You are on page 1of 14

Original Article

Proc IMechE Part D:


J Automobile Engineering
1–14
Robustness optimization of engine Ó IMechE 2018
Reprints and permissions:
mounting system based on Six Sigma sagepub.co.uk/journalsPermissions.nav
DOI: 10.1177/0954407018755247

and torque roll axis decoupling method journals.sagepub.com/home/pid

Jian Wu1,2, Xiandong Liu1,2 , Yingchun Shan1,2 and Tian He1,2

Abstract
Robustness optimization to enhance the reliability of an engine mounting system (EMS) is necessary in the decoupling
design of the EMS, because the stochastic factors or uncertainties will have a significant influence on the decoupling state
of the system and decrease the reliability of the system. This paper carries out the robustness optimization on the
decoupling design based on torque roll axis decoupling theory. After building the dynamic model of the EMS, the con-
straints and objective functions of the optimization are improved by the principle of Six Sigma. A complete process of
robustness is built and the examples for three different types of EMS are calculated and analyzed. The results show that
the quality levels have been enhanced by the robustness optimization, which means the design of EMS is more reliable
than deterministic design. Although the degree of the decoupling state lowers slightly, by taking all the performance of
the system into consideration, the decrease of decoupling is acceptable and the robustness design is valuable in the real
engineering.

Keywords
Robustness optimization, engine mounting system, Six Sigma, torque roll axis, decoupling

Date received: 4 July 2017; accepted: 19 December 2017

Introduction mounting system was used to validate the proposed


method and the results showed that the optimum
An engine mounting system (EMS) is a device including robust method provided a higher sigma level than tradi-
several mounting components connecting an engine to tional deterministic design. However, almost all those
the frame (or sub-frame) in order to isolate the vibra- existing researches on the robustness of the EMS are
tion of the engine. By adjusting the parameters of the based on the energy decoupling theory, which takes the
mounting components, the vibration of the engine may modal kinetic energy as the criterion of decoupling
be decoupled very well. Until now, several theories state. In the past 20 years, the TRA decoupling method
about the decoupling of an EMS have been well devel- has been well studied and used. In 1999, Jeong and
oped. The torque roll axis (TRA) decoupling method, Singh established the model of the EMS and deduced
which theoretically can achieve complete decoupling, is the decoupling conditions based on TRA theory for a
one of the most effective methods. In 2002, Qatu et al. proportionally damping system.4 Their work laid the
studied the effect of the powertrain mount stiffness var- foundation for later research. In 2008 and 2012, Hu
iations on idle noise, vibration, and harshness (NVH) and Park separately made great improvement on Jeong
performance of vehicle with a simple six-degrees-of-
freedom (6-DOF) model and general experience-based 1
School of Transportation Science and Engineering, Beihang University,
criteria.1 The conclusions were obtained that the varia-
Beijing, China
tion of the mount stiffness would yield gross influence 2
Beijing Key Laboratory for High-efficient Power Transmission and
on the NVH performance of the vehicle on idle and fur- System Control of New Energy Resource Vehicle, Beihang University,
ther research on the robustness of NVH of the vehicle Beijing, China
was recommended. In 2008, Wu and Shangguan carried
Corresponding author:
out robustness optimization on the EMS to increase the Xiandong Liu, School of Transportation Science and Engineering, Beihang
robustness of the EMS natural frequencies, frequency University, Beijing, 100191, China.
separation, and decoupling ratio.2,3 A powertrain Email: liuxiandong@buaa.edu.cn
2 Proc IMechE Part D: J Automobile Engineering 00(0)

and Singh’s work by taking non-proportional damping adapted in this paper, the optimization process is also
and compliant base into consideration and made the effective for the design with other quality level.
model of NVH more authentic.5–7 Wu and Liu dis-
cussed the effect of damping and compliant base by
Six Sigma optimization method
analyzing and comparing three typical models,8 and the
decoupling condition illustrates that the TRA decou- Figure 1 shows that deterministic optimization is
pling method is effective for all these models. These always close to the constraint boundary, which leads to
studies, which are based on Geck and Patton’s early high risk of failure when the design variable gets a
research on engine mount optimization, 9 have illu- small change. The goal of the robust optimization
strated that the TRA decoupling method can obtain a design is not only to optimize the performance, but also
more thorough decoupling state and even a theoretical to desensitize the performance to the fluctuations of
complete decoupling for an EMS. the design variables.2 A robustness solution may not
However, in reality, the parameters of the mounting lead to the best performance, but it possesses both high
components are difficult to be exactly realized due to reliability and relatively good performance.
some uncertain reasons in manufacture and assembly. The Six Sigma method is one of the robustness opti-
For example, the stiffness of the mounting components mization methods, where sigma refers to standard
may fluctuate around a certain value due to the instabil- deviation s. It brings the probabilistic model into the
ity of the rubber materials, and the real location of the optimization model in the beginning of the design, and
mounting may be different from the optimal value evaluates the reliability of the system by measuring the
because of installation errors. These uncertainties may probability distribution of performance index around
have a significant influence on the decoupling state of the mean value.14–16 Generally, it can be assumed that
the system, so the fluctuation of the parameters should the parameters of the system conform to normal distri-
be taken into consideration to make the optimization bution. Thus, the probability of failure, which can
model more authentic and reliable. Obviously the reflect the reliability of the system, is related to the s
robustness optimization of the EMS is critically neces- level
sary. Studies have been done about the robustness of an As shown in Figure 2, different sigma levels repre-
EMS and some progress has been achieved. The sent different probabilities of normal distribution.
research of Sirafi and Chang on the robustness of For example, 3 s is equivalent to a probability of
mount system did a valuable analysis about the NVH 99.73%, and 6 s is equivalent to the probability of
performance on idle condition.10 The optimization 99.9999998%. So 6 s represents a very high probability
method used in their paper was well studied by Jiang that the objective function value is in feasible region
et al.,11 who did research about the programming even in a long term or massive production.
method for uncertain optimization problems, and Qiu
and Wang,12 who compared the dynamic response of
structures with uncertain non-probabilistic interval
analysis method and probabilistic approach. The
research of Vochteloo et al. about a multidisciplinary
optimization method was also a great supplement of
robustness optimization.13
This paper aims at the application of robustness
optimization on TRA decoupling. Three typical kinds
of EMS are used to validate the method. The best com-
bination of parameters of the mounting components
can be obtained by the TRA decoupling process. After
conducting the Six Sigma analysis, it is found that the
Figure 1. Process of deterministic optimization and robustness
reliability of the original system is pretty low. So the
optimization.
robustness optimization is carried out on the determi-
nistic optimized system. First, the sensitivity analysis is
made to choose the most influential parameters as the
optimization variables. Then the Six Sigma optimiza-
tion process is carried out by Isight software and new
optimum parameters are obtained. Finally, the quality
level and frequency response functions (FRFs) of the
new EMS are calculated to compare with the determi-
nistic optimized system. The results show that although
the decoupling state of EMS is slightly damaged, the
reliability of the system get enhanced, which makes the
design more feasible. Although Six Sigma principle is Figure 2. Probability of different levels of s.
Wu et al. 3

Figure 3. Model of the engine mounting system.

In the optimization process, the objective function KqTRA = lk MqTRA ð3Þ


and the constraint of the model should be adjusted
For a non-proportional damped EMS with rigid base,
according to the Six Sigma principle. The optimization
the decoupling conditions can be written as
formulation can be written as

KqTRA = lk MqTRA
min F(my (X), sy (X)) ð1Þ ð4Þ
 CN qTRA = lc MqTRA
g (m (X), sy (X))40(i = 1, 2, :::, m)
s:t: i y ð2Þ For a non-proportional damped EMS with compliant
XL + nsX 4mX 4XU  nsX
base, the decoupling conditions can be written as
where X is the solution of the problem, m and s are the 8 p
mean value and standard deviation, respectively, F is >
> K q = lpk Mp qTRA
< p TRA
the objective function, and g is the constraint of the C qTRA = lpc Mp qTRA
ð5Þ
problem. When n = 6, the constraint of 6 s is added to >
> Kbp q =0
: bp TRA
the optimization problem, and in the objective func- C qTRA = 0
tion, not only the minimum mean value, but also the
Among these equalities, K, M, and C are the stiffness,
minimum standard deviation, is required.
mass, and damping matrices of the system, qTRA is the
TRA vector, the subscript N refers to non-proportional
Modeling and 6s optimization of an EMS damping, and in equation (5) the different matrices of
based on TRA theory powertrain and the base are distinguished by super-
scripts p, b, and bp.
It has been well illustrated in the literature about These TRA decoupling conditions shown above are
how to build the model for the EMS based on TRA equalities, and include many variables connected by a
theory,4–6 and a similar method will be used in this complicated function. Obviously, it is very difficult to
paper. As shown in Figure 3, different coordinate sys- obtain a set of parameters strictly satisfying these con-
tems are built in the model of the EMS, e.g., mounting ditions due to the interaction among the parameters.
coordinate system built on the mounting components, Therefore, these decoupling conditions are adjusted,
inertial coordinates system built on the center of grav- and the objective functions suitable for the optimiza-
ity of the engine, and the TRA coordinate system built tion are established as follows.
by the TRA. The details of these coordinate systems For an undamped EMS with rigid base, the objective
are well illustrated elsewhere and will not be repeated function can be written as
here.4–6
The literature shows that the decoupling states of an f1 = 1  cos \ qmode , qTRA . ð6Þ
EMS are affected by many factors,4–6 while non-
where \  ,  . is the angle between these two vectors.
proportional damping of the mounting components
For a non-proportional damped EMS with rigid
and the elasticity of the foundation are two of the most
base, the problem can be viewed as multi-objective opti-
important factors. Thus, three types of model, includ-
mization and the objective functions can be written as
ing undamped EMS with rigid base, non-proportional
damped EMS with rigid base, and non-proportional 
f1 = 1  cos \ KqTRA , MqTRA .
damped EMS with compliant base, are built and dis- ð7Þ
f2 = 1  cos \ CqTRA , MqTRA .
cussed in this paper. Meanwhile, the decoupling condi-
tions of these three types of model can be also obtained where KqTRA , MqTRA , CqTRA are the vectors in the
as follows. decoupling condition, respectively.
For an undamped EMS with rigid base, the decou- For a non-proportional damped EMS with compli-
pling conditions can be written as ant base, the problem also can be viewed as
4 Proc IMechE Part D: J Automobile Engineering 00(0)

Figure 4. Process of Six Sigma analysis in Isight.


Figure 5. Process of sensitivity analysis in Isight.

multi-objective optimization and the objective func-


tions can be written as this process, the coefficient of standard deviation (the
8 ratio of standard deviation to mean value) is set to be
>
> f 1 = 1  cos \ Kp qTRA , Mp qTRA . 10%.18 The flow chart in Isight is shown in Figure 6.
< f 2 = 1  cos \ Cp q p
 bp  TRA , M qTRA . Finally, the robustness of the system after uncertain

f 3 = K qTRA   ð8Þ
>
> optimization is also evaluated with Six Sigma analysis
:
f 4 = Cbp qTRA  and the result is compared with the deterministic opti-
mized system. Meanwhile, the FRF of the new system
where kk refers to Euclid norm.
can be calculated to illustrate the decoupling state of
The objective functions above have been improved,
EMS and the results can be compared to the former
and smaller value of the function corresponds to the
system.
higher degree of the decoupling. According to the the-
ory of Six Sigma method, the normal distribution of
objective function can be obtained by mean value and
standard deviation, and the sigma level will be deter-
mined by the limit range of the objective function. For
equations (6) and (7) and f1 and f2 of equation (8), the
limit range of the functions is decided as [0,2] based on
the physical meaning of cosine of angle between the
two vectors. For f3 and f4 of equation (8), the limit
range of the function is over 0.
All the Six Sigma analyses and optimizations of each
type of the model are carried out on the basis of the
objective functions above. Six Sigma analysis is the eva-
luation process of the robustness of these objective
functions, and in the Six Sigma optimization, an addi-
tional constraint is that the quality levels of these objec-
tive functions is 6 s at least.
In the Isight optimization platform, the whole pro-
cess can be realized by several steps. First, for a deter-
ministic optimized EMS, the Six Sigma analysis is
carried out and the quality level of the system is
obtained, as shown in Figure 4.
If the quality level is \ 6 s, the Six Sigma optimiza-
tion can be carried out to enhance the robustness of the
Figure 6. Process of Six Sigma optimization in Isight.
system. Based on the deterministic optimized para-
meters, a Mont Carlo analysis is carried out as in the
Optimization results and analysis of three
Figure 5 to analyze the sensitivity of the parameters.17
The most sensitive parameters influencing the objective
types of EMS
function strongly are chosen as the variables of the Six In this section, three models for the three different types
Sigma optimization. of EMS are chosen, and all the steps mentioned above
Then, the Six Sigma optimization is applied to the are carried out. The results and the analyses are as
EMS by adjusting the selected sensitive parameters. In follows.
Wu et al. 5

Table 1. Stiffness and coordinates of the mounting components in deterministic design.

Direction Stiffness Coordinates (m)


(105 N m21)
Mount 1 Mount 2 Mount 3 Mount 4

x 9.3499 0.2518 0.2518 20.2518 20.2518


y 3.7645 20.1713 0.1824 20.1491 0.1446
z 3.6774 20.0595 20.0741 20.0820 20.0900

Figure 7. Frequency response function of Model A in deterministic design.

Undamped EMS with rigid base model However, after carrying out Six Sigma analysis on
First, for the example of undamped EMS with rigid the deterministic optimized system, it is found that the
base denoted as Model A, the mass matrix is4 quality level (Figure 8) of the system is only 1.683s,
which is much smaller than 6s, indicating that the sys-
2 3
73:2 0 0 0 0 0 tem has low reliability.
6 0 73:2 0 0 0 0 7 Therefore, it is necessary to apply the robustness
6 7
6 0 0 73:2 0 0 0 7 optimization to the system to enhance the reliability of
M=6 6 7
6 0 0 0 1:94 0:129 0:415 77 the design. Before the optimization, a Mont Carlo anal-
4 0 0 0 0:129 3:43 0:073 5 ysis is applied to the system to test the sensitivity of the
0 0 0 0:415 0:073 3:39 parameters. The result is shown in Figure 9, and in this
ð9Þ model the most sensitive parameters x, y1, y2, y3, y4, kx,
and kz are chosen as the design variables of the robust-
Based on TRA decoupling theory, the deterministic ness optimization. Here, x is the X coordinate of the
optimization is carried out on Model A, the optimized mounting component, y1, y2, y3, and y4 are the Y coor-
values of the stiffness and coordinates of the mounting dinates of the first, second, third, and fourth mounting
components are obtained and listed in Table 1. components, and kx and kz are the stiffnesses of the
With the deterministic design of the EMS, the FRF mounting components in the X and Z directions.4–6
(Figure 7) can be calculated and the results show that In the Isight software, the Six Sigma optimization is
the responses in other directions are much smaller than carried out on the system with the constraint of 6 s
the response in ux direction, which means that the sys- quality level. The result is shown in Figure 10 and a
tem is well decoupled. comparison of the parameters with those from the
6 Proc IMechE Part D: J Automobile Engineering 00(0)

Figure 10. Quality level of Model A in robustness design.


Figure 8. Quality level of Model A in deterministic design.

Table 2. Comparison of parameters before and after


robustness optimization.

Parameters Deterministic Robustness


optimization optimization

x 0.2518 0.4378
y1 0.1713 0.1772
y2 0.1824 0.1180
y3 0.1491 0.1008
y4 0.1446 0.1726
kx (3106) 9.3499 10.5318
kz (3106) 3.6774 3.6505

As the price of enhancing the reliability of the design,


the lower decoupling degree of the system is foresee-
able. By taking all the performance of the system into
consideration, the robustness optimized system is more
practical and feasible for the design of an EMS.

Non-proportional damped EMS with rigid base


Figure 9. Result of sensitivity analysis of Model A.
Secondly, the example of a non-proportional damped
deterministic optimization are listed in Table 2. The EMS with rigid base denoted as Model B is studied
obvious conclusion is that the quality level has been and optimized in robustness. The results are listed as
enhanced to 6.019s, which satisfies the 6s requirement. follows. It shows that the system remains in a good
In order to observe the change of the decoupling decoupled state (better than Model A) when the 6s
state of the system, FRFs of the robustness optimized constraint is satisfied.
model are calculated, as shown in Figure 11. For the example non-proportional damped EMS
It can be seen from Figure 11 that the responses in with rigid base, the mass matrix is4
X, Y, Z, and uy increase, which means that the decou- 2 3
73:2 0 0 0 0 0
pling state of the system is not as good as before. 6 0
6 73:2 0 0 0 0 7 7
However, it also can be found that the responses in X, 6 0 0 73:2 0 0 0 7
Y, Z, and uy directions appear in almost the same fre- M=6 6 0
7
6 0 0 1:94 0:129 0:415 7 7
quency range with the response in ux direction, but the 4 0 0 0 0:129 3:43 0:073 5
amplitudes are much smaller than that in the ux direc-
0 0 0 0:415 0:073 3:39
tion. Therefore, the system can be viewed as decoupled
approximately. ð10Þ
Wu et al. 7

Figure 11. Frequency response function of Model A in robustness design.

Table 3. Stiffness and coordinates of the mounting components in deterministic design.

Direction Stiffness (105 N m21) Damping (102 N s m21) Coordinates (m)


Mount 1 Mount 2 Mount 3 Mount 4

X 8.4699 14.8187 0.3225 0.3225 20.3225 20.3225


Y 3.3120 5.7590 20.1800 0.1900 20.1466 0.1479
Z 2.5308 10.0010 20.0600 20.0800 20.0850 20.0997

For this model, the deterministic optimized parameters


are listed in Table 3, and FRFs of the system are shown
in Figure 12, which also indicates the system is well
decoupled.
After the Six Sigma analysis, the quality levels for
two objective functions of the deterministic design are
found to be 4.479 s and 4.914 s, respectively, as shown
in Figure 13. The 6 s requirement is not satisfied, and
the Six Sigma optimization is necessary.
The sensitivity analysis of the parameter is carried
out and the results are shown in Figure 14. For this
model, parameters x, y4, y1, y3, and z4 are chosen as
the design variables of the robustness optimization and
the optimized results are shown in Figure 15. Here, x is
the X coordinate of the mounting components, y1, y3,
and y4 are the Y coordinates of first, third, and fourth
mounting components, and z4 is the Z coordinate of
Figure 12. Frequency response function of Model B in
the fourth mounting component.4–6
deterministic design.
8 Proc IMechE Part D: J Automobile Engineering 00(0)

Figure 13. Quality level of Model B in deterministic design.

Figure 14. Results of sensitivity analysis of Model B.

Figure 15. Quality level of Model B in robustness design.


Wu et al. 9

Figure 16. Frequency response function of Model B in robustness design.

The quality levels of the objective function f1 and f2


Table 4. Comparison of parameters before and after
are enhanced to 6.318 s and 7.018 s, respectively. A robustness optimization.
comparison of the parameters with the deterministic
optimization is listed in Table 4. FRFs in Figure 16 Parameters Deterministic Robustness
show that the coupled responses in other directions are optimization optimization
much smaller than that in ux direction, so this design of
x 0.3225 0.2556
EMS is acceptable in the decouple state (better than y1 0.1800 0.1641
Model A) and feasible in real design. y3 0.1466 0.1966
y4 0.1479 0.1787
z4 0.0997 0.0786
Non-proportional damped EMS with compliant base
Finally, for the example of a non-proportional damped
EMS with compliant base, the mass matrix is6
deterministic design are shown in Table 5 and
2 3 Figure 17, respectively.
73:2 0 0 0 0 0
6 0 73:2 0 0 0 0 7 The results of Six Sigma analysis of the system are
6 7 shown in Figure 18, and the quality levels for two of the
6 0 0 73:2 0 0 0 7
Mp = 6
6
7
6 0 0 0 1:94 0:129 0:415 7
7
four objective functions do not meet the criterion of 6s.
4 0 0 0 0:129 3:43 0:073 5 To make the system more reliable, the Six Sigma opti-
0 0 0 0:415 0:073 3:39 mization is carried out.
ð11Þ The sensitivity test results are shown in Figure 19,
and cz, xp0, kpx, kbx, yb0, and zb0 are chosen as the design
2 3
52:1 0 0 0 0 0 variables. Here, cz is the damping in the Z direction of
6 0 52:1 0 0 0 0 7 the mounting components, xp0 is the X coordinate of
6 7
6 0 0 52:1 0 0 0 7 the powertrain in the inertia coordinates system, kpx is
M =6
b
6
7 ð12Þ
6 0 0 0 3:8 0:35 0:85 7
7 the stiffness in the X direction of mounting compo-
4 0 0 0 0:35 1:6 0:25 5 nents, kbx is the stiffness in X direction of bushing com-
0 0 0 0:85 0:25 3:1 ponents of base, and yb0 and zb0 are the Y and Z
coordinates of bushing components in base coordinates
where Mp and Mb refer to the inertia matrix of engine system, respectively.4–6
and base, respectively. The parameters and FRF of
10 Proc IMechE Part D: J Automobile Engineering 00(0)

Figure 17. Frequency response function of Model C in deterministic design.

Table 5. (a) Stiffness and damping of components in Model C. (b) Coordinates of components in Model C.
(a).

Direction Stiffness of mounting Damping of mounting Stiffness of bushing Damping of bushing


(105 N m21) (102 N s m21) (105 N m21) (102 N s m21)

X 5.2402 7.7527 2.0198 7.7527


Y 3.8046 12.2066 1.7028 12.2066
Z 2.9469 8.8425 4.2945 8.8425

(b)
Direction X Y Z

Coordinates of mounting (m) Mount 1 0.4909 0.2802 20.0117


Mount 2 20.4909 0.2802 20.0117
Mount 3 0.4909 20.2802 20.0117
Mount 4 20.4909 20.2802 20.0117
Coordinates of bushing (m) Bushing 1 0.5663 0.8892 20.0562
Bushing 2 20.5663 0.8892 20.0562
Bushing 3 0.5663 20.8892 20.0562
Bushing 4 20.5663 20.8892 20.0562
Coordinates of mounting in bushing CS (m) Mount 1 0.7660 0.2606 20.2066
Mount 2 20.2158 0.2606 20.2066
Mount 3 0.7660 20.2997 20.2066
Mount 4 20.2158 20.2997 20.2066

After robustness optimization and recalculating FRFs the robustness of the system has been enhanced a lot.
of the system, a comparison of the parameters with the The 6s constraint is satisfied for each objective function
deterministic optimization is listed in Table 6, and con- and the FRFs are very close to Figure 17. The result is
clusions can be obtained from Figures 20 to 21 that the even better than both Model A and Model B, which
decouple state of the system remains good enough while means this design is also effective in a real situation.
Wu et al. 11

Figure 18. Quality level of Model C in deterministic design.

Figure 19. Results of sensitivity analysis of Model C.


12 Proc IMechE Part D: J Automobile Engineering 00(0)

Figure 20. Quality level of Model C in robustness design.

Figure 21. Frequency response function of Model C in robustness design.


Wu et al. 13

Table 6. Comparison of parameters before and after robustness optimization.

Parameters Deterministic optimization Robustness optimization

kp0 (3106) 5.2402 8.9559


kb0 (3106) 2.0198 4.6116
xp0 0.4909 0.7729
yb0 0.8892 0.6469
zb0 0.0562 0.4542
cz (3102) 8.8425 19.9366

The three examples above take different factors into Declaration of conflicting interests
consideration for the EMS models, such as non- The author(s) declared no potential conflicts of interest
proportional damping and compliant base, and the with respect to the research, authorship, and/or publi-
similar optimization process is carried out. From above cation of this article.
results, some conclusions about the robustness optimi-
zation can be obtained, which may provide an impor-
tant guidance to the engineering design. Funding
The author(s) received no financial support for the
research, authorship, and/or publication of this article.
Conclusions
In this paper, the robustness optimization is combined ORCID iD
with TRA decoupling method to enhance the reliability Xiandong Liu https://orcid.org/0000-0002-0111-3221
of the decoupling design of EMS. Firstly, on the basis
of TRA decoupling theory, some improvements are
References
made to the objective functions of the optimization
including Six Sigma principle constraint. Then, by the 1. Qatu M, Sirafi M and Johns F. Robustness of power-
combination of MATLAB and Isight software, a com- train mount system for noise, vibration and harshness at
idle. Proc IMechE Part D: J Automobile Eng 2002; 216:
plete process of the robustness optimization is designed
805–810.
for the EMS. Finally, three types of EMSs are respec- 2. Wu J and Shangguan WB. Robust optimization design
tively optimized and the results are analyzed. The fol- method for powertrain mounting systems based on six
lowing conclusions may be noted. sigma quality control criteria. Int J Automot Technol
2010; 11: 651–658.
1. Generally, the results from the deterministic design 3. Wu J. A robust optimization for the frequency and
for EMS cannot be directly used since the reliabil- decoupling ratio of a powertrain mounting system based
ity of the design may be pretty low and the risk of on interval analysis. Int J Automot Technol 2012; 13:
409–422.
failure may exist in the design.
4. Jeong T and Singh R. Analytical methods of decoupling
2. The proposed robustness optimization process
the automotive engine torque roll axis. J Sound Vib 2000;
based on TRA decoupling method and Six Sigma 234: 85–114.
principle can obviously enhance the reliability of 5. Hu JF and Singh R. Improved torque roll axis decou-
the system and make the quality level satisfy the pling axiom for a powertrain mounting system in the
6s principle. presence of a compliant base. J Sound Vib 2012; 331:
3. The imperfections existing in this optimization is 1498–1518.
that the decoupling state may not be as good 6. Park JY and Singh R. Effect of non-proportional damp-
enough as the deterministic design. However, the ing on the torque roll axis decoupling of an engine
results of the three examples show that the decou- mounting system. J Sound Vib 2008; 313: 841–857.
pling state is acceptable. Therefore, the robustness 7. Park JY and Singh R. Role of spectrally varying mount
design should be more feasible in the real engineer- properties in influencing coupling between powertrain
ing than the deterministic design. motions under torque excitation. J Sound Vib 2010; 329:
2895–2914.
8. Wu J, Shan YC, Liu XD, et al. Parameter optimization
The Six Sigma principle is adopted in this paper, but design of TRA decoupled engine mounting system. Int J
other quality levels can also be used as the criterion Veh Des 2017; 74: 41–61.
with the proposed optimization process. Meanwhile, 9. Geck PE and Patton RD. Front wheel drive engine
this robustness optimization process can also provide mount optimization. SAE paper 840736, 1984.
guidance for other dynamic system design. 10. Sirafi M, Chang YP and Qatu MS. Robustness of mount
systems for idle NVH, part I: centre of gravity (CG)
mounts. Int J Veh Noise Vib 2006; 2: 317–333.
14 Proc IMechE Part D: J Automobile Engineering 00(0)

11. Jiang C, Han X, Liu GR, et al. A nonlinear interval num- 15. Koch PN, Evans JP and Powell D. Interdigitation for
ber programming method for uncertain optimization effective design space exploration using iSIGHT. Struct
problems. Eur J Oper Res 2008; 188: 1–13. Multidiscip Optim 2002; 23: 111–126.
12. Qiu Z and Wang X. Comparison of dynamic response of 16. Koch PN and Kodiyalam S. Variable complexity struc-
structures with uncertain non-probabilistic interval analy- tural reliability analysis for efficient reliability-based
sis method and probabilistic approach. Int J Solids Struct design optimization. In: Structures, structural dynamics,
2003; 40: 5423–5439. and materials conference and exhibit, St. Louis, MO,
13. Vochteloo AJ, Moerman S, Tuinebreijer WE, et al. Mul- April 12–15, 1999.
tidisciplinary design optimization of engine mounts with 17. Lai YY. Isight parameter optimization theory and example
consideration of the driveline. Proc IMechE Part D: J explanation. Beijing: Beihang University Press, 2012.
Automobile Eng 2003; 217: 107–114. 18. Wang XY, Liu XD, Shan YC, et al. Lightweight design
14. Koch PN, Yang RJ and Gu L. Design for Six Sigma of automotive wheel made of long glass fiber reinforced
through robust optimization. Struct Multidiscip Optim thermoplastic. Proc IMechE Part C: J Mech Eng Sci
2004; 26: 235–248. 2016; 230: 1634–1643.

You might also like