The transmission loss of curved laminates and sandwich composite panels

Sebastian Ghinet, Noureddine Atalla, and JGMHaisam Osman and JGM

Citation: The Journal of the Acoustical Society of America 118, 774 (2005); doi: 10.1121/1.1932212
View online: http://dx.doi.org/10.1121/1.1932212
View Table of Contents: http://asa.scitation.org/toc/jas/118/2
Published by the Acoustical Society of America

Articles you may be interested in

Experimental investigation of the accuracy of a vibroacoustic model for sandwich-composite panels
The Journal of the Acoustical Society of America 137, (2015); 10.1121/1.4908239

New Wall Design for High Transmission Loss or High Damping

The Journal of the Acoustical Society of America 31, (2005); 10.1121/1.1907780

Transmission of sound through sandwich panels

The Journal of the Acoustical Society of America 56, (2005); 10.1121/1.1903474

Wave propagation in sandwich panels with a poroelastic core

The Journal of the Acoustical Society of America 135, (2014); 10.1121/1.4869090
The transmission loss of curved laminates and sandwich
composite panels
Sebastian Ghineta and Noureddine Atalla
Department of Mechanical Engineering, Universit de Sherbrooke, 2500 Boulevard Universit, Sherbrooke,
QC, J1K 2R1, Canada

Haisam Osman
The Boeing Company, 5301 Bolsa Avenue, Huntington Beach, California 92647

Received 30 August 2004; revised 7 April 2005; accepted 26 April 2005

The paper describes a model to calculate the transmission loss of both curved laminate and sandwich
composite panels within statistical energy analysis SEA context. The vibro-acoustic problem is
developed following a wave approach based on a discrete lamina description. Each lamina is
considered to consist of membrane, bending, transverse shearing and rotational inertia behaviors.
Moreover, the orthotropic ply angle of each lamina is considered. Using such a discrete lamina
description, the dispersion behaviors of the panel are correctly represented. Using the dispersion
curves, the radiation efficiency, the modal density, as well as, the nonresonant and the resonant
transmission are computed. Moreover, expression for the evaluation of the ring frequency and the
critical frequencies of such panels is given. The described model is shown to handle accurately, both
laminate and sandwich composite shells. Additionally, a transmission loss test is presented to
confirm the validity of the presented model. 2005 Acoustical Society of America.
DOI: 10.1121/1.1932212
PACS numbers: 43.40.r, 43.40.At JGM Pages: 774790

I. INTRODUCTION laminate symmetric. More advanced approaches accounting

Laminate and sandwich composite panels and cylinders
have increasingly found application in modern aerospace and
aeronautical structures. The composite materials used for
these constructions are generally lighter and stronger than
the most advanced aluminum alloys, which are prevalent in
aerospace constructions. These qualities lead however to in-
creased radiation efficiency and lower nonresonant transmis-
sion loss which unfortunately leads, in some instances, to
higher interior noise levels. In consequence, there is a need
for robust and fast numerical tools to efficiently estimate and
optimize the vibroacoustic behaviors of large laminate and
sandwich panels.
A large amount of work has been devoted to the model-
ing of laminate and sandwich panels. The published models
handle two classical types of constructions: laminates and
sandwich. The first class refers to lay-ups of composite plies
of similar or close physical properties. In general the lami-
nate is symmetric and each layer is modeled using a thin
plate theory13 or thin shell theory.49 The laminate is repre-
sented by an equivalent set of variables. It will be referred to
in the paper by a smeared thin laminate. Other authors con-
sider generalizations in which each layer is modeled using
thick plate theory and the laminate is planar10,11 or curved.12
Still, an equivalent set of variables is used for the whole
laminate. In this latter case, the proper estimation of the
equivalent shear correction coefficient is of paramount im-
portance. The majority of these formulations111 assume the
a
Author to whom correspondence should be addressed. Tel: 1 819 821 8000
# 3773; Fax: 1 819 821 7163; e-mail: Sebastian.Ghinet@USherbrooke.ca
for through-thickness deformations are presented in Refs. 13,
14, and 40. Reference 14 uses spectral finite element to
handle flat laminates; that is 1D finite element are used to
model through-thickness deformation and exponential func-
tions propagating wave shape are used to handle the in-
plane displacements. Reference 40 uses a similar method and
considers laminated cylindrical shells and pipes. It uses axi-
symmetry to handle the circumferential direction, exponen-
tial functions to model the axial displacement and solid finite
element for the radial displacement. The model is applied to
the calculation of the dispersion curves of the first modes of
laminated cylindrical shells and pipes.
For sandwich constructions a trilayer arrangement is
classically used; the core is generally soft and thick com-
pared to the skins. Each layer can be of a composite con-
struction. The classical models are based on two main as-
sumptions: i the skins are thin and work in bending; ii the
core is relatively thick and handles shearing effects only.
Following these assumptions and assuming the sandwich
symmetric, two classes of models are used. The first smears
the elastic constants of the panel through the
thickness.3,1521,37,39 The second uses a discrete layer
representation.22 This model uses a complete and mathemati-
cally coherent discrete layer theory for sandwich-type panels.
The theory is developed for a symmetric singly curved sand-
wich made up of a bottom skin laminate, a shearing core, and
a top skin laminate. Finally, note that in the context of lami-
nated composite cylinders modeling, two models were pre-
sented and compared by Ghinet et al.;12 symmetrical lami-
nate composite and discrete thick laminate composite. The
latter was shown to encompass the first and to handle accu-

774 J. Acoust. Soc. Am. 118 2, August 2005 0001-4966/2005/1182/774/17/$22.50 2005 Acoustical Society of America
rately, as a particular case, sandwich composite shells. In
these two models, membrane, bending, transverse shearing
as well as rotational inertia effects and orthotropic ply angle
of the layers were considered.
The thickness and structural complexity of the sandwich
panel could influence in varying degrees, the transverse
shear, the rotational inertia as well as the separate bending
motion of the skins at the mid-to-high frequencies. At these
frequencies it was observed12 that the two classes of model-
ing approaches, smearing and discrete layer, may lead to
different behaviors depending on the nature of the construc-
tion sandwich, thick laminate, etc.. It is important that a
model be devised that can handle both configurations. This
paper presents such a model. It is based on a discrete layer
approach. Each layer is allowed to exhibit bending, shearing,
and membrane behaviors. The construction can be none sym-
metric, laminates or sandwich. It will be referred to in this
text by a general laminate.
The numerical estimation of the transmission loss of
elastic structures can be accurately performed using finite
elements FE and boundary elements BE methods.23,24
These approaches require extensive computing resources and
are inappropriate for large geometrical-scale structures and
high frequencies calculation, when the vibration wavelength
becomes much smaller than the structural dimensions. In
contrast, statistical energy analysis is commonly character-
ized as much simpler to apply than FE/BE methods but it is
known to fail at low frequencies where the number of modal
resonance frequencies in the analysis band is low.25 There
are two methods for applying the SEA methodology. The
first is referred to by the modal approach and is based on
modeling each subsystem as a superposition of the resonant
responses of the set of uncoupled modes of the system. The
second method is based on a wave approach and models
each subsystem as a superposition of waves travelling
around the subsystem. It consists of deriving and solving a
dispersion system of equations between wave numbers and
frequency for the subsystem of interest assuming simple geo-
metrical configurations plates, shells, etc.. At interfaces, the
coupling loss factor is usually expressed in terms of the
semi-infinite system wave impedances. At low frequencies
where size effects are important, it is essential to include
corrections. One approach, for the specific problem of air-
borne transmission loss, is based on the application of spatial
windowing.26,27 An asymptotical approach is presented in
Ref. 28.
One particular problem of interest in the paper is the
estimation, using SEA, of the transmission loss of laminate
composite and sandwich curved structures. The principal
phenomena concerning the resonant and nonresonant trans-
mission of shells as well as a comprehensive review were
presented by Szechenyi.29,30 The problem of the nonresonant
transmission of a cylinder was solved by a simple geometri-
cal argument.30 The contribution of the stiffness-controlled
region was neglected and the subcoincident region was as-
sumed to be a circular sector for frequencies below the ring
frequency. The problem of the transmission loss into finite
cylinders was also discussed by Pope et al.3133 and
Lesueur35 using a modal approach. A similar approach33,35 is
J. Acoust. Soc. Am., Vol. 118, No. 2, August 2005 Ghinet, Atalla, and Osman: Curved laminate and sandwich composite panels
FIG. 1. The composite shell coordinates.
successfully applied here to calculate the low frequency
transmission loss of laminate and sandwich composite
curved panels.
This paper describes the SEA modeling of the transmis-
sion loss through finite laminate and sandwich composite
singly curved panels. Both laminate composite and sandwich
composite are modeled using a discrete thick laminate com-
posite theory. The studied transmission problem has three
primary resonant systems: two reverberant rooms separated
by the composite curved panel. The dispersion curves of the
structure are derived and solved for the modal density and
the radiation efficiency. Several models to compute the ra-
diation efficiency were tested.26,27,34 Identical results were
obtained and the model of Leppington34 was selected due to
its accuracy and fast convergence. These parameters allow
for the calculation of the radiation loss factor and also the
resonant contribution of the transmission loss. The standard
flat panel theory35 is used to compute the nonresonant trans-
mission but it is adapted here to the particular vibration be-
haviors of the curved panels see Sec. VI. In particular, a
subcoincident modes selection method is used to compute
the nonresonant transmission contribution. Moreover, the
classical wave approach nonresonant contribution is cor-
rected using the spatial windowing method presented in Ref.
27. Finally, a transmission loss experimental result of a
curved sandwich composite panel is successfully compared
with numerical estimations.
II. GEOMETRY AND COORDINATE SYSTEM
Figure 1 represents the global geometrical configuration
of the composite shell, where R is the curvature radius and h
is the total thickness. The layered construction is considered,
in general, asymmetrical as represented in Fig. 2a. The ori-
gin for the z axis is defined on a reference surface passing
through the middle thickness of the shell.
FIG. 2. The discrete laminated composite shell thickness constitution a
and interlayer forces b.
775
III. DISPERSION RELATION

The dynamic behavior of the curved panel is modeled

using a discrete layer theory which allows for both thick
laminate composites and sandwich shells. The displacement
field of any discrete layer i of the panel is of Mindlins
type: FIG. 3. The propagative wave number and the heading direction.
uix,y,z = ui0x,y + zixx,y,
U = u1 ; v1 ;w1 ; 1x ; 1y ;u2 ; v2 ;w2 ; 2x ;
vix,y,z = vi0x,y + ziyx,y, wix,y,z = wi0x,y. 1
2y ; uN ; vN ;wN ; Nx ; Ny T ,
For any layer of the shell, Flges theory36 is used to describe
the strain-displacement relations. Rotational inertia, in-plane,
F = F1x ;F1y ;Fz1 ;F2x ;F2y ;Fz2 ; FN1
x ;F y ;Fz .
N1 N1 T
3
bending as well as transverse shearing effects are accounted
for in each layer. Also, orthotropic ply angle is used for any The associated 5N + 3N 1 equations are composed of 5
layer. The resultant stress forces and moments of any layer equations of dynamic equilibrium for each of the N layers
are defined in Appendix A Eqs. A1 and A2. There are plus 3 equations of interlayer continuity of displacements for
three interlayer forces between any two layers, as repre- each of the N 1 interlayer surfaces.
sented in Fig. 2b. The total number of interlayer forces is To solve for the dispersion relations, the system of dy-
3N 1, where N is the number of layers. For any layer i namic equilibrium equations is expressed in terms of a hy-
there are five equilibrium equations: brid displacement-force vector e defined as:
i
Nx,x + Niyx,y + Fix Fi1

= ms +
Iz2
x

u,tt +
Iz

+ Iz2 x,tt ,
i
e =
U
F
. 4

R R Assuming a harmonic solution e = eexpjkxx + jky y jt,

the system is expressed in the form of a generalized poly-
Qiy nomial complex eigenvalue problem:
Niy,y + Nixy,x + + Fiy Fi1
y
R k2c A2e ikcA1e A0e = 0, 5

= ms +
Iz2
R

v,tt +
Iz
R
i
+ Iz2 y,tt , where, kc = k2x + k2y
is represented in Fig. 3, with kx
= kc cos and ky = kc sin , i = 1 and A0, A1, A2 are

real square matrices in the absence of damping of di-
Niy Iz2 i
mension 5N + 3N 1 defined in A13. Relation 5 has
i
Qx,x + Qiy,y + Fzi Fzi1 = ms + w,tt ,
R R 25N + 3N 1 complex conjugate eigenvalues and repre-
sents the dispersion relations of the laminated composite
i
M x,x + M iyx,y Qix + ziFix zi1Fi1 shell. In the context of the present approach, at any head-

= Iz x,tt +
u,tt
R
i
+ Iz2u,tt ,
ing direction the curved panel has two propagating solu-
tions below the ring frequency. At the ring frequency a
third solution becomes propagating thus, in the dispersion
field context the ring frequency is mathematically per-
M ixy,x + M iy,y Qiy + ziFiy zi1Fi1 ceived as a cut-off or transition frequency. Two other cut-

= Iz y,tt +
v,tt
R
i
+ Iz2v,tt . 2
off frequencies appear at high frequencies where two ad-
ditional solutions become propagating.

The subscript notation , in relations 2 and in the fol-

lowing indicates partial derivatives of with respect to
and . The external and internal surfaces of the shell are
considered stress-free so that F0x = F0y = Fz0 = 0 and FNx = FNy
= FzN = 0.
The expressions of the transverse shear stress forces Qi,
the in-plane stress forces Ni, the inertial terms Ii, and the
stress moments M ij of each layer are presented in Eqs.
A4A7. For any layer, the dynamic equilibrium equations
can be rewritten, using Eqs. 2 and relations A4A6 with
appropriate algebraic manipulations, as presented in Eq.
A12. The resulting dynamic equilibrium system, presented
in that form, has 5N + 3N 1 variables regrouped in two
vectors; a displacement-rotation vector U, and an interlayer
forces vector F:
IV. RING AND CRITICAL FREQUENCIES
The ring and critical frequencies of symmetrical lami-
nate composite cylinders were presented in Ref. 12. In this
section the ring frequency and the critical frequencies rela-
tions are presented in the context of the presented curved
discrete laminate model. As mentioned above, the ring fre-
quency can be considered as the first cut-off or transition
frequency of the dispersion equation. Here a simple approach
is used to estimate it. Considering that at the ring frequency
ring, the shell displacement is characterized by a breathing
mode shape, the derivatives of the hybrid vector e along x
and y directions are equal to zero. Consequently, Eq. 5
simplifies to

776 J. Acoust. Soc. Am., Vol. 118, No. 2, August 2005 Ghinet, Atalla, and Osman: Curved laminate and sandwich composite panels
 A01e = ring
2
A02e,
which represents a generalized eigenvalues problem with
A01 and A02, real square matrices of dimension 5N
+ 3N 1 defined by the expression A0 = A01 2A02
knowing that A0 is defined by expression A13.
By analogy with plates, the critical frequency limits of
the laminate composite curved shell are given by the particu-
lar solution of the dispersion equation 5 at coincidence; that
is when the structural wave number kc matches the acous-
tic wave number k0
6
n = 0
n, d . 12
The structural wave number of the shell k , and the
group velocity are computed numerically from the solution
of the dispersion relation 5.
The radiation efficiency of the panel k , for a
given frequency and heading is computed from Leppingtons
analytical formulas.34 Assuming energy equipartition
amongst the resonant modes equal modal energy, the radia-
tion efficiency of the composite panel is given by

k2c A2e ikcA1e A0e = 0, k0 = kc = /c0 .

7
rad =
1
n
0

k, n, d . 13

The corresponding band averaged values of the modal den-

In the case of a discrete laminated composite flat panel R sity and radiation efficiency are given, respectively, by
, the critical frequencies are computed numerically
from 7 using A0 = A01 2A02, kkmax0n, kdkd
n = min
,

kkmaxkdk
A2 A1 min
2c 2 + A02 e ic e A01e = 0, 8
c0 c0 kkmax0, n, kdkd
rad = min
, 14
which is a second order polynomial eigenvalues problem kkmax0n, kdkd
min
with A1 and A2 defined by the relations A13. Assuming
where, kmin and kmax are the wave number bounds of the
c = ic, Eq. 8 can be expressed in the form, studied frequency band. The radiation efficiency and the

modal density of each of the first three solutions are com-

A1 A2 puted using relation 14. Each solution is then considered a
+ A02 e A01 0 e resonant SEA subsystem.
c c0 c20 = ,
ce 0 I ce
1 0 VI. NONRESONANT TRANSMISSION
9
In general, for a complex construction, the nonresonant
transmission is heading dependent. Thus for or a given exci-
to obtain a generalized eigenvalue problem, where I is the
tation frequency band with cen the center band frequency
identity matrix and 0 a zero matrix of dimension 25N
and 1, 2 the frequency limits of the band, and an inci-
+ 3N 1. This problem has 25N + 3N 1 complex con-
dence direction , the structural and the forced wave
jugate eigenvalues. The critical frequencies correspond to so-
numbers are calculated from the dispersion relation 5 and
lutions which satisfies the condition c = ic, purely
the following conditions is checked to ensure that the forced
imaginary. This heading dependency will lead to a critical
modes are nonresonant,
frequency region given by
k0censin ks1 or k0censin ks2.
ic 15
f c = . 10
2
This accounts for both mass and stiffened controlled non-
It is found that the limits of the critical frequency region are resonant modes. Usually, stiffness controlled modes contri-
defined by f c1 = f c = 0 and f c2 = f c = / 2. bution is neglected and the mass-controlled non resonant
transmission coefficient is given by
1
nr =
V. THE MODAL DENSITY AND THE RADIATION cos2 min cos2 max

EFFICIENCY 2 max
nr, , sin cos dd , 16
The angular distribution of the modal density is classi- 0 min
cally expressed in terms of the ratio of the structural wave
number and the group velocity1 where
4Z20
A k, nr, , = , 17
n, = 11 ims + 2Z02
22 cg,
and min , max are the limit incidence angles describing
with A the area of the panel. The modal density is obtained the diffuse field, Z0 = 0c0 / cos is the specific acoustic
numerically by integrating over all headings directions, impedance of the medium and ms is the surface mass of

J. Acoust. Soc. Am., Vol. 118, No. 2, August 2005 Ghinet, Atalla, and Osman: Curved laminate and sandwich composite panels 777
the panel, and is the heading direction limited to non-
resonant modes. The allowable heading directions are ob-
tained using the dispersion equation 5 and the first con-
dition in Eq. 15.
In order to improve the low frequency predictions of the
non-resonant transmission coefficient, a geometrical win-
dowing correction method is also used. The correction
method used here, is detailed in Ref. 27 and examples of its
validation are given in Refs. 27,37 and 39. According to this
correction, the relation 17 changes as follows:
4Z20
nr, , = , , cos , 18
ims + 2Z02
where, , , is the geometric radiation efficiency of
the finite baffled window defined by27
, , = R jk0
A
S S
eikpx cos +y sin Gx,y,x,y
e jkpx cos +y sin dSx,ydSx,y . 19
In the above relation, Gx , y , x , y is the Greens function,
A is the area of the panel, and S denotes the radiating surface
of the panel. Relation 19 is evaluated using a semianalyti-
cal algorithm as presented in Ref. 27. Since this correction
only affect the low frequencies, it is postulated to remain
applicable for singly curved panels even if its derivation is
strictly limited to flat systems. The comparison with the ex-
perimental measurement presented in Sec. VIII C corrobo-
rates the validity of this correction.
Alternatively, a modal method can also be used to cal-
culate the diffuse field incidence transmission coefficient for
mass controlled modes33,35
16 2 2
Rmn 32 A 2
nr = c0 2 4 =2 j2 2 .
A mn ms ms2 mn mn
In the above equation, Rmn = mn0c0 denotes the modal ra-
diation resistance and jmn2
the joint acceptance. Symbol
mn indicates that the summation is limited to nonreso-
nant modes in the band of interest. In order to apply the
modal approach, the shell is assumed simply supported. The
modes are related to the wave number components by, kx
= m / Lx; ky = n / R. The corresponding natural frequencies
mn are obtained from the solution of the dispersion equation
5 recast in the symbolic form
kmn
2
A2 ikmnA1 A01e = mn
2
A02e,
with, kmn = k2x + k2y 1/2 and the matrices A1, A2, A01, and
A02, real square matrices of dimension 5N + 3N 1 de-
fined Eq. A13 with A0 = A01 2A02. The modal ra-
diation resistance or equivalently the modal joint accep-
778 J. Acoust. Soc. Am., Vol. 118, No. 2, August 2005
tance is calculated analytically assuming the classical spa-
tial separate form of the Greens function.31
VII. DIFFUSE FIELD TRANSMISSION LOSS
The modal density and the radiation efficiency presented
in the above sections are used within a SEA framework to
compute the transmission loss of the laminated composite
shell. A simple SEA acoustic transmission scheme consists of
two reverberation rooms separated by the studied curved
panel. One of the rooms is excited by a diffuse field and the
acoustic transmission problem is assumed to encompass two
transmission contributions: resonant and nonresonant trans-
mission. As a first approximation, and for the sole calculation
of the transmission loss, the first solution wave of the disper-
sion relation is used to represent the dynamics of the curved
panel.
The acoustic rooms cavities are described by the sys-
tems 1 and 3 while the curved panel is identified as system 2.
Using the classical SEA equations, the noise reduction of the
panel is given by
NR = 10 log10
A3
A
+ nr + r
2 1
rad
, 22
where, A3 is the surface absorption of the receiving room,
2 = 2 + 2rad is the sum of the space and band averaged
radiation loss factor35 rad of the panel and its averaged
structural loss factor 2; nr is the field incidence nonreso-
nant transmission coefficient mass controlled, r is the
diffuse field resonant transmission coefficient, = nr + r,
and finally is the random incidence absorption coeffi-
cient of the panel seen from the receiving room. Explic-
itly, in terms of modal densities n1 , n2 , n3, damping loss
factors 1 , 2 , 3 and coupling loss factors,35 21 = 23
= rad , 13, Eq. 22 is equivalent to
n2 rad
2
13 +
n1 2
NR = 10 log10 . 23
n n
3 + 1 13 + 2 rad
n3 n3
20 The resonant transmission coefficient31 is calculated from
the radiation efficiency of the panel and its modal density
using
8c20n2 rad
2
r = . 24
A 2
The SEA total transmission loss of the panel including reso-
nant and non resonant contributions is expressed as
TL = NR + 10 log10
A
A3
, 25
where, NR is the noise reduction, A the area of the panel and
21 A3 the absorption area of the receiving room.
A richer model was also studied. The panel was assumed
to be composed of three subsystems corresponding to each of
the first three propagating solutions of the dispersion rela-
tion. It was assumed that each of the three subsystems couple
with the acoustic cavities. As expected, the contribution of
Ghinet, Atalla, and Osman: Curved laminate and sandwich composite panels
 TABLE I. Materials properties for diffuse field transmission loss validations.
Material #1 Material #2
EL Pa 7.1 1010 1.25 1011
ET Pa 7.1 1010 1010
GLT Pa 2.67 1010 5.9 109
GLZ Pa 2.67 1010 3 109
GTZ Pa 2.67 1010 5.9 109
vLT 0.33 0.4
kg/ m3 2700 1600
the first wave solution was found to be dominant. The com-
puted resonant contributions of the other solutions were
found to be unimportant for this air-borne transmission case.
However, it is worth recalling that these propagating solu-
tions resonant subsystems as well as the evanescent com-
ponents are important in structural transmission problems
e.g., plate to plate or plate to beam junction.
VIII. NUMERICAL RESULTS AND VALIDATION
In this section, results of the acoustic transmission prob-
lem applied to curved composite laminate and sandwich pan-
els are presented. The problem is solved in a SEA context,
using a wave approach. An alternate modal approach is also
used to validate the wave approach. The properties of the
materials used in this study are presented in Table I. The
associated notations concerning the orthotropic directions
L , T are presented in Fig. 4.
A. Dispersion curves
In this section, the presented general discrete laminate
model is compared to a discrete sandwich model22 and to a
symmetrical laminate model.12 Recall that symmetrical lami-
nate model assumes each layer thick and uses equivalent
properties including a shear correction factor for the whole
structure smeared physical properties over the total thick-
ness of the panel. On the other hand, the sandwich model is
based on a discrete approach using the classical assumptions
for a laminate sandwich e.g., thin laminate skins, a shear
bearing core. It leads to a 47 order dispersion system. On
the other hand, the presented discrete general laminate model
is also based on a discrete layer approach but allow all layers
skins and core to be thick laminate e.g., with smeared
properties through each layers thickness, orthotropic and
thick. It leads to a 42 order dispersion system for the particu-
lar case of a curved sandwich panel. It allows for both sym-
metrical and asymmetrical laminated composite and/or
FIG. 4. Orthotropic directions of a ply.
J. Acoust. Soc. Am., Vol. 118, No. 2, August 2005 Ghinet, Atalla, and Osman: Curved laminate and sandwich composite panels
Material #3 Material #4 Material #5
0.48 1011 0.1448 109 3.0 107
0.48 1011 0.1448 109 3.0 107
0.181 1011 0.5 108 1.25 107
0.2757 1010 0.5 108 1.25 107
0.2757 1010 0.5 108 1.25 107
0.3 0.2 0.2
1550 110.44 48
sandwich-type composite panels with thin or thick laminate
skins. For the particular case of sandwich composite panels,
the present discrete laminate model and the sandwich
model22 lead to identical results.
The comparison is shown here for a singly curved sand-
wich composite panel. It has a 2 m radius of curvature and a
projected area of 2.43 m 2.03 m. The thickness of the skins
is 1.2 mm and that of the core is 12.7 mm. It is made up of
Graphite/Epoxy Material #3 face sheets and of a rigid foam
core Material #4; the panels orthotropic layout is 0 / 0 / 0.
Figure 5 illustrates the propagating solutions of the disper-
sion system at three selected heading directions: = 0,
= 45, = 90. It is observed that the presented model and
the sandwich model lead to identical results. The dispersion
relation, for this case, is of the 42nd order. In Fig. 5 the five
propagative solutions are represented. As it can be observed
in Fig. 5, below the ring frequency f ring = 401.8 Hz the dis-
persion relation has two propagating solutions. Note that the
ring frequency is the first transition frequency of the panel.
At this frequency a third solution becomes propagating. At
very high frequencies two supplementary core transition fre-
quencies are defined and at these frequencies two supple-
mentary solutions become propagating.
The same behaviors are present for laminate composite
panels. A seven layers graphite/epoxy Material #2 laminate
panel is considered as a second example. The layers have
equal thickness hi = 2 mm and the panel has the orthotropic
layout: 0 / 45/ -45/ 90/ -45/ 45/ 0. Comparisons are made here
between the present discrete lamina model and a simple sym-
metrical laminate model.12 The solutions of the dispersion
relations in these cases are represented in Fig. 6. It is ob-
served that the two models lead to identical results. The core
transition frequencies appear in that case at frequencies
above the audible range and are not represented in Fig. 6.
In order to identify the asymptotic tendencies and the
solution type of the dispersion problem, a three layer isotro-
pic thin curved panel is considered in Fig. 7. The thickness
of each layer Material #1 is 1 mm. The corresponding flat
panel has the classical wave number solutions: bending,
shear, and membrane. These asymptotes are also represented
in Fig. 7 and are calculated using the following relations:
kbending =
ms
D
; kshear = ms
Gh
;
779
FIG. 5. Dispersion curves of a sandwich composite shell for different head- FIG. 6. Dispersion curves of a laminate composite shell for different head-
ing directions: a = 0; b = 45; c = 90. Modeling type: Discrete ing directions: a = 0; b = 45; c = 90. Modeling type: Discrete
laminate composite ; sandwich composite . laminate composite ; symmetrical laminate composite .

kmembrane = ms
1 2
Eh
, 26
wich panel has three different behaviors: pure bending of the
whole panel at low frequencies; shearing of the core at mid-
to-high frequencies, and pure bending of the skins at very
with, ms the surface mass, D the bending stiffness, G the core high frequencies. This is illustrated in Fig. 8. It represents the
shear modulus, and E the Youngs modulus. first solution of the sandwich panel employed for the disper-
The first propagating wave number solution of a sand- sion studies represented in Fig. 5 and its three dispersion

780 J. Acoust. Soc. Am., Vol. 118, No. 2, August 2005 Ghinet, Atalla, and Osman: Curved laminate and sandwich composite panels
 FIG. 8. Dispersion asymptotes of the first propagative solution for a com-
posite sandwich curved panel at a heading angle = 90.

B. Vibroacoustic indicators and experimental

validation
The dispersion results represented in Fig. 5 are used in
this section to compute the associated vibro-acoustic indica-
tors modal density, radiation efficiency, and transmission
loss following the methodology of Secs. VVII. Figure 9
presents the results obtained for the modal density computa-
tion. Three models are compared: the presented general dis-
crete laminate model, the sandwich composite theory,22 and
the laminate panel theory.12 The asymptotic behaviors ob-
served in Fig. 8 are also seen in the modal density curve. It is
observed that at low frequencies below the ring frequencies
all three models lead to the same result. Moreover, at high
frequencies the modal density of a sandwich panel is accu-
rately calculated using a sandwich assumption modeling or a
general discrete laminate modeling. However, the laminate
model,12 which uses equivalent physical properties with
shear corrections, fails at high frequencies; its dispersion
equation has just two correct asymptotes e.g., pure bending

FIG. 7. Dispersion curves of a thin isotropic shell for different heading

directions: a = 0; b = 45; c = 90. Analytical thin plate solu-
tions: Bending ; shear ; membrane .

asymptotes. For a heading direction set to 90, the dispersion

relation has equivalent flat panel behaviors; that is, the same
first propagating solution is obtained for a heading of 90,
using R . FIG. 9. Modal density of a sandwich composite curved panel.

J. Acoust. Soc. Am., Vol. 118, No. 2, August 2005 Ghinet, Atalla, and Osman: Curved laminate and sandwich composite panels 781
 FIG. 10. Radiation efficiency of a sandwich composite curved panel.
of the panel and shear. It is not able to correctly capture the
separate skins asymptotes at high frequencies. The separate
bending of the skins behavior becomes especially important
for the radiation efficiency computation. As it can be ob-
served in Fig. 10, the composite laminate model could result
in large errors in the radiation efficiency estimation for fre-
quencies above the critical frequency zone of the panel.
The proposed general discrete laminate approach uses
individual first-order shear displacement fields for each layer.
While the use of high order displacement fields may seem
more appropriate for very high frequencies, it is worth show-
ing that the proposed approach is sufficient to capture the
physics. An example comparing the modal density of a typi-
cal flat thick sandwich panel modeled by the proposed ap-
proach and a laminate model based on spectral finite
elements14 in which the through-thickness deformation is
captured using finite elements is represented in Fig. 11. Also
the analytical modal density tendencies for low frequencies
pure bending of the panel and high frequencies pure bend-
FIG. 11. Discrete laminate composite modeling validation. Comparison of
the present approach with spectral finite elements prediction.
782 J. Acoust. Soc. Am., Vol. 118, No. 2, August 2005 Ghinet, Atalla, and Osman: Curved laminate and sandwich composite panels
FIG. 12. Contributions of nonresonant and resonant transmissions to the
total transmission loss of the structure.
ing of skins are computed and represented. It can be ob-
served in Fig. 11 that the present approach is accurate. The
studied panel has skins made up of Material #1 0.001 m of
thickness and a core made up of Material #5 0.003 m thick-
ness. The materials physical properties are presented in
Table I. The dimensions of the panel are 2 2.4 m2.
In passing, it is worth noting that the presented general
discrete laminate model can also be refined using a finer
subdivision of the layers to capture complicated behavior
through the thickness.
Finally, the physical properties of the lay-up studied in
Fig. 5 are used to illustrate the contribution of the nonreso-
nant and resonant transmission to the total transmission loss.
This time, the thickness of a skin is 1 mm while the thick-
ness of the core is 3 mm. The lateral dimensions of the panel
are 2 2.4 m2 and the radius of curvature is 2 m. In Fig. 12
are represented the total transmission loss, the resonant, and
the nonresonant contributions. The classical tendencies are
observed: a combined contribution of resonant and nonreso-
nant transmissions below the ring frequency, a nonresonant
contribution between the ring and the critical frequencies,
and a resonant contribution above the critical frequency.
C. Experimental validation
Transmission loss tests were performed on the singly
curved sandwich composite panel described in the previous
section. The tests were performed at the Canadian National
Research Center transmission loss facility located in Ottawa.
The test panel was installed in the opening window of a
transmission loss suite comprising two reverberant cham-
bers. It should be noted that the two chambers are mechani-
cally isolated, and that the test specimen is mounted into a
movable heavy frame. The tests were performed according to
the specifications of ASTM-E90-97 and ISO 140-1:97. Both
chambers are equipped with automated moving microphone
position systems, which allow sampling a large volume of
the rooms. The volumes of the source and receiving rooms
are 140 m3 and 250 m3, respectively. The receiving room is
 FIG. 14. Orthotropic ply angle influence on the transmission loss of a lami-
FIG. 13. Total transmission loss of a sandwich composite curved panel.
nate composite curved panel.

equipped with both stationary and rotating diffusers. TL mea-

influence of physical parameters on the transmission loss.
surements in this facility are considered valid down to the
Orthotropic ply angles, layers thickness and constant mass
100 Hz 1 / 3 octave band. Considerable care was taken in the
per unit area effects are studied. Moreover, this study will
design of the supporting frame for the panel and the method
also allow for a numerical validation of the ring and critical
of mounting the panel. The curved edges were inserted in a
frequencies expressions of the panel as calculated from re-
groove supported by closed cell foam and sealed by caulk.
lations 6 and 10.
The straight edges were also supported by closed cell foam
The transmission loss of a symmetrical laminate sand-
gasket attached to wood chocks and sealed by caulk and
wich composite panel is presented in Fig. 14. Each skin is a
tape.
composite made up from three lamina of equal thickness
Measured transmission loss and predictions with both
hskin = 0.003 m made of Material #2. The core has a single
the wave approach and the modal approach are given in Fig.
layer of Material #4 and thickness hcore = 0.01 m. The ply
13. In the latter method, the modes of the panel computed
angle of the core is 0 but each skin has the following ortho-
from the dispersion equation Eq. 5 with the assumption of
tropic layout: 90/ 45/ 45 for configuration A; 0 / 45/ 45 for
a simply supported panel are used in conjunction with Eqs.
configuration B, 0 / 30/ 30 for configuration C and 0 / 0 / 0 for
20 and 21 to compute the nonresonant transmission loss
configuration D. The computed values of the panels ring and
while the resonant transmission loss is calculated with Eq.
critical frequencies are as follows: f ring = 458.55 Hz; f c1
24. It should be noted at this stage that the mounted panel
= 1938.55 Hz; f c2 = 4586.99 Hz for Configuration A; f ring
damping was not measured, and that a nominal modal damp-
= 334.39 Hz; f c1 = 1920.76 Hz; f c2 = 4587.01 Hz for Configu-
ing ratio of 2.5% was assumed in the analysis. This is ac-
ration B; f ring = 242.61 Hz; f c1 = 1751.74 Hz; f c2
ceptable since the damping of the panel in free-free condi-
= 5056.42 Hz for Configuration C and f ring = 191.66 Hz; f c1
tions is around 1% and increases to 3% around the critical
= 1546.72 Hz; f c2 = 5398.25 Hz for Configuration D. Their
frequency. The wave approach with the geometrical correc-
transmission loss is represented in Fig. 14. It is observed that
tion leads to a very good agreement throughout the fre-
the panel in Configuration A has the shortest region con-
quency range of the test apart from the ring frequency re-
trolled by the mass between the ring frequency and the first
gion. On the other hand, the modal approach shows a higher
critical frequency and consequently will better react to a
transmission loss than the test below the panel ring fre-
damping treatment. A symbolical simplification of the ring
quency. Above the panel critical frequency, both the wave
frequency relation 6 shows that it is a function of the
and modal approaches yield almost identical results. Note
y-direction circumferential elastic behaviors, the radius and
that the number of resonant modes is not sufficient in the first
the surface mass of the panel. For this reason, the orthotropic
1 / 3 octave bands for the modal method to be reliable at low
arrangement with the highest y-direction elongation stiffness
frequencies less than 1 mode at 100 Hz. In summary, the
has the highest ring frequency. Moreover, the width of the
wave approach leads to excellent results, especially at low
critical frequency zone is strongly dependent on the ortho-
frequencies, with the benefit of applying the geometrical cor-
tropic arrangement of the laminas. As an example, for a per-
rection Eq. 18, even if the latter is based on the flat
fectly equilibrate orthotropic arrangement = same number
baffled window theory.
of plies oriented along and the width of the critical
zone tends to zero. The most equilibrate arrangement in Fig.
D. Parametric study
14 is described by configuration A while the reference con-
Laminated composite and sandwich composite curved figuration D shows the largest critical zone that the panel
panels of surface 2.438 2.032 m2 are used here to study the could have.

J. Acoust. Soc. Am., Vol. 118, No. 2, August 2005 Ghinet, Atalla, and Osman: Curved laminate and sandwich composite panels 783
FIG. 15. Skins thickness influence on the transmission loss of a sandwich FIG. 16. Cores thickness influence on the transmission loss of a sandwich
composite curved panel. composite curved panel.

see Case A in Fig. 17. Next, the skins density is multiplied

In Fig. 15 the influence of the skins thickness is illus- by two Case B. The results are plotted in Fig. 17 for the
trated in the context of a sandwich panel. The skins are made two cases to highlight the phenomena appearing in the trans-
of one layer of Material #2 while the core has a single layer mission loss nonresonant mass controlled zone between the
of Material #4. The 0 / 0 / 0 ply angle case is considered with ring and the critical frequencies. In Case A the panel does
three different skin thicknesses: Configuration A with the not have a zone mainly controlled by the nonresonant trans-
skins thickness of h1 = h3 = 0.003 m; Configuration B with mission. The region between the ring frequency and the criti-
h1 = h3 = 0.004 m and Configuration C with h1 = h3 = 0.005 m. cal frequency is just a resonant transition zone from mem-
The core thickness remains unchanged at 0.01 m. Figure 15 brane to bending and shearing behaviors. The transmission
shows that the TL increases with the skins thickness. The loss is mainly controlled by the resonant contributions and
increase is in the order of about 3.5 dB in the low frequency for this reason increasing the core thickness results in a deg-
range and about 7 dB in the high frequencies between Con- radation of the transmission loss in low and mid-frequency
figuration A and C. The ring frequencies and the critical fre- ranges below and inside the critical frequency region.
quencies calculated for these three configurations are: f r Finally, a comparison between three typical configura-
= 191.66 Hz; f c1 = 1543.9 Hz; f c2 = 5393.32 Hz for Configu- tions based on previous comparisons and having the same
ration A, f r = 193.38 Hz; f c1 = 1260.14 Hz; f c2 = 4403.52 Hz overall mass per unit area is presented in Fig. 18. The mass
for Configuration B and f r = 194.41 Hz; f c1 = 1050.87 Hz; per unit area of the sandwich composite shell is kept constant
f c2 = 3673.34 Hz for Configuration C. to ms = 17.1044 kg/ m2. As in the previous studies the same
Next the influence of the core thickness is illustrated in
Fig. 16. Three configurations are considered: h2 = 0.005 m
for Configuration A, h2 = 0.01 m for Configuration B and
h2 = 0.015 m for Configuration C. In all the three cases, the
thickness of the skins is h1 = h3 = 0.004 m. It is observed that
the transmission loss decreases between the ring frequency
and the first critical frequency for configurations B and C
compared to Configuration A.
A detailed study of this case reveals that the mass law of
the three configurations is almost identical while the resonant
transmission below the highest critical frequency controls the
tendencies observed in Fig. 16. The radiation efficiency in
this region is found to increase with the core thickness for
this particular construction. Around the highest critical fre-
quency the influence of the separate bending and shearing of
the skins starts for these configurations. In this region, the
modal density is almost identical for the three configurations
and so is the resonant transmission.
To clarify the results of Fig. 16 the same configurations
are reinvestigated but this time the sandwich panel has skins FIG. 17. Cores thickness influence on the transmission loss of a sandwich
made up of Material #3 and a core made up of Material #4 composite curved panel combined with the influence of the skins density.

784 J. Acoust. Soc. Am., Vol. 118, No. 2, August 2005 Ghinet, Atalla, and Osman: Curved laminate and sandwich composite panels
 applicable to both sandwich panels and composite laminate
panels with thin and/or thick layers. Finally, a parameters
study showed that the transmission loss of such panels can
be improved by a judicious choice of orthotropic arrange-
ments of plies and layers thicknesses.

APPENDIX: MAIN EQUATIONS OF THE GENERAL

DISCRETE LAYER MODEL
1. Equilibrium equations
The resultant stress forces and moments of any laminate
layer are defined by, Leissa:36

Ni i
z huk z
Qix = xz 1+ dz = xz
k
1+ dz,
z R k=1 i
hlk R

Ni i
huk
FIG. 18. Constant mass per unit area influence on the transmission loss of a Qiy = yzdz = kyzdz, A1
sandwich composite curved panel. z k=1 i
hlk

materials are used here. For Configuration A the skins thick- Ni i
z huk z
ness is h1 = h3 = 0.0045 m and the core thickness is h2 Nix = x 1 + dz = kx 1 + dz,
= 0.0245 m; for Configuration B the skins thickness is h1 z R k=1 i
hlk R
= h3 = 0.00475 m and the core thickness is h2 = 0.017244 m

Ni
and for Configuration C the thickness of the skins is h1 = h3 z
i
huk z
= 0.005 m and the core thickness is h2 = 0.01 m. The com- M ix = xz 1 + dz = kxz 1 + dz,
z R k=1 i
hlk R
puted ring and critical frequencies of the three configurations
are: f r = 186.577 Hz; f c1 = 748.677 Hz; f c2 = 2621.572 Hz for

Ni i
Configuration A; f r = 190.5255 Hz; f c1 = 941.73 Hz; f c2 huk
= 3289.65 Hz for Configuration B and f r = 194.41 Hz; f c1 Niy = ydz = kydz,
i
z k=1 hlk
= 1050.87 Hz; f c2 = 3673.35 Hz for Configuration C. The
most interesting configuration Configuration C has the

Ni i
huk
yzdz =
thicker skins and the thinnest core. It could be concluded
M iy = kyzdz,
here that the transmission loss of a sandwich composite shell z k=1 i
hlk
can be improved by a judicious choice of the thicknesses of

the layers while keeping constant the mass per unit area. Ni i
z huk z
Nixy = xy 1+ dz = kxy 1 + dz,
IX. DISCUSSIONS AND CONCLUSIONS z R k=1 i
hlk R

An efficient model to compute the transmission loss of Ni i
z huk z
dz =
sandwich and laminate composite curved panels has been
M ixy = xyz 1 + kxyz 1 + dz,
presented. The physical behavior of the panel is represented z R k=1 i
hlk R
using a discrete lamina description. Each lamina is repre-

sented by membrane, bending, transversal shearing and rota- Ni i
huk
tional inertia behaviors. The model is developed in the con- Niyx = yxdz = kyzdz,
i
text of a wave approach. It is shown that the dispersion z k=1 hlk
curves and the panels ring and critical frequencies are accu-

rately estimated. Using the dispersion relations solutions, Ni i
huk
the modal density, the radiation efficiency as well as the M iyx = yxzdz = kyxzdz, A2
i
resonant and nonresonant transmission loss are calculated. z k=1 hlk
The acoustic transmission problem is represented within sta- i
The integral limits huk and hilk in relations A1 and A2 are
tistical energy analysis context using two different schemes computed using the following relations:
for the nonresonant path. It is observed that for the presented
k k1
problem, the modal energy exchange is dominated by the
first wave solution. It is concluded that for classical acoustic
i
huk = zi1 + hij, hilk = zi1 + hij , A3
j=1 j=0
transmission problems, the SEA scheme presented here is
accurate. The results were compared successfully to the where, hij is the thickness of the lamina j of the layer i hi0
transmission loss test of a singly curved sandwich panel and = 0 and zi1 is the position of the i 1 surface delimiting a
to two other models. In particular, the presented model is layer.

J. Acoust. Soc. Am., Vol. 118, No. 2, August 2005 Ghinet, Atalla, and Osman: Curved laminate and sandwich composite panels 785
 The transverse shear stress forces are defined by the fol-
lowing relations:
M ix = B11u,x + B12 v,y + w
R

+ B16u,y + v,x

Qix = F45 w,y + y
v

+ F55w,x + x + D11 u,x
R

+ x,x + D12y,y

+ H55 w,x x
R
+
R
i
, + D16 v,x
R
+ x,y + y,x
i
,

Qiy = F44 w,y + y
v
R

+ F45w,x + x

M iy = B12u,x + B22 v,y + w

+ B26u,y + v,x

R
+ H44 v y w,y

R2 R

R
i
, A4
+ D12x,x + D22 y,y v,y w

R R2

and the in-plane stress forces,

+ D26 x,y + y,x
u,y
i
,

Nix = A11u,x + A12 v,y + w

R

+ A16u,y + v,x + B11

u,x
R

+ x,x + B12y,y + B16
v,x
R
+ x,y + y,x
M ixy = B16u,x + B26 v,y + w
R

+ B66u,y + v,x

+ D11
x,x
R
+ D16
y,x i
R
, + D16 u,x
R

+ x,x + D26y,y

v,x
i

+ x,y + y,x

+ D66 ,
w R
Niy = A12u,x + A22 v,y + + A26u,y + v,x + B12x,x
R

+ B22 y,y
v,y w
R R

2 + B26 x,y + y,x
u,y
R

M iyx = B16u,x + B26 v,y + w
R

+ B66u,y + v,x

y,y x,y u,y
D22 D26 + D26 2 v,y w
R R R + D16x,x + D26 y,y

R R2

i
v,y w i
+ D22 2 + D22 3 ,
u,y
R R + D66 x,y + y,x . A6
R

Nixy = A16u,x + A26 v,y + w
R

+ A66u,y + v,x The inertial terms derived in the equilibrium equations 2
are expressed by the following relations:

+ B16 u,x

+ x,x + B26y,y + B66

Ni Ni

huk
3
h3lk i
x,x y,x = khuk hlk , = k
i
v,x msi i
Izi ,
+ x,y + y,x + D16 + D66 , k=1 k=1 3
R R R

Niyx = A16u,x + A26 v,y + w
R

+ A66u,y + v,x + B16x,x
i
Iz2
Ni
= k huk
2
h2lk i
, A7

k=1 2
v,y w u,y
+ B26 y,y 2 + B66 x,y + y,x
R R R

where, msi is the mass per unit area, Izi and Iz2
i
y,y x,y u,y v,y w i are the rational
D26 D66 + D66 2 + D26 2 + D26 3 , inertia, and k is the mass density of the lamina k of the layer
i
R R R R R i
i. The rotational inertia Iz2 is zero for symmetrically lami-
A5 nated composite sandwich panels. The elastic constants de-
rived in Eqs. A4A6 are defined by the following rela-
as well as the stress moments, tions:

786 J. Acoust. Soc. Am., Vol. 118, No. 2, August 2005 Ghinet, Atalla, and Osman: Curved laminate and sandwich composite panels

Ni Q22
k i
= CLk sin4 k + CTk cos4 k + 2CLT
k k
+ 2GLT
i
A = Q
k
huk hlk i
k=1 sin2 k cos2 ki ,

i
B
Ni
= Q
k

k=1
2
huk h2lk
2
i
, = 1,2,6 Q26
k i
= CLk CLT
+ CLT
k
k k
2GLT sin3 k cos k
CTk + 2GLT
k
sin k cos3 ki ,

Ni 3
huk h3lk i
i
D = k
Q Q66
k i
= CLk + CTk 2CLT
k k
+ GLT sin2 k cos2 k
k=1 3
k
+ GLT cos4 k + sin4 ki , A9

Ni
with
= C

i
F k
huk hlki
k=1 ELk i
, = 4,5. A8 CLki =

,
Ni 3 3 i 1 LT
k k
TL
k huk hlk
i
=

H C
k=1 2 ETk i
LT
k
ETk i
CTki = , CLT
k i
= . A10
In Eq. A8, Q are the elastic constants of the kth lamina
k i 1 LT
k k
TL 1 LT
k k
TL
of layer i and are defined by the following relations:38
k is the orthotropic orientation represented in Fig. 4 and
Q11
k i
= CLk cos k +
4
CTk sin k 4 C are the transverse shear elastic constants of the kth
k i

lamina of the layer i and are defined by38

+ k
2CLT + k
2GLT sin2 k cos2 ki ,
C44
k i
= GTZ
k
cos2 k + GLZ
k
sin2 ki ,
Q12
k i
= CLk + CTk k
4GLT sin2 k cos k
2
C45
k i
= GLZ
k k
GTZ sin k cos ki , A11
k
+ CLT cos4 k + sin4 ki ,
C55
k i
= GLZ
k
cos2 k + GTZ
k
sin2 ki .
Q16
k i
= CLk CLT
k k
2GLT sin k cos3 k
The dynamic equilibrium equations of the shell are rewritten,
+ CLT
k
CTk + 2GLT
k
sin3 k cos ki , after appropriate algebraic manipulations as

A11 +
B11
R

u,xx + 2A16u,xy + A66
B66
R

u,yy + A16 +
B16
R

v,xx + A12 + A66v,xy + A26
B26
R
v,yy + B11 +
D11
R
x,xx
+ 2B16x,xy + B66 D66
R

x,yy + B16 +
D16
R
y,xx + B12 + B66y,xy + B26
D26
R

y,yy +
A12
R
w,x +
A26 B26
R
2 w,y
R
i

+ Fix Fi1
x + ms +
Iz2
R

u+
Iz
R
i
+ Iz2 x 2 = 0,

A16 +
B16
R

u,xx + A12 + A66u,xy + A26
B26
R
u,yy + A66 +
B66
R
v,xx + 2A26v,xy + A22
B22
R
v,yy + B16 +
D16
R
x,xx
+ B12 + B66x,xy + B26 D26
R
x,yy + B66 +
D66
R
y,xx + 2B26y,xy + B22
D22
R

y,yy +
A26 F45
R
+
R
w,x

+ A22 B22 F44 H44

R
2 +
R R
2 w,y
R
F44 H44
R2

3 v+
R
F45
R

x +
F44 H44
R
2 y
R

i
+ Fiy Fi1
y + ms +
Iz2
R
v+
Iz
R
i
+ Iz2 y 2

= 0,

F55 +
H55
R

w,xx + 2F45w,xy + F44
H44
R
w,yy
A12
R
u,x
A26 B26
R
2 u,y
R

A26 F45
R
+
R
v,x R
2 +
R

A22 B22 F44 H44
R
2 v,y
R

B12
R
F55
H55
R
x,x
R

B26 D26
2 F45 x,y
R
B26
R
F45 y,x
B22 D22
R

2 F44 +
R
H44
R

y,y
A22 B22
R
3 w
R
i

+ Fzi Fzi1 + mswi2 = 0,

J. Acoust. Soc. Am., Vol. 118, No. 2, August 2005 Ghinet, Atalla, and Osman: Curved laminate and sandwich composite panels 787
 B11 +
D11
R

u,xx + 2B16u,xy + B66
D66
R

u,yy + B16 +
D16
R

v,xx + B12 + B66v,xy + B26
D26
R

v,yy + D11x,xx + D16x,xy
+ D66x,yy + D16y,xx + D12 + D66y,xy + D26y,yy + B12
R
F55
H55
R
w,x +
R

B26 D26
2 F45 w,y +
R
F45
R
v

F55 +
H55
R

x F45y i
x + Iz x +
+ ziFix zi1Fi1 u
R
i
+ Iz2u 2 = 0,

B16 +
D16
R

u,xx + B12 + B66u,xy + B26
D26
R
u,yy + B66 +
D66
R

v,xx + 2B26v,xy + B22
D22
R
v,yy + D16x,xx
+ D12 + D66x,xy + D26x,yy + D66y,xx + 2D26y,xy + D22y,yy + B26
R
F45 w,x
B22 D22
R
2
R

F44 +
H44
R
w,y +
F44 H44
R
2 v F45x F44
R
H44
R
y i

y + Iz y +
+ ziFiy zi1Fi1
v
R
i
+ Iz2v 2 = 0. A12

2. Dispersion equation matrices

The matrices A0, A1, A2 used in Eq. 5 are real square matrices of dimension 5N + 3N 1 defined as follows:

A01 0 0 0 0 0 F01 0 0 0 0 0
A0 2
0 0 0 0 F01 F02 0 0 0 0
A03 0 0 0 0 F02 F03 0 0 0

0 0 0 0 0
0 0
A0 N1
0 0 0 0 0 F0 N2
F0N1
A0 N
0 0 0 0 0 F0N1
A0 =
0 0 0 0 0 0
0 0 0 0 0
sym 0 0 0 0
0 0 0
0 0
0

A11 0 0 0 0 A21 0 0 0 0
0 A1 1
0 0 0 0 A2 2
0 0 0
A1 = 0 0
0 0 ; A2 = 0 0
0 0 ; A13
0 0 0 A1 N
0 0 0 0 A2 N
0
0 0 0 0 0 0 0 0 0 0
where

i
a11 0 0 a14 0 1 0 0
0 a22 0 a24 a25 0 1 0
A0 = i
0 0 a33 0 0 ; F0 = i
0 0 1 ;
a14 a24 0 a44 a45 zi 0 0
0 a25 0 a45 a55 0 zi 0

0 0 13 0 0 i
11 12 0 14 15 i

0 0 23 0 0 12 22 0 24 25
A1 = 13 23
i
0 34 35 ; A2 = 0
i
0 33 0 0 ; A14
0 0 34 0 0 14 24 0 44 45
0 0 35 0 0 15 25 0 45 55

788 J. Acoust. Soc. Am., Vol. 118, No. 2, August 2005 Ghinet, Atalla, and Osman: Curved laminate and sandwich composite panels
with coefficients ai , i , and i defined as follows:
D16
i

15
i
= B16 + cos2 + B12 + B66i cos sin
Iz2 2 Iz R

i
a11 = ms + ; i
a14 = + Iz2 2; i
R R D26
+ B26 sin2 ,

R
Iz2 2 F44 H44
i
2 + 3;

a22 = ms +
R R R B66 i
22
i
= A66 + cos2 + 2A26
i
cos sin

F45 F44 H44 Iz
i
a24 = ; i
a25 = 2 + + Iz2 2; B22 i
R R R R + A22 sin2 ,
R
i
a33
= ms +
R R

Iz2 2 A22 B22
2 + 3;
R
24
i

= B16 +
D16
i
cos2 + B12 + B66i cos sin
R

H55
i
a44 = Iz2 F55 ; A15 D26 i
R + B26 sin2 ,
R

H44
i
a45 = F45 ; i
a55 = Iz2 F44 ; D66 i
R 25
i
= B66 + cos2 + 2B26
i
cos sin
R
i
a13 = A12
R
i
cos A26 B26
R
2
R
i
sin ,
+ B22
D22
R
i
sin2 ,

23
i
= A22 B22 F44 H44
R
2 +
R R
2
R
i

33
i

= F55 +
H55
R
i
cos2 + 2F45
i
cos sin

sin A26 F45

R
+
R
i
cos ,
+ F44
H44
R
i
sin2 ,

34
i
= B12
R
F55
H55
R
i
cos 44
i i
= D11 cos2 + 2D16
i
cos sin + D66
i
sin2 ,

+ B26 D26
R
i
2 F45 sin ,
R
45
i i
= D16 cos2 + D12 + D66i cos sin + D26
i
sin2 ,

55
i i
= D66 cos2 + 2D26
i
cos sin + D22
i
sin2 . A17

35
i
= B22 D22
R
2 F44 +
R
H44
R
i
sin 1
R. S. Langley, The modal density of anisotropic structural components,

i
J. Acoust. Soc. Am. 996, 34813487 1996.
B26 2
L. A. Roussos, C. A. Powell, F. W. Grosveld, and L. R. Koval, Noise
+ F45 cos , A16 Transmission Characteristics of advanced composite structural materials,
R
J. Aircr. 21, 528535 1984.
3
and J. P. D. Wilkinson, Modal densities of certain shallow structural ele-

ments, J. Acoust. Soc. Am. 43, 245251 1968.
i 4
B11 L. R. Koval, On sound transmission into an orthotropic shell, J. Sound
11
i
= A11 + cos2 + 2A16
i
cos sin Vib. 631, 5159 1979.
R 5
H. C. Nelson, B. Zapatowski, and M. Bernstein, Vibration analysis of

+ A66
B66
R
i
sin2 ,
6
orthogonally stiffened circular fuselage and comparison with experiment,
in Proceedings of the Institute of Aeronautical Sciences National Special-
ists Meeting on Dynamics and Aeroelasticity, 1958, pp. 7787.
L. R. Koval, Sound transmission into a laminated composite cylindrical

12
i

= A16 +
B16
R
i
cos2 + A12 + A66i cos sin
7
shell, J. Sound Vib. 714, 523530 1980.
C. W. Bert, J. L. Baker, and D. M. Egle, Free vibrations of multilayer
anisotropic cylindrical shells, J. Compos. Mater. 3, 480499 1969.

8
A. Blaise and C. Lesueur, Acoustic transmission through a 2D orthotro-
i
B26 pic multilayered infinite cylindrical shell, J. Sound Vib. 1551, 95109
+ A26 sin2 , 1992.
R 9
S. B. Dong, Free vibration of laminated orthotropic cylindrical shells, J.

Acoust. Soc. Am. 44, 16281635 1968.
i 10
D11 S. Ghinet and N. Atalla, Vibro-acoustic behavior of multilayer orthotro-
14
i
= B11 + cos2 + 2B16
i
cos sin pic panels, Can. Acoust. 30, 7273 2002.
R 11
Y. X. Zhang and K. S. Kim, Two simple and efficient displacement-based

+ B66
D66
R
i
sin2 , 12
quadrilateral elements for the analysis of composite laminate plates, Int.
J. Numer. Methods Eng. 61, 17711796 2004.
S. Ghinet, N. Atalla, and H. Osman, Diffuse field transmission into infi-

J. Acoust. Soc. Am., Vol. 118, No. 2, August 2005 Ghinet, Atalla, and Osman: Curved laminate and sandwich composite panels 789

nite sandwich composite and laminate composite cylinders, J. Sound Vib.
to be published.
13
A. Blaise and C. Lesueur, Acoustic transmission through a 3D ortho-
tropic multilayered infinite cylindrical shell, Part I: Formulation of the
problem, J. Sound Vib. 1715, 651664 1994.
14
P. J. Shorter, Wave propagation and damping in linear viscoelastic lami-
nates, J. Acoust. Soc. Am. 115, 19171925 2004.
15
L. L. Erickson, Modal densities of sandwich panels: theory and experi-
ment, The Shock and Vibration Bulletin 393, 116 1969.
16
B. L. Clarkson and M. F. Ranky, Modal density of honeycomb plates, J.
Sound Vib. 91, 103118 1983.
17
K. Renji, P. S. Nair, and S. Narayanan, Modal density of composite
honeycomb sandwich panels, J. Sound Vib. 1955, 687699 1996.
18
G. Kurtze and B. G. Watters, New wall design for high transmission loss
or high damping, J. Acoust. Soc. Am. 316, 739748 1959.
19
M. A. Lang and C. L. Dym, Optimal acoustic design of sandwich pan-
els, J. Acoust. Soc. Am. 576, 14811487 1975. 34
20
E. H. Baker and G. Herrmann, Vibrations of orthotropic cylindrical sand-
wich shells under initial stress, AIAA J. 4, 10631070 1966.
21
E. Nilsson and A. C. Nilsson, Prediction and measurement of some dy-
namic properties of sandwich structures with honeycomb and foam cores,
J. Sound Vib. 2513, 409430 2002.
22
K. H. Heron, Curved laminates and sandwich panels within predictive
SEA, in Proceedings of the Second International AutoSEA Users Confer-
ence, 2002, Detroit, USA.
23
R. Panneton and N. Atalla, Numerical prediction of sound transmission
through finite multilayer systems with poroelastic materials, J. Acoust.
Soc. Am. 100, 346353 1996.
24
F. Sgard, N. Atalla, and J. Nicolas, A numerical model for the low fre-
quency diffuse field sound transmission loss of double-wall sound barriers
with elastic porous linings, J. Acoust. Soc. Am. 1086, 28652872
2000.
25
J. N. Pinder and F. J. Fahy, A method for assessing noise reduction
provided by cylinders, in Proceedings of the Institute of Acoustics,
1993, 195205, Vol. 15, Part 3.
26
M. Villot, C. Guigou, and L. Gagliardini, Predicting the acoustical radia-
tion of finite size multilayered structures by applying spatial windowing
on infinite structures, J. Sound Vib. 2453, 433455 2001.
790 J. Acoust. Soc. Am., Vol. 118, No. 2, August 2005 Ghinet, Atalla, and Osman: Curved laminate and sandwich composite panels
27
S. Ghinet and N. Atalla, Sound transmission loss of insulating complex
structures, Can. Acoust. 29, 2627 2001.
28
F. G. Leppington, K. H. Heron, and E. G. Broadbent, Resonant and
nonresonant noise through complex plates, Proc. R. Soc. London, Ser. A
458, 683704 2002.
29
E. Szechenyi, Modal densities and radiation efficiencies of unstiffened
cylinders using statistical methods, J. Sound Vib. 19, 6581 1971.
30
E. Szechenyi, Sound transmission through cylinder walls using statistical
considerations, J. Sound Vib. 19, 8394 1971.
31
L. D. Pope and J. F. Wilby, Band limited power flow into enclosures, J.
Acoust. Soc. Am. 62, 906911 1977.
32
L. D. Pope and J. F. Wilby, Band limited power flow into enclosures. II,
J. Acoust. Soc. Am. 67, 823826 1980.
33
L. D. Pope, D. C. Rennison, C. M. Willis, and W. H. Mayes, Develop-
ment and validation of preliminary analytical models for aircraft interior
noise prediction, J. Sound Vib. 824, 541575 1982.
F. G. Leppington, E. G. Broadbent, and K. H. Heron, The acoustic radia-
tion efficiency from rectangular panels, Proc. R. Soc. London, Ser. A
382, 245271 1982.
35
C. Lesueur, Rayonnement acoustique des structuresVibroacoustique,
Interactions fluide-structure in French Acoustical radiation of
StructuresVibroacoustics, Interactions Fluid-Structure Editions Ey-
rolles, Paris, 1988.
36
A. W. Leissa, Vibration of Shells, NASA SP 288 U.S. Government Print-
ing Office, Washington, D.C., 1973.
37
H. Osman, N. Atalla, Y. Atalla, and R. Panneton, Effects of acoustic
blankets on the insertion loss of a composite sandwich cylinder, Tenth
International Congress on Sound and Vibration, ICSV 10, Stockholm,
Sweden, July 2003.
38
J.-M. Berthelot, Composite Materials, Mechanical Behavior and Struc-
tural Analysis Springer-Verlag, New York, 1999.
39
N. Atalla, S. Ghinet, and H. Osman, Transmission loss of curved com-
posite panels with acoustic materials, in Proceedings of the 18th Interna-
tional Congress on Acoustics ICA, Kyoto, 2004.
40
Z. C. Xi, G. R. Liu, K. Y. Lam, and H. M. Shang, Dispersion and char-
acteristic surfaces of waves in laminated composite circular cylindrical
shells, J. Acoust. Soc. Am. 1085, 21792186 2000.