Mechanics & Industry 21, 412 (2020)

© AFM, EDP Sciences 2020 Mechanics

[Link] &Industry
Available online at:
[Link]

REGULAR ARTICLE

Multi-objective optimization of a sports car suspension system

using simplified quarter-car models
Salman Ebrahimi-Nejad*, Majid Kheybari, and Seyed Vahid Nourbakhsh Borujerd
Vehicle Dynamical Systems Research Lab, School of Automotive Engineering, IranUniversity of Science and Technology,
Tehran, Iran

Received: 26 July 2019 / Accepted: 5 May 2020

Abstract. In this paper, first, the vibrational governing equations for the suspension system of a selected sports
car were derived using Lagrange’s Equations. Then, numerical solutions of the equations were obtained to find
the characteristic roots of the oscillating system, and the natural frequencies, mode shapes, and mass and
stiffness matrices were obtained and verified. Next, the responses to unit step and unit impulse inputs were
obtained. The paper compares the effects of various values of the damping coefficient and spring stiffness in order
to identify which combination causes better suspension system performance. In this regard, we obtained and
compared the time histories and the overshoot values of vehicle unsprung and sprung mass velocities, unsprung
mass displacement, and suspension travel for various values of suspension stiffness (KS) and damping (CS) in a
quarter-car model. Results indicate that the impulse imparted to the wheel is not affected by the values of CS and
KS. Increasing KS will increase the maximum values of unsprung and sprung mass velocities and displacements,
and increasing the value of CS slightly reduces the maximum values. By increasing both KS and CS we will have a
smaller maximum suspension travel value. Although lower values of CS provide better ride quality, very low
values are not effective. On the other hand, high values of CS and KS result in a stiffer suspension and the
suspension will provide better handling and agility; the suspension should be designed with the best combination
of design variables and operation parameters to provide optimum vibration performance. Finally, multi-
objective optimization has been performed with the approach of choosing the best value for CS and KS and
decreasing the maximum accelerations and displacements of unsprung and sprung masses, according to the
TOPSIS method. Based on optimization results, the optimum range of KS is between 130 000–170 000, and the
most favorable is 150, and 500 is the optimal mode for CS.
Keywords: Multi-objective optimization / suspension system / sports car / quarter-car / numerical solution
unsprung mass / sprung mass / suspension travel

1 Introduction steering stability and good handling in order to keep the

vehicle in control, because all vehicle-ground interaction
One of the most critical factors for assessing vehicle forces rely on the tire contact patch. The automotive
performance is ride comfort and researchers have been suspension system consists of all the parts and components
trying to improve it on each vehicle. Road roughness that connect the vehicle’s body to its wheel assembly and
produces forced vibration which adversely affects ride permit suspension travel. Tires, pressurized air, springs,
comfort and can even result in chaotic motions [1,2]. The shock absorbers, and linkages form the suspension system,
suspension is responsible for minimizing the discomfort for which contributes to both vehicle handling and road holding
passengers through the selection of proper springs and and vehicle ride quality, which are two qualities that are
dampers to reduce vehicle motion including pitch and roll, usually in contradiction with each other. Therefore, tuning
and functions to smooth out the ride and isolate the and design of the suspension involves finding the right
passengers and also protect the vehicle and its cargo from compromise between handling and ride quality, to maintain
vibrational damage and fatigue [3,4]. On the other hand, both safety and comfort. In this regard, automotive
the suspension system is essential for maintaining maxi- companies make available a variety of suspension systems
mum contact between wheels and the road, to provide for their production line, and the design of the front and rear
suspensions of their cars are usually different.
Many factors affect a vehicle’s ride comfort, and the
* e-mail: ebrahiminejad@[Link] suspension can be designed with the best combination of
2 S. Ebrahimi-Nejad et al.: Mechanics & Industry 21, 412 (2020)

design variables and operation parameters to provide and ride characteristics, indicating the effectiveness of the
optimum vibration performance [5]. Therefore, designers of adaptive control in reducing the displacements and thereby
automotive suspension require a deep understanding of the providing better ride comfort for the passengers.
effects of design parameters on the important dynamic Moreover, investigations on simple vehicle models have
response of the system, especially under various loading been performed to consider handling and braking stability.
conditions and road surface inputs. To this end, suspension For example, Hamersma and Els [14] attempted to improve
design and analysis has evolved as an interesting research braking performance using ABS and experimentally
topic. validated simulations of a suspension system and showed
Gohari and Tahmasebi [6] showed that the use of that the semi-active suspension has a significant positive
intelligent active force control (AFC) will have a significant impact on brake performance. Simulation analysis of the
effect on the seat suspension of an off-road vehicle, hence, response of state variables of slip angle and yaw rate was
the neuro-AFC control system improves suspension performed by Shi et al. [15] at various speeds for a step
performance compared to conventional PID control. Seat steering angle input. Simulation results indicated that the
suspension is widely applied to attenuate vibrations, as vehicle’s time-response curves of slip angle and yaw rate
well. Wang et al. [7] studied a scissors mechanism to can intuitively reflect the variations in vehicle handling
improve a vehicle seat suspension while also enhancing the stability. Recently, Maier et al. [16] investigated the
convergence speed of the optimization procedure. Applying braking dynamics of an experimentally validated multi-
light-weight designs, for weight reduction in the vehicle body dynamic model design of a bicycle front suspension,
body and chassis systems, especially for application in incorporating frame geometries and suspension concepts.
future electric vehicles has also been a subject of research. Suspensions with low spring stiffness and low damping
Tobolár et al. [8] presented a split carrier wheel suspension rate have large suspension travels, while high damping rate
system comprising of a three mass suspension design that results in small suspension travel and improves vehicle
enhances ride comfort and road holding. Zhou et al. [5] handling. Most recently, multiple controlled generators
optimized the performance of an electric vehicle’s suspen- have been used for damping-tunable energy-harvesting in
sion system comprised of double suspension arm torsion the suspension system. By designing an energy harvesting
bars, under random vibrations and calculated the suspen- system and altering the number of generators, Xie et al. [17]
sion stiffness and performed sensitivity analyses for design showed that the damping coefficient can be fine-tuned
parameters. according to suspension system requirements.
Suspension vibrations are most comfortable if the Generally, ride comfort and vehicle handling are
frequency of the vibrations is between 1 and 1.5 Hz. Once critically important for automotive engineers. For this
the frequency exceeds this limit and is between 1.5 and 2Hz, reason, engineers are constantly working to enhance ride
the ride becomes harsh for the passengers. To solve this comfort and handling characteristics by improving and
problem, many researchers have focused on the design and optimizing the suspension system. Since the suspension
analysis of active or semi-active suspension systems with system parameters are interdependent, to optimize the
adaptive stiffness and damping parameters, including suspension system, all parameters must be checked in
hydro-pneumatic suspension and magneto-rheological conjunction with each other; hence, the problem should be
(MR) suspension systems in which the elastic and damping considered as a multi-purpose decision-making problem
properties can vary with road conditions under the [18]. Such a problem has been the subject of a plethora of
influence of regulating air pressure or changing the electric research. For example, to improve handling stability,
current supply or electromagnetic field. To improve ride Zhang et al. performed [19] the multi-objective suspension
comfort, Marzbanrad et al. [9] used an active MR controller system optimization for an in-wheel-motor driven electric
in a half-car model and performed a multi-objective vehicle. Using the Taguchi method, Zhang and Wang
optimization procedure. Their results showed significant Conducted a parametric study to optimize a half-vehicle
improvements due to the application of an MR damper suspension system model [20]. Numerical computational
compared to a passive suspension system. studies comprising the multi-objective optimization of a
Automotive control systems monitor a variety of full-vehicle suspension model and non-dominated sorting
operational parameters to control vehicle performance, genetic algorithm II (NSGA-II) has been established by
thereby providing better handling and ride quality. Chen Fossati et al. [21]. Optimization based on neighborhood
et al. [10] conducted a non-linear control of a semi-active cultivation GA and weighting combination method has
variable damping suspension to improve ride comfort, been designed by Su et al. [22] for a minivan.
handling stability, and driving safety. Ozbulur [11] used a Among the different methods to solve multi-objective
fuzzy-logic-based controller and showed significant decision-making problems, the technique for order prefer-
improvements in resonance values and reductions in the ence by similarity to ideal solution (TOPSIS) is simple and
vibration amplitude in comparison with passive suspension efficient [23]. Jiang and Wang used TOPSIS to optimize the
performance active suspension. Gad et al. [12] conducted suspension system of a truck [18] and also to optimize the
another investigation on genetic algorithm (GA) multi- handling stability and ride comfort [24].
objective optimization of a fractional-order PID (FOPID) As mentioned above, suspension stiffness and damping
control of a semi-active MR-damped seat suspension to coefficient highly affect vehicle ride and handling proper-
examine the successful performance of the proposed ties and the study of their effects can be of great
FOPID system. Prassad et al. [13] designed an adaptive significance. The purpose of this paper is to characterize
control system for a suspension to improve vehicle safety the proper spring stiffness coefficient and damping
 S. Ebrahimi-Nejad et al.: Mechanics & Industry 21, 412 (2020) 3

Fig. 2. Two-DOF model of the car with bounce and pitch

Fig. 1. The suspension system of the studied vehicle. motions.

Table 1. Characteristics of vehicle mechanical system.

System Type
Steering Speed-dependent, electro-hydraulic
Steering gear type Rack-and-pinion
Front wheel or axle location Independent Suspension
Rear wheel or axle location Independent Suspension
Front/Rear Springs Coil spring / Coil spring
Front/Rear shock absorber Double-tube gas pressure / single-tube gas pressure
Front/Rear stabilizer type Tubular Torsion bar / Tubular Torsion bar

coefficient for a specific sport vehicle to achieve the best 2 Modeling and methods
ride comfort, considering lowest expenses without funda-
mental changes in the suspension system with multi- Mercedes AMG SLC-43, which is designed with indepen-
objective optimization. The current study is concerned dent front and rear suspensions, is studied in this paper.
with the theoretical mathematical formulation of vibra- Apart from providing a high ride quality, the multilink
tional analysis of a sports vehicle. As the vibrational suspension system of the sports car should be designed close
analysis of a discrete system with high degrees of freedom to the ideal attributes for responsive handling, to provide
requires considerable analytical and computational effort, high lateral acceleration, and to reduce body roll tendency.
in such circumstances, numerical methods are used to The front and rear suspensions are shown in Figure 1.
analyze and predict the behavior of the system. To this end, Stiffer components and firmer tuning of the springs and
a quarter-car model is taken into consideration and its shock absorbers combine to provide greater responsiveness,
equations are derived and solved. Finally, numerical results stability, control, and better driver feedback. Table 1
are obtained and compared in different conditions to presents the mechanical characteristics of the vehicle.
examine the impact of design parameters. On a rough surface, an automobile may undergo pitch,
The remaining sections of our article proceed as follows. bounce, and roll motions. A 2-DOF vehicle model with
Initially, we provide a brief description of the modeling and pitch and bounce motion, as presented in Figure 2, can
methods procedure of the vibrational pitch and bounce provide a preliminary suspension model. In this model, tire
model of the car. Then, the modeling of all effective elasticity and damping properties and those of the
parameters of the components including suspension damp- suspension are collected into the combined equivalent
ing and stiffness, sprung and unsprung masses, tire system of spring and damper for each of the front and rear
damping coefficient, and step height are described in order suspensions.
to be used in simulations. Next, displacement, velocity, and Table 2 lists the dimensions, inertia, damping ratio, and
acceleration in the shock absorber are solved to obtain and spring stiffness values and weights of the vehicle.
discuss the effects of damping CS and stiffness KS in the However, as rolling motion is assumed to be negligible
presented model. Finally, a multi-objective optimization is and we would prefer to incorporate the stiffness, mass, and
performed with the approach of choosing the optimum rolling properties of the tire, individually in the vibrational
values for CS and KS and decreasing the maximum values equations of motion, a quarter-car model is used instead.
of the displacements and the accelerations of the sprung For this purpose, a MATLAB code is generated to estimate
and unsprung masses. These optimum values, obtained the spring stiffnesses and damping coefficients for the
according to the TOPSIS method, ensure significant quarter-car model. The quarter-car model is frequently
improvements in suspension response and handling used in vehicle suspension analyses due to its simplicity,
characteristics.
4 S. Ebrahimi-Nejad et al.: Mechanics & Industry 21, 412 (2020)

Table 2. Dimensions and weights of the Mercedes-AMG The unsprung mass EOM is:
SLC-43.
€ 2 þ kt ðx2 wÞ þ ct ðx_ 2
mu x wÞ
_ ks ðx1 x2 Þ
Dimensions and weights Value cs ðx_ 1 x_ 2 Þ ¼ 0: ð2Þ
Total length, L 4143 mm
Equations (1) and (2) are formulated in state space:
Total height, h 1303 mm
Front Track width, T 1565 mm X_ ¼ AX þ BU
Vehicle width (including side mirrors) 2006 mm V ¼ CX þ DU ð3Þ
Wheelbase, l 2431 mm
Distance of front axle from COG, l1 1.00 m where A, B, C, and D are state space, input, output, and
direct transmission matrices, respectively, and U is system
Distance of rear axle from COG, l2 1.43 m input. Let,
Turning Circle 1052 mm x_ 1 ¼ V 1 ; x_ 2 ¼ V 2 ; D ¼ ðx1 x2 Þ; D_ ¼ ðV 1 V 2 Þ
Curb weight 1580 kg Therefore, equations (1) and (2) can be written as:
Gross vehicle weight, GVW 1890 kg
Moment of inertia about COG, J 2500 kg m2 V_ 1 ¼ ½ks =MŠD ½C s =MŠðV 1 V 2Þ ð4Þ
Stiffness of front spring, K1 29350 N/m
Stiffness of rear spring, K2 24730 N/m V_ 2 ½C t =MŠw_ ¼ ½ks =mŠD þ ½kt =mŠw ½kt =mŠx2
Front damping ratio, C1 3890 N s/m þ½C s =mŠV 1 ½ðC s þ C t Þ=mŠV 2 : ð5Þ
Rear damping ratio, C2 2915 N s/m
Moreover,

T_ ¼ V_ 2 ðC t =M Þw_
T ¼ V2 ðC t =M Þw
V2 ¼ T þ ðC t =M Þw: ð6Þ

Replacing equations (6) into (5), gives:

T ¼ ½ks =MŠD ½kt =MŠx2 þ ½C s =MŠV 1 þ ½ðC s þ C t Þ=mŠT

þ½ ðC s C t Þ=m2 C 2t =m2 þ kt =MŠw ð7Þ

Now,

D_ ¼ ðV 1 V 2Þ ¼ V 1 ½T þ ðC t =mÞwŠ: ð8Þ

Replacing V2 from equations (6) into (4), results in:

Fig. 3. Quarter-car model of the vehicle suspension. V_ 1 ¼ ½ks =MŠD ½C s =MŠV 1 þ ½C s =MŠT
þ½ðC s C t Þ=ðMmފw ð9Þ
examined only vertical vibrations of the car body
(bouncing), but it can provide the main characteristics Based on equation (7), the state variables are D, X2, V1,
of the full model which can be useful for preliminary design. T. The phase space matrix becomes:
The dynamic model, shown in Figure 3, can serve as a basis 2
x_ 2
3
for analyzing the response of the full-vehicle model to road 6D 7
bumps or steps. The model is presented considering that 6
4V15
7
the vehicle body is rigid and each suspension includes a
spring, a damper, the total sprung and unsprung weights T
and tire elasticity, and damping.
2 3
0 0 0 1
Based on information of the vehicle and its suspension 6 0 0 1 1 7
system, presented in Table 3, additional characteristics CS ¼6 7
4 0 ks =M C s =M C s =M 5
of the quarter-car model can be obtained. kt =m ks =m C s =m ½ðC s þ C t Þ=mŠ
The vehicle is considered to be moving at a speed of 2 3
50 km/hr and the path of the road is a sinusoidal path with an
2 3 ðC t =mÞ
x2
amplitude of 50 mm and wavelength of l = 5 m. For the 6D 7 6
6 ðC t =mÞ 7
7
model, moving on a rough surface, the sprung mass EOM 6 7 þ 6 ½ðC s C t Þ=ðMmފ 7:
4V1 5 6 7
becomes: 4 C s C t C t 2 kt 5
T þ
€ 1 þ ks ðx1
ms x x2 Þ þ cs ðx_ 1 x_ 2 Þ ¼ 0: ð1Þ m2 m2 m
 S. Ebrahimi-Nejad et al.: Mechanics & Industry 21, 412 (2020) 5

Table 3. Design information on the suspension system. The mode shapes are calculated as:

Parameter Value
 
A1 ðC 2 l2 C 1 l1 Þs þ k2 l2 k1 l
¼ ¼ 1:1468
Sprung Mass, ms 395 Kg B1 ms2 þ ðC 1 þ C 2 Þs þ k1 þ k2
ð12Þ
Unsprung Mass, mu 38 Kg
 
A2 ðC 2 l2 C 1 l1 Þs þ k2 l2 k1 l
Suspension Spring Stiffness, Ks 29300 N/m ¼ ¼ 0:5507
B2 ms2 þ ðC 1 þ C 2 Þs þ k1 þ k2
Suspension Damping Coefficient, CS 3000 N.s/m
Tire Stiffness, kt 290000 N/m    
A1 1:1468
Tire Damping Coefficient, ct 3100 N s/m ¼
B1 1
Road Step Height, w 50 mm     ð13Þ
A2 0:5507
¼
B2 1
The output matrix becomes:
Next, we extract the natural frequencies and mode
3 21 1 0 0
3 shape, in addition to the stiffness matrix and mass-matrix
x1
2
and the results are:
6V1 7 6 0 0 1 0 7
7 6 7
0 k =M C s =M C s =M 7
6
6 V_ 1 7 6 s

1580 0

6 x2 7 ¼ 6 7
M¼
6 7 6 1 0 0 0 7
0 2500
4V 5 6 40
7
2 0 0 1 5 ð14Þ
D
 
0 1 0 0 4:9457 1:0658 4
K¼  10
1:0658 7:5367
0
2 3
0
2 3
x2 6 7 The resulting natural frequencies and mode shapes
V_ 1
6 7
6D 7 6 7 matrices are:
4 V 1 5 6 ðC s C t Þ=ðMmÞ7:
6 7þ6 7
6 7  
T C t =m 5:4652 0
v¼
4 5
0 0 6:0273
 
Using the above equations, the time-histories of vehicle 0:0143 0:0207
u¼ ð15Þ
unsprung and sprung mass velocities and displacement can 0:0164 0:0114
be studied. Moreover, the overshoot values of unsprung and
sprung mass velocities, unsprung mass displacement, and The natural frequencies are vn = 5.4652 rad/s and
suspension travel of the quarter-car model suspension, with a vn = 6.0273 rad/s and the first mode is:
step input of 0.05 m, are investigated. It is worth noting that  
it is regular to model both road steps and bumps, for 1
: ð16Þ
evaluating the performance of the suspension system. 1:147

The second mode is:

3 Results and discussions
 
1:816
We used MATLAB software package to perform a : ð17Þ
1
numerical solution to find the complex conjugate pair of
characteristic roots as 2.0373 + 5.3958i, 2.0373 MATLAB Simulink is used to verify the results obtained
5.3958i, 1.7500 + 4.9580i, 1.7500 4.9580i. In the above. The elements of the mode shape vectors are set as the
complex roots, the negative real parts values indicate that initial condition for the integrators. Results indicate that the
the oscillation will decay with time. bounce response occurs at a damped frequency of 5 rad/s,
The two damped natural frequencies and mode shapes which is in close agreement with our calculated damped
for the 2-DOF quarter-car model are indicated as: natural frequency of 4.8903 rad/s. The damped natural
qffiffiffiffiffiffiffiffiffiffiffiffiffi frequency is also obtained for the second mode:
vd1 ¼ vn1 1 j21 ¼ 4:8903 rad=s
ð10Þ
qffiffiffiffiffiffiffiffiffiffiffiffiffi
vd2 ¼ vn2 1 j22 ¼ 5:4506 rad=s:
qffiffiffiffiffiffiffiffiffiffiffiffiffi
vd2 ¼¼ vn2 1 j22 ¼ 5:4506 rad=s: ð18Þ

A MATLAB code is used to obtain the system response

Hence:
to road excitation, which will clearly provide the above-
j1 vn1 ¼ 1:7349 rad=s mentioned transfer functions associated with this input. As
j2 vn2 ¼ 2:0630 rad=s: ð11Þ mentioned above, the inputs Y1 and Y2 for the simulations
6 S. Ebrahimi-Nejad et al.: Mechanics & Industry 21, 412 (2020)

Fig. 4. (a) Velocity of unsprung mass (m/s),. (b) Velocity of sprung mass (m/s),. (c) displacement of unsprung mass (m). (d)
suspension travel (m), for various combinations of Ks (N/m) and CS (N-s/m).

are provided based on the assumed road surface amplitude Now to determine the appropriate values of damper
of 10 mm and wavelength of 5 m is the car speed which is coefficient and spring stiffness (KS and CS), we compare
assumed to be 50 km/hr. The time period, cyclic frequency, various values of damper coefficient and spring stiffness in
and phase delay due to Y2 are: order to figure out which one causes the suspension system
to perform better. As shown in Figure 4, we obtained our
l 5 results for the time histories of vehicle unsprung and sprung
T¼ ¼ ¼ 0:36s ð19Þ
V 50=3:6 mass velocities and displacements.
Results for the quarter-car suspension system for step
2p height of 0.05 m (step input) indicates that the greater the
v¼ ¼ 17:453 rad=s ð20Þ
T value of damping coefficient and spring stiffness, the
l1 þ l2 2:431 greater the unsprung and sprung mass velocities and the
f¼ 2p ¼ 2p ¼ 0:97p rad ð21Þ unsprung mass displacement, which is not desirable for the
l 5 system; however, suspension travel decreases slightly and is
Substituting v, f and motion amplitude in the above damped more rapidly, as seen in the output results.
equations result in: The time histories of vehicle sprung mass acceleration is
y1 ¼ 0:05 sin17:453t shown in Figure 5. For instance, although the system has a
ð22Þ steady-state behavior after 1.5 s for all different KS and CS,
y2 ¼ 0:05 sinð17:453t 0:97pÞ
 S. Ebrahimi-Nejad et al.: Mechanics & Industry 21, 412 (2020) 7

method has been developed with the help of the Technique

for Order of Preference by Similarity to Ideal Solution
(TOPSIS) multi-criteria decision-making technique.

Objectivefunction ¼ min
8 9
>
> maximumun sprung mass displacement; >
>
sprung mass displacement;
< =
>
> sprung mass acceleration >
>
unsprung mass acceleration
: ;

subject to : f500  C s  1000; 10000  K s  200000g

For as much as KS and CS are related to each other,

first, for CS, a constant value is assumed, and all values for
KS are investigated, then, a constant value for KS is
assumed and all values for CS are investigated. After that,
the values referring to maximum unsprung and sprung
mass displacements, and unsprung and sprung mass
accelerations during vehicle vibrations have been
Fig. 5. Sprung mass acceleration (m/s2), for various combina- extracted. The values obtained for CS being assigned a
tions of Ks (N/m) and CS (N-s/m). constant value are shown in Table 4.
The values obtained for the state in which KS assumes a
it has an overshoot in the time period of 0–0.75 s, and it is constant value, are shown in Table 5.
clear that the greater the values of KS and CS the more In order to obtain optimal values for KS and CS with the
overshoot will appear in results which is not appropriate. approach of minimizing the maximum values of unsprung
Finally, we specifically analyzed the overshoot values mass displacement, sprung mass displacement, sprung
for the quarter-car model suspension system for step input mass acceleration, and unsprung mass acceleration, the
by obtaining the maximum values of the time history TOPSIS method is applied.
outputs of Figure 4. The resulting maximum unsprung and First, the decision matrix is made according to the
sprung mass velocities, maximum unsprung mass displace- following relation:
ment and maximum suspension travel of the quarter-car
model suspension, with a step input of 0.05 m, are displayed 2 3
in Figure 6. As shown in Figure 6a, the maximum unsprung x11 x12 :: x1n
mass velocity, which indicates the impulse imparted to the 6 x21
D¼6
x22 :: x2n 7
7 ð23Þ
wheel, is not affected by the values of CS and KS and 4 : : : : 5
therefore, the maximum value of unsprung mass velocity xm1 xm2 :: xmn
remains constant. Figure 6b shows that increasing spring
stiffness KS will increase sprung mass velocity for different where xij is value for each criterion., the decision matrix
CS values. On the other hand, increasing the value of CS should be normalized. For normalization of values, rij is
slightly reduces the maximum sprung mass velocity. formed by the vector method. Unlike the simple lineariza-
Figure 6c shows with increasing spring stiffness KS the tion method, normalization is performed according to the
maximum displacement of the unsprung mass increases, following:
whereas, increasing values of CS slightly reduces the
maximum unsprung mass displacement. For the maximum xij
rij ¼ qffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
ffi: ð24Þ
value of suspension travel, indicated in Figure 6d, it is seen
Xm
x 2
that with increasing both KS and CS we will have a small i¼1 ij

suspension travel value. Although lower values of CS provide

The next step is the establishment of a normal
better ride quality, very low values of CS are not effective. On
compatible matrix based on the weights of the criteria.
the other hand, high values of CS and KS result in a stiffer
The weights are multiplied in the normalized matrix as
suspension and it is clear that the suspension will provide
follows:
better handling and agility. It is worth mentioning that the
suspension system should be designed with the best
combination of design variables and operation parameters vij ¼ rij  wj ð25Þ
to provide optimum vibration performance.
where wj is the weighting factor symmetric to the jth
criterion.
4 Application of the TOPSIS method
n
X
In order to find the best value for KS and CS and having a wj ¼ 1 ð26Þ
good ride from the vehicle, a multi-objective optimization j¼1
8 S. Ebrahimi-Nejad et al.: Mechanics & Industry 21, 412 (2020)

Fig. 6. (a) Maximum velocity of unsprung mass (m/s), (b) maximum velocity of sprung mass (m/s), (c) maximum displacement of
unsprung mass (m), (d) maximum suspension travel (m), versus spring stiffness Ks (N/m) for various values of CS (N-s/m).

There are many ways for weighing the criteria, – For positive criteria, the positive desired value is the
including Analytic Hierarchy Process (AHP), least-squares largest value of that criterion.
method, logarithmic least squares method, approximate – For positive criteria, the negative desired value is the
methods, and Shannon entropy. In this optimization, the smallest value of that criterion.
Shannon entropy method is used to correctly measure the – For negative criteria, the positive desired value is the
criteria. The subsequent step is to calculate positive and smallest value of that criterion.
negative desired values based on the following relation- – For negative criteria, the negative desired value is the
ships: largest value of that criterion.
    Here, Vb is related to the positive indicators, while Vc is
max vij min vij
Aþ ¼ j j∈Vb ; j j∈Vc related to the negative indicators.
n i o i In the next step, Euclidean distance from the positive
þ
¼ vj j j ¼ 1; 2; . . . ; n ð27Þ and negative desired values are calculated using the
following formula:
   
min vij max vij vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
A ¼ j j∈Vb ; j j∈Vc u n
uX
2
i i þ
n o di ¼ t vij vþ j ð29Þ
¼ vj j j ¼ 1; 2; . . . ; n ð28Þ j¼1
 S. Ebrahimi-Nejad et al.: Mechanics & Industry 21, 412 (2020) 9

Table 4. The impact of variations of spring stiffness Ks in suspension performance for a given CS = 1000 N-s/m.

CS Ks Maximum unsprung Sprung mass Sprung mass Unsprung mass

mass displacement displacement acceleration acceleration
1000 10 000 0.5530 0.0519 0.2449 0.0948
1000 20 000 0.0543 0.0508 0.345 0.1836
1000 30 000 0.0533 0.0497 0.4253 0.2591
1000 40 000 0.0541 0.0488 0.4924 0.3424
1000 50 000 0.0552 0.0479 0.5500 0.3809
1000 60 000 0.0562 0.0470 0.6004 0.4309
1000 70 000 0.0572 0.0462 0.6452 0.4751
1000 80 000 0.0581 0.0455 0.6852 0.5145
1000 90 000 0.0590 0.0488 0.7213 0.5497
1000 100 000 0.0599 0.0441 0.7541 0.5812
1000 110 000 0.0607 0.0434 0.7841 0.6096
1000 120 000 0.0615 0.0428 0.8115 0.6351
1000 130 000 0.0622 0.0423 0.8368 0.6582
1000 140 000 0.0629 0.0417 0.8601 0.6790
1000 150 000 0.0635 0.0411 0.8817 0.6980
1000 160 000 0.0642 0.0406 0.9017 0.7152
1000 170 000 0.0648 0.0401 0.9203 0.7308
1000 180 000 0.0653 0.0397 0.9377 0.7450
1000 190 000 0.0659 0.0392 0.9539 0.7580
1000 200 000 0.0664 0.0388 0.9691 0.7698

Table 5. The impact of variations of damping coefficient in suspension performance for a given Ks = 100 000 N/m.

Ks CS Maximum unsprung Sprung mass Sprung mass Unsprung mass

mass displacement displacement acceleration acceleration
100 000 500 0.0607 0.0479 0.7837 0.6660
100 000 1000 0.0599 0.0441 0.7541 0.5812
100 000 1500 0.0592 0.0409 0.7327 0.5119
100 000 2000 0.0587 0.0381 0.7188 0.4548
100 000 2500 0.0583 0.0357 0.7124 0.4075
100 000 3000 0.058 0.0336 0.7131 0.3682
100 000 3500 0.0579 0.0318 0.7198 0.3353
100 000 4000 0.0578 0.0302 0.7309 0.3079
100 000 4500 0.0579 0.0288 0.7447 0.2851
100 000 5000 0.0580 0.0275 0.7599 0.2662
100 000 5500 0.0583 0.0264 0.7755 0.2509
100 000 6000 0.5860 0.0254 0.7912 0.2386
100 000 6500 0.0590 0.0245 0.8066 0.2291
100 000 7000 0.0595 0.0236 0.8215 0.2222
100 000 7500 0.0600 0.0228 0.8359 0.2175
100 000 8000 0.0604 0.0221 0.8497 0.2150
100 000 8500 0.0609 0.0214 0.8629 0.2143
100 000 9000 0.0613 0.0208 0.8755 0.2152
100 000 9500 0.0618 0.0202 0.8876 0.2175
100 000 10000 0.0622 0.0197 0.8991 0.2210
10 S. Ebrahimi-Nejad et al.: Mechanics & Industry 21, 412 (2020)

The closeness value Cl is between zero and one and as it

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
u n
uX
2
di ¼ t vij vj ð30Þ approaches unity, it is assumed to become closer to the
j¼1 ideal answer.
Results of TOPSIS optimization with the goal of
The final step is calculating the relative proximity of the minimizing the maximum unsprung mass displacement,
desired solution. Here, the relative proximity of each sprung mass displacement, sprung mass acceleration, and
option, defined in terms of the closeness value Cl is unsprung mass acceleration and weighing these criteria by
considered to be the desired solution using the following Shannon Entropy are presented in this Section.
formula: As shown in Table 6, Cl has been calculated and ranked
for each KS, using the Shannon entropy method, according
di to the rankings, the results of which are also illustrated in
Cli ¼ ð31Þ
di þ dþ
i Figure 7, the optimum range of KS is between 130 000 and
170 000, and the most favorable is 150 000.
Table 6. The closeness values obtained for various spring Also, Cl values have been calculated and ranked for
coefficients and ranking of the resulting performance. each CS, as shown in Table 7. As can be seen in Table 6 and
further depicted in Figure 8 with increasing CS, the optimal
Ks Cl Rank mode moves farther away. Therefore, the optimal mode for
CS is 500.
10 000 0.4930 15
As can be seen in Table 7 with increasing CS, the
20 000 0.3929 20 optimal mode moves farther away. Therefore, the optimal
30 000 0.4205 19 mode for CS is 500.
40 000 0.4556 18 Based on the results of the TOPSIS optimization and
50 000 0.4637 17 the numerical results presented in Tables 6 and 7, it can be
60 000 0.4896 16 expressed that relatively large spring coefficients and low
damping coefficients lead to better isolation of the sprung
70 000 0.5141 14 mass and unsprung mass from the vibrating source by
80 000 0.5361 13 minimizing maximum unsprung mass displacement,
90 000 0.5534 12 sprung mass displacement, sprung mass acceleration and
100 000 0.5656 11 unsprung mass acceleration.
110 000 0.5744 10
120 000 0.5796 7 5 Conclusion
130 000 0.5830 4
140 000 0.5842 2 A vibrational model has been discussed for a sports car
150 000 0.5849 1 suspension system. The output of the model was elaborated
using MATLAB and Simulink in order to obtain
160 000 0.5836 3 characteristic roots, the natural frequencies, mode shapes,
170 000 0.5823 5 and mass and stiffness matrices. We also showed how our
180 000 0.5812 6 suspension system works in variable road conditions; for
190 000 0.5789 8 example, the accelerations and displacements of each shock
200 000 0.5771 9 absorber were presented, when the car passes a step or
bumps on the road. A key finding of the paper is to compare

Fig. 7. The closeness values obtained for various spring stiffness values, using the TOPSIS method.
 S. Ebrahimi-Nejad et al.: Mechanics & Industry 21, 412 (2020) 11

various values of damper coefficient and spring stiffness indicate that the greater the value of KS and CS, the greater
(KS and CS), in order to figure out which one causes better the unsprung and sprung mass velocities and the unsprung
suspension system performance. In this regard, we mass displacement and the more overshoot will appear in
obtained and compared the time histories of vehicle the sprung mass acceleration results, none of which are
unsprung and sprung mass velocities, unsprung mass desirable for the system; however, suspension travel
displacement, and suspension travel for various values of decreases slightly and is damped more rapidly.
KS and CS in a quarter-car model. Time history results The maximum unsprung mass velocity, which indicates
the impulse imparted to the wheel, is not affected by the
values of CS and KS and therefore, remains constant.
Table 7. The closeness value obtained for various Increasing KS, will increase the maximum values of sprung
damping coefficient and ranking of the performance. mass velocity and unsprung mass displacement for
different CS values and increasing the value of CS slightly
CS Cl Rank reduces the maximum values. For the maximum value of
suspension travel, it is seen that with increasing both KS
500 0.9945 1
and CS we will have a smaller suspension travel value. It is
1000 0.9871 2 worth noting that although lower values of CS provide
1500 0.9778 3 better ride quality, very low values of CS are not effective.
2000 0.9699 4 On the other hand, high values of CS and KS result in a
2500 0.9635 5 stiffer suspension and it is clear that the suspension will
3000 0.9581 6 provide better handling and agility and the suspension
should be designed with the best combination of design
3500 0.9537 7 variables and operation parameters to provide optimum
4000 0.9500 8 vibration performance. It was shown that suspensions with
4500 0.9469 9 low spring stiffness and low damping rate have large
5000 0.9444 10 suspension travels, while high damping rate results in small
5500 0.9422 11 suspension travel and improves vehicle handling. However,
it was shown that higher spring stiffness increases the
6000 0.9405 12
maximum displacement of the unsprung mass and also
6500 0.9391 13 sprung mass velocity.
7000 0.9380 14 To increase ride comfort, a multi-objective optimization
7500 0.9372 15 with the approach of reducing maximum unsprung and
8000 0.9367 16 sprung mass displacements, and unsprung and sprung mass
8500 0.9363 18 accelerations, with the help of TOPSIS method has been
implemented. Given that KS and CS are related to each other,
9000 0.9362 19 to find the optimal value for each one, the other is assigned a
9500 0.9362 19 constant value. The multi-objective optimization procedure
10 000 0.9364 17 resulted in a value of 500 for CS and 150 000 for KS.

Fig. 8. The closeness values obtained for various damping coefficient values, using the TOPSIS method.
12 S. Ebrahimi-Nejad et al.: Mechanics & Industry 21, 412 (2020)

References for semi-active magnetorheologically damped seat suspen-

sion, J. Vibration Contr. 23, 1248–266 (2015)
[1] J. Marzbanrad, M. Mohammadi, S. Mostaani, Optimization [13] S.G. Prassad, S. Aakash, K.M. Mohan, Simulation of a
of a passive vehicle suspension system for ride comfort Control System for an Adaptive Suspension System for
enhancement with different speeds based on design of Passenger Vehicles, Simulation 1, 62788 (2017)
experiment method (DOE) method, J. Mech. Eng. Res. 5, [14] H.A. Hamersma, P. Schalk Els, Improving the braking
50–59 (2013) performance of a vehicle with ABS and a semi-active suspension
[2] M. Boreiry, S. Ebrahimi-Nejad, J. Marzbanrad, Sensitivity system on a rough road, J. Terramech. 56, 91–101 (2014)
analysis of chaotic vibrations of a full vehicle model with [15] P.-C Shi, Q. Zhao, S.-S. Peng, Modeling and Simulation of
magnetorheological damper, Chaos Solitons Fractals 127, Linear Two-DOF Vehicle Handling Stability, ITM Web
428–442 (2019) Conf. 1, 07007 (2017)
[3] H. Nazemian, M. Masih-Tehrani, Development of an [16] O. Maier, B. Györfi, J. Wrede, R. Kasper, Design and
Optimized Game Controller for Energy Saving in a Novel validation of a multi-body model of a front suspension bicycle
Interconnected Air Suspension System, Proceedings of the and a passive rider for braking dynamics investigations,
Institution of Mechanical Engineers Part D Journal of Multibody Sys. Dyn. 42, 19–45 (2017)
Automobile Engineering, In press (2020) [17] L. Xie, J. Li, X. Li, L. Huang, S. Cai, Damping-tunable
[4] H. Nazemian, M. Masih-Tehrani, Hybrid Fuzzy-PID Control energy-harvesting vehicle damper with multiple controlled
Development for a Truck Air Suspension System, SAE generators: design, modeling and experiments, Mech. Sys.
International Journal of Commercial Vehicles, In press Signal Process 99, 859–872 (2018)
(2020) [18] R. Jiang, D. Wang, Optimization of Suspension System of
[5] S.-T. Zhou, Y.-J. Chiu, I.H. Lin, The Parameters Optimizing Self-Dumping Truck Using TOPSIS-based Taguchi Method
Design of Double Suspension Arm Torsion Bar in the Electric Coupled with Entropy Measurement, SAE Technical Paper,
Sight-Seeing Car by Random Vibration Analyzing Method, 2016-2001-1385 (2016)
Shock Vibration 22, 1–9 (2017) [19] L. Zhang, S. Zhang, W. Zhang, Multi-objective optimization
[6] M. Gohari, M. Tahmasebi, Active off-road seat suspension design of in-wheel motors drive electric vehicle suspensions
system using intelligent active force control, J. Low for improving handling stability. Proceedings of the Institu-
Frequency Noise Vibration Active Contr. 34, 475–489 (2015) tion of Mechanical Engineers, Part D: Journal of Automobile
[7] C. Wang, X. Zhang, K. Guo, J. Lv, Y. Yang, Hierarchical Engineering 233, 2232–2245 (2019)
optimisation on scissor seat suspension characteristic and [20] R. Zhang, X. Wang, Parameter study and optimization of a
structure, Vehicle Sys. Dyn. 54, 1538–1553 (2016) half-vehicle suspension system model integrated with an
[8] J. Tobolár, D. Baumgartner, Y. Hirano, T. BÛnte, M. Fleps- arm-teeth regenerative shock absorber using Taguchi
Dezasse, J. Brembeck, Model Based Design of a Split Carrier method, Mech. Sys. Signal Process. 126, 65–81 (2019)
Wheel Suspension for Light-weight Vehicles. Paper pre- [21] G.G. Fossati, L.F.F. Miguel, W.J.P. Casas, Multi-objective
sented at the 12th International Modelica Conference, optimization of the suspension system parameters of a full
Prague, Czech Republic, May 15–17 (2017) vehicle model, Optim. Eng. 20, 151–177 (2019)
[9] J. Marzbanrad, P. Poozesh, M. Damroodi, Improving vehicle [22] Z. Su, F. Xu, L. Hua, H. Chen, K. Wu, S. Zhang, Design
ride comfort using an active and semi-active controller in a optimization of minivan MacPherson-strut suspension
half-car model, J. Vibration Contr. 19, 1357–1377 (2012) system based on weighting combination method and
[10] H. Chen, C. Long, C.-C .Yuan, H.-B Jiang, Non-linear neighborhood cultivation genetic algorithm, J. Automobile
modelling and control of semi-active suspensions with Eng. 233, 650–660 (2019)
variable damping, Vehicle Sys. Dyn. 51, 1568–1587 (2013) [23] C.-L. Hwang, K. Yoon, Methods for Multiple Attribute Decision
[11] V. Ozbulur, An adaptive compensator for a vehicle Making, Springer, Berlin, Heidelberg, 1981, pp 58­–191
suspension system, J. Vibration Contr. 21, 3090–3097 (2014) [24] R. Jiang, D. Wang, Optimization of Vehicle Ride Comfort
[12] S. Gad, H. Metered, A. Bassuiny, A.M. Abdel Ghany, Multi- and Handling Stability Based on TOPSIS Method, SAE
objective genetic algorithm fractional-order PID controller Technical Paper, 2015-200-1348 (2015)

Cite this article as: S. Ebrahimi-Nejad, M. Kheybari, S.V.N. Borujerd, Multi-objective optimization of a sports car suspension
system using simplified quarter-car models, Mechanics & Industry 21, 412 (2020)