International Journal of Damage

Mechanics
http://ijd.sagepub.com/

Failure Modes of Composite Sandwich Beams

I. M. Daniel, E. E. Gdoutos, K.−A. Wang and J. L. Abot
International Journal of Damage Mechanics 2002 11: 309
DOI: 10.1106/105678902027247

The online version of this article can be found at:

http://ijd.sagepub.com/content/11/4/309

Published by:

http://www.sagepublications.com

Additional services and information for International Journal of Damage Mechanics can be found at:

Email Alerts: http://ijd.sagepub.com/cgi/alerts

Subscriptions: http://ijd.sagepub.com/subscriptions

Reprints: http://www.sagepub.com/journalsReprints.nav

Permissions: http://www.sagepub.com/journalsPermissions.nav

Citations: http://ijd.sagepub.com/content/11/4/309.refs.html

>> Version of Record - Oct 1, 2002

What is This?

Downloaded from ijd.sagepub.com at UNIV OF VIRGINIA on October 29, 2013

 Failure Modes of
Composite Sandwich Beams

I.M. DANIEL,* E.E. GDOUTOS, K.-A. WANG AND J.L. ABOT

Robert R. McCormick School of Engineering and Applied Science
Northwestern University
Evanston, IL 60208, USA

ABSTRACT: An investigation was conducted of failure modes and criteria for their
occurrence in composite sandwich beams. The initiation of the various failure modes
depends on the material properties of the constituents (facings and core), geometric
dimensions and type of loading. The beams were made of unidirectional carbon/
epoxy facings and aluminum honeycomb and PVC closed-cell foam cores. The
constituent materials were fully characterized and in the case of the foam core,
failure envelopes were developed for general two-dimensional states of stress.
Sandwich beams were loaded under bending moment and shear and failure modes
were observed and compared with analytical predictions. The failure modes
investigated are face sheet compressive failure, adhesive bond failure, indentation
failure, core failure and facing wrinkling.

KEY WORDS: sandwich construction, composite sandwich beams, failure modes,

compressive failure, wrinkling, indentation, core failure, cellular foams.

INTRODUCTION

ANDWICH CONSTRUCTION IS of particular interest and widely used,

S because the concept is very suitable and amenable to the development of
lightweight structures with high in-plane and flexural stiffness. Sandwich
panels consist typically of two thin face sheets (or facings, or skins) and
a lightweight thicker core. Commonly used materials for facings are
composite laminates and metals, while cores are made of metallic and non-
metallic honeycombs, cellular foams, balsa wood or trusses. The facings
carry almost all of the bending and in-plane loads and the core helps to
stabilize the facings and defines the flexural stiffness and out-of-plane shear
and compressive behavior.
*Author to whom correspondence should be addressed. E-mail: imdaniel@northwestern.edu

International Journal of DAMAGE MECHANICS, Vol. 11—October 2002 309

1056-7895/02/04 0309–26 $10.00/0 DOI: 10.1106/105678902027247
ß 2002 Sage Publications
310 I.M. DANIEL ET AL.

The overall performance of sandwich structures depends not only on the

properties of the facings, but also on those of the core and the adhesive
bonding of the core to the skins, as well as on geometrical dimensions.
Sandwich beams under general bending, shear and in-plane loading display
various failure modes. Their initiation, propagation and interaction depend
on the constituent material properties, geometry and type of loading.
Failure modes and their initiation can be predicted by conducting a
thorough stress analysis and applying appropriate failure criteria
in the critical regions of the beam. This analysis is difficult because of
the nonlinear and inelastic behavior of the constituent materials and the
complex interactions of failure modes. For this reason, properly designed
and carefully conducted experiments are important in elucidating the
physical phenomena and helping the analysis. Possible failure modes include
tensile or compressive failure of the facings, debonding at the core/facing
interface, indentation failure under concentrated loads, core failure,
wrinkling of the compression face and global buckling. Following initiation
of a particular failure mode, this mode may trigger and interact with other
modes and final failure may follow another failure path.
A substantial amount of work has been reported on failure of sandwich
panels (Allen, 1969; Hall and Robson, 1984; Zenkert, 1995). The various
modes have been studied separately and both initiation and ultimate failure
have been determined. No studies are known of the behavior following
initiation of a particular failure mode and triggering and interaction with
other failure modes.
In the present work, failure modes were investigated experimentally in
composite sandwich beams under four-point and three-point bending and
in end-loaded cantilever beams. Failure modes observed and studied include
face sheet compressive failure, face sheet debonding, indentation failure,
core failure and face sheet wrinkling.

CHARACTERIZATION OF CONSTITUENT MATERIALS

The sandwich beam facings were unidirectional carbon/epoxy plates

(AS4/3501-6), fabricated separately by autoclave molding. Uniaxial tensile
and compressive tests were conducted primarily in the longitudinal direction
in order to obtain the relevant constitutive behavior of the facing material.
The longitudinal tensile specimens were six-ply unidirectional coupons,
22.9 cm (9 in) long, 1.27 cm (0.50 in) wide and 0.76 mm (0.030 in) thick.
The specimens were tabbed with 3.81 cm (1.5 in) long glass/epoxy tabs at the
ends for better gripping and load introduction (Daniel and Ishai, 1994).
The longitudinal compressive specimens were 20-ply unidirectional coupons
16.5 cm (6.5 in) long, 6.4 mm (0.25 in) wide and 2.54 mm (0.100 in) thick.
Failure Modes of Composite Sandwich Beams 311

They were tabbed with 7.62 cm (3 in) long tabs at the ends and were tested
in an IITRI fixture (Daniel and Ishai, 1994). The longitudinal tensile
and compressive stress–strain behavior for the AS4/3501-6 carbon/epoxy is
shown in Figure 1, where it is seen that the material exhibits a characteristic
stiffening nonlinearity in tension and softening nonlinearity in compression.
Three different core materials were investigated. One of them was
aluminum honeycomb (PAMG 8.1-3/16 001-P-5052, Plascore Co.). The
other core materials investigated were two types of a fully cross-linked PVC
closed-cell foam, Divinycell H100 and H250, with densities of 100 and
250 kg/m3, respectively.
The aluminum honeycomb material is highly anisotropic with much
higher stiffness and strength in the through-the-thickness direction (cell
direction) than in the in-plane directions. The three principal moduli E1, E2
and E3 (along the cell axis) were obtained by means of flexural and pure
compression tests. The out-of-plane shear modulus G13 was obtained by
means of a rail shear test.
The lower density foam core material, Divinycell H100, exhibits nearly
isotropic behavior as illustrated by the stress–strain curves under uniaxial
tension and compression along the in-plane (1) and out-of-plane (3) direc-
tions in Figures 2 and 3. The higher density foam, Divinycell H250, exhibits
pronounced axisymmetric anisotropy with much higher stiffness and
strength in the cell direction (3-direction). Figure 4 shows stress–strain

Figure 1. Stress–strain curves of carbon/epoxy facing (AS4/3501-6).

312 I.M. DANIEL ET AL.

Through-the-thickness
In-plane

1(3)

3(1)

Figure 2. Stress–strain curves of PVC foam (Divinycell H100) under uniaxial tension in the
through-the-thickness (3) and in-plane (1) directions.

Figure 3. Stress–strain curves of PVC foam (Divinycell H100) under uniaxial compression in
the through-the-thickness (3) and in-plane (1) directions.
Failure Modes of Composite Sandwich Beams 313

Figure 4. Stress–strain curves of PVC foam (Divinycell H250).

curves for this material under uniaxial tension and compression along the
in-plane (1) and through-the-thickness (3) directions. The material displays
different behavior in tension and compression with tensile strengths much
higher than corresponding compressive strengths. The uniaxial stress–strain
behavior in tension is nonlinear elastic without any identifiable yield region.
In uniaxial compression the material is nearly elastic-perfectly plastic in
the initial stage of yielding. The shear stress–strain behavior on the
1–3 plane was determined by the Arcan test and is shown in Figure 5. The
shear behavior is also nearly elastic-perfectly plastic. Some characteristic
properties of the sandwich constituent materials investigated are tabulated
in Table 1.
A common failure mode in sandwich construction is the so-called ‘‘core
shear failure,’’ in which the core fails when the shear stress reaches its critical
value. However, although the shear stress is usually the dominant one in the
core, there are situations in which the normal stresses in the core are
of comparable magnitude or even higher than the shear stresses. Under such
circumstances a material element in the core may be subjected to a multi-
axial state of stress. Therefore, proper design of sandwich structures requires
failure characterization of the core material under combined stresses.
The higher density foam core was fully characterized under multiaxial
states of stress in the 1–3 plane (Gdoutos et al., 2002). A number of tests
314 I.M. DANIEL ET AL.

Figure 5. Shear stress–strain curves of PVC foam (Divinycell H250) on 1–3 plane.

Table 1. Characteristic properties of sandwich constituent materials.

Facing Honeycomb FM-73 Foam Core Foam Core

Core Adhesive (H100) (H250)
Density, 1620 (102) 129 (8.1) 1180 (74) 100 (6.0) 250 (15.0)

, kg/m3 (lb/ft3)
Thickness, 1.01 (0.040) 25.4 (1.0) 0.05 (0.002) 25.4 (1.0) 25.4 (1.0)
h, mm (in)
Longitudinal 147 000 (21 260) 9.5 (1.4) 1700 (247) 120 (17.5) 228 (33)
modulus,
E1, MPa (ksi)
Transverse 10 350 (1500) 2415 (350) 139 (20.1) 403 (58.4)
modulus,
E3, MPa (ksi)
Transverse shear 7600 (1100) 580 (84) 110 (16) 48 (6.9) 117 (17)
modulus, G13,
MPa (ksi)
Longitudinal 1930 (280) 0.2 (0.03) 1.7 (0.25) 4.5 (0.65)
compressive
strength, F1c,
MPa (ksi)
Transverse 240 (35) 11.8 (1.7) 1.9 (0.28) 6.3 (0.91)
compressive
strength, F3c,
MPa (ksi)
Transverse shear 71 (10.3) 3.5 (0.50) 33 (4.8) 1.6 (0.23) 5.0 (0.72)
strength, F13,
MPa (ksi)
Failure Modes of Composite Sandwich Beams 315

Figure 6. Failure pattern of a PVC foam (Divinycell H250) tube specimen subjected to
combined axial, torsion and internal pressure loading.

were conducted to define a failure surface for the material. A typical failure
pattern for a thin-wall tubular specimen under axial load, torsion and
internal pressure is shown in Figure 6. Experimental results conformed well
with the Tsai–Wu failure criterion for anisotropic materials as shown in
Figure 7. The Tsai–Wu criterion for a general two-dimensional state of
stress on the 1–3 plane is expressed as follows

f1 1 þ f3 3 þ f11 12 þ f33 32 þ 2f13 1 3 þ f55 52 ¼ 1 ð1Þ

where

1 1 1 1
f1 ¼  , f3 ¼ 
F1t F1c F3t F3c
1 1 1
f11 ¼ , f33 ¼ , f13 ¼  ð f11 f33 Þ1=2
F1t F1c F3t F3c 2
1
f55 ¼ ,
F52

F1t, F1c, F3t, F3c are the tensile and compressive strengths in the in-plane (1,2)
and out-of-plane (3) directions; F5 is the shear strength on the 1–3 plane.
Setting 5 ¼ kF5 , Equation (1) is rewritten as
316 I.M. DANIEL ET AL.

Figure 7. Failure envelopes predicted by the Tsai–Wu failure criterion for PVC foam
(Divinycell H250) for k ¼ 0, 0.8 and 1, and experimental results (k ¼  13/F13 ¼  5/F5).

f1 1 þ f3 3 þ f11 12 þ f33 32 þ 2 f13 1 3 ¼ 1  k2 ð2Þ

The failure surface described by the Tsai–Wu criterion is an ellipsoid in

the 1 , 3 , 13 ð5 Þ space displaced toward the tension–tension quadrant. It is
seen that the material can sustain shear stresses 13 ð5 Þ up to 17% higher
than the pure shear strength (F5). The most critical region for the material
is the compression–compression quadrant. The most critical combination is
compression and shear. Although the Tsai–Wu criterion appears reasonable
and adequate, it might be expected that in the compression–compression
region the material might follow better a maximum stress criterion because
its plastic Poisson’s ratio approaches zero (Fleck, 2000).

EXPERIMENTAL PROCEDURE

The honeycomb core was 2.54 cm (1 in) wide and was machined from
a 2.54 cm (1 in) thick sheet along the stiffer in-plane (E1) direction. The
2.54 cm (1 in) wide composite facings were machined from unidirectional
plates, bonded to the top and bottom faces of the honeycomb core with
FM73 M film adhesive and the assembly was cured under pressure in an
oven following the recommended curing cycle for the adhesive. Sandwich
beams were also prepared by bonding composite facings to foam cores of
Failure Modes of Composite Sandwich Beams 317

2.54  2.54 cm (1  1 in) cross section using a commercially available epoxy

adhesive (Hysol EA 9430) (Daniel et al., 1999).
Special fixtures were fabricated for beams subjected to three-point and
four-point bending and for end-loaded cantilever beams. Five span lengths
were chosen for beams under three-point bending, 10.2, 20.3, 25.4, 40.6 and
76.2 cm (4, 8, 10, 16 and 30 in). The cantilever beam was 46 cm (18 in) long.
In the four-point bending configuration, the distance between supports was
41 cm (16 in) and the distance between the middle loads was 18 cm (7 in). In
studying the effects of pure bending special reinforcement was provided for
the core at the outer sections of the beam to prevent premature core failures.
Also, under three-point bending, the faces directly under concentrated loads
were reinforced with additional layers of carbon/epoxy to suppress and
prevent indentation failure. Only in the case when the indentation failure
mode was studied there was no face reinforcement.
Strains on the outer and inner (interface between facing and core) surfaces
of the facings were recorded with strain gages. Most gages were oriented
along the axis of the beam, but some were mounted in the transverse
direction to record transverse strains. Beam deflections were measured with
a displacement transducer (LVDT) and by monitoring the crosshead
motion. The deflection was also monitored with a coarse moiré grating
(31 lines/cm, 80 lines/in). Longitudinal and transverse strains in the core
were measured with finer moiré gratings of 118 lines/cm (300 lines/in) and
200 lines/cm (500 lines/in). Before applying the moiré gratings, the lateral
surface of the core was coated with a white silicone rubber layer to make the
surface smooth and reflective.
The deformation of the core was also monitored with birefringent
coatings using reflection photoelasticity. Coatings, 0.5 mm (0.02 in) and
1 mm (0.04 in) thick, were used (PS-4D coatings, Measurements Group).
The coating is bonded to the surface of the core with a reflective cement to
insure light reflection at the interface. A still camera and a digital camcorder
were used to record moiré and isochromatic fringe patterns. The fringe
order of this pattern is related to the difference of principal strains as
follows:

N
"c1  "c3 ¼ "s1  "s3 ¼ ð3Þ
2hK

where N is the fringe order,  is the wavelength of the illuminating light, h is

the coating thickness and K is a calibration constant for the coating
material. Superscripts s and c denote specimen and coating, respectively.
318 I.M. DANIEL ET AL.

FAILURE MODES

A number of failure modes were recorded and studied in the composite

sandwich beams tested. They include compressive facing failure, face sheet
debonding, indentation of the loaded face, core failures and wrinkling of the
compression facing. These failure modes are discussed in the following
sections.

Compressive Facing Failure

The tension and compression faces may fail under uniaxial stress (Allen,
1969; Zenkert, 1995). In the case of composite facings, compressive failure is
more likely than tensile failure because the material is appreciably weaker in
compression than in tension. This type of failure occurs in beams under pure
bending or bending and low shear with cores of sufficiently high stiffness in
the through-the thickness direction.
This type of failure was observed in sandwich beams with carbon/epoxy
facings and aluminum honeycomb core loaded in four-point bending
(Daniel et al., 1999, 2000). As mentioned before, special reinforcement was
provided for the core at the outer sections of the beam to prevent premature
core failures under shear or indentation failures under the loads. Figure 8
shows plots of the applied bending moment versus strain on the outer
surfaces of the facings. As expected, the curves show the same stiffening and
softening characteristics on the tension and compression sides as the carbon/
epoxy facing material (Figure 1). Failure was governed by the compressive

Figure 8. Experimental and predicted moment–strain curves for two facings of composite
sandwich beam under four-point bending (dimensions are in cm).
Failure Modes of Composite Sandwich Beams 319

strength of the facing which in this case reached a value of 1930 MPa
(280 ksi), higher than any recorded value for this material under direct
compression. The ultimate compressive strain recorded was 1.6%. This is
attributed to the support provided to the skin by the core which suppresses
the tendency for buckling.
The strain variation through the thickness was checked by embedding
strain gages at the interfaces between the facings and the core and by using
moiré gratings on the core. The moiré pattern corresponding to axial
displacements on the core consists of fringes in the form of hyperbolas,
which is consistent with a linear strain variation through the thickness
(Figure 9).

uðx,zÞ ¼ cxz ð4Þ

@u
"1 ðzÞ ¼ ¼ cz ð5Þ
@x
since u(0, z) ¼ 0 (along vertical axis of symmetry).
The linear strain variation through the thickness of the beam was also
corroborated by the embedded strain gage readings.
The experimentally obtained stress–strain relations of the facing material
in tension and compression (Figure 1) and that of the honeycomb core were
used to obtain moment–strain relations. In the modeling, two cases were
considered one by assuming linear variation of the strain through the facing
thickness and one by assuming constant strain. The nonlinear stress–strain
relations of the facing material were also taken into account. Results
were compared with the experimental ones in Figure 8. The agreement is
satisfactory. The small discrepancies observed are not due to the model but
rather to the difficulty in obtaining reliable stress–strain curves in direct
longitudinal compression.

Figure 9. Moiré fringe pattern in aluminum honeycomb core corresponding to longitudinal

displacement (12 lines/mm; 300 lines/in).
320 I.M. DANIEL ET AL.

In general, for the widely used cores which have much lower stiffness than
the facing material, the contribution of the core is negligible. For relatively
thin skins and relatively low core stiffness, compressive failure of the facing
is satisfactorily predicted by the simple moment equilibrium relation

M ffi Ffc hf ðhf þ hc Þ ð6Þ

where M is the applied bending moment at failure, Ffc is the compressive

strength of facing material, and hf, hc are the facing and core thicknesses,
respectively.
In the equation above it is assumed that the stress in the thin facing is
constant reaching the ultimate value of Ffc at failure and that the core
contribution is negligible. If the core stiffness in the through-the-thickness
direction is not sufficiently high, another mode of failure, face wrinkling,
takes place. This failure mode will be discussed later on in more detail.

Face Sheet Debonding

Face sheet debonding may develop during fabrication of sandwich panels

or may be caused by external loading such as impact. Debonding reduces
the stiffness of the structure and makes it susceptible to buckling under
in-plane compression. Buckling of the debonded region of the face sheet was
analyzed by Hansen (1998). Debonding of sandwich panels with foam cores
and isotropic faces was investigated by Triantafillou and Gibson (1989) in
terms of the critical strain energy release rate of the interface. Fracture loads
of sandwich beams with disbonds were predicted by Zenkert (1991) using
interfacial fracture mechanics principles. The axial compressive strength of
debonded sandwich composite specimens with graphite/epoxy faces and
aramid fiber honeycomb core was determined by Avery and Sankar (2000).
The strength and fracture toughness of the facing/core interface was
determined from tension and double cantilever beam tests. The buckling,
postbuckling and delamination (debonding) growth behavior of a sandwich
beam in bending was studied by Kardomateas (1999).
In the present study, no debonding failures were observed in most beam
specimen and loading configurations. Beams with aluminum honeycomb
cores showed some premature debonding failure in some cases due to the
very small bonded area of the honeycomb cross section. In the case of foam
cores no debonding was observed under quasi-static loading due to the
relatively high interface fracture toughness. Under impact, delamination
failures of the compressive face sheet were observed, but no interfacial
debonding.
Failure Modes of Composite Sandwich Beams 321

Indentation Failure

Indentation failure is a dominant mode of failure in cases of highly

localized external loads, such as point or line loads. The failure process
consists of local yielding and deformation of the core material under the
highly loaded area followed by significant local deformation of the loaded
facing into the core. This results in a complex elastic-plastic multiaxial state
of stress in the vicinity of localized loads.
The classical approach of a beam on a Winkler foundation has been
used by some investigators (Plantema, 1966; Allen, 1969; Zenkert, 1995).
An analysis for an elastic-perfectly plastic and a rigid-perfectly plastic
foundation has been performed (Soden, 1996; Shuaeib and Soden, 1997).
The analysis led to simple formulas for prediction of critical loads for core
yielding and facing failure, length of plastic zone under the load and the
load–displacement relation. A two-parameter elastic foundation model,
which includes normal and shear stresses between the loaded face and the
supporting medium (core) has been proposed by Thomsen (1992). This
model accounts for the interfacial shear stresses in the vicinity of the applied
load. Frostig and Baruch (1990) analyzed the bending behavior of sandwich
beams under concentrated and distributed loads. Photoelastic investigations
of model sandwich beams were performed by Allison (1990) and Thomsen
and Frostig (1997). The indentation of sandwich structures with honeycomb
core was studied from the points of view of test methodology, panel
construction and failure mechanisms by Tsotsis and Lee (1996).

Figure 10. Moiré fringe patterns in sandwich beam with foam core corresponding to vertical
displacements at various applied loads (11.8 lines/mm grating). Loads: (a) 318 N (71.5 lb);
(b) 574 N (129 lb); (c) 812 N (182 lb); (d) 924 (208 lb); (e) 1057 N (237 lb); (f) 1080 N (243 lb).
322 I.M. DANIEL ET AL.

Figure 11. Load vs. deflection under load of sandwich beam under three-point bending
(carbon/epoxy facings, Divinycell H100 core).

Indentation failure was observed in beams under three-point bending

when no special reinforcement of the facing or the core was provided in the
area under the load. Figure 10 shows moiré fringe patterns corresponding to
vertical displacements of a sandwich beam loaded in three-point bending. It
shows the indentation development and the area of the beam affected by
the indentation. Figure 11 shows a plot of the applied load versus the
displacement of the indenting roller at the center of the beam. This
displacement represents the sum of the global beam deflection and the local
indentation, but it is more sensitive to the local indentation. Therefore, the
proportional limit of the displacement curve is a good indication of
initiation of indentation. In the present case the beam had 1 mm (0.04 in)
thick carbon/epoxy facings, a 25 mm (1 in) thick Divinycell H100 core and
a 36 cm (14 in) span. The load at initiation of indentation is 735 N (165 lb).
The peak load measured was Pmax ¼ 1080 N (243 lb).
The initiation of indentation can be predicted by treating the loaded face
as a beam on an elastic foundation (Hetenyi, 1946). According to this theory
the compressive stress at the interface between core and skin is proportional
to the local deflection

 ¼ kw ð7Þ

where  is the interface stress and w is the deflection,

Failure Modes of Composite Sandwich Beams 323
sffiffiffiffiffiffi
3 Ec
k ¼ 0:28Ec , foundation modulus ð8Þ
Df

Df ¼ bh3f Ef =12, skin flexural stiffness,

Ef, Ec are the facing and core moduli, respectively, b is the beam width, and
hf is the facing thickness.
The interfacial stress (core stress) in Equation (7) is obtained by
determining the deflection w(x) for a beam on an elastic foundation

P x
wðxÞ ¼ e ðcos x þ sin xÞ ð9Þ
2k
where
sffiffiffiffiffiffi
1:18 3 Ec
¼ ð10Þ
hf Ef

By equating the interfacial stress under the load to the yield stress of the
foam core, we obtain an expression for the force at initiation of core yield:
rffiffiffiffiffiffi
3 Ef
Pcy ¼ 1:70cy bhf ð11Þ
Ec

For the Divinycell H100 core of the beam tested, we obtain

Pcy ¼ 800 Nð180 lbÞ ðcalculatedÞ

compared to the measured value of

Pcy ¼ 135 Nð165 lbÞ ðmeasuredÞ

Failure of the beam is due to compressive failure on the upper surface of

the skin after indentation. The total compressive stress in the loaded face
can be obtained by superposition of the global stress due to bending of the
beam and the local bending stress due to indentation under the concentrated
load. The local bending stress for a rigid-perfectly plastic foundation is given
by Soden (1996) as
9P2
fi ¼ ð12Þ
16b2 h2f Fc

where Fc ¼ compressive strength of core. The global bending stress is

324 I.M. DANIEL ET AL.

PL
fb ¼ ð13Þ
4bhf ðhf þ hc Þ
Failure occurs when the sum of the stresses above equals the strength of the
facing material

Ffc ¼ fi þ fb ð14Þ

For the beam tested, the critical load at facing failure is calculated as

Pcr ¼ 1310 N ð294 lbÞ ðcalculatedÞ

compared to the measured value of

Pcr ¼ 1080 Nð243 lbÞ ðmeasuredÞ

The difference in the results above may be attributed to Soden’s assumption

of a rigid-perfectly plastic foundation.

Core Failure

The core is primarily selected to carry the shear loading. Core failure by
shear is a common failure mode in sandwich construction (Allen 1969; Hall
and Robson, 1984; Zenkert and Vikström, 1992; Zenkert, 1995). In short
beams under three-point bending the core is mainly subjected to shear and
failure occurs when the maximum shear stress reaches the critical value
(shear strength) of the core material. In long-span beams, the normal stresses
become of the same order of magnitude as, or even higher than the shear
stresses. In this case, the core in the beam is subjected to a biaxial state of
stress and fails according to an appropriate failure criterion. It was shown
earlier that failure of the PVC foam core Divinycell H250 can be described
by the Tsai–Wu failure criterion (Gdoutos et al., 2002).
Sandwich beams with aluminum honeycomb cores under three-point
bending failed due to early shear crimping of the core. The shear force at
failure remained nearly constant for varying span lengths. This means that
as the span length increases, the applied maximum moment and thereby the
maximum face sheet strains at failure increase (Figure 12). The results also
indicate that the bending moment is carried almost entirely by the face
sheets. The average shear stress at failure from the three tests represented in
Figure 12 is  u ¼ 3.59 MPa (520 psi) which compares well with the measured
shear strength of the honeycomb material of Fcs ¼ 3.45 MPa (500 psi).
The deformation and failure mechanisms in the core were studied
experimentally by means of moiré gratings and birefringent coatings.
Failure Modes of Composite Sandwich Beams 325

Figure 12. Applied moment vs. maximum facing strain for beams of different span length
under three-point bending.

Figure 13. Moiré fringe patterns corresponding to horizontal and vertical displacements in
sandwich beam under three-point bending (12 lines/mm; Divinycell H250 core).

Figure 13 shows moiré fringe patterns in the core of a sandwich beam under
three-point bending. The moiré fringe patterns corresponding to the u and w
displacements away from the applied load consist of nearly parallel and
equidistant fringes from which it follows that

@u
"x ¼ ffi0
@x
@w
"z ¼ ffi0 ð15Þ
@z
@u @w
xz ¼ þ ffi constant
@z @x
326 I.M. DANIEL ET AL.

Figure 14. Isochromatic fringe patterns in birefringent coating of sandwich beam under
three-point bending.

Figure 15. Variation with increasing load of shear strain distribution through-the-thickness.
Failure Modes of Composite Sandwich Beams 327

Thus, the core is under nearly uniform shear stress. This is true only in the
linear range as shown by the isochromatic fringe patterns of the birefringent
coating in Figure 14. In the nonlinear and plastic region the core begins to
yield and the shear strain becomes highly nonuniform peaking at the center.
From fringe patterns like those of Figure 14 it was found that the shear
deformation starts becoming nonuniform at an applied load of 3.29 kN
(740 lb) which corresponds to an average shear stress of 2.55 MPa (370 psi).
This is close to the proportional limit of the shear stress–strain curve of
Figure 5. As the load increases, the shear strain in the core becomes
nonuniform peaking at the center as illustrated in Figures 14 and 15.
Core failure is accelerated when compressive and shear stresses are
combined. This critical combination is evident from the failure envelope of
Figure 7. The criticality of the compression/shear stress biaxiality was tested
with a cantilever sandwich beam loaded at the free end. The isochromatic
fringe patterns of the birefringent coating in Figure 16 show how the peak
birefringence moves towards the fixed end of the beam at the bottom where
the compressive strain is the highest and superimposed on the shear strain.
Plastic deformation of the core, whether due to shear alone or a combina-
tion of compression and shear, degrades the supporting role of the core and
precipitates other more catastrophic failure modes, such as facing wrinkling.

Compression Facing Wrinkling

A common failure mode of sandwich beams subjected to compression

or bending is localized short-wavelength buckling (wrinkling) of the

Figure 16. Isochromatic fringe patterns in birefringent coating of cantilever sandwich beam.
328 I.M. DANIEL ET AL.

compression face. Wrinkling may be viewed as buckling of the compression

face supported by an elastic continuum, the core (Hoff and Mautner, 1945;
Plantema, 1966). A wrinkling analysis of honeycomb sandwich beams with
linear and nonlinear stress–strain behavior of the faces was developed by
Gutierrez and Webber (1980) and Ditcher and Webber (1982). An analytical
model and a finite element analysis were presented by Vonach and
Rammerstorfer (2000a,b) for sandwich beams with isotropic or orthotropic
faces and thick transversely isotropic or orthotropic cores. A general theory
for the simultaneous calculation of wrinkling and global buckling for isotropic
faces and orthotropic core was developed by Benson and Mayers (1967).
The theory was recently generalized for anisotropic sandwich panels by
Hadi and Matthews (2000).
Face wrinkling failures were observed in sandwich beams with foam cores
but not in those with honeycomb cores. This behavior is controlled to a great
extent by the core modulus in the through-the-thickness direction.
Figure 17 shows moment versus strain results for two different tests of
sandwich beams with foam cores (Divinycell H100) under four-point
bending. Evidence of wrinkling is shown by the sharp change in recorded
strain on the compression face, indicating inward and outward wrinkling in
the two tests. In both cases the critical wrinkling stress was

cr ¼ 673 MPa ð97:5 ksiÞ

The predicted value based on Heath’s (1960) expression for isotropic

Figure 17. Facing wrinkling in sandwich beam under four-point bending (Divinycell H100
foam; dimensions are in cm).
Failure Modes of Composite Sandwich Beams 329

materials modified here to consider only the face modulus along the axis of
the beam, is
 1=2
2 hf Ec3 Ef 1
cr ¼ ¼ 687 MPa ð99:5 ksiÞ ð16Þ
3 hc ð1 
13
31 Þ
In the case when shear is present in addition to bending, accounting must
be taken of the influence of the transverse shear modulus of the core, Gc13.
An early expression given by Hoff and Mautner (1945) has the form
cr ¼ cðEf 1 Ec3 Gc13 Þ1=3 ð17Þ
where c is a constant usually taken as equal to 0.5, 0.6, or 0.65. Note that
the critical stress in this expression depends only on the elastic moduli of
the facing and core materials. In the relation above the core moduli are the
initial elastic moduli if wrinkling occurs while the core is still in the linear
elastic range. This requires that the shear force at the time of wrinkling be
low enough or, at least,
V < Ac Fcs ð18Þ
where Ac is core cross-sectional area and Fcs the shear strength of the core.
This is the case of long span beams under three-point bending or long
cantilever beams under end loading.
Figure 18 shows moment versus strain curves for an end loaded cantilever
beam where an indication of wrinkling is given by the strain reading at a
local critical stress of

Figure 18. Moment vs. strain curves for cantilever sandwich beam (Divinycell H250 core).
330 I.M. DANIEL ET AL.

cr ¼ 862 MPa ð125 ksiÞ

Using the measured material properties from Table 1 and a value of c ¼ 0.5
in Equation (17) we obtain

cr ¼ 0:5ðEf 1 Ec3 Gc13 Þ1=3 ¼ 947 MPa ð137 ksiÞ

In the case of shorter span beams where the shear loading component is
significant, core failure precedes facing wrinkling. Core yielding and
stiffness loss reduce core support of the facings and precipitate facing
wrinkling failure at a lower stress. The critical wrinkling stress in that case
would be

0
cr ¼ 0:5ðEf 1 Ec3 G0c13 Þ1=3 ð19Þ

0
where Ec3 , G0c13 are the reduced core Young’s and shear moduli in the
through-the-thickness direction.
Figure 19 shows moment versus strain curves for a beam under three-
point bending. Facing wrinkling seems to occur at a local bending moment
of 330 N m (2900 lb in) and a shear force of 2875 N (646 lb). The local critical
facing stress is

cr ¼ 500 MPa ð72:5 ksiÞ

Figure 19. Moment vs. strain curves for sandwich beam under three-point bending
(Divinycell H250 core).
Failure Modes of Composite Sandwich Beams 331

considerably lower than the predicted value of 947 MPa (137 ksi). The
reduction is attributed to the core stiffness reduction, which in this case
would be

0
Ec3 G0

E ¼ ffi c13 ¼ 0:385
Ec3 Gc13

This reduction (in secant or tangent modulus) is quite possible in view of

the high shear strain shown in the photoelastic patterns of Figures 14 and 15
(over 15%) and the elasto-plastic shear behavior of the core material
depicted in Figure 5.
It is obvious from the above that failure modes, their initiation, sequence
and interaction depend on loading conditions. In the case of beams under
three-point bending this is illustrated by varying the span length. For short
spans, core failure occurs first and then it triggers facing wrinkling. For long
spans, facing wrinkling can occur before any core failure. Core failure
initiation can be described by calculating the state of stress in the core and
applying the Tsai–Wu failure criterion. This yields a curve for critical load
(at core failure initiation) versus span length. On the other hand, in
the absence of core failure, facing wrinkling can be predicted by Equation
(17) and expressed in terms of a critical load as a function of span length.
Figure 20 shows curves of the critical load versus span length for initiation
of the two failure modes discussed above. Their intersection defines the
transition from core failure initiation to facing wrinkling initiation. For

Figure 20. Critical load vs. span length for failure initiation in sandwich beams under three-
point bending.
332 I.M. DANIEL ET AL.

a beam with carbon/epoxy facings (8-ply unidirectional AS4/3501) and PVC

foam core (Divinycell H250) of 2.5  2.5 cm (1  1 in) cross section, the span
length for failure mode transition is L ¼ 35 cm (13.8 in).
Although the results above are at least qualitatively explained by available
theory, it is apparent that better theoretical modeling is needed. The
theoretical prediction of facing wrinkling, Equation (17), gives equal weight
to the three moduli involved and is independent of facing and core
dimensions. A more sound theory should take into consideration the
nonlinear and inelastic biaxial stress–strain behavior of the core material
and the stress–strain redistribution following core yielding in addition to
geometrical dimensions.

SUMMARY AND CONCLUSIONS

Failure modes were investigated experimentally in composite sandwich

beams. Failure modes and their initiation, propagation and interaction
depend on the type of loading, constituent material properties and
geometrical dimensions. Failure modes discussed, observed and studied
include face sheet compressive failure, face sheet debonding, indentation
failure, core failure and face sheet wrinkling. Experimental results were
compared with theoretical predictions whenever they were available. A
summary of the failure modes investigated and the appropriate failure
criteria is given in Table 2.
The constitutive materials, core and face sheet composite, were fully
characterized. The high density foam core (Divinycell H250) was found to

Table 2. Failure modes in sandwich beams.

Mode
Face sheet compressive failure
Face sheet debonding
Indentation failure
Core failure
Face sheet wrinkling
Criterion
Composite failure criterion
(e.g., maximum stress, Tsai–Wu)
Maximum shear stress of adhesive,
or interfacial fracture toughness
Core state of stress reaches failure
condition; critical compressive stress
in face sheet under combined local
and global bending
State of stress in foam reaches
failure condition (e.g., Tsai–Wu)
Face sheet stress equals critical
local buckling stress
Failure Modes of Composite Sandwich Beams 333

behave according to the Tsai–Wu interaction failure criterion for a general

two-dimensional state of stress.
Compressive face sheet failure is likely under pure bending conditions or
when the shear load is low enough and the core stiff and strong enough to
avoid core failure. A maximum stress failure criterion for the composite face
material is sufficient to predict this type of failure.
Face sheet debonding is not very common in sandwich beams with foam
cores, unless there are initial fabrication defects. It is more likely under
impact loading.
Indentation failure is a serious problem whenever there is any load
concentration on foam-core sandwich panels. It results from local (multi-
axial) compressive failure of the core under the load and is followed by local
face sheet bending to failure of the face sheet. Experimental results were in
qualitative agreement with calculated ones based on the theory of beams on
elastic or rigid-perfectly plastic foundations for the indentation initiation
and indentation failure loads, respectively.
In the linear range the core shear stress (strain) is uniform through
the thickness. In the nonlinear/plastic range the core begins to yield and the
shear strain becomes highly nonuniform, peaking at the center. Core failure
is accelerated when compressive and shear deformations are combined. Core
yielding and stiffness loss reduce core support for the facings, precipitating
facing wrinkling failure.
Face sheet wrinkling occurs when the local compressive stress reaches a
critical value which depends on the face sheet and core moduli. This critical
stress depends on whether the supporting core is degraded or not. In the
latter case, while the core is in the linear elastic region, experimental results
are in reasonable agreement with predictions. When core failure and
stiffness degradation occur first, the critical wrinkling stress is substantially
reduced. No theory exists currently to give a quantitative prediction of the
critical stress in this case.

ACKNOWLEDGMENTS

This research was sponsored by the Office of Naval Research (ONR). We

are grateful to Dr. Y. D. S. Rajapakse of ONR for his encouragement and
cooperation and to Mrs. Yolande Mallian for typing the manuscript.

REFERENCES

Allen, H.G. (1969). Analysis and Design of Structural Sandwich Panels, Pergamon, London.
Allison, I.M. (1990). In: Proceedings of 9th Int. Conf. Experimental Mechanics, Copenhagen,
Denmark, pp. 20–24, 1604–1609.
334 I.M. DANIEL ET AL.

Avery J.L. III and Sankar, B.V. (2000). J. Comp. Mat., 34: 1176–1199.
Benson, A.S. and Mayers, J. (1967). AIAA, 5: 729–739.
Daniel, I.M., Abot, J.L. and Wang, K.A. (1999). ICCM 12 Intern. Conf. on Composite Mat.,
Paris, France.
Daniel, I.M. and Abot, J.L. (2000). Comp. Sci. and Tech., 60: 2455–2463.
Daniel, I.M. and Ishai, O. (1994). Engineering Mechanics of Composite Materials, Oxford
University Press, New York.
Ditcher, A.K. and Webber, J.P.H. (1982). Aero. Quart, 33: 1–24.
Fleck, N. (2000). Private communication.
Frostig, Y. and Baruch, M. (1990). AIAA, 28: 523–531.
Gdoutos, E.E., Daniel, I.M. and Wang, K.A. (2002). Composites, Part A, 33: 163–176.
Gutierrez, A.J. and Webber, J.P.H. (1980). Int. J. Sol. Struct., 16: 645–651.
Hadi, B.K. and Matthews, F.L. (2000). Comp. Struct., 49: 425–434.
Hall, D.J. and Robson, B.L. (1984). Composites, 15: 266–276.
Hansen, U. (1998). J. Comp. Mat., 32: 335–360.
Heath, W.G. (1960). Aircraft Engineering, 32: 230–235.
Hetenyi, M. (1946). Beams on Elastic Foundation, The Univ. of Michigan Press.
Hoff, N.J. and Mautner, S.E. (1945). J. Aero. Sci., 12: 285–297.
Kardomateas, G.A. (1999). AMD-Vol. 235, Thick Composites for Load Bearing Structures,
ASME, 51–60.
Plantema, F.J. (1966). Sandwich Construction, Wiley, New York.
Shuaeib, F.M. and Soden, P.D. (1977). Comp. Sci. Tech., 57: 1249–1259.
Soden, P.D. (1996). J. Strain Anal., 31: 353–360.
Thomsen, O.T. (1992). In: Olsson, K.-A. and Weissman-Berman, D. (eds.), Sandwich
Construction, 2: 417–440.
Thomsen, O.T. and Frostig, Y. (1997). Comp. Struct., 37: 97–108.
Triantafillou, T.C. and Gibson, L.J. (1989). Mat. Struct., 22: 64–69.
Tsotsis, T.K. and Lee, S.M. (1996). In: Deo, R.B. and Saff, C.R. (eds.), Composite Materials:
Testing and Design, Vol. 12, ASTM STP 1274, ASTM, pp. 139–165.
Vonach, W.K. and Rammerstorfer, F.G. (2000a). Arch. Appl. Mech., 70: 338–348.
Vonach, W.K. and Rammerstorfer, F.G. (2000b). Acta. Mech., 141: 1–20.
Zenkert, D. (1991). Comp. Struct., 17: 331–350.
Zenkert, D. and Vikström, M. (1992). J. Comp. Res. Tech., 14: 95–103.
Zenkert, D. (1995). An Introduction to Sandwich Construction, Chameleon, London.