International Journal of Automotive Technology, Vol. 17, No. 6, pp.

1003−1011 (2016) Copyright © 2016 KSAE/ 093−08

DOI 10.1007/s12239−016−0098−6 pISSN 1229−9138/ eISSN 1976−3832

CROSS-SECTIONAL SHAPE DESIGN AND OPTIMIZATION OF

AUTOMOTIVE BODY WITH STAMPING CONSTRAINTS

W. J. ZUO1, 2)* and J. T. BAI3)

1)
State Key Laboratory of Automotive Simulation and Control, Jilin University, Changchun 130025, China
2)
Key Laboratory of Advanced Manufacture Technology for Automobile Parts (Chongqing University of Technology),
Ministry of Education, Chongqing 400054, China
3)
School of Mechanical Science & Engineering, Jilin University, Changchun 130025, China

(Received 18 January 2016; Revised 30 March 2016; Accepted 31 March 2016)

ABSTRACT−At conceptual design stage, automotive body is usually simplified as a frame structure, which consists of thin-
walled beams (TWBs). Therefore, the most important issue is to determine the cross-sectional shape of TWBs under the
requirement of mechanical properties. However, design engineers mostly depend on their experience or repeated modification
to design the cross-sectional shape of TWBs. So this paper presents a rapidly cross-sectional shape design and optimization
method to satisfy the demand of mechanical properties and meanwhile minimize the weight of TWB. Firstly, cross-sectional
mechanics property formulations are summarized. Especially, the torsional rigidity formulation of three-cell cross section is
derived for the first time in this paper. Secondly, the shape optimization model is created to minimize the weight of TWB and
improve the mechanical properties, which is solved by genetic algorithm. Moreover, three stamping constraints, draft angle,
chamfer radius and assembly, are introduced to promote the cross-sectional shape more practice. Lastly, numerical examples
verify the effectiveness of the optimization model and show the application in structural modification of automotive frame.

KEY WORDS : Automotive body, Thin-walled beam, Cross-sectional shape optimization, Stamping constraints

1. INTRODUCTION stage. The first order analysis (FOA) for automotive body
structure design (Nakagawa et al., 2004; Nishigaki and
At the conceptual design stage, automotive body is usually Kikuchi, 2004; Nishigaki et al., 2001, 2004; Kunishi and
simplified as a frame structure consisting of thin-walled Kikuchi, 2004; Tsurumi et al., 2004) was developed in the
beams (TWBs) with complex cross section, which concept design stage. Design engineers can quickly choose
determine automotive performances such as stiffness and the best cross-sectional shape based on FOA. FOA method
crashworthiness (Zuo, 2013). The frame structure can also can only solve cross-sectional mechanical properties of
provide early-stage evaluations about the global stiffness open cell and single cell, but those of double cell and three
and mass of automotive body. cell can not be solved, which are significant cross section
TWB is a type of lightweight structure, which has the for automotive frame structure. The cross-sectional shape
high stiffness to mass ratio. TWBs are generally fabricated design and optimization for automotive body was also
from pressed metal sheets and spot-welded together (Suh et studied (Yoshimura et al., 2005), but three-cell cross
al., 2006). The cross-sectional shape of TWB determines section was not still considered. Much effort has been
the performances of automotive body. Therefore, at the devoted to optimize the cross section of TWBs, such as
conceptual design stage of automotive body, sufficient topology optimization (Griffiths and Miles, 2003) and
computer-aided-design geometry data of TWBs is shape optimization (Heo et al., 2013). Topology
necessary to design automotive body (Tovey, 2002; Wang optimization method of the cross-sectional shape was
et al., 2010, 2011; Song et al., 2015). studied (Kim and Kim, 2000) at the conceptual design
To date, design engineers mostly depend on their stage, which can not satisfy the stamping constraints of
experience and intuition to repeatedly adjust the cross- thin-walled sheets. During the process of optimization,
sectional shape to satisfy the demands of mechanical when the cross-sectional shape violates stamping
properties in the automotive practice. Furthermore, the constraints, it is difficult or impossible to successfully
cross-sectional area should be minimized to reduce the manufacture (Yu and Yang, 2015). The spot weld layout
mass of TWB. So the most important issue is to acquire the optimization was studied considering manufacturing
optimal cross-sectional shape at the conceptual design constraints including minimum weld pitch, thickness and
curvature of flanges (Zhou et al., 2014). They only
*Corresponding author. e-mail: [email protected] considered spot weld layout constraints and did not

1003
1004 W. J. ZUO and J. T. BAI

Figure 1. Sketch of three-cell cross section.

consider stamping constraints. In this paper, three stamping

constraints including draft angle, chamfer radius and
assembly are studied, which can guarantee the sheets of Figure 2. Seven types of cross sections and their
TWB to be stamped directly. formulations of torsional rigidity J.
Therefore, the purpose of this paper is to optimize the
cross-sectional shape considering cross-sectional properties, n m
⎡⎛ l 3t − l t 3 ⎞ ⎤
such as area, second moments of area, torsional rigidity, I xy = ∑∑ ⎢⎜ ij i ij i ⎟ sin 2θ ij + lijti xCij yCij ⎥ (5)
and stamping constraints. Additionally, another aim is to i =1 j=1 ⎣⎝ 24 ⎠ ⎦
develop SuperSection software for rapid design and where θij is the angle between the positive x axis and the
optimization of the cross-sectional shape of TWBs of j-th segment of i-th sheet. From Ix, Iy and Ixy, the principal
automotive body. moments of area are derived that

2. FORMULATIONS OF MECHANICAL 1 1
I max = ( I x + I y ) + ( I x − I y ) 2 + I xy2 (6)
PROPERTIES 2 2
1 1
A three-cell cross section is shown in Figure 1. Four metal I min = ( I x + I y ) − ( I x − I y ) 2 + I xy2 (7)
sheets are spot welded together as a TWB. Each sheet can 2 2
be regarded as a folding line consisting of rectangle The counterclockwise angle of principle inertia direction
segments. So the cross-sectional area can be defined as with respect to the reference x axes ϕ is calculated using
n m n m the following formulas
A = ∑∑ Aij = ∑∑ lijti (1)
i =1 j=1 i =1 j=1 1 ⎛ −2 I xy ⎞
ϕ = tan −1 ⎜⎜ ⎟⎟ (8)
where n is the number of sheet; m is the number of segment 2 ⎝ Iy − Ix ⎠
of i-th sheet; lij and Aij are the length and area of the j-th
The procedure for calculating the torsional rigidity
segment of the i-th sheet, respectively; ti is the thickness of
depends on the types of the cross-sectional shape, as shown
the i-th sheet.
in Figure 2. The torsional rigidity of open-cell section is
The centroid of the cross section is the weighted average
calculated by
of the centroid of segments, namely
o m
1 n m 1 n m J o = ∑∑ lijti3 (9)
Cx = ∑∑ xCij and Cy = ∑∑ yCij (2) i =1 j=1
A i j A i j
where o is the number of sheet for open-cell section.
where ( xCij, yCij ) denotes the coordinate of the center of the The torsional rigidity of single-cell section J1c , double-
j-th segment of the i-th sheet, as shown in Figure 1. cell section J 2c and three-cell section J 3c can, respectively,
Then, second moments of area Ix, Iy and product moment be written as
of area Ixy with respect to the centroid can be, respectively,
4F12
expressed as J1c = (10)
Ll tl + Lu tu
n m
⎡⎛ l t 3 ⎞ ⎛ l 3t ⎞ ⎤
I x = ∑∑ ⎢⎜ ij i ⎟ cos 2 θ ij + ⎜ ij i ⎟ sin 2 θ ij + lijti xC2 ij ⎥ (3) ⎧ ⎛ L L L − Lu′ ⎞ ⎛ Lr Lu′ ⎞ ⎫
i =1 j=1 ⎣⎝ 12 ⎠ ⎝ 12 ⎠ 4 ⎨ F12 ⎜ l + r + u
⎦ ⎟ − F2 ⎜ + ⎟ ⎬
2
t
⎝ l t t ⎠ ⎝ tr tu ⎠ ⎭
J 2c = ⎩
r u
(11)
n m
⎡⎛ l t 3 ⎞ ⎛ l 3t ⎞ ⎤ ⎛ Lr Lu′ ⎞⎛ Ll Lr Lu − Lu′ ⎞ L2r
I y = ∑∑ ⎢⎜ ij i ⎟ sin 2 θ ij + ⎜ ij i ⎟ cos 2 θ ij + lijti yC2 ij ⎥ (4) ⎜ + ⎟⎜ + + ⎟−
i =1 j=1 ⎣⎝ 12 ⎠ ⎝ 12 ⎠ ⎦ ⎝ tr tu ⎠⎝ tl tr tu ⎠ tr2
CROSS-SECTIONAL SHAPE DESIGN AND OPTIMIZATION OF AUTOMOTIVE BODY WITH STAMPING CONSTRAINTS 1005

J 3c = 4(q1F1 + q2 F2 + q3 F3 ) (12) as shown in Figures 3 (a) and (b). On the other hand, the
intersection of sheets should be prohibited in the
where q1, q2 and q3 are solved by Equation (13).
assembling sheets process for spot welding, as shown in
⎛ q1 ⎞ ⎛ F1 ⎞ Figure 3 (c).
⎜ ⎟ ⎜ ⎟ So, the mathematical form of the optimization model for
B ⎜ q2 ⎟ = ⎜ F2 ⎟ (13)
⎜q ⎟ ⎜F ⎟ cross-sectional shape of TWB can be expressed as
⎝ 3⎠ ⎝ 3⎠ n m
min A = ∑∑ lijti (15)
⎡ Lu Lr Lm − Lm′ Lr Lm − Lm′ ⎤
⎢t + t + t − − i=1 j=1

⎢ u r m tr tm ⎥⎥
⎢ Lr Lr Lm′ L′ ⎥ ⎧X ≤ X ≤ X
where B = ⎢ − + − m ⎥. ⎪
⎢ tr tr tm tm ⎥ ⎪Y ≤ Y ≤ Y
⎢ Lm − Lm′ L′ Ll Lm ⎥ ⎪t ≤ t ≤ t
⎢ − − m + ⎥ ⎪
⎣ tm tm tl t m ⎦ ⎪ I x ≥ Ix
⎪
s.t. ⎨ I y ≥ Iy (16)
F1, F2 and F3 are, respectively, the enclosed area of cell I, ⎪ 
cell II and cell III; Lu, Ll, Lm and Lr are, respectively, the ⎪J ≥ J
length of the upper sheet, lower sheet, middle sheet and ⎪α i ≥ α i = 2,3,..., nd
reinforcement; Lu′ and Lm′ are, respectively, the length of ⎪ 
⎪ Rj ≥ R j = 1,2,..., nr
the part of upper sheet and middle sheet; tu, tl, tm and tr are, ⎪ nisp = 0
respectively, the thickness of the upper sheet, lower sheet, ⎩
middle sheet and reinforcement. It is noted that the where X and X are the lower and upper bounds of
torsional rigidity formulation of three-cell cross section J 3c coordinates X , respectively; Y and Y are the lower and
is firstly derived in this paper. upper bounds of coordinates Y , respectively; t and t are
Furthermore, when a section consists of open-cell and the lower and upper bounds of thicknesses t , respectively;
closed-cell, the torsional rigidity can be defined as
Ix , Iy and J are the specified lower bounds of second
J = J o + J kc k = 1, 2, 3 (14) moment of area and torsional rigidity, respectively; αi is the
i-th draft angle and nd is the number of draft angles; Rj is the
where k is the number of cell on the cross section. j-th chamfer radius and nr is the number of chamfers; α
In conclusion, the formulations of second moments of and R are the specified lower bound of draft angle and
area Ix and Iy are the same for the seven types of cross chamfer radius, respectively; nisp is the number of
sections. However, the formulations of the torsional intersection points. Stress and deflection constraints are
rigidity J for the seven types of cross sections are different, important to achieve structural safety of the automotive
as shown in Figure 2. frame structure. However, the existed beam theories can
not acquire the good solution of the stress and cross-
3. OPTIMIZATION MODEL sectional deflection. Especially, stress concentration usually
occurs at the joint zone of automotive body. Therefore, we
3.1. Optimization Formulations recommend that stress and deflection constraints should be
The moments of inertia Ix and Iy and the torsional rigidity J considered in the detailed design stage, when the automotive
play an important role in the stiffness evaluation of TWBs.
The reduction of the cross-sectional area A can achieve the
aim of lightweight design of TWB. So the optimization
model is formulated to determine the optimal cross section
of TWB, which is to minimize the weight of TWB
constrained with the moments of inertia, torsional rigidity
and stamping constraints. The cross-sectional shape of the
metal sheets is described by a number of shape control
points, as shown in Figure 1. Points are classified as
moving points and fixed points. Moving points are the
design variables of the optimization model. Fixed points
remain unchanged during the optimization process, which
are determined by the limited design space and
manufacturing requirements.
Metal sheet is formed using stamping process, so
stamping constraints should be satisfied. Draft angle and
chamfer radius should be greater than design requirements, Figure 3. Three styles of stamping constraints.
1006 W. J. ZUO and J. T. BAI

body is meshed by plate or shell finite elements. obtain improved solution is

1
3.2. Solved by GA Fitness = >1 (22)
f
The optimization model is a highly nonlinear problem.
Therefore, the effective strategies for this problem are the 3.3. SuperSection Software
sequential linear programming (SLP) (Lamberti and SuperSection software is developed independently by
Pappalettere, 2002, 2005) and evolutionary algorithms ourselves, which can rapidly design and optimize the seven
(EAs) (Zuo et al., 2012, 2014; Zuo, 2015). SLP needs types of cross-sectional shape of TWBs for automotive
sensitivity analysis. Unfortunately, the sensitivity analysis body. The key operation forms of the software are shown in
with regard to design variables is too complicated to derive Figure 4, which demonstrates the cross-sectional shape
the explicit expressions in this problem. More importantly, design form, the cross-sectional shape properties solution
stamping constraints can not be explicitly expressed by form and the cross-sectional shape optimization form,
formulas, so sensitivity analysis from responses of stamping respectively.
constraints with respect to design variables can not be
obtained. GA is a classical EA, and it has been proved that 3.4. Optimization Procedure
GA is particularly effective to solve the nonlinear The optimization processes are summarized in the
optimization problem without sensitivity information following;
(Gauchia et al., 2010; Zuo et al., 2011). Consequently, we Step 1: Define the initial cross-sectional shape using the
directly adopt GA with elitism preservation to overcome
the above difficulties.
To acquire the fitness of GA, we transform the
constrained optimization problem into an unconstrained
one. Meanwhile, the normalized penalty functions are used
to deal with the implicit constraints. Thus, the penalty
function of Ix can be defined as
⎧⎪c ( I − I ) I , for Ix < I x
p( I x ) = ⎨ Ix x x x (17)
⎪⎩0, otherwise
where cI x is a penalty coefficient and assigned the value
103. The mathematical expression of the penalty function
for I y and J is similar to Equation (17). Consequently, the
total stiffness constraints pstiffness can be calculated by the
summation:
pstiffness = p( I x ) + p( I y ) + p ( J ) (18)
The penalty functions of αi can be expressed as
⎧cα (α − α i ) α , for α i < α
p(α i ) = ⎨ i (19)
⎩0, otherwise
where cαi is a penalty coefficient and assigned the value
104, which is larger than the cIx , because draft angles must
be satisfied in optimization process. The mathematical
expressions of penalty function for Ri and nisp are similar
to Equation (19). The total stamping constraint functions
pstamping can be calculated by the summation:
pstamping = ∑ i =1 p(α i ) + ∑ i =1 p( Ri ) + p( nisp )
nd nr
(20)
Consequently, the fitness formulation of GA for the
optimization model can be written as
1 1
Fitness = = (21)
f A A0 + pstiffness + pstamping
where A0 is the initial area to normalize A. When A/A0 < 1,
pstiffness = 0 and pstamping = 0 , an improved solution is
obtained to satisfy f < 1. So the sufficient condition to Figure 4. Key operation steps of SuperSection software.
CROSS-SECTIONAL SHAPE DESIGN AND OPTIMIZATION OF AUTOMOTIVE BODY WITH STAMPING CONSTRAINTS 1007

preprocessor of SuperSection software. sectional shape of TWB for automotive body. The cross-
Step 2: Determine the upper and lower bounds of design sectional shape of B-pillar, as shown in Figure 6, is
variables and the constraints of the optimization model. optimized as a numerical example. The initial and bounding
Step 3: The optimization model is solved by using GA. values of design variables are listed in Tables 1 and 2.
Step 1 and 2 are manually operated by the mouse click Points 44, 48, 53 and 56 are fixed so as to satisfy
and keyboard input on the graphic user interface. Step 3 is manufacturing and designing requirements. The parameters
automatically implemented by the SuperSection software. of GA, namely, the number of generations, the number of
The detailed flowchart, as shown in Figure 5,
demonstrates the solution procedure of optimization model
for TWB structures.

4. NUMERICAL EXAMPLE

4.1. Cross-sectional Shape Optimization

Three-cell with opens section is the most complex cross-

Figure 7. Fitness curve of the optimization model.

Figure 5. Flowchart of the GA optimization method.

Figure 6. Initial cross-sectional shape with variable

boundary. Figure 8. Cross-sectional shapes at some generations.
1008 W. J. ZUO and J. T. BAI

Table 1. Point variables and optimal solutions (mm).

No. Lower bound Initial Upper bound Optimal solution
17 (– 41, – 30) (– 35.33, – 1.17) (– 6, 4) (– 35.98, – 3.19)
1 (– 35, – 24) (– 29.34, 26.76) (22, 32) (– 29.62, 26.49)
5 (23, 35) (29.07, 26.17) (21, 31) (24.45, 28.39)
7 (27, 39) (32.92, 1.28) (– 4, 6) (38.39, 3.59)
39 (– 46, – 34) (– 40.00, – 15.00) (– 20, – 10) (– 38.78, – 17.41)
32 (– 36, – 24) (– 30.00, – 30.00) (– 35, – 25) (– 31.69, – 28.8)
28 (– 21, – 9) (– 15.00, – 30.00) (– 35, – 25) (– 14.64, – 29.51)
27 (– 20, – 8) (– 13.96, – 27.58) (– 33, – 23) (– 10.33, – 30.75)
33 (9, 21) (15.01, – 27.32) (– 32, – 22) (15.17, – 28.7)
31 (14, 26) (20.00, – 35.00) (– 40, – 30) (20.99, – 30.76)
30 (24, 36) (30.00, – 35.00) (– 40, – 30) (27.01, – 36.81)
29 (29, 41) (35.00, – 15.00) (– 20, – 10) (36.42, – 14.95)
43 (– 38, – 26) (– 32.11, – 7.67) (– 13, – 3) (– 35.7, – 4.62)
41 (– 31, – 19) (– 25.00, 20.00) (15, 25) (– 23.43, 21.13)
46 (14, 26) (20.00, 20.00) (15, 25) (23.24, 24.92)
49 (24, 35) (29.42, – 6.65) (– 12, – 2) (32.71, – 9.27)
23 (– 28, – 17) (– 22.72, 5.94) (1, 11) (– 21.07, 6.03)
20 (– 25, – 14) (– 19.24, 13.00) (8, 18) (– 15.34, 10.45)
24 (– 7, 5) (– 1.08, 16.86) (12, 22) (– 6.73, 14.49)
18 (9, 20) (14.36, 13.52) (9, 18) (9.63, 13.53)
19 (7, 18) (12.18, 4.76) (0, 10) (16.46, 4.72)
57 (– 61, – 49) (– 55.00, – 10.00) (– 15, – 5) (– 57.53, – 11.34)
58 (44, 56) (50.00, – 10.00) (– 15, – 5) (50.84, – 13.91)

population, crossover probability, mutation probability and Table 2. Thickness variables and optimal solutions (mm).
elitism probability are 80, 30, 0.55, 0.2 and 0.01,
Lower Upper Optimal
respectively. No. Initial
bound bound solution
The demands of stiffness and stamping constraints are
listed in Table 3. In practice, these constraints are tu 0.60 0.70 0.70 0.69
commonly originated from the benchmarking experiment. tl 1.02 1.20 1.20 1.10
The draft angle at point 18 and chamfer radius at point 28
tm 1.70 2.00 2.00 1.94
are, respectively, less than the demands, which must be
improved during the optimization process. tr 1.28 1.50 1.50 1.28
In the optimization process, optimal cross-sectional
shapes are gradually elite-preserved and the cross-sectional
properties are enhanced at each generation of GA. The When the optimal solution is obtained, the computational
cross-sectional shapes and properties at the 3th, 6th, 28th cost is 104 seconds by using a laptop computer (8 GB
and 46th generation are shown in Figure 8, where αmin and memory and i5 CPU). Although GA usually needs
Rmin are the minimal draft angle and minimal chamfer numerous iterative evaluations, the computational time is
radius of the cross section, respectively. It can be seen from quite short due to the small-scale optimization model.
Figure 8 that: (1) in Figures (a) and (b), the constraints are The optimal design variables are listed in Tables 1 and 2.
not completely satisfied; (2) in Figures (c) and (d), all the The objective and constraints of the optimal cross-sectional
constraints are satisfied and the objective function shape are presented in Table 3, which shows that the area is
decreases. decreased and the moments of inertia Ix and Iy and the
CROSS-SECTIONAL SHAPE DESIGN AND OPTIMIZATION OF AUTOMOTIVE BODY WITH STAMPING CONSTRAINTS 1009

Table 3. Comparison of properties between the initial and optimized to reduce the mass of automotive body and
optimal section. improve the torsional stiffness, bending stiffness and the
dynamic frequency stiffness.
Improved
Properties Demands Initial Optimal The torsional and bending load cases are necessary for
ratio
the test on the static stiffness of a car body in engineering.
A (mm2) Null 681.5 653.4 − 4.12 % The torsional load case often occurs when one of the front
4
Ix (mm ) ≥ 218634.1 212266.2 218815.3 + 3.09 % wheels jumps across a pit, as shown in Figure 9. Therefore,
the torsional stiffness is generally defined as
Iy (mm4) ≥ 563500.3 547087.7 580634.7 + 6.13 %
MT FB
J (mm4) ≥ 304852.5 295973.3 318291.5 + 7.54 % KT = = (23)
Δφ arctan(u / B )
αmin ( ) o
≥ 94 76.0 94.1 Satisfied
where MT is the torsional moment, Δφ is the rotation angle,
Rmin (mm) ≥2 1.3 2.8 Satisfied F is the lumped force, B is the wheel span and u is the
nisp 0 0 0 Satisfied deflection caused by F.
Meanwhile, the bending load step is shown in Figure 10,
the bending stiffness is evaluated from
torsional rigidity J are improved; meanwhile, all stamping nf
1 Fi
constraints are satisfied. Therefore, designer can develop KB =
nf
∑u (24)
high stiffness and lightweight TWB structure of automotive i =1 i

body using the proposed method.

The maximum fitness of each generation gradually
increases and converges until the 46th generation, as shown
in Figure 7. It should be noted that the maximum fitness
exceeds one at the 46th generation. So after that all
constraints are satisfied and the cross-sectional area of
TWB continues to reduce.

4.2. Application in Automotive Frame

The frame structure of automotive body is shown in
Figures 9 and 10, which consists of twenty-two TWBs. In
order to apply the proposed method into the automotive
frame, the cross-sectional shapes of A-pillar, B-pillar, C-
pillar and floor longeron, as shown in Figure 9, are

Figure 9. Automotive frame torsional load step.

Figure 11. Comparison of cross-sectional shapes between

Figure 10. Automotive frame bending load step. the initial and modified solutions.
1010 W. J. ZUO and J. T. BAI

Table 4. Comparison between the initial and modified frame structure of automotive body.
Responses Initial Optimal Improved ratio
6 6
Torsional stiffness (N·mm/rad) 6.349 × 10 6.441 × 10 + 1.45 %
Bending stiffness (N/mm) 3702.8 4294.0 + 15.97 %
First-order frequency (Hz) 36.9 37.6 + 1.90 %
Second-order frequency (Hz) 52.1 53.0 + 1.73 %
Third-order frequency (Hz) 57.1 57.6 + 0.87 %
Structural mass (kg) 229.7 223.9 – 2.53 %

where ui is the displacement caused by the corresponding study:

lumped force Fi and where nf is the number of lumped (1) Three-cell cross section has more potential application
forces. in automotive frame. The torsional rigidity of three-cell
The dynamic frequency stiffness of the whole carbody is cross section is formulated for the first time in this
generally measured by the structural frequency ωi from the paper.
modal analysis by solving the eigenvalue equation (2) The cross-sectional shape optimization model is effective
to achieve a high stiffness and lightweight TWB
(K − ω i2M )ui = 0 for i = 1,..., m (25)
structure of automotive body. It is unnecessary to
where K is the structural stiffness matrix, M is the mass manually modify the optimal cross-sectional shape
matrix, ωi is the i-th angular frequency, ui is the i-th mode because the stamping demands, such as draft angle,
shape and m is the number of specified vibration modes. chamfer radius and assembly constraints, are considered
The dynamic frequency stiffness is crucial for evaluating during the optimization design process.
the vibration performance, so as to avoid resonance. (3) The optimization model solved by the GA is reliable
The mass of the frame structures is calculated as and efficient. When the optimal solution is obtained,
b the computational cost is 104 seconds using a laptop
Mass = ∑ Li × Ai × ρ (26) computer. More importantly, the cross-sectional shape
i =1
optimization method is successfully applied in the
where Li and Ai are the length and cross-sectional area of structural modification of the automotive frame.
the i-th thin-walled beam, respectively. b is the number of (4) Finally, the user manual of SuperSection software can
thin-walled beams. The material density ρ of thin-walled be downloaded from the website: http://mach.jlu.edu.cn/
beams is 7.83 × 10−6 kg/mm3. hb_images/xygk/xssz_sz_js.php?id=395
By using the proposed method, the optimal shapes for
the four TWBs are compared with the initial ones in Figure ACKNOWLEDGEMENT−This work was supported by the
11. Although the enclosed area of the optimal solution National Natural Science Foundation of China (No. 51575226),
increases, the thickness of sheets decreases, so the cross- the Plan for Scientific and Technological Development of Jilin
sectional areas are all reduced in Figure 11. The optimal Province (No. 20140101071JC) and the Foundation of Key
cross sections are adopted to update the initial ones of the Laboratory of Advanced Manufacture Technology for Automobile
Parts, Ministry of Education (No. 2014KLMT01).
automotive frame in Figures 9 and 10. Subsequently, the
performance of the automotive frame is improved, as listed
in Table 4. The structural mass of automotive body is REFERENCES
reduced by 2.53 %. Meanwhile, the torsional, bending, and
frequency stiffness are all improved. During the optimization Gauchia, A., Diaz, V., Boada, M. J. L. and Boada, B. L.
process, the stamping manufacturability including draft (2010). Torsional stiffness and weight optimization of a
angle, chamfer radius and assembly constraints is satisfied. real bus structure. Int. J. Automotive Technology 11, 1,
From these results, it can be concluded that a modified 41−47.
automotive frame with higher stiffness and lighter mass Griffiths, D. R. and Miles, J. C. (2003). Determining the
can be readily obtained by the proposed method. optimal cross-section of beams. Adv. Eng. Inform. 17, 2,
59−76.
5. CONCLUSION Heo, S. J., Kang, D. O., Lee, J. H., Kim, I. H. and Darwish,
S. M. H. (2013). Shape optimization of lower control
The cross-sectional shape optimization and design method arm considering multi-disciplinary constraint condition
of TWBs for automotive frame is proposed in this paper. by using progress meta-model method. Int. J. Automotive
The following conclusions can be summarized from this Technology 14, 3, 499−505.
CROSS-SECTIONAL SHAPE DESIGN AND OPTIMIZATION OF AUTOMOTIVE BODY WITH STAMPING CONSTRAINTS 1011

Kim, Y. Y. and Kim, T. S. (2000). Topology optimization of analysis considering nonlinear behavior. SAE Paper No.
beam cross sections. Int. J. Solids Struct. 37, 3, 477−493. 2004-01-1659.
Kunishi, D. and Kikuchi, N. (2004). Analysis of FEM Wang, H., Li, E. Y. and Li, G. Y. (2011). Probability-based
results based upon FOA. SAE Paper No. 2004-01-1729. least square support vector regression metamodeling
Lamberti, L. and Pappalettere, C. (2002). Design technique for crashworthiness optimization problems.
optimization of large-scale structures with sequential Computational Mechanics 47, 3, 251−263.
linear programming. Proc. Institution of Mechanical Wang, H., Li, G. Y. and Li, E. Y. (2010). Time-based
Engineers, Part C: J. Mechanical Engineering Science metamodeling technique for vehicle crashworthiness
216, 8, 799−811. optimization. Comput. Method. Appl. M. 199, 37-40,
Lamberti, L. and Pappalettere, C. (2005). An efficient 2497−2509.
sequential linear programming algorithm for engineering Yoshimura, M., Nishiwaki, S. and Izui, K. (2005). A
optimization. J. Eng. Design 16, 3, 353−371. multiple cross-sectional shape optimization method for
Nakagawa, T., Nishigaki, H., Tsurumi, Y. and Kikuchi, N. automotive body frames. J. Mech. Design 127, 1, 49−57.
(2004). First order analysis for automotive body structure Yu, K. G. and Yang, Z. H. (2015). Assembly variation
design - Part 4: Noise and vibration analysis applied to a modeling method research of compliant automobile
subframe. SAE Paper No. 2004-01-1661. body sheet metal parts using the finite element method.
Nishigaki, H., Amago, T., Sugiura, H., Kojima, Y., Int. J. Automotive Technology 16, 1, 51−56.
Nishiwaki, S. and Kikuchi, N. (2004). First order Zhou, Q., Wu, X., Xia, Y. and Cai, W. (2014). Spot weld
analysis for automotive body structure design - Part 1: layout optimization of tube crash performance with
Overview and applications. SAE Paper No. 2004-01- manufacturing constraints. J. Manuf. Sci. Eng. 136, 1, 1−
1658. 10.
Nishigaki, H. and Kikuchi, N. (2004). First order analysis Zuo, W. J. (2013). An object-oriented graphics interface
for automotive body structure design - Part 3: design and optimization software for cross-sectional
Crashworthiness analysis using beam elements. SAE shape of automobile body. Adv. Eng. Softw., 64, 1−10.
Paper No. 2004-01-1660. Zuo, W. J. (2015). Bi-level optimization for the cross-
Nishigaki, H., Nishiwaki, S., Amago, T., Kojima, Y. and sectional shape of a thin-walled car body frame with
Kikuchi, N. (2001). First order analysis - New CAE static stiffness and dynamic frequency stiffness
tools for automotive body designers. SAE Paper No. constraints. Proc. Institution of Mechanical Engineers,
2001-01-0768. Part D: J. Automobile Engineering 229, 8, 1046−1059.
Song, J. H., Lee, G. A., Jung, K. H. and Park, S. J. (2015). Zuo, W. J., Bai, J. T. and Li, B. (2014). A hybrid OC-GA
Laser irradiated bending characteristics of the ultra-high approach for fast and global truss optimization with
strength steel sheets. Int. J. Automotive Technology 16, frequency constraints. Appl. Soft. Comput., 14, 528−535.
1, 89−96. Zuo, W. J., Li, W. W., Xu, T., Xuan, S. Y. and Na, J. X.
Suh, J., Lee, J. H., Kang, H. S., Park, K. T., Kim, J. S., Lee, (2012). A complete development process of finite
M. Y. and Jung, B. H. (2006). Optimal processing and element software for body-in-white structure with semi-
system manufacturing of a laser welded tube for an rigid beams in .NET framework. Adv. Eng. Softw. 45, 1,
automobile bumper beam. Int. J. Automotive Technology 261−271.
7, 2, 209−216. Zuo, W. J., Xu, T., Zhang, H. and Xu, T. S. (2011). Fast
Tovey, M. (2002). Concept design CAD for the automotive structural optimization with frequency constraints by
industry. J. Eng. Design 13, 1, 5−18. genetic algorithm using adaptive eigenvalue reanalysis
Tsurumi, Y., Nishigaki, H., Nakagawa, T., Amago, T., methods. Structural and Multidisciplinary Optimization
Furusu, K. and Kikuchi, N. (2004). First order analysis 43, 6, 799−810.
for automotive body structure design - Part 2: Joint