Thin-Walled Structures 179 (2022) 109591

Contents lists available at ScienceDirect

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

Full length article

Crashworthiness analysis of bioinspired hierarchical gradient multicell tubes

under axial impact
Shangan Qin a , Xiaolin Deng b ,∗, Xinyan Liu c
a
School of Mechanical Engineering, Guangxi University, Nanning, China
b
School of Electronics and Information Engineering, Wuzhou University, Wuzhou, China
c Farsoon Technologies, Changsha, China

ARTICLE INFO ABSTRACT

Keywords: Hierarchical structures and gradient structures have become research hotspots due to their excellent crashwor-
Crashworthiness thiness. However, hierarchical gradient structures that combine hierarchical structures and gradient structures
Hierarchical structures are rarely reported. Inspired by the gradient distribution of biological structure vascular bundles and the
Gradient structures
excellent energy absorption capacity of hierarchical structures, this paper innovatively combines hierarchical
Energy absorption
structures and gradient structures to propose a new type of bioinspired hierarchical gradient multicell tube
(BHGMT). Crashworthiness studies have shown that the high-order BHGMT has a much better crashworthiness
than the low-order BHGMT. The specific energy absorption of the 4th order BHGMT under the same mass is
216.26% that of the 0th order BHGMT. A comparison with other existing hierarchical structures under the
same mass conditions shows that the proposed hierarchical gradient structure also has better crashworthiness.
Subsequently, the systematic development of the influence of structural parameters such as hierarchy, section
thickness, and crisscross rib thickness on the parameterization of BHGMT was carried out. Finally, theoretical
research on the mean crushing force of the 3rd order BHGMT was carried out, and the theoretical solution
was shown to be in agreement with the numerical simulation results.

1. Introduction and scholars. For example, Xie et al. [15] analyzed the crashworthiness
of multicell tubes with different geometric configurations and different
With an increasing number of cars, the occurrence of traffic acci- geometric cross-sections and found that the geometrical cross-sections
dents caused by them has increased. Traffic accidents seriously threaten have a greater influence on the energy absorption of the tube than the
the lives of occupants. Therefore, research on the passive safety of geometric configuration. Li et al. [12] designed a series of multicell cor-
automobiles to reduce damage to occupants when an accident occurs rugated square tubes and analyzed the collapse mode under different
is necessary. To improve the crashworthiness of the passive safety of a impact angles. The multicell triangular corrugated square tubes with
car, one of the most common methods is to install an energy absorbing
crisscross ribs and the multicell cosine corrugated square tubes with
device in front of the car [1]. Thin-walled structures are widely used
crisscross ribs were found to have better performance. Qiu et al. [13]
in buffering energy-absorbing structures due to their light weight,
proposed four types of hexagonal multicell tubes. After optimizing the
convenient manufacturing, and low price. In the past few decades,
cross-sectional dimension, the hexagonal multicell tube with the web-
various thin-walled energy-absorbing structures with different cross-
to-web model was found to have the best performance. These studies
sectional configurations have emerged, such as square tubes [2,3],
circular tubes [4], tapered tubes [5,6], and corrugated tubes [7–9]. In have proven that the multicell tube is superior to the single-cell tube in
the early stage, Zhang et al. [10] studied the axial impact of a square terms of resistance to axial impact. To further improve the crashworthi-
tube with a gradient thickness in cross-sections and found that the ness of the multicell tube, a variable thickness gradient tube with more
energy dissipation of the structure mainly occurs near the corners, and materials distributed in the corner area was created. Fang et al. [16]
the number of corners largely determines the overall energy absorption designed a multicell tube with a functional gradient distribution of wall
capacity of the structure. The number of corners in a single-cell tube thickness. The study found that compared to tubes with equal wall
is limited, which means that its energy absorption capacity is not high. thickness, gradient tubes can absorb more energy under the premise
To construct more corners in the hollow part of the structure, various of a similar peak force. Zhou et al. [17] designed a new type of double
multicell tubes [11–15] have been continuously designed by experts surface gradient tube and studied its energy absorption performance

∗ Corresponding author.
E-mail address: [email protected] (X. Deng).

https://doi.org/10.1016/j.tws.2022.109591
Received 28 October 2021; Received in revised form 21 May 2022; Accepted 4 June 2022
Available online 24 June 2022
0263-8231/© 2022 Elsevier Ltd. All rights reserved.
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

and found that the combination of this double surface gradient strategy crashworthiness study also confirmed that the hierarchical structure
and a multicell tube can lead to better energy absorption capacity. In has better crashworthiness.
the early stage, we designed lateral corrugated tubes with a gradient However, most of the abovementioned studies are inspired by the
thickness in the axial direction [18,19]. Crashworthiness analysis and shape of the structure and seldom consider the distribution of the
optimization studies show that the proposed variable thickness lateral structure. Fig. 1 shows several common plant structures. From Fig. 1(a),
corrugated tubes can greatly reduce the initial peak force. The above- the vascular bundles of bamboo are distributed in a sparse to dense
mentioned studies have fully proven that multicell tubes and gradient gradient from the inside to the outside. In addition, from the magnified
tubes have better crashworthiness than single-cell tubes. view of the core skin group and the cross-sectional view of the stem
The hierarchical structure has become a research hotspot in re- of rice, the structure of their vascular bundles also shows a gradient
cent years because of its excellent energy absorption capacity. There distribution from the inside to the outside. This type of biological
are three common hierarchical structures. One uses a substructure structure with a gradient distribution from the inside to the outside has
instead of vertex [20–25]. For example, Wang et al. [24] designed a inspired us to design a new type of energy-absorbing structure. Then,
vertex-based hierarchical structure, which is formed by replacing the if the gradient and hierarchical structure are combined, what is the
vertices of multicell tubes with self-similar subunits. The study found energy absorption effect of the hierarchical gradient energy absorbing
that the vertex-based hierarchical structure can improve the collapse structure obtained? This is the starting point of this paper. However,
mode, and the wall thickness has a more significant impact on the during our large-scale research in the early stage, there were few litera-
mechanical properties of the structure. Zhang et al. conducted in-plane ture reports of such structures. Ha et al. [41] used a similar concept and
compression experiments [21] and out-of-plane [20] compression and designed a bioinspired hierarchical gradient multicell structure. Their
impact experiments on the hierarchical structure, respectively. The research found that the proposed bioinspired hierarchical structure in-
results show that the hierarchical honeycomb has a higher Young’s creases by 178.4% compared to the 0th order specific energy absorption
modulus during in-plane compression and can absorb more energy at the 3rd order, and the undulation of the load-carrying capacity of
during axial compression and impact. The second is the form of sub- the 3rd order structure is decreased by 88.8% compared with that of
structures instead of edges [26–31]. For example, Luo et al. [31] the square tube. This proves that a hierarchical gradient structure has
designed a structure that replaced the square tube with a self-similar great advantages and research potential in energy absorption compared
structure to the original square tube, forming a rectangular self-similar with traditional structures. Inspired by the related research of plant
structure. The mean crushing force of this structure is 2.5 times higher gradient distribution and existing hierarchical structures, a new type
than that of the original square tube. Li et al. [26] compared the of hierarchical gradient multicell tube structure (Fig. 1(d)) is designed
honeycomb structure and the second order structure and found that in this paper. The structure is a self-similar hierarchy from the multicell
the hierarchical structure has a higher mean crushing force. Zhang tube. After such a hierarchical evolution of the multicellular tube, the
et al. [27] designed an edge-based hierarchical structure and found that new structure has the characteristic of gradient distribution on the cross
as the hierarchical order increases, the crashworthiness of the structure section, that is, the distribution of square cells is sparse to dense from
also improves. The 2nd order structure designed by Fang et al. [29] the center to the outside. At the same time, Ha et al. [41] did not study
has a plateau stress that is increased by 2.63 times and 4.16 times the influence of the ribs on the crashworthiness of the structure. This
compared to the traditional honeycomb structure and aluminum foam paper focuses on the impact of different crisscross rib thicknesses on
filling structure, respectively. The third structure replaces subunits in the crashworthiness of the structure.
cells [32–36]. For example, Zhang et al. [36] designed a hierarchical The organization of the paper is as follows: The second section
structure inspired by pomelo peel and found that the specific energy presents materials and methods, which explain the hierarchical form
absorption and equivalent plateau stress of the structure were 1.5 and parameters of the structure, as well as the crashworthiness index
times and 2.5 times higher than those of the traditional honeycomb and the finite element model and validation. The third section analyzes
structure, respectively. Xu et al. [34] designed a new type of hexagonal the crashworthiness of different hierarchical structures under the same
hierarchical structure. Compared with the original structure, the new mass condition and compares the crashworthiness of other hierarchical
type of hierarchical structure showed better crashworthiness as a result structures. The fourth section is the parameterization research that was
of controlling the distribution of materials. Although the three hier- performed. Finally, the theoretical model of the mean crushing force for
archical structures have different hierarchical forms, the various new 3rd order tubes is analyzed.
hierarchical energy-absorbing structures compared with the traditional
single-cell and multicell structures all illustrate the superiority of the 2. Materials and method
hierarchical structure in energy absorption.
After hundreds of millions of years of evolution and natural selec- 2.1. Geometry and parameters
tion, organisms have long possessed excellent performance and clever
structures to deal with the extreme environment of the outside world. This section explains the design and evolution process of the hier-
By placing a material where it is most needed, the structure can archical gradient multicell structure in this paper through bioinspired
achieve the required function with the least material cost. Therefore, design (as shown in Fig. 2). Taking the 0th order bioinspired hier-
a bioinspired structure provides an important source of inspiration archical gradient multicell tube (BHGMT) as the initial structure, the
for the structural design of thin-walled tubes. The energy absorbing 1st order BHGMT and 2nd order BHGMT are based on the previous
structure of bioinspired structures has become a hot research topic. order, and crisscross ribs are used to divide the square cell into 4
For example, Ma et al. [37] designed a multicell corrugated tube smaller square cells. The 3rd order BHGMT and 4th order BHGMT are
inspired by the structure of the Palinuridae tail and found that the constructed by dividing the square cell on the outside of the previous
structure has a lower peak force and higher energy absorption capacity order into 4 smaller square cells with crisscross ribs. Due to structural
than the same traditional tube. Inspired by several different organisms, limitations, only 0th order BHGMTs to 4th order BHGMTs are designed.
Wang et al. [38] designed four types of bioinspired structures and Finally, to ensure the structures have the same mass, the thickness of
obtained more complex structures through a self-similar hierarchy. each order can be expressed as follows:
Through comparative analysis, these types of structures have different
4 × 𝐿0 × 𝑡0
advantages. Inspired by the bamboo structure, Hu et al. [39] designed 𝑡𝑛 = ∑𝑛 (1)
4 × 𝐿0 + 4 × 𝑚 𝑖 × 𝐿𝑖
a bioinspired honeycomb tube nesting structure that exhibits high 𝑖=1
specific energy absorption. Zhang et al. [40] designed a new type n is the hierarchical order, and 𝑚𝑖 is the number of crisscross ribs of
of hierarchical structure inspired by the microstructure of bone. The the 𝑖th order. 𝑡0 and 𝑡𝑛 are the wall thicknesses of the 0th and 𝑛th

2
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

Fig. 1. Microstructure of several plants and evolution of the cross-section of tubes: (a) bamboo’s gradient structure [42], (b) maize’s gradient structure [43], (c) rice’s gradient
structure [44], (d) evolution of the cross section of hierarchy gradient multicell tubes.

Fig. 2. Cross-sectional configuration of the bioinspired hierarchical gradient multicell tube.

order structures, respectively. 𝐿0 is the length of the cross-section of performance of the two structural unit masses. SEA can be expressed by
the square tube, and 𝐿𝑖 is a quarter of the length of the 𝑖th order the following Eq. (4):
crisscross rib. According to the hierarchical rules, 𝐿𝑛 can be determined 𝐸𝐴
as follows: 𝑆𝐸𝐴 = (4)
𝑚
𝐿 where m is the mass of the structure.
𝐿𝑛 = 𝑛0 (2)
2 CFE is the crushing force efficiency. It is defined as follows:
𝐿0 is the length of the cross-section square. In this paper, 𝐿0 is main- 𝑀𝐶𝐹
tained at 160 mm. According to Eqs. (1) and (2), the wall thickness of 𝐶𝐹 𝐸 = × 100% (5)
𝐼𝑃 𝐹
each order tube and the quarter length of each order crisscross rib (the
IPF is the initial peak force of the force–displacement curve. MCF is the
minimum cell length) can be determined.
mean crushing force, defined as follows:
𝐸𝐴
2.2. Crashworthiness parameters 𝑀𝐶𝐹 = (6)
𝑑
To determine the energy absorption capacity and crashworthiness The ideal structure should have high SEA, and low IPF. At the same
of the structure, several classical parameters are recognized, such as time, CFE should be as close to 100% as possible.
energy absorption (EA), specific energy absorption (SEA), initial peak force
(IPF ), crushing force efficiency (CFE), and mean crashing force (MCF ). 2.3. Finite element model construction and validation
EA refers to the energy absorbed by the structure during the colli-
sion, that is, the integral of the force to the displacement during impact. 2.3.1. Finite element model construction
The expression is as follows: Abaqus/Explicit was used for numerical simulation. The constructed
𝑑
numerical model is shown in Fig. 3. The height of the tube is 𝐻 =
𝐸𝐴 = 𝐹 (𝑥)𝑑𝑥 (3) 100 mm, and it is placed between the two rigid walls. To improve
∫0 the analysis efficiency, the upper rigid wall impacts the tube with a
where 𝐹 (𝑥) is the crushing force as a function of displacement 𝑥 during constant speed of 1000 mm/s [46], and the impact distance is set to
the crushing process and 𝑑 is the effective deformation distance [41, 70 mm. In the numerical simulation of the tube, the S4R shell element is
45]. used for simulation, and the default 5 integration points are used in the
SEA represents the energy absorbed per unit mass of the structure. thickness direction. It is worth noting that when using thin plate/thin
For unequal masses, SEA can be used to compare the energy absorption shell theory finite element elements, the thickness divided by length of

3
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

Fig. 3. Finite element model.

2.3.2. Validation of finite element modeling

The finite element model has been widely used in the engineering
field. With reasonable settings, the numerical simulation and experi-
mental results will agree. This not only saves costs, but the result is
acceptable in engineering. The structure proposed in this paper is rela-
tively complex. Although it can be processed by means of 3D printing
or Wire cut Electrical Discharge Machining (WEDM), the overall cost
is higher. Therefore, this paper performs structural analysis by means
of finite element simulation. For the needs of subsequent analysis, the
accuracy of the finite element model is first validated by experiments.
Here we choose 1st order tube to validate the accuracy of the finite
element model. Fig. 6 is the comparison between experiment and
simulation. The length, height, and wall thickness of the experimental
sample are 60 mm, 40 mm, and 1 mm, respectively. The material of
the experimental sample is AA6061 O, and the manufacturing method
adopts WEDM. The sample is shown in Fig. 6(a). The equipment used in
the quasi-static compression experiment is a universal testing machine
SANS CMT5205. The quasi-static compression tester compresses to 70%
(28 mm) of the total tube height at a constant speed of 2 mm/min. After
compression, the deformation mode of the 1st order tube is shown in
Fig. 4. Engineering stress–strain curve of the material [47]. the Fig. 6(a). The geometric parameters of the finite element model are
consistent with the experiments. The material properties are obtained
from the tensile experiments. The engineering stress–strain curve is
shown in Fig. 6(b), and the Young’s modulus is 69 GPa, Poisson’s ratio
element should be less than 1/5. To prevent the tube and rigid wall is 0.33, the yield stress is 60 MPa, the ultimate stress is 171 MPa. Since
from penetrating, the general contact algorithm is used to control the the bottom of the sample is not fixed during the experiment, no tie
contact behavior of the rigid wall and the tube, and the friction factor constraints are imposed here, and the rest of the settings are consistent
is 0.2 [25,30]. with those described in Section 2.3.1. It can be seen from Fig. 6(a)
The material used for the tube is aluminum alloy AA6061O. Fig. 4 that the deformation models after compression in the experiment and
shows the engineering stress–strain curve of the material [47]. The simulation are similar. In addition, from the data in Fig. 6(c) and
material has the following properties: Young’s modulus 𝐸 = 68.0 GPa, Table 1, the error of the experimental and simulated MCF and EA
density 𝜌 = 2700 kg/m3 , Poisson’s ratio 𝑣 = 0.33, initial yield stress is less than 8.5%, and the error of IPF is less than 7%. For further
𝜎𝑦 = 71 MPa, ultimate stress 𝜎𝑢 = 130.7 MPa. To simulate the plastic de- validation, we validate the experiment of Zhang et al. [49], and the
formation of the proposed tubes, the standard elastic–plastic isotropic results are shown in Fig. 7. From Fig. 7, the force–displacement curve,
hardening constitutive model is employed. This paper does not consider energy absorption curve, and final deformation state of the impact have
the failure of the material. Because the strain rate dependency can be high consistency between the numerical value and the simulation. At
neglected here for the aluminum [32,48], the effect of the strain rate about 75 mm, the experimental sample first entered the densification
is ignored in the numerical simulation. stage due to the influence of the bottom welding, so the forces showed
Different mesh sizes have a great influence on the numerical sim- a relatively large difference. This difference is understandable. The
ulation results. In this paper, the 4th order BHGMT was selected for errors of the IPF and EA are shown in Table 1. Experimental and
mesh convergence analysis and testing to find a suitable mesh size. simulated IPF are 30.19 kN and 30.45 kN, respectively, with an error
The numerical model mesh of the 4th order BHGMT is divided into 5 of only 0.46%, while the EA impacting 85.2 mm is 1640.02 J and
1698.11 J, respectively, with an error of approximately 2.98%. The
different sizes: 2.0 mm, 1.8 mm, 1.5 mm, 1.3 mm, and 1.2 mm. Fig. 5
above analysis shows that the numerical simulation can be used instead
shows the numerical simulation results. From Fig. 5, when the mesh
of the experiment to perform the following analysis.
are 2.0 mm and 1.8 mm, respectively, although the deviation of the IPF
was small, the EA and the mesh are 17.89% and 12.22%, respectively, 3. Crashworthiness analysis
when the mesh is 1.5 mm. When the mesh is 1.3 mm and 1.2 mm, the
maximum deviation of energy absorption compared with IPF at 1.5 mm 3.1. Crashworthiness analysis of BHGMT
is only 2.82%. A large amount of computing resources are required
due to the mesh size being too fine. Therefore, the mesh size of the In this section, we conduct a preliminary study on the crashworthi-
subsequent numerical simulation is 1.5 mm. ness of the five structures. These structures are all compared under the

4
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

Fig. 5. Mesh convergence analysis: (a) MCF and computing time of different mesh sizes of BHGMT, (b) deviation of EA and IPF with different mesh sizes compared to 1.5 mm.

Fig. 6. Comparison between experiment and simulation: (a) Schematic diagram of experiment and simulation, (b) Engineering stress–strain curve, (c) Comparison between
experiment and simulation force–displacement curve and MCF.

Fig. 7. Finite element model verification [47]: (a) force–displacement curves and (b) EA curves.

same mass (0.207 kg). The relevant parameters and crashworthiness the energy absorption curve of the five structures. From the force–
data are shown in Table 2. Fig. 8 is the force–displacement curve and displacement curve, it is intuitive to see that the difference in IPF

5
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

Fig. 8. Force–displacement and energy absorption curves of tubes with the same mass under axial impact: (a) force–displacement curve, (b) EA curve.

Table 1 comprehensive comparison with several other cross-section tubes is

Errors in experiments and simulations.
made in this section. Fig. 10(a) shows a tube with a 4-panel bioinspired
Experiments Parameters IPF (kN) EA (J) Errors (%) constituent element designed by Jiang et al. [50] inspired by biological
IPF EA constituent elements. Fig. 10(b) shows a new type of multicell tube
Experiments Experimental 33.11 457.17 – – obtained by Nia et al. [11], a modified 16-cell tube. Fig. 10(c) is a
in this paper rectangular self-similar multicell sandwich-walled tube designed by
Numerical 35.32 495.29 6.67 8.34
Zhang’s [49] Experimental 30.19 1640.02 – – Luo et al. [31]. Fig. 10(d) is a vertex-based hierarchical multicell tube
experiment designed by Wang et al. [24], and its principle is to replace the vertex
Numerical 30.45 1698.11 0.46 2.98
of the original multicell tube with small rectangular cells. Fig. 10(e)
shows the 4th order BHGMT proposed in this paper.
Numerical simulation analysis is performed for these different tubes
between the five structures is not large. The plateau forces are different. while maintaining the same mass, and the results are shown in Fig. 11.
It is obvious from the figure that the plateau forces of 0th order, 1st Fig. 11(a) and Fig. 11(b) are the force–displacement curves and energy
order and 2nd order are not very different, which means that the effect absorption curves of tubes with different structures. There are many
of evolution from 0th order to 2nd order is not obvious. However, when differences in the results obtained with different cross-sections. The
the structure evolves to 3rd order, the plateau force is significantly plateau forces of the 4th order BHGMT are significantly higher than
improved, and the improvement of 4th order is even greater. At an those of the other tubes, indicating that the 4th order BHGMT has better
impact distance of 70 mm, the densification stage is not evident. There energy absorption capacity. Fig. 11(c) shows the comparison between
is a direct relationship between the EA and the plateau force. With the the MCF and IPF of these five structures. The 4th order BHGMT in this
improvement of the 3rd and 4th order plateau force, the corresponding paper obviously has the highest MCF among the five structures. At the
EA has also been significantly improved. From the previous structural same time, the IPF of the 4th order BHGMT is 96.56 kN. Compared
evolution, it is known that at the 3rd order, the gradient characteristic with other structures, there is almost no increase, but compared with
appears in the structure, so it is possible that the introduction of the sandwich-walled tube’s IPF 116.58 kN, there is a 20.73% decrease. In
gradient characteristic leads to the improvement of EA. Subsequent addition, Fig. 11(d) shows that the CFE value of the 4th order BHGMT is
parametric studies will further analyze this change. significantly higher than that of the other tubes. The CFE value reaches
Fig. 9 shows a comparison of crashworthiness between structures. 85.04%, and the CFE of the other structures is only 62.40%. From
Fig. 9(a) is a comparison between IPF. The largest IPF appears in the this analysis, under the same mass, the 4th order BHGMT designed
0th order (60.17 kN), and the smallest appears in the 3rd order (54.26 in this paper does not improve the IPF compared with other structure
kN), and the difference is not large. Fig. 9(b) shows the MCF of the configurations but the MCF is larger than for other structures, which
five structures. The 1st and 2nd orders are only 17.63% and 42.59% indicates that the 4th order BHGMT has more advantages than other
higher than the 0th order, respectively. The 3rd order and 4th order structures (see Table 3).
are 110.95% and 216.26% higher than the 0th order respectively. For
SEA, it has the same growth trend as MCF due to the comparison at the 4. Parametric research
same mass. Finally, the comparison of CFE. The CFE of 0th order is only
20.63%, while the CFE of 4th order reaches 69.57%. Compared with 4.1. Influence of hierarchical order
the 0th order, the 4th order has increased by 237.23%. In summary,
with the evolution of the hierarchy, the performance of the structure This section mainly studies the influence of the number of hierar-
has been greatly improved without increasing the IPF. It is preliminarily chical orders on the crashworthiness of the structure. First, 4 different
confirmed that this form of hierarchical gradient evolution can greatly thicknesses were selected for the 0th order, which are 1 mm, 1.5 mm,
improve the crashworthiness of the structure, and subsequent studies 2 mm, and 2.5 mm, respectively, and the corresponding masses are
will further analyze their performance. 0.173 kg, 0.259 kg, 0.346 kg, and 0.432 kg, respectively. Under the
condition of ensuring the same mass, the following four structures are
3.2. Crashworthiness comparison between BHGMT and other hierarchical obtained respectively according to the evolution form of Fig. 2, and
structures the corresponding thickness can be obtained by Eq. (1). We define
a group of structures of 0th order to 4th order corresponding to the
In previous research, various types of cross-section tubes have same mass. The detailed structural parameters are shown in Table 4.
emerged. To test the performance of hierarchical gradient tubes, a Fig. 12 shows the force–displacement curve from the 0th order to the

6
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

Fig. 9. Crashworthiness analysis of BHGMT under the same mass conditions: (a) MCF, (b) SEA, (c) CFE, (d) IPF.

Fig. 10. Cross-sections of different structures: (a) a tube with a 4-panel bioinspired constituent element [50], (b) a multicell square tube [11], (c) a rectangular self-similar multicell
sandwich-walled tube [31], (d) a vertex-based hierarchical multicell tube [24], and (e) 4th order BHGMT.

Table 2
Structural parameters and crashworthiness data of BHGMTs with the same mass.
BHGMT m t EA IPF MCF SEA CFE (%)
(kg) (mm) (J) (kN) (kN) (kJ/kg)
0th order 0.207 1.200 869.13 60.17 12.42 4.20 20.63
1st order 0.207 0.800 1022.71 57.82 14.61 4.94 25.27
2nd order 0.207 0.480 1239.44 57.52 17.71 5.99 30.78
3rd order 0.207 0.300 1834.26 54.26 26.20 8.86 48.29
4th order 0.207 0.209 2749.26 56.45 39.28 13.28 69.57

Table 3
Detailed crashworthiness data of different structures.
Different tubes m (kg) t (mm) EA (kJ) IPF (kN) MCF (kN) SEA (kJ/kg) CFE(%)
Ref. [50] 0.346 1.0607 2.206 95.99 31.51 6.38 32.83
Ref. [11] 0.346 0.7692 3.364 97.96 48.06 9.72 49.06
Ref. [31] 0.346 0.7619 5.092 116.58 73.74 14.72 62.40
Ref. [24] 0.346 0.5450 4.097 97.96 58.53 11.84 59.75
4th order 0.346 0.3478 5.748 96.56 82.11 16.61 85.04

7
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

Fig. 11. Crashworthiness data analysis of different structures: (a) force–displacement curve, (b) EA curve, (c) MCF and IPF, (d) CFE.

4th order under the same mass. Similar to other thin-walled structures, Another noteworthy phenomenon is the nonprogressive folding de-
BHGMT undergoes an initial rapid ascent stage under axial impact, formation of the 4th order BHGMT during impact. The results of five
followed by a plateau stage, and part of the structure finally enters a deformation modes extracted at different positions are shown in Fig. 15.
densification stage. In general, a longer plateau stage is better for a The number of wrinkles is different for different positions at the same
given impact distance. The larger the order of BHGMT is, the higher time. When the impact distance is 17.5 mm, the number of wrinkles
the force of the structure. In the 3rd order and 4th order BHGMTs, the increases from position 1 to position 5. When the impact reaches
force increased rapidly at the final stage of impact. Fig. 13 is a graph 52.5 mm, positions 2, 3, and 4 begin to show global buckling. At the
of energy absorption. From the figure, energy absorption increases end of the impact, different positions show different deformation states.
substantially as the hierarchical order increases. Taking Group 1 as It is reasonable to believe that different deformation modes in adjacent
an example, the energy absorption of the 1st order to the 4th order positions will inevitably produce interactions. This is also believed to
is increased by 23.93%, 55.61%, 132.67% and 242.74%, respectively, be an important reason for the nonprogressive deformation of the 4th
compared to that of the 0th order. The method shown in Fig. 14(a) order BHGMT.
is used to refer to the cross-sectional schematic diagrams of different Fig. 16 shows the crashworthiness analysis of different orders of
orders of BHGMT. The 0th order BHGMT only has 1 wrinkle, and BHGMT. Fig. 16(a) is a comparison of IPF. Under the same mass
the wrinkles are not completely flattened, which indicates that the (same mass in the same group), the initial peak force does not change
structure can continue to impact. Similarly, the 1st order BHGMT and significantly as the order increases. From the data in Table 4, the
2nd order BHGMT were not completely compacted. The folds of 3rd fluctuation of each group is within 7%. In other words, if the same mass
order BHGMT and 4th order BHGMT obviously touched each other, is maintained, the introduction of a hierarchical gradient structure will
which indicates the structure has entered the densification stage. This is not improve the IPF of the structure. Fig. 16(b) is the result of MCF. As
also due to the high-order structure having more crisscross ribs and the the order increases, the MCF gradually increases. This trend can also be
cross-section configuration being more complicated. It is worth noting clearly seen in the previous force–displacement curve. The SEA of the
that the wavelength of the wrinkles from the 0th order to the 3rd order structure also increases as the order increases. For example, the 4th
becomes increasingly shorter, and the number of wrinkles increases. order BHGMT of each group is up to 2.43 times higher than the 0th
There is only 1 wrinkle in the 0th order BHGMT, 4 in the 3rd order order BHGMT SEA. This is because as the order increases, the wrinkle
BHGMT, and the wavelength becomes shorter. The number of folds is wavelength decreases as the structure deforms and more materials
related to the energy absorption capacity of the structure. The more participate in plastic deformation. Similar to SEA and MCF, CFE also
folds and the shorter the wavelength, the better the energy absorption increases as the order increases. In each group, the CFE of the 0th
capacity. order BHGMT was 17.94%, 24.46%, 27.75%, and 32.90%. When the

8
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

Fig. 12. Force–displacement curves of BHGMT at different orders: (a) Group 1, (b) Group 2, (c) Group 3, (d) Group 4.

Table 4
Detailed crashworthiness data for each order of different groups.
Group BHGMT t (mm) m (kg) EA (kJ) SEA (kJ/kg) IPF (kN) MCF (kN) CFE(%)
1 0th order 1.0000 0.173 0.606 3.504 48.26 8.66 17.94
1st order 0.6667 0.173 0.751 4.340 48.38 10.73 22.17
2nd order 0.4000 0.173 0.943 5.448 47.40 13.46 28.40
3rd order 0.2500 0.173 1.410 8.153 47.05 20.15 42.82
4th order 0.1739 0.173 2.077 12.008 46.77 29.68 63.44
2 0th order 1.5000 0.259 1.291 4.986 75.41 18.45 24.46
1st order 1.0000 0.259 1.484 5.730 72.77 21.20 29.14
2nd order 0.6000 0.259 1.783 6.885 72.41 25.48 35.18
3rd order 0.3750 0.259 2.539 9.802 69.87 36.27 51.91
4th order 0.2609 0.259 3.733 14.413 71.42 53.33 74.67
3 0th order 2.0000 0.346 1.962 5.670 101.00 28.02 27.75
1st order 1.3333 0.346 2.294 6.629 100.27 32.77 32.68
2nd order 0.8000 0.346 2.830 8.179 97.69 40.43 41.38
3rd order 0.5000 0.346 4.124 11.919 96.63 58.92 60.97
4th order 0.3478 0.346 5.748 16.612 106.67 82.11 76.98
4 0th order 2.5000 0.432 2.984 6.908 129.57 42.63 32.90
1st order 1.6667 0.432 3.198 7.403 127.01 45.68 35.97
2nd order 1.0000 0.432 3.993 9.243 125.15 57.04 45.58
3rd order 0.6250 0.432 5.747 13.304 128.13 82.10 64.08
4th order 0.4348 0.432 7.828 18.121 131.91 111.83 84.78

order was increased to the 4th order, the CFE of the 4th order BHGMT 4.2. Influence of wall thickness
reached 63.44%, 74.67%, 76.98%, 84.78%, respectively. The lowest
CFE of the 4th order BHGMT is nearly twice as high as the highest This section mainly analyzes the influence of wall thickness on
CFE of the 0th order BHGMT, and the highest CFE value is 84.78%. the crashworthiness of the BHGMT. The four different thicknesses
In summary, the crashworthiness of structure improves as the order of each order are shown in Table 5. Fig. 17 and Fig. 18 are the
increases. The introduction of hierarchical gradients can effectively force–displacement and energy absorption curves, respectively. The
improve various crashworthiness indicators, such as MCF, SEA and CFE, force–displacement curve of the BHGMT fluctuates as the thickness
without increasing the IPF. increases. This is because the structure with a thinner thickness is more

9
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

Fig. 13. EA curves of BHGMT in different orders: (a) Group 1, (b) Group 2, (c) Group 3, (d) Group 4.

prone to plastic deformation. With a thinner wall thickness, the wrinkle there is a linear relationship between the size of the IPF and the
wavelength is relatively short. Therefore, the force changes by a small mass. Fig. 19(b) is a comparison of the MCF for the same-order tube
amount as wrinkles occur, and the curve fluctuation is not obvious. under four different thicknesses. The MCF increases with thickness. For
When the hierarchical order increases to the 3rd and 4th orders, the example, the MCF of the 0th order BHGMTs with different thicknesses
force fluctuations in Fig. 17(d) and (e) begin to slow down. This is due are 8.66 kN, 18.45 kN, 28.02 kN, and 42.63 kN. The MCF of the
to the large cross-sectional area of the 3rd order BHGMT and the 4th 4th order BHGMT are 29.68 kN, 53.33 kN, 82.11 kN, and 111.83 kN.
order BHGMT, and the thickness being less than that of the 0th order The significant increase in MCF is due to the increased wall thickness
BHGMT under the same mass. The thickness of the 1st order BHGMT is strengthening the resistance to plastic deformation at the initial stage
small, and the thinner wall thickness is more prone to wrinkles, so there of impact, which leads to an increase in plateau force. Similarly, SEA
is not an obvious fluctuation in the curve. From the EA curve in Fig. 18, also increases with wall thickness. For example, the SEA of the first
an increase in thickness will lead to a corresponding increase in energy group of 0th order BHGMTs was 3.504 kJ/kg, while that of the fourth
absorption. This is because the strength of the tube is increased against group increased to 6.908 kJ/kg. The SEA of the first group of 4th order
plastic deformation as thickness increases, and the structure dissipates BHGMT was 12.008 kJ/kg, and the fourth group had an increase of
more energy during plastic deformation. For example, the 0th order 18.212 kJ/kg. The analysis of the CFE in Fig. 19(d) shows that the CFE
BHGMT of Group 4 and the EA of Group 1 are 0.606 kJ and 2.984 kJ, increases with wall thickness. Among them, the CFE of the 4th order
respectively. Compared with Group 1, the EA of Group 4 increased by BHGMT increased by 63.44% for the thickness of the first group and
392.41%. When the hierarchical order is increased to 4th order, the EA by 84.78% for the fourth group. As the thickness of the CFE increases
of 4th order BHGMT Group 4 has increased by 276.89% compared to within a certain range, the growth rate of the MCF is faster than that
Group 1. The energy absorption increases greatly as the thickness of of the IPF. This indicates that when the IPF allows, the thickness can
the tube increases. be increased appropriately to achieve a higher MCF.
Fig. 19 shows the detailed crashworthiness data analysis of the
structure. From the IPF of each order tube in Fig. 19(a), the IPF is 4.3. Influence of crisscross rib wall thickness
shown to increases with the wall thickness. When the thickness of the
first group is increased to that of the second group, the IPF of each Because the hierarchical evolution is actually the addition of these
order tube increases by approximately 50%. When the thickness of crisscross ribs. In Section 4.2, we analyzed the influence of the overall
the second group is increased to that of the third group, the IPF of thickness on the structure. In this section, we want to analyze the
each order tube increases by approximately 35%. When the thickness influence of the crisscross ribs introduced by different orders on the
of the third group is increased to that of the fourth group, the IPF of crashworthiness of the structure. The 4th order structure we designed
each order tube increases by approximately 28%. This result shows is obtained from the 0th order evolution through the hierarchy in Fig. 2.

10
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

Fig. 14. Impact final state of BHGMT at different orders: (a) schematic diagram of cross-sectional extraction, (b) 0th order BHGMT, (c) 1st order BHGMT, (d) 2nd order BHGMT,
(e) 3rd order BHGMT, (f) 4th order BHGMT.

Fig. 15. Deformation modes of 4th order BHGMT (Group 4) in 5 different positions.

11
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

Fig. 16. Crashworthiness analysis of different orders of BHGMT: (a) IPF, (b) MCF, (c) SEA, (d) CFE.

Table 5
Parameters and detailed crashworthiness data of each order of BHGMT under four different thicknesses.
BHGMT Group t (mm) m (kg) EA (kJ) SEA (kJ/kg) IPF (kN) MCF (kN) CFE(%)
0th order 1 1.0000 0.173 0.606 3.504 48.26 8.66 17.94
2 1.5000 0.259 1.291 4.986 75.41 18.45 24.46
3 2.0000 0.346 1.962 5.670 101.00 28.02 27.75
4 2.5000 0.432 2.984 6.908 129.57 42.63 32.90
1st order 1 0.6667 0.173 0.751 4.340 48.38 10.73 22.17
2 1.0000 0.259 1.484 5.730 72.77 21.20 29.14
3 1.3333 0.346 2.294 6.629 100.27 32.77 32.68
4 1.6667 0.432 3.198 7.403 127.01 45.68 35.97
2nd order 1 0.4000 0.173 0.943 5.448 47.40 13.46 28.40
2 0.6000 0.259 1.783 6.885 72.41 25.48 35.18
3 0.8000 0.346 2.830 8.179 97.69 40.43 41.38
4 1.0000 0.432 3.993 9.243 125.15 57.04 45.58
3rd order 1 0.2500 0.173 1.410 8.153 47.05 20.15 42.82
2 0.3750 0.259 2.539 9.802 69.87 36.27 51.91
3 0.5000 0.346 4.124 11.919 96.63 58.92 60.97
4 0.6250 0.432 5.747 13.304 128.13 82.10 64.08
4th order 1 0.1739 0.173 2.077 12.008 46.77 29.68 63.44
2 0.2609 0.259 3.733 14.413 71.42 53.33 74.67
3 0.3478 0.346 5.748 16.612 106.67 82.11 76.98
4 0.4348 0.432 7.828 18.121 131.91 111.83 84.78

The 1st order is obtained by adding the crisscross rib to the 0th order, and the added crisscross ribs are the 4th order-4 red part. This is the
and the added crisscross rib is the purple crisscross rib in the 4th order- relationship between the four structures. As shown in Fig. 20, the black
1 (Fig. 20). The 2nd order is obtained by continuing to add crisscross wall thickness is t, double the thickness of the colored crisscross ribs
ribs to the square cells of 1st order on the basis of the 1st order. The to study the impact of different crisscross rib wall thicknesses on the
added crisscross ribs are the green part of the 4th order-2. The 3rd order crashworthiness of the structure. Here, four different crisscross ribs are
structure is obtained by adding crisscross ribs to the outer square cell selected to change the thickness, which are marked with purple, green,
of the 2nd order, and the added crisscross ribs are the blue part of the blue, and red in the figure, and the corresponding tubes are named
4th order-3. The 4th order structure is obtained by adding crisscross 4th order-1, 4th order-2, 4th order-3, and 4th order-4. At the same
ribs on the basis of the square cells outside the 3rd order structure, time, to compare the crashworthiness of the structure after changing

12
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

Fig. 17. Force–displacement curves of BHGMT with different wall thicknesses: (a) 0th order BHGMT, (b) 1st order BHGMT, (c) 2nd order BHGMT, (d) 3rd order BHGMT, (e) 4th
order BHGMT.

the crisscross ribs, 4th order-0 was used as the original structure for impact reaches 35 mm, the tube is further impacted, and some areas
comparative analysis. do not have wrinkles in 4th order-3 and 4th order-4. Until the impact
Fig. 21 shows the deformation modes of five tubes with different reaches 52.5 mm, all tubes gradually enter the densification stage. From
crisscross rib thicknesses. To better show the details of the structure, the deformation process, the internal deformation of the 4th order-2
only 1/4 of the entire structure is shown. Different crisscross rib thick- (Fig. 21c) is shown to be worse. The outer surface has wrinkles, and
nesses have a great influence on the deformation mode of the structure. only the middle part of the inner part is formed. At the same time,
When the impact distance was 17.5 mm, the folds of the 4th order- it extends outward, causing the wrinkle wavelength to increase. The
0, 4th order-1, and 4th order-2 tubes were basically complete, while structure does not fully participate in energy absorption, resulting in a
the 4th order-3 and part of the 4th order-4 tube did not occur. This decrease in overall energy absorption performance. The deformation of
means that the folds of 4th order-0, 4th order-1, and 4th order-2 are 4th order-3 (Fig. 21d) produces a similar progressive deformation from
not sufficiently folded, and the folds of 4th order-3 and 4th order-4 are top to bottom, which is better for absorbing energy.
folded more thoroughly, which implies that the 4th order-3 and 4th Fig. 22 shows the force–displacement and energy absorption curves
order-4 tubes have better energy absorption performance. When the of the five structures. The detailed crashworthiness data is shown in

13
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

Fig. 18. Energy absorption curve of BHGMT with different wall thicknesses: (a) 0th order BHGMT, (b) 1st order BHGMT, (c) 2nd order BHGMT, (d) 3rd order BHGMT, and (e)
4th order BHGMT.

Table 6. The forces of 4th order-3 and 4th order-4 are significantly the initial deformation of the structure, which leads to a corresponding
increased compared to that of 4th order-0, and their energy absorption increase in IPF. For the MCF, changing the thickness of the high-order
is also greater than that of 4th order-0. This could be attributed to crisscross ribs will increase the MCF accordingly. However, the MCF for
the improvement of mass. However, the deformation process of 4th the 4th order-1 and 4th order-2 only increases by 9.10% and 16.29%
order-3 and 4th order-4 also presents a deformation mode that is more compared to that of the 4th order-0, which is not much. When the
conducive to energy absorption than 4th order-0. hierarchical order is increased to the 3rd order and 4th order, the MCF
Fig. 23 shows the crashworthiness analysis of the structure. The IPF increases by 47.06% and 52.03% compared to that of the 0th order,
of the structure increases as the order increases. This is a result of the which means that increasing the thickness of high-order crisscross ribs
mass of the structuring changing as the thickness of different ribs is can effectively improve the energy absorption capacity. From the SEA
changed. A structure with more mass has a stronger ability to resist analysis in Fig. 23(c), the SEA values for 4th order-1 and 4th order-2

14
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

Fig. 19. Crashworthiness analysis of BHGMT with different wall thicknesses: (a) IPF, (b) MCF, (c) SEA, (d) CFE.

Fig. 20. BHGMT with changing crisscross rib wall thickness.

are less than for 4th order-0. Combined with the data in Table 6, the 5. Theoretical analysis
SEA for the 4th order-1 and 4th order-2 is 8.04% and 9.26% less than
for 4th order-0. Increasing the thickness of the crisscross ribs for the 1st Simplified super folding element (SSFE) theory can predict the
order and 2nd order is shown to decrease the performance of the tube. structure of the MCF. In this section, we use the SSFE theory to
In contrast, the SEA for the 4th order-3 and 4th order-4 is 6.89% and derive the structure of the MCF. By establishing theoretical prediction
6.83% greater than that of the 4th order-0. Fig. 23(d) is a comparison of formulas, it is possible to quickly predict the MCF of the structure. From
CFE. From the figure, the 4th order-2 has the lowest CFE, which means the previous analysis, the deformation mode of the 4th order BHGMT
that more materials being allocated to the 2nd order crisscross ribs will is not the progressive folding deformation mode. However, the SSFE
result in a higher IPF and a lower MCF. The CFE of 4th order-3 is as theory is only applicable to the progressive folding deformation mode.
high as 89.50%, which is obviously the intended result. Therefore, only the 3rd order BHGMT is analyzed.

15
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

Fig. 21. Deformation mode of 4th order BHGMT with different crisscross rib thickness: (a) 4th order-0, (b) 4th order-1, (c) 4th order-2, (d) 4th order-3, (e) 4th order-4.

Table 6
Structural parameters and crashworthiness data of 4th order BHGMTs with different crisscross rib thicknesses.
BHGMT m (kg) t (mm) IPF (kN) MCF (kN) SEA (kJ/kg) CFE (%)
Unaltered Altered
4th order-0 0.346 0.3478 – 96.55 82.16 18.15 85.09
4th order-1 0.376 0.3478 0.6957 106.31 89.64 16.69 84.31
4th order-2 0.406 0.3478 0.6957 120.07 95.54 16.47 79.57
4th order-3 0.436 0.3478 0.6957 135.00 120.83 19.40 89.50
4th order-4 0.451 0.3478 0.6957 147.98 124.91 19.39 84.41

The SSFE theory divides the energy dissipation into two parts. One that the value of k ranges from 0.7 to 0.8 [41]. Therefore, the value of k
part is dissipated through bending, and the other part is dissipated 𝑠𝑢𝑚
in this paper is 0.71. 𝑊𝑏𝑒𝑛𝑑𝑖𝑛𝑔 is the energy dissipated through bending,
through membrane [51]. The expression is as follows: 𝑠𝑢𝑚
and 𝑊𝑚𝑒𝑚𝑏𝑟𝑎𝑛𝑒 is the energy dissipated through the membrane.
𝑠𝑢𝑚 𝑠𝑢𝑚 Fig. 24 shows the plastic hinge lines. As shown in Fig. 24(a),
2𝐻𝑘𝑀𝐶𝐹 = 𝑊𝑏𝑒𝑛𝑑𝑖𝑛𝑔 + 𝑊𝑚𝑒𝑚𝑏𝑟𝑎𝑛𝑒 (7)
when the panel is folded, the bending deformation of the folding is
2HkMCF is the external work, 2H is the length of the entire fold, k is simplified to three horizontal plastic hinge lines. At the same time,
the effective impact distance coefficient [51]. Experiments have shown perfect folding is assumed to occur, that is, two plastic hinge lines are

16
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

Fig. 22. Force–displacement and energy absorption curves of different crisscross rib thicknesses: (a) force–displacement curve, (b) EA curve.

Fig. 23. Crashworthiness analysis of different crisscross rib thicknesses: (a) IPF, (b) MCF, (c) SEA, (d) CFE.

rotated by 90◦ and one plastic hinge line is rotated by 180◦ , as shown in the ultimate stress of the material, and n is the power law exponent.
Fig. 24(b). The value of n in this paper is 0.18 [47]. Rewrite Eq. (8) as follows:
Therefore, the expression for bending energy dissipation is as fol- 𝜎0 𝑡
𝑊𝑏𝑒𝑛𝑑𝑖𝑛𝑔 = 𝜋 𝐿𝑡 (9)
lows: 2
It is worth noting that when a fold occurs in the entire structure, Lt
𝑊𝑏𝑒𝑛𝑑𝑖𝑛𝑔 = 2𝜋𝑀0 𝐿 (8)
is the area of the cross-section. In this paper, the section area of the 3rd
L is the length of the panel. 𝑀0 is a fully plastic bending moment, which order BHGMT is 16𝐿0 t mm2 . Therefore, the sum of the bending energy
𝜎 𝑡2 dissipation is as follows:
can be calculated by 𝑀0 = 04 . 𝜎0 is the flow stress, which can be
√
𝜎𝑦 𝜎𝑢
calculated by 𝜎0 = 1+𝑛
. 𝜎𝑦 is the yield stress of the material, 𝜎𝑢 is 𝑠𝑢𝑚
𝑊𝑏𝑒𝑛𝑑𝑖𝑛𝑔 = 8𝐿0 𝜋𝜎0 𝑡2 = 32𝐿0 𝜋𝑀0 (10)

17
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

Fig. 24. Bending plastic hinge lines: (a) schematic diagram of the plastic hinge lines, (b) angle of rotation for the plastic hinge lines.

𝐻2
𝐸crisscross = 4𝐸𝑜𝑛𝑒𝑝𝑎𝑛𝑒𝑙 = 16𝑀0 (14)
𝑡
In this paper, the 3rd order BHGMT has 4 corner elements, 36 T-
shaped elements, and 33 crisscross elements, so the total membrane
energy of each element can be expressed as follows:

𝑠𝑢𝑚 𝐻2
𝐸𝑐𝑜𝑟𝑛𝑒𝑟 = 4𝐸𝑐𝑜𝑟𝑛𝑒𝑟 = 32𝑀0 (15)
𝑡
𝐻2
𝐸𝑇𝑠𝑢𝑚
-𝑠ℎ𝑎𝑝𝑒 = 36𝐸𝑇 -𝑠ℎ𝑎𝑝𝑒 = 432𝑀0 (16)
𝑡
𝑠𝑢𝑚 𝐻2
𝐸𝑐𝑟𝑖𝑠𝑠𝑐𝑟𝑜𝑠𝑠 = 33𝐸𝑐𝑟𝑖𝑠𝑠𝑐𝑟𝑜𝑠𝑠 = 528𝑀0 (17)
𝑡
𝑠𝑢𝑚
Sum all of the elements to obtain 𝑊𝑚𝑒𝑛𝑏𝑟𝑎𝑛𝑒 :

Fig. 25. 3rd order BHGMT constituent elements. 𝑠𝑢𝑚 𝑠𝑢𝑚 𝐻2

𝑊𝑚𝑒𝑛𝑏𝑟𝑎𝑛𝑒 = 𝐸𝑐𝑜𝑟𝑛𝑒𝑟 + 𝐸𝑇𝑠𝑢𝑚 𝑠𝑢𝑚
-𝑠ℎ𝑎𝑝𝑒 + 𝐸𝑐𝑟𝑖𝑠𝑠𝑐𝑟𝑜𝑠𝑠 = 992𝑀0 (18)
𝑡
Substituting Eqs. (10) and (18) into Eq. (7), the expression is as follows:

𝐻2
2𝐻𝑘𝑀𝐶𝐹 = 32𝐿0 𝜋𝑀0 + 992𝑀0 (19)
𝑡
MCF can be obtained by the following expression:
( )
1 16𝐿0 𝜋𝑀0 𝐻
𝑀𝐶𝐹 = + 496𝑀0 (20)
𝑘 𝐻 𝑡
Fig. 26. Schematic diagram of plastic flow area. The half folding wavelength H is obtained from the stationary condi-
tion. That is,
𝜕𝑀𝐶𝐹
=0 (21)
The 3rd order BHGMT in this paper consists of three basic elements, 𝜕𝐻
namely, corner elements, T-shape elements, and crisscross elements, as Thus, the half folding wavelength H can be expressed as follows:
𝑠𝑢𝑚
shown in Fig. 25. Therefore, we can obtain 𝑊𝑚𝑒𝑚𝑏𝑟𝑎𝑛𝑒 by calculating the √
membrane deformation of the three elements separately. 𝐿0 𝜋𝑡
𝐻= (22)
In the symmetrical deformation mode, the edges of the panel will 31
stretch outward. In SSFE theory, the stretching of the panel is simplified Substituting Eq. (22) into Eq. (20), the MCF is obtained as follows:
as a plastic flow area [52], as shown in Fig. 26. After simplification, the √
178.16 𝐿0 𝜋
plastic flow area (outlined in red in Fig. 26) can be integrated to obtain 𝑀𝐶𝐹 = 𝑀0 (23)
𝑘 𝑡
the energy dissipated by the membrane as follows:
Introduce the dynamic coefficient 𝜆 = 1.30 [51] to modify the expres-
𝐻2
𝐸𝑜𝑛𝑒𝑝𝑎𝑛𝑒𝑙 = 𝜎0 𝑡𝑑𝑠 = 𝜎0 𝑡𝐻 2 = 4𝑀0 (11) sion:
∫𝑆 𝑡 √
178.16𝜆 𝐿0 𝜋
Fig. 27 shows the deformation diagrams of the corner, T-shape and 𝑀𝐶𝐹 = 𝑀0 (24)
𝑘 𝑡
crisscross and the corresponding plastic flow area. Fig. 27 shows that
these elements are all symmetric deformation modes. Therefore, there From this, the theoretical prediction values of MCF for 4 different
are 2, 3 and 4 corresponding plastic flow areas. The membrane energy groups of 3rd order BHGMTs with different thicknesses can be calcu-
of the corner, T-shape and crisscross can be calculated as follows: lated, as shown in Table 7.
Table 7 shows that the maximum error between the results pre-
𝐻2 dicted by the SSFE theory and the numerical simulation results is
𝐸𝑐𝑜𝑟𝑛𝑒𝑟 = 2𝐸𝑜𝑛𝑒𝑝𝑎𝑛𝑒𝑙 = 8𝑀0 (12)
𝑡
only −2.70%, indicating that the theoretical results of MCF are highly
𝐻2
𝐸𝑇 -𝑠ℎ𝑎𝑝𝑒 = 3𝐸𝑜𝑛𝑒𝑝𝑎𝑛𝑒𝑙 = 12𝑀0 (13) consistent with the results of numerical simulations.
𝑡

18
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

Fig. 27. Deformation diagram and plastic flow area of 3 elements: (a) corner, (b) T-shape, (c) crisscross.

Table 7 and in the range of 0.6–0.8, 𝐿0 ∕𝐻 does not change the deformation
Comparison of the MCF of 3rd order BHGMT obtained through simulation and theory.
mode of the tube. While 𝐿0 ∕𝑡4 significantly affects the deformation
Group Numerical (kN) Theoretical (kN) Error (%) mode of the tube, in the case of 𝐿0 ∕𝑡4 <600, the deformation of the 4th
1 20.15 20.27 0.60 order tube is in the transition mode. When 𝐿0 ∕𝑡4 >600, the deformation
2 36.27 37.23 2.65
mode of the tube is the progressive mode. Therefore, when designing
3 58.92 57.33 −2.70
4 82.10 80.12 −2.41 a 4th order tube, 𝐿0 ∕𝑡4 should be greater than 600 to ensure the
deformation mode of the tube.

7. Conclusion
6. Deformation mode analysis of 4th order tube
Inspired by the gradient distribution of biological structure vascular
In the previous analysis, the deformation mode of the 4th order tube bundles and the excellent energy absorption capacity of the hierarchical
is not a progressive deformation mode. Therefore, in this section, we structure, a new type of bioinspired hierarchical gradient multicell
further analyze the deformation mode of the 4th order tube. Generally, tube is proposed that innovatively combines the hierarchical structure
the deformation mode of the tube is related to 𝐿∕𝑡 and 𝐿∕𝐻. Therefore, and gradient structure. The validated finite element model was used
we analyze the deformation mode by changing different 𝐿0 ∕𝑡4 and to systematically research the crashworthiness of the structure, the
𝐿0 ∕𝐻. By finite element simulation, we observed two deformations of theoretical model for MCF was deduced, and the following conclusions
the tube, as shown in Fig. 28. One is that the tube deforms first in the were obtained:
middle after impact, and then the deformation spreads to both ends. (1) While maintaining the same mass, the energy absorption ca-
This deformation can easily lead to the global buckling of the tube pacity of high-order BHGMT is much greater than that of low-order
after impact. Therefore, this deformation mode is called transitional BHGMT. At the same time, compared with other hierarchical tubes
mode. The other is the progressive deformation of the folds from under the same mass, the 4th order BHGMT designed in this paper does
top to bottom, which is called progressive mode. Fig. 29 shows the not produce a large IPF and has better energy absorption capacity.
deformation mode classification diagram for 𝐿0 ∕𝐻 and 𝐿0 ∕𝑡4 in the (2) The parameterization study shows that an increase in the hier-
range of 0.6–0.8 and 200–1200, respectively. It is observed from the archical order can reduce the fluctuation of the force during impact.
figure that 𝐿0 ∕𝐻 has little effect on the deformation mode of the tube, At the same time, the increase in the order will not increase the IPF,

19
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

Fig. 28. Two typical deformation modes of the 4th order tube: (a) transitional mode, (b) progressive mode.

Declaration of competing interest

The authors declare that they have no known competing finan-

cial interests or personal relationships that could have appeared to
influence the work reported in this paper.

Acknowledgments

This work is supported by National Natural Science Foundation of

China (No. 52065059) and Guangxi Science and Technology Base and
Talent Special Project, China (No. 2020AC19011).

References

[1] S. Xie, N. Wang, W. Yang, H. Li, Energy absorption performance of thin-walled

metal plate due to upheaval deformation based on experiments and numerical
simulation, Thin-Walled Struct. 131 (2018) 258–273, http://dx.doi.org/10.1016/
j.tws.2018.07.006.
[2] X. Zhang, G. Cheng, Z. You, H. Zhang, Energy absorption of axially compressed
thin-walled square tubes with patterns, Thin-Walled Struct. 45 (2007) 737–746,
http://dx.doi.org/10.1016/j.tws.2007.06.004.
[3] X.W. Zhang, H. Su, T.X. Yu, Energy absorption of an axially crushed square
Fig. 29. Classification of deformation modes of 4th order tubes.
tube with a buckling initiator, Int. J. Impact Eng. 36 (2009) 402–417, http:
//dx.doi.org/10.1016/j.ijimpeng.2008.02.002.
[4] X. Luo, J. Xu, J. Zhu, Y. Gao, L. Nie, W. Li, A new method to investigate
the energy absorption characteristics of thin-walled metal circular tube using
while the EA, SEA, and CFE will increase significantly. The SEA of the
finite element analysis, Thin-Walled Struct. 95 (2015) 24–30, http://dx.doi.org/
4th order BHGMT is up to 2.43 times greater than that of the 0th order 10.1016/j.tws.2015.06.001.
BHGMT. The 4th order BHGMT has the highest CFE of 84.78%. [5] A. Asanjarani, S.H. Dibajian, A. Mahdian, Multi-objective crashworthiness op-
(3) Through a comparison of 4 groups of different thicknesses, timization of tapered thin-walled square tubes with indentations, Thin-Walled
Struct. 116 (2017) 26–36, http://dx.doi.org/10.1016/j.tws.2017.03.015.
increasing the thickness was found to increase the IPF, but MCF, SEA,
[6] S. Hou, X. Han, G. Sun, S. Long, W. Li, X. Yang, et al., Multiobjective
CFE also increase significantly as thickness increases. optimization for tapered circular tubes, Thin-Walled Struct. 49 (2011) 855–863,
(4) By changing the thickness of the crisscross ribs at different http://dx.doi.org/10.1016/j.tws.2011.02.010.
positions, it was found that when the thicknesses of the 1st order and [7] X. Deng, W. Liu, L. Jin, On the crashworthiness analysis and design of a lateral
corrugated tube with a sinusoidal cross-section, Int. J. Mech. Sci. 141 (2018)
2nd order crisscross ribs are doubled, the SEA of the tube is reduced by
330–340, http://dx.doi.org/10.1016/j.ijmecsci.2018.03.001.
8.04% and 9.26%, respectively, compared with the case when thickness [8] Z. Li, W. Ma, P. Xu, S. Yao, Crushing behavior of circumferentially corrugated
is not increased. When the thickness of the 3rd order and 4th order square tube with different cross inner ribs, Thin-Walled Struct. 144 (2019)
crisscross ribs is doubled, the SEA of the tube is 6.89% and 6.83% 106370, http://dx.doi.org/10.1016/j.tws.2019.106370.
[9] Z. Li, S. Yao, W. Ma, P. Xu, Q. Che, Energy-absorption characteristics of a
higher than when the thickness was not increased.
circumferentially corrugated square tube with a cosine profile, Thin-Walled
(5) A theoretical derivation was used to obtain the MCF prediction Struct. 135 (2019) 385–399, http://dx.doi.org/10.1016/j.tws.2018.11.028.
model of the 3rd order BHGMT. The difference between the theoretical [10] X. Zhang, Z. Wen, H. Zhang, Axial crushing and optimal design of square tubes
solution and the numerical simulation result was only −2.70%, indi- with graded thickness, Thin-Walled Struct. 84 (2014) 263–274, http://dx.doi.
org/10.1016/j.tws.2014.07.004.
cating that the theoretical solution was in good agreement with the
[11] A.Alavi. Nia, M. Parsapour, An investigation on the energy absorption char-
numerical solution. acteristics of multi-cell square tubes, Thin-Walled Struct. 68 (2013) 26–34,
http://dx.doi.org/10.1016/j.tws.2013.01.010.
[12] Z. Li, W. Ma, P. Xu, S. Yao, Crashworthiness of multi-cell circumferentially
CRediT authorship contribution statement
corrugated square tubes with cosine and triangular configurations, Int. J. Mech.
Sci. 165 (2020) 105205, http://dx.doi.org/10.1016/j.ijmecsci.2019.105205.
Shangan Qin: Writing – original draft, Validation, Investigation, [13] N. Qiu, Y. Gao, J. Fang, Z. Feng, G. Sun, Q. Li, Theoretical prediction and
Formal analysis, Data curation, Conceptualization. Xiaolin Deng: Writ- optimization of multi-cell hexagonal tubes under axial crashing, Thin-Walled
Struct. 102 (2016) 111–121, http://dx.doi.org/10.1016/j.tws.2016.01.023.
ing – review & editing, Resources, Project administration, Methodology,
[14] Z. Wang, J. Liu, S. Yao, On folding mechanics of multi-cell thin-walled
Conceptualization. Xinyan Liu: Software, Resources, Conceptualiza- square tubes, Compos. B Eng. 132 (2018) 17–27, http://dx.doi.org/10.1016/j.
tion. compositesb.2017.07.036.

20
S. Qin, X. Deng and X. Liu Thin-Walled Structures 179 (2022) 109591

[15] S. Xie, W. Yang, N. Wang, H. Li, Crashworthiness analysis of multi-cell square [34] X. Xu, Y. Zhang, J. Wang, F. Jiang, C.H. Wang, Crashworthiness design of
tubes under axial loads, Int. J. Mech. Sci. 121 (2017) 106–118, http://dx.doi. novel hierarchical hexagonal columns, Compos. Struct. 194 (2018) 36–48, http:
org/10.1016/j.ijmecsci.2016.12.005. //dx.doi.org/10.1016/j.compstruct.2018.03.099.
[16] J. Fang, Y. Gao, G. Sun, G. Zheng, Q. Li, Dynamic crashing behavior of new [35] Y. Nian, S. Wan, M. Li, Q. Su, Crashworthiness design of self-similar graded
extrudable multi-cell tubes with a functionally graded thickness, Int. J. Mech. honeycomb-filled composite circular structures, Constr. Build. Mater. 233 (2020)
Sci. 103 (2015) 63–73, http://dx.doi.org/10.1016/j.ijmecsci.2015.08.029. 117344, http://dx.doi.org/10.1016/j.conbuildmat.2019.117344.
[17] J. Zhou, R. Qin, B. Chen, Energy absorption properties of multi-cell thin-walled [36] W. Zhang, S. Yin, T.X. Yu, J. Xu, Crushing resistance and energy absorption of
tubes with a double surface gradient, Thin-Walled Struct. 145 (2019) 106386, pomelo peel inspired hierarchical honeycomb, Int. J. Impact Eng. 125 (2019)
http://dx.doi.org/10.1016/j.tws.2019.106386. 163–172, http://dx.doi.org/10.1016/j.ijimpeng.2018.11.014.
[18] X. Deng, S. Qin, J. Huang, Energy absorption characteristics of axially varying [37] W. Ma, Z. Li, S. Xie, Crashworthiness analysis of thin-walled bio-inspired multi-
thickness lateral corrugated tubes under axial impact loading, Thin-Walled Struct. cell corrugated tubes under quasi-static axial loading, Eng. Struct. 204 (2020)
163 (2021) 107721, http://dx.doi.org/10.1016/j.tws.2021.107721. 110069, http://dx.doi.org/10.1016/j.engstruct.2019.110069.
[19] X. Deng, S. Qin, J. Huang, Multiobjective optimization of axially varying [38] Z. Wang, J. Zhang, Z. Li, C. Shi, On the crashworthiness of bio-inspired hexagonal
thickness lateral corrugated tubes for energy absorption, Mech. Adv. Mater. prismatic tubes under axial compression, Int. J. Mech. Sci. 186 (2020) 105893,
Struct. (2021) 1–14, http://dx.doi.org/10.1080/15376494.2021.1924901. http://dx.doi.org/10.1016/j.ijmecsci.2020.105893.
[20] Y. Zhang, T. Liu, H. Ren, I. Maskery, I. Ashcroft, Dynamic compressive response [39] D. Hu, Y. Wang, B. Song, L. Dang, Z. Zhang, Energy-absorption characteristics
of additively manufactured AlSi10Mg alloy hierarchical honeycomb structures, of a bionic honeycomb tubular nested structure inspired by bamboo under
Compos. Struct. 195 (2018) 45–59, http://dx.doi.org/10.1016/j.compstruct. axial crushing, Compos. B Eng. 162 (2019) 21–32, http://dx.doi.org/10.1016/j.
2018.04.021. compositesb.2018.10.095.
[21] Y. Zhang, Y. Lin, Y. Li, X. Li, 3D printed self-similar AlSi10Mg alloy hierarchical [40] Y. Zhang, X. Xu, J. Wang, T. Chen, C.H. Wang, Crushing analysis for novel
honeycomb architectures under in-plane large deformation, Thin-Walled Struct. bio-inspired hierarchical circular structures subjected to axial load, Int. J. Mech.
164 (2021) 107795, http://dx.doi.org/10.1016/j.tws.2021.107795. Sci. 140 (2018) 407–431, http://dx.doi.org/10.1016/j.ijmecsci.2018.03.015.
[22] Q. He, J. Feng, H. Zhou, A numerical study on the in-plane dynamic crushing [41] N.S. Ha, T.M. Pham, H. Hao, G. Lu, Energy absorption characteristics of bio-
of self-similar hierarchical honeycombs, Mech. Mater. 138 (2019) 103151, http:
inspired hierarchical multi-cell square tubes under axial crushing, Int. J. Mech.
//dx.doi.org/10.1016/j.mechmat.2019.103151.
Sci. 201 (2021) 106464, http://dx.doi.org/10.1016/j.ijmecsci.2021.106464.
[23] Y. Zhang, M. Lu, C.H. Wang, G. Sun, G. Li, Out-of-plane crashworthiness of
[42] U.G. Wegst, H. Bai, E. Saiz, A.P. Tomsia, R.O. Ritchie, Bioinspired structural
bio-inspired self-similar regular hierarchical honeycombs, Compos. Struct. 144
materials, Nature Mater. 14 (2015) 23–36, http://dx.doi.org/10.1038/nmat4089.
(2016) 1–13, http://dx.doi.org/10.1016/j.compstruct.2016.02.014.
[43] J. Huang, W. Liu, F. Zhou, Y. Peng, N. Wang, Mechanical properties of maize
[24] Z. Wang, Z. Li, C. Shi, W. Zhou, Mechanical performance of vertex-based
fibre bundles and their contribution to lodging resistance, Biosyst. Eng. 151
hierarchical vs square thin-walled multi-cell structure, Thin-Walled Struct. 134
(2016) 298–307, http://dx.doi.org/10.1016/j.biosystemseng.2016.09.016.
(2019) 102–110, http://dx.doi.org/10.1016/j.tws.2018.09.017.
[44] J. Huang, X. Deng, W. Liu, Bionic design of bend-twist coupled thin-walled beam
[25] D. Zhang, Q. Fei, D. Jiang, Y. Li, Numerical and analytical investigation on
based on the structure of rice stem, Mech. Adv. Mater. Struct. (2021) 1–14,
crushing of fractal-like honeycombs with self-similar hierarchy, Compos. Struct.
http://dx.doi.org/10.1080/15376494.2021.1950244.
192 (2018) 289–299, http://dx.doi.org/10.1016/j.compstruct.2018.02.082.
[45] W. Ma, S. Xie, Z. Li, Mechanical performance of bio-inspired corrugated tubes
[26] S. Li, Z. Liu, V.P.W. Shim, Y. Guo, Z. Sun, X. Li, et al., In-plane compression of
with varying vertex configurations, Int. J. Mech. Sci. 172 (2020) 105399, http:
3D-printed self-similar hierarchical honeycombs – static and dynamic analysis,
//dx.doi.org/10.1016/j.ijmecsci.2019.105399.
Thin-Walled Struct. 157 (2020) 106990, http://dx.doi.org/10.1016/j.tws.2020.
[46] J. Zhang, D. Zheng, B. Lu, T. Zhang, Energy absorption performance of hybrid
106990.
cross section tubes under oblique loads, Thin-Walled Struct. 159 (2021) 107133,
[27] Y. Zhang, T. Chen, X. Xu, Z. Hu, Out-of-plane mechanical behaviors of a side
http://dx.doi.org/10.1016/j.tws.2020.107133.
hierarchical honeycomb, Mech. Mater. 140 (2020) 103227, http://dx.doi.org/10.
1016/j.mechmat.2019.103227. [47] X. Deng, L. Cao, Crushing analysis and crashworthiness optimisation for a novel
[28] W. Li, Y. Luo, M. Li, F. Sun, H. Fan, A more weight-efficient hierarchical bioinspired multicell filled tubular structure, Int. J. Crashworthiness 25 (2020)
hexagonal multi-cell tubular absorber, Int. J. Mech. Sci. 140 (2018) 241–249, 1–16, http://dx.doi.org/10.1080/13588265.2020.1807688.
http://dx.doi.org/10.1016/j.ijmecsci.2018.03.006. [48] X. Deng, S. Qin, J. Huang, Crashworthiness analysis of gradient hierarchical
[29] J. Fang, G. Sun, N. Qiu, T. Pang, S. Li, Q. Li, On hierarchical honeycombs under multicellular columns evolved from the spatial folding, Mater. Des. 215 (2022)
out-of-plane crushing, Int. J. Solids Struct. 135 (2018) 1–13, http://dx.doi.org/ 110435, http://dx.doi.org/10.1016/j.matdes.2022.110435.
10.1016/j.ijsolstr.2017.08.013. [49] X. Zhang, H. Zhang, Energy absorption of multi-cell stub columns under axial
[30] D. Zhang, G. Lu, D. Ruan, Q. Fei, Energy absorption in the axial crushing compression, Thin-Walled Struct. 68 (2013) 156–163, http://dx.doi.org/10.1016/
of hierarchical circular tubes, Int. J. Mech. Sci. 171 (2020) 105403, http: j.tws.2013.03.014.
//dx.doi.org/10.1016/j.ijmecsci.2019.105403. [50] B. Jiang, W. Tan, X. Bu, L. Zhang, C. Zhou, C.C. Chou, et al., Numerical,
[31] Y. Luo, H. Fan, Energy absorbing ability of rectangular self-similar multi-cell theoretical, and experimental studies on the energy absorption of the thin-walled
sandwich-walled tubular structures, Thin-Walled Struct. 124 (2018) 88–97, http: structures with bio-inspired constituent element, Int. J. Mech. Sci. 164 (2019)
//dx.doi.org/10.1016/j.tws.2017.11.042. 105173, http://dx.doi.org/10.1016/j.ijmecsci.2019.105173.
[32] Y. Zhang, J. Wang, C. Wang, Y. Zeng, T. Chen, Crashworthiness of bionic fractal [51] X. Zhang, G. Cheng, H. Zhang, Theoretical prediction and numerical simula-
hierarchical structures, Mater. Des. 158 (2018) 147–159, http://dx.doi.org/10. tion of multi-cell square thin-walled structures, Thin-Walled Struct. 44 (2006)
1016/j.matdes.2018.08.028. 1185–1191, http://dx.doi.org/10.1016/j.tws.2006.09.002.
[33] H.H. Tsang, S. Raza, Impact energy absorption of bio-inspired tubular sections [52] T. Tran, S. Hou, X. Han, W. Tan, N. Nguyen, Theoretical prediction and
with structural hierarchy, Compos. Struct. 195 (2018) 199–210, http://dx.doi. crashworthiness optimization of multi-cell triangular tubes, Thin-Walled Struct.
org/10.1016/j.compstruct.2018.04.057. 82 (2014) 183–195, http://dx.doi.org/10.1016/j.tws.2014.03.019.

21