Mathematical

and Computational
Applications

Article
Multi-Objective Optimization of a Spring
Diaphragm Clutch on an Automobile Based on the
Non-Dominated Sorting Genetic
Algorithm (NSGA-II)
Junchao Zhou 1,2, *, Chun Wang 1,2 and Junjun Zhu 3
1 School of Mechanical Engineering, Sichuan University of Science and Engineering, Zigong 643000, China
2 Sichuan Provincial Key Lab of Process Equipment and Control Engineering, Zigong 643000, China;
[email protected]
3 School of Mechanical and Vehicular Engineering, Beijing Institute of Technology, Beijing 100081, China;
[email protected]
* Correspondence: [email protected]

Academic Editor: Fazal M. Mahomed

Received: 25 September 2016; Accepted: 17 November 2016; Published: 3 December 2016

Abstract: The weight coefficients of the diaphragm spring depend on experiences in the traditional
optimization. However, this method not only cannot guarantee the optimal solution but it is also not
universal. Therefore, a new optimization target function is proposed. The new function takes the
minimum of average compress force changing of the spring and the minimum force of the separation
as total objectives. Based on the optimization function, the result of the clutch diaphragm spring in
a car is analyzed by the non-dominated sorting genetic algorithm (NSGA-II) and the solution set of
Pareto is obtained. The results show that the pressing force of the diaphragm spring is improved by
4.09% by the new algorithm and the steering separation force is improved by 6.55%, which has better
stability and steering portability. The problem of the weight coefficient in the traditional empirical
design is solved. The pressing force of the optimized diaphragm spring varied slightly during the
abrasion range of the friction film, and the manipulation became remarkably light.

Keywords: clutch; diaphragm spring; multi-objective; optimization; NSGA-II

1. Introduction
The diaphragm spring clutch is widely used in automation because of good nonlinear
characteristics [1]. The diaphragm spring load-deformation characteristic curve is illustrated in
Figure 1. The point A represents the wear limit point, the point B represents the new working point,
and the point C represents the separating point. λ1B is the diaphragm spring deformation of the
point B. However, there are some problems with the clutch, such as frequent usage, high labor intensity
of operators, incomplete separation during long term operation and unstable connection with easiness
of slipping [2]. Because of the above results, the operating pressure of the clutch pedal stroke and
many other parameters in the design are often unreasonable. This maymean that clutch separation is
not complete and the clutch slips. Crucially, the performance of the clutch diaphragm spring directly
affects the above parameters and it has important practical engineering optimal design implications.

Math. Comput. Appl. 2016, 21, 47; doi:10.3390/mca21040047 www.mdpi.com/journal/mca

Math.
Math.Comput.
Comput.Appl. 2016,21,
Appl.2016, 21,47
47 22ofof10
10

F1

F1A
F1B

 1F

1B 1H 1N 1

Figure 1. Characteristics
Figure of loading-deflection
1. Characteristics diaphragm
of loading‐deflection spring.
diaphragm A: wear
spring. A: limit
wear point; B: newB:working
limit point; new
point; H: working point; C: separating point; N: lowest point; λ
working point; H: working point; C: separating point; N: lowest1B : diaphragm spring deformation
point; λ1B: diaphragm spring of
pointdeformation
B; λ1H : diaphragm
of pointspring
B; λ1H:deformation of point
diaphragm spring H; λ1N : diaphragm
deformation spring
of point H; λ1N: deformation of point
diaphragm spring
N; F1Adeformation
: diaphragmofspring’s
point N; working pressurespring’s
F1A: diaphragm of working point pressure
working A; F1B : diaphragm
of workingspring’s
point A;working
F1B:
pressure of working point B.
diaphragm spring’s working pressure of working point B.

Many scholars
Many scholars still
stillshow
showakeenakeeninterest
interestininthe thediaphragm
diaphragmspring’s spring’soptimization
optimizationdesigndesignand and
havedone much research work on it. The authors of [3,4] take
havedone much research work on it. The authors of [3,4] take ideal comprehensive performance ideal comprehensive performance
requirementsof of
requirements a clutch
a clutch diaphragm
diaphragm springspring as the function
as the objective objectiveand function
establish and establish
a new a new
mathematical
mathematical
model. model.
The authors The present
of [5,6] authorsaof [5,6] present
sensitivity design a sensitivity
method based design method based
on traditional designontheory
traditional
and
design theory and establish an optimization model, considering the
establish an optimization model, considering the controllable variables and uncontrollable parameters controllable variables and
uncontrollable
variation on the parameters
influence of variation
the diaphragmon thespring influence of the diaphragm
optimization. The method spring optimization.
is used in the parallelThe
method is used in the parallel optimization design of the friction type
optimization design of the friction type diaphragm spring clutch. The balance between each optimal diaphragm spring clutch. The
balanceweight
target’s between each optimal
coefficient target’s
is studied weight
in [7,8], thecoefficient
concept of is studied
parallel in [7,8], theand
optimization concept of parallel
its application
optimization
model and its application
is also proposed. The authorsmodel of [9] set is upalso
a new proposed. The authors
multi-objective of [9] design,
optimization set up taking
a new
the minimum of average compressing force of spring within the scope of the friction slice spring
multi‐objective optimization design, taking the minimum of average compressing force of wear
within
and the the scopeminimum
driver’s of the friction slice wear
manipulating forceand the driver’s
on separating minimum
bearings manipulatingobjectives.
as optimization force on
separating
They also makebearings as optimization
some analysis objectives.
using algorithms They also
combined withmake some analysis
the genetic algorithm. using
Thealgorithms
optimum
combined with the genetic algorithm. The optimum mathematical
mathematical model of a light carrying truck’s diaphragm spring clutch is established in Ref. model of a light carrying truck’s
[10].
diaphragm
The spring clutch
optimal solution set of is established in Ref.
a multi-objective [10]. Themodel
optimization optimal of asolution
diaphragm set of a multi‐objective
spring was gained
optimization
using model of a diaphragm
the multidisciplinary optimization spring
softwarewas ISIGHT
gained using
[11,12].the multidisciplinary
According optimization
to the above, it can be
software ISIGHT [11,12]. According to the above, it can be seen that
seen that many scholars have begun to study the theory of optimization design and its applications. many scholars have begun to
study
In theof theory
respect of optimization
multi-objective optimization designdesign, and its applications.
multi-objective In respect
optimization of multi‐objective
is usually transformed
optimization design, multi‐objective optimization is usually transformed
into single optimization by each optimal target’s weight coefficient. However, there is no exact method into single optimization by
each
to optimal
obtain the truetarget’s
optimal weight
value coefficient.
of each optimal However,
target’sthereweightis coefficient
no exact method to obtain
and it usually the true
depends on
optimal
the value
scholar’s of each optimal
experience and should target’s weight many
be adjusted coefficient
times.andThisitmethod
usuallynot depends on theguarantee
only cannot scholar’s
experience
the and should
optimal solution but be adjusted
also many times. This method not only cannot guarantee the optimal
is not universal.
solution
In thisbut also is
paper, wenotareuniversal.
concerned with the automobile clutch diaphragm spring and current related
In
research. We aim to study theconcerned
this paper, we are separationwith stroketheaverage
automobile of theclutch
minimum diaphragm
separationspring and current
of control force
related research. We aim to study the separation stroke average
and the pressing force change minimum as two multi-objective optimization models. At the same of the minimum separation of
control
time, the force
Paretoand the pressing
solution set of theforce change minimum
multi-objectives as two
is obtained multi‐objective
under the NSGA-II. optimization
The problemmodels.of the
At the same time, the Pareto solution set of the multi‐objectives is obtained
weight coefficient in the traditional empirical design is solved. The pressing force of the optimized under the NSGA‐II. The
problem ofspring
diaphragm the weight
variedcoefficient in the the
slightly during traditional
abrasionempirical
range of design is solved.
the friction Thethe
film, and pressing force of
manipulation
the optimized
became remarkably diaphragm spring
light. This varied
method slightly to
is verified during the abrasion
be correct and effective rangebyofsimulations.
the friction film, and
the manipulation became remarkably light. This method is verified to be correct and effective by
simulations.
Math. Comput. Appl. 2016, 21, 47 3 of 10

This paper is organized as follows: in Section2, we provide a theoretical analysis. Section

Math. Comput. Appl. 2016, 21, 47
3
3 of 10
presents the NSGA and multi‐objective solution. The comparison of the NSGA‐II simulations are
shown in Section 4. Section 5 concludes the paper with possible extensions in the future.
This paper is organized as follows: in Section 2, we provide a theoretical analysis. Section 3
2. Diaphragm
presents Spring
the NSGA andMulti‐Objective
multi-objectiveMathematical
solution. TheModel
comparison of the NSGA-II simulations are
shown in Section 4. Section 5 concludes the paper with possible extensions in the future.
2.1. Multiple Objective Functions
2. Diaphragm Spring Multi-Objective Mathematical Model
The main structural parameters of a diaphragm spring are illustrated in Figure 2. The absolute
value of the pressing
2.1. Multiple Objective force |Fb − Fa| is the lowest force to ensure a reliable pressing force in the range
Functions
of the wear limit. At the same time, the average value of the separation force of bearings in the
The main structural parameters of a diaphragm spring are illustrated in Figure 2. The absolute
separation process should be as small as possible to ensure light manipulation. Therefore, the
value of the pressing force |Fb − Fa | is the lowest force to ensure a reliable pressing force in the
absolute value of the pressing force change to separate the spring is selected as the first optimization
range of the wear limit. At the same time, the average value of the separation force of bearings in the
target F1(X) to transfer torque reliably. The minimum average driven disc spring‐loaded force is
separation process should be as small as possible to ensure light manipulation. Therefore, the absolute
chosen as the second optimization target F2(X). Therefore, the objective function can be expressed as
value of the pressing force change to separate the spring is selected as the first optimization target
[8]:
F1 (X) to transfer torque reliably. The minimum average driven disc spring-loaded force is chosen as
F1  X   Fthe
the second optimization target F2 (X). Therefore, b F
objective function can be expressed as [8]:

a
F ( 1 (1)
F12 ( X
X) = | Fb − F
a | i 1  Fbi  Fb 
n
F  min
F=  N (1)
F2 ( X ) = min N1 ∑in=1 ( Fbi − Fb )
where Fb is the diaphragm spring’s working pressure of working point B; Fa is the working pressure
where
of Fb is the diaphragm
diaphragm spring whenspring’s
the working pressure
wear reaches of working
the limit point, B; FaFis
pointand bi the
is working pressure
the pressing of
force
diaphragm spring
corresponding whenN.the wear reaches the limit point, and Fbi is the pressing force corresponding
to point
to point N.

Figure 2. Main
Figure 2. Mainstructural
structuralparameters
parameters of of diaphragm
diaphragm spring.
spring. H: cone
H: cone height
height of theofdisc
thespring;
disc spring; h:
h: spring
spring diaphragm thickness; R 1: radius of the pressure plate’s loading point; r1: radius of loading
diaphragm thickness; R1 : radius of the pressure plate’s loading point; r1 : radius of loading point of the
point
supportof ring;
the support ring;ofrspring
r0 : diameter 0: diameter of spring diaphragm’s small end; rf: action radius of the
diaphragm’s small end; rf : action radius of the separating bearing
separating bearing force; δ 1: width of the small end; δ2: window width; re: radius of the small end; R:
force; δ1 : width of the small end; δ2 : window width; re : radius of the small end; R: big end radius of
big
the end
disc radius
spring;ofr: the disc
small spring;
end radius r: of
small
the end
disc radius
spring.of the disc spring.

2.2.
2.2. Design
Design Variables
Variables
Diaphragm spring design is complex. For the formula of the loading-deflection diaphragm
spring, six variables are needed [9]. The lateral position of the new point B and corresponding
Math. Comput. Appl. 2016, 21, 47 4 of 10

amount of deformation λ1B has a greater impact on the entire characteristic curve. So the main design
parameters are seven parameters according to the objective function [8], namely, h, R, r, R1 , r1 and λ1B .
The diaphragm spring design variables matrix X is determined as:

X = [ x1, x2 , x3 , x4 , x5 , x6 , x7 ] T = [ H, h, R, r, R1 , r1 , λ1B ] T (2)

2.3. Constraints
(1) The new pressing force of the spring Fa should be equal to the required clamping force FY [3]:

Fa = FY (3)

(2) To ensure the stable operation after the friction plate, the pressing force of the spring work after
the damaged F1A should not be less than the new corresponding parameters F1B [3]:

F1A ≥ F1B (4)

(3) Considering the depth-thickness ratio H/h has impactions on the load-deformation curve of the
diaphragm, it should be met within a certain range. At the same time, the diaphragm spring
initial cone angle α0 should be controlled within a certain range [8]. That is:
(
1.7 ≤ H/h ≤ 2.2
◦ ◦ (5)
9 ≤ α0 ≤ 15

(4) For the friction plate on the pressing force distribution, the radius of the big end of the friction
plate R1 should be taken between the mean radius and the outside diameter of the friction plate,
according to engineering experience [9]:
(
pull: ( D + d)/4 ≤ r1 ≤ D/2 − 3
(6)
push : ( D + d)/4 ≤ R1 ≤ D/2 − 3

(5) In order to meet the structural arrangement of the diaphragm spring actual situation, the big
end of the radius R, the support ring radius R1 , load radius r1 and inner diameter r should be in
a certain range [9]: 
 1 ≤ R − R1 ≤ 7

0 ≤ r1 − r ≤ 6 (7)
 0 ≤ r −r ≤ 6

1 0

(6) To ensure the working point, the wear point and separation point should be distributed relatively
reasonabley the new location λ1B should meet the following conditions [8]:

R−r
 
λ1B
0.8 ≤ ≤ 1.0 (8)
H R1 − r1

(7) The work pressing force of the new designed diaphragm FB should be not less than the force FC
in the separation process [9]:
FB ≥ FC (9)

(8) In order to make diaphragm spring satisfy a certain leverage ratio during the separation, the ratio
of the outer diameter to inner diameter should be met [3]:
(
1.2 ≤ 
R/r ≤ 1.3
 (10)
3.0 ≤ r1 − r f /( R1 − r1 ) ≤ 4.0
Math. Comput. Appl. 2016, 21, 47 5 of 10

(9) To take advantage of spring material, part size should meet certain requirements, according to
engineering experience [8]: 
 1.2 ≤ R/r ≤ 1.35

70 ≤ 2R/h ≤ 100 (11)
 3.5 ≤ R/r ≤ 5.0

0

(10) The highest point of the tensile stress σAmax (σCmax ) in the bottom of the A (or C) dangerous parts
of the diaphragm spring separating finger holes should meet a strength condition [9]:
(
σAmax ≤ [σA ]
(12)
σCmax ≤ [σC ]

(11) During the diaphragm spring manufacturing process, there are some major dimensions
machining errors, the error during assembly process should meet certain requirements [9]:

(∆Fh + ∆Ft + ∆FR + ∆Fr ) /Fb ≤ 0.05 (13)

3. NSGA-II Algorithm and Multi-Objective Solution

In 2002, Deb et al. proposed the NSGA-II, and introduced elite strategies to improve the efficiency
of the algorithm searches and robustness [13,14].
In practical engineering, multiple indicators may be involved in many optimization
problems [15,16], e.g., the diaphragm spring requires that not only the absolute value of the pressing
force |Fb − Fa | is the lowest value that ensures a reliable pressing force in the range of the wear limit
but, in addition, the average value of the separation force of bearings in the separation process should
be as small as possible to ensure light manipulation. These two or more design specifications achieve
an optimal value that is termed multi-objective optimization. In the diaphragm spring multi-objective
optimization problem, each target cannot achieve the optimal simultaneously. In the conventional
processing multi-objective optimization problem, the normalization method is usually used, such as
the weighting coefficient method and the hierarchical sequence method, by building an evaluation
function and eventually transforming the multi-objective optimization problem into a single objective
optimization solution. However, the normalization method often relies on the experience of policy
makers. Since each target objection of diaphragm spring is interrelated and constrained, even a target
is improved by reducing another target as the cost [17,18]. Many objective solutions are typically
non-inferior solution sets. Therefore, it is important to search for the Pareto set in the multi-objective
optimization of the diaphragm spring. Figure 3 shows the flow of the NSGA-II algorithm. The main
steps of NSGA-II algorithm used in this paper are as follows: (Initial population) create a random
population of N chromosome in the population. (Calculate the objective function) calculate the
multi-objective function of spring diaphragm. (Satisfied termination criteria) check whether the
termination criterion is satisfied or not. If yes, the population is terminated, otherwise, adjust the
fitness distribution. (Fitness distribution) evaluate the multi-objective fitness of each chromosome in
the population. (Genetic operation and insert parent) generate a new population.
 Math. Comput. Appl. 2016, 21, 47 6 of 10
Math. Comput. Appl. 2016, 21, 47 6 of 10

Initial population

Calculate the
objective function
value

Satisfied Yes
The termination of
Termination
the population
criteria?

No

Fitness distribution

Genetic operation

Insert parent

Figure
Figure3.3.Non‐dominated
Non-dominatedsorting
sortinggenetic
geneticalgorithm
algorithm(NSGA)‐II;
(NSGA)-II;algorithm
algorithmoptimization
optimizationprocess.
process.

4.4.Results
Resultsand
andDiscussion
Discussion
The
Thedesign
designofofcertain
certainnew newtypetype of of
pull diaphragm
pull diaphragm spring
springclutch optimization
clutch is taken
optimization as an
is taken as
example. The parameters are as follows: Maximum engine torque is 1150 N∙m,
an example. The parameters are as follows: Maximum engine torque is 1150 N·m, reserve coefficient reserve coefficient β
isβ 1.8, the friction factor is 0.3, and the diaphragm spring material is 60 si2MnA. Allowable
is 1.8, the friction factor is 0.3, and the diaphragm spring material is 60 si2MnA. Allowable stress stress is
1500–1700
is 1500–1700 MPa; Poisson’s
MPa; Poisson’sratioratio
is 0.3,
is the
0.3, wear limitlimit
the wear is 3.2ismm
3.2 and
mm separation strokestroke
and separation is 3.5 is
mm.3.5 The
mm.
specific target variables of the experiment are shown in Table 1. The NSGA‐II algorithm
The specific target variables of the experiment are shown in Table 1. The NSGA-II algorithm parameter parameter
setting
settingisisshown
shownTable
Table2.2.

Table
Table1.1.Experimental
Experimentalparameters
parametersdesign
designof
ofoptimization
optimizationvariables.
variables.
Plan Objective Function Algorithm
Plan Objective Function Algorithm
A Penalty function method
A B αF1 ( X ) + βF2 ( X ) Penalty
genetic function method
algorithm
B αF1 ( X ) + βF2 ( X ) genetic algorithm
C
C ,
F1 ( X ), F2 ( X )
NSGA‐IINSGA-II
Notes: α, β: weight coefficients, α+β=1; F1(X): target 1; F2(X): target 2.
Notes: α, β: weight coefficients, α + β = 1; F1 (X): target 1; F2 (X): target 2.

Table
Table2.2.NSGA‐II
NSGA-IIalgorithm
algorithmparameter
parametersetting.
setting.
Population Stop Fitness Function Optimal Front End Maximum
SizePopulation
AlgebraStop Fitness Function
Value Deviation Optimal Front End
Individual Coefficient Maximum
Iterative Algebra
Size Algebra Value Deviation Individual Coefficient Iterative Algebra
100 200 1 × 10−100 0.3 200
100 200 1 × 10−100 0.3 200

The model is solved by using the penalty function method [8], the genetic algorithm [9] and
NSGA‐II.The The
model is solved optimization
constrained by using the problem
penalty function method
is transformed [8],anthe
into genetic algorithm
unconstrained [9] and
condition by
NSGA-II.
the penaltyThe constrained
function method. optimization
The geneticproblem
algorithmis has
transformed
very goodinto an unconstrained
convergences such ascondition
calculationby
oftheaccuracy
penalty function method.
requirements. The genetic
However, the algorithm has very method
penalty function good convergences such as
and the genetic calculation
algorithm areof
used to solve
accuracy the multi‐objective
requirements. However,optimization problem method
the penalty function into a single objective
and the geneticfunction.
algorithmTheareweight
used to
coefficient is needed to balance
solve the multi-objective the relationship
optimization between
problem into multiple
a single objectives.
objective The
function. weight
The weight coefficients
coefficient
usually
is needed depend on experience.
to balance So, thebetween
the relationship multiplemultiple
objectives are solved
objectives. The byweight
the NSGA‐II. The usually
coefficients Pareto
solution
depend set on by NSGA‐II So,
experience. is shown in Figure
the multiple 4.
objectives are solved by the NSGA-II. The Pareto solution set
by NSGA-II is shown in Figure 4.
 Math. Comput. Appl. 2016, 21, 47 7 of 10
Math. Comput. Appl. 2016, 21, 47 7 of 10

Figure 4. Multi-objective optimization Pareto optimal solution set.

Figure 4. Multi‐objective optimization Pareto optimal solution set.

From
From Figure
Figure 4, is
4, it it found
is found that
that thethe Pareto
Pareto frontier
frontier AA andandDD stand
stand forfor
thethe absolute
absolute value
value of of
thethe
difference between the minimum point and the spring pressing force change value, before and afterthe
difference between the minimum point and the spring pressing force change value, before and after
thespring-loaded
spring‐loadedforce
forceof of
thethe
diaphragm
diaphragm spring breaks
spring the disc
breaks wear.wear.
the disc During AB there
During are great
AB there are changes
great
in discs wear spring-loaded forces due to small changes of the spring pressing force.
changes in discs wear spring‐loaded forces due to small changes of the spring pressing force. During During CD there
are great changes of the spring pressing force because of small changes in the
CD there are great changes of the spring pressing force because of small changes in the disc wear disc wear loaded forces.
Therefore,
loaded forces.theTherefore,
midpoint near the the BC is chosen
midpoint near as thea result
BC isofchosen
multi-objective optimization.
as a result The optimal
of multi‐objective
results of the diaphragm spring are shown in Tables 3 and 4.
optimization. The optimal results of the diaphragm spring are shown in Tables 3 and 4.

Table 3. Comparison of optimized structural parameters with original ones.

Table 3. Comparison of optimized structural parameters with original ones.

Plan
Plan H (mm)
(mm) h (mm)
(mm) R (mm)
(mm) (mm)
r (mm) R11 (mm)
R (mm) rr11 (mm)
(mm) λ1B
Original
Original 5.8
5.8 2.93
2.93 145.7
145.7 116.8
116.8 143.66
143.66 116.1
116.1 4.80
4.80
A A 5.24
5.24 2.80
2.80 140.00
140.00 115.00
115.00 138.68
138.68 115.00
115.00 4.21
4.21
B B 5.20
5.20 2.80
2.80 140.04
140.04 115.18
115.18 138.80
138.80 114.00
114.00 4.02
4.02
C 5.21 2.81 140.35 115.48 140.66 114.50 4.01
C 5.21 2.81 140.35 115.48 140.66 114.50 4.01
Notes: H: cone height of the disc spring; h: spring diaphragm thickness; R: big end radius of the disc spring;
Notes: H: cone height of the disc spring; h: spring diaphragm thickness; R: big end radius of the disc
r: small end radius of the disc spring; R1 : radius of the pressure plate’s loading point; r1 : radius of loading point
spring; r: support
of the small end
ring;radius of deformation
λ1B : new the disc spring; 1: radius of the pressure plate’s loading point; r1:
of pointRB.
radius of loading point of the support ring; λ1B: new deformation of point B.
Table 4. Comparison of optimized performance parameters with original ones.
Table 4. Comparison of optimized performance parameters with original ones.
Plan Fb (N) Fa (N) Fc (N) |Fb − Fa | (N) |Fb − Fa |/|Fb | (%)
Plan Fb (N) Fa (N) Fc (N) |Fb − Fa| (N) |Fb − Fa|/|Fb| (%)
Original 5226 5925 3817 699 13.37
Original 5226 5925 3817 699 13.37
A 4834 5185 3709 351 7.23
B A 4834
4757 5185
5016 3709 3715 351 259 7.23 5.44
C B 4757
4422 5016
4603 3715 3567 259 181 5.44 4.09
C
Notes: Fb : working 4422 of 4603
pressure 3567 B; Fa : working
working point 181 pressure of diaphragm
4.09 spring when the
wear reaches the limit point; F : separation force; |F − F |/|F |: relative rate of
Notes: Fb: working pressure of working point B; Fa: working pressure of diaphragm spring when
c b a b change of the steering
the
separation force.
wear reaches the limit point; Fc: separation force; |Fb − Fa|/|Fb|: relative rate of change of the steering
The optimal
separation force. results of diaphragm spring are illustrated in Tables 3 and 4. From Tables 3 and 4
it is found that the smaller the deformation of the big-deformation λ1 , the longer the lifespan of the
clutch. So the lifespan
The optimal results ofofdiaphragm
the diaphragm
springspring by the NSGA-II
are illustrated in Tables is 3better
and 4.than
FromtheTables
penalty function
3 and 4 it
method and the genetic algorithm. The wear range
is found that the smaller the deformation of the big‐deformation |F b − F | = 181 N and relative rate of
a λ1, the longer the lifespan of thechange
|Fb −
clutch. SoFathe
|/|F b | = 4.09%,
lifespan of thewhich is far less
diaphragm thanby
spring thethe
10% [9], improving
NSGA‐II is betterthe compaction
than force
the penalty stability.
function
method and the genetic algorithm. The wear range |Fb − Fa| = 181 N and relative rate of change |Fb −
Fa|/|Fb| = 4.09%, which is far less than the 10% [9], improving the compaction force stability. In
Math. Comput. Appl. 2016, 21, 47 8 of 10
Math. Comput. Appl. 2016, 21, 47 8 of 10

In addition,
addition, thethe steering
steering separation
separation forceofofthe
force theclutch
clutchafter
afteroptimization
optimizationisisthe the smallest.
smallest. The
The relative
rate of change of the steering separation force is |3817
change of the steering separation force is |3817 − 3567|/3817 = − 3567|/3817 = 6.55%, so the operating
flexibility is improved.
The final optimal results of the load deformation characteristic characteristic curve
curve based
based onon the
the Table 2 are
illustrated in Figure 5. From Figure 5, it is found that the friction plate abrasion clamping force of the
diaphragm spring within the limit range in the scheme A is greatly reduced, but the separation of
driver control force average change is increased slightly. slightly. So, the optimization of the scheme A is not
ideal. Compared
Comparedwith thethe
with original scheme,
original the friction
scheme, plate abrasion
the friction clampingclamping
plate abrasion force of the diaphragm
force of the
spring
diaphragmwithin the limit
spring range
within thein scheme
limit rangeB inis greatly
schemereduced too. reduced
B is greatly In addition,
too. separation
In addition, ofseparation
the driver
control force average
of the driver control also
forcedecreased,
average also so the effect is good.
decreased, so theCompared to theCompared
effect is good. scheme C with to thethe original
scheme C
scheme,
with the the friction
original plate the
scheme, abrasion
frictionclamping force of
plate abrasion the diaphragm
clamping force of spring within the
the diaphragm limitwithin
spring range
is
thedrastically
limit range reduced, and separation
is drastically reduced, andof the driver control
separation of theforce
driveris greatly
control reduced. The separation
force is greatly reduced.
performance
The separation performance improved significantly compared with the first A,
improved significantly compared with the first two schemes. Comparing twoB and C, the
schemes.
three schemes
Comparing A, improve
B and C,athe little, butschemes
three clamping force inascheme
improve C isclamping
little, but more gentle.
forceAt in the sameCtime,
scheme the
is more
pedal force that needs to be overcome to separate the drives control force is
gentle. At the same time, the pedal force that needs to be overcome to separate the drives control smaller, so separation is
more
force portable
is smaller,andsotheseparation
comprehensive performance
is more portable and of thethe
lastcomprehensive
program is the most comprehensive
performance of the and
last
comparative
program is the one of the
most three integrated
comprehensive andprograms.
comparative one of the three integrated programs.

Figure 5. Characteristic curve of the diaphragm spring.

Figure 5. Characteristic curve of the diaphragm spring.

From Tables 3 and 4, and Figure 5, in can be seen that the pressing force of the optimized
From Tables 3 and 4, and Figure 5, in can be seen that the pressing force of the optimized
diaphragm spring varied slightly during the abrasion range of the friction film, and the
diaphragm spring varied slightly during the abrasion range of the friction film, and the manipulation
manipulation became remarkably light. The problem of the weight coefficient in the traditional
became remarkably light. The problem of the weight coefficient in the traditional empirical design is
empirical design is solved. The new optimization model is proved correct and effective, and the
solved. The new optimization model is proved correct and effective, and the purpose of optimization
purpose of optimization is achieved.
is achieved.
5. Conclusions
5. Conclusions
In this paper, the multi‐objective optimizations of a clutch diaphragm spring are carried out
In this paper, the multi-objective optimizations of a clutch diaphragm spring are carried out based
based on NSGA‐II. A new optimization target function is proposed. The new function takes the
on NSGA-II. A new optimization target function is proposed. The new function takes the minimum
minimum average change in compress force of the spring and the minimum force of the separation
as total objectives. The results show that the pressing force of the diaphragm spring by the new
Math. Comput. Appl. 2016, 21, 47 9 of 10

average change in compress force of the spring and the minimum force of the separation as total
objectives. The results show that the pressing force of the diaphragm spring by the new algorithm is
improved by 4.09% and the steering separation force is improved by 6.55%, which has better stability
and steering portability. The results show that the optimal purpose can be achieved in the given case.
Comparing the NSGA-II with the penalty function method and the genetic algorithm indicates that the
multi-objective can lead to a more desirable design than the single-objective optimization. Future work
may include further development of NSGA-II by adding other constraints, e.g., reliability, or other
objective functions, e.g., weight; use of these algorithms on other clutch optimization problems may
also be valuable.

Acknowledgments: This research is supported by the research project of the Sichuan Provincial Key Lab of
Process Equipment and Control Engineering (GK201508).
Author Contributions: Chun Wang analyzed the data; Junjun Zhu contributed analysis tools; Junchao Zhou
wrote the paper.
Conflicts of Interest: The authors declare no conflict of interest.

References
1. Xia, C.; Zhu, P.; Zhang, M.; Gao, X.; Lu, L. Dispersion of the mechanical parts performance indicators based
on the concept of random vector. Chin. J. Mech. Eng. 2012, 25, 153–159. [CrossRef]
2. Haugen, E.B. Probabilistic Mechanical Design; John Wiley & Sons: New York, NY, USA, 1980.
3. Wu, Y.-H. Multi-objective optimization design of vehicle clutch diaphragm spring. In Proceedings of
the International Conference on Intelligent Computation Technology and Automation, Changsha, China,
10–11 October 2009; pp. 193–197.
4. Tripathi, K. Some Design-Objectives and Design-Guidelines for Automotive Friction Clutch Based on Clutch
Engagement Dynamics. J. Inst. Eng. 2014, 95, 51–61. [CrossRef]
5. Amisano, F.; Ercole, G.; Mattiazzo, G.; Mauro, S.; Serra, G.; Velardocchia, M. The influence of the push
plate mechanical characteristic on torque transmissibility in diaphragm spring. In Proceedings of the
3rd International AIMETA Tribology Conference, Vietri sul Mar, Italy, 18–20 September 2002; Volume 24,
pp. 98–108.
6. Xi, J.-Q.; Lu, J.; Pan, C. Modeling and simulation of the static characteristics of diaphragm spring clutches.
J. Beijing Inst. Technol. 2015, 24, 188–193.
7. Kaya, N. Optimal design of an automotive diaphragm spring with high fatigue resistance. Int. J. Veh. Des.
2006, 40, 126–143. [CrossRef]
8. Zhao, L.-J.; Liu, T.; Song, B.-Y. Optimum design of automobile diaphragm spring clutch. In Proceedings of
the IEEE Vehicle Power and Propulsion Conference, Harbin, China, 3–5 September 2008; pp. 1421–1427.
9. Ran, Z.Y.; Zhao, S.E.; Yu, L.I. Optimum Design for the Diaphragm Spring Clutch on Automobile with Genetic
Algorithms. J. Chongqing Univ. 2003, 26, 77–81.
10. Mohamad, N.; Iovenitti, P.; Vinay, T. Modelling and Optimisation of a Spring-Supported Diaphragm
Capacitive MEMS Microphone. Engineering 2010, 2, 762–770. [CrossRef]
11. Ye, Z.; Kaiyuan, Y. A Study of Belleville Spring and Diaphragm Spring in Engineering. J. Appl. Mech. 1990,
57, 1026–1031.
12. Chan, C.-K.; Lai, W.-C.; Wu, M.; Wang, M.-Y.; Fang, W. Design and Implementation of a Capacitive-Type
Microphone with Rigid Diaphragm and Flexible Spring Using the Two Poly Silicon Micromachining Processes.
IEEE Sens. J. 2011, 11, 2365–2371. [CrossRef]
13. Pasandideh, S.H.R.; Niaki, S.T.A.; Asadi, K. Bi-objective optimization of a multiproduct multi-period
three-echelon supply chain problem under uncertain environments NSGA-II and NRGA. Inf. Sci. 2015, 292,
57–74. [CrossRef]
14. Brownlee, A.E.I.; Wright, J.A. Constrained, mixed-integer and multi-objective optimization of building
designs by NSGA-II with fitness approximation. Appl. Soft Comput. 2015, 35, 114–126. [CrossRef]
15. Li, Y.; Liao, S.; Liu, G. Thermo-economic muti-objective optimization for a solar-dish Brayton system using
NSGA-II and decision making. Electr. Power Energy Syst. 2015, 64, 167–175. [CrossRef]
Math. Comput. Appl. 2016, 21, 47 10 of 10

16. Bandyopadhyay, S.; Bhattacharya, R. Solving multi-objective parallel machine scheduling problem by
modified NSGA-II. Appl. Math. Model. 2013, 37, 6718–6729. [CrossRef]
17. Nourbakhsh, A.; Safikhaniand, H.; Derakhshan, S. The comparison of multi-objective particle swarm
optimization and NSGA II algorithm: Applications in centrifugal pumps. Eng. Optim. 2011, 43, 1096–1113.
[CrossRef]
18. Murugan, P.; Kannan, S.; Baskar, S. Application of NSGA-II Algorithm to Single-Objective Transmission
Constrained Generation Expansion Planning. IEEE Trans. Power Syst. 2009, 24, 1790–1797. [CrossRef]

© 2016 by the authors; licensee MDPI, Basel, Switzerland. This article is an open access
article distributed under the terms and conditions of the Creative Commons Attribution
(CC-BY) license (http://creativecommons.org/licenses/by/4.0/).