INFERENTIAL MEASUREMENT OF AERODYNAMIC FORCES FOR USE IN WING DIVERGENCE ANALYSIS by James Smith McDonnell, III S.B., Princeton University (1958) SUBMITTED IN PARTIAL FULFILIENT OF THE REQUIREMENTS FOR THE DEGREE OF MASTER OF SCIENCE at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY September, Signature of Author 1959 Signature redacted Department of Aeronatics and Astronautics, July 15, 1959 Certified by Signature redacted Thesis Supervisor Signature redacted Accepted by Chairman, Departmental Commi tee on Graduate Students , OF TECHNo OCT 8i19 INFERENTIAL MEASUREKENT OF AERODYNAMIC FORCES FOR USE IN WING DIVERGENCE ANALYSIS by James Smith McDonnell, III Submitted to the Department of Aeronautics and Astronautics on July 15, 1959, in partial fulfillment of the requirements for the degree of Master of Science. ABSTRACT A method of divergence analysis has been formulated in which strain measurements are utilized to determine aerodynamic influence coefficients. Using this method, a divergence analysis of a cantilever delta.-type wing was carried out, and the predicted values of divergence speed compared with the actual values. All work was performed at low subsonic speeds. Thesis Supervisor: Associate Professor of Aeronautics and Astronautics - ii - Title: Erik L. Mol1d-Christensen July 15, 1959 Professor Alvin Sloane Secretary of the Faculty Massachusetts Institute of Technology Cambridge 39, Massachusetts Dear Professor Sloane: In accordance with the regulations of the faculty, I hereby submit a thesis entitled Inferential Measurement of Aerodynamic Forces for Use in Wing Divergence Analysis in partial fulfillment of the requirements for the degree of Master of Science. Respectfully, James S. McDonnell, III iii - - ACKNOWLEDGEMENTS The author wishes to express his appreciation to the following persons: Professor Erik Moll-Christensen for his advice and generous assistance throughout this work; Oscar Wallin for his continual help in constructing the test apparatus; Al Stone who mounted the strain gages in the wing model; Professor Eugene Larrabee for his assistance with the Student Wind Tunnel; and Mrs. C. E. Smart who typed the - iv - manuscript. TABLE OF CONTENTS Chapter No. Page No4 Object 1 Introduction 2 Derivation of Method for Predicting Divergence Speed of Wing 5 Test Apparatus 14 Experimental Procedure 18 Summary of Data 21 Calculations and Results 26 Discussion of Results 34 Conclusions 37 Accuracy of Angle of Attack Obtained from Strain Measurements 38 Testing Wing for Actual Divergence Speed 39 Design Calculations of Spar - Jones's Low Aspect Ratio Theory 40 Calibration of Wind Tunnel 46 References 47 Appendices A B C D PRINCIPAL SYMBOLS [B] matrix of integration weights C(x, p) flexibility influence function E modulus of elasticity EI bending stiffness [G matrix of integration weights diagonal matrix of integration weights section moment of inertia lift per unit chord ) L ( M Ql(' ..r), Q2( U bending moment r) aerodynamic influence functions velocity applied wing load c maximum chord of wing e bending strain h spar thickness length of spar q dynamic pressure r streamwise coordinate 5 distance in z-direction measured from neutral axis of wing w width of spar x streamwise coordinate - - vi - 7 spanwise coordinate y linear deflection of wing Ax chordwise distance o( angle of attack wing semispan taper ratio of spar Subsc d denotes conditions at divergence 0 denotes initial conditions t denotes values at spar tip t.e. denotes values at wing trailing edge a,b,...; 1,2,3,... indicate chordwise wing stations Superscript ( )' indicates chordwise differentiation - vii m OBJEOT The object of this thesis was to determine if a divergence analysis of a wing could be performed by utilizing strain measurements to evaluate aerodynamic influence coefficients. - - 1 CHAPTER 1 INTRODUCTION Static divergence of a wing may be described as "the instability that results when the structural restoring forces arising from a small structural deformation (of the wing) are overpowered by the aerodynamic forces arising from the deformation" (reference 1). For the usual type straight or sweptback wing, divergence will ordinarily occur in a torsional mode; that is, the wing is no longer able structurally to withstand the torsional loads acting on it. In the majority of cases, however, divergence of this type wing is of secondary importance in comparison to other aeroelastic phenomena such as flutter and aileron reversal. On the other hand, with the use of low aspect ratio wings and control surfaces, in particular the Canard-type configuration, on some of the present missiles, there is the possibility that the problem of wing divergence will assume renewed importance. Consequently, then, it will become necessary to be able to predict at what speed, or more specifically, at what dynamic pressure a given wing will diverge. And to be of maximum value, such divergence predictions not only must be reliable, but also must be accomplished while the missile or aircraft is still in the design stage. Before any sort of divergence analysis can be performed on a wing, however, an estimate of the aerodynamic forces that the wing will encounter =2 -Th must be made.* Naturally, the accuracy of the ensuing divergence analysis will be no better than that of the force estimation. For wings of simple shape and thickness distribution, expressions for the lift forces acting on the wing can be obtained analytically (eg., see references 1, 2, 3, or 4). Such expressions, however, besides being quite complex at times, are usually limited in usefulness and accuracy because of restrictions and simplifying assumptions necessary in their formulation. Moreover, for wings of complex shape and thickness distribution, it becomes no longer possible to obtain expressions for lift forces by analytical methods without having to make drastic approximations0 A more reliable approach to the divergence problem, then, would appear to be the utilization of experimental data (obtained from a wind tunnel model) in the estimation of the aerodynamic forces to be encountered by a given full scale wing. Such an approach is undertaken in this thesis, the object of which was to determine if the divergence speed of a wing could be predicted by utilizing strain measurements. The wing model con- sidered was a delta-type wing whose apex pointed upstream and whose trailing edge was restrained from deflecting. For such a wing divergence occurs not in torsion as described previously, but instead in a streamwise bending mode. This is the simplest type of divergence and lends itself well to the use of strain gages for determining bending stresses. With the use of strain measurements taken on the previously mentioned wing model, two aerodynamic influence functions were determined which could * In the more general case of dynamic divergence, inertial loads would also have to be considered. Here, however, only static divergence will be dealt with. Hence, inertial loads will be omitted from the analysis. be used to predict the lift forces the wing would encounter in any given deflection shape. Once determined, these influence functions are used to predict the divergence speed of the wing for any change in bending stiffness distribution without the need of additional wind tunnel tests. The advantage in this approach to the divergence problem lies in the fact that the previously described aerodynamic influence functions can be determined for any given wing; no restrictions or assumptions concerning shape or thickness distribution are necessary. To simplify construction, however, the wing model analyzed in this thesis was of constant thickness, and a simple bending stiffness distribution was employed. But as will be seen in the following sections, these simplifications were by no means necessary. All experimental work performed in conjunction with this thesis was carried out in the M.I.T. Student Wind Tunnel. Using the method of divergence analysis formulated in the following sections, divergence speeds were predicted within eight percent of their actual values and dynamic pressure at divergence within seventeen percent of true value. Although all work was performed at very low subsonic speeds in order to concen. trate on the method itself, the successful results obtained indicate the possibility of extending this type of approach to the divergence problem to much higher speeds, supersonic and possibly even hypersonic. CHAPTER 2 DERIVATION OF METHOD FOR PREDICTING DIVERGENCE SPEED OF WING Formulation of Theory (2.l) Consider a delta-type wing whose trailing edge is rigidly restrained from deflecting and whose apex is pointing upstream. (See figure 1.) The origin of an xyz-coordinate system will be located at the wing trailing edge with the maximum chord line coinciding with the x-axis. its undeflected position the wing will lie in the xy-plane. velocity U emanates from the positive x-direction. In A steady For purposes of gen- erality, the wing here will be assumed to have an arbitrary thickness distribution. Assuming the wing geometry to be symmetric with respect to the x-axis, the lift distribution will also be symmetric about the x-axis. Conse- quently, deformations of the wing will occur in bending only and only in a chordwise direction. Then assuming the wing structure to be linear, the deflection of the wing at any point x due to aerodynamic lift forces can be obtained using a one-dimensional influence function. Thus, C. z(x) (x( L = 0 deflection at point x due to a unit lift force at point ' ; and L( ( ) where C(x, g ) is a flexibility influence function and equals the linear is the aerodynamic lift per unit chord acting on the wing. For sufficiently small deflections, the angular deflection of the wing caused by the aerodynamic lift forces acting on it may be approximated by the slope of the wing. Hence, C the angular deflection of the wing at x due to a where unit lift force at The aerodynamic lift per unit chord acting on the wing will now be expressed asC 0 0 where q is the dynamic pressure, and , r) and Q2 ( ( dynamic influence functions, Q: (j , r) is the lift ( , r) are aero- coefficient induced at by the downwash (or upwash) resulting from a unit angle of attack at r; Q2 ( , r) is the lift coefficient induced at ( by the downwash (or upwash) resulting from a unit change in angle of attack at r. 2 and 3.) Notice that from their definitions both Q1 (( , (See figures r) and Q 2( will depend only on wing geometry (i.e., shape, thickness, etc.). they will be independent of wing bending stiffness and dynamic pressure. Introducing equation (2.3) into equation (2.2) and remembering that * Justification for expressing lift as a function of both o( and do/dr is implied by Jones's low aspect ratio theory. (See reference 2; also reference 3, pages 247 and 248.) ** In the more general case of compressible flow, Mach number must also be included. - - 6 ,r) Q( cX 4 W(x) = C((X) -cx(x) (2.4) where co(x) is initial (zero velocity) angle of attack, there follows: r - 0((X 00 Letting [H] be a matrix of integration weights, equation (2.5) may be expressed in matrix form as: {I} - < = g [C'][ H]([QiH1{) + [Q]FH where the primes denote chordwise differentiation. consist of values of o(, o( , and o(' (2.{6') (' The column matrices taken at discrete points along the chord. Equation (2.6) will be the underlying equation in the method of divergence analysis presented in the following two sections, After [C1] has been determined, either analytically or by static measurement, equation (2.6) is used to determine [Q,] and [Q 2] . Then once these two matrices are known, (2.6) is used to predict divergence speeds for any desired distribution of wing bending stiffness. (2.2) Determination of the Aerodynamic Influence Coefficients Referring to equation (2.6), the elements of [Qlj and [Q2] are determined by first giving the wing model an initial angle of attack dis. tribution {o(0), then measuring values of {o(o and {o(') with the wind tunnel running at a velocity less than the divergence speed of the model. if [Q] and [Q 2] are to be nxn matrices, then a total of 2n wind tunnel runs will be required to determine all of their elements. In each run the wing must be given a different initial angle of attack distribution, As will be shown later, these initial angle of attack distributions must exhibit distinct linear independence. In chapter 4 the method of obtaining the initial wing deflections will be outlined. Since a series of wind tunnel tests are to be run, and not necessarily at the same tunnel speed, equation (2.6) must be rewritten as: [o(] - [] = ['J[HJ (EQaEH][cK]Jr_ Each column of of {o( (o(.3 [o(, [o(j, and , and (o('} + [Q1J]W) [o('] will now correspond to the values obtained in one of the tunnel tests. The problem now is to set up some means of measuring o(, &( (.i o( , and at given chordwise locations. This may be done by taking strain measurements at discrete points along the chord. time being, wing bending stress - a- Neglecting signs for the may be expressed as: Ms I (x) where M is the bending moment; s is the distance from the neutral axis to the point at which stress is being computed; and I is the section moment of inertia. Since the wing model will deflect only in chordwise bending modes, it is assumed that it will obey the mechanics of a simple beam. Hence, (.9) M =Ex) where E is the wing modulus of elasticity. Combining equations (2.8) and (2.9) there results: 0~ = E sJz jy For sufficiently small deflections (2.10) 2 .c((x)~ d= (X) , dO(X) = d' W and Hence, equation (2.10) may be rewritten as: dl(() a sE But 7= eE , where e is bending strain. Substituting this into equation (2.11), there results, in matrix form- [o j =[T]Le] (2.12) where the prime again denotes chordwise differentiation. [~~] 2.1'3) Likewise, e0 is the bending strain due to the initial angle of attack dis- tribution. A positive lift force will now be defined to produce positive where strain; and positive strain in turn defined to produce positive Using the trapezoidal rule for numerical integration, determined from values of o(' (x). o(. &((x) can be Denoting the station at the wing trailing edge by the subscript 1 and progressing in the positive x-direction, there results, for an equal interval AX 4(K &x between stations: + _ (z14) o(), and so forth. Notice that positive values of c( will produce positive values of o( Defining [B] to be the matrix: 0 0 0 0 0.5 005 0 0 0.5 1 005 0 0.5 1 1 0.5 (2.15) equations (2.14) may be written in matrix form as: [j] = zX [B][ Substituting equation (2.12) into (2.16): B [ e] - 10 - [c(] = Ax (2.17) . + o o( + + 0(4 =AX( In like manner: [] [o( [o( ] , , and 6 [(' (2.18) can now be computed from strain measure- ments through the use of equations (2.12), (2.13), (2.17), and (2.18), then placed in equation (2.7) to determine method used to solve (2.7) for Q1] and [Q2 ]. The [Q1] and [Q2] will be outlined in chapter 6. (2.3) Method of Divergence Analysis Letting {o() - 0, equation (2.6) becomes: Cj= Once [C ] 1([Q jH o(] + [H] .o(.) [Q 1] and [Q2] have been determined, this is the basic equation from which the divergence analysis of the wing will be performed. But first it must be put in a more workable form by expressing fcOG in terms of {o(} Since the wing model is restrained so that its trailing edge will not deflect, the elements of will be zero. [C] Consequently, elements of pertaining to the trailing edge Q] and [Q2] are indeterminate there, and the corresponding row and column of these two matrices must be omitted from the divergence calculations. So that the matrices in equation (2.19) remain conformable, the term in {o(} and edge also must be deleted. {o'( at the wing trailing Consequently, equation (2.16) loses its con- formability, and a new relation between {c*} and f is necessary. Henceforth, for the purposes of divergence calculations, the following expressions will be employed: a 'O (AX),+ ~ and so forth. where the subscripts denote wing stations progressing in the positive xdirection; (,6x) 8 is the chordwise distance between the trailing edge and station a; and (1 X)b the distance between stations a and b. wing stations can be added in a similar manner. Additional Equations (2.20) are obviously a cruder approximation than equations (2.14); but for the divergence mode of the wing, they should still give reasonably accurate results since o((x) increases from zero at the trailing edge to its maximum value at the apex, and o(' (x) remains positive over the entire chord. LG] to be the matrix: 0 L&X (AX)b] equations (2.20) may be written in matrix form as: f - [G] {&X3 (.? 2 and subsequently substituted into equation (2.19) to yield, after rearranging: = [H] [ H[G - 12 - [G]o + [_j('H]) C(2.3) Defining The two quantities in equation (2.23) which a divergence analysis will be concerned with are q and [C9] If . [Cf] is known, then (2.23) can be solved for qd, the dynamic pressure necessary for divergence, by either of two methods. For a nontrivial solution of equation (2.23) to exist, its principal determinant must vanish. WG] CHI ( G] - Hence, +[Q][H1) (2.24) = 0 Or alternatively, qd can be determined through matrix iteration by expressing equation (2.23) in the form: {o. [T c' [H[J) fo( HI L1[]G (2.2S) In general the latter method will probably be the easiest to carry out. Before carrying out the actual divergence analysis of the wing, equation (2.24) or (2,25) should first be used to predict the divergence speed of the wing model used in the wind tunnel tests. Its predicted divergence speed can then be compared to its observed divergence speed so that the over-all accuracy of the method can be determined. The object of the divergence analysis is to predict what changes will occur in divergence speed when the bending stiffness distribution of the wing is altered. Since [Q 1] and [Q2] are independent of wing bending stiffness, a change in bending stiffness will show up only in [C . Hence, a new value of IC corresponding to the new bending stiffness distribution can be calculated, then introduced into equation (2.24) or (2.25) to determine the new divergence speed of the wing. No additional wind tunnel tests are necessary, - - 13 CHAPTER 3 TEST APPARATUS (3.1) Wing Model The wing model used in the wind tunnel tests was built by the Aeroelastic and Structures Research Model Shop. As mentioned previously, it was triangular in shape and cantilevered at its trailing edge. figures 4 and 5.) (See The wing itself was constructed of balsa wood and divided into chordwise sections for reasons to be explained shortly. Through the center of the wing (in the streamwise direction) ran an aluminum alloy spar. If so desired, this spar could be given a shape and thickness distribution to simulate the bending stiffness distribution of an actual wing. A change in bending stiffness could be accomplished by simply altering the shape of the spar. The entire bending stiffness of the wing model was concentrated in the spar by dividing the body of the wing into chordwise sections, each of the sections being clamped to the spar at only one point. Sufficient room was also left within each of the sections to allow the spar to bend freely without touching or rubbing one of them. (See figures 6, 7 and 8.) Consequently, the balsa would contribute no bending stiffness to the wing. As a result, the bending stiffness distribution of the model could be accurately set, Also, as will be discussed in the next chapter, the static measurement of [Cj was simplified considerably since it could be performed on the spar alone. The maximum chord of the wing was set nominally at twenty-seven inches. After rounding off the apex, an actual length of 26 7/16 inches resulted which included a 1/32 inch gap between the sections to prevent them from touching when the wing deflected. There were a total of seven sections, the tip section being 3 3/8 inches in chordwise length and the other six 3 13/16 inches. The spanwise edges of the sections were also rounded off, but the wing trailing edge was left blunt. On both sides of all sections, mounted flush with the wing surface, was a small aluminum alloy plate into which were anchored screws to clamp the sections to the spar. hooks could also be screwed into each of the plates. Small To the hooks were attached restraining strings which, when secured to the walls of the wind tunnel, prevented the model from deflecting past the elastic limit of the spar. Also attached were strings by which the model could be given an initial angle of attack distribution. with more fully in chapter 4. This latter aspect will be dealt The model span (i.e., width at trailing edge) was set at 131 inches which resulted in an aspect ratio of unity. To simplify construction of the wing, its thickness was made 5/8 inches throughout. Since the gaps between the wing sections were considerably smaller, it was assumed that viscosity would hinder air flow between the sections sufficiently so as to prevent an abnormal lift distribution from occurring. The wing spar was constructed of 75ST aluminum alloy. For its modulus of elasticity, a value of 10.4 x 106 pounds per square inch was The spar had a constant thickness of 0.10 = 15 - used in all calculations. inches, and its width tapered from 1.69 inches at the wing trailing edge to 0.10 inches at the forward tip. The length of the spar within the wing model was 26 inches, and an additional 6 inches protruded aft of the trailing edge for purposes of mounting the model in the wind tunnel. In designing the spar, Jones's low aspect ratio theory was used to obtain an initial estimation of model divergence speed. These calculations appear in section C of the appendix. Commencing at the wing trailing edge, a pair of strain gages (Baldwin SR-4 type A-7, gage factor: 1.96) was mounted every 6 inches along the spar. (See figure 6.) A total of five pairs was required, the strain gages in each pair being located on opposite sides of the spar. tional dummy gages were used. No addi- The leads from the strain gages were run out through the wing trailing edge, then along the model support and through the tunnel floor. (See figures 5 and 6.) The wing was mounted in a vertical plane so as to simplify the problem of giving it initial angle of attack distributions. tion of the wing due to its own weight was eliminated. Also deflec- The model support consisted of a length of four-inch steel channel welded to a steel floor plate which was bolted to the floor of the test section. The flanges were removed from the top part of the channel so that two steel blocks could be bolted there, The portion of the spar protruding from the wing trailing edge was then clamped between the blocks. 5.) (See figures 4 and A section of an old helicopter rotor blade was used as fairing around the lower portion of the channel. Additional Apparatus All experimental work was performed in the M.I.T. Student Wind - 16 - (3.2) This tunnel has a test section six feet wide by four and one- Tunnel. half feet high. In the floor of the test section is a turntable on which a model can be rotated about a vertical axis. the wind tunnel was recalibrated. Before test work was begun, The results appear in section D of the appendix. The leads from the strain gages were connected through a switch-box to an SR-4 Model K Strain Indicator. Because of model oscillations caused by tunnel turbulence, it was impossible to zero out the strain indicator directly with any degree of accuracy. Consequently, a resistance- capacitance filter was installed between a vacuum tube voltmeter and the output terminals of the strain indicator. This filter effectively reduced the amount of a-c signal entering the voltmeter so that it was possible to zero out the strain indicator by using the voltmeter as a null indicator. Through experiment it was found that a resistance of ten kilohms and a capacitance of 790 microfarads produced optimum response in the voltmeter; that is, the oscillation lowest in frequency and amplitude while at the same time maintaining sufficient sensitivity in the voltmeter response for reasonable accuracy in zeroing it out. 17 - - CHAPTER h EXPERIMENTAL PROCEDURE (4.1) Static Measurement of [C'] ) is the angular deflection of the As mentioned previously, C'(x, ' wing at x due to a unit lift force at measuring unit lift [C] , C'] was determined by first , where C(x, 4 ) is the linear deflection at x due to a force at 4 * Then using the elements of corresponding deflection curves, the elements of by measuring slopes of these curves. [C] [C1] to draw the were determined Since the balsa wing sections made [C'J no contribution to the bending stiffness of the wing, was deter- mined using only the spar. The linear deflections were obtained by hanging scales from the spar and using a surveyor's level to take the readings. were made every three inches along the spar. Measurements The equation governing the spar deflections can be expressed in matrix form as: [c] where [z is the matrix of spar deflections, and of applied loads. (4.1) [W] is the matrix In order to increase the accuracy of the static tests, W] was made a unit diagonal matrix. - 18 - L z]vj (See reference 4.) The resulting values of [C] (4.2) and [C'j appear in the following chapter. Wind Tunnel Test Procedure The wing model was given initial angle of attack distributions by hanging weights from strings attached to the model, These strings were run through small holes drilled in the tunnel walls and then passed over pulleys to minimize friction. wing from deflecting. Thus, the weights did not restrain the By varying the sizes of the weights and their points of application on the wing model, different deflection shapes could be obtained. But as mentioned previously, considerable care must be taken to make sure that the initial angle of attack distributions are distinctly linearly independent. chapter 6. lift Reasons for this will be discussed in Since equation (2.7) was derived in terms of aerodynamic forces only, the forces due to the weights used to obtain the initial deflections of the wing do not enter into the equation. Because of the necessity that there be distinct linear independence among the initial angle of attack distributions, only four test runs in the wind tunnel could be made. ] mine elements of [Q 1 and As a result, it was possible to deterat only two wing stations. [Q 2] However, strain readings were taken at all five strain gage locations in order to improve the accuracy of equations (2.12), (2.13), (2.17), and (2,18). Following is a summary of the four runs; the minus signs denote weights acting in the direction of negative lift. of the various wing sections. Figure 10 shows the location The corresponding initial deflection shapes are shown in figure 11. 19 - - Summary of Wind Tunnel Runs Run No. Tunnel Speed (ft./sec.) 1 73.9 2 77*3 Locations and Approximate Sizes of Weights -2 lbs. from section D, and 3/4 lb. from section G. 21 lbs. from section C, -2 lbs. from section E, and } lb. from section G. 3 6o.5 h 6o..5, 1 1b. from section G. a5.53* (o.o965o radians) In runs 1-3 the test section turntable was clamped so that in its undeflected position, the wing was at zero angle of attack. In run 4 the turntable was rotated so that a constant initial angle of attack was obtained over the entire wing. The tunnel speed in each test run was made high enough so that a significant difference was obtained between all strain readings from any one strain gage. errors in the readings would be minimized, In this way the percent On the other hand, the tunnel speeds had to be kept low enough so that the wing would not deflect against the restraining strings, the presence of the weights in some cases lowering the divergence speed considerably. run 3, for example.) (See figure 11, In order to minimize changes in tunnel temperature from one set of readings to another, the zero velocity readings were always made at the conclusion of a test run, And to eliminate friction forces within the pulleys, they were always tapped before the zero velocity readings were taken. 20 - - CHAPTER 5 SUMMARY OF DATA (5.1) Flexibility Influence Coefficients Spar deflections were measured every three inches along the spar and at the spar tip. 0 0 (See figure 9.) 0 0.02 0 zI = The results were, in inches: 0 0 0 0.05 0.07 0.10 0.15 0.24 0.34 0.52 0.70 0 0.10 0 0.16 o.49 0,85 1.20 0 0.19 0.68 1.26 1082 0 0.25 O0186 1071 2053 0 0029 1.04 2.15 3.34 0 0034 1.23 2.61 4.21 0 0.37 1.35 2.89 4,77 The columns in the above matrix correspond to the successive loadings. The numbers underlined are the elements of [C] , i.e., the deflections at strain gage locations. The other numbers are the deflections at intermediate points (i.e., spar stations - M 21 a - e). The resulting matrix of C '(x, g ) turned out to be, in radians per pound: 0 - [Cv] 0 0 0 0 0 0.O155 0.0427 0.0768 0.1073 0 0.0155 0.0607 0.1325 0.1902 0 0.0155 0.0607 o.1475 0.2583 0 0.0155 0.0607 O.1475 0.2801 I (5.2) Wind Tunnel Data The strain readings from the wind tunnel tests were as follows, where R0 is the zero reading, Ri the reading with the wing initially deflected, and Rf the final reading (tunnel running): Run #1 Run #2 Sta.2 Sta. 3 Sta. 4 Sta. 5 Rf : 5,590 4,835 6,995 7,720 5,490 Ri: 5,770 5.9010 7,020 7,670 5,1465 Ru: 6,450 5,9250 6,285 6,705 5,o55 Rf: 6,370 49700 5,735 7,355 5,9375 Ri: 6,570 4,970 5,930 7,430 5,380 RO: 6,490 5,280 6,300 6,710 5,060 Refer to figure 9 for location of wing stations. 22 . * Sta.l* Run #3 Run #4 R : 7,950 6,900 7,575 7,560 5,30 R : 7,2h0 6,150 7,080 7,370 5,330 R 6,530 5,320 6,315 6,715 5,o6o R 8,025 6,600 7,030 6,940 5,080 5,300 6,310 6,710 5,060 Ri: --- ,: 6,520 Since a pair of strain gages was located at each wing station, the strain readings must be halved to obtain the actual values of strain. Hence, e.o= where e 2 (s.i) e = R_-R. ; R-R. is the final bending strain (in microinches per inch), and eo is the initial bending strain (due to the initial angle of attack distribution), The strain matrices then become, with their columns corresponding to the successive tunnel runs: [e 0 ] = 10-6 x -340 4o 355 0 -120 -155 415 0 367 -185 383 0 483 360 327 0 205 160 135 0 -23 so [e] 10- 6 -430 -60 710 753 -207 -290 790 650 355 -283 630 360 507 323 423 115 217 157 140 10 As mentioned previously, the bending stiffness of the wing is concentrated in the spar. Consequently, equations (2.12), (2.13), (2.17), and (2.18) may be expressed as: [9c]= [e] 5.2) h [() = 2 h&X [B][e] (5.4) [cO] = 24 X[B] [e-J (5.5) where h is the spar thickness, and Ax- 6 inches. be used in conjunction with runs 1 must be made for run 4, where ao ~c,0 4+ A X[B] {o(' = These equations can 3; but the following modifications - 0.09650 radians: + {o(o( After carrying out the preceding calculations, where (2.15), there results, in radians: - -24 (2.X) [B]{S) [B] is given by [o(0 * [(j = = 0 0 0 0,09650 -0,02760 -0.00690 0.04620 0.09650 -0.01278 -0.02730 0.09408 0.09650 0.03822 -0.01680 0.13668 0.09650 0.07950 0.o1440 o.1644o 0.09650 0 0 0 o.09650 -0.03822 -0.02100 0.09000 0.18068 -0.02934 -0.05538 0.17520 0.24128 0.02238 -0.05298 0,23838 0.26978 0.06582 -0.02418 0.27216 0.27728 -0.00860 -0.00120 0,01420 0.01506 -0.00414 -0.00580 o.o580 0.01300 0.00710 -0.00566 0.01260 0.00720 0.01014 0.00646 0,00846 0.00230 0.00434 0.00314 0.00280 0.00020 -25- CHAPTER 6 CALCULATIONS AND RESULTS (6.1) Solving Matrix Euation for the Aerodynamic Influence Coefficients In order to solve equation (2.7) for [Q1] and [Q2] it will be partitioned to form two separate equations; half the test runs will form one equation, the other half the second equation. = (-( - 0( ) ;4 Letting: (.) eK] [N"] = [-c4 = OH]L 0'O and ; [K(2 ; [NZ] A] [H1[C(2) H [C 'IJH (6.4) where the superscripts refer to the two equations, there results after some rearrangement: Q and [A] GP'jr~f] Kj + Q][NG] + [QJ [N'zJ =QJ[K(2)] - 26 - A (6.5) Postmultiplying equation (6.5) by [N(2)] [A] - [N(l)] -1 nd equation (6.6) by .1 [()[N Q [K (' N = LQ][K2[Nz)] + [QJ (6.7) [Qz] (6.8) + Subtracting equation (6.8) from (6.7): - [K (]EN [K MN Letting [5] = [Kj [N' T = [(A' N - [ - (6./0) [s][ ~N) In like manner, after postmultiplying equation (6.5) by equation (6.6) by [K(2)] -1 there follows: [A] -( K"' = [i.]=[A]-[e [N[NK0) N[K i' LN - '[K - 27 L and [[1 TL K~ '(1 - - T [K()] ~1 and subtracting the resulting equations, QLetting (6.1i) (6.) K][K0.K) K T (6.I4) During the course of the work presented in this thesis, it was observed that if the initial angle of attack distributions of the wing model were not distinctly linearly independent, the columns of [o] would be almost simple multiples of one another; and similarly, also, the columns of [(' . [N] would be small. Consequently, the determinants of As a result, any errors in [o(] and be greatly magnified in the inverses of jK] and [N] 4 for a numerical example.) [a] . The ensuing values of [Q1 ] [K] and would (See reference and [Q2] would then be totally unreliable and the results of any divergence analysis useless. (6.2) Numerical Values of the Aerodynamic Influence Coefficients As mentioned previously, two wing stations. h were selected. [Q,] and [Q 2] can be determined at only To obtain an optimum spread in data, stations 2 and (See figure 9.) Wind tunnel runs 1 and 3 will be paired together and denoted by superscript 1; runs 2 and h by superscript 2. There then follows from the results of chapter 5: ) _0.02238 0.23838 -28 L -0.01584 0.10170 -o.14lo 0.08418 4-0,03618 0.17328 = - 0.09000 4380 L] [a - [ = 0.03822 -0.01062 -0.02100 0.18068 -0.05298 0.26978 L( j -0.00414 0.01580 L.01014 0 , 00 846 o C' .00580 0.01300 _0.0646 0.00230 0.0155 0.0768 O.0155 0.1475_ The matrices of dynamic pressures are, in pounds per square inch: [1. yi 004509 0O= 0.0 4 934 0 O.03O22 0 0 O.03022 The matrix of integration weights will be determined using the trapezoidal rule. fr( 0 Referring to figure 9, = f,+ ()+-6+f ( 2 ,+ + 10. 218?5 where f(x) is an arbitrary function and apex. +) + 4 + -Ft (81) dx + 4-. ?18 t 657/5 (6.15) denotes the wing tip or From equation (6.15), for wing stations 2 and 4: [H] oooo -9 1- 0 O 10.21875 Then after carrying out the calculations in equations (6.1 - 6.4), (6.1Q - 6.11), and (6.13 - 6.14), there results: - - 29 [Q, IQ2] 2.71376 -2.29502 1.16918 0.96033 -9.79559 2.95760 -3o76830 -1.13971 Notice that the signs of the elements of [Q 2] appear to be reversed. This arises from the fact that both angle of attack and do(/dx have been defined to be positive as shown in figure 3. But as seen by the approach- ing velocity, do(/dx in figure 3 is actually negative since the angle of attack changes from a positive value to zero. Hence, in a more consistent treatment, equations (2.16) and (2.22) should respectively be changed to: c] = -A [BJ[oC] ; f 0() = - [G ]fo(' where it is understood that a negative dc4/dx would produce a positive angle of attack, (6.3) Divergence Analysis of Wing Model Divergence speed will first be calculated for the original bending stiffness distribution of the model (i.e., the spar used in the wind tunnel tests). In actual practice this step would be carried out to determine the over-all accuracy of the analysis. Then, as a practical example, and also as an additional check, the divergence speed for a different bending stiffness distribution will be calculated. cases the method of matrix iteration will be used. - 30 L In both Rewriting equation (2.25): . o) [G] = C]H]( [Q,][H][G] + [QJ[HI) {oQJ (2.25) where qd is the dynamic pressure at divergence. [Gi is given by (2.21), where the subscripts a and b will now denote wing stations 2 and 4. Referring to figure 9, [G] [Z 2 ] S[Q] LH] and [C] 6 o 6 12 for the original spar were determined in the preceding section of this chapter, Substituting these matrices in equation (2.25) and performing the indicated calculations, there results-, Fo0( _ 9.6094 8.0392 (.' I L_ _5.302 6.3887 O After carrying out the iterative process (outlined in reference 3), equation (6.16) converges to, S[(K] 14.725 LAi 9,369 Ud = 1 14.725 14.725 1.0000 0.6363 0.067912 psi, and 90*7 ft* s*c* - 31 - Hence, qd 1 The actual divergence speed was observed to be 84 feet per second (dynamic pressure 0.0583 pounds per square inch). The method of testing the wing model for divergence is discussed in section B of the appendix. A different distribution of bending stiffness in the wing model was obtained by changing the shape of the original spar. The thickness was left the same, but the spar width at the wing trailing edge was narrowed down from 1.69 inches to 1.03 inches. The width of the spar at its tip was also unchanged, but in between sufficient material removed so as to obtain a constant taper from the trailing edge to the tip. EC' CX for the new spar was calculated from the equation: = IS X (- (EI)(x-).= e [_ where (EI)te,, f. is the value of bending stiffness at the wing trailing edge; ) is the spar length; and width at tip X A is the spar taper ratio (i.e., 4 width at trailing edge). in section C of the appendix. values of x and Equation (6.17) is derived After working out the calculations for at wing stations 2 and h, there is obtained for the new spar: LC1 = 0.0220 0.1126 0.0220 0.2394 Substituting this value of [CI] into equation (2.25) and performing the necessary calculations, there results: - - 32 . ' 14l1784 9.5130 (' 12.059][o] 1.4579 (,.8) d( After carrying out the iterative process, equation (6.18) converges to: 23.615 - 01 Hence, od - 1 - 23.615 078249 18*479 - 1.00000 0.042346 psi, and Ud a 71.6 ft./sec. The actual divergence speed was observed to be 69 feet per second (dynamic pressure 0.0393 pounds per square inch). this measurement would not have been made. In actual practice Here it was done to check the reliability of the method as a whole, not the specific calculations. - - 33 CHAPTER 7 DISCUSSION OF RESULTS Comparing the results obtained in the preceding chapter, divergence speed was predicted within eight percent of actual value and dynamic pressure at divergence within 16.5 percent for the original bending stiffness distribution of the wing model. It is of interest to note that the same quantities predicted by Jones's low aspect ratio theory were, respectively, ten and twenty-one percent in error. (These calcu- lations are contained in section C of the appendix.) For the second bending stiffness distribution, the divergence speed and dynamic pressure at divergence predicted in the preceding chapter were within four and eight percent, respectively, of their observed values. Although these results do represent an improvement in accuracy, their real significance lies in their substantiation of the foregoing method of divergence analysis. Since the principal objective of the experimental work undertaken in this thesis was to determine if strain measurements could be effectively used in performing a divergence analysis, the need of obtaining precise data was not paramount. evaluation of the method. More important was the over-all On the other hand, if the method of diver- gence analysis presented herein were to be applied to an actual wing, - - 34 and especially if extended to supersonic flow, certain refinements in technique should be incorporated in order to improve the accuracy of the data obtained. The primary source of error encountered in the present work was model oscillations, caused mainly by tunnel turbulence, which resulted in considerable fluctuation of the strain readings. It is estimated that the majority of these readings were obtained within eight percent of their true steady state values although some were probably as much as fifteen percent off. As mentioned in chapter 3, the use of a resistance- capacitance filter in conjunction with a separate null indicator did help considerably in zeroing out the strain indicator, but its usefulness was limited by the loss in sensitivity which resulted. A more direct method of improving the accuracy of the strain measurements would naturally be to minimize tunnel turbulence. This could be accomplished by installing additional screens in the return section of the wind tunnel to smooth out the flow, provided the accompanying loss in tunnel performance could be tolerated. In addition, however, it is recommended that in any future work, strain measurements be recorded on an oscillograph trace instead of being obtained directly by visual means, This would not only yield a better over-all picture of the strain fluctuations, but would also reduce tunnel operating time by eliminating the process of zeroing out the strain indicator for each reading. In section A of the appendix is a comparison of measured values of angle of attack with values calculated from strain measurements. The largest errors are seen to occur in the vicinity of the wing trailing edge where naturally the rate of change of angle of attack will be greatest. - - 35 This situation can easily be remedied, though, by the use of additional strain gages. The distance between strain gage locations would then be reduced and, hence, the numerical integrations improved. - - 36 CHAPTER 8 CONCLUSIONS A method of divergence analysis has been presented in which strain measurements are utilized to determine aerodynamic influence coefficients. Using this method, favorable results were obtained in a divergence analysis of a cantilevered delta wing. Although the work was carried out at low subsonic speeds, the successful application of the method points up the feasibility of extending it to supersonic speeds and possibly, also, to more complex wings. But regardless of the application, care must be taken to ensure that the strain measurements will possess sufficient linear independence. Otherwise, the resulting divergence calculations will be meaningless. -37 APPENDIX A ACCURACY OF ANGLE OF ATTACK OBTAINED FROM STRAIN MEASUREMENTS As mentioned earlier, the value of [C'3 used in the determination of the aerodynamic influence coefficients was measured from static deflection curves of the wing spar. But to obtain an estimate of the accuracy of values of angle of attack calculated from values of bending strain, the elements of wing station 5 [CI] pertaining to a unit lift were also determined from strain readings. was used in these calculations. force acting at Equation (5.5) A comparison of results follows; the percent errors are based on the values obtained from the static deflection curves. Values From Deflection Curves Values From Strains C15 0 Percent Error 0 C 0.1073 0.0938 C35 0.1902 0.1815 h.57 C# 0.2583 0.2553 1.16 C o.2801 O.2877 2.72 38 12.6 APPENDIX B TESTING WING FOR ACTUAL DIVERGENCE SPEED In order to prevent destruction of the wing model in divergence, its maximum deflection was limited by restraining strings. The wing was then considered to be in a diverged state when it remained deflected against one of these restraints, that is, the string remained taut. But because of tunnel turbulence, the precise speed at which the wing did diverge was difficult to determine. Moreover, it was observed that if the restraining strings were not made sufficiently long, the wing would not remain diverged at speeds at which it had remained so when allowed to deflect more. As divergence speed was neared, however, the natural frequency of oscillation of the wing in response to an applied disturbance approached zero. Consequently, divergence speed was determined to be that speed at which the frequency became vanishingly small even though the wing did not necessarily remain indefinitely deflected against the restraints. - - 39 DESIGN CALCULATIONS OF SPAR - APPENDIX C JONES'S LOW ASPECT RATIO THEORY In order to obtain a preliminary estimate of the divergence speed of the wing model for purposes of designing the spar, Jones's low aspect ratio theory was used to predict the lift forces on the model. A summary of the calculations follows. The angular deflection of the wing due to aerodynamic lift given by equation (2.2). is Letting o( 0 (x) equal zero, this equation may be rewritten as: (x) =fC(x,4) L(g)Jd ) is the lift per unit chord. Using Jones's low aspect ratio theory (see reference 2 or 3), L ( may be expressed as: .x -h4TrWre /! ithlo& o) 1h is the local semispan of the wing. - 40 - where ( L(g) = n(0-2) ) where, again, L ( (c-i) Substituting equation (C - 2) into equation (C - 1) and rewriting in matrix form, there results: {o(I a)&]I[&]{o fe (c-.3) 2TICJH( 2 LHJ where is a matrix of integration weights, and the primes denote chordwise differentiation. The positive x-direction will again be taken as in figure 1. {(3 may be expressed in terms of {oc) using equation (2.16). For wing stations at the trailing edge, mid-chord, and apex there results: {f where now 4) =~C [B -1(C 0 0 0 0.5 0.5 o 0.5 1 o.5 EB] (C Introducing equation (C - 4) into equation (C [C] as c 1 , - -5) 3) and expressing where (EI)te is the value of bending t.e. stiffness at the wing trailing edge, there results after rearrangement: K [B] -ll() - [c'1[H]( ' H. C [g[,][j + 62]] =- For equation (C - 6) to possess a nontrivial solution, its principal determinant must vanish. Hence, for divergence: i41l- + P, t9CZ[H( cJ9 tBI- From wing geometry, and letting c j ) (C--7), be twenty-seven inches, there results: 6.750 0 0] 3.375 0 0 0 45.6 0] -n 3 0 0 0 0.25 - 0 0 11.4 o 0 0 0 0 0.25 0 0 -0.25 0 Using Simpson's rule for numerical integration: 0 [CI] 04 0 4.5 for the wing spar will now be determined by analytical Assuming the spar obeys the mechanics of a simple beam, there results at point x for a unit load at I i-x El (x) =C x) - 42 - methods. 18 - EH] 0 : ) , x (c-8) Since the spar thickness is constant: EIr(Y) = I. where and - = = I- __ -t.e. (EI).e. I- __ w is the local width of the spar, the length of the spar. (C - 8) and letting A Y W.e./ T Wt.e. wt (c -s) the width at the spar tip, Substituting equation (C - 9) into equal wt/wte, there results: C" (x, ) = (EI).t. I -(,-X (c -io) T '..-. After carrying out the indicated division: (c-u) - C"(X4 (EI)~ t~e. -X (-X + "X After integrating both sides of equation (C - 11) with respect to x and applying the boundary conditions for a cantilever beam, there results: C'(x,4) = (EI)t -, - h (c-I?) To simplify the design calculations, the length of the spar was considered the same as that of the wing (i.e., twenty-seven inches). X equal to 0.058 and evaluating equation (C - 12) at the -W43 - Setting wing trailing edge, mid-chord, and apex, there results: 1 ECIu I (E) 0 0 0 0 109 356 0 109 635 Then after substituting the various matrices into equation (C - 7) 0 0 07 6.75 6.75 0 6.75 - and carrying out the calculations, the following is obtained: 6.751 13o5 (E), e 0 0 0 22,300 0 0 22,300 0 0 a 0 (C - 13) 6.75 + The condition for divergence thus becomes: 22,300,. 6.75 - 0 6.75 whereM for P4, 22,300)A 2 is equal to 7T qd (C - 14) 13.5 (in.-4 ). Solving equation (C - 14) there results: S-0.000303 in. (C - 15) where the minus sign originates from the direction of the velocity being taken in the negative x-direction. equal to 1,465 lb.-in. 2 , * Taking (EI)t (See figure 1.) equation (C - 15) gives, *It was originally planned to design the spar so that the wing model would diverge at approximately eighty miles per hour (117 feet er second). The required value of (EI)t.e. was thought to be 1,465 lb.-in. However, an error in the calculations was discovered after construction of the spar had already begun. disregarding the minus sign: qd a 0.0707 psi; and hence The actual value of Ud - 92.5 ft./sec. X, in equation (C - 12), turned out to be 0.0592 (instead of the 0.058 used in the calculations). The effect on the preceding values of qd and Ud should be slight, however. - .1 45 APPENDIX D CALIBRATION OF WIND TUNNEL Permanently installed in the Student Wind Tunnel is a manometer which measures the difference between the static pressure in the test section and that in the settling chamber. Since the air in the settling chamber does not actually come to rest, there is a slight difference between the pressure in the settling chamber and true stagnation pressure. To measure true dynamic pressure, a second manometer was connected to a Pitot-static tube installed in the test section. From the dynamic pressure readings, tunnel velocities were computed and plotted against the readings of the tunnel manometer (figure 12). Runs were made with the Pitot-static tube installed both immediately forward and immediately aft of the test section turntable. There was found to be negligible difference between the corresponding velocities. All runs were made with the model support removed from the tunnel. The specific gravity of the alcohol in the two manometers was measured to be 0.807 (at 700 F). temperature change were neglected. Variations in this value due to Compressibility effects were also neglected. - - 46 REFERENCES 1.) D.J. Martin and C.E. Watkins, "Transonic and Supersonic Divergence Characteristics of Low-Aspect-Ratio Wings and Controls," Institute of the Aeronautical Sciences report no. 59-58, 1959. 2.) R.T. Jones, "Properties of Low-Aspect-Ratio Pointed Wings at Speeds Below and Above the Speed of Sound," NACA TR No. 835, 1946. 3.) R.L. Bisplinghoff, H. Ashley, and R.L. Halfman, Aeroelasticit, Addison-Wesley Publishing Company, Inc., Cambridge Mass. 1 5. 4.) E. Mollb-Christensen, "Utilization of Experimental Results in . Flutter Analysis," Journal of the Aero/Space Sciences, Vol. 25, No. 10, October, 19 5.) F.B. Hildebrand, Methods of Applied Mathematics, Prentice-Hall, Inc., Englewood Cliffs, N.J., 1952, 6.) W.M. Murray, "The Wire-Resistance Strain Gage (SR-h Type)," notes prepared for Testing Materials Laboratory, M.I.T., course 2.37, Feb. 2, 1953. 7.) D.J. Peery, Aircraft Structures, McGraw-Hill Book Company, Inc., New York, N.Y., 1950. 8.) S. Timoshenko and G.H. MacCullough, Elements of Strength of Materials, Third Edition, D. Van Nostrand Company, Inc., New York, N.Y., 1949. 9.) S. Timoshenko and J.N. Goodier, Theory of Elasticity, Second Edition, McGraw-Hill Book Company, Inc., New York, N.Y., 1951. 10.) A.H. Shapiro, The Dynamics md Thermodynamics of Compressible Fluid Flow, Vol. II, The Ronald Press Company, New York, N.Y., 1934. U .P r z L(x) U z(x /,r Fig. 1 - Coordinate system. U Fig. 2 - Unit angle of attack at r, positive as shown. z U rr Fig. 3 - Unit change in angle of attack at - - 48 r, positive as shown. Fig. ii. - View of wing model and support looking downstream. 1' I I Fig. 5 - View of wing model and support looking upstream. Fig. 6 - Wing model with spar in place; top halves of wing sections removed. Fig. 7 - Wing sections; top halves and spar removed. iig.8.- Typical wing section, other half similar. 3t- 'I -3" 3 3" -- 3 3". 3. 2" 16 ac)d Q U3) Fig. 9 - Spar stations at which deflections were measured for the determination of [CI]. X's denote strain gage locations. - 52 - () 311 ) F L Fige 10 - Position of wing sections. z z x x b.) Run No. 2 a,,) Run No~ l z x c,) Run No, 3 d.) Run No. Fig. 11 - Initial deflection shapee of wing model. - 53 - 4 1' t .i 4'-.114:Ii1414--fh iii T 4 4 1 '1- -i_14 1 - L 4-4 iI 4 I~4 T-'r - -1 4- Tht r 4 I4 -t ff4" 14- 'I-44- i I 4 4171 ,-t _-4L 1 + . 44h 'I 7j -__ 77- 4 . 1-J 4-1 r ~~ TI1 4~2 J-41 ~ t4 1:~i-_ '4 - 1 ______________7- 4+ t Tr j t l~ LiFT ~ + -1. . 1,~tf 4 ~~4r~ I4t t I ;4I t I. ~ -- iT 7 1 J1 J J~ 44 - -1 - 4--i 4 .:_ -v ' T LPrT' y q Ift- -41