Engineering Structures 169 (2018) 201–215

Contents lists available at ScienceDirect

Engineering Structures
journal homepage: www.elsevier.com/locate/engstruct

Crashworthiness optimization of automotive parts with tailor rolled blank T

a a a b a,c,⁎
Guangyong Sun , Jun Tian , Tangying Liu , Xiaolei Yan , Xiaodong Huang
a
State Key Laboratory of Advanced Design and Manufacture for Vehicle Body, Hunan University, Changsha 410082, China
b
School of Mechanical and Automobile Engineering, Fujian University of Technology, Fuzhou 350108, China
c
Faculty of Science, Engineering and Technology, Swinburne University of Technology, Hawthorn, VIC 3122, Australia

A R T I C LE I N FO A B S T R A C T

Keywords: Tailor rolled blank (TRB) as an advanced manufacturing process developed recently has broad application
Crashworthiness prospects in automotive and aerospace industries for its significant advantages in better load-carrying ability and
Tailor rolled blank (TRB) lighter mass. TRB is a typical custom product, whose performance is closely related to its thickness variation,
Optimization thus how to obtain the optimal thickness variation of a TRB part becomes a challenging task, especially for
Nonlinear structures
crashworthiness design. To address this issue, an optimization method based on the variation of wall thickness is
presented to maximize the energy absorption capacity of TRB parts. Firstly, finite element (FE) models of two
typical automotive parts, TRB top-hat column and TRB bumper beam, are established and validated through
crushing experiments. Then a heuristic optimization method is proposed based on the general assumption that
more mass enables to absorb more energy. The elemental energy density is used to optimize structural thickness
distribution of geometrical and material nonlinear structures. Numerical results demonstrate the capability and
effectiveness of the proposed optimization method for achieving the best thickness layout of automotive TRB
parts for crashworthiness.

1. Introduction
Since thin walled structures have the advantages of light weight,
low price, high strength and stiffness, high reliability, excellent loading-
carrying efficiency and energy absorption capacity, they have been
widely used as energy absorbers in crashworthiness applications such as
automobile, train and aeronautical industries to protect passenger from
severe injuries or fatalities [1]. To better understand the energy ab-
sorption mechanism and improve the design quality of energy absor-
bers, exhaustive studies have been conducted by using analytical, ex-
perimental and numerical approaches for thin-walled structures with
various sections [2–5], different cell numbers [6–10], different tapered
angles [11,12] under single/multiple loading cases [13,14]. Generally
speaking, all crashworthiness structures are expected to absorb max-
imum kinetic energy with minimum mass. In order to achieve this goal,
the optimization technique is the best choice. In this regard, Bi et al.
[15] adopted the response surface methodology to optimize the
crashworthiness of foam-filled thin-walled structures. Liu [16] opti-
mized the crashworthiness of straight and curved octagonal section
columns. In which, maximizing the specific energy absorption (SEA) is
set as the optimization objective and the side length of cross-sections
and wall thickness are selected as design variables, and maximum
crushing force is set as the design constraint. Zhang et al. [17] improved
the bending resistance and crashworthiness of the beam through opti-
mizing the shape of the arc-like rib. All these thin walled structures
mentioned above are based on the uniform material and/or the uniform
wall thickness. In fact, energy absorbers often experience very complex
loading, which implies that different regions of the structures should
have different roles to maximize usage of materials [18]. On the other
hand, since the design optimization methodologies are widely adopted,
the capacity of such conventional materials and structures could have
been pushed to its limits unless some new manufacturing process or
configurations can be introduced.
Under such circumstances, some advanced manufacturing pro-
cesses, e.g. tailor welded blank (TWB) and tailor rolled blank (TRB),
have been developed and widely applied in automotive industry.
Compared with TWB, TRB varies the blank thickness continuously
through adjusting the roll gap (see Fig. 1), which leads to better
formability and greater weight reduction [19]. The advantages of TRB
are as follows [20]: (1) production cost of TRB does not depend on the
number of thickness transitions; (2) any thickness transition can be
chosen within the process limits; (3) there are no stress peaks across the
transition due to the smooth thickness variation; (4) TRBs compared
with TWBs have better forming characteristics because of the elimina-
tion of welding seams and corresponding heat-affected zones in TWBs.
Due to the advantages of TRB, some studies have been conducted

⁎
Corresponding author at: Faculty of Science, Engineering and Technology, Swinburne University of Technology, Hawthorn, VIC 3122, Australia.
E-mail address: xhuang@swin.edu.au (X. Huang).

https://doi.org/10.1016/j.engstruct.2018.05.050
Received 21 March 2018; Received in revised form 13 May 2018; Accepted 15 May 2018
Available online 19 May 2018
0141-0296/ © 2018 Elsevier Ltd. All rights reserved.
G. Sun et al.
Fig. 1. The schematic of rolling process for TRB [20].
[21–24]. For example, Yang et al. [25] pointed out that the sheet with
varying thickness could be more promising structure in the area of
crashworthiness. Sun et al. [1] developed TRB thin-walled tubes with
varying wall thickness along the longitudinal direction in certain gra-
dient, which showed that the TRB tube is superior to its uniform
thickness counterparts in overall crashing behaviors. Besides of variable
thickness in axial direction, Zhang et al. [26] introduced variable
thickness in cross-section of square tubes, which can increase up to
30–35% energy absorption than the traditional uniform thickness
square tube under axial loading. Lately, Sun et al. [27] first investigated
the crashworthiness of TRB tubes under dynamic bending load. The
aforementioned investigations have shown that the TRB structures have
excellent crashworthiness characteristics.
The performance of TRB parts are closely related to the thickness
variation, thus it is critical to obtain the optimal thickness distribution
under crashworthiness [23,28,29]. As an effective alternative as the
design of conventional thin-walled structures, the engineering optimi-
zation methodology is inevitably used to design the TRB parts. In this
regard, Chuang et al. [30] adopted a multidisciplinary design optimi-
zation methodology to obtain the optimal thickness profiles of under-
body parts. Duan et al. [18] combined the support vector machine
(SVR) surrogate models with artificial bee colony algorithm to obtain
the optimal thickness distribution of front longitudinal beam inner.
Although the optimization methodology promotes the development of
TRB design to some extent, the design optimization mentioned above
mainly focused on few parameters used to adjust the thickness varia-
tion, which greatly limited the performance of TRB parts. In other
words, the material utilization is not exhaustively exploited, thus there
is an urgent need to develop a novel design method to fine adjust
thickness variation of TRB structures.
Topology optimization, as an efficient design approach of struc-
tures, can handle nonlinear large-scale combinatorial optimization
models with multi-decision-making variables. In order to improve the
performance of thin-walled structures, the design obtained from to-
pology optimization can be further modified by the parameter optimi-
zation. In fact, much research has been done to improve the crash-
worthiness of thin-walled structures by topology optimization [31–34].
Fox example, Huang et al. [35] investigated topology optimization for
the crashworthiness of continuum structures using the bi-directional
evolutionary structural optimization (BESO) approach. Ortmann and
Schumacher [36] developed a graph and heuristic based topology op-
timization technique for designing the cross-section profile of crashing
structures. The modified hybrid cellular automaton (HCA) methods
202
Engineering Structures 169 (2018) 201–215
were also used for topology optimization of crashworthy structures
[37–39]. Sun et al. [40–42] proposed the structural topology optimi-
zation method based on intelligent algorithm for the crashworthiness
design of multi-cell tubes. This paper aims to establish an optimization
approach for large deformation of energy-absorbing TRB structures
where only allow to change the thickness of structures.
It is well known that B-pillar and bumper beam are the most sig-
nificant deformable parts under vehicle side and frontal impact, re-
spectively. Their crashworthiness and deformation modes can greatly
influence the vehicle safety. In addition, the B-pillar and bumper beam
are mainly subjected to some local lateral loads, which lead to that the
material utilization of different regions has significant difference. Thus
this kind of parts should be very suitable to be designed with TRB. To
the authors’ best knowledge, there have been very limited reports
available on the crashworthiness design of B-pillar and bumper beam
with TRB. Therefore, the paper aims to present a novel optimization
method to design the simplified B-pillar (top-hat column) and bumper
beam with TRB under the bending collapse loading conditions.
Following this introduction, an optimization model for energy absorp-
tion is first developed in Section 2. Based on the proposed optimization
method, the two TRB automotive parts are analyzed and discussed in
Section 3. Finally, some conclusions are drawn in Section 4.
2. Design problem and optimization method
2.1. Design problem statement
The design is to improve the structural crashworthiness and energy
absorption characteristics through gradually optimizing the thickness
distribution in the design domain. When designing the energy absorp-
tion structures, certain constraints should be considered, such as force
limitation and deformation limitation. Typically, a maximum allowable
crushing distance is required to retain sufficient space for survival in
automotive collision. To simulate the crush behavior of a structure,
nonlinear finite element analysis is conducted by gradually increasing
the displacements of impact points from 0 to the maximum allowable
crushing distance U ∗ as shown in Fig. 2.
To obtain such an efficient energy absorption design that dissipates
as much energy as possible during a maximum allowable crushing
distance, the natural choice of the optimization objective is to maximize
the total external work shown as the area enclosed by the curve in
Fig. 2, which is equal to the total internal energy under quasi static
conditions. Considering the mass constraint and displacement con-
straint, the optimization problem can be formulated using the elements
thickness as the design variables,
Fi
Crushing force
Fi-1
Ui-1 Ui U*
Crushing distance
Fig. 2. Schematic diagram for the force-displacement relationship by gradually
increasing the crushing distance.
G. Sun et al. Engineering Structures 169 (2018) 201–215

Fig. 3. (a) Geometry description of the TRB top-hat column; (b) The true stress-true plastic strain curve of steel HSLA.

Fig. 4. (a) Experimental set-up for the specimens; (b) Experimental validation via three point bending tests.

⎧
N M ∗ represent the mass of the i-th element and the mass constraint of the
⎪ Maximize f (x ) = ∑ Ei TRB structure. The continuous design variable x i is the thickness of the
⎪ i=1
i-th element, which can continuously change from the minimum
⎪ Subject to Umax = U ∗
thickness x min to the maximum thickness x max . Since the design vari-
⎨ N
ables are based on elements, the optimization problem in Eq. (1) is
⎪ ∑ mi−M ∗ ⩽ 0
⎪ i=1
largely similar to a topology optimization problem based on finite
⎪ element analysis (FEA).
⎩ x i ∈ [x min ,x max ] (i = 1,2…,N ) (1)

where Ei is the i-th element internal energy, Umax and U ∗ are the max-
imum controlled displacement and its constraint, respectively. mi and

203
G. Sun et al. Engineering Structures 169 (2018) 201–215

Fig. 5. Evolution histories of the absorbed energy and mass fraction for optimized top-hat columns with thickness varying along (a) circumferential direction (CTC)
and (b) axial direction (CTA).

Fig. 6. Force-Displacement curves, deformation and stress field of the initial

and optimized structure (UT = 1.8 mm).

2.2. Optimization method and variables update rule

For structures involving large deformation or nonlinear problems,

the traditional gradient-based optimization methods are hardly adopted
directly since obtaining the gradient information of the objective and
constraint functions is a challenging task. Even successful, a large
number of local optimum due to geometrical and material non-
linearities will cause the optimal design highly dependent on initial
guess. Meanwhile, computing the sensitivity information needs a large
amount of computation resource. Therefore, an alternative optimiza-
tion method is proposed based on the relationship between the energy
absorbing in an element when loaded and the elemental mass.
Obviously, elements with large ratios between energy and mass in-
dicate that materials of those elements have been fully exploited for
absorbing external kinetic energy. So it can be assumed that the more
the energy absorbed by an element, the greater the mass needed to
assign.
Here the element-based energy information is used to optimize
structural thickness distribution, which is similar to the bi-directional Fig. 7. Thickness variation of optimized top-hat columns along (a) circumfer-
evolutionary structural optimization (BESO) method [43–46]. When a ential direction (CTC) and (b) axial direction (CTA).
structure is discretized with finite elements, the internal energy level in
each element of the structure is determined. The internal energy can be of the factors influencing elemental internal energy. Consequently the
used as an evaluation indicator of the importance of the element con- elemental internal energy can be taken as the index for the design of
tribution to the total absorbed energy. And the element thickness is one element thickness. Since the thicknesses of elements are different, the

204
G. Sun et al. Engineering Structures 169 (2018) 201–215

Table 1 effect of the elemental mass should be taken into account when eval-
Results comparison of top-hat columns made of TRB and UT. uating its energy absorption. In such a case, the indicator can be re-
Cases Mass Mass (kg) Thickness EA (J) IED (J/kg) Energy placed with the elemental internal energy density (IED).
fraction (mm) increase rate
Ei Ei
compared to αi = = (i = 1,2,...,N )
UT (%) Mi ρi si x i (2)

UT 0.60 2.16 1.35 655.30 303.38 0 where ρi is the density of the i-th element, si is the area of the middle
CTA 0.65/2.25 1128.12 522.28 72.20 surface in the i-th element, αi is the IED of the i-th element.
CTC 0.65/2.25 1030.45 477.06 57.30 Like the idea of fully stressed design, ideally the IED should be
UT 0.70 2.51 1.57 879.95 350.58 0 evenly distributed among the various finite elements of the design do-
CTA 0.65/2.25 1452.33 578.62 65.00 main. This concept leads to a criterion based on the local IED level,
CTC 0.65/2.25 1410.92 562.12 60.20
which is determined by comparing the IED of the i-th element αi with
UT 0.80 2.89 1.80 1142.14 395.20 0 the threshold α c . If the element has smaller IED than the control value,
CTA 0.65/2.25 1720.26 595.25 50.60
i.e.
CTC 0.65/2.25 1633.95 565.38 43.10

UT 0.90 3.22 2.00 1393.83 432.87 0

αi < α c (3)
CTA 0.65/2.25 1713.23 532.06 23.00
CTC 0.65/2.25 1703.30 528.98 22.30
it is assumed to be inefficient and the removal of material can be
conveniently undertaken by reducing the thickness of the element. On
the contrary, if the element has greater IED than the threshold value,
i.e.
αi > α c (4)

it is assumed to be efficient and then the thickness of the element is

increased.
Based on the above concept, the thickness of element i for the new
design is determined according to Eq. (5).
Fig. 8. Schematic of the three-point bending deformation mode. k+1
⎧ xi = x ik + Δx ik
αik − αck
⎨ Δx ik = k k
Δx 0
αmax − α min (5)
⎩

(a) Hat-shaped structure

(b) Bottom sheet

Fig. 9. Stress contours for the CTC, CTA and UT top-hat columns under the three-point bending conditions.

205
G. Sun et al. Engineering Structures 169 (2018) 201–215

(a) Hat-shaped structure

(b) Bottom sheet

Fig. 10. Energy distribution of the CTC, CTA and UT top-hat columns under the three-point bending conditions.

Fig. 11. Thickness distribution of (a) CTA hat-shaped structure from axial direction; (b) CTC hat-shaped structure from circumferential direction.

k
where k is the iterative number, x ik + 1 and x ik are the thickness of ele- k
positive number. α max and α min are the maximum and minimum ele-
ment i in finite element analysis during the k and k + 1 iterations, re- mental IED, respectively. And α c is the threshold that can be determined
spectively. Δx ik is the variation of the thickness for element i in k-th by the following bi-sectioning algorithm:
iteration. Δx 0 is the thickness change-limit per iteration, it is a small

206
G. Sun et al. Engineering Structures 169 (2018) 201–215

Fig. 12. Thickness distribution of (a) CTA bottom sheet from axial direction; (b) CTC bottom sheet from circumferential direction.

Fig. 13. Geometry description of the bumper beam.

Fig. 14. True stress-plastic strain curves under different strain rates of (a) B410LA; (b) B260LYD [47].

1. Initially, let αlower = min [αi],αupper = max [αi];
2. Estimate α c = (αlower + αupper )/2 ;
3. Attempt to update design variable x i and calculate the corre-
sponding mass of the design, ∑ mi . If ∑ mi < M ∗, assign αupper = α c ,
otherwise assign αlower = α c ;
4. Repeat 2–3 until αupper −αlower < 10−5 which means the mass con-
straint M ∗ is properly satisfied. Meanwhile, the final value of the
threshold α c is determined accordingly in Step 2 and then the
elemental thicknesses are finally updated according to Eq. (5).
In general, the actual crushing distance may not always reach the
maximum design displacement U ∗. To provide a robust design that is
still viable for any crushing distance within the allowable limit, the
evaluating indicator of the IED should be extended to multiple crushing
distance cases as follows:

207
G. Sun et al. Engineering Structures 169 (2018) 201–215

Fig. 15. (a) Bending test of steel bumper beam [47]; (b) Bending simulation of steel bumper beam.

| ∑ mi−M ∗|
⩽τ
∑ mi (8)
where τ is the tolerance parameter which is a specified small value.
The cycle of finite element analysis and element thickness decrease
and increase continue until the above two convergence criterions are
both satisfied.

2.4. Smoothing the iteration process

To avoid large oscillations of the algorithm and make the objective

function converge steadily, here a stabilization scheme is employed by
updating the IED of i-th element at the current iteration k, αik , by a
weighted sum of the three previous iterations:
1 k 1 1
αik = αi + αik − 1 + αik − 2
2 3 6 (9)

Fig. 16. Experimental validation via three point bending tests [47]. 2.5. Manufacturing possibility

D Conventional applications of thickness optimization optimizing the

αi = ∑ αi j thickness distribution in the design domain to obtain a concept design
j=1 (6)
which is very difficult to manufacture. Considering the manufactur-
where D is the number of crushing distance cases. The elemental IED for ability of shell based structures, a uniform thickness constraint is in-
each case will be calculated as soon as the crushing distance arrives at troduced to make the designed structure have constant thickness dis-
the corresponding displacement design point. Therefore, the overall IED tribution along a desired direction. Therefore elements in each list
calculated through Eq. (6) implies that the thickness change of an along the direction perpendicular to the designed direction are as a
element at any crushing distance will have effects on the objective design variable, which greatly reduces the number of design variables.
function. As a result, the final structural has high efficiency for ab- In this paper, consistent thickness is constrained along the axial or
sorbing energy for all design points. circumferential direction by making elemental IED in each column or
row been in the same level.
P
2.3. Convergence criterion ∑ Ei
i=1
αil = P
(i = 1,2,...,P )
Typically, the convergence criterion is defined in terms of the
change of the objective function as the following: ∑ mi
i=1 (10)
g g
| ∑ fk − e + 1 − ∑ fk − g − e + 1 | where αil is the IED of i-th element in the l-th column or row, P is the
i=1 i=1 total number of element connected to l-th column or row.
error = g ⩽ε
∑ fk − e + 1
2.6. Optimization process
i=1 (7)

where g is an integer number and ε is the convergence tolerance
parameter. Normally, g is selected to be 5 which implies that the change
in the objective function over the last 10 iterations is small enough.
Further details on the equation can be found in BESO convergence
criterion [35].
Besides, to ensure the final solution satisfies the constraint in the
optimization problem a second convergence criterion should be in-
troduced. The mass convergence iteration can be expressed as:
The overall optimization iteration procedure is given as follows:
Step 1: Discretize the design domain based on FE mesh for the given
boundary and loading conditions and assign the initial elemental
thicknesses.
Step 2: Perform nonlinear finite element analysis using FEA soft-
ware such as ABAQUS.
Step 3: Calculate and update the elemental IED by using Eqs. (2) and
(10), respectively.

208
G. Sun et al.
(a) CTC MF=0.9
(c) CTC MF=0.8
Fig. 17. Evolution histories of the absorbed energy and mass fraction.
Step 4: Renew the thickness values of elements according to Eq. (5).
Step 5: Construct a new design for the next iteration.
Step 6: Repeat steps 2–5 until the constraint mass (M ∗) is reached
and the convergence accuracy of the objective function is satisfied.
3. Numerical examples and discussion
In order to illustrate the efficiency of the optimization method and
to guide crashworthiness design of the TRB structure of automotive
parts, here choose the simplified B-pillar (top-hat column) and bumper
beam under the bending collapse loading conditions to test.
3.1. TRB top-hat column
Firstly, a top-hat column made of TRB is investigated to demon-
strate the capability of the proposed crashworthiness optimization ap-
proach. Its geometrical structure and dimensions are shown in Fig. 3(a).
The thickness is 1.8 mm. A three-point bending top-hat column can be
used to approximate the serving states of the automotive B-pillar in
aside collision. Before conducting the demonstration, an accurate finite
element (FE) model should be established.
3.1.1. Finite element model and experimental validation
The top-hat column is assumed to be made of High-strength Low-
alloy Steel (HSLA) with mechanical properties of density
ρ = 7.85 × 103 kg/m3, Young’s modulus E = 210 GPa, and Poisson’s
ratio μ = 0.3. To establish the relationship between stress and strain for
209
Engineering Structures 169 (2018) 201–215
(b) CTA MF=0.9
(d) CTA MF=0.8
High-strength Low-alloy Steel (HSLA), specimen with thickness of
1.8 mm were examined by quasi-static tensile tests. The geometry of the
specimen used in static tensile tests (INSTRON-5985) and test equip-
ment are shown in Fig. 3(b). The tensile tests were performed with a
constant velocity of 2 mm/min. The true stress-true plastic strain curve
is shown in Fig. 3(b).
An elastic-plastic material model is adopted in the finite element
analysis based on software ABAQUS. The FE model of top-hat column
was verified against the quasi static 3-point bending test. The PLS-
L50B4 electrohydraulic servo system for component impact was used to
carry out the bending test. Two supports were placed 50 mm off each
end of the top-hat column, as shown in Fig. 4(a). The cylindrical im-
pactor with a diameter of 25 mm moved down over a displacement of
60 mm at the velocity of 0.001 m/s. The corresponding numerical 3-
point bending simulation was performed in the FE software ABAQUS.
Fig. 4(b) displays the force-displacement curves and final de-
formation modes that were obtained from the experimental tests and
numerical simulations. Similar local bending deformation mode ap-
peared in both the experiment and simulation tests, which was con-
centrated in the mid-section area where the impactor contacted. And it
can be seen that the force-displacement curves of simulation agree well
with the experimental results, which ensures the accuracy and relia-
bility of the FE model for the further optimization investigation.
3.1.2. Optimization and discussion
The initial thickness of the top-hat column is set to distribute uni-
formly. Fig. 5 shows the iteration histories of the objective function
G. Sun et al.
Fig. 18. Thickness distribution of (a) CTA and (b) CTC (MF = 0.8).
when mass constraint is 80%. From the results, it can be seen that the
optimized top-hat columns with thickness varying along axial direction
(CTA) and circumferential direction (CTC) both have a great improve-
ment in crashworthiness compared to the uniform thickness (UT) ones
under the condition of the same mass.
By comparing the corresponding deformations and the force-dis-
placement curves of the initial and optimized top-hat columns as shown
in Fig. 6, it can be found that the optimized results have a more uniform
stress distribution and a higher plateau force meaning a greater struc-
tural stiffness. Fig. 7 shows the corresponding thickness distributions of
the CTC and CTA top-hat columns. The results indicate that in order to
210
Engineering Structures 169 (2018) 201–215
absorb more energy thicker material is mainly distributed in deformed
regions for both of CTA and CTC columns. Conversely, thinner material
is distributed in the un-deformed or slightly deformed areas. Ad-
ditionally, the internal energy and bending resistance of the CTA top-
hat column are larger than that of the CTC top-hat column shown in
Figs. 5 and 6, which demonstrates that the CTA top-hat column can
absorb more energy than the CTC one, i.e. for the top-hat column,
material thickness varying along axial direction (CTA) is more effective
than that varying along circumferential direction (CTC).
Detail optimization results of the top-hat columns made of TRB
(CTA, CTC) are compared with the one having uniform material
G. Sun et al. Engineering Structures 169 (2018) 201–215

3.1.3. The stress and energy distribution of the CTC, CTA and UT top-hat
columns
To further explore why CTC and CTA top-hat columns result in
better crashworthiness against the UT counterpart, a more detailed
analysis on the stress and energy distribution was carried out in this
section. Fig. 8 showed the schematic of the three-point bending de-
formation mode. The specific parameters of the top-hat columns are as
follows: L represents the distance to the center line of the top-hat
column (left to the indenter is negative, and right is positive); l1 re-
presents the cross-sectional length of hat-shaped structure from the
point (a) to the point (b); l2 represents the length of bottom sheet from
the point (c) to the point (d).
Taking mass fraction MF = 0.8 as an example, the stress contours of
all the top-hat columns under the three-point bending were plotted in
the developed planes, which are displayed in Fig. 9. Because of the
symmetry of the top-hat columns, only half of the total field was dis-
played. It can be concluded from these plots that the stress were mainly
concentrated around the loading indenters, and the stress gradually
Fig. 19. Force-Displacement curves, deformation and stress field of the initial
and optimized structure. decreased along the axis from the indenter to the ends. Note that a large
stress range can be clearly observed for the CTC and CTA top-hat col-
umns. This implies that compared with the UT counterpart, the CTC and
Table 2 CTA top-hat columns engaged more material to participate in the
Automotive bumper detail comparison among TRB and UT.
bending deformation.
Cases Mass Mass (kg) Thickness EA (J) IED (J/kg) Energy The energy distribution for the CTC, CTA and UT top-hat columns
fraction (mm) increase rate under the three-point bending conditions were plotted in Fig. 10.
compared to
Fig. 10(a) shows the energy distribution of hat-shaped structures from
UT (%)
axial and circumferential direction. From axial direction, noted that the
UT 0.60 3.44 1.00, 1.20 689.56 200.45 0 energy mainly distributes around the loading position and the energy
CTA 0.60/2.00 1856.20 539.59 169.40 gradually decreased along the axis from the loading position to the
CTC 0.60/2.00 1045.69 303.98 51.70
ends. Does it increase the energy absorption of the structure by in-
UT 0.70 4.06 1.20, 1.40 903.44 222.52 0 creasing the thickness of main energy absorbing parts? Fig. 11(a) shows
CTA 0.60/2.00 1865.89 459.58 106.50
the material distribution of CTA hat-shaped structure from axial di-
CTC 0.60/2.00 1231.60 303.35 36.30
rection. It is revealed that increasing the material thickness of main
UT 0.80 4.62 1.36, 1.60 1124.78 243.46 0 energy absorbing parts does increase the absorption of energy. From
CTA 0.60/2.00 1854.55 401.42 50.60
CTC 0.60/2.00 1380.72 298.86 22.80
circumferential direction, it can be seen that the energy absorption is
mainly concentrated in four parts, and the energy absorption in the
UT 0.90 5.20 1.53, 1.80 1376.48 264.71 0
middle part is relatively small. Fig. 11(b) shows the thickness dis-
CTA 0.60/2.00 1868.59 359.34 35.80
CTC 0.60/2.00 1636.71 314.75 18.90 tribution of CTC hat-shaped structure from circumferential direction.
From Fig. 11(b), we found that the distribution of material is consistent
with the energy distribution curve. The same situation occurs at the
thickness (UT) under the same mass constraint in Table 1. It is shown bottom sheet, as shown in Fig. 12. The energy distribution of bottom
that the absorbed energy of the CTA and CTC columns increases aver- sheet from axial direction corresponds to Fig. 12(a). And the energy
agely 52.7% and 45.7% respectively compared with the UT column distribution of bottom sheet from circumferential direction corresponds
with the same mass fractions. The CTA top-hat column is better energy- to Fig. 12(b).
absorbing structure than the CTC one. Furthermore, the energy increase
rate is getting higher and higher with the decrease of the mass fraction.
3.2. Automotive TRB bumper beam
In general, through the top-hat column example, it sufficiently illus-
trates the validity and efficiency of the proposed non-linear optimiza-
In this section, crashworthiness design for the practical automotive
tion method in designing energy-absorbing structures. Moreover,
bumper beam with TRB is investigated by using the proposed optimi-
adopting the TRB technique, the structure crashworthiness can be
zation method. As shown in Fig. 13, the automotive bumper is mainly
greatly improved by rationally distributing the material or thickness.
composed of a curved bumper beam (1) and crash boxes (2). The
bumper beam is welded by a 2.0 mm outer sheet (1a) of and a 1.7 mm

Fig. 20. Schematic of the bumper beam under three-point bending load.

211
G. Sun et al.
(a) Outer sheet
(b) Inner sheet
Fig. 21. Displacement contours for the CTC, CTA and UT bumper beam under the three-point bending conditions.
inner sheet (1b). The two crash boxes are made of 1.6 mm steel sheet.
Other geometrical dimensions are also given in Fig. 13.
3.2.1. Finite element model and experimental validation
The finite element analysis is conducted in explicit finite element
code ABAQUS. The thin-walled structures of bumper beam are modeled
with the S4R shell elements. The material of the bumper beam is steel
B410LA with mechanical properties of density ρ = 7.8 × 103 kg/m3,
Young’s modulus E = 210 GPa, Poisson’s ratio μ = 0.30, initial yielding
stress σ0 = 433 MPa and the material of crash boxes is steel B260LYD
with initial yielding stress σ0 = 344 MPa. The constitutive relation is
based on an elastic–plastic material model with von Misses isotropic
plasticity and plastic hardening. Strain rate effects are considered by
defining stress-plastic strain curves of steel sheet which obtained by
tests. Intermediate values can be obtained by interpolation method. The
stress–plastic strain curves of these two materials were obtained by
tests, as shown in Fig. 14 respectively.
The FE model of bumper beam was validated against the quasi static
3-point bending test on the PLS-L50B4 electrohydraulic servo system, as
shown in Fig. 15(a). The cylinder impactor moved down in a velocity of
0.001 m/s to press the bumper beam until the cylindrical impactor
reached 100 mm displacement. According to the experimental test, the
FE model of bumper beam was set up for a 3-point bending test in
ABAQUS, as shown in Fig. 15(b). In order to reduce CPU time, the
impactor velocity was set to 1 m/s in the FE model.
Fig. 16 displays the force-displacement curves and final deformation
modes of the experimental tests and numerical simulations. The bumper
beam undergoes in a localized bending deformation mode in both the
experiment and simulation. And good consistency of the experimental
and simulation force-displacement curves validates the accuracy of the
212
Engineering Structures 169 (2018) 201–215
bumper beam FE model.
From Fig. 16, it should be noted that the bending mode involves
formation of local hinge, in which only small amount of material par-
ticipates in deformation and energy absorption, thus it is a lower energy
absorption mode. To improve the localized bending mode, it would be
beneficial to optimize the bumper beam with TRB technology, applying
in optimization, which will be discussed next section.
3.2.2. Optimization and discussion
The design domain includes an outer sheet of bumper beam with
initial thickness of 2 mm and an inner sheet with initial thickness of
1.7 mm. Fig. 17 shows the evolution histories of the objective function
and mass when the mass constraint is 90% and 80%. It is obviously the
absorbed energy has been greatly improved.
Fig. 18 shows the optimized axial and circumferential thickness
distribution of half of the symmetric CTA and CTC bumper beam when
the mass fraction is 0.8 and Fig. 19 displays its corresponding force-
displacement curves, deformation and stress field. CTC bumper beam
adds a plastic hinge, which implies the energy absorption capacity will
be increased. While the more uniform stress distribution of the CTA
bumper will results in the more efficient material utilization. And the
force-displacement curves of the CTA and CTC bumper beam are
compared with that of the UT bumper beam because of the increase of
their platform force.
Table 2 lists the optimization results of three kinds of automotive
bumper beam with TRB (CTA and CTC) and UT under the same mass
constraints. The results indicate that the absorbed energy of TRB
bumper beam increases averagely 28% compared with the UT bumper
beam with the same mass fractions. It can also be seen that the CTA
bumper beam absorbs more energy than the CTC one. In conclusions,
G. Sun et al. Engineering Structures 169 (2018) 201–215

(a) Outer sheet

(b) Inner sheet

Fig. 22. Energy distribution for the CTC, CTA and UT bumper beam under the three-point bending conditions.

Fig. 23. Thickness distribution of (a) CTA outer sheet from axial direction; (b) CTC outer sheet from circumferential direction.

213
G. Sun et al.
Fig. 24. Thickness distribution of (a) CTA inner sheet from axial direction; (b) CTC inner sheet from circumferential direction.
the proposed optimization approach is effective to handle the crash-
worthiness design of structures with TRB. Furthermore, through con-
straining the thickness distribution along a desired direction can im-
prove the design manufacturability greatly, which will promote the
practical engineering application of the optimization method.
3.2.3. The stress and energy distribution of the CTC, CTA and UT bumper
beam
To illustrate why CTC and CTA bumper beams result in better
crashworthiness against the UT counterpart even further, a more de-
tailed analysis on the stress and energy distribution was carried out in
this part. Fig. 20 showed the schematic of bumper beams under three-
point bending load. The specific parameters of the bumper beam are as
follows: L represents the distance to the center line of the bumper beam
(left to the indenter is negative, and right is positive); l1 represents the
cross-sectional length of outer sheet from the point (a) to the point (b);
l2 represents the cross-sectional length of inner sheet from the point (c)
to the point (d).
Taking mass fraction MF = 0.8 as an example, Fig. 21 shows the
stress contours in a developed plane for CTC, CTA and UT bumper
beam. As in TRB top-hat column, the stress was mainly concentrated
around the loading indenters. Note that a large stress range can be
clearly observed for the CTA bumper beam. This implies that compared
with the UT counterpart, the CTA bumper beam engaged more material
to participate in the bending deformation. Although the stress range of
CTC bumper beam is not larger than that of UT bumper beam, a large
main stress range around the loading indenter (red box markings) can
be clearly observed for the CTC bumper beam. And as noted in Fig. 19,
the deformation is mainly concentrated around the loading indenter. So
this is why CTC bumper beam result in better energy absorption ca-
pacity than UT bumper beam.
Fig. 22 shows the energy distribution of the CTC, CTA and UT
bumper beam under the three-point bending. Figs. 23 and 24 shows the
corresponding thickness distribution of CTC and CTA bumper from
circumferential and axial direction. Again, the distribution of material
is consistent with the energy distribution curve. This is yet another
example makes clear our earlier claim that increasing the material
thickness of main energy absorbing parts does increase the absorption
of energy.
4. Conclusions
A optimization approach for large deformation of energy-absorbing
TRB structures is established in this paper. Practical engineering
214
Engineering Structures 169 (2018) 201–215
problems, e.g., TRB top-hat column and automotive bumper beam are
investigated by applying this optimization method. Optimized results of
the two examples demonstrate the capabilities of the proposed opti-
mization method for the crashworthiness design of structure with TRB
and the structural energy absorption can be greatly improved by ra-
tionally distributing the material thickness when applying the TRB
technique. Based on the preceding discussion, some concluding can be
drawn as follows:
(1) From the design analyses, it can be seen that the automotive parts
with tailor rolled blank can provide better crashworthiness than
those of uniform thickness. From this perspective, adopting the TRB
technique, the structure crashworthiness can be greatly improved
by rationally distributing the material or thickness.
(2) Note that the distribution of material for the optimal structures are
consistent with the energy distribution curve. This revealed that
increasing the material thickness of main energy absorbing parts
does increase the absorption of energy.
(3) These optimization results demonstrated that the proposed struc-
tures and optimization method in this paper was fairly effective and
can provide a useful tool for generating novel energy absorption
structures. Moreover, manufacturing difficulty that is usually pro-
duced by using optimization can be overcome through constraining
the thickness distribution along desired direction.
Appendix A. Supplementary material
Supplementary data associated with this article can be found, in the
online version, at http://dx.doi.org/10.1016/j.engstruct.2018.05.050.
References
[1] Sun G, Xu F, Li G, Li Q. Crashing analysis and multiobjective optimization for thin-
walled structures with functionally graded thickness. Int J Impact Eng
2014;64:62–74.
[2] Abramowicz W. Thin-walled structures as impact energy absorbers. Thin-Walled
Struct 2003;41:91–107.
[3] Abramowicz W, Jones N. Dynamic axial crushing of circular tubes. Int J Impact Eng
1984;2:263–81.
[4] Al Galib D, Limam A. Experimental and numerical investigation of static and dy-
namic axial crushing of circular aluminum tubes. Thin-Walled Struct
2004;42:1103–37.
[5] White MD, Jones N. Experimental quasi-static axial crushing of top-hat and double-
hat thin-walled sections. Int J Mech Sci 1999;41:179–208.
[6] Alavi Nia A, Parsapour M. An investigation on the energy absorption characteristics
of multi-cell square tubes. Thin-Walled Struct 2013;68:26–34.
[7] Chen W, Wierzbicki T. Relative merits of single-cell, multi-cell and foam-filled thin-
G. Sun et al.
walled structures in energy absorption. Thin-Walled Struct 2001;39:287–306.
[8] Kim H-S. New extruded multi-cell aluminum profile for maximum crash energy
absorption and weight efficiency. Thin-Walled Struct 2002;40:311–27.
[9] Najafi A, Rais-Rohani M. Mechanics of axial plastic collapse in multi-cell, multi-
corner crush tubes. Thin-Walled Struct 2011;49:1–12.
[10] Zhang X, Cheng G. A comparative study of energy absorption characteristics of
foam-filled and multi-cell square columns. Int J Impact Eng 2007;34:1739–52.
[11] Nagel GM, Thambiratnam DP. A numerical study on the impact response and en-
ergy absorption of tapered thin-walled tubes. Int J Mech Sci 2004;46:201–16.
[12] Song X, Sun G, Li G, Gao W, Li Q. Crashworthiness optimization of foam-filled
tapered thin-walled structure using multiple surrogate models. Struct Multidiscip
Optim 2013;47:221–31.
[13] Qiu N, Gao Y, Fang J, Feng Z, Sun G, Li Q. Crashworthiness analysis and design of
multi-cell hexagonal columns under multiple loading cases. Finite Elem Anal Des
2015;104:89–101.
[14] Yang S, Qi C. Multiobjective optimization for empty and foam-filled square columns
under oblique impact loading. Int J Impact Eng 2013;54:177–91.
[15] Bi J, Fang H, Wang Q, Ren X. Modeling and optimization of foam-filled thin-walled
columns for crashworthiness designs. Finite Elem Anal Des 2010;46:698–709.
[16] Liu Y. Crashworthiness design of multi-corner thin-walled columns. Thin-Walled
Struct 2008;46:1329–37.
[17] Zhang Z, Liu S, Tang Z. Design optimization of cross-sectional configuration of rib-
reinforced thin-walled beam. Thin-Walled Struct 2009;47:868–78.
[18] Duan L, Sun G, Cui J, Chen T, Cheng A, Li G. Crashworthiness design of vehicle
structure with tailor rolled blank. Struct Multidiscip Optim 2016;53:321–38.
[19] Meyer A, Wietbrock B, Hirt G. Increasing of the drawing depth using tailor rolled
blanks—Numerical and experimental analysis. Int J Mach Tools Manuf
2008;48:522–31.
[20] Kopp R, Wiedner C, Meyer A. Flexibly rolled sheet metal and its use in sheet metal
forming. Adv Mater Res 2005;6–8:81–92.
[21] Sun G, Zhang H, Lu G, Guo J, Cui J, Li Q. An experimental and numerical study on
quasi-static and dynamic crashing behaviors for tailor rolled blank (TRB) structures.
Mater Des 2017;118:175–97.
[22] Zhang H, Sun G, Xiao Z, Li G, Li Q. Bending characteristics of top-hat structures
through tailor rolled blank (TRB) process. Thin-Walled Struct 2018;123:420–40.
[23] Sun G, Zhang H, Fang J, Li G, Li Q. Multi-objective and multi-case reliability-based
design optimization for tailor rolled blank (TRB) structures. Struct Multidiscip
Optim 2016;55:1899–916.
[24] Fang J, Gao Y, Sun G, Zheng G, Li Q. Dynamic crashing behavior of new extrudable
multi-cell tubes with a functionally graded thickness. Int J Mech Sci
2015;103:63–73.
[25] Yang Z, Peng Q, Yang J. Lightweight design of B-pillar with TRB concept con-
sidering crashworthiness. In: Proceeding ICDMA '12; 2012. pp. 510–513.
[26] Zhang X, Wen Z, Zhang H. Axial crushing and optimal design of square tubes with
graded thickness. Thin-Walled Struct 2014;84:263–74.
[27] Sun G, Tian X, Fang J, Xu F, Li G, Huang X. Dynamical bending analysis and op-
timization design for functionally graded thickness (FGT) tube. Int J Impact Eng
2015;78:128–37.
215
Engineering Structures 169 (2018) 201–215
[28] Sun G, Pang T, Zheng G, Song J, Li Q. On energy absorption of functionally graded
tubes under transverse loading. Int J Mech Sci 2016;115–116:465–80.
[29] Zheng G, Pang T, Sun G, Wu S, Li Q. Theoretical, numerical, and experimental study
on laterally variable thickness (LVT) multi-cell tubes for crashworthiness. Int J
Mech Sci 2016;118:283–97.
[30] Chuang CH, Yang RJ, Li G, Mallela K, Pothuraju P. Multidisciplinary design opti-
mization on vehicle tailor rolled blank design. Struct Multidiscip Optim
2008;35:551–60.
[31] Sun G, Liu T, Huang X, Zheng G, Li Q. Topological configuration analysis and design
for foam filled multi-cell tubes. Eng Struct 2018;155:235–50.
[32] Fang J, Sun G, Qiu N, Kim NH, Li Q. On design optimization for structural crash-
worthiness and its state of the art. Struct Multidiscip Optim 2016;55:1091–119.
[33] Guo L, Tovar A, Penninger CL, Renaud JE. Strain-based topology optimisation for
crashworthiness using hybrid cellular automata. Int J Crashworthiness
2011;16:239–52.
[34] Lee H-A, Park G-J. Nonlinear dynamic response topology optimization using the
equivalent static loads method. Comput Methods Appl Mech Eng 2015;283:956–70.
[35] Huang X, Xie YM, Lu G. Topology optimization of energy-absorbing structures. Int J
Crashworthiness 2007;12:663–75.
[36] Ortmann C, Schumacher A. Graph and heuristic based topology optimization of
crash loaded structures. Struct Multidiscip Optim 2013;47:839–54.
[37] Duddeck F, Hunkeler S, Lozano P, Wehrle E, Zeng D. Topology optimization for
crashworthiness of thin-walled structures under axial impact using hybrid cellular
automata. Struct Multidiscip Optim 2016;54:415–28.
[38] Bandi P, Schmiedeler JP, Tovar A. Design of crashworthy structures with controlled
energy absorption in the hybrid cellular automaton framework. J Mech Des
2013;135:091002.
[39] Zeng D, Duddeck F. Improved hybrid cellular automata for crashworthiness opti-
mization of thin-walled structures. Struct Multidiscip Optim 2017;56:101–15.
[40] Sun G, Liu T, Fang J, Steven GP, Li Q. Configurational optimization of multi-cell
topologies for multiple oblique loads. Struct Multidiscip Optim 2017;57:469–88.
[41] Fang J, Sun G, Qiu N, Steven GP, Li Q. Topology optimization of multicell tubes
under out-of-plane crushing using a modified artificial bee colony algorithm. J
Mech Des 2017;139:071403.
[42] Qiu N, Gao Y, Fang J, Sun G, Kim NH. Topological design of multi-cell hexagonal
tubes under axial and lateral loading cases using a modified particle swarm algo-
rithm. Appl Math Model 2018;53:567–83.
[43] Liu B, Huang X, Huang C, Sun G, Yan X, Li G. Topological design of structures under
dynamic periodic loads. Eng Struct 2017;142:128–36.
[44] Chen W, Xia L, Yang J, Huang X. Optimal microstructures of elastoplastic cellular
materials under various macroscopic strains. Mech Mater 2018;118:120–32.
[45] Meng F, Huang X, Jia B. Bi-directional evolutionary optimization for photonic band
gap structures. J Comput Phys 2015;302:393–404.
[46] Huang X, Zhou S, Sun G, Li G, Xie YM. Topology optimization for microstructures of
viscoelastic composite materials. Comput Methods Appl Mech Eng
2015;283:503–16.
[47] Xiao Z, Fang J, Sun G, Li Q. Crashworthiness design for functionally graded foam-
filled bumper beam. Adv Eng Softw 2015;85:81–95.