Professional Documents
Culture Documents
Multi-Objective Optimization
of Automotive Front Rail Based
on Surrogate Model and NSGA-II
Abstract With the rapid development of automotive industry, more and more atten-
tion has been paid to the lightweight and safety design. Crashworthiness optimiza-
tion is an essential part in automotive design. In this study, a non-dominated sorting
genetic algorithm II (NSGA-II) based on Kriging model is proposed to optimize
the structure of automotive frontal rail to meet the requirements of crashworthiness
and lightweight. The material of frontal rail is mild steel, which performs well in
strength. Kriging surrogate model is employed to replace traditional finite element
model, which will reduce much computational time and improve the efficiency. Then
NSGA-II is applied to solve the multi-objective optimization problem. The results
illustrate that the Pareto optimal front obtained by NSGA-II exhibits good perfor-
mance on convergence and diversity. And then the optimal scheme for design is
selected, the accuracy is proved to be high. Compared to baseline model, the opti-
mized automotive frontal rail shows significant improvement on crashworthiness and
achieves 7.5% weight reduction.
20.1 Introduction
Y. He (B)
Mechanical Engineering, Rensselaer Polytechnic Institute, Troy, NY 12180, USA
e-mail: YiweiHEEE@163.com
W. Xu
Mathematics and Computer Science, University of California-San Diego, La Jolla, San Diego, CA
92122, USA
F. Gu
Faculty of Arts and Sciences, College of William & Mary, Williamsburg, VA 23187, USA
© The Author(s), under exclusive license to Springer Nature Switzerland AG 2021 251
J. Xu and K. M. Pandey (eds.), Mechanical Engineering and Materials,
Mechanisms and Machine Science 100,
https://doi.org/10.1007/978-3-030-68303-0_20
252 Y. He et al.
mass points.The reduced FE model contains 428,263 elements and 451,820 nodes
while the percent of triangular elements less than 5%. According to the standards
of frontal impact test against deformable barrier with 40% overlapping in C-NCAP,
the speed of impact is set as 64 km/h while the simulation time is set as 100 ms.
LS-DYNA is utilized for collision simulation. The contact force in a collision is
necessary to be considered in modeling [12], therefore the contact force is detailed
construction in FE model.
The structure considered in this study is a front rail that consists of several thin-
walled parts, and the material is mild steel, then the properties of material are listed
as Table 20.1. And this material has a piecewise linear elastic-plastic behavior with
strain hardening. Figure 20.2 illustrates the relationship between the plastic stress
and plastic strain of mild steel [13].
The process of optimization design on lightweight and crashworthiness can be
draw as Fig. 20.3.
the steering column D2 and the deformation of front rail D3 are considered as the
constrains.
The principal parts of crumple zone on the automotive frontal rail are selected
as design parameters, and the explosive view of thin-walled parts in front rail for
optimization design is shown in Fig. 20.4.
Fig. 20.4 The explosive view of design parts in front rail for optimization design
20 Multi-Objective Optimization of Automotive Front Rail … 255
Then the thickness of six components t1 , t2 , t3 , t4 , t5 and t6 are set as the design
variables. Considering the actual manufacturing of the thin-walled parts, the thick-
ness should be rounded up to 2 decimal places in this study. The ranges for design
variables are listed in Table 20.2.
Design of experiment (DoE) is the first step to build a surrogate model. In order
to determine a set of training samples, an Optimum Latin Hypercube Sampling is
employed to generate 200 sample points in this study.
The Kriging model uses the training samples to construct the surrogate model for
predicting the output. It consists of approximate model and a stochastic process as
shown in Eq. (20.1):
k p
(l)
cor Y x i , Y x l = exp(− θ j x (i)
j − x j ) (20.2)
j=1
where the θ vector represents the width of each basis function, the parameters p
controls the smoothness of the approximation in the proximity of the given sample
points, and k is the dimensionality of the problem.
Then an n × n correlation matrix of all the observed data can be conducted as
Eq. (20.3):
256 Y. He et al.
⎛ ⎞
cov Y (1) , Y (1) · · · cov Y (1) , Y (n)
⎜ .. .. .. ⎟
cov⎝ . . . ⎠ (20.3)
(n) (1)
(n) (n)
cov Y , Y · · · cov Y , Y
These correlations
depend on the absolute distance between the sample points
(i) (l)
xj − xj and the parameters θ and p.
The likelihood function is defined as Eq. (20.4):
(i)
1 (Y − μ)2
L Y (1) , · · · , Y (n) |μ, σ = n exp − (20.4)
2π σ 2 2 2σ 2
where Ψ is the correlation matrix between the random variables. Then we take the
natural logarithm to simplify the likelihood function as Eq. (20.6)
n n (y − 1μ)T Ψ −1 (y − 1μ)
ln(L) = − ln(2π ) − ln σ 2 − ln|Ψ | − (20.6)
2 2 2σ 2
And we can obtain the maximum likelihood estimates (MLE) for μ (Eq. 20.7)
and σ 2 (Eq. 20.8):
1T Ψ −1 y
μ̂ = (20.7)
1T Ψ −1 1
(y − 1μ)T Ψ −1 (y − 1μ)
σ̂ 2 = (20.8)
n
Then the concentrated ln-likelihood function [14] can be deduced as Eq. (20.9)
n 1
ln(L) ≈ − ln σ̂ 2 − ln|Ψ | (20.9)
2 2
The value of this function depends on the unknown parameters θ and p. Therefore
we are supposed to find values for these parameters which maximize Eqs. (20.3),
(20.9) to build the best Kriging model with high accuracy. According to previous
study [15], we can fix the parameter at p = 2, which shows the best smooth correlation
with a continuous gradient,
To test the error of surrogate
model,
such metrics as maximum relative error
(emax ), average relative error eavg and R-square (R2 ) are employed [16]. And the
results of error analysis are shown in Table 20.3.
20 Multi-Objective Optimization of Automotive Front Rail … 257
It can be concluded from Table 20.3 that the Kriging models show high accuracy,
therefore, the surrogate models can be applied in next optimization.
As mentioned above, the maximum acceleration of B-pillar amax and mass of design
components m are considered as objectives while the rearward intrusion of A-pillar
D1 , the steering column D3 and the deformation of front rail are determined to be
the constrains, the value of D1 and D2 are supposed to be less than 300 mm on the
demand of C-NCAP and the value of D3 should be less than 1106.43 mm that is the
value of baseline. The aim is to find the optimum design variables t1 , t2 , t3 , t4 , t5 and
t6 to minimize the values of amax and m. The mathematical models can be expressed
as Eq. (20.10):
Find t1 , t2 , t3 , t4 , t5 , t6
⎧
⎪
⎪ Minimi zeamax (t1 , · · · , t6 ), m(t1 , · · · , t6 )
⎨
s.t.D1 , D2 300mm
(20.10)
⎪
⎪ D3 1106.43mm
⎩
0.6mm t1 , · · · , t6 3.2mm
index for crossover and mutation are 10 and 20. The fitness function and constrains
function can be determined by the Kriging metamodels above.
The POF of the NSGA-II based on Kriging surrogate model is shown in Fig. 20.5.
Refer to Fig. 20.4, the Pareto optimal front of NSGA-II based on the Kriging model
performs well in convergence and diversity.
In real engineering optimization design, the lightweight is as important as the
crashworthiness. Therefore, one of the optimal points in POF was determined as the
optimum solution. Then the optimum solution is compared to baseline model and
the accuracy is verified by calculating the objective values in FE model, the results
are shown in Table 20.4.
To compare the optimum results in FE models, the deformation of the front rail
in frontal impact test against deformable barrier with 40% overlapping in Ls-Dyna
software are shown in Fig. 20.6. It is obvious that the optimized front rail has smaller
deformation after impact.
20.6 Conclusion
In order to obtain the optimum structure of the parts in front rail, a surrogate-based
optimization is developed a frontal rail. The multi-optimization problem is defined to
minimum the peak accelerate amax and the mass of frontal rail m with the constraints
of the rearward intrusions of A-pillar D1 and the steering column D2 .To reduce the
calculation time, the Kriging surrogate model is proposed to provide high accuracy
20 Multi-Objective Optimization of Automotive Front Rail … 259
Fig. 20.6 Deformation of baseline model (left) and optimized model (right)
surrogate model for NSGA-II. Optimized by the proposed approach, the crashworthi-
ness improved significantly and achieve 7.5% weight reduction than that of baseline
model.
References
1. Ma, Y., Chen, R., Bai, J., Zuo, W.: Shape optimization of thin-walled cross section for automo-
bile body considering stamping cost, manufacturability and structural stiffness. Int. J. Automot.
Technol. 21, 503–512 (2020)
2. Ma, Y., Wang, X., Zuo, W.: Analytical sensitivity analysis method of cross-sectional shape for
thin-walled automobile frame considering global performances. Int. J. Automot. Technol. 21,
1207–1216 (2020)
3. Lu, S., Ma, H., Xin, L., Zuo, W.: Lightweight design of bus frames from multi-material topology
optimization to cross-sectional size optimization. Eng. Optimiz. 51(6), 961–977 (2019)
4. Bai, J., Li, Y., Zuo, W.: Cross-sectional shape optimization for thin-walled beam crash-
worthiness with stamping constraints using genetic algorithm. Int. J. Veh. Des. 73, 76–95
(2017)
5. Zuo, W.: Bi-level optimization for the cross-sectional shape of a thin-walled car body frame
with static stiffness and dynamic frequency stiffness constraints. Proc. Inst. Mech. Eng., Part
D: J. Automob. Eng. 229(8), 1046–1059 (2015)
260 Y. He et al.
6. Zuo, W., Yu, J., Saitou, K.: Stress sensitivity analysis and optimization of automobile body
frame consisting of rectangular tubes. Int. J. Automot. Technol. 17(5), 843–851 (2016)
7. Chen, W., Zuo, W.: Component sensitivity analysis of conceptual vehicle body for lightweight
design under static and dynamic stiffness demands. Int. J. Veh. Des. 66(2), 107–123 (2014)
8. Qiao, H., Wang, S., Zhao, T., Tang, H.: Topology optimization for lightweight cellular material
and structure simultaneously by combining SIMP with BESO. J. Mech. Sci. Technol. 33,
729–739 (2019)
9. Guo, S., Ang, H., Cai, H.: Construction of an adaptive sampling surrogate model and application
in composite material structure optimization. Acta Materiae Compositae Sinica 35(8), 2084–
2094 (2018)
10. Liu, H., Zhang, X., Xiao, L.: Collision simulation analysis of 7075 aluminum alloy car bumper
based on LS-DYNA. J. Machine Des. 2, 18–23 (2011)
11. China Automotive Technology and Research Center, C-NCAP Management Regulation http://
www.c-ncap.org/cncap/pjgz (2018)
12. Ma, Y., Chen, X., Zuo, W.: Equivalent static displacements method for contact force
optimization. Struct. Multidiscipl. Optimiz. 62, 323–336 (2020)
13. Technical Transfer Dispatch#6-Body Structure Materials, ULSAB-AVC Consortium (2001)
14. Raponi, E., Bujny, M., Olhofer, M., Aulig, N., Boria, S., Duddeck, F.: Kriging-assisted topology
optimization of crash structures. Comput. Methods Appl. Mech. Eng. 348(1), 730–752 (2019)
15. Forrester, A., Sobester, A., Keane A.: Engineering Design Via Surrogate Modelling: A Practical
Guide. John Wiley & Sons, Inc (2008)
16. Tian, L., Mehari, M., Santi, S., Latré, S., Poorter, E., Famaey, J.: Multi-objective surrogate
modeling for real-time energy-efficient station grouping in IEEE 802.11ah. Pervasive Mobile
Comput. 57, 33–48 (2019)
17. Deb, K., Pratap, A., Agarwal, S., Meyarivan, T.: A fast and elitist multi-objective genetic
algorithm: NSGA-II. IEEE Trans. Evol. Comput. 6(2), 182–197 (2002)
18. Ardakan, M.A., Rezvan, M.T.: Multi-objective optimization of reliability–redundancy alloca-
tion problem with cold-standby strategy using NSGA-II. Reliab. Eng. Syst. Saf. 172, 225–238
(2018)