You are on page 1of 8

2.4.

Acoustic excitation of structures


Marold Moosrainer Institute of Mechanics, University of the Federal Armed Forces Munich, Neubiberg, Germany

Summary: During the launching phase of space vehicles structures are subjected to high noise levels of 140 dB and more transmitted via the surrounding air. This acoustic loading is particularly critical for lightweight structures with large surfaces, e.g. solar panels. Thus they have to be tested in reverberant chambers, where a diffuse sound field of comparable power distribution over the frequency axis is generated to approximate the launch conditions. This paper gives insight into the numerical simulation techniques of acoustic excitation. After outlining the main features of the structural (FEM) and the acoustic equations two different discretization schemes are presented for the resulting boundary integral equations governing the fluid domain. The first, based on element shape functions, leads to the indirect boundary element method (BEM), whereas the second, based on global trial and weighting functions (Ritz approach), yields a very efficient solution for rectangular plates. Due to the inherent symmetry of both formulations the coupling with the structural FEM matrices is straightforward. An additional difficulty is the stochastic character of diffuse sound fields. Applying random vibration theory, a power spectral density (PSD) description of the sound field is used to compute the PSD of the structural acceleration. Results are presented for solar generators and simplified benchmark plate structures. A discussion of the accuracy and the efficiency of the mentioned approaches concludes the paper.

Keywords: FEM, BEM, fluid-structure interaction, acoustic excitation, diffuse sound field, random vibration, plates

18. CAD-FEM Users Meeting Internationale FEM-Technologietage

20. - 22. September 2000 Graf-Zeppelin-Haus, Friedrichshafen

Introduction
[C]: Sound pressure level at 482.5 Hz [dB] 80 75 70 65 60 55 50 45 Z X Y 40

For the numerical solution of acoustic radiation problems the boundary element method (BEM) has gained more and more recognition during the last decade. Today the available hardand software allows the computation of sound spectra or directivities of musical instruments (fig. 1) or engines, just to give two examples. Both stress the point that usually the focus is put on force excited structures that radiate sound originating from their vibrating surfaces. A two-step analysis is state of the art for this purpose. First, the in vacuo vibration of the structure is computed by means of the finite element method (FEM). Then, in a second step, the sound field is computed via BEM using the surface velocities from the structural results as the acoustic boundary conditions.

Fig. 1: BEM computation of a Yet, this approach neglects the fluid-structure coupling effects guitars 3D directivity [10]. that are essential for lightweight structures or heavy fluids. For example the classical in vacuo analysis of an A-tuned kettledrum gives a frequency of 171Hz for the main tone whereas the correct value of 110 Hz can only be determined if the surrounding air as well as the air enclosed in the kettle is taken into account [10]. This fluid loading shows a strong frequency dependence and can be interpreted as an additional mass, additional stiffness (enclosed fluid volume) and additional damping (sound radiation). In contrast to pure structural problems one now has to deal with the sound pressure as an additional degree of freedom (DOF) due to the acoustic part of the problem. Now, both sets of variables (structure and fluid) have to be computed simultaneously by solving the coupled system of equations. Apart from those radiation problems assuming force or base excitation the coupled fluid-structure analysis allows for the modeling of acoustic sources. Depending on their frequency and spatial distribution they more or less excite structural vibration. This mechanism proves to be rather cumbersome to deal with analytically even for simplified configurations like plane structures inserted into a rigid wall (baffled) that suggest the use of the Rayleigh integral or Fourier techniques in the spatial domain [5] for the description of the acoustic part. Relevant examples for this kind of excitation are the numerical simulation of sensor design (microphones etc.) or the acoustic excitation of space structures during the launching phase, where critical vibration amplitudes can be observed for unfavorable load cases. The coupled FEM/BEM analysis [1]/[3] is very versatile with respect to arbitrary geometries and boundary conditions. However, some experience and skill is required from the user to get reliable results. Further on the CPU time for a coupled analysis is substantially higher compared to the uncoupled in vacuo analysis, particularly for high ratios of a characteristic structural length to the acoustic wavelength considered. Therefore an alternative approach based on admissible global trial and weighting functions (Ritz) for rectangular plates was implemented into an easy-to-use Windows program (WinFloaP) which significantly reduces the CPU time.

2
2.1

Structure
Variational formulation

A common starting point for the derivation of the equations of motion for a linear elastic structure is Hamiltons principle [6], a variational scheme where a solution in the form of the displacement distribution is found by minimizing the Lagrangian functional L, also defined as a potential s L = 0, L = s = T U + W , (2.1) where T represents the kinetic energy, U the strain energy and W the potential of the external forces, each depending on the displacement field u. The integration with respect to time has already been cancelled in eq. (2.1) because of the Fourier transformation yielding equations in the frequency domain. Therefore the terms in eq. (2.1) and the displacement u are complex from now on. The energy expressions can be found in standard textbooks (e.g. [6]) for different structural components. The sound pressure p is an external loading for the uncoupled case and has to be considered in W. 2.2 Discretization

The discretization of eq. (2.1) by means of element shape functions Ns leads to the well-known linear system of equations of the FEM

18. CAD-FEM Users Meeting Internationale FEM-Technologietage

20. - 22. September 2000 Graf-Zeppelin-Haus, Friedrichshafen

[ 2 Ms + K s ] U = Fs , (2.2) where Ms is the mass matrix, Ks the stiffness matrix, U the vector of nodal displacements and Fs the vector of external forces including the sound pressure. It is advantageous to introduce modal coordinates at this point. Firstly, modal truncation is possible for weakly damped structures in order to reduce the number of DOF and moreover the interpretation of the results will be supported. Taking U as a superposition of uncoupled in vacuo (`dry') modes j of the structure weighted by modal participation factors (or modal coordinates) j leads to the approach of modal superposition U = i i i = . (2.3) Multiplication of eq. (2.2) by the transpose of the modal matrix where the eigenvectors j are arranged in columns yields  +K  ]  . = [ 2 M (2.4) s s s
  Z s

Here,
 Z

 = M s

s

 = = diag(m j ) is the modal mass matrix, K s


T

T s

= diag(k j ) = diag( 2 j m j ) the

 = modal stiffness matrix and F s


s

s

the vector of modal forces. The modal dynamic stiffness matrix

 = diag( c ), assuming a viscous damping may be extended by a modal damping matrix C s j cr

mechanism with j as the percentage of critical damping ccr,j = 2 k j m j . Alternatively, hysteretic or structural damping is modeled by replacing the stiffness matrix in eq. (2.4) by the modified term  [1 + i diag( )] , where are the loss factors of the distinct modes. Thus, for the complete descripK s j j tion of a structures in vacuo vibration it suffices to know its modal parameters (eigenvectors, eigenfrequencies and modal damping factors). The superposition eq. (2.3) holds also for the case of a structure immersed in a fluid. Then, however, additional fluid terms have to be added (see sec. 4).

3
3.1

Acoustics
Basic equation and boundary conditions (3.1)

The governing equation of linear acoustics in the frequency domain is the Helmholtz equation 2 p + k 2 p = 0

for the sound pressure p, where k = /cf is the wavenumber. To solve this partial differential equation one has to prescribe the acoustic boundary conditions. The most important ones for an external fluid volume are the normal surface velocity vn of the radiator and the so called Sommerfeld radiation condition p + ikp = 0 , lim r (3.2) r r which states that no waves are traveling from infinity towards the radiator. The linearized momentum equation p (3.3) = i f v n n relates the normal surface velocity to the normal derivative of the pressure and therefore is used for the coupling of the fluid and the structural equations. 3.2 Boundary integral equation

The proceeding of section 2 directly leads to a FE formulation of the acoustic problem. However, the satisfaction of the Sommerfeld radiation condition is rather cumbersome for domain methods such as FEM because of the numerical reflections at the truncated volume discretization. As a first workaround a large fluid volume (several acoustic wavelengths) enclosing the radiatior is discretized and then a plane wave damper (impedance Z = f cf) is applied at this artificial boundary. An improvement of this situation is the introduction of wave-envelope elements , i.e. semi-infinite elements that satisfy the Sommerfeld radiation condition by decaying, oscillating shape functions [2]. A boundary integral equation (BIE) of the governing differential equation (3.1) is frequently preferred for exterior problems, because only the surface of the radiating body has to be discretized and the Sommerfield radiation condition is implicitly satisfied due to the use of a Greens function. Textbooks mainly deal with the direct BIE [7] where the physical quantities (pressure and velocity) themselves are involved in the integrand. It can be derived by using either Greens integral formula, a reciprocity

18. CAD-FEM Users Meeting Internationale FEM-Technologietage

20. - 22. September 2000 Graf-Zeppelin-Haus, Friedrichshafen

statement (Bettis principle) or the method of weighted residuals yielding the Kirchhoff-Helmholtz integral. However, for many cases the indirect BIE G( x , y ) p( x ) = pin ( x ) + p q G ( ) ( ) ( , ) (3.4) y y x y dAy ny
A

is superior. According to fig. 2 p(x) is the unknown pressure at an arbitrary field point x, pin(x) is an incident pressure from -ikr an acoustic source, G(x,y) = e /(4r) is the fundamental solution (Greens function), where r is the distance between the field point x and the considered surface point y. Before determining the left hand side of eq. (3.4) it is necessary to compute the unknown surface potentials

p( y ) = p( y + ) p( y ) and q ( y ) = i f (v n ( y + ) v n ( y )).
They represent the jump of pressure and the jump of pressure gradient (~vn, see eq. (3.3)) across the thin surface A. For plates, e.g., the velocities of the back and front sides are equal and therefore q = 0 in the following. The lack of physical meaning of the surface variables has lead to the denotation indirect. However, this form of the BIE is very versatile with respect to the geometry considered. The direct BIE requires a closed body (fig. 2 (left)) and is restricted to the computation of either an interior or an exterior problem. Moreover, thin structures cause numerical difficulties. The indirect BIE on the other hand can handle thin radiating bodies and even open surfaces. Fig. 2: Configuration for direct (left) and indirect (right) BIE. To compute the surface potentials the normal derivative of eq. (3.4) is executed. Taking x to the boundary in the limit then yields p ( x ) 2G( x , y ) i f v n ( x ) = in + FP p( y ) dAy . (3.5) n x n x n y
A

Now only surface variables are addressed and the only unknown quantity is p . The solution of this integral equation posed many questions in the past because the Hadamard integral (Finite Part) is strongly singular due to the second order derivative. Dating back to the work of Hamdi [8] and others some vector algebraic transformations lead to a weak singularity that easily can be integrated using standard techniques [11]. Using a Galerkin approach one ends up with a variational form of the BIE f = 0, where f = 1 1 i 1 pin ( x ) 2G( x , y ) p p( x )dAy dAx v n ( x )p( x ) dAx p( x ) dAx . (3.6) ( ) y 2 nx ny A 2 f f 2 n x
AA A

The minimization of the fluid potential f gives the unknown jump of pressure. Indeed, the Galerkin approach requires the evalution of a twofold surface integral but the inherent symmetry of the fluid matrix makes this formulation particularly attractive for the coupling with the structural eq. (2.4). 3.3 Discretization

3.3.1 Indirect Boundary Element Method


The subdivision of the boundary surface into elements together with the approximation of the continuous p( y )
1 1

p ( y ) = Nf ( y ) P (3.7) 0 0 2 -1 by a matrix of element shape functions Nf,j (see fig. 3) and the 10 discrete vector P of nodal values yields a linear system of equations 1 (3.8) Fig. 3: BEM shape function Nf,j Af () P = Ff LT f 2 for the acoustic problem with the fully populated admittance matrix Af of the fluid. The system is excited by external acoustic sources Ff and by the velocity vector Vn of the vibrating surface. The latter is expressed by modal participation factors j. L is a geometrical coupling matrix combining fluid and structural shape functions. This formulation is implemented in SYSNOISE [3].

18. CAD-FEM Users Meeting Internationale FEM-Technologietage

20. - 22. September 2000 Graf-Zeppelin-Haus, Friedrichshafen

3.3.2 Ritz approach


As for every other discretizing method the practical applicability of the indirect BEM is limited by the mesh density increasing with frequency. Another variational approach introduced by Pierce [13] and refined by others [11,12,14] makes use of global trial functions j. This way eq. (3.7) is replaced by p( y ) = ( y ) , (3.9) resembling eq. (2.3). The j are chosen from a set of admissible functions, i.e. they satisfy the boundary condition p = 0 along the contour line of any open part of the boundary 0 b/2 (e.g. the edges of a plate). Of course this approach is only -a/2 x2 possible for simple geometries, but as for the case of the x1 solar panels frequently rectangular plates are considered, where trigonometric functions (see fig. 4) defined on the whole surface do an excellent job. For this particular selecFig. 4: Ritz trial function j tion the inner integral of the twofold surface integration in eq. (3.6) can be computed analytically, yielding a fluid matrix Af with many zero entries because of the orthogonality of the trigonometric functions. This leads to a block structured topology of the matrix, where the resulting system of equations is 1   T . Af () =  (3.10) f f 2
1

Fluid-structure coupling

When fluid-structure interaction has to be accounted for both the displacement and the pressure are the degrees of freedom. To accomplish the coupling of the equations one has to shift these terms on the left hand side of eqs. (2.4) and (3.8) or (3.10) which results in the coupled systems
   Z L L s  F s = , 1 1  T  Af P Ff L Af f 2 f 2 Y X Z for the BEM case and the Ritz approach, respectively. Advantage can be  if the structural variable is eliminated in eq. (4.1). The result is of Z s  Z s T L  

 F s  Ff

(4.1)

taken of the diagonal shape

 T  -1  1    T  -1   . Af =  f s s L Z s L + 2 f The size of the problem is reduced now without loosing the symmetry of the system.

(4.2)

Diffuse Sound Field

Experimentally a diffuse sound field is simulated in a reverberant chamber. The large number of reflections approximates the ideal diffuse sound field where an infinite number of plane propagating waves with random phase relationships from uniformly distributed directions arrive at the test object. Because of this random characteristic a stochastic approach [9] is best suited to model the diffuse sound field. It is based on the power spectral density (PSD) representation of the acoustic loading 2 (unit Pa /Hz). This quantity expresses the acoustic energy per frequency band and is given from the experimental data of a former launch measured in the payload bay of the space vehicle. If the PSD matrix SX of the input of a discrete linear system is known the matrix equation SY = H * S X H T (5.1) -1 results in the PSD matrix SY of the output. In our case the complex frequency response matrix H = Z is the inverse of the coupled impedance matrix. Coyette [4] presented an efficient scheme to do this task numerically with respect to the coupled system eq. (4.1). The excitation is driven by the pressure gradient of the incident wave (last term in eq. (3.6)). Based on the PSD of pin at a single point in the reverberant room (this is the quantity known from measurements in the payload bay) the input PSD matrix SX relating the nodes of the mesh via cross PSDs can be derived analytically [9,11].

18. CAD-FEM Users Meeting Internationale FEM-Technologietage

20. - 22. September 2000 Graf-Zeppelin-Haus, Friedrichshafen

6
6.1

Applications
Solar panels

Fig. 5 (left) presents a sketch of a solar generator in the stacked launch configuration. The result quantity of interest for the design of the panel is the PSD of the acoustically excited acceleration ex2 pressed in g /Hz at two points A4 and A5 (fig. 5 (right)). To give a feeling what this frequency response
100. 10. 1. 0.1 S_y [g**2/Hz] 0.001 1.E-4 1.E-5 1.E-6 20. 50. f [Hz] 100. 150. A4 A5

Fig. 5: Solar panels in the stacked configuration (left) and PSD of acceleration in 2 units g /Hz vs. frequency at the selected points A4 and A5 (right) (SYSNOISE). says quantitatively it should be noted that sitting in a civil airliner one would experience a vibration of 2 2 about 0.001 g /Hz. A PSD of 0.1 g /Hz is typical of military vehicles and (mechanically induced) base excitation of satellite equipment. A comparison with fig. 5 (right) reveals that acoustically induced vibrations can even exceed these amplitudes. In fact acoustic loading means a critical load case for large and lightweight structures such as solar panels. The analysis, i.e. a frequency by frequency computation of eq. (5.1) in steps of 1 Hz was done within SYSNOISE [3] on the basis of the coupled system eq. (4.1) (left). A modal basis of 142 in vacuo modes and a fluid mesh with 3220 nodes was used to obtain convergence of the results up to 200 Hz. A rule of thumb suggests six elements per acoustic wavelength. This proved to be a good estimate for closed structures. However, benchmarks have shown that about 12-18 elements per wavelength are necessary for the convergence of the current problem. The CPU time on a SGI/origin platform was several hours per frequency step. 6.2 Isotropic plate benchmarks

6.2.1 Mechanical excitation


To speed up the coupled analysis of rectangular plates the alternative Ritz approach of section 3.3.2 and 4 was implemented in a Windows program (fig. 6) with a graphical user interface. It covers all issues of fluid loaded panels (coupled, uncoupled, baffled, unbaffled) and was named WinFloaP (see [11] for more details). To validate the results some simplified benchmarks were considered. An isotropic, simply supported, quadratic steel plate (a = 1 m) was immersed in water, a situation where heavy fluid loading is expected. The thickness h of the plate has been set to 1 cm and a loss factor of 1 % was assumed for all modes. A unit point force was applied at (0.2m;0.3m) with respect to a centered coordinate system.

Fig. 6: Screenshot of the program WinFloaP.

18. CAD-FEM Users Meeting Internationale FEM-Technologietage

20. - 22. September 2000 Graf-Zeppelin-Haus, Friedrichshafen

Fig. 7 shows the surface averWinFloaP SYSNOISE in vacuo aged root mean square (RMS) 1E-2 velocity as a structural response quantity of the plate resulting from WinFloaP. A 1E-3 comparison with a FEM/BEM based computation by means of SYSNOISE proves the ac1E-4 curacy of the approach. Also plotted is the result of the classical in vacuo analysis. It can be noticed that the ex1E-5 0 50 100 150 200 250 300 350 400 450 500 pected behavior, i.e. deFrequency [Hz] creasing vibration amplitudes and decreasing characteristic Fig. 7: Surface averaged RMS velocity of a water loaded unbafffrequencies, is well reproled steel plate versus frequency compared to the result in vacuo. duced by the results. The large shifts of the eigenfrequencies are a result of the additional mass due to the heavy fluid loading caused by water. A coarse estimate [11] for the fluid loaded characteristic frequencies based on the radiation of a baffled circular piston yields 14 Hz for the first in vacuo mode at 50 Hz which is in good agreement with the result of fig. 7.

6.2.2 Acoustic excitation

RMS velocity [m/s]

60

85

Now the same structural 1E-2 model is excited by plane waves in air impinging on the 1E-3 plate. The inclination angle of the incident wave with respect O 1E-4 to the plate normal (0 corresponds to the normal direction) is modified whereas the 1E-5 wavefronts are parallel to a plates edge (y-axis) for all O 1E-6 cases. For 0 only the res0 100 200 300 400 500 600 700 800 900 1000 ponse of some modes is noFrequency f [Hz] ticed. They prove to be the Fig. 8: Surface averaged RMS velocity versus frequency of an unsymmetric ones. On the other O O O baffled steel plate excited by plane waves (0 , 60 , 85 ). hand all the plane waves with wavefronts parallel to the y-axis cannot excite modes that show anti-symmetry with respect to the xaxis. Therefore, e.g. the 22-mode at 197 Hz and the 24-mode at 493 Hz are missing in fig. 8 for all angles of incidence (compare with the in vacuo curve of fig. 7). To explain the behavior of the velocity amplitudes with respect to frequency two adverse effects have to be considered. For low frequencies the response of the plate is decreasing with increasing angle of incidence because the surface averaged RMS jump of pressure p is decreasing, too. For higher frequencies this effect isnt as pronounced as before because the radiation efficiency of a plate reaches its maximum at the coincidence frequency [5,11] which is about 1200 Hz for the current case. As a consequence of the acoustic reciprocity [5] also the sensitivity of a plate to incident waves is higher for frequencies above the coincidence frequency. Moreover the trace wavelength of the acoustic wave on the plate is shorter for high incidence angles what means the same effect as increasing the frequency. Again, comparison with FEM/BEM results (not presented here) show good agreement [11]. The comparison of the convergence between the two approaches of eq. (4.1) revealed that in order to obtain a certain level of accuracy up to 1000 Hz 81 DOFs are sufficient for the Ritz approach whereas 2600 DOFs are necessary for the BEM analysis when a pure acoustic scattering problem (rigid plate) is considered [11]. Therefore the reduction of the CPU time is about the factor 20 (BEM on SGI/origin, WinFloaP on PC/Pentium platform). This factor is even increasing for higher frequencies. For small systems with several hundreds degrees of freedom the bottleneck of the solution time is the quadrature of the fluid matrix. Thus, further reduction of the CPU time can be obtained by neglecting its offdiagonal terms, i.e. the cross modal coupling. For the acoustic quantities larger errors are observed after this approximation, but the structural results only revealed minor differences [11].
RMS velocity [m/s]

18. CAD-FEM Users Meeting Internationale FEM-Technologietage

20. - 22. September 2000 Graf-Zeppelin-Haus, Friedrichshafen

Fig. 9 (left) presents the distribution of the jump of pressure (amplitude) at 980 Hz and an inclination O angle of 30 . This periodical pattern of the loading is a typical feature for the excitation with plane waves at higher frequencies. Fig. 9 (right) shows the response of the structure, i.e. the amplitudes of the displacement distribution. On its basis one can easily compute the stress distribution in the plate and the object, namely to design a structure subjected to acoustic loading, is achieved.
Jump of pressure (amplitude [Pa])
4.272E+00 4.005E+00 3.738E+00 3.471E+00 3.204E+00 2.937E+00 2.670E+00 2.403E+00 2.136E+00 1.869E+00 1.602E+00 1.335E+00 1.068E+00 8.010E-01 5.340E-01 2.670E-01

Displacement (amplitude [m])


7.937E-09 7.441E-09 6.945E-09 6.449E-09 5.953E-09 5.456E-09 4.960E-09 4.464E-09 3.968E-09 3.472E-09 2.976E-09 2.480E-09 1.984E-09 1.488E-09 9.921E-10 4.960E-10

f = 980 Hz, 30 degrees

f = 980 Hz, 30 degrees

Fig. 9: Jump of pressure (left) and displacement amplitude (right) for a plane wave excitation in air O at 980 Hz with an angle of incidence of 30 (WinFloaP).

Conclusion

The paper presented two approaches to compute the response of a structure excited by acoustic sources. The FEM/BEM technique implemented in SYSNOISE has been compared to a variational approach based on global trial functions for rectangular plates (WinFloaP). Besides the fact of producing reliable results the efficiency of the proposed algorithm has been demonstrated for the examples of a force excitated plate with heavy fluid loading and a plate subjected to acoustic plane waves.

8
[1] [2]

References
ANSYS, Inc., Canonsburg, PA: ANSYS Basic Analysis Procedures Guide (V5.5). 3 ed. (1998). Astley, R. J. et al.: Three-dimensional wave-envelope elements of variable order for acoustic radiation and scattering. Part 1. Formulation in the frequency domain. J. Acoust. Soc. Am. 103(1), pp. 49 63 (1998). Coyette, J.-P. and van de Peer, J.: SYSNOISE Users Manual, Rev. 5.3. LMS Numerical Technologies, Leuven, Belgium (1996). Coyette, J.-P. and Lecomte, C.: An efficient computational scheme for evaluating the response th of a structure subjected to random acoustic excitations. In: Proc. 15 International Modal Analysis Conference, Orlando (1997). nd Cremer, L. and Heckl, M.: Structure-Born Sound. Springer Verlag, Berlin, 2 ed. (1988). Hurty, W. C. and Rubinstein, M. F.: Dynamics of Structures. Prentice-Hall, New Jersey (1964). Ciskowski, R. D. and Brebbia, C. A. (ed.): Boundary Element Methods in Acoustics. Computational Mechanics Publications, Southampton (1991). Hamdi, M. A.: Formulation variationelle par quations intgrales pour le calcul de champs acoustiques linaires proches et lointains. PhD thesis, Universit de Compigne (1982). Jacobsen, F.: The diffuse sound field. Technical report 27, Technical University of Denmark, The Acoustics Laboratory (1979). Moosrainer, M. and Fleischer, H.: Application of BEM and FEM to musical instruments. In: von Estorff, O. (ed.): Boundary Elements in Acoustics Advances and Applications, Chapter 12, WIT-Press, Southampton (2000). Moosrainer, M.: Fluid-Struktur-Kopplung Vibroakustische Lsungsmethoden und Anwendungen. In: Fortschr.-Ber. VDI, Reihe 11, VDI-Verlag, Dsseldorf (to appear). Nlisse, H. et al: Fluid-structure coupling for an unbaffled elastic panel immersed in a diffuse field. J. Sound Vib. 198(4), pp. 485 506 (1996). Pierce, A. D.: Stationary variational expressions for radiated and scattered acoustic power and related quantities. IEEE J. Oceanic Eng. 12(2), pp. 404 411 (1987). Wu, X. F. et al: Variational method for computing surface acoustic pressure on vibrating bodies, applied to transversely oscillating disks. IEEE J. Oceanic Eng. 12(2), pp. 412 418 (1987).
rd

[3] [4]

[5] [6] [7] [8] [9] [10]

[11] [12] [13] [14]

18. CAD-FEM Users Meeting Internationale FEM-Technologietage

20. - 22. September 2000 Graf-Zeppelin-Haus, Friedrichshafen

You might also like