Automotive Innovation (2019) 2:1–13

https://doi.org/10.1007/s42154-019-00051-7

Crashworthiness Optimization Design of Thin‑Walled Tube Filled

with Re‑entrant Triangles Honeycombs
Fangwu Ma1 · Ying Zhao1 · Guowang Wang1 · Liang Wu1 · Yongfeng Pu1

Received: 30 December 2018 / Accepted: 17 January 2019 / Published online: 5 February 2019
© The Author(s) 2019

Abstract
A novel re-entrant triangles-filled tube (RTT) is proposed through decoupling structural stiffness and energy absorption.
Inner re-entrant triangles are employed to satisfy energy absorption, and outer thin wall is used to acquire high stiffness.
This paper starts from establishment of theoretical models between geometric parameters of re-entrant triangles and relative
density, equivalent elastic modulus and energy absorption characteristics, which are validated by experiments. On this basis,
the optimal geometric parameters of unit cell are sought to maximize unit volume energy absorption and minimize relative
density by adopting NSGA-II method. Subsequently, the cross-section of tube with optimal stiffness is obtained with targets
for maximizing axial stiffness and lateral stiffness by employing static topology optimization method. To verify the pro-
posed optimization method, RTT is analyzed and compared with positive Poisson’s ratio foam-filled tube (PFT), non-filled
traditionally optimized tube (NTT) and pre-optimized square tube (PST). The results show that the novel RTT can improve
stiffness and energy absorption performance simultaneously. Compared with the positive Poisson’s ratio material, re-entrant
triangles honeycomb shows superior advantages in energy absorption. In comparison with the PFT, energy absorption of the
RTT increases by 17.23%, and the peak crush force reduces by 5.04%. Therefore, the proposed decoupling design method
demonstrates superiority in satisfying various performance requirements simultaneously.

Keywords  Thin-walled tube · Re-entrant triangles honeycomb · Multi-objective optimization · Energy absorption ·
Structural stiffness

1 Introduction Rossi [9] carried out the study on the crushing behavior
among square, hexagonal and octagonal tubes. Gao et al.
With increased customer demands and tough vehicle safety [10] carried out simulations on crushing impact of thin-
regulations, increasing attentions have recently been drawn walled tubes with various cross-sections, which provided
to achieving better crashworthiness of automobile [1–3]. reference for the selection of automotive cross-section.
Moreover, environmental sustainable developments and Thin-walled tubes should be equipped with energy absorp-
energy crisis have forced engineers to design increasingly tion performance and possessed with stiffness requirements.
lighter structures [4–6]. Structures that possess with char- For example, outer thin wall of front side rail should be
acteristics and superior crashworthiness have drawn abun- equipped with high stiffness. It is, however, not always wise
dant attention for their better performance [7, 8]. Owing to to enhance crashworthiness by sacrificing stiffness when
outstanding capabilities of energy absorption, thin-walled using the aforementioned optimization methods. Therefore,
structure is regarded as the most potential energy absorber. this paper proposes the concept of decoupling to satisfy
To enhance energy absorption performance, substantial the contradictory requirements of structural stiffness and
efforts on design shape optimization have been expended. energy absorption, i.e., the outer thin wall needs high stiff-
ness, while the thinner filler is responsible for strengthen-
ing energy absorption. Substantial research on thin-walled
* Ying Zhao
[email protected] tubes with fillers has been conducted. Axial crushing analy-
sis on foam-filled thin-walled structure was carried out by
1
State Key Laboratory of Automotive Simulation and Control, Ghamarian [11] and Zarei [12]. Ahmad et al. [13] emulate
Jilin University, 5988, Renmin Ave., Changchun 130025, crushing behaviors of foam-filled conical tubes adopting
China

13
Vol.:(0123456789)
2 F. Ma et al.

numerical methods. Whereas, the above-mentioned studies continuous, parameters can be utilized to represent perfor-
only involve single performance and conventional positive mance of cellular structure. The equivalent description is
Poisson’s ratio foam fillers. Positive Poisson’s ratio fillers called as asymptotic homogenization method, put forward
usually possess unstable platform stress and higher peak by Benssousan [15] and Sanchez-Palentia [16]. Asymptotic
stress which limit the improvement in energy absorption and homogenization methods [17–19] have been regarded as
further application of impact protection. Various re-entrant effective means for multiscale modeling, while these meth-
structures have also been studied. Ma et al. [14] proposed a ods are always neglected in the aspect of mechanics during
type of re-entrant square cellular structure and illustrated the the homogenization process and widely used in the predic-
effects of cellular structural parameters on in-plane dynamic tion of equivalent performance of composites. Asymptotic
performance. Compressing this kind of structure results in homogenization method focuses on the theory where main
structural contraction, and platform stress improves, which concepts and derivation of the equations for computation of
contributes greatly to enhancing the capability for energy effective constitutive parameters of complex materials with
absorption. Herein, re-entrant triangles honeycomb acts as the unit cell. It acts as a bridge-linking microscale analysis
a filler, and the octagonal and cruciform cross-section is and macroscopic properties of cellular structures; structural
regarded as the stiffener. response is obtained from the perspective of macroscopic
To satisfy the requirements of structural stiffness and scale. If size of macroscopic structure is larger compared
energy absorption simultaneously, the novel tube utilizes with microscopic unit cell and the number of unit cell is
re-entrant triangles as filler based on the concept of decou- sufficient, more accurate results can be expected by utilizing
pling structural stiffness and energy absorption. Inner re- homogenization method.
entrant triangles are utilized to meet the demands of energy
absorption, while the octagonal and cruciform cross-section 2.1 Relationship Between Relative Density
is designed to achieve higher stiffness. and Geometric Parameters
Firstly, relationships among geometric parameters and
relative density, equivalent elastic modulus and energy Based on the homogenization method, Fig. 1 exhibits peri-
absorption characteristics are established, which are vali- odic re-entrant triangles and representative unit cell.
dated by experiments. The optimal geometric parame- To date, much pioneering work has been conducted to
ters of unit cell are then sought to maximize unit volume set up the connections among cell geometry and mechanical
energy absorption and minimize relative density utilizing properties of cellular structure. Relative density is consid-
the NSGA-II method. Subsequently, the cross-section of ered as a critical parameter to illustrate its features [20].
tube with optimal stiffness is obtained by employing static For two-dimensional (2D) unit cell, relative density 𝜌2D is
topology optimization method with targets for maximizing described in Eq. (1) and Fig. 2.
axial stiffness and lateral stiffness. To certify the optimiza-
S1
tion method, RTT, PFT, NTT and PST are analyzed and 𝜌2D = (1)
S2
compared in light of stiffness and energy absorption. The
proposed RTT demonstrates superior energy absorption and It can be seen that 𝜌2D is dominated by S1 and S2 , which
structural stiffness, and this work proposes a design method represent the areas of internal cell wall and 2D unit cell,
of decoupling structural stiffness and energy absorption and respectively.
applying re-entrant triangles on tube inside. The novel RTT Figure 3 exhibits geometric variables. Where L , M and N
presents prominent effects on energy absorption and struc- stand for the lengths of long cell beam, short cell beam and
tural stiffness. The results are encouraging in the sense that horizontal beam accordingly,H indicates distance between
it offers another potential material for engineers to address
the issue and the novel-type RTT with re-entrant triangles
can be widely promoted and achieve further applications on
new energy absorbent structures.

2 Theoretical Models of Re‑entrant Triangles

Honeycomb

In order to set up the relationship among structural param-

eters and relative density, equivalent elastic modulus and
energy absorption, theoretical models of the honeycomb
are established. On the hypothesis that cellular material is Fig. 1  Periodic re-entrant triangles and representative unit cell

13
Crashworthiness Optimization Design of Thin-Walled Tube Filled with Re-entrant Triangles… 3

where V1 stands for the volume of cellular structure, V2 is

volume of unit cell, b manifests cell thickness along Z direc-
tion, giving
( )2 ( )
b
V1 = S1 b − T L + TM (4)
2
( )
1 1
V2 = S2 N + N + L sin 𝜑 = S2 (𝛽L + L sin 𝜑) (5)
2 2
Similar to 𝜌2D and width coefficient 𝛾 , 𝛾 = Lb , relative den-
sity 𝜌3D yields
b2 𝛼
𝛾S1 − 𝛾
𝜌3D = 2
≈ 𝜌 (6)
(sin 𝜑 + 𝛽)S2 sin 𝜑 + 𝛽 2D
Fig. 2  Relative density of 2D unit cell
2.2 Mathematical Descriptions of Equivalent Elastic
Modulus and Plateau Stress

Linear elastic response of this re-entrant triangle is primar-

ily caused by the bending cell walls. Axial extension or
compression and shear deformation of the cell beams can
be ignored, and strains are small enough to avoid the gross
changes in geometry. To simplify the analysis, following
hypothesis is set that cellular angle 15◦ ≤ 𝜑 ≤ 40◦ , thick-
ness coefficient 0.05 ≤ 𝛼 ≤ 0.25 , length coefficient 𝛽 is set
from 0.1 to 0.3; the short cell beam should be possessed
with re-entrant characteristics, i.e., sin 𝜑 < K < 1 . To reduce
the complexity of optimal design, the width coefficient 𝛾
and thickness coefficient 𝛼 are kept the same: 𝛾 = 𝛼 . The
Fig. 3  Geometric description of 2D unit cell
mechanical model can be exhibited in the following Fig. 4.
According to standard rail theory [21], long cell beam lL
two horizontal beams, 𝜑 stands for the intersection angle deflects by
between long cell beam and Y -axis, while 𝜃 is the angle
of short cell wall and Y -axis, TL and TM represent cellular P sin 𝜑L3 TL2
𝛿L = − (7)
thicknesses of long beam and short beam, respectively. Most 3EIL 2EIL
importantly, 𝜃 must be greater than 𝜑 in light of re-entrant
where T stands for the applied moment tending to bend the
characteristics. Meanwhile, thickness coefficient 𝛼 and
wall, IL represents the second moment of inertia of the long
length coefficient 𝛽 are introduced, 𝛼 = LL  , 𝛼 = MM  , 𝛽 = NL  .
T T
beam lL,E denotes the elastic modulus of the material used
To simplify the expression 𝜌2D , parameter K = ML  , scaling
and P is given by
factor of length, is introduced.
Using the latter facts and substituting the series into ( )
1 1
P = 𝜎b N + N + L sin 𝜑 = 𝜎b(𝛽L + L sin 𝜑) (8)
Eq. (1), relative density 𝜌2D is finally obtained. 2 2
𝛼 + 𝛼𝛽 + 𝛼K 2 + 𝛼𝛽K Of which, 𝜎 is the applied stress along the Y direction.
𝜌2D = � √ � Hence, the deflection of long cell beam lL yields
(2)
cos 𝜑 − K 2 − sin2 𝜑 + 𝛼 + 𝛼K (sin 𝜑 + 𝛽)
P sin 𝜑L3 TL2 𝜎L(𝛽 + sin 𝜑) sin 𝜑
𝛿L = − = (9)
Similarly, the relative density of two-dimensional (3D) 3EIL 2EIL E𝛼 3
cells is given by Similarly, the deflection of short cell beam lS is given by
V
𝜌3D = 1 (3) P sin 𝜃M 3 TM 2 𝜎L(𝛽 + sin 𝜑) sin 𝜃
V2 𝛿M = − = (10)
3EIM 2EIM E𝛼 3

13
4 F. Ma et al.

Fig. 4  Axial compression
mechanical model

Strain 𝜀x and 𝜀y along X and Y axes, respectively, are then stress limit, a series of failure modes including elastic buck-
given by ling, plastic collapse and brittle fracture will occur.
According to the pioneering work [23], plastic collapse
𝛿M cos 𝜃 − 𝛿L cos 𝜑
𝜀x = (11) and elastic buckling are the two main failure modes for cel-
𝛽L + L sin 𝜑
lular structure comprising elastic–plastic materials. Elastic
buckling comes up when the loading is greater than Euler
𝛿L sin 𝜑 + 𝛿M sin 𝜃 buckling load Fcr [24]. When component force PL along the
𝜀y = (12)
H long cell beam is equal to Fcr , elastic bulking occurs. Plastic
Combined with the above-mentioned equations, the collapse comes up as bending moment reaches or beyond the
equivalent elastic modulus Eeq and equivalent Poisson’s ratio fully plastic moment. Under these conditions, the work of
𝜈eq are given by the long and short cell beams WL and WM are, respectively,
expressed as
� √ �
E𝛼 3 cos 𝜑 − K 2 − sin2 𝜑 1
𝜎 WL = 𝛾 ⋅ 𝜎(𝛽 + sin 𝜑)L2 b sin 𝜑 (16)
Eeq = = � � (13) 2
𝜀y (𝛽 + sin 𝜑) 1 + K12 sin2 𝜑
1
WM = 𝛾 ⋅ 𝜎b(𝛽L + M sin 𝜃)M sin 𝜃 (17)
Similar to 2D cells, applied stress 𝜎3D of 3D cells is then 2
𝛾 where 𝛾 implies the angle through which all joints of the cell
𝜎3D =
𝛽 + sin 𝜑
𝜎 (14) wall rotate. Besides, the work done by the plastic hinges can
be described as
Thus, the equivalent elastic modulus E3D of 3D cells is
described as 1
WP = 𝜎 bT 2 𝛾 (18)
𝛾 2 ys
E3D = E
𝛽 + sin 𝜑 eq (15)
1
WPL = 𝜎 bT 2 𝛾 (19)
Average stress of plateau region is regarded as plateau 2 ys L
stress in stress–strain curve [22]. Plateau stress acts as a key
factor to improve capabilities of energy absorption effec- 1
WPM = 𝜎 bT 2 𝛾 (20)
tively, that is, energy absorption attributes significantly to 2 ys M
plateau stress. Plateau stress also has a close relationship The critical stresses 𝜎L and 𝜎M of long beam and short
with the structural failure mechanism. Hence, it is of signifi- beam are given by the following equations. Among this, 𝜎ys
cance to further study the predominant deformation mode of means yield stress.
these re-entrant triangles. Because the cellular structure is
subjected to axial compression, the structural deformation 𝜎ys 𝛼 2
is uniform in the elastic stage. As loading stress exceeds the 𝜎L = (21)
(𝛽 + sin 𝜑) sin 𝜑

13
Crashworthiness Optimization Design of Thin-Walled Tube Filled with Re-entrant Triangles… 5

absorption W comprises three parts: the elastic region W1 ,

𝜎ys 𝛼 2 sin 𝜑 plateau region W2 and plateau stress enhancement region W3.
𝜎M = (22) In the elastic region, the following assumption is set that
(𝛽 + sin 𝜑) sin2 𝜃
the relationship between stress and strain is linear. Unit vol-
In view of structural characteristics, sin 𝜑 < sin 𝜃 . Hence, ume energy absorption in the elastic region is written as
sin2 𝜃
> 1,  𝜎L > 𝜎M.  Therefore, the critical stress 𝜎M of short
𝜎L
𝜎M
= sin 2
𝜑
� �
cell wall contributes greatly to the cellular stress limit, giving 2
2
𝜎pl,3D sin2 𝜑(𝛽 + sin 𝜑)2 1 + K12
𝜎
W1 = = � √ � (27)
2E3D 2E𝛼 4 cos 𝜑 − K 2 − sin2 𝜑
K2 2 𝛼 𝜎ys
𝜎pl = 𝜎M = (23)
(𝛽 + sin 𝜑) sin 𝜑
Total energy absorption in the plateau region comprises
The elastic bulking 𝜎eb exceeds the plastic collapse 𝜎pl , energy absorbed under constant plateau stress and energy
giving absorbed in plateau stress enhancement region. Under con-
𝜎ys stant plateau stress, energy absorption can be written as
n2 𝜋 2 𝛼 sin2 𝜃
< (24) ( )
E 6 sin 2𝜑
W2 = 𝜎pl,3D 𝜀E − 𝜀0 (28)
For metals and polymers belonging to conventional elas-
𝜎 where 𝜀0 depicts the strain at the end of the elastic region,
tic materials, order of magnitude of Eys is 10−3 and 10−2 . 𝜎
𝜀0 = Epl,3D  . 𝜀E is plateau stress enhancement strain, 𝜀E = A𝜀D ;
Through calculation, plastic collapse proves to be the domi- 3D

nant failure mode. Similar analysis on equivalent plastic col- coefficient A can be obtained by simulation analysis.
lapse stress can be conducted for 3D cells. As the concen- The following empirical formula exhibits densification
trated force resulting in complete plastic hinges of 3D cells strain 𝜀D [21]. Where,B , C and D are the coefficients, which
is equal to that of 2D cell, applied stress on 3D re-entrant can be determined by the simulation approach.
cells is acquired ( )
𝜀D = B 1 − C𝜌3D + D𝜌33D (29)
K 2 𝛼 2 𝜎ys K 2 𝛼 3 𝜎ys The volume of the cellular structure as it reaches the pla-
𝛼
𝜎pl,3D = ⋅ = teau stress enhancement region is given by
(𝛽 + sin 𝜑) sin 𝜑 𝛽 + sin 𝜑 (𝛽 + sin 𝜑)2 sin 𝜑
( )
(25) VE = L2 H 𝜀D − 𝜀E (30)
2.3 Characterization of Energy Absorption Under
The instantaneous volume of the cellular structure can
Quasi‑Static Compression
be expressed as
As cellular structures with re-entrant features are subjected ( )2 ( )
V(𝜀) = VE − L + 𝜈𝜀0 L 𝜀H − 𝜀E H (31)
to external force, energy absorption can be illustrated by
the work of external force, i.e., the area surrounding the where 𝜀E < 𝜀 < 𝜀D.
collision force–displacement curve and the displacement In plateau stress enhancement region, instantaneous rela-
axis. The collision force–displacement curve is associated tive density can be given by
with the overall size of structure, so it can be transformed
into a stress–strain curve in order to show different energy 𝜌(𝜀) 𝜀D (1 − A)
absorption characteristics for structures of various sizes. 𝜌E
= ( )2 ( ) (32)
𝜀D (1 − A) − 1 + 𝜀0 𝜈3D 𝜀 − A𝜀D
Unit volume energy absorption W is denoted by the area
under stress–strain curve and horizontal axis, given by For 3D cellular structure composed of elastic–plastic
𝜀D materials, relative density is proportional to the square of
∫0 (26)
W= 𝜎(𝜀)d𝜀 𝛼 , and equivalent plastic collapse stress is proportional to
vertical component of 𝛼 . Thus, the instantaneous stress in
where 𝜀D and 𝜎(𝜀) are the densification strain and instantane- plateau stress enhancement region is expressed as:
ous stress, respectively.
√
The stress–strain curve can be utilized to analyze energy √( )3
√
absorption characteristics and evaluate cell shape of the √ 𝜀D (1 − A)
𝜎(𝜀) = √ ( )2 ( ) 𝜎pl,3D
re-entrant cellular structure. Effective unit volume energy 𝜀D (1 − A) − 1 + 𝜀0 𝜈3D 𝜀 − A𝜀D
(33)

13
6 F. Ma et al.

Table 1  Three groups of geometric parameters

Thickness Length Cellular Scaling
coefficient 𝛼 coefficient angle 𝜑 factor of
𝛽 length K

First group 0.1 0.2 35° 0.6

Second group 0.08 0.2 35° 0.7
Third group 0.06 0.2 30° 0.7

√
√( )3
√
𝜀D
√ 𝜀D (1 − A)
∫A𝜀D
W3 = √ ( )2 ( )
𝜀D (1 − A) − 1 + 𝜀0 𝜈3D 𝜀 − A𝜀D
K 2 𝜎ys 𝛼 3
× d𝜀
(𝛽 + sin 𝜑)2 sin 𝜑 (34)

3 Certification of Theoretical Models

Fig. 6  Comparisons of theoretical and experimental stress–strain
curves
With the purpose of verifying the theoretical models
obtained above, three groups of parameters are listed in
Table 1.
Based on the selected three groups of parameters, theoret- Figure 6 shows that the experiment results of the re-
ical results are obtained. Besides, the quasi-static compres- entrant triangles honeycomb are fluctuating up and down;
sion mechanical testing on re-entrant triangles is performed. this is due to the plastic collapse deformation mode of the
The re-entrant triangles structure is made of aluminum structure layer by layer. It also shows that the results of
alloy with density 𝜌 = 2.7e−6 Kg/mm3 , Young’s modu- experimental analysis match well with theoretical results
lus E = 68.21 Gpa , Poisson’s ratio 𝜈 = 0.3 , yield strength and varying tendencies of the curves tend to be consistent.
𝜎ys = 80 Mpa and ultimate strength 𝜎us = 173 Mpa . The There exist relations between unit volume energy absorp-
cellular structure is constrained at top and bottom by steel tion Wu and stress–strain curve, and Wu can be described by
plates. The axial loading is applied on the top of the cellular the surrounded area between stress–strain curve and X -axis.
structure with the velocity of 1 mm/min. Figure 5 presents Hence, Table 2 can be obtained, and it proves that theoretical
the pre-compression and post-compression of the specimen results of Wu maintain a high agreement with that of experi-
under quasi-static loading. Then, the experimental and theo- mental results. Theoretical models derived above prove to
retical stress–strain curves are achieved and compared. be reasonable and appropriate by means of test.

Fig. 5  Pre-compression and
post-compression of the speci-
men under quasi-static loading

13
Crashworthiness Optimization Design of Thin-Walled Tube Filled with Re-entrant Triangles… 7

Table 2  Comparisons of theoretical and experimental results for unit

volume energy absorption
Theoretical Experimental Relative
results (J/m3) results (J/m3) error
(%)

First group 9347.872 10,162 8.01

Second group 13,720.62 14,318 4.17
Third group 10,470.34 10,792 2.98

4 Optimization of Unit Cell

4.1 Single Objective Optimization on Maximizing

Unit Volume Energy Absorption
Fig. 7  Optimization process of unit volume energy absorption Wu for
unit cell
With the purpose of achieving the optimal geometric param-
eter, multi-objective optimization on unit cell is executed.
The NSGA-II method is adopted with the objectives for Similarly, Fig. 8 exhibits the specific optimization process
maximum Wu and minimum 𝜌RD. of relative density for unit cell. It shows that the optimal 𝜌RD
While re-entrant triangles act as an energy absorption can be achieved through 350 iterations.
equipment, maximization of unit volume energy absorption
occupies a leading position. Wu is closely related to thick- 4.3 Multi‑objective Optimization of Unit Cell
ness coefficient 𝛼 , length coefficient 𝛽  , cellular angle 𝜑 and
scaling factor of length K  . The corresponding optimization Generally, several objectives should be taken into account in
model is given by structural design simultaneously. Hence, it is indispensable
to search a balance among various objectives. The multi-
Objective ∶maxWu (𝛼, 𝛽, 𝜑, K) objective optimization model can be constructed as
⎧ ⎫ �
⎪ 0<𝛼<1 ⎪ maxWu (𝛼, 𝛽, 𝜑, K)
⎪ 0 < 𝜑 < 90◦ ⎪ Objective ∶
(35) min𝜌RD (𝛼, 𝛽, 𝜑, K)
s.t.⎨ sin 𝜑 < K < 1 ⎬
⎪ 𝛼(1+K) ⎪ ⎧ ⎫
⎪ + 𝛽2 < sin 𝜑 ⎪ ⎪ 0<𝛼<1 ⎪
⎩ cos 𝜑
⎭ (37)
⎪ 0 < 𝜑 < 90 ◦
⎪
s.t.⎨ sin 𝜑 < K < 1 ⎬
⎪ 𝛼(1+K) ⎪
⎪ + 𝛽2 < sin 𝜑 ⎪
The optimization process is shown in Fig. 7, where the ⎩ cos 𝜑
⎭
optimal Wu is obtained after 400 iterations.

4.2 Single Objective Optimization on Minimizing

Relative Density

Lightweight plays an essential role in automotive design pro-

cess. The corresponding optimization model of minimizing
relative density can be expressed as follows.

Objective ∶min𝜌RD (𝛼, 𝛽, 𝜑, K)

⎧ ⎫
⎪ 0<𝛼<1 ⎪
⎪ 0 < 𝜑 < 90◦ ⎪ (36)
s.t.⎨ sin 𝜑 < K < 1 ⎬
⎪ 𝛼(1+K) ⎪
⎪ + 𝛽2 < sin 𝜑 ⎪
⎩ cos 𝜑
⎭
Fig. 8  Optimization process of relative density 𝜌RD for unit cell

13
8 F. Ma et al.

5 Topology Optimization of Thin‑Walled

Tube

Based on above analysis, the re-entrant triangles honey-

combs with the optimal parameters will be embedded into
thin-walled tube to improve energy absoprtion performance,
whereas the cross-section of tube should be determined
firstly. An automotive front side rail is a critical component
of load bearing and safety protection. This paper takes front
side rail as an example. Based on the decoupling thought,
filler re-entrant triangles honeycomb is utilized to satisfy
energy absorption performance, and outer thin wall should
meet high structural stiffness. In order to obtain the cross-
section shape profile, static topology optimization is con-
Fig. 9  Multi-optimization of unit volume energy absorption Wu and
relative density 𝜌RD for unit cell
ducted with targets for maximizing axial stiffness and lat-
eral stiffness. On the basis of static topology optimization,
dynamic topology optimization of impact energy absorption
of thin-walled tube is also conducted to compare the RTT
where Wu represents unit volume energy absorption, 𝜌RD with the NTT design.
denotes relative density of re-entrant triangles structure,
K stands for scaling factor of length, 𝜑 is cellular angle 5.1 Theories of Static Topology Optimization
and𝛼 and 𝛽 are thickness coefficient and length coefficient,
respectively. Variable density method is the most commonly utilized
Multiple Pareto non-inferior solutions can be acquired method of continuous topology optimization [25, 26]. On
through multi-objective genetic algorithm optimization. the hypothesis that the material in structure is changeable, as
An evaluation function Q is proposed herein to search the topology optimization model is established, the elements
for the optimal result. Based on the definition of evalu- with the density less than the threshold value governed by
ation function, the solution improves as the satisfaction certain rules are omitted [27]. Then topological structure is
decreases: realized and optimal allocation of structural materials can be
( ) achieved. With the target for maximum stiffness, the relative
𝜌(𝛼, 𝛽, K, 𝜑) − 𝜌min (Wmax − Wu (𝛼, 𝛽, K, 𝜑))
Q = 𝜔1 + 𝜔2 density of material xi ( i = 1, 2, … , n ) is set as a variable of
(𝜌max − 𝜌min ) (Wmax − Wmin ) topological design. The optimization objective is structural
(38)
minimum compliance. In view of equilibrium conditions and
where 𝜌max and 𝜌min represent the maximum and minimum mass constraints, mathematical models of topology optimi-
of relative density in optimal solutions, respectively. Wmax zation are given by
and Wmin stand for maximum and minimum unit volume
energy absorption in optimal solutions and𝜔1 and 𝜔2 denote findxi , (i = 1, 2, … , n);
the weighting coefficients of relative density and unit vol- objective ∶ minC = F T S = ST KS
ume energy absorption. The multi-optimization process is
⎧ ∑ mi� ≤ M f
N
explained in Fig. 9. The optimal evaluation function Q is (39)
⎪ M
⎪ 0 < xmin ≤ xi ≤ xmax ≤ 1
obtained through 400 iterations. Weighting coefficients are s.t. ∶ ⎨ i=1
𝜔1 = 0.4 and 𝜔2 = 0.6 , respectively. Through the optimiza-
⎩ F = KS
tion process, the optimal structural parameters of unit cell
are displayed in Table 3.
Here, C is compliance,F stands for applied load, S repre-
sents structural displacement, K expresses structural stiff-
Table 3  The optimal geometric parameters
ness, xmin denotes lower bound of relative density, xmax
Optimal geometric parameters Values is upper bound of relative density, mi indicates the mass of
Thickness coefficient 𝛼 0.1 unit i and M and M f are initial total mass and target mass
Length coefficient 𝛽 0.2 fraction. Among this, for the� thin-walled tube with re-entrant
triangles structure, M f = and for the one without re-
M −MACM
Cellular angle 𝜑 33.5° M �
Scaling factor of length K 0.56 entrant triangles structure, M f = M
 . M represents the mass
′

13
Crashworthiness Optimization Design of Thin-Walled Tube Filled with Re-entrant Triangles… 9

of PST, and MACM is the total mass of re-entrant triangles

structure fully filled with solids.
Static topology optimization method is composed of sev-
eral steps; the first step is to define design space of topology
optimization. Then finite element model is constructed; load
and boundary conditions are also established. Optimization
parameters are subsequently set up. Finally, the calculation
is implemented in HyperWorks software. Based on conver- Fig. 11  Boundary conditions for static topology optimization
gence situations, if the results cannot meet the convergence
criterion, recalculation should be executed by modifying
the design variables. Otherwise, the optimal results are The static topology optimization is conducted using the
obtained. A flow diagram of static topology optimization is OptiStruct solver of HyperWorks software; final optimiza-
described in Fig. 10. tion results are displayed in Fig. 12.
The topological optimization results are subsequently
5.2 Static Topology Optimization pruned according to the requirements of the actual manu-
facturing process. The optimization result of maximum
Based on the profile of a car model’s front side rail, a rec- axial stiffness can be seen as quadrilateral embedded with
tangular structure of 100 mm × 100 mm × 300 mm is adopted an octagon, and material in the four corners is very thin.
as initial design space for optimization. The front side rail The cross-section is interpreted as an octagon embedded
is meshed with hexahedral elements, and the bottom is with cruciform structure if the four corners are removed. To
clamped with all the degrees of freedom restricted. Fig- achieve maximum lateral stiffness, the octagonal structure is
ure 11 exhibits boundary conditions of the traditional one. obtained in consideration of loading direction and structural
Uniform distributed pressure is applied to both axial and side symmetry. Final simplified shape and cutting graph of front
loadings. Element density is set as design variable, minimum side rail are exhibited in Fig. 13.
strain energy is regarded as objective function, and mass It can be seen that the optimal cross-section is the com-
fraction is considered as constraint. Meanwhile, extrusion bination of octagonal and cruciform structure. The join of
molding process constraints are also considered. cruciform structure contributes greatly to improving energy
absorption capabilities and static stiffness including axial
Define Space of stiffness and lateral stiffness. If the optimal cross-section
Topology Optimization without cruciform structure is filled with re-entrant triangles
structure, subjected force along inner re-entrant triangles
will be partially or totally counteracted during impact com-
Finite Element Mesh
Generation
pressive process, and the characteristics that shrinking under
compression load will not occur, resulting in the decrease
in energy absorption capabilities, as illustrated in Fig. 14.
Define Applied Load & Owing to the existence of the cruciform structure, the octag-
Boundary Conditions
onal structure is divided into four parts, each achieving supe-
rior performance of stiffness and energy absorption. Besides,
Set up Optimization use of symmetric re-entrant triangles will also enable the
Parameters
convenience of replacement. Although one or two filled re-
entrant triangle structures may fail, the thin-walled tube still
xi t +1 ≤ xi t + ∆x
Structural Analysis works in various loading and material cases. Hence, it is
vitally meaningful to study this novel RTT.
Modify Design Parameters
5.3 Dynamic Topology Optimization of Thin‑Walled
Convergence ? Tube Without Re‑entrant Triangles
N
Sensitivity Analysis
Y
On the basis of the static topology optimization, dynamic
topology optimization of the tube without re-entrant tri-
Optimal Results angles is executed. The optimized octagon shape of static
topology optimization is utilized herein with a length of
300 mm. A rigid wall impacts the tube with mass of 1000 kg
Fig. 10  Flow diagram of static topology optimization and initial velocity of 10 m/s. All the degrees of freedom

13

10 F. Ma et al.

Fig. 12  Optimal results of static

topology optimization

Fig. 13  Optimal cross-section
and cutting graph of front
side rail using static topology
optimization

Fig. 14  Counterbalance of force
along inner re-entrant triangles
without cruciform structure

at the bottom of this tube are constrained, and maximum energy absorption. Eight rectangular induction slots with
effective energy absorption is no less than that of PST [28]. length of 20 mm and width of 5 mm are formed in the eight
The materials used are consistent with that of static topology edges of the octagon at a distance of 15 mm from collision
optimization. The model of the dynamic topology optimiza- end, as exhibited in Fig. 17.
tion is exhibited in Fig. 15.
The dynamic topology optimization model is carried
out in LS-TaSC software; the optimal result is expressed 6 Analysis Results and Discussion
in Fig. 16.
Subsequently, dynamic topology optimization result is To validate the superiority of proposed optimization design
pruned. The outer contour retains octagon shape. The opti- method, the RTT designed based on the decoupling con-
mal shape is thin-walled structure with an inner quadrangle cept, NTT designed using traditional optimization method
and an outer octagon, possessing capabilities of two-stage and PST are analyzed and compared in light of structural

13
Crashworthiness Optimization Design of Thin-Walled Tube Filled with Re-entrant Triangles… 11

Table 4  Stiffness analysis results of four thin-walled tubes

PST (mm) NTT (mm) PFT (mm) RTT (mm)

Axial stiffness 14.60 16.21 13.37 12.95

(maximum
displace-
ment)
Lateral 13.99 15.72 13.30 13.55
stiffness
(maximum
displace-
ment)

Fig. 15  Model of dynamic topology optimization

herein. The most widely used filler with positive Poisson’s
ratio and most excellent performance is aluminum foam
[29]. Here, structural parameters of PFT are the same with
that of RTT, and the density of filled foam is similar to that
of re-entrant triangles. During simulation analysis, the foam
material model is established based on the isotropic uniform
material model proposed by Deshpande and Fleck [30, 31].
Figure 18 exhibits the specific structures of the four thin-
walled tubes. Stiffness analysis and energy absorption com-
parisons are executed below.

6.1 Stiffness Analysis

All tubes are meshed with 2-mm hexahedral elements.

Fig. 16  Optimal result of dynamic topology optimization Loading conditions and boundary conditions are the same
as those of Fig. 12, and Table 4 exhibits the analysis results.
stiffness and energy absorption performance. To verify
outstanding energy absorption capabilities of the proposed 6.2 Energy Absorption Capabilities
re-entrant triangles structure, RTT is compared with PFT
In this paper, finite element models of PST and NTT are
developed using HyperWorks software. The tubes are
meshed with quadrilateral elements, with size of 5 mm.
MAT 24 piecewise linear plasticity model is adopted to
define the parameters of material AA6060T4. In crushing
process, automatic single surface contact is applied on tube
to prevent any interpenetration. The rigid wall impacts the
tube with an initial velocity of V = 10 m/s ; the bottom of
tube is clamped with all degrees of freedom constrained,
as illustrated in Fig. 19. The static and dynamic friction
coefficients are set as 0.3 and 0.2, respectively. In addition,
Fig. 20 exhibits analysis results of energy absorption for four
Fig. 17  Finite element model of thin-walled tube without re-entrant thin-walled tubes and peak crush forces of the four tubes are
triangles
obtained and compared in Table 5.

Fig. 18  Four types of thin-

walled tubes

13

12 F. Ma et al.

Fig. 19  Impact models of PST and NTT

Table 5  Comparisons of peak crush force for four tubes

PST NTT PFT RTT​

Peak crush force (KN) 212.10 176.32 156.06 148.19

Fig. 20  Analysis results of energy absorption for four thin-walled

It can be seen from above results that although traditional tubes
optimization design method can enhance the capabilities of
energy absorption and reduce peak crush force effectively,
absorption performance simultaneously. Compared with the
structural stiffness is weakened, while the proposed decou-
positive Poisson’s ratio material, re-entrant triangles show
pling method can improve both energy absorption and struc-
superior advantages in energy absorption; this is mainly
tural stiffness simultaneously. In comparison with positive
due to the plateau stress enhancement region. In com-
Poisson’s ratio material, re-entrant triangles designed herein
parison with PFT, energy absorption of RTT increases by
exhibit outstanding advantages in energy absorption. In the
17.23% and peak crush force reduces by 5.04%. The results
aspect of stiffness, RTT and PFT show little difference.
are encouraging in the sense that it offers another potential
The reason why lateral stiffness of RTT is slightly lower
material for engineers to address the issue and the novel-
is that re-entrant triangles are possessed with anisotropic
type RTT with re-entrant triangles can be widely promoted
characteristics; stiffness along Y direction is higher than that
and achieve further applications on new energy absorbent
along X direction. When the strain exceeds 0.57 and plateau
structures.
stress increases, energy absorption of the RTT has been sig-
nificantly enhanced. In comparison with the PFT, energy Acknowledgements  The authors are highly appreciated for the finan-
absorption of RTT increases by 17.23% and peak crush force cial support from the National Nature Science Foundation of China
reduces by 5.04%, which is mainly because the re-entrant (No. 2016YFB0101601), Jilin Province Scientific Research Program
triangles structure is more regular and impact deformation (No. SXGJQY2017-7). In addition, we also express our great gratitude
to ZD Ma’s team support in University of Michigan, Ann Arbor, USA.
process is more stable relative to that of aluminum foam.
OpenAccess  This article is distributed under the terms of the Crea-
tive Commons Attribution 4.0 International License (http://creat​iveco​
7 Conclusions mmons​.org/licen​ses/by/4.0/), which permits unrestricted use, distribu-
tion, and reproduction in any medium, provided you give appropriate
credit to the original author(s) and the source, provide a link to the
This paper takes front side rail to introduce decoupling Creative Commons license, and indicate if changes were made.
thought. Filler re-entrant triangles are utilized to satisfy
the requirements of energy absorption, and outer thin wall
should meet high structural stiffness. This work initially References
exhibits theoretical models of re-entrant triangles in terms
of relative density, equivalent elastic modulus and energy 1. Miller, W.S., Zhuang, L., Bottema, J., et al.: Recent development
absorption. Then, experimental test is executed to validate in aluminum alloys for the automotive industry. Mater. Sci. Eng.
the accuracy and reasonability of mechanical models. On A 280, 37–49 (2000)
2. Abramowicz, W., Jones, N.: Dynamic progressive buckling of
this basis, the optimal geometric parameters of unit cell are circular and square tubes. Int. J. Impact Eng. 4(4), 243–270 (1986)
achieved by utilizing NSGA-II method in terms of maxi- 3. Nahas, M.N.: Impact energy dissipation characteristics of thin-
mum unit volume energy absorption and minimum relative walled cylinders. Thin Walled Struct. 15, 81–93 (1993)
density. 4. Seitzberger, M., Rammerstorfer, F.G., Gradinger, R., et  al.:
Experimental studies on the quasi-static axial crushing of steel
The results show that the novel RTT designed based columns filled with aluminum foam. Int. J. Solids Struct. 37(30),
on decoupling thought can improve stiffness and energy 4125–4147 (2000)

13
Crashworthiness Optimization Design of Thin-Walled Tube Filled with Re-entrant Triangles… 13

5. Langseth, M., Hopperstad, O.S.: Static and dynamic axial crush- orthogonally stiffened composite laminates. Compos. B Eng. 105,
ing of square thin-walled aluminum extrusions. Int. J. Impact Eng. 116–131 (2016)
18(7–8), 949–968 (1996) 20. Kim, J. H., Kim, D., Lee, M.: Mechanics of cellular materials and
6. Abramowicz, W.: Thin-walled structures as impact energy absorb- its applications. Multiscale Simul. Mech. Biol. Mater. (2013)
ers. Thin Walled Struct. 41(2–3), 91–107 (2003) 21. Odacı, I.K., Güden, M., Kılıçaslan, C.: The varying densification
7. Limam, A.: Experimental and numerical investigation of static and strain in a multi-layer aluminum corrugate structure: direct impact
dynamic axial crushing of circular aluminum tubes. Thin Walled testing and layer-wise numerical modelling. Int. J. Impact Eng
Struct. 42(8), 1103–1137 (2004) 103, 64–75 (2017)
8. Mirzaei, M., Shakeri, M., Sadighi, M., et al.: Experimental and 22. Sun, D.Q., Zhang, W.H.: Mean in-plane plateau stresses of hex-
analytical assessment of axial crushing of circular hybrid tubes agonal honeycomb cores under impact loadings. Compos. Struct.
under quasi-static load. Compos. Struct. 94(6), 1959–1966 (2012) 91(2), 168–185 (2009)
9. Rossi, A., Fawaz, Z., Behdinan, K.: Numerical simulation of the 23. Zhang, W., Ma, Z., Hu, P.: Mechanical properties of a cellular
axial collapse of thin-walled polygonal section tubes. Thin Walled vehicle body structure with negative Poisson’s ratio and enhanced
Struct. 43, 1646–1661 (2005) strength. J. Reinf. Plast. Compos. 33(4), 342–349 (2014)
10. Gao, G.J., Yao, S.: Study on the energy absorption characteristics 24. Nistor, M., Wiebe, R., Stanciulescu, I.: Relationship between
of vehicle thin-walled structures. Railw. Locomot. Car 02, 8–10 Euler buckling and unstable equilibria of buckled beams. Int. J.
(2002) Non Linear Mech. 95 (2017)
11. Ghamarian, A., Zarei, H.R., Abadi, M.T.: Experimental and 25. Jiao, H.Y., Zhou, Q.C., Li, W.J.: Periodic topology optimization
numerical crashworthiness investigation of empty and foam- using variable density method. J. Mech. Engineering 49 (13),
filled end-capped conical tube. Thin Walled Struct. 49, 1312–1319 132–137 (2013) (in Chinese)
(2011) 26. Ghasemi, H., Park, H.S., Rabczuk, T.: A multi-material level set-
12. Zarei, H.R., Ghamarian, A.: Experimental and numerical crash- based topology optimization of flexoelectric composites. Comput.
worthiness investigation of empty and foam-filled thin-walled Methods Appl. Mech. Eng. 332, 47–62 (2018)
tubes with shallow spherical caps. Exp. Mech. 54, 115–126 (2014) 27. Eschenauer, H.A., Olhoff, N.: Topology optimization of con-
13. Ahmad, Z., Thambiratnam, D.P.: Crushing response of foam-filled tinuum structures: a review. Appl. Mech. Rev. 54(4), 1453–1457
conical tubes under quasi-static axial loading. Mater. Des. 30, (2001)
2393–2403 (2009) 28. Tang, Z.L.: Analysis and improvement design of axial impact
14. Ma, F., Zhao, Y., Liang, H., et al.: Effects of cell micro topology energy absorption of thin walled structure. Dalian University of
on the in-plane dynamic crushing analysis of re-entrant square Tech, Dalian (2012)
cellular material. Automot. Innov. 1(1), 24–34 (2018) 29. Song, X.G., Sun, G.Y., Li, G.Y., et al.: Crashworthiness optimi-
15. Papanicolau, G., Bensoussan, A., Lions, J.L.: Asymptotic analysis zation design of foam-filled tapered thin-walled structures using
for periodic structures. North Holland Publ, Amsterdam (1978) multiple surrogate models. Struct. Multidiscip. Optim. 47(2),
16. Sanchez-palencia, E., Zaoui, A.: Homogenization techniques for 221–231 (2013)
composite media. Springer, Berlin (1987) 30. Yin, H.F., Wen, G.L., Hou, S.J., et al.: Multiobjective crashworthi-
17. Cai, Y., Liang, X., Cheng, G.: Novel numerical implementation of ness optimization of functionally lateral graded foam-filled tubes.
asymptotic homogenization method for periodic plate structures. Mater. Des. 44, 414–428 (2013)
Int. J. Solids Struct. 51(1), 284–292 (2014) 31. Zhang, Y., Sun, G., Xu, X., et al.: Multiobjective crashworthiness
18. Artioli, E.: Asymptotic homogenization of fiber-reinforced com- optimization of hollow and conical tubes for multiple load cases.
posites: a virtual element method approach. Meccanica 1, 1–15 Thin Walled Struct 82, 331–342 (2014)
(2018)
19. Zhou, X.Q., Yu, D.Y., Shao, X.Y., et al.: Asymptotic homogeni-
zation analysis of the dynamics properties of periodically and

13