Professional Documents
Culture Documents
a r t i c l e i n f o a b s t r a c t
Article history: As a class of important lightweight structural components, sandwich panels have gained considerable
Received 27 October 2012 popularity in a range of engineering applications. The crashworthiness of sandwich structures, which sig-
Accepted 26 April 2013 nifies a key mechanical property under impact loading, is found largely related to shape and dimensional
Available online 7 May 2013
parameters, such as cell width, wall thickness, structural angle and core height. This study exemplifies
corrugated sandwich panels with trapezoidal and triangular cores to determine the relationship between
Keywords: the structural parameters and the crashworthiness under low-velocity local impact and planar impact,
Crashworthiness
further optimizing these structural parameters with the crashworthiness criteria by using multiobjective
Optimization
Sandwich panel
optimization techniques. The configurations of trapezoidal and triangular core cells are firstly optimized
Corrugated core for maximizing energy absorption. The wall thickness of sandwich panels with optimal trapezoidal core
Energy absorption shape is then optimized for crashworthiness. Finally, the crashworthiness of these two types of corru-
gated sandwich panels is compared with each other under the identical face sheet thickness and core
density and it is found that the triangular configuration has better performance.
Ó 2013 Elsevier Ltd. All rights reserved.
0261-3069/$ - see front matter Ó 2013 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.matdes.2013.04.086
1072 S. Hou et al. / Materials and Design 51 (2013) 1071–1084
energy absorption of sandwich honeycomb panels by systemati- loading process were transferred from the sensor mounted in the
cally exploring the effects of such geometric parameters as thick- test machine to the computer.
ness of face sheet, wall dimension of cell, and core depth, on
specific energy absorption (SEA). It is noted that these previous 2.3. Numerical modeling
studies were mainly focused on the failure mechanisms and dam-
age development of sandwich panels. In spite of few studies look- In our study, the corrugated sandwich panels were also ana-
ing at the effects of impact conditions and geometric parameters lyzed by using the explicit FE code ANSYS/LS-DYNA. Identically
on the dynamic responses of sandwich panels, there have been to the experimental study, the face sheets and sandwich cores
very limited resports available on the optimal design for sandwich were made of Al-2024 aluminum alloy and 5052-0 aluminum al-
panels with crashworthiness criteria [13]. loy, respectively. The bilinear isotropic plastic material model of -
This paper aims to investigate the effects of the key shape and MAT-RIGID was employed in the ANSYS/LS-DYNA program. For the
dimensional parameters on the crashing behaviors of corrugated quasi-static loading process, the effect of strain rate was not con-
sandwich panels and optimize the sandwich cores with the trape- sidered here. The detailed material parameters are summarized
zoidal and triangular configurations for crashworthiness criteria. It in Table 1.
will also compare the optimized crashworthiness of these two dif- Belytschko-four-node shell element [14] was used to discretize
ferent corrugated sandwich panels, thereby providing some guides the face sheets and the sandwich core members. The adhesive bond-
to design of sandwich structures. ing between the core and face sheets was assumed to be perfect. The
upper indenter was simulated by using eight-node solid elements,
2. Correlation of numerical modeling with experimental tests and the lower platform was defined to be rigid [15]. Parameters of
rigid material are also shown in Table 1. An automatic surface-
2.1. Experimental specimens to-surface contact was defined between the upper indenter and
the sandwich panel. Meanwhile, an automatic single surface contact
Two types of corrugated configurations with trapezoidal and was considered to simulate self-contact of core sheets during defor-
triangular cores are considered in this study. The basic form of cor- mation. An automatic one-way surface-to-surface contact was
rugated sandwich panel is depicted in Fig. 1, in which Hf1 ; Hf2 , Hc, a defined between the face sheets and core members.
and b denotes the upper face sheet thickness, lower face sheet It was understood that use of the same speed (2 mm/min) as in
thickness, core height, total length and width, respectively. Two the experiments would result in prohibitively long computational
different core cell configurations with trapezoidal and triangular time, while the quasi-static loading process could be simulated
cells are illustrated in Fig. 2. sufficiently well through a uniformly slow motion of the upper in-
Aluminum 5052-O plates with a thickness of 0.2 mm are used denter [16]. For this reason, a speed of 1 m/s was adopted in the
to be the upper and lower face material of the sandwich panels. simulation. As a result, the FE models of these two corrugated
Aluminum 2024 sheet with a thickness of 1 mm were used to be sandwich panels are shown in Fig. 5.
the core cell material. According to the fabrication process of spec-
imen, the trapezoidal and triangular cores were prepared by using 2.4. Comparisons of experimental and numerical results
the self-made molds respectively. An epoxy adhesive was used to
bond the face plates and the cores for no less than 24 h under con- 2.4.1. Load–displacement curves
stant pressure at the room temperature. The prepared specimen Fig. 6 plots the experimental and modeling results of load–dis-
with trapezoidal and triangular cores are displayed in Fig. 3, placement curves of the two corrugated sandwich panels, respec-
respectively. tively. From Fig. 6, we can see that despite some difference, the
In the experiments, two specimens were tested for each type of simulation curves correlate to the experimental curves reasonably
sandwich panels with an identical dimensions. For these two dif- well. It is noted that the load–displacement curves exhibit three
ferent configurations of specimens, the overall structural size was clear characteristic stages, namely elastic deformation, compres-
a b = 70 mm 80 mm. For the trapezoidal cores, L1 = 10 mm, sion-shear coupled progressive buckling and densification, which
L2 = 10 mm and x = 90°. For the triangular cores, L = 10 mm and is the same as the experimental results in Ref. [17] and as the
x = 60°. numerical results in Ref. [13]. In the elastic stage, the sandwich pa-
nel was compressed vertically with slight core damage. When
loaded up to a peak value elastically, the buckling of core led to
2.2. Experimental set-up
load drop. As the core sheets were further squeezed to contact with
each other, the core member started densification.
Compression tests were carried out at a low speed of 2 mm/min
In the experimental process, the loading condition was com-
through an upper smooth indenter in the INSTRON-5984 test ma-
pletely the same as the simulation process and there was no
chine in accordance with the ASTM: C365/C365M-11a, which pro-
pre-loading imposed. In the preparation of specimen, an epoxy
vides a standard method of obtaining the flatwise compressive
structural adhesive was used to combine the face plates with the
strength. The specimen was placed on the lower platform as
cores, which followed the same procedure as the industrial speci-
showed in Fig. 4. The data of the force and displacement in the
fication. The adhesive combination was however not very firm on
the edge of the corrugated sandwich panels, especially for the tri-
anglular configurations, in which there existed some form of
b a abscission between the face plates and the cores when the impact
loading was imposed, and then, the cores slipped out of the edge of
Hf1 the structure, as shown in Fig. 7b. That is why the densification
time appears to be delayed in the experiments, making the curves
Hc shift toward right, as shown in Fig. 6. Except for this region, the
Hf2 numerical curves correlated to the experimental curves fairly well,
which validated that the FE simulation is applicable to the subse-
quent design optimization for these two corrugated sandwich
Fig. 1. The configuration of corrugated sandwich panel. panels.
S. Hou et al. / Materials and Design 51 (2013) 1071–1084 1073
L2
t
t L
L1
ω ω
Fig. 4. Experiment set-up of quasi-static loading. The surrogate model method has proven fairly effective to re-
late the crushing responses to design variables for optimization
problems involving highly nonlinear mechanics, in which as expli-
cit relation may not be easily to establish mathematically (if not
Table 1 impossible) [18–25]. It fits a surrogate function to the true struc-
The parameters of the materials.
tural responses generated from finite element analysis (FEA) at a
Property Face sheet Al-2024 Core 5052-O Indenter rigid limited number of sampling points in the design space. The surro-
aluminum alloy aluminum alloy material gate model can be used for structural optimization after verifying
q (kg/m3) 2700.00 2685.00 7830.00 the modeling accuracy.
E (GPa) 72.40 69.60 210.00
ET (GPa) 28.00 2.50
ry (MPa) 75.80 65.50 3.1. Design of experiment (DoE)
m 0.33 0.33 0.30
There are many different DoE methods available, such as facto-
rial design, optimal Latin Hypercube and central composite design
[26,27]. The full factorial design and the Optimal Latin Hypercube
2.4.2. Deformation process Sampling (OLHS) techniques have proven effective in our previous
Comparison of the deformation processes of these two corru- studies and will be used in this paper for different problems,
gated sandwich panels is shown in Fig. 7. From Fig. 7, it is clear that respectively.
the core structure mainly underwent the buckling deformation. By The rn factorial design generates a mesh of sampling points (an
comparing the numerical with the experimental snapshots, one n-dimensional hypercube) consisting of r points spaced regularly
can see that the deformation processes and patterns were fairly in each variable interval. To create an approximation of order r,
similar in these two sandwich panels. This further confirmed that at least (r + 1)n factorial designs are needed [26]. Therefore, the full
1074 S. Hou et al. / Materials and Design 51 (2013) 1071–1084
" #2
factorial design is suitable for the problem with a few design vari- X
M X
M XN
2 ðiÞ ðiÞ
ables and lower level design. Generally, the OLHS method is capa- EðaÞ ¼ e i ¼ aj uj ðx Þ y ð3Þ
ble for capturing the higher order of nonlinearity issues with much i¼1 i¼1 j¼1
t=0min
t=2min
t=3min
t=4min
t=5min
(a) trapezoid
t=0min
t=1min
t=2min
t=3min
t=3.5min
(b) triangle
Fig. 7. Comparison of the deformation of experiment and simulation.
R2 ¼ 1 SSE=SST ð7Þ X
M
SSE ¼ ~ i Þ2
ðyi y ð11Þ
M1
R2adj ¼ 1 ð1 R2 Þ ð8Þ i¼1
Mp1
sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
SSE where y is the mean value of FEA results yi.
RMSE ¼ ð9Þ Generally speaking, the larger the values of R2 and R2adj , and the
Mp1
smaller the value of RMSE, the better the RS accuracy [21,26]. In
where p is the number of non-constant terms in the RS model, SSE addition to these statistics, the accuracy of the RS model can also
and SST are the sum of squared errors and the total sum of squares, be measured by checking the relative errors between FEA results
respectively, calculated as and the RS fitting values [18,19].
1076 S. Hou et al. / Materials and Design 51 (2013) 1071–1084
For clarification, a flowchart is provided in Fig. 8 to depict the properties, and thus the bilinear, isotropic and plastic constitutive
procedure of DoE, RSM and FEA based crashworthiness model was used in the FE simulation [8].
optimization. The FE models were generated by ANSYS Parametric Design
Language (APDL), and the rigid floor and rigid plate in the planar
4. Optimization and comparative study of the two corrugated impact were defined by the internal rigid-wall feature provided
sandwich panels in LS-DYNA [15].
4.1. Loading and boundary conditions for optimization design 4.2. Shape optimization of trapezoidal core for sandwich panels
The core densities of trapezoidal and triangular sandwich struc- In the crashworthiness design of sandwich panels with trape-
tures are formulated respectively as zoidal core, the shape of core cell was firstly optimized for the en-
ergy-absorption criterion (i.e. internal energy IntEn). Then the
2ðL1 þ L2 Þt
qc ¼ q for trapezoidal core ð12Þ structural size of sandwich panels with the optimal trapezoidal
2L1 L2 sin x þ L21 sin 2x core cell shape was further optimized under those two different
2t impact loadings with different crashworthiness criteria.
qc ¼ q for triangular core ð13Þ
L sin 2x
where q is the density of the base material of the core sheets, 4.2.1. Shape optimization of trapezoidal core cell
L1 = Hc/sin x for trapezoidal core and L = Hc/sin x for triangular core The thickness of face sheets Hf1 ¼ Hf2 ¼ 1 mm; Hc ¼ 10 mm and
as in Fig. 2. the core density qc = 0.05q were considered to be constants in the
These two types of corrugated sandwich panels are designed for shape optimization of core cell. From Fig. 2a and Eq. (12), it was
the crashworthiness criteria under two loading conditions of (1) known that L1 = Hc/sin x and that any of the three parameters L2,
the low-velocity local impact and (2) the planar impact as illus- t and x can be obtained by the other two parameters. In this case,
trated in Fig. 9. In the local impact, the impactor is a hemispherical- the base angle x and the top side length L2 of trapezoidal core cell
ly-ended cylinder with a diameter of 50 mm that carries the were selected to be the design variables with the space interval of
impact energy of 15 J (Fig. 9a), and the four sides of sandwich pan- 0.5236(30°) 6 x 6 1.5708(90°) and 3 mm 6 L2 6 9 mm. In the de-
els were fixed. In the planar impact, the impact energy was 600 J, sign ranges of x and L2, four-level full factorial design was used
and the sandwich panel was placed on a rigid floor and was im- for sampling. The structural responses under the two impact load-
pacted by an upper rigid plate (Fig. 9b). The lower face sheet was ings are listed in Table 2, in which IntEn, Peak F and Def denotes the
fixed so as to avoid sliding between the rigid floor and sandwich internal energy, the peak crushing force and the deformation of the
panels. upper face sheet, respectively.
For the local impact, the response curves and values were cap- As shown in Table 2, the boldface values are the maximum and
tured at the time when the impactor begins to rebound and the minimum values of IntEn, Peak F and Def, respectively. For the low-
crushing force reduces to be zero. For the planar impact, the re- velocity local impact, it is seen in Table 2 that the difference be-
sponses were captured at the time when the sandwich core began tween the maximum and minimum values is relatively small for
to be densified and the crushing force begins to increase sharply. all these three crashing responses. This indicates that the cell shape
The face sheets and sandwich cores were made of Al-2024 alu- of trapezoidal core has limited effects on the responses of sand-
minum alloy and 5052-0 aluminum alloy, respectively [6,28]. wich structures under the given constant face sheets thicknesses,
These two aluminum alloys have similar strain-hardening material core height and core density.
S. Hou et al. / Materials and Design 51 (2013) 1071–1084 1077
0.06
250
0.04 220
0.02 190
Relative error
IntEn /J
160
0
130
-0.02
100 L2=3mm
L2=5mm
-0.04
70 L2=7mm
L2=9mm
-0.06 40
0 2 4 6 8 10 12 14 16 0.6 0.8 1 1.2 1.4 1.6
Sample points ω
Fig. 10. Relative errors of response surface function. Fig. 11. Influences of x and L2 on IntEn.
1078 S. Hou et al. / Materials and Design 51 (2013) 1071–1084
crushing force and internal energy curves arrive at the peak values 4.2.2. Multiobjective size optimization for the trapezoidal core
at the same time. However, it can be seen from Fig. 14 that in the 4.2.2.1. Case 1: low-velocity local impact. As shown in Fig. 2, the
planar impact loading case the crushing force occurs in the begin- upper and lower face sheet thickness Hf1 ; Hf2 and the core cell wall
ning and then decreases sharply. When the sandwich core was fur- thickness t were selected to be the design variables in this case. The
ther deformed and started largely contacting with the upper face design ranges were 0:5 mm 6 Hf1 6 2 mm, 0:5 mm 6 Hf2 6 2 mm
sheet, the crushing force increases again to another lower peak and 0.2 mm 6 t 6 0.8 mm, respectively. The optimized shape
(from 1.2 ms to 1.4 ms). When the sandwich core beginned to com- parameters of x = 0.9482( 55°) and L2 = 3.06 mm were consid-
pletely contacting with the lower face sheet, the crushing force in- ered as the constants in this stage, then the core internal density
creases sharply (from 2.4 ms), which indicates that the sandwich was 3%q 6 qc 6 12%q based on Eq. (12). Twenty-seven sample
core has been crushed completely (Fig. 12b). points were selected by the OLHS method to sample the design
Fig. 13. Crushing force–time and internal energy-time results for local impact – trapezoidal sandwich panels.
Fig. 14. Crushing force–time and internal energy-time results for planar impact – trapezoidal sandwich panels.
S. Hou et al. / Materials and Design 51 (2013) 1071–1084 1079
space. The cubic polynomial RS functions of IntEn, Peak F and Def the sandwich panel under the planar impact loading. Therefore,
are presented in Eqs. A.2a, A.2b, and A.2c respectively. the thicknesses of the upper and lower face sheets are considered
The maximum relative errors of these three RS functions in Eqs. to be the constants of Hf1 ¼ 0:5 mm and Hf2 ¼ 2 mm in this design
A.2a, A.2b, and A.2c are in range of [2%, 2%] and the other error case, which were the optimal solutions for the local impact case
measures in Eqs. (8)–(10) are listed in Table 3, which shows good (Case 1). In this case, the cell wall thickness t was considered to
accuracies of these RS models. be the design variable and 15 sample points were selected in the
Table 4 summarizes the results based on the single-objective range of 0.1 mm 6 t 6 0.8 mm (correspondingly, 1.5%q 6 qc
optimization problems for three different crashworthiness objec- 6 12%q). The structural responses are listed in Table 5.
tives, respectively. It can be seen that the internal energy of From Table 5, it can be found that the maximum absorbed en-
9.3084 J, which was generated from minimizing Defl and Peak F, ergy is about 590 J which is close to the impact energy of 600 J.
respectively, is just a slightly lower than 9.8119 J which was gen- Note that the energy-absorption does not vary remarkably when
erated from maximizing the energy criterion of IntEnðHf1 ; Hf2 ; tÞ. t P 0.55 mm, which indicates that the impact energy of 600 J has
However, when the internal energy was used as the objective, been almost absorbed at t = 0.55 mm before the sandwich core is
the corresponding value of Defl = 4.39 mm is much greater than completely crushed. From Table 5, It is very clear that Peak F and
its minimum value (1.13 mm). Therefore, H f1 ¼ IntEn both increase as t increases, when t 6 0.5 mm.
0:5 mm; Hf2 ¼ 2 mm and t = 0.2 mm could be considered to be an However, the crushing structures are expected to be able to ab-
optimal solution for the trapezoidal sandwich panel under the sorb as much energy as possible with as low crushing force as pos-
low-velocity local impact. The true response values obtained by sible for a safety consideration. Therefore, we formulated the
FEA at this optimal point were 9.2992 J, 4.0259 kN and design problem in the multiobjective optimization framework as
1.1179 mm, in which the relative error was accepted. 8
< Maximize : IntEnðtÞ
>
Minimize : Peak FðtÞ ð15Þ
>
:
4.2.2.2. Case 2: planar impact. As shown in Fig. 14b, the energy- s:t: 0:1 mm 6 t 6 0:8 mm
absorption of the upper and lower face sheets is almost zero for
Table 3 -20
Evaluation of fitting accuracy of 3 RS for size optimization of trapezoidal sandwich
panel under local impact.
-40
R2 R2adj RMSE
-60
-Peak F /kN
-100
Table 4
The results of three single objective optimization problems (the boldface number in
-120
each row highlights the objective function value in the corresponding single
optimization solution).
-140
Optimization Optimal designs (mm) Response values by 0 100 200 300 400 500 600
problems RSM IntEn /J
IntEn Peak Def
(J) F (mm) Fig. 15. Pareto front of the multi-objective optimization for trapezoidal sandwich
(kN) panels under the planar impact.
Max IntEn Hf1 ¼ 0:638; Hf2 ¼ 0:722; t ¼ 0:800 9.81 4.82 4.39
Min Peak F Hf1 ¼ 0:500; Hf2 ¼ 2:000; t ¼ 0:200 9.31 4.03 1.13
Min Def Hf1 ¼ 0:500; Hf2 ¼ 2:000; t ¼ 0:200 9.31 4.023 1.13 Table 6
Structural response values of sandwich panels with triangular core (the boldface
numbers highlight the maximum and minimum values in each column).
Fig. 16. Influence of x on IntEn and Peak F under the planar impact.
Fig. 18. Crushing force–time and internal energy-time results for local impact – triangular sandwich panels.
where IntEn(t) and Peak F(t) are cubic polynomial RS functions con- a b = 200 mm 200 mm were constants as illustrated in
structed in Eqs. (A.3a) and (A.3b). The R2 and R2adj of these two RS Fig. 2b. The base angle x was considered to be a design variable
models approximately equal to 1, indicating sufficient accuracies. with the range of 0.5236(30°) 6 x 6 1.309(75°). Here, the core
By using the multiobjective particle swarm optimization (MOP- density was defined as a constant of qc = 0.05q. Thus, the depen-
SO) algorithm as done before [29–33], the Pareto front of the opti- dent geometric parameters can be derived as L = Hc/sin x and
mization problem defined in Eq. (15) can be obtained as shown in t ¼ q2cqsin 2x
Hc according to Eq. (13). In this case, 19 sample points
sin x
Fig. 15. The designers can select the optimal point according to
were selected in the design range of x. And the corresponding re-
their emphasis on these two different objectives.
sponse values of the triangular sandwich panels were obtained
from FEA under the two different impact loadings as summarized
4.3. Sandwich panels with triangular core in Table 6.
Similarly to the trapezoidal core panel, the maximum and min-
As shown in Fig. 2b, the triangular core cells were optimized in imum values of IntEn and Peak F under the local impact loading ex-
this case, in which Hf1 ¼ Hf2 ¼ 1 mm, Hc = 10 mm and hibit no much difference. Except for x = 0.7854(45°) and
S. Hou et al. / Materials and Design 51 (2013) 1071–1084 1081
Fig. 19. Crushing force–time and internal energy-time results for planar impact – triangular sandwich panels.
x = 1.309(75°), the values of Def at the other design points also By using the multiobjective particle swarm optimization (MOP-
show no clear difference. It can be concluded that the core cell SO) algorithm [29–32], the Pareto front of the optimization prob-
shape still has minor effect on the crashworthiness of the triangu- lem defined in Eq. (17) was obtained and shown in Fig. 20 for
lar sandwich panels under the low-velocity local impact with the the triangulare sandwich panels. The designers can select the opti-
given constant thicknesses of face sheets, core height and core den- mal point according to their emphasis on these two objective.
sity. On the contrary, the responses of the triangular sandwich pan-
els were influenced remarkably by the core cell shape under the 4.4. Comparison of the two corrugated sandwich panels
planar impact loading. Therefore, we performed the design optimi-
zation for the planar impact loading condition. From Table 6, it is 4.4.1. Low-velocity local impact
seen that the variations of IntEn and Peak F along x are not mono- As above mentioned, for the two corrugated sandwich panels,
tonic. To reflect this, the RS models of IntEn and Peak F were estab- the shape of core cell barely has much effect on the crashworthi-
lished as shown in Eqs. (A.4) and (A.4b). The accuracy measures by ness of the structures under the low-velocity local impact loading.
Eqs. (7)–(11) are R2 = 0.9930, R2adj ¼ 0:9807 and RE interval is In order to analyze the effect of trapezoidal and triangular config-
[2.94%, 3.42%] for IntEn, which satisfy the accuracy requirements. urations on crashworthiness, the optimized trapezoidal panel is
At the same time the accuracy measures for Peak F are R2 = 0.989, compared with the optimized triangular panel with the same face
R2adj ¼ 0:9673 and RE interval is [5.25%, 8.12%], which also satisfy sheet thicknesses and core density. Fig. 21 shows the comparison
the accuracy requirement. The curves of IntEn(x) and Peak F(x) are of the time histories of crushing force, internal energy and
plotted in Fig. 16. It is clear that IntEn firstly increases and then de- displacement.
creased with the increase of x. However, Peak F always increased From Fig. 21a and b, it is seen that there is a little difference
with the increase of x. In order to obtain the maximum internal between the two corrugated sandwich panels in the aspects of
energy and the corresponding optimal base angle x, the single- the peak crushing force and the energy absorption. Nevertheless,
objective optimization was defined as the displacement of triangular sandwich panel is smaller than
that of the trapezoidal counterpart. It is thus concluded that
Maximize : IntEnðxÞ
ð16Þ the triangular core is slightly better than the trapezoidal one
s:t: 0:5 6 x 6 1:3
The optimal solution is x = 0.8023 46° with IntEn =
253.2929 J. The FEA result of IntEn is 264.156 J at the optimal angle.
The relative error between FEA and RSM is 4.11%. The deformation
patterns of the triangular sandwich panels with optimal core cell
shape under the two impact loading are shown in Fig. 17, which
is similar to the patterns of the trapezoidal sandwich panel. The
crushing force vs time and internal energy vs time curves of trian-
gular sandwich panel with optimal cell shape for the two impact
loading are given in Figs. 18 and 19, respectively.
Meanwhile, it is expected that the structures with good crash-
worthiness is able to absorb as much energy as possible with rela-
tively as low crushing force as possible. Thus, the multiobjective
optimization problem was defined as
8
< Maximize :
> IntEnðxÞ
Minimize : Peak FðxÞ ð17Þ
>
: Fig. 20. Pareto front of the multi-objective optimization for trapezoidal sandwich
s:t: 0:5 6 x 6 1:3 panels under the planar impact.
1082 S. Hou et al. / Materials and Design 51 (2013) 1071–1084
Fig. 21. Compared results of two types of sandwich panels for local impact.
180 600
Triangular Triangular
160 Trapezoidal
Trapezoidal 500
140
400
120
Peak F /kN
IntEn /J
100 300
80
200
60
100
40
20 0
0.01 0.03 0.05 0.07 0.09 0.01 0.03 0.05 0.07 0.09
ρc ρ ρc ρ
(a) Peak crushing force (b) Absorbed energy
Fig. 22. Compared results of two types of sandwich panels for planar impact.
in terms of crashing stiffness under the low-velocity local impact have the identical face sheet thickness of Hf1 ¼ 0:5 mm and
loading. Hf2 ¼ 2 mm. The diagrams of structural responses vs q ¼ qc =q
are plotted in Fig. 22 for these two different corrugated sand-
4.4.2. Planar impact wich panels.
The peak crushing force and energy absorption of the trian- From Fig. 22, we can see that the peak crushing force of the tri-
gular configuration under the planar impact are also summa- angular sandwich panels is somewhat smaller than that of the
rized in Table 5, when qc/q 6 8.3%. The triangular panels also trapezoidal panels, and also the internal energy of triangular pan-
S. Hou et al. / Materials and Design 51 (2013) 1071–1084 1083
els is higher than that of trapezoidal panels. When the relative core IntEnðHf1 ; Hf2 ; tÞ ¼ 8:3175 þ 1:0540Hf1 þ 1:3075Hf2
density q is small, there is a little difference between the internal
þ 2:4477t 1:4787H2f1
energy of the two sandwich panels. However, the internal energy
of the triangular panels is somewhat greater than that of the trap- 1:5962Hf1 Hf2 þ 0:8579Hf1 t
ezoidal ones with the increase of q , before the impact energy is to-
0:3882H2f2 1:2614Hf2 t 1:3802t2
tally absorbed. Therefore, the crashworthiness of triangular
sandwich panels is better than trapezoidal sandwich panels under þ 0:4636H3f1 þ 0:2263H2f1 Hf2
the planar impact. þ 0:5091Hf1 H2f2 þ 0:07241H3f2
References [18] Hou SJ, Li Q, Long SY, Yang XJ, Li W. Design optimization of regular hexagonal
thin-walled columns with crashworthiness criteria. Finite Elem Anal Des
2007;43:555–65.
[1] Heimbs S. Computational methods for bird strike simulations: a review.
[19] Hou SJ, Li Q, Long SY, Yang XJ, Li W. Multiobjective optimization of multi-cell
Comput Struct 2011;89:2093–112.
sections for the crashworthiness design. Int J Impact Eng 2008;35:1355–67.
[2] Meo M, Vignjevic R, Marengo G. The response of honeycomb sandwich panels
[20] Jansson T, Nilsson L, Redhe M. Using surrogate models and response surfaces
under low-velocity impact loading. Int J Mech Sci 2005;47:1301–25.
in structural optimization-with application to crashworthiness design and
[3] Aktay L, Johnson AF, Holzapfel M. Prediction of impact damage on sandwich
sheet metal forming. Struct Multidiscip Optim 2003;25:129–40.
composite panels. Comput Mater Sci 2005;32:252–60.
[21] Fang H, Rais-Rohani M, Liu Z, Horstemeyer MF. A comparative study of
[4] Tilbrook MT, Radford DD, Deshpande VS, Fleck NA. Dynamic crushing of
metamodeling methods for multiobjective crashworthiness optimization.
sandwich panels with prismatic lattice cores. Int J Solid Struct
Comput Struct 2005;83:2121–36.
2007;44:6101–23.
[22] Forsberg J, Nilsson L. Evaluation of response surface methodologies used in
[5] Dharmasena K, Queheillalt D, Wadley H, Chen Y, Dudt P, Knight D, et al.
crashworthiness optimization. Int J Impact Eng 2006;32:759–77.
Dynamic response of a multilayer prismatic structure to impulsive loads
[23] Hajela P, Lee E. Topological optimization of rotorcraft subfloor structures for
incident from water. Int J Impact Eng 2009;36:632–43.
crashworthiness considerations. Comput Struct 1997;64:65–76.
[6] Liu YC. Optimum design of straight thin-walled box section beams for
[24] Liao XT, Li Q, Yang XJ, Zhang WG. A two-stage multi-objective optimisation of
crashworthiness analysis. Finite Elem Anal Des 2008;44:139–47.
vehicle crashworthiness under frontal impact. Int J Crashworthines
[7] Meguid SA, Stranart JC, Heyerman J. On the layered micromechanical three-
2008;13:279–88.
dimensional finite element modeling of foam-filled columns. Finite Elem Anal
[25] Hou SJ, Li Q, Long SY, Yang XJ, Li W. Crashworthiness design for foam filled
Des 2004;40:1035–57.
thin-walled structures. Mater Design 2009;30:2024–32.
[8] Zarei HR, KrÖger M. Optimum honeycomb filled crash absorber design. Mater
[26] Montgomery DC. Design and analysis of experiments. 6th ed. New
Design 2008;29:93–204.
York: Wiley; 2005.
[9] Dear JP, Lee H, Brown SA. Impact damage processes in composite sheet and
[27] Myers RH, Montgomery DC. Response surface methodology. New York: Wiley;
sandwich honeycomb materials. Int J Impact Eng 2005;32:130–54.
1995.
[10] Zhou G, Hill M, Hookham N. Investigation of parameters governing the damage
[28] Zhu F, Zhao LM, Lu GX, Wang ZH. Deformation and failure of blast-loaded
and energy absorption characteristics of honeycomb sandwich panels. J Sandw
metallic sandwich panels – experimental investigations. Int J Impact Eng
Struct Mater 2007;9:309–42.
2008;35:937–51.
[11] Othman AR, Barton DC. Failure initiation and propagation characteristics of
[29] Yin HF, Wen GL, Hou SJ, Chen K. Crushing analysis and multiobjective
honeycomb sandwich composites. Compos Struct 2008;85:126–38.
crashworthiness optimization of honeycomb-filled single and bitubular
[12] Atli-Veltin B, Gandhi F. Energy absorption of sandwiched honeycombs with
polygonal tubes. Mater Design 2011;32:4449–60.
facesheets under in-plane crushing. In: Annu Forum Proc-AHS Int Inc; 2010. p.
[30] Sun GY, Li GY, Zhou SW, Li HZ, Hou SJ, Li Q. Crashworthiness design of vehicle
206–21.
by using multiobjective robust optimization. Struct Multidiscip Optim
[13] Hou SJ, Ren LL, Dong D, Han X. Crashworthiness optimization design of
2011;44:99–110.
honeycomb sandwich panel based on factor screening. J Sandw Struct Mater
[31] Hou SJ, Han X, Sun GY, Long SY, Li W, Yang XJ, et al. Multiobjective
2012;14:1–24.
optimization for tapered circular tubes. Thin Wall Struct 2011;49:855–63.
[14] Hallquist JO. LS-DYNA theoretical manual. Livemore Software Technology
[32] Song X, Sun GY, Li GY, Gao WZ, Li Q. Crashworthiness optimization of foam-
Corporation, California; 1998.
filled tapered thin-walled structure using multiple surrogate models. Struct
[15] Hallquist JO. LS-DYNA keyword user’s manual v. 971. Livemore Software
Multidiscip Optim 2013;47:221–31.
Technology Corporation, California; 2007.
[33] Sun GY, Li GY, Stone M, Li Q. A Two-Stage Multi-fidelity Optimization
[16] Wang QC, Fan ZJ. The improved method in simulating the response of the
Procedure for Honeycomb-Type Cellular Materials. Comput Mater Sci
structure under quasi-static load by using LS-DYNA. Mech Eng 2003;25:20–3.
2010;49:500–11.
[17] Giglio M, Manes A, Gilioli A. Investigations on sandwich core properties
through an experimental–numerical approach. Compos Part B: Eng
2012;43:361–74.