Vous êtes sur la page 1sur 16

Original Article

Proc IMechE Part D:


J Automobile Engineering
2016, Vol. 230(2) 273–288
Design and application of lightweight Ó IMechE 2015
Reprints and permissions:
multi-objective collaborative sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0954407015581937

optimization for a parametric pid.sagepub.com

body-in-white structure

Chuan-Qing Wang, Deng-Feng Wang and Shuai Zhang

Abstract
The static bending and torsional stiffnesses, the lower-order modal frequencies of a body-in-white structure and the full
frontal-crash and side-impact passive safety performances are simulated with finite element models which are generated
on the basis of the implicit parametric model. The implicit parametric model is established through SFE CONCEPT soft-
ware. The simulation results are compared with tests to validate the simulation analysis results. It is proposed that the
multi-objective optimization is divided into non-safety parts optimization, frontal-crash safety parts optimization and
side-impact safety parts optimization, which is computationally more efficient than optimizing the non-safety parts, the
frontal-crash safety parts and the side-impact safety parts simultaneously. In this paper, the lightweight multi-objective
collaborative optimization design of the body-in-white structure is conducted for a passenger car by optimizing the thick-
ness, the beam section shape and the size; while maintaining the performances of the static bending and torsional stiff-
nesses, the lower-order modal frequency decreases to less than 5% of the initial value, and the full-frontal-crash and
side-impact passive safety performances remain almost the same. Structural modifications are applied by means of impli-
cit parametric technology, providing changes in the geometry in a fully controllable manner. After comparison between
the optimized body-in-white structure and the initial structure, the mass decreased in total by 32.41 kg (i.e. by as high as
7.63%). The decreases in the performances of the bending and torsional stiffnesses are less than 2.54%; the bending and
torsional frequencies increased a little, and the frontal-crash and side-impact passive safety performances underwent
almost no change.

Keywords
Implicit parametric body-in-white structure, lightweight design, multi-objective collaborative optimization

Date received: 27 October 2014; accepted: 20 March 2015

Introduction contains topology optimization, size optimization,


shape optimization, single-objective optimization and
Recently, laws and regulations on the emissions, the multi-objective optimization.
fuel efficiency and the protection of the environment The aim of topology optimization is to improve the
have become stricter. Design of a lightweight vehicle is allocation efficiency of the material of the structure
one of the easiest methods to preserve energy and to within the design domain. In recent decades, much
reduce gas emissions. The mass of the BIW structure is
about 30% of the mass of the whole vehicle.1
Therefore, a reduction in the mass of the BIW structure State Key Laboratory of Automotive Simulation and Control, Jilin
plays an important role in decreasing the mass of the University, Changchun, Jilin, People’s Republic of China
whole vehicle.
Although multi-material mixes, assembling tech- Corresponding author:
Deng-Feng Wang, State Key Laboratory of Automotive Simulation and
niques and new manufacturing are promising areas,
Control, Jilin University, Changchun, Jilin 130022, People’s Republic of
one of the main approaches to lightweight design is China.
structure optimization.2 Structure optimization often Email: 419170738@qq.com
274 Proc IMechE Part D: J Automobile Engineering 230(2)

work has been carried out by topology optimization.3–5 detailed-design multi-disciplinary optimization. In the
However, topology optimization is mainly focused on conceptual-design optimization, they determined the
the concept design phase, for which the size of the make-up of the local composite laminate and the
structure is not known accurately. Thus, topology opti- boundaries between different laminate patches using
mization, size optimization and shape optimization the free element sizing technique. In the detailed-design
work together to determine the structure.6 The size multi-disciplinary optimization, they took the car-door
optimization and shape optimization methods have panel mass as the objective and made sure that the
been extensively studied and utilized.7 These methods NVH, the crashworthiness and the durability met the
consider parameters such as the plate thickness and the requirements. Finally, they obtained the ply thickness
beam cross-section (e.g. the height and the width) as and the inter-patch boundaries of the door panel.
the design variables to conduct optimization design.8 In These research studies are very helpful in developing
the process of optimization, some geometrical shapes a new part structure of new material. So far, most stud-
may be changed. In order to solve this problem, the ies of BIW structure optimization have been mainly
morphing approach has been applied through software limited to the field of size optimization and utilize the
such as MeshWorks. The morphing approach takes the performance of crashworthiness as the constraint for
dimensions of the geometrical shape as variables in the single-objective or multi-objective optimization.17–19
process of optimization. It is used as a significant per- Studies which employ the thickness, the size and the
formance enhancement tool, which was originally used beam section shape as variables and synthetically con-
in the aerospace domain,9–11 and then subsequently in sider the performances of the static stiffnesses, the low-
the vehicle industry, which attracted much attention. In order natural modal frequencies and the frontal-crash
the vehicle industry, it has been used to improve the and side-impact passive safety performances of the
performances of the aerodynamics, the static stiffness BIW structure are rare. This may obviously cause some
and the crashworthiness.12,13 This technique is rela- performances of the BIW structure to decrease. While
tively limited since the quality of the finite element (FE) this paper is based on a parametric BIW model in order
mesh rapidly decreases as the amount of shape varia- to conduct lightweight design. The aim is obtained
tion increases, and the adjacent parts cannot change through seeking the optimal combination of the thick-
with the changing parts. Although remeshing technol- ness, the local section and the geometrical shapes. In
ogy is provided for the changed shape, the quality is the optimization, the BIW model is divided into non-
poor in comparison with that of the parametric model. safety parts, frontal-crash safety parts and side-impact
The implicit parametric model allows larger geometri- safety parts. The optimization includes three phases
cal modifications, and the adjacent parts change with based on these parts. This method can improve the
the shape variation.14 The mesh of the implicit para- computational efficiency and is described in more detail
metric model is generated at each new generation, and in the third section.
so the quality is good. Therefore, the body-in-white After each optimization, the optimized model is
(BIW) model used in this paper is an implicit para- compared with the initial model. Also, by maintaining
metric model which is established by SFE CONCEPT the performances of the static bending and torsional
software. stiffnesses, the lower-order modal frequencies decrease
Grujicic et al.15 introduced topology optimization, by less than 5% of the initial values, and the full-fron-
size optimization and shape optimization. Then they tal-crash and side-impact passive safety performances
obtained a new optimal structure with a polymer–metal remain almost the same. However, in the process of
hybrid material based on the stiffness and the strength, multi-objective optimization, conflicting objectives
which was subject to bending. They also obtained the arise mostly. Therefore, one performance is increased
optimal structure by linearized eigenvalue buckling while another performance may decrease. The non-
analysis and non-linear buckling analysis based on the dominated sorting genetic algorithm (NSGA-II) when
buckling resistance, which was subject to axial com- used for the multi-objective optimization problem
pression loading. They found that the difference in the (MOP) is a quite important method, which can deal
geometrical structure obtained from the linearized anal- with the above contradiction efficiently.
ysis and the non-linear analysis is relatively small.
Grujicic et al.16 presented a multi-disciplinary optimiza-
tion methodology, and they considered an inner door Performance verification of the implicit
panel as an example for carrying out lightweight design parametric BIW model
based on this methodology. In the process of optimiza-
tion, the number and orientation of the composite In the process of establishing the parametric BIW
piles, the thickness of the local laminate and the shape model, it is usually divided into three processes. The
of the panel were taken as variables, and the perfor- first process is to define the base points and the base-
mances of noise vibration and harshness (NVH), the lines to indicate the position of the component ledges.
durability, the crashworthiness and the manufacturabil- The second process is to define the beam section and to
ity were the constraints. Their research studies were assign the lines to it. Finally, the beams and compo-
divided into conceptual-design optimization and nents are created, and the corresponding components
Wang et al. 275

are mapped; then the implicit parametric model is fin- implicit parametric BIW model was correct. In testing,
ished. The structure can be easily modified by changing the air pressure of the air spring was adjusted to make
some base point coordinates, baselines and local sure that the rigid vibration frequency of the BIW on
sections.20 the air spring brackets (shown in Figure 2) was less
The implicit parametric BIW model is shown in than 3 Hz. The air spring was placed on the testbed,
Figure 1. Before multi-objective collaborative optimiza- and the BIW structure was made horizontal using a
tion design, the FE model was generated on the basis wood block.
of the implicit parametric model. The low-order modal The BIW object was tested in the frequency range
frequencies, the static bending and torsional stiffnesses, for a burst random signal of 1–256 Hz which was gen-
the full-frontal-crash and side-impact passive safety erated with an electromagnetic vibrator and amplified
performances were simulated and compared with tests with a power amplifier. It helped to analyse the global
to verify that the BIW model was correct. vibration frequencies and modes through LMS
Test.Lab structure testing software.19 The vibrator
force was given to two points of the left front longitudi-
Verification of the low-order modal frequencies of the nal beam and right rear longitudinal beam, which are
implicit parametric BIW model shown in Figure 3. The rear excitation force was verti-
cally upwards, and the front excitation force had a gra-
The low-order modal frequencies of the BIW model
dient angle in the lateral direction and the longitudinal
were analysed by NASTRAN, and the simulation
direction, which can fully indicate the three direction
results were compared with the tests to verify that the
modes of the BIW structure.
The vibration acceleration response of the BIW
structure was collected through 180 standard acceler-
ometers which were pasted on the BIW. The geometry
of the physical BIW structure was created according to
the coordinates of 180 accelerometers, which are shown
in Figure 4. The BIW geometry in Figure 4 was able to
exhibit vibration modes in the tested BIW. These accel-
erometers provided signals in three directions to LMS
Test.Lab. The ‘PolyMAX’ of LMS Test.Lab used the
Figure 1. Implicit parametric BIW model. poly-reference least-squares complex frequency-domain
method which could construct a stabilization diagram
(shown in Figure 5) and identified the vibration modes
through stable poles.21 A comparison of the results on
the principal modal frequencies for the tests and the
simulations are shown in Table 1. As can be seen from
Table 1, the relative errors of the simulations and the
tests were less than 7.00%. Thus, it is acceptable to use
this parametric BIW model to conduct modal analysis.
The rear torsional mode and the first-order bending
mode were overall modal, and the relative errors were
very small. Therefore, they were used as constraints for
lightweight optimization design.
Figure 2. The free support of the BIW structure.

Figure 3. The vibrator location of the BIW structure.


276 Proc IMechE Part D: J Automobile Engineering 230(2)

Table 1. Comparisons of the results for the major modal frequencies obtained from the tests and the simulations on the BIW
structure.

Value for the following modes


Rear torsional mode Frontal torsional mode First-order bending mode

Major modal frequency, simulations (Hz) 30.33 36.11 51.54


Major modal frequency, tests (Hz) 30.26 38.83 51.84
Relative error 0.23% 7.00% 0.58%

Figure 6. Loading forces of the bending mode.

Figure 4. The BIW geometry in LMS Test.Lab.

Figure 7. Loading forces of the torsional mode.

Figure 5. The stabilization diagram.


Figure 8. Locations of the constraint points.

Verification of the static bending and torsional suspensions (points C and D). The equal but opposite
stiffnesses of the implicit parametric BIW structure forces exerted on the front suspensions (points A and
B) are shown in Figure 7. The force was 1703.5 N,
The static bending and torsional stiffnesses of the BIW
which is equivalent to a torque of 2 kN m. The loca-
structure were analysed using NASTRAN, and the
tions of points A, B, C and D are shown in Figure 8.
simulation results were compared with the tests to ver-
The bending stiffness and the torsional stiffnesses are
ify that the implicit parametric BIW structure was cor-
expressed as
rect. In the simulations of the static bending stiffness,
the front suspensions (points A and B) and the rear sus- 4000
pensions (points C and D) were all constrained. Static Bending stiffness = ð1Þ
jZ1 j + jZ2 j
vertical forces of 2 kN were exerted on the body floor
around the B pillars, as shown in Figure 6. In the simu- and
lations of the static torsional stiffness, the BIW struc-   
ture was constrained except for the Z direction of the 1 jZ3 j + jZ4 j
Torsional stiffness = 2000 tan ð2Þ
front suspensions (points A and B) and all the rear L
Wang et al. 277

Figure 9. Constraints of the BIW structure when testing.

Table 2. Comparisons of the results for the stiffnesses


obtained from the tests and the simulations.

Bending stiffness Torsional stiffness


(N/mm) (N m/deg)

Simulations 16 071.86 18 308.32 Figure 10. The vehicle model, the chassis and the powertrain.
Tests 15 052.24 18 937.79
Relative error 6.70% 3.30%

Car Assessment Program (C-NCAP). A full frontal


respectively, where Z1, Z2, Z3 and Z4 are the displace- crash is a vehicle crash into a full-width barrier with a
ment values of the loading location and L is the lateral velocity of 50 km/h. In the simulations, the parametric
distance between A and B. BIW structure was connected with the chassis and the
The test analyses were conducted under the same powertrain, as shown in Figure 10. The simulation
restrictions and loadings as the simulation analyses. analysis was carried out using LS-DYNA software.
When testing the static bending and torsional stiff- The vehicle needed a counterweight so that it could
nesses, two adjustable height loading brackets with be considered without a body trim decoration, a
force sensors were used to constrain the front- dummy, etc. After the counterweight was added, com-
suspension shock tower of the BIW structure, and two parisons of the masses and the mass centre positions
fixed brackets were used to constrain the rear- for the physical vehicle and the FE vehicle are shown
suspension spring holder, as shown in Figure 9. The in Table 3. As can be seen from Table 3, the relative
loading brackets were linked with the front suspension errors of the masses and the mass centre positions were
through bolts. The fixed brackets were linked through less than 2.01%, which accorded with the full-frontal-
bolts with the rear longitudinal beam, which was adja- crash requirement.
cent to the rear suspension. When testing the static tor- The upper left front-door hinge (ULH) intrusion,
sional stiffness, the left loading bracket was adjusted to the lower left front-door hinge (LLH) intrusion, the
an upper location, and the right loading bracket to a upper right front-door hinge (URH) intrusion, the
lower location, to make sure that the values shown by lower right front-door hinge (LRH) intrusion, the
the two force sensors were 1703.5 N. When testing the deformation mode and the acceleration curves of B pil-
static bending stiffness, it was important to make sure lars of both sides of the BIW structure are extracted in
that the BIW structure was horizontal, which kept the the simulations and compared with the tests. Figure 11
values shown by the two force sensors equal. shows a comparison of the deformation modes for the
A comparison of the stiffness results for the simula- simulations and the tests. It can be seen that the defor-
tions and the tests are shown in Table 2. As can be seen mation modes were identical. This comparison was
from Table 2, the relative errors of the static bending only visual. It showed only that the frontal crash had
stiffness and the torsional stiffness were 6.70% and no problems of fatalities. More detailed comparisons
3.30% respectively. Therefore, the implicit parametric are shown in Figure 12 and Table 4.
BIW model is acceptable for conducting bending and As can be seen from Figure 12, the acceleration
torsional analyses. curves of both sides of the vehicle for the simulations
and the tests fitted each other well. The peak value of
Verification of the full-frontal-crash safety the acceleration was 33g, where g is the acceleration
performance of the implicit parametric BIW model due to gravity. The crash time was 78 ms.
In this paper, both simulations and tests on a full fron- Comparisons of the front-door hinge intrusions for
tal crash were carried out according to the China New the simulations and the tests are shown in Table 4. As
278 Proc IMechE Part D: J Automobile Engineering 230(2)

Table 3. Comparisons of the masses and the mass centre positions of the physical car and the FE model.

Value for the following


Physical vehicle FE model Relative error

Mass (kg) 1603.80 1606.01 0.14%


Mass centre coordinate X (mm) 1040.76 1042.49 1.65%
Mass centre coordinate Y (mm) 7.00 6.86 2.01%
Mass centre coordinate Z (mm) 203.76 203.86 0.05%

FE: finite element.

Table 4. Comparisons of the front-door hinge intrusion values obtained from the tests and the simulations.

Intrusion (mm) of the following hinges


UHL LHL UHR LHR

Simulations 2.86 3.16 1.30 1.77


Tests 2.68 3.51 1.19 1.96
Relative error 6.72% 9.97% 9.24% 9.69%

UHL; upper left front-door hinge; LHL: lower left front-door hinge; LHR: lower right front-door hinge; UHR: upper right front-door hinge.

was again needed. After the counterweight was added,


comparisons of the masses and the mass centre posi-
tionsc for the physical vehicle and the FE model are
shown in Table 5. As can be seen from Table 5, the rela-
tive errors of the masses and the mass centre positions
were less than 4.89%. Therefore, the implicit para-
metric BIW model and the connected parts are accepta-
Figure 11. Comparison of the deformation modes for the ble for conducting side-impact analysis.
simulations and the tests.
In this paper, the moving deformable barrier is a
commercial model (shown in Figure 13) from
Engineering Technology Associates, Inc.
Figure 14 show a comparison of the deformation
modes for the simulations and the tests. This was a
visual comparison only to show that there was no prob-
lems of fatalities. Therefore, comparisons of the intru-
sion velocity curves and the acceleration curves were
needed. The intrusion velocity curves, the acceleration
curves of the head location, the beltline location, the
chest location and the H point of the B pillar and the
deformation modes were extracted from the simula-
Figure 12. Comparison of the acceleration curves of the B tions and compared with the tests.
pillars for the simulations and the tests. A comparison of the corresponding points of the B-
pillar intrusion velocity curves for the simulations and
can be seen from Table 4, the maximum intrusion was the tests are shown in Figure 15. As can be seen in
the lower left door intrusion, and the relative errors Figure 15, the maximum deviation was about 1.5 m/s
were less than 9.97%. Therefore, the implicit para- at the location of the H point. The minimum deviation
metric BIW model and the connected parts are accepta- was at the head location of the B pillar. The curves
ble for conducting full-frontal-crash analysis. fitted each other well.
A comparison of the corresponding points of the B-
pillar acceleration curves for the simulations and the
Verification of the side-impact safety performance of
tests are shown in Figure 16. As can be seen in Figure
the implicit parametric BIW model 16, before 30 ms all the acceleration curves fitted each
Side-impact analyses of both simulations and tests were other well, then the peaks and the troughs did not coin-
made according to C-NCAP. Because the types and the cide and finally they all tended to zero. The implicit
numbers of dummies in a side impact are different from parametric BIW model and the connected parts are
those in a full frontal crash, therefore, a counterweight acceptable for conducting side-impact analysis.
Wang et al. 279

Table 5. Comparisons of the masses and the mass centre positions of the the physical car and the FE model.

Value for the following


Physical vehicle FE model Relative error

Mass (kg) 1551.80 1548.90 0.19%


Mass centre coordinate X (mm) 1150.55 1137.67 1.12%
Mass centre coordinate Y (mm) –100.94 –96.64 4.26%
Mass centre coordinate Z (mm) 234.28 222.83 4.89%

FE: finite element.

Lightweight optimization design for the


implicit parametric BIW model
This paper divided the components of the BIW model
into non-safety parts, frontal-crash safety parts and
side-impact safety parts. Thus, the optimization can be
divided into three phases: non-safety parts optimiza-
tion, frontal-crash safety parts optimization and side-
Figure 13. The moving deformable barrier model. impact safety parts optimization. This method can
reduce the calculation time in comparison with the cal-
culation time when these parts are considered together.
For example, in this paper, the method takes only
21.17% of the time requires when the parts are consid-
ered together. The number of non-safety part variables
is 73, the number of frontal-crash part variables is 13
and the number of side-impact part variables is 11. The
following times for calculation are required: 1.5 h for
the static bending and torsional stiffnesses; 2 h for the
low-order modal frequencies; 15 h for the frontal-crash
Figure 14. Comparison of the deformation modes for the variables; 20 h for the side-impact variables. If the
simulations and the tests. radial basis functions (RBFs) surrogate model is

Figure 15. Comparison of the corresponding points of the B-pillar intrusion velocity curves for the simulations and the tests.
280 Proc IMechE Part D: J Automobile Engineering 230(2)

Figure 16. Comparison of the corresponding points of the B-pillar acceleration curves for the simulations and the tests.

adopted, at least 2n + 1 sample points are required.


Therefore, this method needs a calculation time of only
1589.5 h, in contrast with the calculation time of
7507.5 h when the parts are considered together.
For comprehensive consideration of the structural
optimization design for crashworthiness, static stiff-
nesses and natural vibration properties, the computa-
tion time was extremely expensive. The surrogate model
can reduce the computational cost.22 The Latin hyper-
cube samples (LHSs) were considered to have good
space-filling properties and some degree of symmetry.23
When the number of samples ranged from 20 to 500,
the optimal LHS algorithm was recommended.24 The Figure 17. The sampling strategy: (a) the LHS method; (b) the
optimal LHS method can represent the higher order of optimal LHS method.
non-linearity using relatively fewer sample points. To
ensure the uniformity of the sampling points, a combi-
natorial optimization algorithm was considered with an X
n
f0 ðxÞ = li uðkx  xi kÞ
entropy criterion to minimize the bias of the mean i=1
square error.25 Figure 17 shows a comparison between
the LHS method and the optimal LHS method to illus- where n is the number of sampling points, x is the vec-
trate the sampling strategy. tor of design variables, xi is the vector of design vari-
Response surface methodology (RSM) and RBFs ables at the ith sampling point, kx  xi k is the
are two commonly used methods in constructing a sur- Euclidean distance, u is a basis function and li is an
rogate model. Fang at al.26 compared RSM and RBFs unknown weighting coefficient. Therefore, an RBF is
for multi-objective optimization of a BIW structure in in fact a linear combination of n basis functions with
the field of impact and found that RBFs performed bet- weighted coefficients. Some of the most commonly used
ter for optimization of highly non-linear objectives. The basis functions include thin-plate spline, Gaussian,
RBFs are a series of basic functions that are symmetric multi-quadric and inverse multi-quadric functions.
and centred at each sampling point and were originally In recent years, the NSGA-II has been widely used
developed for scattered multi-variate data interpola- when dealing with multi-objective optimization.27
tion. Let f(x) be the true value of the response function Therefore, the optimal LHS was adopted to build RBF
and f 0 (x) the approximate value obtained from a classi- neural network surrogate models for each optimiza-
cal RBF with the general form tion. The MOP presented in this paper used a
Wang et al. 281

non-normalized method, which was NSGA-II. NSGA-


II employs both the elite-preserving strategy as well as
an explicit diversity-preserving mechanism. The algo-
rithm first randomly generates a population of prede-
fined size N, which undergoes conventional selection,
crossover and mutation procedures to produce off-
spring for the next generation with better solutions.
From the second generation, the parent generation
(size N) competes with its offspring generation (size N) Figure 18. Non-safe part design variables.
to introduce elitism.
Multi-objective optimizations of the three phases
were carried out through the software ISIGHT. All the fðxÞ = fmin½fðMÞ, max½fðTÞg ð4Þ
solutions were searched for using NSGA-II, which
and the design variables are given by
could be evaluated and compared graphically with
Pareto frontier sets.28,29 ISIGHT can also recommend  23t0i 4ti  t0i 413t0i ð5Þ
one solution through a scale factor which was deter-
mined by users.  13t0i 4ti  t0i 423t0i ð6Þ

subject to
Non-safety part optimization for the implicit
0:95ff1 ðBÞ, f1 ðBMÞ, f1 ðTMÞg4fðBÞ, fðBMÞ, fðTMÞ ð7Þ
parametric BIW model
The variable parts were selected through relative sensi- where f(M) and f(T) are the mass of the BIW and the
tivity analysis which was represented by the ratio of the mass of the static torsion stiffness respectively, ti is the
direct sensitivity of the stiffness and the modal fre- thickness of the components, t0i is the initial thickness
quency to the direct sensitivity of the mass, as given by of the components, f1(B), f1(BM) and f1(TM) are the
initial static bending stiffness, the initial first-order
 
  Sb , St , Sfb , Sft bending modal frequency and the initial first-order tor-
RSb , RSt , RSfb , RSft = ð3Þ sional modal frequency respectively and f(B), f(BM)
Sm
and f(TM) are the static bending stiffness, the first-
where RSb, RSt, RSfb and RSft are the relative sensitiv- order bending modal frequency and the first-order tor-
ities of the static bending stiffness, the static torsional sional modal frequency of the sample points. The
stiffness, the first-order bending modal frequency and design space is shown in equation (5); the smaller rela-
the first-order torsional modal frequency respectively, tive sensitivities of 12 design variables were chosen to
Sb, St, Sfb and Sft are the direct sensitivities of the static increase the thickness, and the design space is shown in
bending stiffness, the static torsional stiffness, the first- equation (6). The non-safety parts were mainly the
order bending modal frequency and the first-order tor- panels of the passenger compartment and had little
sional modal frequency respectively and Sm is the direct influence on the crashworthiness. When optimizing, the
sensitivity of mass. The non-safety part design variables objective functions were defined as the minimum mass
were selected according to the following criteria. and the maximum torsional stiffness in equation (4).
The constraint conditions were defined as the static
1. Select the intersection set of all the relative sensitiv- bending stiffness, and the first-order bending modal
ities of these performances. frequencies were not less than 95% of the initial values,
2. Increase the thickness of the design variables to as shown in equation (7).
enhance the stiffness and the modal frequency per- The generation process of the sample points by the
formances if the relative sensitivities are high. design-of-experiments (DOE) method is shown in
3. Decrease the thickness of the design variables to Figure 19. The frontal crash and the side impact were
reduce the BIW mass if the relative sensitivities are non-linear, and the maximum acceleration peak and
low. the maximum intrusion were changing with time.
4. Do not select as a design variable if the mass of a Therefore, the performances were summarized into
single part is less than 0.5 kg. text, and the multi-objective optimization design pro-
cess was based on the text, as shown in Figure 20.
Through relative sensitivity analysis, the design Although non-safety parts optimization did not con-
variables were divided into variables increasing the sider crashworthiness, the multi-objective optimization
thickness and variables reducing the thickness. The design process in Figure 20 was also adopted. Through
design variables are shown in Figure 18. The larger iterations the front Pareto sets were obtained. Owing to
relative sensitivity of 61 design variables were chosen frontal-crash safety parts optimization and the subse-
to reduce the thickness. The MOP for non-safety quent side-impact safety parts optimization, the static
parts optimization can be expressed as follows. The stiffnesses were reduced more. Therefore, in this paper
objective is the middle location of the torsional stiffness in the
282 Proc IMechE Part D: J Automobile Engineering 230(2)

Table 6. Comparisons of the masses and other results for the N-S model and the initial model.

Mass (kg) Bending stiffness Torsional stiffness Bending frequency Torsional frequency
(N/mm) (N m/deg) (Hz) (Hz)

Initial model 424.75 16 071.86 18 308.32 51.54 30.33


N-S model 398.26 16 061.49 18 243.47 52.02 30.39
Change –6.24% –0.06% –0.35% + 0.93% + 0.20%

N-S model: non-safety parts optimized model.

Figure 21. The variables of the frontal-crash safety parts: (a)


Figure 19. The generation process of the sample points. thickness variables; (b) shape variables.
DOE: design of experiments.

part variables were the parameters for the bumper, the


frontal longitudinal beam, etc., and were chosen mainly
because of their contributions to frontal-crash safety
based on experience. The number of thickness variables
of the frontal-crash safety parts is 13, which are shown
in Figure 21(a). The MOP for the frontal-crash safety
parts can be expressed as follows. The objective is

fðxÞ = fmin½fðMÞ, min½fðLHLÞ, min½fðLHRÞg ð8Þ

and the design variables are given by

Figure 20. The multi-objective optimization design process.  204xi  x0i 420 ð9Þ

Pareto sets was chosen as the optimization design solu- subject to


tion. After optimization, the optimization results were
0:95ff1 ðBÞ, f1 ðTÞ, f1 ðBMÞ, f1 ðTMÞg4fðBÞ,
assigned to the corresponding parts and the perfor- ð10Þ
mances were analysed. Comparisons of the masses and fðTÞ, fðBMÞ, fðTMÞ
other results for the non-safety parts optimized model 1:05ffðUHLÞ, fðUHRÞg4f1 ðUHLÞ, f1 ðUHRÞ ð11Þ
(the N-S model) and the initial model are shown in
Table 6. As can be seen from Table 6, the BIW mass where f (LHL), f (LHR), f (UHL) and f (UHR) are the
was reduced by 26.49 kg (i.e. by 6.24%), the static lower left front-door hinge intrusion, the lower right
bending stiffness was reduced by 0.06% and the static front-door hinge intrusion, the upper left front-door
torsional stiffness was reduced by 0.35%. Although the hinge intrusion and the upper right front-door hinge
static bending and torsional stiffnesses decrease, the intrusion respectively of the sample points, f1(UHL)
BIW mass decreased more. Thus, the first-order bend- and f1(UHR) are the upper left front-door hinge intru-
ing modal frequency and the first-order torsional sion and the upper right front-door hinge intrusion
modal frequency increased. In comparison with the tor- respectively of the initial model, xi is the shape variable
sional modal frequency, the bending modal frequency coordinate value of the beam section in the x-axis direc-
increased more. tion and x0i is the initial coordinate value of the beam
section in the x-axis direction. The design space is indi-
cated in equation (5). The number of section shape
Frontal-crash safety parts optimization for the
variables of the longitudinal beam is 2, as shown in
parametric BIW model Figure 21(b), and the design space of the section in
Based on the N-S model, the frontal-crash safety parts the longitudinal direction (the X direction) was from
optimization was conducted. The frontal-crash safety –20 mm to 20 mm, as shown in equation (9).
Wang et al. 283

Table 7. Comparisons of the masses and other results for the F-S model and the initial model.

Mass (kg) Bending stiffness Torsional stiffness Bending frequency Torsional


(N/mm) (Nm/deg) (Hz) frequency (Hz)

Initial model 424.75 16 071.86 18 308.32 51.54 30.33


F-S model 394.62 15 902.94 18 232.83 52.65 30.76
Change –7.09% –1.05% –0.41% + 2.15% + 1.42%

F-S model: frontal-crash safety parts model.

Figure 23. Comparison of the acceleration curves for the F-S


model and the initial model: (a) left side of the vehicle; (b) right
side of the vehicle.
F-S: frontal-crash safety parts (model).

Figure 22. The generation process of sample points.


DOE: design of experiments.

optimized model (the F-S model) and the initial model


are shown in Table 7. As can be seen from Table 7, the
The objective functions were defined as the mini-
mass was reduced by 30.13 kg (i.e. by 7.09%). The
mum BIW mass and the minimum intrusion rate of the
maximum reduced performance was that of the bend-
lower front-door hinges, as shown in equation (8). The
ing stiffness, which changed by 1.05%; the maximum
acceleration peaks of both sides of the vehicle were not
increased performance was that of the bending fre-
more than 35g. Generally, these values are not large for
quency, which changed by 2.15%.
a passenger car. Therefore, this paper was more con-
Comparison of the acceleration curves of both sides
cerned with the intrusion of the door. The constraint
of the vehicle for the F-S model and the initial model
conditions were defined as the static bending and tor-
are shown in Figure 23. In comparison with the initial
sional stiffnesses, and the first-order bending and tor-
model, the acceleration curves of the left side fitted each
sional modal frequencies were not less than 95% of the
other well; however, the peaks and the pulses of the
initial values, as shown in equation (10); the intrusion
right side were larger for the initial model. However,
values of the upper front-door hinge were not
the peak value was smaller than the initial value, and so
more than 105% of their initial values, as shown in
it is acceptable.
equation (11).
Comparison of the intrusion rates for the F-S model
The generation process of the sample points by the
and the initial model are shown in Figure 24. As can be
DOE method is shown in Figure 22, and the multi-
seen from Figure 24, all curves fitted each other well.
objective optimization design process is shown in
Therefore, the frontal-crash safety parts optimization
Figure 20. In Figure 22 the modules of SFE DYNA
was acceptable.
and SFE Nastran were used for the implicit parameter
BIW model to generate the FE BIW model in the for-
mat of k files and bdf files. The modules of LS-DYNA
Side-impact safety parts optimization for the
and Nastran in Figure 22 were used to analyse the
safety performances, the stiffnesses and the lower-order parametric BIW model
modal frequencies. Based on the F-S model, the side-impact safety parts
After the frontal-crash safety parts were optimized, optimization was conducted. The frontal-crash safety
in this paper the most lightweight solutions of the part variables were the parameters associated with the
Pareto sets were chosen as the optimized results, the B pillar, the threshold beam, etc., which were chosen
results were assigned to the corresponding parts and mainly because of their contributions to side-impact
the performances were analysed. Comparisons of the safety based on experience. The number of thickness
masses, the bending and torsional stiffnesses and the variables of the side-impact safety parts is 11, which are
modal frequencies for the frontal-crash safety parts shown in Figure 25(a). The MOP for the side-impact
284 Proc IMechE Part D: J Automobile Engineering 230(2)

Figure 24. Comparison of the intrusion rates for the F-S model and the initial model.
LHL: lower left front-door hinge; LHR: lower right front-door hinge; UHL; upper left front-door hinge; UHR: upper right front-door hinge; F-S:
frontal-crash safety parts (model).

f1(chest) and f1(H point) are the intrusion velocities of


the beltline location, the chest location and the H point
respectively of the initial model. The design space is
indicated in equation (13). The section shape variables
are the parameters of the reinforcement plate of the
threshold beam, which are shown in Figure 25(b).
The objective functions were defined as the mini-
mum BIW mass and the minimum intrusion velocity of
the head location, as shown in equation (12). The con-
straint conditions were defined as the static bending
and torsional stiffnesses, and the bending and torsional
Figure 25. Side-impact safety parts variables: (a) thickness modal frequencies were not less than 95% of the initial
variables; (b) shape variables. values, as shown in equation (10); the intrusion veloci-
ties of the head, the beltline, the chest and the H point
safety parts can be expressed as follows. The objective were not more than 105% of the initial values, as shown
is in equation (14).
The generation process of the sample points by the
fðxÞ = fmin½fðMÞ, min½fðheadÞg ð12Þ
DOE method is shown in Figure 22, and the multi-
and the design variables are given by objective optimization design process is shown in
Figure 20.
04yi  y0i 420 ð13Þ After the side-impact safety parts were optimized, in
subject to this paper the most lightweight solutions were also cho-
sen as the optimized results, and the optimized results
1:05ffðbeltlineÞ, fðchestÞ, fðHpointÞg were assigned to the corresponding parts. The bending
ð14Þ and torsional stiffnesses, the modal frequencies and the
4f1 ðbeltlineÞ, f1 ðchestÞ, f1 ðHpointÞ
frontal-crash and the side-impact performances were
where f(head) is the intrusion velocity of the head analysed. The side-impact safety parts were mainly
point, yi is the coordinate value of the shape variable of focused on the B pillar, the threshold beam and the
the reinforcement plate in the lateral direction of the door’s anti-collision beam. Comparison of the masses,
sample points, y0i is the initial coordinate value of the the bending and torsional stiffnesses and the bending
reinforcement plate in the lateral direction, f(beltline), and torsional modal frequencies for the side-impact
f(chest) and f(H point) are the intrusion velocities of safety parts optimized model (S-S model) and the initial
the beltline location, the chest location and the H point model are shown in Table 8. As can be seen from
respectively of the sample points and f1(beltline), Table 8, the mass was reduced by 32.41 kg (i.e. a
Wang et al. 285

reduction as high as 7.63%), and the bending and tor- model, the peak values of both sides remained the same.
sional stiffness were reduced by less than 2.54%; the There was a small pulse on both sides for the period
bending and torsional frequencies increased a little. 20–40 ms. The pulse decreased for the period 40–60 ms.
Figure 26 shows a comparison of the acceleration Then the acceleration curves of both sides became close
curves of both sides of the vehicle for the S-S model to zero.
and the initial model. In comparison with the initial Figure 27 shows a comparison of the intrusion rates
for the S-S model and the initial model. All curves fitted
each other well. After the three phases were optimized,
the BIW mass was reduced. The initial energy was also
reduced, and so the intrusion rate was reduced.
Figure 28 shows a comparison of the intrusion velo-
city curves for the S-S model and the initial model. The
intrusion velocities of the optimized BIW structure in the
head location and the beltline location were almost the
same. The peak intrusion velocity of the chest location
was more than 105% of the corresponding value for the
initial model, but the steady intrusion velocity still met
Figure 26. Comparison of the acceleration curves for the S-S the constraint conditions. The steady intrusion velocity
model and the initial model: (a) left side of the vehicle; (b) right of the H point also met the constraint conditions.
side of the vehicle. Figure 29 shows a comparison of the acceleration
S-S: side-impact safety parts (model). curves for the S-S model and the initial model. All the

Figure 27. Comparison of the intrusion rate for the S-S model and the initial model.
LHL: lower left front-door hinge; LHR: lower right front-door hinge; UHL; upper left front-door hinge; UHR: upper right front-door hinge; S-S: side-
impact safety parts (model).

Table 8. Comparisons of the masses and other results for the S-S model and the initial model.

Mass (kg) Bending stiffness Torsional stiffness Bending frequency Torsional frequency
(N/mm) (N m/deg) (Hz) (Hz)

Initial model 424.75 16 071.86 18 308.32 51.54 30.33


S-S model 392.34 15 664.0 18 224.66 52.05 30.36
Change –7.63% –2.54% –0.46% + 0.99% + 0.10%

S-S model: side-impact safety parts optimized model.


286 Proc IMechE Part D: J Automobile Engineering 230(2)

Figure 28. Comparison of the intrusion velocity curves for the S-S model and the initial model.
S-S: side-impact safety parts (model).

Figure 29. Comparison of the acceleration curves for the S-S model and the initial model.
S-S: side-impact safety parts (model).
Wang et al. 287

peak values of the acceleration curves for the optimized Funding


model remained the same as for the initial model; how- The author(s) disclosed receipt of the following finan-
ever, the phase and the amplitude had large differences cial support for the research, authorship, and/or publi-
during the period 60–80 ms, and then finally they again cation of this article: This work wast supported by
had only a slight difference. Chongqing Basis and Leading-edge Research Program
After comparisons of the bending and torsional stiff- Projects (grant number cstc2013jcyjC60001).
nesses, the bending modal frequencies, the torsional
modal frequencies and the frontal-crash and side-
impact safety performances, the optimized BIW struc- References
ture all met the requirements. Therefore, this optimiza- 1. Pan F, Zhu P and Zhang Y. Metamodel-based light-
tion method was efficient and correct. weight design of B-pillar with TWB structure via support
vector regression. Comput Structs 2010; 88: 36–44.
Conclusions 2. Hunkeler S, Duddeck F, Rayamajhi M et al. Shape opti-
mization for crashworthiness followed by a robustness
Through the multi-objective optimization design of the analysis with respect to shape variables. Struct Multidis-
non-safe parts, the frontal-impact safety parts and side- ciplinary Optimization 2013; 48: 367–378.
impact safety parts of the BIW model, the following 3. Bendsøe MP. Optimal shape design as a material
conclusions can be obtained. distribution problem. Struct Optimization 1989; 1:
193–202.
1. From comparisons of the analysis of the bending 4. Chen CJ and Usman M. Design optimization for auto-
mobile applications. Veh Des 2001; 25: 126–141.
and torsional stiffnesses, the modal frequencies and
5. Bendsoe MP, Guedes JM, Haber RB et al. Analytical
the full-frontal-crash and side-impact safety perfor-
model to predict optimal material properties in the con-
mances for the implicit parametric BIW model with text of optimal structural design. Trans ASME, J Appl
the corresponding tests, the parametric BIW model Mech 1994; 4: 930–937.
was found to be correct. 6. Torstenfelt B and Klarbring A. Conceptual optimal
2. The non-safety parts multi-objective optimization design of modular car product families using simulta-
was carried out. After optimization, the mass was neous size, shape and topology optimization. Finite Ele-
reduced by 26.49 kg (i.e. by 6.24%). The maximum ments Analysis Des 2007; 43: 1050–1061.
reduced performance was that of the torsional stiff- 7. Lee KH, Joo WS, Song SI et al. Optimization of an
ness, namely a reduction of 0.35%; the maximum automotive side door beam, considering static require-
increased performance was that of the bending ment. Proc IMechE Part D: J Automobile Engineering
modal frequency, namely an increase of 0.93%. 2004; 218(1): 51–57.
8. Mundo D, Hadjit R, Donders S et al. Simplified model-
3. Furthermore, on the basis of the non-safety parts
ing of joints and beam-like structures for BIW optimiza-
optimization, the frontal-crash safety parts optimi-
tion in a concept phase of the vehicle design process.
zation was carried out. After optimization, the Finite Elements Analysis Des 2009; 45: 456–462.
mass was reduced by 30.13 kg (i.e. by 7.09%). The 9. Diaconu CG, Weaver PM and Mattioni F. Concepts for
maximum reduced performance was that of the morphing airfoil sections using bi-stable laminated com-
bending stiffness, namely a reduction of 1.05%; the posite structures. Thin-Walled Structs 2008; 46: 689–701.
maximum increased performance was that of the 10. Sofla AYN, Meguid SA and Tan KT. Shape morphing of
bending frequency, namely an increase of 2.15%. aircraft wing: status and challenges. Mater Des 2010; 31:
4. Also, on the basis of the frontal-crash safety parts 1284–1292.
optimization, the side-impact safety parts optimi- 11. Portela P, Camanho P and Weaver P. Analysis of morph-
zation was carried out. After optimization, the ing, multi stable structures actuated by piezoelectric
mass was reduced in total by 32.41 kg (i.e. a reduc- patches. Comput Structs 2008; 86: 347–356.
12. Daynes S and Weaver PM. Review of shape-morphing
tion as high as 7.63%). The reductions in the bend-
automobile structures: concepts and outlook. Proc
ing and torsional stiffnesses were less than 2.54%,
IMechE Part D: J Automobile Engineering 2013; 227(11):
the bending and torsional frequencies slightly 1603–1623.
increased, and the side-impact performance under- 13. Xu Z, Xu X, Wan X et al. Structure optimal design of
went almost no change. aluminum alloy bumper anti-collision beam (in Chinese).
J Mech Engng 2013; 49: 136–142.
Acknowledgements 14. Shi G, Chen Y, Yang Y et al. BIW architecture multidis-
ciplinary light weight optimization design (in Chinese). J
The authors would like to express their appreciation for Mech Engng 2012; 8: 110–114.
the funding support. 15. Grujicic M, Arakere G, Pisu P et al. Application of topol-
ogy, size and shape optimization methods in polymer
Declaration of Conflicting Interests metal hybrid structural lightweight engineering. Multidis-
cipline Modeling Mater Structs 2008; 4: 305–330.
The author(s) declared no potential conflicts of interest 16. Grujicic M, Sellappan V, Arakere G et al. Multi-
with respect to the research, authorship, and/or publi- disciplinary design optimization of a composite car door
cation of this article. for structural performance, NVH, crashworthiness,
288 Proc IMechE Part D: J Automobile Engineering 230(2)

durability and manufacturability. Multidiscipline Model- 23. Ye KQ, Li W and Sudjianto A. Algorithmic construction
ing Mater Structs 2009; 5: 1–28. of optimal symmetric Latin hypercube designs. J Statist
17. Hou S, Li Q, Long S et al. Multi-objective optimization Planning Inference 2000; 90: 145–159.
of multi-cell sections for the crashworthiness design. 24. Liefvendahl M and Stocki R. A study on algorithms for
Impact Engng 2008; 11: 1355–1367. optimization of Latin hypercubes. J Statist Planning
18. Xing R, Kong F and Lin S. Lightweight design and anal- Inference 2006; 136: 3231–3247.
ysis of BIW (in Chinese). J Shenyang Aerospace Univ 25. Liao X, Li Q, Yang X et al. Multiobjective optimization
2012; 3: 43–47. for crash safety design of vehicles using stepwise regres-
19. Yang Y, Zhao G and Meng F. Structure sensitivity analy- sion model. Struct Multidisciplinary Optimization 2008;
sis and optimized design of a certain white body work (in 35: 561–569.
Chinese). J Northeastern Univ 2008; 8: 1159–1163. 26. Fang H, Rais-Rohani M, Liu Z and Horstemeyer MF. A
20. Zuo W, Li W, Xu T et al. A complete development pro- comparative study of metamodeling methods for multi-
cess of finite element software for body-in-white structure objective crashworthiness optimization. Comput Structs
with semi-rigid beams in NET framework. Adv Engng 2005; 83: 2121–2136.
Software 2012; 45: 261–271. 27. Deb K, Pratap A, Agarwal S and Meyarivan T. A fast
21. Peeters B, Guillaume P, Van der Auweraer H et al. Auto- and elitist multi-objective genetic algorithm: NSGA-II.
motive and aerospace applications of the PolyMAX IEEE Trans Evolutionary Comput 2002; 6: 182–97.
modal parameter estimation method. In: 22nd IMAC con- 28. Konak A, Coit DW and Smith AE. Multi-objective opti-
ference & exposition on structural dynamics, Dearborn, mization using genetic algorithms: a tutorial. Reliability
Michigan, USA, 26–29 January 2004, Vol 1, pp. 38–48. Engng System Safety 2006; 91: 992–1007.
Bethel, Connecticut: Society for Experimental Mechanics. 29. Coutu D, Brailovski V and Terriault P. Optimized design
22. Adl AH and Panahi MS. Multi-objective optimal design of an active extrados structure for an experimental
of a passenger car’s body. In: ASME 2010 10th biennial morphing laminar wing. Aerospace Sci Technol 2010; 14:
conference on engineering systems design and analysis, 451–458.
Istanbul, Turkey, 12–14 July 2010, Vol 3, paper
ESDA2010-25124, pp. 277–286. New York: ASME.

Vous aimerez peut-être aussi