- SLENDER SHIPS WITH FORWARD SPEED A NEW APPROACH AND A NEW THEORY by Sea Heon Kim B.S., Seoul National Uhiversity, Seoul, Korea Department of Naval Architecture and Marine Engineering (1976) M.S., Massachusetts Institute of Technology, Cambridge, Mass. Department of Ocean Engineering (1980) SUBMITTED IN PARTIAL FULFILLMENT OF THE REQUIREMENTS FOR THE DEGREE OF DOCTOR OF PHILOSOPHY at the MASSACHUSETTS INSTITUTE OF TECHNOLOGY December 1982 Massachusetts Institute of Technology 1982 Signature Redacted Signature of Author Department of Ocean Engineering December 1982 Signature R(edacted Certified by Roniald W. Yeung, Supervisor Signature Redacted Accepted by Chairman, Departmental Coafi.ttei--nraduate MAS;t3L'upr, INSTiTUTE OF TECHNOLOGY JUN 0 2 1%J3 LIBRARIES Archives Students SLENDER SHIPS WITH FORWARD SPEEDA NEW APPROACH AND A NEW THEORY by Sea Heon Kim Submitted to the Department of Ocean Engineering in February, 1983, in partial fulfillment of the requirements for the Degree of Doctor of Philosophy ABSTRACT A linear theory is developed for the heave and pitch motions of a slender ship moving with constant speed in calm water. The ship beam-to-length and draft-to-length ratios are assumed to be small. The method of matched asymptotic expansions is applied to obtain a new slender-ship formulation. Novelty lies in the fact that a complete linearized free-surface condition is used in the formulation of the inner problem, which contains forward-speed effects not heretofore accounted for in other theories. Genuine three-dimensional effects are also incorporated by introducing a generalized inner Green function that matches with the inner limit of the outer Green function. It is shown in the course of development of this theory that the inner problem, interpreted in a pseudo-time sense, is completely compatible with the three-dimensional outer problem. The new theory is expected to be more valid for general frequencies of oscillation of the heave and pitch motions. Several special cases of simplification of this theory are In particular, the present formulation yields a considered. rather promising approach to the steady forward motion problem of a slender ship, which is simply a limiting case of the more comprehensive theory presented here. Thesis Supervisor: Title: Ronald W. Yeung Professor of Naval Architecture, Visiting (Univ. of Calif., Berkeley) -3- ACKNOWLEDGEMENT S I would like to express my sincere gratitude to my advisor Professor Ronald W. Yeung not only for his advice and guidance during the course of this work but also for his continuous support and encouragement during my stay at M.I.T.. Without his constant support my graduate studies would not have been completed. I am also indebted to Professors P. D. SclavounOs, T. F. Ogilvie, J. N. Newman, and M. S. Triantafyllou for having served on my thesis committee. Special. thanks go to Professor Sclavounos for his valuable discussions. Financial support during this study was provided primarily by the Office of Naval Research under Task # NR62-611, Contract # N00014-78-C-0390 and in part by the National Science Foundation under Grant ENG 77-17817. Both supports are gratefully acknowledged here. Thanks are also due to Miss Beth Germanotta for her proficient and patient typing of this manuscript. Finally, I would like to thank my wife Kyong-Ho for her endless encouragement and assistance. I would also like to thank my parents and parents-in-law for their continuous interest and support throughout my education. -3a- TABLE OF CONTENTS Page ABSTRACT ....................... 2 ACKNOWLEDGEMENTS ............. 3 ,................ TABLE OF CONTENTS ...... I. INTRODUCTION ., II. THE BOUNDARY-VALUE PROBLEM ...................... ................................ ... .... ... ... ......... .. .. .... 111.2 111.3 IV. IV.1 IV.2 V. VI. 30 Forward-Speed Transformation and Resulting Representations .. *............. Inner Behavior of the Modified Green Function .. ........................ Simplification of the Surface Integrals THE INNER PROBLEM .. 4 20 III. THE OUTER PROBLEM ............ III.1 3a . The Pseudo-Time Inner Problem ... .......... Integral-E uation Formulation in terms of Gi) ........................... MATCHING OF INNER AND OUTER SOLUTIONS ........... .................... The Matched Solution V.2 Behavior of Interaction Functions F and F ................................. 35 38 44 *.*....... V.1 33 47 49 57 57 61 V.2.1 Zero-Speed Forced Oscillation ..... 64 V.2.2 Steady Forward Motion ... 70 *...... V.3 The High-Frequency Approximation ......... ... 75 V.4 Alternative Representation of the Solution in terms of a Line Distribution of Wave Sources ..... ...... ...... . . . ADDED MASS AND DAMPING ......................... VII. DISCUSSIONS AND CONCLUSIONS ............. 78 83 90 -3b- Page .- REFERENCES ... .. .. e. .. . .. L ..... t . * .. .... S a 99 a . APPENDIX 1 : Derivation of Modified Greenks Theorem 103 APPENDIX 2 :. Derivation of the Pseudo-time-dependent ......... Green Function .......... 105 APPENDIX 3: 111 APPENDIX 4: Alternative Expressions and Asymptotic Approximations of the Pseudo-time.dependent Green Function The Inner Expansion of h (x,y,z) ..... APPENDIX 5: Reduction of the Kernel APPENDIX 6: Inner Expansion of the Outer Solution O(x,y..Z) APPENDIX 7: .. . . . . . 116 ................ . . . . . - . .. Relationship between a Line Distribution and a Surface Distribution of the ....... Generalized Inner Green Functions . -------a 120 128 130 -4- I. INTRODUCTION One of the primary: goals of the naval architect is to be able -to design ships which operate satisfactorily in a wave environment that is frequently uncomfortable and quite irregular in nature. The accurate prediction of the ship motions and the dynamic sea loads is of unquestionable importance for a safe and economical operation. However this is such a complex problem that reliable methods of making these predictions have only been available for the last thirty years. One of the major difficulties in develop- ing a suitable ship motion theory was the lack of any technique which could describe the complex nature of the random ocean waves. Since the well-known paper of St. Denis and Pierson (1953) on the application of spectral analysis to ship motions, there has been much development in both experimental and theoretical efforts for the ship-motion problem in regular waves, which, by linear superposition, is ultimately applicable to an irregular seaway. With the assumption of small unsteady motions of the ship and of the surrounding fluid, the ship motion problem can be decomposed linearly into two problems; the radiation problem where the ship undergoes prescribed oscillatory motions in otherwise calm water, and the diffraction problem where the ship is kept fixed at its mean advancing position but is otherwise free -5- to interact with the incident wave train. The present study is concerned only with the former problem., and is restricted to the solution of the forced heave and pitch motions of the ship which is also advancing at a constant forward speed. The first theoretical analysis of the ship motion problem is connected with the works of Froude Krylov (1896). (1861) and Froude and Krylov derived differential equations for the motion of a ship in waves. The inertial and hydrostatic restoring forces for the radiation problem were evaluated without attempting to analyze the hydrodynamic disturbances associated with the ship's oscillatory motion. Furthermore, the exciting force for the diffraction problem was evaluated using only the pressure field of the undisturbed incident wave under the hypothesis that the presence of the ship did not alter its pressure field. The resulting exciting force has become known as the Froude-Krylov exciting force. Michell (1898) made the first significant step to account for the hydrodynamic disturbance due to a realistic ship hull with his steady-state wave-resistance theory. He developed a so-called thin-ship theory of wave-resistance under the assumption that the ship has a small beam compared to its length and wavelength. Although Michell recognized the possibility of extending his theory to include unsteady -6- motins, a promised paper was never published. workers, however, Subsequent did take up his ideas for bQth. wave- resistance and ship motion studies. A comprehensive analysis of the linearized ship motion problem was made by Haskind (1946). Green's theorem was used to construct the velocity potential due to the presence of a ship hull and the necessary Green's function or source potential was derived. The thin-ship approximation was then invoked to solve the resulting integral equation. He also introduced the now widely adopted procedure of decomposing the velocity potential into a canonical form so that thp radiation problem and the diffraction problem could be solved independently. Throughout the development of ship motion theory, certain geometrical characteristics of a typical ship has been taken advantage of. Thin-ship theory is the result of assuming the beam is small compared to the length, with the draft being taken as arbitrary. Slender-body theory, how- ever, assumes that the beam and draft are the same small order of magnitude by comparison to the length. Since Michell's pioneering effort, the thin-ship theory has been extended and refined by many investigators, Haskind (1946), Peters and Stoker (1957), Havelock (1958), and Newman (1961). In spite of the fact that real ships are hardly thin, study of the thin ship prospered, and the theory provided some -7- .guidance in the reduction of wave resistance, see Wehausen (1973). But corresponding results for ship motion problem have not been so useful. A breakthrough was made by Korvin-Kroukovsky and Jacobs (1957). Using the concepts of the slender-body theory of aerodynamics supplemented by substantial physical insight, they derived a "three-dimensional" strip theory for the approximation of ship motions in head waves. They assumed that the hydrodynamics associated with the ship could be represented by a series of two-dimensional transverse ship elements or 'strips'. They solved two-dimensional boundary- value problem for each cross-section of the ship by neglecting hydrodynamic interactions between adjacent ship sections. The two-dimensional solutions were then adjusted to include certain three-dimensional forward-speed effects based on intuitive, physical arguments. Even though their theory was derived mainly from 'physical intuition' rather than rational mathematics, there is no doubt that this original strip theory deserves its recognition as one of the most significant contributions in the field of seakeeping. The Korvin- Kroukovsky and Jacobs theory has since been modified and extended. For example, Smith (1967) has shown that a modified strip theory by Gerritsma and Buekelman (1967) predicts the head-seas motions for a high-speed destroyer hull which agree quite well with experiments. Here the word 'modified' -8- is used to represent a modification leading to a speeddependence of some of the hydrodynamic coefficients which was not found with. a 'pure' strip theory calculation. Even though the agreement between experiments and the KorvinKroukovsky and Jacobs strip theory has usually been satisfactory, a major objection to this theory has been that the forward-speed dependent terms in the cross-coupling coefficients do not satisfy the symmetry relationships proved by Timman and Newman (1962)'. Subsequent work was concentrated on providing a rigorous derivation of the ship motion theory based on the slenderbody approach of aerodynamics. This has resulted in a so- called long-wavelength ordinary slender-body theory of ship motions. This was done initially for the steady-state wave- resistance problems by Cummins (1956). Later, Tuck (1963) reformulated this theory by using rigorous singular perturbation and matched asymptotic expansions techniques, with the assumptions that the beam and draft are small compared to the wavelength scale as well as the ship length. Using a similar slender-body assumption, Newman and Tuck (1964) derived a long-wavelength slender-body theory .for unsteady motions of the ship. In this case, however, most nontrivial hydrodynamic effects as well as the inertial force due to body mass are higher order by comparison to the hydrostatic restoring force and the Froude-Krylov exciting force, which depend primarily -9- on waterplane area. This implies that the resulting leading-order equationa of motion are nonresonant. Thus, their unsteady solutions only gave adequate results for low forward-speed and low frequencies of oscillation. In order to improve the predictions of the slender-body formulation the inertia terms must be retrieved into the lowest order theory in such a way that added mass and damping also appear. This can be done under the assumption that the wavelength is the same small order of magnitude as the transverse dimensions of the ship. A systematic analysis of the short-wavelength slender-body theory for the radiation problem was carried out by Ogilvie and Tuck (1969), where a strip theory approximation was shown to be an adequate representation of the flow field adjacent to the ship hull. By consistently retaining the higher order terms of relative magnitude e1/2 in their perturbation analysis, e being the slenderness parameter defined as the ratio of transverse dimension to ship length, Ogilvie and Tuck provided a rational approximation for the effects of the ship's forward speed. Their analysis gave some integral terms over the free surface as being an additional contribution to the cross-coupling coefficients. These additional contributions are essentially due to the forward-speed effects on the free surface condition. Faltinsen (1974) noted that better agreement with experiments was achieved by using the Ogilvie and Tuck's cross-coupling coefficients. A cQmprehensive derivation of a strip theory, valid for short waves, was carried out by Salvesen, Tuck and Faltinsen (1970) in the context of the prediction of ship motions and sea loads. Even though this formulation was derived less rigorously compared with Ogilvie and Tuck's (1969), the results were in good agreement with experiments. theory of Salvesen, et al. The strip (S.T.F.) agrees in the zero-speed case with Ogilvie and Tuck's short-wavelength slender-body theory, which is known as the 'rational strip theory'. However, the SFT'theory included in the forward-speed case the forward-speed squared terms that were discarded as higher-order effects in the particular perturbation scheme of Ogilvie and Tuck. In spite of the work of Ogilvie and Tuck (1969), all strip theory formulations have deficiencies in one way or another. The principal questions are the validity of the solution at lower frequencies, and the validity of treating forward-speed effects as simple corrections to the zero-speed solution or only as higherorder corrections. Newman (1978) derived a unified slender-body theory, which embraces the ordinary slender-body theory and the strip theory, under the single assumption of small beam and draft to length ratio. The ordinary slender-body theory and the strip theory were shown to be the long- and the -11- short-wavelength liMit of this theory respectively, Computa- tions for the radiation prob.lem wi.th. ?k unified 4lendernbody theory were presented by Mays Sclavounos' 1980). CI9781 Newman and The theory has been extended to the diffraction problem by Sclavounos 19811, Throughout the analysis of the above slender-kbody theories, the technique of matched asymptotic expansions, described by Ogilvie (1977), was used to split the problem into an inner problem and an outer problem, The inner problem, which describes the flow field in an inner region close to the ship hull, involves two-dimensional flow about a long slender body with some unknown three-dimensional interactions. This inner problem is governed by the two- dimensional Laplace equation subject to the linearized time-harmonic free-surface boundary condition and the hull boundary condition. The outer problem which applies far from the hull surface is fully three-dimensional, and is governed by the three-dimensional Laplace equation subject to the complete linearized free-surface boundary condition and the proper radiation condition at infinity. The unknown three-dimensional interactions in the 'inner problem are determined by requiring the two solutions to be compatible* in a suitably defined overlap region. It is worthwhile to point out that the free-surface boundary condition in the inner problem has be'en customarily -12- taken as time-harmonic so that it was free of forward-speed dependence,' while thle hull boundary conditiQn did include forward-speed effects. However, one will not expect the time-harmonic free-'surface houndary condition in two dimensions to remain physically valid if the characteristic time associated with the ship speed is comparable to the period of oscillation. In the rational strip theory of Ogilvie and Tuck (1969), the forward-speed effects were taken into account in the free-surface boundary condition as higher order effects, which provided the cross-coupling coefficients with the additional contributions represented by some integral terms over the free surface. The speed- squared terms were, however, discarded as being higher order, while most of the intuitive versions of the theory retained these terms. The conventional strip theory is deficient not only for low frequencies but also for high speeds. A complementary approach, which includes the forward-speed effects on the free surface condition, was initiated by Chapman (1976, 1977). The flow at each section along the ship was analyzed in a quasi-three-dimensional manner such that a part of longitudinal interactions could be incorporated through the free surface condition. Here the interactions were assumed to propagate downstream only. Chapmanls computations for the sway and yaw response of a vertical surface-piercing flat -13- plate were suppQrted by an impressive agreement with experiment, But the 'roude number was extraordinarily Using a similar idea, Yeung and Kim (1981) derived a high.. formulation for the radiation problem of a slender ship. By letting the ship pierce through a sequence of control planes defined in a fixed frame of reference and by solving the resulting initial boundary-value problem on a control plane, the forward-speed effects on the free-surface boundary condition could be easily included in a quasi-three-dimensional manner. The computational results from the resulting intlegral equation of Volterra type showed some improvement over the strip theory predictions at low and moderate frequencies. Since the solution satisfied a two-dimensional Laplace equation and no outer matching was used in the above approach, there could be no representation of transverse waves. A direct numerical solution of the linearized threedimensional ship motion problem was carried out by Chang (1977) and Inglis (1980). It was found from the computa- tional results that the simple speed-correction normally used in strip theory does not adequately characterize the actual speed effects observed in the measurements, and that the speed terms in the free-surface boundary condition have to be included in order to obtain reasonable quantitative and qualitative 'predictions. Even though this three- -14- dimensional approAch is derived in. the Most 5reneral way within linear theory, one 'MAy be faced with the need to assume that the ship is either thin, slender or completely submerged to justify the linearization of the free-surface boundary condition, for the disturbances due to the steady forward motion of the ship are not necessarily small compared to the unsteady ones unless the geometrical restrictions are imposed. -15In the present study a new slender-body formulation of the ship motion is derived by using the technique of matched asymptotic expansions under the single assumption of small beam and draft to length ratio. It is a well-known fact that the hydrodynamic disturbances due to ship motions are represented by a velocity potential which satisfies the three-dimensional Laplace equation, the complete linearized free-surface condition, the hull boundary condition, and the radiation condition of outgoing waves at infinity. It is also known that the velocity potential can be constructed by using Green's theorem and appropriate Green functions. Conventional ship hulls are slender in the sense that the geometry variation in the longitudinal direction is small compared to the corresponding variation in the transverse directions. This geometric characteristic of the slender ship has been used for the simplification of the fully three-dimensional boundary-value problem. In what follows, the flow pattern and rationals behind the present theory are briefly described from the physical standpoint. The theoretical framework of the present approach is built from mathematical analysis presented in the subsequent chapters. Far away from the ship, at distances comparable to ship length, the flow is fully three-dimensional and depends mainly on the elongated nature of the ship hull, being relatively insensitive to its local geometric details. The velocity potential in this outer region, therefore, satisfies -16the three-dimensional Laplace equation, the complete linearized free-surface condition, and the radiation condition at infinity. The outer solution is either obtained by using Green's theorem, or approximated by a distribution of freesurface wave sources and dipoles along the ship's centerSince the hull boundary condition is not imposed on line. the outer problem, the precise quantitative information is not available until the flow close to the ship hull is analyzed. Near the ship, at transverse distances of the order of the ship's lateral dimension, the flow gradients in the longitudinal direction are small by comparison to those in the transverse directions due to the slenderness of the ship, which justifies the reduction of the three-dimensional Laplace equation to the two-dimensional one in the inner region. It is common also to simplify the free-surface boundary condition by relating the wave parameters and the forward speed to the slenderness parameter - e . In the zero forward-speed case of forced oscillations of a slender ship, there is only one wave system, whose wavelength depends on the oscillation frequency w , far away from the ship. In the unsteady forward-speed case, however, Faltinsen (1981) showed that there may be up to 5 wave systems with different wavelengths behind a translatory harmonic oscillating source. Furthermore, the wavelengths depend on the forward speed and the wave propagation direction as well as on the oscillation -17frequency, which makes the unsteady forward-speed problem more complicated than the zero-speed case. The forward- speed terms in the free surface condition have been usually either neglected completely or included as higher order effects under the assumption that U = 0(1) and w = 0(C-1/2) However, it is difficult to know how to relate the different wavelengths to the slenderness parameter in a way that leads to meaningful results. The three-dimensional approaches of Chang (1977) and Inglis (1980) also suggested the importance of the forward-speed terms in the free-surface condition for the reasonable quantitative and qualitative predictions. From the above reasoning the present formulation employs the complete linearized free-surface condition which includes the forward-speed terms under the simple assumption that the forward speed is appropriately limited so as to justify the linearization. Even though these forward-speed terms may be considered higher order effects, including terms of higher order should not significantly affect the results in a strict sense. In reality, the present approach provides physically meaningful results. Thus, in view of the above arguments, we adopt the following formulation in the inner region. The velocity potential satisfies the two-dimensional Laplace equation, the complete linearized free-surface condition, and the hull boundary condition. Since a radiation condition cannot be imposed, the inner solution is not complete, unless the qualitative information is -18- provided from the outer solution, appropriately approximated in an overlap region. The mathematical analysis of the present study is largely devoted to the derivation of this approximation, which is known as the inner expansion of the outer solution, and to its subsequent effects on the inner problem. As it turned out, the general solution of the inner problem requires the knowledge of a Green function which cannot be determined uniquely from the inner problem alone. It is determined uniquely and completely by examining the outer expansion of the inner solution and requiring it to be compatible with the inner expansion of the outer solution. Owing to the forward-speed dependence of the free surface condition in the inner problem a line distribution of Green functions along the ship's centerline cannot produce the same wavemaking effectiveness even in the far field as that of a surface distribution. In other words, the line distribution should be extended over an infinite length in order to produce the same far-field wave effects as that of the surface distribution. The outer solution, therefore, is also represented in terms of the surface distribution of Green functions in the present approach. In Chapter II, the exact linearized boundary-value problem for the unknown radiation potential is formulated, and the fundamental assumptions of the present approach are explicitly stated. The method of matched asymptotic -19- expansions is used for the solution of the problem. Thus, the outer solution is firstly derived in Chapter III together with its inner expansion. The inner problem is formulated in Chapter IV and then the matching is carried out in Chapter V. In Chapter VI, the pressure on the hull and the resulting added mass and damping coefficients are determined. Finally, the present formulation is compared with the existing theories, and conclusions and future directions of work are described in Chapter VII. -20- II. THE BOUNDARY-VALUE PROBLEM We consider a ship which moves in the positive x- direction with constant forward speed U, while performing small harmonic oscillations of frequency pitch. It w in heave and is - convenient to introduce three Cartesian coordinate systems, with x =(x , y , z ) fixed in space, x =(x,', y', z') fixed with respect to the ship, and x=(x, y, z) moving in steady translation with the mean forward-speed of the ship. These coordinate systems are illustrated in Figure 1. The z0=0 plane is taken as the undisturbed free surface, the positive x,-axis in the direction of the ship's forward speed, and the positive zo-axis upward. The steady-moving coordinate system X, which is defined by (2.1), is an inertial frame of reference in which the motions of the ship are periodic. coordinate system x', The ship-fixed which is defined such that x'=x in steady-state equilibrium, is the best to derive the hull boundary condition on the ship's wetted surface. x= (xo-Ut, y 0, z ) (2.1) -21- , z Z X .x. Figure 1. Coordinate System The fluid is assumed to be ideal, incompressible with constant density p, and its motion to be irrotational. Surface tension effects are neglected. With these assumptions, the fluid velocity vector V (X ,t) is represented as the gradient of the velocity potential ND(xo ,t) which satisfies the three-dimensional Laplace equation, V20 ( ,t) = 0 (2.2) The fluid pressure throughout the fluid domain. p(x.,, t) is determined by Bernoulli's equation, p (xO, t) = Here p - p (t + V + gz0 ) + a is the fluid density, g is the gravitational (2.3) -22- acceleration, and pa is the atmospheric pressure which is In (2.3) and hereafter, when the assumed constant. independent variables (x, t) appear as subscripts, partial differentiation is indicated, i.e., tt = at/at, etc. On the submerged portion of the ship's wetted surface S, the normal velocity of the ship is equal to that of the adjacent fluid such that on S, V-n = Vson where Vs (2.4) is the local velocity of the ship's wetted surface, and n is the unit normal vector pointing out of the fluid domain. On the free surface, whose elevation is given by z (x, the kinematic boundary condition is y , t), expressed by means of the substantial derivative D/Dt E/at + V-V , (D/Dt)( in the form - zo) = 0 on z = (2.5) An additional dynamic boundary condition is obtained by requiring that the pressure on the free surface is atmospheric. t + Since Bernoulli's equation (2.3) then gives V2 + gz 0 on z0 = (2.6) holds on the free surface for all time, its (2.6) -23substantial derivative can be set equal to zero. This leads to an exact nonlinear free-surface boundary condition, 1 vo-V(Vo-vo) + go 0 = 0 on z = , 0tt + 270-VOt + (2.7) with Z 12 (1/g) [P +1 V2 (2.8) 0 The fluid velocity V is assumed to vanish at z + - and a radiation condition has to be imposed such that the energy flux associated with the disturbance of the ship is directed away from the ship at infinity. The set of equations (2.1)-(2.8) formulates the exact boundary-value problem within the limitation of an ideal incompressible fluid. However, the nonlinear free-surface condition precludes any solution of the unsteady motion problem without further simplifications. In the theory of ship motions it is customary to assume the amplitude of the incident wave system to be small by comparison to its wavelength, which justifies the smallness of the oscillatory motions of the ship and the surrounding fluid. However, the steady disturbances associated with the ship's forward motion are not necessarily small compared to the unsteady ones unless geometrical restrictions are imposed. Under the previous assumptions the total velocity potential can be decomposed linearly into a steady and an -24unsteady part. Thus, following Newman(1978), we write (t) (Nt) + )=U(x) UT(N) is the velocity potential due to the steady where i(N,t) is the unsteady one forward motion of the ship, and resulting from ship's oscillatory motions. i (2.9) ( x, t) and Both potentials n'1satisfy.Laplace's equation (2.2) subject to the boundary conditions that are appropriate to them. The velocity vector of the steady flow relative to the moving reference frame is = UV(T - x) (2.10) The boundary condition on the hull surface in its steadystate position S takes the form n = 0 q (2.11) on In the moving reference frame the nonlinear free-surface condition for becomes - (W + gUoz = 0 on z = I i where the steady free-surface elevation (2.12) , is given by (g/2) (W2 - U2 (2.13) Neglecting second order terms in T and using a Taylor-series expansion, Newman (1978) showed that the unsteady velocity potential T satisfies the following boundary condition on the steady-state free surface z=i. -25- -V)[I -V + Ttt + 2A.V'ft + vT-vW 2 + g z = 0 + ]-,z + l/[g (Vw2) + gt - + on z = (2.14) This boundary condition is a combination of steady and unsteady effects, and it is impossible at present to derive a velocity potential which satisfies (2.14). Using a similar analysis, Newman obtained the linearized boundary condition on the ship's wetted surface, n= ( + (t-v) (a-.)% AL n on S, S (2.15) where S denotes the instantaneous position of the hull surface, S the steady-state mean position, and the overdot signifies time differentiation in the reference frame of the ship. Here a = x - x is the local oscillatory 4. displacement of the ship's surface. a Since is a small oscillatory quantity and each member of (2.15) is 0(a), this boundary condition can be applied either on S or S with the difference 0 (a2 . From a vector identity the equation (2.15) is expressed in a more compact form 4. "I 'n = E onF + Vx(a x W)]-n (2.16) If the perturbation of the flow due to steady forward motion of the ship is neglected, reduces to = - Ut, and (2.14) -26- tt - + U2 x 2Ux + g on z = 0 = 0 (2.17) This assumption is justified for a slender ship at moderate speeds. Since the unsteady motions are assumed small, the vector can be expressed as a displacement = where I 1 and denote the unsteady translation and rotation of the ship relative to the origin P X' = 0. The potential in (2.9) can thus be decomposed linearly into separate components due to each of the six rigid body motions of the ship. With the restriction that the unsteady motions are sinusoidal in time with the frequency w, the motions of the ship are denoted by + & = Re{(%1, = Re{( 1 , 2' 3 e iwt 22' a 3 S Re{(&4 ,41' 6)it (2.19) 1 t (2.20) Here the Re stands for the real part of the expression. accordance with this notation, the unsteady component of the velocity potential can be expressed as S(X, t) = Re{ Tx)eit} ( jJJ (2.21) In -27Under the assumptions stated above the linear boundaryvalue problem satisfied by T(N) and P. (') in the moving J reference frame is summarized as V2= 0 (i V? 22 + V) 2 + 0 on z = 0, w = 0 on +0 at z = and f (2.22) + 0 (2.23) (2.24) = f * where U~ - in the fluid domain -c (2.25) denote the corresponding velocity potential and the normal velocity, respectively, which are defined in (2.9), (2. 21), (2.11) , and (2.16) . Here we introduce an artificial Rayleigh viscosity parameter P for the purpose of specifying an appropriate radiation condition at infinity. From (2.11), (2.16), and (2.18) the hull boundary conditions for (2.24) S= mio take the forms, n +Um.) V.Fjn == iwn. 1w Here the components n (n4 , n 5 , n 6 ) (x (2.26) on~ (2.27) are defined as n. (nl, n 2 , n 3 ) on (2.28) x n) (2.29) and, following Ogilvie and Tuck (1969), U(m1 , m 2 , m 3 ) UM = -(n-V)W U(m4 , m 5 , M 6 ) = -(,-V) (, x ') (2.30) (2.31) -28In a strict sense, a theoretical model which satisfies (2.23) and (2.24), where f is given by (2.26)-(2.31), is inconsistent since the influence of the perturbation of the steady flow has been excluded from the free-surface boundary condition but retained in the hull boundary condition. For a slender ship at moderate speeds, however, the disturbances due to the steady forward motion of the ship can be assumed small compared to the unsteady ones. Thus this model is certainly valid for a slender ship at moderate speeds since retaining higher order quantities that are mathematically of a higher order in the formulation could possibly improve the results. For an arbitrary geometry the solution of the above boundary-value problem can be obtained using Green's theorem, which results in an integral equation for the unknown potential on the wetted surface. Such a Green function should satisfy the three-dimensional Laplace ecuation, the linearized free-surface condition, and the radiation condition at infinity. The solution by this approach involves, in general, lengthy computations of the kernel associated with this fully three-dimensional Green function as well as the inversion of a big matrix. Conventional ship hulls are slender and this characteristic will be used in the present approach to reduce the complex three-dimensional problem into the simpler one through the method of matched asymptotic expansions. Before -29proceeding any mathematical manipulations we state our The beam and draft are assumed small basic assumptions. compared to the length such that B/L, T/L = 0(), (2.32) E. .<< where B, T, and L are the beam, the draft, and the length of the ship respectively. The ship's forward speed U is restricted only to the extent that interactions between the steady and unsteady disturbances remain relatively unimporThus we set tant. U (2.33) 5 0 (1) The frequency of oscillation w varies independently of the ship dimensions. W We will introduce the parametrization of for the ease of analysis in the application of the method of matched asymptotic expansions. O= O(j) ,0 Thus we set S y. S 1/ 2 (2.34) Since only the radiation problem for heave and pitch is analyzed in the present study, the subscript used hereafter with the understanding that j j will be = 3 or 5. -30- III. THE OUTER PROBLEM r =(x2 +Y2 +z 2 1/2 The outer solution, which applies for 0 (1) , satisfies the following set of equations, in the fluid domain V24= 0 (iw V 0 U-ax +1)2) + g 3 - + -0 0 on z =0, at z = + 0+ (3.1) (3.2) (3.3) -- Here the radiation condition of outgoing waves at infinity is incorporated by introducing an artificial Rayleigh viscosity parameter p . An appropriate outer solution can be either exactly obtained by using Green's theorem, or approximately by a distribution of wave sources and dipoles along the ship's centerline. The former representation is used in the present study for reasons that will be readily apparent. Applying Green' s second identity to 4)(Q) and G (0) (PQ) in Q-space, which is bounded by the ship's wetted surface a large surface surface 1, S, , and the portion of the undisturbed free SF between S and S. as shown in Figure 2, we can express the outer solution as 4[f (P) ds(Q) a) - S+S +S F G(0) (PQ) )n an (3.4) -31- Sao Figure 2. where nQ Q-space. (o) Domain of application of Green's theorem is the unit normal vector out of the fluid in The Green function G (o) (P,Q), with superscript denoting "outer", is defined as G(O)(P,Q) = - l/r + 1/r1 + h(P,Q) 1 lim h (P,Q)= .- ~ 1+o+ dkj x d exp[k2+2 2 1/2 2 2 1/2_ exp[-ik(x-E)-it(y-n)J (3.5) 2 (3.6) where r =(x-) 2 + (yn) 2 + (Z-0 2 1/2 (3.7) -32- r1= with (x-02+ P = (x,y,z) and 2+ Q = (+ 2 1/2 (,n, I) . (3.8) , The function h is defined such that it is harmonic in the fluid domain and satisfies conditions (3.1)-(3.3) with respect to P-variable. The details of the derivation can be found in Wehausen and Laitone (1960). The integral over large surface vanish if G(O and can be shown to Since both tend to infinity. S, we let S. satisfy the free surface condition, the 4 bracketed quantity in (3.4) evaluated on =0 can be written as ) [- 1 1- {2iwU[$hJ+U2[4 a 3h} C=O ;=O (3.9) Making use of (3.9) and Stokes's theorem, we obtain 47r4 (P) dS(Q)[4,(Q) dn where C G - Q ()(P(,Q) Q [2iwU4(Q)+U2 (0 )h(P - ,Q) (3.10) represents the intersection of the hull surface and the undisturbed free surface. Since the hull boundary condition is not imposed on the outer problem, the precise quantitative informations for 4 and an will be determined -33from the matching with the inner solution. III.1 Forward-Speed Transformation and Resulting Representations It is convenient in the matching procedure to introduce the following "forward-speed transformation" = *(x,y,z) -0 = 0 (U - )2 VO (3.11) ei(ox/U 0 (x,y,z) then becomes * The boundary-value problem for 2+ defined by in the fluid domain + g a + on z=0, P at z= -w 0+ (3.13) (3.14) represent generalized gradient v+ where the special symbols + (3.12) (vector) operators which are defined as S i(+( a= 0 ) + + k (3.15) in rectangular coordinates. The modified Laplacian operators 2 V+ are, therefore, of the forms, 2 2+ , 2 - iw + a2 Dy a2 z Dz (3.16) Following a similar analysis to what we did a little earlier, the solution 0 can be obtained from a "modified Green theorem" proved in Appendix 1. More specifically, -34- if and 0 are any differentiable scalar functions, *0 then J JJ S = - - 0 TdS 3n03n S (3.17) where the normal-derivative operators in (3.17) are defined as (n-V +) = n( + n 2- - + n3az (3.18) If the function *0 Green function G 0 () in (3.17) is chosen as a 'modified' then the following expression for , results 47r () n 0 -- 0(G ) (P,-) @nQ+ (Q) dS(Q) [ fJ = DnQ- T - U2 g ra Cdn[f 0 C (Q) ~ - Here the particular Green function V+GO a 2aG 0h (3.19) (P,Q) p. G (0) has the properties, in the fluid domain (PQ) = 6 (P-Q) - (Uax _ 3 (0) 2 G 0-(o)+ g G0z and is in fact related to (3.5) 0 on z=0, P such that + (a) G( 0+ is given (3.20a) (3.2 Ob) -35by the following expressions, (o) Go ((PQ) = (-/r + 1/r 1 )e h (PQ) = e U 1 tim eO((k'-*/U)+2 2- (Uk-i )2 (x-0) U + h 0 (PQ) (3.21) h(PQ) dk' -*++ dt exp[(k'-w/U) 2 [ exp [-ik' 2 1/2 (x- )-iZ(y-n) (3.22) It is of interest to notice that the line integral in (3.10) has been reduced to a simpler form in (3.19), and that the expression in (3.19) will be seen to be easily matchable to the inner solution later on. Inner Behavior of the Modified Green Function 111.2 If we are later going to match the outer solution in (3.19) with an inner solution, which will be derived in Chapter IV, it will be necessary to obtain an inner expansion We shall first examine the expansion of this outer solution. of ho for small values of R( y2 + z2) . In the next section, we shall derive the corresponding inner approximation . of the integrals in (3.19) The expansion of h will be derived in the Fourier space because of the relative simplicity of the algebraic -36manipulations that are needed in the transform space as opposed to the more cumbersome manipulations in the physical We define a Fourier-transform pair as follows space. Odx e F{P (x)} p*(k) F P(x) 1 (3.23) P(x) (3.24) Cdk e-ikxp*(k) {P (k)} The Fourier transform of (3.22) then gives h 0 (y*, z;k) Jt k1 exp(k-w/U) 2 1/2z costy d0 [ (k-w/U) 2 +t2 11/2- (Uk-ip )2 im -& -4 0+ / = (3.25) The expansion of for small values of h k0 R0 (=k-w/UIR) is derived in Appendix 4, and is given by h O(k R , (y,z;k) = 2[H (y,z;k)+ F (y,z-k)]+ 0 K R2 ) (3.26) 00 where -2 H00 (y,z;k) *1 FF*-(y,z;k) 0 (y2z7k =-- + 0 dZ e costy t- (Uk-ip) *-* (3.27) /g * ~- f *0 (kK0 ) [H*0 (y, z;k)-H 0 (y,z;k) ] (3.28) -37- f (k, 0 0 ) ag sn (k) + cosh" (K /k )22 1/2 -sgn (k) - + sgn(k) Zn(2c /k ) + 1 (3.29) i * is the complex conjugate of H Here I (/k 0o 0 + * H , and the upper or lower expression in (3.29) is applicable according as K 0 /k0 Z 1, with K 0 U 2k2 /g k0 and (k-w/U). Transforming 2R , R ) (3.30) back to a physical space, we obtain h (x,y,z)=2[H (y, z;x)+F0 (XyZ)+0 where F 1 (x,y,z)= * * F1 F-1 {f9 (k, K )[H* (y,z;k)-Hf0 (y,z;k)]} * /K.. -l-k It will be seen later in Chapter IV that if (3.31) we interpret x/U as time t,- H (y,z;x) can be considered as a two-dimensional time-dependent source potential which satisfies a twodimensional Laplace equation condition K 0 H oxx + H oZ = 0 V22D H 0 = 0, the free surface on z=0, and the initial (i.e., upstream) conditions of vanishing motion x > 0 on z=O. H *(y,z;k)0 , H = H ox= 0 for Furthermore, the inverse Fourier transform of designated as H0 (y,z;x) , will represent the corresponding source potential that satisfies a reverse flow problem such that = H(c) . H(C) Ox = 0 for x S 0 on z=0. -38- Simplification of the Surface Integrals 111.3 On account of slender-body approximations of the ship geometry, we have = 0( e) n and 0(1) n2 , n 3 (3.32) Thus, to leading order, the normal derivatives in (3.19) reduce to a anQ+ 4. a nQ n (3.33) aNQ a anQ 1 + O(weR 0 , eR0 ) , with an error factor where NQ is the two-dimensional unit normal vector out of the fluid and the vector operators V are defined in (3.15). From (3.33) we obtain the slender-body approximation of (3.19) as rL/2 ( da d' L/2 + Q) 3N.. ds S( W -L/2 dsQ[I (h)(PrQ 0 N0 /2 U I g C dnli 0II-.... T -o 1 ( d U N os[--Ni+ a. + (P) (p/ = ) 4wo '' - ~ - a h MQ0 0(PQ) (3.34) -39with an error factor , 0) 1 + O(weRQ .R where . the transverse section of the hull surface at B(e) is Now we consider the first term of (3.34) defined by L/2 OA (P) s df f/ -L/2 ds N [0(Q) - -1W.(x-0U - r + ] (- e 8 U) (3.35) It is seen that the kernels in (3.35) for small values of are approximately 1/r , 1/r Ix- I except for small R = I 1/x- (3.36) + 0(R ) R0 = 0(R,). Therefore, we recast (3.35) for as follows AL/2 d9 -L/2 ds 0 (x [ 8(x) +0) ( - lnC) Q Q L/2 Q 8(x)Q 1 )e r 1 + L) (-l + ar (- - f ds [0 (x,n,t) 1 ) aN Q - B( - ) -L/2 jri,) - d{J dsQH[ 0 ( (x-) I (337 (3.37) Now we can apply the approximations (3.36) to the second term of (3.37) because the curly bracketed quantity in (3.37) tends to zero for -+ x. Thus A becomes approximately -40- OA ( L/ d9 ds [ (x,n,)dQoa(r.-NQ - - -J-Q -L/2 .8(x) with an error factor Completing the integration 1 + 0(R ). 9 j ~~ jjI dsQ[00(x B(x) we obtain , x. - - aN.] )3* x 2. x (3.38 Q1 in (3.38) with respect to OA(P) = ] (-QL + 1-) 2 2 1/2 tn (x-L/2)+[*(x-L/2) +(y-n) +(z 2+(Z-4 2 11/2 (x+L/2) + [ (.x+L/2 )2+Yn2 + t 2 1/2 2 (x+L/2)+ [ (x+L/2) 2(-) 2+ N-n 2+(Z+4) 2 11/2 (X-L/2)+[ (x-L/2)2 (3.39) For small values of R, and -L/2 < x < L/2, it is also seen that /2.2 (x+L/2) + (x+L/2)+R (x+L/2)+ (x+L/2) [1+ 2 - +...] 2 (x+L/2) =2(x+L/2) (x-L/2) + (x-L/2) 2 +R 2 + 0(R 2 22 (3.40a) - (L/2-x)+ (L/2-x) [l+ ] 2 (L/2-x) 2 + O(R ) 2(L/2-x) (3.40b) -41From (3.39) M Q 8(x) (3,41) ]nR/R - ds0(xn) = 2 A(p) and (3.40) we obtain where R = [ (y-n) 2+ R = ( (y-n) 2+ (z-)2 11/ 2 (3.42a) (z+;)2 1/ 2 (3.42b.) with the cumulative error factor collecting all 1 + 0 (R ) Therefore, the terms in (3.34), the inner expansion of the outer solution in (3.19) takes the form 2ir (p) =f ds 0 0 0 (x,n,,)- ]Q N nR/R1 8(x) + L/2 d ds10( (Q) - 0t ][H (PQ)+F (PQ)] -L/2 8(E) - - [H (PQ)+F (PQ)J with a cumulative error factor 1 + 0(w2R 2, weR 0 , - (Q) ldn[ (3.43) IR, 2 We include in (3.43) the higher-order terms represented by a line integral over the free surface. The justification for this comes from the same arguments we made in Chapter I for the inclusion of speed-dependent terms into the free surface condition. Actually, the contribution from this -42line integral turned out to be important in the computational results of Yeung and Kim (1981). The main,result so far is the inner expansion of the outer solution defined in (3.19). approximation for 0 The corresponding may be obtained using a similar procedure to what we did above (see Appendix 6 for details). 2-rr 0 (P) = T, ) ds 1s0[(XI Q 5(x)Q f QL + +L d ds QN [O(E, -L/2 S(E) -nR/R a 3N Q1 T,) Q- 1N NQ [H (PIQ)+F(PQ)]I dn[2iwU(E,n,O) + U 2(0,.- g The results are 0)] C [H(P,Q) + F(PQ)J with an error factor 1 + O(w 4 R, (3.44) R , R ) x The same result will be obtained if we substitute into (3.43) the following relations, O(P) = eiwx/U(P) H(P,Q)= e U HO (PQ) (3.11) (3.45) -W(x-0) F(P,Q)= eU Fi(PQ) (3.46) which are consequences of the forward-speed transformation. It is obvious that (3.11) should not be used in the case of -43zero forward-speed forced oscillation prohlem. Instead (3.44) will be used without the line integral over the free surface. It is seen that the forward-speed transformation in (3.11) simplifies this line integral in the non-zero forward-speed case. The inner expansions of the outer solution in (3.43) or (3.44) will be eventually matched with the corresponding expressions of the inner solution in a suitably defined overlap region. -44- IV. THE INNER PROBLEM The inner problem is defined in an appropriately restricted region close to the ship hull, at transverse distances of the order of the ship's lateral dimension. The basic assumption in the inner region is that the flow gradients in the longitudinal direction are small by comparison to those in the transverse direction due to the slenderness of the ship. of 0(1) y and that Assuming that the x coordinate is and z coordinates are both of 0(R0 ), a coordinate stretching suggests that -z 0 , ax, ay0(1) =0(R7 1) (4.1) Exploiting the slenderness of the ship, it has also been assumed that ),n2,n 3=0(1) n (4.2) = 0( n e is the slenderness parameter. where Now the flow equations (2.22)-(2.25) will be approximated using (4.1) and (4.2). The three-dimensional Laplace equation is approximated as the two-dimensional one, + ay 0 in the fluid domain (4.3) az with an error factor 1 + O(R 0 ). The body boundary condition -45- reduces to (n2 2ay + n 3 az with an error factor onn N(y,z;x) Nyzx aN 1 + O(eR), where VN (4.4) is the inner approximation of the normal velocity of the ship's wetted Thus, surface. N = N5 = (2.28)-(2.31) (N2 ,N3 ) (4.5a) -xN3 (4.5b) = -(N- + N 3a7 M 3 32ay M5 (4.5c) z (4.5d) -xM3 + N3 with an error factor 2 1 + O(e ), the inner approximations of T reduces to nj N. where and J m. and M. are v) respectively, and is the velocity potential due to the steady forward motion of the ship. The reduction of the free-surface boundary condition needs further comments. Neglecting influences of the perturbation of the steady flow, we obtain the linearized , free-surface condition in the form (iso-U)2 ax I + g a = 0 on z=O If we were to further neglect the forward-speed terms in (4.6) , it would reduce to (4.6) -46- a- az vV0 on z;= , v 2/g (4.7) with a resulting error factor 1 + 0 (wR 0). Then, the inner problem becomes a truly two-dimensional one, being free from any explicit U dependence. In the strip theory of Salvesen et al.(1970), (4.7) is used for the .free surface condition so that the forwardspeed effects are involved only through hull boundary condition. However, the three-dimensional approaches of Chang(1977) and Inglis(1980) pointed out the importance of the forward-speed terms in the free surface condition for reasonable quantitative and qualitative predictions. In a systematic perturbation analysis of Ogilvie and Tuck(1969), these terms are included as higher order effects of relative order e /2 For reasons described in Chapter I, we simply use (4.6) for the free surface condition in the present work under the assumption that the forward speed is appropriately limited so as to justify the relations (4.6). Even though these forward-speed terms may be considered higher order effects, it is possible that their inclusion can significantly improve the results. In summary, the boundary-value problem satisfied by the inner solution takes the form, -47- a2 + 0 az = 0 U I) 2 ++ ga ax (iW with the cumulative error factor (4.9) in z=0 97 on VN (yZ;X) =N IV.1 (4.8) in the fluid domain (4.10) 1 + 0 (eR, R ) ay 2 0 0 The Pseudo-Time Inner Problem In accordance with the formulation of the outer solution, we set (y,z;x) = eiwx/U ( (y'Z;X) (4.11) The corresponding boundary-value problem for 40 then takes the form, a2 2 2 4 1 + U2 2 0 aN = a -w/ 2 +g = in the fluid domain (4.12) on z=0 (4.13) on (4.14) 0 ---= 0 z NYzx with the cumulative error factor F 1 + 0( eR, R2 0 0 -48- A similar boundary-value problem to (4.11) - (4.14) has been studied by Yeung and Kim(1981) based on a "transient This is readily apparent formulation" of radiation problem. if one simply replaces the x/U by time t. For more details Yeung and Kim(1981) should be referred to. Since no outer matching was used in the analysis of Yeung and Kim, initial (i.e., upstream) conditions of vanishing motion were imposed for the uniqueness of the solution, which could be justified for T = wU/g > 1/4. We will use the method of Green function to solve the above inner problem. (y,z,n,rC;x,5), with super- G Let script (i) denoting "inner", be an appropriately defined Green function for (4.12)-(4.14), of the form, G0 G Here "initial" = (4.15) G 00(y,z,n,C;x,F) + E (yznI;xr) is defined such that it satisfies the upstream conditions of G0 = G The boundary-value problem for = G0 0 for and its x > E on z=0. solution are derived in Appendix 2, of the form, G= H 6(x-&)tnR/R1 + H (y,z,n,r ;xI) u (E -x) d (4.16) e(z+r) cost (y-n) sin//7 a (E-x) (407 (4.17) -49- where 2 (y-n) 2 + (Z R = 1/2 (4.18) , R= [(y-n) 2+ (z+;) ]l/2 (4.19) a = U 2 /g (4.20) 6(x-) u(E-x) and are the Dirac delta function and the Heaviside step function respectively. Since the inner problem defined by (4.12)-(4.14) has no E corresponding initial conditions, the harmonic function cannot be determined until we proceed the matching with the outer solution. However, it is expected that and H0 some modulated combination of E will be (c) for both of Hc, them satisfy the free surface condition similar to (4.13). The boundary-value problem for H(c) 0 and its solution are also derived in Appendix 2, of the form, H c 0 = - afp (x- e(Z+,) cost (y-n) sin/Z7a / dZ 0 Here Hc) 0 (4.21) is defined such that it satisfies the downstream initial conditions of IV.2. 0(x-E) H(c)= H(c) ox 0 - 0 for x < E on z=0. Integral-Equation Formulation in terms of G In order to solve the problem at hand, we start out by applying Green's second identity to t (Q;E) 0 and (PIQ;X. G(i) 0 -50- S (c), a large in the region bounded by the body contour semicircle CR, and the portion of the undisturbed free F(E) surface as shown in Figure CR and B(E) between Thus we obtain 3. 2n6 (x-E) 00 (P; E)= f ds Q [ 0o3N (Q;E) 8 (E) +F W) Q - (4.22) normal vector out of the fluid in Q-space. over CR CR vanishes if we let (PQ;x-) 0-4G 9 is the two-dimensional unit N P=(y,z), Q=(rj,,), and where - The integral tend to infinity. E ting both.sides of (4.22) with respect to Integra- from -L/2 to L/2, we obtain ds [ (P;x)= (Q;x) -- 0Q + JL/2 -L/2 ds d I[ 4,DG Q B(E) x[H fL/2 -L/2 0- nR/R Q -Q 0] x 00 (P,Q;x- ) + E (P,Q;x-F)] dn[ $ (Q;.) 0 - 1; ]G 0i) (P,Q;x- ) 27r F(E) (4.23) -51- AL .a FW N FC Figure 3 Domain of application of Green' s theorem Since both and G(1) 0 0 satisfy the free surface condition, it follows that [( a 0 a=O 0 )G (i) - 2 9~G5E -) 0E 3E (H +E) 00r (4.24) Making use of (4.24) takes the form, and Stokes's theorem, equation (4.23) -52- ds0 [% (Q;x) 27r 0 (P;x)= 0 nR/R - 8 (x) L/2 + as [ 1 (Q; ds Qs[t d() 0 8 (W -L/2 x Q [H 0(P,Q;x-E) + E (P,Q;x-)] a 2 -U g 0 0 H [ (Do (Q; 0 Cd x N a NQ (P,Q;x-F)+E (PQ;x- )I 0 (4.25) where C is the intersection of the hull surface and the undisturbed free surface. An integral equation for 5 (x) on 00 can be readily obtained by letting P that the factor in (4.25) is now replaced by 2w approach The net effect is . B(x) 7. Thus we obtain 7 0 (P;x) - ds [ (Q ;x) 0]ZnR/R - 5 8(x)0 f L/2 d- f -L/2 ds0 I B (9) Q (Q;) 0 aN Q as 2 g x + E (PQ;x-V)] x[H 0Qx-V U as -0] 3NQ -0] Cdo [ D (Q; [H0 (PQ;x-C)+E (PFQ;x-0) J C for P(y,z) e 8(x) (4.26) -53- t, which satisfies the boundary- The inner solution value problem defined in (4.8)-(4.10), can be obtained from (4.25) and (4.11), or alternatively it can be expressed as 21r t(P;x) = ' x ds It(Q; x)'--- a ao InR/R 1 Q Q 5(x) ds [(Q;1) +(L/2d4 -L/2 - + E(PQ;x-)] x[H(P,Q;x-E) id [2iwU a (Q;E)+U2 x[H(P,Q;x-) where H and E Q Q S(E) + E(PQ;x-)] (4.27) are defined such that H(P,Q;x-E) = e U -(x-o) H (PQ;x-) (4.28a) E (PQ;x-) (4. 28b) -(x-) E(P,Q;x-E) = e U which are consequences of the forward-speed transformation. Thus, an integral equation for form, t on 8(x) now takes the -54- 7 t (P;x) i - dsQ [ 8 (x) f 2d -L/2 (g) x aQ Q I [H (PQ;x-)+E (P Q;X-0)I ds [ + (Q; ) a NQ aNQ x [2iwU(DQ;)+U2 (0 a [H (P, Q; x-g ) Cdn S D]ZnR/R aNQ - Q;x)- + E(PQ;x-)] for P (y, z) e 8 (x) (4. 29) D is obtained from the integral equation for t (P;x) - f 8(x) - MQ { L/2 ds 8( -L/2 I [tlnR/R [ (Q;) 0a + H2D(PIQ)I QN0 - aD]E(PQ;x-) I ) = a ds Q 0 , For the zero speed case, where (4.11) is not applicable, for P(y,z) e 8(x) (4.30) where H 2 D (PQ) - e (z+)cost(y-1) m i-2 2+O 0f at t-(W-iv) 2/g (4.31) -55- Here the limiting behavior, SL/2d -L/2 = f dsQ[V(Q;H) ,Q;x-E) - Q 8(t) dsQ[O(Q;x) - 0H (PQ) (4.32) Q 8(x) has been used to obtain (4.30) from (4.29). It is worthwhile to discuss the character of the motion furnished by (4.25) in more detail. The first term in (4.25) is evidently the fluid motion generated by an impulsive body motidn characterized by the hull boundary condition at x, assuming no previous disturbances existed in the fluid. However, the ship moves with forward speed U, and therefore, each cross section of the ship will pass through the flow field disturbed by the rest of the ship. The second term represents the memory effects due to hull boundary conditions on the rest of ship sections. Since the inner solution satisfies the two-dimensional Laplace equation, effects due to the presence of any transverse waves are absent unless the three-dimensional interactions characterized by E0 is included. The third term of (4.25) has been resulted from the inclusion of the forward-speed effects in the free surface condition. If the ship's forward speed approaches zero, the fluid motion -56- becomes time-harmonic so that it is free from any explicit dependence on U. Owing to the transient disturbances generated by a finite forward speed, however, we can hardly expect the two-dimensional time-harmonic representation of the solution to always remain physically valid. -57- V. V.1 MATCHING OF INNER AND OUTER SOLUTIONS The Matched Solution We are now in a position to match our inner and outer solutions in a suitably defined overlap region e << R << 1 to determine the unknown source and normal dipole strengths of the outer solution, and the three-dimensional interaction function in the inner solution. The process of matching involves equating terms of similar form from the inner expansion of the outer solution with those from the outer expansion of the inner solution. The outer and inner problems were analyzed in Chapter III and IV respectively, and the corresponding solutions were expressed by (3.19) and (4.25) respectively. Equating the inner expansion of the outer solution in (3.43) with the corresponding expansion of the inner solution, we obtain 4 (x,y,z) = 0 (y,z;x) (5.1) N ,z;x) ON (yD (5.2) E 0 (y,z;x) = F (x,y,z) (5.3) where the cumulative error factor in (5.1)-(5.3) takes the form, Err = 1 + 0(W R2, weR , 0 0 R , eR 0 ) 0 (5.4) -58- The approximations (3.43) and (4.25) are valid in a region where the corresponding error factors approach one as Let -i+-O. R = O(eB) 0 < B < , in the matching region. (5.5) Thenthe error factor in (5.4) takes the form, Err = 1 + 0(e 2 $-2y I w where 1+$-y 1+a) 20 is parameterized in (5.6) as W = O(J T ) (5.7) Hence, the matching conditions only if (5.6) 2U-2y > 0 and (5.1)-(5.3) are legitimate 1+0-y > 0. Therefore, y < a < 1(5.8) defines the matching region. Alternatively, if we do not apply the forward-speed transformation, we will obtain from (3.44) and the corresponding expression of (4.26) the matching conditions, = .(x,y,z) ON (x,y,Z)= 0(y,z;x) PN (y,Z;X) (5.9) (5.10) -59- E(y,z;x) = F(x,y,z) (5.11) with the cumulative error factor, 42 2 Err=l + o(w R ,R , 0 0 eR ) (5.12) 0 Thus, the direct matching provides the matching region, 2y < 0 < 1 (5.13) For sufficiently large values of w such that + (e-1/2 a separate matching must be conducted from the high-frequency inner approximation of the outer solution. This will be discussed in more detail in section V.3. By comparing (3.43) and (4.25), it is now clear that the wave kernel H in the pseudo-time inner problem, i.e., the portion associated with homogeneous "initial conditions", forms only a part of the inner expansion of the outer Green function. of E0 The remaining part is accounted for by the presence in the inner Green function. It is precisely due to this reason that we need to introduce a generalized Green function in formulating the inner problem in Chapter IV. The matching procedure thus defines E 0 , which contains the relevant three-dimensional characteristics of the problem. In terms of the inner variables, we have therefore the following integral equation for t (P;x) to solve: -60- Tr 0f - (?;x) ds [P Q 8 (x) Q 0 Cd -~ (Q;x) - aNQ a - o (; ZnR/R1 [H0 (P,Q;x-)+F0 (P,Q;x-)] )(a BC -L/2 0 Q( [H (P,Q;x-()+F(P,Q;x-E)], - for P(y,z) e B(x) (5.14) which represents the final results of the matching analysis. 0 must be solved numerically from this integral equation. The behavior of the kernal function The potential F0 is of obvious importance and will be examined in the next section. For completeness, we will also provide the corresponding integral equation for which is obtained from a 4D(P;x) formulation without the forward-speed transformation: , 4 (P;x) - -]nR/R 1 ds [((Q;x) - 8(x) =L/2d -L/2 -1 f ds [(Q;-) - [H(P,Q;x-E)+F(P,,Q;x-E) I 8(E) dn[2iwU (Q; )+U2 (, - for P(y,z) f][H(PQ;x-E)+F(PQ;x-)],, e 8(x) (5.15) -61- F. Behavior of Interaction Functions V.2 The kernel function and. F in (5.14) essentially contributes F a "three-dimensional correction" to the pseudo-time inner problem, which would otherwise have no upstream effects because of the step-function property of (4.17). H as given by Considerable amount of analyses were necessary to extract the essential behavior of out in Appendix 5. This was carried The final results may be summarized as 1 F F (x,y,z) o0 ~ Fo. 2 ) (x,y,z) + F ( (x,y,z) , x < 0 , x > 0 (5.16) 0 (x,y,z) F where F (x,,Z) kldk e-ikxeKozcosK y {l-[l-k2 K2 -1/2 = + 2i[ fk - 2i k2dk e-ikxeKoxcosK y 0 2 W/Udk e-ikxe Kozcos K y {l-[l-k /K 2 -l/2 k2 (5.17) F 0((x,y,z) k3 + =-i[ ici/U r-ikx 0dk k KozCOS KY 0 x{l-[l-k2 0 -i 4 k dk e eK 0 X cOsK y 2 -1/2 0 {l-ik 2 /K 2 0 0 - 1 /2} 3 0 < -T < 1/4 (5.18) -62- Zcos K y {1- [1-k /K dk e-kXe 0 f W/U T where K = U2 k2/g , root singularities of (k 2 -1/2 - 2 k (2 (xyz) F > k0 = |k-w/U|, and (5.19) 1/4 are the square- k. F 0 (y,z;k) defined as W/U)2 =.U k /g2 i=1, . . ., 4 (5.20) T = oU/g > 0 (5.21) Thus, k i2 {1 ~l-+4r} = ,F 2U 0 < k = T < 1/4 (5.22) 9 1 2U, 1 + i T > 1/4 F (x,y,z) The previous expression indicates that regular, being finite for all totic behavior of x. for large F is The leading-order asymp- lxI can be obtained from the theorem on Fourier integrals (see Copson,1976), and the results are of the following forms, 21r 1 7X- /2 (1+4-c1 4 x ipix/6-i32/4 2 P2 e 2 e-i2x+i 3 / 4ep )}x~ + 0(1x|) e 22/6~cos(p 22y/M Ix I-* , C > 0 , / F 0 )x~~)[ (5.23) -63- F(2) 2 4x-i3 w/4, 42/ -i -1/4 1/2 2 -1 cos(p 4 y/6) + O(IxI X+-co, 0 < < 1/4 T (5.24) 1/2 y Z/ F o( ()21r &/4 2 x 1 /4 - x 3e~ P 3x/d+i37r/4 e z/6 x cos(p 2y/6) + O(IxKI') x+ 0 < T < 1/4 +00, (5.25) F (2)(x,y,z)=(e- V4T-lx|/6) 6= U2/g where and p , xf+o , T > 1/4 (5.26) = k 6. The corresponding expressions for F(x,y,z) in (5.15) can be easily derived from (5.16)-(5.26) by using the following relations, F(x,y,z) = e1wx/U F0 (x,y,z) (5.27) F (y,z;k)= F 0 (y,z;k+w/U) (5.28) and we will not repeat the expressions here. for T We note that > 1/4, the interaction function (5.26) is exponentially negligible upstream. We will now discuss some special cases of the parameters and w U in some detail. -64- V.2.1 Zero-Speed Forced Oscillation; U=0 , w#O This limiting case corresponds to the physical problem of a ship with U=O performing forced oscillations at a w on the otherwise calm free surface. finite frequency Since the forward-speed transformation is not applicable in this special case, we will analyze the kernel function F(x,y,z) instead of F (x,y,z). The corresponding expression for F(x,y,z) can be obtained from (5.16)-(5.22) by using (5.27) and (5.28). f"dk e-ikx F (y,z;k+/U) F(x,y,z) =1 - Here K = Thus, we have (W+Uk) 2 /g dk e-ikxD(k)eKz cos Ky and D (k) is defined such that D(k)=Zn[2K/k]+7ri sgn(w+Uk)-7 x (5.29) 1 - k2 i sgn (w+Uk) -cosh -(K _ + cos 2 (K/JkJ) -1/ 2 /Ikl) k (5.30) where the upper or lower expression in the bracket is applicable according as K/|kj '< 1 (5.31) -65- F The corresponding square-root singularities of are obtained from (5.21), (5.22), and (5.28). , k 1,2 = k 1,2 - w/U = (y,x;k) Thus, we have - 2U22 {1+2T + /l+4T }, T=wU/g 0 > W/U 1 = k 4 111-2T 2U 21-2T 1-41 + -i/4-r-1 0 < T 1/4 < T > 1/4 , k , (5.32) (5.33) Since T+O U-0, the upper expression in for applicable in this limiting case. timk (5.33) is It is easily proved that (5.34) -O = - v imk 23 U 2,3 2 v W /g (5.35) Replacing the integrand and the limits of integral in (5.17)-(5.18) by the corresponding ones in (5.29)-(5.35), k = k/v and making use of a coordinate transformation we obtain F F(x,y,z) = (1) (x,y,z) + F (2) (x,y,z) (2) (xyz) x < 0 (5.36) x > 0 where F(1) (x,y,z) = -2iv eVz Cos vy {dk e -2iv e VZ Cos vy J dk e -ikvx ikvx 1 1-k 2 1-.1/27 (5.37) -66- F (2) (x,,z) = -iv eVzcos vy 1 dk e-ikvx{l-l-k 2 -l/ 2 Q -iv e'Vzcos vy dk e-ikvx{1-i[k2_1]-1/ 2 (5.38) By using integral representations of the Bessel and Struve functions defined in Abramowitz and Stegun(1964), the x)r--x) F (x,y,z)=e V z cosvy{-l/ lx|+(7rv/2) [H (v ) expressions in (5.36)-(5.38) can be reduced to (5.39) + 2iJ 0 (vIxj)]} where J0 , Y , and H of zeroth order. +e are the Bessel and Struve functions The resulting expression (5.39) shows a symmetric disturbance in x. Now the integral equation for O (P;x) corresponding to this limiting case can be obtained from (5.15) and (5.39) as 1 aD(Q;X) - - [tnR/R1 + H2D 8(x) -L/2 dsQ[ 1(Q;) ]F(PQ;x- ) - Q 8(e) for P(y,z) e (x) (5.40) ( =L/2 d , ift (P;x) -67- where 2 Zim H2 +(Z+c)cost.Zy-n) H 2 D (PQ) = -2 V.+ (5.41) 2/g Here the limiting behavior, L/2 dE -L/2 dsQ[tO(Q;) - 1H(P,Q;x-E) 8(E) (5.42) ,r H2(PQ) - fxdsQ[O(Q;x) has been used to obtain (5.40) from (5.15). However, the integral equation (5.40) is not complete, defined in (5.39) is singular at F(x,y,z) for the kernel If x=0, and furthermore the singularity is not integrable. we integrate the right-hand side of (5.40) by parts, we will obtain Wi (P;x) - f ds [ (Q;x) MQ Q - a] [ZnR/R +H 2 D(PQ)I 12 NQ B (x) -L/2 B( - ev (z+r:) cosv(y-n), ) ds [ (Q; )0 a - L/2dFA (X-) for P(y,z) e 8(x) (5.43) FA(x) = sgn(x){[Zn(2vjx + y + 7il - where 7r IXIdt [IH(t)+Y (t)+2iJ( t) ]} I (5.44) -68- Here y " 0.577... that the kernel is Euler's constant. FA (x) It is observed in (5.44) is identical to the corresponding one in the unified slender-body theory (Newman and Sclavounos (1980)). In this special limiting case, the three-dimensional interaction function F(x,y,z) in (5.39) is represented simply as the product of a homogeneous Green function times a function of x such that G M= 6(x-E)G 2 D(PQ) + F(PQ;x-) = 6 (x-") G2 D - [(X-[G G2 D1 (5.45) where G 2 D(P,Q) = ZnR/R1 + H (PQ) (5.46) Far from the ship in the inner region, the inner solution 0 can be written as (see Appendix 7) L/2 O(y,z;x) i) d~a(E)G = (y,z,0,0;x-E) L/2 = a(x)G 2 D(y,z,O,O)- [G2 2D] L/2 da(E)f(x-E) -L/2 (5.47) Here a (x) is an effective source strength which is defined -69- It is Qf interest to compare (5.47) with the in (A7.13). corresponding far-field expression of the unified slenderbody theory, c(y,z;x){ (s) - x) + C(x) [a (s)x) C (x) a (x) + a (S) (5.48) [G 2 D~ U 2 D] where CCX) 1 L/2 d( o( )f(x-) = (5.49a) , -L/2 a (x) = a() and a (S) + C x) [ (S (X) + (5.49b) (s)(x) 1 is an effective source strength which is derived from the strip theory solution ds a (S)(X) t (s) such that [t (S) (N,(;x) TIF I 6(x) a-s) -ev( +in) aN 0 NQ (5.50) From (5.47) and (5.48) we can draw a conclusion that the present formulation and that of the unified slenderbody theory have the same far-field behavior of the inner solution. In principle, both theories are identical in this special limiting case, even though their approaches to the solution are quite different. More specifically, -70- the former approach includes the three-dimensional interactions through the construction of a generalized inner Green function, while the latter approach does through the construction of a general inner solution. Both approaches involve unknown interaction functions to be determined by outer matching. In a computational standpoint, the unified slender-body theory approach (Newman and Sclavounos,1980) is more economical than the present one especially for the zero-speed problem, in which the two-dimensional time-harmonic free surface condition appears to be a good approximation. How- ever, it should be noted that the present work is directed to the development of a new approach to the more general non-zero speed problem, and that the expression in (5.43) represents simply a limiting form of the present more general results. V.2.2 Steady Forward Motion; w=0 t0 T, This limiting problem is of particular interest in the sense that the present formulation may provide a promising approach to the steady forward-motion problem of waveresistance. It has long been known that quantitative predictions from the ordinary slender-body theory are poor (see Newman,1970). This has generally been attributed to -71- the fact that the flow caused by a line distribution of sources has wavemaking effectivenesses that are different from those of a surface distribution. In addition to the above fact, Ogilvie (1977) pointed out that ordinary slenderbody theory at the lowest order is deficient in the sense that it does not include any effects of the diffraction by the ship of its own waves. Waves are generated at each cross section of the ship, and these waves must undergo diffraction by the rest of the ship. In other words, these waves have an associated velocity field which violates the body boundary condition. Since the interference among cross sections appears in the form of a function of x alone, these diffraction effects cannot be accounted for in the ordinary slender-body theory. Daoud' s (1975) and Chapman' s (1976) analyses, in which the speed-dependent terms in the free surface condition were retained, both included the diffraction of the diverging waves by the body itself. However, since their solutions satisfied a two-dimensional Laplace equation and no useful outer matching was used, there can be no representation of transverse waves, which, on the other hand, did appear in the ordinary slender-body theory as a function of x alone. Recently, Maruo(1982) derived a new slender-body formulation by approximating the kernel function of the Neumann-Kelvin problem near the ship axis. It is of interest -72- to notice that Maruo' s analysis turns out to be a rather crude approximation of the present more comprehensive results. First, it can be noticed from (5.27) that F(x,y,z) = FO(x,y,z) for w=0 (5.51) The corresponding square-root singularities of F (y,z;k) are, therefore, obtained from (5.21) and (5.22) as = k2 3 k1 ,4 - im i m k = 2 g (5.52) 0 (5.53) Replacing the integrand and the limits of integral in (5.17)(5.18) by the corresponding ones, and making use of a coordinate transformation k (xyZ) F (x,y,z) kd = + 6 = U 2 /g, , F(2) (x,y,z) we obtain , x < 0 , x > 0 (2) F (0 (5.54) (x,y,z) where i F~1 )(xyz) dk e-ikx/6e k 2 z/6cos(k2y/ 6 [lI-k f4dk ei +~ (2) -ikx/S6 = - (1 dk e 0 e cos~ky/S) 7k2 (5.55) k 2Z/6cos(k2y/) ek2 Z/6cos (k2y/6 ) [1-ik k2 (5.56) -73- These integrals in (5.55) and (5.56) are greatly simplified compared to those which we started with. If we simply set y and equal to zero (i.e., z evaluated at the x-axis), the resulting expression reduces F (x,0,0)=-2+sgn(-X)J/x + {H(-x/6) + to the one derived by Tuck (1963): (5.57) + 2(2+sgn(-x)]Y 0 (x|/6)} which shows that Tuck's three-dimensional interaction function is a special limit of the present one. was a very much simplified one, Tuck's inner problem viz. the rigid free-surface condition was assumed, in contrast to our more elaborate development. Replacing T, we obtain the inner in (3.43) by cO expansion of the outer solution for this limiting problem as ds [T(xn, ) 27rT (6)= InR/R - 6(x) L/2 d + dsQ[T( -L/2 x - B(E) x U F,) Cdn [T( [H 0(PQ;x-E)+FO(Q0x-E In ,)' -E I [H (P,Q;x-E)+F0 (PQ;x-9)] (5.58) with an error factor 1 + 0 (eR , R2), where i represents the velocity potential due to the steady forward motion of the ship, and F is defined in (5.54). Similarly, the -74inner solution can be obtained from (4.25) ds [T(Q;x) Q 5(x) L/ l + .aNQ ]nR/R1 - -L/2 f dsQ 5(eM x - a[ I(Q;) - - -f 21r(P;x) as [H 0(P,.Q; X-) (Q;() a- 9 C [H x QN +E (,0 I (P,Q;x-E)+E, (PQ;X-M) (5.59) Therefore, the matching procedure described in section V.1 will provide the matching conditions, N(x,y,z) = T(y,z;x) (5.60) TN(y,Z;X) (5.61) (x,y,z) (5.62) E0 (y,z;x) = F with the cumulative error factor integral equation for = - DQ B(x) L/2dt -L/2 g j(P;x) takes the form, ds [ET(Q;x) f B( C Now the as [I (Q; 0) -nR/R aNQ - Q 9NQ 1 I (H0 (P,Q;x-E.)+F0 (PFQ;X-M~ 0 ) WiT(P;x) 1 + 0 (eR ,R ). 0 (PQ;x-E)+Fo( . ,0X-0 for P (y, z) S W(x) , = T(x,y,z) (5.63) -75- We should note that the combination of and F (x-g,O,O) H (P,Q;x-g) in (5.58) turns out to provide the similar expression as Maruo's(1982), which is, strictly speaking, an inconsistent approximation in the present context. V.3 >,>l The High-Frequency Approximation; The high frequency approximation of the inner expansion of presents a particular interest because of the h(x,y,z) simplicity of the final expressions. cR 0 and kR , where 0 For small values of K=(w+Uk) 2 /g, we obtain from the results of Appendix 6 2 2 2 2 * * * h (y,z;k) = 2[H (y,z;k)+F (y,z;k)] + O(k R 2,K R ) (5.64) where ,,,~ (5.65) F (y,z;k) = 1 - * f kK H (y,z;k)J H (y,z-k) = 2(1+Kz)[enKR 0 + y + ni sgn(w+Uk)-KZ+KyO] + 0 (K2 * ( ) = ) (5.66) , -sgn~w+Ukfi sgn(w+Uk)+cosh f -)f + cos 1 11-k2 /K2 (K/k) + sgn(w+Uk)Zn(2K/Ikj) + x i and (K/lkI) y = 0.577... is Euler's constant. , (5.67) -1/2 -76- For high frequencies of oscillation such that and K>[kj, f is f*(k,K) (k,K) 0 = (k2 w+Uk>O approximated as (5.68) K2) Here the upper expression in the bracket of (5.67) is only applicable and its limiting behavior, 1-k2 I -1/2cosh~I(K/|kI) = Zn(2K/jkJ)+0(k2 /K2 (5.68) from (5.67). is used to obtain KR and large 0 K, (5.69) Therefore, for small h (y,z;k) reduces to ** 2 2 2 h (y,z;k) =.2H (y,z;k) + O(k /K , k R /K, 2 2 R 2) K (5.70) The inverse Fourier transform of (5.70) then gives = 2H(y,z;x) + 0(w~ , R-2R, w R (5.71) ) h(x,y,z) Since the approximation (5.71) has been derived based on the assumption that KR0 << 1, a complementary analysis is required for sufficiently large values of + of (F- -1/2 h and for large integration of h*(y,z;k) >> 1. KR h = -4 K w such that The precise form of the expansion will be obtained by using a contour defined by +J. I+>0 fo dZ exp[k 2 + 2 1/2z cos y [k2 + 2 1 2 - (w+Uk-ip ) 2/g (5.72) -77- of and KR for large values h The asymptotic approximation of has been derived in Newman(1978), and is K given by h 2 -1 / 2 (x,y,z) = 47i(1-k rexp[Kz-i ly(K 2 - k 2)1/23 2 cose 0 + (5." 73) 0 (K-2R2 0 From the results of Appendix 3, we also obtain the asymptotic approximation of H H for large (y,z;k) = 27i exp[Kz-iK|yl ] + cOs 6 0 KR as 0 (5.74) + 0 (K-2 R-2 0 Comparison of (5.73) and (5.74) then gives the result, h (y,z;k) = 2H (y,z;k) + 0(k2 K 2, k2 R/K, Thu0 K- 2R- ) (5.75) Thus, h(x,y,z) = 2H(y,z;x) + O(w , -R 0 , 4R-2) (5.76) Therefore, in this special limiting case, we obtain = $N(xryz) = E(y,z;x) D(y,z;k) D (y ,z;x) = F(x,y,z) = 0 ( t(x,y,z) 5.77) ( from (5.9)-(5.11) and (5.76) the matching conditions, 5.:78) (5.79) -78- with the cumulative error factor, - Err = 1 + '0 (W -4R- , R , R ) (5.80) Thus, in the matching region, we will have (5.81) 0 < $ < 2y Since F(x,y,z) vanishes for w>>l in accordance with (5.71), the results (5.9)-(5.11) are consistent with the high frequency results. For this reason (5.1)-(5.3) are valid, in general, for all frequencies of oscillation. V.4 Alternative Representation of the Solution in terms of a Line Distribution of Wave Sources In Chapter III, an appropriate outer solution has been obtained by using Green's theorem. In other words, the outer solution is expressed in terms of a surface distribution of wave sources and normal dipoles over the ship's wetted surface. It is possible to show that the matching procedure is not fundamentally changed if of wave sources is used instead. a line distribution To do this, we consider the outer solution being represented as follows: (x,y,z) = d q( ) G ( (x-F,y,z) (5.82) -79- Here q(x) is the unknown source strength which will be determined by matching with the inner solution, and G source located on the x-axis at From (3.5), x= . is the potential of a translatory harmonic oscillating is expressed as G h(x-_,yz) G(O) (x- ,y,z) (5.83) In order to match the inner expansion of (5.82) with the outer expansion of the inner solution, an asymptotic expression of G (0) for small R (= y +z 2)is required. From the results of Chapter III, the inner approximation of G(O) in (5.83) can be expressed as G(O) (x-E,y,z) 2(H(y,z;x-E)+F(x-Ey,z)] = R<<l , (5.84) The inner solution has been derived in Chapter IV and given in (4.27) as 2Trr(P;x) ds Q(IQ;x)D- -9 = ..nR/R B(x) [ L/2 d + -L/2 - ) (-] d D x B(E) Cdn[2iwU('(Q; )+U2 ((D x X [H(P(4;X-2)+E(P,7Q;x-) (4.27) -80- E Here the harmonic function is unknown and will be determined by matching with the outer solution. Far from the ship in the inner region, the inner solution in (4.27) can be expressed in terms of a line distribution of effective wave sources, OdE a (E) (y,z;x) = G (y,z;x-) (5.85) R 0 >> , where the exact relation between the effective source strength derivative a(x) DN and the body potential will be addressed later. D and its normal In (5.85) G is the generalized inner Green function which is defined as G(i) (y,z;x-E) = H(y,z;x- ) + E(y,z;x-E) (5.86) Here the relations (4.28a,b) are used to obtain (5.86) from (4.15). Equating the inner expansion of the outer solution in (5.82) with the outer expansion of the inner solution in (5.85), we obtain (5.87) q(x) = a(x)/2 E(y,z;x) = F(x,y,z) (5.88) Now the matching procedure defines the unknown function by (5.88). Thus, the integral equation for (P;x) E can be -81- obtained from (4.27) in terms of inner variables, WO (P;x) ds [QD(Q;x) . - -1nR/R - a(x) f L/ aI[H(P ,Q;x-)+F(P ,Q;X-) 2 df -L/2 ds [4(Q;) Q ~~~ - awQ N 8(C) dn [2iwMUcNQ; )+U 2 (H(pQ -1 X ) F(pL C for P (y,z) e B (x) The unknown source strength q(x) (5.89) in (5.82) is also' determined from this matching procedure, provided a(x) is known. The effective source strength a (x) has been derived in terms of inner variables in Appendix 7, L/2 ds [ . d -L/2 (D , ;) a;xQ B(t) C.dn [2iwMD (n,0;g)+U2( M0 - -W < ) M(T ,0 ;x-g) x < , 2ra (x) = (5.90) W where S(-4U 2 / (~n,,C 04U g 4U 2 (+in)]exp[ +) 2 in+ U (5.91) -82- It is of some interest to point out that for the special case of zero forward-speed, U=O, . ds [c(Dn, ;x) 2ra (x) (x) simplifies to - 3 e.v (C+in), |xl< L/2 lx> L/2 (x) 0 , (5.92) where v 2 /g The representation of the outer solution in terms of a line distribution of wave sources turns out to provide the same results as those of surface distribution of wave sources and normal dipoles. However, it is worthwhile to note that this line distribution, in general, should be extended over an infinite length in order for the outer solution to be completely compatible with the outer expansion of the inner solution. From a computational standpoint, the representation (5.82) would probably be more economical than that of surface distribution if farfield behavior of the outer solution is desired for a given q(x). -83- VI. ADDED MASS AND DAMPING We consider the unsteady component of the hydrodynamic pressure force, with the usual assumption that the oscillatory motions of the ship and the fluid are small. Neglecting second-order terms in Bernoulli's equation (2.3), the unsteady hydrodynamic pressure is given by -p7iWV+-Vi+l/2$-'VW 2 ] P(x)= on (6.1) The last term in (6.1) gives a force proportional to the unsteady displacement of the ship, and hence an additional buoyancy force to the hydrostatic restoring force. Since we are now interested in the linearized unsteady pressure force associated with the added mass and damping, we will use (6.1) without the last term. Following the conventional definition of added mass and damping coefficients, the unsteady hydrodynamic pressure force can be written in a form, F = dSpn= p dS(iw.+W'.-V? j -w 2 1(a +b t/io) )n J , i=3,5 j=3,5 (6.2) Cb -84- where the quantities a are the added mass and b and damping coefficients associated with the force (moment) in the i-th direction due to the j-th mode of motion. The term in (6.2) proportional to the steady velocity can be transformed by means of a theorem due to field (Ogilvie and Tuck, 1969) , Tuck dS(n-V)T 1 - = dS( -vVT)i - (6.3a) (dxp*) C N fdS(v') (Zxn) = (dZx I)x x dS(-V) (nxW) - (6.3b) C K where the line integral is over the intersection of the ship hull with the undisturbed free surface. It is noted that the line integral in (6.3) may be ignored for a slender ship, for it is of higher order than the remaining surface integral by the relative order of in (2.30) m F= and (2.31), . From the definition (6.2) can be transformed to e (6.4) dS(iwn. - Um )T.] p E. E [ S.. 3 and a. + b 1) /iw = 13 (P/ 2) fdS(iwn - 1 )T (6.5) -85- Let * (yz;x) and $ (y,z;x) be the numerically determined solutions of the Inner problem defined in (4.8)(4.10), with the following hull boundary conditions on S(x) =n (6.6a) =m (6.6b) aN aN i From the boundary condition for T. J in (2.27), the unsteady velocity potential can now be expressed in terms of and '. V such that = From (6.5) and (6.7) the added mass and damping coefficients can be obtained in the form, a + b 1)i/iw13 )1= f) + f!I) + f i) (6.8) where f = p fdS n *. (6.9) J ij W j i -86- U )2 f II) (6.11) i=3,5 dS m j3,5 =f Now the expressions in (6.8)-(6.11) will be discussed From Green's theorem and the boundary in more detail. conditions (6.6), we can show that dS[n 4i 0= where ES dS[* = i ff ES ] - ni rsJ - DNaN represents a closed surface bounded by the ship's wetted surface F, a closure surface portion of the undisturbed free surface and S, as shown in Figure 2. be shown to vanish if we let * both (6.12) -] and SF S., and the between The integral over S, tend to infinity. So S can Since satisfy the free surface condition (4.9), ii it follows that )- fI) FfdS[n i = (p/g) -n dy[2iwU* ] + U2 (4i a - f C (6.13) It is of interest to notice that the line integral in (6.13) has the similar form as that of (4.27). Actually, this line integral results from the inclusion of the forward -87- speed effects in the free surface condition. It is also seen that the line integral in (6.13) is of higher order than the surface integral by the relative order -ew. If we further neglect the contribution due to this line integral, we would have ) f = (6.14) P By using a similar analysis to what we did above, we can also show that (II) f(II) (6.15) J. 1) (6.16) !111) f III)= f ji :LJ with the resulting error terms represented by a line integral similar to the one in (6.13). Since the velocity potential in the present formulation retains the similar line integral (see equation 4.27), we do not expect that the relations (6.14)-(6.16) will generally hold. In the derivation of the reverse-flow theorem of Timman and Newman (1962), it was assumed that = G(O +(+ where the superscripts ) x (6.17) ( ) denote directions of the forward speed, and the Green functions G ~ the linearized free surface conditions g 3G () /ax] = 0 conditions aG on /an z=0 are defined to satisfy [(iw UB/3x) 2G () + G and the homogeneous body boundary on the body surface. Strictly speaking, -88- the reciprocity relation (6.17) was obtained from the approximation which is equivalent to a neglect of the forward-speed effects on the free surface condition (see Since these more details in Timman and Newman, 1962). forward-speed effects are, however, included in the present formulation, the applicability of the reverse-flow theorem to the present formulation is questionable. In an intuitive approach, if the gradients of the steady-state disturbance velocity field are considered negligible, then the only nonzero element of m5 which is identified to 3* 5 and n3 . m (j=3,5) is Thus it leads to *3 = 0 With these simplifications in (6.8)-(6.11), we obtain a + 3 3 /i a53/iA a = (I) b = -(U/ii) + b3 5 a5 5 + b/iW = (6.18a) f (6.18b) M + (U/iw) f(I) M - 35 (6.18c) 53 I (6.18d) -(U/iw)2 fI) These equations are 'structurally' very similar to the strip theory results for heave and pitch derived by Salvesen et al.(1970). But the coefficients f1J) in (6.18) are now three-dimensional ones which include the three-dimensional interactions, while those of the traditional theories are -89- their corresponding two-dimensional values. In the next chapter we will compare the results in (6.8)-(6.11) with those of the existing theories. -90- VII. DISCUSSIONS AND CONCLUSIONS A linear theory has been presented for the heave and pitch motions of a slender ship, moving with forward speed in calm water. Using the complete linearized free-surface condition in the formulation of the inner problem, the velocity potential includes the forward-speed effects in a physically meaningful manner. Three-dimensional interactions are also incorporated through the appropriate outer matching. The resulting slender-body formulation is quite different from the existing theories of the slender ship, and the new results are valid for more general frequencies of oscillation. Furthermore, the present approach has provided a quite promising slender-body formulation for the steady forward motion problemwhich is simply a limiting case of the more general problem being examined. The resulting expressions for the added mass and damping coefficients are 'structually' very similar to those of the existing strip theories even though the origin of the corresponding physical quantities is quite different. In the process of deriving the present formulation we introduce a forward-speed transformation by which we can show very clearly how the inner problem reduces to an initial-boundary value problem. From a physical standpoint we can interpret the present work as a transient formulation -91- of the time-harmonic motion problem. With the restriction that the unsteady motions of the ship are sinusoidal in time, the unsteady components of the velocity potential can be expressed as time-harmonic solutions. Due to the "transient" nature of the free surface condition in the forward-speed case, however, the resulting spacedependent velocity potential, which is obtained by factorizing out the time-dependent component of the unsteady velocity potential, has a "transient" character along the ate. x coordin- These transient or memory effects are adequately considered in this work. Before making any further comments on the present work, we shall describe briefly other slenderbody approaches together with their corresponding results. Strip Theory of Salvesen et al. (1970) : This formulation assumes that there are no longitudinal interactions regardless of frequency. This assumption is generally valid for high frequencies of oscillation. The added mass and damping coefficients of the ship are given by integrating corresponding sectional coefficients with an appropriate weighting factor, a333+ 3 3 + b 333/w a + b5 3 i viz. () (7.la) (0 . (U/iw) f (0 (7.lb) 3 33-/f -92- a 3 5 + h35) + (U/i) a5 5 + b 5 55 5 f (7.lc) (U/i) 2 (O) =f 55"w 55 (7.ld) 33 where f dx[A 3 3 (x) + B 3 3 (x)/iwJyi (x) , i=3,5 (7.2) 1=3,5 1 i i=j=3 -x i or j=5 x2 Here A33 and (7.3) i=j=5 B33 represent the sectional added mass and damping coefficients respectively, and the superscript (o) is used in (7.1) to denote that the corresponding quantities are zero speed results. Rational Strip Theory of Ogilvie and Tuck (1969): This formulation uses a systematic perturbation analysis under the assumption that the frequency of oscillation S= O(C-1/2) and the forward speed U=0(1). By consistently retaining the higher order terms of relative magnitude e 1/2 this formulation provides the results, a 3s3 +33/w + b 3 3 /io a 3 + b5 3 /iW = = f 3(3 = f 0 (7.4a) Ui)f(0) l(2pwg) s (7.4b) -93- a 3 5 + b3 () +(U/iW)f (-(2ipwU/g) a5 5 +55b 555/iw 5/ where f !0? ds 3 (7. 4c) f555 (7.4d) are defined in (7.2)-(7.3). Here the integral over the free surface in (7.4b)-(7.4c) has a special interpretation which was described in more detail in Ogilvie and Tuck(1969). This integral actually results from the inclu- sion of the forward-speed effects on the free surface condition of the inner problem. Since the forward-speed effects are accounted for only as higher-order corrections, the term proportional to has been discarded in (7.4d). (U/W) 2 Comparative merits of this formulation to the strip theory of Salvesen et al. (1970) have not been resolved yet, although Faltinsen(1974) has shown that the comparison with experiments is improved by using the Ogilvie-Tuck cross-coupling coefficients. Unified Slender-Body Theory of Newman(1978): In this formulation the velocity potential for the inner problem is represented as the superposition of a particular solution similar to that of the high-frequency strip theory, plus a homogeneous solution which accounts for interactions along the length in an analogous manner to the low-frequency ordinary slender-body theory. The -94- resulting expressions are, therefore, valid for more general frequencies of oscillation. The homogeneous solution is constructed such that [ ts) 1 Dhomo = C(x)D where the superscript (7.4) (s) is used to denote that the corresponding quantity is a strip theory solution, and C(x) represents the longitudinal interaction function to be determined from outer matching. The resulting expressions for the added mass and damping coefficients take the forms, a = f.f(O) + b. ./i () (2) + f.. (1) + f.. + f.. (7.5) where s (7.6b) f~o n = (s) ) ( 5 = (p(-fm. -f~ C (x)[n f ()= p s (7.6c) (U/iw)m 1](s)+ (s))ds, i=3,5 j=3, 5 (7.6d) -95(S) where J and theory ;(s) are the corresponding strip solutions which satig fy the hull boundary conditions (6.6a) and (6.6b) respectively. is of interest to compare the results in (6.8)-(6.11) It with those of the foregoing existing theories. If the gradients of the steady-state disturbance velocity field are neglected, the resulting simplified expressions in (6.18) will be 'structurally' very similar to the corresponding expressions of Salvesen et al. (1970). f!) in (6.18) But the coefficients are now three-dimensional ones which do include the effect of longitudinal interactions, while the corresponding strip theory coefficients .f! do not. At high frequencies of oscillation the longitudinal interaction function (5.79). F(x,y,z) vanishes in accordance with In such a case, longitudinal interactions still occur via the free surface condition and are of a downstream type, i.e., only the upstream sections have effects on a given section. This is precisely the approach adopted by Yeung and Kim(1981), who in fact devoted considerable effort to give a physical interpretation of the formulation by using a fixed-frame of reference. To a certain extent, the strip theory formulation of Salvesen et al. (1970) represents a further approximation of Yeung and Kim(1981), since neither upstream nor downstream influence is accounted for in constructing the local solution at a given ship -96- section. it is important.to note that forward-speed effects on the free surface condition are not accounted for in the derivation of the strip-theory potentials whereas it is here. The rational strip theory of Ogilvie and Tuck(1969) do include the forward-speed effects in a more systematic way than the present analysis. Due to the high-frequency assumption which is made in the systematic derivation of the rational strip theory, however, the solutions in (7.4) do not have any longitudinal interactions either. Furthermore, the term proportional to (U/w)2 has been discarded as being higher order, while the other theories including the present approach keep this. The present formulation and the unified slender-body theory formulation have many common viewpoints. The unified slender-body theory includes the longitudinal interactions through the construction of a general solution for the inner problem based on the two-dimensional time-harmonic solution. The present formulation includes such interactions through the construction of a generalized Green function for the inner problem. Thus both theories are expected to be valid for more general frequencies. However, the general solution of the unified slender-body theory is so constructed that it satisfies the two-dimensional time-harmonic free surface condition, with which no direct forward-speed effects on -97- the free surface condition can be accounted for. In the present formulation the complete linearized free-surface condition is used for the inner problem so that the forwardspeed effects can be included in both body and free surface conditions. It is also worthwhile to note that the general solution of the unified slender-body theory is not matchable with the outer solution unless some further approximations of the inner expansion of the outer solution are made. More specifically, it is assumed in the unified slender-body theory that the longitudinal interactions are simply some modulations of the two-dimensional time-harmonic solution However, the inner expansion of the along the x-axis. outer Green function clearly shows that this is not the case, and that the longitudinal interactions of the unified slender-body theory are simply approximations of the real interactions. The principal complication of the present approach is that the factor m3 must be determined, and that the solution involves the inversion of a big matrix and lengthy computations for the kernel even though it is greatly simplified compared with that of the fully three-dimensional approach. From a computational standpoint the present formulation is not expected to be more economical than the existing theories. However, the present study has provided -98- a new insight on the slender-body formulation and would likely lead to the reasonable quantitative and qualitative predictions of ship motions and wave resistance in the future. -99- REFERENCES "Handbook of (1964). Abramowitz, M., and Stegun, I., eds. Mathematical Functions." 'U.S. Gov. Print. Off., Washington, D.C. (1977). Computations of three-dimensional ship Chang, M.-S. Proc. Int..Conf. Numer. motions with forward speed. Ship Hydrodyn., 2nd, 124-135. Univ. of California, Berkeley. (1976). Free-surface effects for yawed Chapman, R. B. surface-piercing plate. J. Ship Res., 20, 125-136. (1977). Survey of numerical solutions for Chapman, R. B. Proc. Int. Conf. Numer. Ship free-surface problems. Hydrodyn., 2nd, 5-16. Univ. of California, Berkeley. (1965). "Asymptotic Expansions." Cambridge Copson, E. T. Univ. Press, Cambridge. (1956). The wave resistance of a floating Cummins, W. E. slender body. Ph.D. Thesis, American University, Washington, D.C. (1975). Potential flow near to a fine ship's bow. Daoud, N. Rept. No. 177. Dept. Nav. Archit. Mar. Eng., University of Michigan, Ann Arbor. (1974). A numerical evaluation of the OgilvieFaltinsen, 0. Tuck formulas for added mass and damping coefficients. J. Ship Res. 18, 73-85. (1981). Bow flow and added resistance of Faltinsen, 0. slender ships at high Froude number and low wave lengths. (to be published). Froude, W. (1861). On the rolling of ships. Inst. Nav. Archit., Trans. 2, 180-229. (1967). Analysis of the Gerritsma, J., and Beukelman, W. modified strip theory for the calculation of ship motions Int. Shipbuild. Prog. 14: (156), and wave bending moments. 319-337. The hydrodynamic theory of ship (1946a). Haskind, M. D. Prikl. Mat. Mekh. oscillations in rolling and pitching. (Engl. transl., Tech. Res-. Bull. No. 1-12, 10, 33-66. Soc. Nav. Archit. Mar. Eng., New York, 1953.) pp. 3-43. -100- Haskind, M. D. *(1946b.). The oscillation of a ship in still water. Izv. Akad. Nauk. SSSR, Otd. Tekh. -.Nauk 1, 23-34. (Engl. transl., Tech. Res. Bull. No. 1-12, pp. 45-60. Soc. Nav. Archit. Mar. Eng., New York, 1953.) Havelock, T. H. (1958). The effect of speed of advance upon the damping of heave and pitch. Inst. Nav. Archit., Trans. 100, 131-135. Inglis, R. B. (1980). A three dimensional analysis of the motion of a rigid ship in waves. Ph.D. Thesis, University of London. Korvin-Kroukovsky, B. V., and Jacobs, W. R. (1957). Pitching and heaving motions of a ship in regular waves. Soc. Nav. Archit. Mar. Eng., Trans. 65, 590-632. Kriloff, A. (1896). A new theory of the pitching motion of ships, and of the stresses produced by this motion. Inst. Nav. Archit., Trans. 37, 326-368. Lighthill, M. J. (1958). "An introduction to Fourier analysis and generalized functions. " Cambridge Univ. Press, Cambridge. Maruo, H. (1982). New approach to the theory of slender ships with forward velocity. Bull. Faculty Eng., Yokohama National University 31, Yokohama, Japan. Mays, J. H. (1978). Wave radiation and diffraction by a floating slender body. Ph.D. Thesis, Massachusetts Institute of Technology, Cambridge, Massachusetts. Michell, J. H. (1898). The wave resistance of a ship. Philos. Mag. [5] 45, 106-123. Newman, J. N. (1961). A linearized theory for the motion of a thin ship in regular waves. J. Ship Res. 3:(1), 1-19. Newman, J. N., and Tuck, E. 0. (1964). Current progress in the slender-body theory of. ship motions. Proc.- Symr>. Nav. Hydrodyn., 5th ACR-112, 129-167. Off. Nav. Res., Washington, D.C. Newman, J. N. (1970). Application of slender-body theory in ship hydrodynamics. Ann. Rev. Fluid Mech. 2, 67-94. -101- (1978). The theory of ship motions. Newman, J. N. Appl. Mech. 18,. 221-283. Adv. (1980). The unified Newman, J. N., and Sclavounos, P. D. Proc. Symp. Nav. Hydrodyn., theory of ship motions. 13th, Tokyo, Japan. (1969). A rational strip Ogilvie, T. F., and Tuck, E. 0. theory for ship motions. Part 1, Rept. No. 013. Dept. Nav. Archit. Mar. Eng., University of Michigan, Ann Arbor. (1977). Singular-perturbation problems in Ogilvie, T. F. ship hydrodynamics. Adv. Appl. Mech. 17, 91-188. Peters, A. S., and Stoker, J. J. (1957). The motion of a Commun. ship, as a floating rigid body in a seaway. Pure Appl. Math. 10, 399-490. (1970). Ship Salvesen, N., Tuck, E. 0., and Faltinsen, 0. Soc. Nav. Archit. Mar. Eng., motions and sea loads. Trans. 78, 250-287. Sclavounos, P. D. (1981). On the diffraction of free surface Ph.D. Thesis, Massachusetts waves by a slender ship. Institute of. Technology, Cambridge, Massachusetts. (1967). Computation of pitch and heaving Smith, W. E. Int. Shipbuild. motions for arbitrary ship forms. Prog. 14: (155), (1953). On the motion of St. Denis, M., and Pierson, W. J. ships in confused seas. Soc. Nav. Archit. Mar. Eng., Trans. 61, 280-354. (1962). The coupled damping Timman, R., and Newman, J. N. coefficients of symmetric ships. J. Ship Res. 5:(4), 34-55. (1963). The steady motion of a slender ship. Tuck, E. 0. Ph.D. Thesis, University of Cambridge. (1962).. - Slender oscillating ships at zero forUrsell, F. ward speed. J. Fluid Mech. 19,' 496-516. Surface waves. (1960). Wehausen, J. V., and Laitone, E. V. In "Handbuch der Physik" (S. Flugge, ed. ), Vol. 9, pp. 446-778. Springer-Verlag, Berlin and New York. -102- (1973). The wave resistance of ships. Wehausen, J. V. Adv. Appl. Mech. 13, 9-3-245. (1981). Radiation forces on Yeung, R. W. and Kim, S. H. Proc. Int. Conf. Numer. ships with forward speed. Ship Hydrodyn., 3rd, (in press). -103- APPENDIX 1: 0 If Derivation of Modified Green's Theorem are any differentiable scalar 0 and functions of position, the following is true - 0 a 0 11 2 -2. JJI I .eyIdV V - - 0 .1JdS (Al.1) S Here the operators in (Al.1) are defined by + -) +L ) i(3x V Vnijn-3 where + + + k (Al.2) ) + a ay n3 + n (Al.3) 4 3aza an n- _= n lax + -n) 23y. + n an a = n(a ne~V -l 2aY + n 3az 1ax + a) + na to the surface Proof. theorem, S and (l5 (l6 represents the unit vector normal (n1 , n 2 , n 3 ) n (Al.4) a is an arbitrary constant. Making use of Green's theorem and Stokes's (Al.1) can be proved as follows: -104- a I, S 11 an+ iiD- 1-dS I ~ go, i ff go, an + 2a(yoi n 1JdS S =1~ f 1~ =1 fr I [ i V20 - Oilv 20) + 2ca ( 0 )tdV ( V (1 (V +ct2 '1 +2ct V a ax + 0o 2il - =111 *2 2 a 20 1 ) dV - 20 dV (Al. 7) V If Jf - V +11 S ian + V = 0 1~-d in V, there follows from (Al. 1) (Al.8) -105- Derivation of the Pseudo-time-dependent APPENDIX 2: Green Function We construct a Green's function which satisfies the following set of equations, 2 2 + -z-2)G (y,z,n,;x,() = 6 (x- ) 6 (y-n)S(z-c) ( ay az (U ax VG 0 a G0 + g a ) - = on z=O 0 , x+0+ (A2.2) at z= -c 0 + (A2.1) (A2.3) where the initial conditions of vanishing motion upstream are incorporated introducing an arbitrary Rayleigh viscosity parameter V such that aG G0 for x > t, z=0 (A2.4) It is of some interest to notice the similarity between G and the time-dependent Green function corresponding to an unsteady two-dimensional free surface flow. we replace x Green function. by Ut, G0 Actually, if will reduce to the time-dependent The solution for G0 can be found in Wehausen and Laitone(1960) in the form, G0 (y,z,n,ic;x,t) = 6(x-C)ZnR/R1 + H (y,z,n,C;xE) H = - u(E-x) Jf (A2.5) e(Z+c) cost(y-n )sin / 0 ( -x) 0 (A2.6) -106- where R = (y-n) 2 + (z-)2 1 1 / 2 R (y-n) 2 + (z+0)2 &= 1/2 (A2.8) (A2.9) 2/g are the Dirac delta function and the u(g-x) and 6(x-) (A2.7) , Heaviside step function respectively. Here the first term in (A2.5) represents an impulse at the point when E=x, while Q = (n,) describes the disturbance at H due to this impulse as it propagates away from course of pseudo-time Q (y,z) P in the x/U. the flow direction is reversed, the corresponding If conjugate Green function will satisfy the following set of equations, ( + ay az ) (c) (y,z,n,;x,) = (x- (y- (z(A2.10) 3G (c) 0 (ax + P)2G(c) VG 0(c) 0 + g a 0Gc = 0 on z=0, (+0+ A2.ll) at z= (A2.12) -- where the initial conditions of vanishing motion are reserved such that 9G(c) G0 Q -107is easily derived in the form, G(c) 0 The solution for G(c) (y,z,n,?;x,g) = 6(x- )nR/R 1 + H c) (y,z,r,C;x,5) . 0( (A2.l14) where H (c)= 00 - dZ u(x-) (Z+C) cost(y-n)sinF7-C) o V/a0 %J (A2.15) It is instructive to derive the Green function which satisfies the following set of equations, (-L 2 2 + 1y )G(yz,r, ;x,) 77 - 2 + (A2.16) = 6(x-)6(y-n)6(z-) G on z=O, P+0+ VG + 0 at z= -= (A2.17) (A2.18) If we set is to satisfy the following set of equations, H then (A2.19) , G = 6(x-E)ZnR/R1 + H(y,z,n,i;x,) a2 2 ( + ay (iW - 3 )H(y,z,n,4;x,) = 0 az UaAx + -P)2H + gz -6 (X-0) g[nR/R a2z. (A2.20) ] on z=0, -yO+ (A2. 21) VH +- 0 at z= -m (A. 22) -108- We define the double Fourier transform of respect to H ** and x y H with as + ity (Z,z,ri,iyk,5) dy H(y,z,n,;x,)e dx (A2.23) The double Fourier transform of (A2.20)- (A2.22) gives d2 H t 2 H ** = 0 (A2.24) dz ** dH (c+Uk-iu2 )H = Ij I c+ikE+tn 2wg on z=0, 1 +0+ (A2. 25) ** dli at + 0 z = -w (A2. 26) From (A2.24) -(A2.26) H we obtain I (z+c)+ik +itn] (Z,z,n,?;k,&) = (-2r) exp[ It I I- (w+Uk-ipi) 2/g (A2.27) and )0 = 2 ,C;x, 4 7r l_ im - df U .+0+ dt 2 -- 0 (t ,z, T,.c; k, e-ikx-ity 0 dk exp [Z (z+c) -ik (x- ) -iZ (y-n ) I Z-(w+Uk-iu) 2 iw a T dk H ) H(y,z, ** o 1 0 dZ x o Z/a0 et(z+)cost(y-n)sin Z (E-x) 0 A2..28) -109- Here the contour integral in the complex k-plane, which is shown in Figure A2-l, is used to obtain (A2.28). Im k when x- E< 0 Re k U U 0 when when Figure A2-l. x-E> x-E>00 Integral paths for the k-integral We consider a Green function G(c) which satisfies a , reverse-time and reverse-flow problem of (A2.16)-(A2.18) a2 2 + ay )G(c) (y,z,n, ;x,) = a z2 (iW-Ua - VG (c )2 G (c) + (c) = 0 (x-E) 6 (y-n) 6 (z-c) (A2.29) on z=0,; y+40+ (A2.30) at z= -w (A2. 31) -110- If we set G(c) = 6(x-C)ZnR/R1 + H(c) (y,z,,;x,) , (A2.32) the similar analysis as above will provide the solution for H(c) in the form, H (c) = (X-C) 00t(+ -2-u_(x-o)e U ds ao (X-) cost(y-n)sin a fo /1 (A2.33) It is worthwhile to note that G(y,=z,n,;x,C) = eU G (A2.34) iuJ G(c) (y,z,n,4;x,C) = eU G(c) 0 (A2.35) -111- APPENDIX 3: Alternative Expressions and Asymptotic Approximations of the Pseudo-time-dependent Green Function The Fourier transforms of the homogeneous terms of the pseudo-time-dependent Green functions derived in Appendix 2 take the forms, H 0 (y,z;k) - = 1-+ +O Zim H(c) * (y,z;k)= 0 .l et(e Jo ez r a+0 e 'ly) ty+ -(Uk-ip) 2 ieZy+ e ZIyI) Z-(Uk+iu ) /g (A3. lb) * = H H (y,z;k) =- (y,z;k) fd '+ + (A3.la) e 0 (eZItY+ 2 t-(w+Uk-ip) /g . * = H~(y,z;k+w/U) H c)*(y,z;k)= H 0(c) *-* (y,z;k+w/U) = H (y,z;k) (A3.lc) (A3.ld) where the overhead bar denotes the complex conjugate of the expression involved. ; = z - ilyl , , , = z + ilyl Making a change of variables, (A3.2a) WA. 2b) -112- to the form, (A3.la) H 0 (y,z;k) = [eKOC - Ko( 0 (I ds + eK dse 1 ) we may reduce (A3.3) where K = U 2 k 2 /g. Here the integral paths are sketched easily shown that the integral over as R+ e, and that (A3.3) H (y,z;k) = - [eoc It can be k < 0. in Figures (A3-1)-(A3-2) according as r = R0 e vanishes reduces to ds K0 ds + e oC - { K0 e 0 + 27ri sgn(k) (A3. 4) eK o where the upper and lower terms in brackets are applicable according as k. 0. The equation (A3.4) can be expressed in terms of the exponential integral function, CO E l(z) = fzds -s s , (jargzf defined in Abramowitz and Stegun(1964). < 7r) (A3.5) -113- K K 0 pnr 0 s-plane .- plane j 2 r K0 Figure A3-1 Integral paths for k>O -fexp (K0c) (y, z;k)-e- 0 E (0c)+e CE -2Re {e'< (z+2.y)E 1(K (KO) +27Ti sgn(k) exp(K 0 0Z+i K y) }+27ri sgn(k) x(0 exp (K0 c (A3.6) The asymptotic properties of (A3.6) can be obtained from the corresponding approximations of the exponential integral. For small values of K the sourcelike logarithmic singularity is displayed in the approximation -114- K0C 0 snr s-plane t-plane I2r 0 t Figure A3-2 Integral paths for k<0 H 0 (y,z;k) = 2(1+K 0 z) [ZnK 0 R 0 + y + wi sgn(k) + Here 0 (K 2 R 2 (A3. 7) 0 0 y = 0.577... is Euler's constant, and polar coordinates defined such that z = -R 0cosG.. Koz + K 0 ye] For large values of (R ,9) y = Rosine K0 R are and (A3. 6) is approximately exp iK0 (z-i jyf) I + 0(-) H (y,z;k) = 2ni sgn(k) exptrO (z+iIy 00 (A3. 8) The corresponding expressions for H. can be obtAined froM (A3,61-CA3,81 (c* Q * -115- rI~ and conbined with (A3.1) as exp H 0* (y,z;kY=-c2 Re{eg Z+0y) E C,0 z+iK 0 (K 0) }y1-21"'1 sgn (k) exp (K )} (A3. 9) exp (Kr) H (y,z;k) =-2 Re{e K(Z+iy) E (KZ+iKy) }+2wi sgn (w+Uk) ) exp (Oc (A3.10) exp (Kr) (y,z;k)=-2 Re{eK (z+iy)E (KZ+iKy)1-2ti sgn-(w+Uk) (A3.ll) where K = (w+Uk) 2/g exp (KcI -116- APPENDIX 4: h (xyfz) The Inner Expansion of In this appendix we shall derive the inner expansion of the modified Green function defined in (3.22). h in (3.25) . with the Fourier transform of 2 i-4"+fd 2 11exp[(k-w/U) (k-w/U) 2 +Z /2-(Uk-ii) 2g = h*(yz;k) 0 ykO =0 d We start 21 2z cosy (A4 .1) Making a change of variable, (A4.2) t = jk-w/Ulsinh v we may reduce (A4.1) to the form, coshx +imf0dv h (y,z-k) = -4 0 Vt).+ o coshv -(Uk-i ) 2/g |k-w/U| x exp(zIk-w/Ulcosh v)cos(yjk-w/Ujsinh v) (A4.3) Ursell(1962) derived a series representation of the expression similar to that of (A4.3). Ursell's analysis to (A4.3) provides The application of -117- coshm 0 h 0 (y,z;k) = 4 [1 0 (k0 R ) + 2 M m=. (-1) mm (k R ) cosme{ 1 (X i sgn(k) I + a) - 4 4K (kR) sinhma cotha + 8m ~a 1 m v (k 0R 0)cos v e)] v =m sin ma* cota* (A4. 4) In' K where are the modified Bessel functions defined in Abramowitz and Stegun(1964), y = R sinO , z = -R 0 cose and the upper and lower terms in curly brackets are applicable cosh a > K /k a* 0Cos where . according as K0 U 2 k 2 /g <1 and I (A4. 5) k0 Ik-w/U| Using the ascending series expansions for and neglecting terms of In K0 0 (k2R2), we obtain -l 0~ h* y~zk) (lK Z)4 &iisnkcohlI/,O i sgn (k)+cosh h *(y,z;k)=4(1+K -W + cos (K /o) 11-k /K 21/2 (K 0 /k) + 4[Zn(k 0 R 0 /2)+ y] + 4K 0 z[Zn(k0 R0 /2) + 4K ye + O(k2R2 where and , Kc2R) y = 0.577... is Euler's constant. + y-1] (A4.6) -118- The Fourier transform of a pseudo-time-dependent Green ing asymptotic expansion for small values of KOR , function has .been derived in Appendix 3 with the correspond- of the form H 0 (y,z;k) = 2(1+K 0 z) [n K R + y + 7i sgn(k) - + z (A447) 2 O(K 2R 0 0 + K 0 yel and 47ri sgn(k) H where (1+K denotes the complex conjugate of the expansion of h (yz;k) = 2 H (y,z;k) Thus, KOR0 and 0 0 - . H and H can be expressed in terms of H k0 R0 for small values of h (A4.8) z) 4 sgn(k) (1+K0 z)f 0 (k,K ) H (y,z;k) - H (y,z;k) 0 0 [H- (f /2ini) (H0 - H0 ) ] (A4.9) where 0 -sgn (k) n) sgn(k)+cosh -7r+cos 1 (K (i,/k 0 /k) + sgn(k)tn(2K /k0 ) + in with an error factor 1 + O(k2R2 , K2R ). 21-k 2/K 2-/ ) *i )= f 0 (k,r )= ( (A4.10) -119- If we define = -2Tri f 0 (k,K )[H (y, z;k) * -2 h0 0 (y,z;k) sgn(k)f (k,K )e0zcos - 0 (y, z;k)] Koy (A4.11) , S* F in (A4.9) reduces to * 22 h (y, z;k) = 2[H (y, z;k) + F 0 (y, z;k) ] + 0 (k2R 0 0 2 2 0 0 (A4.12) Therefore, the inverse Fourier transform of (A4.12) yields an inner expansion of ho (x,y,z) as h (X,y Z) = 2[H (y,z;x) + F (x,y,z)J 1+ 0 1 + (m 22 2 e , with an error factor 0 -120- Reduction of the Kernel APPENDIX 5: In this appendix, we shall make a reduction of the kernel defined in (3.31) following a similar analysis as Sclavounos(1981) Using (3.27)-(3.29) did. and the results of Appendix 3 followed by a coordinate transformation, k=k6 , 6= U2/g, we obtain F (x,y,z) = J dk e -ikx/6D -- (k)ek2Z/6 cos(k 2 y/6) (A5.1) + in sgn(k) - i sgn(k)+cosh 1 (k2 / |k-Ti) %. -7+cos xl - x 1(k 2/|jk-Tj ) D (k) = Zn(2k 2 /jk-Tj) + where (k--t) 2/k41-1/ 2 (A5.2) where the upper or lower expression in bracket is applicable according as k2 /jk-T Since the kernel 1 , D 0 (k) (A5.3) T= WU/g defined in (A5.2) similar to that of Sclavounos(1981), analysis for the present work. is structurally we will repeat his -121- cosh- (x) = Zn[x + (x2 _l) 1/2 , We use the definition and start by analyzing the function x 5 1 (1-z f(z) 1/2 2 - 2 -1/2 ) n[z+(z -1) ] on the complex z-plane, where Znz and (A5.4) z1 / 2 are said to be the principal branches of the corresponding multivalued functions with -7 < arg (z) < i. The function f (z) is analytic on the finite complex z-plane with a branch cut along (--,-l]. The values of shown in Figure (A5-l), where 0 <lxi < f(z) along the x-axis are 0< cos 1 |x 1, and the previous definition of understood for < 7r/2 cosh LFr is l- ~v-- -II- /1-2 0 Figure A5-1 1 x1 JxJ> 1. Cx, ~,r for 1. Icsxiii--76 The values of f(z) along the x-axis COSh -122- f(k 2/ (k-r)] The analytical structure of the function on the complex k-plane can be determined if we first consider the mapping, z = x + iy = k 2 /(k-T) T>O , (A5.5) k Solving the resulting quadratic equation for gives [z-(z 2 -4tz) 1/2] k = u + iv = (A5.6) Here the branch of the square root is chosen so that the The complex z-plane is mapped onto the lower k-plane. branch point at equation z=-l corresponds to the roots of the k2/(k-t)=-l, defined by , -[l+(1+4t) 1 /2 ] , (A5.7) z =.-liO The corresponding roots for the which are real for all T. singular point at are defined, according as z=1 4T 5 1, by 1 ;(1-4) 1 /2] , z = l iO , 4t< 1 (A5.8) P3 , 4 [1-i (4T-1) I , z = 1 iO , 4T> 1 It is not difficult to show that the mapping function z (k) is a singlevalued analytic function of lower k-plane. The inverse mapping function k in the k(z) is also -123- 7 .,P t\ A c A Ir l E D . T <1/4 & gKDu N T L 4%+ K 0 ------ A 1 EB X. -- PB F V. L/4 Figures Al -2, A5-3 Mappings between z-plane and k-plane -124- a singlevalued analytic function in the finite z-plane except branch cuts along (-w,O] and t4T,<o). Thus the mappings between the corresponding domains in the complex k-plane -and z-plane are one-to-one. The previous considera- tions are schematically shown in Figures (A5-2) and (A5-3) according as 4T 1. Next we define R(k) = [l-(k-r) 2 /k 2 /(k-r)+[k 4 /(k-t) 2_ 1 -1/2Ztn{k 1/2 (A5.9) The function Dku and R(k) is analytic in each of the domains DkL, its values being determined from the corresponding values of where f(z) is analytic. f(z) in Dzu and DzL Since the values of respectively, along R(k) the curves CD and IJ correspond, in the limit, to the values of f(z) along the segment (0,4T) on the x-axis where is analytic, R(k) f(z) can be analytically continued across the dotted semicircle from the domain Dku to DkL and vice versa, using definition (A5.9) uniformly in the lower half k-plane. Furthermore, R(k) takes real values along the real axis except the branch cuts, one located along the segment (-o ,plJ and the other along be analytically continued for Im k > 0+ [p 2 ,T], and it can using Schwartz's -125- reflection principle, R (k) = (A5.10) where the overhead bar denotes the complex conjugate of the expression involved. Consequently R(k) is analytic in the finite k-plane apart from the two branch cuts. Now we define W(k) = Zn[2k 2/(k-r). Since the function - R(k) (A5.ll) Zn[2k 2 /(k-T)] is analytic in the finite k-plane except for a branch cut along (-ct], W(k) is an analytic function in the finite k-plane with a branch cut along (-c, T] . Thus the values of D (k) along the real k-axis can be expressed in terms of the values of W(k iO), where the upper or lower sign is applicable according as x 5 0 Figures in (A5.1). Using (A5.4), (A5.8), (A5.10),, and (A5-1) - (A5-3) , we obtain Do (k) = W(k iO) + iri sgn(k) 7ri[u(T-a)-2u(-a)] + (|1-(k-T)2 /k 4 -1/ 2 )g (k) where u(k) (A5.12) is the Heaviside step function and g+(k) = 2ni , g+(k) = 0 , k < p1 (A5.13a) p. < k < p2 (A5.13b) -- < -126- g_ (k)= 0 -00 < g (k) = -'ri T g (k) = -ni k. < (A5.13d) * k < p3 , , (A5.13c) < t p3 < k p4 < - < k k T. < 1/4 (A5.13e) < p4 T < 1/4 (A5.13f) < T < 1/4 (A5.13g) T > (A5.13h) 0w < co , g (k) = -ni , < k = , -27i p2 g+(k) , 1/4 It can be seen that W(k) = 0(nlkl) ) W(k) = 0(tnlk|/k 2 as k 0 (A5.14a) as k 00 (A5.14b) Applying Jordan's lemma, we obtain dk -ikX/6 W(k iO) = 0 , x >0 (A5.15) -00 Using (A5.15) and the convolution theorem, we can also show that f 00 2 Ok eik/ W(k iO)e kz6cos(k 2Y/6) = 0, x > 0 (A5.16) Combining (A5.1), (A5.12), and (A5.13) we obtain IF~ F(1) (X ,Z 0 F (x,y,z) = 1 F (2) + F (2) (x,y,z) x < 0 (x,y,z) (x,y,z) (A5. 17) , x > 0 -127- F 0 (x,y, Z) 2i P di e-ikx/6ek2Z/6 cos(k 2y/6) {1-[l-(k-t) 2/k 4 J + :2 e-ikx/5 ek 2 z/6 cos (k2 y16) - 1 2 4 dk e-ikx/6ek 2 z/6 cos (k2 y/ )'U1- [ 1- (k-r) /k ] -1/2k 2i 6 f P (A5.18) F (2) (XIyTrZ) + f43 i (P f 2 4 &c e -ikx/Sek Z/ cos (k2 y/6)'{1-[l-(k-T) 2 /k 4 ]-1/2} 2 dk e -ikx/6 ek z/6 cos(k 2y/c)x 4 J P3 - x {1-i[ (k-t) 2/k 4-1]-1/2 . < 1/4 (A5.19) and F 0(2)xyz) dk e-ikx/e k2Z/6 cos (k2Y/S) =-o x {1- [l-(k--c) 2 /k 4 1- 1/2 T where 6 = U2/g. > 1/4 Th a corresponding expressions for (A5.20) F (x,y,z) can be obtained from (A5.17)-(A5.20) by using the relations, F(xyz) = e x/' F (x,y,z) F (y,z;k)= F 0 (y,z;k+w/U) (A5.21a) (A5.21b) -128- Inner Expansion of the Outer Solution APPENDIX 6: 4(x,y,z) In this appendix we shall derive the inner expansion of the outer solution $ The outer solution 4 ur$(p) for the purpose of comparison. 0 has been derived in (3.10) as ds(Q) [(Q) = Q - C g dn[2iwUt r[(-+ ) + h(pQ)j a Q (Q) + U2 W n - (A6 .1) ")I)h(p,Q) C The Fourier transform of 4m(m h+0 h(yk)= h 0 is obtained from (3.6) as 2 + 21/2z cos ty exp[k [k 2+2212(w+Uk-iz) 2/g A6.2) Appedix4 fr te drivtionof he nne exanson * Making use of a similar analysis to what we did in Appendix 4 for the derivation of the inner expansion h,0 we will obtain ** * 2 2 2 2 h (y,z;k)= 2[H (y,z;k)+F (y,z;k)]+O(k R ,K R 2) (A6.3) where H (y,z;k)= H (y,z;k+w/U) = 2(1+KZ) [Zn KR0 + y + 'ri sgn(w+Uk)-Kz+Ky61 + 0(K2R ) (A6.4) * -129- * F (y, z;k) = F (y,Z;k+w/U) * 1 f (k,K) * -* (A6.5) f (k,i) [H (y,z;k) -H (y,z;k)] - = = -sgn(w+Uk) t sgn(w+Uk)+cosh 1 1(K/Ik|) _-1 (< I in -V + Cos I 11-k 2 K2 -1/2 (A6. 6) + sgn(w+Uk)tn(2K/fkl) + 7i Here the upper or lower term in bracket is applicable according as ic/|kI > 1 I K = (A6.7) (w+Uk) 2/g The inverse Fourier transform of (A6.3), therefore, provides (A6.8) h(x,y,z) = 2[H(y,z;k) + F(x,y,z)1 1 + O(w 4,R ) . with an error factor Thus, an application of the similar analyses in section 111.3 to (A6.1) will give rise to the inner expansion of (A6.1) as 2n$(p) = ds [$(Xn,) t ]ZnR/R - a 3Q1 B (x) ds Q[ d [2iwU ( -1 (, , ) - Q DNQ O'n,0)+U 2 9C with a resulting error factor 1+0(W4R 2, - 1 [H(PQ)+F(PQ)] )- [H (PQ)+F(PQ) I (A6.9) R0 ,R ) -L/2 f ( j ) + L/2 /2 -130- APPENDIX 7: Relationship between a Line Distribution and a Surface Distribution of the Generalized Inner Green Functions In this appendix we shall derive an effective line distribution of wave sources which produces far-field wave effects comparable to those of a surface distribution. The solution of the inner problem in (4.8)-(4.10) takes the form, 2ff(P;x) = - IG (n-&) L/2 ds -L/2 Q 8(e) - dn[2iwU0(n,;)+U 2 (u- - )]G (yzrn,;x--) (A7.1) where G ()= 6(x-E)ZnR/R1 + H(y,z,rn,C;x-E) + F(y,z,n,r;x-E) (A7.2) Here H and F are defined in Chapter III. The Fourier transform of (A7.1) gives L/2 -L/2 ds [ B( G x - ;) (, a-x NQ aN Q (ylzjn,C;k)eik Cdn[2iwU((n,0;E)+U2 (x C Y a 0;k - (y, z;k) ) 2* aE k (A7.3) -131- (y,z,n, C;k) = e K G (W (;C+in) G (1)* (y,z,O,O;k) + 01( ) From the results of Appendix 3 we know that (A7.4) where R0 = (y2+z2)1/2 and K = (w+Uk) 2 /g The velocity potential due to a free-surface line distribution of wave sources can be expressed as t(y,z;x) Here a(x) =f da(&)G(i) (y,z,,O;x-&), R>>1 is the effective source strength. transform of (A7.5) (A7.5) The Fourier gives $*(y,z;k) = a (k)G ('(y,z,0,0;k) (A7.6) Far from the ship in the inner region, therefore, the effective source strength of inner variables. 2Tra -1 can be expressed in terms From (A7.3)-(A7.6) we obtain ai,;l) K(c+in)+ik& ds, [ (k) =(L/2d -L/2 a(x) 8(&) Q Cdn [2iUt (n , 0 ; &) +U2 (t Q ol ign+ik t (A7.7) Thus, we can immediately write down the effective source strength as -132- f -L/2 B(E) ds [4( Q [2iwo(n,0; -T -O; - NQ ) fL/2 Jd 2ra (x) = DQ (n,C;x-) )+U2 Cd -0 < X < 0 (A7.8) where M(-n,, ;x-V 7~1 dk e-ikx e (C+in)+ik -47rU2 (+in) 9 F2 I -1/2 I4U ( +in) U (A7.9) For the special case of zeror forward-speed, U=O, we recast (A7.9) as 7(a/1/2 exp[-a (x-+a2 ) 2]exp(aa 2 ) M(n,c;X-) (A7.10) where a2 4U (+in) 2iwU (c+in) (A7.llb) . 1 (A7.lla) - = a From the definition of a generalized function Lighthill(1958), we know that 6 (x) (see j -133- tim Thus, *dx e-ax 2 (a/Tr) 1 /2F(x) = F (0) a (x) (A7.12) simplifies to ds [I( x e ,x . L/2 (x) 21ra(x) = Ix|> L/2 0 (A7.13) where v = obtain 2 /g. (A7.13) Here (A7.10)-(A7.12) have been used to from (A7.8). From (A7.8) and (A7.13) we can draw a conclusion that a free-surface line distribution of sources, in general, should be extended over an infinite length in order to produce far-field wave effects comparable to those of a surface distribution, while the line, distribution over a ship length sufficies for the case of zero forward-speed.