International Journal of Impact Engineering 64 (2014) 62e74

Contents lists available at ScienceDirect

International Journal of Impact Engineering

journal homepage: www.elsevier.com/locate/ijimpeng

Crashing analysis and multiobjective optimization for thin-walled

structures with functionally graded thickness
Guangyong Sun a, Fengxiang Xu a, Guangyao Li a, *, Qing Li b
a
b

State Key Laboratory of Advanced Design and Manufacturing for Vehicle Body, Hunan University, Changsha, Hunan 410082, PR China
School of Aerospace, Mechanical and Mechatronic Engineering, The University of Sydney, Sydney, NSW 2006, Australia

a r t i c l e i n f o

a b s t r a c t

Article history:
Received 15 February 2013
Received in revised form
25 September 2013
Accepted 12 October 2013
Available online 19 October 2013

Thin-walled structures have exhibited signicant advantages in light weight and energy absorption and
been widely applied in automotive, aerospace, transportation and defense industries. Unlike existing
thin-walled structures with uniform thickness, this paper introduces functionally graded structures with
changing wall thickness along the longitudinal direction in a certain gradient (namely, functionally
graded thickness e FGT). Its crashing behaviors are the key topics of the present study. We examine the
crashing characteristics of functionally graded thin-walled structures and evaluate the effect of different
thickness gradient patterns on crashing behaviors. It is shown that the gradient exponent parameter n
that controls the variation of thickness has signicant effect on crashworthiness. To optimize crashworthiness of the FGT tubes, the Non-dominated Sorting Genetic Algorithm (NSGA-II) is used to seek for
an optimal gradient, where a surrogate modeling method, specically response surface method (RSM), is
adopted to formulate the specic energy absorption (SEA) and peak crashing force functions. The results
yielded from the optimization indicate that the FGT tube is superior to its uniform thickness counterparts
in overall crashing behaviors. Therefore, FGT thin-walled structures are recommended as a potential
absorber of crashing energy.
Crown Copyright
2013 Published by Elsevier Ltd. All rights reserved.

Keywords:
Multiobjective optimization
Tailor rolled blank (TRB)
Crashworthiness
Non-uniform thickness sheet
Functionally graded thickness

1. Introduction
Over the past years, research interests in crashworthiness have
resulted in a series of systematic investigations into crash responses of various thin-walled structures via analytical, experimental and numerical approaches [1e8]. Two signicant aspects,
namely weight and crashworthiness, have drawn primary attention
in these studies. As an effective structure, thin-walled components
have showed signicant advantages over other solid elements and
are capable of carrying substantial loads with desired deformation,
which could be appreciably higher than the corresponding ultimate
or bulking loads [9,10]. In reality, thin-walled structural members
play a critical role on enhancing the capability of energy absorption
in impact engineering.
The automobile body in white (BIW) is mainly composed of
thin-walled structural parts, which are made by stamping or
forming process of traditional metal sheets with uniform thickness
[11e19]. It is of great interests in investigating the crashworthiness
of thin-walled structures for improving the vehicle safety and light

* Corresponding author. Tel.: 86 731 8882 1717; fax: 86 731 8882 2051.
E-mail addresses: sgy800@126.com (G. Sun), gyli@hnu.edu.cn (G. Li).

weight. In this regard, Zhang et al. [12] evaluated the energy absorption characteristics of regular polygonal and rhombic columns
under quasi-static axial compression. Song et al. [13] introduced
origami patterns into thin-walled structures and minimized the
initial peak force and subsequent uctuations. Tang et al. [20]
presented a cylindrical multi-cell column to improve energy absorption. Naja and Rais-Rohani [21] proposed a sequentially
coupled nonlinear nite element analysis (FEA) technique to
investigate the effects of sheet-forming process and design parameters on energy absorption of thin-walled tubes made of
magnesium alloy. Acar et al. [22] studied the crashing performances
of tapered tubes using multiobjective optimization. Although such
thin-walled structures have been extensively used as energy absorbers for their high energy absorption capacity, light weight and
low cost [23], all these thin walled structures were based upon the
uniform material and/or the same wall thickness. The inherent
shortcoming resides on that such structures may not exert their
maximum capacities in crashworthiness, and furthermore, a uniform wall thickness does not necessarily make best use of material
for meeting the requirements of vehicular light weight [24e26]. So
there is an urgent need to develop new structural conguration
with different material and/or thickness combinations for maximizing crashworthiness and material usage.

0734-743X/$ e see front matter Crown Copyright
2013 Published by Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.ijimpeng.2013.10.004

G. Sun et al. / International Journal of Impact Engineering 64 (2014) 62e74

According to Yang et al. [27], a metal sheet with varying thickness could be a more desirable structure because it not only uses
material more efciently, but also increases design exibility
considerably. It has been demonstrated that with an optimal choice
of different materials grades (e.g. via tailored welded blanks (TWB),
or hybrid blanks) and/or thicknesses (e.g. via tailor rolling blanks),
crashing performance of the combined components can be
improved to a higher extent. Indeed, design of specic thin-walled
components with desired materials/thicknesses in a more efcient
manner could represent new potential for further reducing weight
and enhancing performance of the products. Of these components
with variable material/thickness, the TWB structures, which consists of laser-welded sheet metals of different thicknesses and
materials, provide a exible combination of component materials
and thicknesses, which has been adopted in vehicular oor
component [28], B-pillar [29], front-end structure [30], and door
inner panels etc. [31,32]. The main shortcoming of those blanks lies
in that it consists of discrete thickness sections and may lead to
stress concentration in the interfaces. To overcome such defects of
TWB, a relatively new rolling process, named tailor rolled blank
(TRB), has been developed. In the newly developed TRB process, the
rolling gap can be varied, which leads to a continuous thickness
variation in the workpiece. Applications of such a rolling process
allow reducing more weight compared with traditional stamping
or forming processes. As such, varying sheet thickness can better
meet more and more demanding design requirements, thereby
enhancing utilization of material and/or thickness comparing with
traditional stamping uniform sheets.
There have been some reports on TRB in sheet metal forming.
For example, Zhang et al. [33] investigated the effects of transition
zone length, blank thickness variation, friction coefcient and die
clearance on the springback of TRB component. Meyer et al. [34]
used TRB to increase the maximum drawing depth compared to
the blanks with constant thickness. Urban et al. developed a design
tool by combining numerical simulation and optimization algorithm to improve the formability of TRB [35]. To the authors best
knowledge, however, very limited studies on crashworthiness
design of thin-walled TRB structures have been performed to date.
To make use of TRB thin-walled structures with functionally
graded thicknesses (FGT) for impact engineering, it is essential to
understand the energy absorption characteristics in comparison
with those well-studied uniform thickness (UT) thin-walled
structures. More importantly, it is of particular importance to
seeking the best possible thickness gradient for crashing performance with different measures. Thus the objective of this paper
resides in quantifying and improving crashing behaviors of thinwalled structures with functionally graded wall thickness. For this
reason, two critical issues need to be addressed in this paper: (1) a
direct problem that quanties the crashing characteristic of functionally graded thin-walled structures with variable wall thickness
and evaluates the effects of the different thickness pattern on both
specic energy absorption (SEA) and peak impact forces; (2) an
inverse problem that seeks optimal gradient for maximizing the
specic absorption energy (SAE) and minimizing peak crashing
force (Fmax).
As for the functionally graded thickness structure, the thickness
of thin-walled varies throughout the depth in an ascending or
descending gradient. It is expected that the gradient exponential
parameter (n) has a signicant effect on crashworthiness. To
represent such complex crashworthiness objective functions with
respect to gradient parameter, which has not been explored in
literature before, a surrogate model technique, namely specically
response surface method (RSM), will be attempted here. To maximize the energy absorption and minimize the peak crashing force,
the multiobjective optimizations for the FGT structures are

63

formulated and the Non-dominated Sorting Genetic Algorithm

(NSGA-II) is applied for its proven effectiveness in crashworthiness
design [36,37].
2. High-strength steel column structures with functionally
graded wall thickness
2.1. Geometrical description and material properties
Dynamic axial crushing simulation was performed in the square
tubes which were made of high-strength steel grade DP800 [38].
The dynamic procedure was conducted at velocities of 5 m/s, 10 m/s
and 15 m/s, respectively, with an impacting mass of 600 kg in order
to assess the crash behaviors measured in the impact force and
energy absorption. Fig. 1 illustrates the geometry of thin-walled
square structure in the dynamic tests. These specimens have a
nominal square core cross-section with rounded corners and the
average dimensions of 60 mm  60 mm. The strain-rate dependent
properties of DP800 are considered herein and the true-stress
versus true-plastic strain curves at different strain rates
(0.000903/s, 1.029/s, 278/s and 444/s, respectively) are plotted in
Fig. 2. To characterize the material behavior, an empirical constitutive equation for the effective yield stress as a function s of the
effective plastic strain is t to the following formulae [39]:

s0

2
X
i1

!
q
_
Qi 1  expCi
1
_ 0

(1)

where s0 is the initial yield stress, and Qi and Ci denote strain

hardening coefcients, q represents a material constant and _ 0 is a
user-dened reference strain rate. The relevant material properties
are summarized in Table 1. It is assumed that the material properties remain the same regardless of variation in sheet thickness.
2.2. Structural crashworthiness criteria
The design optimization aims to generate a controllable
crashing pattern for maximizing energy absorption and minimizing
the peak forces during collapse [7]. There are several key indicators
to evaluate crashworthiness of a structure, e.g. energy absorption
(EA), specic energy absorption (SEA), average crash force (Favg),
and crash force efciency (CFE), as given in Eqs. (2)e(5) respectively
(Fig. 3), are widely used in the measurement.
As a key indicator, the energy absorption (EA) of a structure
measures the capacity of absorbing impact energy, which can be
determined mathematically as,

Zd
EA

Fddd

(2)

where F(d) is the instantaneous crashing force with a function of

the displacement d.
The specic energy absorption (SEA) assesses the absorbed
energy per unit mass of a structure as,

SEA

EA
m

(3)

where m is the total mass of the structure. In this case, a higher

value indicates a higher energy absorption efciency of material.
The average crashing force Favg for a given deformation also
indicates the capacity of energy-absorption of a structure, which is
calculated as EA divided by the compressive displacement d as [35],

64

G. Sun et al. / International Journal of Impact Engineering 64 (2014) 62e74

V0=5 m/s, 10 m/s, 15 m/s

Unit: mm

t=1.2

R=3

a=60
Seam weld
L2=100 mm

Impacting
mass=600 kg

L1=310 mm

b=60

(a)

(b)

Fig. 1. Schematic of (a) experimental set-up for dynamic tests and (b) geometry description of square tube [39].

Favg

EA

(4)

From Zarei and Kroger [41], the crash force efciency (CFE) and
specic energy absorption can increase simultaneously. The CFE
indicates the uniformity of forceedisplacement curve, meaning
that the higher the CFE, the more efcient the structure. Thus, CFE
of a structure is formulated as other key indicator,

CFE

Favg
 100%
Fmax

(5)

2.3. Discrete functionally graded column

It is assumed that the square column is fabricated with functionally graded thickness sheet whose direction of thickness
gradient is identical to the direction of the axial impact velocity.
Fig. 4 illustrates the key geometrical features (in a side view) of the
square column under consideration, in which we ignore the effect
of changing inner surface and outer surfaces as this difference is not
required in nite element modeling. The idea is similar to the
density gradient of the functionally graded foam materials developed by us previously [42]. The top wall thickness ttop is chosen at
the origin of gradient axis, whilst the bottom wall thickness tbot
corresponds to the farthest layer from the origin. In the axial
grading case, the thickness gradient function tf(x) can be dened in
terms of the following power law:

1200

True stress (MPa)

1000
800
600

Strain Rate (1/s)

0.000903
1.029
278
444

400
200
0
0.00

0.05

0.10
0.15
Plastic Strain

0.20

0.25

Fig. 2. True-stress versus true-plastic strain curves at different strain rates for DP800
high-strength steel [39].

tf x

xn

for an ascending pattern

 n
tmax  tmax  tmin xL

for a descending pattern

tmin tmax  tmin

(6)
where x is the distance from the origin (top) of the column, and n is
the grading exponent and assumed to vary between 0 and 10, L
denotes the total length of column. When the total layer Ns is
approaching to innite, the wall thickness will be graded continuously. Thus, the wall thickness increases along the length with a
gradient function changing from convexity to concavity when the n
value varies from less than 1 to greater than 1, as shown in Fig. 5a.
On the contrary, the wall thickness decreases along length, an
opposite tendency is observed as shown in Fig. 5b. Herein, the
ascending pattern is considered to perform the crashworthiness
design with the FGT structures.

3. Numerical modeling
3.1. Finite element (FE) modeling
The FE model used to simulate the crashing process was based
upon the experimental tests conducted by Tarigopula et al. [39]. An
explicit FE code, LS-DYNA, was used to implement the parallel
computing. Fig. 6 shows the FE mesh using the 4-node shell elements with 6 degrees of freedom at each node. In this model, the
bottom end is xed and a rigid mass-block of 600 kg was linked to
the loading end through a master node. In the crash scenario, the
mass block is assigned an initial velocity of 10 m/s through the
master node. The contact between the mass block surface and the
FGT tubes was modeled with a friction coefcient of 0.2, while for
the self-contact of the tube, frictional effects were neglected [43].
To nd the optimum mesh size for the numerical simulation, a
convergence test with ve different mesh sizes was carried out. In
comparison, a UT column having the same mass as its FGT counterpart was also employed to perform the mesh convergence.
Therefore, element size 3 mm  3 mm is adopted in the nite
element model. The energy absorption characteristics predicted by
different element sizes (exponent n 0) are summarized in Fig. 7,
which shows that there is very small difference between element
sizes 2 mm  2 mm and 2.5 mm  2.5 mm. It can be seen that the
mesh size adopted for all models of 2.0  2.0 mm is sufcient.
As for FGT, it is assumed that the depth of each layer is the same and
dened as Le, an equivalent thickness of UT column that has the same
volume of high-strength steel material can be calculated as follows,

tavg

NS
X
i1

ti Le =NS Le

NS
X
i1

ti =NS

(7)

G. Sun et al. / International Journal of Impact Engineering 64 (2014) 62e74

Table 1
Material properties for DP800 high-strength steel.
E(GPa) v
195

r (kg/m3) s0 (MPa) Q1(MPa) Q2(MPa) C1 C2 q

0.33 7850

495

200

233

_ 0 (1/s)

76 10 0.0116 0.001

65

presented in Fig. 8, in which the grading exponent (n) in Eq. (6) is

dened as 0.2 and the top (incident end) thickness ttop and the
bottom (distal end) thickness tbot are set to 0.8 mm and 2.2 mm,
respectively. There is no signicant difference between the
maximum peak loads of these three congurations and the curves
of 31 and 52 layers exhibit considerable consistency. Therefore, we
chose 31 horizontal layers for modeling the thickness gradient.

1000
3.2. Validation of the numerical models

900

Fmax

800

In order to validate the developed FE models, the modeling results of the UT square tubes under axial dynamic loading are rst
compared with the theoretical solutions available in the literature
[44], i.e.

Force F (kN)

700
600
500

1=3

Favg 13:06s0 bw t 5=3

(8)

400
where s0 is the characteristic stress of tube material, bw and t are
the width and wall thickness of tube, respectively. The comparison
of the crashing force versus displacement of the UT square tubes
(thickness t 1.2 mm) under an axial dynamic loading rate of 10 m/
s is shown in Fig. 9. It is seen that the present FEA results are in good
agreement with the theoretical solutions for the UT square tubes.

Favg

300
200

Ea

100
0

50

100

150 200 250 300

Displacement (mm)

350

400
1.0

Fig. 3. A typical relationship of force versus displacement of axial crashing behavior

with progressive folding [40].

Thickness ratio tmin/tmax

where NS denotes the total number of layers of FGT thin-walled

square column, and ti is thickness of the ith layer. As such, such
gradient exponent n would also affect the crashing results of the UT
column with uniform wall thicknesses but different values.
In an ideal functionally graded continuous model, the column
wall should be divided to an innite number of layers. In the FE
framework, the minimum depth of layer would be equal to the size
of each shell element, which could however lead to very costly
computational time. Furthermore, increasing the number of layers
could increase the risk of numerical instability in the model
resulting from the use of smaller element sizes. In this respect,
another convergence test was performed for three differently
layered congurations of 21, 31, and 52 layers, respectively, in order
to determine the optimum number of layers. Comparison of the
load-displacement response for various congurations of layers is

0.8

n=0.1
n=0.2

0.6

n=0.5
n=1

0.4

n=2
n=5

0.2

n=10

0.0
0.0

0.2

0.4
0.6
Normalized distance x/L

0.8

1.0

(a)
1.0
n=10
0.8

Bottom end (tbot)

Crash speed v0
1
2
3
4
5
6
7
8
9
10
11

...

Direction of thickness grading

Crash speed v0
Top end (ttop)

NS-11
NS-10
NS-9
NS-8
NS-7
NS-6
NS-5
NS-4
NS-3
NS-2
NS-1
NS

Fig. 4. Schematic showing thickness grading patterns in the axial direction.

Thickness ratio tmin/tmax

Loading plate (impacting mass)

n=5
n=2

0.6
n=1
0.4
0.2

n=0.5
n=0.2
n=0.1

0.0
0.0

0.2

0.4
0.6
0.8
Normalized distance x/L

1.0

(b)
Fig. 5. Variation of thickness vs normalized distance. (a) ascending gradient pattern,
and (b) descending gradient pattern.

66

G. Sun et al. / International Journal of Impact Engineering 64 (2014) 62e74

Fig. 6. Finite element model for dynamic impact simulation. (a) Experimental test; (b) Finite element model (2D); (c) 3D model.

In addition to the theoretical validation, in which the thickness

of t 1.2 mm was adopted, the FE model is also veried against the
experimental results in literature [39]. Fig. 10 illustrates the comparison of the force versus displacement curves between the
experiment and simulation at the velocity of 10 m/s. In terms of the
peak force (Fmax) and specic energy absorption (SEA), the experimental and simulation results are 135.06 kN and 130 kN, 10.6 kJ/kg
and 9.7 kJ/kg, respectively. It can be seen that there is also
reasonable agreement between the simulation and experimental
results. The validated numerical model provides considerable
condence for us to explore the energy absorption capacity of thinwalled sections with different gradients of wall thickness.
3.3. Numerical results of functionally graded column
In the present work, a group of graded and uniform thickness
thin-walled columns with the same weight were compared for
energy absorption and peak impact force (SAE and Fmax) respectively. The effects of thickness ranges and design parameter n (Eq.
(6)) on the resulted crashing characteristics of the axially thicknessgraded column are explored.
3.3.1. Effect of thickness range
Thickness range has a notable effect on the crashworthiness
performance. Herein, the value of ttop was varied with 0.8 mm,
1.2 mm, and 1.5 mm while tbot was xed at 2.2 mm in order to
quantify the effect of selecting different thickness range (Dt) on SEA
and Fmax. Variations in SEA and Fmax for various values of Dt and n
are illustrated in Fig. 11. It is noted again that equivalent wall
thickness was calculated as Eq. (7) for the corresponding UT tube to
keep the same weight as the functionally graded thickness (FGT)
tube with different parameter n. In other words, the wall thickness
of UT tube can also be expressed in term of n hereafter. Two
important points can be emphasized: (1) SEA and Fmax of both GT
and UT columns decrease with the increased exponent n; (2) SEA of
FGT column is generally higher than that of UT column at the same
exponent n whilst Fmax of FGT is smaller. This indicates that the FGT
column is superior to the UT column. In addition, the changes in
SEA and Fmax at different thickness ranges are summarized in
Table 2. For 0  n  1, the DFmax has a maximum value at range
Dt 1.4 mm, whilst for 1 < n  10 DSEA reaches a maximum value
at range Dt 1.4 mm except for n 8. Note that the DFmax and DSEA
are the amplitudes of variation in the peak force and specic energy
absorption when the FGT and UT tubes are compared with each

other. Overall, FGT with larger thickness range is more effective in

minimizing Fmax for 0  n  1; as well as in maximizing SEA for
1 < n  10.
The value of grading exponent (n) in Eq. (6) is tested in a series of
parameters of n 0, 0.2, 0.4, 0.6, 0.8, 1.0, 2.0, 4.0, 6.0, 8.0, 10.0, while
the top (incident end) thickness ttop and the bottom (distal end)
thickness tbot are set to 0.8 mm and 2.2 mm (i.e. Dt 1.4 mm),
respectively. As justied before, the total number of layers in the
FGT thin-walled square column is prescribed to be 31, which was
considered sufcient for modeling axially graded FGT structures
accurately. Note that each layer has the different color representing
different thickness sections for easy observation in the schematic
diagram (Fig. 12a). It is worth mentioning that the total weight of
thin-walled tube increases with the decreasing of gradient
parameter n (Fig. 12b). This change would be directly related to the
variation in SEA and Fmax.
3.3.2. Effect of different deformation distances
The relationship of SEA and Fmax versus mass of FGT and the
corresponding UT is shown in Figs. 13e15 at the timeframes of
15 ms, 20 ms, and 25 ms, respectively. Importantly, it can be
observed that the SEA of FGT columns is not necessarily better at
the initial stage of crashing. In a shorter crashing instance (e.g. at
15 ms in Fig. 13), the SEA of UT is even greater than that of the
corresponding FGT columns for all gradient parameters considered
here. However, FGT becomes more and more superior to the corresponding UT counterpart as the time increases (from Fig. 14 to
Fig. 15). It can be also observed that Fmax of the FGT column is much
smaller than that of the corresponding UT column no matter what
crashing displacement is. Note that gradient exponent n controls
the variation of thickness distributions and has a signicant effect
on SEA for the FGT columns. As seen in Fig. 15, SEA monotonously
decreases with increase of n in the region of n [0, 10] considered.
The reason lies in that the total weight is to decrease as the
exponent n increases. The similar trend can be seen for the UT
column.
Fig. 16 depicts the deformation modes of the FGT column and
the corresponding UT column with the same weight at different
timeframes and different exponents n. It can be seen that the
deformation of the FGT column seems to be more stable and advantageous over its uniform counterpart, especially with increased
parameter n.
To be brief, the columns with graded wall thickness have a
better crushing performance than the uniform thickness columns

G. Sun et al. / International Journal of Impact Engineering 64 (2014) 62e74

Dynamic crashing force by FE simulations

Dynamic average crashing force by FE simulations
Theoretical solution of Favg by Eq. (8)

140
120

Crashing force (kN)

67

100
80
60
40
20
0

in terms of SEA and Fmax in the later stage of impact (e.g. Fig. 15).
Therefore, the FGT component has signicant potential and seems
to be superior and ideal to the crashworthiness design. While
knowing considerable effect of the exponent of thickness gradient
function (Eq. (6)) on crashworthiness of FGT column, it remains a
question how to seek for best possible n by optimizing the crashing
characteristics of corresponding FGT columns, which forms another
goal of this study below.

4. Multiobjective optimization for functionally graded

thickness column
4.1. Optimization methodology
4.1.1. Denition of optimization problem
Design optimization is further applied for the crashworthiness
criteria in the FGT columns. In general, the SEA and force peak
(-Fmax) both are the key indicators [45,46]. An overly high Fmax
often leads to severe injury or even death of occupant. From the
vehicle safety point of view, the smaller the Fmax, the lower the
deceleration, thus the higher the safety. In this study, the optimal n

40

60
80 100 120
Displacment (mm)

140

160

180

is obtained based on the same impact initial velocity to further

perform the comparisons. However, SEA and Fmax often conict
with each other, i.e, the increase in SEA often leads to the increase
in Fmax. Therefore, to account for both different design criteria and
impose the optimum in a Pareto sense, the optimization problem
can be formulated mathematically in a multiobjective framework
as follows:

MaxSEAn; Fmax n
s:t: nLower  n  nUpper

(9)

4.1.2. Surrogate model and error metrics

It is often difcult to mathematically derive the analytical
objective functions for SEA and Fmax that involve highly nonlinear
contact-impact and large deformation mechanics. As an effective
alternative, the surrogate techniques, e.g. response surface method
(RSM), have proven particularly effective and been widely adopted
in crashworthiness design [45,47].
To estimate the tting accuracy of these surrogate models, such
error metrics as R square (R2), Relative Average Absolute Error

120

O riginP ro

E valuation

O riginP ro

21 layers
31 layers
E v a l u a52
t i o layers
n

O riginP ro

E valuation

O riginP ro

E valuation

O riginP ro

E valuation

O riginP ro

E valuation

O riginP ro

E valuation

O riginP ro

E valuation

O riginP ro

E valuation

O riginP ro

E valuation

O riginP ro

E valuation

O riginP ro

160
140
120

E valuation

100
80
60
40
20
0

50

100
150
200
Displacement (mm)

(a)

250

300

O riginP ro

E valuation

O riginP ro

E valuation

O riginP ro

E valuation

O riginP ro

E valuation

O riginP ro

E valuation

O riginP ro

E valuation

O riginP ro

E valuation

O riginP ro

21 layers
31 layers
52 layers
E valuation

40

O riginP ro

E valuation

O riginP ro

E valuation

20

O riginP ro

E valuation

O riginP ro

E valuation

100

Average crash force Favg (kN)

180

Crash force F (kN)

20

Fig. 9. Crashing force and displacement curves for square tubes under axial dynamic
loading.

Fig. 7. Internal energy versus crashing displacement (n 0).

80
60

50

100
150
200
Displacement (mm)

(b)

Fig. 8. Force-displacement responses for various numbers of horizontal layers (n 0.2).

250

300

68

G. Sun et al. / International Journal of Impact Engineering 64 (2014) 62e74

(RAAE) and Relative Maximum Absolute Error (RMAE) are adopted

in Eqs. (10)e(12), respectively [48e50].

Pm 

2
yj  b
yj
R 1
2
Pm 
j1 yj  y
j1

Pm
b
j1 yj  y j

RAAE P

m
j1 yj  y

RMAE

maxfjy1  b
y 1 j; :::; jym  b
y m jg

Pm

b
j1 yj  y =m

(10)

4.1.4. Procedure of multiobjective optimization for FGT structures

As in Fig. 17, a owchart is provided to clarify the procedure of
multiobjective optimization for thin-walled structures with
functionally-graded thickness (FGT). It is noted that in this case, the
gradient parameter(s) (e.g. exponent n, if a power law formula is
adopted herein as in Eq. (6)) should be one of the design variables in
order to reect their signicant roles in crashworthiness design.
4.2. Results and discussion

(11)

(12)

where b
y j and y are the corresponding predicted (or surrogate) and
mean values for the FEA value yj at each checking point j, respectively, m represents the number of these checking points. In general, the larger the R2 values, the more accurate the surrogate
model. The smaller the RAAE and RMAE, the better the metamodel.
4.1.3. Multiobjective genetic algorithm
Genetic algorithm (GA) is a popular global optimization tools
that was originated from mechanisms of natural evolution and
genetic principles, which is superior to many traditional optimization algorithms because of capability of avoiding trapping in local
optima for searching an optimum [51]. To deal with multiobjective
optimization, the non-dominated sorting GA (NSGA) algorithm has
proven effective by ranking the solutions with non-dominated
sorting and assigning the tness with the ranking. As an
improved version of NSGA, NSGA-II has been demonstrated to be
one of the most efcient algorithms for multiobjective optimization
in a number of benchmarking problems [52]. The NSGA-II algorithm was used in this study and the relevant NSGA-II parameters
are listed in Table 3.

4.2.1. Comparison of different RSM

Accuracies of the surrogate models can be measured in different
ways [50]. Note that it is difcult to check all metamodels tness
accuracies based on the existing sampling points properly. Hence, a
series of new checking points should be randomly generated to
better verify the accuracy of the constructed metamodels. For
design variable n, 10 additional checking points are obtained in the
design ranges of 0  n  1 and 1 < n  10, respectively.
Based on these experimental results, the polynomial RSM
models (Linear, Quadratic, Cubic, Quartic) of SEA and Fmax are
established as follows. In Eqs. (13)e(16), superscripts G and U
represent the functionally graded tubes and uniform tubes,
respectively. Thus, SEAG and FG
max indicate the specic energy absorption and maximum crashing force of functionally graded tube.
Correspondingly, the SEAU and FU
max indicate the specic energy
absorption and maximum crashing force of uniform tube.

  
8
23:29696  2:80979n 0  n  1
>
< SEAG n
   21:14585  0:56497n 1 < n  10
222:37182  114:94364n 0  n  1
*>
:FG
max n
141:69533  3:31388n 1 < n  10



Linear : 8
22:58031  4:98897n 0  n  1
>
< SEAU n
16:71194
 0:42712n 1 < n  10
  
304:14318  121:91000n 0  n  1
>
:FU
max n
181:59733  6:94079n 1 < n  10
(13)



  
23:487984:08330n1:27351n2 0  n  1


SEAG n
21:539870:76198n0:01791n2 1 < n  10


  
2
286:80818544:51939n429:57576n 0  n  1
>
G
*>
: Fmax


n
2
148:558676:74555n0:31197n 1 < n  10


  
Quadratic : 8
23:223369:27598n4:28701n2 0  n  1
>
U n


>
SEA
<
19:005771:57404n0:10427n2 1 < n  10


  
>
328:04650281:26548n159:35548n2 0  n  1
>
U
: Fmax


n
211:4640021:87412n1:35758n2 1 < n  10

(14)



 

23:26355  0:51731n  8:07786n2 6:23425n3 0  n  1


SEAG n
2
3
19:43504 1:10489n  0:38697n 0:02453n 1 < n  10


 

316:18392  1011:26717n 1653:56469n2  815:99262n3 0  n  1
>
G
*>
: Fmax


n
153:91233  11:49396n 1:34152n2  0:06240n3 1 < n  10


 

Cubic :
8
23:30785  10:61836n 7:80724n2  2:34682n3 0  n  1
>
U n


>
SEA

<
20:01888  2:47261n 0:29909n2  0:01181n3 1 < n  10


 

>
335:79308  404:34992n 482:12937n2  215:18260n3 0  n  1
>
U
: Fmax


n
231:33800  39:50129n 5:17950n2  0:23163n3 1 < n  10

(15)

8
>
>
<

8
>
>
<

G. Sun et al. / International Journal of Impact Engineering 64 (2014) 62e74

69



 

23:06291 6:44927n  42:91078n2 61:96692n3  27:86634n4 0  n  1


SEAG n
22:29852  2:56623n 0:95643n2  0:15902n3 0:00834n4 1 < n  10


 

>
330:47713
 1507:55942n 4135:02593n2  4786:33061n3 1985:16900n4 0  n  1
G
*>
: Fmax


n
2
3
4
199:33333  69:72601n 22:64916n  2:97400n 0:13235n 1 < n  10


 

Quartic :
8
23:33132  11:43353n 11:88308n2  8:86816n3 3:26067n4 0  n  1
>
U n


>

SEA
<
21:74548  4:68621n 1:10907n2  0:12249n3 0:00503n4 1 < n  10


 

>
335:39238  390:43675n 412:56352n2  103:87723n3  55:65268n4 0  n  1
>
U
: Fmax


n
245:57500  57:75386n 11:85828n2  1:14426n3 0:04148n4 1 < n  10
8
>
>
<

160
Experiment (ds04)
Simulation

140

Force (kN)

120

Table 2
Change range of SEA and Fmax at different thickness range (Unit: %).

80
60
40
20

20

40

60

Dt 1.0 mm

DSEA

DSEA

DFmax

DSEA

DFmax

13.54
6.11
7.65
13.45
14.26
22.44
33.86
19.40
1.08
20.71

45.72
39.36
38.09
30.47
23.21
29.59
11.64
13.51
14.18
8.52

4.56
3.11
12.00
8.68
13.31
10.15
21.07
16.22
14.85
16.52

41.14
30.55
24.29
15.58
14.93
15.2
11.91
9.52
8.03
6.88

19.10
10.89
6.44
7.84
11.75
1.09
11.67
8.67
6.56
3.34

16.10
13.64
12.63
10.11
7.98
8.38
7.05
5.31
4.60
3.90

DSEA (SEA of FGT-SEA of UT)/SEA of UT.

DFmax (Fmax of FGT-Fmax of UT)/Fmax of UT.

80 100 120 140 160 180 200

Displacement (mm)

1.3 0

Fig. 10. Comparison of force versus displacement curves between experiments [39]
and FE simulations.

Graded exponent n (1<n<10)

4
6
8

10

12

1.0

1.2

1.2
0<n<1

1.1

Weight (kg)

Tables 4 and 5 summarize the modeling accuracy, which enable

us to compare different forms of RSM surrogate functions using the
same set of checking points. The higher the order, the better the
modeling accuracy in such cases. In addition, they will be further
compared through the results of multiobjective optimization later.

1.0
0.9
0.8
0.7

4.2.2. Multiobjective optimization results

From NSGA-II, the Pareto fronts of the FGT column and UT column are plotted in Fig. 18 based upon different metamodels. It can
be clearly seen that the Pareto fronts of the FGT columns are much
more predominant than those of the UT columns. Specically, the
SEA of FGT is better than that of the UT counterpart at the same
level of Fmax. For the UT columns, the optimal SEA results are rather
close for different orders of polynomial RSM.

Dt 0.7 mm

Fmaxb

Dt 1.4 mm

0.2
0.4
0.6
0.8
1
2
4
6
8
10

100

(16)

1<n<10

0.6
0.5

0.0

(a)

0.2
0.4
0.6
0.8
Graded exponent n (0<n<1)

(b)

Fig. 12. (a) Schematic of FGT thin-walled square column (31 layers); (b) Mass variation
with different gradient exponent n.

Fig. 11. Variation of SEA and Fmax for various values of Dt and n.

70

G. Sun et al. / International Journal of Impact Engineering 64 (2014) 62e74

18
350

n=0

UT
FGT

16

Peak force Fmax (kN)

SEA (kJ/kg)

14
n=0.2
n=0.4
n=0.6
n=0.8

12
10

n=1.0

n=2.0
n=4.0

n=8.0

n=0

UT
FGT

300

n=0.2

250

n=0.4
n=0.6
n=0.8
n=1.0

200
n=2.0
n=8.0

150

n=4.0

100
n=10.0 n=6.0

6 n=10.0 n=6.0
50

0.5

0.6

0.7

0.8 0.9 1.0

Mass (kg)

1.1

1.2

0.5

1.3

0.6

0.7

(a) SEA

0.8 0.9 1.0

Mass (kg)

1.1

1.2

1.3

(b) F max

Fig. 13. Comparisons of SEA and Fmax of FGT and UT columns at the timeframe of 15 ms.

18

350
UT
FGT
n=0.4

14

n=1.0

12

n=0.6
n=0.8

n=4.0

10 n=8.0

n=2.0

8 n=10.0 n=6.0
0.5

0.6

n=0.2

250

n=0.4
n=0.6
n=0.8
n=1.0

200
n=2.0
n=8.0 n=4.0

150
100

0.7

0.8 0.9 1.0

Mass (kg)

1.1

1.2

n=0

UT
FGT

300

n=0.2

Peak force Fmax (kN)

SEA (kJ/kg)

16

n=0

n=10.0 n=6.0

1.3

0.5

0.6

(a) SEA

0.7

0.8 0.9 1.0

Mass (kg)

1.1

1.2

1.3

(b) F max

Fig. 14. Comparisons of SEA and Fmax of FGT and UT columns at the timeframe of 20 ms.

However, the SEA of FGT may not be consistently better when

the order of polynomial functions is higher. In Fig. 18(a), for
instance, there are higher energy absorption capacities with higher
order of the polynomial on the left of intersection A (i.e. n 0.3047,
SEA 22.5568 kJ, -Fmax 136.7385 kN). This is to say that the
Pareto front of the 4th order polynomial function locates above the
others when the peak force level is higher than 136.7385 kN. It
indicates that the 4th order polynomial function leads to better
optima in this range. However, this becomes different when peak
force is lower than 136.7385 kN, which shows that the 3rd order
polynomial function may provide a better Pareto solution. In
addition, on the right of intersection B (n 0.4983,

SEA 21.7712 kJ, Fmax 121.8934 kN), the 2nd order polynomial
function may yield better optimal results. It is thus concluded that
the higher metamodeling accuracy of RSM does not necessarily
yield better nal optimal results.
In order to further discuss the results of the two single objective
optimizations, SEA and Fmax predicted by the different RS models
are listed in Tables 6 and 7, respectively. These single objective
optima given in both tables correspond to the special (or idealized)
points in the Pareto space, which lie at each end of the Pareto
curves as in Fig. 18. From both tables, the optimum exponents of the
FGT tube for the same objective function generally differ from the
different RS models. To minimize Fmax, for instance, the optimal

24

350

n=0

UT
FGT

22

14

n=8.0 n=6.0
n=10.0

Peak force Fmax (kN)

SEA (kJ/kg)

18
16

300

n=0.2
n=0.4
n=0.6
n=0.8
n=1.0

20

UT
FGT

n=2.0

n=4.0

12
10
0.4

0.6

0.8
1.0
Mass (kg)

(a) SEA

1.2

n=0

1.4

n=0.2

250

n=0.4
n=0.6
n=0.8
n=1.0

200
n=2.0

150

n=4.0
n=6.0
n=10.0

100

n=8.0

0.4

0.5

0.6

0.7

0.8 0.9 1.0

Mass (kg)

1.1

(b) F max

Fig. 15. Comparisons of SEA and Fmax of FGT and UT columns at the timeframe of 25 ms.

1.2

1.3

1.4

G. Sun et al. / International Journal of Impact Engineering 64 (2014) 62e74

Fig. 16. Deformation modes of FGT and UT columns with different graded parameters n at different time steps.

71

72

G. Sun et al. / International Journal of Impact Engineering 64 (2014) 62e74

Table 3
Parameters for NSGA-II algorithm.
Parameters

Value

Population size
Number of generations
Crossover probability
Crossover distribution index
Mutation distribution index

20
50
0.9
10.0
20.0

Define design problem: Design variables

(with gradient parameter n Eq. (6)),
objectives, constraints

Sample design space;

Create FE model with graded thickness
Conduct finite element analysis
Construct metamodels for objectives
and constraints (e.g. RSM, Eqs. (13)-(16)
Evaluate modeling error (Eqs.(10)-(12))

Perform multiobjective optimization

(Eq. (9)) (e.g. using NSGA-II)
Obtain Pareto solutions

Yes

Satisfy?

No

Fig. 17. Flowchart of multiobjective optimization for thin-walled structures with

functionally-graded thickness (FGT).

gradient yielded from the cubic RS model is 0.4686

(Fmax 121.440 kN), whereas from the quartic model it is 0.8239
(Fmax 133.173 kN).
In addition, the ideal optima of such two single objective functions in the FGT columns can be obtained from the different RS
models. For example, in the region of (0  n  1), the maximum SEA
(23.488 kJ) is obtained from the quadratic RS model, and the
minimum Fmax (107.428 kN) is yielded from the linear RS model. On
the other hand, the maximum SEA (20.796 kJ) can be generated
from the quadratic RS model in the region of (1 < n  10), and the
minimum Fmax (106.6 kN) can be obtained from the quartic RS
model. This indicates again that an optimal solution for a higher
order RS model does not necessarily yield the better solution.

5. Conclusions
This paper proposed to characterize and optimize the crashing
performance of thin-walled structures with functionally graded
thickness (FGT). The crashworthiness of the FGT structures was
examined and the effects of the different thickness gradients on
both specic absorption energy (SEA) and peak force (Fmax) levels

24

Table 4
Accuracy assessment of the RSM metamodels (0  n  1).

SAE

FG
max

SAEU

FU
max

Order

1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4

R2

0.95296
0.95335
0.95394
0.95496
0.88159
0.89639
0.91112
0.95557
0.99374
0.99478
0.99512
0.99513
0.99473
0.99929
0.99976
0.99976

23

Fitting indicators
RAAE

RMAE

0.15160
0.14830
0.14912
0.14650
0.27700
0.23092
0.27446
0.16865
0.06528
0.06732
0.06412
0.06376
0.06879
0.02005
0.01266
0.01285

0.75046
0.75292
0.74572
0.74061
0.66889
0.67663
0.53360
0.45332
0.15181
0.11677
0.11813
0.12022
0.10292
0.05267
0.02571
0.02571

22
SEA (kJ/kg)

Objective

21
20
FGT from Eq. (13)
FGT from Eq. (14)
FGT from Eq. (15)
FGT from Eq. (16)

19
18
17
-360

-320

21

SAEG

FG
max

SAEU

FU
max

Order

1
2
3
4
1
2
3
4
1
2
3
4
1
2
3
4

0.98745
0.99113
0.99373
0.99525
0.98875
0.98928
0.98943
0.99017
0.92491
0.92566
0.93214
0.93311
0.98995
0.99912
0.99939
0.99984

19

Fitting indicators
RAAE

RMAE

0.07209
0.06297
0.05478
0.05980
0.07963
0.08803
0.08366
0.07934
0.16616
0.16722
0.20066
0.16312
0.09155
0.02413
0.02299
0.01013

0.29503
0.25297
0.18699
0.19098
0.21789
0.19839
0.21692
0.24795
0.70909
0.72494
0.65886
0.66710
0.13862
0.06688
0.04259
0.02590

SEA (kJ/kg)

Objective

-240 -200
-Fmax (kN)

-160

-120

-80

FGT from Eq. (13)

FGT from Eq. (14)
FGT from Eq. (15)
FGT from Eq. (16)

20

Table 5
Accuracy assessment of the RSM metamodel (1 < n  10).

-280

UT from Eq. (13)

UT from Eq. (14)
UT from Eq. (15)
UT from Eq. (16)

18
17
16
15

UT from Eq. (13)

UT from Eq. (14)
UT from Eq. (15)
UT from Eq. (16)

14
13
12
-200

-180

-160
-140
-Fmax (kN)

-120

-100

Fig. 18. Pareto fronts of FGT thin-walled columns and corresponding UT columns.

G. Sun et al. / International Journal of Impact Engineering 64 (2014) 62e74

Table 6
Ideal optimums of the single objective functions for the FGT and UT columns
(0  n  1).
Columns

RS order

Single objective

Grade n

SEA (kJ/kg)

Fmax (kN)

UT

Linear

Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal

2.25e-08
1.0000
1.92e-10
0.8826
6.34e-09
1.0000
5.19e-11
1.0000
7.24e-10
1.0000
7.51e-10
0.6341
8.23e-09
0.4686
0.0924
0.8239

22.580
17.591
23.223
18.376
23.308
18.150
23.331
18.173
23.297
20.487
23.488
21.411
23.264
21.889
23.339
21.064

304.143
182.233
328.047
203.937
335.793
198.390
335.392
197.989
222.372
107.428
286.808
114.253
316.184
121.440
222.855
133.173

Quadratic
Cubic
Quartic
FGT

Linear
Quadratic
Cubic
Quartic

Max SEA
Min Fmax
Max SEA
Min Fmax
Max SEA
Min Fmax
Max SEA
Min Fmax
Max SEA
Min Fmax
Max SEA
Min Fmax
Max SEA
Min Fmax
Max SEA
Min Fmax

Table 7
Ideal optimums of the two single objective functions for the FGT and UT columns
(1 < n  10).
Columns

RS order

Single objective

Grade n

SEA (kJ/kg)

Fmax (kN)

UT

Linear

Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal

1.0000
10.0000
1.0000
8.0568
1.0000
10.0000
1.0000
9.4953
1.0000
10.0000
1.0000
10.0000
1.7073
10.0000
1.0000
8.9055

16.285
12.441
17.536
13.092
17.834
13.394
18.051
13.277
20.581
15.496
20.796
15.711
20.316
16.329
20.538
15.461

174.657
112.189
190.947
123.352
196.785
122.643
198.577
123.946
138.381
108.557
142.125
112.300
137.889
110.728
149.415
106.600

Quadratic
Cubic
Quartic
FGT

Linear
Quadratic
Cubic
Quartic

Max SEA
Min Fmax
Max SEA
Min Fmax
Max SEA
Min Fmax
Max SEA
Min Fmax
Max SEA
Min Fmax
Max SEA
Min Fmax
Max SEA
Min Fmax
Max SEA
Min Fmax

were evaluated. As for such FGT structures, the gradient parameter

in terms of power law exponent n has signicant effect on crashworthiness analysis and design. It was found that SEA of FGT column is superior to that of the UT counterpart, when the crashing
time increased. In addition, Fmax of the FGT column is always
smaller than that of the corresponding UT column. To formulate
sophisticated crashworthiness objective functions, the response
surface method (RSM) was employed after validating the modeling
accuracy. In order to maximize the SEA and minimize the Fmax, the
multiobjective optimizations were formulated and the Nondominated Sorting Genetic Algorithm (NSGA-II) was adopted. The
optimal results demonstrated that in general, the FGT column does
provide better crashing performance than the UT column. Interestingly, a higher metamodeling accuracy of the RSM does not
necessarily imply that the nal optimal results would be the better.
It is concluded that FGT structure is of considerable implication and
benets with signicant potential in crashworthiness applications.
It must be noted that in this study the impact follows the axial
direction of thickness gradient. When oblique loads are involved,
the results would be different and a new design optimization for
multiple load cases will be needed, which is however beyond the
scope of this paper. In addition, Nagel and Thambiratnams [53] and
others works quantied the sensitivity of wall thickness to tapered
column. From functional point of view, the tapered column might
have some similar behaviors to the FGT columns. However, spatial

73

and/or weight constraints sometimes may not allow to use large

tapered angle in practice.
Acknowledgments
This work was supported from National 973 Project of China
(2010CB328005), The National Natural Science Foundation of China
(61232014, 11202072), The Doctoral Fund of Ministry of Education
of China (20120161120005), The Hunan Provincial Science Foundation of China (13JJ4036), and The Open Fund of Traction Power
State Key Laboratory of Southwest Jiaotong University (TPL1206).
The authors would like to thank the Graduate Student Innovation
Project of Hunan province, China (521298760).
References
[1] Wierzbicki T, Abramowicz W. The mechanics of deep plastic collapse of thinwalled structures. Struct Fail 1989;9:281e329.
[2] Abramowicz W. Thin-walled structures as impact energy absorbers. ThinWalled Struct 2003;41:91e107.
[3] Hou SJ, Li Q, Long SY, Yang XJ, Li W. Multiobjective optimization of multi-cell
sections for the crashworthiness design. Int J Impact Eng 2008;35:1355e67.
[4] Seitzberger M, Rammerstorfer FG, Gradinger R, Degischer HP, Blaimschein M,
Walch C. Experimental studies on the quasi-static axial crushing of steel
columns lled with aluminium foam. Int J Sol Struct 2000;37:4125e47.
[5] Mirzaei M, Shakeri M, Sadighi M, Akbarshahi H. Experimental and analytical
assessment of axial crushing of circular hybrid tubes under quasi-static load.
Compos Struct 2012;94:1959e66.
[6] Al Galib D, Limam A. Experimental and numerical investigation of static and
dynamic axial crushing of circular aluminum tubes. Thin-Walled Struct
2004;42:1103e37.
[7] White MD, Jones N. Experimental quasi-static axial crushing of top-hat and
double-hat thin-walled sections. Int J Mech Sci 1999;41:179e208.
[8] Aktay L, Toksoy AK, Gden M. Quasi-static axial crushing of extruded polystyrene foam-lled thin-walled aluminum tubes: experimental and numerical
analysis. Mater Des 2006;27:556e65.
[9] Abramowicz W, Wierzbicki T. Axial crushing of multicorner sheet metal columns. J Appl Mech 1989;56:113.
[10] Wierzbicki T, Abramowicz W. On the crushing mechanics of thin-walled
structures. J Appl Mech 1983;50:727e34.
[11] Ghamarian A, Tahaye Abadi M. Axial crushing analysis of end-capped circular
tubes. Thin-Walled Struct 2011;49:743e52.
[12] Zhang X, Zhang H. Experimental and numerical investigation on crush resistance of polygonal columns and angle elements. Thin-Walled Struct 2012;57:
25e36.
[13] Song J, Chen Y, Lu GX. Axial crushing of thin-walled structures with origami
patterns. Thin-Walled Struct 2012;54:65e71.
[14] Zhao H, Abdennadher S. On the strength enhancement under impact loading
of square tubes made from rate insensitive metals. Int J Sol Struct 2004;41:
6677e97.
[15] DiPaolo BP, Tom JG. A study on an axial crush conguration response of thinwall, steel box components: the quasi-static experiments. Int J Sol Struct
2006;43:7752e75.
[16] Gu GY, Xia Y, Zhou Q. On the fracture possibility of thin-walled tubes under
axial crushing. Thin-Walled Struct 2012;55:85e95.
[17] Reddy TY, Al-Hassani STS. Axial crushing of wood-lled square metal tubes.
Int J Mech Sci 1993;35:231e46.
[18] Sun GY, Li GY, Stone M, Li Q. A two-stage multi-delity optimization
procedure for honeycomb-type cellular materials. Comput Mater Sci 2010;49:
500e11.
[19] Sun GY, Li GY, Zhou SW, Li HZ, Hou SJ, Li Q. Crashworthiness design of vehicle
by using multiobjective robust optimization. Struct Multidiscipl Optim
2011;44:99e110.
[20] Tang ZL, Liu ST, Zhang ZH. Analysis of energy absorption characteristics of
cylindrical multi-cell columns. Thin-Walled Struct 2013;62:75e84.
[21] Naja A, Rais-Rohani M. Sequential coupled process-performance simulation
and multi-objective optimization of thin-walled tubes. Mater Des 2012;41:
89e98.
[22] Acar E, Guler MA, Gerceker B, Cerit ME, Bayram B. Multi-objective crashworthiness optimization of tapered thin-walled tubes with axisymmetric
indentations. Thin-Walled Struct 2011;49:94e105.
[23] Alghamdi AAA. Collapsible impact energy absorbers: an overview. ThinWalled Struct 2001;39:189e213.
[24] Davies RW, Grant GJ, Khaleel MA, Smith MT, Eddie Oliver H. Forming-limit
diagrams of aluminum tailor-welded blank weld material. Metal Mater Trans
A 2001;32:275e83.
[25] Davies RW, Vetrano JS, Smith MT, Pitman SG. Mechanical properties of
aluminum tailor welded blanks at superplastic temperatures. J Mater Process
Technol 2002;128:38e47.

74

G. Sun et al. / International Journal of Impact Engineering 64 (2014) 62e74

[26] Zadpoor AA, Sinke J, Benedictus R. Mechanics of tailor welded blanks: an

overview. Key Eng Mater 2007;344:373.
[27] Yang RJ, Fu Y, Li G. Application of tailor rolled blank in vehicle front end for
frontal impact. SAE Tech Paper 2007-01-0675.
[28] Hariharan K, Kalaivani K, Balachandran G. Foil optimization in tailor welded blank
of an automotive oor component. Mater Manuf Process 2012;27:936e42.
[29] Pan F, Zhu P, Zhang Y. Metamodel-based lightweight design of B-pillar with
TWB structure via support vector regression. Comput Struct 2010;88:36e44.
[30] Pan F, Zhu P. Lightweight design of vehicle front-end structure: contributions
of multiple surrogates. Int J Vehicle Des 2011;57:124e47.
[31] Shin JK, Lee KH, Song SI, Park GJ. Automotive door design with the ULSAB
concept using structural optimization. Struct Multidiscipl Optim 2002;23:
320e7.
[32] Xu FX, Sun GY, Li GY, Li Q. Crashworthiness design of multi-components
tailor-welded blank (TWB) structures. Struct Multidiscipl Optim 2013;48:
653e67.
[33] Zhang HW, Liu LZ, Hu P, Liu XH. Springback characteristics in U-channel
forming of tailor rolled blank. Acta Metallurgica Sinica-English Lett 2012;25:
207e13.
[34] Meyer A, Wietbrock B, Hirt G. Increasing of the drawing depth using tailor
rolled blanks-numerical and experimental analysis. Int J Machine Tools Manuf
2008;48:522e31.
[35] Urban M, Krahn M, Hirt G, Kopp R. Numerical research and optimization of
high pressure sheet metal forming of tailor rolled blanks. J Mater Process
Technol 2006;177:360e3.
[36] Liao XT, Li Q, Yang XJ, Li W, Zhang WG. A two-stage multiobjective optimization of vehicle crashworthiness under frontal impact. Int J Crashworthiness
2008;13:279e88.
[37] Liao XT, Li Q, Zhang WG, Yang XJ. Multiobjective optimization for crash safety
design of vehicle using stepwise regression model. Struct Multidiscipl Optim
2008;35:561e9.
[38] Zhang Y, Sun GY, Li GY, Rungsiyakull C, Li Q. Identication of material parameters for high strength steel under impact loading. Adv Sci Lett 2011;4:
708e14.

[39] Tarigopula V, Langseth M, Hopperstad OS, Clausen AH. Axial crushing of thinwalled high-strength steel sections. Int J Impact Eng 2006;32:847e82.
[40] Paik JK, Kim BJ, Park DK, Jang BS. On quasi-static crushing of thin-walled steel
structures in cold temperature: experimental and numerical studies. Int J
Impact Eng 2011;38:13e28.
[41] Zarei HR, Krger M. Crashworthiness optimization of empty and lled
aluminum crash boxes. Int J Crashworthiness 2007;12:255e64.
[42] Zhang X, Zhang H. Optimal design of functionally graded foam material under
impact loading. Int J Mech Sci 2013;68:199e211.
[43] Kazanc Z, Bathe KJ. Crushing and crashing of tubes with implicit time integration. Int J Impact Eng 2012;42:80e8.
[44] Abramowicz W, Jones N. Dynamic progressive buckling of circular and square
tubes. Int J Impact Eng 1986;4:243e70.
[45] Sun GY, Li GY, Hou SJ, Zhou SW, Li W, Li Q. Crashworthiness design for
functionally graded foam-lled thin-walled structures. Mater Sci Eng A
2010;527:1911e9.
[46] Hou SJ, Han X, Sun GY, Long SY, Li W, Yang XJ, et al. Multiobjective optimization for tapered circular tubes. Thin-Walled Struct 2011;49:855e63.
[47] Kurtaran H, Eskandarian A, Marzougui D, Bedewi NE. Crashworthiness design
optimization using successive response surface approximations. Comput
Mech 2002;29:409e21.
[48] Hamad H, Al-Smadi A. Space partitioning in engineering design via metamodel acceptance score distribution. Eng Comput 2007;23:175e85.
[49] Sun GY, Li GY, Gong ZH, He GQ, Li Q. Radial basis functional model for multiobjective sheet metal forming optimization. Eng Optim 2011;43:1351e66.
[50] Jin R, Chen W, Simpson TW. Comparative studies of metamodelling techniques under multiple modelling criteria. Struct Multidiscipl Optim 2001;23:
1e13.
[51] Franulovic M, Basan R, Prebil I. Genetic algorithm in material model parameters identication for low-cycle fatigue. Comput Mater Sci 2009;45:505e10.
[52] Murugan P, Kannan S, Baskar S. NSGA-II algorithm for multi-objective generation expansion planning problem. Electric Power Syst Res 2009;79:622e8.
[53] Nagel GM, Thambiratnam DP. A numerical study on the impact response and
energy absorption of tapered thin walled tubes. Int J Mech Sci 2004;46:201e16.