Optimization of Foam-Filled Bitubal Structures For Crashworthiness Criteria

You might also like

Download as pdf or txt
Download as pdf or txt
You are on page 1of 12

See

discussions, stats, and author profiles for this publication at: https://www.researchgate.net/publication/257086255

Optimization of foam-filled bitubal structures


for crashworthiness criteria

Article in Materials and Design June 2012


DOI: 10.1016/j.matdes.2012.01.028

CITATIONS READS

80 225

5 authors, including:

Yong Zhang Guangyong Sun


National Huaqiao University Hunan University
26 PUBLICATIONS 375 CITATIONS 121 PUBLICATIONS 1,920 CITATIONS

SEE PROFILE SEE PROFILE

Zhen Luo Qing Li


University of Technology Sydney University of Sydney
70 PUBLICATIONS 1,416 CITATIONS 314 PUBLICATIONS 4,972 CITATIONS

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Biomechanics of Periodontal Ligament and Bone Remodelling during Orthodontic Tooth Movement
View project

Implant Surface Morhphology Design and Optimisation for Enhanced Osseointegration View project

All content following this page was uploaded by Guangyong Sun on 10 February 2014.

The user has requested enhancement of the downloaded file. All in-text references underlined in blue are added to the original document
and are linked to publications on ResearchGate, letting you access and read them immediately.
Materials and Design 38 (2012) 99109

Contents lists available at SciVerse ScienceDirect

Materials and Design


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

Optimization of foam-lled bitubal structures for crashworthiness criteria


Yong Zhang a, Guangyong Sun b, Guangyao Li b,, Zhen Luo c, Qing Li d,
a
College of Mechanical Engineering and Automation, Huaqiao University, Xiamen 361021, China
b
State Key Laboratory of Advanced Design and Manufacture for Vehicle Body, Hunan University, Changsha 410082, China
c
School of Electrical, Mechanical and Mechatronic Systems, University of Technology, Sydney, NSW 2007, Australia
d
School of Aerospace, Mechanical and Mechatronic Engineering, The University of Sydney, Sydney, NSW 2006, Australia

a r t i c l e i n f o a b s t r a c t

Article history: Thin-walled structures have been widely used as key components in automobile and aerospace industry
Received 27 July 2011 to improve the crashworthiness and safety of vehicles while maintaining overall light-weight. This paper
Accepted 16 January 2012 aims to explore the design issue of thin-walled bitubal column structures lled with aluminum foam. As a
Available online 4 February 2012
relatively new ller material, aluminum foam can increase crashworthiness without sacricing too much
weight. To optimize crashworthiness of the foam-lled bitubal square column, the Kriging meta-model-
Keywords: ing technique is adopted herein to formulate the objective and constraint functions. The genetic algo-
Foams
rithm (GA) and Non-dominated Sorting Genetic Algorithm II (NSGA II) are used to seek the optimal
Impact and ballistic
Plastic behavior
solutions to the single and multiobjective optimization problems, respectively. To compare with other
thin-walled congurations, the design optimization is also conducted for empty bitubal column and
foam-lled monotubal column. The results demonstrate that the foam-lled bitubal conguration has
more room to enhance the crashworthiness and can be an efcient energy absorber.
2012 Elsevier Ltd. All rights reserved.

1. Introduction the effects of different lled tubes on quasi-static axially crushing


behavior; Shariati et al. performed the design optimization for foam
Crashworthiness of structure has been a major concern in vehi- lled spot-welded column [6], and Niknejad et al. presented a
cle design to ensure safety of passenger. In practice, thin-walled theoretical model to predict the instantaneous folding force of a
structures have been widely adopted as main energy absorber for polyurethane foam-lled square column [7]. Moreover, to study
crashing protection attributable to their high stiffnessweight ra- the inuence of the thin-walled structures on energy absorption of
tio, deformation pattern and energy absorption capacity. As higher vehicle, Fang and Solanki [8] and Kamel et al. [9] studied the
and higher design requirements being placed, increasing attention crashworthiness design of full-scale vehicle and truck chassis,
has been paid to new congurations for enhancing vehicle crash- respectively. In this respect, numerical modeling has played a signif-
worthiness and structural performances. In this regard, design icant role. Abramowicz and the coworkers were probably amongst
optimization techniques have been extensively applied to the the earliest researchers to develop numerical models for simulating
development of novel and efcient structures over the past decade. the axial crashing response of thin walled columns [1012].
As one of the most typical structures, thin-walled metallic col- One of the effective ways to improve crashworthiness of thin-
umns have gained considerable popularity for their low manufac- walled structures without increasing too much volume and weight
turing cost and outstanding crashworthiness over the years. is perhaps to use light structures or materials as ller in hollow
Exhausted studies have demonstrated that different thin-walled chambers. It has been shown that ultra-light cellular structures
structures are of rather different capacities of energy absorption, and metal foams are of superior characteristics to absorb energy
which have been motivating materials and design engineers to and increase deformation stability under severe crushing. For
search for more efcient materials and structural congurations to example, some natural porous materials like agglomerate cork
enhance the crashworthiness. For example, the crashworthiness of [13] as well as articial materials like PVC [14] and polyurethane
the fender [1], tapered thin-walled [2] and multi-corner thin-walled foams [15] were attempted as the llers to enhance energy absorp-
[3] structures were studied by using experimental and numerical tion. Such lightweight periodic structures as honeycomb have also
simulation methods. Meantime, to improve the energy absorption, been adopted [16,17] for circular tubes [18] and bitubal hexagonal
Seitzberger et al. [4] and Ahmad and Thambiratnam [5] investigated columns [19,20], which exhibited signicant increases in specic
energy absorption (SEA).
Corresponding authors. Tel.: +86 731 8882 1717; fax: +86 731 8882 2051 (G. Li), As the rapid development of metallic foam materials, signicant
tel.: +61 2 9351 8607; fax: +61 2 9351 7060 (Q. Li). research efforts have been devoted to more thorough understand-
E-mail addresses: gyli@hun.cn (G. Li), Qing.Li@Sydney.edu.au (Q. Li). ing of crashing behaviors and energy absorption characteristics for

0261-3069/$ - see front matter 2012 Elsevier Ltd. All rights reserved.
doi:10.1016/j.matdes.2012.01.028
100 Y. Zhang et al. / Materials and Design 38 (2012) 99109

various metallic foam-lled structures, e.g. monotubal circular force and utilization rate (UR) of progressive deformation length.
[21], square aluminum columns [22,23], patterned square steel Taking an axial impact of column as an example, the absorbed en-
tubes [24], squared and circular boxes [2527], variable sectional ergy (E) is determined theoretically by integrating the crashing
square tubes [28], multi-cell hexagonal columns [18], conical tubes force with respect to displacement as:
[2931], boxes [32,33], and rectangular [34,35] and circular [36] Z d
bending beams, etc. To understand the difference with and without Ed Fxdx 1
aluminum foam-ller, the studies were conducted to compare the 0

structural responses for the empty and foam-lled aluminum where d denotes deformation distance and F is the axial impact
tubes. It was found that the foam-lled tubes could substantially force.
enhance the energy absorption [24,37]. The studies on such differ- Often, the specic energy absorption (SEA) is considered a more
ent congurations as monotubal, bitubular and multi-tube-packed critical criterion to measure energy absorption capability of unit
hexagon and square structures identied that the friction between material and structure, which is dened as the ratio of total ab-
multitube and foam materials contributed energy absorption con- sorbed energy Etotal to total mass of structure Mtotal as:
siderably [5,15]. In a comparative study, Jones suggested that
SEA Etotal =M total 2
foam-llers could be used for signicantly increasing the effective-
ness factor of energy-absorption [38]. The utilization rate (UR) of progressive deformation length can
For the cellular materials or foam-lled thin-walled structures, be dened as the ratio of the deformation distance (d) to the initial
the specic energy absorption (SEA) has been often taken as the de- length (l) of the thin-walled structure:
sign criteria for lightweight requirements in the literature [39,40].
UR d=l 3
It was reported that foam density is of considerable inuence on
energy absorption of various thin-wall tubes [41]. In general, the However, these crashworthiness criteria cannot be quantied
higher the cellular or foam density, the higher the energy absorp- analytically because crashing solution involves highly nonlinear
tion. However, use of high density foam materials may somewhat and time dependent process. Therefore, experimental studies
lose its lightweight feature. As reported by Reyes et al., a high-den- and/or nite element (FE) simulation are needed in order to deter-
sity aluminum foam can increase the energy absorption [42], but mine some approximate solutions to these criteria.
may lower the specic energy absorption (SEA) compared with
the empty tubes [29], which raises a question on how to optimize 2.2. Material properties of aluminum foam and tubes
foam-llers for different structures [19,25].
For this reason, design optimization of thin-wall structures lled The mechanical properties of aluminum foams are of great
with foam or cellular materials has drawn increasing attention re- importance for ensuring the accuracy of numerical simulations.
cently. For example, Zhang et al. optimized the honeycomb llers There have been several constitutive models available for metallic
for enhancing energy absorption of hexagonal columns [20]. Yin foams, which are based on experimental tests under a range of
et al. optimized the honeycomb llers for various polygonal-type loading conditions. Deshpande and Fleck proposed two models
monotubal and bitubal structures [43]. Zarei and Kroger [25] used for mechanical properties of metal foams, which were based on a
multicriteria optimization to maximize the energy absorption and self-similar yield surface and the hardening effects from hydro-
minimize the weight of foam-lled aluminum tubes. Hou et al. static stress of yield surface, respectively [46]. The models were
studied the monotubal thin-walled square column lled with alu- updated over the years and had been implemented in commercial
minum foam by using multiobjective optimization methods [27]. code LS-DYNA, in which the yield function (U) of the constitutive
To minimize the peak impact force and maximize the energy model is described as follows [47]:
absorption, Sun et al. recently proposed an optimal functionally
U re  Y 6 0 4
graded foam-ller for thin walled columns [41]. In these optimiza-
tion studies, response surface method (RSM) has been widely where re is the equivalent stress, and yield stress Y can be dened
adopted to establish suitable surrogate models for various objective as,
and constraint functions. To seek optimal solutions, such different
Y rp Ree 5
approaches as mathematical programming [26], genetic algorithms
(GAs) [44], and multiobjective particle swarm optimization where rp denotes the plateau stress, R(ee) is the strain hardening
(MOPSO) [40,45] were applied to different problems. term and ee is the equivalent strain. The equivalent stress re can
Although the abovementioned signicant work, limited knowl- be expressed as:
edge has been available for such more sophisticated foam-lled
congurations as bitubal structures [5,15]. It is worth exploring its 1
r2e r2v m a2 r2m 6
structural features to gain further insights into the design problems. 1 a=32
This paper aims to address the design issues for foam-lled bitubal
structures by using both single and multiobjective optimization where rvm represents the von Mises stress, rm is the mean stress,
procedures. The optimal foam-lled bitubal structure is compared and parameter a is related to the shape of the yield surface as
with the optimized empty bitubal and foam-lled monotubal below,
structures for exhibiting the crashworthiness difference, thereby 91  2mp
showing its advantages for potential engineering applications. a 7
21 mp
where mp is the plastic coefcient of contraction, and in most cases
2. Material and methods the value of mp equals to zero [48]. The following strain hardening
rule is implemented in this material model by using such material
2.1. Crashworthiness criteria parameters as the density and Youngs modulus of the metal foam.
Thus, the yield stress Y can be expressed in terms of [47]:
In general, there are a number of typical characteristic parame-  b !
ee ee
ters to measure crashworthiness of a structure, namely energy Y rp Ree rp c aln 1  8
absorption (E), specic energy absorption (SEA), maximum crash
ep eD
Y. Zhang et al. / Materials and Design 38 (2012) 99109 101

where rp, a, c and b are the material parameters that can be ob- 250
tained from the stressstrain data via the uniaxial compression
tests with curve tting and they can be expressed as the functions
of the metal foam density in the following consistent form [47]: 200

Engineering stress (Mpa)


  !n
1 qf t=0.9mm
rp ; a; c; C 0 C 1 9 t=1.2mm
b qf 0 150
t=2mm
where qf is the foam density, and qf0 is the density of the base mate-
rial. C0, C1 and n are the constants listed in Table 1 for the different 100
material parameters. The densication strain eD in Eq. (8) is deter-
mined from qf and qf0 as,
! 50
q
eD  ln f 10
qf 0
0
Aluminum alloy AA6063-T6 is used for the outer and inner tubes, 0 0.02 0.04 0.06 0.08 0.1
whose material properties are density q = 2.7  103 kg/m3, Youngs Engineering strain
modulus E = 70 GPa, Poissons ratio m = 0.30, initial yielding stress
Fig. 1. Stressstrain curves of AA 6063 T6 with the uniaxial tensile.
ry0 = 162 MPa, and the ultimate strength ru = 192 MPa [51]. The
constitutive model of the aluminum tube is adopted in the plastic
kinematic material model, Mat 3, in LS-DYNA. The yielding function integration. The shell elements are slightly offset from the top sur-
(U) is dened by von Misses yield criterion as [52]: face of brick elements to avoid the initial penetration. Moreover,
r2y the hourglass control is used to eliminate spurious zero-energy
1
U Sij Sij  11 modes for all the reduced integration elements. A rigid body is cre-
2 3
ated using the brick elements to model the compression block. An
where Sij and ry are the deviatoric stress and the current radius of automatic surface to surface contact algorithm is dened between
the yield surface. In plastic kinematic plasticity criterion, ry is de- the aluminum foam elements and outer and inner tube shell ele-
ned as: ments. To take into account the contact between the tube walls
during deformation and avoid interpenetration of folding mode
ry r0 bEp epeff 12
generated during axial collapse, an automatic single surface con-
where r0 is the yield strength, b describes the isotropic and kine- tact algorithm is used to the walls of outer and inner tubes. Lastly,
matic hardening, epeff is the effective plastic strain and Ep is the plas- the node to surface contact is modeled between the impacting sur-
tic hardening modulus, and Ep can be described as: face of the compression block and the thin-walled tube. Contact
stiffness for all the contact interfaces are selected to avoid inter-
EEt
Ep 13 penetration resulted from low stiffness or instability occurring in
E  Et excessively high stiffness. All degrees of freedom of the left ends
where E and Et are Youngs modulus and tangential modulus, of the tubes (Fig. 2) are constrained to represent kinematic bound-
respectively. ary of a xed wall. The nite element model of the bitubal struc-
The stressstrain curves of this material model are given in ture is shown in parts in Fig. 3.
Fig. 1 based on the uniaxial tension tests [51].
2.4. Kriging approximation model
2.3. Finite element model
As abovementioned, the crashworthiness design can involve a
To model the typical front side rail of a vehicle, a foam-lled bit- range of criteria. First, the specic energy absorption (SEA) indi-
ubal square column of length l = 250 mm is considered. As shown cates the energy absorption per unit mass and has been often used
in Fig. 2, the sectional dimensions of the outer tube and inner tube as one of the most important criteria in crashworthiness design.
are given as b1  b2  t = 82  82  2.0 mm and c1  c2  Second, the maximum crashing force, Fmax, directly determines
t = 42  42  2.0 mm, respectively, for the baseline design. It is as- the force transferred to occupants and can be critical to the occu-
sumed that a compression block impacts onto the foam-lled bit- pant survival rate when crash occurs, which is usually chosen as
ubal column at an initial velocity of 15 m/s. To generate enough a design criterion or constraint. Third, utilization rate (UR) of pro-
initial kinetic energy, an additional mass of W = 400 kg is attached gressive deformation length reects the deformation pattern for
to the compression block as shown in Fig. 2. The commercial nite energy absorption characteristics and is sometimes chosen as a de-
element code, LS-DYNA, is used to simulate crashworthiness of sign constraint.
foam-lled bitubal column under impacting load. One major challenge relies on that often, these abovementioned
The thin-walled tubes are modeled using the BelytschkoTsay criteria cannot be formulated analytically. As an effective alterna-
reduced integration thin shell elements. The foam-ller is modeled tive, various surrogate model techniques have been widely used
using the eight-node brick elements with one-point reduced in crashworthiness design. The response surface method (RSM),
moving least square method (MLS), Kriging and feed-forward neu-
ral networks [53] are some examples. In this regard, the Kriging
Table 1
Material constants for aluminum foam [42,49,50]. method has exhibited a fairly good accuracy for highly nonlinear
design responses [54,55] and will be employed to the construct
rp (MPa) a (MPa) c (MPa) 1/b
surrogate models for design criteria of SEA, Fmax and UR in this
C0 (MPa) 0 0 0 0.22 paper.
C1 (MPa) 590 140 40 320
Typically, the Kriging model can be expressed mathematically
n 2.21 0.45 1.40 4.66
as [54]:
102 Y. Zhang et al. / Materials and Design 38 (2012) 99109

V=15m/s
A-A section A
outer
tube
inner Mass
42mm foam filler z
foam tube block
filler W=400kg
84mm A
l = 250mm

Fig. 2. Dimensions, boundary and loading condition of the foam-lled bitubal column.

Fig. 3. Finite element models of foam-lled bitubal structure (a) assembly of bitubal column, (b) thin-walled square inner and outer tubes, and (c) aluminum foam-ller.

yx f x Zx 14 ^ f T R1 f1 f T R1 y


b 19
where y(x) is an unknown function of interest, f(x) is a known poly- The estimation of variance of the sample data from the global
nomial and Z(x) is stochastic component with zero mean value and model is:
variance r2 as well as a non-zero covariance. Note that f(x) repre-
^ T R1 y  f b
y  f b ^
sents the global approximation to the design response, while Z(x) r^ 2 20
denotes the covariance of the local deviations that can be formu- L
lated as The maximum likelihood estimates for hk can be found by solv-
Cov Zx ; Zx  r RRx ; x 
i j 2 i j
15 ing the following constrained maximization problem:
(
^2
If L is the number of sample points, R is the L  L symmetric positive Maxmize : Uh  Llnr 2lnjRj
21
denite matrix with unit diagonal, and R(xi, xj) is the correlation Subject to : h P 0; h 2 Rn
function between these two sample points xi and xj, which can be
exponential, Gaussian, cubic or other nonlinear functions. The where both r^ 2 and R are the functions of h. R is adaptively regular-
Gaussian correlation function was used in the study as, ized to avoid potential ill-conditioning.
" #
X
nv
2.5. Optimization algorithm
i j
Rx ; x exp  hk jxik  xjk j2 16
k1
2.5.1. Optimization formulation
where nv is the number of design variables, hk (k = 1, 2, . . ., nv) are the The aim of this study was to optimize foam-lled bitubal struc-
unknown correlation parameters, and jxik  xjk j is the distance be- tures with respect to structure thickness (t), yield strength (Y) and
tween the kth components of points xi and xj. Once the correlation foam density (qf) to improve crashworthiness in terms of SEA, peak
^x can be gi-
function is selected, the estimation of the response at y impact force and utilization rate. To do so, single and multiobjec-
ven as: tive optimization methods will be applied, respectively.
^ rT xR1 y  f b
^x b ^ In general, multiobjective optimization can be mathematically
y 17
formulated as follows:
T
where r (x) is the correlation vector between a predicted x and
these L sample points, vector y represents the responses at each Minimize : Fx f1 x; f2 x; . . . ; fn xT

sampling point and f is an L-vector. Vector r and scalar parameter g j x 6 0 j 1; 2; . . . ; m 22
Subject to :
^ are given by:
b hl x 0 l 1; 2; . . . ; k

rT x Rx; x1 ; Rx; x2 ; . . . ; Rx; xL T 18 where n is the number of objective functions (note that n = 1 stands
for single-objective problem), fn(x) is the nth objective function, m
Y. Zhang et al. / Materials and Design 38 (2012) 99109 103

and k are the numbers of inequality and equality constraints,


Random initialization
respectively, and x denotes the vector of design variables.
Typically, there are two methods to solve the above multi objec- of population
tive optimization problem. One is to formulate the different objec-
tive functions into a single cost function, for example, by using
Evaluate the Fitness
weight coefcients, hereby allowing to apply various mathematical
value of individual
programming procedures. However, it can be very difcult to se-
lect these weights for different objectives properly. Furthermore,
for each set of weights, only a single solution can be obtained un- YES
less carefully varying the weights for more solutions. The second Convergence criteria Optimal result
method does not formulate a single-objective cost function. In-
stead, it explores all the objectives independently to search for NO Selection
the Pareto optimal solution. Some heuristic approaches like genetic New generation by
algorithm (GA) [56,57] and particle swarm optimization (PSO) choosing the best Crossover
[40,41] are of such a feature and have been well developed for
applications. In this paper, the second method will be employed. Mutalation
Output theoptimal
2.5.2. Single and multiobjective genetic algorithm (GA) result of current
Practical engineering problems may not be always smooth and generation
continuous, so it can be very difcult (if not impossible) to obtain
the derivatives. This is one of the main reasons why GA is superior
to many traditional mathematical programming algorithms. Ge-
netic algorithm is a stochastic optimization approach based on Fig. 4. Flowchart of single objective GA.
the mechanism of natural selection and natural genetics and it
has become a popular choice to search for a global optimum
[58,59].
The main difference of GA and other algorithms is that GA uses
a random set of solutions called population, instead of only one
single search point. After creating initial solutions randomly, each
one is evaluated based on its tness to the specic criteria. GA will
evolve a population through use of such genetic operators as selec-
tion, mutation and crossover, which makes this numerical proce-
dure similar to the process of natural evolution. Selection is a
process of articially simulating the natural selection, and the indi-
viduals are reproduced according to their tness function values.
The individuals with a higher tness function value have a higher
chance of contributing the offspring in the next generation. After
performing the selection operator, the crossover operator will be
applied to generate new individuals for the next generation. It ex-
changes and recombines the genes of two good individuals. The
high performance individuals are repeatedly selected and ex-
changed for the better and better performance by using selection
and crossover operators. However, it can result in premature con-
vergence of GA to a local optimal point if the selection and cross-
over operators are repeated alone. Therefore, the mutation
operators need to be employed to alter the individual values with Fig. 5. Flowchart of NSGA II.

a very small probability, thus simulating the gene mutation pro-


cess in nature. Hence, the operators allow reintroducing diversity
crashworthiness design problems [56,57]. The owchart for NSGA
into a convergent population and avoiding local optimums. To
II algorithm is briey described in Fig. 5.
achieve fast convergence toward global optimum, the proper evo-
To clarify the whole design optimization procedure, a ow chart
lution of population must be assured by choosing adequate genetic
is provided as in Fig. 6, in which LS-DYNA is used as FEA simula-
operators. The cycle of evolution is repeated until the desired crite-
tion. In-house Matlab programs are coded to conduct the Kriging
rion is reached. The owchart for single objective GA is shown in
modeling as well as the GA and NSGA II algorithms.
Fig. 4.
Since GA searches from population to population rather than
from one solution to another, which makes it particularly suitable 3. Results and discussion
for multiobjective optimization problems. To conduct multiobjec-
tive optimization as formulated in Eq. (22), the Non-dominated 3.1. Single-objective optimization
Sorting Genetic Algorithm (NSGA II) [60] was used herein. NSGA
II is a multiobjective genetic algorithm based on a fast non- In this section, the aluminum foam-lled bitubal column
dominated sorting principle [61]. The non-dominated set can be and empty bitubal column (Fig. 3b) are respectively studied by
obtained by using the elitist non-dominated sorting along with using the Kriging models and GA method to compare their
sorting the crowding distance. The non-dominated fronts are ob- crashworthiness.
tained after each generation. Finally, the non-dominated set is Firstly, to validate the accuracy of nite element model and
obtained with convergence of the iteration to form the Pareto ensure the correctness of material behaviors in simulations for
fronts. This algorithm has proven rather effective for solving design optimization, a comparative study was performed via
104 Y. Zhang et al. / Materials and Design 38 (2012) 99109

thickness t, yield strength r of thin-walled tube and density q of


Initialization of design D-optimal experiment aluminum foam are adopted as the design variables. Thus, a con-
parameters design sampling strained single-objective design optimization of foam-lled bitubal
column can be formulated as follows:
8
> Maximize : SEAt; q; r
>
> 8
>
> > F max t; q; r 6 160 KN
Finite element Construct Kriging meta- >
> >
>
>
< >
>
evalulation models of design responses < UR P 0:67
>
23
>
> Subject to : 0:6 6 t 6 2 mm
>
> >
>
>
> >
> 2:5  102 6 q 6 5  102 kg=m3
>
> >
>
: :
140 6 r 6 230 MPa
Optimization by single For comparison, a constrained single-objective design optimiza-
objective GA or tion of empty bitubal column is similarly formulated as follows:
Converge?
mutiobjective GA 8
> Maximize : SEAt; r
NSGA- II >
> 8
>
> > F max t; r 6 160 KN
NO < >
>
< UR P 0:67 24
> Subject to :
>
>
> >
> 0:6 6 t 6 2 mm
>
: >
:
Add sample points and Optimal solutions or YES
140 6 r 6 230 MPa
update the meta models Pareto fronts
in which the constraints are set up by referring to the literature data
Fig. 6. Flowchart of crashworthiness optimization process. [5,27,46].
The design objective and constraint functions need to be formu-
lated by using Kringing meta-models, which provide an approxi-
experimental test and simulation herein. The compressive test re- mation to the design space in terms of function values at certain
sults of foam-lled double tubes were reported in the literature sampling points. The selection of sample points is a critical step
previously [51]. To use the same testing conditions given in the to ensure a good approximation with minimum computational
reference, the foam-lled bitubal structure is modeled as the cost. In the study, the D-optimal design of experiments (DOEs)
length of 114 mm and the sectional dimensions of the outer and method is used, as it can reduce the number of required sample
inner columns of b1  b2  t = 38  38  0.9 mm and c1  c2  t = points [65] for constructing approximate Kringing functions.
25  25  0.9 mm, respectively. The experiments yielded the Note that the accuracy of meta-models needs to be validated.
specic energy absorption (SEA) of 12.15 J/g and utilization rate Nevertheless, the Kriging model allows the approximate functions
(UR) of progressive deformation length of 0.61, respectively [51], passing through the sample points exactly. For this reason, ve ex-
while our simulation led to 11.98 J/g and 0.59 for these two tra random points are generated in the design domain for the val-
corresponding responses, which represent 1% and 3% differ- idation. Tables 2 and 3 summarize the errors between the FEA
ences, respectively. Furthermore, Fig. 7 compares the progressive values and Kriging model values for these two structures, respec-
deformation mode between the test and simulation. From the tively, which demonstrated that the errors of these Kriging approx-
comparisons, the numerical modeling technique adopted here is imations are in a range of 14%. The similar accuracy was justied
considered in rather good agreement with the experiment and in the literature [54,55], thus the Kriging models established are
can be used for the design optimization. considered fairly adequate for the subsequent design optimization.
Secondly, it is essential to identify the most important parame- By using the single-objective GA, the optimal results for the
ters for the crashworthiness design. Following a variable screen problem dened in Eq. (23) and Eq. (24) are summarized in Table 4.
analysis, it is rst found that the thicknesses of different structures From initial and optimum values of the foam-lled bitubal column,
have a signicant effect on energy absorption and deceleration it can be seen that the optimal SEA is 9.8% higher than the initial
[62,63]. As shown in Fig. 1, the energy absorptions differ when value. The optimal utilization rate of progressive deformation
the thickness of thin-walled tube varies. Second, many experimen- length (UR) increases about 4% and meets the constraint require-
tal studies showed that the yield strength of the tube material is of ment, making the foam-lled bitubal column deform more ideally
signicant effect on energy absorption [35,51,64]. Third, the
stressstrain curves of aluminum foam material a as plotted in
Fig. 8 for two different foam densities [51], also indicated a signif- 30
icant inuence in energy absorption. Thus, in the present study,
25 0.44g/cm3
Engineering stress (Mpa)

0.36g/cm3
20

15

10

0
0 0.2 0.4 0.6 0.8 1
Engineering strain

Fig. 7. Deformation pattern of foam lled bitubal test (left) and simulation (right). Fig. 8. Stressstrain curves of different foam densities under uniaxial compression.
Y. Zhang et al. / Materials and Design 38 (2012) 99109 105

Table 2
Validation of Kriging meta-models for foam-lled bitubal column.

No. t (mm) q (kg/m3) r (MPa) SEA (kJ/kg) Fmax (kN) UR


FEA Kriging FEA Kriging FEA Kriging
1 0.7295 373 175.9 7.02 7.058 73.9 71.92 0.698 0.6983
2 1.0078 278 197.2 10.03 10.46 114.5 114.4 0.689 0.688
3 1.7067 448 177.0 10.85 11.13 173.7 176.9 0.664 0.665
4 0.7896 337 214.8 8.50 8.26 89.56 92.95 0.695 0.6948
5 1.3152 475 229.3 10.33 10.61 174.0 174.58 0.673 0.6742

Table 3 er design constraints even for an experienced designer. Peak im-


Validation of Kriging meta-models for empty bitubal column. pact force Fmax is indeed a key criterion in crashworthiness
No. t (mm) r (MPa) SEA (kJ/kg) Fmax (kN) UR design and can be rather critical for preventing occupants and ob-
FEA Kriging FEA Kriging FEA Kriging
jects from severe injury and damage. To account for these two dif-
ferent design criteria simultaneously, it is more sensible and
1 0.7295 175.9 8.0287 8.2718 73.82 74.38 0.705 0.706
meaningful to formulate the design problem in a multiobjective
2 1.103 191.6 10.971 11.205 123.5 122.13 0.692 0.691
3 1.5067 152.0 12.365 12.097 131.8 132.5 0.647 0.658 optimization framework. Hence, we revise the design problem in
4 0.8152 150.3 7.921 7.8541 72.4 71.32 0.703 0.704 Eq. (23) to a multiobjective problem, as follows:
5 1.3837 179.2 12.451 12.348 143.2 144.42 0.657 0.667 8
> Minimize : SEAt; q; r; F max t; q; r
>
> 8
>
> > UR P 0:67
< >
>
for energy absorption via a longer deformation process. The < 0:6 6 t 6 2 mm 25
> Subject to :
>
maximum impact force of the initial baseline design (192.3 kN) >
> >
> 2:5  102 6 q 6 5  102 kg=m3
>
: >
:
exceeded the corresponding constraint, while the optimum leads 140 6 r 6 230 MPa
to a 17% reduction of Fmax and meets the requirement (159.3 kN).
In addition, both the thickness t of thin-walled tube and density where Fmax is taken as the other objective function. The Kriging
q of aluminum foam are reduced, making the whole weight of meta-models presented in the previous section are used here again
structure decrease from 1095.8 g initially and to 756.9 g nally, a for the multiobjective optimization.
signicant reduction of 31%. A major difference of single-objective and multiobjective design
For comparison, the empty bitubal column is studied herein. optimization is that the multiobjective optimization can provide
Although the optimal SEA is 7.2% lower than the initial baseline va- the designers with a range of optimal solutions by plotting a Pareto
lue, the initial impact force Fmax (189.78) is far above the constraint front as shown in Fig. 9 for objectives SEA vs. Fmax of the foam-lled
(160 kN). The optimum resulted in a 20.6% reduction in Fmax to bitubal column.
meet the constraint requirement. Moreover, it makes UR improve Fig. 9 indicates that the NSGA II method worked effectively and
about 6% and satisfy the constraint. Interestingly, it can be seen generated a well-distributed Pareto front. It is noted from Fig. 9
that the SEA of optimized foam-lled bitubal column is about 3% that SEA and Fmax are strongly competing with each other and can-
higher than that of the optimized empty bitubal column. More not reach a common optimum. In other words, to some stage any
signicantly, the total energy absorption (E) of foam-lled bitubal further improvement in one objective must worsen the other
column is 22.3% higher than that of the empty bitubal column, objective, which leads to a series of solution points in the design
indicating that more energy can be absorbed by the foam-lled space as reported in [27]. It is interesting to note that the single-
structure. Overall, it is clearly demonstrated that the foam-lled objective optimization presents a special point in the Pareto front,
bitubal column has a better crashworthiness than the empty representing a special case in the multiobjective optimization,
bitubal column, as reported in the literature [20]. while the latter indeed provides designer with a broader range of
design options.

3.2. Multiobjective optimization 3.3. Comparison between optimal foam-lled monotubal and bitubal
structures
Although the constrained single-objective optimization can
generate a better design, it may not be always easy to impose prop- To compare the crashworthiness of other foam-lled structures,
the multiobjective optimization, as dened in Eq. (25), is per-
Table 4
formed for traditional foam-lled monotubal column as well. The
Initial and optimal designs of single-objective optimization for two structures. sectional dimensions of foam-lled monotubal column are the
same as the outer tube of foam-lled bitubal column with
Bitubal t q r SEA Fmax UR E (J)
foam- lled (mm) (kg/m3) (MPa) (kJ/kg) (kN)
b1  b2  t = 82  82  2.0 mm and the same design variables are
column considered. For comparison, the same boundary and loading condi-
tions, as in Fig. 2, are applied.
Initial value 2.0 380 165 12.04 192.3 0.65
Lower bound 0.6 250 140 0 0.67 The Kriging models are constructed for new objective and con-
Upper bound 2.0 500 230 160 straint functions, SEA, Fmax and UR, with respect to design variables
Optimal values 1.413 253 195 13.17 159.3 0.675 9911.4 of t, q, r for the foam-lled monotubal column. Table 5 summa-
Empty bitubal t (mm) r SEA Fmax UR E(J) rizes the errors between the FEA results and the Kriging models
column (MPa) (kJ/kg) (kN) in ve new random sample points. Obviously, the Kriging models
Initial value 2.0 165 13.83 189.78 0.64 provide sufcient accuracy for the design optimization.
Lower bound 0.6 140 0 0.67 By using the Kriging meta-models and NSGA II optimization
Upper bound 2.0 230 160
method, the Pareto fronts of foam-lled monotubal and bitubal
Optimal values 1.814 159.8 12.85 149.8 0.6794 8109.2
columns are plotted in Fig. 10. Interestingly, these two Pareto
106 Y. Zhang et al. / Materials and Design 38 (2012) 99109

-6 -4
double filled tube double filled tube
-7 single objective optimal point single filled tube
-6
-8

-9
SEA (kJ/kg)

SEA (kJ/kg)
-8

-10

-11 -10

-12
-12
-13

-14 -14
50 100 150 200 20 40 60 80 100 120 140 160 180 200
Fmax (kN) Fmax (kN)
Fig. 9. Pareto front of SEA vs. Fmax. Fig. 10. Pareto fronts of Fmax vs. SEA for the foam-lled monotubal and bitubal
columns.

fronts have an intersection, which divides the fronts into two parts.
In the right part of the intersection point, the Pareto front of the
2
bitubal column is lower (i.e. better) than that of the monotubal col- double filled tube
umn. From the safety point of view, this indicates that the bitubal single filled tube
1.8
column allows absorbing more energy under the same level of
peak impact force, or lower peak impact force for absorbing more
1.6
energy. While in the left part of intersection point, the monotubal
Thickness t (mm)

column outperforms. Nevertheless, Fmax is less than 70 kN in this


region, which could be well placed within a safety zone [6668]. 1.4
Thus a higher design demand could be placed for more energy
absorption and allows pushing the peak impact force to a higher le- 1.2
vel (e.g. more to the right part of Pareto front). In this circum-
stance, the bitubal column seems to be of greater potential. 1
Furthermore, since the thickness of tube wall has a signicant
inuence in objective function Fmax, the relationship between Fmax 0.8
and tube thickness t for the Pareto fronts is plotted in Fig. 11. It can
be seen that in order to make Fmax less than the peak force con-
20 40 60 80 100 120 140 160 180 200
straint of 70 kN, the thickness of tube wall should be less than
Fmax (kN)
0.8 mm, while the real thickness of tube wall is typically greater
than 0.8 mm in the automotive applications. From the practical Fig. 11. Curves of Fmax vs. t for foam-lled monotubal and bitubal columns.
point of view, the Pareto fronts of the left part of intersection point
in the Pareto curves (Fig. 10) actually have less engineering impli-
cation. Hence, it allows us to draw a conclusion that the foam-lled the single-objective optimization. Their energy absorption capaci-
bitubal column has a better crashworthiness than traditional ties are further investigated. Fig. 12 compares the deformation pat-
foam-lled monotubal column from the industrial perspective. terns of these three different structures (i.e. Fig. 3ac) under the
Lastly, to explore the interactive inuence between the foam- same boundary and loading conditions. It can be seen that the
ller and empty bitubal column, one optimal design is selected foam-lled bitubal column (Fig. 12a) and empty bitubal column
from the Pareto set for further study. Since multiobjective optimi- (Fig. 12b) both generate some progressive folding as reported in
zation generates a series of optimums, if the designer wishes to the literature, e.g. [5], while the foam-ller can be only compressed
emphasize on the energy absorption, the single-objective optimum in the longitudinal direction with very limited expansion in the lat-
presented in Section 3.1 can be one of such options. The foam-lled eral directions (Fig. 12c).
bitubal column is divided two parts in this optimum point. As Table 6 summarizes the results of the total energy absorption
shown in Fig. 3, the models of the empty bitubal column (E), specic energy absorption (SEA) and maximum impacting force
(Fig. 3b) and foam-ller (Fig. 3c) are considered separately. The de- (Fmax) for the (i) empty bitubal column (Fig. 12b), (ii) foam-ller
sign parameters and load conditions are the same as the solution to (Fig. 12c), (iii) summation of (i) and (ii), and (iv) foam-lled bitubal

Table 5
Accuracy of the Kriging meta-models for monotubal foam-lled column.

No. t (mm) q (kg/m3) r (MPa) SEA (kJ/kg) Fmax (kN) UR


FEA Kriging FEA Kriging FEA Kriging
1 2.0 323 158.9 9.35 8.99 123.7 122.0 0.68 0.681
2 1.587 325 143.5 8.17 7.90 88.46 90.77 0.69 0.689
3 1.123 280 180.0 8.19 8.382 77.54 77.92 0.695 0.6948
4 0.900 480 176.0 5.12 5.35 61.28 60.23 0.698 0.6987
5 0.765 290 160.0 6.91 6.814 47.56 47.88 0.701 0.7016
Y. Zhang et al. / Materials and Design 38 (2012) 99109 107

Fig. 12. Deformation of three different components: (a) foam-lled bitubal column, (b) empty bitubal column, and (c) foam-ller.

Table 6 to deal with multiple loading cases simultaneously. To take into ac-
Crashworthiness results of different separation structures for optimal point. count various uncertainties, such non-deterministic methods as
E (J) SEA (kJ/kg) Fmax (kN)
probability based design optimization (PBDO) and non-probability
based design optimization (NPBDO) can be employed, as reported
(i) Empty bitubal column (Fig. 12b) 6435.7 10.67 159.43
(ii) Foam-ller (Fig. 12c) 2263.3 7.996 67.42
in the literature e.g. [62].
(iii) Empty bitubal column + foam-ller 8699 11.56 226.85
(iv) Foam-lled bitubal column (Fig. 12a) 9911.4 13.17 159.3
4. Conclusion

This paper explored the crashworthiness design for a special


thin-walled structure made of foam-lled squared bitubal col-
umns. The design criteria of specic energy absorption (SEA), peak
impact force (Fmax) and utilization rate (UR) of progressive defor-
mation length were taken into account under dynamic crushing
loading. The Kriging modeling technique was adopted for approx-
imating the response functions. Firstly, to optimize crashworthi-
ness the genetic algorithm (GA) and Non-dominated Sorting
Genetic Algorithm (NSGA) II were applied for the single-objective
and multiobjective optimization, respectively. It is noted that the
optimal result of the single-objective optimization represents a
special point in the Pareto front. From this perspective, the multi-
objective optimization appears more suitable for the crashworthi-
ness design problems that often require addressing a number of
different criteria. Secondly, to compare with the optimized foam-
lled monotubal column, the multiobjective optimization was also
performed and the results showed that the foam-lled bitubal col-
umn has a better crashworthiness than the foam-lled monotubal
column for the wall thickness available practically in vehicle engi-
Fig. 13. Internal energy curves of different components of the structures.
neering. Thirdly, the crashworthiness of empty bitubal column and
foam-ller was separated to explore the role of foam-ller in such
column (Fig. 12a). Fig. 13 plots the energy absorption curves for a more sophisticated bitubal structure. The comparative study
different structural components under the same conditions. demonstrated that the foam-lled bitubal structure is better than
From Table 6, it is noticed that the peak impact force Fmax of the empty bitubal column as well as the sum of two separate com-
foam-lled bitubal column is much lower than the sum of the ponents of the empty bitubal column and foam-ller as their own.
other two components (159.3 kN vs. 226.9 kN). Table 6 and It is concluded that the optimized foam-lled bitubal structure
Fig. 13 also show that the energy absorption of foam-lled bitubal could provide a better crashworthiness performance than the
column (iv) outperforms the sum (iii) of bitubal empty column (i) foam-lled monotubal column and empty bitubal column. It can
and foam-ller (ii) (13.17 vs. 11.56 kJ/kg), mainly due to the fric- be a potential structural component for vehicle engineering
tional interaction between the foam-ller and the inner/outer tube applications.
walls. Therefore, it can be claimed that the combined structure (i.e.
foam-lled bitubal column) has a higher energy absorption capac- Acknowledgments
ity and lower peak crushing force, thereby suitable for improving
the crashworthiness performance for thin-walled structures in This work is supported by The National 973 Project of China
vehicle design. (2010CB328005), The Fundamental Research Funds for the Central
Note that although beyond the scope of this paper, several Universities (JB-ZR1157), The Open Fund of the State Key Labora-
important issues need to be addressed in the future studies. First, tory of Advanced Design and Manufacture for Vehicle Body at Hu-
the uncertainties of foam density, tube materials and geometric nan University 30915003, The Open Fund of Shanghai Key
imperfection could largely affect the crashworthiness responses Laboratory of Digital Autobody Engineering (201004), The Open
and design stability. Second, the impact velocity and direction in Fund of The State Key Laboratory of Vehicle NVH and Safety Tech-
real impact scenarios may not be known and the design may need nology (NVHSKL-201002) and The Open Fund of Key Laboratory for
108 Y. Zhang et al. / Materials and Design 38 (2012) 99109

Automotive Transportation Safety Enhancement Technology of the [32] Toksoy AK, Guden M. Partial Al foam lling of commercial 1050H14 Al crash
boxes: the effect of box column thickness and foam relative density on energy
Ministry of Communication, PRC (CHD2011SY008).
absorption. Thin-wall Struct 2010;48:48294.
[33] Toksoy AK, Guden M. The optimisation of the energy absorption of partially Al
foam-lled commercial 1050H14 and 6061T4 Al crash boxes. Int J
References Crashworthiness 2011;16:97109.
[34] Chen W, Wierzbicki T, Santosa S. Bending collapse of thin-walled beams with
ultralight ller: numerical simulation and weight optimization. Acta Mech
[1] Jiang ZY, Gu MT. Optimization of a fender structure for the crashworthiness
2002;153:183206.
design. Mater Des 2010;31:108595.
[35] Guo LW, Yu JL, Li ZB. Experimental studies on the quasi-static bending
[2] Nagel GM, Thambiratnam DP. A numerical study on the impact response and
behavior of double square tubes lled with aluminum foam. Acta Mech
energy absorption of tapered thin-walled tubes. Int J Mech Sci 2004;46:
2010;213:34958.
20116.
[36] Guo LW, Yu JL. Dynamic bending response of double cylindrical tubes lled
[3] Liu YC. Crashworthiness design of multi-corner thin-walled columns. Thin-
with aluminum foam. Int J Impact Eng 2011;38:8594.
Wall Struct 2008;46:132937.
[37] Song HW, Fan ZH, Yu G, Wang QC, Tobota A. Partition energy absorption of
[4] Seitzberger M, Rammerstorfer FG, Gradinger R, Degischer HP, Blaimschein M,
axially crushed aluminum foam-lled hat sections. Int J Solids Struct
Walch C. Experimental studies on the quasi-static axial crushing of steel
2005;42:2575600.
columns lled with aluminium foam. Int J Solids Struct 2000;37:412547.
[38] Jones N. Energy-absorbing effectiveness factor. Int J Impact Eng 2010;37:
[5] Ahmad Z, Thambiratnam DP. Crushing response of foam-lled conical tubes
75465.
under quasi-static axial loading. Mater Des 2009;30:2393403.
[39] Bi J, Fang HB, Wang Q, Ren XC. Modeling and optimization of foam-lled thin-
[6] Shariati M, Allahbakhsh HR, Saemi J, Sedighi M. Optimization of foam lled
walled columns for crashworthiness designs. Finite Elem Anal Des
spot-welded column for the crashworthiness design. Mechanika 2010;3:106.
2010;46:698709.
[7] Niknejad A, Liaghat GH, Naeini HM, Behravesh AH. Theoretical and
[40] Sun GY, Li GY, Stone M, Li Q. Application two-stage multi-delity optimization
experimental studies of the instantaneous folding force of the polyurethane
procedure for honeycomb-type. Comput Mater Sci 2010;49:50011.
foam-lled square honeycombs. Mater Des 2011;32:6975.
[41] Sun GY, Li GY, Hou SJ, Zhou SW, Li W, Li Q. Crashworthiness design for
[8] Fang H, Solanki K, Horstemeyer MF. Numerical simulations of multiple vehicle
functionally graded foam lled thin-walled structures. Mater Sci Eng A
crashes and multidisciplinary crashworthiness optimization. Int J
2010;527:19119.
Crashworthiness 2005;10:16171.
[42] Reyes A, Hopperstad OS, Hanssen AG, Langseth M. Modeling of material failure
[9] Kamel H, Sedaghati R, Medraj M. Crashworthiness improvement of a pickup
in foam-based components. Int J Impact Eng 2004;30:80534.
trucks chassis frame using the Pareto-front and genetic algorithm. Int J Heavy
[43] Yin H, Wen G, Hou S, Chen K. Crushing analysis and multiobjective
Veh Syst 2011;18:83103.
crashworthiness optimization of honeycomb-lled single and bitubular
[10] Wierzbicki T, Abramowicz W. On the crushing mechanics of thin-walled
polygonal structures. Mater Des 2011;32:444960.
structures. J Appl Mech 1983;50:72739.
[44] Nariman-Zadeh N, Darvizeh A, Jamali A. Pareto optimization of energy
[11] Abramowicz W. Simplied crashing analysis of thin-walled column sand
absorption of square aluminium columns using multi-objective genetic
beams. Eng Trans 1983;29:327.
algorithms. Proc Inst Mech Eng Part B J Eng Manuf 2006;220:21324.
[12] White MD, Jones N, Abramowicz W. A theoretical analysis for the quasi-static
[45] Sun GY, Li GY, Gong ZH, He GQ, Li Q. Radial basis functional model for
axial crushing of top-hat and double-hat thin-walled sections. Int J Mech Sci
multi-objective sheet metal forming optimization. Eng Optim 2011;43:
1999;41:20933.
135166.
[13] Gameiro CP, Cirne J. Dynamic axial crushing of short to long circular
[46] Deshpande VS, Fleck NA. Isotropic constitutive models for metallic foams. J
aluminium tubes with agglomerate cork ller. Int J Mech Sci 2007;49:
Mech Phys Solids 2000;48:125383.
102937.
[47] Reyes A, Hopperstad OS, Berstad T, Hanssen AG, Langseth M. Implementation
[14] Meguid S, Attia MS, Monfort A. On the crush behavior of ultra-light foam-lled
of a constitutive model for aluminum foam including fracture and statistical
structures. Mater Des 2004;25:1839.
variation of density. In: 8th International LS-DYNA users conference,
[15] Aktay L, Kroplin BH, Toksoy AK, Guden M. Finite element and coupled nite
Michigan; 2004.
element/smooth particle hydrodynamics modeling of the quasi-static crushing
[48] Zhang ZH, Liu ST, Tang ZL. Crashworthiness investigation of Kagome
of empty and foam-lled single, bitubular and constraint hexagonal- and
honeycomb sandwich cylindrical column under axial crushing loads. Thin-
square-packed aluminum tubes. Mater Des 2008;29:95262.
Wall Struct 2010;48:918.
[16] Santosa S, Wierzbicki T. Crash behavior of box columns lled with aluminum
[49] Reyes A, Hopperstad OS, Berstad T, Hanssen AG, Langseth M. Constitutive
honeycomb or foam. Comput Struct 1998;68:34367.
modeling of aluminum foam including fracture and statistical variation of
[17] Santosa S, Wierzbicki T. Effect of an ultralight metal ller on the bending
density. Eur J Mech 2003;22:81535.
collapse behavior of thin-walled prismatic columns. Int J Mech Sci
[50] Hanssen AG, Hopperstad OS, Langseth M. Validation of constitutive models
1999;41:9951019.
applicable to aluminum foams. Int J Mech Sci 2002;44:359406.
[18] Zarei HR, Kroger M. Optimum honeycomb lled crash absorber design. Mater
[51] Guo LW, Yu JL. Experimental investigations on the quasi-static crushing of
Des 2008;29:193204.
foam lled double square columns. J Exp Mech 2010;25:2718.
[19] Alavi NA, Sadeghib MZ. The effects of foam lling on compressive response of
[52] Wang Q, Fan Z, Song H, Gui L. Experimental and numerical analyses of the axial
hexagonal cell aluminum honeycombs under axial loading-experimental
crushing behaviour of hat sections partially lled with aluminium foam. Int J
study. Mater Des 2010;31:121630.
Crashworthiness 2005;10:53543.
[20] Zhang X, Cheng GD, Wang B, Zhang H. Optimum design for energy absorption
[53] Fang H, Rais-Rohani M, Liu Z, Horstemeyer MF. A comparative study of
of bitubal hexagonal columns with honeycomb core. Int J Crashworthiness
metamodeling methods for multi-objective crashworthiness optimization.
2008;13:99107.
Comput Struct 2005;83:212136.
[21] Hanssen AC, Langseth M, Hopperstad OS. Static and dynamic crushing of
[54] Stander N, Roux WJ, Giger M, Redhe M, Fedorova N, Haarhoff J.
circular aluminium extrusions with aluminium foam ller. Int J Impact Eng
Crashworthiness optimization in LS-OPT: case studies in meta-modeling and
2000;24:475507.
random search techniques. In: proceedings of the 4th European LS-DYNA
[22] Hanssen AG, Langseth M, Hopperstad OS. Static crushing of square aluminum
conference, Ulm, Germany; May 2223, 2003.
extrusions with aluminum foam ller. Int J Mech Sci 1999;41:96793.
[55] Simpson Timothy W, Mauer M, Korte John J, Mistree Farrokh. Kriging models
[23] Hanssen AG, Langseth M, Hopperstad OS. Static and dynamic crushing of
for global approximation in simulation-based multidisciplinary design
square aluminium extrusions with aluminium foam ller. Int J Impact Eng
optimization. AIAA J 2001;39:223341.
2000;24:34783.
[56] Liao XT, Li Q, Yang XJ, Zhang WG, Li W. Multiobjective optimization for crash
[24] Zhang X, Cheng GD, You Z. Energy absorption of axially compressed thin-
safety design of vehicle using stepwise regression model. Struct Multidiscip
walled square tubes with patterns. Thin-Wall Struct 2007;45:73746.
Optim 2008;35:5619.
[25] Zarei HR, Kroger M. Crashworthiness optimization of empty and lled
[57] Liao XT, Li Q, Yang XJ, Li W, Zhang WG. A two-stage multi-objective
aluminum crash boxes. Int J Crashworthiness 2007;12:25564.
optimisation of vehicle crashworthiness under frontal impact. Int J
[26] Zarei HR, Kroger M. Optimization of the foam-lled aluminum tubes for crush
Crashworthiness 2008;13:27988.
box application. Thin-Wall Struct 2008;46:21421.
[58] Franulovic M, Basan R, Prebil I. Genetic algorithm in material model
[27] Hou SJ, Li Q, Long SY, Yang XJ, Li W. Crashworthiness design for foam lled
parameters identication for low-cycle fatigue. Comput Mater Sci 2009;45:
thin-wall structures. Mater Des 2009;30:202432.
50510.
[28] Zhang X, Huh H. Energy absorption of longitudinally grooved square tubes
[59] Ledoux Y, Sebastian P, Samper S. Optimization method for stamping tools
under axial compression. Thin-wall Struct 2009;47:146977.
under reliability constraints using genetic algorithms and nite element
[29] Ahmad Z, Thambiratnam DP. Dynamic computer simulation and energy
simulations. J Mater Process Technol 2010;210:47486.
absorption of foam-lled conical tubes under axial impact loading. Comput
[60] Deb K. Multiobjective optimization using evolutionary algorithms. John Wiley
Struct 2009;87:18697.
& Sons; 2001.
[30] Ahmad Z, Thambiratnam DP. Application of foam-lled conical tubes in
[61] Deb K, Pratap A, Agarwal S, Meyarivan T. A fast and elitist multiobjective
enhancing the crashworthiness performance of vehicle protective structures.
genetic algorithm: NSGA-II. IEEE Trans Evolut Comput 2002;6:18297.
Int J Crashworthiness 2009;14:34963.
[62] Sun GY, Li GY, Zhou SW, Li HZ, H SJ, Li Q. Crashworthiness design of vehicle by
[31] Ahmad Z, Thambiratnam DP, Tan A. Dynamic energy absorption characteristics
using multiobjective robust optimization. Struct Multidiscip Optim 2011;44:
of foam-lled conical tubes under oblique impact loading. Int J Impact Eng
99110.
2010;37:47588.
Y. Zhang et al. / Materials and Design 38 (2012) 99109 109

[63] Mirfendereski L, Salimi M, Ziaei-Rad S. Parametric study and numerical [66] Kurtaran H, Eskandarian A, Marzougui D. Crashworthiness design optimization
analysis of empty and foam-lled thin-walled tubes under static and dynamic using successive response surface approximations. Comput Mech 2002;29:
loadings. Int J Mech Sci 2008;50:104257. 40921.
[64] Zhang Y, Sun GY, Li GY, Li Q. Identication of material parameters for high [67] Hou SJ, Li Q, Long SY, Yang XJ, Li W. Multiobjective optimization of multi-cell
strength steel under impact loading. Adv Sci Lett 2011;4:70814. sections for the crashworthiness design. Int J Impact Eng 2008;35:135567.
[65] Redhe M, Forsberg J, Jansson T, Marklund PO, Nilsson L. Using the response [68] Hou SJ, Li Q, Long SY, Yang XJ, Li W. Design optimization of regular hexagonal
surface methodology and the D-optimality criterion in crashworthiness thin-walled columns with crashworthiness criteria. Finite Elem Anal Des
related problems. Struct Multidiscip Optim 2002;24:18594. 2007;43:55565.

View publication stats

You might also like