Академический Документы
Профессиональный Документы
Культура Документы
Thin-Walled Structures
journal homepage: www.elsevier.com/locate/tws
a r t i c l e i n f o
abstract
Article history:
Received 9 March 2012
Received in revised form
10 May 2012
Accepted 14 May 2012
Available online 14 June 2012
In this paper, a class of axisymmetric thin-walled square (ATS) tubes with two types of geometries
(straight and tapered) and two kinds of cross-sections (single-cell and multi-cell) are considered as
energy absorbing components under oblique impact loading. The crash behavior of the four types of
ATS tubes, namely single-cell straight (SCS), single-cell tapered (SCT), multi-cell straight (MCS) and
multi-cell tapered (MCT), are rst investigated by nonlinear nite element analysis through LS-DYNA. It
is found that the MCT tube has the best crashworthiness performance under oblique impact regarding
both specic energy absorption (SEA) and peak crushing force (PCF). Sampling designs of the MCT tube
are created based on a four-level full factorial design of experiments (DoE) method. Parametric studies
are performed using the DoE results to investigate the inuences of the geometric parameters on the
crash performance of such MCT tubes under oblique impact loading. In addition, multiobjective
optimization design (MOD) of the MCT tube is performed by adopting multiobjective particle swarm
optimization (MOPSO) algorithm to achieve maximum SEA capacity and minimum PCF with and
without considering load angle uncertainty effect. During the MOD process, accurate surrogate models,
more specically, response surface (RS) models of SEA and PCF of the MCT tubes are established to
reduce the computational cost of crash simulations by nite element method. It is found that the
optimal designs of the MCT tubes are different under different load angles. It is also found that the
weighting factors for different load angles are critical in the MOD of the MCT tubes with load angle
uncertainty.
& 2012 Elsevier Ltd. All rights reserved.
Keywords:
Tapered tube
Crashworthiness
Energy absorption
Multiobjective optimization
Oblique impact
Load uncertainty
1. Introduction
Thin-walled metal tubes have been widely used as energy
absorbing devices for decades in trains, passenger cars, ships and
other high-volume industrial products since they are relatively
cheap and weight efcient. For instance, the crash box of an
automotive body in white (BIW) is often made of thin-walled
tubes which can absorb the kinetic energy of the vehicle through
plastic deformation during an impact event.
Extensive efforts have been made to investigate the crushing
behavior of thin-walled tubes through analytical, numerical and
experimental methods. Among those, some focused on tubes with
various cross-sections including circular [13], polygonal (e.g.
square, rectangular, etc.) [46] and top-hat like [7], while other
researchers tried to improve the energy absorption of thin-walled
tubes by lling them with different materials including metallic
[811] and polymer foams [12,13]. Besides that, thin-walled
0263-8231/$ - see front matter & 2012 Elsevier Ltd. All rights reserved.
http://dx.doi.org/10.1016/j.tws.2012.05.008
104
established. Design information for such tubes as energy absorbers in oblique impact applications are developed through parametric study based on the sampling designs. A four-level full
factorial design of experiments (DoE) method is used to determine those sampling design points. Also based on the DoE results,
quartic polynomial functions are used to build response surface
(RS) surrogate models that relate SEA and PCF to the geometric
design variables associated with the MC tapered square tubes
under oblique impact loading. Multiobjective particle swarm
optimization (MOPSO) algorithm is used as the optimizer for
solving the MOD problems both with and without load angle
uncertainty.
2. Problem description
2.1. Structures of the crushing analysis under oblique impact loading
The structures considered in this study are tapered axisymmetric
thin-walled square (ATS) tubes with single-cell (SC) and multi-cell
(MC) cross-section congurations. For simplicity, they are named as
SCT (abbreviate of single-cell tapered) tubes and MCT (abbreviate of
multi-cell tapered) tubes afterwards, respectively. For comparison
purposes, their counterparts of straight tubes are named as SCS and
MCS tubes, which stand for single-cell straight and multi-cell straight
tubes, respectively. All tubes have the same baseline geometry with
length L 250 mm, largest cross-sectional side length A80 mm and
wall thickness t 2.0 mm besides that the tapered tubes have an
angle y 51. This geometry is determined from typical dimensions of
the front side rail of a passenger car [15] as well as similar thinwalled structures investigated in the literatures [25,26,37]. The
schematic diagram of the computational model is shown in Fig. 1.
105
3. Numerical simulations
Fig. 2. Cross-section of the ATS tube: (a) SC and (b) MC.
The large end of the tube is clamped, while the other end is impacted
by a rigid wall with a constant velocity of V10 m/s, which is
typically encountered in passenger vehicle crash event. The normal of
the rigid wall is in the XZ plane and has an oblique angle a with the
axis of the tube. The SC and MC cross-sections of the ATS tubes are
shown in Fig. 2(a) and (b), respectively.
2.2. Structural crashworthiness indices under oblique loading
Generally, there are four key crashworthiness indices for
evaluating the performance of the energy absorbing structure,
i.e., mean crushing force (MCF), peak crushing force (PCF), energy
absorption (EA) and specic energy absorption (SEA). The MCF for
a given structural deformation is dened as
Z
1 d
MCF
Fxdx
1
3.1.2. FE models
The FE models of the four types of ATS tubes, namely SCS, SCT,
MCS, MCT tubes, are shown in Fig. 4. The tubes are modeled with
BelytschkoTsay four-node shell element with three integration
points through the thickness and one integration point in the
element plane.
To nd the optimum mesh size for the numerical simulation, a
mesh convergence analysis is performed. The peak crushing force
(PCF) of the four types of ATS tubes under axial impact loading
predicted by different element size and the relative errors with
respect to the results of element size 2.5 mm are summarized in
Table 1. Fig. 5 further depicts the convergence of the PCF with
106
Fig. 4. FE models of the ATS tubes (Overall view and top view): (a) SCS, (b) SCT, (c) MCS and (d) MCT.
Table 1
PCF of the four types of ATS tubes under axial impact loading predicted by different element size.
ATS tubes
SCS
SCT
MCS
MCT
4.0
3.5
3.0
2.5
PCF (kN)
PCF (kN)
PCF (kN)
PCF (kN)
PCF (kN)
60.677
33.792
114.693
91.184
3.82
5.53
3.91
6.43
59.612
33.157
113.273
87.558
1.99
3.55
2.62
2.19
58.767
32.979
111.472
86.911
0.55
2.99
0.99
1.44
58.539
32.255
110.903
85.902
0.16
0.73
0.48
0.26
58.446
32.02
110.376
85.678
0
0
0
0
MCF 13:06s0 b
Fig. 5. Convergence of peak crushing force versus mesh density for ATS tubes
under axial loading.
MCF s0 t
p
Nc 4No 2NT pLc t
p
N c 4No 2NT pLc t
107
Fig. 6. Forcedeection curves for straight square tubes under axial dynamic loading: (a) SC section and (b) MC section.
Table 2
Comparison of the FEA results for SCT tube under quasi-static oblique loads with
the experimental results [25].
Load
angle
(deg)
10
30
Nagel
Present FE Relative
et al. [25] model
error (%)
Nagel
Present
et al. [25] FE model
Relative
error (%)
64.8
39.4
46.7
21.4
0.86
0.93
66.4
39.8
2.47
1.02
47.1
21.6
108
Fig. 8. Effect of load angle on the crushing forcedeection response of ATS tubes: (a) SCS, (b) SCT, (c) MCS and (d) MCT.
Fig. 9. Final deformed proles of ATS tubes under oblique impact with different load angles.
Nevertheless, the critical load angle for the SCT tube is in the range
of 301401, implying that tapering the square tube helps to stabilize
the tube under oblique loading. Same observations have also been
made for both single and double tapered rectangular tubes [25].
Moreover, for the straight tube, the PCF occurs at the beginning of the
crushing when the tube is loaded axially, i.e., a 01, corresponding to
the initiation of buckling at the impacted end of the tube. As the load
angle increases, the PCF reduces in magnitude. In contrast, there is no
predominate initial PCF for the SCT tube due to the tapering feature of
the tube which will be further discussed in a later section. Furthermore, it is seen from Fig. 8(c) and (d) that both straight and tapered
MC tubes show higher crushing force level than their SC counterparts
for all the simulated load angles. It is also seen that both MCS and
MCT tubes exhibit progressive collapse when the load angle is less
than 101. The force histories of the MC tubes show more uctuations
due to the constraining effect of the crisscross part in the tubes
resulting in more lobes during the impact. Similar to the SCT tube, the
MCT tube shows no obvious initial PCF.
The above observed load-deection characteristics of the ATS
tubes can be further understood by looking at the nal stage
deformation modes of the tubes at typical load angles as shown in
Fig. 9. All four types of tubes for a load angle of 101 deform by
progressive buckling initiating at the impacted end. Folds in all
tubes form via the global symmetric (Type I) collapse mode, as
dened in Ref. [6]. Different from the SC tubes with unchanged
central axes, the MC tubes show a tendency to slide sideways
slightly in a direction towards the oblique loading direction due
to their relatively higher stiffness. This is discussed in detail later
on. In fact, for a load angle smaller than 101 each tube fails via
progressive collapse. When the load angle reaches 201, the SC
tubes still fail via progressive collapse but the MC tubes fail via
global bending collapse. Initially the tension ange buckles at the
impacted end, after which a plastic hinge or lobe forms on the
compression ange either close to the clamped end (MCS tube) or
a little bit higher (MCT tube). When the load angle reaches 301,
only the SCT tube fails via progressive collapse, and when the load
angle nally reaches 401, all tubes fail via global bending collapse.
A single bending mode is exhibited for all tubes, in which the
compression ange buckles inward while the two side-walls
buckle outward to form the lobe. A similar mode has been
observed on the oblique loading of square [18] and rectangular
[25] tubes. Thus, there exists a critical load angle for each tube
conguration which makes a transition of the deformation mode
from progressive to global bending collapse. This is further
examined with additional data at 11 increments of the load angle
in the transition range, to determine the critical load angle of each
tube conguration.
Fig. 10 shows the effect of load angle on the SEA calculated up
to a deection of 150 mm for each tube under oblique impact
loading. As seen from the gure, for each tube conguration, three
regions arise from the relationship between the SEA and the load
angle: a region (1) where progressive axial collapse dominates
(progressive region), a region (2) where there is a transition from
progressive to global bending collapse (transition region), and
nally a region (3) where global bending collapse dominates the
response (global region). This relationship has also been found for
the oblique loading of mild steel straight square columns [21],
tapered rectangular tubes [25] and empty and foam-lled conical
tubes [26]. From the detailed observation, the critical load angle
lies within the transition range of 253 r a r 273 for the SCS tube,
393 r a r 403 for the SCT tube, 153 r a r173 for the MCS tube and
143 r a r 163 for the MCT tube. It is evident that the critical load
angles of the SC tubes are generally higher than that of the MC
tubes. The reason is that, with the same wall thickness, the SC
tubes relative lower stiffness causes its loaded end more likely to
deform under oblique loading instead of being pushed laterally
away from the central axis of the tube as in the case of the MC
tube. Thus, the deformed portion of the loaded end serves as a
crush trigger to initiate the progressive collapse of the
109
Fig. 10. Effect of load angle on the SEA of different types of ATS tubes.
110
Table 3
Range and step size of design variables and load angle in the oblique impact
analyses of MCT tubes.
Fig. 11. Normalized SEAa of different types of ATS tubes under oblique impact
loading.
Parameter
Lower bound
Upper bound
Step size
1.0
1
0
2.5
7
30
0.5
2
5
For the purpose of doing this, FEA are performed for the selected
sampling design points determined through the design of experiments (DoE) method since the selection of sampling design points
can be critical for the parametric study as well as the construction
of the responses surface (RS) models for MOD. Many different DoE
methods are available, such as the factorial, Koshal, composite,
Latin Hypercube, D-optimal, etc. [45]. In the following study, a set
of sampling design points are created using a four-level full
factorial DoE method. Consequently, 4 4 sampling design points
are selected, which evenly distribute in the design domain. The
responses of SEA and PCF of the MCT tubes are evaluated at these
16 sampling points with the load angle varying from 01 to 301 at a
step size of 51 such that a total number of 112 FEA are carried out.
The range and step size of the two design variables of the MCT
tubes and the load angle for the DoE are listed in Table 3. The
dimensions are chosen to cover the typical range of tube sizes
commonly used in crashworthiness applications, such as in
automobiles [21,25].
4.2. Parametric study
Fig. 12. Effect of load angle on the peak crushing force of different ATS tubes.
From the results shown above, it appears that the MCT tube
conguration shows a better performance in both SEA and PCF
under oblique impact loading, compared with the other three
structural congurations of the ATS tube. The MCT tube is thus
selected as the best structural component in the present study. In
the following, the inuences of important geometric parameters
on the crushing response of MCT tube are rst examined through
parametric study, followed by multiobjective optimization design
(MOD) to further improve its crashworthiness under oblique
impact loading.
111
Fig. 13. Effect of wall thickness and load angle on tube response: (a) SEA and (b) PCF.
Fig. 14. Effect of taper angle and load angle on tube response: (a) SEA and (b) PCF.
Fig. 15. Relative effect of wall thickness, taper angle and load angle on the
crushing response of MCT tube under oblique impact.
112
12
where F denotes the matrix consisting of basis functions evaluated at the M sampling design points as
2
3
j1 t, y1 jN t, y1
6
7
^
&
^
F4
13
5
j1 t, yM jN t, yM
By substituting Eq. (12) into Eq. (10), the RS models are created
and the response functions SEA(t,y) and PCF(t,y) can be fully
dened.
The accuracies of these RS models can be measured using
several criteria. The maximum absolute error (MAX), the relative
error (RE) and the R square values are written as
i
14
yi y^ i
RE i ,
y
15
R2 1 PM
PM
i1
i1
yi yi 2
i 1,2,. . .,M
16
i
where yi is the FEA result, y^ is the approximation based on the
i
regression function, and y is the mean value of yi. The smaller the
values of MAX and RE, or the larger the value of R2, the more
accurate the RS model.
Table 4
Sampling design points and crashworthiness indices of MCT tubes under different impact angles.
No.
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
Thickness t
(mm)
1.0
1.0
1.0
1.0
1.5
1.5
1.5
1.5
2.0
2.0
2.0
2.0
2.5
2.5
2.5
2.5
Taper angle y
(deg)
1.0
3.0
5.0
7.0
1.0
3.0
5.0
7.0
1.0
3.0
5.0
7.0
1.0
3.0
5.0
7.0
Load angle a 03
SEA
(kJ/kg)
PCF
(kN)
SEA
(kJ/kg)
PCF
(kN)
SEA
(kJ/kg)
PCF
(kN)
SEA
(kJ/kg)
PCF
(kN)
11.020
11.820
12.608
14.087
14.511
15.472
16.967
19.884
17.736
19.223
21.666
27.512
21.065
22.678
27.091
34.329
41.180
34.066
27.011
25.472
70.421
58.021
49.293
49.413
101.30
85.012
85.678
83.106
136.65
123.26
129.56
107.77
9.013
10.402
10.632
12.824
11.544
12.974
14.958
15.266
14.532
16.053
18.670
18.658
17.676
19.235
21.902
16.034
27.446
24.854
20.492
20.312
48.224
43.877
39.264
39.482
77.078
68.147
66.498
63.518
108.19
97.826
99.601
88.439
3.926
7.412
8.104
5.695
4.705
4.929
4.904
8.808
5.677
5.802
5.879
8.782
6.826
6.554
6.080
9.215
23.732
22.031
17.502
15.420
44.728
37.146
32.698
23.516
58.064
54.232
48.244
30.481
76.910
73.698
61.988
38.087
2.672
2.504
2.766
3.221
3.218
3.360
3.403
6.567
4.022
4.182
3.988
5.553
4.652
4.607
4.338
5.916
15.610
15.309
14.573
10.042
28.704
27.081
23.109
15.473
39.991
36.132
33.055
20.198
50.647
51.309
42.102
25.331
polynomial functions for SEA and PCF prediction for the four
representative load angles are provided in the Appendix.
The constructed quartic response surfaces for SEA and PCF
prediction at different impact angles are depicted in Figs. 16 and
17, respectively. It is observed from Fig. 16 that the SEA favors
different taper angle and wall thickness for different load angles.
The PCF, however, favors large taper angle and small wall
thickness for all load angles as shown in Fig. 17. In general, it is
hardly possible to have these two objectives be optimized
simultaneously. Such a typical feature calls for different multiobjective optimization schemes. In this study, the MOPSO algorithm is used to determine the optimum tube congurations
under various design circumstances, which is discussed in the
following section.
Table 5
Accuracy of different polynomial RS models for the MCT tube under load angle
a 01.
Response surface model
SEA
Linear polynomial
Quadratic polynomial
Cubic polynomial
Quartic polynomial
PCF
Linear polynomial
Quadratic polynomial
Cubic polynomial
Quartic polynomial
Maximum RE (%)
R2
4.2336
0.8677
0.3426
0.3031
24.111
7.8736
1.3891
1.3989
0.9206
0.9924
0.9994
0.9996
10.991
10.240
6.8526
2.1966
24.1280
7.9036
12.817
4.1530
0.9808
0.9885
0.9917
0.9993
MAX
113
min
SEAai t, y, PCF ai t, y
17
Fig. 16. Quartic RSs of SEA for the MCT tube under different load angles. The bold black dots in the plots show the training points. (a) a 01, (b) a 101, (c) a 201 and
(d) a 301.
114
Fig. 17. Quartic RSs of PCF for the MCT tube under different load angles. The bold black dots in the plots show the training points. (a) a 01, (b) a 101, (c) a 201 and
(d) a 301.
115
Table 6
Ideal optimums of the two single objective functions for the MCT tube under
different load angles.
Load angle
(deg)
0
10
20
30
Fig. 18. Flowchart of the implementation of the MOD for the MCT tubes with load
angle uncertainty.
Fig. 19. Pareto fronts of PCF and SEA for the MCT tube under different load
angles.
Single objective
t (mm)
y (deg)
SEA (kJ/kg)
PCF (kN)
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
Ideal
2.5000
1.0000
2.5000
1.0000
2.2228
1.0000
1.5742
1.0000
7.0000
6.2925
4.7611
5.9742
7.0000
7.0000
7.0000
7.0000
34.403
13.351
21.701
11.247
9.384
5.910
6.293
3.352
108.270
23.496
99.138
18.453
34.213
15.381
16.318
9.937
Max. SEA
Min. PCF
Max. SEA
Min. PCF
Max. SEA
Min. PCF
Max. SEA
Min. PCF
116
SEAa for each design case as shown in Fig. 20. It should be noted
again that the maximum PCF occurs when the tube is loaded pure
3
axially (a 01), i.e. PCF a PCF 0 . Therefore, the RS models for PCFa
prediction are the same as that shown in Fig. 17(a) for both
design cases.
By using the RS models and the MOPSO algorithm, Pareto
fronts of the two design cases are obtained and shown in Fig. 21.
It is clearly shown that a concave Pareto front curve exists for
each design case, which means the trade-off relationship between
SEAa and PCFa. On the other hand, the range of SEAa is different in
the two design cases, the SEAa value in design case 2 is relatively
small since more weights were given to larger load angles under
which the tube has less SEA.
The optimum tube congurations (see columns 3 and 4) and
the corresponding ideal optimal values of the two single objective
functions SEAa (see column 5) and PCFa (see column 6) predicted
by the RS models are provided in Table 8 and marked as solid star
in Fig. 21. Notice that, for both design cases, the maximum SEAa of
the tube occurs at the value of taper angle locating at its upper
boundary (y 7.01). This indicates that larger taper angle is
benecial for increasing SEA of MCT tubes under oblique impact
loading as demonstrated previously by parametric study. This
observation is the same as the optimal results of the tapered
circular tubes of Refs. [37] and [38]. It is interesting to note that
even though equal weighting factors were assigned to the four
load angles to calculate SEAa in design case 1, the optimum tube
conguration (y 7.01 and t 2.5 mm) to maximize SEAa in design
case 1 is the same as that to maximize SEA when the load angle
a 01 (refer to Table 6). This can be explained as follows: since the
absolute value of tube SEA at a small load angle is much higher
than at a large load angle (e.g., the MCT tube with y 7.01 and
t 2.5 mm gives SEA of 34.329 and 5.916 kJ/kg at load angle a 01
and 301, respectively), the SEA value obtained at a small load
angle plays a leading role in SEAa calculation and dominates the
Table 7
Optimal designs of the MCT tube with PCF constrained under 70 kN at different
load angles.
Load angle
(deg)
t
(mm)
SEA (kJ/kg)
0
10
FEA
Fig. 21. Pareto fronts of the MCT tube for the two design cases with load angle
uncertainty.
Table 8
Ideal optimums of the two single objective functions for MCT tube with load angle
uncertainty.
PCF (kN)
(deg)
RS
model
Error
(%)
RS
model
FEA
Error
(%)
Design case
Single objective
t (mm)
y(deg)
SEAa (kJ/kg)
PCFa (kN)
Ideal
Ideal
Ideal
Ideal
2.5000
1.0000
2.3717
1.0000
7.0000
6.1725
7.0000
6.2118
16.422
8.544
11.790
6.7759
108.270
23.397
102.860
23.433
Max. SEAa
Min. PCFa
Max. SEAa
Min. PCFa
Fig. 20. Quartic RSs of SEAa for the MCT tube with load angle uncertainty. The bold black dots in the plots show the training points. (a) design case 1 and (b) design case 2.
Table 9
Optimal designs of the MCT tube with PCF constrained under 70 kN with load
angle uncertainty.
Design
case
t (mm)
1
2
1.8060
1.8067
7.0000 14.283
6.9943 11.077
FEA
PCFa (kN)
Error
(%)
14.518 1.62
11.246 1.50
RS
model
FEA
Error
(%)
69.676
69.741
69.029 0.94
69.442 0.43
angle uncertainty. In designing energy absorbers in real applications such as the automobile front side rails, the weighting factors
for different load angles should be decided based on the crash
probabilities of each angle in a statistical sense. However, this is
out of the scope of the current paper and could be studied
separately.
Similarly, a larger taper angle is preferred for minimizing PCFa
in both design cases as listed in Table 8. In contrast, the minimum
PCFa occurs at the value of wall thickness locating at its lower
boundary (t 1.0 mm). This conrms the ndings in the parametric study that SEA and PCF show obvious trade-off characteristics with respect to the wall thickness of the MCT tubes.
The Pareto optimal designs of the MCT tube with load angle
uncertainty and PCF constrained under the level of 70 kN are
marked as solid square and solid circle in Fig. 21 for design case
1 and 2, respectively. Meanwhile, the detailed design parameters
for these optimal designs and their corresponding FEA results are
summarized in Table 9. We can nd that the errors between FEA
results and the results obtained by using the RS models are less
than 2%. The error of SEAa is bigger than that of PCFa for the
optimal design of the MCT tubes in both design cases. This implies
3
that the RS model for PCFa, i.e., PCF 0 (Fig. 17a) is more accurate
than those for SEAa (Fig. 20) in the current design cases. Moreover, the optimal design parameters in the two design cases are
very close to the optimal values in the case when load angle a 01
3
(refer to Table 7). This again conrms that SEA0 is the major
contributor in SEAa calculation and the PCF at load angle a 01 is
the maximum among PCF ai for all four load angles considered in
3
the current study, i.e., PCF a PCF 0 .
117
also been presented for the MCT tube with and without load angle
uncertainty effect. It is found that SEA and PCF could be conicting with each other and generally not obtain a simultaneous
optimum as demonstrated in other studies, e.g. Refs. [29,38]. It is
also found that the optimum tube congurations are in general
different for different load angles, either with or without constraint on the PCF level. Therefore, it is necessary to include the
load angle uncertainty effect in the optimization process of such
energy absorbers to obtain a more robust design under oblique
impact loading. Last but not the least, the MOD results of two
design cases with load angle uncertainty show that the selection
of weighting factors for different load angles can affect the
location of the optimal solution to maximize SEAa of the MCT
tube, and thus plays an important role in the crashworthiness
optimization design of such energy absorbers with load angle
uncertainty.
Acknowledgments
This work is supported by the National Natural Science
Foundation of China (no. 50905024, 51105053), Liaoning Provincial Natural Science Foundation of China (no. 20102026) and the
Research Fund for the Doctoral Program of Higher Education of
China (no. 20090041120032, 20110041120022). The nancial
supports are gratefully acknowledged.
Appendix
The quartic polynomial functions of SEA and PCF for the MCT
tube under different dynamic load angles in this study are
provided below
3
A:1
2
2
6. Conclusions
A:2
A:3
2
A:4
A:5
2
2
118
A:6
A:7
2
A:8
A:9
Design case 2:
2
A:10
References
[1] Alexander JM. An approximate analysis of the collapse of thin cylindrical
shells under axial loading. Quarterly Journal of Mechanics and Applied
Mathematics 1960;13(1):105.
[2] Abramowicz W, Jones N. Dynamic progressive buckling of circular and square
tubes. International Journal of Impact Engineering 1986;4(4):24370.
[3] Guillow SR, Lu GX, Grzebieta RH. Quasi-static axial compression of thinwalled circular aluminum tubes. International Journal of Mechanical Sciences
2001;43:210323.
[4] Wierzbicki T, Abramowicz W. On the crushing mechanics of thin-walled
structures. Journal of Applied Mechanics 1983;50:72734.
[5] Mamalis AG, Manolakos DE, Baldoukas AK, Viegelahn GL. Energy dissipation
and associated failure modes when axially loading polygonal thin-walled
cylinders. Thin-Walled Structures 1991;12:1734.
[6] Abramowicz W, Jones N. Dynamic axial crushing of square tubes. International Journal of Impact Engineering 1984;2(2):179208.
[7] White MD, Jones N, Abramowicz W. A theoretical analysis for the quasi-static
axial crushing of top-hat and double-hat thin-walled sections. International
Journal of Mechanical Sciences 1999;41:20933.
[8] Hanssen AG, Langseth M, Hopperstad OS. Static and dynamic crushing of
circular aluminium extrusions with aluminium foam ller. International
Journal of Impact Engineering 2000;24:475507.
[9] Hanssen AG, Langseth M, Hopperstad OS. Static and dynamic crushing of
square aluminium extrusions with aluminium foam ller. International
Journal of Impact Engineering 2000;24:34783.
[10] Seitzberger M, Rammerstorfer FG, Degischer HP, Gradinger R. Crushing of
axially compressed steel tubes lled with aluminium foam. Acta Mechanica
1997;125(14):93105.
[11] Seitzberger M, Rammerstorfer FG, Gradinger R, Degischer HP, Blaimschein M,
Walch C. Experimental studies on the quasi-static axial crushing of steel
columns lled with aluminum foam. International Journal of Solids and
Structures 2000;37(30):412547.
[12] Aktay L, Toksoy A, Guden M. Quasi-static axial crushing of extruded
polystyrene foam-lled thin-walled aluminum tubes: Experimental and
numerical analysis. Materials and Design 2006;27(7):55665.
[13] Ghamarian A, Zarei HR, Abadi MT. Experimental and numerical crashworthiness investigation of empty and foam-lled end-capped conical tubes. ThinWalled Structures 2011;49(10):13129.
[14] Chen WG, Wierzbicki T. Relative merits of single-cell,multi-cell and foamlled thin-walled structures in energy absorption. Thin-Walled Structures
2001;39(4):287306.
[15] Kim H-S. New extruded multi-cell aluminum prole for maximum crash
energy absorption and weight efciency. Thin-Walled Structures 2002;40:
31127.
[16] Zhang X, Cheng G. A comparative study of energy absorption characteristics
of foam-lled and multi-cell square columns. International Journal of Impact
Engineering 2007;34(11):173952.
[17] Tang Z, Liu S, Zhang Z. Energy absorption properties of non-convex multicorner thin-walled columns. Thin-Walled Structures 2012;51:11220.
[18] Reyes A, Langseth M, Hopperstad OS. Crashworthiness of aluminum extrusions subjected to oblique loading experiments and numerical analyses.
International Journal of Mechanical Sciences 2002;44:196584.
[19] Reyes A. Square aluminum tubes subjected to oblique loading. International
Journal of Impact Engineering 2003;28(10):1077106.
[20] Reyes A, Hopperstad OS, Langseth M. Aluminum foam-lled extrusions
subjected to oblique loading: experimental and numerical study. International Journal of Solids and Structures 2004;41(56):164575.
[21] Han DC, Park SH. Collapse behavior of square thin-walled columns subjected
to oblique loads. Thin-Walled Structures 1999;35(3):16784.
[22] Reid SR, Reddy TY. Static and dynamic crushing of tapered sheet metal tubes
of rectangular cross-section. International Journal of Mechanical Sciences
1986;28(9):62337.
[23] Nagel G. A numerical study on the impact response and energy absorption of
tapered thin-walled tubes. International Journal of Mechanical Sciences
2004;46(2):20116.
[24] Nagel G, Thambiratnam D. Computer simulation and energy absorption of tapered
thin-walled rectangular tubes. Thin-Walled Structures 2005;43(8):122542.
[25] Nagel G, Thambiratnam D. Dynamic simulation and energy absorption of
tapered thin-walled tubes under oblique impact loading. International
Journal of Impact Engineering 2006;32(10):1595620.
[26] Ahmad Z, Thambiratnam DP, Tan ACC. Dynamic energy absorption characteristics of foam-lled conical tubes under oblique impact loading. International Journal of Impact Engineering 2010;37(5):47588.
[27] Hou S, Li Q, Long S, Yang X, Li W. Design optimization of regular hexagonal
thin-walled columns with crashworthiness criteria. Finite Elements in
Analysis and Design 2007;43(6-7):55565.
[28] Hou S, Li Q, Long S, Yang X, Li W. Multiobjective optimization of multi-cell
sections for the crashworthiness design. International Journal of Impact
Engineering 2008;35(11):135567.
[29] Hou S, Li Q, Long S, Yang X, Li W. Crashworthiness design for foam lled thinwall structures. Materials and Design 2009;30(6):202432.
[30] Liu Y. Optimum design of straight thin-walled box section beams for crashworthiness analysis. Finite Elements in Analysis and Design 2008;44(3):13947.
[31] Liu Y. Crashworthiness design of multi-corner thin-walled columns. ThinWalled Structures 2008;46(12):132937.
[32] Zarei H, Kroger M. Optimization of the foam-lled aluminum tubes for crush
box application. Thin-Walled Structures 2008;46(2):21421.
[33] Zarei H, Kroger M. Optimum honeycomb lled crash absorber design.
Materials and Design 2008;29(1):193204.
[34] Yin H, Wen G, Hou S, Chen K. Crushing analysis and multiobjective
crashworthiness optimization of honeycomb-lled single and bitubular
polygonal tubes. Materials and Design 2011;32(8-9):444960.
[35] Zarei H, Kroger M. Bending behavior of empty and foam-lled beams: Structural
optimization. International Journal of Impact Engineering 2008;35(6):
5219.
[36] Zhang Z, Liu S, Tang Z. Design optimization of cross-sectional conguration of
rib-reinforced thin-walled beam. Thin-Walled Structures 2009;47(89):86878.
[37] Acar E, Guler MA, Gerc- eker B, Cerit ME, Bayram B. Multi-objective crashworthiness optimization of tapered thin-walled tubes with axisymmetric
indentations. Thin-Walled Structures 2011;49(1):94105.
[38] Hou S, Han X, Sun G, Long S, Li W, Yang X, et al. Multiobjective optimization
for tapered circular tubes. Thin-Walled Structures 2011;49(7):85563.
[39] Hallquist J. LS-DYNA keyword users manual, version: 970. Livermore Software Technology Corporation; 2003.
[40] Kurtaran H, Eskandarian A, Marzougui D, Bedewi NE. Crashworthiness design
optimization using successive response surface approximations. Computational Mechanics 2002;29(45):40921.
[41] Santosa SP, Wierzbicki T, Hanssen AG, Langseth M. Experimental and
numerical studies of foam-lled sections. International Journal of Impact
Engineering 2000;24(5):50934.
[42] Langseth M, Hopperstad OS. Static and dynamic axial crushing of square
thin-walled aluminium extrusions. International Journal of Impact Engineering 1996;18(78):94968.
[43] ANSYS I: Ansys documentation. ANSYS LS-DYNA users guide, version 8.1;
2004.
[44] Zhang X, Cheng G, Zhang H. Theoretical prediction and numerical simulation
of multi-cell square thin-walled structures. Thin-Walled Structures
2006;44(11):118591.
[45] Myers RH, Montgomery DC. Response surface methodology:process and
product optimization using designed experiments. New York: Wiley; 1995.
[46] Yang RJ, Gu L. Experience with approximate reliability-based optimization
methods. Structural and Multidisciplinary Optimization 2004;26(12):
1529.
[47] Papalambros PY, Wilde DJ. Principles of optimal design: modeling and
computation. New York: Cambridge; 2000.
[48] Coello Coello CA, Pulido GT, Lechuga MS. Handling multiple objectives with
particle swarm optimization. IEEE Transactions on Evolutionary Computation
2004;8(3):25679.
[49] Deb K, Pratap A, Agarwal S, Meyarivan T. A fast and elitist multiobjective
genetic algorithm: NSGA-II. IEEE Transactions on Evolutionary Computation
2002;6(2):18297.
119
[50] Knowles JD, Corne DW. Approximating the nondominated front using the
Pareto archived evolution strategy. Evolutionary Computation 2000;8:
14972.
[51] Raquel CR, Naval Jr. PC. An effective use of crowding distance in multiobjective particle swarm optimization. Association for Computing Machinery;
2005:25764.