Professional Documents
Culture Documents
The Mechanical Behaviour of Corrugated-Core Sandwich Panels
The Mechanical Behaviour of Corrugated-Core Sandwich Panels
Composites: Part B
journal homepage: www.elsevier.com/locate/compositesb
a r t i c l e i n f o a b s t r a c t
Article history: A series of experimental investigations and numerical analyses is presented into the compression
Received 23 July 2012 response, and subsequent failure modes in corrugated-core sandwich panels based on an aluminium
Received in revised form 5 October 2012 alloy, a glass fibre reinforced plastic (GFRP) and a carbon fibre reinforced plastic (CFRP). The corru-
Accepted 25 October 2012
gated-cores were fabricated using a hot press moulding technique and then bonded to face sheets based
Available online 22 November 2012
on the same material, to produce a range of lightweight sandwich panels. The role of the number of unit
cells and the thickness of the cell walls in determining the overall deformation and local collapse behav-
Keywords:
iour of the panels is investigated. The experiments also provide an insight into the post-failure response
A. Carbon fibre
A. Glass fibre
of the sandwich panels. The results are compared with the numerical predictions offered by a finite ele-
B. Strength ment analysis (FEA) as well as those associated with an analytical model. Buckling of the cell walls has
C. Finite element analysis been found to be initial failure mode in these corrugated systems. Continued loading resulted in fracture
of the cell walls, localised delamination as well as debonding between the skins and the core. The predic-
tions of the FEA generally show reasonably good agreement with the experimental measurements.
Finally, the specific compressive properties of the corrugated structures have been compared to those
of other core materials where evidence suggests that these systems compare favourably with their more
conventional counterparts.
Ó 2012 Elsevier Ltd. All rights reserved.
1359-8368/$ - see front matter Ó 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.compositesb.2012.10.031
268 M.R.M. Rejab, W.J. Cantwell / Composites: Part B 47 (2013) 267–277
In recent years, a number of researchers have investigated the materials. Along with the usual length-to thickness ratio of the
behaviour of corrugated-core panels under lateral compression composites, the number of plies per unit thickness is also impor-
loading. Côté et al. [7] investigated corrugated and diamond lattice tant. Although the former governs the likelihood of global buckling,
materials (prismatic cores) based on stainless steel, manufactured the latter influences failure at a microscopic scale, which is more
by slotting and brazing methods, and then compared their re- unpredictable. This can be understood from the fact that failure in
sponses with analytical and FE predictions. They found that the thick composites often involves more complex fracture modes than
measured strengths were below the analytical predictions, due to in thin composites. The reason for this is that a thick composite con-
the presence of imperfections. The authors concluded that corru- tains a greater number of plies. Both the number of ply interfaces
gated-cores offer significant potential for applications in sandwich and the total interfacial area are increased in proportion to the
beam construction, since they exhibit what they consider to be a thickness of the laminate, both increasing the possibility of interfa-
near optimal longitudinal shear strength and energy-absorption cial defects and delamination. From a statistical point of view, the
capacity. The transverse shear strengths of corrugated composite possibility of a composite containing both more and larger defects
structures will clearly be lower than their longitudinal values, increases with thickness. The compressive strength of a composite
although they remain well above those of diamond cores and foam tends to decrease with increasing thickness, despite the fact that
cores [7]. Recently, Kazemahvazi et al. [10,11] proposed a novel accurate experimental validation is difficult, due to the wide level
corrugation concept where each core member is made in a sand- of scatter in the results [16]. Compressive failure occurs as a result
wich configuration, creating a hierarchical sandwich structure to of fibre instability, due to misalignment of the reinforcement.
eliminate buckling effects. The core structures were manufactured Thicker composites contain more layers, leading to a greater possi-
by combining a unidirectional carbon fibre prepreg with a PMI bility of fibre misalignment [17].
foam and were then modelled using both analytical and FE tech- In this paper, the compressive properties of corrugated-core
niques. They showed that if designed correctly, hierarchical struc- sandwich panels based on three different materials are investi-
tures can offer flexural strength to weight ratios that are up to gated. Particular focus is placed on identifying the influence of
seven times greater than that of their monolithic counterparts. the number of unit cells and the thickness of the cell walls in deter-
Zhang et al. [15] discussed the crush behaviour of four types of cor- mining the overall deformation and local collapse behaviour of the
rugated core (V-type, U-type, X-type and Y-type cores) sandwich panel. Attention also focuses on establishing the fracture modes
panels made from mild steel, in which the corrugated-cores were during compression loading. Finally, the properties are compared
connected through a double-angle welding technique. By ignoring to those of other more conventional core materials in order to
imperfections, the finite element models over-predicted the peak gauge their overall potential.
crushing strength of most of the core designs.
The failure mechanisms observed in composites can vary greatly
2. Experimental procedure
from thin to thick laminates. For composites, the concept of being
thick should be defined in a different manner more than for uniform
2.1. Materials and fabrication process
(a)
2cm
Deetail C
(b)
All units in mm
Fig. 1. (a) Corrugated mould is made from steel and, (b) the profile angle of the mould.
of the core and HS is the overall height of the sandwich panel; HU bonded to the upper and lower skins using a two part epoxy adhe-
and HL are the upper and lower thicknesses of the skins, respec- sive (Araldite 420 A/B). The corrugated sandwich panel was then
tively; H is the average thickness of inclined core member i.e. the heated in an oven to a temperature of 120 °C for approximately
wall thickness; x is the length of the core; and w is the width of 1 h in order to cure the adhesive. As shown Fig. 2b–d, the structure
a sample. For the current mould design, the value of x was was cut into test specimens with dimensions of 100 mm (length)
20 mm while h and b were set to 45° and 90°, respectively. During and 25 mm (width). The specimen density, q was determined by
preparation of the test specimens, the value of the width, w was dividing the specimen mass (including the epoxy adhesive) by
maintained constant at 25 mm. The relative density of the corru- the specimen volume (Hs.w.x). Two types of rectangular test spec-
gated-core is given as [7]: imen were prepared. Initially, the effect of increasing the number
of cells in the core was investigated for a constant cell wall thick-
2H
q ¼ ð1Þ ness. Here, the number of cells, n was varied between 1 and 5. Fol-
L sin 2h lowing this, the effect of varying the cell wall thickness, H was
investigated for a constant value of n = 5. Table 2 summarises the
2.3. Sandwich panels various geometries and properties of the test specimens.
A hot press was used to produce all of the corrugated sheets for 2.4. Compression tests
the sandwich panels. In order to manufacture the composite corru-
gations, the composite prepreg was placed between the upper and Compression tests were conducted on the corrugated-core
lower moulds, and then cured according to the manufacturer’s rec- sandwich panels using an Instron series 4204 testing machine.
ommended processing cycle. Once the hot press had cooled to a Fig. 3 shows a specimen placed between the platens of the test ma-
temperature below 60 °C, the sheet was removed from the mould chine, where it was deformed by applying a uniform lateral com-
and visually inspected for defects. The corrugated sheet was then pression at a nominal displacement rate of 1 mm/min. At least
270 M.R.M. Rejab, W.J. Cantwell / Composites: Part B 47 (2013) 267–277
HU
L H
β
T
θ
HL w
x
(a)
(b)
(c)
(d)
Fig. 2. (a) Unit cell geometry of the corrugated-core sandwich panel (b) photo of an aluminium corrugation, (c) a GFRP corrugation and (d) a CFRP corrugation.
Table 2
Properties and characteristics of the corrugated-core sandwich panels.
Material Core ID No. of unit No. of Average wall thickness, H Average panel density, q(kg/ Relative core density, q
cell plies (mm) m3) (%)
Aluminium AL1 1 0.50 465 10.0
AL2 2 0.50 462 10.0
AL3 3 0.50 457 10.0
AL4 4 0.50 455 10.0
AL5 5 0.50 454 10.0
Glass fibre reinforced plastic GF2U1 1 2 0.19 228 3.8
GF2U2 2 2 0.19 218 3.8
GF2U3 3 2 0.19 213 3.8
GF2U4 4 2 0.19 210 3.8
GF2U5 5 2 0.19 202 3.8
GF3U5 5 3 0.31 220 6.2
GF4U5 5 4 0.43 238 8.6
GF5U5 5 5 0.50 261 10.0
GF7U5 5 7 0.77 292 15.4
GF10U5 5 10 0.98 363 19.6
Carbon fibre reinforced plastic CF2U1 1 2 0.45 260 9.0
CF2U2 2 2 0.45 259 9.0
CF2U3 3 2 0.45 256 9.0
CF2U4 4 2 0.45 248 9.0
CF2U5 5 2 0.45 244 9.0
CF3U5 5 3 0.89 296 17.8
CF4U5 5 4 1.03 340 20.6
CF5U5 5 5 1.10 350 22.0
CF6U5 5 6 1.25 390 25.0
three tests were carried out for each sample type. A high-resolu- 3. Finite element analysis
tion displacement measurement system was used to record the
displacement data as well as to identify the failure modes. The The responses of the corrugated-core sandwich panels under
load–displacement trace was recorded until the specimen had compression loading were modelled using the ABAQUS/Standard
been fully crushed. finite element software package. In these lateral compression tests,
M.R.M. Rejab, W.J. Cantwell / Composites: Part B 47 (2013) 267–277 271
Specimen
Recorder
Fig. 3. Photograph of the compression test set-up with the displacement measurement system.
the face sheets do not carry significant load and it is therefore suf- over-estimate both the peak load and the stiffness of the panel.
ficient to model the cores without skins, whilst applying the appro- Previous workers, for example Côté et al. [7] and Kazemahvazi
priate boundary conditions. In Fig. 4, three-dimensional shell and Zenkert [10] introduced imperfections in their models to more
elements (S4R) with six degrees of freedom were used. The core accurately predict the load–displacement behaviour of corrugated
and the platen were connected using a contact interaction formu- core geometries. The initial imperfection, in the shape of a buckling
lation. The nodes along the upper and lower edges were fully con- mode, was imposed on each strut. Here, the initial imperfection, a
strained, except in y-direction at the upper edge. The edges in the was described according to:
corrugations were modelled as perfectly. Clearly, this can influence
both the strength and stiffness predictions, although it does result
nH 2ps
in a more simple FE model. Displacements were applied uniformly aðsÞ ¼ 1 cos ð2Þ
2 L
to the nodes at the apex of the unit cell to simulate compression of
the core. The model comprised of 24 linear shell elements across
the width direction and 14 elements along the length of the struts, where n is a dimensionless imperfection parameter and s is the arc
giving a total of 3360 elements. Given that the corrugated-core was length along the wall thickness, H measured from one end. In ABA-
bonded to the skins, there is a softening interaction associated with QUS/Standard [18], this was achieved using the IMPERFECTION
the epoxy adhesive. To account for such effects, a softened contact function in an analysis step. An elastic buckling analysis was ini-
pressure-overclosure formulation was employed during the initial tially conducted to retrieve the eigenmodes of the core. From the
stage of contact between the sharp edge of the core and the platen. analysis, a selected eigenmode with similar behaviour of the exper-
For the aluminium corrugation, an isotropic hardening model, with imental was used to introduce a small imperfection in the straight-
true stress–strain data, was used. Hashin’s damage model assum- ness of the wall, which could then be used in the step procedure. An
ing, an initial linear elastic behaviour followed by evolution of imperfection sensitivity analysis was then performed, during which
damage, was used to predict the behaviour of the composite. The the imperfection amplitudes were varied systematically between 0
models were used to simulate initial failure and to predict the (a perfect corrugated geometry) and 0.5. The IMPERFECTION took
compression strength and stiffness of the panels. the form:
Corrugated-core =
Deformable body
Fig. 4. Mesh, boundary and loading conditions used in the corrugated-core model.
272 M.R.M. Rejab, W.J. Cantwell / Composites: Part B 47 (2013) 267–277
3.0 AL5, H=0.50mm due to interactions between the surfaces of the cell walls and the
uppermost skin. Finally, in Region IV the corrugated-core has been
Compression Stress (MPa)
(a) I
(b)
II 1mm
(c)
III
IV
1mm
Fig. 6. (a) Photographs of progressive damage development in a five unit cell aluminium system. (b) Debonding image at an end edge of the core after process (b – Region III).
(c) Crumpling when the corrugated-core is almost completely flattened.
M.R.M. Rejab, W.J. Cantwell / Composites: Part B 47 (2013) 267–277 273
0.20 GF2U5, H=0.19mm indicates that the influence of the cell number is accurately
predicted. Clearly, the compression strength of the aluminium cor-
Compression Stress (MPa)
II
rugation is much higher than that of the GFRP structure, due to the
0.15 fact that the aluminium alloy is significantly thicker than its com-
posite counterpart. Fig. 11b shows the variation of the strength and
III stiffness of the CFRP-based corrugations as a function of cell num-
0.10 ber. Here, the model over-estimates the experimental strength and
stiffness values, particularly at intermediate values of cell number.
I A comparison of the Figs. 11a and b indicate that the strengths of
0.05
the aluminium and carbon fibre corrugations (HAL = 0.5 mm and
IV HCFRP = 0.45 mm) are similar, in spite of the fact that the density
0.00 of the composite is significantly lower than that of the alloy.
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 Fig. 12 shows the deformed shapes of GFRP structures based on be-
Compression Strain tween one and five unit cells. All of the structures have been sub-
jected to a similar level of crosshead displacement. An examination
Fig. 7. Typical compression stress–strain curve for a GFRP sandwich corrugated-
core panel.
of the images indicates that the failure modes are similar in all five
samples, will buckling and subsequent fracture of the cell walls
being in evidence.
model. It should be noted that the initial FE predictions, based on
imperfection-free corrugations, grossly over-estimated both the 4.6. The effect of cell wall thickness
strength and stiffness of all three types of corrugation. Following
a detailed imperfection sensitivity analysis, the FE predictions The final part of this study investigated the influence of varying
were modified to introduce initial imperfections of amplitude the thickness of the cell wall on the compressive properties of the
n = 0.01, for the aluminium corrugation and n = 0.05 and n = 0.1 corrugations. Fig. 13 shows the variation of strength and stiffness
for the GFRP and CFRP corrugations, respectively. The appropriate of the GFRP and CFRP corrugations as a function of wall thickness.
scale factor was obtained through an imperfection sensitivity pro- As expected, the strength increases in a highly non-linear fashion
cedure whereby the factor was varied until the predicted maxi- with wall thickness. The compression strength of the CFRP corru-
mum force agreed with the experimental value. Applying these gation with a wall thickness of 1.25 mm is impressive, with a value
initial imperfections yields reasonably good predictions between in excess of 18 MPa being recorded. The stiffness properties of the
the numerical and experimental data, especially for the all-metal corrugations also increase rapidly with wall thickness. Closer
system. It is evident that the FE model of the GFRP corrugation fails examination of Fig. 13a and b indicates that the properties associ-
to predict the early instabilities in the load–displacement trace and ated with the thickest cores tend to be lower than expected. These
subsequently over-predicts the softening phase of the deformation discrepancies are associated with difficulties in manufacturing
process. thicker corrugations. It was observed that once the wall thickness
of the GFRP system exceeded 1 mm, the apexes became progres-
sively rounded and defects were introduced at the bends in the
4.5. The effect of varying the number of unit cells corrugation. This is shown schematically in the top right hand cor-
ner of the figures. The presence of these manufacturing defects
Clearly, one would intuitively expect the performance of multi- greatly reduced both the strength and stiffness of the corrugations.
ple unit cell corrugations to accurately reflect that of a single cell In order to facilitate a fairer comparison of the compression prop-
system. However, difficulties associated with the manufacture of erties of the corrugations, the data in Fig. 13 were normalised by
small single cell samples as well as problems with test alignment the density of the core to yield specific compression strength
may introduce unwanted anomalies into the test data. In order properties, rsp, and these values are shown as a function of cell
to establish more accurately, tests were undertaken on samples wall thicknesses in Fig. 14. An examination of the figure indicates
based on between one and five cells. The effect of varying the num- that the specific strength increases rapidly with wall thickness
ber of unit cells on the compression strength of the aluminium and with the specific strength of the 1.2 mm panel being roughly three
GFRP sandwich panels is shown in Fig. 11a. A comparison of the times that of the 0.4 mm thick structure. The figure suggests that
finite element predictions and the experimental measurements the CFRP composites out-perform their GFRP and aluminium
(a) I (b)
Delamination
II
Fibre
breaking
III
Debonding
IV 0.5mm
Fig. 8. (a) Photographs of progressive damage development in a five unit cell GFRP corrugation. (b) Failure mode images after crushing process completed from fibre breaking
in the middle of the core after buckling load, delamination occurred at the apex and debonding at both ends.
274 M.R.M. Rejab, W.J. Cantwell / Composites: Part B 47 (2013) 267–277
(a) I
(b)
Debonding
II Fibre
breaking
Delamination
III
IV
1mm
Fig. 9. (a) Photographs of progressive damage development in five unit cell CFRP corrugation. (b) Failure mode images after crushing process completed from formation of
hinges in the middle of the cell wall after fibre fracture, delamination and debonding at both ends.
2.0
2 0.2
1.5
1.0
1 0.1
0.5
0.0
0 0.05 0.1 0.15 0.2 0.25 0.3
0 0
Compression Strain 0 1 2 3 4 5
Number of unit cells
Expt.k k
Expt.
FE- σ max
FE-Fmax
Compression Strength (MPa)
0.15 120
FE- k
FE-k
Stiffness, k (MPa)
100
2
0.10
80
60
0.05
1
40
0.00 20
0 0.02 0.04 0.06 0.08 0.1 0.12 0.14 0.16 0.18 0.2
Compression Strain 0 0
0 1 2 3 4 5
Fig. 10. The measured and predicted responses for (a) a five unit cell aluminium
Number of unit cells
corrugation with H = 0.5 mm n = 0.01 and (b) a five unit cell glass fibre/epoxy
corrugation with H = 0.19 mm n = 0.05.
Fig. 11. Comparison between measured and FE predictions for compression
strength for (a) AL5 corrugated-core sandwich panels (H = 0.50 mm) with GF2U5
corrugated-core sandwich panels (H = 0.19 mm), and compression strength and
counterparts at low wall thicknesses, whereas there appears to be stiffness responses to the different number of unit cells for (b) CF2U5 corrugated-
little difference between the two composite systems as the wall core sandwich panels (H = 0.45 mm).
thickness increases.
Considering equilibrium of loads in y-direction, it can be shown
that:
4.7. Prediction of the compression strength of the corrugations
P
N sin h þ R cos h ¼ ð4Þ
Consider the model of the corrugated-core structure with an ap- 2
plied vertical load as shown in Fig. 15. If the sandwich panel is sub- and the bending moment M and the shear load R are related
jected to a compressive load 5P, then a single cell is subjected to a through:
load P as shown in Fig. 15b. Due to the symmetry of the triangular
RL
profile, each of the core members can be considered as a cantilever M¼ ð5Þ
beam subjected to the same axial stretching load N, bending mo- 2
ment M and shear load R as shown in Fig. 15c. Here, the lower The deformation d then can be written in terms of the displace-
end of each core member is fixed. ment parameters of U1 and U2 as:
M.R.M. Rejab, W.J. Cantwell / Composites: Part B 47 (2013) 267–277 275
n=20mm 20mm
2n=40mm
3n=60mm
4n=80mm
5n=100mm
Fig. 12. Photographs of damage in GFRP samples based on an increasing number of units cells.
(a)
Expt. σmax
Fmax
11 700
kExpt. k
10
FE - σmax
FE-Fmax
Stiffness, k (MPa)
500
7
6 400
5
300
4
3 200
2
100
1
0 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1 1.1
Wall thickness (mm)
(b)
Expt. σmax
Fmax
kExpt. k
18 600
FE - σmax
FE-Fmax
Compression Strength (MPa)
16 FE-k
FE - k
500
14
Stiffness, k (MPa)
12 400
10
300
8
6 200
4
100
2
0 0
0 0.2 0.4 0.6 0.8 1 1.2 1.4
Wall thickness (mm)
Fig. 13. Comparison of the experimental data and FE predictions for compression strength with stiffness as a function of cell wall thicknesses of (a) GFRP sandwich panels and
(b) CFRP sandwich panels.
GFRP ρ *= 25 5. Conclusions
60 AL
Triangular corrugated-cores, manufactured using a 45° profiled
50
mould, have been used to produce a range of lightweight sandwich
40 structures. The compressive behaviour and resulting failure mech-
anisms in structures based on three different materials have been
30
investigated both experimentally and numerically. An examination
20 of the corrugations during testing indicated that initial failure was
ρ *= 3.8 ρ *= 10 dominated by instabilities as the cell walls begin to buckle. Beyond
10
this buckling threshold, the aluminium corrugations deformed
0 plastically, resulting in the formation of localised plastic hinges.
0 0.2 0.4 0.6 0.8 1 1.2 1.4
In contrast, the composite corrugations exhibited zones of fibre
Wall thickness (mm)
fracture, delamination and debonding.
Fig. 14. Comparison of the specific strength as a function of cell wall thicknesses of The predictions offered by the numerical models were found to
the GFRP, CFRP and aluminium cores. be in reasonably good agreement with the experimental data,
M.R.M. Rejab, W.J. Cantwell / Composites: Part B 47 (2013) 267–277 277
Applied load, 5P
(a)
Unit cell
H
θ
P N
(b) (c) R
M
Φ2
δ
δ L
Φ1
Fig. 15. Free-body diagram of (a) a five unit cells, (b) a single unit cell and (c) a cell wall under compression loading.