Composite Structures 101 (2013) 204–214

Contents lists available at SciVerse ScienceDirect

Composite Structures
journal homepage: www.elsevier.com/locate/compstruct

Damage and failure mode maps of composite sandwich panel subjected

to quasi-static indentation and low velocity impact
Zhu Shengqing, Chai Gin Boay ⇑
School of Mechanical & Aerospace Engineering, Division of Aerospace Engineering, Nanyang Technological University, Singapore

a r t i c l e i n f o a b s t r a c t

Article history: In this contribution, modes of damage and modes of failure of composite sandwich panel subject to quasi-
Available online 26 February 2013 static indentation and low velocity impact were investigated. This study was conducted through the use
of the finite element analysis, existing formulae and experiments. The damage and failure modes of com-
Keywords: posite sandwich panels were analyzed with some deliberation. With these results, damage and failure
Composite mode maps were compiled to facilitate the design of composite sandwich structures. Existing formulae
Sandwich that predict limit loads for damage modes which includes matrix shear/tension damage and core buck-
Damage
ling, were reviewed and compiled to plot the damage mode map. The failure modes of facesheet cracking,
Failure
Mode maps
overall core crushing and core shear fracture, were investigated to configure the failure mode map. Pre-
ferred damage and failure modes for impact resistance were also discussed in term of energy absorption.
Extensive quasi-static indentation and low-velocity impact tests were carried out to verify these modes.
The results of the finite element analysis are also included to complement the work reported in this
contribution.
Ó 2013 Elsevier Ltd. All rights reserved.

1. Introduction has been extensively investigated [8,9]. Since the impact duration
is quite short and impact damages are difficult to detect, Finite Ele-
Composite sandwich panels are increasingly used in aircraft and ment Analysis (FEA) has been widely used to simulate the damage
aerospace industries due to their high specific strengths and stiff- evolution and failure mode of sandwich panels [10–12].
nesses. However, composite sandwich structures are susceptible Definition of damage initiation has been defined phenomeno-
to damage and failure due to transverse contact and to impact by logically according to the law of damage evolution [13]. From the
foreign object. The failure mode map [1–7] is found to be a good extensive experimental investigations of the quasi-static indenta-
technique in designing composite sandwich structures to optimize tion and impact tests [8,14–19], damages are initiated in sandwich
their energy absorption properties and load carrying capacity with panels when the impact energy has exceeded a limit damage value.
minimum weight to satisfy the quasi-static and impact loading The contact force for damage initiation PI may not be the maximum
requirement. The introduction of a damage initiation map in this load but it can be used as a valuable index to investigate the energy
contribution would complement the failure modes map. The com- absorption and impact resistance of composites. To investigate the
bination of damage and failure mode maps would be of great re- onset of damage, damage initiation modes in composite sandwich
search interest and they will serve the needs of the structural structures are classified as: (i) core buckling and core crushing, (ii)
designers. delaminations in facesheet or debonding along facesheets and the
Failure mode maps of sandwich beam have been extensively core, (iii) matrix cracking in facesheet due to contact pressure, (iv)
investigated since the late eighties. Investigations on the formation fiber breakage in the facesheet.
and implementation of the failure mode maps of sandwich struc- When damage occurs in a sandwich structure, its stiffness will
tures were pioneered via three-point bend tests on sandwich degrade and its load capacity decrease significantly up to the final
beams [3,4]. Experimental and theoretical work on three-point point of failure [18]. The failure mode of wrinkling of facesheets
bend tests [2–6] and four-point bend tests [5] are readily available. was reported for sandwich beam structures in bending [2,4] but
Failure modes on sandwich structures subjected low-velocity im- there is no report of this mode for sandwich panels subjected to
pact [6], blast loading [1] and uniform pressure [7] were also pro- low velocity impact of a rigid projectile. This mode, thus, will not
posed. Research work on sandwich panel with isotropic facesheets be discussed in the present work. The failure modes of sandwich
structures subjected to low velocity impact load are reported to
⇑ Corresponding author. Tel.: +65 67905756; fax: +65 67924062. be similar to those of the sandwich structures subjected to quasi-
E-mail address: [email protected] (G.B. Chai). static indentation [9,20]. Therefore, the experimental results from

0263-8223/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compstruct.2013.02.010
S. Zhu, G.B. Chai / Composite Structures 101 (2013) 204–214 205

the quasi-static indentation tests [21,22] and the low velocity im- experiments were recorded to validate the damage and failure
pact tests will be combined in this present work. In this contribu- mode maps developed here and these will be deliberated later.
tion, damage and failure mode maps for composite sandwich The densities of the honeycomb core are 45 kg/m3 for S1 and S2,
panels are developed using published explicit formulae for their and 75 kg/m3 for S3 to S5.
corresponding limit loads. The mapping process itself is a not dif-
ficult one but the process of reviewing and summarizing the exist- 2.2. Finite element model
ing published expressions has to be thorough and carefully done
such that the results of the mapping is a useful and practical one. Finite element method and analysis is a powerful numerical
tool that is frequently used to simulate the structural response,
2. Experimental investigation and finite element model damage modes and failure modes of sandwich panel subject to var-
ious load conditions [10,11,23]. A commercially available finite ele-
The configurations of the test specimens fabricated for the ment software ABAQUS/Explicit is used to simulate the damage
experiments are summarized in this section. The details of the test initiation, evolution and failure modes. A finite element model of
specimen’s geometries and properties are required as inputs in the the actual specimen geometry was developed as shown in Fig. 3b.
finite element model and the subsequent analysis. However the The core and facesheets were meshed with S4R shell elements.
details of the complete test procedures and the finite element Elastic–plastic properties were specified for the Nomex core while
modeling/analysis were discussed in other contributions [12,29]. Hashin criterion [24,25] was used in the prediction of the damage
A brief description of the experimental procedures will be dis- initiation and evolution of facesheets [25,26]. The impactor was
cussed in the following section for completeness. specified to be a rigid body with a mass and initial velocity as listed
in Table 1. The results of the contact force between impact, dis-
2.1. Experimental investigations placement of strategic nodes and damage initiation of laminates
were output for analysis.
The sandwich panels for experiments were fabricated using Fib- The respective damage/failure modes of facesheet cracking, ma-
reduxÒ 913C-HTA carbon fiber-epoxy composites bonded to Hex- trix cracking, matrix compressive and shear are evaluated using a
WebÒ Nomex honeycomb core with Redux 335K-300gsm damage variable di which ranges between 0 and 1 [25]. The value
adhesive films. The quasi-static and drop-weight impact tests were of di = 0 means the material is intact and behaves elastically while
conducted on Instron 5500R and Instron Dynatup 8250 respec- di > 0 means the material is damaged due to one of the damage/
tively. The test jig that holds the test specimens is identical in de- failure modes. When di = 1, the material has failed and the affected
sign for both the quasi-static and the low-velocity impact tests. The element will be deleted from the finite element model in the next
details of the experimental set-up for quasi-static test and low- iteration.
velocity impact test are shown in Figs. 1 and 2 respectively.
The specimen is fixed between the steel clamps with a hole of
3. Damage modes and limit load prediction
76.4 mm diameter in the center as shown in Fig. 3a. The specimen
is indented with a hemispherical-end steel tup of radius of
The competing damage/failure modes and corresponding limit
6.55 mm. The indentation loading was controlled by displacement
loads are tabulated clearly in Fig. 4. When a sandwich panel is sub-
at a constant cross-head speed of 0.5 mm/min. During testing, a
jected to transverse loading, it is widely accepted that the displace-
linear variable displacement transducer mounted under the speci-
ment of impactor is the sum of the indentation of top facesheet and
men recorded the displacement of the center of the indenter. For
the global bending deflection [15,25–27].
each test, the load versus indenter displacement data is collected
The limit loads for damage initiation are reviewed here. The fol-
via a digital data acquisition system. For the low-velocity impact,
lowing assumptions are used: (i) The indentation of the top face-
the free-falling impactor is allowed to fall along two smooth
sheet is small and thus the deformation of the top facesheet in/
guided columns upon released. After the first impact, the rebound
around contact region are bending dominated, (ii) Hertz contact
brake is activated to ensure that the specimen is impacted only
law can be applied to analyze the contact between impactor and
once. The total displacement of impactor and top skin deflection
the top facesheet [27]. These assumptions will be not valid after
were recorded as a function of time with a data acquisition system.
the core has buckled as the top facesheet may undergo large
The limit load for damage initiation and failure obtained from the
indentation.
To explore the limit loads, it is assumed that bending response
dominates the deformation of the top facesheet during small
deflection and membrane response becomes dominant during
large deflection. The responses are regarded as quasi-static and
strain-rate effect is ignored in the present work. All damages
occurring in the experimental results are found to initiate in local-
ized region. Thus the damage modes presented in this section are
limited to only the localized region. This means that the damage
initiation is independent on the boundary condition while depen-
dent on the radius of impactor, dimensions in transverse direction
of the paper, mechanical properties of the facesheets, the core and
the adhesive.

3.1. Matrix shear damage

Shear stress in matrix of the top facesheet is highly concen-

trated in the localized area of contact. Thus shear damage in the
matrix may occur for thick facesheet combined with a soft core
Fig. 1. Quasi-static experiment set-up with Instron 5500R machine. when the shear stress exceeds the yield shear stress sf. Since this
206 S. Zhu, G.B. Chai / Composite Structures 101 (2013) 204–214

Fig. 2. Low-velocity impact test set-up with Instron Dynatup 8250 machine.

13.1 mm
Sandwich panel Clamp
v0

∅ 76.4 mm
100 mm

(a) (b)
Fig. 3. Illustration of drop-weight test for (a) experimental set-up and (b) Finite element model.

where E3c and hc are the transverse modulus and thickness of the
Table 1 core respectively. The symbol Df is the effective bending rigidity
Configuration of sandwich panels and their properties. of facesheet and this can be derived as Df ¼ 18 ð3D11 þ 3D22 þ
Sandwich panel configuration Impact energy (J) Eff. bending
2D12 þ 4D66 Þ where Dij are the laminate bending stiffnesses.
rigidity, Df Substituting Eqs. (2) and (3) into Eq. (1), the limit load for shear
ID Facesheet Core Facesheet
(N mm) damage can be derived as:
layup height thickness
(mm) (mm)
Ri ðpt f sf Þ2
S1 [+45/45/0/90/0]s 20 1.25 14.1 10.32 Pshear
I ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð4Þ
Df E3c =hc
S2 [0/90/0/90/0]s 15 1.25 13.7 10.31
S3 [0/90]8 25 2.0 24.8 42.28
As can been seen in Eq. (4), limit load Pshear
I is a function of the
S4 [45/45/0/90]2S 30 2.0 24.8 42.29
S5 [0/90]10 15 2.5 7.2 77.44
radius of impact, the stacking sequence of the laminate and thick-
ness facesheets, and the fundamental mechanical properties of the
facesheets and the through thickness modulus of the core.
damage mode usually occurs in a sandwich panel with a soft core
and thick facesheets, the support from the core can be ignored and
3.2. Core bucking
the limit load Pshear
I is given as [28]:

Pshear
I ¼ 2pRc t f sf ð1Þ Results obtained from the quasi-static tests showed that the
indentation failure of the sandwich panel is mainly core dependent
where tf is the thickness of the facesheets. The contact radius, Rc, [22,30]. Core buckling is widely reported as the first damage mode
can be expressed as a function of the radius of the impactor Ri that occurs in experiments involving low impact energy levels [31].
and the indentation a [29]: Since honeycomb core is made of hexagonal interlinked thin-
pffiffiffiffiffiffiffiffiffiffi walls, this damage mode will be the dominant mode for large core
Rc ¼ 2Ri a ð2Þ
thickness of hc. For this sandwich panel, the threshold load has
And the indentation is given as: been developed [32]:
P pffiffiffi 2Df 3
2
a ¼ pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð3Þ
Pbuckling ¼ 3 3rcu ð5Þ
8 Df E3c =hc I
E3c
S. Zhu, G.B. Chai / Composite Structures 101 (2013) 204–214 207

Top/bottom facesheets Honeycomb core

Composite sandwich panel

Damage initiated

Facesheet shear/tensile • Core buckling

damage • Debonding between Damage modes
facesheet/the core

Damage interaction,
Damage evolution,
Damage progression,
Damage propagation

• Top facesheet cracking • Core shear Failure modes

• Core crushing

Fig. 4. Damage/failure modes of sandwich panel subjected to impact loading.

where rcu is the compressive strength in the thickness direction of 9p3 Y 3t R2i
the core. As shown in Eq. (5), the limit load for core buckling is a Ptension
I ¼ 3
ð8Þ
2ð1 2m12 Þ E2
function of the bending rigidity of the facesheet, compressive
strength and thickness direction modulus of the core. The limit load The limit load Ptension
I in Eq. (8) is highly dependent on the radius
of Eq. (5) is independent of impactor/indenter radius while experi- of impactor as well as the tensile strength in transverse fiber-direc-
mental results show that the radius Ri may affect the limit load for tion. Based on the assumption of small deflection of top plate and
core buckling [15,22,33–35]. Hertz contact law, the limit load is independent of the dimensions
of the sandwich panel.
3.3. Matrix tension damage
3.4. Delamination, debonding between facesheets and the core
Matrix damage may decrease the localized stiffness of the pan-
els but is usually not catastrophic until the matrix of the facesheet
Delamination is one of the most common damage initiation
cracks and finally the fibers in the facesheet breaks. Experimental
modes in laminated composites and in sandwich structures. Delam-
results showed that the matrix in the top facesheet is prone to ten-
ination between plies in laminates and debonding between face-
sion crack as a result of tensile stress caused by contact [36]. The
sheets and core will occur when the inter-laminar shear strength
tension crack perpendicular to the fiber-direction due to the stres-
or bond strength is exceeded as illustrated in Fig. 6. Although not
ses of r11 and r13 throughout the impact event. The maximum ten-
a catastrophic failure mode, debonding is also found to decrease
sile stress in the top side of top facesheet is located in the contact
the stiffness and strength significantly and these were proven via
periphery. The tensile stress rr is maximum at the contact periph-
experimental results and theoretical predictions [34,40].
ery as shown in Fig. 5.
There are already many literatures published on delamination
For laminated composite, the tension strength, Yt, in transverse
of composite laminates [41–45] and many conclusive deductions
direction is much less than that of the fiber direction, Xt. As such,
can be drawn from their results. For the damage mode of debond-
the matrix tension damage is the most common damage mode
ing, the limit load is not only dependent on geometry but also on
and there are many supportive experimental evidences
the fracture energy of Mode-II inter-laminar fracture energy GIIc
[12,20,36,37]. The damage initiation by tensile stress in the matrix
[41–50]. It is also widely agreed that debonding of sandwich panels
at point A shown in Fig. 5b is based on the maximum stress crite-
is highly dependent on the Mode II inter-laminar fracture energy,
rion [20]:
GIIc [42,46,47,49,51]. Mode-II inter-laminar fracture toughness en-
1 2m12 ergy values were significantly higher than mode-I values [52,53].
rr ¼ p0 Y t ð6Þ
Analytical models [50,52,54] are also available to predict the value
3
of GIIc. And experimental methods such as 3/4-ENF(End-Notched
where m12 is the in-plane Poisson’s ratio of the lamina and p0 is the
Flexure), End Loaded Split (ELS), Single Leg Bending (SLB) tests
maximum contact pressure given by [38]:
[34,35,38,39], are available to measure the value of GIIc. Some val-
!13
ues of GIIc tested under laboratory conditions are reviewed in Ta-
1 6PE2
p0 ¼ ð7Þ ble 2 for reference.
p R2i The analytical prediction of threshold load for delamination of
The details of the derivation of the parameter E can be found in laminates was originally given by Davies and Zhang [57]:
the Appendix A of Ref. [39]. sffiffiffiffiffiffiffiffiffiffiffiffiffiffi
Substituting Eq. (7) into Eq. (6), the limit load for matrix tension 2p Et3 GIIc
Pdel ¼ ð9Þ
damage becomes: 3 1 t2
208 S. Zhu, G.B. Chai / Composite Structures 101 (2013) 204–214

Localized region
Ri Rc
tf Contact region
Ri

hc Rc
Honeycomb core

A
Rp Fibre

(a) Side view of deformation (b) Point A in top facesheet

Fig. 5. Damage mode initiated by matrix tension.

Debonding Table 2
Delamination Mode II inter-laminar fracture energy, GIIc, for different adhesives by experiment.

Adhesives XD4600 Araldite AS4/BP-907 AS4/PEEK ESP110 AF126

@ 2015
GIIc (kJ/m2) 5.40 [55] 4.67 [56] 2.33 [53] 2.01 [53] 5.87 [52] 3.58 [52]
Facesheet Core

Fig. 6. Delamination and debonding in sandwich panel. properties are still not yet matured. Therefore, the threshold load
of delamination and debonding for sandwich structures is still
not well-predicted. For this reason, explicit formula for delamina-
where E is the mean elastic modulus of the laminate in the 0° and tion and debonding of sandwich panel cannot be included in the
90° direction and Poisson’s ratio t is approximated as 0.3. Experi- present contribution and need future investigations.
ments conducted by Yang and Cantwell in 1995 [58] showed that
3
the limit load for delamination is proportional to t 2 which is consis- 3.5. Damage mode map configuration and limit load prediction
tent with theoretical solution shown in Eq. (9). In 2006, Olsson [43]
developed another formula for threshold load for delamination: The various explicit equations to predict the different damage
rffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi modes reviewed so far are tabulated in Table 3 for easy reference
32Ddel GIIc and these will be used to plot the damage mode map in the subse-
Pdeb
I ¼p ð10Þ
3 quent section.
where Ddel of the laminates is defined as
qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 4. Failure modes and limit load prediction
1þg
Ddel ¼ 2
D11 D22
pffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi ð11Þ
In this section, the various theories and corresponding explicit
g ¼ ðD12 þ 2D66 Þ= D11 D22
equations to predict the various failure modes pertaining to sand-
3
Note that the threshold load in Eq. (10) is also proportional to t 2 wich structures during bending, and also considering the low
since Ddel is proportional to t3 with a given stacking sequence of velocity impact, are reviewed.
laminates.
However, the case of debonding in sandwich panels may be 4.1. Facesheet cracking (facesheet fracture)
more complicated than delamination in composite laminates since
the core is much softer than the facesheet. The debonding along Another common failure mode for sandwich panels subjected to
the facesheets and the core interface of sandwich panels is found indentation test and to low velocity large mass impact test is the
to be predominated by a sliding mode [46]. It was found that the fiber breakage in the facesheet or facesheet cracking
critical energy released rate is influenced by the honeycomb prop- [12,19,27,34], this is illustrated in Fig. 7. When the thickness of
erties and there are experimental proofs [47]. Numerical simula- facesheet is thin compared to the other dimensions of the panel
tions using Irwin’s crack closure integral approach revealed that and considering that the core is of sufficient stiffness, the global
increasing the cell wall thickness induces a corresponding increase deflection of the sandwich tends to be large and as a result there
in the energy release rate [51]. Furthermore, the effect of elastic arises high in-plane tensile forces that will cause tensile cracking
modulus of the core on debonding has been found to be significant, [9,12]. A large diameter hemispherical-end impactor can cause
and low relative density of the core will results in more critical plugging in thin facesheet if the radial tensile strain reaches the
debonding [46]. tensile rupture strain of the facesheet. For sandwich panel with
In contrast to debonding in sandwich beams [42,49], assuming thick facesheets, the global deflection will be small and transverse
small deflection analysis the theoretical threshold load of a sand- shear force in regions that surround the impactor can be very high
wich panel is independent of the delamination radius and there- and this can cause transverse shear cracking.
fore remains constant under static axisymmetric delaminations The bottom facesheet may failed if the core crushing resistance
growth. Despite extensive experimental work on delamination is high enough to resist local deformation of top facesheet [28].
and debonding of sandwich panels [42,45–47,49,51], their impact This failure mode was however not proven yet in any reported
S. Zhu, G.B. Chai / Composite Structures 101 (2013) 204–214 209

experimental results. In the current work, the failure mode of face- thick skin and weak core as shown in Fig. 8. But for sandwich panel
sheet cracking was considered to occur only in top facesheet due to with a stiff core and thin facesheet, the failure mode is likely to be
localized deformation. facesheet cracking as discussed earlier. To investigate the core
Based on the maximum strain criterion, a theoretical prediction shear failure mode of sandwich panel with the core made of Core-
for sandwich panels with cross-plied facesheets was proposed to mat, Fatt and Sirivolu [59] developed an explicit formula for the
predict the damage initiation load for circumferential cracking limit load:
[28]: 2
pffiffiffiffiffiffiffiffiffi 4phc c2c
Pshear ¼ ð16Þ
PfII ¼ 2pRe A11 ecr 2ecr þ K c pqR2e ð12Þ II
q
and the damage initiation load for radial cracking is: where q is the crushing stress and cc is the shear yield stress of the
pffiffiffiffiffiffiffiffiffi core. However, it may not be applicable to the sandwich panel with
PfII ¼ dA11 ecr 2ecr þ K c pqR2e ð13Þ honeycomb [16]. Taking into account of the localized region of the
where A11 is the laminate extensional stiffness, q is the crushing isolated honeycomb core by a rigid impactor, when the core shears
strength of the core, Re is the critical radius of contact onset of face- and the deformation of top facesheet is dominated by bending, the
sheet cracking which can be approximated as Re ¼ 25 Ri , ecr is the ten- crushing load is given by [28]:
sile fracture strain for facesheet and the coefficient Kc can be taken 1
512 pffiffiffiffiffiffiffiffiffi 255D1 c4c 6
an average value of Kc = 2, and d is the total damage length. The lim- Pshear
II ¼ D1 q þ pqR2e ð17Þ
it load of Eq. (12) can be re-written as 15 q
pffiffiffiffiffiffiffiffiffi and when the deformation of top facesheet is dominated by mem-
PfII ¼ 2pRe A11 ecr 2ecr þ 2pqR2e ð14Þ
brane, the crushing load is:
Ignoring the second part of Eq. (14) and assuming that the lam-
A21 c6c
inated facesheet is composed of n layers, Eq. (14) can be re-written Pshear
II ¼ 72 þ pqR2e ð18Þ
as q
pffiffiffi where D1 ¼ 1024 ½D11 þ D22 þ 47 ðD12 þ 2D66 Þ, A1 ¼ 45
8 90
½A11 þ A22 þ 49
4 2p 3 1575
PfII ¼ Q 1 t f Ri e2cr ð15Þ ðA12 þ 2A66 Þ, Dij and Aij are the laminate bending and membrane
5
Pn stiffness of the facesheets.
where the coefficient Q 1 ¼ 1n k¼1 ðQ 11 Þk . For a sandwich panel with a soft core, the top facesheet is prone
to large deflection and thus Eq. (18) is valid. Neglecting the second
4.2. Core shear yield/fracture part of Eq. (18) yields:

The low-density Nomex cores are particularly susceptible to the A21 c6c
Pshear
II ¼ 72 ð19Þ
failure mode of core shear [2], especially for sandwich panel with q

4.3. Overall core crushing

Table 3 For sandwich panels consisting of very stiff facesheets com-

Summary of damage modes and limit loads. bined with a very soft core, the impactor and the top skin may
Damage Limit loads Conditions for damage initiations move down rigidly while the core crushes uniformly. This is re-
initiation ferred as overall core crushing as shown in Fig. 9. In this mode of
Matrix shear R ðpt s Þ2 Small impactor size failure, the honeycomb core is essentially subjected to uniaxial
P shear
I ¼ pi ffiffiffiffiffiffiffiffiffiffiffiffiffiffi
f f

damage Df E3c =hc

transverse compression. The load capacity of this mode is thereby
Thin facesheets simply given as:
Core of large relative density
Low shear strength in Pcrush
II ¼ prcu R2p ð20Þ
transverse fiber-direction
Core bucking pffiffiffi 23 Thin facesheets
P buckling ¼ 3 3rcu
2Df where Rp is the radius of the sandwich panel.
I E3c
High core thickness
Core of low relative density 4.4. Failure mode map
Matrix tension P tension ¼
9p 3 Y3
t R2i Small impactor size
I 2ð12m12 Þ E 3 2
damage The explicit equations to predict the modes of failure reviewed
Hard matrix
Core of large relative density
so far are tabulated in Table 4 for easy reference and these will be
Low tension strength in used to plot the failure mode map towards the end of the paper.
transverse fiber-direction

Fiber breakage Core shear

Core bucking

Facesheet Core Facesheet Core

Fig. 7. Facesheet cracking of sandwich panels. Fig. 8. Core shear yield/facture of sandwich panels.
210 S. Zhu, G.B. Chai / Composite Structures 101 (2013) 204–214

5. Energy absorption Table 4

Summary of failure modes.
From the vast experimental database available on sandwich
Failure Limit loads Conditions for failure initiations
panels subjected to quasi-static indentation and low velocity im- modes
pact, a typical load–displacement curve of a sandwich panel before pffiffi
Facesheets P fII ¼ 4 2p
3
e Low tensile strength in fiber-direction by
failure can be divided into two stages: the 1st stage is linear and 5 Q 1 t f Ri cr
2

cracking low strength fiber and/or low fiber

the 2nd stage begins after some damage initiation as shown in content
Fig. 10. Note that if any one of the damage modes in Fig. 4 is initi- Small impactor size and thin facesheet
ated, other damage modes maybe triggered and there maybe cou- Core shear P shear ¼ 72
A21 c6c Soft core combined with thick and stiff
II q
yield/ facesheets
pling of damage modes. As a result, the damage interaction and
facture
damage progression in 2nd stage are coupled until final failure. Low shear strength of the core
The area under this curve is approximately equal to the energy Overall core P crush ¼ prcu R2p High bending stiffness facesheets with
II
absorbed Eabs during impact [34]. To design a sandwich panel with crushing soft core
higher impact resistance, the energy absorption is a vital parame- Short boundary condition

ter used as a gauge.

6. Results and discussions The limit loads predicted using the equations of Table 3 are
listed in Table 6 for specimens S1–S5 and these predictions are
The limit loads corresponding to the damage and failure modes plotted in Fig. 12. As shown in Fig. 12, the damage initiation of
are calculated with the formulae tabulated in Tables 3 and 4. These specimens S1 and S2 are predicted to be core buckling and the
modes are also simulated using finite element analysis as de- damage initiation of specimens S3, S4 and S5 are predicted to be
scribed earlier. A typical illustration of the response and the dam- matrix tension damage.
age/failure points of specimen S2 are plotted in Fig. 11. Damage The observed damage initiations from the experimental results
variables at the contact edge are also plotted to illustrate when are compared with those predicted using the finite element analy-
the damage modes are initiated. Take matrix cracking for example, sis. A summary of the comparison is tabulated in Table 7. The dam-
this damage mode occurs when the top facesheet deflects by about age modes predicted using finite element analysis and that
0.73 mm while core buckling occurs at a deflection of 0.22 mm, predicted using the equations of Table 3 gave consistent results
which means core buckling is the damage initiation mode for spec- for specimens S1 and S2, but gave different modes for specimens
imen S2. S3, S4 and S5. This is could be due to the different damage criteria
used in deriving the equations of Table 3. The finite element anal-
6.1. Damage modes initiation map ysis used the 3-D Hashin damage criteria.

The equations for the damage mode listed in Table 3 are plotted 6.2. Failure modes map
t
and configured in Fig. 12 in which the Rfi hRci plane is divided into
three sections corresponding to the three damage modes. The var- The competing failure modes using the equations listed in Ta-
ious necessary inputs listed in Table 5 were used in plotting the ble 4 are mapped out in Fig. 14 using the input parameters of Ta-
map of Fig. 12. ble 5, and the value of the maximum loads are tabulated in
Core buckling and matrix tension damage were found to be Table 8. The failure mode of overall core crushing was located in
dominant in the map. Matrix tension damage may be initiated a narrow region at the bottom of Fig. 14 and this condition is not
when the thickness of facesheet exceeds a critical value. Matrix a common practical structural configuration. The subsequent dis-
shear damage is found to occur only for small core thickness. Note cussion will focus on the failure modes of facesheet cracking and
that the value of Df is highly dependent on thickness of laminates core shear.
and mechanical properties of the laminated facesheets and less As shown in Table 8, the predicted limit load that cause the fail-
dependent on the stacking sequence. This means that thicker face- ure mode of facesheet cracking is overestimated while that of core
sheet will yield a higher limit load when applying Eqs. (4) and (5), shear is underestimated. This can be explained by the fact that
and different stacking sequence of the laminates yields insignifi- even when a sandwich panel is subjected to load that cause the
cant difference in the limit loads of Table 3. From the results of failure mode of core shear, any further increase in load can be sup-
the indentation tests, core buckling is the most likely damage ported until the facesheet cracks in the top facesheet.
mode as the load increases as shown in Fig. 13. Damage mode of The maximum loads predicted using finite element analysis and
core buckling is the preferred initiation mode of damage among the equations of Table 4 are compared with the experimental re-
the three damage modes as this means higher impact resistance. sults. This comparison is shown in Table 9. The percent difference

Overall core crushing

Facesheet Core

Fig. 9. Overall core crushing of sandwich panel.

S. Zhu, G.B. Chai / Composite Structures 101 (2013) 204–214 211

Damage initiation 2nd stage

P P
2nd stage
1st stage Failure
1st stage Damage progression

Damage initiation

Energy
absorbed
Energy absorbed, E abs
E abs
3rd stage
δ δ

Fig. 10. Energy absorbed by sandwich panel and damage progression.

5
6 2
Expt Core buckling Matrix
4 tension damage
FEA
Fiber breakage 1.5
3
h c /R i
Matrix cracking
4
Damage variables, di .

Shear damage 2
1
Force (kN)

1
Matrix shear damage
0.5
0
2 0.1 0.2 0.3 0.4 0.5
t f /R i
0
Fig. 12. Damage modes for composite sandwich panel Symbols for specimens: 4:
Onset of core buckling by expt S1; : S2; h: S3; +: S4; s: S5.

Onset of core buckling by FEA

0 -0.5
0 1 2 3 4 5 6 Table 5
Displacement (mm) List of important parameters used in the maps.

Fig. 11. The response and damage initiation/evolution of specimen S3 during Parameter Values
impact. Yield shear strength of facesheet sf = 101.2 MPa
Yield shear strain of core cf = 0.018
Tensile fracture strain of facesheet ecr = 0.0 126
between finite element results and those using the equations of Ta- Transverse modulus of core E3c = 150 MPa
ble 4 are within 30%. However, the failure modes predicted using Transverse compressive strength of core(75 kg/m3) rcu = 3.2 MPa
the equation of Table 4 are different from those predicted using fi- Transverse compressive strength of core(45 kg/m3) rcu = 2.6 MPa
Transverse yield strength of facesheet Yt = 65.5 MPa
nite element simulation for specimen S1 and S2 since different fail- Poisson’s ratio of facesheet m12 = 0.263
ure criteria were used. Effective stiffness at contact E = 2.5 MPa
The analysis using Eqs. (15) and (19) concludes that as the face-
sheet thickness increases, the limit loads for both modes will also
increase. This means increasing the facesheet thickness will signif- The damage and failure mode maps gave consistent trends of
icantly increase the ultimate loads. Likewise, increasing the core the effect of the size of the impactor. Smaller diameter of the
density will definitely leads to higher compressive strength. But impactor tends to yield predominantly facesheet damage in the
the analysis using Eqs. (15) and (19) will not result in higher limit form of matrix cracking/shear and facesheet cracking while the lar-
loads. The effect of facesheet thickness and core density on limit ger diameter of the impactor produces core buckling/crushing and
loads correlates well with those of the experimental results by core shear/fracture that dominated until the facesheet fracture ini-
Zhou and Hill [34]. tiated. These trends have been reported in a review of the experi-
However, the failure mode map in Fig. 14 is highly dependent mental results by Raju et al. [14].
on strength of the fiber-reinforced plastic and core density, and
also on the dimensions of the sandwich panel. In general, the face- 7. Concluding remarks
sheet thickness will not only affect the failure modes but also the
limit loads. In contrast, core density affects both the damage Damage initiation and failure modes were investigated theoret-
modes and failure modes but have little effect on limit loads. ically, experimentally and numerically. The damage initiation
212 S. Zhu, G.B. Chai / Composite Structures 101 (2013) 204–214

(a) Damage initiation (b) core crushing when facesheet cracking

Fig. 13. Damage initiation and core crushing of sandwich panel of specimen S2.

Table 6
Prediction of limit loads for damage mode initiation with minimum underlined.

Specimen ID Limit loads predicted by damage modes (kN) Minimum of predictions (kN)
Matrix shear damage Core buckling Matrix tension damage
S1 0.997 1.01 4.14 1.01
S2 1.60 1.45 3.89 1.45
S3 2.09 1.14 3.75 1.14
S4 2.08 1.25 2.30 1.25
S5 2.41 2.40 2.68 2.40

Table 7
Critical load by experiment compared with prediction by damage mode, FEA.

Specimen ID Critical load by experiment (kN) Prediction by FEA Prediction by damage modes
Limit load(kN) Damage initiation Error (%) Limit load (kN) Damage initiation Error (%)
S1 0.949 1.76 Core buckling 85.3 1.01 Core buckling 6.43
S2 1.50 2.62 Core buckling 74.7 1.45 Core buckling 3.27
S3 0.855 0.88 Core buckling 2.92 1.14 Matrix tension 15.6
S4 0.921 1.25 Core buckling 35.7 1.25 Matrix tension 18.3
S5 2.07 3.91 Core buckling 88.9 2.40 Matrix tension 15.8

5 Table 8
Comparison of limit loads for failure mode by experiment and prediction.
Core buckling Matrix
4 tension damage Specimen Limit loads predicted by failure modes(kN) Minimum of
ID predictions (kN)
Facesheet Core shear/ Overall core
cracking fracture crushing
3
hc /Ri

S1 3.32 3.24 9.26 3.24

S2 3.87 2.07 9.26 2.07
2 S3 5.26 5.28 9.26 5.26
S4 4.74 7.10 9.26 4.74
S5 6.58 8.26 9.26 6.58
1
Matrix shear damage

0
0.1 0.2 0.3 0.4 0.5
t f /Ri The failure modes of facesheet cracking, core shear fracture and
overall core crushing were presented and discussed in some de-
Fig. 14. Failure mode map for sandwich panel subjected to impact loading Symbols tails. It was found that overall core crushing is seldom seen in prac-
for specimens: 4: S1 and S2; h: S3 and S4; s: S5. tical design. The effect of facesheet thickness was found to be vital
contribution to the failure mode and its corresponding ultimate
load while core density affected the failure mode but not its corre-
modes such as the matrix shear/tension damage mode and core sponding ultimate load.
buckling mode were investigated. The conditions of each mode Damage mode and failure mode maps were presented and dis-
were reviewed and deliberated. Damage initiation was found to oc- cussed with deliberations. Predicted modes of certain sandwich
cur as a direct result of localized deformation and was independent configurations were compared with the experimental results and
of the boundary condition and stacking sequence of the laminated the overall predictions are satisfactory but more work are needed
facesheets. to be done to validate these maps.
S. Zhu, G.B. Chai / Composite Structures 101 (2013) 204–214 213

Table 9
Maximum loads by experimental results compared with FEA and failure mode prediction.

Specimen ID Max.load by experiment (kN) Prediction by FEA Prediction by failure modes

Limit load (kN) Failure modes Diff. (%) Limit load (kN) Failure modes Diff. (%)
S1 3.32 2.64 Facesheet cracking 20.5 3.24 Core shear 2.41
S2 3.06 4.02 Facesheet cracking 12.2 2.07 Core shear 32.4
S3 4.63 4.41 Facesheet cracking 4.75 5.26 Facesheet cracking 13.6
S4 4.72 5.28 Facesheet cracking 23.7 4.74 Facesheet cracking 0.424
S5 5.96 5.85 Facesheet cracking 1.85 6.58 Facesheet cracking 10.4

Acknowledgements [24] Hashin Z. Failure criteria for unidirectional fibre composites. J Appl Mech
1980;47:329–34.
[25] Lapczyk I, Hurtado JA. Progressive damage modeling in fiber-reinforced
The authors want to acknowledge the financial supports of materials. Composites: Part A 2007;38(11):2333–41.
School of Mechanical and Aerospace Engineering, Nanyang Tech- [26] Hibbitt K, Karlsson B, Sorensen D. ABAQUS Analysis User’s Manual. ABAQUS,
Inc. vol. version 6.10. 2010.
nological University, Singapore in the form of Research Student
[27] Abrate S. Impact on composite structures. Cambridge University Press; 1998.
Scholarship for this PhD. programme, and for their permission to [28] Fatt MSH, Park KS. Dynamic models for low-velocity impact damage of
use the computing and laboratory facilities. composite sandwich panels – part B: damage initiation. Compos Struct
2001;52(3–4):353–64.
[29] Zhu S, Composite Sandwich Panels Subjected to Impact of a Foreign Body, PhD
References Thesis 2013, Nanyang Technological University, Singapore.
[30] Tsotsis TK, Lee SM. Characterization of localized failure modes in honeycomb
[1] Andrews EW, Moussa NA. Failure mode maps for composite sandwich panels sandwich panels using indentation. ASTM Spec Tech Publ 1996;1274:139–65.
subjected to air blast loading. Int J Impact Eng 2009;36:418–25. [31] Nettles AT, Hodge AJ. Impact testing of glass/phenolic honeycomb panels with
[2] Petras A, Sutcliffe MPF. Failure mode maps for honeycomb sandwich panels. graphite/epoxy facesheets. In: 35th International SAMPE symposium,
Compos Struct 1999;44:237–52. Anaheim, CA; 1990.
[3] Triantafillou TC. Failure mode maps and minimum weight design for structural [32] Olsson R. Engineering method for prediction of impact response and damage
sandwich beams with rigid foam cores. Massachusetts Institute of Technology; in sandwich panels. J Sandwich Struct Mater 2002;4:3–19.
1987. [33] Zhou G, Hill M, Loughlan J. Damage characteristics of composite honeycomb
[4] Triantafillou TC, Gibson LJ. Failure mode maps for foam core sandwich beams. sandwich panels in bending under quasi-static loading. J Sandwich Struct
Mater Sci Eng 1987;95:37–53. Mater 2006;8(1):55–90.
[5] Zenkert D, Burman M. Failure mode shifts during constant amplitude fatigue [34] Zhou G, Hill M. Investigation of parameters governing the damage and energy
loading of GFRP/foam core sandwich beams. Int J Fatigue 2011;33:217–22. absorption characteristics of honeycomb sandwich panels. J Sandwich Struct
[6] Lim TS, Lee CS, Lee DG. Failure modes of foam core sandwich beams under Mater 2007;9(4):309–42.
static and impact loads. J Compos Mater 2004;38(18):1639–62. [35] Zhou G, Hill MD. Impact damage and energy-absorbing characteristics and
[7] Raja SV, Ghosnb LJ. Failure maps for rectangular 17-4ph stainless steel residual in-plane compressive strength of honeycomb sandwich panels. J
sandwiched foam panels. Mater Sci Eng: A 2008;474(1–2):88–95. Sandwich Struct Mater 2009;11(4):329–56.
[8] Crupi V, Epasto G, Guglielmino E. Collapse modes in aluminium honeycomb [36] Richardson MOW, Wisheart MJ. Review of low-velocity impact properties of
sandwich panels under bending and impact loading. Int J Impact Eng composite materials. Composites: Part A 1996;27:1123–31.
2012;43:6–15. [37] Hampson P, Moatamedi M. A review of composite structures subjected to
[9] Zhu F, Lu G, Ruan D, Wang Z. Plastic deformation, failure and energy absorption dynamic loading. Int J Crashworth 2007;12(4):411–28.
of sandwich structures with metallic cellular cores. Int J Protect Struct [38] Johnson KL. Contact mechanics. Cambridge: Cambridge University Press; 1985
2010;1(4):507–41. [V.P.W. Shim].
[10] Nguyen M, Elder D, Bayandor J. A review of explicit finite element software for [39] Swanson SR. Contact deformation and stress in orthotropic plates. Composites:
composite impact analysis. J Compos Mater 2005;39:375–86. Part A 2005;36:1421–9.
[11] Fan X, Verpoest I, Vandepitte D. Finite element analysis of out-of-plane [40] Zhu S, Chai GB. Effect of adhesive in sandwich panels subjected to low-velocity
compressive properties of thermoplastic honeycomb. J Sandwich Struct Mater impact. Proc Institut Mech Eng, Part L: J Mater Des Applic
2006;8:437–58. 2011;225(3):171–81.
[12] Chai GB, Zhu S. A review of low-velocity impact on sandwich structures. Proc [41] Elder DJ, Thomson RS, Nguyen MQ, Scott ML. Review of delamination
Institut Mech Eng, Part L, J Mater: Des Applic 2012;225(4):207–30. predictive methods for low speed impact of composite laminates. Compos
[13] Chow CL, Lu TJ. On evolution laws of anisotropic damage. Eng Fract Mech Struct 2004;66(4):677–83.
1989;34(3):679–701. [42] Olsson R. Closed form prediction of peak load and delamination onset under
[14] Raju KS, Smith BL, Tomblin JS, Liew KH, Guarddon JC. Impact damage small mass impact. Compos Struct 2003;59:341–9.
resistance and tolerance of honeycomb core sandwich panels. J Compos [43] Olsson R. Delamination threshold load for dynamic impact on plates. Int J
Mater 2008;42(4):385–412. Solids Struct 2006;43:3124–41.
[15] Raju KS, Tomblin JS. Damage characteristics in sandwich panels subjected to [44] Zhang C, Wang J. Delamination of layered structures on elastic foundation. Eng
static indentation using spherical indentors. In: 42nd AIAA/ASME /ASCE/AHS/ Fract Mech 2011;78(5):1055–66.
ASC structural dynamics and materials conference and exhibit Seattle, WA; [45] Zheng D, Binienda WK. Effect of permanent indentation on the delamination
2001. p. 103–11. threshold for small mass impact on plates. Int J Solids Struct
[16] Bernard ML, Lagace PA. Impact resistance of composite sandwich pates. J Reinf 2007;44:8143–58.
Plast Compos 1989;8(5):432–45. [46] Goswamia S, Becker W. The effect of facesheet/core delamination in sandwich
[17] Schubel PM, Luo J-J, Daniel IM. Impact and post impact behavior of composite structures under transverse loading. Compos Struct 2001;54(4):515–21.
sandwich panels. Composites: Part A 2007;38(3):1051–7. [47] Guédra-Degeorges D. Recent advances to assess mono- and multi-
[18] Tomblin JS. Review of damage tolerance for composite sandwich airframe delaminations behaviour of aerospace composites. Compos Sci Technol
structures. U.S. Department of Transportation; 1999. 2006;66(6):796–806.
[19] Kim JK, Yu TX. Indentation, penetration and perforation of composite laminate [48] Khan SU, Alderliesten RC, Benedictusa R. Delamination growth in fibre metal
and sandwich panels under quasi-static and projectile loading. Key Eng Mater laminates under variable amplitude loading. Compos Sci Technol 2009;69(15–
1998;141:501–52. 16):2604–15.
[20] Stronge WJ, Matemilola SA. Rate effect for impact damage initiation in CFRP [49] Okada R, Kortschot MT. The role of the resin fillet in the delamination of
laminates. In: 4th International conference on mechanical and physical honeycomb sandwich structures. Compos Sci Technol 2002;62(14):1811–9.
behaviour of materials under dynamic loading, Paris, France; 1994. p. 225–30. [50] Wang J, Zhang C. Energy release rate and phase angle of delamination in
[21] Olsson R, McManus HL. Improved theory for contact indentation of sandwich sandwich beams and symmetric adhesively bonded joints. Int J Solids Struct
panels. AIAA J 1996; 34(6): 1238–44. 2009;46:4409–18.
[22] Lee SM, Tsotsis TK. Indentation failure behavior of honeycomb sandwich [51] Goswami S, Becker W. Analysis of debonding fracture in a sandwich plate with
panels. Compos Sci Technol 2000;60(8):1147–59. hexagonal core. Compos Struct 2000;49(4):385–92.
[23] Hanssena AG, Girardc Y, Olovssond L, Berstadb T, Langseth M. A numerical [52] Blackman B, Kinloch A, Paraschi M. The determination of the mode II adhesive
model for bird strike of aluminium foam-based sandwich panels. Int J Impact fracture resistance, GIIC, of structural adhesive joints: an effective crack length
Eng 2006;32(7):1127–44. approach. Eng Fract Mech 2005;72(6):887–97.
214 S. Zhu, G.B. Chai / Composite Structures 101 (2013) 204–214

[53] Chai H. Interlaminar shear fracture of laminated composites. Int J Fract [56] Moura M, Campilho R, Gonçalves J. Pure mode II fracture characterization of
1990;43(2):117–31. composite bonded joints. Int J Solids Struct 2009;46(6):1589–95.
[54] Schellmann N, Taylor M. The effect of consolidants on the mechanical fracture [57] Davies G, Zhang X. Impact damage prediction in carbon composite structures.
behaviour of gesso-type foundations in multilayered decorative coatings. In: Int J Impact Eng 1995;16(1):149–70.
Proceedings of symposium 2011 – adhesives and consolidants for [58] Yang FJ, Cantwell WJ. Impact damage initiation in composite materials.
conservation, Ottawa; 2011. Compos Sci Technol 2010;70(2):336–42.
[55] Yang QD, Thouless MD, Ward SM. Elastic-plastic mode-II fracture of adhesive [59] Fatt M S H, Sirivolu D, Impact perforation of composite sandwich panels, in
joints. Int J Solids Struct 2001;38(18):3251–62. 16th international conference on composite materials, Kyoto, Japan, 2007.