incompressible Fluid The Integral Representation Formula For writing this subsection we used the paper (7.3]. Since we study the same problem like in the previous subsections, we shall utilize the repre- sentation (4.6.1). the equations (4.6.2), the boundary condition (4.6.3) and the condition to infinity (4.6.4). The difference is now that the prob- lem is three - dimensional. In (4.6.3), C will be replaced by E, the surface of the perturbing body B. For avoiding the singular integrals, we shall replace the equations (4.6.2) by the equations div(v-c)=O, rot(v-c)=O, (7.3.1) 272 BIEM. THREE - DIMENSIONAL AIRFOIL c being a vectorial constant. The system (7.3.2) div v' = 6(x - xo) , rot v' = 0 for every point xo E D + E, D representing the domain occupied by the fluid (the exterior of the body B), defines the fundamental solution V* = x - xO 41rIx-x013 1 (7.3.3) From the equations (7.3.1) we deduce the identity 4, [ f div (v - c) + g rot (v - c))d v = 0 (7.3.4) for every two functions, or regular distributions, f and g. We denoted by Do the exterior of B, bounded by a sphere S(O, R), R being great enough, such that the body B is included into the interior of the sphere. Utilizing the identity (4.6.9)1 and the identity rot [g x (v-c)] (7.3.5) and applying Gauss's formula, from (7.3.4) we deduce ( v - c) (grad f - rot g)d v = (7.3.6) (v - c) (fn - (n x g)id a, E n being the outward pointing normal on Do. Substituting successively (f, g) -' (7 - v*, -9 x VI) (7.3.7) we find the projections on the axes of coordinates of the following identity: J (v - c)div v'd v = LER {n (v - c)v' + [n x (v - c)] x v' }d a , Do xoEDo+E (7.3.8) 273 THE DIRECT METHOD. THE INCOMPRESSIBLE FLUID proving in this way that it is correct. Taking (7.3.2) into account, it results V(x0) - c = f {n (v - c)v* + [n x (v - c)j x v' }d a . (7.3.9) +ER The integral on ER may be written as follows v)v+ (n x v) x vjd a - hR c)v' + (n x c) x v*] d a . IS,, T he first term vanishes when R oo, because of the condition (4.6.4). For calculating the second term we use the spherical coordinates: R, 9, cp: x = R sin O cos V y = R sin O sin p 0<9<7r. z=Rcos9, We obtain lim. JR [(n c)v' + (n x c) x v')d a = C 47C f 0 2A f A sin 9d 9d W = c. 0 Now, the representation (7.3.9) becomes v(xo) = j{n (v - c)v+ [n x (v - c)] x v}d a , xoED+E. Setting c = v(xo) = vo we obtain the representation formula V(x0) = j{n. (v - vo)v* + [n x (v - vo)) x v' }d a . (7.3.10) This is a regularized representation, valid both for xo in D and for xo on E. Obviously when xo E D, the integrand has no singularity. This property is also valid when xo E E because of the factor v - vo which vanishes for x -- xo. This representation formula is fundamental [7.1j. Utilizing the boundary condition (4.6.3), the formula (7.3.10) becomes vo=1 E (v-vo)) x v*}da. (7.3.11) 274 7.3.2 RIEM. THREE - DIMENSIONAL AIRFOIL The Integral Equation The vector F- nxv (7.3.12) which intervenes in the surface circulation C= J nxvda. (7.3.13) E introduced by Pascal and utilized again by R.von Mises 11.271, [1.20], has a great importance herein. We shall deduce the equation for P. To this aim we write the formula (7.3.11) as follows: (7.3.14) vo = After the vectorial multiplication by no one obtains Fo = J[_(nz + n - vo)(n° x v') + (n° v')F(7.3.15) -(n° F)v - (n v')Fo + (vo . v')(n(' x n)ld a. But, from the boundary condition n v = -it, and from (7.3.12) we deduce v = -nxn - n x F such that n - vo = -non ' no - (n, n°, Fo) (v' vo)(n° x n) = -n°(v` n°)(n° x n) - (n°, Fo, v*)(n° x n). (7.3.16) One obtains therefore Fo = J(._(ni - nn n°)(n° x v°) + (n, n°, Fo)(n° x +(n° v')F - (n° - F)v' - (n v')Fo-n°(v' n°)(n° x n) - (n°, Fo, v)(n° x n)Jd a . Utilizing the double vectorial product formula (UxV)xW (7.3.17) 275 THE DIRECT METHOD. THE INCOMPRESSIBLE FLUID it results (n, n°, x v°) - (n°, Fo, v*)(n° x n) _ = no x ((n°, Fo, n)v' - (n°, Fo, v')nJ = (7.3.18) = no x [(n x v') x (n° x Fo)J = -(no ,n,v')(no x Fo). Hence, the equation (7.3.10) becomes F°+ J, [(n° F)v' - (n° v')F + (n v')Fo] d a+ + no x Fo) f (n° x n) v'd a = j {n(v" n°)(n x n°)+ + (7.3.19) this representing the integral equation of the problem. This is a regularized equation. The integrals are not singular. Although the denominators of the distribution v' vanish for x -+ x°i we have to take into account that in these points the numerators vanish too. Indeed, no -.0, 7.3.3 Kutta's Condition In the sequel we shall study the flow around lifting bodies (with smooth surfaces). In order to solve this problem, the classical theory, beginning with the lifting line theory (Prandtl 1918), introduces a vortical surface behind the body. In this way the flow in no longer irrotational in the exterior of the body, hence the above solution cannot be utilized. For defining the fluid How with the aid of the solution given herein, we shall consider a discrete distribution of vortices located in certain points from the interior of the body [7.3]. In this way, the external flow remains everywhere irrotational and it will be characterized by the solution given herein. Nloreover, the circulation C defined by (7.3.13) does not vanish on the trailing edge. It remains to discuss the location of the vortices. The Kutta-Joukowskicondition concerning the continuity of the pres- sure on the trailing edge will be imposed, like in the two-dimensional BIEM. THREE - DIMENSIONAL AIRFOIL 276 case, writing that the velocity in plane cross-sections, perpendicular to the trailing edge, is continuous when P, and P, - Pf (4.6.21). Considering in figure 7.3.1 the point P, and the local frame n, a and Fig. 7.3.1. (3 = n x s and knowing that V = V,s, we deduce n x V = V,/3. (7.3.20) Hence, the vectorial form of (4.6.21) is: (n x V)(P,)+(nx V)(P,)-o 0 (7.3.21) P P; - PI. Taking into account that V has the form (4.6.1) it results the condition <F>+<n,> j- <ny>k=0, (7.3.22) <¢>=q(P,)+ b(P.). (7.3.23) with the notation 7.3.4 The Lifting Flow One knows from the theory of the two - dimensional potential in compressible flow that the lift is generated by a vortex of intensity rk placed in a point from the interior of the body. For example, in the case of the flow past the circular obstacle with the center in ;.o (see [1.11) p. 117), the potential Ua, f(z - zo)e-io + l R2 _ Z-20 1l et° I characterizes the non-lifting flow and the potential r 27ri In (z - zo) 277 THE DIRECT METHOD. THE INCOMPRESSIBLE FLUID defines the lifting flow, the lift being -pU,:1'. According to this model, we shall try to generate the lift by means of some vortices from the interior of the body, the number of vortices being determined by the number of pairs of panels adjacent to the trailing edge for which we have to write Kutta's condition (7.3.22). To this aim we need at first the velocity field w' generated by a vortex line of constant intensity r. This will be determined by the system div w' = 0, rot w' = r6(x) , (7.3.24) w' being a distribution (for this reason it was marked by *). For determining the solution we apply the Fourier transform, like in the case of the system (7.3.2). Using the formulas (A.6.5) we deduce axiv'=ir, whence, utilizing (A.6.9), w'=rx(- w' =rxVF-' 2 and then (A.7.10) and (7.3.3) w` = v' x r. (7.3.25) If the vortex is located in the interior of the body, in a point having the vector of position xi, then w(x) = v'(x - xi) x r, (7.3.26) and if there are L lines, since the equations (7.3.24) are linear, it results L w'(x) = E v'(x - xk) x rk . (7.3.27) k=i Returning to our problem and considering, for the sake of simplicity, a single line, we shall write the formula (4.6.1) as follows: V=U_ (i+w'+v). (7.3.28) From div v = 0 and rot r = 0, taking (7.3.24) into account, it results div v = 0, rot v = -ri(x - xh), (7.3.29) 278 BIEMM. 'THREE . DIMENSIONAL AIRFOIL and the equations (4.6.5) will be replaced by div(v-c) =0 rot(v - c) = -ra(x-xl). (7.3.30) The identity (7.3.4) will be written as follows: Joo [fdiv(v-c)+g.rot(v-c)]ctv= ,xl E D° ,x1 D°. 0 (7.3.31) Since the point having the vector of position xl is in the interior of the body, we shall obtain a second equality identical to (7.3.4), such that all the consequences follow like above. We have therefore (7.3.10). From the boundary condition v - n = 0 we deduce (7.3.32) such that the formula (7.3.11) becomes vo= R xv'}da (7.3 .33) xoED+E. The relation between v and F is now v=-(n,, +wn)n-nxF, (7.3.34) and the integral equation (7.3.19) becomes Fo+ JN [(n° - F)v* - (n° - v')F + (n v')Fo]da+ + (n° x Fo) x n) v'd a + wn(xo) I [(n° v')(n° x n)- -(n°.n)(n° x v`)]da+J wn(x)(n° x v')da = E = f, { n2(n° v*) (n x n°) + [n0. (n - n°) - n.,] (n° x t v')}da, (7.3.35) where, taking (7.3.27) into account, we have: wn(xo) _ I no x v'(xo - xi)] r (7.3.36) w,,(x) = [n° x v(xo - xi) ] . r. L THE DIRECT METHOD. THE INCOMPRESSIBLE FLUID 279 In (7.3.35) the dependence of F and r is linear. In the case of L vortices, the equation (7.3.35) remains unchanged and the expressions (7.3.36) become L Wn(x0) _ F [no X v'(x0 - xk)] r 11 k=1 (7.3.37) L W4(z) _ E In X v'(xo k=1 - xk)1 r. In equation (7.3.34), F, r1i rL are unknown. For specifying the conditions (7.3.21) we shall write the product n x V as a function of the principal unknown F. From (7.3.28) and (7.3.34) it results nxV=Uoo(F+nzj-nyk). (7.3.38) The condition (7.3.21) will be written as follows 0 F + (nx -ny 0 (P,) + [F+ (n, (P;) = 0, (7.3.39) -ny for every pair of panels from the trailing edge. It will result therefore L equations. These ones, together with the equation (7.3.35) discretized below, will determine the solution F, F1,... , r,. The numerical results are more accurate when the points P, and P; are closer from the corresponding point Pf from the trailing edge (see [7.31, p. 364). 7.3.5 The Discretization of the Integral Equation Like above, one approximates the surface of the body by N panels (triangles and quadrilaterals) and we approximate on every panel III the function F by the constant value F,,, that it has in the center G, (x9) (the collocation method). If we impose to the equation just obtained from (7.3.35) to be verified in all the centers G,(x°) and we denote X;;=J v'(x-x°)da (7.3.40) 280 BIEM. THREE - DIMENSIONAL AIRFOIL we deduce, for i = 1 N 1 Fi+ L.. [(ni Fj)Xji - (iii X ji)Fj + (nj Xji)Fi] + j=1 N L + (ni x Fi) J(ni x nj) Xji + j [(ni x x;k) . rk]' k=1 1 [(ni x nj)(ni Xjj) - (ni x X ji)(ni ' nj)] + E j=1 [(nj x xjk) rk] (ni x xji) = + j x1 k=1 { (7.3.41) - ni(ni. ,j=1 X ji)(ni x nj) + {n(rsi.nj)_n}(ni x Xji) }, In this system, the singular terms Xii disappear, because the integrals from (7.3.35) are regularized. Taking into account the relations (nj Fj)X3i - (ni ' Xji)Fj = ni x (Xji x Fj) (ni nj)X ji - (ni Xji)nj = nj x (Xji x N), we deduce that the system (7.3.41) may be written as follows N L LA,,F,f+EBk{lrk=Ci, i= 1,...,N j=1 (7.3.42) k=1 where AiiFi = Fi+FnEnj Xji+(ni x Fi)E(ni x nj) Xjie j#i j#i AijFj = ni x E(Xji x Fj), j#i Bki?rk = [rk(na x Xik)j E(ni Xji)(ni x nj)+ ;oi + > {r. j#i [ - (ni x Xik)(ni nj) + nj X Xjk] }(ni x Xji), 1 THE DIRECT METHOD. THE INCOMPRESSIBLE FLUID 281 Ci= nixE[n (njxXji)xn1-n Xji! i0i the unknowns being F1,... , FN, r1, .... r1. To this system of N equations, we add the equations (7.3.39) written for every pair of panels adjacent to the trailing edge. The number L of vortices equals the number of pairs; in this way, the number of (N+L) equations obtained from (7.3.42) and (7.3.39) equals the number of unknowns F1,... , FN, rl,... , ri. For calculating the coefficients Xji (7.3.40) one utilizes either the quadrature formulas (for example, the Gauss-type formulas), or the analytical formulas given by Hess and Smith in [7.10]. Denoting by x, =21, x23, (24) the vectors of position of the vertices of the panel lIl which is a triangle (or a quadrilateral), with the definitions Vill - y)2 + ( k+1 - 4)2 (xik)2+(yiyk)2+ (Z, -k)2, Tk= mk,k+1 = '.)I + (yll + (k+1 dk,k+1 = +1 - , ek = (xi-2'k)2+(zi-k)2 , hk = (xi- )2(Yi-yk)2 , k the formulas of Hess and Smith are Xji = (Xji,Yji,Zji), +1 - Ylk Xji = dk,k+l k=1 Yji = - 3(4) -4+1 4s k=1 4 = - - ` k In rk + rk+i - dk,k+1 dk'k+l rk + rk+1 + dk,k+1 3(4) 1 41r rk + rk+1 + dk,k+1 arctan k=2 'mk,k+lek - hk (zi -arctan mk,k+1 + ek+1 '- hk+1 (zi - k)rk+1 k)rk In fact, Hess and Smith deduced the formulas assuming that the panels IIj are situated in planes parallel to xOy (4 = 0). An example is presented in [7.3]. Chapter 8 The Supersonic Steady Flow 8.1 8.1.1 The Thin Airfoil of Infinite Span The Analytical Solution In this subsection we study the same problem like in 3.1. The only difference is that the unperturbed flow is assumed to be supersonic (M > 1). Hence the uniform flow, having the velocity Ui, the pressure p,. and the density p,, is perturbed by the presence of an infinite cylindrical body. The xOy reference plane represents a cross section and the leading edge is the origin of the axes of coordinates. The length of the projection of the airfoil on the direction of the unperturbed stream (which does not always coincide to the length of the chord) will be taken as reference length (fig 8.1.1). Utilizing the coordinates defined Fig. 8.1.1. by (2.1.1), we shall denote the equations of the profile determined by the xOy plane y = hf(x), (8.1.1) 284 THE SUPERSONIC STEADY FLOW the functions h..(x) being defined on the interval (0, 1J and and possessing first order derivatives. Taking into account the formulas (2.1.3), we deduce that the perturbation (which is obviously steady), will be defined by the system (2.1.32) and by the boundary conditions (2.1.33). From this system, taking into account that the perturbation is plane, it results: u = -p, v= + py = 0, M2pz + ut + vy = 0 , (8.1.2) whence one obtains k2vtt - vyy = 0, (8.1.3) with the usual notation k = M - 1. The boundary condition (2.1.32) becomes v(x,±0)=ht(x), 0<x<1, (8.1.4) and the damping conditions at infinity lim (p, u, v) = 0. (8.1.5) The equation (8.1.3) is hyperbolic and the families of characteristics C+ and C_ are defined by the equations x-k-y=c+, x+ky=c_. (8.1.6) Obviously, the first family consists of parallel half-lines which make the angle µ with the Ox axis, the second consists of parallel half-lines which make the angle -µ with the Ox axis, where tan p =1 (sins = , oos p = (8.1.7) For determining the characteristics passing by a certain point, we write that the coordinates of the point verify the equations (8.1.15) and and then we determine the constants c+ and c_. For example, the characteristics detaching from the leading edge have the equations x - Icy = 0 and x+ky = 0. We assume that the profile is in the interior of the angle having the opening 2µ which is determined by these characteristics. It is well known that the equation (8.1.3) has the general solution v(x, y) = F(x - ky) + G(x + ky), (8.1.8) such that from (8.1.2) and (8.1.5) it results kp(x,y) = F - G (8.1.9) ku(x, y) _ -F + G. 'I'III: TIIIN AIRFOIL OF INFINITE SPAN 285 'I'll(- function F is defined on the characteristics from the superior half-plane, and the function G on the characteristics from the inferior half-plane. We have in view the characteristics detaching from the segment [0, 1) which replaces the profile that represents the source of perturbations. The solution from the strip I will be determined by F(x - ky) which will be denoted by F+(x - ky) and the solution from the strip II will be determined by C(x + ky) which will be denoted by F_. (x + ky). Hence the solution has the shape vf(x,y) = FF(xT ky), kpf(x, y) = ±F±(x T ky), (8.1.10) ku±(x,y) = T Ff(xT ky), the upper sign corresponding to the solution from the y > 0 half-plane and the lower sign to the solution from the y < 0 half-plane. In fact one may demonstrate that if upstream of AOA' the flow is not perturbed, then the solution of the equation (8.1.3) has necessarily the forum (8.1.10). Indeed, if the assumption is true, it means that OA and OA' are lines of discontinuity where we have to impose the jump relations (1.3.8), (1.3.9) and (1.3.10) Op1V1nO = 0, Opl'IVln +P1nO = 0, (8.1.11) with the notations (2.1.1) and (2.1.3). Taking into account the order of magnit.ucle of the perturbations, by linearization it results =0, Ovn,, +pnO =0. (8.1.12) On the line OA we have n = (sin p, - cos µ). Taking into account that in front of OA and OA' the perturbation vanishes, from (8.1.12) we obtain: 0=using-vcosa+psina (8.1.13) U=vsina - pcosa, the relationships being imposed on the line x = ky. Replacing here u, v and p given by (8.1.8) and (8.1.9) and taking (8.1.7) into account, from the second condition we find C = 0. Since in the linear steady theory we have p = M2p, it follows that the first condition is identically verified. In the sauce way one demonstrates that in the zone II v = G(x + ky). Hence the solution of the equation (8.1.3) has the form (8.1.10). 286 THE SUPERSONIC STEADY FLOW Imposing the boundary conditions (8.1.4), it results F±(x) = 14(x). 0 < x < 1. (8.1.14) These relations determine the functions F±(x) on the segment 10, 11. In the exterior of this segment we take F±(x) = 0 because v and p are continuous functions on the Ox axis in the exterior of the segment [0,11 (it results from (8.1.12), setting n = (0, 1)), such that F+ = F_, F+ = -F_. Hence the perturbation zones are I and II. For c E [0, 1], on the half-line having the equation x - ky = c we deduce: v(x, y) = h+(c), and on the half-line x + ky = c kp(r, y) = h+(c), (8.1.15) (8.1.16) u(x, y) = h' (c), kp(x, y) = h' (c) . Along a characteristic, the velocity and the pressure take the values that they have in the point where the characteristic intersects the segment [0,11. In the sequel we shall prove that all these results may be obtained directly with the fundamental solutions method, as we could expect, taking into account that this is a global method. 8.1.2 The Fundamental Solutions Method We know that. the perturbation produced in the uniform stream hav- ing the velocity Ui, the pressure p,., and the density po, by a force density (fl, f2) uniformly distributed along an axis which is parallel to Oz and intersects the xOy plane in the point of coordinates (F, 0), is given by the formulas (2.3.30) and (2.3.31). Replacing the profile from figure 8.1.1 with such a density applied on the segment [0, 1) from the Ox axis (i.e. on the strip whose cross sect ion is the segment [0, 1] from the Ox axis), the perturbation will be given by the formulas i 2kp(x, y) = in [ - fl(E) + kf ()sign y] b(xo - klyl)d (8.1.17) 2v(x, y) = J0 [ - fl ()sign y + k f 6(xo - klyl)d Taking into account the property of the distribution f f (t)6(x - )d = f (x), 0, if: ifx E (a, b) ifx E C(a, b), (8 . 1 . 18) 287 THE THIN AIRFOIL OF INFINITE SPAN we deduce that if x - klyl = c E (0,1), then we have: 2kp(x, y) = -ft (c) + k f (c)sign y (8.1.19) 2v(x, y) = - f, (c)sign y + k f (c) . Conversely, if c E C(0,1), then p = 0, v = 0. The set of points for which x - kly) = c determines the characteristic half-lines x f ky = c (8.1.20) which detach from the point x = c belonging to the segment (0,1). On these half-lines, the pressure and the velocity have constant values, equal to their values in c. 8.1.3 The Aerodynamic Action From the representation (8.1.10) we deduce that the jump of the perturbation pressure on the profile p(x, -0) - p(x, +0) is given by the formula kOP1 = -[F+(x) + F-(x)] = -[h+(x) + h'-(x)] (8.1.21) which results from (8.1.14). In this way, the lift and moment coefficients CL = L CM = (1/2)po,U2Lo' M (l/2)poU2Lo' (8.1.22) are given by the formulas CL = - 2 1'[W+ (x) + h'_ (x)Jd x CA! = - 2 k (8.1.23) rx[h+(x) t + h'(x)]d x the moment M being calculated with respect to the origin of the axes of coordinates. We notice from the formulas (8.1.10) and (8.1.23) that the linear theory cannot be applied in the vicinity of the value M = 1(k = 0). 288 THE SUPERSONIC STEADY FLOW As a first example, we shall consider again the case of the flat plate having the angle of attack e with E < y (fig. 8.1.2). Since the equation of this profile is y = -Ex, we deduce h.4. = h' = -E whence v = -E, ku = ±e, kp = T- e, (8.1.24) the total velocity and pressure being A Pig. 8.1.2. vl U [CI± ) i - Ej] P1 = poc , _T P00U2 k . (8.1.25) For CL and cM we obtain cL = 4e , T 2s CM=T. (8.1.26) These formulas have been obtained for the first time by Ackeret in 1925. The torsor of the aerodynamic forces reduces to a lifting force L, applied in the middle of the chord of the profile, force which tends to rotate the airfoil in the trigonometric sense. The graphic representation of the function cL(M) (fig. 8.1.3) shows that the linear theory is not valid in the vicinity of the value M = 1, because the lift cannot be infinitely great. One estimates that the validity of this theory begins from approximately M = 1, 2 and finishes to approximately M = 2, 5, because the lift cannot be extremely small. For the zone M = 1(0,8 < M < 1,2) one elaborates the theory of transonic flow (see Chapter IX), and for M > 2, 5, the theory of hypersonic flow. 289 THE THIN AIRFOIL OF INFINITE SPAN CO O Fig. 8.1.3. 8.1.4 The Graphical Method This method relies on the following fundamental theorem: in every point the perturbation velocity is perpendicular to Mach's line (the characteristic line) which is passing through that point. The demonstration results at once if we take into account that the characteristic lines have the versors i of coordinates (cos p, ± sin µ) and we utilize the representation (8.1.10) and the formulas (8.1.7). Indeed, we have v-if=uoosµfvsinµ= kF* ±Ftj=0. (8.1.27) Let us show for example how we can utilize this theorem for obtaining the graphical representation of the perturbation in the case of the flat plate. We consider at first the region I from the 8.1.2. In an arbitrary point M from OA we draw the vector MP = Ui and a vector V1 parallel to the plate and having an arbitrary magnitude; the perpendicular on OA which passes through the vertex P of the vector Ui determines the magnitude of the total velocity V 1 = MQ; the perturbation velocity Uv is given by the vector PQ. For determining the perturbation velocity from the region II, in a certain point M' from OA' ones draw the vectors M'P' = Ui and V'1, the last being parallel to the plate and having the magnitude unprecised yet; drawing from P (the vertex of the vector Ui) the perpendicular line on OA' we determine the vector V' = M'Q' and then the perturbation Ua,v=P'Q'. Beyond FB and FB' the flow becomes again parallel to Ox. One knows the non-linear solution of the supersonic flow past a convex dihedron(the Prandtl-Meyer fan). See for example [1.21], [1.34). In the framework of the non-linear theory, the velocity which is parallel 290 THE SUPERSONIC STEADY FLOW to Ox before OA , becomes at last parallel to the plate, crossing the Prandtl-Meyer fan. In the framework of the linear theory this fan is reduced to the half-line OA. 8.1.5 The Theory of Polygonal Profiles In the case of polygonal profiles, the solution may be easily determined starting from the solution for the flat plate. The solution, which obviously is piecewise constant, may be analytical, graphical or mixed. Ui Fig. 8.1.4. We shall consider at first the profile from figure 8.1.4 (E < p) with and e1CO8E1 + e2COSE2 = 1. According to the linear theory, the last relation will be replaced by el + t2 = 1. From the formulas (8.1.14) it results OP = el, PQ = e2 F+=-El, 0<T<t cosEl =el, F+ = -E2, f1COSE1^'e1 <x<1, (8.1.28) FI=E, 0<x<1. We deduce, k Up0 = E+E1, C + E2, (8.1.29) 291 THE THIN AIRFOIL OF INFINITE SPAN whence CL = I(e+E1e1 +E2e2), CM = a kE2(1- e,). (8.1.30) Further we shall determine the pressure in different zones starting from the solution for the flat plate. Taking (2.1.3) and (8.1.24) into account, we deduce in zone I that P1 = Poo - POQU2 k . (8.1.31) In zone II we take into account that the deviation with respect to the direction of the uniform stream from zone I is 62 - E1. Using again the formula (8.1.24) it results P2=PI-PI V (8.1.32) Using the formulas (2.1.3), we deduce P1V2 = p ,,,,U2 +... Taking into account that the factor (E2 - E1) has the order of magnitude of the perturbations, in the framework of the linear theory we have to retain E E E P2=P1-Po,U2 2k 1 =Poo_PooUT In zone III we have: p3 = poo + po0U2 . (8.1.33) (8.1.34) For the jump of the pressure we get the formulas (8.1.29). In fact the formulas (8.1.33) and (8.1.34) coincide with the formulas given by the analytic method. We shall calculate now, in a different way, the resultant of the pressure on the wing. The perpendicular to the plate OP oriented towards the body has the director cosines nl = coos (El + 2) , - sin (El + 2 )) _ (-E1i -1). (8.1.35) In the linear approximation, the perpendicular to the plate PQ has the director cosines (-E2, -1), and the perpendicular to OQ pointing towards the body (e, 1). Hence the resultant of the pressure is R = (pit, n 1 + p2e2n2 + p3n3)Lo (8.1.36) We took into account that the real lengths of the plates are I, LO, 12L0, L0. For the lift we obtain the formula 2 L = Ry - paoUk'0(e+f1e1 +E2e2). (8.1.37) 292 THE SUPERSONIC STEADY FLOW In the framework of the linear theory the drag (8.1.38) R. = (E - 1191 - e2E2)LoPoo vanishes because, from the projection of the contour OPQ on the Oy axis we deduce E = 11E1 + 12E2 In fact this is a known result. The drag has the order of magnitude two CD = 2 (92 + E1t + 6212) k (8.1.39) In the sequel we shall consider the polygonal profile from figure 8.1.5, symmetric with respect to the OR axis. We assume that all the E-s are small. Obviously, the profile is considered to be in the interior of Mach's angle with the vertex in the origin. We deduce like above that in zone I the pressure is given by the formula P1 = Poo + p00U2 (8.1.40) E, in zone II by the formula 86 Fig. 8.1.5. k E2 p - p00U2 , (8.1.41) P3=Pa-PzV2E3 kE2 =Poo-PooU2k , (8.1.42) P2 = P1 - P1VzE1 k in zone III by 293 THE THIN AIRFOIL OF INFINITE SPAN in zone I', P1 =Poo+p00U2!1 , Ei = 2E+e1, ( 8.1.43) in zone IF, -P'1V1' k Ps=Pi =Poo+p,,U2k2,E'2=2E-E2 (8.1.44) 2kes"'p.-.U2k3,E's=ES-2E (8.1.45) and in zone III', P'P2P',Vi2C1 We denote by n 1, n2, n3, n1, n; and n3 the perpendiculars to OP, respectively PQ, QR, OP, P'Q' and Q'R pointing towards the body. In the framework of the linear theory they will have the coordinates (E1,-I),(-E2r-1),(-E3,-1),(es, 1),(E'2,1) and (-E'3i1). The resultant of the pressure is R = Lo(£1pini + t2P2n2 + 13p3n3 + tlp/ini + t21/2ns + tsp3n3) . (8.1.46) The lift is p _ Poo 2 Lo [6 (E' - E1) + 12(C'2 + E2) + t3(E3 - 6'3)) . (8.1.47) Taking into account El, c'2, r'3, and the relation t1 + t2 + t3 = 1 , we deduce CL = (8.1.48) In the framework of the linear theory, the drag Rt = 2p ,Lo(E + t1E1 - t2E2 - t3e3) (8.1.49) vanishes, because from the projection of the polygonal contour ORQ'P' on the Oy axis we deduce E = titJl + t2bJ2 - tstJ3 (8.1.50) whence, taking into account the expressions E'1, E12, E13 we get E + ties - t2E2 - 1363 = 0. The formula (8.1.39) is very important. It shows that cL has the same expression like in the case of the wing consisting only of its skeleton OR. The result is natural, because of the symmetry of the wing with respect to OR. 294 THE SUPERSONIC STEADY FLOW 8.1.6 Validity Conditions We shall investigate, for the flat plate, the limits of validity of the linear theory. Since the flow is defined by the formulas (8.1.25) in I and II (fig. 8.1.1) and by Vt = Ui, pt = p in the regions upstream of AOA' and downstream of BQB', we find that pi < pil and pi < p,,. Hence, in order to have a positive pressure all over the fluid (this is a physical condition), we must have (8.1.51) Pi > 0, this representing a first condition of validity. Taking into account the expression of pf given in (8.1.25), we deduce, with the notation kX = = M2e, 0<ryX<1. (8.1.52) A second condition must ensure us that the flow is everywhere supersonic. Since from (8.1.25) or from Bernoulli's integral it results IVtlf'2 < (Vt1 12, we deduce that we must have ( 1)2 < (V ')2, whence '7Pf'/pfl = (cl'2 < U2 [(i - k)2 + x,2] . (8.1.53) Taking into account that in the framework of the steady linear theory we have p = M2p (it results from (2.1.7)), we deduce the inequality 11+X equivalent to k)2 <M2 (1 +s2, 1 +yX < (1+X)(M2 - 2X + X2). (8.1.54) (8.1.55) In the linear approximation with respect to e, this inequality becomes 1-M2<(M2-2--y)X. (8.1.56) Further we shall deal with the inequalities (8.1.52) and (8.1.56) when the fluid is the air ('y = 1, 405). For M2 < 3,405, the two inequalities are satisfied in the shaded region from the (M2, X) plane (fig. 8.1.6), where M2 - 1 2 f (Af 2) = 3,405 - M (8.1.57) 295 GROUND AND TUNNEL EFFECTS x+ 0.711 ............. ......... ........... ............ 0 Fig. 8.1.6. This result shows that the linear theory is valid only in the zone 1 < M2 < 2. For a certain value given to M2 in this zone, E is subjected to the restriction (M2 e< - 1)3/2 M2(3,405 - M2) (8.1.58) The last ratio increases when M2 increases, hence the superior limit of the angle of attack is increasing with M2. Other results concerning this subject are given in [8.17] [8.32]. 8.2 8.2.1 Ground and Tunnel Effects The General Solution We assume now that the wing lies between two parallel planes having the equations a y=-2 and y=Z (8.2.1) In this case, the general solution (8.1.8) and (8.1.9) ku(x, y) _ -F(x - ky) + G(x + ky) v(x, y) = F(x - ky) + G(x + ky) (8.2.2) P(x, y) = -u(x, Y), with the damping condition lim F(x - ky) = lim s--00 x--ooG(x + ky) = 0, (8.2.3) 296 THE SUPERSONIC STEADY FLOW has to satisfy besides the slipping conditions on the profile, v(x, ±0) = h4 (0), x E [0,1] , (8.2.4) the slipping conditions on the planes ,-a) = 0, v v (x, 2 1 = 0, x E R. (8.2.5) We assumed that the equations of the profile are y = ht(x) (8.2.6) In the sequel we have to calculate the velocity field in the domain D-={(x,y):x+kyE[0,1),0>y>-2}. (8.2.7) We consider the natural number n such that nka < 1 < (n+ 1)ka and the segments (0, ka), (ka, 2ka), ..., ((n - 1)ka, nka), (nka, 1) on the chord of the profile which is the segment [0,1] from the Ox axis. We consider the characteristic having the equation - kq = C E (0, ka). 2). For n=-2 wehave The characteristic having the equation x + ky = E (- ka ka 2, 2 ka 2, t E does not intersect the chord of the profile, i.e. the segment y = 0, x E (0, 1]. We have, by virtue of condition (8.2.3) G ka) = G(x + ky) = 0, t E lim ka, ka (8.2.8) From (8.2.5) it results 0=va) 2 and from (8.2.8) and (8.2.9) F(f+ Z)=0, 2 ) t E(-2, 2j. (8.2.9) (8.2.10) Utilizing the notations G(x+ky) = G- (x + ky), F(x - ky) = F"(x-ky),(x,y) E D" (8.2.11) GROUND AND TUNNEL EFFECTS 297 we have on the upper surface of the profile F- (x) = 0, x E (0, ka) . (8.2.12) From the previous equation and from the slipping condition on the profile (8.2.4) h'_ (z) = v(x, -0) = F- (x) + G` (x), x E (0,11 (8.2.13) it follows that G-(x) = h'_(x) x E (0, ka) . (8.2.14) Further one demonstrates by induction that for x E (jka, (j+1)ka) , j E N, j < n, we have: G-(x) = h'_(x) +h'_(x-ka)+h'_(x-2ka)+... ... + h'_ (x - jka), F- (x) _ -h'-(x - ka) - h' (x - 2ka) - ... - h'_ (x - jka) . (8.2.15) (8.2.16) Indeed, for j = 0 the relations are true. Assuming that the relations are true for j, we shall demonstrate that they are also true for j + 1. Let us consider x E ((j + 1)ka, (j + 2)ka). From (8.2.5) we obtain: 0=v(x-k2, -G'(x-ka)+F-(x). Since x - ka E (lka, (l + 1)ka), by the induction hypothesis we have F-(z) =-G- (x-ka)=-h! (x-ka)-h! (x-2ka)-... (8.2.17) ...-h'(x-(j+1)ka). From the slipping condition on the profile h'_ (x) = v(x, -0) = p-(x) + C-(z) and from (8.2.14) it follows: G-(x) = h' (x) +h'_(x-ka)+h'_(x-2ka)+...+h'_(x- (j+1)ka), (8.2.18) whence we deduce by induction the validity of the relations (8.2.15) and (8.2.16). Taking into account that P(x, y) = -u(x, y) = k [F(x - ky) - G(x + ky)) , (8.2.19) 298 THE SUPERSONIC STEADY FLOW we may calculate the velocity field on the lower surface of the profile [h'_ (x) + 2h' (x - ka) + ... + 211- (x - jka)] p(x, -0) , (8.2.20) x E (jka.(j+1)ka)n[0,1]. Considering the domain D= (x, y) : x - ky E [0,1), 05Y:5 Z (8.2.21) and utilizing the notations G(x + ky) = G+(x + ky), F(x - ky) = F+(x - ky), (x, y) E D+, (8.2.22) we can prove, like in the previous case, that for x e (jkb, (j + 1))kb) fl [0, 1) we have G+(x) = -h+(x - kb) - h+(x - 2kb) - ... - h+(x - jkb), (8.2.23) F+(x) = h+(x)+h+(x-kb)+h+(x-2kb)+...+h+(x- jkb), (8.2.24) whence it results P(x, +0) = k [h+(x) + 24' (x - kb) + ... + 2h+(x - jkb)] (8.2.25) x E (jkb,(j+1)kb)n[0,1). 8.2.2 The Aerodynamic Coefficients As we know from section 8.1, the aerodynamic coefficients are given by the formulas cL = j rt [p(x, -0) - p(x, +0)) d x (8.2.26) cA( = J x [p(x, -0) - p(x, +0)1 d x. ..110 In the sequel we shall consider some examples. For the profile in a free stream (a = oo, b = oo), it results the already known formulas (8.1.23). 299 GROUND AND TUNNEL EFFECTS For the thin profile in ground effects (b = oo), taking for example 1 > ka > 1/2, we deduce 1 CL = j[14(x) + h'_ (x)] d x- k jh..dx (8.2.27) 1 2 1 cM= -- x[h+(x)+h'_(x)]dxo k f 1-ka (x+ka)h'dx. o For the thin profile in a wind tunnel, taking for example 12: ka > 1>_kb>1/2,we get CL = - I J 1 [h+ (z) + h' (x)] d x - k -2 JI h+(x)d x, k 0 1 CM = - k1 f 1-ka h'_ (x)d x- 1 2 j1-ka x (h'+ (x) + h'_ (x)] d x - 2 (x + ka)h'_ (x)d x- -2 11-kb (x + kb)h+(x)d x. k (8.2.28) For the flat plate with the angle of attack e, we have h+(x) _ = h_ (x) = -Ex, x E (0,1], whence 2E E L= k, CM=k7 (8.2.29) for the profile in a free stream, (2 - ka), cm = k (2 - k2a2) , CL = (8.2.30) k for the profile in ground effects (12: k > 1/2) and cL = (3 - ka - kb), cu= k (3 - k2a2 - k2b2) , for the profile in tunnel effects (1 > ka > 1/2, 12: kg > 1/2). This subsection was written following the paper [8.8]. (8.2.31) 300 8.3 8.3.1 THE SUPERSONIC STEADY FLAW The Three-Dimensional Wing Subsonic and Supersonic Edges In this subsection we present the general theory of the thin wing in a supersonic stream. We shall utilize the coordinates (2.1.1) and the fields (2.1.3). The free flow is by hypothesis supersonic. Like in the subsonic case, we shall denote by z = h(x, y) ± hi (x, y) (8.3.1) = er(x, y) ± hi (x, y)} the equations of the upper and lower surfaces of the wing. The projection of the wing on the xOy plane will be the domain D, assumed to be simple connected. On the boundary r of this domain we have: hi (x, y) = 0. We assume that r is smooth. Then there exist a point F where the tangent to r makes with the direction of the stream at infinity the angle of Mach p defined by the formulas (8.1.7) and a point A, where the tangent to r makes with the direction of the stream at infinity the angle -µ (fig. 8.3.1). The point of intersection of these tangents will be considered the origin of the frame of reference. There also exist two Fig. 8.3.1. Fig. 8.3.2. points B and E where the tangents are parallel to the direction of the stream at infinity. As we know from the subsonic case, the points B and E separate the boundary r in two portions: the leading edge 301 THE THREE-DIMENSIONAL WING EFAB and the trailing edge BCDE (C and D are the points where the tangents make the angles p respectively -p with the direction of the free stream). Obviously, we assume here again that every parallel to the direction of the stream at infinity intersects the edge r in at most two points at a finite distance. Definition. We name supersonic (subsonic) part of the leading or trailing edge, the part for which the absolute value of the component normal to the edge of the velocity of the free stream is greater (smaller) then the sound velocity. We shall prove that this definition is equivalent to the following one: If in a certain point of the leading or trailing edge the angle of the tangent to the edge with the direction of the unperturbed stream is greater (respectively smaller) than Mach's angle, then in that point the edge is supersonic (respectively subsonic). Indeed, from figure 8.3.2 it results that the component normal to the edge of the velocity of the free stream in the generic point P has the magnitude U sin p1. If this is greater than the velocity of the sound in the unperturbed flow we have U sin p > c, whence sin pl > 1 - = sin p Utilizing the second definition, it results that the edge FA1 A from figure 8.3.1 is a supersonic leading edge, the edges AB and FE are subsonic leading edges, the edges BC and DE are subsonic trailing edges, and the edge CB'E'D is a supersonic trailing edge. It is known from the theory of hyperbolic partial differential equations (see also the plane problem from 8.1 and 8.2) that the zones of influence are the zones delimited by the characteristic lines. For example, in figure 8.3.1, the zone of influence of the subsonic leading edge FE is FF'E'E, FP and EE' being parallel to OA. Definition. We name wing with independent subsonic leading or trailing edges, a wing for which the zones of influence of these edges are disjoint. It results therefore that a wing has independent subsonic leading edges if the Mach lines AN and FP do not intersect in the domain D and independent subsonic trailing edges if BY and EE' do not intersect in D. For example the wing from figure 8.3.1 has dependent subsonic leading edges and independent subsonic trailing edges and the wing from figure 8.3.3 has only independent subsonic edges. THE SUPERSONIC STEADY FLOW 302 Fig. 8.3.3. 8.3.2 The Representation of the General Solution Like in the subsonic case, we shall replace the wing with a continuous distribution of forces having the form f = (fl, 0, f) defined on D. We shall see that we may determine such a structure of f, in order to satisfy the boundary conditions. The perturbation of the pressure determined in the uniform stream by the distribution f will be, according to the formula (2.3.32), P(x,y,z) = -T" fi)a-+f(Tl)8 'ID J G(ro,yo,z)ddq, (8.3 2) where G(zo, yo, z) = H(zo - s) ' z0-9 s = k ya + za., zo=z-t, yo = Y-17. (8.3.3) For the velocity field, from (2.3.12) and (2.3.34), it results v(z, y, z) = 6(z) lID f (rl)II (zo)6(yo)ddi+ V , (8.3.4) where ( X, y, z) = 2Tr ff fi n) - f (t+ o G(zo, yo, z)dc dTl z2 J (8.3.5) Obviously, the perturbation is potential excepting the trace of the domain D in the uniform stream, where the first term from the expression of v does not vanish. 303 THE THREE-DIMENSIONAL WING For the component w from (2.3.12) and (2.3.34) it results: fJ f (, n)H(xo)6(yo)de dt7+ w(x, y, z) = 6(z) + 2a 8z JJD [h (t, n) - f rl) 0 + y G(xo, I, z)dtdo z2 J (8.3.6) For w we also have the representation: w(x, y, z) = -'ID'n) 2 8 G(xo, y, z)d (8.3.7) +2x 11 f(t,n)N(xo,yo,z)dt dtl, D where N(xo, yo, z) = k2 G(xo, yo, z) + 8 xo+G (xo, Y0, [-yol-+-Z2 z)(8.3.8) J which results from (2.3.37) and the representation w(x, y, z) = ' AD ft (t, n) 8z G(xo, yo, z)dF dn+ 1(2_02 _ (k 2n 8x2 2 2) AD f (C, r( ) H -7.1 [1-0*077 d11 d drl , (8.3.9) which results from (2.3.38). Each of these representations determine an integral equation for the function f (x, y). All the known representations (Evvard (8.9], Ward (8.34), Krasilscicova (8.20], Heaslet and Lomax (8.15], Homentcovschi (8.16] and Dragoq (8.7)) are found in the formulas (8.3.6)-(8.3.9). Prolonging the functions fl and f with 0 in R2\\, the above representations may be written as convolutions. For example, p(x, y, z) and rp(x, y, z) have the following form: p=-2A 8 fl*G+BZf*G if h #G-f * y2+z2G) 7r where the sign x, y. * (8.3.10) , indicates the convolution relative to the variables 304 8.3.3 THE SUPERSONIC STEADY FLOW The Influence Zones. The Domain Di First of all we must notice that the perturbation my be represented by integrals whose integrand contains the factor H(xo - 8). Indeed, for (8.3.5) this is obvious. Taking into account the formulas (2.3.35) and (2.3.36), it results that the assertion is also valid for (8.3.2) and (8.3.7). In (8.3.9) we have f 7 f zp H(r - s) d-r = H(xo - s) J 2 T -s .ll: rdT-s (8.3.11) Since the above integrands contain the factor H(xo - s) we deduce that for a point M(x, y, z) from the domain occupied by the fluid, the integrals on D are in fact calculated only on the domain Dl where we have: xo > s (8.3.12) This inequality implies 4 < x (x - 4)2 > k2[(y - 11)2 + z2]. (8.3.13) The points from D verifying these inequalities are situated between the leading edge and the hyperbola C which has the equation (x - )2 = k2[(y - q)2 + z2) (8.3.14) and the branches to -oc because < x (fig. 8.3.4). The hyperbola C (the variables are and t) has the axis parallel to Ox. In fact, C represents the intersection of the cone having the equation (x - C)2 = k2 [(y - t1)2 + (Z - ()2) with the plane C = 0. This is Mach's cone. It has the vertex in Al and the axis parallel to Or. From the mechanical point of view, this result represents a consequence of a fact known from the hyperbolic partial differential equations theory [1.6] namely the fact that in Al one can receive only the perturbations produced in the points belonging to the interior of Mach's cone with the vertex in Al. When Al will be on the wing (z = 0), the hyperbola C will be reduced to the half-lines x-==±k(y-,). These are the characteristics issuing from M (fig. 8.3.5). (8.3.15) THE THREE-DIMENSIONAI. WING FIg. 8.3.4. 305 Fig. U.S. We may easily explain why the points from D - D1 do not affect the perturbation in M if we have in view (the significance of the fundamental solution) that the perturbation produced in a point Q E D propagates only in the interior of Mach's cone with the vertex in Q. The point 141 is in the interior of all the cones with the vertices in Dl and in the exterior of all the cones with the vertices in D - D1. It also results that in the fluid exterior to the envelope of the posterior cones with the vertices on D, the perturbation is zero. Hence we can give up the factor H(xo-s) in the integrals expressing the perturbation if we replace the domain D by D1. Prolonging the functions f, and f in the exterior of D with the value zero, for a given i; , ij will vary between Y_ and Y+ defined by (8.3.14) through kYf = ky ± xo - kszs . (8.3.16) The vertex of the hyperbola C has the coordinates tI = y, = x - kjzj (obtained for Y+ = Y_). 8.3.4 The Boundary Values of the Pressure For the integrals having the form: I (x, y, z) = JJ rl) 8zG(xo, yo, z)d>; dry (8.3.17) 306 THE SUPERSONIC STEADY FLOW we have: Jz-kJ=I 1 _ 8z d f Y+ fy_ r1) ro-s (8.3 dt) . Performing the change of variables rq -+ 8: kn = ky - Jxo - k2 z2 cos 0 (8.3.19) we obtain jx_&.izi I (x, y, z) = k. z d/f rR ! f y - rx.2 - k2z2 cos 9) d8 = C z -L-1-TI _ -sign z !O f (x - kjzj, y)d©+ JJ dC oo Of Jo 8rr noose xo - k z dO. Hence, I (x, y, f0) = F it f (x, y) . (8.3.20) Using this formula, from (8.3.2) we obtain Ax, y, ±) = - 2I j jD f1 (e, n) G(xo, No, 0)d do f 2 f (z, y) whence, (8.3.21) f (x, y) = F(x, y, +0) - P(x, y, -0) . This result puts into evidence the significance of the function f. 8.3.5 The First Form of the Integral Equation The simplest way to obtain the lifting surface equation relies on the representation (8.3.7). Taking into account the derivation formula (2.3.35), we may write the kernel (8.3.8) as follows: N(xo, yo, z) = -k2 H(xo - s) _ xo(yo - p2) H(xo - s) xo - s yo + z2 (xo - s2)312 (yo x022 (8.3.22) We notice that for z = ±0 it appears the singular line n = y. Detaching from D the domain DE defined by y - e < n < y + E in Dl - D, (fig. 8.3.5), it is possible to simplify by yo after putting z = ±0. Performing this.operation we deduce N(xo, yo, f0) xO H(xo o xo so) so H(xo, yo) . (8.3.23) 307 THE THREE-DIMENSIONAL WING Adopting the definition f-'0 11. f- D2 (8.3.24) , we shall prove that j = lim lID. f ( r1)N(xo, Uo, z)ddq = 0 . (8.3.25) Fore small enough, we may perform the replacement n = y in the integrand. Hence, rx = C-0 0o lirn ( x-kIzl ao [f-" f (F, n)N(xo, yo, z)di7 I d _ c v-c d= 0 [f()N(zotz)jdfll 1 - . Using the form ula (8.3.20) from (8.3.7) we deduce: w(x, y, +0) = T 2 f1(x, y) + 27r JJ f(f, n)N(xo, yo)dtdo . (8.3.26) D Adding and subtracting the boundary conditions w(x, y, f0) = h..(x, y) ± h1x(x, y) (x, y) E D , (8.3.27) one obtains: h71 f(f,n)N(xo,yo)dt do = h.(x,y), ft (x, y) = -2h1x (x, y) , (x, y) E D. (8.3.28) (8.3.29) The equation (8.3.28) is the lifting surface equation in the supersonic stream. It can be also written as follows: f(") _ , yo o::ll xox k (8.3.30) o D1 representing the shaded domain from figure 8.3.5. The analogy of this equation with equation (5.1.28) is obvious. The equation (8.3.30) was given in [8.7). For the sake of simplicity we shall name it the equation D. 308 THE SUPERSONIC STEADY FLOW The Equation D in Coordinates on Characteristics 8.3.6 We know from (8.3.15) that if the current point M is on the wing (x, y) E D, z = 0, the hyperbola C from (8.3.4) degenerates in the characteristics MMl and MM2 (fig. 8.3.5) having the equations t -- krt = x - ky respectively t; + kr' = x + ky ( and q are the variables, x and y are the coordinates of the point M). Performing the change of variables C, ,q - a, Q, x, y -+ a, b defined by the formulas - kn = a x-ky=a t;+kn =Q x+ky=b, (8.3.31) we deduce 2kdt; drt = da d#. The characteristic MAi1 has the equation a = a and the characteristic MM2, the equation Q = b. The domain Dl given in the old variables by the inequalities t-k, <x-ky, +krl<x+ky (8.3.32) will be characterized in the new variables by a < a; f3 < b . (8.3.33) The axes OA and OF will be the new axes of coordinates because on OA we have /3 = 0 and a is variable, and on OF, a = 0 and 0 is variable. The coordinates of M with respect to the new axes are (a, b) (fig. 8.3.6). Since the new axes are characteristic lines, the coordinates a and /3 will be named coordinates on characteristics. In these coordinates the equation (8.3.30) becomes f(a, k 2rr Dl a - a + b -,8 (a - a - (b - Q)(2 dadQ = H(a b), (a a)(b - /3) (8.3.34) where we denoted by f (a,,6) the unknown f in the new variables, H(a, b) the function given by h=(x, y) in the new variables and Dl, the domain D in the new variables, i.e. the shaded domain from figure 8.3.6 (a<a,#<b). Obviously, the integrand has in M (a = a, Q = b) a strong singularity. We shall isolate therefore this point drawing the parallel Q = b - f THE THREE-DIMENSIONAL WING 309 Fig. 8.3.6. and denoting by D6 the parallelogram indicated on figure 8.3.6. We adopt in (8.3.34) the definition Jf , C0 (8.3.35) if, -n. * In this way, the non - integrable denominator from (8.3.34) does not vanish (even if a = a). In Dl - D6 we may utilize the identities a - a+b _ 1 -a-(b- /3)]2 n-a b- 2 0 a s a- a -(b-Q)- P-i a-a+_/_-_ 82 -2 8a8b In _ I a --a - VS --- I I ' (8.3.36) such that the equation (8.3.34) may be also written as follows k0 ('Pv, f(a, Q) dadf T-,6 a-a- (b-Q) it 8a k 82 ;r 8aOb a --a JID,Di H(a, b) , a --a + V- da d,B = H(a, b) (a, Q) In I a --a - ../b-/31 (8.3.37) (8.3.38) 310 THE SUPERSONIC STEADY FLOW in order to have weaker singularities. In [8.7) one gives another form for equation (8.3.37). The Plane Problem 8.3.7 We act like in 5.1.7. In order to have a plane-parallel flow, we must consider the case when the conditions that determine the flow are iden- tical in every plane parallel to xOz. We assume therefore that the equations (8.3.1) have the form z = h(x) ± h1(x), (8.3.39) and the domain D is rectangular, such that the span 2b tends to infinity (b - oo). Hence we assume that D is characterized by the inequalities 0 < x < 1, -b < y < b. These conditions imply ff = fi(t) and f = f (t;). Noticing that 2 100 H(xo +0o Ed xp k 1.00 _ - u + k z) du -k'z2 u °-k I 2T 2 kH(x0 - lID f()8!! (y02 (8.3.40) xo - k222 - u o = kH(xo - kizi) du ldv v 10 = -H(xo - kjzi), +z2E/dj?= f [.1:(j2 E)cli}df 1 we deduce the representation (8.1.17). (8.3.41) =0, THE THREE-DIMENSIONAL WING 311 8.3.8 The Equation of Heaslet and Lomax (the HL Equation) This equation may be deduced from the representation (8.3.6). To this aim we denote L(x, y, z) 8z (/ f f fx° Di 6z . !l 2 do = 82.+8 + (8.3.42) d +z2 Y_ 747-77 the functions Y+, Y_ being defined in (8.3.16). Here the change of variable (8.3.19) is not indicated, because the term that one obtains by deriving the superior limit of the first integral becomes oo for z --+ 0. We act therefore as follows: r-k L = limm Y+ as J-oo f (4, q) x0z a yp + z2 fY_ (8.3.43) so Ll (x, y, z) + Twki,1 xot(x0, Y, 44, where, after performing the change of variable 9 -, it : q - y = u, we have: V+cf(x-k[+ £ q) L1=-k2Jim c-+o lyt - e- do + z2 +s _ -kz2f (x - kizl, y) hm t(xo, y z ) F, f ft -ki f (x - kjzj, y), e - u (u2 + z2) = (8.3.44) Y+ z f (F, n) Y_ x0 - a Y + z d ry = a(s) 8 as a(:) F(u, z)zdu, +z (8.3.45) with the notations a(z) _ x8 - k2z 2 , F(u z) , M,u+y) (8 . 3 . 46) x - k u2 + z) We deduce therefore L(x, y, ±0) _ -ka f (x, y) + f xot(xo, y, ±0)d.. 00 (8.3.47) 312 THE SUPERSONIC STEADY FLOW For determining the function L(xo, y, t0) we notice that a'(+-0) =0. It results that the derivative from (8.3.45) and the integral interchanges. We have therefore: l(xo, y, ±0) = _ limo sL m 8z +* OIF f 8uOz 8 zUM +° 8.. J C r a { [F(a, z) + F(-a, z)] arctan z-»f0 J +a OF f4 = }- U2 + z-.f0 lim a z2 + a2 z +a OF arctan z -d u + yli lim (F(a, z) + F(-a, z)( u F(u, z)d arctan u +0 f + Z2 8F du= z 87Z u2 + ..2 d u+ c u2+z2du. 8u But lim z* .O lim z 0 r+n OF f a J du= z J z-.f0 8z 8Z u2 + z2 +° OF 8F (0,z): /+` du lim u du a 8u u2 + z2 _ - el 0 z 0 f7rBF f u2 + z2 = 8z ( +c OF r (-a j+d019F ) + Ja u 8; u2 + z2 0, 0) du= du (u, U) u where ao = a(0). Finally, taking the formula (D.3.7) into account, we obtain: t( x0, y, f0 ) _ + W OF ( 0 , 0) = F(oo, 0) - F(-ao, 0) + ao ao r '/ +°O OF (u, 0) d 1 8u u± u (8.3.48) F(u2 0)d u -,-0 f BF(0 , 0) . Having in view the expression of F(u, z) (8.3.46), we deduce: L ( x, y, f0 ) = -k rf (x, y) + f oo x0 T y f(t, n) do x =k-2-y- d on d t (8.3.49) 313 THE THREE-DIMENSIONAL WING Hence, w(x,y,±0) _ T 2ft(x,y)+ Zf(x,y)(8.3.50) [* T7r v- s f xk'y yo L771 d such that, imposing the boundary conditions (8.3.27), we find: f kf (x, y) - ! J xo [ T ok2yo 24(x, y), (8.3.51) d ao J I1 (x, y) = (h' - (8.3.51) is the HL equation. Integrating in (8.3.42), at first with respect to respect to y, we obtain w(x,y,±0) = T 2ft(x,y) - oo « 2zr (8.3.52) (x, y). . ° r-kh#o) d2y b0 and then with [.! ro-k o dl . 8.3.9 The Deduction of HL Equation from D Equation This deduction was performed by V. Iftimie. To this aim, we shall extend the integral from (8.3.30) from the shaded domain Dl, in figure 8.3.5 to the infinite domain D&, situated between the characteristics MMl and MM2 extended to infinity, putting f = 0 in the exterior of the arc M1 M2. We denote by I the integral obtained in this way and we perform the change of variables t, n -+ u, v: u=x-Z;, v=y-y. (8.3.53) In the new variables, the domain Df, will be transformed into the domain Du,, (fig. 8.3.7) determined by the angle Mi MMz. Denoting g(u, v) = f (x - u, y - v), we deduce f(x,y) = g(0,0). (8.3.54) THE SUPERSONIC STEADY FLOW 314 Fig. 8.3.7. The HL equations result from the formulas g(u' v) u dudv v2 u-k _ d =-k,rg(0,0)+or duTAOU,v 00 = it v) J u u (8.3.55) k akv which have to be demonstrated (the demonstration was given by V.1ftimie). For proving the first formula, we isolate the singularity by two parallels to the Mu axis (fig. 8.3.7) at the distance e from this DD axis. We denote by D' and D" the domains composing and by I' and I" the integrals I corresponding to these domains. Using the definition (8.3.24) we get I = limo(I, + 1") . Putting v = tu, we deduce: I'- 11 JfJJJJ g(u,v)u v2 dudk u _ r0O 1 =J. u[u /k oc v -lip uI,J (8.3,58) r JJ t2 g(u,tu) 1-k t dt du. u 1 dv'du 315 THE THREE-DIMENSIONAL WING We expand g in a Taylor series: g(u,tu) = g(u,0) + tug"(u,0) + t2u2m(u,tu). and we introduce this series in the above integral. Acting analogously for 1", we deduce: - °° u F f-` I" 9(u, v) d vl d v2 u -k J 1-v-1g(u,tu) kE u l Jl t2 1- k t u= dt1du whence, f 00 1' + 1" = g(u, 0) [ JiY U Jibe r t2 1- k t +Jits g°(u,0)LI=" t lack t + i m(u, tu) uf LJ4 +r dt 1-k t d The factor which multiplies L J_ +J-I t+!J_ 1t dt du+ t2 1- k "t t I- m(u, tu) 1 t t,du+ d t d u. is zero if we replace in the second integral t by -t. An elementary calculus gives rk1-k t dt 411 _r dt _ u -k E t2 1- k t - J_ t2 1- k t whence, P+ I" E e k2E2d u+ u (8.3.57) +fJ-u(j1+j-'U) m(u,tu) dtdu When a -+ 0, the first integral is a Finite Part. Hence 2 /'O g(u, 0) C JkE U 1 9(u1 0) - 9(01 0) u2 - k2E2d u E fkC U u2 k2E2+ 316 THE SUPERSONIC STEADY FLOW u u +E g(0,0} ke du+E foo 9( L+0) u2 - k2F2 d u . JEc The first and the last integrals from the right hand member are finite. When we consider the Finite Part, we neglect these integrals because they h a v e the factor E t . One calculates the integral from the middle and one obtains FP 2 0o E Jke u2 - k2e2d u = -kirg(0, 0) . g(u' 0) U Hence, from (8.3.57) we deduce: r00 I= lim(I`+I") =-k7rg(0, 0)+J o oJ I m1(uk tut) , d t]d u. (8.3.58) We set now J= f 00 [1" 9(u,v)v f dUk Jdu. u (8.3.59) Using the same Taylor expansion we deduce: J' _ J6 u v) g(' [Ilk u d v= - Js v2v Y d tk 9(u. 0) t2 VI U t +9u(u1 0)J° f I m(u, tu) +u f 1 - 00 0 J =Jr 9(u, v) dv u - kv +gv (u, Y 0) J dt g(u, tu) u t2 1 - k t t latkt - + d t, u dt t 1-k t f_1 -{- t2 1-k t+ .. . whence, J' + J" = 2 g(u, 0) u - k e + u{ m(u, tu) U 43 1-kt dt. 317 THE THREE-DIMENSIONAL WINC When we consider the Finite Part (FP) we neglect the first term and we get li m (J' + J") = u m(u, tu) f J= / 1-k t u Jo [ JI m(u, tu) d t d u. 1-k t (8.3.60) From (8.3:58) and (8.3.60) it results I = -k7rg(0, 0) + J, (8.3.61) which demonstrates the first equality from (8.3.55). For proving the second equality, we denote: 00 K I T0, J u g(u,v) kv dudv= lim(K`+K"), where (we may change the order of integration) K' - r-E r r J_O0 sr fI g(u, v) d v2 7u!=-=kFJug(u, v) kv v2 u - u v 0o v du = I', (8.3.62) K" = 100 [/ r - Jkr d ul d v = ky v2 u- JC v h 00 u Jk uy(u, v) du] d v = c u g(u,v) i v2Vu- 72 -k v dv du = l" Hence K = I and the demonstration is finished. The HL equations resulting from (8.3.27) are kf (x, y) - a -1 f a'° 00 l vklvo! co p o° f(") d? l d t = 2h.(x, U), k Uo V5 o-n)k o F--7 d t (8.3.63) = 2h= (x, y) These are the equations (10-21) and (10-22) from (8.151. (8.3.64) 318 THE SUPERSONIC STEADY FLOW 8.3.10 The Equation of Homentcovschi (H Equation) This equation results from the representation (8.3.9). Taking into account (8.3.20), we deduce: w(x, y, t0) = T 1h (x, y)+ +2 f l rr (k2 52 +J 2/ _ =o Id f H(T ^s)drdn, L .0 o (8.3.65) such that imposing the boundary conditions (8.3.27), we obtain the equation (k 2 ll axe - r7y2) ff f n) f r:o d T d dn = 2 (x,b) J H(T - so) J (8.3.66) which was given for the first time (in a different way) by D.Homentcovschi (8.16]. We have to notice that after evaluating the interior integral with the formula (8.3.11) one obtains the equation ! .. (k2..02 _ vx-!:: 02 aye) Jf, f (E, n) In xa + 40 , sod d r1= 2h (x, b) (8.3.67) In the variables on characteristics this equation becomes (8.3.38). However, Homentcovschi does not follow this way. Inspired probably by the papers of Evvard (8.9], Ward (8.33] and Krasilscicova (8.20], he intro. duces the unknown N, related to f by the equality N(x, y) = 1-00 f (t, y)d t, (8.3.68) f (x, y) = N= (x, y) . (8.3.69) whence it results We have seen in (5.4.5) that o function similar to N also intervenes in the theory of low aspect wings in a subsonic stream. Taking (8.3.69) into account, the integrand from (8.3.66) becomes: N :° H(T - so)dT J - = 0 [NJ° H(T - so)dT 9[NH(x(, -4o) f°u x2 -4 o dT a_ so) 0J V0- o _ 319 THE THREE-DIMENSIONAL WING [N(f , n) In 'To + sp o - so] + = H(xo - so) { F(I,,,) 1. 0 0 In this way, after applying Green's formula, the equation (8.3.66) becomes 111I N(C, g) d i d n+ a_ 82 (kz xo-8o 81l + POD N(t, Y)) In sod xo + so 1 (8.3.70) n} = 2h.(x,1!) In variables on characteristics this equation becomes k a dQ 02 8 AD l - 0) + V'r(a - C v1_a --a + JeD; v(a' Q) In a-a + VT--71 d ( - a) } = H(a, b). (8.3.71) From (8.3.68) it follows that N(£, n) vanishes on the leading edge, hence on EFAB. N = 0, (8.3.72) Indeed, for every point n) E EFAB, the integral (8.3.68) is calculated on the half -line having the applicate q upstream of the wing, where f = 0. We also notice that the curvilinear integral from (8.3.71) vanishes on the characteristics a = a and (3 = b (fig. 8.3.6), because the logarithm vanishes. Hence, for a given M(a, b) , the curvilinear integral from (8.3.71) will differ from zero only in those points of the trailing edge belonging to 0DI. For example, for the wing from 8.3.1, the curvilinear integral will differ from zero only for M situated in the zones of influence of the subsonic trailing edges BC and DE, respectively in the domains BCB' and EE'D. Utilizing the definition (D.3.6) for (8.3.71) we also have the equivalent forms: k 4ir N(a) p)d a d.8 f JD1 (a - or)-111(b - 0)312 + (8.3.73) +_7r 8a JD a - a - (b, Q) -d(A-a) =H(a,b) 320 THE SUPERSONIC STEADY FLOW k "f 4;r J N(a,A)da dA 7D) (a - c')3/2(b - 10)3/2 8 f N(a, A) d (8.3.74) +k 8b oD1 b-,C-(a-a) a-a (A - ) = H(a, b), the curvilinear integral being different from zero only in those points of the trailing edge where a < a and A < b (see the definition of D1 in coordinates on characteristics (fig. 8.3.6). The equation (8.3.71) is the equation of Homentcovschi (H equation). In the following subsection we shall present the solutions of this equations as they were determined by the author himself. The Theory of Integration of the H Equation 8.4 8.4.1 Abel's Equation Abel's equation has the form x f"Jdx=h(Y)1 -x a< y. (8.4.1) This was the first equation encountered in applications. Multiplying (8.4.1) by (A - y)-1/2 and integrating with respect to y on the interval (a, A) we obtain: dy J rV Ja f(x) dx y-x rA h(y) dy -JQ Vx --y Changing the order of integration in the first member (fig. 8.4.1), we have: A A f (x) J= dy (A - y) (y - x) and then f f(x)dx = I fA dx= Jay h dy. A h(y) dy Deriving this relation and utilizing the definition of the Finite Part (D.4.3), we deduce: d f (A) = a d A f'\ hid y I 21r , {A h(y)3/2 d y. (8.4.2) THE THEORY OF IN BGRATION OF THE X EQUATION 321 y a x 1) Pt 8.4.1. It results that the solution of the equation (8.4.1) is (8.4.2) and conversely, the solution of the integral equation (8.4.2), where h is the unknown, is given by the formula (8.4.1). 8.4.2 The Solution of the H Equation in the Domain of Influence of the Supersonic tailing Edge If M is in the zone of influence of the supersonic trailing edge (fig. 8.4.2), as we noticed at the end of the section 8.3., the curvilinear integral from the equations (8.3.73) and (8.3.74) disappears. We see in fact on the figure 8.4.2 that on the arc BM2 we have a > a and on the arc Ml E, /3 > b. It remains to integrate the equation: k 4a N(a, /3)d a d,6 JDI (a - a)3/2(b - = H(a, b) . (8. 4 . 3) /3)3/2 Denoting: a = A(,B), the equation of the arc DEFA and (8.4.4) /3 = B(a), the equation of the contour FABC, t h e equation (8 .4 . 3 ) may be wr itten as follows k'b Ni (a, fl) 4ir JBi,l (b -8)3/2 d = H( at b) , (8. 4 . 5) THE SUPERSONIC STEADY FLOW 322 Fig. 8.4.2. where r Nt (a, Q) _ N(a' a) A(p) (a - a)3/2 (8.4.6) d c t. Utilizing (8.4.2) and (8.4.1), we deduce from (8.4.5) 2 _ (a, Q) J °) H(Nl 3d (8.4.7) br In this way, the equation (8.4.6) becomes _ 1 27r N(a,fl) da = 1A(8) (a - a)3/2 T11 H(a,b')db'. (8.4.8) $(°) The equation (8.4.8) is similar to (8.4.2), so that 1V(a, A) ra d a' JA(9) a--a, 1 r H(a', b') dg. Irk ,lg(a') 13 b' For a = a, 6 = b we obtain N(a,b) _ 1 " Irk IA(b) da' H(a',b')db, b - b' a --a' J Br(.') b (8.4.9) 1 If r(ab) H(a',b')da' db' a - a' (b - d) THE THEORY OF INTEGRATION OF TIIE H EQUATION 323 This is the solution in the zone of influence of the supersonic leading edge. 8.4.3 The Solution in the Domains of Influence of the Subsonic Leading Edge We assume that M(a, b) is situated in the zone of influence of the subsonic leading edge AB (fig. 8.4.3). In this case, the curvilinear integral also vanishes, because on BM2 we have a > a, and on M1E, ,8 > b. The equation which has to be integrated is also (8.4.3). Obviously, A(b) < a1 < a, B(a) < b < b1. The equation (8.4.3) may be written as follows k 47r J N(a, /3) da (b) (a - a)3/2 B(a) (b - 0)112 d p = H(a , b) , (8.4.10) such that we denote N2 ( a, b) ( 2 JB(a) (b we obtain the integral equation Fig. 8.4.3. 0Q/2 d 1 (8 .4 11) . 324 THE SUPERSONIC STEADY FLOW k N2(a, b) JA(b) (a - a)3/2 da = H (a, b ) ( 8 . 4 . 12) whose solution is N2(a, b) = -1 k H(a'' b) J A(b) a- a' d a (8.4.13) for a > A(b). From (8.4.11) we obtain for N(a,/3) the following integral equation 2 r r l B(o) (b (IB 2 d 1 3 = 7rk IA' (b) Haas' d a' (8.4.14) Utilizing again the solution of the equation (8.4.2), we deduce: N(a, R) = irk 1 J B(a) for a > A(b) and N(a,6) - d &' H(O' b') d a , f(- b JA(bl) a - a' > B(a). Putting a = a, Q = b, we obtain: 1 db' JB(4) b - b' (8.4.15) H(a', b')d a' d b' D] Irk fN)a=' H(a,b')dIrk (0) V'r(a, - a) (b - b') Dl being the shaded domain from figure 8.4.3 We notice that in this case D1 is not the entire domain determined by the leading edge and the characteristics issuing from M. From this domain one eliminates the strip where b' < B(a). This result was obtained for the first time in 1949, independently, by Evvard and Krasilscicova and it is called in some books the theorem of Evvarrl and Krasilscicoua The solution is obtained analogously when M is in the zone of influence of the edge FE, with the difference that in this case one eliminates from Dl a strip parallel to the 0/3 axis. 8.4.4 The Wing with Dependent Subsonic Leading Edges and Independent Subsonic Trailing Edges For a wing with dependent subsonic leading edges and independent subsonic trailing edges (fig. 8.4.4) the solution in the domain bounded TIE THEORY OF INTEGRATION OF THE H EQUATION 325 by the curve AHFA is given by the formula (8.4.9), the solution in the domain bounded by ABH'HA - by the formula (8.4.15), and the solution in the domain bounded by FHH"EF - by a formula analogous to (8.4.15). The case when M is in the common zone of influence HH'B'E'H"H of the subsonic leading edges is presented in the sequel. Fig. 8.4.4. We notice at first that in this case the curvilinear integral from (8.3.73) also vanishes, because on BM2 we have a > a, and on All E'3 > b. Hence, we have to integrate the equation (8.4.3). Denoting by R1, R2 the domains bounded by the curves FM"RF, respectively FMIt'M"F and by Q1, Q2 the domains bounded by the curves AQA1', respectively AM'M'A, we notice that on Rl and Q, we have N = 0 (it results from (8.3.68)). Hence, the domain of integration from (8.4.3) may be prolonged to the domain bounded by MRFAQM, i.e. to the domain D2+R1+R2+QI +Q2, where D2 is the shaded region from figure 8.4.4, i.e. the region bounded by the contour Iii"'AMA! A.!"Al't'. The leading edge of this domain is entirely supersonic, such that the solution of the equation (8.4.3) is given by the formula (8.4.9) N(a,b) k [Jf.+JJRi+fJR,+ff, +ff. (8.4.16) H(a', b') d a' d b' . (a - a)(b - b') THE SUPERSONIC STEADY FLOW 326 Let us consider now that M belongs to the zone of influence of the subsonic leading edge AB. From formula (8.4.15) we deduce: 1V(a'b) irk IJJ + f+[ , R, , (a-a)(b-b') da'd (8.4.17) Similarly, taking into account that M belongs to the zone of influence of the leading edge FE, we deduce: N(a, b) Tk- f ffDz +If +f fQz, Q1 H(a', b') (a - a) (b - b') dadbl. (8.4.18) From the formulas (8.4.16) - (8.4.18) it results N(a,b) irk ff bi)da'db', (8.4.19) (aH(a',b') this representing the solution in the common zone of influence of the two subsonic leading edges. We notice that the solutions from the previously considered domains have the same form, differing only the domains of integration. But we may establish a common rule for determining the domains of integration. They are bounded by the parallels to the characteristics issuing from the points where the first parallels intersect the leading edge and by the remaining portion of the leading edge. In the first case, when M is in the zone of influence of the supersonic edge, the parallels to the characteristics issuing from the points where the parallels from M towards infinity upstream intersect the leading edge, do not intersect any longer this edge. 8.4.5 The Wing with Dependent Subsonic Trailing Edges We consider now a wing for which the parallels M'M", M"Mf V intersect in a point P belonging to the interior of the domain D (fig. 8.4.5). In this case, denoting by R2 the domain bounded by FM'PM"F and by Q2 the domain bounded by AM'PMIVA, (the domains R1 and Ql keeping the same definition) and performing the same reasoning like in the previous subsection, we obtain successively N(a,b)=1(D3+R1+R2+D4+Q2+ Q1) N(a, b) = I(Ds + Ri + R2) , N(a, b) = I(D3 + Q2 + Q1) , 327 THE THEORY OF INTEGRATION OF THE H EQUATION R'%G A Fig. &4.6. I representing the symbol for the integral appearing in (8.4.19). It results the solution N(a, b) AD, (a H(a,' '){ b d a' d b'(8.4.20) kiDD. H(a',b') (a-a' (b-b') dadb' The result may be generalized for every wing having finite dimensions. For example, for the wing from figure 8.4.6 the solution in the point M(a, b) is N(a,b)=I(D3-D4+Ds-Ds). (8.4.21) We stop here the presentation of the solutions in the zones of influence of the leading edge. 8.4.6 The Solution in the Zone of Influence of the Subsonic Edges under the Hypothesis that the Subsonic Leading Edges are Independent For the wing from figure 8.4.7 the solution is determined in the domain bounded by the curve ABB'E'EFA. It remains to determine the THE SUPERSONIC STEADY FLOW 328 solution in the zones BCB'B and E'DEE'. To this aim we shall use the method of Homentcovschi [8.16]. Fig. 8.4.7. Let M(a, b) be situated in the last zone. In this case, the curvilinear integral from (8.3.73) does not vanish. More precisely, it is zero on BM2, where a > a, it is zero on M1M", where 6 > b, but it is not zero on WE. We assume at first that N is known on the trailing edge. Since M is in the zone of influence of the subsonic leading edge FE, the double integral may be inverted with a formula similar to (8.4.15). We have therefore N(a ,b) = fl EM" 1 Irk AD, (a,b) /a'-a H(a',br)da'db' (a ---a') -(b- b' b'-p a'-a-(b'-p) + 1 W2 d(!g-a)1 AD ' 8a'. da'db' (a-a')(b-b') (8.4.22) where D1 (a, b) is in the shaded domain from figure 8.4.7, i.e. the domain A(b) < a' < a, b' < b. In front of the last integral we have the 329 THE THEORY OF INTEGRATION OF THE H EQUATION sign + because we have changed the sense on WE to EM". On WE we shall denote N(a,10) = N(A(Q), Q) = N(3) , (8.4.23) because the equation of the edge DEFA is a = A(13). The curvilinear integral imposes to eliminate from D1 (a, b) the points where b' < p. D1 from the second integral (8.4.22)is therefore the shaded domain from figure 8.4.8. Denoting this integral by T and interchanging the curvilinear integral and the integral on the domain we get: T= --In, a 8a' a[,! a da' -a')(b _ )(b-b') rb [ a' - A(/3) / N(Q) V 1 J [I-A'((3),d0 1. a'-A(l3) b'-(3 a -A(Q)-(b'-0)J + bN(f3)[1-A'(A)]dfJ 2 db' (b' -)3)(b - b') u - b', da' a A(b) 7a =' r?a' 1 , (8.4.24) where u=a'+A-A(/3). Fig. 8.4.8. 330 THE SUPERSONIC STEADY FLOW For calculating the interior integral we shall prove first that b < u, i.e. that b<a'+l3-A(13). (8.4.25) Denoting by P, R, S the points having the ordinates 3 namely, P on the trailing edge, R on the la intersection with the direction of the unperturbed stream and S on the boundary of the domain Dl (fig. 8.4.8), we have R(A(fl),,3), S(A(b), (3), 3 RM"S =3 SRM" = it, p representing Mach's angle (the angle made by the characteristic lines with the direction of the unperturbed stream). It results PS = A(b) - A(#). RS = SM" = b - f3, Since M"P is an are on the trailing edge we have RS < PS whence 6 < A(b) + A - A(#) < a' +,0 - A(/3) , (8.4.26) because a' > A(b). Employing the substitution b'-b+A+b-Qcose, 2 u=b+A+b-Qs 2 2 2 we obtain b db' 1 d0 2 b-# 0 cosO-s' (b'_/3)(b--b') u-b' the inequality b < u implying 1 < s. This integral has the form (B.6.1) and the solution is given by (B.6.4). One obtains: Il = (u -b (u--/B) 7r 7r a'-A(A)][a'+Q-A(Q)- (8 . 4 . 27 ) The following integral has the form ° 8( 1 1 a' -v J 8a' da '=-1 Q 1 2 JAM (a'-v)3/a da' a-a, where v = b + A(fl) - Q < A(b). This inequality results from the first inequality (8.4.26). Noticing that 1 ZA(b) (a'-v)3/2 da' _ 2 fa - A(b) a-a' a-v A(b)-v 331 THE THEORY OF INTEGRATION OF THE H EQUATION we deduce T=- 1- A'(#) N(/3) a - A(b) ) 6 f3-A(f3)+A(b)-b 13-A(a)+a-b 62 (8.4.28) whence N(a,b) _ H(a',b)da' dN +T 1 Irk (8.4.29) ADi(a,b) -vf(-a - a' (b - 6') in T, N(fl) being unknown. The form (8.4.29) is given in [8.16]. We consider-that M tends to the position M" on the boundary. It means that we make in (8.4.29) the substitution, a integral is 1 1= - ll H (a', b')d a' d b' r( ab) - (a - a')(b - b') where L(a') = jb A(b). The first L(a')d a' I irk JA(b) a - a' H(al'b')db' (8.4.30) (8.4.31) (a) b = B(a) representing the equation of the edge FABC. Obviously, this integral vanishes when a - A(b). We shall perform in the expression of T the substitution 0 - t 13 - A(/3) = t and we shall denote N(Q) = N1(t). We also denote a-A(b)=E, b-A(b)=c, b2-A(b2)=c2. (8.4.32) From the first inequality (8.4.26) it results 13 - A((3) > b - A(b) whence C2>C. Hence, T,= ff Ni(t) 1r c dt t - C+E We shall integrate by parts setting u=N1(t), dv= 7tIm td(t-c) - c+E It results V= arctan rLe--f = The integrated term vanishes because t = c2 implies A = b2 and . N1(c2) = N(b2) = 0. 332 THE SUPERSONIC STEADY FLOW r We obtain therefore: T=--J and 2 f_m T = -- 2 f Ni (t) arctan Vt e cd t (8.4.33) C2 Ni(t)d t = Ni (c) = N(b). In this way, passing to the limit in (8.4.29) we obtain N(A(b),b) = N(b), i.e. an identity. This means that for an arbitrary given N(b) , the function N(a, b) given by (8.4.25) is a solution of the integral equation (8.3.73). One obtains an indetermination like in the subsonic case). This indetermination exists in the zones of influence of the subsonic edges EE'DE and BCB'B. In these zones the integral equation of the problem is not sufficient for determining the solution. Like in the subsonic case we remove this indetermination imposing the Kutta-Joukovsky condition. Imposing a finite velocity on EE', it results that the jump of the velocity on EE' is finite. Since the jump of the component u is given by the jump of the pressure with the changed sign, and the jump of the pressure by f (x, y), it results that it is sufficient to impose for f to have finite values on the subsonic trailing edge. We have f(x, y)=8xN(x,y)_ \8a+ bN(a'b)=(8a+gb)(I+T). (8.4.34) Performing in (8.4.30) the change of variable a' -- s : a - a' = _ (a - A(b)]s and keeping the notation a - A(b) = s we deduce I I= rkoJ L(a - es) f = 2 L(a) - O(£3/2) _ a V - A(b)L(a) - O(e312) , whence a + ) I k .1-AI(b) = a fk Ate)) L(a) + jb (a) H(a, b')d b' + 0(E) b - b' (8.4.35) . 333 THE THEORY OF INTEGRATION OF THE H EQUATION From (8.4.32) we deduce T=- 2 J 7t (b2-A(bz) t- b+ A(b) NfI (t) arctan &-A(b) a - A(b) dt whence, s a as + ab b:-A%) 1- Al(b) T= t- b+ A(b) N' (t) Jb ?r/e- t-b+a b-A(b) V d t + O(,/E . But t-+A -b (b) t-b+a t --b + A(b) = 1 t-b+A(b)+e t-b+A(b) [1 + O(e)) , such that finally we have: a + :b) T (aa N'(t)dt 1 - Al(b) f*2-A(b2) 7r f It - b + A(b) -A(b) +O(f). (8.4.36) We obtain therefore f(x,3!) = A'(b) 1 1 [k b B(a) H(a, ) db,+ %lb (8.4.37) -A(b z) LAb) N(t)d t 1 t --b + A(b) + O(f) . The function f (x, y) has finite values on the trailing edge (a -+ A(b), e -+ 0) when the square bracket vanishes i.e. when Ni (t)d t pb:-A(b,) 1 fb t - b + A(b) Jb-A(b) k JB(A(b)) H(b(b), ) d b' -G(b) . (8.4.38) This condition is an integral equation (of Abel type) for determining the unknown N' (t). Using the notations (8.4.27) we may write this equation as follows j2 Ni (t)d t (8.4.39) J v where Gl (c) = C1 (b - A(b)) = G(b). We deduce f'2 z dc C (C3 N'(t)dt -xJc t-C - t z C, (C) xdC. 334 THE SUPERSONIC STEADY FLOW Changing the order of integration in the left hand member we get: C2 NI(t)dt J do (c - x)(t - c) - _ f C2 Gi(c) dc. c-x s . 1. Since Nl(c2) = N1(b2 - A(b2)) = N(b2) = 0 (on the leading edge N vanishes), we obtain j Glt f t --X d t. 1 N, (x) 7r For x = b - A(b) we deduce l Jbz-A(bz) _A(b) Gl dt (t) t - b + A(b) or, using the change of variable t i3 : 3 - A(,3) = t, N(b) G(i3)[1 1 N(b) _ R A'(/3)l d /3 VP - A(/3) - b + A(b) rb (8.4.40) K(fl)[1 - A'(i3)ld/3 (3 - b + A(b) - A(I3) I f' bz where 8 H(A(#), b') db k D(A(#)) VF-7 1 (8.4.41) . This formula determines N in every point of the subsonic trailing edge ED. Replacing this expression in (8.4.28), we may find out T. In the sequel we shall give an explicit expression of T. Changing the order of integration (fig. 8.4.9), we deduce: T= 12 --A b dJ3 ,3 - A(/3) + a b Jbzb2 A (fl) + A(b) - b a _b. [1 K(E3')(1-A'(f)ldA' -b+A (b) -A(f3)J _- (8.4.42) b _ 2 a --A (b) b K(A')I (1Y)(1 - A'(Q'))d // , where, using the substitution p -+ t:,3 - A(0) = t,/3' - A(i3') = t', IV) = f 1 (t - c)(t'- dt (8.4.43) (b - a) THE THEORY OF INTEGRATION OF THE H EQUATION 335 0 b be I 0 be Fig. 8.4.9. From (8.4.26) It results b - A(b) < i - A(8) for every j6, hence for _ P. This implies c < t'. The integral from (8.4.43) has the form (B.6.11). It results Ir 1=- [a-A(b)] -A(.')-b+a whence T=_1 K ) [1-A'(B)]d3 (8.4.44) rb, /3-A(P)-b+a Using this form of T, we express the solution in the domain bounded by the curve EE'DE by means of the formula (8.4.29). 2. If M(a, b) is in the domain bounded by the curve BCB'B (fig. 8.4.10), then we shall utilize the integral equation (8.3.74). Obviously, the curvilinear integral does not vanish on the arc BM'. Utilizing the solution (8.4.15), we deduce: N(a, b) where Tl = plc 1 JJD1(.,b) H(a', b') a - a' b - d a' d b' + Ti (8.4.45) a' d b' =- 2 JJJD, ad-a(b-Y) (8.4.46) ef° 8b' f FL- B(a) N(°`) 1]d a -a b'-a'-B(a)+a THE SUPERSONIC STEADY FLOW 336 Taking into account that the equation of the edge BC is Q = B(a), we denoted N(a, B(a)) = N(a). Dl (a, b) is the shaded from figure (8.4.10), i.e. the domain bounded by MM', the parallels 0 = B(a) and A = b and the leading edge included between these parallels. D] is the domain bounded by the contour MPQM' imposed by the condition a < a'. Interchanging in (8.4.45) the curvilinear integral and the double in- 337 THE THEORY OF INTEGRATION OF THE H EQUATION tegral we obtain JaN(a)[1 T, = T1 a 5-b' - B'(a)] az f ( a) b - b' do ({/_B(a)j ' w-all db' d1 I (8.4.47) where w=b'+a-B(a). The similarity of the expressions (8.4.47) and (8.4.24) is obvious. TI may be obtained from T replacing b, /3 and A by a, a and B and conversely. It results therefore N(a) 1-B(a) da T,=- Ir b-B(a)f \/a - B(a) + B(a) - a a - B(a) + b - a az (8.4.48) and then 1 T1 = - - r° Kl (a) (1 - B'(a))d a aZ where R 1 (a ) I a-B(a)-a+b k A(B(.)) H(a B( )) d a ' Q (8.4.49) (8 . 4 . 50) The solution in the domain bounded by BCB'B is (8.4.45) where Tl is given by (8.4.49). 8.4.7 The Wing with Dependent Subsonic Trailing Edges For a wing with subsonic dependent trailing edges (fig. 8.4.11) the solution in the zone AHF is determined by the formula (8.4.9), the solution in the zone ABPH by the formula (8.4.15), the solution in the zone FHA'E by a formula analogous to (8.4.15), the solution in the zone HF'IA' by the formula (8.4.19), the solution in the zone BCE'IB by the formula (8.4.45), where T1 is (8.4.48), and the solution in the zone E113'DE by the formula (8.4.29), where T is (8.4.44). It remains to determine the solution in the zone IE'B'I, i.e. in the common zone of influence of the subsonic trailing edges. Noticing that the curvilinear integral does not vanish on BM' and WE and utilizing in the first case the expression from (8.3.74), and in 338 THE SUPERSONIC STEADY FLOW the second case the expression from (8.3.73), we deduce that the integral equation has the following form: N(a , Q) k TJ , (a - k Of +7r 8b 0)3/2(b - /3)3/2 a d p+ N( a) al,a) +k f ir 8a AruE 0 - B(a) (B'(a) - IId a a-a b-a-B(a)+a+ a - A(#) (8.4.51) [1 - A'(Q)]d 0 b-(3 a-b-A(O)+# the integral on BM' representing in fact the integral with respect to a on the interval (a2, a) and the integral on WE representing the integral with respect to 13 on the interval (b, b2). 03 b b3 M' a.b) O- Fig. 8.4.11. In this case, the point M(a, b) is in the common zone of influence of the subsonic leading edges AB and FE. Hence the double integral may be inverted according to the formula (8.4.19), D2 representing the shaded domain from 8.4.11, i.e. the domain bounded by the curve THE THEORY OF CONICAL MOTIONS 339 Af AI"AI""A1I,` AI'A1. N(a,b)= 1 %TA' ff , n V H (a' ) da'db'+Tt+T, (a - a')(b - b') (8.4.52) where Tt has the expression (8.4.43), and T, (8.4.27). Setting Al AI"(a -+ A(b)), the integral on BM' vanishes (as we can see on the figure), such that from (8.4.52) one obtains N(A(b), b) = = N(b). Imposing the Kutta-Joukovsky condition, we deduce that T has the form (8.4.43). Similarly we deduce that Tt has the expression (8.4.49). Now the problem is completely solved. In the end, it is at pleasant duty for me to mention that for elaborating this section I utilized especially Homentcovschi's paper [8.16] and the license thesis of my former student Luminita Berechet [8.2]. 8.5 8.5.1 The Theory of Conical Motions Introduction The theory of conical motions was initiated by Busemann in 1943, 18.31. It refers to wings bounded by conical surfaces with the vertex in the origin of the system of coordinates, the body being placed downstream. The surface of such a body is a smooth surface consisting of half-lines issuing from the origin and leaning on a closed curve situated in the plane x = 1(xi = L I). According to the boundary conditions the velocity is constant along every half-line passing through the origin and belonging to the boundary of the body. The hypothesis of conical flow leads to the assumption that the velocity has everywhere in the fluid this property. We have therefore v(mx, my, mz) = v(z, y, z) (8.5.1) for every in . real and positive. It means that the velocity is a homogeneous function having the zero degree. Under this assumption the equation of the potential becomes simpler, the unknowns depending not on three but on two variables. After Busemann, many authors (Langerstrom [8.22]. Germain [8.11], Poritsky [8.28], Ward [8.34], Heaslet and Loniax [8.15], Iacob [8.18], Carafoli [1.5] $.a.) have contributed decisively to the development of this theory. In all this theory, which will be called the classical theory, we make the hypothesis that the motion is conical. 340 THE SUPERSONIC STEADY FLOW Starting from the lifting surface equation in a supersonic stream, one may prove that if the wing is conical, then the solution of the integral equation is conical. For the equation (8.3.30) this thing is done in (8.5], and for equation (8.3.71), in [8.161. In the present subsection, utilizing the solution from the previous subsection, we shall give the solution of the conical motions by particularization. We shall also give the basic elements of the classical method, because they may be obtained directly, without knowing the solution of the lifting surface equation. 8.5.2 The Wing with Supersonic Leading Edges We assume that the surface of the wing is a conical surface. From z = h(x, y) it results that Az = h(ax, ay) and, with A = (l/x), h(x ,y ) = xh (1 , x ) = xg (x ) (8 . 5 . 2) We deduce therefore hr =.9 (x) - r9 \x) and then H(a, b) = F (a ) Fig. 8.5.1. . 8.5.3) 341 THE THEORY OF CONICAL MOTIONS We shall consider now a wing with supersonic leading edges (fig. 8.5.1). Denoting by b = mla the equation of the edge OA and by b = m2a the equation of the edge OF, it is obvious that m1 < 0, m2 < 0, because on OA we have a > 0 , b < 0, and on OF, a < 0 , b > 0. Since the entire domain D is only in the zone of influence of the supersonic leading edge, for every M(a, b) the solution is given by the formula (8.4.9) where D1(a,b) is the domain limited by the curve OAMFO, and H(a',b') will be replaced by F(µ), where b' = µa'. We have therefore to put b = ma and to replace the variables b and b` by m and p. For M in the zone OCD we shall denote N = N2 N2 (a, ma) = N21 + N22 + N23, (8.5.4) where k 1., Y-f*Jo N21(a,m) -- 1 - ad 'a' \J° k + N23 (a, m) m F() 1 N22(a, m) = aa')(al-c)}dµ, c= (a Pa,. 'dµ+ (a - a) (c - a) r°O F(µ) r °'d a' 1dµ , Fv 14lJco oo aa''-c))df (a (8.5.5) N21 representing the integral on OAA', N22 on OA'MF' and N23 the integral on OF'F. Performing the calculations we find: a irk N21 = - µ m+11 C, arctan it m - VM--\ Jt /I F(µ)d µ N23 = - a /"'.W 2 /m+p ak J7-71 arccot rr m - N22 a irk f o µ , )F(P)d, (8.5.6) \m+µ '+' -'1F(µ)dµa. 2µf If - /I it It is obvious that f(x,Y) = (Oa + 8b)N2 (8.5.7) 342 THE SUPERSONIC STEADY FLOW is constant on every half-line issuing from the origin. The flow is conical. If M(a, b) is in the zone OBC, then the solution is kJmt F(-) Jco N2(a, m) (a aa')(a' - c) ) dy (8.5.8) a f m F(p) m+1dp M17-p A and if M is in the zone ODE, then Ns(a, na) - 1 wk -a 8.5.3 (p) z FlN f. V---14 a d a' C (Z (a - a')(a' - c) (8.5.9) "'2 P((µ) m+{A dµ, Im 11 fl V- The Wing with a Supersonic Leading Edge and with Another Subsonic Leading or Trailing Edge Further we shall consider a wing having a supersonic leading edge (the edge OB from figures 8.5.2) and another subsonic leading edge (fig. 8.5.2a)), or a subsonic trailing edge (the edge OE from fig. 8.5.2b)). In this case, the solution is obtained with the formula (8.4.15) where Dl is the domain limited by MFF'AM. For M belonging to the interior of Mach's cone i.e. M in the zone OCEO, noticing that the equation of the line FF' is a= d (it is obtained from the intersection rn of b' = b with b' = m2a') where d = a, we deduce N2 (a, m) w i m2 F'(µ) d ak mi +.1, J0 1 + ;k J,,, µ ad a' (a - a')(a' - c) d 7 F)dµJ ' F(p)d µ rjA (a sa a' + + (8.5.10) a'd a' Jd (a- a) (c - a') ' c being defined in the previous section. The interior integrals are elementary. THE THEORY OF CONICAL MOTIONS 343 Oa b) a) Fig. 8.5.2. For M in the zone OBC noticing that the intersection of the line b' =pa' with b' = b has the abscissa c, we obtain Ni(a,m) - ak a ffl dJ , (8.5.11) rm F(p) m + µd /ml 7 8.5.4 (a-&)(a'-c) L The Wing with Subsonic Leading Edges When the two leading edges are subsonic, it is difficult to utilize the solution from the previous subsection. We shall use therefore Homentcovschi's idea concerning the direct integration of the equation (8.3.73). Assuming that N(a, µa) has the form N(a, µa) = aN(µ), (8.5.12) 344 THE SUPERSONIC STEADY FLOW the equation we have in view reduces to k a2N(µ)d a d µ _ 47r JJD (a - a)3/2(ma - Na)3/2 - F(m) (8.5.13) D being the shaded domain from figure 8.5.3. With the same notation for c (= ma/p), the equation (8.5.13) may be written as follows /° , µ3/2 fmjV('2 4a 0 a2da d is (a - a)3/2(c - a)3/2 J (8.5.14) N(p) J.r` +k 47 r a2d a {a - a)3/2 (c - a)3/2 µ'/2 Obviously, in the first integral a < c, and in the second, c < a. The interior integrals are considered in Hadamard's Finite Part sense. Taking into account the formula da ml f aa d,,=-'2 a a 0 (a f a)3/2 d a, (8.5.15) given in (D.4.3), the equation (8.5.14) may be written as follows k a2do I -N(p)(02 )dp+ 3/2 f a a (a a)(c - a) ,ni o k, 14 (8.5 .16) ml N( 492 cf 143/2 0 2 a2da ld it (a - a)(c - a) / µ We notice that in the first case (c > a), we have f - 4ac In fc + f - 34 (a + c) fac+ 3(a + c)2 8 f - fa a2d a (a - a)(c - a) Jo and in the second (c < a), f V(a -a)(c-a) a2d a ` _ 3 4(a+c) ac+ 3(a + c)2 - 4ac f- + f In 8 . Vc- (8.5.17) The results are the same if we put under logarithm (/ - . Performing the calculations, it results that the integral equation (8.5.16) may be written as follows k T' N) In -2 mµ(M +14)2}dµ=F(m) 1 (8.5.18) THE THEORY OF CONICAL MOTIONS 345 0, (N Fig. 8.5.3. for ml < m < m2. We put N(p) = µNI(µ). Denoting H(m) = k ., In .IA /./ + ///7;, t%' V'' d µ, on the basis of the equation (8.5.18) we obtain the following differential equation m2H"(m) + mH'(m) - 4H(m) = -F(m). (8.5.20) The homogeneous equation has the linearly independent solutions v/ and l//. Hence, the general solution of the equation (8.5.20) is: H(m) = 2cl frra -- 202 + Fo(m), (8.5.21) cl and CG2 representing constant which have to be determined, and Fo representing a particular solution of the non-homogeneous equation. From (8.5.19) and (8.5.21) we deduce the following integral equation for Ni k fm Nl(N) In / + `dµ = 2c1 VG + Fo(m). (8.5.22) Deriving with respect to m we obtain k Ni (l+) d 7r,'µ-m p = cl + - + /o (m) , m (8.5.23) THE SUPERSONIC STEADY FLOW 346 which is the classical equation of the thin profiles. As we have already observed, from the definition of N it results that Nl (N) vanishes on the leading edge. The solution of the equation (8.5.23) which vanishes for m = ml and m = m2 has the form (C.1.14) with a condition having the form (C.1.13). Taking also into account (B.5.8) it results ' mZ 1 fFa(µ) Nl(m) = -7r (m - ml)(m2 - +c2 m1){m2 - dp N-m+ (m - ml)(m2 - m) m mlm2 (8.5.24) The condition (C.1.13) will give x cl + rm C2 mlm2 Voµ +f m, ( nl)(m'l -P) d y= 0 (8.5.25) and will be useful for the determination of the constant cl, after determining the constant c2. In fact, the constant cl is of no interest. The constant c2 which intervenes effectively in the solution (8.5.24) will be determined imposing for the solution to verify the integral equation (8.5.18). This condition is necessary because the solution was determined after some derivations. Writing the equation (8.5.18) as follows 2 T", Nl (µ) d µ+ N1(l1) d 1 µ-m 2m (8.5.26) - 4m jrn, ' Nl (i)K(m, it)d /1 = - k F( m) M3/2 where K(m, µ) is the symmetric kernel K(m, u) = + 1 Inl%FM-VjAI' (8.5.27) we notice that the equations (8.5.24) and (8.5.26) will determine the un- knowns Nl and c2. The replacement of Nl from (8.5.24) in equation (8.5.26) leads to difficult calculations. It is necessary, for example, to know the formulas for interchanging the FP (Finite Part) with PV (Principal Value) and PV with FP (the formula of Poincar&Bertrand [A.27]). FLAT WINOS 347 The equation (8.5.26) may be also solved numerically using the Gauss-type quadrature formulas (because N has the form N1(µ) = (8.5.28) (µ - ml)(m2 - A)n(p), n(p) representing the new unknown). Flat Wings 8.6 The Angular Wing with Supersonic Leading Edges 8.6.1 For the flat wings having the angle of attack e we have F = -e. The theory of the angular airfoil with supersonic leading edges may be obtained from 8.5.2, putting F = -2. Since m 6 m=-, U= b,W, c=-a, a it (8.6.1) it results ° ad a' (a-a')(a'-c) Jo N21 (a, m) a r 1 + m) arctan F-!A+aFT, ( 8.6.2) - k-v [ m -mli arctan - (8.6.3) Similarly one obtains N23 m) 2ac (M -m 2 arctan (8.6.4) We have also a L a'd a' (a-a')(c-a') -a Iv- m + a rI + ml 1i) V u 2` VrM + and then, from (8.5.5) or directly from (8.5.6), N22(a,m)ael,I J' ffm-1 kzr 2 . m in Vm- +f ddµ 1 + µ/ i i- V M (8.6.5) THE SUPERSONIC STEADY FLOW 348 In I we make the change of variable In this way one obtains l1 = x and we denote vM- = q. I=I1+g212-2g13, (8.6.6) where r4 11= J In x+qdx+ i:° In x+qdx=111+112, 12- Ix-g1 ix - qj v 2In x 13= 11 d. + f x Ix - q) =131+132. The integral 11, is elementary (it has an integrable singularity). One obtains (8.6.8) 111 = 2q In 2 . The integral 121 has a strong singularity in x = 0. It must be considered in the Finite Part sense. On the basis of the formula (D.2.2) we have 12i = f9ln(x+q)-1n(q-x)-2x/q + 2Inq x2 q Integrating by parts, one obtains 121 = 2(ln2q+ q 1). (8.6.9) Using (D.2.3) we deduce (8.6.10) 131 = In q. For calculating 112,122 and 132 we make the substitution x = 1/y and we utilize the results (8.6.8) - (8.6.10). One obtains 13 = 0, 11 2 q( -+1 q =2g1u2+2 In (8.6.11) 2 12=2g11n2+1) +g1n2, such that N22(a, m) = -, /m-(2 2 + 1) . (8.6.12) 349 FLAT WINGS In this way, taking into account that m = b/a, the formula (8.5.4), together with (8.6.3), (8.6.4) and (8.6.12) give a b N2(a, b) _ - - Zee ( -m VM1 am b _n2-aarctan v72 a -b - arctan 2 - )! - ab) v a (2 in 2 + 1) . (8.6.13) This is the solution when M is in the zone limited by the characteristics OC and OD . If M(a, b) is in the domain limited by OD and OE, then we use (8.5.9). We deduce that N3 (a, b) k £ 7n2 (m2a - b), (8.6.14) and if M is in the zone OB, OC (fig. 8.5.1) N1 (a, b) _ - k 8.6.2 b mla (8.6.15) The Triangular Wing. The Calculation of the Aerodynamic Action In order to obtain a finite action, it is necessary to consider a wing having a finite area. We assume that in the physical plane it has the triangular form from figure 8.6.1. In order to obtain a well determined wing we must give the coordinates of the points A and F. Let mi be the inclination of the line OA and ai the ordinate of the point A in the frame of reference Oa,J and m2 the inclination of the line OF and a2 the ordinate of the point F in the same frame of reference. Then the equations of the edges OA and OF will be b = mia respectively b = n2a, and the coordinates of the points A and F respectively (ai, bi = ml al) and (a2i b2 = m2a2). The equation of the line is b = (m1 - m3)a1 + mia, where m3 _ 7Tb2a2 - miai a2 - a1 (8.6.16) (8.6.17) 350 'rilE SUPERSONIC STEADY FLOW F(a,.b,) Ap .e .0 D(O.b4) A(a,.b,) Fig. 8.6.1. Denoting by (a3, b3) the coordinates of the point C and by (a,,, b4) the coordinates of the point D, we deduce a3=1-m1, b3=0; a4=0, b4=(m1-9n3)a1. (8.6.18) As we already know, the lift is given by the formula = L - ffD(x.y) Bpi I d x, d yi = -PmU,2,,,Lo J1D(=.y) epO d x d y, and the lift coefficient cL, by the formula CL = 1 L 2P0U00A , (8.6.19) where A is the area of the wing and Lo, the reference wing. Taking (8.3.69) into account, passing to coordinates on characteris- FLAT WINGS 351 tics and applying Green's formula, it results CL = - A rf 2 o JJD(= v) 2L A f f (x, y)d x d y = 8 N(x, y)d x d y ax (8.6.20) A IJD(a, b) (8a + ab )N(ab)da d b = - - Ac /eD(a,bjM (b - a) IA 1, where 1=11 + 12 + 13, It = JOA+AC+cd"1 d(b - a), (8.6.21) 12 13 = JoD+DF+F 3 d (b - a). - JO C+CD+DO-2 d (b - a), Taking (8.6.15) into account, it results Nid(b-a)loA=0 N1d(b-a)IAC_N1d(b-a) b=(ml -m3)a1+msn -E(m3 - m1)(m3 -1) (a k m1 N1d(b-a)I co=Nid(b-a)lb_o- - ai)da kmlada, such that I1 = _E(m3 km1)(m3 - 1) at)da - £'nl 3 ada = To S(m3 - m1)(m3 - 1) (a3 - at)2 k -m1 2 E -m1 a3 k 2 (8.6.22) In the same way one calculates 13. Taking into account (8.6.14), one 352 THE SUPERSONIC STEADY FLOW obtains 13 - k -m2 2 E 443 E(n3 - 1)(m2 - 1n3) k --r n2 a!+ 2 (8.6.23) +Eata2(m3 - 1)(m1 - m3) k -rn2 and the problem of calculating c1, is solved. 8.6.3 The Trapezoidal Wing with Subsonic Lateral Edges We assume that the projection of the wing (which is flat and has the angle of attack s) on the plane xOy is the isosceles trapezoid ABEF from figure 8.6.2, having the bases 21, 2L and the height h (dimensionless quantities). The direction of the unperturbed stream is perpendicular to the bases. Fig. 8.6.2. We consider the case a <,u. The leading edges AB and FE are subsonic and the edge AF is supersonic. We also assume that h is such that the subsonic edges are independent (for the sake of simplicity). Obviously, in this case the fluid motion is not conical. In the domain D2 the solution may be obtained from (8.4.9), and in D1 and D3 from (8.4.15). For developing these solutions, we must characterize the domains in coordinates on characteristics. We must specify at first the FIAT WINGS 353 physical coordinates. We introduce the parameter m by the formula 0< h tangy = < tan p . (8.6.24) It results therefore 0 < m < 1. The distances dl and d2 are defined bu the formulas d1=lk,d2=d1+h. (8.6.25) The equations of the straight lines AB and FE are respectively -y=l+ m(x - dj),y = t + m(x - d,). (8.6.26) The equation of the wing (see figure 8.6.3) is d12d2 ) x (8.6.27) the function being defined on the domain D + D1 + D2 + D3 from the XOy plane: d1 <x<d2 (8.6.28) -yi(x) < y < y1(x), where (x - d1). y1(x) = e1 (8.6.29) k Fig. 8.6.3. It obviously results H(x, y) = -c. (8.6.30) Passing to coordinates on characteristics we put b-a a+b 2 'y 2k (8.6.31) It results H(a, b) = -e (8.6.32) 354 THE SUPERSONIC STEADY FLOW and the following equations for the sides of the domain D AF:a+b=2d1, BE:a+b=2d2 (8.6.33) AB:a=2d1+mob,FE:b=2d1+moa, where we denoted mo = i+ m (8 . 6 . 34) For the vertices we deduce the following coordinates: A = (2d1, 0), B = (d2 + kL, d2 - U), (8.6.35) F= (0,2d1),E= (d2 -kL,d2+kL). Taking (8.4.9) into account, we deduce for the solution in D2 a kir N2(a,b) da' 1-b db' f a - a' J2d1 -a, %Fb - b' (8.6.36) =-1(a+b-2d1). The solution in D1 is given by (8.4.15). Noticing that A(b) = 2d1 - b, B(a) = 2d1 - a, (8.6.37) it results Nl (a' er b) krr d b' , -4 b a da' b' Jet -V a - a' (8.6.38) Obviously, this expression may be obtained from (8.6.36), changing a with b. Hence N1(a,b) = -1(a + b - 2d1) (8.6.39) and an identical expression for N3(a, b) (because of the symmetry). According to (8.6.20), for cL we have the expression 2 CL = - kA (8.6.40) where I= POD N(a + b - 2dj)d (b - a) - (a + b - 2d1)d (b - a) = (1- m02)(d2 - kL)2. AB+BE+EF FLAT WINGS 355 The Trapezoidal Wing with Lateral Supersonic Edges 8.6.4 We consider an isoeoelea trapezium with the bases having the length 2L respectively 2t perpendicular on the direction of the stream at infinity (fig. 8.6.4). We Introduce here again the parameter m defined by the relation: k-tana=Lht>tanw=k (8.6.41) F. 8.6.4. Obviously, m > 1. Denoting by dl the distance from the small basis to 0 and by d2 the distance corresponding to the great basis, we obviously have d2 = dl + h, and for the c arte ian coordinates of the vertices of the trapezium A = (d,, t), B = (da, -L), F = (dl, t), E = (d2, L) (8.6.42) The coordinates on characteristics are obtained from the formulas, a=x-ky, b=x+ky. No ti c i ng th a t (8.6.43) d2 = kL, we ded uce A = (di + kt, dl - kt), B = (2kL, 0) F = (dl - kt, dl + kt), E _ (0, 2kL) (8.6.44) 356 THE SUPERSONIC STEADY FLOW Using the notation d1-k£ d1+2h-k8 k 2 k-1 m+1 (8.6.45) and observing that from figure 8.6.4 we have the compatibility condition ao < p, which implies dl = tan ao < tan p = 1 dl > kt, (8.6.46) we deduce 0 < k < 1. Now, the equations of the sides of the trapezium may be written as follows AB :a+kb=2d2iFE: k-la +b=2d2, (8.6.47) BE:a+b=2d2,AF:a+b=2d1. The entire leading edge is supersonic. The solution may be expressed by means of the formula (8.4.9). To this aim it is necessary to specify the functions a = A(#) and A = B(a). We have: - on the edge BA, b .- BI(a) = -- on the edge AF, 2d2 a, a = Bi 1(b) = 2d2 - kb, (8.6.48) b = B2(a) = 2d1- a, a= t 1 }=Li_b. We must also observe that for determining the lift coefficient we do not need N(a, b) on the entire wing, but only on the trailing edge BE. Indeed, this may be expressed with the formula (8.6.20), and N on the leading edge BAFE vanishes as we have already mentioned in formula (8.3.68). The domains of influence are (fig. 8.6.3) : D1 = ABA'A, D2 = AA'A"A, Do = AA"F"A, D3 = FF"F'F, D4 = FF'EF Hence, we shall put I=11-+-12+I3+ 14, (8.6.49) FLAT WINGS 357 where r it = J BA' I. = r JA"F" 14 = JF"F. 12 = f Ni (a, b)d (b - a), JA'A" No(a, b)d (b -- a). 13 = J Ni (a, b)d (b - a) "F' N2 (a, b)d (b - a), N3(a, b)d (b - a), (8.6.50) . Using the formula (8.4.9) and the equations (8.6.47), we deduce __ 8 N1 (a'b)IBA' = N2 (a, b)IA'A#l a k ,l Bt I(b) E Tr - Bi (b) No= -h, N4= -,-(I - da' Ilrb dillu a -a' B,(u') vb - v 1,9A' da' a - a' b JB2(n') _eb(l-k) db' b - b' L'A" - ... , )a After elementary calculations we deduce: E(I - k) 11 = 14 and finally, k2 k. f d2-"-h bdb e(1 - k) (dl - kf) = k2 Vk- 2 4Eh = - E(1- k) (d1 -2kt)2 ' to =-k2 (k£ - h), 2' kv-. Chapter 9 The Steady Transonic Flow The Equations of the Transonic Flow 9.1 9.1.1 The Presence of the Transonic Flow We call transonic flow the flow which is subsonic in a domain of the space and supersonic in the adjacent domain. One demonstrates (for the potential flow - see [1.21] pp 517, 518) that the equality v = c comes true in E2 only on curves separating the domains where the flow is subsonic from the domains where the flow is supersonic, and in E3 on the surfaces which separate such domains. The name of transonic flow was introduced by Th. von in 1947. In the present paper the transonic flow has been encountered in several situations. At first, we have to mention the one-dimensional flow [1.11] §4.5. The formula (4.5.8) which gives the variation of the velocity against tile variation of the cross section indicates that, in the subsonic flow (Al < 1). the velocity increases when the area decreases and decreases when area increases (like ca in the incompressible fluid), while in the supersonic flow (M > 1) the variations are produced in the same sense. This circumstance leads to the conclusion that in a tube having the shape from figure 9.1.1 the flow may become transonic. To this aim it is sufficient for the upstream subsonic velocity to have the critical value in the section of minimum area. Further since the area of the section increases, the velocity also increases, remaining supersonic. In the linearized theory we deduced for the aerodynamic action the formulas (3.1.33) and (3.1.34) in the subsonic case and (8.1.9) in the supersonic case. It is obvious that these formulas are not valid in the vicinity of A! = 1. For the flat plate these formulas become (3.1.35) and (8.1.22). The figures (3.1.3) and (8.1.3) are very suggestive. In the cause of the subsonic flow with great velocity past thick bodies like in figure 9.1.2. the flow may become transonic. Indeed, considering THE STEADY 'TRANSONIC FLOW 360 Fig. 9.1.1. the flow between the streamline which includes the boundary and a neighbor streamline L, we shall find that the flow is like in a tube. Since the domain between these lines narrows because of the body, it Fig. 9.1.2. follows that in the vicinity of the body the flow nay become supersonic. The transition from the supersonic flow to the subsonic flow is performed by a shock wave S according to the scheme described in 1.3.6. Until S the flow is transonic. We shall deduce in the sequel the equations which describe this flow. The flow with great subsonic velocity past thick bodies is described by the scheme from 9.1.3. -> V<C y« r- Loctaau Fig. 9.1.3. Finally, in the supersonic flow, for great velocities, practically in the hypersonic regime, it appears, as we noticed in 1.3.6, detachedor at- 361 THE EQUATIONS OF THE, TRANSONIC FLOW tached shock waves (figure 1.3.5). Behind these waves the flow is transonic (it passes from the subsonic regime (A,12 < 1) to the supersonic one (A12 > 1)). As we could see, in modern aerodynamics the transonic regime is frequent. So one explains the great number of papers devoted to this subject in the last years. We mention especially the papers of Bauer, Garabedian and Korn [9.1] devoted to the theory of minimum drag wings. There are three dominant methods for studying the transonic flow, namely: 1° the hodagmph medwd, suitable only for the plane steady jet flow (see for example Ferrari and Tricomi [9.11], Manwell [9.30], [9.31] etc.); 2° direct analytical methods, based on the semi-linearized equation of the potential. They lead to integral equations which may be solved numerically; 3° numerical methods applied directly to the system of equations which describes the fluid flow (we mention especially the finite elements method). In this chapter we present some direct analytic methods. 9.1.2 The Equation of the Potential The reasoning based on the assumption that the independent variables x, y, z have the same role in the structure of gyp, (utilized for deducing the equation (2.1.39)), is not valid for the flow in the vicinity of M = 1. Indeed, in this vicinity M2 - 1 becomes itself a small parameter. If, for example AI2 - 1 = O(E), then for V_: = 0(cp) it results +p.y and W.'. = 0(E2). One imposes an analysis of the order of magnitude of the perturbations depending on the geometry of the body and the conditions which determine the flow (Mach's number Al, the thickness and length parameters, the angle of attack, etc.). In fact, the idea that the variables y and z do not behave like the variable x. results from the special property of the Ox axis (which is parallel to the direction on the unperturbed stream). We shall introduce therefore the variables y = u(E)y. 4 = V(C)Z, (9.1.1) expecting for vv(e), like for q(r) from the expansion Or, Y. Z' `) = U (x + T (E),(x, J, <_) + ... ] to be determined by comparing the orders of magnitude. (9.1.2) 362 THE STEADY TRANSONIC FLOW Coining back to the linearized theory 2.1, we notice that for Al = = 1 a catastrophe is produced (it disappears terms from the equations). The lift and moment coefficients become therefore infinite. But this catastrophe has only a mathematical nature, not a physical one. It is determined by the fact that in the vicinity of the value Af = 1, the order of magnitude of all the first order derivatives is not the same (E). One imposes therefore (9.1.1). It results 0r=U(1+1150:+...),OS,=U1)v +...,O:=Urpi (9.1.3) Orr = UtWrr, Ory = UTJVP,,y, 0, = Ugv2 ;OW . From (1.2.17) we deduce c2 = c2 - (7 - l )U2>);pr + 0(112) , (9.1.4) and from (1.2.16) written explicitly as follows (r.2¢r).o +(C?-0y)Oyy-20=.ysOv+...=0, (9.1.5) we deduce [1 - M2 -(y + 1)(M2 - (y + + [1 _ (y - 1)(Af 0(7)2)l(p=r+ (y - I)W=)v2 y- (9.1.6) - 2M2vrl2"o + 0(1j2v2) + ... = 0. For a fixed Al , we see that the equation is consistent if ii --+ 0 when q-'0,so v2, t), 1- M2_ 71. (9.1.7) We introduce now the boundary condition. If z = eh.(x, y) (9.1.8) is the equation of the perturbing surface, imposing the condition to be a material surface i.e. Eh=dOr + Eh,,Oy = & which implies, taking into account (9.1.2) Eh. = (9.1.9) THE EQUATIONS OF THE TRANSONIC FLOW 363 whence E = 9V. (9.1.10) Taking (9.1.7) into account, we deduce t)=E2'3 V=Et13 (9.1.11) When M - I we have to compare. in (9.1.6) the terms of order immediately superior to those which gave (9.1.7). It results I - M2 = = Kv2 whence K- _ hl2 1 . (9.1.12) . K is called the parameter of the transonic similitude. In this way, the first approximation from (9.1.6) (the dominant equation) is [K - (7 + rpyp + 4p: = 0. (9.1.13) This is the equation of the transonic flow (the equation of the potential). It is elliptic if Cpl < K/(7 + 1) and hyperbolic if gyp= > K/(y + 1). The relation V_ = If/(-y + 1) is verified on the surface where V2 = c2. Indeed, using the notations (2.1.3), and taking (1.3.32) into account, the condition V2 = c2 becomes V2=c2=co-7 v;2=cam- 1 o (Vt2-U2). (9.1.14) Here, the dominant relation is u2 1+2 u, } = c2 - (7 - 1)U2vu, (9.1.15) T whence ii = K/(7 + 1). Now it is clear that the non - linearity is necessary for making this transition possible. The first study of the transonic flow has been performed by von Karmsui [9.26). By various methods the problem was investigated by Ovsiannikov [9.531, Guderley [9.15), Cole & Messiter [9.6[ etc. Cole's study from 1975 relying on the method of perturbations was continued by the same author in 1978. In the last study one proves that if we denote by s the thickness parameter and we set for the cross sections y-Et13y, =-E113' , then the potential 0 has the following structure [9.54]: 4)(i',y,4;Al ,Q,b,b) = U[z+E2/39(x,V,z;K,A,B)+ (9.1.16) +E413 t'2(x, y, Z; K, A, B) + ..., , 364 TilE STEADY TRANSONIC FLOW where K is the transonic parameter (9.1.12), A, the parameter of the angle of attack = aft, and 13, the span parameter = 6e113. For p one obtains the equation (9.1.13). 9.1.3 The System of Transonic Flow It is rigorous to perform the asymptotic analysis on the system of equations and not on the equation of the potential which has been obtained from the system by derivation with respect to the x, y, z coordinates. We present here such an analysis which was performed together with professor A. Halanay in the years '80. We utilize the coordinates y and r in the form (9.1.1) and we denote E) =v \x' V(E) , +'7 (E) )' h(x, y, E) = h (x, (9.1.17) It results r"(X,j/,E) = it , x, h, V(£) (2-, y(E)/ V(E) and the boundary condition Elts(x, y) 11 + u(x, y, Eh(x, y))] + Eh' ,(x, y)v(x, y, eh(x, y)) _ = w(x, y, Eh (x, y)) becomes X. (r, y, c) [1 + u (1', lI, - h(x, P, E), E)] + +sv(E)hV(x, N, 0V (x, /, TU _ S,fj, E _ v(E)-(x'v, ll v(s)h(x,p,E),EJ The dominant term in the first member would be Ehx(x4, E) if 9 would not disturb. But for a small p we have hr (x, y) E) =1t= (T, vVE) ) = hx (x, 0) + Y( h=y(x, 0) + r) ... From the physical conditions of the problem it results that le, (x, 0) 74 0. 365 THE EQUATIONS OF THE TRANSONIC FLOW The condition (9.1.18) suggests that the right hand and member has the order of c. Hence, w (x, v(E) h(x, y, e) J = Eii (X, h(x, b, v{s) E)> E (9.1.19) We assume that this is valid in the entire domain occupied by the fluid, i.e.: i (x,y,',E) = ew(x,y,;F,E). (9.1.20) Taking (9.1.19) into account, from (9.1.18) we retain in the first approximation, under the hypothesis that v(E) --+0 111 +u(x,y,0,e)] = ii (x,y,0,E) (9.1.21) Using the notations (2.1.3) the system which determines the perturbation produced by a fixed body in the uniform flow of a compressible fluid characterized by M is determined (see (2.1.10) - (2.1.13)) by the system (9.1.22) (1 + p)M2p = (1 +7M2p)p M2p + (1 + yM2)div v = 0 (9.1.23) (1+p)v+gradp=0 (9.1.24) where [(1+u a 8 )8x+vp,.... (9.1.25) We notice now that from the structure p x, (9.1.26) v(E) it results the formulas Op Ox _ Op ap _ Ox ' 8g Op Op v(e) 8y ' 8~ I 1 Op v(e) Oz ' (9.1.27) which will be replaced in the projections of the equation (9.1.24) on the axes of coordinates. In this way, the projection on Oz gives r here Comparing thu8x 8w Op +evVOw +e2vw +v !=0. (9.1.28) the dominant terms we deduce that P(x, y, =, E) = v(E) 'lx, , z, £) 1 (9.1.29) THE STEADY TRANSONIC FLOW 366 and from (9.1.28) one retains (1+p)(1+u)8 +-=0. (9.1.30) Analogously, from the projection of the equation (9.1.24) on the Oy axis, it results +v(E)v-+EV(E)w-J +E v8/ =0, (9.1.31) From this equation it follows v(x, y, z, E) = 6(x, T, z, E) and then (1 +;5)(1 + u) (9.1.32) e + 5i = 0. (9.1.33) At last, the projection of the equation (9.1.24) on the Ox axis gives (1 + P) }- EYi 1(1 + u) GIV zi + EvUI v(E) = 0, whence we obtain 11(X, Y(E) u(x, (9.1.34) and then 89 = 0 . (9.1.35) + P) Fx + 8x The behaviour (9.1.34) determines for (9.1.33) and (9.1.30) the forms: (1 8; + 8y =0, (1+P)LW (1+P)WV +-=0, 8x (9.1.36) and the boundary condition (9.1.21) determines the equality K(x,y,E) = 10(x,9,0,E) which implies hx(x, y) = w(x, y, 0) . (9.1.37) Knowing that M2 = 1 constitutes a singularity, we shall consider in (9.1.22) and (9.1.23) M2 = I + µ and we shall keep the dominant 367 THE EQUATIONS OF THE TRANSONIC FLOW terms for a small p . Utilizing the previous results, the equation (9.1.22) becomes l+YU1a+£vu+evrI [1+-1(1+µ)EPI 11 p +EVWF K whence we deduce 4 7 v e Pox, , (9 . 1 . 38) and then 89 = ap . (9.1.39) 8x 8x Having in view the damping condition at infinity for the perturbation, from the last equation we deduce p =P- (9.1.40) Taking the relation (9.1.38) into account, it results that the dominant parts in the equations (9.1.35) and (9.1.36) are 8 +8 =0, 8 + =0, 8 F=0, (9.1.41) whence it results u=-p, ex -=0, -=0. (9.1.42) Z Finally, from the equation (9.1.23) written as follows (1+µ){I1+ _u) 2ff +EVv V ax +£VW + +[l+ry(1+µ)ipj[+2+v] =0 ax ay 6F we obtain, if we have in view (µ+vu+µv +'Y(1+µ)vpe (9.1.41)1, +(1+µ)EV(v +10 Lv) + ax 8Y ft 4rV &0 0. 368 THE STEADY TRANSONIC FLOW The dominant part is obtained from the linear terms. We may write therefore P \µ+;u} 49V +7vpax+uj( O-V + E E v v az- J 0, (9.1.43) and the residual equation (-K+u) 2E +YpBx+ a +a =0. Bxp (9.1.44) At last, from (9.1.43), (9.1.44) and (9.1.42) one obtains K= v(£) _ £1/3 try r?x 1 - M2 £2/3 tr =0. (9.1.45) (9.1.4G ) This equation, together with the equations (9.1.42) constitutes the gen- eral system of equations of the steady transonic flow. In the x, V, z space the equations (9.1.42) give the irrotational conditionof the velocity of coordinates (u, u, is). Introducing the potential jp(x, ji, -_,F) by means of the formulas u = c0*, V = cpy, w = Pz' (9.1.47) one obtains (9.1.13) from (9.1.46). 9.1.4 The Shock Equations In the case of the flow with shock waves, from the integral form of the equations of motion (9.1.42) and (9.1.46), written in the conservative form (by means of the div operator), vi+(-u)y=0, vas+(-u): =0 r f Kii I (9.1.48) - L+-'iP 2 +i%+urf=0, 1 : 369 THE PLANE FLOW integrating on every domain which contains the shock surface and passing to the limit as usually, it results [i1n+: [IKu - - Qulny = 0, 9wf ns - QiiOnr = 0, 7+ 2 (9.1.49) 421% + Ovlny + OwOny = 0, where n=, ny, n1 are the coordinates of the normal to the shock surface, i.e. n;r = (d7jdz) ny = (d-zdx) ns = (dxd-y), . (9.1.50) If, for example, the parametric equations of the shock surface are x = x(A1, A2), 11= A2), z = {ai, 2) , then from n = da1z x da3X, it results n= 9.2 "2 8a1 812 "Z 8a1 dJ11da2i ... . (9.1.51) The Plane Flow 9.2.1 The Fundamental Solution \Ve consider. like in Chapter 3, that an uniform stream, having the Mach number M is perturbed by the presence of an infinite cylindrical body, with the generatrices perpendicular on the direction of the stream which coincides with the Ox axis. The Oy axis is in the section perpendicular to the generatrix. Let y=h*(x), 1xI <1 (9.2.1) be the equations of the profile determined by the cross section. Our aim is to determine the perturbation and the action of the fluid against the profile. The flow is obviously plane. The velocity will lie in the xOy plane and will not depend on the variable z. Taking into account the orientation of the axes, we deduce from (9.1.37) the boundary condition v(x, ±0) = h' (x), jxI < 1. (9.2.2) 370 THE STEADY TRANSONIC FLOW According to (9.1.42) and (9.1.46), the perturbation will be determined by the equations il = -n i y - vs = 0, (9.2.3) Kit= + ij = ( + 1) uit= . (9.2.4) Using the change of variables y=fl?i, u'=VY u, (9.2.5) we reduce the system (9.2.4) to NO 8 ay-ax=0' ax a 86 NO 2 gay=kaxu, 9.2.6 ) where we denote ry + I = 2kK3/2. In the sequel we shall integrate the system (9.2.6) without writing the marks *;' any longer. The fundamental solution of this system is determined by the equations au ay av 8u a 8u - ax = ld(x, Y), ax + ay = k ax u2 + m (x, y) . (9.2.7) For the Fourier transforms u and v one obtains the formulas -ial£+ia2n+kala2F tl = a2 a2 , (9.2.8) where F = F[u2 ], a2 = a + a2. We take into account the formulas F-' a2, - i _ 2a In ro - e, ro = !LA x2 + y2 (9.2.9) F-1 [F] 4", u2 * e 11 t d ri, tt where u2 * c is the convolution product and r = + yo, xo = x - C YO=Y-71- (9.2.10) From (9.2.8) it results 1 mx + ly u(x, y) = 21r x2 + 2 + k 8J ax xo rJ) r2 d (9.2.11) 1 My-ex v(x, y) = 2ir x2 + y2 u2 s + 2 ax fJasr *1) d d il . 371 THE PLANE FLOW Obviously, the last integrals are singular. They have the shape (E.3). Isolating the singular point (x, y) with a circle having the radius a and setting f = xo/r and respectively yo/r we deduce that the condition (E.5) is satisfied and the integrals are singular. Also, from the formula (E.10) we getr I = Ox / f z = +7) r0 d d rl = 2U2 JJ. J=a JIz (n)df dn, II2 (9.2.12) or, performing the calculations, =Jju2(,q)dxd,7_,ru2(x,y) 17y0r4x0J I (9.2.13) U2(t,11)x2 =-ff 9.2.2 2dOdx dn. The General Solution Replacing the profile by a perturbing distribution defined on the segment (--1, +1) (the chord of the profile), it results the following general representation of the perturbation: u(x, y) = 1 1 27r 40Y + m(Oxo P-1 O2 -}- y 2 k d t + 2n I , (9.2.14) 1 v(x, y) = 2a 1 +1 m(E)v - f(F)xn 1 0 - k - T2r Taking into account that xo + y2 = (xo - i y)(xo + i y), xo + yo = (xo - i yo)(xo t i yo) , for complex velocity w(z) = u(x, y) - i v(x, Y), 372 TUE STEADY TRANSONIC FLOW it results the following formula: W(y) = 1 +i 2k L(-), M(C) +i 2 (9.2.15) where ata JJ = u2( F''1)L(z) Z-( 1, and _=x+ir1, In order to impose the conditions (9.2.2) we shall pass to the limit on the segment [-1, +11. Using Plemelj's formulas we obtain from (9.2.15) u(x, f0) - i v(x, f0) = T 2 [rrt(x) + i e(x)] + (9.2.16) 1 + 'IrJr+l m(O + i e(f) xo k 0 If r u2(t, 71) d d +27r OxJf T x- do whence k 0 f/ 2 +2;r 8a R su xa 4 ,dl; r1` (9.2.17) ' +1 v(x, ±0) = ± m(x) - 2n 11 zn) d £- 49 -2 a_JJ 2d dn. It results e(x) = u(x, +0) - u(x, -0) = K (p(x, -0) - p(x, +0)1. (9.2.18) The function l(x) will determine therefore the lift. Imposing the conditions (9.2.2) we shall obtain m(x) = h+(x) - h' (x) twt h`(x), lxj < 1 (9.2.19) 373 THE PLANE FLOW ' +t w 8x fir ul{t,*)) J-t dd dn+ o (9.2.20) C(Z), IxI < 1, +h+(x) where we have utilized the notation h+(x) = h+(x) + h'-(x). The relation (9.2.19) determines the unknown m(z). Considering that G(x) is known, the equation (9.2.20) determines f with the aid of the formula (C.1.9) t(x) -_ 1 Vi- a l+z 7r 1-z 1 /;:E t h+(t) d t- l+t 8 +1 l+xJ_ n 1 -t t - x k d 1-t 8t )2+tp} dt t-x' (9.2.21) Noticing that according to the formulas (B.5.3) and (B.5.4) we have f+l 1-t dl: t-.z(1+ t-z dt 1 (9.2.22) -J 1-4 df ) a 1+4f-t t-z +1 1 VEZ+ 1 and, after replacing (9.2.21) in (9.2.15), it results W(Z) +t 1 C (9.2.23) z- 1 1 tai z+ +l 1 1+ t h+(t) 1-t dtk k t-z+2L(z)-wM(z), where we denoted M(z) -M z +1J 1 1 1+trd fr 1 n dCdr1 ( - t)2 +V2 z t d t Jf u ('*1) (9.2.24) dt --: the formula coinciding with (4.11) from [9.21]. TIIE STEADY TRANSONIC FLOW 374 In the sequel we shall deal with M(z). We may write M(') 7ri z + I f12 u2(f, q) ltd f+1 { , q 1-t8t (t-t)2+q2, dt t-z1dq 1 7n x 1 z+1 +1 ILl (F, q)' 2 The last integral is calculated by Homentcovschi by means of the residue theorem. It may be calculated elementary noticing that we have q 1 1 1 t-z - 2i((-z) t - (- (t-e)2+ rt 1 1 1 2i((-z) t-(+((-z)2+q2 t-z and taking (B.5.1) into account. It results _ M(z) 1 z+lIir z 1 u2((,ri)8 1 1 I(1_ + We write the first integral as follows If u'(t,n) I 1 dt drt_ 1 z C1z s±i u2(E,17)-u2(t,-17) C-z JJR2 r 1 _ and the second 8z 11.2 u2(t,q)( (1-z (1 z}d(dq 1 }dtdn, THE PLANE FLOW 375 = Jf/f u2( ((- z)2 JJR2 We have therefore +00 M(z) = 2 If [u2(.7,) - u2(C, -t1)] K(z,C)dC d tj-00 (9.2.25) 1 +00u2(t, n) -2rJ -00 u2(C, -n) ((-z)2 d d tj, where K(z' } _ z 1 z+1 (+1 1 _ C - 1 L(S - z)2 1 (9.2.26) (C - z)(C2 Taking also (9.2.13) and (9.2.14) into account, we have that y02 -x02 +2ix°y° 1 (C - z)2 , H L(z) = I - i J = -7ru2(x, y) - IL (9.2.27) ' 'ddrj. (9.2.28) (C - Z)2 We replace the expressions (9.2.25) and (9.2.27) by (9.2.23) and we obtain the complex velocity. Separating the real part of the complex velocity, we obtain the integral equation of the problem. We have: 'j, +1 u(x, y) = --fl 1 h 1 l 4tr l+th+(t) t x2 + y2 d z+l 1- z+l z 1t-z VE 1 i t-; jdt+ (9.2.29) +00 +2I(x,Y)-$ JJ q) [K(z,c) -00 +K(,)]d drj+ !T[u2(1 yor4x°dC dn This equation, was given in a slightly different form by Homentcovschi in [9.211. 376 THE STEADY TRANSONIC FLOW 9.2.3 The Lift Coefficient From (9.2.18) we deduce the following formula for the lift coefficient +1 1 c +I V=x 1 fKl-te(x)dy t+( )dt ' +t 1+ t r l+z .!_t dx- ;2/ k J +t v=1 x l+x i-t (9.2.30) We have J R3u2(frn)5 [(t-t)2+ n21dfdtl u2(f,8tn)t-( '( 1-t--( 1)dtdn 2i1 JJ .2 1 -2i J2 (9.2.31) u2(f.n)-u2(f,-n)df do (t-()2 From (B.5.1), derivating, it results: l+t t 1 dt 1 1-t (t-f)2 it 1 (9.2.32) f-1 (2_1 Using these results we obtain cp 7K 1 2i 9.2.4 j k I-t hf (t)d t(9.2.33) rJ u2(f+n) - u2(f, -n) d f do S-1 JJets f-1 . The Symmetric Wing If the wing is symmetric, then the equations (9.2.1) have the form y = ±h(x) (9.2.34) 377 THE PLANE FLOW From the boundary conditions (9.2.2) we have the relation v(x, +0) = -v(x, -0) which suggests that the solution of the system (9.2.6) has the property v(x, y) = -v(x, -y) . (9.2.35) We easily check that if u(x, y), v(x, y) is a solution of the system (9.2.6), then u(x, -y), -v(x, -y) also have this property. By virtue of u(x, y) = u(x, -y), the uniqueness theorem it results (9.2.35) whence u2(x, rj) = u2(t, The integral equation (9.2.29) receives the form +1 ff(t) -To d4+ (9.2.36) +2 Z if dn-u2(x,y), JJ00 given for the first time by Oswatitsch (9.491. Obviously, the lift coefficient vanishes. The result is natural because the angle of attack of the wing is zero. 9.2.5 The Solution in Real We shall present a solution which is different from Hamentcovschi's solution which utilizes the complex velocity. Using the formulas (3.1.19) and (3.1.20), one obtains from the general solution (9.2.14): u(x, ±0) = ±-t(x) + Zx I (ssd s + J v(x,f0) = f2m(x) - 2R ,r+t k I (X, ±0), T7r ds+ 2 J(x, ±0), (9.2.37) -1<s<1, where I(x, t0) .2 J(x, ±0) = 2JJu 2- Z 0° + n2)2 d t d it - iru2(x, f0) , dri, -00 < ,n < 00. (9.2.38) 378 THE STEADY TRANSONIC FLOW We obtain the formula (9.2.18), and from the boundary conditions (9.2.2) m(x) = h- (x) 1 e(s) JJ s-x 7r (9.2.39) d s = h+(x) - kJ(z, 10), w 1xI < 1. (9.2.40) By means of the formula (C.1.9) we deduce r+ £(z) - -V1:11 x l+x n + t h+(t)d 1-tt-x t+ (9.2.41) 1+ t J(t, f0) +as 1+z 1-i t-x 1 dt . For obtaining the integral equation of the problem, we shall replace m and t in (9.2.14)1. We shall denote by s the superposition variable from the first integrals (9.2.14) and xt = x - s. Changing the order of integration, we deduce 1+t 1&+(t)N(3, t)d t- +1 u(x, y) = 27ra J-1i y -27r y T2 , f 1 ± t J(t, ±0)N(x, y, t)d t+ (9.2.42) k +2R I+1x2+ 2ds+2-I(x,y), x1 r -- s, 1 I 1 where we denoted N(x, y, t) = f+i V1 -s 1+s J 1 Since 1 x I ds y2 s-t (9.2.43) 1 fl is 1J + s-t z-z t- s-z t-z s-x 1 1 1 1 1 1 (x-t)2+y2 s-t' taking the formulas (B.5.3) and (B.5.4) into account, we deduce N n x-z ( 1 t-I 'z-1+1 t 1 `+111. z -x (9.2.44) 379 THE PLANE FLOW Let us calculate nw the term T =1 i + J(t, f0)N(x, y, t)d t . t i (9.2.45) To this aim we shall specify J(t, f0). Taking into account (9.2.13) and the identity 2h7 1 t-C - t (t-w + 1 we deduce J(x, ±0) = - JJ rlLe2(F, 18 iT 3i 1 1+I 2i 1C-t C)dCdq= f _ and then I ldfdq= v))! IL ifm (9.2.46) u2(Cn)-UU(E,-11)ded+l= t-( u2(E,11) - U'(C, -WdC dq ( t-C)2 dil} (t- u2(e,-17)dt - C)2iy l +t VI u2(F,17) 1-t 1 , x . (9.2.47) s 1 l 1 1 t -z z+l t--l+1 cit. Changing the order of integration and denoting IPTE To(C, Z) = El N 1 1 t, t (t i()2 t zd (9.2.48) we deduce T If 4Y JJ*3 [ j2((, +1) - U2(t, -n)1 [To((z)J4_ (9.2.49) -To(CIT) FT 711 3+1 jddti 380 THE STEADY TRANSONIC FLOW for specifying the function To we notice that I t 1 (t-C)2 j---Z 1 +[Z 1 _ 1 (t ((Z 1 - ()2 1 1 1 z-( (z-S)2 t-C+(z-()2 Taking (B.5.1) into account , one obtains 1+1 - A To(C,z) _ (z - ()2 ,r - 1 - z+1 n (z - C)2 z 1 - 1 - 1+S d z-C dC C-1' (9.2.50) The term T which intervenes in the integral equation (9.2.42) has therefore the form (9.2.49) where To It is given by the formula (9.2.50). Using the expression (9.2.41) for the lift coefficient cL = JJt(x)dx (9.2.51) one obtains the formula 1 CL k +i ai -VrJ1 FI+- ft+(t)dt+7Vn f i+tJ(t,±0)dt. 1 (9.2.52) The function J(t, f0) is given in (9.2.46). For calculating the last term we change the order of integration and we take the formula (9.2.32) into account. One obtains the formula (9.2.33). 9.2.6 The Symmetric Wing We gave in (9.2.36) the equation of Oswatitsch for the symmetric wing. We obtained this equation from the general theory presented in (9.2.2). We present here the direct deduction based on the equation of the potential cpn + lpyy = lops which may be obtained from (9.1.13) or (9.2.6). (9.2.53) 381 THE PLANE FLOW The fundamental solution is defined by the equation Eu+EI,o, =k 49 E.3 +m6(x,y). (9.2.54) Applying the Fourier transform it results (a2+a4)E=iaikF-m, F- 9e ], whence ia1F al m aFI+ a and then -F rl -k 2 ax F f f ll - mF-1 [- ] . (9.2.55) 1 al +a2 -71+072 Applying the convolution theorem (A.6.13) and taking (A.7.11) into account, we obtain 8 F-l r. (_ ' [-'r a14J s 47r (+ +ao --4, f f E=(t,q)ln(xo+14)dCdn, -00 such that +*0 E = + 7r JJ -6. 2(4, n) ax f d f d n + 4n ln(x2 + y2). (9.2.56) -00 Replacing the wing with a continuous distribution of perturbation sources defined on the chord of the profile (only in the symmetric case the sources distribution is sufficient), from (9.2.56) it results, in the domain occupied by the fluid, the following general representation of the solution of the equation (9.2.53): 'P(z,y) = I fl m(f)ht{so+y2}dC + j; ffR?u2(t,n) X0 PO + y dt dn, (9.2.57) where in(k) is a function which has to be determined by the shape of the profile, i.e. by the boundary condition v(x, f0) = ±h'(x), x E [0, 1], (9.2.58) 382 THE STEADY TRANSONIC FLOW which is imposed by (9.2.34). From (9.2.57) we deduce 1+1 0 + y2 (9.2.59) f+1 v=SPY= 2rr xo y2 where I = ff po d 4' d ii u2 (4, n) x2 (9.2.60) . The singular integral I has the shape (E.9). Applying the formula (E.10) we deduce 1= fju2(f,rl)8x (x o+ f2fR2 x2 - u2(4, 11) Jdt;dn-aru2(x,y) o 2 / (xd { d *1- 7ru2(x, y) + y02)2 (9.2.61) , I=-2 ffzu2(t,,1)(xo Hence we have the following representation of the solution u(x, y) = 1 +1 /-1 xo z0 + y2 d 2 f fR2 u2 v, r+t v(x, y) = T !-1 - +y 2 rl) (x o z d Cd +1- k d t- fJ uz m(C) xo 2uz(x, y) , x2y2 n) (xo + y02)2 d dn (9.2.62) which is also given in (9.2.14). The boundary values of the first integrals are given in (3.1.19) and (3.1.20). We obtain therefore v(xi f 0) = f2rn(x) + 11 ir u2(f,rl) 2 (xz + 2)2dt d>l. o (9.2.63) 383 THE THREE-DIMENSIONAL FLOW Imposing the boundary conditions (9.2.58) we find +oo II u2V,77) 22'°'1d d drl> (ro+r1) (9.2.64) u2(x, y) = u2(x, -y) (9.2.65) whence we deduce 17) = u2(s, -+l) Taking into account (9.2.63) and the previous relation, it results m(x) = 2h'(x). (9.2.66) In this way we determine the distribution m. Coming back to (9.2.62), we obtain the equation: u(x, y) + k u2(x, y) + 2 11.2 2_ 2 (xUO+ y02)2 +i d d j? (9.2.67) rr -1 which coincides with (9.2.36). 9.3 9.3.1 The Three-Dimensional Flow The Fundamental Solution In the last 40 years, a great number of papers was devoted to the steady transonic flow past thin bodies. Usually one assumes that the flow is irrotational, the potential satisfying a non-linear equation having the form (9.1.13). For deducing the integral equations of the problem, we apply Green's formula to the equation of Poisson and we assume that a vortices layer is present downstream the wing. Derivating, we obtain the non-linear integral system for the components of the velocity (see for example [9.36]). In the case of the symmetric profiles the system reduces to a single equation for the component u(x, y, z). In this sense, after the initial paper of Oswatitsch [9.50] where one defines a principal value for the singular integral which intervenes in the representation, it followed the paper of Heaslet and Spreiter [9.17] where one gives a general representation which in the symmetric case reduces to an equation. The 384 THE STEADY TRANSONIC FLOW representation is valid both for the flow with shock waves and the flow without shock waves. For the lifting wings the forms of Norstrud [9.41] and Nixon [9.39] are available. In this case, the problem reduces to a system of two nonlinear integral equations. At last, we mention the paper of Ogana [9.471 where one shows how the integral equations depend on the definition given to the principal value of the singular integrals. A new point of view, belonging to D. Homentcovschi [9.20], [9.21] and L. Dragog [9.8] ]9.9] does not assume that the flow is potential. Utilizing the system of equations of motion it is necessary to assume the existence of the vortices layer downstream. In the sequel we shall utilize the method of fundamental solutions [9.8]. The system which determines the perturbation is (9.1.42) and (9.1.46). Performing the change of variables u`= , (9.3.1) and omitting the marks * and A, the system becomes uy-v==0 U,-w2=O (9.3.2) us + vy + w. = k(u2)s, where k has the same significance like in (9.2.6). We shall see further that employing a fundamental solution similar to the fundamental solution of the system uy-uz=eo(x,y,z), us-ws=0 (9.3.3) u. +vy+w2 = k(u2)=+mb(x,y,z) we may satisfy all the conditions of the problem. This solution will be determined in the manner described in 2.3. Applying the Fourier transform, solving the algebraic system just obtained and considering the inverse Fourier transform, on the basis of the formulas from appendix 385 THE THREE-DIMENSIONAL FLOW A, we obtain: u(x,y,z) )r 41r (m&x + t v(x,y,z) - 4w 8z k 4w 8y r 82 _ M 49 -4w 8 r, 8zsuz r 8 as 8z alai 1 4w 8xOy w(z, y, z) 82 4w 8 I_ m 8 1+ t _ __ us r, k 2 02 2 1 (r} ^18z [ate, - 4 8x8zu * r, 1 (9.3.4) where r = + y + z and, -. I 8, C) dv , (9.3.5) u r Iffits I z - 41 with the notation dv = d£ d, d(. This integral is called the aonvolntion of the functions u2 and 1/r. TaIdng M = 0, from the formulas (2.3.11), (2.2.6) and (2.3.27) it results us al l = -.f`1 8 1 J i alas [ala2J = (9.3.6) 4Ir 8y r j. 4w y2 + z2 11 +r and a similar formula. In fact one obtains the following form of the fundamental solution: u(2,y,z)_-4w (mf +e ) r 1 v(x, y, z) = 4w K jr02 w(z,y,z) = 8 r 8z [y2+z2 (i. + r Io, r- -m 1 4 t 8 + 4w 8a r), r 4w 8xjo' y +z (1 x) +w k 8K° 47r 8x, (9.3.7) 386 THE STEADY TRANSONIC FLOW where we denoted lo= au2*1, r Jo= ay TX au2#1, Ko= aZ r r (9.3.8) P. and m being constants. 9.3.2 The Study of the Singular Integrals The integral (9.3.5) has a weak (integrable) singularity. The integral exists, (it is convergent) (u2 is zero far away) and it may be derived (the convolution, if it exists may be derived (A.3.7)), such that we have I° = u2 * - -u2 * Ix13 = J u2(4) 14 - I3dv -r= and similar expressions for J0 and Iio. Since the integral has the form (E.3), it is convergent. With the notation E -- xl, x f __ IC has the form (E.9) and may be derived according to the formula (E.10). For calculating the last term one utilizes the spherical coordinates with the center in the point having the vector of position x: 1° -x=sin9c s rl - y=sinOsinV ( -z=cos9. One obtains Jfces(n,x)dn = Ir whence it follows the formula 8xlo = J ul()FZ (::,1) dv - 43 u2(x) _ (9.3.9) U2(t)2=02 0 "°dv - 4xu2(x) 3 . THE THREE-DIMENSIONAL FLOW 387 Analogously one demonstrates that J a to = 1, 49X u2(,) Ix?o lsdv, (9.3.10) K = ±Ko =JR3 u2(F)Ixxo{Isdv, where xo = x - t, yo = y- rt, zo =z-(. The integrals we have obtained are convergent if u2 satisfies Holder's condition and if its behaviour at infinity is u2(C) = O(ItI-`) with I > 1. 9.3.3 The General Solution Denoting by D the projection of the wing on the xOz plane and by y = h(x, z) ± hl(x, z), (x, z) E D (9.3.11) the equations of the wing (which is assumed to be thin), we shall be able to satisfy the conditions of the problem with a continuous superposition of solutions having the form (9.3.7), defined on D. It results the following general representation: (1R) u(x,y,z) =-4a JJ [M(C 0ax dt dC (9.3.12) -4 J(x,y,z), v(x,y,z) = 47r 11D [e(C) ax - (R) d d(- {2: 47rf (9.3.13) k ir- Ax' Y, z), u !(x, Y, Z) (R) + T" AD f (9.3.14) 1 D K I y +:.p Y dt d(, THE STEADY TRANSONIC FLOW zo=z-(, R- xo+y2+zo. (9.3.15) Taking the formulas (5.1.16), (5.1.18) and (5.1.24) into account, it results u(x,10,z) =±t(x,z)+4Ao,, m(E,() D v(x,±0, z) = f1m(z, z) + 4-r no dEdC- 4x1(x't0,Z) (9.3.16) 1. (z((,C)2 (i + x0)d{ d(- - 4 J(x, ±0, z) (9.3.17) where Ro= zo+zo (9.3.18) and the mark * indicates the Finite Part like in (5.1.24). From (9.1.42) and (9.3.16) we deduce the significance of the function t(x, z) : t(x, z) = p(x, -0, z) - p(x, +0, Z). (9.3.19) Hence, t(x, z) gives the jump of the pressure. This function will be utilized for calculating the aerodynamic action. From the expression of v(x, ±0, z) and from the boundary condition v(x, f0, z) = h'(x,z) ± h' (x, z) (x, z) E D, (9.3.20) where the mark "prime" indicates the derivative with respect to the x variable, it results after subtracting and adding m(x, z) = 2h' (x, z), 4ir,1Dt( (9.3.21) (1+)dd(+ + 2k s u2(4) (xo + q2 + Z02 T,,-, d v = h'(x, z), (x, z) E D. (9.3.22) 389 THE THREE-DIMENSIONAL FLOW The formula (9.3.21) determines directly the unknown m(x, z). In the equation for t(x, z) it intervenes the values of u2 in R3. They are obtained from (9.3.12) after replacing m by (9.3.21). We deduce u(x)- 3u2(x) + k 4I U2(4)2xix MI5 -o d v(9.3.23) -41T f t((,()R3d(dq 2,1 D f Dh'(k,()R3dCd( Hence, for determining the unknown f(x, z) on D we have to solve the system consisting of the equations (9.3.22) and (9.3.23) where u(x) is defined on R3. For u(x, f0, z) we shall utilize the values (9.3.16). The mathematical problem is extremely difficult and there are not known any attempts for solving it. For the symmetric wing (h = 0), the solution is obtained for t = 0 and u(x, y, z) = u(x, -y, z) if k u(x2)- u2(x) + 4 / 2x x05 d v = U 3 (9.3.24) 3 = 21r JDhi((,()Rd(d 9.3.4 Flows with Shock Waves In the case of the flow with shock waves, the general solution has also the form (9.3.12)-(9.3.14). We can see it in the simplest way if we utilise the notion of Fburier transform for bounded domains, introduced by D.Homentcovschi 19.191. Indeed, in the fluid domain D the equations vx-uy=0, wz-uz=0, (9.3.25) uz + vy + w: = k(u2). , with the notations from (9.3.2) have to be satisfied. On the shock waves E one imposes the relations OvOnz - OuOny = 0, llwOn: - 0, (9.3.26) uOnz + [lvjny + Uu'Qn: = kjJu21nz, THE STEADY TRANSONIC FLOW 390 deduced from (9.1.49), and on the borders S+ (upper surface) and S_ (lower surface), the conditions (9.3.20). Applying the Fourier transform for bounded domains, we shall utilize the formulas of the type (A.8.1). From (9.3.25) we deduce -iaiv+ia2u=S1+T1i -iaiw+ia3u=S2+T2, (9.3.27) -ialu - ia2v - ia3w+kialu2 = S3 +T3i where, taking into account that on S+ we have n = (0, 1, 0), and on S_, n = (0, -1, 0) (vnz - uny)e' a'xd a = S1 = s++s_ I Juleiaxda = -JDt(x,z)e'("'+"'s)da, _- D S2 = J S3= ; +s_ s++s_ (wns - un=)e' a'Zd a = 0, (9.3.28) f(urn+vny+wn;-ku2nr)e'c.'da= =f OvOe1a xda _ fom(x,z)e'(",+"':)da. t(x, z) and m(x, z) having the signification from (9.3.19) and (9.3.21). The integrals Ti = J (OvOn. - Julnw) T3 = JE T2 = f£ (OwOn.= (Duin. - Ovlny + Owonz - Juln=) et°r'ada - klu2Onz) vanish because of the relations (9.3.26). Hence, the system (9.3.27) reduces to -ia1v+ia2u=S1 -ialw+ia3u=0 -i&- ia2i;-ia3w= S3-kialu2, (9.3.29) 391 THE THREE-DIMENSIONAL FLOW which has the solution u= `ia2sas al Ss + i a2 S3 V= op, + 10 S1+ka t W= where a2 = ari + d u2, 3s u, (9.3.30) a S1 + ka y- i 1 + &23. Considering the inverse Fourier transform and utilizdng the formulas (A.6.9) we obtain 1 [83] - k-t-r-I [a,] (9.3.31) Cf2 f is a2] +8 -k.Ozox jr8a 1 us 8° - u2 1 i-k_ 8 air By direct calculations, we deduce .F-1 f (2703 OY J22 = t la {atE+wt)d dC L_ L f e '(a1x+a*v+o*z)d a da]dt ds fD[(211)3mnhJP.3 47s f a _ O R ()dd(. (9.3.32) where, with the notation R= V;i. , (9.3.33) THE STEADY TRANSONIC FLOW 392 we utilized the formula (A.7.10). From (2.3.11) and (9.3.27), it also results a 'r- j ( I_ 1 8 1- 1 iata2 t?z 1 47r 4ir Oz J oo V2.2 ao + y2 + z-2 _ zd x 1 f dx (x3 + y2 + z2)3/2 1 z 4-x y= + z2 (1 + x) r (9.3.34) 'On the basis of this formula we deduce _ 1 8 jI (27r)3 Vz st3 I e't(alx-+a,y+ass)da= P i0la2 OZ .3 (2n)3 1L 4;r i ala2 dct]d d( = 'y2+z2 (1+)ded(. (9.3.35) At last, taking into account the definition of the convolution product, it results u2 i u2(t) d (9.3.36) a2 4:r a3 Ix - f 1 With these formulas and with the similar ones it is not difficult to see that in (9.3.31) we have just the solution (9.3.12)-(9.3.14). 9.4 9.4.1 The Lifting Line Theory The Velocity Field The lifting line theory in the transonic flow is studied in (9.55) and [9.$). In the last reference, it is obtained, as it is natural to do, from the Lifting surface theory. This method is also utilized herein. We shall deduce the equations of the lifting line theory from the lifting surface equations using the assumptions 10,V,3" (Prandtl's 393 THE LIFTING LINE THEORY hypotheses) from 6.1. Hence we shall take hl = 0 and we shall consider that the unknown is the circulation C(c) = + (9.4.1) e(4, ()d C(±c)=0 (9 .4.2) and we shall utilize the formula tim = f ()k(x, y, z, 4, ()d e d ( = fJ s-(()-O-s+(() (9.4.3) }r. C(C)k(x, y, z, 0, <)d 2c representing the span of the wing on the direction of the Oz axis. On the basis of the first hypothesis (hi = 0), from (9.3.21) it results m = 0 and the representation u(x, y, z) = T" fJ C) R3 d d( T- I (x, y, z) v(x,y,z) =- 1 ID f t 4sr (1+ j)]ded(- - J(x,y,z), y (i+ az y2 + s02 R) ,did( - I(x,y,z) a ly + AD f 0 ev,oa[ 1 ffD w(x, y, z) = 4z (9.4.4) or, using also the formulas (9.4.1) and (9.4.3): u(x, y, z) = 1 4zr +` J k C(() R3 d _4 I (x, y, z) , 1 v(x, y, z) _ - c C(S) d I - 4?r -C C`(C) yz zo (1 + +C w(x,y,z)-:i!. f C(S) 2+x2 y o Ri) d( 4 J(x, y, z) , k (1+R )dC- K(x,y,z), 1 (9.4.5) THE STEADY TRANSONIC FLOW 394 where we denoted x2 +y2+; 2. R.1 = (9.4.6) The Integral Equations 9.4.2 Using the identity I(1+xo\^8 xo+Ro Ro J 8z (9.4.7) rozo where Ro is (9.3.18), we deduce for (9.3.22) JJn 1+'0 r(io RD / d d td (= 8z AD e(c o X0 0ze td C = FC(C)Zd -d Using the calculations (6.1.11) - (6.1.13) we deduce in the sequel T -j C'() d C az ff4;)s)flzOd t d ( = e +(1) t(co )d( 10 Je ' +` C(() d Lc zp L(() e a(z-S)d( (9.4.8) 2 , _+(_) t(F, z) d J=-(z) xo Substituting this in (9.3.22), multiplying the obtained equation by x - x_ (z) Vx+(z) - x and, integrating with respect to x on the interval (x- (z), x+ (z)), one obtains the following equation: C(z) = a(2) ' '((z d(+ is u2(C)S(=, F, n, ()d v+ H(z) , (9.4.9) THE LIF"TINC LINE THEORY 395 where x - x_(z) j_+(`) 3xogdx S(z, , , C) = flV x+(z) - x (xo +, 2 + 4)512 , (9.4.10) H(z) = 2 s+(_ - x-(Zh (x, z)dx . x 77 The equation (9.3.23) for the unknown C(z) is u(z) - 3 u2(x) + 4J [ y +` -^ 4 2X.2 U - x 2 y 2 zo is dv (9.4.11) C(C)d C , (x2 + y2 + 40)3/2 We have therefore the equations (9.4.9) and (9.4.11) for the unknowns C(z) and u(x, y, z), C being defined on [-c, +c], and it, on R3. The continuation of this reasoning may be found in (9.8). Obviously, we have to consider only numerical solutions. Chapter 10 The Unsteady Flow 10.1 10.1.1 The Oscillatory Profile in a Subsonic Stream The Statement of the Problem As we have already mentioned (see Chapter 2), the general problem of aerodynamics, i.e. the problem concerning the determination of the perturbation produced by an arbitrary moving body, in a fluid whose state is known, is very difficult. A presentation of this subject can be found in the papers of Kiissner [10.37], [10.38), [10.39]. We consider in this chapter, the particular case of the wing which is oscillating harmonically in an uniform stream. The problem is important in the flutter theory. For the incompressible fluid the plane problem was investigated by Theodorsen (10.75). For the compressible, subsonic fluid in a plane flow, the problem was studied at first by Possio [10.59) and then by Dietze [10.14) and Haskind 110.30). These authors used the potential of accelerations i1', replacing the body by a doublets distribution defined on the chord of the profile. The integral equation which was obtained made the object of many researches [10.22], [10.23]. In [1.1], [1.3], [1.8], [1.18] one may find syntheses and references. Considering that the replacement of the wing by a doublets distribution has no physical justification, in [10.15] we deduced the integral equation starting from the idea that the wing must be replaced according to Cauchy's principle by a distribution of forces. In the same paper, we gave a solution for the integral equation for small values of the frequency (w K 1). The present' section is written on the basis of this paper. 10.1.2 The Fundamental Solution We utilize the dimensionless variables x, y, z, t introduced by (2.1.1), and for the velocity field V 1 and pressure P1 resulting from the super- 398 THE UNSTEADY FLOW position of the perturbation over the basic flow, we set V1 =UO +V), Pi = P.+p,0U;pP. (10.1.1) Like in the previous sections Umi is the velocity of the unperturbed stream, and p,,. and pa, the pressure and the density in that stream. Imposing the condition that these fields verify the equations of motion, it results (see the equations (2.1.26) and (2.1.27)) the system oV /at + 9V/Ox + grad P = F, A2(aP/at + OP/ax) + div V = 0, (10.1.2) lim (V, P) = 0, ixl where F is the force density assumed to be small (the system is the result of a linearization), and M is Mach's number for the unperturbed stream. If the uniform flow of the fluid is perturbed by harmonic forces having the form f Cos (iwt), f Bill (iwt), applied in the origin of the axes of coordinates, then we shall put in (10.1.2) F= (10.1.3) f5(x)ek'Ji . Solving the system, we shall obtain that the real part of the solution is determined by the force f eos(iwt) , and the imaginary part by f Bin(iwt) . The perturbation produced by (10.1.3) will obviously have the shape V = v(x)ei"' P = P(x)e" (10.1.4) Replacing in 10.1.2 one obtains the system iwv + av/ax + grad p = f 5(x) M2(iwp + ap/ax) + div v = 0, , (10.1.5) lim (v, p) = 0. Ixi-. Q The solutions of this system are given in 2.4 (they are the solutions of the system (2.4.3)). 399 THE OSCILLATORY PROFILE IN A SUBSONIC STREAM In the case of the subsonic, two-dimensional flow, assuming that f has the form (0, f) , it results from (2.4.6) (10.1.6) P(x,y) _ -fly Go(x,y), and from (2.4.16) and (2.4.21) v(x, y) = f e-"' [2iwG - f32G= + w21 x G(7-, y) d 71 . (10.1.7) 00 We denoted Go (x, y) - 4I G (x, y) = 4if3 ) H ° ( k/ x2 + My2) e» H(2) (k (10.1.8) x2 + /32y 2) e1 where k=wM, 0 =w/$2, a=kM (10.1.9) Hoe) being Hankel's function. 10.1.3 The Integral Equation Assuming that the perturbation of the fluid is determined by an oscillatory wing having the equation y = h(x)e'`'t , 1x1:51, (10.1.10) from (2.1.29) it results the following boundary condition v(x, 0) = h(x) + iwh(x) = H(x), jx) < 1, (10.1.11) Let us replace the action of this wing against the fluid by the action of a continuous forces distribution having the form (0, f) defined on the interval (-1,+i] . From (10.1.6) and (10.1.7) it results the following general representation of the perturbation P(x, y) = - J +1 v(x,y) = f f(t)e 11 f (t) Go (xo, y) d r.,xo(2LG(xo,y) , (10.1.12) - f32GX(xo,y)+ (10.1.13) +w2J G(r,y)dr]d, THE UNSTEADY FLOW 400 where xo = x - f . The function f has to be determined from the condition (10.1.11). For obtaining the limit values p(x, 0) and v(x, 0) , we shall take into account that Hankel's functions H0(2)(u) and H12)(u) satisfy the relation (10.1.14) Hoe) (u) = (u) , du and have the following asymptotic behaviour for small values of the argument [1.16], [1.40]: H421(u) . I - 1n u , r = 1 - (10.1.15) ?l 2(C- 1n 2) , Hi2l() _i u , (10.1.16) where C(= In y) is Euler's constant (= 0.577215) . We deduce y P(x, y) = i::' f + p2y2) (k exp (iaxo) o + 2y d£ . When we calculate p(x, 0) we observe that, because of the presence of the factor y, the product y f+1 vanishes excepting the vicinity (x - e, x + t) where the integrand becomes infinite (for t = x ). If a is (iaxo) may be apsmall enough, in this vicinity the function f proximated by its value in the middle of the interval (we assume that f (C) is continuous), i.e. with f (x). It results rte , f (x) P(x, ±0) = y 1 Hill(k xo + [32y2) vlli 4i x+y _e -+ t : and, performing the substitution into account, dt +C AX, f0) _ f (x)Y--+O limy 2w J -x = t , and taking (10.1.16) t2 + /32y2 1 f (x). 2 (10.1.17) Hence we obtain the significance of the function f (x) . We have P(-T' + 0) - P(x, - 0) = f (x) . (10.1.18) The component v(x, y) is the sum of three terms vi , V2, V3, which are represented as follows +i v1 (x, y) = J f j f We- G(xo, y)d ""r° Hoe) (kx2 +/32y2)ei0xo d 401 THE OSCILLATORY PROFILE IN A SUBSONIC STREAM r+1 1 VI (X'0) = H(21 4iQ 1-1 f ( (kjxol)e0'0 dE, (10.1.19) the singularity from j 2) being integrable. We have also, v2 (s, Y)- J +1 f(e)e-I° 8 G(xo, II) d t = k 4iQ +i f 1 (k xq + Qom) x -07 0 °d t+ f+1 + 4Q3 fl f (k and taking (10.1.16) into account 1 k V2 (x, 0) Q J-1 f (klxol) f!l e i2-0 d E OHo (k1xo1)e f+' f(t)- Odd, [JZ0G(r,y)dr] 0 00 1 V3 (r,0) (10.1.20) I From one obtains , , (f)e-6-0 H(2) +403 f-+1 v3(x,U)= xo + Q2y2)e'D*° d dt, 1 f(t)e- '-O [JXOJ;42)(kfrI)ev1d1] dt, _!_ f where one performs the change of variable r -+ u : for = u and one takes into account the formula J J Hoe) ( M [u1) eh'd u = 2 n 1 Mf I , (10 . 1 . 21) given in [1.2[. One obtains V3(X,0) 26; If(4)e! 0 [! 1M6 + Zr o (M1u[)ej"duJd J (10.1.22) After all, from (10.1.13) it results I v(x,0) = T f(C)Ni(xo)dC+J 11 f(F2(xo)dt, (10.1.23) 402 THE UNSTEADY FLOW where we denoted ' Nk(xo) = nk(xo) e' nl ( xo ) = 4 pM k''X0 , k=1,2 (10.1.24) wr0 Hi2)(k (xoI) } o_ e n 2 ( Xo ) = A`-'A HO(2) (kl xol)e °''z0 fQC) 4 - 2W- In 1+,3 (10.1.25) M H2)(MIuI)eudu. Imposing the boundary condition 10.1.11 it results the following singular integral equation r I i f d H(x) , IxI < 1 (10.1.26) where N(xo) = N,(xo) + N2(xo) (10.1.27) In fact the kernel N depends on the variable xo and on the parameters M and w. We deduce therefore that the kernel has the following explicit expression (a = kM) N(xo, _iw M'w) e_ -- --Hi2)(klxol) Ixol etQTp + °r QHQ2)(klxol)e' O- e'i'' 0 In (10.1.28) The equation (10.1.26) is Possio's equation [10.591. There are a lot of papers devoted to the kernel (10.1.28) in the literature 110.5], [10.22), [10.301. 10.1.4 Considerations on the Kernel Taking (10.1.16) and (10.1.16) into account, we deduce N(xo, M, 0) = urn N(xo, M, w) = 2 n 403 THE OSCILLATORY PROFILE IN A SUBSONIC STREAM which is the kernel for the steady flow. With this kernel the integral equation (10.1.26) reduces to (3.1.15). The kernel for the incompressible fluid is obtained from (10.1.28) calculating (10.1.29) N(xo, 0, w) Mhm.(xo, M, w) . Integrating by parts we transform the formula (10.1.28) into et '° {Mi42)(kIxoI) N(xo. Al, w) = - iHi2)(k(xol) (X(, + 4A + t!ft e-6'0 lim Hoe) (Maul) 4 - W4M e-6'0- Utilizing now the relations (10.1.15) and (10.1.16) the notations (1.16) 00 J; Ci(z) =lnyz+ u-Idu =lnryz+-1)"( JJ Si(z) = 1sill u u n)t , n-1 °O du = > (-1)r ( n=O 2n+1 + 1)(2n + 1)! ' (10.1.30) called, the first, integral cosine, and the second integral sine, we obtain N(xo' 0, W) 2 2 + Si(wxo), } . (10.1.31) This is the kernel for the incompressible fluid. One demonstrates in (10.15] that for small values of the frequency (w a 1), the integral equation (10.1.26) has the form a xd w + mQJ_11f(t)(In (lx-t(+r))dt=2H(x), (10.1.32) where r is a constant. This kind of equations are solved in (A.16). We leave to the reader the task of writing explicitly the solution. In (10.22) one shows that the general kernel (10.1.28) has the form N(xo,Af,w) = Ao(xo)+Ai(xo,M,w)In(lxol)+A2(xo,M,w), (10.1.33) 404 THE UNSTEADY FLOW where Ao = _0 2, A,=- 2 CBI(xc,M,w)e ,mxo (10.1.34) A2 -iwxo 2. B2 (xo, M, w)e , A, and A2 being analytic functions with respect to x0. 10.2 10.2.1 The Oscillatory Surface in a Subsonic Stream The General Solution The problem presented in this subsection was studied in many papers (10.86], (10.87), (10.45], (10.35], (10.83] where the integral equation was obtained by means of the potential of accelerations, replacing the wing by a distribution of doublets. A slightly different investigation was given in (10.12]. We studied this problem in (10.6] utilizing the fundamental solutions method which will be presented in the sequel. The problem is the following-, an uniform stream having the velocity the pressure p and density po , is perturbed by a surface, oscillating according to one of the laws z = ho(x, y) coo (wt), z = ho sin (wt), (x, y) E D. (10.2.1) One requires to determine the perturbation. One utilizes the dimensionless variables introduced in (2.1.1) and the notations (10.1.1). The problem is simplified if we replace the laws (10.2.1) by z = ho(x, y)e"" , (x, y) E D . (10.2.2) In this case the real part of the solution will give the perturbation produced by (10.2.1) and the imaginary part the perturbation produced by (10.2.1b). The boundary condition (2.1.20) and the linearized system (10.1.2) lead to solutions having the form (10.1.4) where the functions v and p are determined by the system (10.1.5) and by the boundary conditions w(x, y, 0) = 8 ho(x, y) + iwho(x, y) = H(x, y), (x, Y) E D. (10.2.3) The solution of the system (10.1.5) under the assumptions that f = _ (0, 0, f) and the unperturbed stream is subsonic (M < 1) is obtained from (2.4.7) and (2.4.17) as follows a (10.2.4) P( x,Y,z _ - f 8z 0o(x,Y,z), 405 TILE OSCILLATORY SURFACE IN A SUBSONIC STREAM w(x, y, z) = f e i"x l -2 . ) G + (w2 [(2i ) f G(T, Yz) d r 00 J where 1 exp [io(x - M RI )l (10.2.5) R, x + RI(x,y,z) = (y T . As we already know, the formulas (10.2.4) define the perturbation produced in the uniform stream by the force (0, 0, f) exp(iwt) applied in the origin of the axes of coordinates. Replacing the wing with a continuous distribution of such forces, defined on the domain D, we obtain the following general representation of the perturbation P(x,y,z) _ w(x, y, z) = J J° f(f,rr) a8 G (10.2.6) G(xo, yo, z)+ f (t, -T) e-'"' [(2iw - 021) ox 0-2 (10.2.7) +(w2-a-y2) where, as usually, xo = x-t, yo = y-7) .The function f is the unknown. 10.2.2 The Integral Equation In order to determine the unknown f , we shall impose the conditions (10.2.3). At first we shall prove that if f (x, y) is a continuous function, then Zlim0JJ f(E,rr)a Go(xo,yo,z)dfdi' = r2f(x,y), (10.2.8) D Indeed, we have j f fl) iwM + Ro j )q aik(Mx°-RO) d e d >) rJ° a P(x, y, f0) = 4Z J z = 0, the integrand will be zero excepting the point Q(x, y) E D. Denoting by DE the disk where Ro = Rt (xo, yo, z) . We notice that if we /set having the center Q and the radius E and assuming that t is small 406 THE UNSTEADY FWW enough in order to approximate f t)) with f (x, y) (this is possible if f is continuous) and the exponential with the unity, it results P(x, y, ±0) 4lymo Jf 4 (ic?vf + )ddr . Performing the change of variables t, q -* r, 0: -x = Qrcos0, r/- y = rsin0, 0<r<e,0<0<27r, we deduce P(x, y, ±0) = 2p f (x, y) zluuo z i (iwM + r .8 z } r2+ r z2 = /! =fif(x,y), (10.2.9) and then P(x, y, +0) - p(x, Y. -0) = f (x, y) , (10.2.10) The formula (10.2.9) proves (10.2.8) and (10.2.10) gives the significance of the function f (x, y) . In (10.2.7) we may interchange the limit and the derivations (with respect to x and y). It results therefore w(x,y,0) = nj ' fj f 26 -,62 x (,)enw [n+n2]ddtl, (10.2.11) eR , (10.2.12) oo e(1iEl=d,, F00 E =j R, d dT ,02X = (xo R,. = =J T2 + MR), R = xo + /yo ya (10.2.13) 407 THE OSCILLATORY SURFACE IN A SUBSONIC STREAM In E1 we perform the substitution r A: r + AIRT = /32A , (10.2.14) Using the notation Iyot = r, we deduce dr dA 7,\T+:+ vrr2 -+#2r2 (10.2.15) 7, and then r El = 00 a-WA Mr/ A +r a wA +r 00 da = fo Mr lp dA A a-iWA A +r o r M lo = Ko(wr) - 2 [Io(wr) - Lo(wr)] - f dA dµ, el + µ2 (10.2.16) Ko and to being Bessel functions, and Lo the Struve function [1.16]. The expression of El may be derived taking into account that 02/0y2 = c72/c')r2 . Using the relations between Bessel's functions and their derivar tives, we deduce -w2 rtir/$ 1 + µze''wrs`dµ . 0 In E2 we make the notation iIyoj = u and then the change of variables r--# B . r=usinh9. (10.2.17) Observing that 82/8y2 = (3282/&u2 , we deduce 8 e&oRX =7 W2 1 /32 u2 + j92 T3 e x + iwM R2 e-x - yo R2 e wx K2- /'s0 (r - MR,)2efr-Mx.) dr J0 Rr where / 2X = ao - MX, R = xo + u2 , Rr = r2 -+U2. 408 THE UNSTEADY FLOW Performing inside the integral the change of variable T --- A:: r - MR. = Q2A , (10.2.18) we notice that the formula (10.2.15) remains valid because it depends only on A,12. One obtains the identity 1 I F2+ (j4 s0 {r - dT R yo Jo = (10.2.19) 1 z/Q' A2+yoe'"'xdA, Y6 -Mlyoi/$ Taking these results into account we obtain xoe ,x n(xo,yo)=ni+n2=-Q2 R +iw R-xoM R K1(wr) 2 w,x #2y02e - III (wr) - L1(wr)) +W2 -w2 jJ11/1) 1 + µ2ek`'r0dp f ar s A, A2 + r2e d (10.2.20) where, for the sake of simplicity we maintain the notation r = 1yol, and X and R are given in (10.2.13). Obviously the line yo = 0 is singular. Employing this form of the expression nl + n2 in (10.2.11) and imposing the boundary condition (10.2.3) we get the following integral equation 4a jf f n)N(xo, ?ro)d d r) = H(x, y) , (x, y) E D, (10.2.21) where N(xo, yo) = e-"n(xo, yo) (10.2.22) is the kernel. Since w 4z 1 we have the asymptotic expansions Ki(wr) = 1-, +O(w), Il(wr) = O(w), Ll(wr) = W 7r +O(w) (10.2.23) we deduce limn=-4(1+...), t/0 R (10.2.24) 409 THE OSCILLATORY SURFACE IN A SUBSONIC STREAM i.e. just the kernel from the steady case. In the case of the incompressible fluid one obtains the kernel setting M = 0 in (10.2.20). From (10.2.13) it results X=r0, R= xo+yo, (10.2.25) whence, n(xo,yo) =-XoeR +iw y-'e4'0 - iKI(wr)2 -2w (I1(wr) - LI (wr')) + Uf ;07 J 0 so ,\2 + r2eWX d A, (10.2.26) This is the kernel for the ineompressible fluid 10.2.3 Other Expressions of the Kernel Function Because of the Bessel functions, the expression (10.2.20) is considered to be complicated. However it does not contain divergent integrals. The idea concerning the introduction of these functions may be found in [10.86] and it was also taken into consideration in (10.17J. We obtain another expression of the kernel function, starting from the fundamental solution (2.4.13). Indeed, with this one, the component w(x, y, z) is written as follows w(x,?!,z) = a fj f (.,?1)n(xo,yo,z)d dn, where 2 ,2 n(xo,yo,z) = a f =o 000 R1T dr (10.2.27) ( 10.2.28 ) with the notation Rir = Vr2 + /32(y02 + z2) . (10.2.29) The passage to the limit and the derivation with respect to z do not interchange (only the passage to the limit and the derivations with respect to y interchange; for this reason we gave the expression (2.4.23)). Therefore we shall derivate at first with respect to z. The expression 410 THE UNSTEADY FLOW (10.2.28) may be found at Watldns [10.87], Williams [10.89], Dowell and others. Performing the change of variable r --t A : ,Q2A = T - MRir, (10.2.30) we deduce as we have already seen, dT dA ' A' +s' s2yp+z2. (10.2.31) Hence, E(xoyo, a) 110 eMR,) R11. 0o /X1 dr=J dA el" (10.2.32) A+ oo where 32X1 = xo - M xo + p2s2. (10.2.33) Derivating we obtain 8E _ _ 8z ek'X, Me R /X+ (10.2.34) iwX1 :-.0 8z2 dA I+s xo + E fim Ma ewa JA1 0 dA W (A2 + y02)3/2 where R V xo+/32Uo, A2X1 =xo--MR. (10.2.35) Imposing the boundary condition (10.2.3), we get the integral equation (10.2.21), where M ew'X1 R Xl + n(xo, yo) x, J-oo (A2 + g2)3/2 d A, (10.2.36) Obviously the line yo = 0 is singular. This expression of n(xa, yo) is simpler than the expression (10.2.20), but here the integral is no longer convergent. We must therefore consider the Finite Part. The expression (10.2.36) is utilized by Ueda and Dowell (10.80), Ando [10.3], etc. Utilizing the identity 1 X?+yo Mxo + R M xo+yo -x0 -X1, (10.2.37) 411 THE OSCILLATORY SURFACE IN A SUBSONIC STREAM one obtains two other expressions for n(xo, yo) . For the incompressible fluid (M = 0), we deduce eirra 1XO n(xo.ieo)=- J (,\2 + y02)3/2 To d a. (10.2-38) If the unperturbed flow has the sound velocity, we obtain the kernel considering M - 0. Since Al lim X = lim x° - At--1 Xo + yo 02 AM-1 = xo - yo _X1 (10.2.39) 2r it results xl, xo 2+ 2 e a nw N(xo, yo) = -e a ' ' + J oo (A2 + y 2)3/2 dA 1 . J (10.2.40) This is the kernel for the sonic flow. One may verify the similarity of the representations (10.2.20) and (10.2.40), if we employ the relation eiA X e" J1 = Ix j,\2 + yo)3/2 d J1 + Kl (wr) + r (10.2.41) +i 2 r ill (wr) - Li (wr)J which can be easily proved if we take into account the formulas 1 °° ( t 2 CD s 2r ) 3 / 2 d t = r KI wr) , (10.2.42) sill Wt Joy ir W (t2 + ,.2)s/2 d t = - 2 r [11 - Ll (fir)] given in the tables dedicated to Fourier transforms [1.16]. From the integral representation of the function Ko we deduce the identity rX ei'.3(r-MRjT) 1-,0 (t2 + r2)3/2 dT 2Ko(w yo + z2) jf oo R1,. d T' (10.2.43) with the notation (10.2.25). Putting this in 10.2.24) and deriving according to the formula 8z2 82 X32 8 s 8s ' (10.2.44) THE UNSTEADY FLOW 412 we obtain to the limit 2w -iWxo K t (wr) - /x (iWN1 + eo(r-rlRa.) Q2 dT Ix (10.2.45) where r2 + 012 Ror = (10.2.46) This is another expression of the kernel. The Structure of the Kernel 10.2.4 The identity (10.2.41) makes possible to deduce from (10.2.32) the structure of the kernel in the vicinity of the singular line yo = 0. Indeed for a small r we have (1.40] 1 K1(wr) = 1+ wr 2 In r + ... , Ii (wr) = wr + ... , Lt(wr) = 2 a (10.2.47) + ... , such that X eiWX sadA+iI - Jo (ACos Jo + yo) /2 J'00 (A + yo) 3/2dA= +72 - tr (A2 tn AdA+ 3/0)3/2 + 2 + 2 IAr + ... , (10.2.48) the points representing series of integer powers of r. If X > 0 the last two integrals have no singularities. Indeed, using the expansion formulas for LA d a .. . (A2 + yo2)v+1/2 given by Ueda (10.81], we deduce X °T (A 10 )3/2 d x sinwA J0 00 (A2 + r2)3/2 A =E ( (2n)1 2n n=0 00 (_1)nw"+i d A - nL-.o (2n + 1)! 12n+1, (10.2.49) 413 THE OSCILLATORY SURFACE IN A SUBSONIC STREAM where =1rX (A2 + r2)3/2 d A = Atm Im Xm-I 1))(,-3 X2 + r2 - (m - 1)(m - 3)J,.-4. (M X +* (10.2.50) Using the notations Jm= o 0 +r2dA=X--i(X2+r2)312-(m-1)r2Jm-2, Am/A2 (10.2.51) we may calculate the coefficients I,,, step by step, noticing that IX 1 r2=rs + 0= 11 = VTr+--r-l -r. The formulas (10.2.48) and (10.2.49) give the structure of the kernel. 10.2.5 The Sonic Flow If the velocity of the stream equals the sound velocity (U. = ca ), then one obtains the solution of the problem from (10.2.6), (10.2.7) and (10.2.12) setting M --# 1. We have shown in (10.2.27), that in this case there are perturbations only in the x > 0 zone. When we shall pass to limit we shall consider therefore x0 > 0. We have limp Go (xo, Spin, z) = exp 4 [iW Jim M2x0 _MR,1 z0exp-iw 2 +2xo = i - Go (10.2.52) Z2 X02 lime G(x0, y0, z) = 4 lim X = lim M-.1 M-.1 exp I -i. w J z0-MR = hml E2 = 110 exp T- u. 2( (xo 2 = Gl , - 14), )T (10.2.53) 414 THE UNSTEADY FLOW From (10.2.16) it obviously results lim E1 = 0 . Hence, the representation obtained from (10.2.6), (10.2.7) and (10.2.52), gives the perturbation in the sonic flow and (10.2.12) gives the kernel of the integral equation. This is s r nl+n3 = oexp 1 2 (x0 - zo 92 J L 10.2.6 iW 1 2 (T - y02 )1 dT . (10.2.54) The Plane Flow Acting like in 5.1 we obtain the formulas for the plane flow from the formulas which characterize the three - dimensional flow. We assume that in the equation of the perturbing surface (10.2.2), ho(x, y) has the form ho(x), i.e. it has the same form for every y. With a suitable choice of the reference length Lo, the domain D will be rectangular with -1 < x < 1, -b < y < b. In (10.2.6) and (10.2.7) we shall have f = f (r;) . Considering b - oo we obtain +00 go (X0, z) = J Go(xo, yo, z)d r1= 00 eik(Mxo-Ro) 1 =97r ED dr1= Ro e QHoe)(k Vxo+ 62z2)= 00 G(xo, yo, z)d r1= g(xo, z) = f xo + Q e(xo, z) = r-00 (E1 + E2)d r1= roo g(r, z)d r , +00 02 81/2G(xo, yo, z)d q = co +oo 02 J G(xo, yo, z)d r1= 0 . ao X7 (10.2.55) 22 (10.2.56) Using these results we obtain from (10.2.6) and (10.2.7) the representations (10.1.12) and (10.1.13). Taking into account the formulas of Fresnel 1 sin x2d x = fo 0 00 oos x2d x YY OSCILLATORY PROFILE IN A SUPERSONIC STREAM we deduce 415 r+ e-In"'du= Vjj(1-i), / . (10.2.57) o0 and then, with (10.2.54) +00 = it -oo (ni + rn)d q _ + i) Trw OX-0 iw exp(xo) - wi2 JL Zo iw exp(2 r) d7 r (10.2.58) This is the kernel of the two-dimensional sonic flow. 10.3 10.3.1 The Theory of the Oscillatory Profile in a Supersonic Stream The General Solution One considers that Carrick and Rubinow 110.241 have investigated for the first time this problem. They have utilized the method of the pulsating sources potential. The presentation which follows relies on the fundamental solutions method [10.17]. The integral equation is solved explicitly. One obtains finite formulas for the lift and moment coefficients. One utilizes the dimensionless variables (x, y) defined like in (2.1.1) and the representations (2.1.3) for the velocity and the pressure, in order to deduce that in (2.4.8), (2.4.18) and (2.4.21) the perturbation pressure and velocity determined by the action of an oscillatory force having the form felt, applied in the origin of the axes of coordinates, have the form (10.1.4), where P(x,y) = _(f V)Go(x,y), v(x, y) =f (10.3.1) 1e-k.% [G(r, y)d r] + (10.3.2) r] +f2e-k''x [2k4G + k2Cr + w2 f C(r, y)d r] Utilizing the notations W = w1k2 , v = wAf and H for the function of THE UNSTEADY FLOW 416 Heaviside, in (10.3.1) and (10.3.2) we have Go(x, y) = H(x - klyi)so(x, y), (10.3.3) G(x, y) = H(x - klyl)g(x, y), where a = vM 2kgo(x, y) = Jo (''- k2y2) e`ia" 2k9(x, y) = Jo (vVx2 - k ) (10.3.4) e-k:,x. As it is shown in (2.4.22) the perturbation (10.3.1) and (10.3.2) is zero outside Mach's dihedron with the edge on the Oz axis and the opening 21,t . This is the fundamental solution of the problem. Let us consider now that the uniform flow having the Mach number M, is perturbed by the presence of a profile whose equation is y = h(x)exp(k it) . (10.3.5) Taking the origin of the reference frame on the leading edge and the length of the chord as reference length L0 , the function h(x) will be defined on the interval [0,1] . Replacing the profile by a forces distribution (0, f)(t)e''t defined on [0,1] , one obtains the following general representation of the perturbation P(x,y) _-/ f o +1 i f (f )e iWxo L2iwG(xo, y)+ (10.3.6) -G(xo, Y) + W2 f G(r, y)d rJ d 00 . J Taking into account the definition of the function H(xo - kIWI) and the formulas (A.3.15), with the notation X = x - kIyI, (10.3.7) 417 OSCILLA'T'ORY PROFILE IN A SUPERSONIC STREAM we deduce jf()Go(xoiY)d t _ = H(X) y f t f (E)9o(xo, y)d = H(X) Jo, f W `""°G(xo, y)d t = H(X) f f 9o(xo, y)d , °g(xo, y)d t , o 0 f (xo, y)d = =f1 f f J= X f' f A [H(x) I f (t)9o(xo y)d (t)e-'.."`og(xo, ff(t)-iW8 ax (H(xo-kI yI)9(xa, y)ld y)d(xo - kjyj)d4 + H(X) f f (f )e-k"19. (xo, y)dk, o :o = H(X) f to9(r,y)dr. G(r,y)dr, kM 00 (10.3.8) Hence the solution (10.3.6) is P=O, u=0, ifX<0, (10.3.9) X P(x,y) = -Jo u(x,y) = fo f if X > 0, (e)e-iwxonl(xo,y)dt+ Zf(x)exp(ialy)), ifX > 0, (10.3.10) where xo n i (xo, y) = 2iwg(xo, y) + k29=(xo, y) + a'Z f 9(r, y)d r . The formulas (10.3.9) show that the perturbation produced by the profile propagates only in the interior of Mach's angle with the vertex in 0, and (10.3.10) that in a point M(x, y) from the interior of this angle one receives only the perturbation produced by the segment OMo (fig. 10.3.1). THE UNSTEADY FLOW 418 Fig. 10.3. 1. The Integral Equation and Its Solution 10.3.2 For 0 < x < 1 we deduce v(x, 0) p(x, +0) - p(x, -0) = f (z) (10.3.11) = if (x) + 2 I f (t)N(xo)d t, (10.3.12) where N(xo) = e-" = ym n1(xo, y) _ s k (Jo(vzu) + iMJ1(vxo)J e +ke J Jo(v-r)e rd z , 0 (10.3.13) Imposing the boundary condition (10.1.11) one obtains the following integral equation k f (x) + J f (t)N(xo)d t = 2H(z) , 0 < x < 1. (10.3.14) 0 This is a Volternz type integral equation of first order. We solve it using the Laplace transform. One knows (see for example (1.32)) that the Laplace transform of a certain function g(x) is the function g(p), defined by the operator £(g) = jg(x)e_Pxdz (10.3.15) 419 OSCILLATORY PROFILE IN A SUPERSONIC STREAM where p can be a complex number(p = pi + ip2) whose real part is positive. [JX] Applying the operator G in (10.3.14) we obtain kf + fo aPxdx= 2A(p) . Here we shall change the order of integration. In figure 10.3.2 we observe that the domain of integration is D (for a given x , C goes from 0 to x). But D can be also covered integrating at first with respect to x to oo and then with respect to . We have therefore from k7+ f (t;) [f°°e_P0tN(xo)dx]d C= 2R(p) . (10.3.16) XAK 0 F'ig. 10.3.2. Using the change of variable z -+ u : x - = u, we deduce from (10.3.16) (k+R)7=2f1. (10.3.17) In order to determine the transformation R we shall utilize the formula 1 (10.3.18) G(Ja(vx)1 = which may be found in the tables with Laplace transforms (1.161, [1.32], 11.331. So, using the notations Nl (x) = Jo(vx)e-i.x , N2(x) = Ji (vx)e ix" , (10.3.19) N3(x) = e-16'" I = J0(vr)e W-dr, 0 THE UNSTEADY FLOW 420 with a = vM , we obtain 00 C(Nl} = f Jo( vx )a- (p+")xd x = (p + ia} + v2 p + is + 00 £(N3) = (p+ ia)2 + v'j' v 1 G(N2) = 1 e-pie-"[f dxJ (p + ia) + v2v2 ' Jo(pT)e-a d r ] 0 = (10.3.20) °O fc* Jo(vr)e`+E [je ir-"'i"dxI dr r00 e-(P+",)udu = Jo(t r)e_u/M dT 0 0 1 1 (p+ivM) +v P+iw With the notations iwM ' Pt = M+1' UJAf Af-1' (10.3.21) it results 1 (p + ivM) + v = (P + P1)(P + p2) k(k+N) [(P+iw) (P+Pt)(P+P2), =k(P+Pi)(P+p2), A= +k[p+ivM-Vll»PI)(P+P2),+ P11 + )(P + P2) w2 1 + kP+iw} (10.3.22) and then OSCILLATORY PROFILE IN A SUPERSONIC STREAM not g _ k _ p+P2-P2+iw p+iw k + N (p + p1)(p + p2) M1)1/2(p+ 421 (p + P1)(p + p2) M+M1)-1/2_ (p+ M 1w+1) M -1/2 iw M - I (p iwM 1/2 (p+M+1) M whence kf = H+ g. (10.3.23) From the tables with Laplace transforms [10.581 it results g = L-1(g) and then with the convolution theorem x kf (x) = 2 f H (xo)g()d = 2H(x) - 21-w f H(xo) [Jo(ve) + iMJ1(v4)1 a `t d t. 0 (10.3.24) This is the solution of the integral equation (10.3.14). It was given in [10.171. 10.3.3 Formulas for the Lift and Moment Coefficients The lift and moment coefficients have the form CL = cLexp(iwt), CM = cmexp(iwt), (10.3.25) where, because of the formula (10.3.11), CL = -2 f 1 1 f (x)d x, cM = -2 fo x f (x)d x . (10.3.26) 0 We considered that the length of the chord Lo is the reference length and we defined CL = P ' CAt = (1/2)poU*20Lo (1/2)poU.2Lo ' P being the lift and M the moment on the direction Oz z. (10.3.27) 422 THE UNSTEADY FLOW Utilizing ((10.3.24) we find 4iv f1 kAl 0 CL = -4 CM = -411 xH(x)dx + J0 H(x)dx + 1 4iv e 'a`d (Jo(vl) - iMJI , f 1 G(t) (Jo(vt) - iMJ1(ve)J a '&(d 0 (10.3.28) where J 1 it x H(x - )dx H(x - t)d x, 1 E (10.3.29) . The coefficients (10.3.28) may be calculated numerically on a computer. Another method consists in approximating the function h(x) by polynomials whence one deduces that CL and cAf may be expressed by means of the terms having the form f" (M, a) =10 "Jo(vl;)e'°(d 1 (10.3.30) 9n(M,a)= frJi(z)e'd7 Taking into account that Jj(z) = -Jo(-) and integrating by parts one obtains that vg, _-Jo(v)eis+nfi_1-isf", n=1,2,..., (10.3.31) vgo = -Jo(v)e is + 1- lab. These formulas show that g" may be expressed in terms of f,,,. Integrating f by parts, we deduce laf" = -Jo(v)e-' + of"-1 - v ivM j ' f 1 0 F"Jlv )e-'f d , (10.3.32) _ -Ji(v)e+ of+ (10.3.33) +(n - 1)11 0 Substituting (10.3.33) in (10.3.32) we find for f" an expression which contains the last term from (10.3.33). This may be eliminated with the 423 OSCILLATORY PROFILE IN A SUPERSONIC STREAM aid of the relation (10.3.32) where n was replaced by n - 1. After all one obtains na + (n a Jo(v)e-i. - 1 Ji(P)e ia+ 1) (10.3.34) 1)2A-2 + 1(1 - 2n)fn-1 This formula shows that all the terms fn may be expressed by means of fo. This result was given for the first time by Schwartz in [10.70]. In the same paper one gives the following expansion for fo 00 [, fo=e n=O IMM1:Jn(a)+iJn.f.1(a)w". (10.3.35) 2"ni(2n + 11 In [10.701 one gives tables with the numerical values of fo, with eight exact decimals, for 1 < M < 10 and 0 < a < 5. In [10.33] one gives the numerical values of the functions fn for n = 0,... , 11. 10.3.4 The Flat Plate For the flat plate having the angle of attack -E (h = -ex) we deduce H = -2E (1 + iu;x) such that it results s CL = - E [ 2 f2 + iw(2 + iw) fi - (1 + 2iw - s , 2 2e CM = - )f of (10.3.36) z [f3+f2_2+2)f1 - 2(iw - 3 )fo] w2 These formulas are sufficient if we utilize the tables for , f1fo, f2' fs For w -+ 0 one obtains the well known formulas of Ackeret 4e 2E CL = k , cM= k . Obviously, cL and cm may be expressed only by means of fo if one utilizes (10.3.34). Noticing that f = fn + ifn , (10.3.37) where f1 f, = J 1 fn = -f eJo(4)sin(a)de, 424 THE UNSTEADY FLOW we deduce from (10.3.36) CL = c'L + icL , cAf = (.! + icM , (10.3.38) If the equation of the plate has the form y = -ex coswt = Re [-ex exp(iwt)J , (10.3.39) then CL = crL cos wt - CIL sin wt, (10.3.40) CAS = cos wt - ciAf sin wt, these formulas give the variation of the lift and moment coefficients versus the time. For example, for w = 7r and M = 2, we obtain CL = e(-9.2060 cos art + 11.8941 sin zrt) , (10.3.41) CM = e(-6.8779 cos in + 17.5209 sin irt) 10.3.5 , The Oscillatory Profile in the Sonic Flow We are interested in the behaviour of the formulas of Nl and N2 when M 1 (k -- 0). It results v -+ oo such that we shall utilize the well known asymptotic expressions Jo(z) = F2 Coo 7r (z - )+0(.-,), 4 Jl (z) = (10.3.42) r2z ooe (z - 34) + O(z-1) for great values of z. In this way, we deduce [Jo(vxo) + iMJI (vxo)J exp(-iaxo) NJ = _ k 1 =w xo [(1- iM) cos axo cos vxo + (M - i) sin axo sin vxo- -i(1 - iM) sin axo cos vxo + i(M - i) oos axo sin vxoJ , (10.3.43) 425 OSCILLATORY PROFILE IN A SUPERSONIC STREAM and analogously N2(x) = k _ 10" Jo(vre)e'du 1-i coe0(M-1)u+isinW(M- 1)ud u +I, 2fJo Mu (10.3.44) where, with the change of variable u -+ t : u = (M -1)t, we have _ cosw(M+1)u-isin'(M+1)u duI _ 2l+i two Mu (10.3.45) xo l+i M-1 IMexp(-iwt)dt 2 xrw ft M Jo Taking into account that we also have xo f M -1 gyp( Ld)dt slim Ja - Jo )d t = w (1 + it results that , No(xo) = J m1 N(xo) =(i+1) we r 1xoexp(2iwT O ) wi L YYY iw 2 Jexp(2 u) I du , (10.3.46) he. exactly (10.2.54). The integral equation (10.2.14) reduces to fo f (4)No(x - 4)d4 - 2H(x) (10.3.47) This is also a Volterra-type equation of first kind. 'Lbt integrating it we shall use again the Laplace transform. Applying this transformation we deduce (1 + i)% 7 = 2$b0, where § o'= (P+ 1w 2) 1f2 iwf + 2 (P+ iw 2 }` (10.3.48) 1/2 (10.3.49) 426 THE UNSTEADY FLOW From tables (see for example [10.581) we have that G p+iw/2,=exp(-)G[vii=-2exp (-) (10.3.50) such that we obtain go 2 Rx(i``' - x)exp ( i2 x) (10.3.51) and using the convolution theorem, from (10.3.48) we deduce After determining 1(x), the lift and moment coefficients result from (10.3.25) and (10.3.26). We shall give calculation formulas in 10.5.2 when we shall consider again this problem. 10.4 10.4.1 The Theory of the Oscillatory Wing in a Supersonic Stream The General Solution The theory of the oscillatory wing in a supersonic stream, was conceived according to the model of the theory in the subsonic stream. The papers of Kussner [10.37J, [10.38] represent the starting point of this theory. We mention then, the study of Garrick and Rubinow [10.25] where the potential of the pulsating source is determined, the paper of Miles where one considers the symmetric arrow - like wing, having the leading edges outside Mach's cone [10.53, the paper of Nelson for the triangular wing [10.57], etc. But the fundamental work in this domain is the paper of Watkins and Berman [10.85). Here one may find for the first time the integral equation of the problem and various forms of the kernel. The method is similar to the method from the subsonic case. From the potential of accelerations of a pulsating source, one obtains, deriving with respect to z the potential of accelerations of a pulsating doublet. The potential of the flow is obtained superposing the doublet potentials. The boundary condition gives the integral equation of the OSCILLATORY WING IN A SUPERSONIC STREAM 427 problem. In the following papers, due to Ashley, Windall and Landahl [10:4], Landahl [10.44), Stark [10.72], Harder and Rodden [10.29], Ueda and Dowell [10.81] the theory was developed and numerical methods for the integrations of the equation of Watkins and Berman were given. We shall indicate in this subsection how one may also solve this problem by means of the fundamental solutions method. Assuming that the equation of the wing is (10.2.2), we shall use distributions having the shape fe".,c = (0, 0, f)e' . (10.4.1) Utilizing (2.4.9)we deduce that the perturbation of the pressure determined by such a force applied in the generic point (t, n, 0) is given by the formula (10.4.2) p(x, y, z) = f azGo(xo, yo, z). For the component w, it results from (2.4.13) w(x, y, z) = - f e "" `0H(xo)b(1M)6(z)+ 82 +f :o 8z2a-""`O, co (10.4.3) G(r,yo,z)dr, and from (2.4.23) fe",,xa[(2iw+k28x) G(xo,yo,z)+ w(x,y,z) = 8z 02) + (10.4.4) IZO G(7-, yo, z) d r ] , 00 where we denoted Go(xo, yo, z) G(xo, 3fo, z) = 2A = H(xS s) H(xSo s) cos (LS)e-iaxo , (10.4.5) cos (vS)e-o" 1 k= M -1, 1=w/k2, v=OM, a=vM, s=k yo+x , S= xo-s , ST= 'r -3 , (10.4.6) H being the function of Heaviside. One may prove, taking into account the formulas (2.3.35) and (2.3.36), that the perturbation given by (10.4.2)-(10.4.5) vanishes in the exterior 428 THE UNSTEADY FLOW of Mach's cone with the vertex in the point ((, i, 0) and with the axis on the direction of the unperturbed stream (the Ox axis). Using a forces distribution having the form (10.4.1), applied on the domain D - the projection of the wing on the rOy plane, the perturbation will be given by the formulas p(x, y, z) =1 JD w(x, y, z) = -6(z) +21- f (C OF Go(xo, yo, z) d d rl, J JD f (t, i7)e-' °H(xo)b(yo)d (10.4.7) d q+ (10.4.8) f f f(e,q)e D w(x, y, z) = 2 . wh ere nl (xo, yo, z) _ IL f (t, q)e-'"" °n2(xo, yo, z)d t d i 82 f.0 z2 H(T - s) 0o co$ (vSr) e S r (10.4.9) , dr= (10.4.10) 82 _ z2 H(xo - s) J n2(xo, yo, z) = (2iw + k2 40) +[w2- \ 10.4.2 02 2 Cos (VSr) H(xS- s) r, cos (vS)e"'Ww'+ )H(xn-s) r a °sr)e-OrdT, Cos Sr (10.4.11) The Boundary Values of the Pressure They may be obtained writing p(x, y, z) = o - s) Cos (vS) a ; , d t d q, 27r C7z S (10.4.12) and noticing that because of the presence of the factor H (xo - a), the integrand differs from zero only in the domain DI defined by the inequality xo > s for a given M(x, y, z). This inequality is equivalent to (t -x)2-k2(q-y)2> k2`z2, 429 OSCILLATORY WING IN A SUPERSONIC STREAM which are solved in 8.3.3. Denoting by M'(x, y, z) the projection of the point M on the xOy plane and X = t - x, Y = rI - y we deduce that Dl is the foregoing branch of the hyperbola X2 - k2Y2 = k2z2 (fig. 8.3.4). When M' is in D, the hyperbola degenerates into the half-lines X = ±kY (fig. 8.3.5) Since the function f is defined only on D, we shall prolong it in the outer region taking it equal to zero. It follows that in the perturbed region from the fluid we have I f (C, n)Go(.To, yo, z) d d q= Qi l J ly- f() cos (vk (Y+-,)(,-(Y+ - 77)(9- Y_) (10.4.13) where Yf = y xok-2 With the change of variable q - 0: 2(Y++Y_)- 2(Y++Y_)cos0 = y- - z2 . (10.4.14) P- ok-2 - z2cos0, (10.4.15) we deduce 21rk JO cos a-iwco 110 f (C y - xpk-2 z2cos0) . (- z) 2sin 0d BJ d whence, if f (x, y) is a continuous function, P(x, y, ±0) _ _] az I = Tf(x, Y) (10.4.16) P(x, y, +0) - P(x, y, -0) = f (x, y) . (10.4.17) and then Hence, like in the previous sections, f represents the jump of the pressure on the wing. 430 THE UNSTEADY FLOW 10.4.3 The Boundary Values of the Velocity. The Integral Equation For z 0 0 the first term from (10.4.8) vanishes (5(z) = 0). It has to be considered in the same way in the limit values for z - ±0. The remaining term is the kernel given by Watkins and Berman [10.85). Elementary calculations give 0 az _ k2 z e 82 8s 22 8 k4z2 82 - k2s (1 _ k2z2 -s2-)T. + 2 8s2 . (10.4.18) In the cited papers one considers that the terms which contain the factor z2, vanish when z - 0. But this is not always true (see for exampie (3.1.20)). This is true when the factors which multiply z2 remain bounded when passing to the limit. In the following we shall see that for (10.4.10) the form obtained under this assumption is correct. Hence we shall consider the kernel Cos(1S')e 'dTJ 82 nI(xo, yo, z) °-` 8 {H(xo_s) (10.4.19) s The derivation is performed according to the formula (A.3.15), but we have to take care that for s = xo the integrand is unbounded. We eliminate this inconvenient writing (20Cos('1r2 -s')e '''rd7= T -8 is 0 e-0 - e+ /s T -s V7-r- S Js d T + CO° TO 1 e' d-r+ dr T -8 12 (10.4.20) After all k2 ni (xo, yo, z) = s H(xo - s)1, (10.4.21) s ) e "°TdT , (10 .4 .22 ) where 0 8s, =o Cos ( r r the derivation being possible if we utilize the equality (10.4.20), but we have no interest to do it. The integral may be calculated with the OSCILLATORY WING IN A SUPERSONIC STREAM 431 substitution r - A : r = scoshA. Deriving one obtains 1 - -rpcos(vxo-3 )e S O V xp zp -.%1 -r e'"''r s d dr [sin (v r2 - s2)] d r- esin(v r2 - s2)d r (10.4.23) . We integrate by parts in the second term from the right hand side. Passing to the limit in (10.4.21) we notice that, like in the steady case, it appears the singular line yo = 0. After eliminating from D the domain D, defined by the inequalities y - E < tj < y+E we shall put in the remaining domain z = 0. One obtains the following singular kernel It 1(ro,yo) = lim nt(xo,I/o,z) _ Yo + ff e-V"o sin (vS) + M Cos(,S)e_k-xo+ u) -H(x J 1X0 e iur sin (vSS)d r = = H (xo - u)n(xo, yo) , (10.4.24) where, u = kIyol and S= x0-u2, Sr = Jr2-u2. (10.4.25) For w = 0 one obtains (8.3.23). This will be the kernel of the integral equation. In the sequel we shall give a demonstration where the terms which contain tht factor :2 are not neglected. As we have already noticed, acting in the classical manner, we have to calculate the limits for z --+ f0 of some kernels which contain derivatives with respect to this variable (see Nlangler [10.52] for the subsonic steady flow, Heaslet and Loomax for the supersonic steady flow, Watkins, Runyan and Woolston [10.86} for the oscillatory subsonic flow, Watkins and Berman for the supersonic flow, etc.). Since generally, for performing this calculation we have to evaluate at first the derivatives, the passage to the limit becomes difficult. In order to avoid this, we gave other expressions to the component w 432 THE UNSTEADY FLOW ((2.3.29), (2.3.37), (2.4.23)). In the general solutions built on the basis of these expressions it appears only the derivatives with respect to y. The passage to the limit interchanges with these derivatives. In the actual case from (10.4.11) we obtain n2(xo, yo) = line n2(xo, yo, z) _ 2-.o _ (2iw + k2) H(x u) cos (vS)e-'"'%O+ (10.4.26) cos vS,. ecWrd T, 82 +(w2 - 8y2 )H(xo - u) S,) where u = k1yoI . Since we have 02/0y2 = k282/c9u2 , with the notation J_ cos(vST)e-,,tdT, 49 au (10.4.27) r u we deduce 1,2 -y2 H(xo - u)J = k2 82 H(xo - u)J = k2 0 H(xo - u)J (10.4.28) where J, calculated like 1 ,is J = - xo caos (ys)e-'ixo S u -! e MU sin (vS)(10.4.29) 2° W Mu j a w''r sin (vST)d r. For determining H(xo - u)J we take (2.3.35) into account. In this way, from (10.4.25) one obtains rigorously (10.4.24). If the equation of the oscillatory surface is z = h(x, y)e' , (x, y) E D then one imposes the boundary condition w(x, y, 0) = 8 h(x, y) + iwh(x, y) _- G(x, y), (x, y) E D (10.4.30) One obtains the following integral equation J 1n, f n)e`"n(xo, yo) d 4 d n = 21rG(x, y), (10.4.31) DI being the domain marked in figure 8.3.5 (the domain where xo > u). OSCILLATORY WING IN A SUPERSONIC STREAM 433 Other Expressions of the Ker e1 10.4.4 We have rro L =1 e-' sin (vST )d r = 2i (L_ - L+) . (10.4.32) u where we denoted Tn = (10.4.33) U In L+ we perform the substitutions r --- A:: T MS, = kiA. (10.4.34) Taking into account that T is positive in both cases we deduce kT=-uA+uM 1+A2, (10.4.35) such that LT- = tyo( f/ 1 1+A' e-'+Iyolad A (10.4.36) e-i:wlyola d A. (10.4.37) whence L = -- j (yo' 2i (=o+MS)/ku AtA z o-AjS)/ku 1+A -1 ) Since, on the other side, (x +MS)/ku eWroad A = ' 4yo jxo_MS)/ku sin (PS) . (10.4.38) from (10.4.24) we deduce Cos (VS) a-iUX0- iw n(Xo' yo) yo S 2Iyo) =o+MS)/ku L0MS)/kU A e-i&+lYOI-% dA. V1--+-A2 (10.4.39) This is the kernel given by Watkins [10.851. Obviously for w = 0 one obtains the steady kernel. 434 THE UNSTEADY FLOW Performing the change of variable A -> v : Jyp)a = v and and integrating by parts, we deduce ' _IYoEa (x04-MS) f ku a 2iyol =- 1+A .Y+ 1 y2+X+ 2y0 - d.1= iirJ JX* ye-n.rv +L dt - 2-y6' X- - e--iwx_ + TY0127X2 c- iwv 1# 1 e-'wx+ e -d v' 2 x_ {Up + v2)312 (10.4.40) where k2X,=xo± MS. (10.4.41) Observing now that 2e-0'0 coss(vS)=e-+e-"'x+ (10.4.42) utilizing (10.4.40) and the identities xo yQ + .X+ - SX+ = xo yo + X? + SX_ = Myc' (10.4.43) , we obtain for n(x0, yo) the following form given by Harder and Rodden [10.20) 2YL(xo, IM) a-iwx.. e--W,x} M e-Iwv . V7=` 0+ 1j,x +2 (YO-1 +172)3/2 (10.4.44) Another form of the kernel is obtained if one utilizes the identities _ yO+X2 Mxo:F S xp+yo _ Al xe+X (10.4.45) One obtains the relation L12 2n.(ro,rlo) S a-ivx« a swx o+X_ + xo+X+) X.. a-iwv {3lo+tr2)3/2dv, (10.4.46) utilized by Ueda and Dowell [10.811 for obtaining the numerical solution of the integral equation. OSCILLATORY WINC IN A SUPERSONIC STREAM 435 One obtains the sonic limit at once from (10.4.44), or (10.4.46) notic- ing that lim A!-l X_=-12 (xOlxo/ -X, limN--1X=00. (10.4.47) One obtains the following kernel H(ro) 2 2 12g + yo (xu, Ilo) +e iWx la + J/ oo e'"" (Jp + t')3/2d vJ (10.4.48) which coincides with (10.4.26). A New Form 10.4.5 We utilize the formuhas (see for example 11.30], pp. 406, 422) with real parameter O° cos, cos "i (pr)dT = ko(u v2 -12), (10.4.49) °° cos (vSr ) i-Sr sin (pr)d r = 0. For p = 0 we obtain the identity r° ST)e-wTd r. _ Ko(wlyol) T u 00 Cos(VST)e - f.0 rrrdT, (10.4.50) T as follows 2 fi Cw2 - k2 0' f J rn Cos (VS_) u a lord r = r 1 Iw--k2ou2/ [Ko(-ku)- f T7r .J=f T J In the last part we derive without any difficulty. Deriving, the kernel (10.4.25) becomes n2(xo, yo) = H(xo - u)n(xo, yo) n(a'o,yo) _ r - wM kkt(Au) u T J=° . Cos(yST)e_Ord -k.2 To ST sin(vST)e-` S2 r, TdT- (10.4.51) 'TIE UNSTEADY FLOW 436 where u = kJy I . This is the new form of the kernel. Having in view the behaviour of lit , for small values of the argument, this is 1 - 22 (I,, Iyul + r1) , r, = in 2 + -y 2 (10.4.52) , yo -y being Euler's constant. An additive constant r2 also appears from the two integrals (10.4.51). The kernel of the integral equation in the case Al = 1 is obtained from (10.4.51). We have illl A2 I\1u) = 1yu1Rt(wIyoI) Denoting rao It = 2iJ Sr - 7)]d T ao I2-2i f exp(-i (MSr+r)1S 1117- we have. °O sin (yS1) xp r 2 _rd T = It + I2. But It _ 1 exp 2l - 1 21 nhinl It = f.0 [1'A12S2 - r2 Al ST + r r2 exp iw AI (IT S; = dr Aft 2 r2 - k2y + T T2 , - k2y0 li 10"exp Pww (r - TA) ] T L. One obtains after all (10.6.13). 10.4.6 The Plane Problem such that in the repIn this case, the density f q) becomes f resentation (10.4.7)-(10.4.9) we can calculate the integral with respect 437 OSCILLATORY WING IN A SUPERSONIC STREAM to r l. We have 90 = f 00 Co(xo,Uo,z)d9= +oo 1 J = 27 H(xo - k r2 + z2) cos[v xo-k (r +x - k (r + z) zo )J dr. (10.4.53) Because of the presence of the function H, the integrand differs from zero only for xo > k r + z . This inequality implies zo > 0 and k2r2 < x2 k2z2, whence xo > kizl and -c < kr < c, where c = xo - k z2 . After all - oos{v 2-e'H(xo-klzl)f 90 - C -k r dr. r (10.4.54) Utilizing the formula 11.16] ` Cos (p c- x-x) (10.4.55) gxdx= 2Ja(cjp2+g2), it results go = H(xo - klzl)Jo(v xo - k2z2 (10.4.56) and analogously 9 Gd q = gH(xo - klzl)Jo(v roo +00 e = J Edrt = ao 1 jr-a° H(r - kjzj)Jo(v _ H(xo - kjzj) 2k k2z2)e' xo Jo(v 0, (10.4.57) r2 - k2z2)e''O'd r = r2 - k2z2)e rdr. fk1Z1 (10.4.58) For obtaining the results from 10.3 we have to consider the chord of the profile on the Ox axis (0 < x < Lo) and to take Lo as reference length. Observing that j +oc &2 00 OY 2E(xo,yo,z)dq=-J +00 02 8 Edt =-AEI 00=0, 438 THE UNSTEADY FLOW we deduce 1 p(x,z) = - I f(e)e sodt, w(x, z) = f (10.4.59) 1 f we-k""° [(2iw+k2)g+,2c]d which is exactly the solution (10.3.6). We obtain too +00 n(xo) = J n2(xo, yo)d tl = +m G(xo, = 2iw yo, 0) d >1 + k2 J00 +w2J a TX +x G(xo, y , 0)d n+ 100 G(r,yo,0)dil = 2a(xo)Jo(z'xo)eui""`0+ 00 +H(xo)no(xo), (10.4.60) where no(xo) = T- T [Jo(vxo) + 121 Jt(aod e1 .lo(vr)ed T , o (10.4.61) i.e. (10.3.13). 10.5 10.5.1 The Oscillatory Profile in a Sonic Stream The General Solution. The Integral Equation We proved in (10.2.54) that there exists the limit of the subsonic solution for M / 1, and in (10.2.45) that there exists the limit of the supersonic solution for M \ 1, and in addition, the two limits coincide. We shall prove now that there exists also the solution for M = 1, and this one coincides with the two limits. It will result therefore that the flow is continuous to the passage past the sonic barrier, unlike the case of the steady flow. We shall consider therefore the oscillatory profile (fig. 10.5.1) of the equation y = h(x)exp(iwt), (10.5.1) 439 OSCILLATORY PROFILE IN A SONIC STREAM Yt 0x I Fig. 10.5.1. which perturbs the uniform flow which has the velocity Ua, = cm (M = 1). With the notations (10.1.4), the boundary condition (2.1.27) gives v(z, t0) = h'(x) + iwh(z), 0 < x < 1. (10.5.2) The fundamental solution in the two-dimensional sonic flow is given by the formulas (2.4.32)-(2.4.34). If the profile is reduced to the skeleton like in figure 10.5.1 it is sufficient to replace it by a forces distribution having the form (0, f)exp(iwt). (10.5.3) It results therefore P(x, y) = --f a Go, (10.5.4) v(x, y) = f ei`''°` [(21&G + w2 T G(r, y)d rj o where r Go(x, U) = H(x) exp I _ !(X +s )J E H(z)So(x, p) , (10.5.5) 1 G(a, Y) = H(x) -OM [(x - )} H(x)9(z, v) , with the notation 2f/ = 1. A continuous superposition of forces having the shape (10.5.3) on the segment [0, 11, will give the perturbation i P(X, U) I v(x, p) = J0 1 f (() gpGo(xo, y)df , f (C)e^'"'° [(2iWG(Xo, y) + w2 Jzo 0 G(r, y)d rJ dl; 440 THE UNSTEADY FLOW Taking the significance of the function H(x) into account. it results that for x < 0 we have P=O, v = O, (10.5.7) and for x > 0 f 2fV)ryyo(ro.y)d P(x,y) To fr v(x, t!) = (10.5.8) (2iw9(xo y) + w2 f 9(T, y)d r d ti fWe-ten I . 0 (10.5.9) This is the general solution of the problem. It was given in (10.18. Performing the change of variable F -« u : u = y'/xo one obtains °G iw1Z Jy2 P(x. ,) = f (x - y-' )exp ;r. [_!f + u)] ll Ou- and then P(x, ±O) _ T-iwfl f (x) f exp -'2 It ` = +,l f (x ) (10.5.10) It results therefore p(x, -0) - p(x, +0) = f (x) (10.5.11) . For 0 < x < 1 we deduce v(x,±0) = jj f(Z;)e-'"`°n(xo)dy. (10.5.12) ro 27 [ exp(2 xo) - 2 ( } J exp 'T) 1 . (10.5.13) Imposing the boundary condition (10.5.2) it results the following integral equation jf()n(xo)d A(x) (10.5.14) where A(x) = h'(r) + iwh(x). The kernel (10.5.13) coincides with (10.2.54) and (10.3.46). 441 OSCILLATORY PROFILE IN A SONIC STREAM 10.5.2 Some Formulas for the Lift and Moment Coefficients Taking into iweount that 1-i = I D- 2 (10.5.15) 4 fa-w the solution (10.3.52) may be written as follows 3(x) = 2iwft fA )exp(- Z l;)d 10.5.16) -2fl J e 12 ex p(- 2 t)u , the sign * indicating the Finite Part (Appendix D). Denoting f F = Ir A(xo)eXp(- +t)dt A(,) 2 xo o P(_!2xo)df , (10.5.17) with the definition formula (D.4.2), we deduce 1z o 5.7r [A(') l so gyp( _ iw 2 ] xo)d= dx ' (10.5.18) and then -J exp( - 2 )d t - iw F+2df . (10. 5 . 19) After all the Jforinula (10.5.16) becomes 1(x) = ti (ii' + dx (10.5.20) The lift and moment coefficients are given by the formulas (10.3.29) with (10.3.30). Utilizing (10.5.20) we find cL - SS2 J 1 A(1 - 0 + WB(t) 0 P(- 2 (10.5.21) c,tt = 80 ri A{1- ) + iwD(>;) 2 )d where we denoted I 1 B(4) _ I A(xo)dx, D(t) = J xA(xo)dx. (10.5.22) 442 THE UNSTEADY FLOW Approximating the function h(x) by polynomials, we deduce that the integrals from (10.5.21) have the form 1 f( 2exp(-2 )d , it = 1, 2, ... (10.5.23) Integrating by parts we obtain -.2 exp(-12)+ 2n 1W n=1,2,... iW (10.5.24) It results that all the integrals from (10.5.21) may be expressed as functions of Io = J exp(- ) Vq [c(/) - iS( =2 11 7r )] , (10.5.25) where C(x) and S(x) are the integrals of Fresnel [1.30): C(x) = fS(x) = 1 2. j (10.5.26) - the notation z = irx2/2. In the case of the flat plate having the angle of attack c (h = -E.r. ) one obtains CG = -4f k(1 + iw) CM [2exP(_) +(I+ iW)I01 =-aQEK-iw+3tw+e P(12)+(2i 4)i]. +1 (10.5.27) 10.6 The Three-Dimensional Sonic Flow 10.6.1 The General Solution In the three-dimensional sonic flow the fundamental solution is (2.4.36) and (2.4.37). A force having the shape (0, 0, f)exp(iwt) , applied in the origin, will produce the perturbation P = H(x)p, W = H(x)w, (10.6.1) where P(x,y) _ -f8 go, w(x, y) = f e'w,x [2iw9 + (w2 - z) / JJJ (10.6.2) g(r, y, z)d Tl J TIIE THREE-DIMENSIONAL SONIC FLOW 443 with the notations 1 go (x, y, z) = , zy)_ 9(x, 4;rrexp 1 .rr (- 2r.' 2 e`priw IL 2 (10.6.3) y2+z z- x The perturbation produced by a distribution of forces having the form (0, 0, f (s, r1))cxp(iwt) , defined on the domain D (the projection of the wing on the xOy plane), will be characterized by the formulas P(x,y,z) w(x, y, z) = 8 J JD f (E, r?}e-"'"`" 1(2iwg(xo, +(wl - ay ,)J 10.6.2 yo, z)+ (10.6.4) 9(r,yo,z)drJdedr. The Integral Equation Assuming that D is such that every parallel to the span (the Oy axis) intersects the boundary OD in at most two points and denoting by y_ and y+ the ordinates of these points (fig. 10.6.1), we obtain 110 With the change of variables x- s1=y+l:I v; f V}}"'d'1 dC (10.6.5) r1) -p u, v : zJdudv (10.6.6) THE UNSTEADY FLOW 444 Fig. 10.6.1. we deduce P(x,y,±O) = 47r iw lim x lim fp tE) f(c 17) + yo + ~ zo 2 / d dn 22 Z 47r :-.to 2 (X0 I xo 7- (F) exp f (x - -, y + I= Iv) I U (f)-SI/I=I exp iw z2 2 u +u(v2+1)]}dudv f / = exp [ - Z uv2)dv] du. Utilizing the integrals of Fresnel 00 I cosx2dx =2 f 0 20 Cost dt= °sinr2dx = 1 f s'n dt = 2 2 (10.6.7) we deduce P(x,y,f0)_T2f(x,y) (10.6.8) Hence P(-T, y, -0) - P(x, y, +0) = f (x, y) . (10.6.9) 445 THE THREE-DIMENSIONAL SONIC FLOW Assuming that the equation of the oscillatory surface is z = h(x, y)exp(iwt), (10.6.10) we obtain the boundary condition w(x, y, ±0) = h'(x, y) + iwh(x, y) _- A(x, y) , (x, y) E D. (10.6.11) It results the following integral equation 2n lID f( t, ri)N(xo, j) d t d 0 = A(x, y), (x, y) E D , (10.6.12) where N(xo, yo) = amp [iw( 2x o y1 + (10.6.13) \W2 1 02 'O=p iw yl dr 2 We have also obtained this kernel from the kernel corresponding to the subsonic flow (10.2.58). On this formula we cannot observe yet the singular part of N We shall calculate therefore the last term. Using the formulas I 0 zo J exp 2 Ocein coo (r - (ax- z )dx _ {2Ko(2v) \l LT (10.6.14) 2K0(11p0D)-1°O exp f2 (r }] dr The last integral may be derived with respect to y interchanging the derivation with the integration. One obtains b Texp ( 2r0)8'r' lrexp ( "4 exp 2 In this way we deduce the final form of the kernel (.r N(xo,yo)=-1 ! K1(wlllol)+ 2 _)1]J 7d 1, 2 (10.6.15) 446 THE UNSTEADY FLOW The principal part is in the first term. Taking into account that for small values of the argument we have +zlnz+..., K,(z)= (10.6.16) it results N(xo,yo)=----w2(lnlyol+r)+.... (10.6.17) A The singularity has therefore the same order like in all the other spatial problems. 10.6.3 The Plane Problem We remind that one obtains the solution of the plane problem if we assume that D is a rectangle having the dimensions Lo and bL0 and we consider b --+ oo. Moreover we assume that every section with a plane parallel to xOz determines the same profile, hence in (10.6.10) h depends only on x. Taking Lo as a reference length, the domain D will be defined by 0 < x < 1, -b < y < b. Considering b oo we have to obtain +00 n(xo) _ j N(xo,yo)dri, (10.6.18) o0 where n(ro) must be (10.5.13), and N(xn, yo) (10.5.15). Indeed, utilizing the representation 11.30] !OD J (t2 C} ``'dt, (10.6.19) -2)3/2 we deduce rc+Oht(_u)du=2 w)+'Kt(wlyol)dyl_2w 00 o lyol u 0 =2'(r0°coswt-1+Idt-7rw Jo t2 zrw+2J o t J" coswtdt t. +2J, t- 2 16 f00dt + 10 7rw 2 = 447 THE THREE-DIMENSIONAL. SONIC FLOW Utilizing Fresnel's fortnulas (10.6.7) and integrating by parts, we obtain exp T2 011171d" T =21 cxp(i"T) [f (1-i) 2 7.ro= Pxp( ` 2 IIIJII -10 4: u2JduJ tw expTol+ T)dr + iw J exp (t2 =27ri+ 27- vfT - tw I exp (t2 T) vfT-1 (1-i)[ 2 exp(t-xo)-iwJrexp 2TdTr 7x=o `\'2 o We deduce therefore -'.w [ ri(x0)=(1+i 1 yrr-o- x p o iw 2x o1 2 ex p iw ) d T 1 of 2T sfJ ' (10.6.20) i.e. just (10.5.13). 10.6.4 Other Forms of the Kernel From (2.4.13) and (2.4.35) it results the following representation of the component w of the velocity w(X , , z) = 5(`) JJ f(. rt)H(xo)6(yo) d d i+ t f .{.i- (10.6.21) n Jf where 2 11(x0. yo. a'2 J exp 2 (Ir _ yQ+z2/J T LT (10.6.22) One may demonstrate that this integral is convergent. Denoting r = iyol and performing the change of variable r --+ A: 7. T - - = 2A dT -r dA (10.6.23) 77 7 448 THE UNSTEADY FLOW one obtains n(xo,yo,z) = 8 J-. d.1, +r (10.6.24) where X=2(TO-xo). (10.6.25) Observing that 02 _1 1_z29+za -2 r Z2 ( r2) th r2 5r2 we deduce that the line r = 0 is singular. Eliminating a vicinity of this line we obtain lim 82 _ 10 r Or O 0z2 whence n(xo, yo, z) =:-to lim n(xo, ?b, z) _ 2 fA (1u.o.lo) eWA 00pt2+r2)3/2da. This kernel was obtained in (10.2.36) as a limit of the subsonic kernel and in (10.4.48) as a limit of the supersonic kernel (Ueda and Dowell [10.80]). This fact proves that the oscillatory perturbation is continuous to the passage of the sonic barrier. f0, the first term from (10.6.17) Passing to the limit, when z vanishes because 6(f0) = 0. Chapter 11 The Theory of Slender Bodies 11.1 11.1.1 The Linear Equations and Their Fundamental Solutions The Boundary Condition. The Linear Equations In this chapter we study the aerodynamics in the presence of slender bodies (fig. 11.1.1). The axis of the body is considered the Ox axis and the Oz axis lies in the plane determined by the velocity of the unperturbed stream V. and by the axis of the body. We denote by a the angle of attack of the stream and we assume that a = e, where e characterizes the thickness of the body. We employ the cylindrical coordinates x, r and 0 which are related to the cartesian coordinates x, y, z by the formulas x=x,y=rcosO,z=rsinO (11.1.1) xER,rE (0,oo),0E (0,21r). Fig. 11.1.1. The equation of the body has the form r = h(x,0) = eh(x,0). (11.1.2) 450 THE THI3OKY OF SIZNDER BODIES Denoting by i, j, k the versors of the Ox, Oy and Oz axes and i,., ie (fig. 11.1.2), we shall have the following formulas i,. = jcos0+ksin8 j = ircos9--ipsin0 (11.1.3) ie=--jsind+kcos8 k=i,.sin8+iscosO. Fig. 11.1.2. The velocity of the unperturbed stream is V,, - U,,,(i cosa + ksina) = UU(i + ak) + 0(a2). (11.1.4) Denoting by V1 = Uvv, P1 = Poc + P-U«,P, P1 = Poo(1 + P) (11.1.5) the fields which characterize the perturbed flow and using the cylindrical coordinates V=Ui+Vi,.+Wi$ (11.1.6) v=Ui+I ,. +Wie, we deduce U=1+u, V =asin8+v, W =acoa8+w. (11.1.7) THE LINEAR EQUATIONS AND THEIR FUNDAMENTAL SOLUTIONS 451 Obviously, the perturbed flow will be steady because the conditions which determines it do not vary time. On the boundary we shall impose the condition V grad F = 0, (11.1.8) where F = Eh(x, 0) - r. Taking into account that we have gradF= 5s+ Vii.+Tiei (11.1.9) from (11.1.8) we deduce the condition (1 + u)F as + (a cos 0 + w) r L = a sin 0 + v, (11.1.10) which must be satisfied when r = h. Comparing the orders of magnitude we deduce: v(x, h, 0) = ev(x, h, 0) . (11.1.11) We assume that this structure is valid everywhere in the fluid. We have therefore v(x,r,9)=e'v(x,r,0). (11.1.12) From (11.1.10) and (11.1.11) we deduce the condition v(x,r,0) +asin0 = hz(x,0), (11.1.13) which will be imposed for r = h. In fact, this condition must be imposed for r = 0, but here v is not defined. In cylindrical coordinates the equations of motion are (1.11]: AP [Up.. + Vp. + (W/r)pg)+ +(1 + ryM2p)(U,+Vr + (1/r)V + (1/r)We] = o (11.1.14} (1+p)(UU=+VU.+(W/r)Ue]+p,, = 0 (I + p)[UV + vv. + (W/r)Ve - (W2/r)] +p. = 0 (11.1.15) (1 + p)[U11=+VW.+(W/r)We+VW/r]+(1/r)pe =0, (11.1.17) (11.1.16) where U, V, W will be replaced by (11.1.7), and P. =00/8x'... With the reasonings from 2.1, the equation (11.1.16) gives p(x,r,9)=E (x,r,0), v=+pr=0, (11.1.18) the equation (11.1.17) w(x, r, 9) = sw(x, r, 9), rwz + pe = 0 , (11.1.19) 452 THE THEORY OF SLENDER BODIES and the equation (11.1.15) 7t(r,t,B)=SYf{e,r,B), 71r+p. =0. (11.1.20) Keeping the terms having the order of E, from (11.1.14) we deduce AM2pr+u,.+vr+(1/r)v+(1/r)wo 0. (11.1.21) One observes that in the linearized system o does not intervene. The system coincides with the system for cr = 0. It. is the system (2.1.32) in cylindrical coordinate:. One may also obtain the equation of the potential. Indeed. from (11.1.18) -- (11.1.20) it results: yr - ur = 0, rw;r - uy = 0. P = --u. (11.1.22) The last two equations prove the existence of the function V(:r, r, 0), a.I. 7r. _ (j., V = 4'r i to = (1/r);pe , (11.1.23) and the equation (11.1.21) gives (1 - Al `) 11.1.2 a"- 10 t1:r.2 + r Or r a(pl C Or 1 + j92 = 0. (11.1.24) Fundamental Solutions We shall utilize for the solution of the system (2.3.4) the intrinsic form (2.3.8), (2.3.12) which will be written in cylindrical coordinates. From the equality ft: + f29 + f3k = f1 + frtr + fOZO, (11.1.25) taking (11.1.3) into account , we deduce f2 = frcus0 - fosin0, fr = f2cas0+ f3 sill 0 (11.1.26) fi = fr S1118 - fB COS B, f0 = -12 sin 04- f 1 cos 0 . Writing the inner product in cylindrical coordinates, in the subsonic ease and Taking (11.1.26) into ac( -ount, from (2.3.4) it results r) _ - I +fri0) ` ) 1 (11.1.27) 1 THE LINEAR EQUATIONS AND THEIR FUNDAMENTAL SOLUTIONS 453 where R, = x2 +#2r,2, (11.1.213) From (2.3.13) we deduce _ 1 47r _ / dx fr J x (x2 + #2x2)3/2 2r f= R, - (11.1.29) 1 4:s fr Rl - fr 1 + x r(R,) Taking into account the expression of the distribution 6(x) in cylindrical coordinates [A.7J, [A.10). we obtain yr = f'. L. a H(x)d(r) 21rr + 4r, 8r fr a 1 x 1 (11.1.30) R, } 4;r 8r r In the supersonic flow we have P(x,r) _ -i-. (fr_8x +fr where E(x. r) _ i l J E(x ,r), (11.1.31) H( x-kr) x- -k- r (11.1.32) Since a f" H(r-'-r) d r =H (x- kr) 0 8r oo T- k r dT = ,r r- - k r 2:E (x,r), r from (2.3.13) it results p(x, r) = 1 (f= - x f,.) E(x, r) (f.. 19 + r fr - r fr5T) E. (11.1.33) (11.1.34) Taking into account the formula (2.3.35), p and Vr will be: x Vr _ - fr H(x - A-r) r H(x)b(r) 27rr x 81 + 27r h 8 + 1kr, x r2 fr- (11.1.35) 1 fr 5T x -k r These formulas show that the perturbation propagates only in the interior of the cone x = kr. -r 454 THE THEORY 0FSLENDER BODES 11.2 The Slender Body in a Subsonic Stream 11.2.1 The Solution of the Problem In the case of slender bodies of revolution, the equation (11.1.2) has the form r=h(r). 0 < x < 1. (11.2.1) Considering that the unperturbed flow has the angle of attack z in the xOz plane, we deduce that. this plane will be the plane of symmetry of the flow. We shall replace the body with a distribution of forces defined on 10.11 with f2 = 0. From (11.1 26) we deduce fr = f sin0 (we denoted f3 = f). Taking (11.1.27) (11.1.30) into account, we deduce that perturbation produced by this distribution may be represented by means of the formulas p(.r, r, 0) . " { f, Rd Jx + f (l;) Sin A J 47r- (11.2.2) W(arrtd)= xj 1 'r(X, r, O) =d(l) R sin 0 d 4irr sin 0 ,-TI ' x0 (1+ R)f(E)d Jhit(0 0 ) (11.2.3) sill 0 y rl (1 + x0) d 4 it 8r Jo r R where R= 4+Ir'r2. .T.ox- Imposing the boundary condition (11.1.13), we notice that d(h) = 0 because h does not vanish for 0 < x < 1. Separating the variables we obtain the following integral equations: f f 0(I)d 41rh'(x) (11.2.4) (11.2.5) for O < x < 1 and r = h(x). In order to solve the first equation we shall utilize the identity v fir 1 R) 1 8 ((xol _ 1 5-x iI_ ((xo R} ra(RI (11.2.6) 455 THE SLENDER BODY IN A SUBSONIC STREAM Integrating by parts, we obtain: x - 1 47rh(x)h'(x) = ft(l) x -1) + Q h (11.2.7) -fi (0) r2 + = - [fl(1) + fr(o)] [1 + 0(h2)] h where _ t I=f' xo h xo+ fi(:;)dt. (11.2.8) For calculating the principal part of this integral, we notice that we have xo + lio' [i + 0(h2)] h (11.2.9) excepting the vicinity of the point £ = x where xp = 0. In [1.1] one utilizes this approximation on the entire interval (0, 1). Correctly, the integral I must be written as follows (/ I= + / 7+1 J (11.2.10) In the first and last integrals we may utilize the approximation (11.2.9). For ,i small enough, in the second integral one may replace fl (t) by f ' (x). We obtain therefore I = lro [ r:o fi (t)d t - J:+n fi (t)d C] [1 + 0(h2)] + 1 + f (x) lim +'' ,j- .0 q dt p +h We calculate the last integral with the substitution x - = u and we observe that it vanishes. Hence, I = [fl (x) - fi(0) - ft(1) + fl(x)J [1 + 0(h2)] . (11.2.11) Neglecting the terms of order h2 with respect to 1, from (11.2.8) and (11.2.11) we obtain fi(x) = -27rh(x)h'(x) = -S'(x), (11.2.12) with S(x) = 7rh2(x) the area of the crass section a of the body in the point having the abscissa x. 456 THE THEORY OF SLENDER BODIES Deriving in (11.2.5), we find j' (1 + R) f(E)d F + f32h2 "' f J. 1T4 -4iah2 . (11.2.13) Calculating the integrals by means of the formula (11.2.10), we have J1(1+ O)f(:)dF=21 f(t)dF[1+0(h2)] Hence, neglecting 0(h2) with respect to 1, we obtain f(e)d t = -2Tah2 = -2nS(x) 10 or, deriving, f (x) = -2aS'(x) . (11.2.14) For the profile with zero angle of attack (a = 0) we deduce f = 0 whence N(x,r) _ -4 0, xo+ r dF. (11.2.15) This representation of the potential is known in the literature [1.11, (1.38]. We have also, from (11.2.2) (for r # 0), (11.2.16) P(x, I') = -4r, Vr = SFr 11.2.2 The Calculus of Lift and Moment Coefficients We shall calculate at first the pressure for r = h. It can be obtained from (11.2.2), (11.2.12) and (11.214). Utilizing the identity (11.2.6) and the calculations (11.2.8) - (11.2.11), we deduce jf(0 . d (R) (11.2.17) [1 + O(e2)] . Hence, sin 0 P(x, h(x), e) = pi (x, h(x)) - a Wh(x) S`(x), (11.2.18) 157 THE SLENDER BODY IN A SUBSONIC STREAM where r fl (E)' (1?) P1 (X. h(.r)) _ - 1 J 1r,,iid E (11.2.19) 1f1 4r :rOS'(E)d E (%2+132,12)3/'2 Taking into account that fr = O(V2), we calculate the principal part of pl as follows: I 4;rP1(x, h(x)) W) r_I,d fl (S)3 _ 11(1) (x11-) f1(0) x'-+f32h2 -j3 h= 9 r, +J fl(E) I-E S'(1) + S'(0) +f d z*F7 + fi(x) lim r. 1-x T J0 x - - 01 S"(r) Er > r0+th` +O(Fd)- - S"(E)d x-E x t J S"(x.) E--x 1 + S(x) In 1-c r. + O(c4) , the principal part which was written being O(_2). The lift coefficient may be calculated with the formula cL= (11.2.21) - As where S is the surface of the body, and n, the outer normal and with the notation F = r - h(x), given by the formula grad F - -h'i + i, (11.2.22) n = Egrad FI V1-1+ h'2(a) Taking into account the element of area on the surface S, and the relation (11.2.13), we obtain r1 A CL =- / / ,r p(x, h(x), O)h(x) sin 0d rd 0 = o =(I Es(1)-s(0)j. (11.2.23) 458 THE THEORY OF SLENDER BODIES For the drag coefficient one obtains CD =- Jjpn id a = jj(x, 1 h(x), O)h(x)W (x)d xd 0 = (11.2.24) pl(x, h(x))S'(x)d x = O(E4) . The drag coefficient does not depend on a. At last, the moment coefficients are the scalar components of the product -2 /f xxpnda, S where x = xi + h(x)ir. Taking into account (11.2.2) and (11.2.22), we obtain cr =0, cs=0, (x + hh')h(x)p(x, h(x), 0) sin Od x d 0 = f1I -2a / o 1 Lx +' (11.2.25) S'(x)d x = O(E3) . J Obviously, for a = 0 one obtains cy = 0. Neglecting the term (S')2, for the moment coefficient c, we obtain the approximate value ri c, = -2a J xS'(x)d x = 2aV, 0 b ecause S(1) = 0 and the term of the body. 11.3 11.3.1 j S(x)dx represents the volume V The Thin Body in a Supersonic Stream The General Solution In this subsection we consider the same problem like in the previous subsection, but now the unperturbed (free) flow is supersonic (M > 1). 459 THE THIN BODY IN A SUPERSONIC STREAM titre consider again that in this case the xOz plane is a symmetry plane, such that we have f; = 0 whence fr = f sin 0. The fundamental solution is (11.1.31), (11.1.34). and the corresponding potential is (11.1.33). For a continuous superposition of forces on the segment [1).11. the perturbation will be given by 1 +f w sin ©J E(.,ro, r)d P(x. r, 0) = - T f p(x, r, 0) = [f=w 2-r J0 - rl f (5) sin 0] E(r(j, r)d i X t'r(X, r, 0) = (5(1') in0 2'r JU .E(xa)d ,r + sin0 ff U 1 err f0 roI 0Y . l - J)F(xor)d r 9r, (11.3.1) The derivatives may interchange with the integrals and. taking (A.3.15) into account, we have for example f 1 f W E(xo, r)d t= ax f _ a x-kr H(.r-kr) o f x 1 f E(xo, r)(1 _ FL-r H(x-k'')r3 4. `I f (4) == zo k-12-7-11 Hence, the solution (11.3.1) maybe written as follows _ 2r TX o = r-kr 1 r _ yr x-kr 8 1 P x sin 0_ 8 d42z Or x2 - k r(4) dE- in xo - k 2 rI sin 0 2rrr f()cI+ 2 r b(r) +siI 0 27tr 1 Of sin 0 8 f V) dC xo - k2r2 o r-kr xQ 2,rr Jo 1 r r-kr fx (50) ro r- d jr-kr x xo xo-k'r k2r2 d (11.3.2) 460 THE THEORY OF SLENDER BODIES valid for x > kr and p=W=vr=0 forx<kr. (11.3.3) Imposing the boundary condition (11.1.13) we shall notice again that 6(h) = 0, because h # 0. Separating the variables we deduce the following integral equations 8r 0 1 Cr r x0 _ Vd t) rah = 27rh'(x) x-kr a x0 ar} Jo xo - k ref (t)dtIrah = -27rah(x) . (11.3.4) (11.3.5) In order to put into evidence the principal part of the integrals for r = h = O(e), we must notice that we cannot derive directly the integrals, because the integrands become infinite for { = r - kr. For avoiding this situation, we shall perform the change of variable 4 -- u: x - a=krcoshu. (11.3.6) One obtains a of=-kr a' 8 fr (S) xo - k r d = 8r x r x -k r x f:(0) fx(0) x- k r , -kJ f. (x UCh fi(x - kr cosh u)cosh u d u o _ _r 1 r f'(0) [1 + O(ff)] - krcashu)du = rx-kr 1 0 -r 0 xo ' xo - k r f (t)d -kr f=(t) [1 + o(e2)] d. (11.3.7) Neglecting O(e) with respect to 1, we deduce that the principal part of the integral from (11.3.4) is - f=(x)/r. Hence, from (11.3.4) we deduce ff(x) = -2whh' = -S(x), (11.3.8) S(x) giving the area of the cross section of the body in the point x. TILE THIN BODY IN A SUPERSONIC STREAM 461 Similarly we obtain 0 r-Ar xO Or Jo t-tr 1 r X0 r \/x - 1 1 r2 z-kr 0 1f(01d --1 x - 0 + f(0) :c2 f(c do :c3 - k2r2 r ra u f'(Od (11.3.9) Neglecting O(r-2) with respect to 1 we find at first 1 (7 z-kr xp .r f,-Ir f f (E)d -r-kr rx`t -rI 1 x-kr f ()d + r f (0) fz_kr f'(e)d r1-jZkr Ja f(4)d4. The equation (11.3.5) becomes S kr kr f (x - kr) + 0 f -2-xnrh. (11.3.10) where we put r = h. In the left hand part of the equality, we expand into series for obtaining the principal part. We have r kr f (:x) + 0(r2) + kr f (x) + 0(r2) = -2 rorh . f 0 Hence, s J f --2riS(x) whence f (x) = -2aS'(x), (11.3.11) the analogy with the subsonic case being obvious. 11.3.2 The Pressure on the Body. The Lift and Moment Coefficients In order to put into evidence the principal part and the expression of the pressure, we notice that utilizing the change of variable (11.3.6), we TILE THEORY OF SLENDER BODIES, 462 deduce x-kr 8 o -krx f=(t) Ox x -k r r-kr +f=(x) xo-k r o f=(a) o-k r 0 x +f:(x)arccosh kr __ rs - x x - f., t to 2s = fs(x)ln kr xo -kr Jo f'(C) - fr(x)d4+ df f ff(t)d t r=-kr -k r + f=(0) + f1(0) d t= -J + Jo r f.'W - f.'W d t + x- d t+fx(x) In x + x- k r e kr fi(x) - fY(0dt. x-t Taking into account the calculation performed at (11.3.7), we deduce p(x, h, 0) = pi (x, h) - a sin 0 S'(x), (11.3.12) where pi (x, h) = - S"(x) 21r kh _ 2x 1 S"(x) - S"(4) 27r fo X- d (11.3.13) The lift, drag and moment coefficients may be calculated with the formulas (11.2.23) - (11.2.25). We have therefore, like in the subsonic case, CL =a [S(1) - S(0)1 = 0(e3) CD =Ji p,(x,h)S'(x)dx=0(e4) 4 (11.3.14) Gr=az=0, cy = - Cr f ' 0 Ix + 2(x) J S'(x)d x = O(s3) . THE, THIN BODY IN A SUPERSONIC S1'IREAAI 11.3.3 463 The wing at zero angle of attack For a = 0, we deduce f = 0 whence P('r., r) = 9(r, r) t'r(x, 7.) a 1 S'() /'x-rrr 2;, Tx Jo zo ` J. r d =-kr S -o xo - kr' 1 of r-kr 2 j S (e)d o d S'(S)dS .4-k r' x-kr 1 2rr J0 Sit ( )d k2r2 (11.3.15) This is the solution between the wave from the louring edge and the wave from the trailing edge i.e. in the region denoted by II, where a > kr (fig. 11.3.1). In I (:c < kr) p = yr = 0, and in IN the solution is p(x, r) _ 1 oz-kr 2r, 2irr (11.3.16) z-kr 1 Vr(x,I-) -- - S"()d t xo-k-r'2 0 dc. x _ 13-1 1A w 1 Fig. 11.:3.1. This explicit manner of representing the solution is also valid in the general case. 11.3.4 Applications At first we shall consider the thin body having the shape of a cone (fig. 11.3.2) of equation r = ex, 0 < t < 1. We deduce S = 1re2x2 THE THEORY OF SLENDER BODIES 464 whence 2 CL = JrE at cy = - 21u E2 3 ti , CD = -7rE' ln(k/2). (11.3.17) Fig. 11.3.3. Fig. 11.3.2. For cy and CD we have retained only the principal part. If the angle of attack is zero, then in II we have d _d C = 62 arccosh j _ z kr p(x, r) = E2 E2 In yr (x, r) = - xo - k"jr Jo x+ x2-k r x-kr EZ r (11.3.18) kr o x0 xkr-d t _ -8 2 :r - k r r We notice that on the radius vectors r = cx, c < 1/k, the pressure and the velocity are constants. The flow is conical. The second example the double cone from figure is 11.3.3 Since Ex, h(x) = 0 < x < 1/2 (11.3.19) r(1-x), 1/2<x<1, it results that S' has a discontinuity in the point x = 1/2. We suggest to the reader to establish the formulas of derivation for fW dta rx-kr xo-k r and to write the solution. ' arJ10 xo -k'r d (11.3.20) Appendix A Fourier Transform and Notions of the Theory of Distributions A.1 The Fourier Transform of Functions The following definitions will be given in R3. Their expressions in R1, R2 or RI will be easily deduced. The Fourier transform of a function f : R3 --+ R is denoted .F[f], or f and it is defined by the formula (A.1.1) F[f](a) = f(a) = j3 where x = (x, y, z), a = (al, a2, 03) (A.1.2) a.x=alx+a2y+a3z, dx =dxdydz. The operation F is called the Fourier transform. We notice that i f (x) is absolutely integrable on R3, he. if 3 (f )d x < oo, (A.1.3) then the Fourier transform exists. Moreover, if f satisfies certain conditions of regularity, for example, if f E C0(R3), or if f is piecewise smooth with respect to every variable [A.10], then f may be obtained from f using the following inversion formula (2;r)3 f (x) _ (2-,,)3-F-1 [j(x) f a= (A.1.4) _ F[fl(-x) _ [f (-a)J(x) , where ci a = d a1 d 02 d a3. The last two expressions show that the inverse of a Fourier transform may be obtained by a direct transformation. 466 FOURIER TRANSFORM. THEORY OF DISTRIBUTIONS One proves (Lebesque's theorem [A.9], (A.51) that j tends to zero when ICkI -+ .x}. The condition (A.1.3) is restrictive enough. For example, the function f = 1 does not verify this condition. The theory of distributions allows to enlarge the class of the functions which admit a Fourier transform. Notions of the Theory of Distributions A.2 The Spaces D and S The theory of distributions relies on the notion of space of the test functions. The space V = D(R3) consists of the set of functions W(x) infinitely derivable, with compact support in R3. We say that a sequence of functions {V,} from D, converges to a function S' E D, if for every multi-index k = (k1, k2, k3), k1, k2, k3 being non-negative integers, we have: DkW11 rya Dkcp, (A.2.1) where, with the notation IkI = k1 + k2 + k3, we have put Dk -0jkj = --,- k, (A.2.2) The set D endowed with this convergence law is a linear space- Ale may we that the Fourier transform of a function from D is not a function from D. Indeed, the Fourier transform is an analytical function [A.12), but the support of an analytical function which cannot be compact. There exists another space which is invariant to the Fourier transform and plays a basic role in the definition of the Fourier transform of a distribution. This is the space S = S(R3), i.e. the space of the rapidly decreasing test functions. We call rapidly decreasing function an infinitely derivable function V(z) in R3 which, together with its derivatives decreases when Ixl - oo faster then every power of Exl`t. This means that (A.2.3) IxkD'v(x)i < CL-j, for every nmlti-indices k and 1. We denoted xk = xklyk'2zka. The convergence in S is defined in the following manner: the sequence of 467 DISTRIBUTIONS functions {V } from S converges to the function cp E S if for every multi-indices k and 1 , we have: X, xkD'pn(X) xkDrv(x). (A.2.4) The set of the rapidly decreasing functions, endowed with this convergence law. is a linear space. The convergence in D implies the conver- gence in S. We have D C S. but S does not coincide with D; For example. the function exp. (-x2) E S, but A.3 D. Distributions The distributions are linear and continuous functionals on D or on S. This means that a distribution f detennines the correspondence between a test function V and a number denoted by (f, q) and we have: (A.3.1) (f,Anpi +A 2) = an (f,Sit) +A2(f,p2) , for every two real or complex numbers an, A2 and that if P>s I V in D (or S), then (f, °Pn) - (f, (P) . (A.3.2) One denotes by D' the set of the distributions defined on D and by S' the set of the distributions defined on S. S' is the space of the temperate distributions. Obviously, S' C D'. We say that the distribution f is equal to zero in the domain f2 (it is denoted by f = 0) if (f, gyp) = 0 for every p from D (or S) with the support in 11. The distributions fn and f2 are equal in n if fl - f2 = 0 in 11. i.e. if (f1, 0 = (f2. (P) , (A.3.3) for every v E D. We call the support of the distribution f and we denote it by supp f the set of the points which have a vicinity where j is not equal to zero. Every locally integrable function f (x) defines a distribution by means of the functional U. P) = j f (x)4%(x)d x . (A.3.4) A distribution f is it regular function-type distribution if it may have the form (A.3.4). Every other distribution is singular. The best known 468 FOURIER TRANSFORM. THEORY OF DISTRIBUTIONS example of singular distribution is the distribution of Dirac. The distribution of Dirac with the support in 4 is denoted by bo(x) or b(x and is defined by the functional (b(x - V, (x)) = rW . (A.3.5) Formally, it may be written as follows J v(x)6(x - 01 x = pW . (A.3.6) If rn(x) is an infuiitely derivable function, then mv is infinitely the distribution f and the function m., denoted by tnf, by means of the formula derivable and we may define the product rnv) (inf, 4,) (A.3.7) It results m(x)b(x - ) = (A-3.8) whence, ni(x)b(x) = 0, if rn(0) = 0. (A.3.9) For the existence (A.3.8) the continuity of the function is sufficient in in t;. We shall consider the three-factor c and the homothety cx = c3z). If h(x) is a locally integrable function, then h(cz) is also locally integrable, such that the distribution corresponding to h(ex) is defined by the functional _ (cix. (h((x), 4"(x)) =f h.(cx)p(x)d x Setting cm = and observing that the integration limits interchange when c; is negative, we deduce, with notation lei = (clc-2cy) IcC (h(cz), 5'(x)) - j h(E),p(k/c)d _ (h(x). 4p(x/c)) For o distribution f (x), one defines the distribution f (cx) by means of the formula ICI (f (cx) s'(x)) = (f (x), cp(xl (:)) . (A.3.10) The definition may be extended to a non-singular linear transformation. As a particular case, it results: Ic16(cI x. C 2Y, car) = b(x ah z) . (A.3.11) 469 DISTRIBUTIONS f The derivative of the distribution determined by the formula is denoted by f' and is (A.3.12) (f', V) _ - (f, ) . Defining the function of Heaviside H(x) by the formula x<0 H{r.) = 10, 1, (A.3.13) x > 0, it results 00 (x)dx=5P(0) i 0 Taking into account (A.3.5) we deduce H'(x) = 6(x). (A.3.14) Let us establish now the formula (m(x)H(x)' = m(0)6(x) + in'(x)H(x) for a function E C'(R). We have ((mH)'. y,) = = m(0M0) + f - (mH. vp') x (A.3.15) r x= m'(x)cp(x)d x = (m(0)6(x) + m'(r)H(x), v') . 0 On the basis of (A.3.8) and (A.3.15) we deduce that the solution of the differential equation Lu = 6(x), (A.3.16) where Lu = u("')(x) + O, (X)u('"-I)(x) + ... + am(x)u(x). is u(x) = H(x)v(x), (A.3.17) where v verifies the equation Lv = 0 and the conditions v(O) = v'(0) ... = v,(m-2)(0) = 1 0, We call the direct product of the distributions f(x) E D'(R") and g(y) E D'(Rm) the distribution f (x) g(y) E D'(R"+') defined by the functional (f (x) g(y), r'(x, y)) = (f (x), (g(y), W(x, y))) , (A.3.18) 470 FOURIER TRANSFORM. THEORY OF DISTRIBUTIONS where V E D(R"+'"). The definition is valid also and when D' is replaced by S. The direct product is commutative, i.e. f (x) g(y) _ = g(y) f (x). As an application, we shall prove that in R2, we have (A.3.19) b(x) = a(x) - a(y) Indeed, applying the definition, we have: (OX), AX, y)) = 4'(0, 0) . But we have also, (a(x) a(y), Ox, y)) = (a(x), (a(y), (x, y))) _ = (6(x),;?(x,0)) = V(0,0) A.4 The Convolution. Fundamental Solutions For two functions mi(x) and m2(x) absolutely integrable in R3, we define the convolution m(x) = (ini * in2)(z) by the formula (ml * MAX) = =1 3 J pa "11)mx - t)d : = m( m2(x - t)dt = (in2 * ni)(x) The function m(x) is absolutely integrable, such that it generates a function-type distribution (A.3.4). Writing explicitly that functional, we are determined (A.51 to define the convolution 0= fi * f2, of two distributions fi and f2 by means of the formula (fi * f2, ') = (fi (x) f2(O, V(x + f)) , (A.4.1) fi(x) f2(4) representing the direct product of the distributions fi and f2. The convolution (A.4.1) exists if, for example, one of the distributions have a compact support (A.5j. The equality f j *f2 = f2 *fi results from the commutativity of the direct product. As an application, we shall calculate b * f. We have (6 * f, P) = (6(x) . f W,'?(x + 0) = (f (0, (6(x),'P(x + i))) _(fW"PW) THE CONVOLUTION. FUNDAMENTAL SOLUTIONS Hence, a*f=f*6=f. 471 (A.4.2) We shall establish now the formula D(fi*f2)=Df,*f2=fi*Df2, (A.4.3) the operator D being defined in (A.2.2). We have (D(f1 * f2), Sp) = (-1)Ikl (fh * ff, DW) _ = (_ 1)lkl (fi (x) - ME), Dpp(z E)) _ (-1)Ikl (f2(E), (fi(x), Dip(x + i))) _ _ (f2(E), (Dfh(x), D& + F))) = (Dfi * f2, sp) , and the result proves the first equality from (A.4.3). The second equality results from the commutativity of the convolution. We shall consider now the linear differential operator of order m, M L = E a*Dk , Ikl-O where ak are constant. We call fundamental solution of this operator, the distribution t E D' which verify the equation LE = 6(c). (A.4.4) Obviously, the fundamental solution is defined with the approximation of an arbitrary solution of the homogeneous equation. We have the following theor_mr. If f e D' and there exists in ?Y the convolution f s E, then the solution of the equation Lu = f (A.4.5) exists in D', is given by the formula u = f *.C (A.4.6) and it is unique in the set of distributions from D', for which there exists the convolution (A.4.6). Indeed, utilizing (A.4.2) and (A.4.3), we deduce Lu=L(f*E) a1Dk(f*E)=f*LE= f$6= f. lkI-O 472 FOURIER TRANSFORM. THEORY OF DISTRIBUTIONS In order to demonstrate the uniqueness, we denote by u another so. lution. Obviously, no = u - u satisfies the homogeneous equation Luo = 0, such that we have: uo=uo*6=uo*Le=L(uo*e)=(Luo)*e=0. A.5 The Fourier Transform of the Functions from S Since the functions from S are absolutely integrable, they have Fourier transforms F[ pj defined by (A.1.1) and these are bounded continuous functions on R3. Integrating by parts and taking into account (A.2.3), we deduce: -F(VACt) = (-ial)FIVJ(a). (A.5.1) On the basis of the same condition, we may derive inside the integral, such that it results (O/Oai).('PJ(a) = .F(ixtipj(a). (A.5.2) By recurrence, it results: .F(Dk(pJ(a) = (-(a)k.F(V](a), f((ix)'4 j(a), (A.5.3) (A.5.4) for every multi - indices k and L. The last relation show that F(VI is infinitely derivable. We shall prove in the following F(cpJ E S. To this aim we replace cp with (ix)rcp tin (A.5.3) and taking into account (A.5.4), we deduce F(Dk((ix)`p)J(a) = (-ia)k.F[(zx)l ,](a) = = (-i)IkjakDI,F[Wj(a) whence ]akD'F('j(a)I <_ fx$ IDk(xt4o)IdZ < 00, i.e. Y[w] satisfies (A.2.3). Taking into account that the Fourier transform F(spj is integrable and continuously differentiable, we deduce that cp may be obtained from cp sing a formula like (A.1.4). THE FOURICR TRANSFORM OF THE TEMPERATE DISTRIBUTIONS 473 A.6 The Fourier Transform of the Temperate Distributions The Fourier transform of the distribution and it is defined by formula f = .F[ f l f E S' is denoted by (f, 0) = (f. y i) , for every V E S. Since (A.6.1) E S. it results [A.121 that the Fourier transform of a temperate distribution is also it temperate distribution. For f E S' we define the inverse Fourier transform F--I, by the formula(27r)'-F-' [f (x)1 = -F[f (-x)] . (A.6.2) We shall prove that f j:-- I [.[fl] (A.6.3) . Indeed, taking (A.3.10) into account, we deduce for c = -1 (27r)3 (,F-1[.[f]h (.F[.F[f](--c )l, F-' (-a)) = (2r)3 (27r)3 (f,.Fp,-' [.,III) _ (f ) Similarly we prove the second equality. We shall establish now the operations (A.5.3) and (A.5.4) for temperate distributions. We have: (F[f=], } = (f=, ) _ - (f, a,.F[pl) = [iaip]) _ (F[f].-iniy,) _ and we deduce, whence (_iaiJ,) .F[f] =-iaif, (A.6.4) T[grad f) = -i a f , F[div f ] = -i a f (A.6.5) .F[rotfl=-iax f. From (A.6.4) we obtain by recurrence a similar formula to (A.5.3). Analogously, we have: ( 0 _771fl, V(a) Ylfb f..F an., 474 FOURIER TRANSFORM. THEORY OF DISTRIBUTIONS (f, i x.F[d) = (i xf, Y[w)) = (f [i xf ], io) hence, (A.6.6) .F[i:rf) and, by recurrence, a similar formula to (A.5.4). From the definition (A.6.2), it results that relations similar to (A.5.3) and (A.5.4) are available for the inverse transformation too. Hence, f-1 (i x)" f (x) ia)!f](x) = D'f(x). (A.6.7) (A.6.8) We shall frequently utilize relations of the form .F-t[--t«tf1 = f:, F-t[cto2f) _ -f=y. (A.6.9) We shall prove in the following the formula (A.6.10) 17-t(f(ca))(x) = ICIf 1 cl c representing a three-factor. Performing the change of variable a = = cf3, we have for a function lip\alJ (x)=,l ( cxd A = lclw(cx). lei For a distribution f we shall utilize the formula (A.3.10). We have lei (F-t [f(Ca)1 P) (f(a),F-tso(a/c)) lei (f(ca), -F- 1,P) _ (f(x), 0(cx)) = (f (x/c), w) whence (A.6.10). For the direct product f (x - g(y)) we have: (A.6.11) F[f(x) - 9(y)] = Y[f)(a) - F[9)(0) , and for the convolution, if one of the distributions has a compact sup- port, we deduce .Fff * 9) _ F[f).F[9] . From (A.6.11) it results f-I [f (a) 4(/3)) = f (x) . g(y) . (A.6.12) (A.6.13) THE CALCULUS OF SOME INVERSE FOURIER TRANSFORMS 475 A.7 The Calculus of Some Inverse Fburier TSransforms We shall calculate at first the Fourier transform of the distribution of Dirac. For cp E S we have (716(x - 4)], Sp) _ (a(z - ), r) ='P(A) (A.7.1) (el a-4, SP) , Hence .F[a(x - a; )] = exp (i a ), b(a: - F) _ '-t [exp {i a }j , whence, for (A.7.2) = 0, 7A =1, _ 1 [1] _ (A.7.3) It results therefore F[1](a) = (2a)3d(a) . (A.7.4) Hence the temperate distribution I has a Fourier transform. Denoting by F2 the Fourier transform in R2 and by f'1 the transformation in R1, we have on the basis of the formula (A.6.13): .Fz 1 [1/i a] _ .F_ [l/i a1](x) ' t [1]{3l) (A.7.5) Applying the Fourier transform to the equation du/dx = b(x), (A.7.6) we deduce -i o1i = 1 whence: U = -X1[1/ialj(x). (A.7.7) The equation (A.7.6) has the form (A.3.16) and its solution has the form (A.3.17). It results s=X(x). (A.7.8) From (A.7.2) - (A.7.8) we deduce T1 [1/i all = -H(x)6(y). (A.7.9) For the determination of the fundamental solution of the equation of Laplace we need the following results (demonstrated, for example, in [A.12], §97, [A. 131, §6.6, or [A.71, §5.3). I = 4x [ [' n = 3, (A.7.10) 476 FOURIER TRANSFORM. THEORY OF DISTRIBUTIONS YV [i'i'_} n=2, 2 (ln lxl + C), (A.7.11) C being a constant determined in [A.13], but without any importance here. Taking into account (A.6.10) with c = (,3,1,1.) we deduce, on the basis of the formulas (A.7.10) and (A.7.11): 1 Jrz 1 _ 1 #2aI+a2+a3 - -] I FP fl202 t 1 + 4n (A.7.12) (J +z ) = - 2n (ln x'- + 02y= + C - In (i) , 2] 2J i 1 (A.7.13) ,6 FP representing the symbol for the Finite Part (Appendix E), and p, a positive constant. Also, for determining the fundamental solution of the wave equation we need the formulas f-1 sinalalt 6(at - Iii), 4,rat lal T-I sin aloe lt1 H(at - l xI ) I Ial 1 21r a t- -- Ixl n = 2, (A.7.15) demonstrated, for example (A.7.14) in [A.5] Chapter 2, §3.4, and (A.7.15) in (A.12) §9.7. We shall prove in the sequel the following formulas 1 _ 1 H(x - k f2 + z2) 1 [-k2a"i + a2 + a3 J F2 1 -k2 2a x-k (y + a2, = 2k H(x + s) kl>!i). , (A.7.16) (A.7.17) In order to prove (A.7.16), one considers the following partial differential equation: k2u." - uyy - u=, = 6(x) . (A.7.18) Applying the Fourier transform, we deduce (-k2a? + az + a3)i = 1, whence, u [_ k2al2 +a2+a321 1 . (A.7.19) It results therefore that the first member from (A.7.16) is just the solution of the equation (A.7.18). We shall determine in a different manner this solution, namely applying the Fourier transform only with THE CALCULUS OF SOME INVERSE FOURIER TRANSFORMS 477 From (A.7.18) one obtains the respect to the variables y and equation (k2d2/dx2 + w2)ta = d(x) , (A.7.20) where w = + , and u represents the two-dimensional transformation. The equation (A.7.20) is like (A.3.16), and its solution has the form (A.3.17). We deduce u.- H(x.) sin(wx/k) k w Utilizing (A.7.15) one obtains (A.7.16). Acting similarly, in the two-dimensional case, we find: u= ii H(x) sin(a2x/k) k (Y2 t being the notation for the Fourier transform with respect to the variable y. Utilizing the formula F1 sinXax11 7r (A.7.21) I(Inl>aa , 0 given for example in (A.1], p.202, we obtain (A.7.17). In the unsteady aerodynamics we shall meet the following type of formulas co alc tl Mat ct2 47r Ix 1 - cosalaltl U2 - Iml), (A.7.22) a t - Ix+ Ix() 2r = 3, (at + (A.7.23) IxI n=2. We shall prove these formulas (following an idea suggested by V.Iftiinie) using the results concerning the Cauchy problem for the non-homogeneous wave equation. To this aiin we shall denote: v(t, a) = l - cosalalt Ck2 We deduce u.(0, a) = 0, u((0, a) = 0 , irtc(t, a) = a' arsa.lalt = -(i2a2ii(t. a) + (L2. 478 FOURIER TRANSFORM. T'IIEORY OF DISTRIBUTIONS Applying the operator F-', we deduce: 11(0, x) = 0, ut(0. x) = 0 utt = a2Aat + u26(x) . For determining u we have therefore to solve a Cauchy problem for the non-homogeneous wave equation. The solution of this problem in the three- and hi-dimensional cases is given by Poisson's formula [A.12]. Utilizing this formula, we find (A.7.22) and (A.7.23). At last, replacing x by x. - t in the formulas (A.7.9), (A.7.14), (A.7.15), (A.7.22) and (A.7.23), we obtain in the three-dimensional Case, with the notation R = F-I (c - t) + y' + = 5(x - t) - a(t,.z) Ceinit. I sin aIctIt einat f-I _ I a5(at. - R), )' = TI-rat (a1 n247rR 1 - c)sajait iaIt - II(at (A.7.24) R) and in the two-dimensional case, with the notation I# F-a ei,«t] f_1 Il I b(x - t) - (5(71), I ial JI 27 (1 - cosa'aHt dolt I rte _ H(at - R) a' '' --V H(at -- 7) 2-r In at + n."2t2 - f2 R (A.7.25) In [1.101 we may find direct. demonstrations of the formulas (A.7.25). A.8 The Fourier Transform in Bounded Domains In this last part, we return to the Fourier transform of the functions and we give, following Homentcovschi's idea [A.61, the transformation formulas in case that the function f (x) is defined on a bounded domain D. We assume that D is bounded by a surface S which closes a domain D' and by a surface of discontinuity E . We prolong f in DY, giving 479 THE FOURIER TRANSFORM IN BOUNDED DOMAINS it the value zero. We make the same thing in the domain D° which closes E. Applying the flux-divergence formula we obtain: fxID=fx6:' =fTIR3 rrfn - JS fie' f,,n fnleia'xda- J'; Of llnleiaxda. (A.8.1) F(grad f[D = -ic"f - 1 fne'o F[div fJD = -ice f F[rot f]D = -ia x where (A =f+ -f-. J xda- r f nei°t'xda - J QfIne'a-xda f f x ne`a'xda- J s DfOne`a'xda, E (A.8.2) Appendix B Cauchy-type Integrals. Dirichlet's Problem for the Half-Plane. The Calculus of Some Integrals B.1 Cauchy-type Integrals We consider in the z = x + i y complex plane a smooth curve r, i.e. a curve which has the parametric equations x = x(8), y = y(s), sl a 82, (8.1.1) where x(s) and y(a) are continuously differentiable functions, whose derivatives do not vanish simultaneously in the same point. The curve r may be closed or open; if it is dosed, then z(81) = z(82); if it is open, then we assume z'(sl) = z'(sl + 0) and z'(s2) = z'(32 - 0). By definition the positive sense on r is the sense corresponding to the increase of the parameter s. The smooth curve is obviously rectifiable, such that we may consider as parameter 8 the length of the arc measured from sl (= 0) to 82(= 1). In this case, we obviously have zi2 + y'2 = 1. Let f(t) be a complex function depending on the complex variable 1, defined on r and Riemann integrable. The integral F(z) = 2Ai J t (t) d t (B.1.2) is called Cauchy-type integral. As we lmow from the books of complex analysis, the function F(z) is holomorphic in the interior of the contour r. if r is at a finite distance, then F(z) behaves at infinity like 1z4'1. We shall investigate, in the following, what happens with the integral (B.1.2) if z = to E r. In this case, the integrand has obviously a nonintegrable singularity in to and generally the integral has no sense. There exists however a large class of functions (we are not interested here in the largest class) for which we may give a definition to the integral, 482 CAUCIIY-TYPE INTECRALS namely the class of the functions which satisfy the so called Holder's condition. We say that the function f (t) satisfies Holder's condition on r if there exist two positive constants (different from zero) A and µ(µ < 1), such that, for every two points tl and t2 E r sa we have (B.1.3) If (ti) - f(t )I < Ajtl - t2VY. Obviously, the functions f which satisfy Holder's condition are continuous on r. If u = 1, the functions satisfy Lipschitz's condition. B.2 The Principal Value in Cauchy's Sense We shall give now the definition that we have mentioned before. We assume at first that to does not coincide with any extremity of the arc r (if it is open). We consider the are of circle with the center in to and the radius s which cuts the curve F in two points tl and t2 and (t) d t we denote by - the are t1t2. If for a -+ 0 the integral Jr-y tf to has a finite limit, then this limit will be called the principal value in Cauchy's sense. We denote f(t) Iifot--f(t) dt - frto t - d t. Y to (B.2. t. The principal value is it distribution (A.12], [A.14]. We shall prove in the sequel the following theorem: "If f (t) satisfies Holder's condition in the vicinity of the point to, then the limit (B.2.1) exists and it is unique. For the proof we shall write: f f(t) dt= tto fr_1, f(t)-f(to)(I t+f(tu) f t - to r1F t dt - to (B.2.2) Having (B.1.3) in view, the limit of the first integral from the right hand member exists and it equals the usual improper integral on r. The last integral is calculated as follows: f tt -7 t o = 111(t - t)la` + ln(l. - to)It = In u - to+ + ln(tl - to) - 111(t2 - to) . Diu tl - to = Iti - tolein, t2 - to = It2 - tole'' and Itl - toI = It2 - tot. 483 PLEMELJ'S FORMULAS It results In(tt - to) - ln(t.2 - to) = i (a - 0). Passing to limit, when E-0, ct - W=ir,weget dt Inn _ dt e- 0 = In b - to a-t0 + i a = In b- to to - a (B.2.3) Since the last integral from (B.2.2) has a well determined limit, the theorem is demonstrated. The case when to coincides with one of the extremities of the are F. depends on the behaviour of the function f in that point (,See for extunple [A.27], §29--32). If t9 coincides with an extremity and f (to) = 0, we are in the previously considered case., because we may extend arbitrarily the contour t beyond to, setting f = 0 on the extension. B.3 Plemelj's Formulas We shall investigate the behaviour of the Cauchy-type integral in the vicinity of the curve F. To this aim we shall give at first the following definition [A.271: we say that F(z) is continuously prolongable on r in to (different from the extremities) at left (right), if F(z) tends to a well determined limit F+(to)(F_(to)) when z - to on every path situated at left (right). With this definition we may give the following fundamental theorem : verifies or, t Holder's condition. then F(z) is continuously If f (t) prulongable on rat left and at right, excepting the extremities where f (to) 76 0 and Fi. (to) = t2f(to) + 2ri I tf (t ndt. (13.3. 1) formulas. They have been The formulas (B.3.1) are called given in 1908 [A.29]. Their demonstration may be found for example in [A.18], [A.271. B.4 The Dirichlet's Problem for the Half-Plane tine shall solve in the. sequel the following problem: We seek for the function (B.4.1) F(z) = u(x y) + i v(x. y). 484 CAI;CIIY-TYPE, INTEGRALS holornorphic in the half-plane y > 0 and continuously prolongablct on the Ox, axis. which reduces at infinity, to an imaginary constant i C (C may have the value zero) and whasc ir_al part is imposed on the above mentioned axis, i.e. u(x,O) = f(x), (13.4.2) where f is a function with a compact support compact which satisfies Holders condition. At first we have to mention that there exists a single function with the above mentioned properties. Indeed, assuming that there exists two functions F1 and r2 with these properties. their difference F = = F1 - F2 is holotnorphic in the superior half -plane and it vanishes at infinity. The real part of the function F is therefore harmonic in the half-plane y > 0. zero on the boundary y = O and zero at infinity. According to the maximum principle for the harmonic functions, the real part of the function is identical zero. F reduces therefore to an imaginary constant which is zero because F is zero at infinity. We shall prove that the function F(z) 7i f }a t (t) d t + i C (13.4.3) CO satisfies the conditions of the problem and it is therefore the solution we are looking for. Indeed. the function F(z) defined by (13.4.3) is holomorphic in the superior half-plane because it is a Cauchy-type integral and it is continuously prolongable on the real axis. At infinity it reduces to the constant i C because, if we denote by (a, b) the support of f(x),wehave I-M t f (I f () dt (t) dt /- tl dt. The integral is therefore zero at infinity. Passing to the limit with z x a point from the real axis, and using Plemelj's formulas we obtain: 11(.x..0)+11'(x,0)= f(.1;)+ 11 1 J+x -x ., tf(t)dt+1C. Taking the real part of this relation we obtain (B.4.2). Hence the problem is solved. The real part of the solution (13.4.3), i.e. UI(X. Y) y ,or fa+OC f(t) (t - T)2 + ry2 dt (B.4.4) THE CALCULUS OF CERTAIN INTEGRALS IN THE COMPLEX PLANE. 485 determines the harmonic function in the half-pl.uie y > 0, vanishing at infinity, continuously prolongable on the Or axis and satisfying the condition (13.4.2). This is the solution of Dirichlet's problem for the half-plane concerning the harmonic function u. B.5 The Calculus of Certain Integrals in the Complex Plane At first we shall prove that G 1 a tit t. a (B.5.1) b - tt where the determination of the radical is the positive one for z = r > b, and (a, b) is an interval on the real axis. Indeed, with the mentioned determination, we have x<a. b<x (B.5.2) a<:x<b. rb6 After all. r<n. b<x (). Re(i r-a b..... z a<x<b. Since the real part of the holomorphic function i 1/: - n is known on G the real axis, the function will be determined by the formula (13.4.3). We have therefore Urb-t t.. i a -zzz- b -t--a dt -ya' +C. Considering z oc it result., C = 1 and then the formula (B.5.1). Analogously we deduce that ' f it 1 b t a at dt = -l + JfT b (8.5.3) 486 CAUCHY-TYPE INTEGRALS Passing to the limit and taking into account Plemelj's formulas; (B.5.1) and (B.5.3) we obtain that: 'b fl -a dt i t- x- b :r 1j V a t- t a - b=-t dt (B-5-4) -1 We must notice now that the above considered function i z-b Y YYY reduces to an imaginary constant at infinity (this is i). If we want to apply the same procedure for the function (z - a)(z - b), we must consider the expression (z-a)(z-b)-z. To this aim we have x>b, x<a n, Re [i ( (z - a)(z - b) - z)} = -(x-a)(b-x), a<x<b whence (z - a)(z - b) - z = +- r b (.t a)(b - t) d t + C. The integral becomes zero to the infinite. This imply a+b 2 We have therefore 1 (t - a)(b 7r,"b t-z - t) d t = V"(-z-- a) (z - b) - z + a +b (B.5.5) and, passing to the limit and applying Plemelj's formulas, rrt x 1 L (t - a)(b - t) d t = -x + a + b z (B.5.6) The following example is studied starting from the expression i (z - a)(z - b) ' which vanishes at infinity. Hence, the constant C will be zero. One obtains the integrals rb V a (b - )(t - a) t tz- a)(; - b) (B.5.7) THE CALCULUS OF CERTAIN INTEGRALS t\ '1'flE COMvLI;Y PLANE 487 6 dt = 1 ir Ja f h - t) (t - a) t - (B.5.8) Xt) . We consider now two intervals (a, b), (c, d) on the real axis. The method can be extended to an arbitrary number of intervals, or even disjoint arcs from the complex phuie, in the last case the integral being solved in another way in [A.18], p.88. The result is useful in the study of the grids of profiles. Denoting Q(z) = (z - a)(-- - b)(z - c)(z - d), (B.5.9) for P(z) and R(z) arbitrary polynomials, we have P(..) n<x<b - i li'(z) Q(z) C<x<d. The function from square brackets is holomorphic in the half-plane y > 0. Utilizing the formula (B.4.3), we deduce - R( _ V Q(=) t -r J d tz PQ(t)I t a t4 +C. -Q(t)I (B.5.10) Taking into account. the behaviour of the integrals for z - oc, we deduce that at; infinity we must. have the expansion P(z) Q(Z) R(z) C+ a2+.... (B.5.11) This formula allows to determine R(Z) when P(z) is known. For example, for P(z) = (z - b)(z - d) it results R = 0 and C = 1. and for P(z) = z(z - b)(z - d) it results n+c-b-d 2 CAUCIIY-TYPE INTEGRALS 488 After all, we have the results (b - t))("_t) (1_ (d - t) ° 1 dt (t - b)(d - ') t) d t rt 1 (t. t -z (B.5.12) - (z - b)(z - d} - (z-a)(z-c) r: (b - t)(d - t) 1 rr, _ td t t- 1 +%r l d 1 (t - b)(d - t) td t (t-a)(t-c) t -, (B.5.13) a+c-b-d (z-b)(z-d) Passing to the limit for z x E ((L, b), we apply Plemelj's formulas to the first integrals. One obtains: jd /(t-b)(d-t),t) dt (b-t)(d-t) d t (r- -C.) t-x (t-a)(C-t} t-x}7f 6 1 q b tt JG (b - t)(d - t) td t (t - b) (d - t) td t (tc) t-x 1 (t-a)(c-t) t - -1 x+ -x- a+c -2 b - d (B.5.14) and for z - x E (c,d), employing Plemelj's formulas for the integrals on this interval, we obtain from (1.5.12) and (13.5.13): (t-b)(d-t) dt _ -1 1' J (t-b)(d-t) (t-a)(t- c) s,f J(t_a)(c-t)t_-x +r, t _ t-xdt- 1 ti 1 (t-a)(c-t) t-x + r 7r 1 (b-t)(d-t) dt b ' `t (t-a)(t-c) t-x (b-t)(d-t) tdt a+c - b - d 2 489 t:LAUERT'S INTEGRAL B.6 Glaucrt's Integral. Its Generalization and Some Applications We shall calculate the integral 1 _n f -do e ne f _ 27r f r cos e - 8 os ne cos 9 - s do, (B.6 .1) where s is a real number. For -1 < s < +1 the integral was calculated by Glauert in [3.19]. The general case is considered in [5.38]. The method consists in passing to the complex variable. Denoting exp (i 0) and noticing that we have 1 +" sin no d8=0, , cos B - s because the integrand is an even function, it results 1 27r ++ f x -J=1_tz2-2sz+1 e ne oos9-sd 1 9 /' z"d z iri (B.6.2) z"d z in where lsl=1 (z - 0Z - Y) 1 a=s+ s2-1, Q=s - 82-1. The last integral is calculated with the residue theorem. Since aj3 = 1, It results that the poles of the integrand are situated either one in the interior and the other in the exterior of the circle z = 1, or both of them on the circle. In the case s < -1, the pole z = a is interior and according to the residue theorem, we have: I = 2tr1 " Q'" 1 = (3 + S - 1)" MQ (B.6.3) and for s > 1, the pole z = j3 is interior and it results I. 1)" (B.6.4) At last, in the case -1 < 3 < 1, using the substitution s = cosa, we have a = exp (i o), 0 = exp (-i a) and from the semi-residue theorem [A.221. p.320 (the poles are situated on the integration path) f (z) d z = ai f (zo), z - :w (B.6.5) 490 CAUCIIY-TYPE INTEGRALS we deduce In = an /i" sin na a-,Q jj-rt sin a, Hence, Glauert's formula is 1 r* rr ffff0 Cos no Sill I cos 0 - cos a d 0 sin o n=0,1,2 (B.6.6) Glauert's formula has many applications in aerodynamics. For example, using the substitutions t=c+ecosO, x=c+ecoso, (B.6.7) 2c=a+b, 2r.=b-a, (B.6.8) where one obtains the formulas (8.5.4), (8.5.6), (8.5.8) and other similar ones, like for example Ft- t Jt/---dt=ir(x+e). (B.6.9) b Using the substitutions t=c+ecosO, x=r.+es (8.6.10) and the formulas (B.6.3) and (13.6.4) we also obtain l b t-a dt 71' o b-t t-x 4b-Xx 1, l x-a x - b x<a a<x<b x>b b-x - 1 1 7rI'. b t dt t-a t-x I va - x 1 1 n fL a dt (b--t)(t-a) t - x 1 (a - x)(b - r,) 0. 1 Vf(x etc. x<a 491 OTHER INTEGRALS B.7 Other Integrals In the boundary elements methods one may encounter the following integrals cosne+i sin no 1k -1 (B.7.1) for k = 1, 2. They are also calculated with the residue theorem putting z = ei0. It results sin8=(z-z) ,oos8(z+z), 21 Ii 12 - _ z"d z 2 i b + Jc I:(z - zi)(z - z2) ' 4i (i z"f 1d z rJi:i_1 (z - zt)2(z - z2)2 where, with the notation r = a - b zi _ -a+r b- ic' z2 _ , (B.7.2) , we have. -a-r b- is (B.7.3) Obviously, J--1z21 = 1, such that either a root is in the interior of the unit circle and the other in the exterior or both of them are on the circle. We are interested in the case when a2 > b2 + c2 and a > 0. In these conditions, the root z1 is in the interior of the unit circle, such that, utilizing the residue theorem, one obtains: 71= F(r), 12 = G(r), (B.7.4) where F(r) = 27r(-1)" (b+ic\" G(r) = 2w(-1)" a+nr (b+ic)" r3 a+r r r , M IC) (B.7.5) Separating the real part from the imaginary one, we obtain a series of 492 CAUCIIY-TYPE INTEGRALS integrals, like, for example 2a d0 r A a+bcos0+csin0 I '" a+bcos0+csiuO o f- f I cos 0 r d0 sin 0 cos 20 a+bcos0+csin9 ' =- 2:rr b c d0=-'';r r a+r' rr+bcos0+csin0 'T Jo 2. d 0 = 2;. b-' r (B.7.6) d0 27ra. Jo (a+bcos0+csin9)2 - r=s .3T Jci (a+bcos0+csin0)2 o (a+bcos0+csin0)2 ' d0=-rbi sin 0 2:r J0 - c2 cos 20 (a+bcos9+csin9)2 c -; a} 2r r 2+r1 r2 \d0 b2 - c2 (a+r)2. The case a2 > b2 + c2 and a < 0 is studied in the same manner (in this situation, z2 i3 in the interior of the circle). One obtains Il = F(-r). 12 = G(-r). (8.7.7) If n2 < b' + c2, the two roots are on the unit circle, such that the integrals are calculated by means of the semi-residue theorem. One obtains 21, = F(r) + F(-r), 212 = G(r) + G(-r). (B.7.8) At last, if a2 = b2 + c2, the roots coincide and they are situated on the unit circle. The integral will be calculated using the Finite Part (see Appendix D). If a = -s, b = 1. e = 0 and k = 1, one obtains the re cults from the previous section. _ Appendix C Singular Integral Equations C.1 The Thin Profile Equation The thin profile equation in a free stream has the form 1 'b f (t) d t= h(x), 7rt-x a < x< b, (C.1.1) where 1(t) is the unknown, and h(x) is a given function. It is sufficient to assume that f satisfies Holder's condition on the interval (a, b) for deducing the existence of the principal value from (C.1.1). The solution of the equation also depends on the behaviour imposed to the unknown at the extremities of the interval [a, b]. From the expression of the solution it follows that h also has to satisfy Holder's condition on the interval (a, b). The equation was pointed out by Birnbaum in 1923 [3.5) and solved for the first time by Sohngen in 1939 [A.34]. An ample study devoted to this equation is due to Schroder [A.33]. The solution was found again, with different. methods by many other researchers (Weissinger [3.49], Cheng [A.30], Homentcovschi [A.19], Carabineanu [A.151 etc.). In the sequel we start from the method of C. lacob [A.20]. For solving the equation (C.1.1) we have to determine the function F(z) = u(x, y) + i v(x, y), z = x+iy, (C.1.2) holomorphic in the superior half-plane y > 0, vanishing at infinity and continuously prolongable on the real axis, with the conditions u(x, 0) = 0, for x E (-oo, a) U (b, oo) (C.1.3) v(x,0) = -h(x), for x E (a,b). In order to ensure the uniqueness of F(z) we have to impose its behaviour in a and b. The function F may be bounded or not in these points. 494 S1N ULAR IN1 CRAIG l:QU VI`IONS We w.,sume. for example that F is bounded in b, and it behaves in the vicinity of a like F(_)=0<cr<1. (C.1.4) where F` is bounded We must notice that if we have determined such a function and if we denote by f (x) the boundary values of a on (a, b), then It results that the function f is just the solution of the equation (C.1.1). Indeed, the function F(:), holomorphic in half-plane y > 0 and vanishing to infinity, whose real part on Ox is u =0, for :t E (-nc,a)U(b,x) (C.1.5) it = f (.r.), for :r E (a, b), is given by the formula (8.4.3). i.e. F(z) = i r I (t) t - dt (C.1.6) . Utilizing Pleanelj's formulas %,.-c obtain for ar< :r. < b it (X, +0) + i v(r. +0) = AX) + J t, t ! cl (C.1.7) t.. Separating the imaginary part and taking (C. 1.3) into account.. Ave obtain (C.1.1) for v We shall solve in the sequel the problem (C.1.2) - (C.1.4). Having (C.1.4) in view, we shall consider the function (8.5.2). With the determination of the radical precised there and with the conditions (C.1.3), we deduc c: Re _ tt 1 rE 0.. [r(z) Y-bJ -- T---a b-x a) U (b, ;,c) x E (I a, b) Hence, we may utilize the formula (13.4.3). It results a a1tV) Cit. tt rri F(`) z-Q`-1 the constant C vanishing because I' must v:uaish for z -- x. We deduce rz --b t (C. 1.8) 495 TIIE THIN PROFILE EQUATION Applying Plemelj's formulas and separating the real part we obtain Ax) 1 n bx- a a s _ta b t IL (1) t, - x d t. (C.1.9) This is the solution of the integral equation (C.1.1) which satisfies the boundedness condition in x = b. The solution which is bounded in a and unbounded in b may be -b obtained starting from the function We obtain f(x)=- ' W r-a b 6-x f 7r . 4 - a c v (C.1.1O) (txat. If we wish to obtain an unbounded solution the both end-points, - a)(z - b), we have using the function Re [F(z) e (z - a)(z - b)j = x E (-oo, a) U (b, oo) ol (x-a)(b-x.), xE(a,b) h(x) whence ! f (x) 'b r (x-a)(b-x) Ja ( 1 i (t) d t+ t - a)( b - t ) t-x a (C.1.11) C jr - a)(b - ar) At last, for the solution bounded in the both end-points, we deduce b (z - a)(z - b) 1. (t -a)(b - t) (C.1.12) t t, When z is great enough, this becomes 1 F(") _ i (z - a) (z - b) ri (1-a)"2 (i_)"jb /b Jn h(t) (t - a)(b - t) dt 1- t/ _ - h(t)(1+...)(lt 496 SINGULAR INTEGRAL EQUATIONS such that imposing the zero value for F at infinity we have f h(t) d t = 0. (t - a) (b - t) (C.1.13) If this condition is satisfied, the solution of the equation (C.1.1) (bounded at both end-points), is obtained passing to the limit in (C.1.12). Taking into account the determination of the radical, we find f(x)=-1R (a-a)(b-x)Ja b h(t) (t - a)(b - t) t dt -r (0.1.14) Because of the restriction (C.1.13), this solution cannot be utilized in aerodynamics. C.2 The Generalized Equation of Thin Profiles This has the form: b - Ja (tad t + f (t)K(t - x)d t = h(x) a < x < b, (C.2.1) n where K is a non-singular kernel. We assume that f (x), K(x) and h(x) satisfy Holder's condition on (a,b). This type of equations are encountered in the theory of thin profiles in ground effects, in the theory of grids of thin profiles, in the theory of grids of thin wings etc. From the mathematical point of view, the equations (C.1.5) are extensively studied in (A.27). In order to ensure the uniqueness of the solution we have to impose the behaviour of the unknown f at the extremities of the intagration interval. The method of investigation consists In their regularization, i.e. they are reduced to Fredholm-type equations for which existence and uniqueness theorems are available. For the equation (C.2.1) this may be performed utilizing the solution (C.1.1). If we are interested by the solution of the equation (C.2.1) which vanishes in the trailing edge b, we shall utilize (C.1.9). Assuming that the last two terms in (C.2.1) are known and knowing that a singular integral interchanges with a non-singular integral (A.27), we obtain the following Frcdholm-type integral equation f(x) + ! f f (C)M(t, r)dt = H(x), a < x < b, (C.2.2) 497 THE GENERALIZED EQUATION OF THIN PROFILES where b =x f z) _ -1 t wa k(t - L)dt' aQ x b -1 bz [ t-a _x- s, W t- x t (C.2.3) h(t) -C--t t - x dt. Sometimes we may specify the kernel M . For example, in the case of the thin profile in ground effects, we have: k{t -) = (t -)l + (C.2.4) mz . The integrals t 1_ a t-t dt ) (C.2.5) b-t (t-t)2+m2 t-z a may be calculated with the residue theorem [A.20], or considering the integral _ = f b-a m dt b-t (t-4)2+m t-x 8 and noticing that if we denote S = t + i m, we have: 6 t dt a Fa dt V=t t (t-() l - (dt f n VLb_at Utilizing (B.5.1) and (B.5.4) it results 1+iJ=z*C C-b )b (C.2.6) t-z (C.2.7) whence 1- a J 2 1 S-a+ 1 z-C-b x-S x1 C C- -a -b c -b (C.2.9) 498 SINGULAR INTEGRAL EQUATIONS and con."juently, b= x 1 1ll(f r) = - a+ 1 1 C/z:). C.3 The Third Equation We shall consider the equation [A.1G]: - b 1 J f (t)(In it - xj + ro)d t = g(x) , (C.3.1) defined on the interval (a, b), where f is the unknown and r o is a constant. We intend to determine the general solution and to investi. gate the conditions which are necessary for the solution to satisfv the restrictions f(a) = f(b) = 0. (C.3.2) With the change of variables t, x -. 0, a defined by the relations t = c + e cos 9, x = c + e cosa , where c= a+b -2' e= b-a 2 , (C.3.3) (C.3.4) the equation (C.3.1) becomes 1 - f F(9)(lnelcos0-cosaI +1'o)csinOd0=G(o)7T With the notation (t - a)(b - t) f (t), (C.3.5) 'fF'(0)(lncjcm0-cosOl+l'o)d0=G(a), (C.3.6) F'(0) = eF(O) sin 9 = this may be written as follows 1 where we have 1 coe m9 cos crier . In 2l cos O - cos al = -2 m>1 (C.3.7) 499 THE TN1R1) EQUATION The function F''(6) may be prolonged on the interval (-ir,O) such that the result is an even function on (-pr, +7r) and it may be therefore expanded into a trigonometric series, with the aid of the even functions F' (9) = ao + an cos no, (C.3.8) F` (9) cos nod 0, (C.3.9) n>1 where an = ao r ao = I f,P(O)dOi f F(9)e sin 9d 9= 1Jaf f (t)d t. 7r o Analogously, the function G(a) may be expanded into the series G(a) = bo + E bn aog no, (C.3.10) n>1 where b,, = - rx G(o) coo nod a = - 21 " G,(0,) I sin nad Q , 0 9(x) dx.. (C.3.11) 4 1o (x a) (b - x) a Substituting these series into the integral equation (C.3.6), it results bo G(a)da = 1 A r* ao+E an coo no x 0 n>I x n>1 (_2!cosmOcosmo+F)d9= (C.3.12) m>_ 1 m -aor + n>1 i a. cos no, where we denoted r = ro + In e = ro + I n (b 4 a) . ( C . 3 . 13) Identifying the coefficients we find: I bo = -aor, bn = + an . (C.3.14) 500 SIN(:ULAIt INTEGRAL EQUATIONS In the sequel. we have a) (b - t) g,(t)d T t.. (t 1 u)(b - 1 1 ( l ( fb x 77 sin0 1 tt r Jot COS a -, Cclti 0 _ 1 sin a r 0 COS a - C ci sin0 (C(4)) cos a - cos 0 ( (g)d0= sin n0 d0= 7 cos(n + 1)0 - cos(n - 1)0 d O cos or - Cos 0 n or, utilizing Glauert's integral (B.6.6), i (t.-a)(b-t),(t)dt %b ra t -- x sngit - 1)a Stn(n + 1)a n>1 n>1 -E a,, cosna+-F'(a)+tatt = V12!1 (x - a)(b - x)f(x) - bo/r, = -- (C.3.15) where bo is (C.3.11). From (C.3.15) it results f (s) = - 1 rb y/ (t - (z)(G - 1 (x, - (1) (b t-r X) JJ4 t)9 (t)d t,(C.3.16) _ 1 r1' 1 (x - a)(b - x) 6 g{t) dt. (t -- -a) (b - t) This is the first form of the general solution of the equation (C.3.I). - T11F,ru1RD EQUATION 501 One obtains another form if one utilizes the identity (t - a) (b - t) (x - a)(b - x) (x -- a)(b - x) (t - a)(b - t) (C . 3 . 17) (x-t)(x+t-a-b) (r. - a)(b - x)(t - a)(b - t) Substituting the first ratio in (0.3.16), it results the final solution f(x) (z g'(t) dt + -a)(b-x.) (t-a (b-t) t-x a 1 1 1 -a)(b-a) J g(t)1 I' J [(. + t - a - b)g'(t)- (C.3.18) dt (t-«)(b--t) We notice that from the relations (C.3.9), (C.3.14) and (0.3.11) it results 1 Jn bf(t)dt=TJ b -a)(b-r)clx, ( (C.3.19) which is an useful relation in applications. We shall determine in the sequel the conditions which have to be satisfied by g, such that the relations (C.3.2) are satisfied by the solution (C.3.1). At first we notice that when the parameter r vanishes, a necessary condition for the existence of the solution (C.3.18) is b L g(t) d t = 0. (t - a)(b - t.) (0.3.20) The last integral from (C.3.18) defines for x real a polynomial of the first degree. f vanishes in a, is this polynomial has the root a (the first term from (0.3.18) vanishes for x = a). In this case, the last integral has the order of (x - a), while the denominator of the fraction has the order of (x - a) 1/2. Hence, we must have -",)(b f' [(t - b)g'(t) grt)J (t - t) = 0. (C.3.21) 502 SINGULAR INTEGRAL EQUATIONS Analogously, f (b) = 0 implies fb [(t - u)9'(t) - 9(t) I (t - u)(b - t) 0. (C.3.22) Imposing the both conditions, subtracting and adding, we obtain 91 (t) L b J [(t a 9rt - - tad - c)9 (t) (C.3.23) (t-a)(b-t)dt __ n, ( t/b )1 t) 0. (C.3.24) This is the answer for the proposed problem. C.4 The Forth Equation At least in nragnetoaerodynarnics [3.9] [1.9) p.208, in the theory of oscillatory wings [10.15) and in the theory of the wing in fluids with chemical reactions [3.10) [3.24], it intervenes the following singular integral equation [A.16) fb 1 f(t)dt+- j f(t)(lnt-x)+ro)dt=h(x), a<x<b, (C.4.1) where f is the unknown, ro an arbitrary constant and It, a given function. The solution is unique if one imposes the behaviour of f in one of the end-points of the interval. We shall impose the boundedness of f sa in b. Denoting 1b g(x) = -,- f f(t)(ha it - xI + ro)d t, (C.4.2) a we shall notice that g'(x) = r ad t . t So, the equation (C.4.1) is reduced to the differential equation (C.4.3) a (C.4.4) whose general solution is 9(x) = Aee('-`) +go(x), (C.4.5) 503 THE FOIri'H EQUATION where A is an undetermined constant, c is (C.3.4), and (C.4.6) The unknown f may- be found from the equation (C.4.2) which is exactly the equation (C.3.1). Its solution is therefore (C.3.18) where g will be replaced by (C.4.5). For specifying it, one observes that performing the substitutions (C.3.3), denoting by lo(w), Il (w), .. . Bessel's functions of imaginary argument and taking into account that II(w) = Il (w), one obtains ( CW(t T1 ) a)(b -- t) dt rr" e a , x b T2 '" °d 6 = 7rIo(M) (t - a)(b - t) dt (c+ecosO)e 06Od0 (C.4.7) o = irc1o (w) + rrel i (w) . From the expansion exp(0cos9) = lo(Z3) + (C.4.8) n>l and from Glauert's formula (B.G.6) we deduce e`'(t-c) b dt 1 ' -x - 2-,r e Sill it (T n>1 smo ejjr ewcoo0 cos0-cosad8 (C.4.9) - 2-,r V-r(b E 1,, (0) sin laor - x)(x - a) ,>I where w = ew. For T; and x we have the parametric representation (C.4.9) and (C.3.3). We deduce therefore f(.x)=-'A (x-a)(G-x)7'3+ x [w(x_2c)_]Ti+_ 17 7,- (x-a)(G-x) x --a) (x - (b - x) (C.4. 10) 504 SINGULAR INTEGRAL EQUATIONS fo being obtained from f by replacing g with go. In aerodynamics f represents the jump of the pressure and we need therefore the integrals j rb b f x f (x)d x f (x)d x, (C.4.11) 4 which may be easily calculated without utilizing (C.4.9). Directly, the first integral is obtained from (C.3.18). In the sequel we shall determine the constant A. Imposing the condition f (b) = 0, we have from (C.3.22) and (C.4.5) A [ (aw + ,) Tt - wT2] (C.4.12) = GO, where b Go [(t - a)9o(t) - Tgo(t) (t u)(b t) (C.4.13) Now the problem is solved. C.5 The Fifth Equation In the theory of oscillatory wings [10.20) it intervenes the following integral equation (A.16[: b(tf(t))2dt+ 2 Jf(t)(lnIt-x[+I'o)dt=h(x), a < x < b, b (C.5.1) which has to be solved together with the following conditions f(a) = f(b) = 0. (C.5.2) The sign "' from (C.5.1) shows that. one considers the Finite Part. Introducing the notation (C.4.2) and taking into account the formula (D.3.6), from (C.4.3) we deduce . b o (t - x)2 d t , (C.5.3) a such that the integral equation (C.5.1) reduces to the following differential equation g" - w2g = h. (C.5.4) 505 TILE FIFTH EQUATION The homogeneous equation has the solution g = Aocosh w(x - c) + Bosinh w(x - c) . Applying Lagrange's method of variation of constants, one obtains for the solution of the differential equation (C.5.4), g = Acosh w(s: - c) + Bsinh w(x - c) + go(x) , (C.5.5) where h(E)siuhw(x - E)dt; r(l(x) (C.5.6) The solution of the integral equation (C.5.1) may be obtained from (C.3.18) where one replaces (C.5.5). For specifying it we notice that setting t = c+eu, u E f-I,+1J, we deduce (t - a)(b - t) = e2(1 - u2) and then sinhw(t - c) d t = r+1 sinh (Foil) So n J (t - a)(b - t) 1 d u = 0, (1 -- uu the integrand being an odd function, r6 Coshw(t - c) (t - a)(b - t) n b tcosh w(t - c) C1 (t - a) (b - t) Ja ar. = Cosh Pu} _rdu=glo(w) fJ J1 - u2 1 dt (C.5.7) Yt J Sl /b (c+eu) cosh (Cu) 1-u tsinhw(t - c) (t-a)(b-t) du=crrlo(w) 1 usinh (Ou) dt=e f-i 1-u d u = eirl1( ,.r ) . Also, using the change (C.3.3) and taking into account (C.4.8), Clauert's formula (B.6.6) and the relations (-1)' I,,(w), (C.5.8) SINGULAR INTEGRAL EQUATIONS 506 we deduce ?,s b sinh w(t - c) dt - a)(b - t) r' sinh (iv cos 9) e o cos0-cosy d8 sina (b - x)(x - a) f[ 1 sinner =E [1. (0) e n>1 TC t-x - Eli n>1 cashw(t - c) ner (C.5.9) dt (t-a)(b-t) t-x n sin nay. (b - x) (x - a) n>1[1 - These formulas, together with (C.3.3), give the parametric representation of the integrals T., TT and the variable x. With these results, the solution of the equation (C.5.1) is f (x) _ + (x - a)(b - x)(ATT + BTT)+ (x - a)(b - x)I A(wS1 \\ 11,Co) +wB (x-2c)Co+C1I I, L (C.5.10) where fo is obtained from (C.3.18), replacing g(t) by go(t) given by (C.5.6). As we have already mentioned in the case of the equation (C.4.1), in aerodynamics we need the coefficients (C.4.11). They may be obtained easier utilizing the form (C.3.18). For determining the constants A and B we impose the conditions (C.3.23) and (C.3.24) where g' has the expression (C.5.5). One obtains the system wCOA=Gm, (C.5.11) w(C1- cCo)A+ (wS1 -- r-1Co)B = G2, 507 THE FIFTH EQUATION where '' C1 -L 9-'o(t) dt (t-a)(b-t), (C.5.12) rb G2 = - JJ [(t - 09' W) - l TYO(t)J (t.- a)(b - t) Taking into account (C.5.7), we deduce G1 A- w7rlj1(D) B, r-110(w)[ (C.5.13) In many applications we encounter the situation when h = -c (the case of the flat plates with the angle of attack e). In this situation, from (C.5.6) it results w290(r) = e[1 - cosh w(X - c)), such that G1 = 0, W2G2 = Frr[wl1(i) - r-`lo(ay) + r-.1[ whence, A=0; 13= 4 [1+zr11(0)-IQ(W)1 . (C.5.14) Appendix D The Finite Part D.1 Introductory Notions The notion of "Finite Part" of a improper integral has been introduced by Hadamard in 1923 [A.39], in order to give a significance to the divergent integrals which appear in applications and to utilize them. Hadamard studied integrals having the form b f Ja fW d (b-X)*,+1/lr, (D.1.1) where n = 1,2,3,.... There exists however many integrands with non-integrable singularities which appear in applications especially in aerodynamics. It exists therefore different manners for treating this problem. In the subsonic aerodynamics one utilizes especially the definition of Mangler [A.45], but we had not the possibility to read this paper. A less cited contribution, but very adequate to aerodynamics belongs to Ch. Fox [A.37]. Here, the notion of "Finite Part" appears like a natural extension of the concept of "Principal Value" in Cauchy's sense. We shall present in the sequel some results of this author. For the integrals having the shape (D.1.1) we shall utilize the paper of Heaslet and Lomax [A.44]. These ones appear in the supersonic aerodynamics. Important results concerning the notion may be found in the papers of Kutt [A.42) and Kaya and Erdogan [A.41]. At lasst, the theory of distributions give an unitary method for the study of this notion [A.5]. D.2 The First Integral We shall consider at first the integral Il - fa dx, n=0,1..... (D.2.1) 510 THE FINITE PART If f admits derivatives up to the order n + 1 in the origin, then we may write =Jara [f(x) - (dx I yt fl')(0) + ft'?(o)} d+1+ [.rx)-E o ft,,(0)] -, xi-^ f(i)(0) a i - it = i=0 r=0 = dx it 14) xla + n! In 0 The integrated part for x = 0 becomes infinite. Neglecting these infinite constants, one obtains the so called "Finite Part" of the integral 1. Hence, indicating by an "asterisk" the Finite Part, we have: f(X) xn{i(Ix = in [1(x) - "- t JO dx .n+1 + im0 " f(i)(O) (D.2.2) a' f In) (0) In a. i-0 For f = 1, it results Td'x =lna,I-!+i J » 1 it a (n.> 1). (D.2.3) D.3 Integrals with Singularities in an Interval We shall consider the integrals having the form 12 rb I( X) +idx, 11 =0,1,..., (D.3.1) where a < u < b. For n;-- 0, we consider the "Principal Value" of the integral in Cauchy's sense. ,(J G-E + c-» u \ , x-u d x = lien lb 1L f (x) r1 f (?) d x xu . (D.3.2) 511 I? TECRALS WITH SI`CULARITIES IN AX INTERVAL «Vc know from (B.1.2) that this limit exists if f satisfies Holder's condition in the interval (a, 6). Let us derive now (D.3.2) with respect to the variable u. In the right hand part, the derivation is performed according to the derivation formula for integrals containing the variable in the limits. We have therefore d(,( b d r =1im U C-0 d u Jn X - u f 0x) t. s+ b f (X) Ju+E) (:r - fl):. - d x- -f(u-0 _ f(u+=) or, expanding into a Taylor series the functions f (u - E), f (it + E), , + x (xatd x Tit J en o (.1: L+J (z )2J (D .3.3) If the limit from the right hand part exists. we denote it by Ja T. f 0-) (T-u)2 (D.3.4) d X. and we have f d x def (x - u)2 lim s-'0 l f u-e + fb l C a d rh f(x) dA, J du a X - It a f (X) d (:C - u)2 2--1 x- (Xf(T) -.11)2 dx. (D.3.5) E (D.3.6) The limit (D.3.5) defines the Finite Part of the integral from the left hand part. The Finite Part is a distribution {A.14]. Ex. 11. One proves [A.37{ that if there exists f'(x) on (a, 6) and this function satisfies holder's condition, then the limit from (D.3.5) exists. We notice that this theorem constitutes the extension of the theorem of existence of the limit (D.3.2). We indicate now how one may reduce the calculation of the integral (D.3.1) to the calculation of an integral with a weaker singularity. We consider the case n = 1. Hence, we demonstrate that in the same conditions like above (f' is defined and satisfies Holder's condition on (a, b) ), we have f(T) , (x-n)2 f(u) f(6) a-u 6-it f' , :r - u THE FINITE PART 512 for every of from (a, b). Indeed, employing for the left hand side member the definition (D.3.5) and integrating by parts, we obtain r f () a + ii (x-u)2 fb t+ U -C d x = lim e-'0 CL I- + Zb +t ) a f 2f (u)} XU d a - c f (a) f f (x), d x+ a; x - rt - f (b) + a-u b-u f.b t(x) f da X-u . The extension of the definition (D.3.5) and theorem (D.3.7) to an arbitrary value of n , is performed in (A.37]. In the same paper one gives the respective definitions in the complex plane and also Plemelj's formulas for integrals having the form F(z) 7ri I 21 (t 1(t +I d t (D.3.8) . Utilizing (D.3.6) we may calculate (by derivation) the Finite Part when we know the Principal Value. So, from (D.4.1) it results the integral often used in Appendix C, 1 7r (t-a)(b-t)dt=-i, u<s<b. (t - X)2 (D.3.9) From (B.5.3) we deduce 1`/b ia dt 1 0 (b - t) (t -a) (t - x)2 (D.3.10) , and from (B.6.9), t 1 it Ja t a b - t (t x)2 dt = 1, (D.3.11) and the sequence of examples may be enlarged (see also (A.41]). We also observe that we have day (x - u)2 _ 1 a - it - b - it 1 which may he obtained considering f = 1 in (D.3.7). (D.3.12) 513 HADAMARD-TYPE: INTEGRALS D.4 Hadamard-Type Integrals We shall consider the integrals having the form I3=Ja a (s x)'+tf2dx, n=1,2,... (D.4.1) which intervene in the supersonic flow. As we have already seen in (D.2.1), the basic idea in the definition of the Finite Part consists in leaving apart the infinite values from the structure of the integrals. In order to see the significance of the integral denoted by (D.4.1) in the case n = 1, we shall observe that d ds f' 78 { (s s(x)x d. = sic dds - e) f +J a -C 8 a f (x) s-xJ dxJ a8 . If f is continuous in a, the first term from the right hand part becomes infinite such that we must leave it apart. We shall consider by definition B(x)xl d x = ds. Ja a I s(x) d x (D.4.2) or like in (D.3.6). For f = 1 we deduce dx a (s-x)3/2 - 2 s-a (D.4.4) Let. us prove now that if f is continuous in s and admits bounded first order derivatives on [a, s), then we may give a formula for calculating the member from the right hand part of (D.4.3). Indeed, we have d ds f f(x))s f(a)dlmf(x)-f(a)+ [1(x)-f(s)dx. ifs s - x Ja as ,/ -x 3-x J (D.4.5) But, with the above hypotheses, on the basis of Lagrange's formula f(x) = f(s)+(x-s)f`[s+9(x-s)), THE FINITE PART 514 we deduce that the limit from (D.4.5) is zero. Hence, d f(x) dx= ds / s-x is f(x)-f(s)dx+ ,/Rx dx R-x J = !( s ),. dx --1JNf(x)-f(s)dx- f(s) f 2 a (s - x)3/' (s - x)3/2 (D.4.6) 2 If we also utilize (D.4.4), then (D.4.3) becomes: ' AX) (s - x)3/2 d x= fa f(x) - f(s) dx- 2f(s) s --a (s - x)3/2 (D.4.7) ' this representing the calculation formula for the Finite Part. It is similar with the formula (D.3.5). From (D.4.5) it also results a derivation formula, in fact, the formula analogous to (D.3.7), which reduces the calculation of the integral with a strong singularity to the calculation of an integral whose singularity is weaker with an unity. In the case we had in view (n = 1), the weaker singularity will be integrable. Indeed, taking (D.4.4) into account, we have: ds f" f (r) - f (s) fa s-x d x= d ds f (x) d x-- s-x d (D.4.8) x - 77=7= f (_) lax V-9 - i Noticing now that 8/8R = -8/8x, integrating by parts and taking into account (D.4.6), we obtain: 1f(z) - f(s) 8s s-x dx S^x, d a -1NIf(x)-f(s)]dx (7s l a f (g) Z N dx 797-77 x )dx= +f(a)-f(s)+ s-a +f8 f'(x) dx. s -x. (D.4.9) 515 GENERALIZATION Equating the first members from (D.4.8) and (D.4.9) according to the formula (D.4.5), we obtain: d dx_ f(a) +1fs fI (x) dx,. 8-x 8-at a s - x f(.T) as ds (D.4.10) This is the formula of reduction to a weaker singularity, analogous with (D.3.7). It also proves that the term from the right hand part of the equality (D.4.2) is finite. Analogously one obtains d ds x(x)8dx = I' - f (b)s fx(x)sdx, . +b (D.4.11) The generalization of the definition (D.4.2) is performed as follows I's 88" an [ f(x) s-x ]d, = dn (D.4.12) d8" Q s-x if f is continuous in a, n times derivable and with the derivative of order n bounded in (a, s) (A.40J. Generalization D.5 In the theory of oscillatory wings one may encounter integrals having he shape (D.4.1) where f depends also of 8. We shall establish for these ones the derivation formula (i rb AX'S) dx = - f (b, 8) + J/'b fs (x, 8) + fi (x, 8) {D.5.1) x x-e b-s s Vime from which one obtains (D.4.11) in case that f does not depend on s. d4 Indeed, we have d dsJ, xx-s fdx 1 ds[rf(x's L +f(8,s) x fss's)dr.+ X JJ,I - (D.5.2) sJ But, d bf(--, cly y s,s)dx=\sf(z,8)x fs8,d)+ z (D.5.3) + /'b 8 [f(x1s)_f(ss)1dx8x-8 J THE FINITE PART 516 If f (x, s) is continuous in x = s and admits the derivative f= bounded in the interval Is, b], then, from f(x,s) = f(s,s) + (x - s)f [s,s+d(x - s)] we deduce that the limit from (D.5.3) is zero. After all, changing 8/8s by -a/Ox and integrating by parts, we obtain f f (x, s r d s J,b + Jb ff(x,s) f"[f(x.s) - f (s, s) d - f(s,s))sdx = -f(x,s) - x-s + _ g s' v ad x+ x fb f=(x's)dx+ 777- x-s J3 Xd-f(s,4) j xs b fi(x, s) + fr(x, 9) f (b, s) - f (s, s) +le b-s 1 s)1849 s b e x dx-8sr(s'8) x-s Replacing in (D.5.2) and taking into account that d s TS dx x-s b - s' we deduce (D.5.1). The established formula (D.5.1) shows that the first member is finite. Hence, we may set by definition r O f(x ss ' d ds= ds,f f (D.5.4) d x, for every function f (x, s) continuous in the point (s, s), derivable, with bounded partial derivatives in x and a, where s < x < b. In the general case we shall define I n f(x,8)dx.. dsb f (x,s)dx. (D.5.5) Appendix E Singular Multiple Integrals In the euclidean space with n dimensions E, one considers a domain D bounded or not ( D may coincide with and a function If F(,c) defined on D. We denote t = (fi, ... , r;,,), a = (XI, ... , 1) is unbounded we shall assuine that F tends to zero when Itl -+ 00 in a certain iuanner which will be precised in the sequel. We admit that there exists a point Q(r) in D, such that in every vicinity D. (having the diameter E) F is unbounded. while, in D - DE, F is bounded mica integrable in the usual seas. Then we set E-U D-D, IIDFdC=Urn If this limits exists, it is finite and does not depend on the shape of D,, then the integral will be convergent. Otherwise the integral will be diverrgeiit. We consider now a divergent integral. If there exists a certain shape of D, (for example, sphere or cube) for which the limit exists and is finite (hence it is unique for every sequence of spheres or cubes contracting towards Q). then the integral will be called semi-convergent. The limit (which depends on the shape of DE) will be called principal value of the integral. Utilizing spherical coordinates one demonstrates that the integrals having the form JD'' f (E.2) where r = IC - xi, and is bounded in D, are convergent for a < n and divergent for a > n. The case a = n will be investigated separately. We shall consider the integral having the form V(X) = JD (X, M) where m = r T, (E.3) SINGULAR MIXTIPI.E IN1'E<;RALS 518 where f will be called the characteristic, u will be called the density and x the pole of the integral. The ratio K(x1) = f /r" will be called kernel. All the integrals that we utilize in this book have the form (E.3). The first who studied this type of integrals was Tricomi [A.351 who gives some results in the case a = 2. An ample presentation, which will guide us in the sequel, of the theory of integrals having the forth (E.3), may be found in the books of Mihlin (A.251 and [A.26]. In this presentation. D, will be spheri cal the convergence of the integral will be investigated with respect to this form. We assume that: satisfies Holder condition in D; if D is unbounded, we 10 assume that u(t;) = 0(It;l-t), k > 0. Holder's condition means that there exists two positive constants A and a, 0 < a < 1, such that for every two points 1;1 and 42, front D sa we have ju(f1) - Ajf 1 - 21' ; (E.4) 2° The characteristic f (x, in) is hounded and for a fixed x it is continuous in in. Under these assumptions, we have the following theorem: The necessary and sufficient condition for the existence of the inteyrnl (E.3) is to have ' is f(x,rn)dS=0, (E.5) where S is the surface of the unit sphere centered in x. In order to make the demonstration, we isolate the pole with a sphere included in D, having the radius 6 and the center in x. Obviously, f (r, n)d, 1) 1 Litre <K6 r JII--Ui u(h)f r" (u(E) - u(x)J f (T.t-) d o + u(x) l t o J t<r<6 f (rnm) The first two integrals from the right hand part of the equality are absolutely convergent. In the third one utilizes spherical coordinates relative to the pole x. Since d = r" 1 d rd S, we obtain J Cr" f (x, err) iss f (x, rn) J = In r f f (x. rn)d S. F fs It results the necessary and sufficient condition (E.5). 519 SINGULAR MULTIPLE INTEGRALS if this condition is satisfied, the integral (E.3) may be represented as follows L v.(o r" (E.6) r L - nj u(h)f (x,m)(it + inh Mu(d) - u(x)]f (x,m)dt. r As an application, we shall consider the integral: f 0 J p 1(c,n) (E.7) 1 Ox Rcwhere I? is (5.1.11). With the change of variables q - y = sin 0, the condition (E.5) gives -x= 008 0, 21r X (19 I2-.T 0 = Ro J0 cos 8d 9 = 0. (E.8) Hence, the integral (E.7) exists. The second important theorem that we utilize (for the transonic flow) is the following: If, in addition to the hypotheses 10 and 20, we assume that grad.K(:r. ) = O(r-"° 1), then the singular integral (E.3) (as a function of x) satisfies Holder's condition, with the same exponent like u, in every domain which is bounded, closed and included in D. The theorem was proved for the first time by Giraud in 1934, and for n = 1 by Privalov in 1916. As we can see in (A.251-and (A.261, the demonstration is not simple. The last theorem refers to the derivation of the integrals with weak singularities having the shape v(x) = JD u(h)f (E.9) which lead to integrals of the type (E.3). Like in the previous theory, D may be a domain bounded or not of the space E,,, or it may coincide with the entire space. W e assume that the function f (x, m) is continuous and bounded together with its first derivatives (the first order derivatives with resped to the cartesian codrdiriates of the points x and na). We also assume that u(i:) satisfies Holder's condition and at infinity (if D is unbounded) o('t;-'J), t > 1. Under these 520 SINGULAR MULTIPLE INTEGRALS assumptions, there exist the first derivatives of the integral (E.9), and they are given by the formula axk = ID xk [1t] d , - u(x) j s f(x, m) cos(n, xk)d S, (E.10) where, like above, S is the surface of the unit sphere centered in x, and n, the outer normal to the sphere. Olviously, the first integral from the right hand part of the equality is singular. Appendix F Gauss-Type Quadrature Formulas F.1 General Theorems This appendix relies on the paper [A.49] and it is completed with some results due to Monegato [A.52). Gauss-type quadrature formulas give exact evaluations for the integrals of polynomial functions, multiplied by a weight function w. In aerodynamics it also meet integrals with singularities. Approximating an arbitrary function (according to Weierstras's theorem) by a polynomial function, we may utilize these evaluations. Practically, the approximation is performed by a Lagrangetype interpolation formula. We shall consider in the sequel that to : 1-1, +11 - R is a positive integrable function. THEOREM 1. We have the exact evaluation +i 1 n Af (x.) f (x)w(:c)d x (F.1.1) 0=1 if. 10 f is a polynomial of degree < 2n -- 1; 2e the points x = x,,, a = 1, n are the n zeros of the polynomial P,,(x) of degree n from the orthogonal system of the weight w(x) on [-1, +1j w(x)P,(x)Pj(x)dx = 0, i 0j; (F.1.2) 30 Using the notation Wt) = fP(x)!.Q3dx, (F.1.3) the coefficients A. are given by the formulas Qn(2a) 4a = Pn(xa) . (F.1.4) CAUSS-TYPE QUADRATURE FORMULAS 522 Proof. Taking into account that f is a polynomial of degree 2n-1, and P is a polynomial of degree n, we may write f n a, E 27 -xa + Fn_1, T- (F.i.5) where Fi_1 is a polynomial of degree < n - I. We determine the coefficients a° multiplying (F.1.5) with x - x° and putting x = xa. It results (F.1.6) as = f(z°)/PP(xa) whence n __ f P. f(xa) =1 Pn(X-)(Z - xa) + A, Since dx = J- 11 n( ) )(x ) ( )`vn(xa) a) = and because F,,-, may be written as a linear combination of P0,.. . , Pi_1, such that +1 Pn(x)FF-1(x)w(x)dx = 0, we deduce (F.1.1). THEOREM 2. We have the following exact evaluation: rt J- t x; d x = E A° f (x° n f (x) a=1 . (F.1 (F.1.7) x where j = 1, 2, ..., if 10 f is a polynomial of degree < 2n; 2° the points x = x°t a = T-, n are those defined in Theorem 1; 3° the points t = t,, j = 1;2,..., are the zeros of the function Qn defined by (F.1.3); 40 the coefficients A are defined by (F.1.4). 523 GENERAL THEOREMS Prof. Reasoning like before, we have: f P_E ac. +Fn, X X0, A-1 where the degree of F, < n. It results that as has the form (F.1.6). Setting F = (x - tj)Fn_1(x) + A, where the degree of Fs_1 < n - I, we deduce f P' L-i whence +1 P"(x)f(x0)w(x) LI f (xa Pn(x.)(x-xQ)(x-tj)dx__ '(+1 P,(x)w(x) P., (X-) A P (xQ)(x - x.) + (x - 1 A. (:r.) _ QXQ - tj xQ - tj [_jdx= 1 1 f (zz)Qn(t5) 1n(xa)(xa-tj) _ A. f (x,) xQtj because Q,, (t,) = 0. THEOREM 3. We have the exact evaluation I (f(x) (F.1.9) t where the "asterisk" is for the Finite Part (I).3.5), if.- 10 f is a polynomial of degree < 2n + 1; 20 the points x = x.., a = 1, n are those defined in Theorem 1; 30 The points t = tj, j = 1, 2.... are those defined in Theorem 1; 4e the coefficients An are given by (F.1.4), and A = Qgn(tj) Pn(tj) (F.i.10) Proof. We shall notice at first that, on the basis of the definition (D.3.G), we have for t E (-1, +1), t0adx, (F.1.11) 524 GAUSS-TYPE QUAUR .ATURE FORMULAS whence t Q;,(tJ) = jPn(x)(')2dz. (F.1.12) We shall write m above is PA f - P" LF- (F.1.13) )x-) being a polynomial of degree :5 n - I and B, C, constants. For x = tj it results I(ti) C F& f(3.-*) (F.1.14) 0-1 Fnxa)(tj - xa) P"(tj) Since _ 1 (x - xn)(x - tj)2 2 1 1 tj)2 x (rct 1 x,. 1 1 (z _tj)2 x-tj + tj -xo (x-tj)2 Taking into account that 0 and that Ave have (F.1.12),G we deduce 1+` P"(x)f(x.)w(x) f-1 Pn(x.)(x - X.)(x - t j)2 = f(x.) (x.-tj)2( x-x P (x.) (' +1 P"(x)w(x) 1 A 1 dx= - r. 2- t3/ d x+ ' +1 +J-1 (F.I.15) dx ty-xo (x-ti)2 f(xa) Q.(xrx) + Q;z(tj) P.(x0) (x,, - tj)2 tj - Xw Having in view the definitions of the Principal Value and Finite Part, it results t T'I x - tj fl+' P.(r)u'(x)dx x - tj 4n (tj) = 0 Utilizing (F.1.14) and (F.1.10), we deduce 1Pn(xQ)(tj +! C + _1 P"(x)(xu,(x) - tj) 2dx = AI(tj) - y 1{XQ}'^Ln(tj) - xa) 525 FORMULAS OF INTEREST IN AERODYNAMICS whence it results (F.1.9). The integrals having the form wx ( )n : +1 1 f (x) (x n>2 ' (F.1.16) 1 are studied in [A.49] and [A.52]. F.2 Formulas of Interest in Aerodynamics It is well known that on the interval (-1, +1] Jacobi's polynomials (x) constitute an orthogonal basis with the weight function w(x) _ (1 - x)°(1 + x)' ( see, for example, (A.561). Obviously, the zeros xo and t) from the theorem--, from F.1 do not depend on the factor of normalization of the polynomial P,,. One can simplify this factor in A,, from (F.1.4) and A from (F.1.10). We shall utilize therefore Jacobi's polynomials without the constant factor. 1°. For the weight w(x) = (1-x2)-1/2, Jacobi's polynomials reduce to Chebyshev's polynomials x = cos9. TT(x) = cosn9, (F.2.1) F o r 0 < 0 < jr, the polynomials T. (n = 1, 2, ...) vanish when 9° - 2a-1 w , n 2a-lir ' a = Tn. xQ_ - oos 2 n Utilizing the notation t = coax 2 and Glauert's integral (B.6.6), we deduce: sin or (F.2.3) Qn(t) = x sin or with the zero a,=? , tj=cos r(x) _ - Since 1 -1-,n n (F.2.4) sinn9 sin9 dTn sing dO from (F.1.4) it results that A. = 7r/n. Since n(t)) V (F.2.2) d sin na - -slam do sino' = nir cos ja cf sill2 526 GAUSS-TYPE QUADRATURE FORMULAS from (F.1.10) it results that A = -n7r/(1 - t). We deduce therefore the following formulas: tt 411 f(x) dx = T Ef(xn), it 1-x 0 r J- i +I +1 f (x) dx 7r c- f (x, ) y:"]X-tj dz f(x) (x _ tj)2 1-Z J-1 (F.2.5) 1 n i-. (F.2.6) n.Lt f (v) tt7r (xa - t;)2 1 -- t=f(tj) (F.2.7) for j = -, n- 1, where .r are given by (F.2.2) and tj, by (F.2.4). 211. For the weight function w(x) = (1- x2)1/2, Jacobi's polynomials reduce to Chebyshev's polynomials of second order sin( e1)9 sin Cyr + e" I' 1 x= xa - coy re[+ 1 ' (F.2.8) a = 1, nrt . (F.2.9) Using the notation t = cos a and Glauert's integral (B.6.6), we deduce: -ircos(n j- I)a (P.2.10) whence 2, j=1,ri+1. tj=cvs't (F.2.11) 1 Utilizing the formulas given in the theorems F.1 for A,, and A, on the basis of Glauert's integral, we obtain: - n4-1 (1 -" xa), A = -r (n + 1) . Hence, the following quadrature formulas are established: n +' 1-1`.f(x)dx=n+1E(I- +1 xt f(rn), (P.2.12) es=1 f(r.) x -tj dx= 7r I - xp f( :ra ), -t; (F.2.13) 527 FORMULAS OF INTEREST IN AERODYNAMICS +1 1- xa ir f (x) 1-z (x-tj)2dx=n+1 J-1 (xQ-tj)2f(xo) awl (F.2.14) -ir(n + 1)f (tj), where j = 1, n + I, xa are given by (F.2.9) and tj by (F.2.11). We notice that the numbers tj given by (F.2.4) are the zeros of the and the numbers ti given in (F.2.11) are the zeros of the polynomials Tn i(t). x)1/2(1 + x)''/2, Jacobi's 3°. For the weight function w(x) polynomials polynomials are [A.56]. Pn(x) = with the zeros sinj(2n + 1)8/2] _ 2aar x = cos 0, , a=1-,n. (F.2.15) (F.2.16) With the aid of Glauert's integral, we deduce /2j' -,'[(2n Qn(t) = cos( CO6C /Z) whence 9=l . (F.2.17) Utilizing (F.1.4) and (F.1.10), it results: A. = 2n 2n+1(1-xa), A _ 2n+1x 1+tj 2' Hence, we established the following formulas: +1 J- 1 r+i . +1 _0 2w fin 1f(x)dx 2n+1` a-i J - x f (x) _ (1-x.)f(x.), 27r +x x-tjdx 2n+1 a-i -a 1-z f(x) -I V 1+x (x - tj)2 n (F.2.18) (F.2.19) 13 dx= _ (F.2.20) j 2f(tj) 2n+1r(Q-tj)2f(x.)- + 528 GAUSS-TYPE QUADRATURE FORMULAS j = 1 n, for being given by the formula (F.2.16) and tj by x,, (F.2.17). 40. For the weight function w(.r.) = (1 - x)'"112(1 + r)h/2, Jacobi's polynomials are 'n(s) with the zeros cos((2n + 1)9/2J cos(0/2) 2a -1 r x = cos9, o= ln. 2n+1T, (F.2.21) (F.2.22) Utilizing Glauert's integral, we deduce: Qn(t) _ sin[(2n+ 1)Q/2] s in(a/2) , t=CADS a, polynomials which have the zeros t'- (F.2.23) 2n+1' On the basis of the formulas (F.1.4) and (F.I.10) we obtain 2'7r A"2n+1(1+xn), 2n + I An=-1-t, ar such that one establishes the following formula, +1 J 1+x 1 ' n L(1 +x,,)f(F.2.24) f (x)dx = 2n + 1 27 J1 + x f (x) 21r 1-xx-tjdx=2n+l 1 + x0 1+x ar 2n+1 it 1-t; ?f(tJ) s"-tjf(n), (F.2.25) 1 f 1--x {x-tJ)2 (F.2.26) tar n 2n+I owl (ca-tf)-f(x°)' for j = 1, n, the zeros xa being given by the formula (F.2.22) and tj by the formula (F.2.23). We have to notice the relations between the formulas from 30 and 40. The points x,, from 3° coincide with the points t1 from 40, and x,,, from 40 with t,, from 3°. 529 TILE MODIFIED MONECATO'S FORMULA F.3 The Modified Monegato's Formula It is preferable sometimes to replace the formula (F.1.9) which contains the numbers f (t;) by the formula given by Monegato (A.52] p. 279. t1,(r) f(r) dx=> (F.3.1) (x - t)2 J-1 where Q1. (Z.) - Q. (t) - Q/n(t)(x° - t) Pn(x°)(x° W (F.3.2) , the formula containing only the numbers f (x°). The formula (F.3.1) is exact, i.e. Rn (f) = 0. if f (x) is a polynomial of degree n - 1. In fact, in applications one utilizes not the formula (F.3.1) [5.10), [6.5], but another one which may be obtained as follows. In (F.3.1) we isolate the term corresponding to a = j and we pass to the limit, +1 t -. We find n, ut(x) (x f(x) dx = E wa' (xj)f(x°)+ - xj)2 °=1 +wj(xj)f(xj) + Rn(f), (F.3.3) . where the mark at F, means. that one excepts the term corresponding to a = j. The factor w'°(xj) is obtained from (F.3.2). Using the rule of I'Hospital u J(xi) we find that q., (xj) (F.3.4) In applications one utilizes (F.3.3) for w(x) = (1-x2)1/2. The numbers xo are given therefore by (F.2.9). From an elementary calculation it results Qn(s'a) = -7r(-1)° , Q'n(xj) = 0, + 1)(-1)° X'2 , Qn(xj) = 70 + 1)2 01 s 1-x12 530 GAUSS-TYNE QUADRATURE FORMULAS whence f(x) +' 1 _-2 E [1 (-1)j] (xa - 0=1 2f(:rn) - n2 1f(xr)+ R,,. (F.3.5) This is the modified formula. From f d_ +t w(x) 1.-t uwa(t)f (xo) + Rn(f) a (F.3.6) a=1 where (A.50J page 275, ww(t) Rn (xo) Qn(T ) - Q.t (t) (F.3.7) , isolating in E the term for which a = j and passing to limit t --+ xj, one obtains the formula ry1 -t Tl u w(x) f (x) d x = x - xj ',)f (.r.)+ owl xa) (F.3.8) f(x ' For the weight function w(x) = (1 - x2)1/2 we find: +t 1 n 1 - x2 1(x) cl x = amt n+1 2 f( X( -xj x° ) (F.3.9) F.4 A Useful Formula We shall establish the following series expansion: = -2E(j + 1)Uj(y)Uj(ri) 1 (n X1)2 1 (F.4.1) 531 A USEFUL FORMULA where U (x) are Chebyshev's polynomials (F.2.8). The series is divergent, but it has a first Cesaro finite sum: n (tl -2 1 y)2 I U + 1)Ui (y)Ui(q) . (F.4.2) Indeed. setting q = ooeO and y = oosa on the basis of Glauert's integral, we deduce r+1Ua(q)dq=-aoos(n+1)o, (F.4.3) and deriving and taking (D.3.6) into account, we obtain the formula J (q - y)z d q = -w(n + 1)U*(y). (F.4.4) 1 Expanding now the function (q - y)'2 on the interval [-1, +11 in a Chebvshev-tvpc polynomial series, we have: 1 aJ(y)Uitq). (F.4.5) (q - y)" = Taking into account (P.4.4), and the orthogonality condition 2! UJ(1)Uk(q)dq = 10 JiAk J7r ,1 = k, (F.4.6) we obtain: a,(y) = -2(j + 1)UJ(y). Replacing it in (F.4.5), we obtain the formula (F.4.1). (F.4.7) Bibliography Chapter 1 [1.1) H. Ashley, M. Landahl: .4er dynamics of wings and bodies, Publ. Co.. Readings, Mass.. 1965. 11.21 R.L. Bisplinghoff, H. Ashley, R.L. Ilalftnan: Aervelasticity, Addison-Wesley Publ. Co., Inc.. Cambridge, Mass., 1955. [1.3) R.L. Bisplinghoff, H. Ashley: Principles of Acroclasticity, J. Wiley, New York, London, 1962. (1.4) P.F. Byrd. H.D. Friedman: Handbook of eliptic internals for engineers and physicists, Springer, Berlin, 1954. (1.5) E.Carafoli, V.N. Constantincscu: Dinarnica fluidelor comprrsibile, Ed. Acadetniei, Bucureeti, 1984. [1.6] R. Courant: Partial differential equations, New York, London. (1.7) G.G. Chernyi: Gas Dynamics (in Russian), Moscow, 1988. [1.8) E.H. Dowell, H.C. Curtis .R.H. Jr., Scanlan and Sisto: A modern course in aeroelasticity, Sijthotf and Noort.dhoff, the Netherlands, 1978. [1.9) L. Drago§: Magnetofluid dynamics. Ed. Acadetniei. Bucharest, Abacus Press, Tunbridge Wells. Kent, 1975. [1.10) L. Drago§: Principiile rnecanieit mediilor continue, Ed. Tehnicti, Bucureeti, 1983. [1.11) L. Drago§: Afecanica fluidelor. I. Editura Academiei Romine. Bucure.Iti. 1999. [1.12) W.F. Durand (Editor): Aerodynamic theory, Springer, Berlin, 1935. (1.13) Van Dyke: Perturbation methods in fluid mechanics, Academic Press, N.Y., 1964. 534 BIBLIOGRAPHY [1.14) F.L. Frank!, E.A. Karpovich: Gas Dynamics of slender bodies, Moskow, 1948. 11.15) J.P. Gilly, L. Rosenthal, Y. Seutezis: Aerodynamique hypersonique, Gauthier Villars, Paris, 1970. (1.161 I.S. Gradthteyu, I.N. Ryznik: Tables of integrals, series and prod- ucts, Academic Press, New York, 1965; in Russian 5"' edition Moskow, Nauka, 1971. 11.171 W. Grobner, N. Hofreiter: Integrnltafel, Ted II, Bestinite integrale, Wien, Innsbruck, Springer-Verlag, 1958. [1.181 L. Howarth: Modern developments in Fluid Dynamics, High Speed Flow I, Oxford, 1953. [1.19] W.R. Hayes, R.F. Probstein: Hypersonic flow theory, Academic Press, New York, 1962, 1966. Intrnducere: maternatic6 in mecanica fluidelor, Ed. Academiei R.P.R., Bucureti, 1952. 11.201 C. lacob: [1.211 C. Iacob: Introduction mathernatique a la mecanique des fluides, Ed. de l'Academie de R.P.R., Bucharest, Gauthier-Villars, Paris, 1959. [1.22] C. lamb: Matematici clasice fi nuxlerne, Editura Tehnici , Bu- cure§ti, Vol 2, 1979; vol 3, 1981. [1.23] Janke-Emde-Li sch: Tafeln hohcrcr funk-tionen, B.G. Toubner, Stuttgart, 1960. [1.241 W.P. Jones (Gen. Ed.): Manual on Acroelasticity, vol. 11, NATO AGARD. [1.25] K. Karauncheti: Principles of ideal fluid aerodynamics, John Wiley and Sons, Inc. N.Y., 1966; Krieger, Malabar, FL, 1980. and J.M. Burgers, General Aerodynamic theory(1.26] Th. von Ks perfect fluids, in [1.11). (1.27) R. von Mises: Mathematical theory of compressible fluid flow, Academic Press, N.Y. 1958. [1.28] J. Katz, A. Plotkin: Low-speed aerodynamics (frmn wing theory to panel metho(ks), Me Grow-Hill, N.Y., 1991. 535 BI 3LIC)C.RAPF1Y 11.29] I.K. Lifanov: Singular integral equations and discrete vortices, VSP. Utrecht. The Netherlands. 1996. F. Oberhettinger. R.P. Soni: Formulas and thearerns for the special functions of Mathematical Physics, third Ed., [1.301 W. Springer-Verlag. Berlin, 1966. [1.31] L.M. Milne-Thomson: London, 1949. Theoretical hydrodynamics, 11Ic Millan, [1.32] F. Oberhettinger, L. Radii: Tables of Laplace Transforms, Springer-Verlag, Berlin Heidelberg, New York, 1973. 11.33] A.N. Pancenkov: )xxlvodnovo kr-dla, Kiev, Naukovo Duneka, 1965. [1.341 A.N. Paneenkov: Teorija potenjiala. uskorenii, Nauka, Novosibirsk, 1975. [1.15] A.P. Prudnikov, Yu.A. Brychkov, D.I. Maricev: Integrals and Series (in Russian), Nauka, Moskow, 1981; in English, Cordon and Breach Sci. Publ.. New York. 1986. [1.361 H. Schlichting, E. Truckenhrodt: Aerodynarnik der F ztgzcnge, Springer, Berlin. 1960. 11.371 W.R. Sears (ed): General theory of high speed aerodynamics, in High Speed Aenxlynantics and Jet. Propulsion, vol. 6, Princ. Univ. Press, 1959. [1.38] L.I. Sedov: Plane problems of hydrodynamics and aerodynamics (in Russian). Ed. 2, Nauka, Moskva, 1966. 11.391 J. Serrin: Mathematical principles of classical fluid mechanics, in Handbuch der Phystk, VIII/1, Springer, Berlin, 1959. [1.40] B. Thwaites: Incompressible aerodynamics, Oxford Univ. Press, Oxford, 1960. [1.41] A.N. Tihonov. A.A. Samarski: Ecuaeiilc fizicii matematice, Ed. Tehnica. tiiniifica .ji Enciclopedicu, Bucureiti, 1956. [1.421 C.N. Watson: A treatise on the theory of Bessel functions, 211d ed., Univ. Press Cambridge. Mac Millan Co., New York. 1945. 11.431 J. \Vcissinger: Theoric des Tmgfiiigels in Handbuch der Physik, B 8/2. Springer-Verlag, Berlin 1963. 536 13IBL100-11Al'ltY [1.44] J. Zierep: Vorlesung uber theoretische Gasdynamik, Verlag, G.Braun, Karlsruhe, 1962. Chapter 2 [2.1] H.R. Aggarwal: Derivation of the fundamental equation of sound generated by moving aerodynarnir surfaces, AIAA Journ. 21(1983) 1048; Reply by the author to F. Farassat, idein 22(1984) 1184. [2.2] R. Belasubranianyam: Aerodynamic sound due to a point source near a half plane, IMA Journ of Appl. Math., 33(1984) 71. [2.3] A. Barsony-Nagy, M. Hanin: Aerodynamics of wings in subsonic shear flow, AIAA Journ, 20(1972) 451. (2.4] A. Barsony-Nagy, M. Hanin: Aerodynamics of wings in supersonic shear flow, AIAA Journ, 22(1984) 453. (2.5) P. Ciobotaru: Asupra ecualiilor fluidelor in distribuf ii, Studii §i Cercetbri Mateinatice, 35(1983) 173. [2.6] L. Drago§: Fundamental matrix for the equations of ideal fluids, Acta Mechanica, 33(1999) 163. [2.7] L. Dragon: Fundamental solutions for ideal fluids in uniform motion, Quart. Appl. Math., 36(1979) 401. [2.8] L. Drago§, D. Honientcovschi: Stationary fundamental solution for an ideal fluid in uniform motion, Zeitschr. f. Angew. Math. a Mech., 60(1980) 343. (2.9] L. Drago§: The fundamental plane matrix in the uniform flow of the ideal compressible fluids, An. St. Univ. "A1.1. Cuza", I+i, Sup] t. 27(1981) 143. [2.10] L. Dragoq: Fundamental solutions in linear aerodynamics, Rev. Roum. Math. Pures et Appl., 27(1982) 313. [2.11] L. Dragog: Fundamental solutions of the linearized Steichen equation, Rev. Roum. Sci. Teciin., M 6c. Appl. 30(1985) 173. (2.12] F.C. Friedlander: Sound pulses, Cambridge Univ. Prms, 1958. [2.13] H.F. Goldstein: Acroacoustics. Mc Graw-Hill, New York, 1980. BIBLIOGRAPHY 537 12.141 M iidaiu, A. Barsonv-Nagv: Slcndkr frinq the.onj for nnnunifonn stream, AIAA Journ. 18(19$O) 381. [2.15] D. Homentcovschi, A. Barsonv-Nagy: A linearized theory of three - dimensional air foils in nonuniform flow, Acta Mechanica, 24(1976) 63. [2.16] M.J. Lighthill: The fundamental solutions for small steady three dimensional disturbances to a two-dimensional parallel shear flow, J. of Fluid Mech., 3(1957) 113. [2.17] D.S. Jones: Aerodynamic sound due to a source near a half plane, J. Inst. Math. Appl. 9(1972) 114. [2.18] M. Morse, K.U. Ingard: Theoretical Acoustics, Mc Graw-Hill, New York, 1968. [2.19] N. Rott: Das Feld einer raschbewcg Schall (The field of a mottin.g sound source), Milt Inst. f. Aerod. RTH (1945) nr. 9. [2.20] P.N. Shankar, U.N. Sinha: Weakly sheared two-dimensional inviscid flow past bodies, J. Mecaniyue, 19(1980) 125. [2.21] L. Sirovich: Initial and boundary value problem in disipative gas dynamics, Pliys. of Fluids, 10(1967) 24. [2.22] C.S. Ventres: Shear flow aerodynamics. Lifting surface theory, AIAA Journ. 13(1975) 1183. Linearisierte Profiltheorie bei ungleichfonniger Anstromung. Teil I, Unendlich diinne Profile, Acta Mechanica, 10(1970) 207; Ted 11, Schlanke Profile, Idein 13(1972) 133. [2.23] J. VVeissinger: Chapter 3 [3.1] J.A. Bagley: Pressure distribution on two-dimensional wings near the ground, RAE Rep.Aero 2625(1960). The unsteady motion of twoBasu, G.J. Hancock: dimensional hydrofoil in incompressible inviscid flow, J. of Fluid Mech. 87(1978) 159. [3.2] B.C. [3.3] A. Betz: Singularitatenver f ahren zur Bcstimmung der Krafte and Momente auf Karper in Potential stri mungen, Ingen: Arclt, 3(1932) 454. 538 [3.4] BIRLIOGRAI'll Y A. I3etz: Applied airfoil theory, Section .1 in [1.12], vol. 4. [3.5) W. Birnbaum: the trugcnde Wirbelftiiche As Hilfsinittel zur liehandlung des cbenen Problems der T'rngflugelthcoric, Z. Angew. Math. u. Mech. 3(1923) 290. [3.6) W. Birnhautn: Das abene Problem des schlagenden Fluyein, Z. Angew. Math. u. Mech. 4(1924) 277. [3.7] A. Carabineanu, L. Dralo§, 1. Oprea: The linearized theory for the submerged thin hydrofoil of infinite span, Arch. Mech., 43(1990) 25. [3.8] Inviscid analysis of two-dimensional D.H. Choi, L. airfoils in unsteady motion using conformal mapping, AIAA Journ. 28(1990) 2025. [3.9) L. Drago§: On the motion past. thin airfoils of incompressible fluids of incompressible fluids with conductivity tensor, Quart. Appl. Math., 28(1970) 313. [3.10] 1.. Dragofl: On the. theory of thin airfoils in nonequrlibrium ideal fluids, ZAMM 64(1984) 17. [3.11] L. Dragoei: Method of fundamental solutions in plane steady linearaerodynamics, Acta Mechanica 47(1983) 277. [3.12) L. Drago§: On the integration of equation of thin airfoils by quadrature formulae of Causs type. Bull. wath. Soc. Sci. Math. Route., 34(1990) 227. [3.13] L. Drago§: A numerical solution of Me: equation of thin airfoil in ground effects, AIAA Journ. 28(1990) 2132. [3.14) L. Dragog: Subsonic flow past thin airfoil in wind tunnel, Mech. Res. Comm. 18(1991) 129. [3.15] L. Drago§: New methods in the theory of subsonic flow 1xist thin airfoil configurations, Rev. Roust. Sci. Techn. Sen. M- c. Appl., 35(1990) 179-196. [3.16] L. Drago§, N. Marcov: Exact solutions of the problem. of incom- pressible fluid flows with circulation past a circular cylinder in ground effect, Rev. Rouen. Sci. Techn., Ser. We. Appl., 38(1993) 25. BIBLIOGRAPHY 539 13.171 D.A. Efros: Hydrodynamic theory of two dimensional flow with cavitation Dokl. Akad. Nauk SSR 5(1946) 267. [3.18J H. Garner, E. Rogers, W. Acu!n and E. Maskell: Subsonic wind tunnel wall corrections, AGARDograph 1966, 109. 13.19] J.P. Giesing. A.M.O. Smith: Potential flow about two-dimensional hydrofoils, J. Fluid Mecli. 28(1967) Part 1, pp. 113. [3.20] D. Gilberg and J. Serrin: Free boundaries and jets in the theory of cavitilion, J. Math. and Phys. 29(1950) 1. [3.21] H. Glauert: The elements of acrofoil and airscrew theory, Cainbridge Univ. Press 1926 (21d ed. 1947). [3.22] 11. Glauert: The effect of compressibility on the lift of an aerofoil, Proceed. Roy. Soc., A, 118(1928) 113. 13.231 N.D. Halsey: Potential flow analysis of multielement airfoil using conformal mapping, AIAA Journ 17(1979) 1281. [3.24) D. Honientcovschi: Non-equilibrium flow of an inviscid gas past a thin profile, ZAMM, 57(1977) 461. [3.251 W.H. Isay: Kavitation, Schiffahrst-Verlag "Hansa" Hamburg 1981. mapping including multi[3.26] D.C. Ives: A modern look at ple connected regions, AIA Journal 14(1976) 1006. [3.27) R.M. James: Analytical studies of two-clement airfoil systems, Douglas Aircr. Comp. Rep MDC JS 831, 1974; A new look at two-dimensional incompressible airfoil theory, Dougl. A. Comp. Rep MDC JO 918/01.1971. [3.28] D.S. Jones: Note on the steady flow of a fluid past a thin aerofoil, Quart. J. Math. 6(1955) 4. [3.29] M.J. Lighthill: A Mathematical method of cascade design, Rep. Memor. aero. Res. Coin. London. 2104(1945); A new method of two-dimensional acnxlynamic design, Idem 2112(1945). [3.30] M.J. Lighthill: A new approach to thin air foil theory, Aeronaut. Quart. 3(1951) 193. [3.311 L.C. Malvard: The use of nceo-electrical analogies in certain aerodynamic problems, J.R. acro Soc. 51(1947) 734; Recent develop- 540 131131.10(:Rt1I'IIY inents in the method of the rheeo-electric analogy applied to aerodynamics, J. Aero. Sci. 24(1957) 321. [3.32] N. Marcov: Ali§cari poten(iale ale flutdclor ideate i incompresibile in prczen;a a doud obstacoic cirrulare, St. Cerc. Mec. Apl. 55(1996) 29. [3.33] M. Mokry, D.J. Peake. A.J. I3owker: Wall interference on twodimensional supercritical air foils, using wall pressure measurement to determine the porosity factors for tunnel floor and ceiling, High Speed Aerodynamics Laboratory, L.H. Ohman. [3.34] M.M. Munk: General theory of thin wing sections, Rep. Nat. Ads'. Coin. Aero., Washington 142(1922). [3.35] T. Nishiyaina: Hydrodynamical investigation in the submerged hy- drofoil, ASNE pI, August 1958, pp. II Nov. 1958, pp. III Febr. 1959. [3.36] A.N. Pancenkov, V.M. Iveenko: Zadaci i metodii ghidro - dinarniki podvodnih 1-rilov i vintov, Naukova Dumka, Kiev, 1966. [3.37] A. Plotkin, C. Kennel: Thickness-induced lift on a thin airfoil in ground effect, AIAA Journ. 19(1981) 1984. [3.38] A. Plotkin: Thickness and camber effects in slender uriny theory, AIAA Journ 21(1983) 1755. [3.39] A. Plotkin, S.S. Doubele: Slender wing in ground effect, AIAA Journ, 26(1988) 493. [3.40] L. Prandtl: Uber Strornungen deren Geschurindegkeiten mit dcr Schallgcschwindigkeit vcrghcichbar sind, J. Aero. R.es. Inst., Tokyo University 63(1930) 14. [3.41] E. Reissner: On the general theory of thin airfoils for non-uniform motion, Tech. Notes NACA. Washington 946(1944). [3.42] M. Shiffman: On free boundaries of an ideal fluid 1, Comm. Pure & Appl. Math. 1948, pp. 89. [3.43] M. Shiffman: On five boundaries of an ideal fluid 11, Comm. Pure & Appl. Math. 1949, pp. 1. [3.44] S. Tomotika, T. Nagamiya, Y. Takenouti: The lift on a flat plate placed near a plane wall with special reference to the effect on the 541 BIBLIOCRAPHY ground upon the lift of a monoplane aerofoil, Rep. Aero Res. Inst. Tokyo 97(1933). (3.45) S. Tomotika, Z. Hasimoto, K. Urano: The forces adding on an aerofoil of approximate Joukowski type in a stream bounded by a plane wall, Q.J.Meeh. Appl. Math., 4(1951) 289. [3.46] E.O. Tuck: Steady flow and static stability of airfoils in extreme ground effect. J. Engng. Math. 15(1981) 89. (3.47] J.S. Uhlinan, Jr.: The surface singularity method appl. to partially cavitating hydrofoils, 511' Lips Propeller Sypp. Drunen, The Netherlands, May 1983. [3.48] B.R. Williams:An exact test case for the plane potential flow about two adjacent lifting airfoils, RAE Tech. Rept. 71,1971. [3.49] J. Weissinger: Ein Satz uber Fourierreihen and seine Anwendung auf die Tragfli yeltheorie, Math. Zeitschr. 47(1941) 16. 13.501 L.C. Woods: Generalized aerofoil theory, Proc. Roy. Soc. A, 238(1956) 358. Chapter 4 (4.11 P. Bassanini, C.M. Casciola, M.R. Lancia and R. Piva: A boundary integral formulation for the kinetic field in aerodynamics, Part.I. Mathematical Analysis, Eur.J.Mech.B/Fluids, 10(1990) 605: Part.II. Application to unsteady 2 D flows, idem 11(1991) 69. [4.2] C. Bindolino, P. Mantegazza and L. Visintini: A comparison of panel methods for potential flow calculation, Proc. of 8th' Nat. Sympos. AIDAA (1985) 59. 14.31 C.V. Canip: Direct evaluation of singular boundary integrals in 2D biharrnonic analysis. Int.J.Nuuner Methods Engng. 35(1992) 2067. [4.4] A. Carabineanuu, A. Dinu: The study of the incompressible flow past a smooth obstacle in a channel by the boundary element method, Rev. Roum. Sci. Tech.-&W. Appl. 38(1993) 601. [4.5] V. Cardr : A method with boundary elements t. the study the subsonic floras. ZAMM 75(1995) 483. 542 BIBLIOGRAPHY [4.6] L. Drago.j: Boundary Elemcnt Alethods (BEM) in the theory of thin airfoils, Rev. Rouen. Math. Purees et. Appl., 34(1989) 523. [4.7] L. Dragoj, A. Dinu: Application of the Boundary Element Method to the thin airfoil theory, AIAA Journ. 28(1990) 1822-1924. [4.8] L. Dragei, A. Dinu: Application of the Boundary Integral Equations Method to subsonic flow last thin smooth airfoil in wind tunnel, Proceed. of the 2`1 Nat. Conf. on Boundary and Finite Elements 13-15 May 1992 Sibiu, pp. 75-84. [4.9) L. Drago, A. Dinu: The application of the Boundary Integral Equations Method, to subsonic flow with circulation last thin airfoils in wind tunnel, Acta Aiechanica 103(1994) 17-30. [4.10] L. DrLgo§, A. Dinu: A direct Boundary Integral Equations Methods to subsonic flow with cirt:ulation past thin airfoils in grnnnd effect, Comput. Methods Appl. Mech. Engrg., 121(1995) 163. [4.11] L. Drago§, A. Dinu: A direct Boundary Integrrl Method for the two-dimensional lifting flout, Arch. Mech., 47(1995) 813. [4.12] L. Dragoq, N. Marcov: Solutions of the boundary integral equations in the theory of incompresstble: fluids, Analele Univ. Bucurerit i, 46(1997) 111-119. [4.13] L. Dr go§, A. Dinu: A boundary regularized integral method for plane potential flow past airfoil configumlions (in press). [4.14] .Ciesing: J.P. Potential flow about lwo-dimensional airfoils. Dou- glas Aircraft Comp. Rep. No LB 31946, May 1968. [4.15] C. Griello, G. Iatmelli and L. Tulumello: An altci7hativc boundary element method approach to the 2-D potential flow problem around airfoils, Eur..l.Mech.B/Fluids 9(1990) 527. [4.16] J. Hess, A.M.O. Smith: Calculation of potential flows about arbitrnry bodies, Progress in Aeronautical Sci., 8, Ed.D. Kiichemanu, Pergamon Press, N.Y. 1967. 14.171 J. Katz, A. Plotkin: Low-slx:ed aerodynamics. From wing theory to panel methodls, McCrow-Hill Inc., N.Y., 1991. [4.18) S.V. K(Yzlov, I.K. Lifanov, A.A. Mikhailov: A new approach to mathennatical modelling of flow of ideal fluid around bodies, Sov. J. Numer. Anal. Math. Modelling, 6(194)1) 209. BIBLIOGRAPHY 543 [4.19] M. Mokry: Integral equation naethod for subsonic flow past airfoils in ventilated wind tunnels, AIAA 3.13(1975) 47. [4.20] L. Morino. C.C. Luo: Subsonic potential aerodynamics for complex configurations, A general theory, AIAA J.12(1974) 191. [4.21] S.W. Zhang. Y.Z. Chen: New integral equation in the thin airfoil problem, Acta Mech.87 (1991) 123. Chapter 5 [5.1] H. Ashley. S. Windnall and M.T. Landahl: New directions in lifting surface theory, AIAA Journ., 3(1965) 3. (5.2] S.M. Belotserkovskii: The theory of thin wings in subsonic flow (translated from Russian), Plenum Press, New York 1967. (5.3] J. Boersma: Note on the lifting-surface problem for a circular wing in incompressible flow, Q.J. Mech., Appl. Math., 42(1989) 55. (5.4] Von M. Borja. H. Brakhage: Uber die numerische Behandlung der Tragfldchengleirhung. Z. Angew. Math. u. Mech., 47(1967) T.102. Borja, H. Brakhage: Zur numerischen Behandlung der Tragflachengleichung, Zeitschr. f. Flugwissenscli. 16(1968) 349. [5.5] 111. (5.6] P. Cociirlan: Cercetdri asupra tcorici aripii portanle, teza de doctoral, Universitatea din Bucurejti, 1971. [5.7] L. Drago§: Method of fundamental solutions. A novel theory of lifting surface in subsonic flow. Arch. of Mechanics, 35(1983) 579. [5.8] L. Drago§: Subsonic flow post thick wing in ground effect. Lifting line theory, Acta Mechanica. 82(1990) 49. [5.9] L. Dragoq: Arbitrary wings of low aspect ratio in subsonic flour, Z. Angew. Math. u. Phys, 38(1987) 648. (5.10] L. Drago§, A. Carabine:una: New quadrature formulae for the integration of the lifting surface equation, (in press). (5.11] L. Dragoq, A. Carabineanu: A numerical solution for the equation of the lifting surface in ground effects, Commun. Nuuner. Meth. Engng. 2002, 18 2002. 177. 544 13I111,Icx;ItAI'i1Y [5.12] H. Flax, H.R. Lawrence: The aeraadynamres of loin aspect ratio wings and wing body combinations. Proc. of the and Anglo-Amer. Aero. Conf. R. Aero. Soc. (1951) :363. (5.13] J.L. Hess: The problem of three-dimensional lifting potential flow and its solutions by means of surface singularity distribution, Comp. Meth. in Appl. Mech. & Engng., 4(1974) 283. [5.14] J.G. Hess: Review of integral cquatcon techniques for solving potential flow problems with emphasis on the surface-source m.ethcxl, Comp. Meth. Appl. Mech. & Engng., 5(1975) 145. [5.15) P.T. Hsu: Sortie recent developments in the flutter analysis of low - aspect - ratio wings, Proc. Nat. Spec. Meeting on Dynamics and Aeroelasticity, Forth Worth, Texas 1958, pp. 7-26. [5.16) D. Homentcovsclii: Aerodynatnique stationairv linearisee, Arch. of Mechanics, 27(1975) 325. (5.17] R.T. Jones: Properties of low-aspect-ratio pointed wings at speed below and above the speed of sound NACA Rep. 835(1446). (5.18] R.T. Jones: Some recent deuuelopments in the aerodynamics of wrings for high speeds, Z. Flugwiss, 4(1956) 257. (5.19] P.F. Jordan: On lifting wings with parabolic tips, Z. Angew., Math. u. Mech., 54(1974) 463. (5.20] M. Landahl and V.J.E. Stark: Numerical lifting surface theoryProblems and Progress, AIAA Journ, 6(1968) 2049. [5.21] H.R. Lawrence: The lift distribution on lour aspect ratio wrings at subsonic speeds, J. Aero. Sci., 18(1951) 683; The aerodynamic charvetcristics of low-a.cpxrt-tntio urinq-body combinations in steady subsonic flow, ideni 20(1953) 541. [5.22) K.W. Mangler, B.F.R. Spencer: Some remarks on Mutthopp's subsonic lifting-surface theory, Aero. Rcs. Count. It. & M. 2926 (1952). [5.23] L. Morino, K.C. Chang: Subsonic potential aeraxlyn.atnic for cornpler eonfiqurtction: a) general theory, AIAA .lourn., 12(1974) 191. [5.24] II. Multhopp: Methods for calculating the lift distnbutcan of wings (subsonic lifting - surface theory). Aero. Res. Coun. R & N I 2884 (1960,1955). 545 BIBLIOGRAPHY (5.25] J.N. Newnian: Analysis of small-aspect-ratio. Lifting surfaces in ground effect, J. Fluid Mech., 117(1982) 305. [5.26] L. Prandtl: Theorie des Flugzenghng fltigels im zu.cammen - driickbaren Medium, Luftfahrt.forsehung 13(193(;) 313. [5.27] L. Prandtl: Beitrag --ur theorie der tragenden Fhiche, Z. Angew., Mat. u. Mech., 16(1939) nr.6. [5.28] J.W. Purvis, J.E. Burkhalter: Simplified solution of the compress ible lifting surface pmblerrrs, AIAA .lourn., 20(1982) 589. [5.29] E. Reissner: Note on the theory of lifting surface, Proc. Nat. Acad. Sci., USA, 35(1949) 208. [5.30] A. Rosen, A. Isser: Recent studies on the lifting surface theory. Computers Math. Appl. 17(1989) 1115. 15.311 G.H. Sounders: Aerodynamic characteristics of wings in ground proximity, Canad. Aeronaut a. Space Journ. 1965, pp. 185. [5.32] E. Truckenbrodt: Das Ceschwindigkeits potential der tnngenden Fldche bei Inkompressiblen Strornung, Z. Angew. Math. it. Mech., 33(1953) 165. [5.33] A.I. Van de Vooren: An approach to lifting surface theory, Rep. Nat. Luchty. Lab.. Amsterdam, F 129(1953). [5.34] J. Weber: The calculation of the pressure distribution of thick wings of small aspect ratio at zero lift in subsonic flow, Rep. Memor. Aero. Res. Counc., London 17522(1954). [5.351 J. Weissinger: The lift distribution of swept - buck wings, NACA 1'N 1120(1947). [5.36] J. N eissinger: Theorie des 7ingflrigels bei stationiirrrr Bewequng in reibungslesen Inkompressiblen Medien, in Handbuck der Physik, bd 8/2, Springer-Verlag, Berlin, 1963. An analytic solution for two and three [5.37] S. Widnall, T.M. - dimensional wing in ground effect. J. Fluid Mech. 41(1970) 769. [5.38] I.T. Wu: Hydrofoils of finite spun, Math. Phys.. 33(1954). 546 BIRLIOCRAPIIY Chapter 6 [6.1) E. Carafoli: Stir les caracteristiques des wiles trapezoidales et rectangulares, Corn. Academici de 5tiinte din Romania 1943; TIdorie des wiles monoplanes d'envergure fine, Ann. de I'Academie Roumaine 1945. [6.2) H.K. Cheng: Lifting-line theory of obligue wings, AIAA Journ., 16(1978) 1211. [6.3] H.K. Cheng, R. Chow, R.E. Melnik: Lifting-line theory of swept wings based on the full potential theory, Z. Angew Math. u. Phys. 32(1981) 481. [6.4) L. Drago§: Asupra ecua4iei integro - diferr..nf isle a lui Prandtl, Com. Acad. R.P.Romane, 8(1958) 451. [6.5) L. Dragon: Integration of Prundtl's equation by the aid of quadrature formulae of Gauss type, Quart. Appl. Math. 52(1994) 23. [6.6] L. Dragoq: A colocation method for the integration of Prandtl's equation, Z. Angew. Math. U. Mech., 74(1994), 289 [6.7] I. Filimon: On Prandtl's integro - differential equation, But. $tiint,ific Acad. R.P.R., 9(1957) 981. [6.8) Jean-Luc Cuertnond: A generalized lifting-line theory for curved and swept wings, J. Fluid Mech., 211(1990) 497. [6.9) D. Homentcovschi: On the deduction of Prandil's equation, Z. Angew. Math. u. Mech. 57(1977) 115. [6.10) Th von H.S. Tsien: Lifting-line theory for a wring in non-uniform flow, Q.Appl. Math. 3(1945) 1. [6.11) H.R. Lawrence, A.H. Flax: Wing-body interference at subsonic and supersonic speeds. Survey and new detrelopments, J.Aero. Sci., 21(1954) 289. [6.12) J.H. de Leeuw, W. Echans, A.I. van de Veoren: The solution of the generalized Prandtl equation for swept wings, Rep. Nat. Luchty Lab., Amsterdam F. 156(1954). [6.13] J. Legras: Mcthodes et techniques de ]'analyse numerique, Dunod, Paris 1971. BIBLIOGRAPHY 547 [6.14] I. Lighthill: Nydromechanics of aquatic animal propulsion, An, Rev. Fluid Mach., 1(1969) 413; Aquatic animal propulsion of high hydrornechanical efficiency, J. Fluid Mech.. 44(1970) 265. [6.15] 1. Lotz: Bcrechnung der Auftriebsverteilung beliebig geformter Fliigel, Z. Flugtech. 22(1931) 189; The calculation of the potential flow past airship bodies in yaw, Tech. MMfemor. NACA, Wachington 675(1931). (6.16] L.G. Magnaradze: Asupru unei not ccuatii intcgrale in teoria aripii de anion (in lb. rush) Coman. AN Gruz SSSR 3(1942) 503. (6.17] H. Multhopp: Die Berrchnung der auflriebsverteilung von Trugfliigeln. Luftfahrtforschung, 15(1938) 153, Translation ARC 8516. [6.18] T. Nishiyama: Lifting - line theory of the submerged hydrofoil at finite span, ASNE p. I. Aug. 1959, p. II, Nov. 1959, p. III. Febr. 1960, p. IV, May 1960. [6.19] A. Pfiuger: Ein Spannungsgletchnis --um Problem der tragenderr Linie. Z. Angew. Math. u. Mech., 25-27(1947) 177. (6.20] A. Plotkin, C.H. Tan: Lifting-line solution for a syrninetrical thin wing in ground effect, AIAA Journ 24(1986) 1193. [6.21] L. Prandtl: Trugflrigeltheoric 1, II, Mitteilung Nachr. Ges. Viss., Gottingen, 1918/1919. [6.22] S. Pr6ssclorf. D. Tordella: On an Extension of Prundtl's Lifting Line Theory to Curved Wings, Impact of Computing in Science and Engineering. 3(1991) 192. [6.23] M. Schleiff: Untersuclurng ciner linearen singularien Inteypodiferential gleiehung der Trogflr7gcltheorie, Wiss. Z. Univ. Halle 17(1968) 44: Ober Nahcrungsverfahren zur Losung einer singularen linearrn Integro-diff'errntialgleichung, Z. Angew. Math. u. Mech., 48(1968) 477. [6.24] H. Schmidt: Strenge Losungen zur Prnndtlschen Theorie der trngenden Linie, Z. Angew. Math. u. Mech., 17(1937) 101. [6.25] K. Schroder: Ubcr die Prundtlsche Inlegrrxliff errentialgleichruig der Trngfliigeltheorie. Abhandl Preuss Akad. Wis.S., Math.-Naturwiss E'Llasse 1939. 548 I3II3LICCRAPHY (6.26) P.D. Sclavounes: An unsteady lifting-line theory. J. Engrrg Math., 21(1987) 201. [6.27) E. Trefftz: Zur PrmndtLschen Tragjtuchen-theorie, Math. Atrnaletr 82(1921) 306; Z. Angew Math. u. Mech. 18(1938) 12. [6.28) I.N. Vekua: On Prandil's integro - differential equation (in Russian), Priki. Mat., Mekh., 9(194)) 143. [6.29) A.I. van de Vooren: The generalization of Prundtl's equation for yawed and swept wings, Rep. Nat. Ludrty Lab., Amsterdam, F 121(1952). [6.30] J. Weissinger: Uber eine Erweiterung tier PzmndtLschen Theoric der tragonden Linie, Math. Nach.. 2(1949) 46. (6.31) J. Weissinger: Uber Intcgrndiferentialgleichungen von Typ der Prandtlschen T'ngftiigelgleichung, Math. Nachr., 3(1950) 316. (6.32) D.E. Williams: The effect of wing - tailplane aerodynamic interaction on taie flutter, Techn. Rep. ARC R & M nr. 3065. (6.331 J. De Young, C.W. Harper: Theoretical symmetrical span loading at subsonic speedds for wings having arbitrary planform, NACA Rep. 921(1948). Chapter 7 (7.1) J.H. Argyris and D.W. Scharph: Two and three-dimensional potential flow by the method of singularities, Aero.J.Roy. Aero Soc. 73(1969) 959. (7.2) L. Drago$, A. Dinu: Application of the boundary integrul equation method to subsonic flow past bodies and wings, Acta Mechanica 86(1)(1991) 83. (7.3) L. Dragon, A. Dinu: A direct boundary integral method for the three-dimensional lifting flow, Comput. Methods Appl. Mech. Engng. 127(1995) 357. (7.4) L. Drago§, N. Marcov: Solutions of the boundary integral equations in the theory of incompressible fluids, Analele Univ. Buc. XLVI(1997) 111-119. BIBLIOGRAPHY [7.5] 549 L.J. Gray, E. Lutz: On the treatment of corners in the boundary element method, J.Comnput.Appl..\latli.32(199J0) 369. [7.6] equation method for the potential E. Grodtkjaer: A direct flow about arbitrary bodies, The Danish Center for Appl. Mathematics and Ateclianics. Rep.21. Dec.1971. [7.7] F. Hartmann: Introduction to boundary clentents. Theory and ap. plications, Springer - Verlag. Merlin. 1985. [7.8] Heshan Liao. Zhixin Xu: Methoxl for direct evaluation of singular integral in direct BEAT, Int.J.Nu mer Met hods Engng, 35(1992) 1473. [7.9] J.L. Hess. A.M.O. Smith: Calculation of tonlifting potential flow about arbitrary throe-dimensional bodies, J.Ship.Res. 8(196.1) 22. [7.10] J.L. Hess. A.M.O. Smith: Calculation of potential flows about arbitrary bodies, Progress in Aeronautical SCI., 8(1967) Pergamon Press N.Y. Ed.D.Kuchemann. [7.11] J.L. Hems: The problem of thrrc-dimensional lifting flow and its solution by means of surface singularity distribution, C'omput. Methods Appl.Alech.Engrg.. 4(197.1) 183. [7.12] J. Katz. A. Plotkin: Low-speed nermlynainies. From wing theory to panel methods, Me Gray-Hill Inc., N.Y.. 1991. (7.13] S.V. Kozlov. L.K. Lifanov and A.A. Mikhailov: A new approach to mathematical modelling of flour of ideal fluid around bodtc Soy. J. Numer. Anal. Math. Modelling. 6(1991) 2(1<J. (7.14] R.I. Lewis: Inverse method fur the design at bodies of revolution by boundary integral modelling. Proc. Inst. Mech. Engng., 205(1991) 91. [7.15] R. Srivastava, D.N. Contractor: Efficient evaluation of integrals in 3D BEAM using linear shape functions over plane triangular elements, Appl.Alatlt.Model. 16(1992) 282. (7.16] James S. Uhlman. Jr.: The surface singularity method applied to partially cavttating hydrofoil.,. 5th' Lips Propeller Symporium, Druueu - The Netherlands 198:3. [7.17] «'u Zltang. Huan-Ran Xit: G'cnernl and effective way for evaluating the integrals with various order of singularity in the di- 550 BIBLIOGRAPHY rcct boundary element method, Int. J. Numer. Methods Engng. 28(1989) 2059. Chapter 8 Uber Luftkrtifte auf Fliigel, die roil grosseres als Schallgeschwindigkeit bewegt weiden, Zeitschr. f. Flugtechuik Motorluftschiffahrt, 16(1925), 72. [8.1] J. Ackeret: (8.2] L. Berechet: Lifting surface theory in supersonic flow, Lucrare de licentti la Facultatea de Mateinatica, Universitatea din 1984. (8.3] Busemann: Infinitesimale heegelige Uberschallstromung, Schriften der Deutschen Akademie dr. Luftfahrtfaschung 713(1943) pp 105-122. A. [8A] V. Cardoti: Stir les ecoulements supersoniques autour des obstacles coniques aplatis on de faible ouverture, Rev. Roum. Math. Pures et Appl., 27(1982) 285. (8.5] V. Cardo§: Asupnri ,nz carilor supersonice cu undd de ,hoc, St. Cerc. Mee. ApI., 52(1993) 355. [8.6] V. Cardo§: Le probleme aux lirnites simplifie pour les mouvements coniqucs supersoniques, Rev. Roum. Sci. Techu. Mlcc. Appl., 42(1997) 59. [8.7] L. Dragon: The method of fundamental solutions in aerodynamics 11. Applications to the theory of lifting surfaces in supersonic flow, Arch. Mechanics 37(1985) 221-230. [8.8] L. Dragoq, A. Carabineanu: The supersonic flow past a thin profile in ground and tunnel effects, ZAMM, 82(2002) 9, 649-652. [8.9] C. Evvard: Use of source distributions for evaluating theoretical aerodynamics of thin finite wings at supersonic speeds, NACA Rep. 951(1950). [8.10] C. Ferrari: Interaction Problems in vol. VII of Princeton Series in High Speed Aerodynamic and Jet Propulsion (A.F. Donavan and H.R. Lawrence) Princ. Univ. Press, Princeton 1957. BIBLIOGRAPHY 551 [8.11] P. Germain: La theories gene le ties rnouvements coniqucs et ses applications a l'aenxlynarnique supersoniques, Office Nat. Etude et Rech Aeronaut. 34(1949). [8.12] P. Germain: Sur le minimum de trainee dune aile de form en plan donnee Compt.. Rend. 244(1957) 2691. Aile syrrrraetrTque u 1xirtance nulle et de volum donnee realisant le rninimuru de trainee en ecoulemeut supersonique, Compt. Rend. 244(1957) 2691. [8.13] P. [8.14] H. Glauert: Theory of thin airfoils, British A.R.C., R & H 910(1924) sau The clcmcnts of aerofoil and airscrew theory, sec. ed. Cambr. Univ. Press 1947. [8.15] M.A. Heaslet. H. Supersonic and transonic small perturbation theory, in General theory of High Speed Aerodynamics eel. by W.R. Sears. Oxford University Press, London, 1955. 18.16] D. Homentcovschi: Aerndynanuque stationnaire Iineare. a 1!. Supersoniquc. Arch. Mechanics. 29(1977) 41-51. [8.17] C. lacob: Validity conditions in plane supersonic lineal aenxlynamics, Studii §i Cercet. Mat. 32(1980) 649. [8.18) C. lacob: Teoria aripii ungh.iulare la viteze supersonrce, Analele Ac:ademiei R.P.R. T13, Memoriul 15. [8.19) T. von Karman: Supersonic aerodynamics, Principles and Appl. J. Aeron. Sci. 14(1947) 373-409. [8.20] E.A. Krailseicova: Wing of finite average in compressible flow (in Russian), Fizmatghiz, Moscow, 1952. [8.21] M.N. Kogan: On bodies of minimum drag in supersonic gas stream, Prikl. Mat. Mekh., 21(1957) 207. [8.22] P. A. Lagerstrom: Linearized supersonic theory of conical wings, NACA Techn. Note 1685. 1948. [8.23) H.W. Liepmann. A. Ro hko: Elements of gas dynamics, John Wiley. New York, (1957). [8.24) M.J. Lighthill: On boundary layer and upstream rnfluenee II_ Supersonic flour without separation, Proc. Roy. Sex:. London, Ser A, 217(1953) 478. 552 BIBLIOGRAPHY [8.25] J.M. Lighthill: The conditions behind the trailing edge of the supersonic aerofoil, ARC Rep and Memo. 1930, 1941. [8.26] H. Loniax, M.A. Heaslet: Recent developments in the theory of wing-body wave drag, Journ Aero. Sci., 23(1956) 1061-1074. [8.27] W. Miles: Potential theory of unsteady supersonic flow, Camp. Univ. Press, London 1959. [8.28] H. Poritsky: Homogeneous harmonic functions, Quart. Appl. Math., 6(1949) 379-390. [8.29] A.E. Puckett: Supersonic wave drug of thin airfoils, J. Acro. Sci. 13(1946) 475-484. [8.30] A. Robinson: On source and vortex distributions in the linearized theory of steady supersonic flow, Quart. J. Mech. Appl. Math., 1(1948) 402. [8.31] D.A. Spence: Prediction of the characteristics of two-dimensional airfoils, J. Aer. Sci., 21(1954) 577-587. [8.32] Irinh Dinh An: Validity conditions in supersonic flow past a rectangular wing, St. Cerc. Mat., 33(1981); The validity conditions in three-dimensional supersonic lineal aerodynamics, Rev. Rourn.t Math. Pures et Appl., 29(1984) 215; Sur les conditions de validity physique en eeoulemnent supersonique autour dun obstacle conique de fatble ouverture, idem, 291. [8.33] G.N. Ward: Supersonic flow past thin wings 1. General theory, Quart. .1. Mech. Appl. Math., 2(1949) 136-152. [8.34] G.N. Ward: Linearized theory of steady high - speed flow, Cainbridge University Press, 1955. [8.35] G.W. Ward: On the minimum drag of thin lifting bodies in steady supersonic flow, British. A.R.C. Rep. 18(1956). Chapter 9 [9.1] F. Bauer, P. Carabedian & D. Korn: A theory of supercritical wing section with computer pragnuns and examples, Springer, 1972. 553 BIBLIOGRAPHY [9.2] Y.T. Chan: An asymptotic analysis of transonic wind - tunnel interference based on the full potential theory. 'LAMP 36(1985) 89, 85. [9.3] J.D. Cole: Modern developments in transonic flow, SIAM J. Appl. MIath., 24(1975) 763, 27(1978). [9.4) J.D. Cole: Drag of finite wedge at high subsonic speeds, J.Math. and Phys 30(1951) 79-93. [9.5) J.D. Cole, L.P. Cook: Tnunsome aerodynamics, North - Holland, Amsterdam, 1986. [9.6) J.D. Cole and A.F. hlessiter: Expansion procedures and similarity laws for transonic flow, Z AMP 8(1957) 1-25. 19.71 J.C. Crown: Calculation of transonic flow over thick airfoils by intemnl methods 4' ; A Journ 6(1968), no. 3, 413-423. . Theory of three - dimensional wing and lifting line theory in transonic steady flow, Rev. Rouen. Sci. Math. Purer et Appl., 32(1987) 137-154. [9.8) L. Dragon;: [9.9) L. Dragog: Theory of three - dimensional wing and lifting line theory in transonic flow, Preprint Ser in Math., INCREST nr. 62/1984. [9.10] T. Evans: An approximate solution for two dimensional transonic flow, Proc. Cambr. Phil. Soc. 61(1965) 573. [9.11) C. Ferrari, F. Tricomi: Aerodinamica transunica, Rome, Edit. Crenionesse 1962, English translation: Transonic Aerodynamics, Acad. Press N.Y. 1968. 19.12) F.I. Frankl: Theoretical research on the wing of infinite span at sonic speed, Dokl. Akad. Nauk. SSSR, 57(1947) 661-664- [9.13] P.R. Garabedian & D.G. Korn: Analysis of transonic airfoils, Comm. Pure Appl. Math. 24(1971) 841. [9.14] K.G. Guderley: Theorie Schallnaher Stromunyen. Springer- Verlag 1957: Theory of transonic flow. Internat. Series of Monographs in Aeron. a. Astron., Division 11 Aerodynamics, vol. 3, Pergamon, New York 1962. 554 BIBLIOGRAPHY [9.15] K.G. Guderley: Theoric schallnaher Strornungen, Springer, Berlin, 1957. [9.16] K.G. Guderle,y and Yoshihara: The flow over a wedge profile at Much number 1, J. Aero Sci., 17(1950) 723 - 735. Three-duncnsional transonic J.R. Spreiter: flow theory applied to slender wings and bcxlie:s, NACA Rep. (9.17] M.A. Iiea-slet, 1318(1956). [9.18] M.A. Heaslet.: Thv e-dirnensioual transoms flour theory applied to slender wings and bodies, NACA Rep 1318(1956). [9.191 U. Homenteovschi: An introduction to BEAT by integral transforms, Preprint Series in Math. INCREST 25 1987. for the lifting profile in [9.20] I). Homentcovschi: Integral steady transonic flows, Preprint Series in Mathematics INCREST, 63, 1983. [9.21] U. lIomentcov di : Stradej Arch., Mech., 42(1990) 3. t.erudynar tics III Transonic. [9.22] 1. llosokawa: A rrfincruent of lxnearised transonic flow theory, J. Phys. Soc. Japan 15(1960) 149. (9.23] I. Hosokawa: Transonic flow past a wavy wall, J. Phys. Soc. Japan 15(1960) 2080 - 2086. 19.241 1. Hosokawa: Theoretical prediction of the prr.ssurc distribution on a non - lifting, thin symnructricul aervfoil at varzous transonic speeds, J. Phys Soc. Japan, 16(1961) 546 - 558. [9.25] 1. }Josokawa: Unified formalism of the lineariscd romprrssible flow fields, Quart Appl. Math., 22(1964) 133-142. [9.26] Th. von The sirnilanty low of transonic flow, .1. Math. Phys., 26(1947), 182-190. [9.27] M.T. Landahl: Unsteady trunsoni.c flow, Pergacuon, 1961. [9.28] K. Leelavathi, N.R. Subramrueian: l'rIcssure. distribution in inviscid trmasonic flow past axisynunetric bodies (AID = 1), AIAA .Iourn 7(1969) 1362 - 1363. [9.29] H.W. Liepmann, A.E. Bryson: Transonic flow past wedge sections .1. Aero. Sci. 17(1950) 745. BIBLIOGRAPHY 555 (9.30) A.R. Manwell: The holograph equations, Oliver and Boyd, Edinburgh, 1971. [9.31) A.R. Manwell: The Tricomi equation with applications to the theory at plane transonic flow, Pittuan Adv. Publ. Pr. San Francisco, 1979. [9.32] C. Masson, 1. Paraschivoiu: Integral method for transonic flow, Can. Aeronaut. Space J. 36(1990) 18. (9.33) C.S. Morawetz: The mathematical approach to the sonic banier, Bull. Amer. Math. Soc., 6(1982) 127 - 145. [9.34] T.H. Moulden: Fundamentals of transonic flow, John Wiley & Sons, New York, 1984. [9.35] E.M. Murmann, J.D. Cole: Calculation of plane steady transonic flows, AIAA Jouni. 9(1971) 114. [9.36] P. Niyogi: Integral equation method in transonic flow, Springer Verlag, Berlin 1982. (9.37] D. Nixon: Calculation of transonic flows using and extended integral equation method, AIAA Journ., 15(1977) 295. [9.38] D. Nixon: The transonic integral equation method with curved shock waves, Acta Mechanica, 32(1979) 141. [9.39] D. Nixon(ed.): Transonic aerodynamics, Amer. Inst. Aeron. a. Astron., New York, 1982. [9.40] D. Nixon: Observations on the occurence of multiple solutions in transonic potential theory, Acta Mechanica, 68(1987) 43. (9.41] H. Norstrud: Transonic flow past lifting wings. AIAA Journ. 11(1973) 754. [9.42] W. Ogana: Numerical solutions for subcritieal flow by a transonic integral equation method, AIAA J. 15(1977) 444. [9.43] W. Ogana: Derivation of an integral equation for three dimensional transonic flours, AIAA J.. 17(1979) 305. [9.44] W. Ogana: Transonic integro-differential and integral equations with artificial viscosity, Eng. Anal. Boundary Elem. 6(1989) 3. [9.45] W. Ogana: Boundary element solution of the transonic integra differential equation, Eng. Anal. Boundary Elem. 556 BIBLIOGRAPHY [9.46] W. Ogana: Analysis of transonic integral equation, fart. lI Boundary element methods, AIAA J. 28(1990) 3. [9.47] W. Ogana: Solution of the transonic intcgrn - difcrrntial equation using a decay functions, Appl. Math. Model 14(1990) 30. [9.48] W. Ogauna, J.R Sprieter: Derivation of an integral equation for transonic flows, AIAA J., 15(1977) 281. Die Gesc lawtndtgl,:catsvcrtcrlurag bei healer (9.49] K. Oswatitseh: Uberschallgebeiten an flachen 1'rofilen, ZAMM, 30(1950), 17-24. [9.50] K. Oswatitsch, S.B. Berndt: Aerodynamic similarity of axisyrnrnetric transonic flow around slender bodies, KTH Aero TN 15, Royal Inst. Technology, Stockholm, 1950. [9.51] K. Oswatitsch, F. Keune: Eirt Aquivalenssatz fur nichtangestellte Flugel Kleiner Spanaweite in Schallnaher Strornung, Z. Flug wiss., 3(1955) 29 - 46. (9.52] K. Oswatitsch, K. and F. Keune: Eire Aquivalenssatz fur nichtangestellte Flugel Kleiner Spanaa'eatc in Schallrtahcr- Stronaung, Zeit. fur Flug wissenschoften, 3(1955) 29. [9.53] L.V. Ovsiannikov: Equation of the transonic: flow of the gases, Vestnik LGY, 1952, nr.6, 47 - 54. [9.54] L. Pamela Cook: Lifting - !iron theory for a swept wring at transonic speeds, QAM, 37(1979) 177. [9.55] L. Pamela Cook: Lifting - line theory for transonic flow, SIAM .1. Appl. Math. 35(1978) 209 - 228. (9.56] M. Schubert, M. Schleiff: Utx-r zwer Randauertprvblerne des inhonaogenen Systems der Cauchy - Rimanschen Differential - gteichungen mit einer Anwendung and ein Problem der stationdr n schallnahcn Strornrnung, ZAMA1, 49(1969) 621. [9.57] J.R. Spreiter: On the application of transonic similarity rules to wings of finite span, NACA Rep. 1153, (1953). (9.58] J.R. Spreiter: On alternative forms for the basic equations of tran- sonic flow theory, J. Acro., SO., 21(1954) pp 70-72, Errata: J. Aero Sci 21(1953) 360. BIBLIOGRAPHY 557 [9.59] J.R. Spreiter, A.Alksne: Therrrrtical pressure distributions on wings of finite span at zero incidence for Mach numbers near 1, NASA TR R-88 (1961). [9.60] J.R. Spreiter, A.Alksne: Slender body theory based on approximate solution of the transonic flow equation, NASA TR R-2 1959. [9.61] Sympomium transonicum, Aachen 1962 ed. by K.Oswatitsch. [9.62] L. Trilling: Tinnsonic flow past a wedge profile at Mach number 1, J. Aero. Sci. 17(1950) 72:1 - 735. [9.63] W.G. Vincenti, C.B. Wagoner: Transonic flow past a wedge profile with detached bow wave, NACA Rep 1095(1952). (9.64] H.J. Wirz, J.J. Smoldern: Numerical methods in fluid dynamics. Mc Graw-Hill, New York. [9.65] J. Zierep: Die Integralgleichungsmethode zur Berechnung schallnaher Stromingen, Symposium Transonicum, Berlin, 1964. (9.66] J. Zierep: Der Aquivalenzsatz and die parabolische Methode fur schallnahe Strornungen, ZAMM 45(1965) 19 - 27. Chapter 10 [10.1] E.A. Acum: Aerodynamic forces on the rectangular wings oscillating in a supersonic air strram, Aero Res Council R §M 2763 (1953). [10.2] E. Albano, W.P. Roden: A doublet - lattice method for calculating lift distributions on oscillating surfaces in subsonic flow, AIAA Journ. 7(1969) 279. [10.3] S. Ando, M. Kato: An improved kernel function computation in subsonic unsteady lifting surface theory, Computer Meth. in Appi. Mech. Enging, 49(1985) 343 - 355. [10.4] H. Ashley, S. Windall. M.T. Landahl: New directions in lifting surface theory, AIAA Journ. 3(1965) 3. [10.5] S.R. Bland: The two dimensional oscillating airfold in a wind funnel in subsonic flout, SIAM Jouru. Appl. Math.,1(1970) 830 848. 558 BIBLIOGRAPHY (10.61 S. Borbely: Acntxlynanuc forres on a harmonica) oscillating wing at in subsonic velocity(2-Dine. arse), ZAMM 22(1942) 190 - 205b. [10.71 M. Borja: Eiric rule MMlethcxle :ur riumerische Bchandlunq harmonisch schwiriger Tingfliigel. doctoral thesis July (1969) 1lniversitiit. Karlsruhe. [10.81 A. Carabineariu: Incompressible flow past oscillatory/ trings of low aspect ratio by the integinl equations method, Int..1. Nunier. Eng. 45(1999) 1187. (10.91 P. Cicala: Lc arioni aenrxlinarniche sui prrifili di a.lla oseilanti in presenza di Conrntc uniforrne. Mean. R. Accad. Torino, Ser. 2. P 168(1934-1935) 73 - 98. [10.10] H..1. Cunningham: Improved numerical proccdun for harmonically deforming lifting surfaces from the supersonic kernel function method, AIAA Journ. 4(1966) 1961 - 1968. supersonic kernel function [10.11] A.M. Cunningham .Oscillatory) Jr.: method for interfering surfaces, .Iourn. Aircraft 11(1974) 664. (10.12( D.E. Davies: Three-dimensional sonic theory, Aero. Res. Couticil, Rep.§Mem.34(1965). [10.131 D.E. Davies: Calculation of unsteadil generalisr:d airforces on a thin wing oscillating harmonically in subsonic flow, Aero. Res. Council, Rep.§Mem. No 3409 (1965). [10.14( F. Dietze: The air forres at the harmonically vibrating wing at subsonic velocity (plane problem), Laftfahrt forvach 16(1939) 81. [10.15] L. Drago : The method of fundamental solutions. The theory of thick oscillatory airfoil in subsonic flour, Rev. Roam. Sci. Tedin.Mec. Appl. 29(1984) 59. [10.16] L. Drago§: On the theory of oscillating thick wrings in subsonic flow. Lifting line theory, Acta Mechanics 54(1985) 221 - 231. (10.171 L. Dragon: On the theory of oscillating airfoils in supersonic and sonic flow, J. Appl. Math. a. Phys.(ZAA1P) 36(1985) 481. 110.181 L. Dragon: On the theory of oscillating airfoils at ATach number one, Mech. Res. Comm. 13(1986) 317 - 324. BIBLIOGRAPHY 559 [10.191 L. Drago§: Calculation formulae for lift and moment coefficients for oscillating arbitrary airfoils in supersonic flow, Rev. RoutnSci. Techn.-Mec. Appl. 31(1986) 345. [10.20) L. Drago§: On the theory of oscilating wings in sonic flow, ZAMM (Z. Angew. Math. u. Mech.) 68(1988) 373. [10.211 L. Drago§: Theory of oscillating wings in supersonic flow (unpublished). [10.22] J.A. Fromme, M.A. Goldberg: Reformulation of Possio s kernel with application to unsteady wind tunnel interference, AIAA Journ. 18(1980) 951. [10.231 J.A. Fromme, M.A. Goldberg: Aerodynamic interference effects on oscillating airfoils with controls in ventilated wind tunnels, Reprint form AIAA Journ. 18(1980) 417. [10.24] I.E. Garrick, S.I. Rubinow: Flutter and oscillating air-forces calculations for an airfoil in two-dimensional supersonic flow, NACA Rep. 846(1946). [10.251 I.E. Garrick, S.I. Rubinow; Theoretical study of air forces an steady thin wing in a supersonic main stream, NACA T. N. 1383(1947). Nonsteady wing characteristics, in High Speed Aerodynamics and Jet Propulsion, vol. VII (ed. A. F. Donavan, H. R. Lawrence) N. Y. (1957). [10.261 I.E. Garrick: [10.271 H. Clauert: The force and moment on an oscillating airfoil, Aero. Res. Counc. R & M 1242(1929). [10.281 J. Grzedzinski: Application of the unsteady lifting-line method to arbitrary configurations of lifting surfaces, Arch. Mech. 41(1989) 165-181. 110.291 R.L. Harder, W.P. Rodden: Kernel function for nonplan.ar oscillating surfaces in supersonic flow, J. Aircraft 8(1971) 677. [10.30) M.D. Haskind: Oscillations in a subsonic gas flow, Prikl. Mat. Mekh. 11(1947) 129 - 146. [10.311 D. Homenteovschi: Chemically reacting flow of an inviscid gas past a thin body, Acta Mechanica 33(1979). 560 11113L1C)C:ItAI'll Y [10.32) V. Huckel, B.J. Durling: Tables for vving-airley-on coefficients of oscillating air forces fwo-dtrnensional supersonic flow. NACA T. N. 2055(1950). [10.33) V. Huckel: Tabulation of the f-functions which Occur in aerodynamic theory of oscillating vnngs in supersonic flow, NACA T. N. 3606(195(1). [10.34) T. Ichikawa, K. Isogai: Note on the aerodynamic theory of oscillating T-tails, Trans of Japan Soc. for Aeron a. Space Sci. 16(1973). [10.35] K. Isogai, T. Ichikawa: Lifting-surface theorrni for a ruing oscillating in Yaw and Sidelip urith an angle of attack, AIAA J. 11(1973) 509 - 606. [10.36] P.F. Jordan: Numerical evaluation of three-dimensional harmonic kernel, Z. Flugwiiss. 24(1976) 205. [10.37] H.G. Kiissner: Allgemeine Trrryflachen theorie, Luftfahrtforschung, 17(1940) 3. [10.38] H.G. Kiissner: A review of flu two-dimensional problem of unsteady lifting surface theory during the last thirty years, Inst. of Fl. Dyn. and Appi. Math. Univ. Maryland, Lecture Series No. 73(1953). [10.39) H.C. Kiissner: A generrd method for solving problems of the unstcrtrly lifting surface theorem in the subsonic range, J A S 21(1954) 17. (10.40) T.P. Kalman, W.P. Rodden, J.P. Giessing: Application of the Doublet Latice, J. Aircraft 8(1971) ur. 6. [10.41) K.R. Kimble, D.D. Liu, S.Y. Ruo. J.M. WVu: Unsteady transonic flow analysis for low aspect ratio. pointed wings, AIAA .1. 12(1974) 516. [10.42) M.T. Landahl: Unsteady transonic flora, Pergamon Press, Oxford (1961). [10.43) M.T. Landahl: Kernel function for nonplane oscillating surface flow, AIAA J. 5(1967) 1045. [10.44) M.T. Landahl: On the pressure loading function for oscillating trings with control sui faces, A IAA J. 6(1968) 345 - :348. IIIIIIJOCRAPHY 561 (10.45] B. Laschka: Das potential and das Gesch.windigkcitsfrld der harrnonis schuringenden trngenden Fliiche bci Uriterschallstromung, ZAAMIRI 43(1963) 325. Zur Theorie der harnioriisch schwingenden tingenden Flache bei Unterschallstrurnunq, Zeitschr. fiir Flugwis- [10.46] B. Laschka: senschaften. 11(1963) 265 - 292. [10.47] H.R. Lawrence, E.H. Gerber: The aen iynamic force on low aspect ratio wings oscillating in an incompressible flow, J.A.S., 19(1952) 769 - 781(Nov.); 20 (Apr. 1953) 296. [10.48] D.E. Lehrian, H.G. Garner: Theoretical calculation of genernlized forces and load distribution on wings oscillating at general frequencies in a subsonic stream, A.R.C. Rep.§Nlem. 3710(1971). [10.49] M.J. Lighthill: Oscillating ail f oils at high Mach number, J. Acton. Sc:i. 20(1953) 402 - 406. [10.50] D.D. Liu, W.H. Hui: Oscillating delta wings with attached shock waves, AIAA Journ. 15(1977) 804. [10.51] D.D. Liu, M.F. Platzer. S.Y. Ruo: Unsteady linearized transonic flour analysis for slender bodies, AIAA Journ. 15(1977) 906 - 973. [10.52] K.W. Mangler: A method of calculating the short-period longitudinal stability derivatives of a wing in linearized unsteady comprrssible flow, Aero. Res. Conlric.Rep.§Mem., London 2924(1952). [10.53] J.W. Miles: Potential theory of unsteady supersonic flow. Cambridge Univ. Press, London (1959). (10.54] Y. Miyazawa. K. WIVashizu: A finite state aerodynamic model for a lifting surface in incompressible flow. AIAA Journ. 21(1983) 163. (10.55] L. Morino, L.T. Chen, E.O. Suciu: Steady and oscillatory subsonic and supersonic aerodynamics around complex. configurations, AIAA .Iourn. 13(1975) 368. (10.56] H.C. Nelson: Lift and Moment on oscillating triangular and related unngs with supersonic edges, NACA T.N. 2494(1951). [10.57] H.C. Nelson, J.H. Berman: Calculations on the forces and moments for an oscillating wing-aileron combination in twodimensional potential flow at sonic ..peed. NACA Rep. 1128(1953). 562 BIBLIOGRAPHY [10.58] K. Oberhettinger. L. Badii: Springer-Verlag, Berlin (1973). Tables of Laplace transforwis, [10.59] C. Possio: L'Azione aemdrnamiea sul pmfilo oscilante in a fluido compressible a velocita iposonorn. L'Aerotecnica, 18(1938) 4. [10.60J E. Reissner: On the application of the Mathieu functions in the theory of subsonic compressible flow past oscillating an f oils, NACA T.N. 2363(1951)- [10.611 J.R. Richardson: A method for calculating the lifting fours on wings (unsteady subsonic and supersonic lifting-surface theory), Acro. Res. Council R. §M. 3157(1955). [10.62] W.P. Rodden, J.Y. Giesing, Ti'. Kalman: New developmnent and applications of the subsonic dublet-latice meth., AGARD SVMP. (1970). [10.63] W.P. Rodden: Comments on the new solution method for lifting surfaces in subsonic flow, AIAA 22(1984) 160. [10.641 N. Rott: Oscillating airfoils at Alach number 1, .1. Aero Sci. 16(1949) 380. [10.65] N. Rott : Fliigclscl:wingungsforYncn in ebener komprr ssible Potential str6mung, ZAMP 1(1950) 380 - 410. 110.661 H.L. Runyan, Ch.E. Watkins: Considcrntions on the effect of the wand-tunnel walls on oscillating air forces for two - dimensional subsonic compressible flow, NACA Rep. 1150(1953). [10.67] H.L. Runyan, D.S. Woolston, A.G. Rainy: A theoretical and experimental investigation of the effect of tunnel walls on the fon-es on an oscillating airfoil in two - dimensional subsonic compressible flow, NACA Rep. 1262(1956). [10.68] H.L. Runyan. U.S. Woolston: Alethod for calculating the acnxlynamic loading on an oscillating finite wing in subsonic and sonic flow, NACA Rep. 1322(1957). [10.69] L. Schwarz: lJerr:chnung tier Druckner teilung ciner har nzonisch. sick Verforma.nden Trngflachc in ebener Strornung, Litft.fahrtforsch 17(1940). (Jndersuchung einiger rnit den Zylinder Funckionen nullter Orrdnuny verwandten Funekionen, Lnftfahrtforsch, [10.701 L. Schwarz: 20(1944) 341. 563 BIBLIOGRAPHY (10.71] S.F. Shen, F.P. Crimi: The theory for an oscillating thin airfoil as derived from. the Oseen equations, J. Fluid Mech. 23(1965) 585609. [10.72] V.J.E. Stark: General equations of motion for an elastic wing and method of solution. AIAA J. 22(1984) 1146. [10.73] H.J. Stewart, Ting-Yi Li: Source-superposition method of solution of a periodically oscillating wing at supersonic speeds, QAXI 9(1951) 31. (10.74] K. Stewartson: On the linearized potential theory of unsteady supersonic motion, Quart. J. Mech. Appl. Math., 3(1950) 182. [10.7 5] T. Theodorsen: General theory of acmdinamic instability and the mechanism of flutter. NACA Rep. 4(1935) 96. [10.76] H. Tijdernan, R. Seebass: Transonic flow past oscillating airfoils, Annual Review of Fluid Mech. 12(1980) 181. [10.77] S. Das St5rpotential fur aperiodisdie instationare schallnahe Stromungen. in Symposium transonicum II, Gottingen (1975). Springer (1976). [10.78] S. Turbatu. J. Zierep: Aperiodischc instationdre schallnahe Strornungen, Acta Mechanica 21(1975) 165. [10.79] T. Ueda: Lifting surface calculations in Laplace domain with application to Root Loc. AIAA Journ. 25(1987) 648. (10.80] T. Ueda. E.H. Dowell: A new solution method for lifting surface in subsonic flow. AIAA Journ. 20(1982) 348. (10.81) T. Ueda. E.H. Dowell: Doublet.-Point method for supersonic unsteady lifting surfaces. AIAA Journ. 22(1984) 179. (10.82] A.I. van Vooreit: Unsteady airfoil theory, pp 36 - 89 in Advances in Appl. Mech. vol. V.(H. L. Dvyden, Th. von Karman, eds) Academic Press N. Y. (1958). [10.83] E.G. Yates Jr.: A Kernel function formulation for nonplanar lifting surfaces oscillating in subsonic flow, AIAA J. 4(1966) 1486 - 1488. (10,84] C.E. Watkins: Three dimensional supersonic theory ... , in Manual on Aeroelasticity. AGRAD. 564 }3I13LIOCRAPIIY (10.85] C.E. Watkins, J. H. Berman: On the kernel function of the integral equation relating lift and doumrnash distributions of oscillating wrings in supersonic flow, NACA Rept. 1257 (1953). (10.86] C.E. Watkins, H. L. Runyan, D. S. Woolston: On the kernel function of the integral equation relating lift and doum.wash distributions of oscillating finite wings in subsonic flow, NA('A T.R. 1234 (1955). [10.87] C.E. Watkins, H. L. Runyan, D. S. Woolston: A systematic kernel function procedure for deteremininq aerodynamic forres on oscillating of steady finite wings at subsonic speeds, NASA Telrn. Rep. R-48 (1959). [10.881 M.H. Williams: Exact solutions in oscillating airfoil theory, AIAA Journ. 15(1977) 875. (10.89] M.H. Williams: The resolvent of singular integral equations, Quart. Applied Math. (1977). [10.901 M.H. Williams: Linearization of unsteady transonic flows containing shocks, AIAA Journ. 17(1979) 394. [10.911 M.H. Williams: Unsteady thin airfoil theory for transonic horn with embedded shocks, AIAA .iourn. 18(1980) 615. [10.92] M.H. Williams, P. Jones: Unsteady supersonic aerodynamic theory by the method of potential gradient, AIAA .iourn. 15(1977) 59 - 65. Chapter 11 [11.11 Mac C.Adatus, W.R. Sears: Slender-laxly theory - review and extension, .1. Aero. Sci. 20(1953) 85. [11.2] C.E. Brown: Aerodynamics of bodies at high speeds. in High Speed Aerodynamics and jet propulsion., vol 7, Editors A.F. Donovan, H.R. Lawrence. [11.3] V. Cardoo: General integral method in the, study of supersonic flows past slender bodies, IRev. Rouen. Math. Purer Appl. 37(1992) 571. BIBLIOGRAPHY 565 [11.4] L. Drago§, A. Carabineanu: On the steady subsonic flow past slender bodies of revolution, Revue Route. Sci. Tekhn, Serie AMIecanique Appl. 34(1989) 453. [11.5[ L. Dragon: A new method in slender body theory in supersonic flow, Rev. Routn. Math. Pures et Appl. 35(1990) 139 - 146. (11.6] A. Ferri: Elements of aerodynamics of supersonic flows, AIacA1illan Co., New York, 1949. (11.7] F.I. Frankl, E.A. Karpovich: Gas dynamics of slender bodies (in russian) Gos. Izd-Vo Tekhn. - Teoret. Lit-Ry, Moskva, 1948, Gas dynamics of thin bodies (in English), Intersci. Publish., Inc., New York, 1953. [11.8] Kuo Chang Wang: A new approach to not-so-slender wing theory, J. Math. Phys., 47(1968) 391. [11.9[ E.V. Laitone: The linearized subsonic and supersonic flout about inclined slender bodies of revolution. J. Aero. Sci. 14(1947) 631. [11.10] E.V. Laitone: The subsonic flow about a body of revolution, Quart Appl. Math. 5(1947) 227. (11.11] M. Lighthill: Supersonic flow past bodics of revolution, Rep. and A Iem., 2003(1945). [11.12] P.R. Maeder, H.U. Thotntnen: Some results of linearized transonic flow about slender airfoils and bodies at revolution, J. Aeron. Sci., 23(1956) 187. [11.13] W.L. Oberkampf, L.E. Watson: fncomprvssible potential flow solutions for arbitrary bodies of revolution, AIAA Journ., 12(1974) 409. [11.14] K. Oswatitsch, F. Keune: Bin Aquivalenzsatz fiir nichtangestellte Fliiyel kleiner Spannuteite in Schallnaher Z. Flugwiss, 3(1955) 29. [11.15] K. Oswatitsch, F. Keune: The flow around bodies of revolution at Mach Number 1, Proc. Coni. on high-speed aeron., Polyt. Inst. of Brooklyn N.Y. (1955) 113. [11.161 H.S. Tsien: Supersonic flow over an inclined body of revolution, J. Aero. Sci. 5(1938) 480. 566 I3IBI JOCRAPIIY (11.17) A. van Miyl: Axially symmetric flow around a new family of half - bodies, Quart. Appl. Math., 7(1950) 395. (11.18] A. Weinstein: On axially symmetric flows, Quart. Appl. Math., 5(1948) 429. (11.19] G.N. Ward: Supersonic flow past slender pointed bodies, Journ. Mech. Appl. Math., 2(1949) 75. [11.20] M.F. Zedan, C. Dalton: Potential flow around axisymmetric bodies: direct and inverse problem, AIAA Journ., 16(1978) 242. Appendices [A.1] I.A. Brychkov, A.P. Prudnicov: Integral transforms of generalized functions (in Russian), Nauka, Moskow, 1977. [A.2) V.A. Ditkin, A.P. Prudnicov: Integral transforms and oper. calculus, 2nd ed. (Russian), Nauka, Moskow, 1974. [A.3) L. Dragog, Al. Nicolau: Partial diferential equations, in [1.22), vol. 3. [A.4J A. Erdelyi a.o.: Tables of integral transforms, Mc Gvaw-Hill, New York, 1954. [A.51 I.M. Guelfand, G.E. Chilov: Les distributions I, Dunod, Paris, 1962. [A.6J D. Homentcovschi: An introduction to BEM by integral transforms, Preprint Ser. in Mathematics 25(1987). [A.7J D. Homentcovschi: Elemente de teoria distribuliilor, in [1.22], vol. 3 and 4. [A.8] Ed. Marie de Jager: Applications of distributions in mathematical physics, Math. Centrum - Amsterdam, 1964. [A.9] L. Schwartz: Mcthodes mathtrnatiques pour les sciences physiques, Hermann, Paris, 1965. (A.10) I. Stakgold: Boundary value problems of mathematical physics, vol 11. MacMillan Co., New York, 1968. (A.11) V. Vladimirov a.o.: Recueil de problemes d'equations de physique rnathcrnatique, Mir, Moscou, 1976. BIBLIOGRAPHY 567 [A.121 V. Vladimirov: Equations of mathematical physics (translated from Russian into Romanian), Editura $tiiutific i §i Enciclopedica, Bucureti, 1980. [A.13] V. Vladimirov: Distributions en physique mathernutique, Edition Mir. Aloscou, 1979. [A. 14] C. Zuily: Distributions et equations aux derrivecs partielles, Hermann. Paris. 1986. Singular Integrals. Integral Equations (A.15) A. Carabineanu. D. Bena: Aletoda transforindrilor conform' pentru donaenii vecine, Ed. Academici Romaine, Bucurcti, 1993. (A.161 L. Dragon, N. Marcov: About some singular integral equations of interest in aerodynamics, Z. Angew. Math. Mech., 79(1999) 3, 205 - 210. [A.17] M. Eichler: Konstruction losender Kerne fur singulire integral gleschungen erster Art insbesondere bei Diferen:. kcrnern, Math. Zs. 48(1941) 503. [A.18] F.D. Gakhoff:Boundary problems (in Russian), Fizmatghiz, Muskow, 1958. [A.19] D. Homentcovschi: Sur la resolution erplicitc du probleme de Hilbert. Application en calcul de la portance dun profile mince dans an fluid electroconducteur, Rev. Rouen. Math. Pures et Appl., 14(1969) 203; Contributions to the study of Hilbert's problem Applications to magnctohydrodynamics. Studii Cerc. Mat., 23(1971) 727. (A.20J D. Homentcovschi: Func;ii complexe cu aplica;ii in §tiinjd ,si tehnicd, Ed. Tehnici . Bucurc ti, 1986. [A.21) C. lacob: Sur uric equation integrede singuliere, Mathematics, 23(1948) 153. [A.22] C. Iacub: Funec;ii complexe in (1.22], vol. 2. [A.23] A.I. Kalandiya: Mathematical methods of two-dimensional elasticity, Mir Publishers, Mocwouw, 1975. 568 BIBLIOGRAPHY [A.24] A. Levinson: Simplified treatment of inteyraLr of Cauchy type. the Hilbert problem and singular integral equations, SIAM Rev. 7(1968) 474. [A.25] S.G. Mikhlin: Multidimensional singular integrals and integral equations (in Russian) Gostekhizdat, Moskow, 1962. [A.26] S.G. Mikhlin: Linear equations t.}:!h partial derivatives (in ltussian ), Moskow, 1977. [A.27] N.I. Muskhclishvili: Singular integral equations (in Russian), TO ed. Fizmatghiz, Moskow, 1962: the first ed. is translated in English, P.Noordhoff, Groninlen, 1953. [A.28] A.S. Peters: Abel's equations and the Cauchy singular integral equation of the second kind, Conun. Pure Appl. Math 21(1968) 51. [A.291 J. Plemelj: Ein Ergnn.-ungssatz zur Cauchy'cchen integraldarstellung analytischer Funktioneri.. Raudwerte betrrfencl, Monatsch. f. Math. u. Phys, XIX, Jahrgang, (1908) 205. [A.30] N. Rott, H.K. Cheng: Generalizations of the inversion formula of thin airfoil theory, J. Rat. Mech. Anal., 3(1954) 357. (A.31] C. Samko, A.A. Kilbas, 0.1. Marichev: Integrals and fractionar products and some of their applications (in Russian), Nauka i Technica, Minsk, 1987. (A.32] W. Schmeidler: Integral gleichungen mit Anurendungen in Physrk and Technik, Leipzig, 1955. Uber eine Integralglcichung erster Art der 7}uglnigeltheorie, Sitzungsber. d. Preuss. Akad. d. Wiss., Phys. [A.33] K. Schroder: Mat. Klasse 30(1938) 345. [A.34] H. Sohngen: Die Losungen der Integral-gleichung ... and dervn Anwendung in der Trag,/lugeltheorie, Math - Zeitschr., 45(1939) 295. [A.35] G. 'fficomi: Formula d'inversione dell'ordinc di due integruzi doppie "con asterisco", Atti della Reale Acad. Naz. dei Lincei, 323(1926) 53. BIBLIOGRAPHY 569 Finite Part [A.36] M.P. Brandio: Improper integrals in theoretical aerodynamics. The problem revisited, AIAA Journ. 25(1987) 1258. (A.37J Ch. Fox: A generalization of the Cauchy principal value, Canadian J. of Math., 9(1957) 110. [A.38] M. Guiggiani a.o.: General algorithm for the numerical solution of hyper - singular boundary integral equations. .1. Appl. Mech. 59(1992) 604. (A.39) J. Hadainard: Lectures on Cauchy's problem in linear partial differential equations, Yale University Press, 1923. [A.40] M.A. Heaslet, H. Lomax: Supersonic and transonic small perturbation theory, Appendix in [1.37]. [A.41) A.C. Kaya, F. Erdogan: On the solution of integral equations with strongly singular kernel, Quart. Appl. Math., 45(1987) 105. (A.42J H.R. Kutt: The numerical evaluation of principal value integruls by Finite Part integration. Nuttier Math., 24(1975) 205. [A.43] E.G. Ladopoulos: Finite - Part singular integro - differential equations arising in two - dimensional aerodynamics, Arch. of Mechanics, 41(1989) 925-936. (A X141 H. Loinax, M.A. Heaslet, F.B. Fuller: Integrals and integral equations in linear*zed unng theory, NACA Rep. 1051(1951). [A.45] K.W. Mangler: Improper integrals in Theoretical aerodynamics, British Aero. Res. Council, R & M 2424(1951). [A.46) D.F. Paget: The numerical evaluation of Hadarnard Finite part integrals, Nuiner. Math., 36(1981) 447. [A.47] S. Voutsianas, G. Bergeles: Numerical calculation of singular integrals appearing in 3-D potential floey problems, Appl. Math. Model, 14(1990) 618. Quadrature Formulas [A.48] S. Ando, A. Ichikawa: Quadraturc formulas for char wise integrals of lifting surface theories, AIAA Journ., 21(1983) 466. 570 BIBLIOGRAPHY [A.49J L. Dragoq, M. Popescu: Certain quadrat*tre formulae of interest in aerodynamics, Rev. Rouen. Math. Pures et Appl, 37(1992) 587. [A.50J D. Elliot, D.F. Paget: Gauss type quadrature rules for Cauchy principal value integrals, Math. Comp. 33(1979) 301. [A.51] D.B. Hunter: Some Gauss - type formulae for the evaluation of Cauchy principal value of integrals, Nunier. Math., 19(1972) 419. [A.521 G. Monegato: On the weights of certain quadratures for the numerical evaluation of Cauchy principal value integrals and their derivatives, Nunter. Math. 80(1987) 273. Constructive function theory (in Russian), [A.53J P. Natanson: Gostekhizdat, Moskow - Leningrad, 1949. [A.541 V.J.E. Stark: A generalized quadrature formula for Cauchy integrals, AIAA Journ., 9(1971) 1854. [A.55] A.H. Stroud, D. Secrest: Gaussian quadrature formulas, Prentioe Hall, N.Y. 1966. (A.56) G. Szego: Ortogonal Polynomials, Amer. Math. Soc. Colloq. Publ., vol. 23, 4th ed. 1975. Index Abel's equation, 320 Acceleration as material derivative , 2 of the particle, I potential, 9, 397, 404 Ackeret's formulas, 288, 423 Acoustics, equation of, 29 Aerodynamic action drag, 171, 172 lift, 76, 170, 172 moment, 76, 170 gyration, 171, 172 pitching, 171, 172 rolling, 171, 172 Arrow of airfoil, 83 shaped wing, 156 Attached shock waves, 361 Complex velocity, 71, 75, 120, 371, 375 Compressibility effects, 86, 148, 231 Conduction law, 13 Cone body's shape, 463 double. 464 Conical flow, 339-347, 464 Conservation laws, 11 Convolution, 33, 370, 470 Critical velocity, 6, 359 Delta wing, 156 Descartes, folium, 18 Detached shock waves, 19, 360 Distributions, theory of, 465.479 Doublet Bernoulli's integral, 5-7 Bessel's functions, 407, 503 Biplane, 92 density, 266 flow induced by, 148 potentials, 426 Drag coefficient co, 181, 203, 204, 213, 229, 231, 241, 292, 458, 462, Body, theory of slender, 449-464 R":rndarv conditions Duhamel principle, 57 airfoils in tandem, 1:,Y airfoils parallel to the undisturbed stream, 93 grids of profile, 99 ground effects, 83 material surface, 24 rest state, 28 tunnel efects, 88 uniform motion, 25, 70 Boundary conditions (nonlinear case), 464 Energy, 3, 4 Enthalpy, 15 Entropy, 4 Euler -Lagrange criterion, 24 constant, 400 equation, 3 formulas, 146 theorem, 2 111, 112 Flutter, 397 Caloric equation, 4 Cauchy integral, 71 principal value, 482, 510 problem, 61 Characteristics coordinates on, 308 variables on, 318 Chebyshev polynomials, 80, 525, 526 Circulation, 125, 126, 274, 393 Clapeyron's equation, 4 Collocation method, 115, 231, 279 Complex potential, 121, 146 Flux-divergence formula, 479 Forces, continuous distribution of, 73 Fourier, transform, 465-479 Fredholm integral equation, 209 Fundamental matrices, 61 Fundamental solutions equation of potential fluid at rest, 42 M = I oscillatory, 34 M = 1 unsteady, 41 subsonic oscillatory, 33 subsonic steady, 31 subsonic unsteady, 37 supersonic oscillatory, 34 INDEX 572 Fundamental solutions equation of potential supersonic steady, 31 supersonic unsteady, 38 oscillatory system pressure formulae, 50 velocity formulae, 52 steady system general form, 45 plane subsonic, 46 plane supersonic, 48 3-D subsonic, 47 3-D supersonic, 48 unsteady system, 57 Causs-type quadrature formulas, 521530 Clauert integral, 489 method, 210 Gothic wing, 156 Green function, 142, 143 Ground effect, 82, 86, 136, 184, 238, 299 Hadamard finite part, 509 type integrals, 513 harmonic forces, 398 Heat, specific c,,, c.,, 4 Helmholtz's equation, 5, 33 Henkel's function, 33 Homentropic motion, 4 Huygens' principle, 42 Ideal fluid, 1 Instable shock, 18 Integral equations steady subsonic flow lifting line, 201, 209, 219 lifting surface, 165, 167, 187 airfoils in tandem, 103 grids of profiles, 99 ground effect, 84 parallel airfoils, 94 tunnel efect, 90 thin profile, 72 steady supersonic flow lifting surface, 307, 313, 320 Integral equations steady transonic flow lifting line, 395 lifting surface, 389 unsteady flow sonic profile, 440 sonic wing, 445 subsonic profile, 402 subsonic wing, 408 supersonic profile, 418 supersonic wing. 432 Invariant, 16 Irrotational flow condition, 368 definition, 5 equation of, 109 Isentropic motion, 4 Joukovaky profile, 86 Lagrange -Cauchy's theorem, 5 interpolation, 223 variation of constants, 218 Leading edge, 62, 156, 301 subsonic, 301 supersonic. 301 Lift coefficient, 129 Lift coefficient CL, 76- 78, 81, 82, 85, 86,90, 96, 100, 104, 107, 171, 180, 183, 193, 195, 202, 204, 213, 229, 231, 237, 241, 287, 288, 291, 293, 298, 299, 350, 354, 357, 380, 421, 423, 441, 457, 462, 464 Mach angle, 40 cone, 31 dihedron, 32 number, 8, 19 Moment coefficient cm, 76-78, 81, 82, 85, 86, 90, 96, 100, 104, 107, 287, 288, 291, 298, 299, 421, 423, 441 Moment coefficient cs, 203, 204, 229, 231, 241, 458, 462 Moment coefficients c1, cv, 171, 181, 183, 19.3, 195, 2(r2, 229, 231, 237, 241, 458, 462. 464 INDEX 573 Plemelj's formulae, 483 Prandtl's theory, 197-205 Prandtl-Mayer fan, 289 Source mass, 43 Pressure coefficient Cp, 120, 129, 148 Swallow wing, 156 Rhombic wing, 156 Thermodynamics, equation of, 3 Trailing edge, 156. 301 Shock waves Hugoniot's equation, 15 jump equations, 13 Prandtl's formula, 16 shock polar, 18 Sonic barrier, 438, 448 Sonic circle, 18, 19 potential, 36 Kutta-Joukovski, 81, 165, 275,332 subsonic, 301 supersonic, 301 Trapezoidal wing, 156 Vortex, 5, 71 KLUWER ACADEMIC PUBLISHERS EDITURA ACADEMIEI ROMANE ISBN 973-27-0986-3