Development of BensonMayers theory on the wrinkling of

anisotropic sandwich panels

B.K. Hadi
a,
*
, F.L. Matthews
b
a
Department of Aeronautics and Astronautics, Bandung Institute of Technology, Jl. Ganesha 10, Bandung 40132, Indonesia
b
Centre for Composite Materials, Imperial College of Science, Technology and Medicine, Prince Consort Road, London SW7 2BY, UK
Abstract
The wrinkling analysis of anisotropic sandwich panels is found by developing BensonMayers unied theory for isotropic
sandwich panels into general anisotropic sandwich panels. Both symmetrical and antisymmetrical wrinkling are analyzed and
calculated simultaneously. The present method has been applied to orthotropic and antisymmetric cross-ply sandwich plates and the
results are in good agreement with the published analytical and experimental results. 2000 Published by Elsevier Science Ltd.
Keywords: Wrinkling; Anisotropic; Sandwich panels
1. Introduction
The wrinkling phenomenon is a form of stability as-
sociated with short wavelengths of the faces. The wrin-
kling may be either symmetrical (out of phase) or
antisymmetrical (in phase) with respect to the center
plane of the core. Many theoretical papers concerning
wrinkling of sandwich constructions have been pub-
lished from about 1940 onward, and most of them lead
to substantially the same result [1]. Some of these earlier
works were presented and compiled by Plantema [1] and
Allen [2]. Ho and Mautner [3] is one of these classic
works. They discussed symmetric and skew (antisym-
metric) wrinkling. They used predetermined shapes for
both symmetric and skew ripple. The agreement be-
tween the theoretical and the experimental buckling
stresses was reasonable.
Another approach in studying wrinkling was to de-
velop a general theory that unied overall buckling and
wrinkling. This is because overall buckling is essentially
a special case of wrinkling, in which the half wavelength
of the buckling mode is equal to the length of the plate.
In fact, the theory provided a link between overall in-
stability on one hand and wrinkling behavior on the
other.
Such an approach was used by Williams, Leggett and
Hopkins in 1941 (Allen [2, pp. 1819]). They included
transverse normal behavior of the core by choosing
suitable stress functions for the core. Benson and Ma-
yers [4] developed a general theory that was able to
calculate overall buckling and wrinkling simultaneously.
The faces were isotropic and the core was orthotropic.
They included normal transverse deformations in the
core. Using straindisplacement and equilibrium equa-
tions, the displacements of the core in transverse
(through the thickness) directions could be calculated.
The displacement continuity between the faces and the
core was obtained by introducing Lagrange multipliers.
Both antisymmetric and symmetric wrinklings were
analyzed. Overall buckling was a special case of anti-
symmetric wrinkling when the half-wavelength was
equal to unity.
Pearce and Webber [5] analyzed overall buckling and
wrinkling of sandwich panels with laminated faces. In
this paper, they dealt only with orthotropic faces. They
separated antisymmetric and symmetric wrinkling mode
and calculated them separately. Numerical results were
given for both overall buckling and wrinkling loads.
Webber et al. [6] developed the theory further into the
wrinkling of honeycomb sandwich columns with un-
balanced laminated cross-ply faces, in which the cou-
pling terms B
11
and B
22
existed. They included also the
adhesive layers and considered it an extra laminate.
They found that wrinkling loads increased substantially
by including the adhesive layers.
The analysis developed in this chapter is a develop-
ment of unied theory of Benson and Mayers [4] into the
www.elsevier.com/locate/compstruct
Composite Structures 49 (2000) 425434
*
Corresponding author.
0263-8223/00/$ - see front matter 2000 Published by Elsevier Science Ltd.
PII: S 0 2 6 3 - 8 2 2 3 ( 0 0 ) 0 0 0 7 7 - 5
buckling of anisotropic sandwich panels. The analysis
will calculate symmetrical and antisymmetrical buckling
simultaneously and thus provides results for overall
buckling and wrinkling. The results of this analysis will
be compared to other methods available in the litera-
ture.
2. Strain energies terms
Benson and Mayers [4] regarded a three-dimensional
sandwich structure as shown in Fig. 1 as a two-dimen-
sional problem by formulating permissible displacement
across the thickness. They utilized a face displacement
description rather than one in terms of middle surface
displacement. This innovation, together with the satis-
faction of continuity requirements between core and
faces, leads to the generation of a two-dimensional
theory in the form of energy equations, which can be
solved by the RaleighRitz method.
The wrinkling phenomenon is associated with thin
faces and thick and soft core. Therefore, in this formu-
lation, we neglect the shear deformation terms of the
faces and bending rigidities of the core.
By separating antisymmetric and symmetric compo-
nents as shown in Fig. 1, Benson and Mayers were able
to calculate the two buckling modes separately. In the
present analysis, we use a face displacement description
as in Fig. 1, but we will not separate antisymmetric and
symmetric components. In this case, the antisymmetric
and symmetric buckling will be calculated simulta-
neously.
By using face displacements in Fig. 1, the strain en-
ergy of upper face, core and lower face can be calculated
as follows.
2.1. Contribution of the faces
The strain energy of the faces due to bending and
membrane strain is
U
fi
=
1
2
_
a
0
_
b
0
e
xoi

T
A [ [ e
xoi

_
2 e
xoi

T
B [ [ j
xi

j
xi

T
D [ [ j
xi

_
dx dy; (1)
where
e
xoi
=
oU
i
=ox
oV
i
=oy
oU
i
=oy oV
i
=ox
_

_
_

_
and
j
x
=
o
2
W
i
=ox
2
o
2
W
i
=oy
2
2o
2
W
i
=oxoy
_

_
_

_
; (2)
and i = 1; 2 are upper and lower faces, respectively. U
1
,
V
1
, W
1
and U
2
, V
2
, W
2
are displacements of upper face
and lower face, respectively.
2.2. Contribution of the core
We assume an anti-plane core, in which the stiness
in the x and y directions is neglected. So the core strain
energies are represented by energies due to shear strains
and normal strains only.
Shear strain
U
C1
=
h
2
_ __
a
0
_
b
0
G
x
c
2
xz
_
G
y
c
2
yz
_
dx dy: (3)
Normal strain
dU
C2
=
E
z
2
_ _
e
2
z
dx dy dz: (4)
First, the distribution of e
z
across the depth of the
core should be calculated and is given by Benson and
Mayers [4]
e
z
= f (x; y; z): (5)
The core stiness in x and y directions is neglected, so
r
x
= 0; r
y
= 0; s
xy
= 0; (6)
and the rest of the stressstrain relations of the core are
s
xz
= G
x
c
xz
; s
yz
= G
y
c
yz
; r
z
= E
z
e
z
: (7a)
The straindisplacement relations are then given by
c
xz
=
oU
oz

oW
ox
; c
yz
=
oV
oz

oW
oy
; e
z
=
oW
oz
; (7b)
Fig. 1. Geometry, coordinate system and face displacements of at
sandwich plates [4]. In the original BensonMayers model, the upper
and lower faces were isotropic and have the same thickness, that is
t
1
= t
2
= t.
426 B.K. Hadi, F.L. Matthews / Composite Structures 49 (2000) 425434
and the equilibrium equations are
os
xz
oz
= 0;
os
yz
oz
= 0;
os
xz
ox

os
yz
oy

or
z
oz
= 0: (7c)
Integrating the rst two equations of Eq. (7c), together
with Eq. (7a) leads to the distributions of c
xz
and c
yz
as
follows,
c
xz
= c
x
(x; y);
c
yz
= c
y
(x; y);
(8)
and the normal strain distribution is given by
e
z
=
z
E
z
G
xz
oc
x
ox
_
G
yz
oc
y
oy
_
g(x; y): (9)
Substituting Eq. (9) into Eq. (7b), we get the displace-
ment functions through the thickness of the core as
follows:
W =
z
2
2E
z
G
xz
oc
xz
ox
_
G
yz
oc
yz
oy
_
zg(x; y) W (x; y) (10)
where g(x; y) and W (x; y) are unknown functions.
Equating core and face displacement at the interfaces,
we get
W
1
=W
z=h=2
=
h
2
8E
z
G
x
oc
xz
ox
_
G
y
oc
yz
oy
_

h
2
g(x; y) W (x; y) (11)
and
W
2
=W
z=h=2
=
h
2
8E
z
G
x
oc
xz
ox
_
G
y
oc
yz
oy
_

h
2
g(x; y) W (x; y): (12)
From Eqs. (11) and (12), we get
W
1
W
2
= hg(x; y): (13)
Then
g(x; y) =
W
1
W
2
h
(14)
and
W
1
W
2
=
h
2
4E
z
G
x
oc
xz
ox
_
G
y
oc
yz
oy
_
2W (x; y): (15)
Therefore,
W (x; y) =
W
1
W
2
2
_ _

h
2
8E
z
G
x
oc
xz
ox
_
G
y
oc
yz
oy
_
: (16)
Inserting Eq. (14) into Eq. (9) we get the distribution of
e
z
across the core thickness as follows:
e
z
=
z
E
z
G
xz
oc
x
ox
_
G
yz
oc
y
oy
_

W
1
W
2
h
_ _
: (17)
Inserting Eq. (17) into the strain energy equation of Eq.
(4) we get the following equation:
U
C2
( ) =
h
3
24E
z
_
a
0
_
b
0
G
x
oc
x
ox
_
G
y
oc
y
oy
_
2
dx dy

2E
z
h
_
a
0
_
b
0
W
1
( W
2
)
2
dx dy: (18)
The rst term of Eq. (18) above is the contribution of
the shear strain and the second term is the normal
strain.
Inserting Eqs. (14) and (16) into Eq. (10) we get the
distribution of W across the core thickness as follows:
W =
1
8E
z
h
2
_
4z
2
_
G
x
oc
xz
ox
_
G
y
oc
yz
oy
_

W
1
W
2
2
_ _

z
h
W
1
( W
2
): (19)
Knowing the distribution of W across the core thick-
ness, we can get the distribution of displacements U and
V across the core thickness by inserting Eq. (19) into Eq.
(7b). The results are:
U = zc
x

1
24E
z
3h
2
z
_
4z
3
_
o
ox
G
x
oc
x
ox
_
G
y
oc
y
oy
_
z
o
ox
W
1
W
2
2
_ _

1
2
z
2
h
o
ox
W
1
( W
2
) U
0
(x; y)
(20a)
and
V = zc
y

1
24E
z
3h
2
z
_
4z
3
_
o
oy
G
x
oc
x
ox
_
G
y
oc
y
oy
_
z
o
oy
W
1
W
2
2
_ _

1
2
z
2
h
o
oy
W
1
( W
2
) V
0
(x; y);
(20b)
where U
0
(x; y) and V
0
(x; y) are displacements U and V in
the (x; y; 0) plane, respectively.
The displacements' continuity in the interfaces be-
tween the core and the upper and lower faces should be
maintained. Thus the following relations occur,
U
1

t
1
2
oW
1
ox
= U
z=h=2
; U
2

t
1
2
oW
2
ox
= U
z=h=2
;
V
1

t
1
2
oW
1
oy
= V
z=h=2
; V
2

t
2
2
oW
2
oy
= V
z=h=2
:
(21)
Inserting Eqs. (20a) and (20b) into the right-hand side of
Eq. (21), and eliminating U
0
(x, y) and V
0
(x, y) from
those equations, we get two independent parameters /
x
and /
y
in the form of
/
x
= U
1
( U
2
)
t
1
2
oW
1
ox
_

t
2
2
oW
2
ox
_

h
3
12E
z
G
x
o
2
c
x
ox
2
_
G
y
o
2
c
y
oxoy
_

h
2
oW
1
ox
_

oW
2
ox
_
hc
x
B.K. Hadi, F.L. Matthews / Composite Structures 49 (2000) 425434 427
= U
1
( U
2
) h
1
oW
1
ox
h
2
oW
2
ox

h
3
12E
z
G
x
o
2
c
x
ox
2
_
G
y
o
2
c
y
oxoy
_
hc
x
; (22a)
/
y
= V
1
( V
2
) h
1
oW
1
oy
h
2
oW
2
oy

h
3
12E
z
G
x
o
2
c
x
oxoy
_
G
y
o
2
c
y
oy
2
_
hc
y
; (22b)
where
h
1
=
h t
1
2
; h
2
=
h t
2
2
: (22c)
Eqs. (22a) and (22b) introduces two degrees of coupling
of the parameters, and, furthermore, are not in integral
form. An uncoupling is achieved by introducing Lag-
range multipliers, k
x
and k
y
. The Lagrange multipliers
are arbitrary functions of x and y. So, the energies due
to /
x
and /
y
are:
U
a
=
_
a
0
_
b
0
k
x
/
x
_
k
y
/
y
_
dy dx: (23)
External energies. The external work done by the in-
plane normal and shear loads is
V
e
=
1
4
_
a
0
_
b
0
N
x
W
1;x
( )
2
_
N
y
W
1;y
_ _
2
2N
xy
W
1;x
W
1;y
N
x
W
2;x
( )
2
N
y
W
2;y
_ _
2
2N
xy
W
2;x
W
2;y
_
dy dx: (24)
Total energies. The total energy of the system is then
given by
P = U
f1
U
f2
U
C1
U
C2
U
a
V
e
: (25)
We will analyze a simply supported plate. The assumed
displacements that satisfy the essential boundary con-
ditions are given as follows:
U
1
=

m=1

n=1
C
1
mn
cosa
m
x sinb
n
y;
U
2
=

m=1

n=1
C
2
mn
cosa
m
x sinb
n
y;
V
1
=

m=1

n=1
C
3
mn
sina
m
x cosb
n
y;
V
2
=

m=1

n=1
C
4
mn
sina
m
x cosb
n
y;
W
1
=

m=1

n=1
C
5
mn
sina
m
x sinb
n
y;
W
2
=

m=1

n=1
C
6
mn
sina
m
x sinb
n
y;
(26)
c
x
=

m=1

n=1
C
7
mn
cosa
m
x sinb
n
y;
c
y
=

m=1

n=1
C
8
mn
sina
m
x cosb
n
y;
k
x
=

m=1

n=1
C
9
mn
cosa
m
x sinb
n
y;
k
y
=

m=1

n=1
C
10
mn
sina
m
x cosb
n
y;
where
a
m
=
mp
a
; b
n
=
np
b
; (27)
where m and n are half-wavelength integers and C
1
mn

C
10
mn
are undetermined coecients. Because the present
analysis calculates symmetrical and antisymmetrical
buckling simultaneously Eq. (26) has 10 undetermined
coecients, compared to seven undetermined coe-
cients of Benson and Mayers [4].
The natural boundary conditions are not fully satis-
ed by Eq. (26) because of the presence of the coupling
terms such as A
(i)
16
; A
(i)
26
; B
(i)
16
; B
(i)
26
; D
(i)
16
and D
(i)
26
. By retaining
a suciently large number of terms m, n, the short-
coming of not satisfying the natural boundary condi-
tions in full can be compensated [7,8].
Taking variations of the total potential energy ex-
pression, Eq. (25), with respect to the basic undeter-
mined coecients C
1
mn
C
10
mn
will give 10 simultaneous
equations as follows:
oP
oC
1
mn
= 0;
oP
oC
2
mn
= 0;
oP
oC
3
mn
= 0;
oP
oC
4
mn
= 0;
oP
oC
5
mn
= 0;
oP
oC
6
mn
= 0;
oP
oC
7
mn
= 0;
oP
oC
8
mn
= 0;
oP
oC
9
mn
= 0;
oP
oC
10
mn
= 0:
(28)
Taking the rst term of Eq. (28), we get
For
oP
oC
1
mn
= 0
C
1
mn
A
(1)
11
a
2
m
_
A
(1)
66
b
2
n
_
C
3
mn
a
m
b
n
A
(1)
12
_
A
(1)
66
_
C
5
mn
B
(1)
12
_ _
2B
(1)
66
_
a
m
b
2
n
B
(1)
11
a
3
m
_
C
9
mn

p=1

q=1
_
A
(1)
16
a
m
b
q
L
mn
_
b
n
a
p
L
pq
_
C
1
pq
A
(1)
16
a
m
a
p
L
mn
_
A
(1)
26
b
n
b
q
L
pq
_
C
3
pq
428 B.K. Hadi, F.L. Matthews / Composite Structures 49 (2000) 425434
B
(1)
16
a
2
p
_ _
B
(1)
26
b
2
q
_
b
n
L
pq
2B
(1)
16
a
m
a
p
b
q
L
mn
_
C
5
pq
_
= 0; (29)
with
L
mn
=
16 mn
p
2
m
2
p
2
( ) n
2
q
2
( )
;
L
pq
=
16 mn
p
2
m
2
p
2
( ) n
2
q
2
( )
;
(30)
and m, n, p, and q are chosen such that m p and n q
are odd integers. The rest of Eq. (28) can be calculated in
the same manner and given in [11].
After expanding for a given range of values of indices
m and n, Eq. (28) give rise to a set of 10mn homogeneous
algebraic equations in C
i
mn
; i = 1; 2; . . . ; 10. The stability
criterion of a sandwich plate reduces to the generalized
eigenvalue problem of the type [P[C = N
x
[Q[C,
where [P] is a square matrix with size 10mn. In the
computational procedure, the values of C
1
mn
C
4
mn
and
C
7
mn
C
10
mn
from Eq. (28) are solved in terms of C
5
mn
and
C
6
mn
. In the present investigation, the Numerical Recipe's
subroutine [9] was used to solved the eigenvalue prob-
lem.
3. Results and discussions
The present analysis will be tested and compared with
analytical and experimental results of sandwich plates
and columns of Pearce and Webber [5] for plates and
Webber et al. [6] for columns. The overall buckling re-
sults of the present analysis will also be compared with
the overall buckling results of Rao [7] and Kim and
Hong [8].
3.1. Wrinkling of honeycomb sandwich columns
Comparison will be made with the analytical and
experimental results of Webber et al. [6]. The mechanical
properties of the faces and core are presented in Table 1,
the column geometry in Table 2 and the results in Table
3.
Table 3 shows that the present analysis is in a good
agreement with the analysis of Webber et al. [6]. Table 3
also shows that the eect of the adhesive layer in the
analysis of wrinkling of sandwich columns is important.
In the present analysis, as well as Webber, the adhesive
layer is treated as an additional layer in the faces with a
thickness of 0.25 mm and mechanical properties as
shown in Table 2. Table 3 shows that the predicted
wrinkling loads increase with the addition of adhesive
layers.
Fig. 2 shows the critical loads versus number of half
waves for aluminium faces and CFRP faces sandwich
columns respectively. Both gures show that symmetri-
cal and antisymmetrical wrinkling loads might occur far
below its overall buckling loads. Overall, buckling loads
are a special case for antisymmetric wrinkling when the
number of half waves (m) is equal to 1. The symmetric
buckling load values are huge for m = 1 (not shown in
Fig. 2).
Table 1
Mechanical properties of sandwich materials
CFRP E
x
= 142; 000 N/mm
2
; E
y
= 9800 N/mm
2
;
G
xy
= 4300 N/mm
2
; v
xy
= 0:34
Al alloy E = 70; 000 N/mm
2
; v = 0:3
Honeycomb
core
E
c
= 109 N/mm
2
; G
xz
= 26:6 N/mm
2
; Density
32 kg/m
3
; cell diameter 5 mm
Adhesive E = 3050 N/mm
2
; v = 0:3
Table 2
Sandwich column geometry
No. Face Length (mm) Core thickness (mm) Width (mm) Face thickness (mm) Lay-up
1 Al 600 75 80 0.325
2 CFRP 600 75 80 0.30 0/90
3 CFRP 400 25 60 0.30 0/90
Table 3
Experimental and theoretical failure of sandwich columns (critical loads in N/mm)
Column number (see Table 2) 1 2 3
Webber Present Webber Present Webber Present
Experimental failure load 183 111 143
Theoretical symmetry 117 116.4 88 89.5 154 154.9
Wrinkling load (102) (101) (59) (58.8) (103) (101.8)
Theoretical antisymmetry 199 198.1 151 153.1 246 247
Wrinkling load (174) (172) (102) (101) (166) (166.7)
( ): Adhesive not included.
B.K. Hadi, F.L. Matthews / Composite Structures 49 (2000) 425434 429
This gure shows that initially the overall buckling
loads increase as the number of m increases, as in the
case of Euler buckling. But, as m increases, the shear
and normal stiness of the core become dominant, and
the buckling loads decrease to a minimum. After
achieving minimum values, the buckling loads increase
again. Those minimum values are called symmetrical
and antisymmetrical wrinkling loads. Because of these
phenomena, it is necessary to increase the values of m in
the buckling analysis of sandwich structures to investi-
gate whether wrinkling will be dominant. As shown in
Fig. 2, in this case, wrinkling can be a more dominant
source of failure than overall buckling. Fig. 2 also shows
that the eect of adhesive layer is to increase the wrin-
kling loads since it increases the eective face thickness.
Note that in the present analysis, the adhesive layer is
considered as an additional layer in the faces.
Ho [3], Plantema [1] and Allen [2] studied wrinkling
of sandwich panels. The formulae that were developed
by those authors are given as follows:
Ho [3]
Symmetric wrinkling
Critical wrinkling loads
P
cr
= 1:82 t

E
f
E
c
G
c
3
_
: (31a)
Critical wrinkling length
L
cr
= 1:65 t

E
2
f
=E
c
G
c
6
_
: (31b)
Antisymmetric wrinkling
Critical wrinkling loads
1:18 t

E
f
E
c
_
t=h
_
0:774 h G
c
: (31c)
Critical wrinkling length
L
cr
= 1:67t

E
f
=E
c
4
_
h=t
4
_
: (31d)
Plantema
Critical wrinkling loads
P
cr
= 1:52 t

E
f
E
c
G
c
3
_
: (32a)
Critical wrinkling length
L
cr
= 1:9 t

E
2
f
=E
c
G
c
6
_
: (32b)
Allen
Critical wrinkling loads
P
cr
= 2 t B
1

E
f
E
2
c
3
_
: (33a)
Critical wrinkling length
L
c
= t C

E
f
=E
c
3
_
; (33b)
where B
1
= 0:570 and C = 2:09 for m
c
= 0:3
The results of dierent authors for Al-sandwich col-
umns are given in Table 4. The material properties are
given in Table 1. The geometry of the columns is given
in Table 2.
Table 4
Wrinkling loads of Al-sandwich columns (N/mm) (L = 600 mm; h = 75 mm; t = 0:325 mm)
Present Ho Plantema Allen Experiment (Webber)
Critical wrinkling loads (N/mm) 116
a
347
a
198
b
1613
b
290 348 183
Critical wrinkling lengths (mm) 9.38
a
5.85
a
7.5
b
10.6
b
6.74 5.86
a
Symmetric wrinkling mode.
b
Antisymmetric wrinkling mode.
Fig. 2. Critical loads vs number of half-waves for CFRP (0/90/core/90/0) sandwich columns.
430 B.K. Hadi, F.L. Matthews / Composite Structures 49 (2000) 425434
3.2. Wrinkling of sandwich panels
Two cases will be investigated. The rst case is as-
sociated with sandwich panels that behave in wrinkling
mode, while the second case is correlated with sandwich
panels having overall buckling mode. Comparisons will
be made with the analytical and experimental results of
Pearce and Webber [5].
3.2.1. Case I
The panel dimensions and properties are given in
Table 5, while the results analyses are presented in Table
6.
Table 6 shows that the present analysis is in a good
agreement with the experimental results of Pearce and
Webber [5]. The incorporation of adhesive layer in the
present analysis increases the wrinkling loads and in-
creases the accuracy of the present analysis compared to
the experimental ones. So the present analysis is gener-
ally in a good agreement with the experimental results
The dimensions and properties of the panels are chosen
such that wrinkling failures are the dominant factor on
the behaviour of the sandwich panels [5].
Fig. 3 shows the behaviour of wrinkling loads as the
number of half waves, m, increases, for the case of (0
+
/
core/0
+
) (Panel 2W in Table 6). This gure shows that
wrinkling is the dominant factor of the buckling loads.
Both symmetrical and antisymmetrical buckling loads
decrease as the number of half waves increases. They
decrease down to certain minimum values. After that,
the buckling loads increase. The theoretical wrinkling
loads are calculated when the symmetrical buckling
loads achieve minimum values. Fig. 3 also shows that
the addition of the adhesive layer into the analysis of
wrinkling loads will increase both the symmetrical and
antisymmetrical loads.
3.2.2. Case II
The panel dimensions and properties for overall
buckling specimens are shown in Table 7. The results of
the overall buckling analysis are given in Table 8.
Table 8 indicates that in general the present analysis
is in good agreement with the theoretical results of Pe-
arce and Webber [5]. The inclusion of adhesive layer will
increase the buckling loads, both for wrinkling and
overall buckling loads. In most cases, the present results
are better than Webber's results in comparison with
experimental failure loads. The accuracy of the results
increases with the inclusion of adhesive layer in the
analysis.
Fig. 4 shows the behaviour of critical loads as func-
tions of number of half waves, m. The sandwich panel is
CFRP-honeycomb panel with stacking sequence (0/90/0/
core/0/90/0) (Panel no. 2B). These curves show that in
that case, the overall buckling load is dominant. The
overall buckling load increases as the number of half
waves, m, increases. The thickness of the core is too thin
to make a signicant contribution in the wrinkling
mode. The symmetric buckling load is very high for
m = 1 (not shown in Fig. 4). This load decreases as m
increases. The load goes down to a certain minimum
value and then increases again as m increases. However,
the minimum symmetric wrinkling load is not low en-
ough when compared with the overall buckling loads.
Therefore, in that case, the overall buckling load is the
dominant factor in the failure mode of this particular
sandwich panel.
Table 5
Panel dimensions and properties
Panel dimensions a = b = 228 mm
Face plates
(CFRP)
E
x
= 142 GPa; E
y
= 9:8 GPa; m = 0:34;
G
xy
= 4:3 GPa; t
ply
= 0:125 mm
Core properties E
c
= 109 MPa; G
xz
= 26:6 MPa;
G
yz
= 15:5 MPa; h = 25 mm
Table 6
Critical loads (N/mm) for wrinkling specimen
Sandwich panel number
1 2W 3 4 5
Lay-up
0
+
45/0/45 45/0
+
/45 0/90
Results Webber Present Webber Present Webber Present Webber Present Webber Present
Experimental
failure loads
361 191 205 371 137
Theoretical
wrinkling
loads
490 497 160 161.3
(183.8)
170 171.7
(203.1)
285 287.6
(315.5)
77 (117) 77.43
(130.4)
Theoretical
overall buck-
ling loads
583 605.6 249 252.8
(286.6)
263 268.5
(313.9)
404 414.8
(450.7)
126
(187)
128.7
(211.2)
( ): Including adhesive layer.
*
Indicates 0.250 mm thick ply.
B.K. Hadi, F.L. Matthews / Composite Structures 49 (2000) 425434 431
3.3. Comparisons with overall buckling solutions
The current methods predict the overall buckling of
sandwich panels when m is equal to 1 for the case of
antisymmetric wrinkling. Comparisons will be made
with the overall buckling solutions of Rao [7], Kim and
Hong [8] and Hadi [10]. The panel dimensions and
properties are given in Table 9. The results are given in
Table 10.
Table 10 shows that the overall buckling result of the
present analysis is in a good agreement with other
methods.
3.4. Eect of core stiness
The eect of core stiness on the wrinkling of
sandwich panels is illustrated graphically in Fig. 5. In
this investigation, we will analyze (0/0/core/0/0) sand-
wich panels with dierent core stiness and thickness
(h). The face thickness is taken as constant, that is 0.2
mm. Fig. 5 shows that for small h/t values (thin core)
the overall buckling loads are dominant and increasing
h/t will increase the buckling loads. The buckling loads
increase up to certain h/t values, where wrinkling loads
are dominant. In this region, increasing h/t will drop
the buckling loads and the wrinkling modes are in
control.
Increasing the stiness of the core will increase the
maximum buckling loads and the wrinkling loads occur
in higher h/t values (thicker core). In these particular
panels, for h/t higher than 40, decreasing the core sti-
ness might change the buckling mode of the panels,
from overall buckling mode to wrinkling mode.
Table 8
Critical loads (N/mm) for overall buckling specimens
Sandwich panel number
1 2B 3 4 5
Lay-up Aluminium
++
0/90/0 45/0/45 45/0
+
/45 0/90
Results Webber Present Webber Present Webber Present Webber Present Webber Present
Experimental
failure loads
234 185 161 262
Theoretical
overall buck-
ling loads
243 302
(298)
152 156
(136.9)
199 203.0
(181.3)
238 241.9
(217.6)
105
(105)
109.3
(93.8)
Theoretical
wrinkling
loads
1806 1833
(1201)
1068 1150
(1076)
629 752
(634)
1053 1165
(1061)
432
(284)
498.9
(322.6)
( ): Without adhesive layer.
**
Indicates 0.650 mm thick faces.
*
Indicates 0.250 mm thick ply.
Fig. 3. Wrinkling loads of CFRP (0
+
/core/0
+
) sandwich panels.
Table 7
Panel dimensions and properties for overall buckling specimens
Panel dimensions a = b = 228 mm
Face plates
(CFRP)
E
x
= 142 GPa; E
y
= 9:8 GPa; m = 0:34;
G
xy
= 4:3 GPa; t
ply
= 0:125 mm
Core properties E
c
= 298 MPa; G
xz
= 60:0 MPa;
G
yz
= 35:2 MPa; h = 5 mm
432 B.K. Hadi, F.L. Matthews / Composite Structures 49 (2000) 425434
3.5. Eect of face lay-up
The eects of face lay-up on the buckling loads of
sandwich panels are investigated in Fig. 6. The face
lay-ups considered are (0/0/0/0/core/0/0/0/0), (90/90/90/
90/core/90/90/90/90), (0/0/90/90/core/90/90/0/0), (0/45/
)45/90/core/90/)45/45/0) and (45/)45/)45/45/core/45/
)45/)45/45). The material properties are given in
Table 9.
Fig. 6 shows that for small values of m (m < 16), the
stacking sequence of (45/)45/)45/45/core/45/)45/)45/
45) gives the highest overall buckling loads. However,
the maximum overall buckling load is achieved with
(0/0/0/0/core/0/0/0/0) stacking sequence, when h=t = 28.
At this value, the other stacking sequences are already
in wrinkling modes.
4. Summary
In this paper, the Benson and Mayers model has been
developed to solve the wrinkling of anisotropic sand-
wich panels. In this analysis, 10 undetermined coe-
Fig. 4. Critical loads vs number of half-waves for CFRP (0/90/0/core/0/90/0) sandwich panels.
Fig. 5. Buckling loads of (0/0/core/0/0) sandwich panels for dierent core stiness vs h/t. The core is honeycomb.
Table 10
Comparisons of overall buckling results of (0/0/core/0/0) sandwich panels by dierent methods (in N/mm)
Present method Hadi and Matthews [10] Rao [7] Kim and Hong [8]
427.2 427.7 423.8 427.6
Table 9
Panel dimensions and properties
Panel dimensions a = b = 225 mm
Face plates (Carbon/
epoxy)
E
x
= 229 GPa; E
y
= 13:35 GPa;
m
xy
= 0:315; G
xy
= 5:25 GPa;
t
ply
= 0:1 mm
Core properties E
zc
= 2 GPa; G
xz
= 1:46 GPa;
G
yz
= 0:904 GPa; h = 10 mm
B.K. Hadi, F.L. Matthews / Composite Structures 49 (2000) 425434 433
cients are taken as compared to 7 coecients of Benson
and Mayers. In the present analysis, both symmetric and
antisymmetric wrinkling are calculated simultaneously.
Benson and Mayers separated both symmetric and an-
tisymmetric wrinkling modes and calculated them sep-
arately.
The predicted symmetric and antisymmetric wrin-
kling loads are compared with the analytical and ex-
perimental results of Webber et al. [6]. The analytical
solutions of Webber only consider orthotropic sandwich
panels. The present solutions agree satisfactorily with
the analytical results of Webber, and in some cases have
better results compared to the experimental data.
The overall buckling load predicted from this analysis
is also compared with the results of Rao [7], Kim and
Hong [8] and Hadi and Matthews [10] and a good
agreement is achieved between these dierent methods.
Acknowledgements
Part of the research was funded by National Research
Council, Republic of Indonesia, through RUT V under
the research contract No. 18/SPK/RUT/BPPT/IV/1999.
The rst author thanks the Council for granting the
research fund.
References
[1] Plantema FJ. Sandwich Constructions. New York: Wiley, 1966.
[2] Allen HG. Analysis and Design of Structural Sandwich Panels.
London: Pergamon Press, 1960.
[3] Ho NJ, Mautner SE. The Buckling of Sandwich Panels. J
Aeronaut Sci 1945;12.
[4] Benson AS, Mayers J. General Instability and Face Wrinkling of
Sandwich Plates Unied Theory and Applications. AIAA J
1967;5(4):72939.
[5] Pearce TRA, Webber JPH. Experimental buckling loads of
sandwich panels with carbon bre faceplates. Aeronaut Quart
1973:295.
[6] Webber JPH, Kyriakides S, Lee CT. On the wrinkling of
honeycomb sandwich columns with laminated cross-ply faces.
Aeronaut Quart 1976:26472.
[7] Rao KM. Buckling analysis of anisotropic sandwich plates faced
with bre-reinforced plastics. AIAA J 1985;23(8):124753.
[8] Kim CG, Hong CS. Buckling of unbalanced anisotropic sandwich
plates faced with bre-reinforced plastics. AIAA J 1988;26(8):982
8.
[9] Press WH et al., Numerical Recipes: The Art of Scientic
Computing. 2nd ed. Cambridge: Cambridge University Press,
1992.
[10] Hadi BK, Matthews FL. Predicting the buckling load of aniso-
tropic sandwich panels: an approach including shear deformation
of the faces. Compos Struct 1988;42:24555.
[11] Hadi BK. Buckling of anisotropic sandwich panels with and
without holes. Ph.D. Thesis. Center for Composite Materials,
Imperial College, London: University of London, 1995.
Fig. 6. Wrinkling behaviour of sandwich panels with dierent face lay-up vs h/t.
434 B.K. Hadi, F.L. Matthews / Composite Structures 49 (2000) 425434