vii
TABLES OF CONTENTS
CHAPTER
1
2
TITLE
PAGE
DECLARATION
DEDICATION
ACKNOWLEDGEMENT
ABSTRACT
ABSTRAK
TABLE OF CONTENTS
LIST OF TABLES
LIST OF FIGURES
LIST OF ABREVIATIONS
LIST OF SYMBOLS
LIST OF APPENDICES
ii
iii
iv
v
vi
vii
xi
xiii
xviii
xix
xxi
INTRODUCTION
1
1.1 Introduction
1
1.2
Research Background
2
1.3
Problem Statements
5
1.4
Objectives and Scope of the Research
6
1.5
Research Contributions
7
1.6
Organization of the Thesis
8
LITERATURE REVIEW
11
2.1 Introduction
11
2.2 Classification of Vehicle Suspension Systems
12
2.2.1
Passive Suspension
12
2.2.2 Semi-Active Suspension
13
2.2.3
15
Active Suspension
2.3 Performance Index
16
2.4
Pneumatic Active Suspension Control Strategies
18
2.5
Active Force Control
19
2.6
Summary
23
viii
3
4
METHODOLOGY
25
3.1 Introduction
25
3.2
Modelling and Simulation
25
3.3
Experimental Study
31
3.4
Summary
33
SKYHOOK ADAPTIVE NEURO ACTIVE FORCE
CONTROL
34
4.1 Introduction
34
4.2 The SAFAFC Scheme
35
4.2.1
Adaptive Fuzzy System
35
4.2.2
Pneumatic Actuator System
39
4.2.2.1
Load Dynamics
40
4.2.2.2
Cylinder Chambers Dynamics
41
4.2.2.3
Valve Model Dynamics
42
4.2.3
Controller Design
43
4.2.3.1
Innermost Control Loop (PI Control)
43
4.2.3.2
Outermost Control Loop (PID
Control)
4.2.3.3
Intermediate Control Loops
(Skyhook and AFC)
47
4.3
Stability Analysis
54
4.4
Simulation
56
4.4.1 Road Profiles as Disturbances
59
Results and Discussions
60
4.5
4.5.1
Sinusoidal Wave Road Profile (Frequency 1.5
Hz, Amplitude 3.5 cm)
4.6
5
45
61
4.5.2 Other Results for Different Road Profiles
65
4.5.3 Effect of Load Variation
68
Summary
71
SKYHOOK ADAPTIVE FUZZY ACTIVE FORCE
CONTROL
72
5.1
72
Introduction
ix
5.2
The SANAFC scheme
72
5.3
Neural Network Model
73
5.4
Controller Design
77
5.4.1 Adaptive Neural Network and Its Application in
the AFC Loop
5.5
Simulation
82
5.6
Results and Discussion
83
5.6.1
Simulation Results in Time Domain
84
5.6.2
Simulation Results in Frequency Domain
96
5.7
6
77
Summary
104
EXPERIMENTAL IMPLEMENTATION OF THE
SANAFC SCHEME
106
6.1
Introduction
106
6.2
Quarter Car Suspension Test Rig
106
6.2.1
Passive Suspension
108
6.2.2
Data Acquisition System
109
6.2.3 Sensors
6.2.4
6.3
6.5
6.6
110
6.2.3.1
Accelerometer
111
6.2.3.2
Displacement Sensor
112
6.2.3.3
Pressure Sensor
113
Pneumatic Actuator System
114
Controller Development
115
6.3.1 Force Tracking Controller
115
6.3.2
Outermost Loop Controller
116
6.3.3
Intermediate Loop Controller
117
6.3.3.1 Skyhook
118
6.3.3.2 AFC scheme
118
Results and Discussions
122
6.5.1
Time Domain Response
123
6.5.2
Frequency Domain Response
131
Summary
134
x
7
DISCUSSION ON SIMULATION AND
EXPERIMENTAL DIFFERENCES
135
7.1
Introduction
135
7.2
General Findings of the Study
136
7.3
Main Simulation and Experimental Differences
140
7.3.1 Values of the Car Model Parameters
141
7.3.2
7.3.1.1
Sprung Mass
141
7.3.1.2
Unsprung Mass
141
Force Tracking Control of the Pneumatic
Actuator
7.3.4
142
Relative Measurement Performance of the
Various Control Schemes Compared to the
Passive Suspension
7.4
8
Summary
143
143
CONCLUSION AND RECOMMENDATIONS
144
8.1
Conclusion
144
8.2
Recommendations for Future Works
146
REFERENCES
147
Appendices A-C
154 - 168
xi
LIST OF TABLES
TABLE NO.
2.1
TITLE
PAGE
The summarised literature review of the active suspension
control strategy
20
2.2
The summarised literature review of the AFC strategy
23
3.1
The vehicle data
33
3.2
The pneumatic data
33
4.1
The RMS error values of the inverse dynamic model
response
54
4.2
Open loop characteristic of a quarter car model
59
4.3
RMS error values of the model response
63
4.4
Percentage improvement of the model compared with passive
suspension
4.5
A summary of results RMS error values of the model
response with different road profiles
4.6
63
67
A summary of percentage improvement results of the model
compared with passive suspension with different road
profiles
4.7
5.1
A summary of results for the SAFAFC scheme with different
road profiles
Error! Bookmark not defined.
RMS error values of the inverse actuator model response
using adaptive NN
5.2
81
RMS error values for the parameters of interest of the
SANAFC and SANAFC schemes
5.3
68
90
Percentage RMS error values for the parameters of interest of
the SANAFC and SANAFC schemes compared with passive
suspension apply different road profiles
5.4
91
A summary of results for the SANAFC scheme with different
road profiles
92
xii
6.1
Responses of the SANAFC, PID and passive suspension
schemes
6.2
Percentage responses of the SANAFC and PID compare with
passive suspension schemes
6.3
130
A summary of results for the SANAFC scheme with different
road profiles
7.1
129
131
A summary of average percentage improvement results for
the SAFAFC and SANAFC schemes compared with passive
suspension using different road profiles
7.2
138
A summary of results for the SANAFC scheme with different
road profiles
139
xiii
LIST OF FIGURES
FIGURE NO.
TITLE
PAGE
1.1
AFC concept applied to an active suspension system
5
2.1
Passive suspension system
13
2.2
Semi-active suspension system
14
2.3
Active suspension system
16
2.4
Power spectral density of various terrains
17
2.5
Human tolerance limits for vertical vibration
18
3.1
The flowchart of the research methodology
26
4.1
The proposed SAFAFC scheme
36
4.2
The structure of a fuzzy logic controller
38
4.3
A representation of the Gaussian membership function
39
4.4
The adaptive fuzzy structure
39
4.5
The pneumatic system
41
4.6
Force tracking control pneumatic actuator
44
4.7
Force tracking control pneumatic actuator
45
4.8
Outermost loop PID controller configuration
47
4.9
Fine tuning the PID controller
48
4.10
The intermediate loops comprising the skyhook and AFC subschemes
49
4.11
Skyhook damper configuration
49
4.12
The skyhook in the intermediate loop controller scheme
50
4.13
Tuning Bsky for the skyhook method
50
4.14
Inverse dynamic of the pneumatic actuator using AF
52
4.15
Membership functions for the input and output parameters of
AF
4.16
4.17
53
Response of the identified inverse dynamic model of the
pneumatic actuator
53
Estimated mass of the body using AF
55
xiv
4.18
The SAFAFC block diagram
55
4.19
Simplification of the SAFAFC block diagram
56
4.20
Simulink block diagram of a passive suspension
58
4.21
Bode plot of a passive suspension
59
4.22
Simulink diagram of the SAFAFC
60
4.23
Road profile inputs as disturbances
61
4.24
The time response simulation of all the control schemes
63
4.25
The frequency domain results for all the control schemes
65
4.26
Desired and actual forces of the SAFAFC scheme
66
4.27
The computed estimated mass of the body of the SAFAFC
scheme
4.28
66
The time domain results of the SAFAFC strategy with load
variation and sinusoidal road profile using f = 1.5 Hz and am =
3.5 cm
4.29
70
The frequency domain results of the SAFAFC strategy with
load variation and sinusoidal input road profile with f = 1.5 Hz
and am = 3.5 cm
71
5.1
The proposed SANAFC scheme
73
5.2
Model of an artificial neuron
74
5.3
The structure of a three-layered multilayer perceptron
75
5.4
The intermediate loop controller scheme
78
5.5
The structure of adaptive NN to identify the inverse
dynamics of the pneumatic actuator
79
5.6
Response of the adaptive NN to estimate the actuator force
80
5.7
Response of the error values the actuator force
81
5.8
A Simulink diagram of the SANAFC scheme
83
5.9
The time domain response of SANAFC scheme compared to
SAFAFC scheme subjected to sinusoidal road profile, f =
1.5Hz, am = 3.5 cm
5.10
85
The time domain response of SANAFC scheme compared to
SAFAFC scheme subjected to sinusoidal road profile, f = 0.8
Hz, am = 1.25 cm
5.11
86
The time domain response of SANAFC scheme compared to
SAFAFC scheme subjected to chirp signal road profile
87
xv
5.12
The time domain response of SANAFC scheme compared to
SAFAFC scheme subjected to sinusoidal wave hole test road
profile
5.13
The time domain response of SANAFC scheme compared to
SAFAFC scheme subjected to sleeper-plate test road profile
5.14
88
89
The time domain response of half-laden condition for the
SANAFC SAFAFC schemes subject to sinusoidal road
profile, f = 1.5 Hz, am = 3.5 cm
5.15
93
The time domain response of half-laden condition for the
SANAFC SAFAFC schemes subject to sinusoidal road
profile, f = 1.5 Hz, am = 3.5 cm as road profile
5.16
The actual force generated by the SANAFC and SAFAFC
schemes for various road profiles
5.17
95
The computed estimated mass the body of the SANAFC and
SAFAFC schemes for various road profiles
5.18
94
96
The frequency domain response of the SANAFC and
SAFAFC schemes subject to sinusoidal road profile, f = 1.5
Hz, am = 3.5 cm
5.19
98
The frequency domain response of the SANAFC and
SAFAFC schemes subject to sinusoidal road profile, f = 0.8
Hz, am = 1.25 cm
5.20
The frequency domain response of the SANAFC and
SAFAFC schemes subject to chirp signal as road profile
5.21
99
100
The frequency domain response of the SANAFC and
SAFAFC schemes subject to a half sinusoidal wave hole test
road profile
5.22
The frequency domain response of the SANAFC and
SAFAFC schemes subjected to sleeper-pate test road profile
5.23
101
102
The frequency domain results for a half-laden condition for
the SANAFC SAFAFC schemes subject to sinusoidal road
profile, f = 1.5 Hz, am = 3.5 cm
5.24
103
The frequency domain results for a full-laden condition for
the SANAFC SAFAFC schemes subject to sinusoidal road
profile, f = 1.5 Hz, am = 3.5 cm
104
xvi
6.1
A quarter car suspension test rig
107
6.2
Configuration of the hardware-in-the-loop simulation
108
6.3
Passive suspension test rig
109
6.4
Pin assignments for the analog and digital I/O connector pins
110
6.5
The DAS1602 card in the PC
110
6.6
Signal conditioning interface
110
6.7
Accelerometer attached to the sprung mass of the rig
111
6.8
Accelerometer attached to the tyre (unsprung mass) of the rig 111
6.9
Simulink and RTW block diagram of the sprung and
unsprung mass accelerometers to measure the acceleration,
velocity and displacement
112
6.10
Suspension deflection sensor attached to the end of actuator
113
6.11
Tyre deflection sensor attached to the base of tyre
113
6.12
Simulink and RTW block diagram of the two LVDTs used
tomeasure the suspension deflection and road profile
113
6.13
The pressure sensor
114
6.14
Simulink and RTW block diagram of the pressure sensor to
measure indirectly the actuated force
114
6.15
Pneumatic actuation system
115
6.16
Simulink and RTW block diagram of the pneumatic actuation
system
115
6.17
Results for the force tacking controller
116
6.18
Tuning of the PID Controller
117
6.19
Intermediate loop configuration
117
6.20
Tuning skyhook parameter Bsky
119
6.21
Identification of the inverse dynamic model of the pneumatic
actuator
120
6.22
Inverse dynamic pneumatic actuator using adaptive NN
120
6.23
The PLC system to generate the road profiles
121
6.24
Simulink and RTW block diagram of the LVDT to measure
the road profile
6.25
122
Road profiles (a) sinusoidal wave f = 1.1 Hz, am = 1.0 cm (b)
chirp signal (c) a half sinusoidal wave hole test (d) sleeperplate test
123
xvii
6.26
The time domain response of the passive suspension subject
to sinusoidal wave road profile, f = 1.1 Hz, am = 1.0 cm
6.27
124
The time domain response of the PID and SANAFC active
suspensions subject to sinusoidal wave road profile, f = 1.1
Hz, am = 1.0 cm
6.28
124
The time domain response of the SANAFC scheme with 40
kg load variation subject to sinusoidal wave road profile, f =
1.1 Hz, am = 1.0 cm
6.29
125
The time domain response of the SANAFC scheme with 75
kg load variation subject to sinusoidal wave road profile, f =
1.1 Hz, am = 1.0 cm
6.30
126
The actual force required by the PID and SANAFC schemes
for various road profiles
127
6.31
The computed estimated mass for various road profiles
128
6.32
The frequency domain response of the vehicle suspension
subject to sinusoidal wave road profile, f = 1.1 Hz, am = 1.0
cm
6.33
The frequency domain response of the SANAFC scheme
with 40 kg as load variation
6.34
132
133
The frequency domain response of the SANAFC scheme
with 75 kg as load variation
133
7.1
Time delay problem in the experiment
140
7.2
Configuration of the front unsprung mass configuration
142
xviii
LIST OF ABBREVIATIONS
A/D
: Analogue to Digital converter
AFC
: Active Force Control
AF
: Adaptive Fuzzy
BP
: Back Propagation
DOF
: Degree of Freedom
D/A
: Digital to Analogue converter
EC
: Evolutionary Computation
FFT
: Fast Fourier Transform
FLC
: Fuzzy Logic Controller
HILS
: Hardware-in-the-loop Simulation
LQR
: Linear Quadratic Regulator
LQG
: Linear Quadratic Gaussian
LM
: Levenberg-Marquardt
MF
: Membership Functions
MRAC
: Model Reference Adaptive Control
NN
: Neural Network
PI
: Proportional Integral
PID
: Proportional Integral Derivative
RTW
: Real Time Workshop
RMS
: Root Mean Square
SANAFC
: Skyhook Adaptive Neuro Active Force Control
SAFAFC
: Skyhook Adaptive Fuzzy Active Force Control
xix
LIST OF SYMBOLS
Aa
-
Piston effective areas a
Ab
-
Piston effective areas b
am
-
Amplitude
b
-
Bias
bs
-
Damping coefficient
Bsky
-
Constant value of skyhook
f
-
Frequency
fs
-
Semi-active damper force
fa
-
Active damper force
g
-
Gravitational acceleration
I
-
Identity matrix
J
-
Jacobian Matrix
Kp
-
Proportional gain
Ki
-
Integral gain
Kd
-
Derivative gain
ks
-
Spring stiffness coefficient
kt
-
Tyre stiffness coefficient
M
-
Mass of the air in the cylinder
ms
-
Sprung mass
mu
-
Unsprung mass
P
-
Pressure of the air in the cylinder
Pa
-
Absolute pressures in actuator’s chambers a
Pb
-
Absolute pressures in actuator’s chambers b
Q
-
Disturbance
R
-
Ideal gas constant
T
-
Temperature
V
-
Volume of the air in the cylinder
w
-
Weight
xx
xil
-
Centre of Gaussian antecedent MF at rule l and input i
yl
-
Centre of l of consequence fuzzy set
zs
-
Sprung mass displacement
zu
-
Unsprung mass displacement
zr
-
Road profile
zs–zu
-
Suspension deflection
zu – z r
-
Tyre deflection
z& s
-
Sprung mass velocity
&z&s
-
Sprung mass acceleration
z&u
-
Unsprung mass velocity
ψ
-
Cost function
σ il
-
Width of Gaussian antecedent MF at rule l and input i
μ
-
LM learning rate
xxi
LIST OF APPENDICES
APPENDIX
TITLE
PAGE
A
List of Publications
154
B
Results for the SAFAFC Scheme Subjected to Various
156
Road Disturbances
C
Results for the Practical Implementation of the
SANAFC Scheme
168