# Analytical modeling of squeeze film damping for rectangular elastic plates using Green's functions

## Analytical modeling of squeeze film damping for rectangular elastic plates using Green's functions

ARTICLE IN PRESS Journal of Sound and Vibration 329 (2010) 4617–4633 Contents lists available at ScienceDirect Journal of Sound and Vibration journa...
ARTICLE IN PRESS Journal of Sound and Vibration 329 (2010) 4617–4633

Contents lists available at ScienceDirect

Journal of Sound and Vibration journal homepage: www.elsevier.com/locate/jsvi

Analytical modeling of squeeze ﬁlm damping for rectangular elastic plates using Green’s functions M.M. Altu˘g Bıc- ak , M.D. Rao Department of Mechanical Engineering - Engineering Mechanics, Michigan Technological University, 1400 Townsend Drive Houghton, MI 49931-1295, USA

a r t i c l e in fo

abstract

Article history: Received 14 August 2009 Received in revised form 11 May 2010 Accepted 12 May 2010 Handling Editor: L. Huang Available online 31 May 2010

An analytical method for the solution of squeeze ﬁlm damping based on Green’s function to the nonlinear Reynolds equation considering an elastic plate is presented. This allows calculating the stiffness and damping forces rapidly for various boundary conditions. The elastic plate velocity is applied to the nonlinear Reynolds equation as a forcing term. The nonlinear Reynolds equation is divided into multiple linear nonhomogeneous Helmholtz equations, which then can be solvable using the presented approach. Approximate mode shapes of a rectangular elastic plate are used, enabling the calculation of the damping ratio and frequency shift for the linear case, as well as the complex resistant pressure, for both linear and nonlinear cases. & 2010 Elsevier Ltd. All rights reserved.

1. Introduction Squeeze ﬁlm damping (SFD) occurs when a plate moves in close proximity to another surface, in effect alternately stretching and squeezing any ﬂuid that may be present in the space between the moving plates. The ﬂuid can act as a mass, spring and damper, having a signiﬁcant effect on the dynamics of the moving plates. The primary goal of a ﬂuid ﬁlm damping system is to limit the vibration of a given structure by dissipating the energy to the ﬂuid within the ﬁlm. However, in micro-electromechanical systems (MEMS) and micro-opto-electromechanical systems, the SFD impacts the operating behavior such as in microswitches, microsensors, microaccelerometers, telescope mirrors , etc. From a vibrational point of view, the SFD is a very useful and cost-effective solution to most vibration and vibration-caused noise reduction problems. Extensive literature has already been developed for the SFD effect relating to air ﬁlm lubrication , which has application in air bearing and levitation systems. The squeeze ﬁlm analysis of the ﬂuid is covered by three classes of models, the standard Helmholtz equation model, the low reduced frequency model and the Navier–Stokes model. Darling et al.  used the linearized Reynolds equation to calculate the additional spring and damping force acting on the plate using Green’s function. Ingard et al.  used the wave equation approach under the small amplitude assumption. Using the wave equation and statistical energy analysis, Chow et al.  predicted the damping well above the critical frequency of the thick plate. However, the loss factor calculations for the statistical energy analysis are based on the impedance approach which needs the pressure distribution to be calculated in advance. Maidanik et al.  used the simpliﬁed Navier–Stokes equation approach with the incompressible ﬂuid assumption. ¨ nsay  and Fox et al.  developed However, the validity of this assumption is restricted to very low frequencies. O fully coupled models including viscothermal effects. Beltman  considered viscothermal effects by the full linearized

 Corresponding author.

ARTICLE IN PRESS 4618

M.M. Altug˘ Bıc- ak, M.D. Rao / Journal of Sound and Vibration 329 (2010) 4617–4633

Navier–Stokes problem and low reduced frequency model, and investigated a spherical resonator . Basten et al.  applied the low reduced frequency solution of Beltman et al.  to calculate the acousto-elastic behavior of double-wall panels. Readers should refer to Refs. [9,11] for an extensive literature review. Moreover, Beltman  developed a viscothermal acoustic ﬁnite element which models the effects of inertia, viscosity, compressibility and thermal conductivity. Akrout et al.  applied this development to two laminated glass plates enclosing a thin viscothermal ﬂuid cavity. Lei et al.  developed a three-dimensional viscous ﬁnite element model for the analysis of the acoustic ﬂuid–structure interaction systems including the cochlear-based transducers which consists of a three-dimensional viscous acoustic ﬂuid medium interacting with a two-dimensional ﬂat structure domain. Akrout et al.  used a modal approach to determine the vibro-acoustic system’s response which shows the importance of the viscothermal effects in the case of thin ﬂuid layers. The Reynolds equation, known from lubrication technology and the theory of rareﬁed gas physics, is the theoretical background to analyze the SFD effect in this paper. The models that account for ﬂexibility are almost exclusively based on the linearized Reynolds equation or its simplest version—the linearized incompressible Reynolds equation . The models that use the nonlinear Reynolds equation however usually approximate a structure as a one-dimensional beam . The nonlinear Reynolds equation is used in conjunction with the plate equation only in Nayfeh’s work . Langlois  found damping and spring forces based on squeeze number using a strip plate. Starr  noticed the nonlinear effects based on amplitude  and gave an approximate formula in order to calculate the nonlinear damping force based on the constant deformation of an oscillating plate. The effects of the boundary condition and the mode shape of the oscillating plate are usually ignored while calculating the viscous forces in MEMS area. The common approach is to minimize the damping force in order to reduce the effect of damping on the operating behavior. Because of this, many researchers analyzed a plate with multiple holes with varying geometry [20–22]. The SFD of parallel plate and basic damping effect can be explained using the modal theory. The resistive force to the plate oscillating normally against the stationary plate is caused by the pressure distribution between plates. Generally, if the plate oscillates with a low frequency, the ﬂuid is not compressed considerably. In this case, there might be an additional mass loading due to the air pumping mechanism. However, if the oscillation frequency of the plate is high, the air fails to escape resulting of the elastic force domination. The Reynolds equation is not valid for high frequencies since the Reynolds equation does not include inertial terms. The measure of applicability of the Reynolds equation can be found by comparing inertial forces with viscous forces. Gross  deﬁned the validity of the Reynolds equation by using the modiﬁed Reynolds number (Re) which is deﬁned by the ratio of the inertial force to the viscous force. According to this approach, the Reynolds equation is valid for Re 51 since the Poiseuille velocity proﬁle along the gap is assumed. The usage of the low reduced frequency model derived from viscothermal models [9,24], covers both the Reynolds equation and the wave equation using shear wavenumbers. This paper reports results of a theoretical analysis for the SFD effect on a ﬂexible plate using Green’s function. The attached plate mode shapes are applied to the nonlinear Reynolds equation as a forcing term for the ﬂuid to calculate the nonlinear spring and damping forces. For the purpose of this investigation, this can be accomplished by afﬁxing a cover plate over the vibrating structure while leaving a thin air gap between the two pieces. Attaching a new elastic system to the vibrating structure can be thought as the tuned mass damper. However, unlike in the tuned mass damper, the SFD is effective over large bandwidth instead of one particular frequency. The harmonic solution of the nonlinear Reynolds equation is presented in this work. The general solution of the nonlinear Reynolds equation is divided into each harmonic subproblem. It is also shown that each harmonic problem is dependent upon the solutions of previous harmonics. Using the general solution of the Helmholtz equation, the solution to each harmonic problem can be found using Green’s functions. Particular solutions of the nonhomogeneous Helmholtz equation require the input mode shape which can be calculated using biharmonic plate expressions. Using the approximate mode shapes of a rectangular elastic plate, the damping ratio and frequency shift for the linear case, as well as the complex resistant pressure, for both linear and nonlinear cases are calculated. 2. The nonlinear Reynolds equation solution The ﬂuid ﬂow in continuum regime is governed by the continuity equation and the Navier Stokes momentum equations which are valid for unsteady, compressible and viscous ﬂow. For a small air-gap separating the two plates, the squeeze ﬁlm ﬂow is predominantly two dimensional (e.g. in the x–y plane). Under following assumptions

 No external forces act on the ﬁlm.  No inertial effects exist.  The structure oscillates with small amplitude and the main ﬂow is driven by pressure gradients in the x and y    

directions. No slip ﬂow occurs at the planar boundaries. No variation of pressure across the ﬂuid ﬁlm. The ﬂow is laminar; no vortex ﬂow and no turbulence occur anywhere in the ﬁlm. Fully developed ﬂow is considered within the gap.

ARTICLE IN PRESS M.M. Altug˘ Bıc- ak, M.D. Rao / Journal of Sound and Vibration 329 (2010) 4617–4633

4619

 Different assumptions can also be considerable for speciﬁc type of ﬂuid. For air, ﬂow is assumed to be isothermal, i.e. ðpprÞ. The nonlinear Reynolds equation is 3

r

ph rp 12m

! ¼

q ðphÞ qt

(1)

where rp is the gradient of the pressure, h is the thickness of the ﬁlm, and m is the ﬂuid viscosity. At low ambient pressure or in very thin ﬁlms, when the mean free path length of the gas is not negligible compared with the ﬁlm gap h, molecular interactions with the surfaces need to be taken into account. The theory of rareﬁed gas ﬂow was developed by Knudsen in the early 1900s. Veijola  has given a function that approximates the pressure and ﬁlm width dependency of the viscosity in narrow gaps by

meff ¼

m

(2)

1 þ 9:658Kn1:159

where Kn is the Knudsen number deﬁned as Kn ¼ l=h, where l is the mean free path of molecules and h is the gap distance. Based on different derivation considerations, one can get a different viscosity deﬁnition such as in . The mean free path of air molecules at ambient pressure Pa is about 65 nm. The Reynolds equation is applicable only in the continuum ﬂow regime; the relationship between the Knudsen number and ﬂow regimes is shown at Table 1. The effective dynamic viscosities are also tabulated to show the effect of the ﬂow regime on the viscosity. Under the assumption of harmonically varying gap thickness with the frequency of o ðrad=sÞ, the following equation is deﬁned: hðx,y,tÞ ¼ h0 ð1þ dFðx,yÞejot Þ

(3)

where F is the function of x,y which is the deﬂection shape of the structure, d is the dimensionless vibration amplitude which is smaller than 1 and o is the angular frequency. Since Eq. (1) is nonlinear, harmonics of o will appear in the pressure solution. So the pressure can be assumed as ! 1 X jnot (4) ar ðx,yÞe pðx,y,tÞ ¼ Pa 1 þ r¼1

where ar(x,y) is the coefﬁcient for the rth harmonic, which is also complex. If Eqs. (3) and (4) are put into Eq. (1), it can be rewritten as 1 X 1 1 X 1 1 q X qa q X qa 1X U rk þ 1 ejrot þ U rk þ 1 ejrot ¼ rjoðar þ ar1 dFÞejrot qx qy sr ¼1 qx k ¼ 1 r ¼ k qy k ¼ 1 r ¼ k

(5)

where s is the squeeze number per unit area,

12meff o h20 Pa

(6)

and

U ¼ ar1 þ 3dFar2 þ 3d2 F2 ar3 þ d3 F3 ar4

(7)

The complex coefﬁcients for the zero and negative values of r are therefore, a0 ¼ 1,

ar ¼ 0 for r o0

(8)

In order to get the squeeze number which is reported in the literature [16,25], Eq. (6) should be multiplied by the total area of the plate, i.e LxLy. The ﬁrst three harmonics of the Eq. (5) can be written as Eqs. (9)–(11) respectively as q2 a1 q2 a1 þ ¼ jsða1 þ dFÞ qx2 qy2

(9)

Table 1 The Knudsen number and the corresponding ﬂow regimes. Knudsen number (Kn)

Flow regime

meff =m

Kn o 0:01 0:01 o Kno 0:1 0:1 o Kno 10 Kn 4 10

Continuum Slip Transitional Molecular

0:956 o meff =m o 1 0:6 o meff =m o 0:956 0:007o meff =m o 0:6 0 o meff =m o 0:007

ARTICLE IN PRESS M.M. Altug˘ Bıc- ak, M.D. Rao / Journal of Sound and Vibration 329 (2010) 4617–4633

4620

    q qa2 qa1 q qa2 qa1 þ ða1 þ 3dFÞ þ þða1 þ3dFÞ ¼ 2jsða2 þ a1 dFÞ qx qy qx qx qy qy

(10)

    q qa3 qa2 qa1 q qa3 qa2 qa1 2 2 þ ð3dF þ a1 Þ þ ð3d F2 þ a1 3dF þ a2 Þ þ þ ð3dF þ a1 Þ þ ð3d F2 þ a1 3dF þ a2 Þ ¼ 3jsða3 þ a2 dFÞ qx qx qy qy qx qx qy qy

(11) These nonhomogeneous Helmholtz equations (9)–(11) can be solved using Green’s functions exactly for the uniform deﬂection proﬁles. However, only the ﬁrst harmonic solution is published in the literature . In the present study, more complicated and realistic plate deﬂections are considered and the modal force approach is presented. Once the ﬁrst harmonic equation is solved using Green’s functions, Fourier series or numerical methods, the second harmonics can be solved using the ﬁrst harmonic solution. Moreover, higher harmonics can also be sequentially solved since all the equations are the nonhomogeneous Helmholtz equations, of which the forcing term is already calculated. 2.1. Solution using Green’s function Darling et al.  showed the use of Green’s function to solve the linearized nonhomogeneous Reynolds equation. Compact analytical models were also presented considering the rigid uniform and the tilting motion of the plate. However, the mode shapes of the rectangular plates are not considered. This present study is the extension of the solution to the elastic models of plates considering the nonlinear Reynolds equation. The general solution to the linearized Reynolds equation can be given using an inﬁnite series as a1 ðx,yÞ ¼

1 X 1 X m

bmn fm ðxÞgn ðyÞ

(12)

n

where fm(x) and gn(y) are the harmonic functions of x and y respectively which satisfy the boundary conditions, and bmn is a complex constant coefﬁcient. The derivative of fm(x) and gn(y) are q2 fm ðxÞ ¼ m2 a2 fm ðxÞ qx2

and

q2 gn ðyÞ 2 ¼ n2 b gn ðyÞ qy2

(13)

where m and n are the integer numbers that can be odd or even based on the boundary conﬁguration, a and b are the multipliers of x and y inside the functions fm and gn. If the deﬂection shape F of the plate is expanded to the same series such as

Fðx,yÞ ¼

1 X 1 X m

emn fm ðxÞgn ðyÞ

(14)

n

the following relationship can be obtained by putting Eqs. (12) and (14) into (9), bmn ¼

jsd 2

m2 a2 þn2 b þ js

emn

(15)

If the same procedure is applied for the second harmonic, considering the expansion on the same fm(x) and gn(y) functions, Rðx,yÞ ¼

1 X 1 X m

rnm ¼

Z

Ly

0

Z

rmn fm ðxÞgn ðyÞ

(16)

Rf m ðxÞgn ðyÞ dx dy

(17)

n Lx

0

where the known forcing function R is R ¼ bmn fðbmn þ3demn Þðfm02 gn2 þ fm2 gn02 Wfm2 gn2 Þ2jsdemn fm2 gn2 g 0

0

2 2

(18)

2 2

where fm and gn are the derivative, w.r.t. its independent variable, and W ¼ m a þ n b . In this case, the complex constant coefﬁcient of the second harmonic turns into cmn ¼

rnm 2

m2 a2 þ n2 b þ 2js

(19)

P P1 where a2 ðx,yÞ ¼ 1 m n cmn fm ðxÞgn ðyÞ. The generalized pressure solution to the rth harmonic can also be given in a similar form as in Eq. (12): ar ðx,yÞ ¼

1 X 1 X m

n

drmk fm ðxÞgk ðyÞ

(20)

ARTICLE IN PRESS M.M. Altug˘ Bıc- ak, M.D. Rao / Journal of Sound and Vibration 329 (2010) 4617–4633

4621

where drmn represents the rth harmonic complex constant coefﬁcient which can be obtained as smn

drmn ¼

(21)

2

m2 a2 þ n2 b þrjs

in which smn represents the coefﬁcients of the residual known term expansion which is dependent upon the solutions of all harmonics up to the rth harmonic. The pressure domain solution for the ﬁrst harmonic equation (12) can be represented in compact form using Green’s function also as Z Ly Z Lx Fðx, ZÞGðx,y, x, ZÞ dx dZ (22) a1 ðx,yÞ ¼ 0

0

where G is Green’s function Gðx,y, x, ZÞ ¼

1 X 1 X m

emn fm ðxÞgn ðyÞfm ðxÞgn ðZÞ

(23)

n

where emn is a multiplier which is dependent upon m,n, o and s. If the deﬂection of the plate F is one of the mode shapes of the plate F ¼ Fm , the modal force can be written using the pressure solution as Z Ly Z Lx Fm ðx,yÞpðx,yÞ dx dy (24) N¼ 0

0

If the mode shape Fm is selected as orthonormal, which satisﬁes Z Ly Z Lx rFm ðx,yÞFm ðx,yÞ dx dy ¼ 1 0

(25)

0

where r is the mass per unit area of the plate. The modal force for the ﬁrst harmonic solution (N1) of the pressure can be obtained as 1 X 1 X

N1 ¼

m

n

jsd 2

m2 a2 þ n2 b þjs

e2mn

(26)

The modal forces for the rth harmonic can now be written as in a generalized form as Nr ¼

1 X 1 X m

n

smn 2

m2 a2 þ n2 b þ rjs

emn

(27)

2.2. Range of applicability In 1962, Langlois  derived the general form of the Reynolds equation based on the Navier–Stokes equations and the general equations of hydrodynamics. The Reynolds equation is obtained under the condition that the modiﬁed Reynolds number is much smaller than unity. The Reynolds equation is not valid for high frequencies since it does not include inertial terms. The measure of applicability of the Reynolds equation can be found by comparing inertial forces with viscous forces per volume , Re ¼

inertial force ruðqu=qnÞ ¼ viscous force mðq2 u=qz2 Þ

(28)

where u is the ﬂuid ﬂow velocity in the direction of n. Assuming the Poiseuille velocity proﬁle along the gap and the direction n as the x-axis, the following formula can be obtained for the ﬂuid ﬂow velocity: u¼

hzðqp=qxÞ  z 1 2m h

(29)

where z is the direction along the gap. The modiﬁed Reynolds number is deﬁned as Re ¼

roh20 51 m

(30)

which should be smaller than unity for any point underneath the plate. The modiﬁed Reynolds number can be rewritten using the squeeze number as Re ¼

rPa h40 s 12m2

The modiﬁed Reynolds number formula considering the journal bearings can be found in Ref. .

(31)

ARTICLE IN PRESS 4622

M.M. Altug˘ Bıc- ak, M.D. Rao / Journal of Sound and Vibration 329 (2010) 4617–4633

3. Calculation of squeeze ﬁlm stiffness and damping According to the analysis of Langlois , and Grifﬁn et al. , squeeze ﬁlm air provides extra damping and stiffness force to the system. For a ﬁnite element modeled system, one can get mass, stiffness and damping matrices as _ Mx€ þ Kx ¼ f ðx, xÞ

(32)

Eq. (32) can be expressed in terms of modal vectors, xðtÞ ¼ Fm jðtÞ in which Fm represents modal vector as follows:

j€ þ o2n j ¼ FTm f ðFm j, Fm j_ Þ

(33)

where on and jðtÞ represent the natural frequency and modal coordinates. For a continuous system, Eq. (33) can be written as Z j€ þ o2n j ¼ Fm f ðFm j, Fm j_ Þ dO (34) O

where O represents the solution domain. After the calculation of the modal forcing term, the imaginary part can be _ ¼ joj. converted into the velocity component using the relationship j The right hand side of Eq. (34), the modal force provides two components the in-phase and the out-of-phase which can _ . Considering a single degree of freedom system, the damping ratio and the natural be decomposed as FTm f ¼ Fk jFd j frequency shift can be found as

z¼ Do ¼

1 F 2on d

(35)

qﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ o2n þFk on

(36)

The damping ratio and the frequency shift formulations are based on the linear pressure solution, hence it cannot be applicable to nonlinear cases. The mass normalized mode shapes are used in order to calculate the modal damping and the frequency shift using Eqs. (35) and (36). The other way to calculate the damping ratio and the frequency shift is to use the time integration technique and then calculating the phase difference. However, the nonlinearity or higher harmonics cannot be captured since the stiffness and the damping are calculated based on the phase difference between the velocity and the pressure. Moreover, the sucking and squeezing motions create unequal ﬂuctuations around the ambient pressure. 4. Examples To illustrate the details of the pressure and modal force solution, the transverse motion of a plate is considered. The problem domain is selected as 0%x%Lx and 0%y%Ly corresponding to a rectangular plate of dimensions Lx,Ly. The proper Green’s function is constructed and used for each different case. For the structural modal solution, approximate mode shapes for Poisson’s ratio is 0.25 which can be found in Ref.  are used. Tabulated mode shapes are calculated approximately using the Rayleigh method. In order to solve the problem deﬁned by Eq. (5), boundary conditions should be deﬁned. For the boundary edges there are two different boundary conditions:

 Boundary point is open to ambient pressure p =Pa.  Pressure gradient is zero, or closed end qp=qn ¼ 0 where n is the outwards normal vector at the boundary. The structural boundary conditions for the plate edges are represented by four letters. For example, CFCS stands for clamped–free–clamped–simply supported plate for bottom (y= 0), right (x= Lx), upper (y= Ly) and left edges (x = 0). It is assumed that the pressure gradient is zero for simply supported and clamped edges, whereas the boundary condition is open to ambient air for the free edge. All investigated mode shapes are shown at Table 2. The ﬁrst harmonic solution to the nonlinear Reynolds equation is presented in the following subsections. The second harmonic solutions can be found in the Appendices. 4.1. Exact solution for clamped boundary conditions (CCCC) For the clamped boundaries case, the approximate normalized mode shape  using the Rayleigh method is given as    2px 2py Fðx,yÞ ¼ cos 1 cos 1 (37) Lx Ly The ﬁrst-order pressure a1 can be written using Eq. (12) where f ðxÞ ¼ cos

2pmx , Lx

gðyÞ ¼ cos

2pny Ly

(38)

ARTICLE IN PRESS M.M. Altug˘ Bıc- ak, M.D. Rao / Journal of Sound and Vibration 329 (2010) 4617–4633

4623

Table 2 Mode shapes. Conﬁguration

Deﬂection shape Fðx,yÞ ¼ Fm ðx,yÞ    2px 2py cos 1 cos 1 Lx Ly    px 2py 1cos 1 cos 2Lx Ly   2px 1cos L   x px sin Lx    px py 1cos 1cos 2Lx 2Ly   px 1cos 2Lx

CCCC CFCC FCFC FSFS CFFC FFFC

Since the mode shape can be represented by using only four terms of the entire series, the solution of the nonhomogeneous Helmholtz equation yields only four terms. Using the solution, the nondimensional pressure can be calculated as follows: ! js js js f ðxÞ þ 2 g1 ðyÞ f1 ðxÞg1 ðyÞ (39) a1 ðx,yÞ ¼ d 1 þ 2 2 a þ js 1 b þ js ða2 þ b Þ þ js where a ¼ 2p=Lx , b ¼ 2p=Ly , f1 ðxÞ ¼ cosax and g1 ðyÞ ¼ cosby. And modal force can be obtained as using the orthonormalized mode shape equation (37) as sﬃﬃﬃﬃﬃﬃﬃﬃﬃ ! 1 Lx Ly js js 1 js Pa d 2 þ 2 (40) þ 2 þ N¼ r 3 a þjs b þ js 2 a2 þ b2 þ js where r is the constant mass density per unit area of the plate. Unlike the solution which is given by Darling et al. , which exhibits the inﬁnite series solution for the uniform plate displacement, the solution to a particular mode shape includes the ﬁrst couple of terms of the entire series due to the expansion of the mode shape. Darling concluded that the damping mechanism does not exist considering the uniform motion. However, even if all the edges are closed, the pumping mechanism due to the mode shape deﬂection, exists which creates damping, which is plausible and can be understood by considering the imaginary part of Eqs. (39) and (40). One can get a similar formula using the low reduced frequency model approach as follows: ! av1 ðx,yÞ ¼

d

2

nðss~ Þ 1 þ

G2 o

2

G o þ c0 a

2

f1 ðxÞ þ

G2 o G2 o g1 ðyÞ þ f1 ðxÞg1 ðyÞ 2 2 2 c0 b þ G o c0 ðb þ a2 Þ þ G2 o

(41)

pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ where av1 is the pressure solution of the ﬁrst harmonic to the low reduced frequency model, s ¼ 12 h0 ro=m is the shear ~ ~ 1 wavenumber, pG ﬃ is pﬃ the propagation constant, c0 is the velocity of the sound, nðss Þ ¼ ð1 þ ððg1Þ=gÞBðss ÞÞ , Bðss~ Þ ¼ tanhðs jÞ=s j1, g is the ratio of the speciﬁc heats, and s~ is the square root of the Prandtl number. More information about the low reduced frequency model can be found in Ref. . The nonlinear (second harmonic) solution for the CCCC case can be found in Appendix A. 4.2. Case: CFCC Considering the same plate geometry as before, but a free edge at y= Lx, the nondimensional pressure a1 can be obtained as   1 X jsdð1Þðm1Þ=2 emn 1 1 2py mpx cos a1 ðx,yÞ ¼  cos (42) mp A1 A2 Ly 2Lx m where 8 < p4  emn ¼ 2 : 4

pﬃﬃﬃﬃﬃﬃﬃ m2 p2 p2 þ4 2 þjs, j ¼ 1 for m ¼ 1,3,5, . . . : 4L2x Ly m41 pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ After using the orthonormalized mode shape with a multiplier 12 3Lx Ly ð3p8Þ=rp, the modal force can be written as pﬃﬃﬃﬃﬃﬃﬃﬃﬃ   1 X j Lx Ly Pa sdemn 1 1 (43) N¼ þ pﬃﬃﬃ p ﬃﬃﬃ ﬃ A1 2A2 2 3p2 Sm2 r m m¼1

and

A1 ¼

m2 p2 þ js, 4L2x

A2 ¼

ARTICLE IN PRESS 4624

M.M. Altug˘ Bıc- ak, M.D. Rao / Journal of Sound and Vibration 329 (2010) 4617–4633

where (

emn ¼

p2 8p þ16 m ¼ 1 32

m 41

and

rﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ 3p8 :

p

The nonlinear (second harmonic) solution for this case can be found in Appendix B.

4.3. Case: FCFC and FSFS Considering the same plate geometry as before, but with the clamped edges, which is the zero ﬂow condition, expressed as x = 0 and x= Lx, and the free edges at y =0 and y= Lx, the nondimensional pressure can be found as   1 X 4jsd 1 1 2px npy sin a1 ðx,yÞ ¼  cos (44) pn A1 A2 Lx Ly n where A1 ¼ n2 p2 =L2y þ js and A2 ¼ 4p2 =L2x þ n2 p2 =L2y þ js for n =1,3,5,y . Using 12 ð6Lx Ly =rÞ1=2 as the orthonormal multiplier, one can ﬁnd the modal force as pﬃﬃﬃ pﬃﬃﬃﬃﬃﬃﬃﬃﬃ   1 X 4 2j Lx Ly Pa sd 2 1 pﬃﬃﬃ þ N¼ p ﬃﬃﬃ ﬃ A1 A2 3p2 n2 r n

(45)

If the simply supported mode shape is used, then the pressure and the modal force turn into a1 ðx,yÞ ¼

1 X 1 X m

n

8jsdemn mpx pny   cos sin Lx Ly m 2 p2 n2 p2 þ L2 þ js L2

p

(46)

2 ðm2 1Þn

x

1 X 1 X m

n

y

pﬃﬃﬃ pﬃﬃﬃﬃﬃﬃﬃﬃﬃ 32 2j Lx Ly Pa sdemn   pﬃﬃﬃﬃ 2 2 2 2 p4 ðm2 1Þn2 r mL2p þ n Lp2 þ js x

(47)

y

where

emn ¼

1

m¼1

2

m 41 pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ and the orthonormal multiplier is Lx Ly =2r for m =0,2,4,yand n =1,3,5,y . The nonlinear (second harmonic) solution for this case is presented in Appendices C and D.

4.4. Case: FFFC In this case, the three edges where x =Lx, y=0 and y=Ly are considered as open to the ambient air, and the left edge is considered as the zero ﬂow condition at x= 0, the nondimensional pressure is a1 ðx,yÞ ¼

1 X 1 X m

n

p



2 mn

jsdemn m2 p2 4L2x

2 2 þ n Lp2 þj y

 cos

s

mpx pny sin 2Lx Ly

(48)

where 8 > < p4 4 emn ¼ > : ð1Þðm þ 1Þ=2

m¼1

for m ¼ 1,3,5, . . . and n ¼ 1,3,5, . . .

m 41

pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ After using the orthonormalized mode shape with the multiplier Lx Ly ð3p8Þ=2rp, the modal force can be expressed as pﬃﬃﬃ pﬃﬃﬃﬃﬃﬃﬃﬃﬃ 1 X 1 X 4 2j Lx Ly Pa sde2mn rﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ (49) N¼   3p8 2 2 pﬃﬃﬃﬃ m2 p2 n2 p2 m n p4 m n r 4L2 þ L2 þ js

p

x

y

The nonlinear (second harmonic) solution for this case is presented in Appendix F.

ARTICLE IN PRESS M.M. Altug˘ Bıc- ak, M.D. Rao / Journal of Sound and Vibration 329 (2010) 4617–4633

4625

4.5. Case: CFFC Considering the same plate as in FFFC case, but with the clamped edge at y= 0, the nondimensional pressure a1 is a1 ðx,yÞ ¼

1 X 1 X m

n

p

2 mn

jsdem en mpx pny   cos cos 2Lx 2Ly m2 p2 n 2 p2 þ 4L2 þjs 4L2 x

(50)

y

where 8 > < p4 4 em ¼ > : ð1Þðm þ 1Þ=2

m¼1

and

m41

8 > < p4 4 en ¼ > : ð1Þðn þ 1Þ=2

n¼1

for m ¼ 1,3,5, . . . and n ¼ 1,3,5, . . .

n 41

The modal force can be obtained after using the orthonormalized mode shape with the multiplier ðð3p8Þ=2Þ pﬃﬃﬃﬃﬃﬃﬃﬃﬃ 1 X 1 X 32j Lx Ly Pa sde2m e2n   N¼ pﬃﬃﬃﬃ m n p3 ð3p8Þm2 n2 r m2 p2 þ n2 p2 þjs 4L2 L2 x

pﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃﬃ Lx Ly =rp as (51)

y

The nonlinear (second harmonic) solution for this case is presented in Appendix E. 5. Comparison of cases 5.1. Linear (First harmonic) comparisons In this section, the pressure values and modal forces are compared for a square plate Lx =Ly = 1 m, considering the orthonormalized mode shapes. Readers should divide the mode shapes with its multiplier to replicate the following results. The real and imaginary components of each modal force are plotted in Figs. 1 and 2 as functions of the squeeze number s deﬁned by Eq. (6) and compared with the solution of Darling et al.  in which deﬂection is taken as uniform. While calculating the complex resistance force using Darling’s formula, the total mode deﬂection is used as the uniform deﬂection amplitude. The mass density per unit area r is selected as 1 kg/m2 in order to compare with the literature results. Moreover, the frequency shift and the damping ratio are also evaluated and plotted in Fig. 3 for each case using Eqs. (35) and (36) considering o is changed inside Eq. (6). The other parameters used in the calculations are Pa ¼ 100 kN=m2 ,

m ¼ 2  105 N s=m2 , h0 ¼ 0:1 mm

(52)



2

and s ¼ ðo=4Þ s=m . The modiﬁed Reynolds number for the examples are Re ¼ s=400, which validates the results up to s ¼ 400 m2 (2.6 in the Log scale). The normalized stiffness force for different boundary conditions are presented with the literature results  in Fig. 1. It is interesting to note that the stiffness force converges to 0.45 for the CCCC case and goes to 1 where the theoretical limit

0 Normalized Stiffness Force Log10 Force/ δ PaLxLy

Normalized Stiffness Force Log10 Force/ δ PaLxLy

0

−2

−4

−6

−8

−10

1 edge is vented* 2 adjacent edges are vented* 2 opposite edges are vented* 3 edges are vented* All edges are vented*

−12

−14 −5

−4

−3

−2

−1 Log σ 10

0

1

2

3

−2

−4

−6

−8 CFFC FFFC FSFS FCFC CFCC CCCC

−10

−12

−14 −5

−4

−3

−2

−1 Log σ

0

1

2

10

Fig. 1. Normalized spring force for different boundary conditions of a square plate: (a)  for constant uniform deﬂection, (b) present study.

3

ARTICLE IN PRESS M.M. Altug˘ Bıc- ak, M.D. Rao / Journal of Sound and Vibration 329 (2010) 4617–4633

4626

for high s values. Fig. 1(b) depicts that the normalized stiffness force remains constant for CCCC unlike vented cases, while vibration frequency o is decreasing. This trend is clearly the same as in Fig. 1(a) and the order of ventilation cases are the same. For the vented cases, the low frequency force exhibits a directly proportional relationship in the Log–Log scale and the following relationship can be extracted log10

stiffness force Clog10 s2 dPa Lx Ly

(53)

The normalized damping force for both the literature results  and the present study results are shown in Fig. 2. It is observable that the damping force increases proportionally with s values up to a maximum level then decreases gradually. The decreasing trend for the realistic cases in Fig. 2(b) is not the same as for the uniform deﬂection case in Fig. 2(a). However, the increasing trend in Fig. 2(a) is nearly the same as in Fig. 2(b). The effects of o and/or s on the frequency shift and damping ratio can be seen in Fig. 3. A Log axis is selected in order to capture the effects. Notice that the plate mode shape is assumed unchanged during the calculations of the frequency shift and damping ratio. It is interesting to point out that for all cases except the CCCC case, the frequency shift is directly proportional with s for low values and inversely proportional for high values of s. Since the stiffness or the real part of the

0 Normalized Damping Force Log10 Force/δ PaLxLy

Normalized Damping Force Log10 Force/δ PaLxLy

0

−1

−2

−3

−4

−5

1 edge is vented* 2 adjacent edges are vented* 2 opposite edges are vented* 3 edges are vented* All edges are vented*

−6

−7 −5

−4

−3

−2

−1

0

1

2

−1

−2

−3

−4 CFFC FFFC FSFS FCFC CFCC CCCC

−5

−6

−7 −5

3

−4

−3

−2

−1

0

1

2

3

Log10σ

Log10σ

Fig. 2. Normalized damping force for different boundary conditions of a square plate: (a)  for constant uniform deﬂection, (b) present study.

0

−1.5

−2

−2 −2.5 Damping Ratio Log10ζ

−1

−3 −4 −5 −6

CFFC FFFC FSFS FCFC CFCC CCCC

−7 −8 −9 −5

−3

−3.5 CFFC FFFC FSFS FCFC CFCC CCCC

−4

−4.5

−5 −4

−3

−2

−1 Log10σ

0

1

2

3

−5

−4

−3

−2

−1

0

Log10σ

Fig. 3. Frequency shift (a) and damping ratio (b) for various boundary conditions of the square plate.

1

2

3

ARTICLE IN PRESS M.M. Altug˘ Bıc- ak, M.D. Rao / Journal of Sound and Vibration 329 (2010) 4617–4633

4627

The normalized midpoint pressure

3.0

2.5 Present study Viscothermal model 

2.0

1.5

1.0

0.5

0 −2

−1.5

−1.0

−0.5 0.0 Frequency Log(kHz)

0.5

1.0

1.5

Fig. 4. The midpoint (x,y= 0.5 m) absolute pressure variation.

reaction force reaches to steady state of which the impact of frequency shift is reduced. Moreover, unlike vented cases such as CFCC, FCFC, etc. the frequency shift of CCCC is constant up to s  0:01 m2 . For high values of s air underneath the plate is getting squeezed, instead of escaping through boundaries which are open to ambient pressure. This phenomenon can also be seen in the damping ratio graph. For very small values of s, the damping ratio stays constant, which means most of the air can be pumped unlike for high frequencies, at which the pumping effect is reduced and damping is said to be inversely proportional with frequency. These damping evaluations can also be used for the statistical energy analysis purposes under parabolic ﬂow restriction. The comparison between the viscothermal model  and the present study is presented in Fig. 4. The variation of the normalized absolute pressure at the center point of the plate is plotted against frequency. The pressure calculations become different starting from 100 Hz, the model presented in this study is no longer capable of producing correct results beyond this frequency. 5.2. Nonlinear (second harmonic) comparisons The capability of the present analysis is further demonstrated by considering the nonlinear analysis of the plate which has the same physical properties as that speciﬁed for the previous linear problem. The three cases CCCC, CFCC, and FCFC are investigated. The inﬂuence of the nondimensional deﬂection multiplier d and s on the total amount of force of the ﬁrst and second harmonics are illustrated in Fig. 5. The isopleth map is used to show the total force acting on the plate. As shown in Fig. 5, the effect of the deﬂection d on the total force is rather signiﬁcant. When the s values are low, the real part of the linear solution Fig. 5(a) is proportionally increasing with increasing d. The real part of the total force tends to shift for higher values of s. The real part of the ﬁrst harmonic total force is nearly directly proportional with the d in the s range investigated. However, the real part of the total force due to the second harmonic Re(Fa2), presents a different distribution. Re(Fa2) is nearly directly proportional with d for high values of s, but nearly constant for the lower values. It can be concluded that the nonlinearity of the real part of the total force is rather important for the high values of s for this case. This phenomenon can also be observed on the imaginary part. Moreover, the optimal damping force of the ﬁrst and second harmonics in Fig. 5(b,d), can be achieved by adjusting the s values around 50 m  2. The plate with CFCC boundary conditions shows different contour lines in Fig. 6. In this case, Re(Fa1) and Re(Fa2) show similar contour proﬁles except the magnitudes of the contours. The increase in the second harmonic force Re(Fa2) is more than that of the Re(Fa1) for high s values. The imaginary parts Im(Fa1) and Im(Fa2) exhibit local extremums around s ¼ 10 m2 and gradually increase for high values of s. Unlike in the CCCC case, the stiffness and the damping force increase in the second harmonic due to high values of d is more than the increase in the ﬁrst harmonic. It can therefore be concluded that the nonlinearity is more severe than that in the CCCC case. The results of the plate with FCFC boundary conditions are presented in Fig. 7. The real forces Re(Fa1) and Re(Fa2), show similar contour proﬁles but the amplitudes of the contours are different as in the previous case. However, the stiffness force nonlinearity is lower compared to the CFCC case. This can also be observed in the damping forces in Fig. 7(b,d).

ARTICLE IN PRESS M.M. Altug˘ Bıc- ak, M.D. Rao / Journal of Sound and Vibration 329 (2010) 4617–4633

0

0.2 0.1

0 05

0

50 σ 0.1

0.2

0.1

0.4

0.25

0.15

5

0.8

0.2 0.15

0.3

5

0.2

0 .4 0.5 0.45 0.35

5 0.2 0.1.1 0 0.05

0.6

0.4

0.3

0.4

100

0.20.3 5

0.8 δ

50 σ

0.6 δ

δ

0.2

0.8

0.1 0.05

0.4

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2

0.3 0.205.2 0.15

0.6

1.1

1

0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2

0.8

0.1

0.6

100

0. 4

0.35 0.3 0.25

0.2 0.15 0.1

0.0

δ

4628

5

0.4

0.05

0.05

0.2

0.2 0

50 σ

100

0

50 σ

100

Fig. 5. Total force acting to the plate for CCCC case, real part of a1 = Re(a1) (a), imaginary part of a1 = Im(a1) (b), Re(a2) (c), Im(a2) (d).

0 .7

0.4

0.4

0.2

0.2 0.15

0.1

0.05

50 σ

100

0

3 40 2 0 1 00.

0 60042

1. 8

1 .6

1.4 1.2

1 0.8

0.25

0.6

0.2

0.1

0

0.3

0.4

0.3

0.2

δ

δ

0.6

0.8

0.35

0.8

0.6

0.8 0.6 0.4

0.4

δ

δ

0.6

1 00012 055

5 0.

00. 130.2

0.8

50 σ

100

0.8 0.7 0.6

0.4

0.3

0.5

0.6

0.2

0.4

0.1

0.2

0.2

0.2 0

50 σ

100

0

50 σ

100

Fig. 6. Total force acting to the plate for CFCC case, Re(a1) (a), Im(a1) (b), Re(a2) (c), Im(a2) (d).

6. Conclusions Compact analytical models for computing the effects of compressible SFD are developed using Green’s function approach. The coupling is handled by applying the structural velocity distribution to the Reynolds equation as the forcing term. The nonlinear Reynolds equation is divided into the harmonics of the oscillation frequency of the structure. Then the

ARTICLE IN PRESS M.M. Altug˘ Bıc- ak, M.D. Rao / Journal of Sound and Vibration 329 (2010) 4617–4633

0.8

25

0.3 0. 0. 2

0.6

0.5

δ

δ

5

0.8

0.6

4 0. 0.3

0.6

0.3

0.7

0.2 0.1

0.8

4629

0.4

0.4

0.2

0.2

0.15 0.1

0

50 σ

100

0.8

0.4

δ

δ

0 .4

1 0.8

0.6

0.2

0.6

50

100

σ 1. 2

0.8

0.05

0

0 .45 0.35 0.3

0.6

0.4

0.4

0.2

0.2

0.25 0.2

0.15 0.1

0.05

0

50 σ

100

0

50 σ

100

Fig. 7. Total force acting to the plate for FCFC case, Re(a1) (a), Im(a1) (b), Re(a2) (c), Im(a2) (d).

solution to the Reynolds equation for each harmonic is presented in order to calculate the nonlinear pressure distribution, the nonlinear damping and stiffness forces, as well as the damping ratio and frequency shift for the linear case. The nonlinearity due to the deﬂection amplitude of the structure and the squeeze number are investigated. It is found that the nonlinearity is directly effected by the boundary conditions for a particular conﬁguration. The presented method allows the rapid calculation of reaction forces using the inﬁnite series which includes the expansion of approximate mode shapes on eigenfunctions. The truncation of inﬁnite series to the ﬁrst few terms can also be represented, which is useful for system simulations. Tabulated examples can be expandable to cover more complicated and higher mode shapes. In conclusion, the present analysis provides an efﬁcient and rapid technique for investigating both linear and nonlinear effects of SFD on rectangular elastic plates. Further, it also provides a powerful, compact and convenient tool to identify the modal damping and frequency shift for linear cases as well as pressure distribution underneath plates in practice. Appendix A. CCCC Considering a1 ðx,yÞ ¼

2 2 X X

cmn cos

mpx npy cos Lx Ly

(54)

emn cos

mpx npy cos Lx Ly

(55)

m¼0n¼0

and a2 ðx,yÞ ¼

4 4 X X m¼0n¼0

the second harmonic complex coefﬁcients are

d

e00 ¼  ðc22 2c20 2c02 þ 4c00 Þ 4 e20 ¼ 

ð6dc20 þc02 c22 þ2c00 c20 2dc22 Þp2 d þ ðc22 2c20 c02 þ2c00 Þ 2 isL2x e40 ¼

e02 ¼ 

2 2 ð3dc22 6dc20 þ2c20 þ c22 Þp2

iL2x s

þ

2c20 c22 d 4

ð2c00 þc20 þ 2c02 c22 ÞidsL2y þ 2ð3dc22 þ c22 c20 þ 2c00 c02 þ 6dc02 Þp2 4p2 þ 2iL2y s

ARTICLE IN PRESS M.M. Altug˘ Bıc- ak, M.D. Rao / Journal of Sound and Vibration 329 (2010) 4617–4633

4630

e04 ¼

e22 ¼ 

e42 ¼ 

e24 ¼ 

2 2 ð2c02 c22 Þ 14 isdL2y þ ð6dc02 2c02 c22 3dc22 Þp2

iL2y s þ8p2

ðc00 c20 þc22 c02 ÞidL2y s þ ðc00 c22 þ c02 c20 3dc20 þ3dc22 Þ2p2

L2 y L2x

 þ1

2p2 þiL2y s

ðc20 c22 ÞL2x L2y ids þ ð6dc20 L2y 6dc22 L2y þ 4c20 c22 L2y þ c20 c22 L2x 3dc22 L2x Þ2p2 2L2x ð2p2 þ iLy2 sÞ

ðc02 c22 ÞL2x L2y ids þ ðc02 c22 Ly2 3dc22 L2y þ 4c02 c22 L2x 6dc22 L2x þ 6dc02 L2x Þ2p2 2L2x ðiL2y s þ 8p2 Þ ðc22 L2y þ 3dL2y þ c22 L2x þ 3dL2x Þ4p2 þ isdL2x L2y

e44 ¼ c22

(56)

4L2x ðiL2y s þ8p2 Þ

Appendix B. CFCC Considering the ﬁrst harmonic solution as a1 ðx,yÞ ¼

 X 2py mpx cos cm þdm cos Ly 2Lx m

(57)

and the second harmonic, a2 ðx,yÞ ¼

X

em þ fm cos

m

 2py 4py mpx cos þ gm cos Ly Ly 2Lx

(58)

the second harmonic coefﬁcients are em ¼

pdð1n2 m2 Þ X ð1þ 8nmÞi4spd 3mn 8L2 x

n

aðn 71 7mÞ

ðdm 2cm Þ þ

3pdmn 32aL2x

ðdm 2cm Þ

X pm2 n

8aL2x n

2 ð2cm þ d2m Þ

where a ¼ 16pL2x =m2 p2 þ 8iL2x s, ðn 71 7 mÞ ¼ ðn þ 1þ mÞðn þ 1mÞðn1 þ mÞðn1mÞ and m,n are odd numbers. ( ) 4  þ 4Ln2 X 1 3p2 2 3p2 L2 ðm þ 1Þ=2 ðc m d Þ d  y 2 x2 pm2 cm dm is þ fm ¼ b m m m dð1Þ 2 2 4 32Lx 2Ly 4m n n   3p 2 2 2 12p ðm 1 þn Þ þ pis ðcm dm Þ L2 dm 8L2x y mnd þ ðm7 1 7nÞ where b ¼ 32L2x L2y =ð8isL2x L2y þ m2 p2 L2y þ 16p2 L2x Þ. ! n2 þ 4m2 is 3p 2 2 12p þ L42 X 1 p þ 16L2x ðn þ m 1Þ þ L2y 3p2 3p2 2 16L2x y ðm þ 1Þ=2 2 2 mnddm m d ð1Þ d  p d þ hm ¼ c m is þ 2 þ m m ðm 71 7 nÞ 8 2Ly 64L2x nð4m2 n2 Þ n

(59)

(60)

(61)

where c ¼ 32L2x L2y =ð8isL2x L2y þm2 p2 L2y þ64p2 L2x Þ Appendix C. FCFC Considering the ﬁrst harmonic solution as  X 2px mpy sin cm þ dm cos L Ly x m

(62)

 X 2px 4px mpy sin em þfm cos þ hm cos Lx Lx Ly m

(63)

a1 ðx,yÞ ¼ and the second harmonic, a2 ðx,yÞ ¼

the coefﬁcients of the second harmonic can be found as ( ) X ðdg dn 2cg cn Þ þcn ðg 2 þ n2 m2 Þðcg þ 1 dg Þ p 3p2 2 2 gmn em ¼ a is þ 2 m dðdm 2cm Þa ðn 7g 7mÞ 2Ly L2y n,g

(64)

ARTICLE IN PRESS M.M. Altug˘ Bıc- ak, M.D. Rao / Journal of Sound and Vibration 329 (2010) 4617–4633

where a ¼ L2y =ð2isL2y þm2 p2 Þ and n,g are odd numbers. ! 1 2 2 2 2 2 4 X 2L2y ðg þ n m Þcg dn 2ðg þ n ÞL2x cg dn 3 12b fm ¼ b 2is þ 2 m2 p2 dðdm cm Þ þ 2 p2 ddm þ 8bp gmn ðn 7 g 7 mÞ Ly Lx n,g where b ¼ L2x L2y =ð2isL2x L2y þ4p2 L2y þ m2 p2 L2x Þ



!

hm ¼ c is þ

1 2 2 2 X L2y ðg þ n m Þ

12 2 3m2 2 p þ 2 p ddm 2cp 2Ly L2x n,g

16 L2x

þ L22

ðn 7 g 7 mÞ

y

4631

(65)

 gmndg dn

(66)

where c ¼ L2x L2y =ð2isL2x L2y þ 16p2 L2y þm2 p2 L2x Þ Appendix D. FSFS Considering the ﬁrst harmonic solution as a1 ðx,yÞ ¼

XX mpx npy cmn cos sin L Ly x m n

(67)

XX r px spy drs cos sin L Ly x r s

(68)

and the second harmonic solution, a2 ðx,yÞ ¼ the coefﬁcients of the inﬁnite series can be found as drs ¼

X dðaÞ rs

(69)

a

where dð1Þ rs ¼

p X

4m2 ng 2 hrscmn cgh 2 Lx m,n,g,h ðs 7 n 7 hÞðr 7 m 7gÞ

dð2Þ rs ¼ 3d

dð3Þ rs ¼ 3d

dð4Þ rs ¼ 

dð5Þ rs ¼

p X m2 cmn ðr2 þ 1m2 Þ L2x

m,n

ðn þsÞðr 7m 7 1Þ

X m2 cmn ð1r 2 m2 Þ pðn þ sÞð 7r þ m 7 1Þ m,n

2 2rgmnhsg cmn cgh 2 Lx m,n,g,h ð 7 s þ n 7hÞð7 r þ m 7gÞ

p2 X

3d X ngshcmn cgh ðs2 h2 n2 Þðm2 þ r 2 g 2 Þ ðs 7 n 7hÞðr 7m 7 gÞ L2y m,n,g,h dð6Þ rs ¼ 3d

dð7Þ rs ¼ 

p X n2 cmn ð þm2 þr 2 1Þ L2y

m,n

ðn þ sÞð 7r 7 1 þmÞ

p2 X 2cmn h2 cgh ðnhsÞðmÞ L2y

m,n,g,h

ðs 7 n 7 hÞðr 7 m þgÞ

X cmn ð1 þ m2 þ r 2 Þ dð8Þ rs ¼ 2jsd pðn þ sÞðr 7 1 7mÞ m,n

(70)

and m,g,r are even numbers, n,h,s are odd numbers. Appendix E. CFFC Considering the ﬁrst harmonic solution as a1 ðx,yÞ ¼

XX mpx npy cmn cos cos 2Lx 2Ly m n

(71)

ARTICLE IN PRESS M.M. Altug˘ Bıc- ak, M.D. Rao / Journal of Sound and Vibration 329 (2010) 4617–4633

4632

and the second harmonic solution, XX r px spy drs cos cos 2L 2Ly x r s P ðaÞ the coefﬁcients of the inﬁnite series can be found as drs ¼ a drs where   m2 g 2 r2 n2 h2 s2 mnghrsc c þ 2 2 gh mn X Lx Ly ð1Þðs þ n þ h þ r þ m þ g6Þ=2 dð1Þ rs ¼ ð 7s þ n 7 hÞð 7 r þ m 7 gÞ m,n,g,h a2 ðx,yÞ ¼

dð2Þ rs

X mnrscmn ð1Þðm þ n þ r þ s4Þ=2

¼ 3d

m,n

dð3Þ rs ¼ 

3pd X mnrcmn ð1þ m2 r2 Þð1Þðm þ r þ n3Þ=2 ðn þsÞð 7r þ m 71Þ 4L2x m,n

dð4Þ rs ¼ 

3pd X mnscmn ðn2 þ 1s2 Þð1Þðs þ n þ m3Þ=2 ðm þ rÞð 7 s 71 þ nÞ 4L2y m,n

dð5Þ rs ¼ 

dð6Þ rs

ð 71 þ n 7 sÞð 7 r þm 7 1Þ

1 þm2 r 2 1 þn2 s2 þ L2x L2y

(72)

!

  3d X 4mnðm2 þn2 Þcmn ð1Þðm þ n þ s3Þ=2 s r þ ð 7 17 m þrÞð 71 þ n 7 sÞ mþr nþs 4L2x m,n

  2js ðm þ n2Þ=2   3p 2 2 X mnd p þ 4L 2 ðm þn Þ cmn ð1Þ 1 1 x ¼ þ ðm þ rÞðn þsÞ 71 7 m þr 7 1 7s þ n m,n X dð7Þ rs ¼ 8jsd m,n

 mncmn ð1Þðr þ m2Þ=2 r 2 ð 71 7 s þnÞð 7 17 r þ mÞ s þ n

p

þ

s mþ r



(73)

and m,n,r,s are odd numbers. Appendix F. FFFC Considering the ﬁrst harmonic solution as a1 ðx,yÞ ¼

XX mpx npy cmn cos sin 2Lx Ly m n

(74)

and the second harmonic solution, XX r px spy drs cos sin 2Lx Ly r s P ðaÞ the coefﬁcients of the inﬁnite series can be found as drs ¼ a drs where a2 ðx,yÞ ¼

dð1Þ rs ¼

dð2Þ rs ¼

1 X mnghrscmn cgh ðm2 r2 g 2 Þ ð1Þðr þ m þ g3Þ=2 L2x m,n,g,h ð 7 s þ n 7hÞðr 7m 7 gÞ

2 X nhrscmn cgh ðn2 þ h2 s2 Þðm2 þ g 2 r2 Þ ð1Þðr þ m þ g3Þ=2 ð 7s þ n 7 hÞð 7r þ m 7 gÞ L2y m,n,g,h

dð3Þ rs ¼ 

mn3 ghrscmn cgh 8 X ð1Þðr þ m þ g3Þ=2 2 Ly m,n,g,h ðs 7n 7hÞð 7 r 7m þ gÞ dð4Þ rs ¼ 3d

dð5Þ rs ¼ 3d

dð6Þ rs

p X mrcmn ðm2 þ 1r 2 Þ 4L2x

m,n

ð 7 r 71 þ mÞðn þsÞ

p2 X 2rðm þrÞ þ ð 71 þ m 7rÞ sm3 cmn ð1Þðm þ r2Þ=2 pðm þ rÞðn þ sÞð 71 þ m 7rÞ m,n

4L2x

p ðm þ r2Þ=2 p2 X ð 7 1 þm 7 rÞ 2 þ2mðm þ rÞð1Þ n2 rscmn ¼ 3d 2 pðm þrÞðn þ sÞð 7 1þ m 7 rÞ Ly m,n

(75)

ARTICLE IN PRESS M.M. Altug˘ Bıc- ak, M.D. Rao / Journal of Sound and Vibration 329 (2010) 4617–4633

4633

p

X 2rðm þ rÞ þ ð 7 1 þm 7 rÞ 2 smc ð1Þðr þ m2Þ=2 dð7Þ mn rs ¼ 2jsd pðm þ rÞðn þsÞð 71 þ m 7rÞ m,n

(76)

and m,n,g,h,r,s are odd numbers. References  W. Jung, Y. Peter, J. Wang, O. Solgaard, Single-crystal-silicon continuous membrane deformable mirror array for adaptive optics in space-based telescopes, IEEE Journal of Selected Topics in Quantum Electronics 13 (2) (2007) 162–167.  W.E. Langlois, Isothermal squeeze ﬁlms, Quarterly Applied Mathematics 20 (2) (1962) 131–150.  R.B. Darling, C. Hivick, J. Xu, Compact analytical modeling of squeeze ﬁlm damping with arbitrary venting conditions using a Green’s function approach, Sensors and Actuators A: Physical 70 (1–2) (1998) 32–41.  U.K. Ingard, A. Akay, On the vibration damping of a plate by means of a viscous ﬂuid layer, ASME Journal of Vibration, Acoustics, Stress and Reliability in Design 109 (2) (1987) 178–184.  L.C. Chow, R.J. Pinnington, Practical industrial method of increasing structural damping in machinery, I: squeeze-ﬁlm damping with air, Journal of Sound and Vibration 118 (1) (1987) 123–139.  G. Maidanik, E.E. Ungar, Panel Loss Factors due to Gas-pumping at Structural Joints, NASA, 1967.  T. Onsay, Dynamic interactions between the bending vibrations of a plate and a ﬂuid layer attenuator, Journal of Sound and Vibration 178 (3) (1994) 289–313.  M.J.H. Fox, P.N. Whitton, The damping of structural vibrations by thin gas ﬁlms, Journal of Sound and Vibration 73 (1980) 279–295.  W.M. Beltman, Viscothermal wave propagation including acousto-elastic interaction, Part I: theory, Journal of Sound and Vibration 227 (3) (1999) 555–586.  W.M. Beltman, P.J.M. Van der Hoogt, R.M.E.J. Spiering, H. Tijdeman, Implementation and experimental validation of a new viscothermal acoustic ﬁnite element for acousto-elastic problems, Journal of Sound and Vibration 216 (1) (1998) 159–185.  T.G.H. Basten, P.J.M. Van Der Hoogt, R.M.E.J. Spiering, H. Tijdeman, On the acousto-elastic behaviour of double-wall panels with a viscothermal air layer, Journal of Sound and Vibration 243 (4) (2001) 699–719.  A. Akrout, L. Hammami, M.B. Tahar, M. Haddar, Vibro-acoustic behaviour of laminated double glazing enclosing a viscothermal ﬂuid cavity, Applied Acoustics 70 (1) (2009) 82–96.  C. Lei, R.D. White, K. Grosh, Three-dimensional viscous ﬁnite element formulation for acoustic ﬂuid–structure interaction, Computer Methods in Applied Mechanics and Engineering 197 (49–50) (2008) 4160–4172.  A. Akrout, C. Karra, L. Hammami, M. Haddar, Viscothermal ﬂuid effects on vibro-acoustic behaviour of double elastic panels, International Journal of Mechanical Sciences 50 (4) (2008) 764–773.  T. Le Van Suu, P. Honzik, S. Durand, N. Joly, Z. Skvor, M. Bruneau, Thin ﬁlm thermo-viscous damping in miniature condenser microphones, The Journal of the Acoustical Society of America 123 (5) (2008) 3229.  Y.J. Yang, M.A. Gretillat, S.D. Senturia, Effect of air damping on the dynamics of nonuniform deformations of microstructures, 1997 International Conference on Solid State Sensors and Actuators, TRANSDUCERS ’97, Vol. 2, Chicago, 1997, pp. 1093–1096.  A.H. Nayfeh, M.I. Younis, A new approach to the modeling and simulation of ﬂexible microstructures under the effect of squeeze-ﬁlm damping, Journal of Micromechanics and Microengineering 14 (2) (2004) 170–181.  J.B. Starr, Squeeze-ﬁlm damping in solid-state accelerometers, 4th Technical Digest, IEEE Solid-State Sensor and Actuator Workshop, 1990, 1990, pp. 44–47.  M.H. Sadd, A.K. Stiffer, Squeeze ﬁlm dampers: amplitude effects at low squeeze numbers, Journal of Engineering for Industry-Transactions of the ASME, Series B (97) (1975) 1366–1370.  M. Bao, H. Yang, Squeeze ﬁlm air damping in MEMS, Sensors and Actuators: A Physical 136 (1) (2007) 3–27.  P.Y. Kwok, M.S. Weinberg, K.S. Breuer, Fluid effects in vibrating micromachined structures, Journal of MicroElectroMechanical Systems 14 (4) (2005) 770–781.  D. Homentcovschi, R.N. Miles, Viscous damping of perforated planar micromechanical structures, Sensors and Actuators A: Physical 119 (2) (2005) 544–552.  W.A. Gross, Gas Film Lubrication, John Wiley, New York, 1962.  S. Beltman, W.M. Basten, T.G.H. Basten, H. Tijdeman, Viscothermal damping in thin gas or ﬂuid layers, Adaptive structures and material systems, American Society of Mechanical Engineers, New York, 1999, pp. 387–395. ISBN 9780791816424. ¨ T. Ryhanen, ¨  T. Veijola, H. Kuisma, J. Lahdenpera, Equivalent-circuit model of the squeezed gas ﬁlm in a silicon accelerometer, Sensors and Actuators A: Physical 48 (3) (1995) 239–248.  Y.T. Hsia, G.A. Domoto, An experimental investigation of molecular rareﬁcation effects in gas lubricated bearing at ultra-low clearance, Journal of Tribology-Transactions of the ASME 105 (1983) 120–130.  W.S. Grifﬁn, H.H. Richardson, S. Yamanami, A study of ﬂuid squeeze-ﬁlm damping, ASME Journal of Basic Engineering (1966) 451–456.  A.W. Leissa, Vibration of Plates, NASA, 1969.