Thin-Walled Structures 122 (2018) 42–51

Contents lists available at ScienceDirect

Thin-Walled Structures
journal homepage: www.elsevier.com/locate/tws

Full length article

Crashworthiness design for bio-inspired multi-cell tubes with quadrilateral, T

hexagonal and octagonal sections
⁎
Linwei Zhanga, Zhonghao Baia, , Fanghua Baib
a
State Key Laboratory of Advanced Design and Manufacturing for Vehicle Body, Hunan University, Changsha, Hunan 410082, China
b
Chongqing Vehicle Test & Research Institute Co., Ltd., Chongqing 401122, China

A R T I C L E I N F O A B S T R A C T

Keywords: Multi-cell tubes have been widely used in vehicle engineering for their excellent energy absorption capacity. In
Microstructure this paper, a group of bionic multi-cell tubes (BMCTs) with quadrilateral, hexagonal and octagonal sections were
Bionic multi-cell tube proposed. The BMCTs were constructed by filling the cylindrical tubes into different position of multi-cell tubes
Theoretical validation (MCTs), which was inspired by the microstructure of beetle forewings. The finite element (FE) models under
Crashworthiness
axial impact loading were established and then validated by the Simplified Super Folding Element (SSFE) theory.
Multiobjective optimization
The crashworthiness of different BMCTs and MCTs was compared, and the results showed that the sixth type of
bionic multi-cell tube with octagonal section (O-BMCT-6) has the best crashing performance. Then, the multi-
objective optimization design of O-BMCT-6 was conducted by using non-dominated sorting genetic algorithm II
(NSGA-II) and radial basis function (RBF) metamodels. The optimal O-BMCT-6 showed superior crashworthiness
and could be used as an energy absorber.

1. Introduction
Thin-walled tubes as effective energy absorbers characterized by
being cost-efficient, lightweight and crashworthy have been widely
utilized in the vehicle engineering [1–3]. Over the past decades, ex-
tensive research efforts were conducted to investigate the crash-
worthiness of thin-walled tubes with various cross-sectional shapes [4],
such as circular [5–7], quadrilateral [8–10], hexagonal [11–13] and
octagonal [14–16], etc. In addition, Yamashita et al. [17], Nia et al.
[18], Fan et al. [19] and Ali et al. [20] compared the crashworthiness of
thin-walled tubes with different cross-sectional shapes. According to
these previous studies, it can be found that the cross-sectional config-
uration has a significant influence on the crashworthiness of thin-
walled tubes.
Recently, multi-cell thin-walled tubes have been widely studied for
improving the crashworthiness [21–29]. Chen and Wierzbicki [30]
proposed the Simplified Super Folding Element (SSFE) theory to predict
the mean crushing forces of single-cell, double-cell and triple-cell tubes
and found that the energy absorption efficiency of triple-cell tubes is
superior to the single-cell tubes. Zhang et al. [31] improved Chen and
Wierzbicki's theoretical solution [30] by dividing the cross-section of
the tube into basic elements and demonstrated that the energy ab-
sorption efficiency of multi-cell tubes is 50% higher than the single-cell
tubes. Zhang and Zhang [32–34] further studied the theoretical models
for different basic elements of multi-cell tubes and indicated that the
theoretical models are agreed well with both numerical and experi-
mental results. Qiu et al. [29] applied the theoretical formulas in [32]
to make a comparative analysis on the crashworthiness of hexagonal
multi-cell tubes with different cross-sections, which showed that the
W2W hexagonal multi-cell tubes (the inner and outer tubes are con-
nected by the ribs at the mid-walls) are the most efficient configuration
among them. Nia and Parsapour [35] investigated the crashworthiness
of single-cell and multi-cell tubes with triangular, square, hexagonal
and octagonal sections by using the experimental and numerical
methods, and revealed that multi-cell tubes with inner ribs connected at
the mid-walls of the outer tubes are more efficient than those at the
corners. Based on the above investigations, it can be seen that the multi-
cell thin-walled tubes have excellent energy absorption capacity and
exhibit better performance with the ribs connected at the mid-walls.
The above studies mainly focus on the conventional single-cell and
multi-cell tubes. However, thin-walled tubes with other complicated
cross-sections may have even better crashworthiness since cross-sec-
tional configurations have a significant influence on crashworthiness.
But, how to design a thin-walled tube with better crashworthiness re-
mains a topic of further studies. Nowadays, many researchers have paid
attention to the use of bionic methods to design engineering structure
[23,36]. Chen and co-workers [37–39] investigated the three-dimen-
sional structures and mechanical properties of beetle forewings and

⁎
Corresponding author.
E-mail address: baizhonghao@hnu.edu.cn (Z. Bai).

http://dx.doi.org/10.1016/j.tws.2017.10.010
Received 8 June 2017; Received in revised form 3 October 2017; Accepted 4 October 2017
Available online 13 October 2017
0263-8231/ © 2017 Elsevier Ltd. All rights reserved.
L. Zhang et al.
found that the beetle forewings have characteristics of high strength
and lightweight. Chen and co-workers [40,41] proposed an integrated
honeycomb structure with reinforced trabeculae inspired by the mi-
crostructure of beetle forewings and confirmed that the integrated
honeycomb structure exhibits excellent mechanical properties through
the experimental study [42]. Xiang and Du [43] employed the micro-
structure of beetle forewings to design a bionic honeycomb structure
and showed that the energy absorption of the bionic honeycomb
structure, which filled columns on its walls is better than that filled
columns in its walls. From the above studies, it can be observed that the
bionic structures have the better energy absorption capacity than tra-
ditional thin-walled structures. However, few studies have been made
to investigate the crashworthiness of bionic multi-cell tubes based on
the microstructure of beetle forewings and systematically compare the
crashworthiness of bionic multi-cell tubes with different cross-section
shapes.
In this paper, eighteen kinds of bionic multi-cell tubes (BMCTs)
based on the microstructure of beetle forewings were proposed. The
finite element (FE) models of these tubes were established by using the
explicit FE code LS-DYNA and then validated by the Simplified Super
Folding Element (SSFE) theory. The comparison was made between the
crashworthiness of traditional multi-cell tubes (MCTs) and BMCTs with
different cross-sectional shapes, and the best possible BMCT structure
was determined. To find the optimal design of BMCT, the multi-
objective optimization was realized by adopting the non-dominated
sorting genetic algorithm II (NSGA-II) and radial basis function (RBF)
metamodels.
2. The bionic multi-cell tube based on the microstructure of beetle
forewing
2.1. Structural characteristics of the beetle forewing
Fig. 1(a) shows the adult Allomyrina dichotoma beetle [41]. Fig. 1(b)
shows the beetle forewing [38]. Fig. 1(c) and (d) show the micro-
structure of the beetle forewing [40]. From Fig. 1(c), we can find that
there exist multi-cell structures in the internal structure of beetle
forewing, and there are many hollow cylindrical tubes located at the
multi-cell structures. Fig. 1(d) shows the microstructure of the cylind-
rical tube. Since natural evolution is a self-optimizing process, the
structural characteristics of the beetle forewing may be reasonable. In
nature, the beetle forewing can resist the impact load represented by
droplets of rain, or produced by the opponent. Therefore, the bionic
thin-walled structure based on the microstructure of beetle forewing
may be an excellent energy absorber when subjected to impact loading.
2.2. Description of the bionic multi-cell tube inspired by the beetle forewing
By imitating the structural characteristics of the beetle forewing,
eighteen kinds of BMCTs with quadrilateral, hexagonal and octagonal
Fig. 1. Microstructure structure of the beetle forewing: (a) the adult Allomyrina dichotoma beetle [41], (b) the beetle forewing [38], (c) internal structure of the beetle forewing filled with
cylindrical tubes [40] and (d) the microstructure of the cylindrical tube in the beetle forewing [40].
43
Thin-Walled Structures 122 (2018) 42–51
sections were designed and compared with the traditional MCTs with
the ribs connected at the mid-walls, as shown in Fig. 2. Fig. 2(a), (h)
and (o) show the cross-section of traditional quadrilateral, hexagonal
and octagonal multi-cell tube, and the rest is bionic multi-cell tube. The
bionic multi-cell tubes were constructed by filling the cylindrical tubes
in the different position of multi-cell tubes. The bionic multi-cell tubes
were named to distinguish from each other, and codes for the simple
specimen were as follows: quadrilateral (Q), hexagonal (H) and octa-
gonal (O). And the numbers denoted different types of bionic multi-cell
tubes. Taking BMCTs with quadrilateral section as an example: (a) Q-
MCT: traditional quadrilateral multi-cell tube, in which the ribs are
connected at the mid-walls, (b) Q-BMCT-1: filling the cylindrical tubes
into the center of the gaps between the inner and outer tube, (c) Q-
BMCT-2: filling the cylindrical tubes into the center of mid-ribs, (d) Q-
BMCT-3: filling the cylindrical tubes into the mid-walls of inner tube,
(e) Q-BMCT-4: filling the cylindrical tubes into the corners of inner
tube, (f) Q-BMCT-5: filling the cylindrical tubes into the corners of outer
tube, (g) Q-BMCT-6: filling the cylindrical tubes into the mid-walls of
outer tube. The length and the circumcircle diameters of the inner and
outer tube for MCTs were 240 mm, 50 mm and 100 mm, respectively.
The diameters of cylindrical tubes for BMCTs were 12 mm, and the wall
thickness for all tubes was set to be 2 mm initially.
3. Numerical simulation and crashworthiness indicators
3.1. Finite element model
Fig. 3 shows the finite element (FE) model of the bionic multi-cell
tube subjected to axial impact loading. The explicit FE code LS-DYNA
was used to simulate the crashing process. The four-node Belytschko-
Tsay shell elements with five integration points through thickness were
employed to model the tube wall. The bottom end of the tube was fixed,
and the top end of the tube was compressed by a rigid plate with a
constant velocity of 10 m/s. When the deformation displacement of the
tube reached 168 mm (70% of the tube length), the tube was unloaded
and the simulation stopped. The contact between the rigid plate and the
tube was modeled as ‘node to surface’. The ‘automatic single surface’
was applied to the tube to avoid interpenetration. The static and dy-
namic frictional coefficients were 0.3 and 0.2, respectively [44,45]. The
element size of the tube was 2 mm × 2 mm in the FE model.
The material of the tube was aluminum alloy AA6060 T4 with the
following properties: ρ = 2700 kg/m3, Young's modulus E = 68.2 GPa,
Poisson's ratio ν = 0.3, the power law exponent n = 0.23, initial yield
strength σy = 80 MPa, and ultimate stress σu = 173 MPa [46]. The
engineering stress-strain curve of the tube material was shown in Fig. 4
[31,46]. The constitutive behavior of the aluminum material was based
upon the piecewise linear elastic-plastic material model. As the alu-
minum is insensitive to the strain rate, the strain rate effect was ne-
glected in the FE model [45].
L. Zhang et al. Thin-Walled Structures 122 (2018) 42–51

Fig. 2. The cross-sectional geometry of MCTs and BMCTs.

Fig. 4. Engineering stress-strain curve for AA6060 T4 [31,46].

Fig. 3. The finite element model of the BMCT subjected to axial impact loading. EA
SEA =
M (2)
3.2. Crashworthiness indicators where M denotes the mass of the structure. Obviously, the higher the
SEA, the better the efficiency of energy absorption.
In order to evaluate the crashworthiness of energy absorbers, it is Mean crush force (Pm) for a given deformation can be obtained as:
essential to define the crashworthiness indicators. In this paper, the
d
following indicators are utilized to evaluate the crashworthiness char- EA ∫ F (x ) dx
Pm = = 0
acteristics of bionic multi-cell tubes and traditional multi-cell tubes. d d (3)
Energy absorption (EA) is the amount of energy absorption during
the plastic deformation, which can be calculated as: Undulation of Load-carrying Capacity (ULC) is used to estimate the
stability of energy absorption [48], given as:
d
EA = ∫0 F (x ) dx (1)
d
∫0 |F (x ) − Pm |dx
ULC = d
∫0 F (x ) dx (4)
where F(x) is the instantaneous crushing force and d is the crush dis-
placement which was set to be 168 mm. When ULC is small, the force-displacement curve is approaching to a
Specific energy absorption (SEA) is defined as the energy dissipation rectangular shape, indicating better energy absorption capacity.
per unit mass, which can be defined as [47]: Besides, Peak crushing force (PCF) is also utilized to evaluate the
structural crashworthiness because it often leads to a high deceleration,

44
L. Zhang et al. Thin-Walled Structures 122 (2018) 42–51

Fig. 5. Scheme of the SSFE [30]: (a) extensional

elements and (b) stationary hinge lines.

3
potentially increasing the probability of occupant injury.
Ebending = ∑ M0 αi Li
i=1 (6)
4. Theoretical validation of the finite element model where M0 = σ0t /4 is the fully plastic bending moment of the flange, αi
2

The FE models of multi-cell tubes with quadrilateral, hexagonal and
octagonal sections under axial impact loading can be validated by using
the Simplified Super Folding Element (SSFE) method developed by
Chen and Wierzbicki [30]. In this theory, a basic element consisting of
three extensional triangular elements and three stationary hinge lines
was proposed, as shown in Fig. 5 [30]. Then, the mean crashing force of
the element can be determined by considering the energy conservation
of the system in one folding wavelength 2 H. According to the energy
balance of the system, the external work done by compression is dis-
sipated by the plastic deformation in bending and membrane:
2HPm k = Ebending + Emembrane (5)
where Pm, k, Ebending and Emembrane denote the mean crushing force, the
effective crashing distance coefficient, the bending energy and the
membrane energy, respectively. In reality, the crushing distance is less
than 2 H. According to previous studies [49,50], the crushing distance
coefficient k ranged from 0.7 to 0.75. Here, the value of k is taken as
0.73.
4.1. The bending energy
The bending energy Ebending can be determined by summing up the
energy dissipation at the three stationary hinge lines:
represents the rotation angle at each hinge line and Li denotes the
length of the flange. While σ0 and t are the flow stress of material with
power law hardening and the wall thickness, respectively. And σ0 can
be calculated as [46]:
σy σu
σ0 =
1+n (7)
where σy, σu and n are the yield stress, the ultimate stress and the strain
hardening exponent, respectively. In the SSFE theory, the flanges are
assumed to be flattened completely after the axial deformation of 2 H
[30]. Therefore, the rotation angles α at the three hinge lines are π/2, π
and π/2, respectively. Then
Ebending = 2πM0 Lc (8)
where Lc denotes the total length of all flanges.
4.2. The membrane energy
In order to analyze the membrane energy dissipation under com-
pression, the multi-cell tubes with quadrilateral, hexagonal and octa-
gonal sections were divided into two kinds of basic elements: corner
element and 3-panel angle element, as shown in Fig. 6. From Fig. 6, it
can be seen that the 3-panel angle elements are all T-shape elements (φ
= 90°). Therefore, the theoretical formula of the 3-panel angle element
I in [29,32] was used in this paper.

Fig. 6. Cross sections of MCTs and the basic ele-

ments.

45
L. Zhang et al. Thin-Walled Structures 122 (2018) 42–51

According to the previous studies [29,32], the membrane energy of dynamic enhancement effects including strain rate effect and inertia
the corner element with central angle θ and the three-panel angle effect should be considered in the theoretical solutions [51]. As the
element I can be evaluated as: aluminum is insensitive to the strain rate [45], the strain rate effect was
neglected in this study. Therefore, the dynamic enhancement factor λ is
corner 4M0 H 2 tan(θ /2)
Emembrane (θ) = introduced mainly for considering the inertia effect. According to pre-
(tan(θ /2) + 0.05/ tan(θ/2)) t /1.1 (9)
vious studies [51,52], λ was recommended to be in the range of 1.3–1.6
3 − panel − I 4M0 H2 tan(φ) for aluminum alloy. In the present work, λ was set to be 1.3.
Emembrane (φ) = ( + 2 tan(φ /2))
t (tan(φ) + 0.05/ tan(φ))/1.1
(10) 4.3.2. The mean crashing force of multi-cell tubes with hexagonal sections
The H-MCTs consist of twelve corner elements with θ = 120° and
Therefore, the whole membrane energy can be calculated as:
twelve T-shape elements. Thus, the total membrane energy is
corner 3 − panel − I
Emembrane = Nc Emembrane + N3 − panel − I Emembrane (11) H − MCT corner T − shape
Emembrane = 12Emembrane (120°) + 12Emembrane (21)
where Nc and N3-panel-I are the numbers of corner element and three-
panel angle element I, respectively. Substitute Eqs. (21) and (8) into Eq. (5), and the mean crushing
force of the H-MCTs can be formulated as:
4.3. The mean crashing force of multi-cell tubes with quadrilateral, 4.328M0 H 2 12.4M0 H 2
hexagonal and octagonal sections PmH − MCT × 2H ⋅k = 2πM0 Lc + 12 × + 12 ×
t t

The constituent elements in the multi-cell tubes with quadrilateral,

200.736M0 H 2
hexagonal and octagonal sections are corner elements (θ = 90°, 120° = 46.392πM0 R +
t
and 135°) and T-shape elements (φ = 90°). Substitute the specific an-
gles into Eqs. (9) and (10), and the membrane energy of these con- (22)
stituent elements can be obtained as: By employing the stationary condition, the half-wavelength H can
4.190M0 H 2 be given as:
corner
Emembrane (90°) =
t (12) H= 0.726Rt (23)
corner 4.328M0 H2 Substitute Eq. (23) into Eq. (22), and the mean crushing force of the
Emembrane (120°) =
t (13) H-MCTs can be described as:
corner 4.363M0 H 2 42.76
Emembrane (135°) = PmH − MCT = λ σ0 R0.5t 1.5
t (14) k (24)

T − shape 3 − panel − I 12.4M0 H 2

Emembrane = Emembrane (90°) =
t (15) 4.3.3. The mean crashing force of multi-cell tubes with octagonal sections
The profile of O-MCTs included sixteen corner elements with θ =
4.3.1. The mean crashing force of multi-cell tubes with quadrilateral 135° and sixteen T-shape elements. Then, the total membrane energy is
sections O − MCT corner T − shape
Emembrane = 16Emembrane (135°) + 16Emembrane (25)
The configuration of the Q-MCTs can be divided into eight corner
elements with θ = 90° and eight T-shape elements. Therefore, the total Substitute Eqs. (25) and (8) into Eq. (5), and the mean crushing
membrane energy is force of the O-MCTs can be calculated as:
Q − MCT corner T − shape
Emembrane = 8Emembrane (90°) + 8Emembrane (16) 4.363M0 H 2 12.4M0 H 2
PmO − MCT × 2H ⋅k = 2πM0 Lc + 16 × + 16 ×
Substitute Eqs. (16) and (8) into Eq. (5), and the mean crushing t t
force of the Q-MCTs can be obtained as:
268.208M0 H 2
4.190M0 H 2 12.4M0 H 2 =51.520πM0 R +
PmQ − MCT × 2H ⋅k = 2πM0 Lc + 8 × +8× t
t t
(26)
132.72M0 H 2
= 39.598πM0 R +
t (17) Applying the stationary condition, we can get the result as:

where R is the radius of the circumcircle of the inner tube and is set to H= 0.603Rt (27)
be 25 mm. Here, Lc can be derived from R, as Lc is the total length of the
Substitute Eq. (27) into Eq. (26), and the mean crushing force of the
wall of the cross-section.
O-MCTs can be obtained as:
According to the stationary condition [30], the half-wavelength H
can be determined as: 52.09
PmO − MCT = λ σ0 R0.5t 1.5
k (28)
∂Pm
=0
∂H (18) The comparison of Pm between FE simulations and theoretical pre-
Then, H can be derived as: dictions for Q-MCTs, H-MCTs and O-MCTs with different thickness is
shown in Table 1. The FE simulation results of the mean crushing forces
H= 0.937Rt (19) were yielded at the deformation displacement of 70% length of the
Substitute Eq. (19) into Eq. (17), and the mean crushing force of the tube. The theoretical predictions of the mean crushing forces were
Q-MCTs can be obtained as: obtained from Eqs. (20), (24) and (28). From Table 1, it can be observed
that the mean crushing forces of FE simulations are in good agreement
32.12
PmQ − MCT = λ σ0 R0.5t 1.5 with the theoretical predictions. The maximum discrepancy of the mean
k (20)
crushing force is less than 4%. Therefore, the numerical model is ade-
where λ is the dynamic enhancement factor. For dynamic loading, quately validated and can be used for further studies.

46
L. Zhang et al. Thin-Walled Structures 122 (2018) 42–51

Table 1
The comparison of Pm between FE simulations and theoretical predictions for Q-MCTs, H-MCTs and O-MCTs with different thickness.

No t (mm) Q − MCT
Pm (kN) PmH − MCT (kN) O − MCT
Pm (kN)

FE. Theo. Diff (%) FE. Theo. Diff (%) FE. Theo. Diff (%)

1 1.6 60.21 61.40 1.94 80.83 81.74 1.11 95.78 99.58 3.82
2 1.8 73.01 73.27 0.35 96.50 97.54 1.07 115.94 118.82 2.42
3 2.0 87.80 85.81 2.32 114.88 114.24 0.56 136.31 139.16 2.05
4 2.2 102.94 99.00 3.98 134.72 131.79 2.22 158.37 160.55 1.36
5 2.4 117.00 112.80 3.72 155.46 150.17 3.52 181.60 182.93 0.73

Table 2
The crashing responses of BMCTs and MCTs obtained from FE simulations.

No Tubes Mass (kg) PCF (kN) Pm (kN) EA (kJ) SEA ULC

(kJ/kg)

1 Q-MCT 0.6415 117.46 87.80 14.75 22.99 0.0991

2 Q-BMCT−1 0.8359 158.48 120.71 20.28 24.26 0.1104
3 Q-BMCT−2 0.7738 146.11 117.74 19.78 25.56 0.0897
4 Q-BMCT−3 0.7428 144.39 114.76 19.28 25.96 0.1149
5 Q-BMCT−4 0.7738 149.29 129.70 21.79 28.16 0.0672
6 Q-BMCT−5 0.7738 154.66 130.60 21.94 28.35 0.0613
7 Q-BMCT−6 0.7428 151.28 132.38 22.24 29.94 0.0576
8 H-MCT 0.7516 148.00 114.88 19.30 25.68 0.0911
9 H-BMCT−1 1.0432 197.83 146.49 24.61 23.59 0.1481
10 H-BMCT−2 0.9501 239.48 169.05 28.40 29.89 0.0653
11 H-BMCT−3 0.9035 194.74 156.85 26.35 29.16 0.0991
12 H-BMCT−4 0.9501 202.25 173.39 29.13 30.66 0.0617
13 H-BMCT−5 0.9501 208.97 182.56 30.67 32.28 0.0617
14 H-BMCT−6 0.9035 210.08 182.08 30.59 33.86 0.0492
Fig. 8. The crashing force-displacement curves of O-BMCT-6, O-MCT, O-BMCT-4 and H-
15 O-MCT 0.8346 165.58 136.31 22.90 27.44 0.0633
16 O-BMCT−1 1.2235 259.06 209.23 35.15 28.73 0.1233 BMCT-2.
17 O-BMCT−2 1.0991 250.92 211.73 35.57 32.36 0.0759
18 O-BMCT−3 1.0369 226.29 196.07 32.94 31.77 0.0728
5. Crashworthiness comparison of different kinds of bionic multi-
19 O-BMCT−4 1.0991 266.11 213.33 35.84 32.61 0.0490
20 O-BMCT−5 1.0991 260.35 219.64 36.90 33.57 0.0792 cell tubes and traditional multi-cell tubes
21 O-BMCT−6 1.0369 251.73 222.74 37.42 36.09 0.0459
In order to compare the crashworthiness of bionic multi-cell tubes
Note that the red font indicates poor crashworthiness indicators, and the blue bold font and traditional multi-cell tubes, FE simulations of different tubes are
indicates the most excellent crashworthiness indicators.

Fig. 7. The deformation modes of BMCTs and MCTs.

47
L. Zhang et al. Thin-Walled Structures 122 (2018) 42–51

Table 3 BMCT-2 and O-BMCT-4 are greater than the other corresponding tubes,
FE simulation data used to construct the RBF metamodels. as shown in Table 2. It also can be observed from Table 2 that the SEA
for the octagonal cross-section are greater than for the hexagonal, and
No D (mm) t (mm) Mass (kg) PCF (kN) EA (kJ) SEA (kJ/kg) ULC
the hexagonal are greater than for quadrilateral, except the H-BMCT-1.
1 8 1.2 0.5809 112.08 15.94 27.44 0.0467 Thus, the octagonal cross-section is the best cross-section configuration
2 1.6 0.7745 167.45 23.98 30.96 0.0565 considering the SEA. From Table 2, it also can be seen that the Q-BMCT-
3 2.0 0.9681 219.36 32.49 33.56 0.0619
6, H-BMCT-6 and O-BMCT-6 have higher SEA and lower ULC compared
4 2.4 1.1617 277.00 41.54 35.76 0.0603
5 2.8 1.3554 357.61 51.00 37.63 0.0751 to the corresponding tubes, indicating that filling the cylindrical tubes
6 10 1.2 0.6016 121.40 16.71 27.78 0.0511 into the mid-walls of the outer tube is the best design. And, the O-
7 1.6 0.8022 183.66 26.07 32.50 0.0462 BMCT-6 has the highest SEA and the lowest ULC among these twenty-
8 2.0 1.0027 240.63 36.33 36.23 0.0484 one designs. To demonstrate the advantages of the O-BMCT-6 more
9 2.4 1.2033 311.96 45.78 38.05 0.0794
intuitively, we compare the crashworthiness performance of O-BMCT-6
10 2.8 1.4038 377.44 56.75 40.43 0.0680
11 12 1.2 0.6222 132.22 17.43 28.01 0.0485 and O-MCT numerically. It can be found that the SEA of O-BMCT-6 is
12 1.6 0.8296 197.06 26.88 32.40 0.0490 31.52% higher than that of O-MCT, and the ULC of O-BMCT-6 is
13 2.0 1.0369 251.73 37.42 36.09 0.0459 27.49% lower than that of O-MCT. This means the O-BMCT-6 is an
14 2.4 1.2444 315.72 48.22 38.75 0.0620
excellent energy absorber. For this reason, the multiobjective optimi-
15 2.8 1.4517 414.18 60.74 41.84 0.0847
16 14 1.2 0.6429 140.10 18.70 29.09 0.0483 zation of the O-BMCT-6 was conducted in next section.
17 1.6 0.8572 208.36 28.96 33.78 0.0544
18 2.0 1.0715 262.45 38.83 36.24 0.0583 6. Crashworthiness optimization
19 2.4 1.2858 336.03 50.28 39.10 0.0700
20 2.8 1.5001 408.24 61.44 40.96 0.0777
21 16 1.2 0.6633 144.88 19.55 29.47 0.0525 6.1. Definition of the optimization problem
22 1.6 0.8844 215.93 30.22 34.17 0.0560
23 2.0 1.1055 281.53 40.60 36.73 0.0609 As an energy absorber, the thin-walled tubes per unit mass are ex-
24 2.4 1.3266 367.20 51.31 38.68 0.0976 pected to absorb energy as much as possible. Thus, SEA should be one
25 2.8 1.5477 414.92 61.89 39.99 0.0827
26 18 1.2 0.6839 147.44 20.09 29.38 0.0493
of the objective functions and maximized in the crashworthiness opti-
27 1.6 0.9118 219.41 31.07 34.08 0.0558 mization problem. Meanwhile, PCF reflects the severity of the collision,
28 2.0 1.1398 293.43 41.19 36.14 0.0688 which is related to the occupant injury [54]. Then, PCF needs to be
29 2.4 1.3678 353.45 52.07 38.07 0.0698 constrained by a certain level. Therefore, the multiobjective optimiza-
30 2.8 1.5957 466.92 64.50 40.42 0.1053
tion of the O-BMCT-6 can be formulated as follows:

⎧ Min {−SEA (D , t ), PCF (D , t )}

carried out under axial impact loading. Table 2 lists the crashing re-
sponses of BMCTs and MCTs obtained from FE simulations. Fig. 7 shows
the deformation modes of BMCTs and MCTs. From Table 2, it can be
found that for each quadrilateral, hexagonal or octagonal cross-section,
the SEA of BMCTs are greater than the corresponding MCTs, except the
H-BMCT-1. The increase in SEA of BMCTs can be attributed to the
filling of cylindrical tubes, which required more energy for plastic de-
formation. However, the SEA of H-BMCT-1 did not see any increase,
which is due to the fact that the cylindrical tubes are simply filled into
the gaps of the multi-cell tube, and global buckling [36,53] occurred in
the cylindrical tubes when impacted, which affects the deformation of
the entire structure, as shown in Fig. 7(i). Furthermore, the occurrence
of unstable collapse modes can also be judged by ULC. It can be found
from Table 2 that the ULC of the H-BMCT-1 is 0.1481, which is higher
than other hexagonal multi-cell tubes. The phenomenon of higher ULC
can also be found in Q-BMCT-1 and O-BMCT-1, and these two tubes
both have unstable collapse modes, as shown in Fig. 7(b) and (p).
Therefore, filling the cylindrical tubes into the gaps of MCTs is not a
good design. In addition, it also can be found from Table 2 that the ULC
of Q-BMCT-1, H-BMCT-1 and O-BMCT-1 is greater than 0.1, and the
ULC of Q-BMCT-3 is also greater than 0.1. From Fig. 7(d), we can find
that the Q-BMCT-3 also has an unstable collapse mode. The reason for
this is that the position of cylindrical tubes is too concentrated and each
cylindrical tube is connected to three ribs, which results in excessive
local stiffness.
Due to the existence of cylindrical tubes, the crashing force-dis-
placement curves of BMCTs would be raised to a higher level compared
to MCTs. For example, the crashing force-displacement curve of O-
BMCT-6 is higher than O-MCT, as shown in Fig. 8. Thus, the Pm and PCF
of BMCTs are greater than the corresponding MCTs, as shown in
Table 2. However, there are two special cases, namely H-BMCT-2 and
O-BMCT-4. Their force-displacement curves are also plotted in Fig. 8.
According to Fig. 8, it can be seen that the crashing forces of H-BMCT-2
and O-BMCT-4 rise sharply at the end of the crash displacement due to
the densification of multi-cell tubes. Therefore, the crashing forces of H-
s. t . D L ≤ D ≤ DU
⎨ L ≤ t ≤ tU
⎩ t (29)
where DL and DU are the lower and upper limits of the diameter of the
cylindrical tubes, respectively. tL and tU denote lower and upper limits
of wall thickness, respectively. In this paper, DL and DU are 8 mm and
18 mm, tL and tU are 1.2 mm and 2.8 mm, respectively.
6.2. Metamodel and optimization algorithm
Since the structural crashworthiness problem involves complex
nonlinear mechanics, direct coupling of the optimization algorithm
with the simulation model may be inefficient. Therefore, metamodels
are extensively used to reduce computational costs [55,56]. The
common methods for metamodels building include the radial basis
function (RBF), Polynomial Response Surface (PRS), Moving Least
Square (MLS), Artificial Neural Network (ANN) and Kriging [57].
Among these metamodels, the RBF shows very good accuracy in the
highly nonlinear impact problem [58,59]. Thus, this approach was
chosen for this study. The mathematical form of RBF can be expressed
as:
n
̂ (x) =
yRBF ∑ λi ϕ (||x − xi||)
i=1 (30)
where xi is the vector of design variables at the ith design point, ||x −
xi|| is the Euclidean distance, λi is the coefficient of the linear combi-
nation of n neurons and ϕ is a basis function. In order to construct the
RBF model, the full factorial method was employed to generate 30
sampling points over the design space. The corresponding numerical
simulations were conducted to evaluate the crashing responses.
The accuracy of the optimization result depends on the fitting ac-
curacy of the metamodel. To examine the accuracy of metamodels, five
extra random validation points in the design space were generated to
assess the relative error (RE) between the simulation results y(x) and
the predicted values ŷ(x) by the RBF model, written as:

48
L. Zhang et al. Thin-Walled Structures 122 (2018) 42–51

Fig. 9. The variation in SEA, PCF and ULC with different diameters and thickness.

Table 4
The comparison of SEA and PCF between FE simulations and RBF predictions.

No D (mm) t (mm) SEA (kJ/kg) PCF (kN)

FE. RBF. RE (%) FE. RBF. RE (%)

1 11.51 2.64 40.53 40.57 0.10 364.24 371.00 1.86

2 8.76 1.58 32.20 31.44 2.36 164.14 170.65 3.97
3 17.42 1.74 35.14 35.13 0.03 249.01 243.86 2.07
4 13.75 1.37 31.32 30.95 1.18 166.61 168.41 1.08
5 10.37 1.82 34.77 34.73 0.12 219.39 217.31 0.95

y (x) − y ̂(x)
RE =
y (x) (31)

To seek the Pareto front, the commonly adopted approaches are

non-dominated sorting genetic algorithm II (NSGA-II) [60–62], multi-
objective particle swarm optimization algorithm (MOPSO) [63], Pareto Fig. 10. The Pareto front and the collapse mode of the O-BMCT-6.
archived evolution strategy (PAES) [64] and et al. In this study, the
NSGA-II algorithm is used to explore the optimal solution due to its fast
Table 5
convergence and well-distributed Pareto front [65]. The optimal design parameters obtained from FE simulation and RBF prediction for the O-
BMCT-6 when PCF ≤ 250 kN.

6.3. Optimization results D (mm) t (mm) SEA (kJ/kg) PCF (kN)

Table 3 is the summary of 30 FE simulation data used to construct FE. RBF. RE (%) FE. RBF. RE (%)
the RBF metamodels. To observe the data in Table 3 intuitively, Fig. 9
10.24 2.04 36.54 36.73 0.52 248.62 247.30 0.53
displays the variation in SEA, PCF and ULC with different diameters and
thickness. It can be found from Fig. 9(a) and (b) that both SEA and PCF

49
L. Zhang et al.
increase as the thickness increases. Therefore, both SEA and PCF reach
a maximum value when the thickness equals to 2.8 mm. In addition, the
PCF exhibits the growth trend with increasing diameter in general.
However, the SEA does not increase noticeably with increasing dia-
meter when the diameter D ≥ 10 mm. Specifically, the difference of
SEA varying with different diameters is not significant, and the SEA
tends to vary within a narrow range when the diameter increases. From
Fig. 9(c), it can be seen that the tendency of ULC is not very regular, but
the value of ULC is relatively small when the thickness is small, which
imply that the energy absorption process is more stable. Since the
tendency of ULC is irregular, it is difficult to construct a metamodel of
ULC with sufficient accuracy. Therefore, ULC is not chosen as the op-
timization objective in this paper.
The comparison of SEA and PCF between FE simulations and RBF
predictions at 5 validation points are given in Table 4. It can be found
that the relative error in terms of SEA and PCF is less than 4%.
Therefore, the RBF models provided high accuracy for design optimi-
zation.
The Pareto front of the O-BMCT-6 which is obtained by the NSGA-II
algorithm and the RBF models is plotted in Fig. 10. From Fig. 10, it can
be observed that -SEA and PCF conflict with each other. This means the
increase in SEA will lead to increase in PCF and vice versa. Actually,
any point in the Pareto front could be an optimum. Which point should
be chosen was fully decided by the engineering design problem. For
example, if the PCF needs to be constrained under 250 kN, the Pareto
point corresponding to the optimal design is marked as the red solid
symbol in Fig. 10, and its detailed design parameters are listed in
Table 5. The FE model of the optimal design is then established to
compare with the RBF model. The SEA and PCF obtained from the FE
simulation and the RBF model are compared in Table 5. It can be found
that the errors are less than 1%, which indicates the optimal results are
accurate enough. It also can be found from Fig. 10 that the collapse
mode of the optimal design is stable, which demonstrates that the O-
BMCT-6 is an excellent energy absorber.
7. Conclusion
In this paper, eighteen kinds of bionic multi-cell tubes (BMCTs)
based on the microstructure of beetle forewings were proposed and
compared with traditional multi-cell tubes (MCTs). The finite element
(FE) models of these tubes were established by using the explicit FE
code LS-DYNA and then validated by the Simplified Super Folding
Element (SSFE) theory. The numerical simulation results agreed well
with the theoretical results. According to the numerical simulation re-
sults, it was found that the crashworthiness of BMCTs is better than
MCTs, except for the BMCTs which filled the cylindrical tubes into the
gaps of MCTs and the Q-BMCT-3, H-BMCT-2 and O-BMCT-4. The oc-
tagonal cross-section is the best cross-section configuration considering
the SEA. The middle wall of the outer tube filled with the cylindrical
tubes is the best position for achieving higher SEA and lower ULC. And,
the O-BMCT-6 has the highest SEA and the lowest ULC among these
twenty-one designs.
To find the optimal design of BMCT, the multiobjective optimization
of the O-BMCT-6 was conducted by using the RBF metamodels and
NSGA-II algorithm. By observing the 30 design points used to construct
the RBF metamodels, it was discovered that both SEA and PCF increase
as the thickness increases. The PCF exhibited the growth trend with
increasing diameter in general, while the difference of SEA varying with
different diameters is not significant. The tendency of ULC is not very
regular, but the value of ULC is relatively small when the thickness is
small. Besides, the Pareto front of the O-BMCT-6 was obtained. It was
observed that -SEA and PCF conflict with each other. And, one optimal
design of the O-BMCT-6 with the constraint of PCF ≤ 250 kN was got
from the Pareto front. The relative error between the simulation result
and the RBF prediction at this optimal design is less than 1%. The
collapse mode of the O-BMCT-6 at this optimal design is stable, which
50
Thin-Walled Structures 122 (2018) 42–51
indicates that the O-BMCT-6 is an excellent energy absorber.
Acknowledgments
The authors would like to thank the financial support from National
Science Foundation (No. 51475153), the Science Foundation of State
Key Laboratory of Advanced Design and Manufacturing for Vehicle
Body (No. 71575004) and Shenzhen Science and Technology Program
(No. JCYJ20160530193357681).
References
[1] W. Abramowicz, Thin-walled structures as impact energy absorbers, Thin-Walled
Struct. 41 (2003) 91–107.
[2] G. Lu, T. Yu, Energy Absorption of Structures and Materials, Woodhead Publishing
Ltd, Cambridge, 2003.
[3] A. Airoldi, G. Janszen, A design solution for a crashworthy landing gear with a new
triggering mechanism for the plastic collapse of metallic tubes, Aerosp. Sci.
Technol. 9 (2005) 445–455.
[4] A. Baroutaji, M. Sajjia, A.G. Olabi, On the crashworthiness performance of thin-
walled energy absorbers: recent advances and future developments, Thin-Walled
Struct. 118 (2017) 137–163.
[5] S. Guillow, G. Lu, R. Grzebieta, Quasi-static axial compression of thin-walled cir-
cular aluminium tubes, Int. J. Mech. Sci. 43 (2001) 2103–2123.
[6] J. Lellep, K. Torn, Plastic response of a circular cylindrical shell to dynamic load-
ings, Int. J. Impact Eng. 30 (2004) 555–576.
[7] C. Reuter, K.H. Sauerland, T. Tröster, Experimental and numerical crushing analysis
of circular CFRP tubes under axial impact loading, Compos. Struct. 174 (2017)
33–44.
[8] M.M. Abedi, A. Niknejad, G.H. Liaghat, M.Z. Nejad, Theoretical and experimental
study on empty and foam-filled columns with square and rectangular cross section
under axial compression, Int. J. Mech. Sci. 65 (2012) 134–146.
[9] X. Zhang, Z. Wen, H. Zhang, Axial crushing and optimal design of square tubes with
graded thickness, Thin-Walled Struct. 84 (2014) 263–274.
[10] R. Lu, X. Liu, S. Chen, X. Hu, L. Liu, Axial crashing analysis for tailor rolled square
tubes with axially graded both wall thickness and material strength, Thin-Walled
Struct. 117 (2017) 10–24.
[11] S. Hou, Q. Li, S. Long, X. Yang, W. Li, Design optimization of regular hexagonal
thin-walled columns with crashworthiness criteria, Finite Elem. Anal. Des. 43
(2007) 555–565.
[12] X. Zhang, H. Huh, Crushing analysis of polygonal columns and angle elements, Int.
J. Impact Eng. 37 (2010) 441–451.
[13] F. Tarlochan, F. Samer, A. Hamouda, S. Ramesh, K. Khalid, Design of thin wall
structures for energy absorption applications: enhancement of crashworthiness due
to axial and oblique impact forces, Thin-Walled Struct. 71 (2013) 7–17.
[14] A. Mamalis, D. Manolakos, M. Ioannidis, P. Kostazos, C. Dimitriou, Finite element
simulation of the axial collapse of metallic thin-walled tubes with octagonal cross-
section, Thin-Walled Struct. 41 (2003) 891–900.
[15] Y. Liu, Crashworthiness design of multi-corner thin-walled columns, Thin-Walled
Struct. 46 (2008) 1329–1337.
[16] Z. Bai, J. Liu, F. Zhu, F. Wang, B. Jiang, Optimal design of a crashworthy octagonal
thin-walled sandwich tube under oblique loading, Int. J. Crashworthiness. 20
(2015) 401–411.
[17] M. Yamashita, M. Gotoh, Y. Sawairi, Axial crush of hollow cylindrical structures
with various polygonal cross-sections: numerical simulation and experiment, J.
Mater. Process. Technol. 140 (2003) 59–64.
[18] A.A. Nia, J.H. Hamedani, Comparative analysis of energy absorption and de-
formations of thin walled tubes with various section geometries, Thin-Walled
Struct. 48 (2010) 946–954.
[19] Z. Fan, G. Lu, K. Liu, Quasi-static axial compression of thin-walled tubes with dif-
ferent cross-sectional shapes, Eng. Struct. 55 (2013) 80–89.
[20] M. Ali, E. Ohioma, F. Kraft, K. Alam, Theoretical, numerical, and experimental
study of dynamic axial crushing of thin walled pentagon and cross-shape tubes,
Thin-Walled Struct. 94 (2015) 253–272.
[21] J. Fang, Y. Gao, G. Sun, G. Zheng, Q. Li, Dynamic crashing behavior of new ex-
trudable multi-cell tubes with a functionally graded thickness, Int. J. Mech. Sci. 103
(2015) 63–73.
[22] J. Fang, Y. Gao, G. Sun, N. Qiu, Q. Li, On design of multi-cell tubes under axial and
oblique impact loads, Thin-Walled Struct. 95 (2015) 115–126.
[23] G. Sun, H. Jiang, J. Fang, G. Li, Q. Li, Crashworthiness of vertex based hierarchical
honeycombs in out-of-plane impact, Mater. Des. 110 (2016) 705–719.
[24] J. Fang, G. Sun, N. Qiu, G.P. Steven, Topology optimization of multicell tubes under
out-of-plane crushing using a modified artificial bee colony algorithm, J. Mech. Des.
139 (2017) 071403.
[25] W. Shen, X. Gu, P. Jiang, J. Hu, X. Lv, L. Qian, Crushing analysis and multiobjective
optimization design for rectangular unequal triple-cell tubes subjected to axial
loading, Thin-Walled Struct. 117 (2017) 190–198.
[26] T.N. Tran, Crushing analysis under multiple impact loading cases for multi-cell
triangular tubes, Thin-Walled Struct. 113 (2017) 262–272.
[27] D. Zhang, Q. Fei, P. Zhang, In-plane dynamic crushing behavior and energy ab-
sorption of honeycombs with a novel type of multi-cells, Thin-Walled Struct. 117
(2017) 199–210.
L. Zhang et al.
[28] S. Pirmohammad, S.E. Marzdashti, Crushing behavior of new designed multi-cell
members subjected to axial and oblique quasi-static loads, Thin-Walled Struct. 108
(2016) 291–304.
[29] N. Qiu, Y. Gao, J. Fang, Z. Feng, G. Sun, Q. Li, Theoretical prediction and optimi-
zation of multi-cell hexagonal tubes under axial crashing, Thin-Walled Struct. 102
(2016) 111–121.
[30] W. Chen, T. Wierzbicki, Relative merits of single-cell, multi-cell and foam-filled
thin-walled structures in energy absorption, Thin-Walled Struct. 39 (2001)
287–306.
[31] X. Zhang, G. Cheng, H. Zhang, Theoretical prediction and numerical simulation of
multi-cell square thin-walled structures, Thin-Walled Struct. 44 (2006) 1185–1191.
[32] X. Zhang, H. Zhang, Numerical and theoretical studies on energy absorption of
three-panel angle elements, Int. J. Impact Eng. 46 (2012) 23–40.
[33] X. Zhang, H. Zhang, Energy absorption of multi-cell stub columns under axial
compression, Thin-Walled Struct. 68 (2013) 156–163.
[34] X. Zhang, H. Zhang, Theoretical and numerical investigation on the crush resistance
of rhombic and kagome honeycombs, Compos. Struct. 96 (2013) 143–152.
[35] A.A. Nia, M. Parsapour, Comparative analysis of energy absorption capacity of
simple and multi-cell thin-walled tubes with triangular, square, hexagonal and
octagonal sections, Thin-Walled Struct. 74 (2014) 155–165.
[36] J. Fang, G. Sun, N. Qiu, T. Pang, S. Li, Q. Li, On hierarchical honeycombs under out-
of-plane crushing, Int. J. Solids Struct. (2017), http://dx.doi.org/10.1016/j.ijsolstr.
2017.08.013.
[37] J. Chen, G. Dai, Y. Xu, M. Iwamoto, Optimal composite structures in the forewings
of beetles, Compos. Struct. 81 (2007) 432–437.
[38] J. Chen, Q.Q. Ni, Y. Xu, M. Iwamoto, Lightweight composite structures in the
forewings of beetles, Compos. Struct. 79 (2007) 331–337.
[39] J. Chen, G. Dai, Y. Xu, M. Iwamoto, Basic study of biomimetic composite materials
in the forewings of beetles, Mater. Sci. Eng. A 483 (2008) 625–628.
[40] J. Chen, C. Gu, S. Guo, C. Wan, X. Wang, J. Xie, X. Hu, Integrated honeycomb
technology motivated by the structure of beetle forewings, Mater. Sci. Eng. C 32
(2012) 1813–1817.
[41] J. Chen, J. Xie, H. Zhu, S. Guan, G. Wu, M.N. Noori, S. Guo, Integrated honeycomb
structure of a beetle forewing and its imitation, Mater. Sci. Eng. C 32 (2012)
613–618.
[42] J. Chen, C. He, C. Gu, J. Liu, C. Mi, S. Guo, Compressive and flexural properties of
biomimetic integrated honeycomb plates, Mater. Des. 64 (2014) 214–220.
[43] J. Xiang, J. Du, Energy absorption characteristics of bio-inspired honeycomb
structure under axial impact loading, Mater. Sci. Eng. A 696 (2017) 283–289.
[44] A.A. Nia, S. Chahardoli, Optimizing the layout of nested three-tube structures in
quasi-static axial collapse, Thin-Walled Struct. 107 (2016) 169–181.
[45] G. Sun, T. Pang, J. Fang, G. Li, Q. Li, Parameterization of criss-cross configurations
for multiobjective crashworthiness optimization, Int. J. Mech. Sci. 124 (2017)
145–157.
[46] S.P. Santosa, T. Wierzbicki, A.G. Hanssen, M. Langseth, Experimental and numerical
studies of foam-filled sections, Int. J. Impact Eng. 24 (2000) 509–534.
[47] H.S. Kim, New extruded multi-cell aluminum profile for maximum crash energy
absorption and weight efficiency, Thin-Walled Struct. 40 (2002) 311–327.
51
Thin-Walled Structures 122 (2018) 42–51
[48] Y. Xiang, T. Yu, L. Yang, Comparative analysis of energy absorption capacity of
polygonal tubes, multi-cell tubes and honeycombs by utilizing key performance
indicators, Mater. Des. 89 (2016) 689–696.
[49] T. Wierzbicki, W. Abramowicz, On the crushing mechanics of thin-walled struc-
tures, J. Appl. Mech. 50 (1983) 727–734.
[50] S. Xie, W. Yang, N. Wang, H. Li, Crashworthiness analysis of multi-cell square tubes
under axial loads, Int. J. Mech. Sci. 121 (2017) 106–118.
[51] T. Tran, S. Hou, X. Han, M. Chau, Crushing analysis and numerical optimization of
angle element structures under axial impact loading, Compos. Struct. 119 (2015)
422–435.
[52] A.G. Hanssen, M. Langseth, O.S. Hopperstad, Static and dynamic crushing of cir-
cular aluminium extrusions with aluminium foam filler, Int. J. Impact Eng. 24
(2000) 475–507.
[53] W. Abramowicz, N. Jones, Transition from initial global bending to progressive
buckling of tubes loaded statically and dynamically, Int. J. Impact Eng. 19 (1997)
415–437.
[54] N. Qiu, Y. Gao, J. Fang, Z. Feng, G. Sun, Q. Li, Crashworthiness analysis and design
of multi-cell hexagonal columns under multiple loading cases, Finite Elem. Anal.
Des. 104 (2015) 89–101.
[55] J. Fang, G. Sun, N. Qiu, N.H. Kim, Q. Li, On design optimization for structural
crashworthiness and its state of the art, Struct. Multidiscip. Optim. 55 (2017)
1091–1119.
[56] H. Fang, M. Rais-Rohani, Z. Liu, M. Horstemeyer, A comparative study of meta-
modeling methods for multiobjective crashworthiness optimization, Comput.
Struct. 83 (2005) 2121–2136.
[57] F. Djamaluddin, S. Abdullah, A. Ariffin, Z. Nopiah, Optimization of foam-filled
double circular tubes under axial and oblique impact loading conditions, Thin-
Walled Struct. 87 (2015) 1–11.
[58] E. Acar, M. Rais-Rohani, Ensemble of metamodels with optimized weight factors,
Struct. Multidiscip. Optim. 37 (2009) 279–294.
[59] X. Song, G. Sun, G. Li, W. Gao, Q. Li, Crashworthiness optimization of foam-filled
tapered thin-walled structure using multiple surrogate models, Struct. Multidiscip.
Optim. 47 (2013) 221–231.
[60] K. Deb, A. Pratap, S. Agarwal, T. Meyarivan, A fast and elitist multiobjective genetic
algorithm: NSGA-II, IEEE Trans. Evol. Comput. 6 (2002) 182–197.
[61] X. Liao, Q. Li, X. Yang, W. Zhang, W. Li, Multiobjective optimization for crash safety
design of vehicles using stepwise regression model, Struct. Multidiscip. Optim. 35
(2008) 561–569.
[62] X. Liao, Q. Li, X. Yang, W. Li, W. Zhang, A two-stage multi-objective optimisation of
vehicle crashworthiness under frontal impact, Int. J. Crashworthiness 13 (2008)
279–288.
[63] G. Sun, G. Li, M. Stone, Q. Li, A two-stage multi-fidelity optimization procedure for
honeycomb-type cellular materials, Comput. Mater. Sci. 49 (2010) 500–511.
[64] J.D. Knowles, D.W. Corne, Approximating the nondominated front using the Pareto
archived evolution strategy, Evol. Comput. 8 (2000) 149–172.
[65] H.S. Kim, J.W. Kang, Semi-active fuzzy control of a wind-excited tall building using
multi-objective genetic algorithm, Eng. Struct. 41 (2012) 242–257.