W

IE
EV
PR

FUZZY LOGIC BASED MULTI-OBJECTIVE

OPTIMIZATION OF ACTIVE SUSPENSION SYSTEM OF
4X4 IN-WHEEL MOTOR DRIVEN ELECTRICAL
VEHICLE

Ömer BİNGÜL

i
 T.C.
BURSA ULUDAĞ ÜNİVERSİTESİ
FEN BİLİMLERİ ENSTİTÜSÜ

W
FUZZY LOGIC BASED MULTI-OBJECTIVE OPTIMIZATION OF ACTIVE
SUSPENSION SYSTEM OF 4X4 IN-WHEEL MOTOR DRIVEN ELECTRICAL
VEHICLE
IE
Ömer BİNGÜL
EV

0000-0002-8113-4988

Dr. Öğr. Üyesi Ahmet YILDIZ

PR

(Danışman)

YÜKSEK LİSANS TEZİ

OTOMOTİV MÜHENDİSLİĞİ ANABİLİM DALI

BURSA – 2021
Her Hakkı Saklıdır

ii
PR
EV
IE
W
 ÖZET

Yüksek Lisans Tezi

FUZZY LOGIC BASED MULTI-OBJECTIVE OPTIMIZATION OF ACTIVE

SUSPENSION SYSTEM OF 4X4 IN-WHEEL MOTOR DRIVEN ELECTRICAL
VEHICLE

Ömer BİNGÜL

Bursa Uludağ Üniversitesi

Fen Bilimleri Enstitüsü
Otomotiv Mühendisliği Anabilim Dalı

Danışman: Dr. Öğr. Üyesi Ahmet YILDIZ

Bu tezde, 4x4 tekerlek içi motorlu bir elektrikli aracın doğrusal olmayan aktif

W
süspansiyon sisteminin bulanık mantık tabanlı çok amaçlı optimizasyonu, süspansiyon
sistemlerinin yuvarlanma açısı ve yük transferi gibi gerçek çalışma koşulları dikkate
alınarak incelenmiştir. Bu bağlamda, on bir serbestlik derecesine sahip ikinci dereceden
lastik ve kübik süspansiyon katılığına sahip doğrusal olmayan bir tam elektrikli araç
IE
modeli ve beş serbestlik dereceli bir koltuk-sürücü modeli oluşturulmuştur. Sürüş ve
sağlık kriterlerini değerlendirmek için ISO 2731-1’de tanımlanan gereklilikler esas
alınmıştır. Seçilen amaç fonksiyonları, weighted root mean square baş ivmesi, root mean
EV

square koltuk ivmesi, crest factor, titreşim doz değeri, root mean square baş ivmesinin
root mean square koltuk ivmesine oranı, root mean square üst gövde ivmesinin root mean
square koltuk ivmesine oranı, ve root mean square üst gövde ivmesidir. Bunlara ek olarak,
nadiren incelenen rollover etkisi araştırılmıştır. Root mean square süpansiyon
deplasmanı, root mean square tekerlek deplasmanı, root mean square tekerlek içi motor
PR

deplasmanı ve yuvarlanma açısı kısıtlar olarak seçilmiştir. Optimizasyon NSGA-II

algoritması ile gerçekleştirilmiştir. Pasif sistem için tasarım değişkenleri; süspansiyon,
tekerlek içi motor ve koltuğun yay katılıkları ve amortisör sönüm katsayılarıdır.
Ardından, en iyi sürüş konforu ve sağlık kriterini sağlamak için proportional derivative
kontrolcü ile birleştirilmiş bir bulanık mantık kontrolcü optimize edilmiştir. Sunulan
optimizasyon sonuçlarının bulanık mantık kontrolcünün pasif sisteme karşılık olarak
önemli bir gelişme gösterdiği ve yük transfer indeksinde devrilme koşuluyla ilgili
olumsuz bir değişiklik göstermediği görülmektedir.

Anahtar Kelimeler: elektrikli araç, tekerlek içi motor, çok amaçlı optimizasyon, genetik
algoritma, yuvarlanma etkisi
2021, xiv + 88 sayfa.

vi
 ABSTRACT

MSc Thesis

FUZZY LOGIC BASED MULTI-OBJECTIVE OPTIMIZATION OF ACTIVE

SUSPENSION SYSTEM OF 4X4 IN-WHEEL MOTOR DRIVEN ELECTRICAL
VEHICLE

Ömer BİNGÜL

Bursa Uludağ University

Graduate School of Natural and Applied Sciences
Department of Automotive Engineering

Supervisor: Asst. Prof. Dr. Ahmet YILDIZ

In this thesis, fuzzy logic based multi-objective optimization of a nonlinear active

W
suspension system of 4x4 in-wheel motor-driven electrical vehicle is studied by
considering real working conditions such as roll angle and load transfer of the suspension
systems. In this regard, a nonlinear full electrical vehicle model with quadratic tire
stiffness and cubic suspension stiffness with eleven degrees of freedom and a seat-driver
IE
model with five degrees of freedom implemented and optimized by the guidelines
introduced in ISO 2731-1 to assess ride and health criteria. Selected objective functions
are comprised of weighted root mean square head acceleration, root mean square seat
EV

acceleration, crest factor, vibration dose value, the amplitude of head root mean square
acceleration to seat root mean square acceleration, the amplitude of upper torso root mean
square acceleration to seat root mean square acceleration, and root mean square upper
torso acceleration. In addition to these, rarely considered rollover effect was investigated.
Root mean square suspension displacement, root mean square tire displacement, root
PR

mean square in-wheel motor displacement, and roll angle were selected as constraints.
Optimization was carried out with NSGA-II algorithm. Design variables for the passive
system are; stiffnesses and dampers of suspension, in-wheel motor, and seat. Then, a
fuzzy logic controller coupled with a proportional derivative controller optimized for best
ride comfort and health criterion. Presented optimization results demonstrated a
significant improvement over the passive system with fuzzy logic controller, and the load
transfer index showed no adverse change between models concerning the rollover
condition.

Key words: electric vehicle, in-wheel motor, multi-objective optimization, genetic

algorithm, rollover effect
2021, xiv + 88 pages.

vii
 ACKNOWLEDGEMENTS

Throughout the writing of this dissertation I have received a great deal of support and
assistance.

First and foremost, I would like to express my sincere thanks to my supervisor, Asst. Prof.
Dr. Ahmet YILDIZ for his guidance, motivation and continual support throughout my
research. The achievement of this work would not be possible without his valuable
expertise and helpful suggestions by making some sense of the confusion.

In addition, I would like to thank my parents for their love, encouragement, wise counsel,
sympathetic ear and support throughout years. I would also like to thank to my friends
who provided stimulating discussions as well as happy distractions to rest my mind
outside of my research.

W
Ömer BİNGÜL
23/09/2021
IE
EV
PR

viii
 TABLE OF CONTENTS
Page
ÖZET . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vi
ABSTRACT . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . vii
ACKNOWLEDGEMENTS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .viii
ABBREVIATIONS . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . x
LIST OF FIGURES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . xii
LIST OF TABLES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . .xiv
1. INTRODUCTION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1
2. LITERATURE REVIEW . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.1. Introduction . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 5
2.2. Vehicle Suspension Systems . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.1. Passive suspension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
2.2.2. Semi­active suspension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 9
2.2.3. Active suspension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
2.3. Electric Vehicle Suspension Systems . . . . . . . . . . . . . . . . . . . . . . . 17

W
2.3.1. Centralized propulsion suspension system . . . . . . . . . . . . . . . . . . . 18
2.3.2. In­wheel motor suspension system . . . . . . . . . . . . . . . . . . . . . . . 20
2.4. Control Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 23
IE
2.4.1. Proportional integral derivative control . . . . . . . . . . . . . . . . . . . . 23
2.4.2. Fuzzy logic control . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 24
2.5. Optimization Methods . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 27
EV

3. MATERIAL AND METHOD . . . . . . . . . . . . . . . . . . . . . . . . . . . . 31

3.1. Full Vehicle and Human Models . . . . . . . . . . . . . . . . . . . . . . . . . 31
3.2. Active Suspension . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2.1. PID controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
3.2.2. Fuzzy logic controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 42
PR

3.3. Multi­Objective Optimization . . . . . . . . . . . . . . . . . . . . . . . . . . 45

3.3.1. Diversity preservation . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 50
3.3.2. Objective functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 53
3.3.3. Population size and stoppage criterion . . . . . . . . . . . . . . . . . . . . . 59
4. RESULTS AND DISCUSSION . . . . . . . . . . . . . . . . . . . . . . . . . . 61
5. CONCLUSION . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 76
REFERENCES . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 77
CURRICULUM VITAE . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 88

ix
 ABBREVIATIONS

Notation Description

ABC Artificial Bee Colony

ABS Anti-lock Brake Systems
ACO Ant Colony Optimization
ADM Advanced Dynamic Damper Mechanism
ANFIS Adaptive Network-based Fuzzy Inference System
ANN Artificial Neural Network
BBO Biogeography-Based Optimization
BP Back Propagation
CD Crowding Distance
CF Crest Factor
CS Cuckoo Search
D Derivative
DE Differential Evolution

W
DOF Degrees Of Freedom
EA Evolutionary Algorithm
EHA Electro-Hydrostatic Actuator
IE
EKF Extented Kalman Filter
EMS Electromagnetic Suspension
ER Electrorheological
ESC Electronic Stability Control
EV

EV Electic Vehicle
FIS Fuzzy Inference System
FL Fuzzy Logic
FLC Fuzzy Logic Controller
PR

FTC Fault Tolerant Control

GA Genetic Algorithm
GP Genetic Programming
HGCZ Health Guidance Caution Zone
HIS Hydraulically Interconnected Suspension
I Integral
ICE Internal Combustion Engine
IWM In-wheel Motor
LHR Likely Health Risk
LQ Linear Quadratic
LQR Linear Quadratic Regulator
LSRA Linear Switched Reluctance Actuator
LTR Load Transfer Ratio
MF Membership Function
MIMO Multiple-Input-Multiple-Output
MOEAs Multi-Objective Evolutionary Algorithms
MOGA Multi-Objective Genetic Algorithm
MOPSO-CD Multi-Objective Particle Swarm Optimization
MR Magnetorheological

x
NHR No Health Risk
NNC Neural Network Control
NPGA Niched Pareto Genetic Algorithm
NSGA Non-dominated Sort Genetic Algorithm
NSGA-II Non-dominated Sort Genetic Algorithm II
P Proportional
PD Proportional Derivative
PHR Potential Health Risk
PI Proportional Integral
PID Proportional Integral Derivative
PMA Permanent-Magnet Actuator
PSA Pattern Search Algorithm
PSO Particle Swarm Optimization
RMS Root Mean Square
SMC Sliding Mode Control
SQP Sequential Quadratic Programming
SRM Switched Reluctance Motor
TCS Traction Control System

W
TWLO Two-Wheel Lift-Off
VDV Vibration Dose Value
WBV Whole Body Vibration
IE
EV
PR

xi
 LIST OF FIGURES
Page
Figure 2.1 Trade­off nature of suspension systems . . . . . . . . . . . . . . . . 5
Figure 2.2 Passive suspension . . . . . . . . . . . . . . . . . . . . . . . . . . . 6
Figure 2.3 Schematic of an anti­roll HIS . . . . . . . . . . . . . . . . . . . . . 8
Figure 2.4 The assembled anti­roll HIS system . . . . . . . . . . . . . . . . . . 8
Figure 2.5 Schematic of a quarter­car suspension with skyhook control . . . . . 9
Figure 2.6 Suspension system model: (a) passive, (b) semi­active . . . . . . . . 10
Figure 2.7 ER damper schematic configuration . . . . . . . . . . . . . . . . . . 11
Figure 2.8 Photograph of ER damper . . . . . . . . . . . . . . . . . . . . . . . 11
Figure 2.9 Schematic of an MR shock absorber . . . . . . . . . . . . . . . . . . 12
Figure 2.10 Photo of an MR shock absorber . . . . . . . . . . . . . . . . . . . . 13
Figure 2.11 Bouc–wen model for MR shock absorber . . . . . . . . . . . . . . . 13
Figure 2.12 Active suspension . . . . . . . . . . . . . . . . . . . . . . . . . . . 14
Figure 2.13 An electrohydraulic actuator schematic . . . . . . . . . . . . . . . . 15
Figure 2.14 Photo of the pressure control unit . . . . . . . . . . . . . . . . . . . 16

W
Figure 2.15 (a) Passive suspension system (b) electromagnetic suspension system 17
Figure 2.16 (a) Schematic configuration of double­sided LSRA module. (b) 3D
model of LSRA . . . . . . . . . . . . . . . . . . . . . . . . . . . . 18
IE
Figure 2.17 TOGG centralized propulsion electrical vehicle . . . . . . . . . . . . 18
Figure 2.18 Tesla model s 85d centralized propulsion electrical vehicle . . . . . . 19
Figure 2.19 REE board electrical vehicle chassis incorporating in­wheel motors . 20
EV

Figure 2.20 Hiriko fold — a vehicle employing in­wheel motors . . . . . . . . . 21

Figure 2.21 Frequency response of sprung mass acceleration and tire deflection . 22
Figure 2.22 Random response of sprung mass acceleration and tire dynamic force 22
Figure 2.23 Simulink model of a pid control system . . . . . . . . . . . . . . . . 25
Figure 2.24 A fuzzy pid controller schematic . . . . . . . . . . . . . . . . . . . 27
PR

Figure 2.25 (a) Schematic of the controller and (b) the general architecture of
the controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 28
Figure 3.1 Full car model of 4x4 in­wheel motor driven electrical vehicle . . . . 31
Figure 3.2 Wan and schimmels human model . . . . . . . . . . . . . . . . . . . 36
Figure 3.3 Wheel load difference and roll angle . . . . . . . . . . . . . . . . . 38
Figure 3.4 Class c road input (velocity 90 kmph) . . . . . . . . . . . . . . . . . 41
Figure 3.5 A controlled system diagram . . . . . . . . . . . . . . . . . . . . . . 42
Figure 3.6 Flc controller . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 44
Figure 3.7 Gaussian type membership functions of error input . . . . . . . . . . 45
Figure 3.8 Gaussian type membership functions of derivative of error input . . . 45
Figure 3.9 Gaussian type membership functions of output . . . . . . . . . . . . 46
Figure 3.10 Flow chart of the multi­objective optimization algorithm based on
fast and elitist non­dominated sorting genetic algorithm . . . . . . . 49
Figure 4.1 Objective values of selected members (1 and 2 scale x10 ­ 5 to 7
scale x100) . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 61
Figure 4.2 Minimum objective values of populations in all generations . . . . . 62
Figure 4.3 kp fuzzy logic controller (a) E, (b) delE, and (c) output membership
functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 65

xii
Figure 4.4 kd fuzzy logic controller (a) E, (b) delE, and (c) output membership
functions . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 66
Figure 4.5 (a) kp and (b) kd surfaces of the fuzzy logic controller . . . . . . . . 67
Figure 4.6 Time response of human model . . . . . . . . . . . . . . . . . . . . 69
Figure 4.7 Comparison of ARh , ARut , V DVh , Aut . . . . . . . . . . . . . . . . 70
Figure 4.8 Sprung mass displacement graphs . . . . . . . . . . . . . . . . . . . 71
Figure 4.9 Suspension deflections . . . . . . . . . . . . . . . . . . . . . . . . . 72
Figure 4.10 Tire deflections . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 73
Figure 4.11 Control forces of active suspension . . . . . . . . . . . . . . . . . . 74
Figure 4.12 Awh graph (health guidance caution zone) . . . . . . . . . . . . . . . 75

W
IE
EV
PR

xiii
 LIST OF TABLES
Page
Table 3.1 Road roughness values classified by ISO 8608 (2016) (degree of rough­
ness S(Ω) × 10−6 ) . . . . . . . . . . . . . . . . . . . . . . . . . . . . 40
Table 3.2 Constants of full vehicle and human model . . . . . . . . . . . . . . . 58
Table 3.3 Design variable range for passive suspension elements . . . . . . . . . 60
Table 3.4 Design variable range of fuzzy logic controlled proportional and deriva­
tive gains . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60
Table 3.5 Design variable range for fuzzy logic controller . . . . . . . . . . . . . 60
Table 4.1 Minimum objective values . . . . . . . . . . . . . . . . . . . . . . . . 61
Table 4.2 Selected passive design variables for passive and fuzzy logic con­
trolled systems . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 63
Table 4.3 Selected membership function design variables of fuzzy logic controller 64
Table 4.4 Fuzzy logic controller kp rules of selected population member . . . . . 64
Table 4.5 Fuzzy logic controller kd rules of selected population member . . . . . 64
Table 4.6 Comparison of selected members . . . . . . . . . . . . . . . . . . . . 68

W
Table 4.7 Vertical wheel forces and load transfer index . . . . . . . . . . . . . . 69
Table 4.8 Time limits for vibration exposure . . . . . . . . . . . . . . . . . . . . 75
IE
EV
PR

xiv
1. INTRODUCTION

Electic Vehicles (EVs) attracted more research interest in recent years as the next gen­
eration transportation vehicles as fossil fuel­powered vehicles wreak havoc on the envi­
ronment by causing global warming. U.S Environmental Protection Agency states that
EVs provide a massive improvement against Internal Combustion Engines (ICEs) from
12­30% to 77% in terms of energy efficiency (DOE and EPA, 2021). More importantly,
tens of millions of people suffer, and millions of people lose their lives due to air pollu­
tion every year. Air pollution increases respiratory morbidity and mortality, which caused
3.7 million people to lost their lives just in 2012 due to poor air quality (Jiang, Mei, &
Feng, 2016). ICEs, as one of the major reasons for poor air quality in cities, presents a big
environmental impact with high emissions and fine particulate matter pollution.

W
As the impact of ICEs on the environment and public health has become more clear, more
IE
effort than ever before is being made to accelerate the development of EVs. As stated by a
report made by Edison Electric Institute, while the US housed 1 million EVs in 2018, 18.7
million EVs expected to be on US roads with annual sales exceeding 3.5 million, more
EV

than 20% of annual automobile sales, by 2030. While there is so much inclining to EVs in
the world, Turkey showed their interest by the foundation of Turkey’s Automobile Joint
Venture Group Inc., also known as TOGG. As the construction of Turkey’s EV production
PR

factory in Gemlik/Bursa continuing and the first production car is expected to hit the roads
in 2022, production capacity is expected to reach one hundred thousand by 2027 (Türkiye
Elektrikli ve Hibrid Araçlar Derneği, 2020). Moreover, as stated in March 2021, the New
Generation Commercial Vehicle and Battery Production facility in Kocaeli, a venture of
Ford Otosan, was announced and expected to reach a production capacity of two hundred
ten thousand commercial EVs and hundred thirty thousand EV battery by 2027 as Turkey’s
second EV production facility (Anadolu Ajansı, 2021). EVs has numerous benefits com­
pared to ICEs such as higher efficiency, enhanced comfort, lower cost of ownership, and
zero greenhouse gas emissions. All these advantages of EVs and environmental effects of
ICEs drive the industry to EVs.

1
While there is so much attention to EVs and their main differences to ICEs are their electric
motor and battery system, their suspension system requires as much research concentra­
tion. A vehicle’s handling capabilities are significantly affected by the suspension sys­
tem’s dynamic behavior, meaning that performance improvements of suspension systems
positively affect the comfort, forestall fatigue, health risks, and decrease traffic accidents
(D. Cao, Song, & Ahmadian, 2011).

It is possible to identify suspensions as passive, semi­active, and active (Hrovat, 1997).

Passive suspensions comprise spring and dampener elements. Passive systems comprise
the superiority of being an unsophisticated design, easy application, no energy consump­
tion, and high­reliability (W. Sun, Gao, & Shi, 2020). However, they are insufficient in
promoting comfort and handling due to the inability of variable spring and shock absorber

W
characteristics to handle different road conditions and conflicting performance expecta­
tions (Naudé & Snyman, 2003).
IE
Semi­active suspensions have shock absorbers that have changeable characteristics. This
means that damping coefficients or spring stiffness can be adjusted at a specific range, and
EV

due to their low energy consumption and high reliability, they are available in numerous
vehicles (Paulides, Encica, Lomonova, & Vandenput, 2006). Nonetheless, the resulting
damping or spring forces have the constraints of a passive suspension system and can
PR

only provide limited improvement in comfort, although they accommodate significant

improvement over passive suspension systems (W. Sun et al., 2020).

Active suspensions consist of the same damper and spring elements as passive systems.
In addition to these, an actuator is present in these types of suspensions, which separates
these systems from the other two. Active suspensions systems have high energy consump­
tion and can produce forces independent of relative suspension displacement or velocity.
Although due to their cost, size, and implementation difficulties, active suspensions are
not popular (Li, Liu, Gao, & Shi, 2012).

Suspension system design is also essential in EVs in precisely the same way they are
in ICEs. EVs also comprise passive, semi­active and active suspensions, and moreover,
they provide advantages over traditional ICE vehicles in terms of efficiency, lightweight

2
designability, and environmentally by zeroing greenhouse gas output. Electric motors,
compared to ICEs, contain a much fewer number of parts which in turn gives them the
advantage of being easier to design and manufacture, more prolonged lifespan makes them
economically desirable, much less noise and vibration creation during operation enhance
comfort, instantaneous speed control provides more performance output.

For EVs two approaches are possible for the positioning the electric motor, just like the
ICEs, the electric motor can be positioned on the sprung mass, vehicle body, and also
due to not having strict size restrictions of ICEs, electric motors can be positioned on the
unsprung mass, inside wheels. In­wheel Motor (IWM) EV design allows the vehicles to be
more compact (Nagaya, Wakao, & Abe, 2003). Nevertheless, this design also increases
the unsprung mass’s weight, drastically affecting comfort, handling, and health criteria

W
(Nagaya et al., 2003).
IE
Many studies have been carried out on the optimization of EVs and IWM configurations
(TOKSOY & YILDIZ, 2020; Yildiz & Özel, 2021). Compared to IWMs, EVs with cen­
tralized propulsion have the advantage of longer motor lifespan due to lower motor vi­
EV

bration, higher comfort, and better road handling due to lower unsprung mass (Liu, Gu,
& Zhang, 2017). Although centralized propulsion systems have these advantages, IWM
propulsion systems have been extensively researched owing to their native advantages
PR

such as simplicity, efficiency, swift and accurate torque generation without adverse ef­
fects on drive­shaft, ease of X­by­wire implementation, enhancing of Electronic Stabil­
ity Control (ESC) system, Traction Control System (TCS), and Anti­lock Brake Systems
(ABS) performance (Murata, 2012a). While IWMs have many upsides, some significant
disadvantages prevent them from being used in production cars, such as an increase in the
wheels’ mass leads to decreased comfort and road­holding abilities. Nagaya et al. (2003)
developed an IWM system which is defined as Advanced Dynamic Damper Mechanism
(ADM) to reduce these adverse effects. The motor acted as a dynamic vibration absorber
isolated from the unsprung mass by attaching the motor to the wheel via a passive suspen­
sion mechanism. It was found that this system reduced fluctuations in dynamic tire loads,
vibrations in the motor, wheel, and vehicle body.

3
While EVs and IWMs provide many improvements, the design phase of any suspension
system contains compromises because comfort, reliability, and safety requirements are
most of the cases trade­off objectives, and determination of suspension characteristics is
a choice that needs to be made in suspension design (H. Chen & Guo, 2005). Due to
suspension systems compromising nature, multi­objective optimization would benefit the
suspension systems development phase in acquiring optimal suspension.

W. Sun et al. (2020) defines the successful development of active suspension systems in
two stages. W. Sun et al. states that foremostly creating a dynamic model of the vehicle,
and secondly designing and optimizing the active system’s control strategy has a crucial
impact on comfort, health, and safety. W. Sun, Li, Huang, and Zhang (2017) states that
three external sources that cause dynamic vehicle responses are the road excitation, inertia

W
accelerations caused by acceleration/deceleration and turning.
IE
W. Sun et al. mentions that comfort and stability are connected with different Degrees
Of Freedom (DOF) movements; for instance, the vehicle body movement can be perilous
when the sprung mass roll causes disproportionate load distribution on tires, resulting in
EV

side­slip or rollover of the vehicle. Considering that active suspension improves stability
on top of the ride comfort, the frequently incorporated quarter or half car models are not
sufficient for active suspension analysis.
PR

This thesis further studies the multi­objective parameter optimization of an EVs active
suspension to address the above issues. In addition to the aforementioned studies, with
facts stated by W. Sun et al. (2017), a nonlinear full electrical vehicle model with quadratic
tire stiffness and cubic suspension stiffness with eleven DOF and a seat­driver model with
five DOF is established. Also, in addition to more traditional comfort and safety objec­
tives widely used in literature, rarely considered real working conditions of the suspen­
sion systems such as roll angle and load transfer are investigated. Then, multi­objective
optimization of a Proportional Derivative (PD) Fuzzy Logic Controller (FLC) is carried
out with a fast and elitist Non­dominated Sort Genetic Algorithm II (NSGA­II). Subse­
quently, responses of optimized passive and FLC controlled systems are matched against
each other. Key findings are presented in the conclusion.

4
2. LITERATURE REVIEW

2.1. Introduction

EVs compared to conventional ICEs provides many benefits such as high driving perfor­
mance, efficiency, and environmental friendliness. IWM configuration with numerous ad­
vantages is attracting more research interests. Regardless of its benefits, IWM technology
has its drawbacks, such as incrementing wheels mass causes critical deterioration in com­
fort and road­holding capabilities. Such compromises in suspension systems, as shown
in Fig. 2.1, affect not only conventional ICE vehicles with their more than a century­long
existence but also EVs regardless of their suspension type.

Extensive research of conventional vehicle suspension systems has been carried out con­

W
cerning control strategies, structural design, and dynamic performance (D. Cao et al.,
2011). Based on the studies related to IWM EVs, active suspension control is impera­
IE
tive for IWM configuration.
EV
PR

R�de Veh�cle
Comfort Stab�l�ty

Low Damp�ng H�gh Damp�ng

Figure 2.1. Trade­off nature of suspension systems (Simon & Ahmadian, 1998)

5
2.2. Vehicle Suspension Systems

Vehicle suspension systems are comprised of springs, shock absorbers, and mechanical
linking parts that connect the vehicle’s body to wheels. Suspension systems provide the
wheel­road contact, also preventing the body from rolling motion in cornering conditions
in addition to its primary purpose of isolating road­induced vibration (Gillespie, 1992).
From the system control perspective, categorization of suspensions is possible in three
distinctive classes: passive, semi­active, and active suspensions.

2.2.1. Passive suspension

Passive suspensions have well­known spring and damper elements that their character­
istics cannot be changed from their designed values. These systems provide numerous

W
advantages over semi­active and active suspensions, such as unsophisticated design, low
cost, and high reliability. A passive suspension schematic is shown in Fig. 2.2.
IE
mb
EV

ks cs
PR

ma

kt Road
Input
Figure 2.2. Passive suspension (Ahmadian, 2001)

Under the specifications of ISO 8608 (2016), vehicle ride comfort is usually assessed with
a random road profile or bump profile. For the prevention of mechanical suspension sys­
tem failures and ride comfort deterioration, suspension deflection constraint is utilized.
Dynamic tire forces are directly linked to road­holding abilities, and tire deflection values

6
are generally constrained and optimized for a good­performing suspension system. For
the improvement of vehicle vibration, passive suspension parameter optimization can be
carried out with a Genetic Algorithm (GA) (Mitra et al., 2016). Factors affecting dynamic
vehicle responses are irregularities in roadways, vehicle speed, and passive systems char­
acteristics.

Regarding cornering conditions, anti­roll torsional bars are implemented to improve ve­
hicle stability, such as roll angle. While torsional anti­roll bars improve roll stiffness
and provide a better cornering dynamic characteristic with a connection between the left
and right wheels, deterioration in road holding ability and ride comfort are undesired side
effects of such systems. Compared to a suspension system with spring elements, an anti­
roll bar comprised suspension system demonstrates improved performance, and indepen­

W
dent suspension systems outperform rigid axle suspension systems (Cole, 2000). D. Cao,
Rakheja, and Su (2010) stated that conflicting parameters such as comfort and handling
IE
performances could be improved with passive interconnected suspension systems.

Interconnected suspension systems can generate forces in all other wheels through me­
EV

chanical or hydraulic links between wheels with the displacement of one wheel. These
systems can potentially overcome the compromising nature of comfort and handling per­
formance (W. A. Smith & Zhang, 2010). N. Zhang, Smith, and Jeyakumaran (2010) re­
PR

searched hydro­pneumatic suspension systems comprehensively. On a different approach

to interconnected suspension systems, Hydraulically Interconnected Suspensions (HISs)
attracted many researchers’ attention. W. A. Smith and Zhang (2010) proposed a HIS
system that replaces the typical shock absorbers. In Figs. 2.3 and 2.4 an anti­roll HIS is
presented. It is seen that four cylinders placed at each wheel are interconnected by hy­
draulic lines, accumulators, damper valves, flexible hoses, and pipelines. W. A. Smith,
Zhang, and Hu (2011); Wang, Zhang, and Du (2012) presented a simulation and experi­
mental study of an SUV comprising HIS system subjected to various steering maneuvers,
and the results indicated that roll performance improvements and prevention of vehicle
rollover are achievable through implementation of these systems.

7
 W
IE
EV

Figure 2.3. Schematic of an anti­roll HIS (W. A. Smith et al., 2011)

PR

Figure 2.4. The assembled anti­roll HIS system (Wang et al., 2012)

8
2.2.2. Semi­active suspension

By comparison to passive suspensions, optimal suspension can be achieved with con­

trolled suspensions, improving comfort and handling. The first semi­active suspension
concept was presented by Karnopp and Crosby (1974) presenting the skyhook algorithm
to enhance vehicle suspension systems. A skyhook quarter car schematic is shown in
Fig. 2.5. Generally, in semi­active suspension systems, shock absorber constitutes a vari­
able characteristic over passive systems fixed­rate shock absorber or active suspension
systems actuator. Force generated by damper can either be regulated by changing area of
the orifice, which in turn changing fluid flow resistance, or by another variation such as
changing the fluid viscosity such as in Electrorheological (ER) and Magnetorheological
(MR) shock absorbers (Gopala Rao & Narayanan, 2009). A comparison of passive and

W
semi­active suspensions is shown in Fig. 2.6. As a competitor to active suspension sys­
tems, semi­active systems provide advantages, including cost savings by simple design
IE
and energy efficiency.
EV
PR

Figure 2.5. Schematic of a quarter­car suspension with skyhook control (Nguyen et al.,
2009)

9
Choi and Han (2007) states semi­active ER suspension systems are capable methods of
promoting comfort and road­handling, and with low latency to electric field stimulation,
ER fluids offer almost instantaneous control times. In their study, they proposed a con­
tinuously variable ER damper controlled by a skyhook controller. To demonstrate perfor­
mance of their suspension, Choi and Han experimented with four independent skyhook
controllers. ER shock absorber proposed in K. G. Sung, Han, Lim, and Choi (2007) is
shown in Figs. 2.7 and 2.8. Choi, Choi, and Park (1998) proposed an Sliding Mode Con­
trol (SMC) for an ER suspension. K. G. Sung, Han, Cho, and Choi (2008) proposed and
experimentally tested a fuzzy moving SMC controlled ER suspension on a quarter car
model. As these studies indicate, ER suspension systems coupled with an efficient con­
troller improve the comfort criteria of vehicles.

W
IE
EV
PR

a b
Figure 2.6. Suspension system model: (a) passive, (b) semi­active (Mihai & Andronic,
2014)

With higher viscosity, MR fluids yields better strength than ER fluids. Choi, Lee, and Park
(2002) presented a cylindrical MR shock absorber based on Bingham model of MR fluid
from several MR fluid models such as polynomial model, LuGre friction model, algebraic
model, Bouc­Wen hysteresis model, and neural network model.

10
Figs. 2.9 and 2.10 shows Choi et al. model’s schematics.

W
IE
Figure 2.7. ER damper schematic configuration (K. G. Sung et al., 2007)
EV
PR

Figure 2.8. Photograph of ER damper (K. G. Sung et al., 2007)

Yıldız, Sivrioğlu, Zergeroğlu, and Çetin (2015) proposed a nonlinear adaptive control of
a semi­active MR damper system with modified dynamic LuGre frictional model, con­
ducted experiments to improve the MR damper, and analyzed the performance of the MR
shock absorber suspension. Balamurugan, Jancirani, and Eltantawie (2014) presented a

11
 modified algebraic model and analyzed the performance of the MR damper, and thus a
controller with low computational complexity is realized through their studies. Du, Yim
Sze, and Lam (2005) proposed a polynomial model to characterize the dynamical behav­
ior of an MR shock absorber, and with comfort, handling, and suspension displacement
objectives.

W
IE
EV
PR

Figure 2.9. Schematic of an MR shock absorber (Choi et al., 2002)

Numerous control methods are deployed as semi­active MR shock absorber suspension

such as nonlinear adaptive control, sky­hook control, model­following SMC, H∞ and
fuzzy Proportional Integral Derivative (PID), etc. Xu, Ahmadian, and Sun (2014) pre­
sented an MR damper capable of variable spring and damping characteristics. Xu et al.
proposed two independent Fuzzy Logic (FL) and an on/off controller to control yaw mo­
ment and stiffness­damping characteristics, respectively. Maciejewski, Glowinski, and
Krzyzynski (2014) presented an adaptive PID capable of self­tuning for semi­active MR
shock absorber with four different controlling schemes for example PID, Proportional (P),

12

Reproduced with permission of copyright owner. Further reproduction prohibited without permission.