THE RIDE COMFORT VS. HANDLING COMPROMISE FOR OFF-ROAD VEHICLES

THE RIDE COMFORT VS. HANDLING COMPROMISE FOR OFF-ROAD VEHICLES
THE RIDE COMFORT VS. HANDLING
COMPROMISE FOR OFF-ROAD
VEHICLES
by
PIETER SCHALK ELS
Submitted in partial fulfilment of the requirements for the degree
Philosophiae Doctor (Mechanical Engineering)
in the
FACULTY OF ENGINEERING, THE BUILT ENVIRONMENT AND INFORMATION
TECHNOLOGY (EBIT)
UNIVERSITY OF PRETORIA
Pretoria
July 2006
© University of Pretoria
ii
THESIS SUMMARY
Title:
The Ride Comfort vs. Handling Compromise for Off-Road Vehicles
Author:
PIETER SCHALK ELS
Supervisor:
Prof. N.J. Theron
Department:
Mechanical and Aeronautical Engineering, University of Pretoria
Degree:
Philosophiae Doctor (Mechanical Engineering)
This thesis examines the classic ride comfort vs. handling compromise when designing a
vehicle suspension system. A controllable suspension system, that can, through the use of
suitable control algorithms, eliminate this compromise, is proposed and implemented.
It is a well known fact that if a vehicle suspension system is designed for best ride
comfort, then handling performance will suffer and vice versa. This is especially true for
the class of vehicle that need to perform well both on- and off-road such as Sports Utility
Vehicles (SUV’s) and wheeled military vehicles. These vehicles form the focus of this
investigation.
The ride comfort and handling of a Land Rover Defender 110 Sports Utility Vehicle is
investigated using mathematical modelling and field tests. The full vehicle, non-linear
mathematical model, built in MSC ADAMS software, is verified against test data, with
favourable correlation between modelled and measured results. The model is
subsequently modified to incorporate hydropneumatic springs and used to obtain
optimised spring and damper characteristics for ride comfort and handling respectively.
Ride comfort is optimised by minimising vertical acceleration when driving in a straight
line over a rough, off-road terrain profile. Handling is optimised by minimising the body
roll angle through a double lane change manoeuvre. It is found that these optimised
results are at opposite corners of the design space, i.e. ride comfort requires a soft
suspension while handling requires a stiff suspension. It is shown that the ride comfort vs.
handling compromise can only be eliminated by having an active suspension system, or a
controllable suspension system that can switch between a soft and a stiff spring, as well as
low and high damping. This switching must occur rapidly and automatically without
driver intervention.
A prototype 4 State Semi-active Suspension System (4S4) is designed, manufactured,
tested and modelled mathematically. This system enables switching between low and
high damping, as well as between soft and stiff springs in less than 100 milliseconds.
A control strategy to switch the suspension system between the “ride” mode and the
“handling” mode is proposed, implemented on a test vehicle and evaluated during vehicle
tests over various on- and off-road terrains and for various handling manoeuvres. The
control strategy is found to be simple and cost effective to implement and works
extremely well. Improvements of the order of 50% can be achieved for both ride comfort
and handling.
iii
SAMEVATTING VAN PROEFSKRIF
Titel:
Die Ritgemak vs. Hantering Kompromie vir Veldvoertuie
Outeur:
PIETER SCHALK ELS
Studieleier:
Prof. N.J. Theron
Departement:
Meganiese en Lugvaartkundige Ingenieurswese
Universiteit van Pretoria
Graad:
PhD in Ingenieurswese (Meganiese Ingenieurswese)
In hierdie proefskrif word die klassieke kompromie wat getref moet word tussen ritgemak
en hantering, tydens die ontwerp van ‘n voertuig suspensiestelsel ondersoek. ‘n
Beheerbare suspensiestelsel, wat die kompromie kan elimineer deur gebruik te maak van
toepaslike beheeralgoritmes, word voorgestel en geïmplementeer.
Dit is ‘n bekende feit dat, wanneer die karakteristieke van ‘n voertuigsuspensiestelsel
ontwerp word vir die beste moontlike ritgemak, die hantering nie na wense is nie, en ook
omgekeerd. Dit is veral waar vir ‘n spesifieke kategorie van voertuie, soos veldvoertuie
en militêre wielvoertuie, wat oor goeie ritgemak en hantering, beide op paaie en in die
veld, moet beskik. Die fokus van die huidige studie val op hierdie kategorie voertuie.
Die ritgemak en hantering van ‘n Land Rover Defender 110 veldvoertuig is ondersoek
deur gebruik te maak van wiskundige modellering en veldtoetse. Die volvoertuig, nielineêre wiskundige model, soos ontwikkel met behulp van MSC ADAMS sagteware, is
geverifieer teen eksperimentele data en goeie korrelasie is verkry. Die model is verander
ten einde ‘n hidropneumatiese veer-en-demperstelsel te inkorporeer en verder gebruik om
optimale veer- en demperkarakteristieke vir onderskeidelik ritgemak en hantering te
verkry. Ritgemak is geoptimeer deur in ‘n reguit lyn oor ‘n rowwe veldterreinprofiel te ry,
terwyl hantering geoptimeer is deur ‘n dubbelbaanveranderingsmaneuver uit te voer. Die
resultaat is dat die geoptimeerde karakteristieke op die twee uiterstes van die
ontwerpsgebied lê. Beste ritgemak benodig ‘n sagte suspensie terwyl beste hantering ‘n
harde suspensie benodig. Daar word aangedui dat die ritgemak vs. hantering kompromie
slegs elimineer kan word deur gebruik van ‘n aktiewe suspensiestelsel, of ‘n beheerbare
suspensiestelsel wat kan skakel tussen ‘n sagte en stywe veer, asook hoë en lae demping.
Dié oorskakeling moet vinnig en outomaties geskied sonder enige ingryping van die
voertuigbestuurder.
‘n Prototipe 4 Stadium Semi-aktiewe Suspensie Stelsel (4S4) is ontwerp, vervaardig,
getoets en wiskundig gemodelleer. Die stelsel skakel tussen hoë en lae demping, asook
tussen ‘n stywe en sagte veer binne 100 millisekondes.
‘n Beheerstrategie wat die suspensiestelsel skakel tussen die “ritgemak” en “hantering”
modes is voorgestel, op ‘n toetsvoertuig geïmplementeer en evalueer tydens
voertuigtoetse oor verskeie pad- en veldry toestande, asook tydens omrol- en
hanteringstoetse. Die beheerstrategie is koste-effektief en maklik om te implementeer en
werk besonder goed. Verbeterings in die orde van 50% kan behaal word vir beide
ritgemak en hantering.
iv
ACKNOWLEDGEMENTS
The research has been made possible through the generous support and sponsorship of the
U.S. Government through its European Research Office of the U.S. Army under
Contracts N68171-01-M-5852, N62558-02-M-6372 and N62558-04-P-6004.
Optimisation related investigations were performed under the auspices of the Multidisciplinary Design Optimisation Group (MDOG) of the Department of Mechanical and
Aeronautical Engineering at the University of Pretoria.
A special word of thanks to the following people who contributed in various ways to the
project:
• Dr. F.B. Hoogterp and Mr. Bill Mackie, US Army Tank Automotive Command
(TACOM)
• Dr. Sam Sampath – European Research Office of the US Army
• Prof. N.J. Theron and Dr. Petro Uys (University of Pretoria)
• Mr. Michael Thoresson, Mr. Werner Misselhorn, Mr. Karl Voigt, Mr. R. Bester
and Mr. B.P. Uys (My dedicated post-graduate students)
• Mr Gareth Thomas from Ford Motor Company (Land Rover) South Africa
• My parents without whose continual support and motivation this project would
not have been possible
Soli Deo Gloria
v
TABLE OF CONTENTS
Thesis summary
ii
Samevatting van proefskrif
iii
ACKNOWLEDGEMENTS
iv
LIST OF SYMBOLS
x
LIST OF ABBREVIATIONS
xiii
LIST OF FIGURES
xviii
LIST OF TABLES
xxvii
1.
INTRODUCTION
1.1
1.1
1.2
1.3
1.4
1.5
1.6
1.2
1.3
1.6
1.7
1.8
1.11
2.
Vehicle models
Controllable suspension system classification
Semi-active springs
Current applications of controllable suspension systems
Global viewpoints
Problem statement
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.1
2.1
Literature
2.1
2.1.1 Ride comfort
2.1
2.1.2 Handling, rollover and stability
2.3
2.1.2.1 Literature survey on handling, rollover and stability 2.5
2.1.2.2 Handling tests
2.11
2.1.2.3 Experimental investigation of handling
2.11
2.1.2.4 Results of experimental investigation
2.12
2.1.2.5 Conclusion from the handling investigation
2.15
2.1.3 Ride comfort vs. handling
2.16
vi
3.
2.2
Case Study 1: Landmine protected vehicle
2.2.1 Vehicle model
2.2.2 Terrain inputs
2.2.3 Results
2.2.4 Conclusions from Case Study 1
2.18
2.19
2.19
2.20
2.25
2.3
Case Study 2: Land Rover Defender 110
2.3.1 Vehicle model
2.3.2 Definition of “design space”
2.3.3 Simulation results
2.3.3.1
Ride comfort
2.3.3.2
Handling
2.3.3.3
Combined ride comfort and handling
2.3.4 Conclusion from Case Study 2
2.3.5 Follow-up work by Uys, Els and Thoresson
2.25
2.25
2.26
2.27
2.27
2.28
2.28
2.28
2.30
2.4
Validated vehicle model
2.4.1 Geometric parameters
2.4.2 Mass properties
2.4.3 Spring and damper characteristics
2.4.4 Tyre characteristics
2.4.5 ADAMS full vehicle model
2.4.6 Baseline vehicle tests
2.4.6.1
Instrumentation
2.4.6.2
Tests
2.4.7 Correlation between ADAMS model and test results
2.4.7.1
Transient response (APG track)
2.4.7.2
Handling (ISO 3888 double lane change)
2.4.8 Simulation results
2.32
2.32
2.32
2.32
2.32
2.32
2.33
2.35
2.35
2.37
2.37
2.37
2.41
2.5
Conclusion
2.42
POSSIBLE SOLUTIONS TO THE RIDE COMFORT VS. HANDLING
COMPROMISE
3.1
3.1
Published literature surveys on controllable suspension systems
3.1
3.2
Controllable suspension system hardware
3.2.1 Semi-active dampers
3.2.1.1 Magneto-Rheological (MR) fluids
3.2.1.2 Hydraulic bypass system
3.2.2 Semi-active springs
3.2.2.1 Air springs
3.2.2.2 Hydropneumatic springs
3.2.2.3 Other semi-active spring concepts
3.2.3 Active suspension systems
3.2.3.1 Electric actuators
3.2.3.2 Hydraulic actuators
3.2
3.2
3.2
3.3
3.6
3.6
3.7
3.8
3.9
3.9
3.9
vii
3.3
4.
Control techniques and algorithms
3.3.1 Combination of input and reaction driven strategies
3.3.2 Linear optimal, skyhook and on-off control
3.3.3 Neural networks and Fuzzy logic
3.3.4 H∞ control
3.3.5 Proportional Derivative (PD) control
3.3.6 Preview control
3.3.7 Model following
3.3.8 Frequency domain analysis
3.3.9 “Relative” control
3.3.10 Traditional controller design on the s-plane
3.3.11 Minimum product (MP) strategy
3.3.12 Roll and pitch velocity
3.3.13 Resistance control
3.3.14 Mechanical control
3.3.15 Steepest gradient method
3.3.16 Use of estimators and observers
3.3.17 Control of handling
3.3.18 Control of rollover
3.3.19 Ride height adjustment
3.3.20 Comparison of semi-active control strategies for ride
comfort improvement
3.10
3.10
3.12
3.16
3.16
3.16
3.16
3.17
3.17
3.18
3.18
3.18
3.19
3.19
3.19
3.19
3.19
3.19
3.20
3.20
3.4
Conclusion
3.21
3.5
Proposed solutions to the ride comfort vs. handling compromise
3.23
3.20
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4S4)
4.1
4.1
Literature
4.1.1 Hydropneumatic springs
4.1.2 Variable spring concepts
4.1.3 Hydraulic semi-active dampers
4.1
4.1
4.2
4.2
4.2
4S4 Working principle
4.2
4.3
Design requirements
4.4
4.4
Space envelope
4.6
4.5
Detail design of 4S4
4.6
4.6
Manufacturing of 4S4 prototypes
4.9
4.7
Testing and characterisation of the 4S4
4.7.1 Gas charging procedure
4.7.2 Bulk modulus
4.7.3 Thermal time constant
4.7.4 Spring characteristics
4.9
4.15
4.21
4.22
4.24
viii
4.7.5
4.7.6
4.7.7
5.
6.
Damping characteristics
Valve response times
Friction
4.25
4.30
4.30
4.8
Mathematical model
4.8.1 Modelling philosophy
4.8.2 Pressure dependent valve switching
4.8.3 Pressure drop over dampers and valves
4.8.4 Flow and pressure calculation
4.8.5 Implementation in SIMULINK®
4.8.6 Validation of the mathematical model
4.33
4.33
4.36
4.37
4.37
4.40
4.40
4.9
Conclusion
4.54
THE RIDE COMFORT VS. HANDLING DECISION
5.1
5.1
Literature
5.1
5.2
Suggested concepts for making the “ride comfort vs. handling
decision”
5.11
5.3
Easily measurable parameters
5.11
5.4
Experimental work on baseline vehicle
5.12
5.5
Evaluation of concepts
5.5.1 Frequency domain analysis
5.5.2 Lateral vs. vertical acceleration
5.5.3 Lateral vs. vertical acceleration - modified
5.5.4 Steering angle vs. speed
5.5.5 Disadvantages of proposed concepts
5.13
5.16
5.16
5.22
5.22
5.24
5.6
Novel strategies proposed
5.6.1 “Relative roll angle” calculated from suspension deflection
5.6.2 Running RMS vertical acceleration vs. lateral acceleration
5.28
5.28
5.28
5.7
Conclusion
5.35
VEHICLE IMPLEMENTATION
6.1
6.1
Installation of 4S4 hardware on test vehicle
6.1
6.2
Control electronics
6.6
6.3
Steady state handling
6.8
6.4
Dynamic handling
6.12
ix
7.
6.5
Ride comfort
6.18
6.6
Mountain pass driving
6.21
6.7
City and highway driving
6.21
6.8
Conclusions
6.21
CONCLUSIONS AND RECOMMENDATIONS
7.1
7.1
7.1
7.2
7.2.
Conclusions
7.1.1 The ride comfort vs. handling compromise
7.1.2 Possible solutions to the ride comfort vs. handling
compromise
7.1.3 The four-state semi-active suspension system (4S4)
7.1.4 The ride comfort vs. handling decision
7.1.5 Vehicle implementation
7.1.6 Final comments
Recommendations
7.2.1 The ride comfort vs. handling compromise
7.2.2 Possible solutions to the ride comfort vs. handling
compromise
7.2.3 The four-state semi-active suspension system (4S4)
7.2.4 The ride comfort vs. handling decision
7.2.5 Vehicle implementation
7.2.6 Additional possibilities
7.3
7.3
7.3
7.4
7.4
7.4
7.4
7.5
7.5
7.5
7.6
7.6
BIBLIOGRAPHY / REFERENCES
BR.1
APPENDIX A: HANDLING CRITERIA
A.1
x
LIST OF SYMBOLS
ENGLISH SYMBOLS:
A
Area [m2]
ay
Lateral acceleration [m/s2]
cv
Specific heat at constant volume [J/kg.K]
C
Damping coefficient [Ns/m]
Cαf
Cornering stiffness of front tyres [N/°]
Cαr
Cornering stiffness of rear tyres [N/°]
f
Fraction
fn
Natural frequency [Hz]
g
Gravitational constant = 9.81 [m/s2]
h1
Distance from centre of gravity to roll axis [m]
i
Index for accumulator or valve to be used (e.g. i=1: small
accumulator, i=2: large accumulator)
ks
Spring stiffness [N/m]
kt
Tyre stiffness [N/m]
Kϕf
Front suspension roll stiffness [N/rad]
Kϕr
Rear suspension roll stiffness [N/rad]
lf
Horizontal distance from front axle to centre of gravity [m]
lr
Horizontal distance from rear axle to centre of gravity [m]
M
Sprung mass [kg]
xi
m
Unsprung mass or mass of gas [kg]
P
Pressure [Pa]
Paccui
Pressure in accumulator i [Pa]
Pbegin
Pressure before valve opens [MPa]
Pend
Pressure after valve is fully open [MPa]
p
Steering Factor
P1
Pressure in small accumulator [Pa]
P2
Main strut pressure [Pa]
P3
Pressure between valve 2 and valves 3 and 4 [Pa]
P4
Pressure in large accumulator [Pa]
q
Flow rate [m3/s]
R
Universal gas constant [J/kg.K]
T
Temperature [K]
V
Volt [V]
V
Volume [m3]
V1
Valve 1 – damper bypass valve on small accumulator
V2
Valve 2 – damper bypass valve on large accumulator
V3
Valve 3 – Spring valve
V4
Valve 4 – Spring valve
v
Specific volume [m3/kg]
W
Vehicle weight [N]
Wb
British Standard BS 6841 vertical acceleration filter
Wf
British Standard BS 6841 motion sickness filter
x
Relative suspension displacement [m]
•
x
Ye
Relative suspension velocity [m/s]
Lateral position error [m]
xii
GREEK SYMBOLS:
ϕ
Roll angle [rad]
β
Bulk modulus of fluid [Pa]
∆
Difference
τ
Thermal time constant [s]
ω
Circular frequency [rad/s]
xiii
LIST OF ABBREVIATIONS
A
AAP
Average Absorbed Power
ABC
Active Body Control (Mercedes Benz)
ABS
Antilock Braking System
ADAMS
Automatic Dynamic Analysis of Mechanical Systems (Computer software)
ADC
Adaptive Damping Control
ADD
Acceleration Driven Damper
APG
Aberdeen Proving Ground
ARC
Active Roll Control
AWD
All Wheel Drive
B
BS
British Standard
BWR
Benedict-Webb-Rubin
C
CATS
Computer Active Technology Suspension (Jaguar)
CAN
Controller Area Network
cg
Centre of gravity
CDC
Continuous Damping Control (Opel)
xiv
CUV
Crossover Utility Vehicle
CVRSS
Continuously Variable Road Sensing Suspension (Cadillac)
D
DADS
Dynamic Analysis and Design System (Computer software)
DC
Direct Current
DFT
Discrete Fourier Transform
DHS
Dynamic Handling System
DIO
Digital Input Output
DRC
Dynamic Ride Control (Audi)
DSP
Digital Signal Processor
DWT
Draw Wire Transducer
E
EAS
Electronic Air Suspension (Volkswagen / Continental)
ECS
Electronic Controlled Suspension (Mitsubishi)
ER
Electro-Rheological
ERM
Electro-Rheological Magnetic
F
FFT
Fast Fourier Transform
Four-C
Continuously Controllable Chassis Concept (Volvo)
G
GA
Genetic Algorithm
GM
General Motors
GPS
Global Positioning System
xv
H
HiL
Hardware-in-the-loop
HMMWV High Mobility Multi-purpose Wheeled Vehicle
HP
Horse Power
HVOF
High Velocity Oxygen Fuel
I
ICS
In Cylinder Sensor
ISO
International Standards Organization
L
LDV
Light Delivery Vehicle
LQO
Linear Quadratic Optimal
LVDT
Linear Variable Differential Transformer
M
MISO
Multiple Input Single Output
MM
Mini Module
MP
Minimum Product
MR
Magneto-Rheological
MTTB
Mobility Technology Test Bed
N
N
Number of points
NAND
Inverted AND gate
NATO
North Atlantic Treaty Organisation
NHTSA
National Highway Traffic Safety Administration (USA)
xvi
NRMM
NATO Reference Mobility Model
P
PC
Personal Computer
PD
Proportional Derivative
PID
Proportional Integral Derivative
PSD
Power Spectral Density
R
ReS
Control strategy proposed by Rakheja and Sankar
RMS
Root Mean Square
RRMS
Running Root Mean Square
S
SSF
Static Stability Factor
SSRT
Steady State Rollover Threshold
SUV
Sports Utility Vehicle
SVFB
State Variable FeedBack
T
TACOM
Tank-automotive and Armaments Command of the US Army
TARDEC
Tank-Automotive Research, Development and Engineering Center of the US
Army
TRW
Automotive Component Manufacturer
U
USB
Universal Serial Bus
xvii
V
VDI
Verein Deutscher Ingenieure (Association of German Engineers)
VDV
Vibration Dose Value
VW
Volkswagen
Z
ZF
German component manufacturer
Other
4S4
4 State Semi-active Suspension System
2WS
Two wheel steer
4WS
Four wheel steer
xviii
LIST OF FIGURES
CHAPTER 1 – INTRODUCTION
Figure 1.1
-
¼ Car suspension system
1.2
Figure 1.2
-
Flow diagram of present study
1.13
CHAPTER 2 – THE RIDE COMFORT VS. HANDLING COMPROMISE
Figure 2.1
-
Test vehicle
2.2
Figure 2.2
-
Handling classification according to Harty (2005)
2.4
Figure 2.3
-
Ride and Handling track
2.13
Figure 2.4
-
Dynamic handling track – light vehicles
2.14
Figure 2.5
-
Figure 2.6
-
Suspension design space according to Holdman and Holle
(1999)
2.17
Photograph of vehicle used in simulation
2.19
Figure 2.7
-
Figure 2.8
-
Improvement in weighted RMS vertical acceleration (ride
comfort – linear spring)
2.22
Improvement in pitch velocity (linear spring)
2.23
Figure 2.9
-
Improvement in roll angle (linear spring)
2.23
Figure 2.10
-
Improvement in roll velocity (linear spring)
2.24
Figure 2.11
-
Land Rover Defender 110 vehicle
2.26
Figure 2.12
-
Results of ride comfort analysis
2.29
Figure 2.13
-
Results of handling analysis
2.29
xix
Figure 2.14
-
Combined ride comfort and handling
2.30
Figure 2.15
-
Path followed by Dynamic-Q
2.31
Figure 2.16
-
Tyre side-force vs. slip angle characteristic
2.33
Figure 2.17
-
Front suspension layout
2.34
Figure 2.18
-
Front suspension schematic
2.34
Figure 2.19
-
Rear suspension layout
2.36
Figure 2.20
-
Rear suspension schematic
2.36
Figure 2.21
-
Belgian paving
2.37
Figure 2.22
-
“APG” Bump
2.38
Figure 2.23
-
Constant radius test
2.38
Figure 2.24
-
Severe double lane change manoeuvre
2.39
Figure 2.25
-
Rough track
2.39
Figure 2.26
-
Rough track
2.40
Figure 2.27
-
Figure 2.28
-
Figure 2.29
-
Model validation results for passing over 100 mm APG bump
at 25 km/h
2.41
Model validation results for a double lane change manoeuvre
at 65 km/h
2.42
Ride comfort vs. gas volume and damping
2.43
Figure 2.30
-
Definition of handling objective function
2.43
Figure 2.31
-
Roll angle vs. gas volume and damping
2.44
Figure 2.32
-
Roll velocity vs. gas volume and damping
2.44
CHAPTER 3 – POSSIBLE SOLUTIONS TO THE RIDE COMFORT VS.
HANDLING COMPROMISE
Figure 3.1
-
Hydraulic two-state semi-active damper with bypass valve
3.3
Figure 3.2
-
Semi-active damper developed by Nell (1993)
3.4
Figure 3.3
-
Figure 3.4
-
Semi-active rotary damper developed by Els and Holman
(1999)
3.5
Operator controlled variable spring as proposed by Eberle
and Steele (1975)
3.8
xx
CHAPTER 4 – THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4S4)
Figure 4.1
-
4S4 circuit diagram
4.3
Figure 4.2
-
Relative suspension velocity over Gerotek Rough track
4.5
Figure 4.3
-
Pressure drop vs. flow rate for SV10-24 valve (Anon, 1998)
4.7
Figure 4.4
-
Operating range for SV10-24 valve (Anon, 1998)
4.7
Figure 4.5
-
Baseline left front suspension layout
4.8
Figure 4.6
-
Baseline left rear suspension layout
4.8
Figure 4.7
-
4S4 Suspension schematic diagram
4.10
Figure 4.8
-
4S4 suspension system – exterior view
4.11
Figure 4.9
-
4S4 suspension system – cross sectional view
4.12
Figure 4.10
-
Front suspension layout with 4S4 unit fitted
4.13
Figure 4.11
-
Rear suspension layout with 4S4 unit fitted
4.13
Figure 4.12
-
4S4 Prototype 2
4.14
Figure 4.13
-
4S4 Prototype 2 (left) compared to Prototype 1 (right)
4.15
Figure 4.14
-
4S4 Prototype 2 on test rig
4.16
Figure 4.15
-
4S4 Prototype 2 on test rig
4.17
Figure 4.16
-
4S4 Prototype 2 on test rig
4.18
Figure 4.17
-
4S4 Prototype 2 on test rig
4.19
Figure 4.18
-
4S4 Strut mounting to test rig
4.20
Figure 4.19
-
Measured bulk modulus
4.23
Figure 4.20
-
Determination of thermal time constant
4.24
Figure 4.21
-
Soft spring characteristic
4.25
Figure 4.22
-
Stiff spring characteristic
4.27
Figure 4.23
-
Soft and stiff spring characteristics
4.27
Figure 4.24
-
Pressure drop over valve 1
4.28
xxi
Figure 4.25
-
Curve fits on pressure drop data
Figure 4.26
-
Figure 4.27
-
Pressure drop over valve 3 (single valve vs. 2 valves in
parallel)
4.29
Damper characteristics for Prototype 2
4.29
Figure 4.28
-
Explanation of valve response time definitions
4.31
Figure 4.29
-
Valve response time for Prototype 1
4.32
Figure 4.30
-
Valve response time for Prototype 2
4.32
Figure 4.31
-
Hysteresis problem on Prototype 1
4.33
Figure 4.32
-
Effect of friction on soft spring at low speeds
4.34
Figure 4.33
-
Effect of friction on soft spring at high speeds
4.34
Figure 4.34
-
Effect of friction on stiff spring at low speeds
4.35
Figure 4.35
-
Effect of friction on stiff spring at high speeds
4.35
Figure 4.36
-
Measured input and output: stiff spring and low damping at
low speed
Comparison between measured and calculated values of P1
and P2: stiff spring and low damping at low speed
Comparison between measured and calculated forcedisplacement curve: stiff spring and low damping at low
speed
Measured input and output: stiff spring and low damping at
high speed
Comparison between measured and calculated values of P1
and P2: stiff spring and low damping at high speed
Comparison between measured and calculated forcedisplacement curve: stiff spring and low damping at high
speed
Measured input and output: stiff spring and low damping at
high speed, larger displacement stroke
Comparison between measured and calculated values of P1
and P2: stiff spring and low damping at high speed, larger
displacement stroke
Comparison between measured and calculated forcedisplacement curve: stiff spring and low damping at high
speed, larger displacement stroke
Measured input and output: stiff spring and high damping at
high speed
Comparison between measured and calculated values of P1
and P2: stiff spring and high damping at high speed
Figure 4.37
-
Figure 4.38
-
Figure 4.39
Figure 4.40
Figure 4.41
Figure 4.42
-
-
Figure 4.43
-
Figure 4.44
-
Figure 4.45
-
Figure 4.46
-
4.28
4.42
4.43
4.43
4.44
4.45
4.45
4.46
4.46
4.47
4.48
4.49
xxii
Figure 4.47
Figure 4.48
-
-
Figure 4.49
-
Figure 4.50
-
Figure 4.51
-
Figure 4.52
-
Figure 4.53
-
Figure 4.54
-
Figure 4.55
-
Figure 4.56
Figure 4.57
-
Figure 4.58
-
Figure 4.59
-
Comparison between measured and calculated forcedisplacement curve: stiff spring and high damping at high
speed
Measured input and output: soft spring and low damping at
low speed
Comparison between measured and calculated values of P1
and P4: soft spring and damping at low speed
Comparison between measured and calculated values of P2
and P3: soft spring and low damping at low speed
Comparison between measured and calculated forcedisplacement curve: soft spring and low damping at low
speed
Measured input and output: soft spring and low damping at
high speed
Comparison between measured and calculated values of P1
and P4: soft spring and low damping at high speed
Comparison between measured and calculated values of P2
and P3: soft spring and low damping at high speed
Comparison between measured and calculated forcedisplacement curve: soft spring and low damping at high
speed
Measured input and output, and valve 3 switch signal:
incremental compression test with low damping
Comparison between measured and calculated values of P1
and P4: incremental compression test with low damping
Comparison between measured and calculated values of P2
and P3: incremental compression test with low damping
Comparison between measured and calculated forcedisplacement curve: incremental compression test
4.49
4.50
4.50
4.51
4.51
4.52
4.52
4.53
4.53
4.55
4.55
4.56
4.56
CHAPTER 5 – THE RIDE COMFORT VS. HANDLING DECISION
Figure 5.1
-
The ride comfort vs. handling decision
5.2
Figure 5.2
-
Figure 5.3
-
TRW’s active roll control system according to Böcker and
Neuking (2001)
5.6
Steering wheel angle vs. vehicle speed
5.8
Figure 5.4
-
Steering wheel rotation speed vs. vehicle speed
5.9
Figure 5.5
-
Dive and squat vs. vehicle speed
5.9
Figure 5.6
-
Accelerator pedal press rate vs. vehicle speed
5.10
Figure 5.7
-
Accelerator pedal release rate vs. vehicle speed
5.10
Figure 5.8
-
City and highway driving route
5.14
Figure 5.9
-
Fishhook test
5.14
xxiii
Figure 5.10
-
Gerotek rough track top 800 m
5.15
Figure 5.11
-
Gerotek Ride and handling track
5.15
Figure 5.12
-
Double lane change test
5.15
Figure 5.13
-
Figure 5.14
-
Figure 5.15
-
FFT magnitude of vertical body acceleration (left and right
rear)
5.17
FFT magnitude of body lateral acceleration (left front and
left rear)
5.17
FFT magnitudes of body roll, yaw and pitch velocity
5.18
Figure 5.16
-
Figure 5.17
-
Figure 5.19
-
FFT magnitude of relative suspension displacement (all four
wheels)
FFT magnitude of steering displacement and kingpin steering
angle
FFT magnitude of relative suspension velocity (all four
wheels)
FFT magnitude of steering velocity
Figure 5.20
-
Strategy proposed by Nell (1993) as applied to city driving
Figure 5.21
-
Figure 5.22
-
Figure 5.23
-
Strategy proposed by Nell (1993) as applied to the rollover
test
5.21
Modified lateral vs. longitudinal acceleration for highway
driving
5.23
Modified lateral vs. longitudinal acceleration for rollover test 5.23
Figure 5.24
-
Steering limits vs. vehicle speed measured during three tests
5.24
Figure 5.25
-
Steer angle vs. speed implemented for city driving
5.25
Figure 5.26
-
Steer angle vs. speed implemented for highway driving
5.25
Figure 5.27
-
Steer angle vs. speed implemented for off-road driving
5.26
Figure 5.28
-
Steer angle vs. speed implemented for mountain pass driving
5.26
Figure 5.29
-
Steer angle vs. speed implemented for handling test
5.27
Figure 5.30
-
Steer angle vs. speed implemented for rollover test
5.27
Figure 5.31
-
Relative roll angle strategy for city driving
5.29
Figure 5.32
-
Relative roll angle strategy for highway driving
5.29
Figure 5.33
-
Relative roll angle strategy for off-road driving
5.30
Figure 5.34
-
Relative roll angle strategy for mountain pass driving
5.30
Figure 5.18
-
5.18
5.19
5.19
5.20
5.21
xxiv
Figure 5.35
-
Relative roll angle strategy for handling
5.31
Figure 5.36
-
Relative roll angle strategy for rollover
5.31
Figure 5.37
-
RRMS strategy for city driving
5.32
Figure 5.38
-
RRMS strategy for highway driving
5.32
Figure 5.39
-
RRMS strategy for off-road driving
5.33
Figure 5.40
-
RRMS strategy for mountain pass
5.33
Figure 5.41
-
RRMS strategy for handling test
5.34
Figure 5.42
-
RRMS strategy for rollover test
5.34
Figure 5.43
-
Effect of number of points in the RRMS on switching
5.36
Figure 5.44
-
Figure 5.45
-
Effect of number of points in the RRMS on the switching
delay
5.36
Effect of number of points in the RRMS on time spent in
“handling” mode
5.37
CHAPTER 6 – VEHICLE IMPLEMENTATION
Figure 6.1
-
Right rear suspension fitted to chassis – front view
6.2
Figure 6.2
-
Right rear suspension fitted to chassis – inside view
6.3
Figure 6.3
-
Right front and right rear suspension fitted to chassis
6.3
Figure 6.4
-
Right rear suspension fitted to test vehicle – side view
6.4
Figure 6.5
-
Right front suspension fitted to test vehicle – side view
6.5
Figure 6.6
-
Assembled hydraulic power pack
6.6
Figure 6.7
-
Control manifold for ride height adjustment
6.7
Figure 6.8
-
Piping, wiring and electronics
6.7
Figure 6.9
-
Control computer schematic
6.9
Figure 6.10
-
Constant radius test results
6.10
Figure 6.11
-
Relative roll angle front – effect of ride height
6.10
Figure 6.12
-
Relative roll angle front – effect of stiffness
6.11
xxv
Figure 6.13
-
Relative roll angle rear – effect of ride height
6.11
Figure 6.14
-
Relative roll angle rear – effect of stiffness
6.12
Figure 6.15
-
Body roll with 4S4 settings compared to baseline at 58 km/h
6.14
Figure 6.16
-
Effect of ride height on body roll at 58 km/h
6.14
Figure 6.17
-
RRMS control at 61 km/h
6.15
Figure 6.18
-
RRMS control at 74 km/h
6.15
Figure 6.19
-
RRMS control at 75 km/h
6.16
Figure 6.20
-
RRMS control at 83 km/h
6.16
Figure 6.21
-
RRMS control at 84 km/h
6.17
Figure 6.22
-
RRMS control compared to “handling mode” at 70 km/h
6.17
Figure 6.23
-
RRMS control compared to “handling mode” at 82 km/h
6.19
Figure 6.24
-
Body roll for “handling mode” at different speeds
6.19
Figure 6.25
-
Body roll for RRMS control at different speeds
6.20
Figure 6.26
-
RRMS control over Belgian paving at 74 km/h
6.20
Figure 6.27
-
Ride comfort of RRMS control compared to “ride mode”
6.21
Figure 6.28
-
RRMS control during mountain pass driving
6.22
Figure 6.29
-
City driving
6.22
Figure 6.30
-
Highway driving
6.23
APPENDIX A: HANDLING CRITERIA
Figure A-1
-
Figure A-2
-
Figure A-3
-
Figure A-4
-
Figure A-5
-
Performance related to driver A – Golf 4 GTI on ride and
handling track
A.2
Performance related to driver B – Golf 4 GTI on ride and
handling track
A.2
Roll angle histograms for drivers A and B – Golf 4 GTI on
ride and handling track
A.3
Lateral acceleration histogram for drivers A and B – Golf 4 GTI on
ride and handling track
A.3
Lateral acceleration, yaw rate and roll angle performance of a
Ford Courier on a dynamic handling track
A.4
xxvi
Figure A-6
-
Figure A-7
-
Figure A-8
-
Figure A-9
-
Figure A-10 Figure A-11 Figure A-12 Figure A-13 Figure A-14 Figure A-15 Figure A-16 Figure A-17 -
Figure A-18 -
Lateral acceleration, yaw rate and roll angle performance of a
Ford Courier on a ride and handling track
Lateral acceleration histogram for a Ford Courier on a
dynamic handling track
Roll angle histogram for a Ford Courier on a dynamic
handling track
Lateral acceleration histogram of a Ford Courier on a ride
and handling track
Roll angle histogram of a Ford Courier on a ride and
handling track
Lateral acceleration, yaw rate and roll angle performance of a
VW Golf 4 GTI on a dynamic handling track
Lateral acceleration, yaw rate and roll angle performance of a
VW Golf 4 GTI on a ride and handling track
Lateral acceleration histogram for a VW Golf 4 GTI on a
dynamic handling track
Roll angle histogram for a VW Golf 4 GTI on a dynamic
handling track
Lateral acceleration histogram for a VW Golf 4 GTI on a ride
and handling track
Roll angle histogram for a VW Golf 4 GTI on a ride and
handling track
Lateral acceleration and yaw rate performance of a Land
Rover Defender 110 on a ride and handling track (roll angle
data not available)
Lateral acceleration histogram for a Land Rover Defender
110 on a ride and handling track
A.4
A.5
A.5
A.6
A.6
A.7
A.7
A.8
A.8
A.9
A.9
A.10
A.10
xxvii
LIST OF TABLES
CHAPTER 1 – INTRODUCTION
Table 1.1 -
Classification of suspension systems
1.4
Table 1.2 -
Applications of controllable suspension systems
1.8
CHAPTER 2 – THE RIDE COMFORT VS. HANDLING COMPROMISE
Table 2.1 -
Summary of measurements
2.12
Table 2.2 -
Limiting parameter values (all vehicles and all drivers)
2.15
Table 2.3 -
Calculated spring stiffness for linear spring
2.20
Table 2.4 -
Natural frequencies for hydropneumatic spring
2.21
Table 2.5 -
Summary of the DADS simulation model
2.27
Table 2.6 -
Summary of results by Thoresson (2003)
2.31
Table 2.7 -
Instrumentation used for baseline vehicle tests
2.35
CHAPTER 3 – POSSIBLE SOLUTIONS TO THE RIDE COMFORT VS.
HANDLING COMPROMISE
Table 3.1 -
Control ideas evaluated by Voigt (2006)
3.22
CHAPTER 4 – THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4S4)
Table 4.1 -
Thermal time constants
4.23
xxviii
CHAPTER 5 – THE RIDE COMFORT VS. HANDLING DECISION
Table 5.1 -
Predictive control as implemented by Hirose et. al. (1988)
5.4
Table 5.2 -
Tracking control as implemented by Hirose et. al. (1988)
5.4
Table 5.3 -
Strategy used by Mizuguchi et. al. (1984)
5.4
Table 5.4 -
Candidate ideas for assisting with the “ride vs. handling” decision 5.11
Table 5.5 -
Directly measurable parameters
5.12
Table 5.6 -
Parameters that can be easily calculated from measurements
5.12
Table 5.7 -
Chosen tests and test routes
5.13
CHAPTER 6 – VEHICLE IMPLEMENTATION
Table 6.1 -
Comparison between baseline and 4S4 relative roll angles through double
lane change at 57 to 61 km/h
6.13
Chapter
1
INTRODUCTION
The main aim of a vehicle’s suspension system is to isolate the occupants from external
terrain induced disturbances, while still allowing the average driver to maintain control
over the vehicle and drive it safely. The design of vehicle suspension systems always
involves a compromise between ride comfort and handling. For good ride comfort a
compliant suspension system is normally required, while good handling demands a stiff
suspension system to control body roll.
With a normal passive suspension system, the characteristics of the springs and dampers
are fixed at the design stage and cannot be changed afterwards. By using controllable
springs and dampers, the suspension characteristics can be changed while the vehicle is
moving. It therefore becomes possible to have soft settings for good ride comfort whilst
traveling in a straight line on a good road, while the suspension characteristics can be
changed to a hard setting moments later to give good handling when the vehicle has to
change direction as required for lane changing or even accident avoidance. Settings can
also be adjusted based on the terrain roughness. With limited suspension travel available,
increased terrain roughness might require an increase in spring stiffness to prevent bumpstop contact and therefore improve ride comfort.
The problem becomes even more severe when the operational requirements of a vehicle
are in conflict with the suspension design. On most off-road vehicles high ground
clearance is required to enable crossing obstacles. Large suspension travel is also required
to keep all wheels in contact with the ground in order to maintain traction. Even load
distribution amongst the different wheels improves traction but requires soft springs. A
problem however arises when these vehicles have to be operated at high speeds on
smooth roads. The high center of gravity, large suspension travel and soft springs create
an inherent handling and stability problem making these vehicles prone to rollover.
A satisfactory solution cannot be obtained with a passive suspension system, but
controllable suspension systems have the potential to reduce or even eliminate the ride
comfort vs. handling compromise. The type of controllable suspension system that is the
topic of the current research, can significantly improve the situation by having a choice of
two discrete spring characteristics, as well as two discrete damper characteristics. The
system also features ride height adjustment capabilities giving control over ground
clearance and center of gravity height. This means that the suspension can be switched to
a setting optimized for off-road use (ride comfort and traction). Another setting,
optimized for high-speed on-road use, is available under conditions where good handling
is required. To obtain maximum benefit, switchover must occur automatically without
driver intervention. This being the main challenge that needs to be addressed before these
1.2
INTRODUCTION
systems can be successfully applied to vehicles:
decision”.
the “ride comfort vs. handling
The rest of this chapter gives background on several relevant topics such as vehicle
models, classification of controllable suspension systems, hydropneumatic springs,
production and concept applications of controllable suspension systems and global
viewpoints. The chapter closes with the problem statement and hypothesis.
1.1
Vehicle models
It can be argued that the simplest model of a car suspension system that is really useful is
the ¼ car representation as indicated in Figure 1.1. Here the vehicle body is represented
by the sprung mass M. The suspension, wheels and tyres are lumped as an unsprung
mass, m. The unsprung mass is connected to the sprung mass via a spring and damper and
to the road input via the tyre, normally represented only by a spring. Tyre damping is
usually small and therefore often neglected in ¼ car analyses. Suspension kinematics is
ignored and the two masses are only allowed vertical translation.
If the spring, damper and tyre characteristics are linear, the analysis can be performed in
either the frequency domain or the time domain. Non-linear characteristics however
require time domain analysis where a weighted root mean square (RMS) value of the
sprung mass acceleration is normally used as a measure of ride comfort.
This model can be useful to obtain first order values of spring and damper characteristics
required to meet a ride comfort specification. Although many authors attempt to get an
indication of handling by looking at the tyre force variations, this is an extreme
oversimplification of the handling phenomena (Miller 1988b). Handling is much more
complicated to simulate or measure on a vehicle, not least because it is very dependent on
the human driver. Human driver preferences and characteristics vary widely making
subjective vs. objective comparison of vehicle handling troublesome.
M
ks
c
m
kt
Figure 1.1 – ¼ Car suspension system
INTRODUCTION
1.3
The ¼ car model can be expanded to a ½ car model considering either roll or pitch
motion, resulting in a model with four degrees of freedom. The next logical step is a full
car model with seven degrees of freedom taking roll, pitch and vertical movement into
account. This type of model usually still ignores suspension kinematics and only allows
for vertical translation of the unsprung mass. However, to obtain useful results for
handling, longitudinal and lateral translations need to be added.
Throughout the present study, use is made of a non-linear full vehicle model. The model
is developed using the ADAMS multi-body dynamics code (Anon, 2002). The model
includes suspension kinematics, a non-linear tyre model and a human driver model.
Simulation is performed over real off-road terrain profiles. The model is validated against
vehicle test results and the full details are discussed in Chapter 2.
1.2
Controllable suspension system classification
Before continuing it is necessary to discuss the main categories of controllable suspension
systems. Confusion has been created by the inconsequent use of the terms adaptive, semiactive and active suspension systems. Suspension systems are classified for the purpose of
the current study as given in Table 1. This classification is based on that proposed by
Decker et.al. (1988). Williams (1994) also gives a description of various concepts.
All the controllable systems have physical limits imposed on them. Maximum force,
displacement, velocity and response times are usually limited by the hardware.
Passive suspension systems are very well known and still used on the vast majority of
new vehicles. Passive implies that the force-displacement and force-velocity
characteristics of the suspension system remain fixed at the design values throughout the
useful life of the components. Some degradation in performance usually takes place due
to wear and fatigue, but the characteristics cannot be modified without replacing or
manually adjusting components. Adjustable dampers that can be switched between
different characteristics using simple tools such as a screwdriver also fall in this category.
Adaptive systems can usually change certain characteristics slowly to adapt to changes in
vehicle load, speed or other operating conditions. These changes may take a few seconds
or a few minutes to have an effect. Self-levelling is the best-known example. Buchholz
(2003b) describes the ZF Sachs Nivomat mono tube damper with self-levelling feature.
Self-levelling is accomplished by using the energy that is generated by relative movement
between the vehicle body and suspension whilst moving. To accomplish levelling,
typically 1.6 to 2.4 km of driving is required. Adding an electric pump will eliminate the
need to drive the vehicle to accomplish self-levelling. The Nivomat system boasts two
features that normal air levelling systems lack, namely a load dependent spring rate
(which controls the ride frequency), and load dependent damping (which provides control
for additional mass/payload in rebound). The conventional Nivomat system first appeared
in the US market in 1996, fitted to a Chrysler vehicle.
Semi-active systems are classified as systems where the characteristics can be changed
rapidly (typically in less than 100 milliseconds). These systems can still only store energy
(springs) or dissipate energy (dampers). The most common example is the semi-active
damper where the damper force-velocity characteristic can be varied either between
1.4
INTRODUCTION
certain pre-defined discrete values (e.g. hard and soft), or continuously between a certain
minimum and maximum boundary.
Table 1.1 – Classification of suspension systems
Class
Passive
Characteristics fixed at design
stage
Adaptive
Characteristics can change
slowly
Semi-active:
Discreet
Characteristics can change
quickly between certain
discrete values
Semi-active:
Continuous
Characteristics can change
quickly and continuously
between certain limits
Active
Characteristics can change
quickly and continuously
between certain limits
Forces
Actuating times
Energy requirements
-
-
Force
•
x, x
Force
•
x, x
Longer than
characteristic
periods of
oscillation
Low
Force
•
x, x
Shorter than
characteristic
periods of
oscillation
Low
Force
•
x, x
Shorter than
characteristic
periods of
oscillation
Low
Force
•
x, x
Shorter than
characteristic
periods of
oscillation
High
According to Harty (2003), “the next step beyond current successful damper designs that
switch between two or a few more discrete calibrations – allows ‘step-less’ damper
characteristic modifications but represents a relatively new production technology.
Adjusting damper characteristics does not alter fundamental handling balance, although it
INTRODUCTION
1.5
can modify transient events (e.g. turn-in) by phasing damper switching front to rear
and/or side to side. Benefits include significant ride enhancements for a given level of
handling performance and the potential for control of transient effects using more
sophisticated algorithms than currently employed. However, handling balance cannot be
continuously altered.” The Magnetic Ride Control dampers on the Cadillac XLR are a
production example. This system employs magneto-rheological (MR) fluid and can
continuously vary damping at speeds approaching 1 millisecond.
The working principles of semi-active dampers can be divided into two main categories
namely: hydraulic dampers with bypass valves, and dampers employing magnetorheological fluids.
The Ohlins dampers fitted by Volvo amongst others, as well as the semi-active dampers
used in the present study are basically normal hydraulic dampers fitted with a valve that
can bypass the damper orifices. This bypass valve can have anything from two states
(open and closed) to being continuously variable.
Lord Corporation is one of the world leaders in the development and manufacture of
magneto-rheological fluids. New fluids have been developed in response to improvements
in MR technology driven by high volume production (Ponticel 2002). Both hydrocarbon
oil and water-based products are available. The viscosity of MR fluids is dependent on the
magnetic field applied. MR materials can change from a fluid to a near-solid state within
milliseconds when applying a magnetic field. System design is simple and control power
requirements are low. MR fluids are used in various applications, including automobile
dampers and seat suspension systems.
Delphi and Lord Corporation co-developed the MR fluid used in the Cadillac Seville STS
dampers (Jost 2002b). The MR fluid consists of iron particles suspended in a synthetic
hydrocarbon base fluid specifically developed for shock absorber application. In the “off”
state the MR fluid is not magnetized and the iron articles are dispersed randomly. When
applying a magnetic field, the iron particles align into fibrous structures, changing the
fluid rheology in the “on” state and thus the damping properties. The damping fluid can
change from a mineral oil type consistency (low damping forces) to a jelly-like substance
(high damping) within one millisecond.
Active suspensions usually replace springs and dampers with fast hydraulic, pneumatic or
electric actuators. A soft spring is often used in parallel with the actuator to reduce power
consumption by carrying the static wheel load. Active systems are still in the prototype
stage and suffer from high power demands and cost. Active suspension has the ability to
add significant amounts of energy to the system.
Low-bandwidth active suspensions can only control primary suspension modes and
front/rear roll moment balance at frequencies typically below 5 Hz. The Mercedes-Benz
Active Body Control (ABC) system is a current production example. Low-bandwidth
active suspension systems can offer ride benefits for a given level of handling
performance. The cost can also be considerably less compared to higher bandwidth
solutions (Harty 2003).
Full-bandwidth active suspensions can provide control up to 25 Hz. High-bandwidth
examples offer control of handling as well as body dynamics and thus theoretically
INTRODUCTION
1.6
improved performance compared to low-bandwidth systems. Several successful motor
sport applications and research vehicles, notably from Lotus, exists but no production
examples are known. The high bandwidth requires expensive hydraulic systems. High
forces and displacements means that power consumption is also high (Harty 2003).
Active suspension control using a linear electromagnetic actuator is described by
Buckner et.al. (2000). The actuator is installed in parallel with a soft spring supporting
the static load. The application is for a high mobility off-road military vehicle
(HMMWV). Decker et.al. (1988) also describes a magnetic spring in parallel with the
conventional steel spring.
A recent variation on the theme of controllable suspension systems is active or semiactive roll control. Kim and Park (2004) describe an electrically actuated active roll
control system. This system uses an electric actuator acting on the vehicle’s anti-rollbar.
The addition of a variable damper is also investigated. Several control strategies are
investigated with simulation and hardware-in-the-loop (HiL) testing. The system is said to
be effective in spite of the limited bandwidth of the actuator. “Active anti-rollbars for
warp control on one or both axles provide some of the platform-levelling benefits of lowband active suspension plus the ability to alter the handling balance continuously”. A
production system is implemented on the 2004 model year BMW 7 Series. Handling
benefits by being continuously adjustable from passive understeer to oversteer. The
technology has minimal package and power requirements, and is lower cost than even
low-band active suspensions. Disadvantages are the lack of pitch control and absolute
authority for roll control, unless both front and rear bars are used.
1.3 Semi-active springs
The two viable options considered for creating semi-active springs are either
hydropneumatic or air springs. Although the rest of this document will refer to
hydropneumatic springs, the same principles can be applied to air springs – the main
difference being the working pressure. Hydropneumatic springs have been applied to
military vehicles for many years. The most well known applications in passenger cars are
by Citroën (Nastasić & Jahn, 2005). The spring force in a hydropneumatic suspension
system is generated by compressing a gas in a closed container. The spring characteristic
is non-linear and governed by gas laws. For low speed excitation, the spring characteristic
can be approximated by isothermal compression, while for very high speeds, adiabatic
compression yields more accurate results. Care must be taken when the ideal gas
approach is used, as this results in significant errors for the typical pressures found in
hydropneumatic suspension systems.
Different models can be used to predict the spring characteristic of hydropneumatic
springs. A thermal time constant model is used by Els (1993), Els & Grobbelaar (1993)
and Els & Grobbelaar (1999). They take heat transfer effects and non-ideal gas
behaviour into account. Another modelling technique is to use an anelastic model
consisting of two parallel springs, one of which is in series with a damper. This approach
is described by Kornhauser (1994) and Giliomee, Els and Van Niekerk (2005) amongst
others.
INTRODUCTION
1.4
1.7
Current applications of controllable suspension systems
Citroën is considered to be the pioneers in the application of controllable suspension
systems to passenger cars. The working principles of the different Citroën
hydropneumatic suspension systems are described in considerable detail by Nastasić and
Jahn (2005). All models since the introduction of the Citroën DS at the 1955 Paris
Motor Show are described. This includes Hydractive I (first used on the Citroën XM),
Hydractive II (launched in 1993), Activa (used on some Citroën Xantia models),
Hydractive 3 and Hydractive 3+ (introduced on the C5).
The number of commercial applications of controllable suspension systems in production
cars is increasing rapidly. Most applications are however still limited to top of the range
models where the cost penalty can be easily justified. An example is Continuous
Damping Control (CDC), as developed by ZF Sachs. CDC was launched in 2001 as
optional equipment on cars such as the BMW 7 Series and Ferrari Modena and is now
available as optional equipment on the new (2004) Opel Astra (Jost 2004). The total
number of CDC equipped vehicles on the road was expected to rise to 225 000 in 2005
(Anon, 2004).
For the present study, application of controllable suspension systems to off-road vehicles
is of special interest. Land Rover implemented a cross-linked electronic air suspension
system on the 2003 model year Range Rover (Mayne 2002). The system joins the left
front and right front air springs via an electronically controlled valve. The same happens
on the rear suspension. When the valve is open, roll stiffness is dramatically reduced as
air flows between the left and right suspension system to equalise the pressure. This
results in low variation between wheel load left and right. The effect is less rocking on
rough terrain and better traction, because all wheels maintain effective ground contact. If
the valve is closed, the left and right suspension systems are isolated from each other –
the setting used for on-road driving. An electronic control unit, that uses vehicle speed
and suspension displacements as inputs, determines valve switching. The system is also
fitted with automatic load levelling and ride height control.
Volkswagen’s SUV, the Touareg, is fitted with air suspension and adjustable damping
(Birch 2002b). Manual adjustment of four different suspension height levels and three
different damper settings is provided. On-road ride height varies automatically according
to vehicle speed. Ground clearance is reduced from 215 mm to 190 mm above 125 km/h
although the driver can manually set other levels. Above 180 km/h the ride height is
automatically reduced to 180 mm.
The SmarTruck II, developed for the U.S. Army Tank-Automotive and Armaments
Command’s National Automotive Centre, is based on a Chevrolet Silverado 2500 pickup.
It features a heavy-duty adjustable air suspension system (Buchholz 2003a).
A summary of current production and prototype applications of controllable suspension
systems to light vehicles is given in Table 1.2. Features include adjustment of ground
clearance, ride height control to compensate for load changes and selectability of different
modes e.g. “Sport” or “Comfort”. This summary does not intend to be complete, but
rather aims to give the reader an overview of the typical applications of controllable
suspension systems.
1.8
INTRODUCTION
Table 1.2 – Applications of controllable suspension systems
Audi Le Mans Concept,
2003
Audi RS6
Bently Continental, 2003
BMW 7-series (2001) opt.
equip.
Cadillac Seville STS, Post
mid 2002
Cadillac Seville STS, Pre
mid 2002
Chrysler Pacifica, 2003
Citroën
Citroën Activa 2 Concept,
1990
Citroën C5
Citroën
C-Airlounge
Concept, 2003
Citroën DS, 1955
Citroën Xantia
Citroën XM
Citroën Xsara Dynactive
Citroën, 1993
Ferrari 306, 2003
Ferrari Modena (2001) opt.
equip.
Ford Visos Concept, 2003
Jaguar XJ, 2003
Jeep Grand Cherokee
Lamborghini Gallardo, 2003
Land Rover Range Rover,
2003
Maserati Quattroporte, 2004
Maserati Spyder, 2002
Mercedes Benz
(2004)
Opel Astra, 2004
A-class
Opel Insignia Concept, 2003
Toyota Soarer, 1986
US Army SmarTruck II,
2003
Volkswagen Phaeton, 2002
Volkswagen Touareg SUV,
2004
Volvo S60R and V70R,
2003
Number of applications
X
X
X
X
X
X
X
Dynamic ride control (DRC)
X
CDC (Continuous Damping Control)
X
MagneRide (Delphi)
X
X
X
X
X
X
X
X
Hydractive 3
X
X
X
X
X
X
X
Hydractive 3+
Hydractive 3
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Birch 2003b
Birch 2003c
Carney 2004b
Birch 2003d
Mayne 2002
Electronic Air Suspension (EAS)
Audi Magnetic Ride Technology
Continuously Variable Road Sensing
Suspension (CVRSS)
Nastasić and Jahn 2005
Birch 2003b
Nastasić and Jahn 2005
Nastasić and Jahn 2005
Nastasić and Jahn 2005
Birch 1999
Nastasić and Jahn 2005
Carney 2003a
Gehm 2004
X
X
X
X
X
X
X
Jost 2002b;
Alexander 2003
Jost 2002b
Gehm 2003
Nastasić and Jahn 2005
Birch et.al. 1990
X
X
Birch 2002c
Birch 2003d
Gehm 2004
Acronym
Semi-active Continuous
X
Semi-Active: Discrete
X
Adaptive
X
X
Class
Passive
X
Magneto-rheologic (MR)
Hydraulic
X
Passive
X
X
Hydropneumatic
Adaptive
Audi All-Road Quattro
Birch 2004a
Birch 2002a;
Birch 2002c
Jost 2002a;
Jost 2005
Birch 2003b
Damper
Load levelling
Audi A6 (2004)
Audi A8, 2003
Class
Ride height control
Spring
Active anti-roll bar
Reference
Air
Manufacturer
Activa
Hydractive I
Hydractive 3
Hydractive II
?
?
X
CDC (Continuous Damping Control)
X
X
Computer
Actives
Technology
Suspension (CATS)
DHS (Dynamic Handling System)
X
X
X
X
X
X
X
X
Carney 2003b
Kelly 2001;
Carney 2003b
Birch 2004b
X
X
X
X
Jost 2004
X
X
X
X
X
Birch 2003b
Hirose et.al. 1988
Buchholz 2003a
X
X
Jost 2002a; Jost 2005
Birch 2002b
X
X
X
X
X
X
X
X
X
X
X
X
X
X
Weissler 2003
4
8
9
15
15
X
X
“Skyhook”
“Skyhook”
X
Continuous Damping Control (CDC);
ISDPlus
X
Electronic Air Suspension (EAS)
X
Continuously
Controlled
Concept (four-C)
X
X
X
3
X
X
7 16
20
2
7
3
Chassis
3 12
It is clear that the application of controllable suspension systems is quickly gaining
ground in new passenger cars. Current fitment is to top-end road vehicles only. Current
systems are mostly semi-active dampers and/or ride height control. A few applications of
active anti-rollbars are also noted. Switching between various gas volumes have been
employed by Citroën and Land Rover. Fully active systems have only been realized on
prototypes. The developments by Lotus (Wright, 2001) are especially notable.
1.5
Global viewpoints
The viewpoints of several global experts in the vehicle suspension field are now discussed
to determine general trends and forecasts.
INTRODUCTION
1.9
Vertical load modification systems (e.g. springs and dampers) have a direct influence on
handling (Harty 2003). The scope for these systems to improve handling and stability is
dependent on the relationship between the vertical force applied to the tyre and the
corresponding lateral and longitudinal forces generated by the tyre.
According to Harty (2003) ride comfort can be significantly improved using adaptive
damper control. Stability can be influenced to a limited extent by damper control, because
the damper can only have an influence during transients. Stability is currently improved
by using brake-based systems. Brake-based systems are proven in the market. Active antirollbar technology can be used to good effect. The other candidate is variable geometry
active suspension, but packaging problems and bandwidth issues hamper progress.
Many of the current technical obstacles centre on sensing difficulties. Reliable sensing of
friction coefficient between the tyre and road is a problem. The body slip angle of the
vehicle also poses problems with many attempts focussed on calculating slip angle using
state estimators and techniques such as Kalman filtering to retrieve robust estimates from
noisy data.
Land Rover’s Director of Product Development, Steve Ross, comments on the use of
advanced technology as follows: “However, for us, technology is used to enhance both
off- and on-road capabilities. Technology is a means to an end in achieving greater safety,
security and refinement” (Birch 2001a). Range Rover’s computer-controlled air
suspension, which can vary vehicle height on and off road, is an added safety element,
lowering the centre of gravity (cg) when necessary, according to Ross.
Birch (2003a) summarized comments from several European industry experts on the
topic of integration of electronically controlled chassis and suspension systems.
Hugh Kemp, Engineering Director of International Automotive Technology
Business at Prodrive regards continuously variable dampers as having the potential to
offer most of the benefits promised by active suspension but at a “more realistic price”.
Prodrive favours mechanical variable orifice technology. Electronic control will also
allow a single damper specification to be used across a vast range of vehicles.
Michael Paul, Main Board Executive Director, Research and Development at ZF,
agrees that there are great opportunities for more intelligent single components. “The new
BMW 7 Series has an intelligent stabilizer system and electronically adjustable dampers.
As for fully active suspension, I would be very hesitant; it requires a lot of power, which
opposes the target of reducing fuel consumption”, he said.
“… for cost reasons, air springs and particularly active body control (ABC) will remain in
‘privileged’ market sectors” according to Hans-Joachim Schöpf, Executive Vice
President for Development, Mercedes Car Group.
Nevio di Giusto, Head of Product Development, Fiat/Lancia Business unit, believes
that any worthwhile project deserves a good chassis. “In the short term, I cannot envisage
a well-designed chassis not using electronic control” he said.
Clive Hickman, Managing Director of Ricardo Vehicle Engineering, regards the
integration of chassis electronics as a significant challenge, with the development of air-
INTRODUCTION
1.10
springs, roll control, adaptive steering and damping systems complementing antilock,
traction and stability programs.
Roberto Fedeli, who is responsible for the development of all Ferrari development
platforms, says that electronic systems should help the driver in some types of conditions
and support safety. On normal road and normal climatic conditions, electronics must not
mask the driver’s enjoyment of driving the car. He considers this to be a tuning problem
that depends on the control algorithms used.
Buchholz (2003c) looks at chassis developments for trucks and SUV’s from a North
American perspective and interviews experts in the field. Scott Bailey, Director of
Engineering for Energy and Chassis Systems at Delphi Corporation says: “With
advanced technologies, [we] can all but eliminate ride and handling compromises”.
Volvo became a first-to-market application example via the 2003 XC90 SUV’s active
Roll Stability Control System, which was co-developed by Ford Motor Company and
Continental Teves. Gyroscopic sensors are used to determine roll speed and roll angle.
The system uses the braking and traction control systems to prevent rollover. Bailey also
comments on trucks: “It’s an entirely different dynamic at play when you’re dealing with
a pickup truck versus a passenger car. A truck can be partially loaded, or fully loaded, and
is often towing something. Any of those conditions can significantly change the mass of
the vehicle, which means the dynamic models have to change. You have to get much
smarter about the different models that are at play to improve upon the control schemes.
The result is a much more sophisticated modelling requirement”. Bob Walker,
Engineering Director of Suspension and Exhaust Product Development for Visteon
Corporation says programs at several companies include the concept of an “active
corner” suspension. One particular active four-corner suspension system involves
“inducing force into the suspension system – typically to a shock absorber or strut – so it
can counteract the energy when the vehicle pitches in a particular direction”, says Walker.
Aly Badawy, TRW Automotive Vice President of Steering, Linkage, and Suspension
Engineering expects a production intent vehicle using the “active corner” concept to be
ready by 2008, but production will be dependent on the development of 42 Volt
technology, which he sees as “a must for an active corner”.
In another article on chassis integration from a North American perspective by
Alexander (2004b), Aly Badawy, TRW Automotive Vice President of Steering,
Linkage, and Suspension Engineering states that rollover avoidance and mitigation is
getting a lot of attention but there is no single technology to make a vehicle safe from
rollover. Active roll control (ARC) is under development and about to go into production.
Active damping control (ADC) is similar but allows for control of different sides of the
vehicle. The key to improvements is to integrate all systems. “Fundamentally, the
industry all has the same technology”. Sensing and control technology is the key to
having the best product.
Delphi has MagneRide in production that uses magneto-rheological fluid. Delphi’s
active stabilizer comes in two variations namely a single channel and a two-channel
version. It operates hydraulically but an electric version is under development according
to Brian Murray, Manager of Delphi’s Innovation Center in Brighton, MI.
It is clear that many industry specialists see the development of controllable suspension
systems as one of the trends that will increase in future.
INTRODUCTION
1.6
1.11
Problem statement
Controllable suspension systems have been implemented successfully in the case of topend passenger cars and is seen by industry specialists as the development trend of the
future. A void exists within the scope of vehicles that require good off-road capability
(high ground clearance, large suspension travel and soft springs), but also good handling
and stability on smooth roads at high speeds (low center of gravity and stiff springs).
Military wheeled vehicles, Sports Utility Vehicles (SUV’s) and Crossover utility vehicles
(CUV’s) all fall within this category.
The following hypotheses are made:
i)
ii)
iii)
iv)
v)
vi)
Ride comfort and handling have opposing requirements in terms of spring and
damper characteristics.
Suspension requirements for off-road use differ substantially from
requirements for high-speed on- road use.
A set of passive spring and damper characteristics, called the “ride comfort
characteristic” can be obtained that will optimise ride comfort over prescribed
off-road terrains at prescribed speeds. Additional improvements might be
possible by using “control” but is not considered for the purposes of this
thesis.
A set of passive spring and damper characteristics, called the “handling
characteristic”, can be obtained that will optimise handling for prescribed
high-speed manoeuvres on good roads. Additional improvements might be
possible using “control” but is not considered for the purposes of this thesis.
Advanced suspension system hardware that can switch between the passive
“ride comfort” and “handling” spring and damper characteristics, can be
feasibly implemented. Response time must be quick enough to enable control
of the sprung mass natural frequencies.
A robust decision can be made whether “ride comfort” or “handling” is
required for the prevailing conditions.
Chapters 2 to 6 of this thesis will investigate the validity of these hypotheses.
Fully active suspension systems have been explicitly eliminated for the purposes of the
current study due to prohibitive cost as well as power requirements during off-road
driving (although both power requirements and cost can be improved).
The purpose of this research is to design, develop, manufacture and test an advanced
suspension system that can eliminate the ride comfort vs. handling compromise for
vehicles that require good off-road capability, but also good handling and stability on
smooth roads at high speeds. The resulting suspension hardware is tested and
characterized to obtain all the parameters required for mathematical modeling.
A Land Rover Defender 110 vehicle was chosen as the platform for simulation and
testing of the controllable suspension concept proposed in the present study. The Land
Rover Defender is still considered by many, not least the marketing division, to be the
“best 4x4xfar” but the design is now dated. Although the vehicle behaves very well offroad, the on-road handling is less than desirable due to the soft suspension with large
suspension travel and the high centre of mass. The vehicle, with its ladder frame chassis
INTRODUCTION
1.12
and boxy styling, makes it relatively easy to change suspension components and
suspension mounting points. The vehicle is also fitted with coil springs. This means that
all the axle-locating functions are performed by suspension links (e.g. leading arms,
trailing arms, Panhard rods etc.) and not the springs, as is the case on many other vehicles
in this class fitted with leaf springs. These factors combine to make the Land Rover
Defender the ideal platform to test the controllable suspension concept.
In order to investigate the feasibility of the proposed suspension system, the project
consists of nine tasks namely:
i)
Develop a full vehicle dynamics simulation model to predict ride comfort and
handling.
ii) Validate the vehicle dynamics simulation model.
iii) Determine the required suspension characteristics for the “best” ride comfort and
“best” handling respectively, using the vehicle dynamics model.
iv) Design a prototype suspension system capable of producing the required
characteristics.
v) Manufacture the prototype suspension system according to the design.
vi) Test and characterise the prototype suspension system to determine feasibility and
conformance to specification.
vii) Develop a mathematical model of the prototype suspension system that can be
incorporated into the vehicle dynamics model.
viii) Develop a decision making methodology that can be used to switch the suspension
system between ride comfort and handling modes. From now on this will be called
the “ride vs. handling decision”. The ride vs. handling decision constitutes the main
challenge for the successful implementation of the controllable suspension system
proposed in this study and is therefore seen as the major contribution of the present
study.
ix) Fit the prototype suspension system to a test vehicle, implement the ride vs.
handling strategy and validate the strategy using vehicle tests.
The nine steps listed are represented graphically in Figure 1.2. The chapter in this thesis
where each step is further discussed is indicated below each block.
In Chapter 2, a validated, non-linear full vehicle model is used to investigate the
“optimal” characteristics for both ride comfort and handling. The conflicts between these
requirements are investigated and analysed using simulation.
The focus of Chapter 3 is on possible controllable suspension solutions to the ride vs.
handling compromise. A possible solution is formulated and investigated in greater detail
in Chapter 4 where the design, manufacturing, testing and mathematical modelling of the
proposed prototype system is described.
Chapter 5 looks at the crucial “ride comfort” vs. “handling” decision. Test data for
different driving conditions is analysed and different decision-making ideas investigated.
Vehicle implementation of the proposed hardware as well as the decision-making strategy
and final test results are discussed in Chapter 6.
1.13
INTRODUCTION
i) Non-linear full
vehicle model
ii)Validate
(Chapter 2)
(Chapter 2)
Ride comfort
criteria (objective
vs. subjective)
Handling criteria
(objective vs.
subjective)
(Chapter 2)
(Chapter 2)
iii) “Optimal”
suspension for ride
comfort
iii) “Optimal”
suspension for
handling
Conflicting
requirements
(Chapter 2)
(Chapter 2)
(Chapter 2)
Assume ride
comfort control
strategy is known
Assume handling
control strategy is
known
(Chapter 3)
(Chapter 3)
Two-state semi-active
hydropneumatic spring
combined with
Two-state semi-active damper
(Chapter 4)
viii) Investigate how the decision
between “ride” and “handling” can
be made based on easily measurable
parameters
(Chapter 5)
ix) Vehicle implementation
(Chapter 6)
Conclusion
Figure 1.2 - Flow diagram of present study
iv) Design
v) Manufacture
vi) Test
vii) Model
(Chapter 4)
Chapter
2
THE RIDE COMFORT VS. HANDLING
COMPROMISE
It is commonly accepted that vehicles with soft suspension systems generally provide
very good ride comfort at the expense of handling. Most sports cars suffer from the
opposite symptoms in that a firm suspension system offers excellent handling up to very
high speeds but then the ride comfort is often described as harsh or rough. It is apparent
that the design of a passive suspension system always involves a compromise between
ride comfort and handling.
In this chapter the ride comfort vs. handling compromise is investigated by means of two
case studies. The case studies are presented after analyses of literature on the subject.
The chapter concludes with the development and experimental validation of an ADAMS
model of a Land Rover Defender 110 vehicle. The ADAMS model is used to determine
spring and damper characteristics optimised for ride comfort and handling respectively.
2.1
Literature
In order to analyse the ride comfort vs. handling compromise, it is important to define the
concepts of “ride comfort” and “handling” separately. In general “ride comfort” is
associated with the vertical dynamics of the vehicle, primarily caused by road input
excitation. “Handling” is usually associated with the lateral, yaw and roll degrees of
freedom that are primarily a result of steering inputs by the driver.
In the remainder of this study, ride comfort is associated with vehicle dynamics caused by
road excitation. Handling is associated with vehicle dynamics due to steering inputs by
the driver.
2.1.1
Ride comfort
Ride comfort is described by Harty (2003) as follows: “Ride comfort is a frequencyweighted measure of vertical acceleration, together with subjective assessments of
harshness over lateral features and other secondary behaviours”.
Four methods to objectively evaluate ride comfort (also referred to as human response to
vibration) are used throughout the world today. The ISO 2631 standard (International
Standards Organisation, 1997) is used mainly in Europe and the British Standard BS
6841 (British Standards Institution, 1987) in the United Kingdom. Germany and
Austria use VDI 2057 (Hohl, 1984) while Average Absorbed Power or AAP (Pradko &
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.2
Lee, 1966) is used by the United States of America and by NATO in the NATO
Reference Mobility Model (NRMM). This presents two questions namely: (i) which
method is most suitable for the evaluation of off-road vehicle ride comfort, and (ii) how
does the results differ if different methods are used.
The correlation between objective methods for determining ride comfort and subjective
comments from crew driving in vehicles was investigated by Els (2005). For objective
measurements, the ISO 2631, BS 6841, AAP and VDI 2057 methods were used. The
emphasis was on the ride comfort of military vehicles operated under off-road conditions
over typical terrains.
An experiment was devised in which a 14-ton, 4x4 mine protected military vehicle (see
Figure 2.1) was driven over seven different terrains, using various vehicle speeds and tyre
pressures. The terrains were chosen to be representative of typical operating conditions in
Southern Africa and excite significant amounts of body roll, pitch and yaw motion. Seven
groups, consisting of 9 people each, were used for determining subjective comments
using a questionnaire, while simultaneously recording acceleration data required for
objective analysis at 11 positions in the vehicle.
Figure 2.1 - Test vehicle
The resulting sets of measured data were converted into objective ride comfort values
according to the ISO 2631, BS 6841, AAP and VDI 2057 methods. The unweighted
values were also used for comparative purposes. Objective values were calculated for all
the relevant parameters and measurement positions and compared to subjective ratings.
It is concluded by Els (2005) that any of the four methods under consideration, namely
ISO 2631, BS 6841, AAP and VDI 2057, could be used to objectively determine ride
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.3
comfort for the vehicles and terrains of importance for the study. The vertical acceleration
measurements give the best, and in fact the only reliable correlation and should be used in
all cases. The RMS values are sufficient for ISO 2631, BS 6841 and unweighted values.
Correlation for roll, pitch and yaw acceleration with subjective values is poor and not
useful.
According to Murphy (1984), a 6-Watt limit is normally assumed to be sustainable while
values as high as 12 Watts can be sustained only for a short period of time. It was found
by Els (2005) that a subjective response of 50% agrees with the 6 Watt AAP limit. Els
reports the corresponding limit for the other methods to be 2.0 m/s2 RMS (according to
ISO 2631 - rated as “very uncomfortable” according to Table 9 and Annex C in
International Standards Organisation, 1997). Limits of 2.8 m/s2 for the unweighted
RMS values, 1.8 m/s2 RMS for BS 6841, a VDI 2057 value of 88 and a 4-hour vibration
dose value (VDV) of 26 m/s1.75 were also found. The 4-hour VDV of 26 m/s1.75 is
significantly higher than the VDV of 15 normally assumed to be the guideline.
Terrain inputs used for measuring or simulating ride comfort are usually described as a
kind of Gaussian random or white noise input, often band limited to the frequency range
of interest (Karnopp, 1968). The power spectral density (PSD or roughness number)
approach is also used (International Standards Organisation, 1995; Gillespie, 1992;
Cebon 1999). Very few studies use physically measured road profile data as input,
because of the difficulties in accurately measuring the profile of rough roads. Discrete
obstacles and sinusoidal inputs are also used.
2.1.2
Handling, rollover and stability
Simulation of vehicle handling is virtually non-existent in literature when analysing
advanced suspension concepts. In isolated cases, handling is evaluated using step inputs
and evaluating transient response.
Figure 2.2 provides a graphical representation of the broader term “handling” as
described by Harty (2005). Harty defines handling using this representation as the
percentage of the available friction or the maximum achievable lateral acceleration
utilised by the vehicle-driver combination. At values lower than the linearity limit,
everybody can control the vehicle and avoid accidents. At values higher than the friction
limit, control over the vehicle is physically impossible and even the most experienced
driver in the best handling vehicle will loose control. He states several reasons why
intervention by electronic stability enhancement systems is required namely:
• Most drivers’ “in-head” model of the vehicle is based on linearity and zero phase
lags.
• Most drivers have no experience of significant loss of linearity.
• Most drivers have no experience of phase lag in yaw/sideslip resonance.
• When the vehicle departs from linearity, the population is very variable in its
ability to retain control of the car.
• There is a group of events where crashes occur even though the vehicle was not
exceeding the friction limit due to the driver’s lack of control skill.
• For road cars we need to match the car to the skill of the population.
• For motor sport we can calibrate the car to the individual driver’s skill level.
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.4
Harty sees the task of the vehicle designer as having two components namely: to raise the
absolute friction limit and to raise the linearity limit. These simplistic statements ignore
some problems namely:
• is the distance between the friction limit and the linearity limit a function of the
population only and not the car?
• are we allowing some people to have crashes at higher speeds than they could
previously have had?
Harty (2005) also compares several advanced methods for improving vehicle handling.
These methods include active front steering, rear wheel steering, brake-based stability
control as well as front / rear torque distribution. No mention is made of semi-active
suspension control.
Figure 2.2 – Handling classification according to Harty (2005)
While human response to vibration (ride comfort) has been extensively researched, a
single, unambiguous objective criterion for handling has eluded the vehicle science
community despite numerous studies pertaining to the topic. As Vlk (1985) notes with
respect to truck-trailer devices: "It is most desirable to define evaluation criteria for the
handling performance of vehicle combinations, both for steady state and transient driving
behaviour”.
For this reason a study was performed in order to establish relationships that can be used
to objectively quantify vehicle handling (Uys, Els & Thoresson, 2006). The idea was to
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.5
identify parameters that can be used to specify handling and that can be used as the
objective function in suspension optimisation studies. Literature on the topic will now be
reviewed, followed by test results for three different vehicles and four different drivers.
2.1.2.1 Literature survey on handling, rollover and stability
Horiuchi et. al. (1989) determined that drivers focus attention on yaw angle rather than
on lateral position error, Ye, for steering a two-wheel active steering vehicle (2WS). For a
four-wheel steer (4WS) vehicle, Ye becomes more important. Handling (steer response) is
measured in terms of yaw rate and lateral acceleration for handling characteristics of
four wheel active steering vehicles over a full manoeuvring range of lateral and
longitudinal accelerations (Masato, 1989).
Sharp and Pan (1991) comment that a vehicle that exhibits no body roll in general has
better steering behaviour than one that rolls.
Crolla et. al. (1998) obtained data using the ISO defined steady state, step input (J-turn)
and impulse steer tests (International Standards Organisation, 1982 and 1988).
Metrics used for their subjective/objective driver-handling correlations include: peak
lateral acceleration response time, peak road wheel steer angle and road wheel steer
angle response time, and peak steering angle torque and steering angle torque response
time. The authors conclude that frequency response results (lateral acceleration gain, yaw
gain, steering gain, steering phase) are of greater value in assessing vehicle response than
has to date been proven to be the case. These metrics, along with the change in sideslip
with respect to the change in lateral acceleration, were rated by drivers as uniform and
unequivocal indicators of steering response required. From an investigation on the
correlation between the different metrics, the authors found that, over smooth roads, the
degree of roll angle correlates with lateral acceleration gain; yaw gain and peak roll rate
and response. The degree of roll angle in transient cornering correlates with lateral
acceleration phase and yaw rate in a J–turn and steady state roll angle at 2 m/s2.
Controllability during a single lane change correlated with the J-turn yaw rate response at
2 m/s2 lateral acceleration (Crolla et. al., 1998).
From these studies it can be concluded that the degree of roll angle is indicative of the
lateral acceleration and yaw rate, which are both effective inputs for driver response.
Lateral transient response to step input is a frequently adopted measure for assessing
handling characteristics according to Reichardt (1991).
Since rollover is to an extent related to handling, although handling capability and
rollover aptitude is not similar, rollover considerations as deducted from the National
Highway Traffic Safety Administration (NHTSA) survey were investigated (National
Highway Traffic Safety Administration, 2000). In a proposed rulemaking exercise
NHTSA was considering a safety standard “that would specify minimum performance
requirements for the resistance of vehicles to rollover in simulations of extreme driving
conditions". The conclusion was that “vehicle rollover response is dominated by the
vehicle’s rigid body geometry with dynamic contributions from suspension effects.”
Analysis of 100 000 single-vehicle rollover crashes eventually focused on two static
measurements: tilt table angle (the angle at which a vehicle will begin to tip off a
gradually tilted platform) and critical sliding velocity (the minimum velocity needed to
trip a vehicle which is sliding sideways) – both measurements address situations in which
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.6
a vehicle encounters something that trips it into rollover (a curb, soft dirt, the tyre rim
digging into the pavement). Taking safety objectives into account, the following vehicle
stability metrics were considered as having a potentially significant role in rollover:
centre of gravity height, static stability factor, tilt table ratio, side pull ratio, wheelbase,
critical sliding velocity, rollover prevention metric, braking stability metric and
percentage of total weight on the rear axle (National Highway Traffic Safety
Administration, 2000).
The following aspects were considered by NHTSA for rollover rating:
a)
Static stability factor (SSF): This is the present static rollover rating calculated
by taking half a vehicle’s track width divided by its centre of gravity height.
b)
J-turn & Fish hook manoeuvres: In order to inform the public about a vehicle’s
stability with specific reference to rollover, NHTSA has chosen the J-turn and the
fishhook manoeuvre to rate a vehicle’s performance. “They are the limit
manoeuvre tests that NHTSA found to have the highest levels of objectivity,
repeatability and discriminatory capability.” The intention is that “vehicles will be
tested in two load conditions, using the J-turn at up to 97 km/h and the fish hook
manoeuvre at up to 80 km/h”. “Light load conditions will be provided by the test
driver who will be the test vehicle’s sole occupant. Heavy load conditions will be
created by adding a 79.5 kg mannequin to each rear seating position”. “The
dynamic manoeuvre test performance will be used to rate resistance to untripped
rollovers on a qualitative scale such as A - for no tip-ups, B - for tip-up in one
manoeuvre, C - for tip-ups in two manoeuvres etc.” (National Highway Traffic
Safety Administration, 2000) “The reverse steer of the fishhook manoeuvre will
be timed to coincide with the maximum roll angle to create an objective ‘worst
case’ for all vehicles regardless of differences in resonant roll frequency".
In response to NHTSA’s request for development of a dynamic test for rollover resistance
(National Highway Traffic Safety Administration, 2000), the following limiting values
for good rollover resistance were mentioned by General Motors: a) quasi-static centrifuge
test tip-up threshold of at least 0.9g; b) maximum lateral acceleration in a circular driving
manoeuvre of at least 0.6g; and c) a stability margin (a)-(b) at least 0.2g or 1.5/wheelbase
[units in m2]. GM estimated that a centrifuge measurement of 0.9g would correspond to a
SSF of 1.06. NHTSA however, estimated the centrifuge measurement as corresponding
closer to a SSF of 1.00, based on comparisons with tilt table tests with an allowance for
the vertical load error inherent with the tilt table. Ford (National Highway Traffic Safety
Administration, 2000) suggested lane change manoeuvres producing a maximum lateral
acceleration of 0.7g.
In the same survey NHTSA posed the question: Should measures of vehicle handling be
reported so that consumers can be aware of possible trade-offs? What indicators of
vehicle handling would be appropriate to measure, and how should this consumer
information be reported? The following responses are documented:
a)
Steady state lateral acceleration and lateral transient response: Nissan
recommended that NHTSA measure handling rather than rollover resistance, on
the basis that the fishhook test may be too severe for the purposes of consumer
information, and that Nissan had no data regarding the correlation of fishhook test
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.7
performance to real-world crashes. It suggested a steady state lateral acceleration
test and a lateral transient response test.
b)
The following comments based on ISO 3888 Part 2 (International Standards
Organisation, 2002) were made:
• Optimised cornering capability and “limit condition performance".
Daimler-Chrysler addressed the question directly by stating that its
recommended ISO 3888 PART 2 test does not give incentives for negative
trade-offs, but rather encourages optimised cornering capability and “limit
condition performance” by giving lower ratings for “bad handling”. In its
recommendation of the ISO 3888 PART 2 test, Continental-Tyres actually
described it as a handling test.
• Entry speed and peak-to-peak yaw rate.
Toyota suggested using the ISO 3888 PART 2 test as a handling test with both
entry speed and peak-to-peak yaw rate as performance criteria. The peak-topeak yaw rate would reflect on the yaw stability of the vehicle.
• Centrifuge and steady state lateral acceleration tests.
General Motors also recommended the centrifuge test, but suggested
combining its results with a driving test of steady state maximum lateral
acceleration to create a stability margin and set a lower limit for handling. In
addition to static and dynamic rollover resistance tests, a steady state lateral
acceleration test on a ski pad and “track-type tests to assess the vehicle’s
controllability, response and grip” is also recommended.
• Evaluation of double lane change
Daimler-Chrysler, Mitsubishi, Volkswagen, BMW and Continental-Tyres
recommended the ISO 3888 PART 2 closed-loop tight double lane change test
as the best dynamic rollover test, but also described it as a handling test.
Toyota, University of Michigan Transport Research Institute, Nissan,
Volkswagen and Ford recommend a separate handling test distinct from the
rollover rating with particular emphasis on yaw stability and Electronic
Stability Control.
• Double lane change vs. fishhook and J-turn.
Although all rollover resistance manoeuvres are influenced by both a vehicle’s
handling characteristics and its resistance to tip-up, it appears that handling
dominates the double lane change manoeuvres but is less important for the JTurn and Fishhook manoeuvres. The double lane change manoeuvres are
better for studying emergency vehicle handling than rollover resistance. Clean
runs of the CU and ISO 3888 tests are not limit manoeuvres in the sense of the
J-Turn and Fishhook because they cannot measure tip-up after the vehicle’s
direction control is lost. One way to characterize manoeuvres is by the number
of major steering movements they involve. The J-Turn has just one major
steering movement, the initial steer. A Fishhook has two major steering
movements, the initial steer and the counter steer. A double lane change has
four major steering movements, the initial lane change steer, the second lane
change steer, the recovery steer, and the stabilization steer, plus some minor
steering movements. These additional major steering movements increase the
influence of handling for Double Lane Change results compared to J-Turn and
Fishhook manoeuvres.
• Highest clean run.
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.8
NHTSA comments: "double lane change manoeuvres scored on the basis of
highest “clean” run speed had no value as dynamic tests of rollover
resistance". For a sample of test vehicles, there was actually an inverse
relationship between double lane change speed scores and the incidence of tipup in more severe manoeuvres that induced tip-up. The test vehicle that
tipped-up the most often in other manoeuvres and at a consistently lower tipup speed than other test vehicles, would be rated the best vehicle for rollover
resistance by ISO 3888 Part 2 double lane change on the basis of maximum
clean run speed. These tests measure a type of handling performance but do
not measure rollover resistance”.
Holdmann & Holle (1999) use effective dynamic wheel loads as a measure of driving
safety. By taking into account the RMS values of the dynamic loads, a hard damper
system assures driving comfort as well as driving safety up to 4 Hz. A soft damper system
assures good results for both at frequencies from 4 to 8 Hz. At higher frequencies, a soft
damper minimises body movement and a hard damper minimises dynamic wheel loads.
They state that different damping systems have a very small effect on lateral dynamics.
Choi et. al. (2001) indicate pitch motion and roll angle as measures of steering stability in
the evaluation a semi-active Electro Rheological suspension system.
For experimental comparison of passive, semi-active on/off and semi-active continuous
suspensions, Ivers and Miller (1989) use RMS tyre contact force as an indication of
wheel hop and road holding capability.
Data and Frigero (2002) note that it is possible to obtain valid objective indications of
vehicle handling behaviour by comparing subjective evaluations by drivers of steady state
circular tests, step steering wheel input and double lane change with objective parameters.
This resulted in the following objective parameters being proposed as representative of
vehicle behaviour:
• Lateral acceleration versus steering wheel angle,
• Yaw velocity versus steering wheel angle,
• Lateral acceleration versus yaw velocity,
• Roll angle versus lateral acceleration and
• Sideslip angle versus steering wheel angle.
Parameters, which are considered functions of lateral acceleration, are standardised with
respect to steering wheel activity, which is strongly influenced by driver activity. The
objective parameters representative of vehicle behaviour are the values of the regression
lines and their angular coefficients at 0.4 g lateral acceleration. It was found that there is
no correlation between a single partial rating and a single objective indicator. Linear
combinations of the objective indicators were used to find a maximum regression
coefficient. This resulted in a series of equations called partial indices that predict a
subjective rating, given objective parameters as input. From this paper the most important
parameters related to handling performance are roll angle, lateral acceleration and roll
velocity, which are related to steering wheel angle, yaw velocity and lateral acceleration.
In its presentation of rollover propensity testing of light vehicles (Forkenbrock and
Garrot, 2001), NHTSA suggests measuring steering wheel angle during a simple step
steer test, a J-turn and a fishhook turn; measuring dynamic weight transfer during a
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.9
double lane change and measuring the roll rate for a steering rate of 1000 o/s in a J-turn
and for 720 °/s during a fish hook turn.
In his studies of the onset of rollover, Dahlberg (2000) states that for the detection of
instability the most frequently used method is in-vehicle measurement of lateral
acceleration, followed by comparison to the steady state rollover threshold (SSRT) where
the accelerometer is mounted on the front axle. SSRT is considered the maximum value
of lateral acceleration that the vehicle may resist during steady state driving not to roll
over. It is a sufficient but not necessary requirement for rollover to occur. The static
stability factor (SSF) = ½ (average front and rear track width) divided by total centre of
gravity (cg) height, is a first order approximation to SSRT. It is the least conservative
estimation of rollover propensity and thus predicts a higher threshold. SSRT becomes
smaller as more flexibility is introduced in the analysis (suspension compliance, lateral
shift of cg, flexibility of tyres, chassis and frame flexibility). Another approach, taking
roll and roll moment into account in addition to lateral acceleration, gives a better
understanding of individual axle roll resistance. From such information it can be
determined that the vehicle can roll over when the lateral acceleration is larger than the
value corresponding to wheel lift. Rollover does not take place during steady state
driving, but during transient manoeuvres. SSRT is a best-case measure of roll stability,
whereas a worst-case measure is needed. Therefore the Dynamic Rollover Threshold is
defined being the minimum absolute peak value of lateral acceleration of all manoeuvres
bringing the vehicle to rollover. This defines a worst-case measure of roll stability. It is a
necessary but not sufficient condition for rollover.
Garrot et. al. (2001) describe experiments to determine untripped rollover propensity.
Different categories of vehicles are used – passenger cars, light delivery vehicles, vans
and sport utility vehicles. Vehicle characterisation is done by means of manoeuvres
designed to determine fundamental handling properties. For vehicles with relative higher
rollover propensity, measures are designed to produce two-wheel lift off. Vehicle
characterisation manoeuvres include: pulse steer, sinusoidal sweep, slowly increasing
steer and slowly increasing speed (at constant steering angle up to 0.7 g lateral
acceleration). Rollover propensity is determined from the following manoeuvres: J-turn,
J-turn with pulse braking, a Fishhook manoeuvre using a fixed 270 degree initial steering
input, a Fish hook manoeuvre using an initial steering angle 7.5 times the overall steering
ratio of a given vehicle and resonant steer. They relate the degree of lift off (minor,
moderate, major) and vehicle manoeuvring steer score, to rollover stability metrics (SSF,
tilt table ratio and critical sliding velocity).
Uffelman (1983) relates handling to the steering factor, p, calculated using the following
equation:
C l
p = αf f
(2.1)
C αr l r
where:
Cαf is the cornering stiffness of the front tyres,
Cαr is the cornering stiffness of the rear tyres,
lf is the distance from the front axle to the centre of gravity and
lr is the distance from the rear axle to the centre of gravity.
The limit of handling instability is considered at the point of a level tangent of the
steering wheel angle versus the acceleration graph. Uffelman considers performance
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.10
characteristics for quasi-steady-state cornering and braking. He shows that for a passenger
car the ratio p and steering wheel angle increase sharply for a lateral acceleration around
5 m/s2 for braking at 1m/s2 and between 4 and 5 m/s2 for braking between 2 and 4 m/s2
where the limit of adhesion is approached. These limits are dependent on braking balance
and load conditions.
El-Gindy and Mikulcik (1993) indicates that yaw rate gain (ratio of yaw rate to steering
angle) increases with increasing speed. The sensitivity of yaw rate gain to steering input
frequency increases with increasing speed, but the sensitivity to and increase in speed,
decreases as speed increases. The effect of mass, moment of inertia, front and rear
cornering stiffness and location of centre of gravity is also addressed. They conclude that
the strongest parameter on the yaw rate gain is the location of the centre of gravity. The
cornering stiffness of the front wheels has a more pronounced effect than the rear
cornering stiffness.
Starkey (1993) derives yaw rate and sideslip frequency response for a highway vehicle
from a yaw-plane handling model valid in the linear range.
Suspension technology capable of reconciling handling, stability and ride comfort has
been designed by Toyota Motor Company. The front and rear suspension settings react to
the lateral force input to the tyres (Kizu et. al. 1989).
In order to objectively evaluate handling performance, Harada (1997) derives stability
criteria for typical lane change cases and running against cross winds, applying a linear
preview control model to the driver and a bicycle model of the vehicle. The performance
index is composed of the weighted mean square values of state variables such as the
course deviation, steering correction angle, yaw velocity and lateral velocity. Stability
criteria consist of the steering control gain and steering time constant, which are obtained
numerically for a closed loop system by the Hurwitz criteria.
In his survey of the handling performance of truck-trailer vehicles, Vlk (1985) mentions
the following criteria that were used: lateral stability and movement, Hurwitz criterion for
stability, yaw angle, lateral displacements in tyre road contact paths, lateral play at the
hitch, side amplitude of trailer, frequency of trailer yaw oscillations, yaw rate gain, lateral
axle deviation, side slip angle, overturning risk, lateral acceleration, change of wheel
vertical loads, longitudinal tyre slip and cornering forces as a result of directional
response due to braking. He also mentions experiments by Zhukov who ascertained that
the roll rotation of a trailer was accompanied by a lateral displacement of both truck and
trailer from their direct path. The most outstanding correlation found was between trailer
roll and yaw.
EL-Gindy and Ilosvai (1983) mention a study of Yim et. al. that indicated that the slipratio of the front wheels relative to that of the rear wheels correlated with stability. ElGindy investigated lane change and braking manoeuvres on dry and wet asphalt and uses
lateral acceleration, yaw rate, lateral displacement and heading angle to determine
stability.
It is apparent from this survey that measurement of vehicle handling is not a clear-cut
matter. The aim of the survey was to determine whether a metric existed that could be
used to decide when a switch over from a soft to a hard suspension setting and vice versa
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.11
should occur. It should also be such that it can be used to optimise the suspension
settings. It is concluded from the information presented here that no such unambiguous
metric is apparent. Different authors use a variety of different metrics. There are,
however, some parameters that are worth considering. For example, the use of roll angle
is frequently encountered. These frequently encountered parameters were used to direct
the experimental investigation discussed in paragraph 2.1.2.3.
2.1.2.2 Handling tests
Handling tests can be divided into two main categories namely steady state handling tests
and dynamic handling tests (also called transient response tests).
The most widely used steady state handling test is the constant radius test, where the
vehicle is driven around a circle with constant radius (e.g. a dry skid pan). The most
important parameters that need to be measured are steering wheel angle and lateral
acceleration. The test starts at the lowest speed the vehicle can drive smoothly. Speed is
gradually increased until the constant radius cannot be safely maintained. A graph of
lateral acceleration against vehicle speed is used to determine whether the vehicle exhibits
oversteer (negative gradient), understeer (positive gradient) or neutral steer (zero
gradient) behaviour (Gillespie, 1992). Variations on this test method are the constant
steering angle test (where speed and radius changes) and the constant speed test (where
steering angle and radius changes).
Dynamic handling tests can be either closed loop where a human driver tries to steer the
vehicle through a prescribed path, or open loop where the steering angle vs. time is
prescribed. Closed loop tests include the severe double lane change test (ISO 3888-1,
International Standards Organisation, 1999), obstacle avoidance test (ISO 3888-2,
International Standards Organisation, 2002) and “Moose” or “Elk” test (Birch, 1998).
Open loop tests can be performed either by an experienced test driver or a computer
controlled steering robot. These include the J-turn (Garrot et. al., 2001), Fishhook
(Garrot et. al., 2001), step steer and pulse steer tests (ISO 7401, International
Standards Organisation, 1988).
2.1.2.3 Experimental investigation of handling
In previous simulation studies by Els and Uys (2003) it was shown that measurements of
roll angle could be used for optimisation of suspension settings. Choi et. al. (2001), Data
& Frigero (2002) and Crolla et. al. (1998) also refer to roll angle as a measure of
handling, as does the NHTSA survey and Vlk (1985) (see section 2.1.2.1). Other
parameters that have prominence in handling quality measurements are lateral
acceleration, dynamic weight transfer, roll rate, maximum entry speed to a clean run on a
double lane change and peak to peak yaw rate. Since dynamic weight transfer is very
dependent on the tyre model used in simulations and direct measurement poses
complications, this property is disregarded. For suspension control it is argued that in
general drivers do not drive vehicles at their performance limits, since they are not trained
to do so. Preferably parameters should be sought that can be measured during regular offroad driving, on highways and over mountain passes requiring greater handling skills.
Also, experience with the optimisation of suspension settings for both handling and
comfort, has indicated that convergence to an optimum can readily be obtained if
optimisation is first performed with respect to handling and then with respect to ride
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.12
comfort, with boundaries set on the handling parameters (Els and Uys, 2003). These
limits of secure handling, as experienced by drivers, have not been quantified as in the
case of comfort (see paragraph. 2.1.1).
With this background, an experiment was designed in which three vehicles were test
driven by four drivers. The vehicles consisted of a Ford Courier LDV, a Volkswagen Golf
1 Chico and a Volkswagen Golf 4 GTI. The drivers included a man in his twenties, a
woman in her forties, a man in his thirties and one in his forties. The vehicles were
equipped with accelerometers, displacement sensors, roll angle sensors and equipment to
measure speed. The measurements taken are indicated in Table 2.1. Measurements were
taken on two tracks at the Gerotek Test Facility outside Pretoria in South Africa: a ride
and handling track and a dynamic handling track for light vehicles. A single run on a
rough track representing off-road conditions was also performed. These tests are
considered preliminary to establish a procedure and base of comparison for future tests
that may also include a constant radius and double lane change test and will be supported
by a larger number of drivers.
Table 2.1 - Summary of measurements
Instrument
Accelerometer
Position
Front centre
Accelerometer
Right rear
Accelerometer
Left rear
Angle sensor
Gyro
Displacement
Speed sensor
Measurement
Lateral acceleration
Longitudinal acceleration
Vertical acceleration
Lateral acceleration
Longitudinal acceleration
Vertical acceleration
Lateral acceleration
Longitudinal acceleration
Vertical acceleration
Roll angle
Yaw angle
Roll rate
Yaw rate
Pitch rate
Steering wheel angle
Longitudinal speed
The Ride and Handling Track, of which a plan view is indicated in Figure 2.3, was
designed to evaluate the ride and handling characteristics and driveline endurance of
wheeled vehicles. The track is 4.2 km long and has 13 left turns and 15 right turns. The
maximum gradient on the low mobility course used for the tests is 15%.
The Dynamic Handling Track for light vehicles, indicated in Figure 2.4, was designed to
evaluate the high speed handling characteristics of light vehicles. The track is 1.68 km
long (excluding the spiral curve) and has an asphalt surface. The coefficient of friction is
0.7 Scrim (average). The track consists of a wave curve, trapezium curve, spiral curve as
well as a kink/hairpin combination.
2.1.2.4 Results of experimental investigation
The results of the experimental investigation into the development of handling criteria
will be discussed in more detail (Uys, Els and Thoresson, 2006). All measured data was
filtered with a 4 Hz low-pass filter so that only low frequency dynamics were observed.
The 4 Hz limit was also used because that is the specified frequency response limit of the
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.13
roll angle sensors used. Due to the large number of graphs, the graphs are given in
Appendix A and only the main conclusions are represented here. Figures A-1 to A-4 in
Appendix A refer to dynamic handling performance of a Volkswagen Golf 4 GTI as
related to two different drivers. Figures A-5 to A-16 are concerned with the performance
of the different vehicles, considering all the drivers, in order to obtain a global impression
of vehicle performance on both the dynamic handling and ride and handling tracks. Only
the results of the Courier and GTI are shown since these indicated the lowest and highest
performance levels. Figures A-17 and A-18 show some results of tests performed on a
Land Rover Defender 110. Unfortunately, roll angle sensors were not installed on the
Land Rover. The Land Rover data is included to determine if the same trends as those
observed for the other vehicle tests apply. The figures relate lateral acceleration, yaw rate
and vehicle speed.
Figures A-1 and A-2 indicate measurements for the VW Golf 4 GTI on the ride and
handling track for two different drivers. The trends in the relationships of longitudinal
acceleration vs. lateral acceleration, yaw rate vs. roll angle, yaw rate vs. lateral
acceleration and roll angle vs. lateral acceleration, are the same. This has been verified for
the other drivers as well. The limiting values do however differ. For example the upper
limit of the lateral acceleration is significantly higher for driver A than for driver B.
The same trends are also observed for different vehicles on different tracks, although the
absolute values differ (compare Figures A-5, A-6, A-11, A-12 and A-17).
Figure 2.3 – Ride and Handling track
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.14
Climbing curve
R 80
Wave curve
R 15
R 75
High speed sweeps
Trapezium curve
R 130
R 100
Spiral curve
R 130
R 22.5
Workshop
Parking area
R 22.5
Kink / hairpin combination
Figure 2.4 – Dynamic handling track – light vehicles
Referring to the yaw rate vs. roll angle, yaw rate vs. lateral acceleration and roll angle vs.
lateral acceleration graphs in Figures A-1 and A-2, linear dependency is observed. The
linear dependency amongst the indicated parameters holds true for:
i)
Different drivers (Figure A-1 compared to A-2),
ii)
Different vehicles (Figure A-5 compared to Figure A-11 and Figure A-6
compared to Figure A-12, refer also to Figure A-17 for yaw rate vs. lateral
acceleration)
iii)
Different test tracks (Compare Figures A-5 and A-6 and Figures A-11 and A-12).
Differences in gradients amongst the vehicles can be attributed to differences in
suspension roll stiffness. This effectively means that the same levels of lateral
acceleration can result in different roll angles for different vehicles, depending on spring,
damper and anti-rollbar characteristics as well as other vehicle parameters such as
suspension kinematics, centre of gravity height etc. This is especially true for off-road
vehicles with high centres of gravity that will normally roll over before the limits of tyre
side force are reached. In these vehicles, body roll and rollover propensity is more
important than the ultimate lateral acceleration that can be generated by the tyre forces.
The limiting hyperbolic tendency between lateral acceleration and vehicle speed is
apparent from Figures A-5, A-6, A-11 and A-12 (see “envelope in figures), confirming
the applicability of the handling control based on these limits (Hirose et. al. 1988).
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.15
The non-linear tendency between yaw rate and lateral acceleration for the Ford Courier
on the dynamic handling track (Figure A-5), is attributed to side-slip of the rear wheels,
since the vehicle exhibits considerable understeer behaviour that goes into limit oversteer.
Limited test results that were available for a Land Rover Defender 110 were also analysed
(Figures A-17 and A-18) and similar trends were observed.
The lateral acceleration and roll angle histograms (Figures A-3, A-4, A-7 to A-10, A-13
to A-16 and A-18) indicate more clearly the limits in lateral acceleration and roll angle
achieved on the various tracks by the different vehicles and drivers. It is clear from
Figures A-3 and A-4 that driver A spent more time at the vehicle limits, while driver B
kept within safer boundaries.
The difference in limiting values for the different vehicles, drivers and tracks can also be
observed. The limits are thus related to the track, driver and vehicle properties. More
noise is observed on the ride and handling track than on the dynamic track. The irregular
surface and bumps induce more high frequency motion.
No relation similar to that observed for yaw rate, lateral acceleration and roll angle is
observed for roll rate.
The limiting values relating to the tracks on which the tests were performed are listed in
Table 2.2.
Table 2.2 - Limiting parameter values (all vehicles and all drivers)
Parameter
Roll angle [°]
Lateral acceleration front centre [g]
Roll rate [°/s]
Yaw rate [°/s]
Steering angle [°]
Vehicle speed [km/h]
Longitudinal acceleration [g]
Ride and handling track
-3.5 to +3.5
-1.4 to +1.0
-32 to +32
-35 to +35
-60 to +130
0 to 120
-0.8 to +0.4
Dynamic handling track
-2 to +2
-1 to +0.7
-10 to +15
-32 to +35
-48 to +48
0 to100
-0.1 to +0.5
Lateral acceleration is often considered by analysts as a measure of handling
performance. The observed relationship between lateral acceleration and roll angle can be
verified by considering the moment distribution of a total vehicle about the roll axis
during steady state cornering (Gillespie, 1992):
Wh 1 /g
ay,
K φf + K φr − Wh 1
i.e. linear dependence determined by the roll stiffness.
φ=
(2.2)
Here ay is the lateral acceleration, Kφf is the front roll stiffness of the suspension, Kφr the
rear roll stiffness, W the weight and h1 the distance from centre of gravity to the roll axis.
2.1.2.5 Conclusion from the handling investigation
A linear relationship between lateral acceleration and roll angle has been observed in the
case of all drivers of different vehicles on a ride and handling as well as a dynamic
handling track. The range of values of roll angle observed for the tracks referred, is
between –3.5˚ and 3.5˚. The same levels of lateral acceleration result in different roll
angles for different vehicles, depending on vehicle parameters. This is especially true for
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.16
off-road vehicles with high centres of gravity that will normally roll over before the limits
of tyre side force are reached. In these vehicles, body roll and rollover propensity is more
important than the ultimate lateral acceleration that can be generated by the tyre forces.
Although the maximum lateral acceleration that can be achieved by a vehicle during, e.g.
a constant radius test, can be used as a measure of handling, at lower speeds the lateral
acceleration will only be a function of the vehicle speed and the radius of the turn, and not
of the vehicle suspension or tyre characteristics. Suspension and tyre characteristics will
however influence the over- or understeer behaviour. If a vehicle was tested through e.g.
a double lane change or constant radius test, at a speed below the maximum capability of
the vehicle, the lateral acceleration will be essentially the same for a wide range of
suspension characteristics, but the body roll angle will differ significantly.
The tests conducted strongly suggest that roll angle is a suitable metric to measure the
effect of suspension stiffness and damping on vehicle stability during handling tests.
From previous results it is known that roll angle is also suitable for the optimisation of
suspension settings given a prescribed road and manoeuvre. If levels of acceptable roll
angle can be determined, this metric can be used as criterion to ascertain the moment of
switchover for a semi-active suspension. Whether the value of the roll angle is a sufficient
indicator to determine suspension settings on rough roads, remains to be verified.
Although this study with four drivers and three vehicles is definitely not exhaustive, and
does not include off-road vehicles driven over rough terrain, it substantiates the use of
body roll angle as a measure of vehicle stability during handling tests. This result is
sufficient for the requirements of this study. Future research should include tests on a
larger number of vehicles and include more drivers to determine the limits of acceptable
roll angle.
2.1.3
Ride comfort vs. handling
According to Harty (2003), controllable suspension systems must be designed to deliver
improvements in ride comfort, handling and stability. These characteristics are to some
extend in conflict with each other. It is also important that controllable systems give the
maximum benefit for the smallest possible actuation forces or energy requirements.
Karnopp (1983) states that for ride comfort, the suspension should isolate the body from
high frequency road inputs. At lower frequencies the body and wheel should closely
follow the vertical inputs from the road to improve handling. Resonance of the body and
wheel should be controlled, so that these disturbances are not excessively amplified and
so that wheel hop and loss of wheel contact with the ground can be avoided. The
suspension must also control forces due to change in payload, forces from braking and
cornering and aerodynamic forces.
Wallentowitz and Holdman (1997) conclude that two spring stages are sufficient to
overcome the compromise associated with passive systems. A soft spring is required to
optimise ride comfort while a stiff spring is only used during cornering and braking when
the soft spring will result in unacceptable body roll and pitch.
2.17
Body acceleration
THE RIDE COMFORT VS. HANDLING COMPROMISE
Constant damping rate
Constant spring rate
• Sports car
Comfort
k
c
Active suspension
goal area
Safety
• Passenger car
Dynamic wheel load
Figure 2.5 - Suspension design space according to Holdman and Holle (1999)
Holdman and Holle (1999) investigate the possibilities to improve ride comfort and
handling of a 3.5-ton delivery vehicle. They illustrate the compromise in terms of the
graph given in Figure 2.5. Any passive system will only resemble one point on this graph
and is thus always a compromise between comfort and safety. They use three different
damping curves namely soft (2/3 of standard), standard and hard (1.5 times standard) and
investigate various skyhook-derived strategies. They find that for the passive damper,
damping should be high at frequencies below 4 Hz to ensure comfort and safety. Between
4 and 8 Hz, low damping gives best results for both comfort and safety. Above 8 Hz a
soft damper improves comfort, but a hard damper improves safety (minimizes the
dynamic wheel load). Different damping systems have a small effect on lateral dynamics
(handling). Additional forces need to be applied between the body and the wheel as a
function of lateral acceleration to reduce body roll angle.
Karnopp and Margolis (1984) discuss the effects of a change in spring and damper rates
on the transfer function of a single degree of freedom suspension system. It is said that
changing the damping alone is not a very efficient way of stiffening or softening a
suspension system. Changing the spring stiffness changes the natural frequency of the
system but the asymptotic attenuation at higher frequencies stays the same. The study
concludes that a system containing variable spring and damper rates can be very
advantageous in improving ride comfort.
A vehicle suspension system must be designed to provide adequate damping over a range
of driving conditions e.g. smooth and rough roads, laden and unladen conditions as well
as good ride comfort and handling according to Hine and Pearce (1988). This leads to
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.18
the well-known conflict between the maximum use of suspension working space for best
ride comfort and the need to provide sufficient displacement for all road conditions and
road surfaces. Their proposed solution is a two- or three-state semi-active damper with
ride height control.
According to Ikenaga et. al. (2000), vehicle suspension system performance is typically
rated by its ability to provide improved road handling and improved passenger comfort.
Automobile suspension systems using passive components can only offer a compromise
between these two conflicting criteria by providing fixed spring and damper
characteristics. Sports cars usually have stiff, harsh suspensions with poor comfort while
luxury cars offer good ride but poor handling. This compromise has existed since the
development of the first automobiles in the late 1800’s.
Nell (1993) states that suspension design involves a compromise between two conflicting
requirements. To ensure good support of the vehicle body, stability, handling and side
wind stability a stiff suspension is required. Good vibration and shock isolation on the
other hand requires a soft suspension. Soft suspension characteristics suffer drawbacks
e.g. when the suspension working space is limited, frequent bump stop contact can occur.
Large static position changes due to variations in vehicle load can also be a problem.
Many investigators therefore agree that ride comfort and handling requirements are often
in conflict with each other. The two case studies that will be presented now analyses the
spring and damper characteristics required for optimum ride comfort and handling
stability as applicable to off-road vehicles.
2.2
Case study 1: Landmine protected vehicle
Els & Van Niekerk, (1999) perform evaluation of the ride comfort and handling of a
heavy off-road military vehicle using DADS (Dynamic Analysis and Design System)
software. The vehicle used for simulation is a 12-ton 4x4 military vehicle, designed for
off-road use over very rough terrain. A photograph of the vehicle is shown in Figure 2.6.
The three-dimensional, multiple degree of freedom, non-linear DADS simulation model
consists of 11 rigid bodies (vehicle body, 4 wheels, front axle, rear axle, ground, 2 front
hubs and steering pivot). The wheels and hubs are connected to the axles using 7 revolute
joints while axle locating rods and steering links are modelled using 10 sphericalspherical joints. Force elements consist of non-linear dampers, springs (linear and
hydropneumatic, depending on simulation), bump stops, as well as a generic non-linear
tyre model.
The resulting model has 66 degrees of freedom but after adding joints, constraints and a
driver model, 14 unconstrained degrees of freedom remain. These consist of the vehicle
body displacements (lateral, longitudinal, vertical, roll, pitch and yaw), wheels (rotation),
front axle (vertical, roll) and rear axle (vertical, roll). Non-linear spring, damper, bump
stop and tyre characteristics are used. The vehicle is steered over a predetermined course
by a simple driver model that estimates the lateral position error based on the yaw angle
of the vehicle body at the current time step and the desired lateral position at the driver
preview time of 0.6 seconds. The driver model is implemented using amplifiers, summers
and input elements.
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.19
Figure 2.6 - Photograph of vehicle used in simulation
2.2.1
Vehicle model
To make provision for controlling the semi-active spring-damper system in future
simulation work, the DADS model is exported for use in Matlab/Simulink by defining
plant inputs and outputs. By doing this, the complete non-linear DADS model is included
as an S-function and solved in Simulink. In this case, the DADS model provides relative
spring displacements, damper velocities and other parameters needed as inputs to the
control algorithm. Simulink then calculates the spring and damper forces according to the
control strategy and outputs these forces to the DADS model.
2.2.2
Terrain inputs
In this study, typical rough terrain inputs, such as a Belgian paving track, were used.
Instead of performing only the normal straight line ride comfort simulation, the vehicle is
also steered over the terrain in order to try and follow a predetermined course. A typical
double lane change manoeuvre was performed over the Belgian paving.
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.2.3
2.20
Results
Simulation was performed with different spring and damper characteristics in order to
complete a sensitivity analysis. A severe double lane change manoeuvre, performed over
Belgian paving at a vehicle speed of 60 km/h, was chosen as representative of high speed
off-road driving on gravel roads and tracks. The speed of 60 km/h is close to the
maximum double lane change speed achievable with the particular vehicle on a level
paved surface.
As the simulation included ride comfort, stability and handling, interpretation of the
results are difficult and it is necessary to define certain performance criteria. For ride
comfort, the vertical acceleration at the vehicle body’s centre of gravity was filtered using
the BS 6841 Wb filter and the RMS value determined. Motion sickness dose values were
determined in a similar fashion using the motion sickness or Wf filter. When driving in
off-road conditions, body roll and pitch usually give the first indication that the vehicle
speed is excessive. Furthermore, it is more difficult to brace the human body against roll
and pitch motion than is the case for yaw or vertical motion. Stability and handling were
therefore evaluated using the RMS roll angle, RMS roll velocity and RMS pitch velocity
of the vehicle body.
Two suspension configurations were simulated namely linear springs (see Table 2.3 for
the stiffness and static deflection) as well as non-linear hydropneumatic springs (see
Table 2.4). In both cases the damper force ratios (damper force normalised to the baseline
damper force at any specific damper speed) were varied between 0.001 and 3. For both
linear and hydropneumatic springs, 9 different spring and 12 different damper
characteristics were simulated, giving a total of 108 simulation runs. The simulation
results are presented as contour plots (Figures 2.7 to 2.10) where the horizontal axis
represents the damper force ratio and the vertical axis the natural frequency (or stiffness)
of the suspension system. The contours represent the percentage improvement in the
respective values relative to that of the baseline suspension (damper force ratio of 1 and
natural frequency of 1.2 Hz). The left hand bottom corners of the graphs (low spring and
damper rates) have no contour lines since the suspension is so soft that the vehicle could
not perform the double lane change manoeuvre and rollover occurred. Figures 2.7 to 2.10
represent the results for the linear springs but the trends are very similar for the
hydropneumatic springs.
Table 2.3 - Calculated spring stiffness for linear spring
Natural Frequency
[Hz]
0.6
0.8
1.0
1.2
1.4
1.6
1.8
2.0
3.0
Required spring stiffness
[N/m]
33 034
60 276
97 486
146 671
210 800
294 324
404 095
551 123
3 985 988
Static deflection
[m]
0.668
0.366
0.226
0.151
0.105
0.075
0.055
0.040
0.0055
2.21
THE RIDE COMFORT VS. HANDLING COMPROMISE
Table 2.4 - Natural frequencies for hydropneumatic spring
Static Gas Volume
[Litre]
0.3
0.5
0.6
0.8
1.0
1.3
1.6
2.0
3.0
Natural Frequency
[Hz]
1.51
1.20
1.10
0.96
0.86
0.76
0.69
0.62
0.50
Stiffness in Static Position
[N/m]
255 100
146 280
120 550
89 200
70 780
54 050
43 710
34 830
23 100
Figure 2.7 indicates the relationship between natural frequency (spring stiffness) and
damper force ratio on the ride comfort for the linear spring configuration. A maximum
improvement of 55% is reached at a natural frequency of 1 Hz and damper force ratio of
0.2. The percentage improvement is calculated using Equation 2.3.
% improvement =
(baseline value − new value)
*100
baseline value
(2.3)
The trend indicates that further improvements in ride comfort may be possible at lower
natural frequencies, but that the vehicle becomes unstable and rolls due to the handling
manoeuvre. The best spring characteristic for ride comfort is therefore as low as can be
tolerated from a stability and handling perspective. Although a reduction in damper force
ratio improves ride comfort, a certain minimum damping level is required.
A maximum improvement of 51% in the motion sickness dose value (not indicated) was
achieved for the lowest natural frequency and highest damper force ratio. The motion
sickness dose value was however much less sensitive to the damping value than to the
natural frequency. It was expected that the motion sickness dose value should increase
with a reduction in suspension natural frequency, but apparently this is offset by the
improved isolation performance of the lower natural frequency suspension.
Figure 2.8 indicates that a maximum pitch velocity improvement of 11% is achievable at
a damper force ratio of 0.7 and natural frequency of 1.2 Hz. This is in close correlation to
the ride comfort optimum although the improvement is not very significant in magnitude
compared to the ride comfort improvement of 55%.
A maximum improvement of 77% in roll angle (Figure 2.9) is achieved at a suspension
natural frequency of 3 Hz. The roll angle improvement is insensitive to the damper force
ratio as can be expected. The optimal characteristics for roll velocity (improvement of
32%) are achieved at a suspension natural frequency of 2 Hz and damper force ratio of
1.8 (see Figure 2.10). Both roll angle and roll velocity are therefore reduced by higher
spring and damper characteristics although the trends indicate that there is little
improvement after the spring and damper characteristics have been doubled from the
base-line values. This is due to the increase in tyre deflection that occurs as the spring
stiffness is increased.
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.22
All the graphs shown are for the linear suspension configuration mainly due to the wider
range of suspension natural frequencies that could be indicated. All the tendencies are
however similar for the non-linear hydropneumatic suspension system.
Figure 2.7 - Improvement in weighted RMS vertical acceleration (ride comfort – linear
spring)
The expected conclusion is made that ride comfort requires opposite characteristics to
handling and stability. The suspension resulting in the best ride comfort, leads to rollover.
This fact gives a good motivation for the use of a semi-active spring-damper system to
improve both ride comfort and handling. The semi-active spring-damper system has to be
designed with natural frequencies of approximately 0.6 and 2 Hz respectively while
damper force ratios of 0.2-0.5 and 2.0 are required. It must be emphasised that these
characteristics are for the passive suspension case only and may change when research on
control strategies is continued. The results are also only valid as long as terrain inputs do
not result in contact with the bump-stops. When bump-stop contact occurs, stiffer
suspension may result in improved ride comfort. The semi-active hydropneumatic springdamper system can however adapt to these circumstances if a suitable control strategy is
employed. Valve dynamics and response times may also affect the results.
Although the simulation results indicate optimal values for the spring and damper
characteristics, these characteristics may not always be obtainable on a practical vehicle
suspension system because of certain physical constraints.
THE RIDE COMFORT VS. HANDLING COMPROMISE
Figure 2.8 - Improvement in pitch velocity (linear spring)
Figure 2.9 - Improvement in roll angle (linear spring)
2.23
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.24
Figure 2.10 - Improvement in roll velocity (linear spring)
The maximum rebound damper force is limited by the pressure difference across the
damper. When the pressure difference becomes so large that the pressure in the hydraulic
strut approaches zero, then cavitation will occur. The oil will boil and apart from the
physical damage that may result, the damper force will stay constant. The minimum
damper characteristic on the other hand is limited by the flow loss through the channels
and hydraulic valves.
The sprung mass natural frequency of a quarter car suspension system is calculated by
approximation according to equation 2.4.
⎛ ks ⋅ k t ⎞
⎟
⎜
1 ⎜⎝ k s + k t ⎟⎠
fn =
2π
M
(2.4)
=
Sprung mass natural frequency [Hz]
Where fn
=
Spring stiffness [N/m]
ks
=
Tyre stiffness [N/m]
kt
M
=
Sprung mass [kg]
For the vehicle under consideration, kt = 1 000 000 N/m and M = 2 250 kg. Table 2.3
indicates the required linear spring stiffness and static deflection for different natural
frequencies.
It is evident that in order to obtain a natural frequency of 3 Hz, the spring stiffness must
be four times higher than the tyre stiffness while the static deflection of the spring is only
5.5 mm. This implies that the tyre will deflect significantly, negating the effect of the stiff
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.25
spring. On the other hand, the static deflection of 0.668 m, required for a 0.6 Hz natural
frequency, is also unreasonable.
Similar limitations are applicable to the hydropneumatic spring except that the spring
stiffness used for calculation of the natural frequency is linearised through the static
suspension position. Table 2.4 indicates the natural frequencies calculated for different
static gas volumes. An accumulator of 2 or 3 litre capacity is bulky and it is doubtful
whether this will fit into the space envelope normally available on a vehicle.
2.2.4
Conclusions from Case Study 1
The results indicated that for best ride comfort, the damper force ratio should be between
20 and 50% of the baseline value and the natural frequency should be in the region of 0.6
Hz. For optimal stability and handling, the natural frequency should be around 2 Hz and
the damper force ratio double the baseline value. These characteristics will be used as the
starting point for determining the best “on” and “off” characteristics for a controlled, twostage semi-active hydropneumatic spring and damper system. It should however be noted
that these values may differ for other vehicle speeds, terrain inputs or handling
manoeuvres. The values might also be vehicle-specific.
2.3
Case Study 2: Land Rover Defender 110
A Land Rover Defender 110 vehicle was chosen for case study 2 as well as for the rest of
the research discussed in this thesis. This vehicle is ideal due to the following reasons:
i)
A used vehicle in good condition could be obtained at an affordable price.
ii)
The vehicle has a good reputation for its off-road capability.
iii)
It is a “low technology” vehicle that greatly simplifies the required modifications.
iv)
The vehicle has a ladder frame chassis that makes it possible to easily modify
mounting points for fitting a controllable suspension system.
v)
The design is square and boxy resulting in enough space in the wheel arches.
vi)
It is easy to mount different springs and dampers to the vehicle.
vii)
Considerable improvements in handling can be obtained.
viii) The vehicle has a high center of gravity that should highlight the improvements
offered by a controllable suspension system.
ix)
The vehicle is fitted with coil springs. The suspension is already located by links
and bars, i.e. spring and damper characteristics can be changed without changing
kinematics. The springs and dampers also perform no axle locating functions.
2.3.1
Vehicle model
In order to simulate the ride comfort and handling of the vehicle, a first order DADS
simulation model, based on a combination of measured and estimated parameters for a
Land Rover Defender 110 sports utility vehicle (see Figure 2.11), was developed. A
second, more detailed model was later developed and is discussed in paragraph 2.4.
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.26
Figure 2.11 - Land Rover Defender 110 vehicle
The DADS model has 81 degrees of freedom, but after adding joints, constraints and a
driver model, 14 unconstrained degrees of freedom remain. These consist of the vehicle
body displacements (lateral, longitudinal, vertical, roll, pitch and yaw), wheel rotations,
front axle vertical displacement and roll and rear axle vertical displacement and roll. Nonlinear spring, damper, bump stop and tyre characteristics are used. The vehicle is steered
over a predetermined course by a simple driver model which estimates the lateral
positional error based on the yaw angle of the vehicle body at the current time step and
the desired lateral position at a specified driver preview time. The driver model is
implemented using amplifiers, summers and input elements. The basic components of the
DADS model are summarized in Table 2.5.
2.3.2
Definition of “design space”
The upper limit of spring stiffness is limited by the tyre stiffness. If the spring stiffness
becomes too high, the tyre stiffness will become dominant, negating the effect of the stiff
spring. As tyre deflection increases, the lateral force capability will decrease as the tyre’s
lateral stiffness decreases. The lower limit of the spring stiffness is limited by two factors
namely motion sickness and restricted suspension movement (“rattle space”). If the spring
stiffness is so low that the suspension natural frequency is well below 1 Hz, the incidence
of motion sickness in the vehicle will increase. A softer spring will also require more
travel than a stiff spring over the same terrain roughness.
The upper limit for damper force in rebound is the cavitation of the oil when the pressure
in the damper drops to the vapour pressure of the oil. There is no physical limit to
compression damping, although the effect of the bulk modulus will increase and
mechanical buckling of the damper rod may arise. The lower limit for damping is
determined by the flow losses through the damper valves, valve block channels as well as
friction.
The coil springs on the baseline suspension were replaced with hydropneumatic springs
where the spring stiffness is determined by the gas volume in the static position. Static
gas volumes were varied between 0.01 litres and 3.0 litres. This gives a range of spring
stiffness from about 10 to 0.1 times that of the baseline coil spring stiffness. To simplify
the damper characteristics, the baseline damper force was scaled with a constant factor
that varied between 0.5 (i.e., softer than baseline) up to 3 (3 times higher than baseline).
Simulations were performed for 7 damper characteristics and 10 spring characteristics
within these ranges, giving a total of 70 simulation runs.
2.27
THE RIDE COMFORT VS. HANDLING COMPROMISE
Table 2.5 – Summary of the DADS simulation model
Model entities
Components
Quantity
Spherical-spherical joints (5)
Vehicle body
Wheels
Front axle
Rear axle
Ground (fixed in space)
Front hubs (left & right)
Anti-rollbars
Front wheels to front hubs
Front hubs to front axle
Rear wheels to rear axle
Body torsional stiffness
Anti-rollbar left and right
Axle locating and push-pull rods, steering links
2
4
1
1
1
2
2
2
2
2
1
2
5
Revolute-revolute joint (1)
Radius rod
1
Revolute-spherical joints (2)
A-arm rear
Panhard rod front
Steering control input
Forward speed
Non-linear dampers
Springs (choice of hydropneumatic and coil springs)
Bump stops
Generic tyres
Body torsional stiffness spring
Anti-rollbar stiffness
Amplifiers
Summers
Inputs
Steering angle limiter
Output torques left and right
1
1
1
1
4
4
4
4
1
1
2
2
2
1
2
Vehicle forward speed
1
Rigid bodies
(13)
Revolute joints
(9)
Constraints
(2)
Force elements
(18)
Control elements
(9)
Initial conditions
(1)
2.3.3
Simulation results
The DADS model was used to predict ride comfort and handling of the vehicle with
different combinations of spring and damper characteristics. Simulation results were used
to determine first order indications of the “best” soft and hard characteristics for both
spring and damper.
2.3.3.1 Ride comfort
Ride comfort was simulated over a typical off-road terrain (Belgian paving block course)
at a vehicle speed of 60 km/h. Ride comfort was evaluated using the vertical acceleration
at the driver position (right front) as well as the left rear passenger position. The vertical
acceleration was weighted using the British Standard BS 6841 Wb weighting filter and
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.28
calculating a weighted root mean square (RMS) value. A three-dimensional plot of
weighted RMS acceleration vs. spring static gas volume and damper scale factor, for the
driver’s seat position, is indicated in Figure 2.12.
The lowest acceleration levels (best ride comfort) are obtained with low damping (damper
scale factor of 0.8) and soft springs (static gas volume > 0.5 litres). Motion sickness
values do however increase with very soft springs (not shown).
2.3.3.2 Handling
Handling was simulated by performing a severe double lane change manoeuvre at a speed
of 60 km/h for the same values of spring and damper characteristics used for ride comfort
analysis. Maximum body roll angle was used as the evaluation parameter of handling and
stability. Figure 2.13 indicates the results of the handling simulations. The smallest body
roll angle is achieved with the stiffest spring (static gas volume of 0.01 litre) while the
roll angle is insensitive to the damper scale factor as could be expected. The “best”
suspension is therefore given by the highest possible spring stiffness. The areas where
there are no data points on the graph are where the vehicle could not complete the lane
change without rolling over.
2.3.3.3 Combined ride comfort and handling
The investigation was further extended by looking at a scenario where ride comfort and
handling were simultaneously required. For this analysis, a double lane change was
performed over the Belgian paving. The result is indicated in Figure 2.14. The infeasible
area where the vehicle rolls over is now significantly enlarged. The suspension design is
forced towards higher spring stiffness to keep the vehicle safe, at the expense of ride
comfort.
2.3.4 Conclusion from Case Study 2
It is concluded that for best ride comfort, a soft suspension is needed and for best
handling a stiff suspension is needed. This is in line with general design rules and was the
motivation for initializing this research project. The simulation results do however
indicate that for the hard suspension setting, a static gas volume of 0.1 litre and damping
scale factor of between 2 and 3 is suitable and for the soft suspension setting, a gas
volume of greater than 0.5 litre and a damping scale factor of 0.8 will be suitable first
order values for the design. The high damper characteristic used in the design of the
suspension system will therefore be between 2 and 3 times the baseline values, while the
low damping should be less than 0.8 times the baseline value.
This also confirms the results obtained for case study 1.
THE RIDE COMFORT VS. HANDLING COMPROMISE
Figure 2.12 - Results of ride comfort analysis
Figure 2.13 – Results of handling analysis
2.29
2.30
THE RIDE COMFORT VS. HANDLING COMPROMISE
Roll angle [°]
25
20
15
10
5
3
0
3
2
2.5
2
1.5
Damper scale factor
1
1 gas
volume
scale
Gas
volume
[l] fact
0.5
0
Figure 2.14 – Combined ride comfort and handling
2.3.5 Follow-up work by Uys, Els & Thoresson
In case study 2, simulations were performed for 7 damper characteristics and 10 spring
characteristics within specified ranges, giving a total of 70 simulation runs. Ride comfort
simulation was only performed for one terrain profile at one speed. Handling was also
simulated only for one handling manoeuvre (the double lane change) at one vehicle
speed. This process was performed manually. The next logical step was to investigate the
applicability of mathematical optimization techniques to the problem in an attempt to
decrease the number of required simulation runs. This should enable the simulation of
more terrain profiles at various speeds, as well as a more in-depth look at handling.
Els and Uys (2003) optimise the handling of a vehicle, equipped with a hydropneumatic
spring-damper suspension system, in conjunction with ride comfort. This is seen as a
challenge due to the fact that the suspension characteristics determining ride comfort and
handling respectively tend to oppose one another. The complexity and non-linearity of the
dynamics of suspension systems impose further difficulty. Furthermore the suspension
characteristics imply a large number of variables. The Dynamic-Q gradient-based
optimisation method is used in conjunction with the dynamics simulation code DADS.
Dynamic-Q is a robust and reliable algorithm particularly suitable for solving engineering
optimisation problems. It applies an existing dynamic trajectory optimisation algorithm to
successive spherical quadratic approximate sub-problems and can be used when
analytical functions are not available and only discrete function values can be obtained
via numerical simulation of engineering processes.
2.31
THE RIDE COMFORT VS. HANDLING COMPROMISE
22
20
18
16
Roll angle [°]
Ro 14
ll
an
12
gle
10
8
6
4
2
0
0.5
1
1.5
2
2.5
3
Gas
volume
[l]
gasvolum
escalefactor
0.5
1
1.5
2.5
2
damper scalefactor
3
Damper scale factor
Figure 2.15 - Path followed by Dynamic-Q
The purpose of the investigation by Els and Uys was to determine whether the DynamicQ method is suitable for optimising the design of a vehicle suspension system modelled in
DADS. Optimisation of the spring and damper characteristic was performed to obtain the
best characteristics for handling on a smooth road, best characteristics for ride comfort
while driving in a straight line and best characteristics for ride comfort and handling
combined (performing a handling manoeuvre on a rough road).
Thoresson (2003) builds on the work by Els and Uys (2003) and performs mathematical
optimisation on a Land Rover Defender 110 vehicle looking at ride comfort and handling
separately (Els et. al., 2003). His results are summarized in Table 2.6.
Table 2.6 – Summary of results by Thoresson (2003)
Two design
variables
Four design
variables
Seven design
variables
Handling
Damper scale factor
Spring static
gas volume [l]
1.35 to 1.99
0.05 (lower
limit)
1.5 to 3 front
0.1 (lower limit)
1.44 to 3 rear
front and rear
Low speed damping
0.1 (lower limit)
must be high
Ride Comfort
Damper scale factor Spring static gas
volume [l]
0.11 to 0.37
1.44 to 3.0
0.18 to 0.48 front
0.1 to 0.35 rear
Low speed damping
must be high
1.32 to 2.08 front
1.6 to 2.18 rear
1.07 to 2.18
For the two design variable case, the spring static gas volume and damper scale factor
were used as design variables. These factors were kept the same for both the front and
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.32
rear suspensions. In the four design variable case, the scale factors for the front and rear
suspension was allowed to differ. For the seven design variable case, the gas volume and
damper characteristics were kept the same front and rear, but the damper was now
approximated by a piecewise linear spline, altering low and high speed damping
separately.
For handling, all the cases required the stiffest spring allowed and damping that is
between 35% and 300% higher than the baseline damping. Ride comfort required soft
springs and damping that is lower than the baseline damping.
2.4
Validated vehicle model
The encouraging results obtained from case study 2 justified the development of an
improved vehicle model that could be used to predict absolute values and not only trends
as was the case with the first model. The fidelity of this model had to be good enough to
accurately predict both ride comfort and handling. This model had to be combined with a
model of the controllable suspension system and control system later in the project.
2.4.1
Geometric parameters
The majority of geometric parameters were obtained by physical measurement on a
vehicle, although some critical measurements were obtained from available drawings.
2.4.2
Mass properties
Mass properties were obtained from physical measurements on a vehicle. The
determination of the centre of gravity position, as well as estimation of the roll, pitch and
yaw mass moments of inertia are described by Uys et. al. (2005).
2.4.3
Spring and damper characteristics
Spring, damper and bump-stop characteristics were obtained by removing the
components from the test vehicle and determining force-displacement and force-velocity
relationships respectively using Schenck Hydropulse test equipment.
2.4.4
Tyre characteristics
Tyre side-force vs. slip angle characteristics were obtained from measurements using a
two-wheeled tyre tester towed behind a vehicle. The measured data was converted to the
coefficients required for the MSC ADAMS Pacjeka ’89 tyre model. The tyres side-force
vs. slip angle characteristics are indicated in Figure 2.16.
2.4.5
ADAMS full vehicle model
i) Front suspension
Figure 2.17 indicates the layout of the front suspension system. The rigid axle is located
longitudinally by leading arms connected to the vehicle body with rubber bushes. The
stiffness of these bushes was measured and included in the ADAMS model. Lateral
location of the axle is via a Panhard rod. The baseline vehicle is fitted with coil springs,
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.33
translational dampers concentric with the coil springs and rubber bump stops. A steering
angle driver is applied directly to the kingpin with a steering link connecting the left and
right wheels. All other steering geometry is ignored in the model. The connections
between the different components are indicated in Figure 2.18. To take the torsional
stiffness of the ladder chassis into account, the vehicle body is modelled as two bodies
connected to each other with a revolute joint along the roll axis and a torsional spring.
Figure 2.16 - Tyre side-force vs. slip angle characteristic
ii) Rear suspension
The rear suspension consists of a rigid axle with trailing arms, an A-arm, coil springs,
translational dampers mounted at an angle outside coil springs and rubber bump stops.
The basic layout is indicated in Figure 2.19. An anti-rollbar is fitted to the rear
suspension. The stiffness of the trailing arm rubber bushes is included in the ADAMS
model. The schematic layout of the rear suspension is indicated in Figure 2.20.
2.4.6
Baseline vehicle tests
A Land Rover Defender 110 SUV was obtained locally for testing purposes. The aim of
the baseline vehicle tests was to validate the ADAMS model of the vehicle. Tests were
performed at the Gerotek Vehicle Test Facility West of Pretoria.
THE RIDE COMFORT VS. HANDLING COMPROMISE
Figure 2.17 – Front suspension layout
Figure 2.18 – Front suspension schematic
2.34
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.35
2.4.6.1 Instrumentation
The instrumentation used for the baseline tests, as well as measurement positions, is
indicated in Table 2.7.
Table 2.7 – Instrumentation used for baseline vehicle tests
No
1
2
3
4
5
6
Parameter
Vehicle speed
Relative displacement
Relative displacement
Relative displacement
Relative displacement
Roll velocity
7
Yaw velocity
8
Relative displacement
9
10
11
12
13
Acceleration
Acceleration
Acceleration
Acceleration
Pitch velocity
14
15
16
Kingpin steer angle
Wheel speed
Driveshaft speed
Position
Roof
Left front suspension
Right front suspension
Left rear suspension
Right rear suspension
Vehicle body between
front seats
Vehicle body between
front seats (close to cg)
Steering arm between axle
and body
Left front lateral
Right rear vertical
Left rear lateral
Left rear vertical
Vehicle body between
front seats
Kingpin
Left rear wheel
Gearbox output rear
Equipment
VBOX GPS
Penny&Giles rope displacement transducer
Penny&Giles rope displacement transducer
DWT rope displacement transducer
DWT rope displacement transducer
Solid state gyro
Solid state gyro
Penny&Giles rope displacement transducer
Solid state accelerometer ±4g range
Solid state accelerometer ±4g range
Solid state accelerometer ±4g range
Solid state accelerometer ±4g range
Solid state gyro
Potensiometer
Turck Banner optical speed sensor
Turck Banner optical speed sensor
2.4.6.2 Tests
The vehicle was evaluated for ride comfort over repeatable test tracks of various
roughnesses at known, repeatable and representative speeds. Weighted root mean square
vertical accelerations were used to quantify ride comfort. Figure 2.21 indicates the vehicle
on the Belgian paving track during testing. Tests also included single discrete obstacles
(locally known as an “APG” bump) as indicated in Figure 2.22. Vehicle handling was
evaluated using a constant radius test (Figure 2.23) as well as a severe double lane change
manoeuvre (Figure 2.24).
Additional tests over typical off-road terrain were performed on the Gerotek rough track
(Figures 2.25 and 2.26) where a combination of ride comfort and handling is required.
The rough track consists of natural terrain features embedded in concrete to give
repeatability.
Test procedures and terrains were chosen to ensure repeatability. Vehicle speed was kept
constant by driving the diesel engine against its governor. This is important, as the
baseline test results will be used later to quantify the improvements offered by the
controllable suspension system.
THE RIDE COMFORT VS. HANDLING COMPROMISE
Figure 2.19 – Rear suspension layout
Figure 2.20 – Rear suspension schematic
2.36
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.37
Figure 2.21 - Belgian paving
2.4.7
Correlation between ADAMS model and test results
To validate the ADAMS model, simulation results were compared to measured results for
two different tests namely the APG bump, and the ISO 3888 double lane change test.
2.4.7.1 Transient response (APG track)
The APG track was chosen to validate the vertical and pitch dynamics of the vehicle. The
road input profile is easily measured and included in a simulation model. Figure 2.27
indicates the correlation obtained between the measured and simulated results.
Correlation is indicated for pitch velocity, spring displacement right front (rf), spring
displacement rear left (rl), steering displacement as well as front and rear vertical
accelerations. Correlation for vertical accelerations is especially good which is important
because vertical acceleration is a direct measure of ride comfort. The model is thus
considered validated for ride comfort simulation.
2.4.7.2 Handling (ISO 3888 Double lane change)
Figure 2.28 indicates the correlation achieved for a double lane manoeuvre performed at
65 km/h. The speed for the baseline vehicle tests varied between 61 and 65 km/h. The
steering input, as measured during baseline testing, was used to drive the vehicle and not
the driver model. The graphs therefore represent the dynamic reaction of the vehicle to
the same input conditions as during testing. Correlation is very good for all the measured
parameters. The model is thus considered validated for handling simulation.
THE RIDE COMFORT VS. HANDLING COMPROMISE
Figure 2.22 - “APG” Bump
Figure 2.23 - Constant radius test
2.38
THE RIDE COMFORT VS. HANDLING COMPROMISE
Figure 2.24 - Severe double lane change manoeuvre
Figure 2.25 - Rough track
2.39
THE RIDE COMFORT VS. HANDLING COMPROMISE
Figure 2.26 - Rough track
2.40
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.41
Figure 2.27 - Model validation results for passing over 100 mm APG bump at 25 km/h
2.4.8
Simulation results
The validated ADAMS model was modified by replacing the coil springs with
hydropneumatic springs. As before the static gas volume was varied between 0.1 and 1
litre while the damper scale factor was varied between 0.5 and 3. The effect of spring and
damper characteristics on ride comfort over the Belgian paving is indicated in Figure
2.29. The conclusion is made that for best ride comfort, the damping scale factor must be
as low as possible and the static gas volume as large as possible although the
improvement is negligible for gas volumes higher than 0.5 litre. For handling, a double
lane change manoeuvre was again performed. In this case both the body roll angle and
body roll velocity was used as a measure of handling and stability. Figure 2.30 indicates
that the maximum body roll angle at the first valley, and maximum roll velocity at the
first peak was used.
Figure 2.31 indicates the maximum roll angle as a function of static gas volume and
damper scale factor. The lowest maximum roll angle (best handling) is obtained with a
static gas volume of 0.1 litre and a damper scale factor of 3. The maximum roll velocity
(Figure 2.32) is more sensitive to the damper scale factor. This result was expected, as
damping force is velocity dependant while spring force is displacement dependant. The
results however point to a damper scale factor of 3.
THE RIDE COMFORT VS. HANDLING COMPROMISE
2.42
All these results indicate that the optimum characteristics for ride comfort and handling
are at opposite corners of the design space. For good ride comfort, low damping and a
soft spring is required. A damper scale factor of less than 0.5 and a static gas volume of
0.5 litre or more will give the best ride comfort. For best handling, a static gas volume of
0.1 litre and damper scale factor of 3 is required.
Figure 2.28 – Model validation results for a double lane change manoeuvre at 65km/h
2.5
Conclusion
The following is concluded based on the evidence presented in this chapter:
a)
b)
c)
A passive suspension system is a compromise between ride comfort and handling
as the respective requirements for ride comfort and handling are at opposite ends
of the design space.
To eliminate the ride comfort vs. handling compromise, two discrete spring
characteristics are required namely:
• A stiff spring for best handling (0.1 litre static gas volume for
hydropneumatic spring in the case of the test vehicle)
• A soft spring for best ride comfort (>0.5 litre static gas volume for
hydropneumatic spring in the case of the test vehicle).
To eliminate the ride vs. handling compromise, two discrete damper
characteristics are required namely:
• High damping for best handling (greater than double baseline damping
value)
• Low damping for best ride comfort (less than ½ the baseline damping
value).
THE RIDE COMFORT VS. HANDLING COMPROMISE
d)
e)
2.43
The capability to switch between the two spring and the two damper
characteristics is required.
A control strategy that can switch between “ride comfort” mode and “handling”
mode in a safe and predictable way is of critical importance.
Figure 2.29 – Ride comfort vs. gas volume and damping
Figure 2.30 - Definition of handling objective function
THE RIDE COMFORT VS. HANDLING COMPROMISE
Figure 2.31 – Roll angle vs. gas volume and damping
Figure 2.32 – Roll velocity vs. gas volume and damping
2.44
Chapter
3
POSSIBLE SOLUTIONS TO THE RIDE COMFORT
VS. HANDLING COMPROMISE
Possible concepts for the improvement or elimination of the ride comfort vs. handling
compromise are investigated in this chapter. Current literature is reviewed, firstly to
determine possible hardware concepts for controllable suspension systems and secondly
to obtain a global view of the technical requirements involved in the development and
implementation of control methodologies. Fully active suspension systems are not
considered mainly due to their large power requirements, especially when applied to
heavy off-road vehicles. For this reason, the literature review is therefore not concerned
with fully active suspension systems in particular, but instead focuses on semi-active and
adaptive systems where spring and damper characteristics can be changed either
continuously or switched between different discrete characteristics. Some active
suspension concepts and control methods are however discussed, as many of these might
be adapted to semi-active suspension systems. In some cases it might be possible to
control a semi-active damper with the same strategy as a fully active suspension system,
but it will only dissipate energy as no energy can be supplied. The damper will therefore
be switched to the low damping state when energy supply is demanded by the control
system. Active suspension systems dissipate energy for a large amount of the time in any
case and semi-active dampers can therefore often approach the results obtainable with
fully active systems.
After briefly discussing published literature on advanced suspension systems, this chapter
deals more thoroughly with the subjects of semi-active dampers, semi-active springs and
active suspension systems, followed by control techniques and algorithms. The chapter
closes with a proposed controllable suspension solution to the ride comfort vs. handling
compromise.
3.1
Published literature surveys on controllable suspension systems
Six published literature surveys concerning advanced suspension systems were found.
Although these surveys do not provide sufficient detail on each topic to be really useful
for the purposes of the current study, they provide a valuable source of references and a
general overview on the specific subject.
Tomizuka and Hedrick (1995) discuss advanced control methods for automotive
applications in general and include a paragraph on suspension systems. Various control
methods are mentioned for fully active systems as well as semi-active dampers. No
mention is made of the existence or control of controllable spring systems.
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.2
Sharp and Crolla (1987) discuss suspension system design in general and include
aspects such as road surfaces, tyres, vehicle models and performance criteria. Passive,
active, semi-active and slow active suspension systems are also included in the survey.
Mention is made of slow active (3 Hz bandwidth) controllable pneumatic and
hydropneumatic systems.
Active suspensions for ground transport vehicles are reviewed by Hedrick and Wormley
(1975). The article does not include semi-active suspension systems and no mention is
made of semi-active or variable springs.
The application of neural networks and fuzzy logic to vehicle systems is reviewed by
Ghazi Zadeh, Fahim and El-Gindy (1997). An introduction to neural networks and
fuzzy logic is given. The techniques have been applied to active and semi-active
suspension systems by various authors.
Elbeheiry et. al. (1995b) give a classified bibliography of advanced ground vehicle
suspension systems. A reference list of 71 papers concerned with semi-active suspensions
and 58 papers concerned with adaptive, actively damped and load-levelling suspensions is
given but not discussed.
Applications of optimal control techniques to the design of active suspension systems are
surveyed by Hrovat (1997). The main emphasis of the survey is on Linear Quadratic
Optimal (LQO) control and active suspension systems, but related subjects such as semiactive suspensions and related control topics are also discussed. Some 256 papers are
included in the list of references.
3.2
Controllable suspension system hardware
Vehicle suspension system configurations vary over a wide spectrum. The most important
variations on the theme will now be discussed.
3.2.1
Semi-active dampers
Semi-active dampers vary from two-state (on/off) to continuously variable. Both linear
and non-linear damper characteristics are considered. The majority of semi-active
dampers are based on either magneto-rheological (MR) fluids or hydraulic dampers with
controllable valves.
3.2.1.1 Magneto-Rheological (MR) fluids
A Magneto-rheological (MR) fluid is used as the damping medium inside a hydraulic
damper and replaces the conventional damper oil. A MR fluid is a dense suspension of
micrometer-sized magnetisable particles in a carrier fluid that solidify to a pasty
consistency in the presence of a magnetic field (Lord Corporation, 2005; Ouellette,
2005). When the magnetic field is removed the fluid returns to its liquid state. Altering
the strength of the applied magnetic field will proportionally control the consistency or
yield strength of the fluid and therefore the pressure required to force the fluid through a
magnetized orifice. MR fluids offer a very fast response time (order of 10 milliseconds)
and have been commercially applied in continuously variable semi-active dampers (see
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.3
Lord Corporation 2005 for a description of MagneRide as fitted to some General
Motors products).
Researchers at the Advanced Vehicle Dynamics Laboratory at Virginia Polytechnic
Institute and State University used controllable MR dampers to control the roll dynamics
of a Ford Expedition SUV. Results of vehicle tests indicated that a velocity based
skyhook control, augmented with steering wheel feedback, outperformed the passive
stock dampers (Simon, 2001).
3.2.1.2 Hydraulic bypass system
Semi-active dampers based on the by-pass principle use a hydraulic valve (mostly
electrically operated) in parallel with a conventional damper orifice and valve assembly.
A two-stage (open-closed) valve is indicated in Figure 3.1. If the bypass valve is closed,
all the flow goes through the conventional damper orifice and valve assembly, giving
high damping or the “on” characteristic. If the bypass valve is open, most of the flow will
pass through the bypass valve due to the lower flow resistance. This results in the low
damping or “off” characteristic. During valve switching some transient response will
result between the “on” and “off” characteristics. The bypass valve can have several
discrete stages, or it can be a servo valve giving continuously variable damping
characteristics.
Figure 3.1 - Hydraulic two-state semi-active damper with bypass valve
The choice of valve is based on the pressure drop and flow rate characteristic, as well as
the required response time.
Examples of two-state semi-active dampers, using the bypass valve principle, are
discussed by Nell (1993) and Nell and Steyn (1994). A picture of their first prototype can
be seen in Figure 3.2 with the bypass valve indicated. This damper was designed for a
maximum flow rate of 1000 l/min, a static wheel load of 3 ton and a response time in the
region of 50 milliseconds. The largest semi-active damper for a wheeled vehicle,
developed by Els and Holman (1999), is indicated in Figure 3.3. This damper has a
maximum damping torque of 150 kN.m and was used on a 46-ton 6x6 vehicle. These
dampers are all applied to off-road military vehicles.
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.4
Solenoid
Bypass valve
Damper
Figure 3.2 – Semi-active damper developed by Nell (1993)
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.5
Damper
Accelerometer
Bypass valve block
Solenoid valve
Hydraulic
bump stop
Trailing arm
suspension
Figure 3.3 - Semi-active rotary damper developed by Els and Holman (1999)
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.6
3.2.2
Semi-active springs
Semi-active springs are based on either air or hydropneumatic springs that are mostly
non-linear due to their operating principles. Hydropneumatic and air springs frequently
incorporate some kind of slow active ride height correcting device. Cases also exist where
an air spring is combined with a normal passive coil spring.
3.2.2.1 Air springs
Decker, Schramm and Kallenbach (1988) describe a prototype adjustable air spring
developed by BOSCH, where the spring characteristic can be changed between several
values by fast (25 milliseconds) switching of different air volumes. The adjustable spring
is used in conjunction with a fast (4 milliseconds) semi-active damper. Very limited
simulation results are included. A closed loop control strategy, of which no details are
provided, is used to switch both the spring and damper during simulation. An
improvement potential of 36% in ride comfort is obtainable from simulation results. The
skyhook control strategy as proposed by Karnopp is also investigated although no further
details are presented. No experimental work concerning evaluation of control strategies is
presented.
An industrialised version of a semi-active suspension developed by Armstrong is
discussed by Hine and Pearce (1988). A two or three state adjustable damper is
combined with an air or oleo-pneumatic spring that is said to offer both height and spring
rate control. It is not clear how the spring rate is changed but it appears as if the spring
rate changes because of the ride height adjustment. The oleo-pneumatic damper can be
pressurised to a maximum pressure of 200 bar (20 MPa) supplied by an oil pump. The
unit is fitted with an external reservoir. The control strategy can be separated into five
components namely ride, handling, acceleration, deceleration (dive), ride frequency
control and vehicle levelling (if required). The system is commercially applied to the
1986 GM Corvette (5.7 litre) and Ford Granada 2.8 Ghia.
Pollard (1983) describes a fully active air actuator fitted to a railway couch. Where most
conventional air suspensions have an auxiliary reservoir to provide the desirable spring
and damping characteristics, the air pump actuator replaces the fixed volume reservoir
with one of continuously variable volume. An electric motor is attached to the diaphragm
via a nut and a lead screw. Operating the lead screw can change the volume. A prototype
has been tested with good success and power consumption is found to be low.
A performance air suspension developed by Bridgestone/Firestone is described by
Alexander (2004a). The system is cockpit adjustable by the driver. Ride height can be
lowered to improve handling or increased to improve ground clearance. Spring rate may
also be reduced to improve isolation or increased for handling. The spring rate can be
changed either with, or independent of height. Roll stiffness distribution between front
and rear can seemingly also be altered.
The suspension system used on the 1986 model Toyota Soarer is described by Hirose et.
al. (1988). This system changes both spring and damper characteristics using direct
current electric motors. The air spring uses main and supplementary air chambers
connected by a disc valve to change the gas volume and therefore the spring
characteristic. Height control is also implemented for which air pressure is supplied by a
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.7
compressor. System response time is 70 milliseconds. The spring and damper rates are
changed simultaneously by a single electric motor. The four struts on the vehicle are also
controlled together. Vehicle speed, throttle position, steering angle, height and other
factors related to vehicle attitude are used to determine the suspension state.
An Electronic Controlled Suspension (ECS) as fitted to the 1984 Mitsubishi Galant is
discussed by Mizuguchi et. al. (1984). A two-stage spring is constructed using an air
spring in parallel with a conventional metal coil spring. The air spring consists of two
chambers connected by a valve. The valve is closed to activate the stiff spring rate.
Vehicle speed, steering wheel speed, sprung mass acceleration, throttle speed and
suspension stroke are used as control parameters. A methodology to determine the spring
and damper rate for the two-state suspension systems is described. The suspension is
either set to “off” (soft spring and soft damper) or “on” (hard spring and damper). The
normal suspension state is soft for good ride comfort but is switched to hard for high
vehicle speeds or during handling manoeuvres.
Karnopp and Margolis (1984) discuss the effects of a change in spring and damper rates
on the transfer function of a single degree of freedom suspension system. It is said that
changing the damping alone is not a very good way of stiffening or softening a
suspension system. A system with two air volumes separated by control valves is
proposed that enables both the spring and damper rates to be adjusted. Air can also be
slowly added to or subtracted from the air volume to enable ride height adjustment. The
proposed system can be adaptively controlled using brake and steering inputs as well as
angular acceleration. Manual overrides can be included to suit personal preference.
Wallentowitz and Holdman (1997) give a frequency domain analysis of the effect of
spring and damper constants on the transfer function of the suspension. It is concluded
that two spring stages are sufficient to overcome the compromise associated with passive
systems. The two-stage spring can be realised in hardware by using two air springs
connected by a pipe and orifice arrangement. The orifice is designed so that the second air
spring is effectively closed off at suspension frequencies higher than 5 Hz. A valve in
series with the orifice can be closed to achieve a high spring rate during handling
manoeuvres. No hardware seems to be available. The study is theoretical only and
includes a suggestion for a possible control strategy based on the frequency response of a
quarter car system.
3.2.2.2 Hydropneumatic springs
Citroën has been applying hydropneumatic suspension systems to their passenger cars for
many years. Nastasić and Jahn (2005) describe the suspension systems fitted to different
models in detail. On the XM model, both the front and rear suspensions consist of three
spheres (bladder accumulators) and four dampers. The system can be switched to a low
spring and low damping state (3 spheres and 4 dampers) or high spring and high damping
rate (2 spheres and 2 dampers). The system reacts in less than 50 milliseconds and is
computer controlled. Inputs to the controller include the angle and angular speed of the
steering wheel, speed of movement of the accelerator pedal, braking effort, rotation of the
front anti-rollbar and vehicle speed. A switch on the centre console enables the driver to
permanently select the high spring and damper state. Another system fitted to Citroën’s
Activa 2 research prototype car is described by Birch, Yamaguchi and Demmler (1990).
The system is an upgrade of that used for the XM and ads an active anti-roll system that
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.8
can double the roll stiffness almost instantly to counter body roll. Roll control is
implemented by adding a fourth sphere and a roll control strategy. Roll is reduced when
this fourth sphere is disconnected from the system. The system absorbs less than 0.375
kW through a fast corner and double that amount for violent emergency avoidance action.
One of the oldest references found for a switchable hydropneumatic spring system is that
described by Eberle and Steele (1975). Their system is indicated in Figure 3.4 and was
intended as an operator controlled system. The operator could choose the spring constant
to suit the vehicle speed and the type of terrain by opening or closing two valves. Four
discrete characteristics are possible namely rigid, firm, medium and soft depending on
valves 1 and 2. The placing of the damping units in the branches to the accumulators
permits matching of the damping to the selected spring constant.
Figure 3.4 – Operator controlled variable spring as proposed by Eberle and Steele
(1975)
3.2.2.3 Other semi-active spring concepts
Semi-active springs may be realized using other methods e.g.:
• Metal springs in combination with air or hydropneumatic springs
• Accumulators with adjustable volume e.g. lead screw connected to an electric
motor
• Compressible fluid suspension systems
• Piezo-electric actuators
• Smart materials
These ideas were not given further consideration for the purposes of the present study.
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.9
3.2.3
Active suspension systems
Active suspension systems have been applied to off-road vehicles with limited success.
Apart from the high cost, power requirements and bandwidth restrictions seem to be the
major obstacles. Both electric and hydraulic actuators have been used.
3.2.3.1 Electric actuators
The design of an electromagnetic linear actuator for active suspension application is
described by Weeks et. al. (1999) and Buckner et. al. (2000). The actuator consists of an
electric motor driving a rack-and-pinion. The actuator was designed to be used in parallel
with an air spring that carries the static wheel load. The actuator was designed for retrofit
to a high mobility multi-purpose wheeled vehicle (HMMWV). It produces a maximum
force of 8896 N, a stroke of 127 mm and a maximum velocity if 1 m/s. The performance
of the actuator was evaluated on a quarter-car test rig and found to meet and even exceed
the design specifications. Very reasonable peak power requirements of about 12 kW were
recorded during some rig tests.
Bose Corporation developed a prototype linear magnetic actuator that was installed at
each wheel of a vehicle in a modified McPherson strut configuration (Anon, 2005b). A
belt-driven alternator and a 12 Volt battery power the system. It is said to improve both
comfort and handling and eliminates the need for anti-rollbars. No quantification of
performance improvements or power requirements is given.
3.2.3.2 Hydraulic actuators
Lotus was one of the pioneers of hydraulic fully-active suspension systems. A concise
summary of the development of the Lotus active suspension system is given by Wright
(2001). The technology was initially developed for use in Formula 1 and quickly banned.
It was used later in various prototype applications to passenger cars as well as military
vehicles (both wheeled and tracked).
Scientists in the Tactical Vehicle Section of the Canadian Army built an active
suspension prototype based on the Iltis truck (Anon, 2005a). The test vehicle has been in
operation since 1995 at the Royal Military College at Kingston, Ontario for it’s training
and testing programmes. The system uses Moog-Lotus servo-controlled actuators with a
20 Hz system response. Power requirements are low (5-10 HP) over moderate crosscountry terrain. Vertical acceleration of the driver is reduced by 10% over discrete bumps
while slalom speed is increased by 20%. The driver is said to feel increased control with
reduced steering effort while rollover is less likely to occur.
Researchers at the University of California (Berkeley) have been involved in research on
the control of fully active, hydraulic suspension systems for many years (Hedrick and
Wormley, 1975 and Hedrick et.al. 1994).
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.10
3.3
Control techniques and algorithms
It seems that the possibilities concerning control strategies are limitless although the
majority of papers use the “Skyhook” strategy proposed by Karnopp et. al. (1974) that
was derived using Linear Quadratic Optimal (LQO) control theory. Other methods
include neural networks, fuzzy logic, H∞ and PD control. Preview control is often
considered.
Control strategies can broadly be classified in two main categories namely input driven
and reaction driven strategies. The control parameters for input driven strategies
usually consist of parameters such as vehicle speed, steering angle and brake pressure.
These strategies therefore react on inputs from the driver or vehicle before the dynamics
of the vehicle changes. Reaction driven strategies react to the vehicle’s dynamic
reaction due to terrain roughness or driver input. Take as an example a vehicle driving in
a straight line, when the driver gives a sudden step input on the steering wheel in order to
avoid an accident. An input driven strategy might use steering angle as input and switch
the dampers to the high damping state as soon as the steering angle or steering velocity
exceeds a predetermined level, while a reaction driven strategy might use lateral
acceleration or yaw rate as input and the dampers will only be switched to the high
damping state after the tyres developed enough side force so that the vehicle will turn. In
this instance it can be seen that the input driven strategy will respond earlier.
Further discrimination must also be made between the terms adaptive, semi-active and
active suspension systems. These terms, as they are used in this study, are defined in
Table 1.1. Adaptive control on the other hand is used for systems where the controller
gains are changed (adapted) according to certain measured parameters i.e. wheel
acceleration as a measure of terrain roughness.
3.3.1 Combination of input and reaction driven strategies
Hine and Pearce (1988) discuss a strategy for obtaining optimum ride comfort and
handling control. The control strategy is separated into six components namely ride,
handling, acceleration, deceleration (dive) as well as ride frequency control and vehicle
levelling (if required). Ride control is initiated by the relative wheel to body
displacement in combination with the vehicle speed. For any particular speed,
displacement limits are established, outside of which the damper is switched to a higher
level. This enables maximum use of available suspension working space while keeping
the damper in the soft state for most of the time. The steering sensor together with the
speed sensor is used to determine when dampers should be switched to a higher state to
improve handling. Dampers are also switched to a higher state during acceleration and
deceleration caused by throttle and brake applications. Levelling is effected by
measurement of relative suspension displacement and compensates for mass and
aerodynamic load changes. It is said that significant improvements in ride comfort have
been achieved while handling is also improved. The system is commercialised and put
into production on the GM Corvette (1986) and Ford Granada 2.8 Ghia.
The hydractive suspension introduced by Citroën in its XM passenger car, and featured in
various other Citroën models, is described by Nastasić and Jahn (2005). The angle and
angular rate of the steering wheel are used together with the car’s speed and the
suspension is switched to firm whenever certain threshold values are exceeded to enable
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.11
handling control. The speed of movement of the throttle, as well as braking effort is
measured and the suspension switched to the firm state when thresholds are exceeded to
enable acceleration and braking control. Roll and yaw control is achieved by measuring
the rotation angle of the front anti-rollbar. Adoption of this control strategy ensures that
the system always works in advance of the dynamic reaction of the car (i.e. input driven
control). This anticipation is said to be of particular advantage during fast driving on
winding roads where it reduces body movement and greatly enhances road holding and
handling, providing the driver with a unique sensation of control. The system is taken one
step further in the Activa 2 concept car (Birch et. al., 1990) by the introduction of an
additional roll control program.
Mizuguchi et. al. (1984) discuss the control system fitted to the Mitsubishi Galant.
Control inputs include steering wheel speed, lateral, longitudinal and vertical
acceleration, vehicle speed and suspension stroke. Test results indicate a significant
improvement in ride comfort, handling and stability. A very similar system is fitted to the
Toyota Soarer (Hirose et. al. 1988). A driver’s selector switch is also included. The
system includes control for anti-dive, anti-roll, anti-squat, anti-bump, response to speed
and response to rough road. Very good ride comfort and stability are achieved while
vehicle attitude changes are remarkably reduced.
Wallentowitz and Holdman (1997) investigate the effect of different spring and damper
characteristics. They suggest that vehicle velocity and steering wheel angle be used to
switch the suspension to the hard characteristics during ambitious driving situations.
Otherwise damper software analyses the excitation frequency and load based on a quarter
car model and switches the damper accordingly. No validation is given.
Hennecke and Zieglmeier (1988) discuss a three-state variable damping system fitted to
the BMW 635 CSi. Sensors used include steering wheel angle, loading condition,
travelling speed, brake pressure, throttle position and vertical body acceleration.
Poyser (1987) describes a system designed by Armstrong incorporating a ride levelling
hydropneumatic spring and a 3-stage controllable damper. Steering wheel angle, vehicle
speed, body roll angle, and suspension travel are used to switch the dampers. For ride
comfort control the dampers are switched to the intermediate and high states when certain
pre-set limits (vehicle speed dependant) are reached.
Pinkos et. al. (1993) investigates the feasibility of a continuously variable semi-active
Electro-Rheological Magnetic (ERM) fluid damper through mathematical analysis,
computer simulation and actual vehicle testing. The control strategy employed is based on
adaptive gain control and vector summation of weighted sensor measurements. Each
corner of the vehicle is treated independently, but the total control output is calculated
from information on vehicle behaviour. Separate calculations are produced for ride
comfort, roll, dive, squat, pitch, heave and yaw. The vector summation of these
calculations produces an output signal to each damper. Algorithm calculations are
prioritised based on safety related vehicle behaviour i.e. any calculations related to
vehicle handling are completed first. Thirteen sensors are used namely vehicle speed,
braking and acceleration, vertical accelerometers on the sprung mass, angular position
between the body and the wheel, lateral acceleration and absolute steering wheel position.
Both analogue (hardware) and digital (software) filters are employed. Quarter car and half
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.12
car models are examined while full-scale vehicle tests are also performed. Good
theoretical and experimental results are obtained.
3.3.2
Linear optimal, skyhook and on-off control
These three control methods are discussed together because skyhook control was derived
using linear optimal control theory and is used for a continuously variable damper. The
on-off strategy is a simplification of the skyhook strategy, adapted for a two stage (onoff) semi-active damper.
Krasnicki (1981) investigates the “skyhook” damping principle applied to a two-stage
(on-off) semi-active damper.
Karnopp (1990) points out that optimal control systems generally require the feedback of
all state variables while passive vibration control elements generate forces related to only
a subset of the system state variables. A quarter car model is used for simulation (four
state variables). Modern control theory suggests that the suspension force should consist
of a weighted sum of any four suitable state variables such as positions and velocities. An
optimum linear active system can thus be designed using Linear Quadratic Gaussian
control theory. According to the author, several other researchers report very similar
results. These control methods result in significantly better control of the body (sprung
mass) natural frequency. Partial state feedback is also shown to offer nearly the same
results as full state feedback. It is concluded that as far as body movement due to terrain
inputs is concerned, semi-active systems can approach the performance of fully active
systems with state variable feedback. It is however necessary to know the sprung mass
absolute velocity in order to apply state variable feedback control. (This cannot easily be
measured and might not be practical for vehicle implementation. It might not even be
possible to accurately estimate (see Hedrick et. al., 1994))
Sharp and Hassan (1987) study two alternative forms of control law. A quarter car
model is used for simulation. The semi-active damper is assumed to be capable of
producing a force that is a linear combination of state variables as long as such a force
opposes the relative motion of the damper. Otherwise it is set to produce no force. The
control laws are derived using stochastic linear optimal control theory. The constants used
in the control laws are obtained by minimising a performance index using two weighting
parameters, one for dynamic tyre load variations and the other for suspension working
space. The results given are for only one road surface roughness and one vehicle speed
but cover a range of suspension working space. It is concluded that semi-active damping
can improve ride comfort significantly but that the constants in the control laws must be
adapted according to the terrain roughness (or available suspension working space). It is
suggested that this adaptation of the coefficients can be achieved by keeping a running
average of the relative suspension displacement or monitoring the number of bump stop
contacts. The maximum use must be made of the available suspension travel while hitting
the bump stops must be avoided.
Margolis (1982a) uses a vehicle model that includes the heave (vertical) and pitching
motions of a vehicle. Controllers are designed for the fully active case and then modified
to be semi-active. Two control strategies are investigated namely the familiar “skyhook”
control (feedback of body absolute velocity and relative damper velocity) as well as
complete state variable feedback (SVFB). It is concluded that SVFB and “skyhook”
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.13
control both give excellent results compared to that of the passive system. Results are not
sufficiently strong in favour of SVFB to justify the increased complexity of measuring all
four state variables.
Margolis (1982b) presents the expected response of a simple vehicle (single degree of
freedom) fitted with an active and semi-active suspension when the control system is
presented with non-ideal feedback information. The control strategies evaluated need
feedback of the absolute velocity of the sprung mass. Determination of this velocity is
quite difficult in a realistic environment where the vehicle has many degrees of freedom,
for example roll, pitch, yaw and heave. This problem is intensified because all
measurements are corrupted by noise. The absolute velocity can be determined by
integrating an accelerometer signal by analog or digital means. It is however very difficult
to produce a drift free pure integrator. A low pass filter is used instead of a pure integrator
with a break frequency much lower than the frequency of interest. This is also very
difficult to realise because huge capacitor and resistor values are needed. Furthermore the
long time constants involved give rise to DC drift. The DC drift is exaggerated by the fact
that an accelerometer that can measure at the very low frequencies is also sensitive to
vehicle orientation (for example driving up a long incline). This necessitates the inclusion
of a high pass filter to eliminate the DC drift or steady state bias. The high pass filter
suffers from the same drawback of an extremely low break frequency. It is indicated that
the provision of acceleration feedback can provide some compensation for the non-ideal
velocity measurement. Significant improvements over the passive system are still
achieved although degraded by non-ideal velocity measurements.
Nell and Steyn (1994) discuss the experimental evaluation of a two-state semi-active
damper for off-road vehicles. Three control strategies available from literature are tested.
The first strategy used is the on-off strategy proposed by Karnopp (see Rakheja and
Sankar, 1985) that switches the damper according to the sign of the product of absolute
body velocity and relative damper velocity. The second strategy uses absolute body
acceleration and relative damper velocity. The third strategy proposed by Rakheja and
Sankar (1985) uses the product of relative damper displacement and relative damper
velocity. Unweighted RMS values of body acceleration, relative displacement and
velocity, absolute velocity and force indicate that the biggest improvement is achieved
using acceleration feedback followed by relative displacement and velocity (Rakheja and
Sankar). The on-off strategy proposed by Karnopp returns unsatisfactory results without
any significant improvements.
Experimental verification of theoretical work is discussed by Rajamani and Hedrick
(1991). A full-scale half-car suspension test rig is used to evaluate semi-active dampers.
High bandwidth (10 ms) 12 state semi-active dampers as well as low bandwidth 3-state
dampers are used. Conventional on-off, optimal on-off, optimal multi-state control and a
robust form of multi-state control are implemented and compared to predicted results.
Good correlation between predicted and measured results is achieved. The semi-active
suspension is found to behave as well as the best of all passive states at every frequency.
Lizell (1988) describes semi-active damper hardware and software that is tested in the
laboratory and on a vehicle. The aim of the control strategy employed is to switch the
two-stage damper to the high damping state in the region of the body resonance and
wheel hop frequencies, while the soft state is used for all other frequencies. This is said to
improve both handling and ride comfort throughout the frequency range. Damping of the
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.14
body resonance frequency is controlled using the Karnopp strategy. The wheel hop
frequency is controlled by calculating a discrete Fourier transform (DFT) around the
wheel hop frequency. The value obtained is compared to a threshold level to determine
damper switching. The damper is switched to the low damping state under all other
conditions. The absolute body velocity is determined from integrating an acceleration
signal after analog low-pass filtering. A digital high pass filter is implemented and drift in
the integration process is controlled by “leakage”. Preliminary test data is promising.
Ivers and Miller (1989) compare experimental results obtained from a quarter car test rig
with simulation data. A semi-active damper with 25 discrete states is used. The control
algorithms used are based on the simple analogy of the skyhook damper. Absolute body
velocity is determined by pseudo-integrating an acceleration signal. Three cases are
investigated namely passive, two stage (on-off) semi-active and continuous (25 stages)
semi-active control. Test results confirm the trends indicated by simulation, but there are
discrepancies due to the fact that valve response times, time delays in the control system,
hysteresis, friction in the test rig and non-linear damper characteristics are ignored in the
simulation.
Miller and Nobles (1988) describes the development and testing of a semi-active
suspension on an M551 military tank. The article gives a good overview of the
development history of controllable suspension systems and presents the basic theory
concerned with optimal control, resulting in the skyhook damper and on-off strategy. The
on-off strategy is implemented for vehicle trials. The determination of absolute velocity is
considered a challenge and is estimated (pseudo integrated) by filtering an accelerometer
signal. The valve configuration in the damper is designed so as to eliminate the need to
measure relative velocity. The control system therefore only has absolute velocity as
input while valve logic takes care of the rest. Vehicle testing is performed on a 10-axis
vertical road simulator. Average absorbed power was used as evaluation parameter and
indicated a measured performance gain between 13 and 43% depending on vehicle speed.
Miller (1988a) investigates the effect of hardware limitations on an on-off semi-active
suspension using a single degree of freedom simulation model and the familiar on-off
control strategy. The effects of non-zero off-state damping, valve dynamics and digital
filter dynamics (used to determine the absolute velocity) are investigated. Results indicate
that the off-state damping ratio should be less than 0.2. Valve response times should be
less than 14 milliseconds and sampling time less than 4 milliseconds. Digital filters
should have a break frequency of approximately 0.1 Hz and a damping ratio of between
0.3 and 1.0.
Temple and Hoogterp (1992) describe simulation and vehicle test results obtained for
the Mobility Technology Test Bed (MTTB) vehicle. The adaptive dampers employ an onoff strategy based on hull and damper dynamics. The damper is turned on only when it
will help to reduce the pitch and roll velocities. Whenever the anticipated jounce or
rebound damping would tend to increase the hull pitch and roll velocities, the dampers are
switched to the low damping state. No further details of the control strategy or
implementation thereof are given. Nearly a 1000 mobility and agility tests were
conducted on 10 vehicle configurations, all indicating noteworthy improvements in ride
comfort, reaction to discrete obstacles, reductions in body roll and reductions in pitching.
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.15
Besinger, Cebon and Cole (1991) tests an on-off semi-active damper in a hardware-inthe-loop (HiL) test setup where a quarter car model is solved by computer simulation
while the damper force is measured directly in real time from the experimental setup. Onoff skyhook control is implemented.
Hrovat and Margolis (1981) describe an experimental heave model of a tracked air
cushion vehicle incorporating an on-off semi-active damper. The control is performed
using a simple analog circuit with operation amplifiers and NAND gates implementing
the on-off strategy. Sinusoidal ground inputs in the range of 2 to 5.5 Hz are used. Results
indicate that significant improvements can be realised using semi-active damping
compared to passive damping. Absolute and relative damper velocities are obtained by
analog differentiation of displacements measured by LVDT’s. It is not possible to
implement this strategy in a real vehicle application.
Soliman et. al. (1996a and 1996b) extend previous work (where linear stochastic
optimal control theory was used to formulate a limited state feedback scheme) to include
adaptive control based on a gain scheduling approach. Results are determined
theoretically and experimentally using a quarter car model and test rig. Two strategies are
investigated using RMS wheel acceleration and RMS of the suspension working space
(relative displacement) respectively. Road surfaces of varying roughness are generated
using Gaussian random distributions and a road roughness number. Based on linear
optimal control theory, the absolute displacements and velocities of the wheel and body
are still required. A look-up table is used to determine the “optimum” gains for the
specific road input conditions as measured by the sensors. Theoretical and experimental
results indicate that the scheme based on the RMS vertical acceleration results in the
highest improvements in body acceleration, suspension working space and dynamic tyre
loads.
Abd El-Tawwab and Crolla (1996) include component limitations in the theoretical and
experimental investigation of a three state semi-active damper in a quarter car model and
test rig. The ideal actuator force is determined from optimal control theory and involves
feedback of absolute displacements and velocities for both the sprung and unsprung mass,
each associated with a control gain. The gains are determined using a gradient search
method. A random road input and a constant vehicle speed of 20 m/s is used. Results
indicate an improvement of between 13 and 17% for sprung mass acceleration and 7 to
8% for dynamic tyre load.
Lieh (1996) studies the application of velocity feedback active suspension systems. No
results are presented.
Petek et. al. (1995) performs vehicle tests using fast, continuously variable, electrorheological (ER) dampers. A modified skyhook algorithm is implemented which include
roll, pitch and heave motion. Accelerometers and LVDT’s are used to determine body
acceleration and relative displacement respectively. Accelerations are integrated (to
obtain absolute roll, pitch and heave velocities) and relative displacements differentiated
to obtain relative velocities. Four gain constants are used to determine the relative
importance of roll, pitch and heave motion. Test results indicate significant improvements
in ride comfort and stability compared to the standard passive suspension.
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.16
3.3.3
Neural networks and Fuzzy logic
An extensive literature survey on the applications of fuzzy logic and neural networks to
vehicle systems, including suspension control, is given by Ghazi Zadeh et. al. (1997).
Chou et. al. (1998) present a new control scheme referred to as the grey-fuzzy control
method that consists of two parts namely the grey predictors and the fuzzy logic
controller. The system is said to be able to control excessive tyre deflection and improve
ride comfort. The Taguchi method is employed to search for the optimal control
parameters and the results, obtained by computer simulation of a quarter car model, is
said to be satisfactory.
Hashiyama et. al. (1995) presents a new method to generate fuzzy controllers through
the use of a genetic algorithm (GA). Appropriate combinations of input variables, number
of fuzzy rules and parameters for membership functions are determined automatically
through the GA operations. A fuzzified version of Karnop’s law of suspension control
was incorporated as the initial fuzzy rules. These initial rules are not modified by the GA
but the GA with a new local improvement mechanism is applied to find additional fuzzy
rules for better performance. The performance index is improved but no comparisons are
given to the passive suspension performance.
Yoshimura et. al. (1997) presents a semi-active suspension controlled by fuzzy
reasoning. The input variables to the fuzzy control rules are the suspension travel and its
derivative. The aim is to minimise body vertical and roll acceleration at the centre of
gravity under the constraints of suspension travel and tyre deflection. A half car
simulation model is used. Simulation results show that the proposed system is very
effective in improving the vertical and rotary accelerations of the vehicle body as well as
tyre deflections.
3.3.4
H∞ control
Palmeri et. al. (1995) describes the application of H∞ optimal control theory to the design
of a fully active suspension system for an experimental Lancia Thema sedan car. The
system functions as a Multiple Input Single Output (MISO) regulator with hub
acceleration, actuator force and actuator position as inputs. The H∞ control strategy has
been chosen to take advantage of the possibility to design a competitive MISO controller
as well as exploit robust disturbance rejection which the H∞ theory grants. Each corner of
the vehicle is modelled as a seventh order state-space model. The H∞ regulator is a modelbased compensator, which means that it contains the system’s state-space model that is
observed and the control compensates for the error. Vehicle tests on a laboratory test
setup indicate that H∞ performs significantly better at all speeds than the skyhook
baseline, especially at low frequencies around the body roll frequency.
3.3.5
Proportional Derivative (PD) control
Esmailzadeh (1979) uses a linear model of a suspension system employing a pneumatic
isolator and a three-way servo valve. Simulation is performed on an analogue computer
and compared to experimental measurements of a quarter car model. Proportional and
derivative feedback control is used.
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.17
3.3.6
Preview control
Currently no feasible mass production preview sensors are available for suspension
control purposes and even if such sensors become available in the near future, it is
doubtful whether they will be of much use on off-road vehicles travelling over rough,
vegetation covered and deformable terrain. Preview control is not discussed in depth due
to this reason.
Soliman and Crolla (1996b) investigate the use of preview or “look-ahead” information
for semi-active damper systems using a quarter car theoretical model. The system is said
to achieve the same performance as a fully active system without preview.
Youn (1991) derives a preview control strategy using optimal control theory with jerk
included in the performance index. Simulation is performed using a two-degree of
freedom quarter car model. The proposed control method is said to improve handling and
ride comfort simultaneously. The jerk controller can determine the damping coefficient or
spring stiffness of the semi-active system.
Crolla and Abdel-Hady (1991) investigates the effect of wheelbase preview (i.e. that the
rear suspension input is just a delayed version of the input at the front) on the
performance of semi-active and fully active suspension systems. A continuous semiactive damper is used which is modelled as having a maximum and minimum damping
constant. Damper response time is modelled as a first order time lag. A simple full
vehicle model with vertical, pitch and roll degrees of freedom is used for simulation. The
control law is based on full state feedback. The conclusion is drawn that semi-active
systems with wheelbase preview can perform better than fully active systems without
wheelbase preview.
3.3.7
Model following
Pollard (1983) adopts a strategy first developed for a maglev train, to control the active
suspension of a normal train. The control system consists of two complementary parts. At
low frequencies the vehicle must follow the tracks and displacements must be maintained
within certain limits. The actuator is then controlled so as to minimise relative
displacements over the secondary suspension. At high frequencies, the acceleration of the
body is fed back to the control system and the system tries to minimise acceleration. The
control system is said to model the ideal suspension while the actuator tries to correct the
error. The bounce and pitch modes of the body are controlled separately.
3.3.8
Frequency domain analysis
Hamilton (1985) proposes to use a Discrete Fourier Transform (DFT) to calculate the
magnitude of vibration levels in different frequency bands in order to control body
resonances.
Kojima et. al. (1991) implement a frequency detection method that changes the dampers
to high damping when suspension inputs are predominantly in the low frequency range.
Low damping is used for suspension movements that are predominantly in the high
frequency range. It is found that the low frequency region is accompanied by large
suspension stroke variation and large variations in distance between the vehicle body and
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.18
ground while damping force variation ratio and bounce down acceleration is small. The
magnitude of these parameters is reversed in the high frequency region, enabling
discrimination between frequency ranges on the basis of the amplitude of these
parameters. A relative position sensor measures suspension movement and piezo-electric
ceramic sensor is used to detect the damping force variation ratio. The suspension
movement sensor does not have an absolute neutral position signal but determines the
neutral position by compensation with learning control. Additional sensors, for example
vehicle speed, steering, brake application and throttle angle are also used.
3.3.9
“Relative” control
Rakheja and Sankar (1985) and Alanoly and Sankar (1987) present an “original”
control strategy employing only directly measurable variables in vehicle applications. A
continuously modulated damper is controlled using only relative damper displacement
and relative velocity as feedback signals. A condition function based on the sign of the
product of relative velocity and relative displacement determines whether the high (on) or
low (off) damping state have to be used. The origin of, or reasoning behind, this strategy
appears to be determined from a thought experiment. There is very little variation
between this scheme and the “skyhook” damping algorithm. Performance approaching
that of a fully active suspension system is achieved from simulation results on a single
degree of freedom system. This system avoids the problem of measuring the sprung mass
absolute velocity, that is said to be a near impossible task, and has never been
implemented on a vehicle (at the time of writing). The same strategy is proposed by Jolly
and Miller (1989) and is termed “relative control”. It is developed by means of intuitive
reasoning. Relative control is found to perform better than the passive system but slightly
worse than skyhook control. At high frequencies, relative control gives results very
similar to skyhook damping, but at low frequencies, relative control performs worse than
the passive system. It is likely that relative control will provide better performance in
applications where most of the disturbance energy is transmitted at higher frequencies.
3.3.10 Traditional controller design on the s-plane
Hall and Gill (1987) depart from the approach of using optimal control theory. Instead
they try to relate the position of the closed loop poles of the system on the s-plane to the
poles of a well-designed “skyhook” system. Not much success is achieved with this
method. The authors then revert to scanning of the s-plane in order to find optimum pole
locations. It is concluded that, although the transmissibility indicates significant
improvements, the phase relationships need to be taken into account.
3.3.11 Minimum product (MP) strategy
Nell and Steyn (1998) develop an alternative control strategy (called the minimum
product or MP strategy) for semi-active dampers on off-road vehicles that takes into
account the pitch and roll degrees of freedom. The strategy selects a combination of
damper settings (all dampers on vehicle taken into account) that minimises roll and/or
pitch acceleration. Both simulation and experimental results, that indicate that this
strategy performs better over off-road terrain in comparison with both the passive and onoff skyhook systems, are given. The damper state that will give the lowest acceleration in
the present direction of movement, or the highest acceleration in the opposite direction, is
selected. Input variables to the control system are relative velocity of each damper as well
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.19
as the roll and pitch accelerations of the vehicle body (calculated from three vertical
acceleration measurements by assuming that the vehicle body is rigid).
3.3.12 Roll and pitch velocity
Salemka and Beck (1975) formulate and test a strategy based on the roll and pitch
velocities of the vehicle body. Terrain parameters, for example the relative amount of roll
and pitch velocities and vertical acceleration generated by the terrain, severely influence
the success of any control strategy.
3.3.13 Resistance control
Fodor and Redfield (1996) implement resistance control semi-active damping on a
1/30th-scale quarter car test rig. Test results are compared to simulation results and good
correlation is found.
3.3.14 Mechanical control
Speckhart and Harrison (1968) perform an analytical and experimental investigation of
a hydraulic damper having internal inertially controlled valves. The valve is purely
mechanical and no “control system” is used. The system claims to improve ride comfort
by reducing jerk. System performance is evaluated by simulation and laboratory testing of
a two degree of freedom system.
3.3.15 Steepest gradient method
Tseng and Hedrick (1994) investigate the optimal semi-active suspension that will
minimise a deterministic quadratic performance index. The optimal control law is a timevarying solution that involves three related Ricatti equations. The constant Ricatti
equation (so-called “clipped optimal” solution) is not optimal. They develop a new semiactive algorithm called the “steepest gradient” algorithm. Performance is shown to be
superior to that of the “clipped optimal” solution.
3.3.16 Use of estimators and observers
Hedrick et. al. (1994) propose a new method for designing observers for automotive
suspensions. The methodology guarantees exponentially convergent state estimation
using easily accessible and inexpensive measurements. It is also demonstrated that the
sprung mass absolute velocity cannot be estimated in an exponentially stable manner with
such measurements. The estimation error is merely bounded and would not converge to
zero. Results are verified on the Berkely Active Suspension Test Rig with excellent
results. The sprung mass velocity is, however, not estimated, but determined by
integrating the body acceleration after passing it through a high pass filter.
3.3.17 Control of handling
No literature proposing any control strategies for specifically improving vehicle handling
was found. In cases where handling is considered, it seems that the authors opted for the
stiffest possible setting when encountering handling manoeuvres.
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.20
3.3.18 Control of rollover
A genetic algorithm predictor for vehicle rollover was developed by Trent and Greene
(2002). They modelled a 1997 model Jeep Cherokee SUV. Their preliminary results
indicate rollover prediction of 400 milliseconds in advance of the actual event. They
suggest that this early warning could be used to prevent rollover by activating other
vehicle systems such as differential braking or suspension control.
3.3.19 Ride height adjustment
A decrease in ride height is generally beneficial for handling as the centre of gravity
height will be decreased. This improves the static stability factor (SSF) and should
therefore reduce the rollover propensity of the vehicle. Care should however be taken not
to change the suspension geometry in a manner that will adversely affect the handling. On
the other hand an increase in ride height might benefit ride comfort over rough terrain
because suspension travel in bump will be increased, thereby reducing the number of
bump-stop contacts. In many vehicles, the ability to maintain constant ride height
independent of load is a major advantage, without necessarily adding the capability to
increase or decrease ride height. The success of ride height control can be judged by it’s
numerous commercial applications.
3.3.20 Comparison of semi-active control strategies for ride comfort improvement
Voigt (2006) studied several control strategies proposed in literature during the last 20
years with the objective of improving ride comfort. The study focussed on on-off control
ideas. The aim of the study was to develop and implement an appropriate ride comfort
control strategy for a 4-state semi-active hydropneumatic suspension system, consisting
of a two-state semi-active hydropneumatic spring and a two-state semi-active damper.
Simulation models of both ¼ car and ½ car (pitch and bounce) vehicles were developed
in Simulink. Typical values for a Land Rover Defender 110 SUV were used in the
models. The suspension model developed by Theron and Els (2005), as described in
paragraph 4.8, was used in the simulation. Hardware-in-the-loop (HiL) testing of a
prototype suspension system was also performed using the techniques developed by
Misselhorn, Theron and Els (2006). The suspension used in the HiL test rig was
Prototype 2 discussed in chapter 4 of the present study. Simulation results and HiL results
were found to correlate very well (within 10%). In order to simulate ride comfort for both
on- and off-road conditions, road inputs included:
i)
sine waves with frequencies between 0 and 30 Hz and amplitudes of 0.001 to
0.015 m.
ii)
Belgian paving (Figure 2.21)
iii)
APG bump (Figure 2.22)
iv)
typical random road profiles ranging from a “smooth runway” to a “ploughed
field” generated from road roughness information obtained from literature.
The following control ideas were evaluated:
i)
ADD – Acceleration driven damper as proposed by Silane et. al. (2004). This
proposed strategy is the same as the strategy proposed by Holsher and Huang
(1991).
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.21
ii)
iii)
iv)
Skyhook – The familiar skyhook damper strategy proposed by Karnopp et.
al. (1973).
ReS – The strategy proposed by Rakheja and Sankar (1985).
MP – The minimum product strategy proposed by Nell (1993).
Table 3.1 indicates all the proposed ideas that were evaluated. No useful semi-active
spring control ideas were found. The springs were controlled using appropriately
modified versions of the damper control ideas. As a comparison, the passive “ride
comfort” mode (soft spring and low damping) of the semi-active hydropneumatic springdamper system was also simulated.
Simulation results indicated that “Spring ADD” performed marginally better than the
passive “ride comfort” mode. No control strategy was able to outperform the passive
“ride comfort” mode by more than 2%, which is within expected simulation error. The
“ride comfort” mode outperformed all control strategies for all HiL tests.
Voigt also investigated why the skyhook strategy performed unsatisfactory. The nonlinearity of the system affects performance. Skyhook performs well at low frequencies
but performance deteriorates at higher frequencies. This indicates that the valve response
time is too slow. Better ride comfort is also achieved by controlling the spring rather than
the damper.
The effect of limited suspension working space was also adressed by including bump
stops in the model. This had the biggest effect on the ride comfort of the passive
suspension. Again the “ride comfort” mode performed the best of all the possibilities. It
seems that the suspension system under consideration exhibits the same useful
characteristic of the twin-accumulator system described by Abd El-Tawwab (1997)
amongst others. Due to the dampers between the accumulators, the large accumulator is
progressively “sealed off” by the increased flow through the damper, i.e. spring stiffness
increases automatically when terrain gets rougher (higher flow rate of oil) thereby
eliminating bumpstop contact. This change is not discrete but happens gradually in
relationship to the suspension velocity.
It is concluded from Voigt’s study that it is not possible to improve ride comfort to any
worthwhile extent by controlling the spring and damper characteristics when the
characteristics have been optimised for ride comfort.
A similar study for handling has not yet been performed.
3.4
Conclusion
The following conclusions are made with respect to possible solutions for the ride
comfort vs. handling compromise:
i)
The ride comfort vs. handling compromise can be eliminated using active
suspension systems. These systems are very expensive and require significant
amounts of engine power. This option is disregarded for these reasons.
ii)
Semi-active suspension systems have the potential to approximate the
performance of fully active systems, but at a considerable reduction in cost
and complexity.
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.22
Table 3.1 – Control strategies evaluated by Voigt (2006)
Control Strategy Description
ADD
Skyhook
ReS
Spring ADD
Spring Skyhook1
Spring Skyhook2
Spring Skyhook3
Spring ReS
Combo ADD1
Combo ADD2
Combo ADD3
Combo Skyhook1
Combo Skyhook2
Combo Skyhook3
Combo ReS1
Combo ReS2
Combo KP
Damper Strategy
&&
x1 ( x&1 − x&2 ) > 0
&&
x1 ( x&1 − x&2 ) < 0
x&1 ( x&1 − x&2 ) > 0
x&1 ( x&1 − x&2 ) < 0
( x&1 − x&2 )( x1 − x2 ) < 0
( x&1 − x&2 )( x1 − x2 ) > 0
Damper
Hard
Spring strategy
Spring
Soft spring
Soft
Hard
Soft spring
Soft
Hard
Soft
Hard damping
Soft spring
&&
x1 ( x&1 − x&2 ) > 0
&&
x1 ( x&1 − x&2 ) < 0
x&1 ( x&1 − x&2 ) < 0
Hard damping
x&1 ( x&1 − x&2 ) > 0
x1 ( x1 − x2 ) < 0
Hard damping
x1 ( x1 − x2 ) > 0
&&
x1 ( x1 − x2 ) > 0
Hard damping
&&
x1 ( x1 − x2 ) < 0
Hard
Soft
Hard
Soft
Hard
Soft
Hard
Soft
Soft
( x&1 − x&2 )( x1 − x2 ) > 0
( x&1 − x&2 )( x1 − x2 ) < 0
&&
x1 ( x&1 − x&2 ) > 0
&&
x1 ( x&1 − x&2 ) < 0
&&
x1 ( x1 − x2 ) > 0
&&
x1 ( x1 − x2 ) < 0
x1 ( x1 − x2 ) < 0
x1 ( x1 − x2 ) > 0
x&1 ( x&1 − x&2 ) > 0
x&1 ( x&1 − x&2 ) < 0
x1 ( x1 − x2 ) < 0
x1 ( x1 − x2 ) > 0
&&
x1 ( x1 − x2 ) > 0
&&
x1 ( x1 − x2 ) < 0
( x&1 − x&2 )( x1 − x2 ) > 0
( x&1 − x&2 )( x1 − x2 ) < 0
&&
x1 ( x1 − x2 ) < 0
&&
x1 ( x1 − x2 ) > 0
&&
zw .&&
zb( 2 ) < &&
zw .&&
zb(1)
Hard
&&
zw .&&
zb( 2 ) < &&
zw .&&
zb(1)
Hard
&&
zw .&&
zb( 2 ) > &&
z w .&&
zb(1)
Soft
&&
zw .&&
zb( 2 ) > &&
z w .&&
zb(1)
Soft
Hard damping
&&
x1 ( x&1 − x&2 ) > 0
&&
x1 ( x&1 − x&2 ) < 0
&&
x1 ( x&1 − x&2 ) > 0
&&
x1 ( x&1 − x&2 ) < 0
&&
x1 ( x&1 − x&2 ) > 0
&&
x1 ( x&1 − x&2 ) < 0
x&1 ( x&1 − x&2 ) > 0
x&1 ( x&1 − x&2 ) < 0
x&1 ( x&1 − x&2 ) > 0
x&1 ( x&1 − x&2 ) < 0
x&1 ( x&1 − x&2 ) > 0
x&1 ( x&1 − x&2 ) < 0
( x&1 − x&2 )( x1 − x2 ) < 0
( x&1 − x&2 )( x1 − x2 ) > 0
( x&1 − x&2 )( x1 − x2 ) < 0
( x&1 − x&2 )( x1 − x2 ) > 0
Hard
Soft
Hard
Soft
Hard
Soft
Hard
Soft
Hard
Soft
Hard
Soft
Hard
Soft
Hard
Hard
Soft
Hard
Soft
Hard
Soft
Hard
Soft
Hard
Soft
Hard
Soft
Hard
Soft
Hard
Soft
Hard
Soft
POSSIBLE SOLUTIONS TO RIDE COMFORT VS. HANDLING COMPROMISE 3.23
iii)
iv)
v)
vi)
vii)
viii)
ix)
x)
xi)
xii)
xiii)
3.5
There are two viable concepts for a semi-active damper namely: Magnetorheological (MR) fluids and hydraulic dampers with bypass valves. Designs
can be continuously variable or discrete.
There are basically two viable concepts for a semi-active spring namely air
springs and hydropneumatic springs.
As far as control is concerned, a myriad of possibilities exist. All ideas can
however not be easily implemented in the vehicle e.g. measurement of
absolute body velocity for full-state feedback.
Ride height adjustment is widely used and offers many possibilities.
Reaction speed needs to be taken into account to determine potential system
performance.
Very little literature exists on semi-active springs.
Most control ideas are developed using ¼ car linear models that do not
sufficiently represent actual vehicle dynamics.
Very limited hardware has been implemented and documented.
Almost no work has been performed on off-road vehicles.
The majority of studies focus on ride comfort, and handling is often neglected.
Preview is a popular research topic, although hardware implementation is
problematic.
Proposed solutions to the ride comfort vs. handling compromise
Based on the ideas and research described in chapters 2 and 3, the proposed solution to
the “ride comfort vs. handling compromise” is to use a twin accumulator hydropneumatic
(two-state) spring combined with an on-off (two-state) semi-active hydraulic damper
(achieved with a by-pass valve), based loosely on idea by Eberle and Steele (1975).
Although more than two spring and/or damper characteristics can be incorporated, two is
considered sufficient based on the simulation results presented in Chapter 2.
Based on the results, presented by Voigt (2006), for ride comfort control, and assuming
that the same trends will be found for handling, if studied, the best practical solution
would be no “control” other than switching between the “ride comfort” and “handling”
modes. The pre-requisite is however that a successfull ride comfort vs. handling decisionmaking strategy can be developed that will automatically switch between the “ride
comfort” and “handling” modes. The switching must be safe and quick enough to prevent
accidents, using only easily measurable parameters.
The proposed suspension system will now be called the 4-State Semi-active Suspension
System or 4S4.
Chapter
4
THE FOUR-STATE SEMI-ACTIVE SUSPENSION
SYSTEM (4S4)
The development of a prototype 4-State Semi-active Suspension System (4S4) is
described in this chapter. Literature appropriate to the development of the suspension
system, and the working principle of the system is discussed. Two prototype suspension
systems (from now on referred to as Prototype 1 and Prototype 2 respectively) were
designed, manufactured, tested on a laboratory test rig and modelled mathematically.
After determination of the space envelope on the proposed test vehicle, Prototype 1 was
designed and manufactured by Hytec, a specialist hydraulic equipment manufacturer.
Prototype 1 suffered from several drawbacks that necessitated a redesign. Prototype 2 is
an in-house design and solved all the problems experienced on Prototype 1.
Detailed test results for Prototype 2 are discussed and interpreted. Test results for
Prototype 1 are only discussed where necessary to motivate some of the decisions made
during development of Prototype 2. Test results include spring and damper characteristics
as well as several parameters required for mathematical modelling of the suspension
system. These parameters include the bulk modulus of the oil, thermal time constant of
the accumulators, valve response times and pressure drops over the valves.
4.1
4.1.1
Literature
Hydropneumatic springs
Hydropneumatic springs are often modelled as polytropic gas compression processes.
With the assumption that the ideal gas law is applicable, this approach gives satisfactory
first order results. The static spring force can be calculated accurately using isothermal
compression. The dynamic force is however time and temperature dependent and requires
a more advanced model to achieve accurate results.
A detailed hydropneumatic spring model is developed and validated by Els (1993) and
Els and Grobbelaar (1993). This model is based on the solution of the energy equation
of a gas in a closed container and therefore takes time- and temperature dependency of
the spring characteristic into account. It is based on a thermal time constant approach and
uses the Benedict-Webb-Rubin (BWR) equation for real gas behaviour (Cooper and
Goldfrank, 1976). The model is verified against experimental results and good
correlation is achieved between measured and predicted spring characteristics. The model
is further developed to include heat transfer effects from the damper that is usually an
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.2
integral part of a hydropneumatic suspension system (Els and Grobbelaar, 1999). This
model was used to predict the 4S4 spring characteristics in paragraph 4.8.
Another approach that can be used to model hydropneumatic springs is by making use of
the so-called anelastic model (Kornhauser, 1994 and Giliomee et. al, 2005).
4.1.2
Variable spring concepts
The concept of making a semi-active spring using accumulators is not new. The
fundamental idea was proposed by Eberle and Steele (1975) as discussed in par 3.2.2.2.
Decker et. al. (1988) also implemented an air spring with various discrete volumes that
can be switched. The design was made for a passenger car, but no quantitative results or
design guidelines are given.
A passive twin-accumulator suspension system is proposed by Abd El-Tawwab (1997).
Two accumulators are connected via an orifice. As the flow rate of oil in the system
increases, damping through the orifices increases thereby resulting in different amounts of
fluid flowing into each accumulator. This results in a speed or frequency dependant
spring characteristic. The 4S4 system incorporates this capability as a function of its
design.
First attempts by the candidate to develop a two-state semi-active spring combined with a
two-state semi-active damper are discussed by Giliomee and Els (1998). The design was
for a heavy off-road wheeled vehicle with a static wheel load of 3 000 kg. Experimental
results included testing the system in a single degree of freedom test rig using various
control methods. Initial results were very promising and warranted further development
of the 4S4 system.
4.1.3
Hydraulic semi-active dampers
Semi-active dampers have been applied widely in prototypes and production vehicles.
The work of Nell (1993), Nell and Steyn (1994, 1998 and 2003) as well as Els and
Holman (1999) is of particular significance to the development of the 4S4 due to the
applications to heavy off-road military vehicles. The applications varied from two-state
translational semi-active dampers for a 12-ton 4x4 vehicle up to a two-state semi-active
rotary damper for a 46-ton self-propelled gun. In all these cases simulation results are
validated using vehicle tests with prototype dampers and control systems fitted. The
results are generally very satisfactory.
All these dampers operate on the bypass valve principle and have valve response times of
between 40 and 200 milliseconds. Large flow rates of up to 1000 l/min can be
accommodated with acceptable pressure drops over the valves.
4.2
4S4 Working principle
The concept behind the 4S4 system is to achieve switching between two discrete spring
characteristics, and between two discrete damper characteristics. The high and low
characteristics for both spring and damper are possible by alternate channeling of
hydraulic fluid with solenoid valves. The basic circuit diagram of the proposed
suspension system is given in Figure 4.1. The strut is fixed between the vehicle body and
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.3
the unsprung mass, replacing both the spring and damper. The strut is connected to two
accumulators via the control valves and hydraulic damper valves. The two-state
hydropneumatic spring can also be used on its own in parallel with an additional semiactive damper e.g. a continuously variable MR fluid based damper, but then some of the
elegance and packaging possibilities of the 4S4 unit will be sacrificed.
The low spring rate is achieved by compressing the combined volume of gas in the two
accumulators. By sealing off accumulator 2 with valve 3, a smaller gas volume is
compressed and a higher spring rate is achieved. Spring rates can be individually tailored
by changing the two gas volumes. For low damping, the hydraulic dampers (dampers 1
and 2) are short circuited by opening the bypass valves (valves 1 and 2). For high
damping these valves are closed and the hydraulic fluid is forced through the dampers
resulting in high damping force. The proposed system therefore achieves its aim to
provide switching between two discrete spring characteristics, as well as switching
between two discrete damper characteristics using solenoid valves.
The concept can easily be extended to more spring characteristics by adding more
accumulators and valves. The two-state dampers can also be upgraded by fitting
proportional or servo valves, thereby achieving continuously variable semi-active
damping. Although these improvements are possible, they will add considerable
complexity and cost and are therefore not considered at present. Adding or extracting oil
from the unit results in ride height adjustment.
Figure 4.1 – 4S4 circuit diagram
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.3
4.4
Design requirements
Before designing the new suspension system, it is necessary to obtain the specifications
for the existing baseline system in terms of wheel load, maximum suspension deflection
and space envelope available.
The maximum static vertical wheel load for the fully laden Land Rover Defender 110 test
vehicle is 800 kg and occurs on the rear wheels. The prototype suspension system is
therefore designed for a static load of 8000 N and a dynamic load of 40 000 N (five times
the static wheel load). Provision is made for a total suspension travel of 300 mm
(maximum compression to maximum rebound). The baseline rear suspension system has
a total travel of 290 mm (170 mm compression and 120 mm rebound). The required
suspension characteristics for the springs and dampers are obtained from the analysis in
chapter 2. Gas volumes of 0.1 litre (Accumulator 1) and 0.4 litre (Accumulator 2) are
used as design values. Provision is made for fitment of a wide range of available
hydraulic damper packs so that the damper characteristics can be fine-tuned before final
vehicle implementation. To enable the use of standard hydraulic seals, valves and fittings,
the system is designed not to exceed a maximum pressure of 20 MPa. A maximum
relative suspension velocity of 2 m/s is assumed to be sufficient for extreme events. This
was determined from simulation results as well as measurements on the baseline vehicle.
It is envisaged that the suspension system must be able to control the body’s natural
frequencies in the region of one to two Hz. This requires a valve reaction of 10 to 20 Hz
or 50 to 100 milliseconds. This was also found to be the case by Nell (1993) and Nell and
Steyn (1994).
The main design specifications for the prototype controllable suspension system are
summarized as follows:
i)
Suspension travel of 300 mm (same as for baseline suspension)
ii) Soft suspension static gas volume of 0.5 litre
iii) Hard suspension static gas volume of 0.1 litre
iv) Maximum system pressure at full bump of 20 MPa
v) Maximum relative suspension velocity of 2 m/s
vi) Maximum suspension force of 40 kN (5x static force)
vii) Valve response time of the order of 50 milliseconds
viii) Must fit into available space envelope without major modifications to vehicle
ix) Low damper characteristic < 0.5 of baseline value
x) High damper characteristics between 2 and 3 times the baseline value
These specifications are for the rear suspension and represent the worst-case scenario.
The only changes required for fitment of the prototype to the front suspension of the Land
Rover 110, is to reduce the total suspension travel to 250 mm.
The piston diameter required to give a maximum pressure of 20 MPa at 40 kN is 50.5
mm. A piston diameter of 50 mm will be used for the design of the prototype. Figure 4.2
indicates the relative suspension velocity over the left rear spring of a standard Land
Rover Defender 110 when driven on the Gerotek Test Facility’s rough track. This
velocity was calculated by differentiating the measured relative displacement. The
maximum extreme event velocity over this type of terrain at representative speeds is 2
m/s. The 50 mm piston diameter will therefore result in a flow rate of 236 litre/min at 2
m/s relative suspension velocity.
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.5
Selection of an appropriate valve was based on the response time of 50 milliseconds,
maximum system pressure of 20 MPa and maximum extreme event flow rate of 236
litre/min. Choice and availability of valves is problematic as a valve with a fast switching
time is required. Standard valves, available off-the-shelf, can meet either the flow or the
time response requirements, but not both. For this flow rate requirement logic element
valves operated by a pilot solenoid valves are usually employed (Nell (1993), Nell and
Steyn (1994), Janse van Rensburg, Steyn and Els (2002), Els and Holman (1999)).
This solution is bulky and expensive, and above all results in response times that are
strongly pressure dependent and very slow at small pressure differences. The design has
therefore been modified to use two smaller, fast switching valves in parallel to handle the
required flow and meet the switching time requirement.
Figure 4.2 – Relative suspension velocity over Gerotek Rough track
The valve selected for the current application is the SV10-24 2-way normally closed
spool valve from HydraForce (Anon, 1998). This valve has previously been characterized
for a different project at the University of Pretoria and information on response times and
pressure drops are available (De Wet, 2000). The valve is actuated by a solenoid that is
available in different voltage ratings. The response time (initial delay) is quoted to be 30
milliseconds when energised (i.e. opening) and 25 milliseconds when de-energised (i.e.
closing). This is the time from the switch signal to the first indication of the change of
state, called the initial delay (see par 4.7.6 for definitions). This response time is quoted at
a flow rate of 80% of the nominal flow rate when the valve is fully open. The valve is
designed for a maximum operating pressure of 20.1 MPa and proof pressure of 35 MPa.
The valve can handle a flow of 113.6 litre/min at a pressure of 6.9 MPa and 37.9 litre/min
at 20.7 MPa (see Figures 4.3 and 4.4). When valve 3 in the proposed concept (see Figure
4.1) is open, the flow will be split between accumulators 1 and 2 but with the higher
portion of the flow going into the bigger accumulator 2. The expected flow is however
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.6
still higher than the maximum capacity of the valve. It was therefore decided to use two
valves in parallel for V3.
Eight standard Land Rover Defender rear dampers were stripped, the damper packs
removed and mounted in the 4S4 units (two damper packs per unit). Due to the difference
in bore size, and thus flow rate, as well as the pressure difference now acting on a larger
area, use of the standard damper packs resulted in the required hard damper
characteristics (see discussion in paragraph 4.7.5 and Figure 4.27).
4.4
Space envelope
The space envelope available for the new suspension was determined by physical
measurement on a Land Rover Defender 110 vehicle. The controllable suspension
system, with the required characteristics, has to fit in the space envelope. The left front
and left rear axle portions and wheel well details were measured and modelled in Solid
Edge for this purpose as indicated in Figures 4.5 and 4.6.
4.5
Detail design of 4S4
The height of the space envelope is the major restricting parameter, followed by the
distance between the fenders and the inside of the wheel arches. Length should not pose
any limitations, as the full tyre diameter is available. To comply with the height
restriction, the two accumulators are mounted to the front and rear of the main strut
respectively as indicated in Figure 4.7. The strut is connected to the two accumulators via
a valve block. All the control valves, hydraulic damper valves, control ports and channels
are accommodated inside the valve block. Piston accumulators are used mainly for two
reasons:
i)
It can be made long and thin compared to bladder accumulators, thereby resulting in
more freedom of packaging
ii) The gas volume can be controlled much more precisely (see paragraph 4.7.1 –
charging of unit).
The choice as far as sealing arrangements are concerned is between sealing in the cylinder
bore and sealing on the piston rod. The rod sealing arrangement was chosen instead of the
more conventional cylinder sealing because it was much easier to finish the rod to the
correct tolerances and surface finish required than the cylinder bore. The options
considered for surface coatings at this stage is the normal hard chroming as well as a
tungsten-carbide-cobalt coating applied with a high velocity oxygen fuel (HVOF)
process. The latter is very resistant to flaking and has extremely good wear resistance. For
both Prototypes 1 and 2, the tungsten-carbide-cobalt coating was used because it is
suggested for the application by one of the world’s biggest seal manufacturers, Greene
Tweede. After coating the rod was ground and superfinished with diamond tape to obtain
a hard, corrosion resistant component with the required surface finish to ensure durability
and low friction. During the design phase, attention was given to minimise friction and
stick-slip. Standard seals from the Busak and Shamban catalogue (Anon, 2005d) were
used throughout the design of the 4S4. A Turcon AQ Seal 5 and two Glydring wear rings
were used on the floating pistons in the accumulators. The main cylinder pressure was
sealed
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
Figure 4.3 – Pressure drop vs. flow rate for SV10-24 valve (Anon, 1998)
Figure 4.4 – Operating range for SV10-24 valve (Anon, 1998)
4.7
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
Figure 4.5 - Baseline left front suspension layout
Figure 4.6 - Baseline left rear suspension layout
4.8
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.9
using a rod sealing arrangement with a triple seal system consisting of a TURCON
STEPSEAL 2K, TURCON RIMSEAL and TURCON EXCLUDER 2 rod scraper.
During testing of the first prototype suspension system the force characteristics exhibited
very high frictional behavior (hysteresis). This was traced to the off-center mounting
arrangement on the first prototype that subjected the cylinder to a moment loading and
caused high seal friction. The mounting arrangement on Prototype 2 was changed to be
concentric with the cylinder. The new mounting arrangement eliminated the hysteresis
encountered on the first prototype (see par 4.7.7. under test results).
On the prototype, provision is made for four pressure transducers (P1 to P4) to measure
pressures in the system.
Two views of the Prototype 2 controllable suspension system are provided in Figures 4.8
and 4.9. Figure 4.8 shows an exterior side view and Figure 4.9 indicates a cross-sectional
view. The suspension system is mounted to the axle and the chassis by means of a
spherical bearing used axially. The spherical bearing, although normally intended for
radial forces, is appropriately sized to handle the axial load. This bearing is used to ensure
pure axial force loading on the suspension system and eliminates any moment loading.
Figures 4.10 and 4.11 depict the 4S4 unit fitted to the vehicle at the front and rear
respectively.
It is concluded that the suspension system can be fitted in the available space although
small changes to the vehicle may be required. The new suspension system is narrower
than the coil spring and this may result in more interior space in the vehicle.
4.6
Manufacturing of 4S4 prototypes
The Prototype 2 controllable suspension system was manufactured according to detail
design drawings. A photograph of the assembled unit is given in Figure 4.12. Figure 4.13
compares Prototype 2 to Prototype 1. Prototype 2 is considerably smaller than Prototype 1
in overall size and weight. Valve positions have been optimised to reduce size. The valve
block requires very few external blanking plugs compared to the first prototype. The
weight of the unit was reduced from 59 kg for Prototype 1 to 40 kg for Prototype 2. The
mounting arrangement to the vehicle chassis has been modified considerably to remove
the moment loading. On Prototype 2, the weight includes all the mounting brackets to the
vehicle, while on Prototype 1 mounting brackets are not included in the quoted weight.
4.7
Testing and characterisation of the 4S4
The 4S4 Prototype 2 suspension was characterized on a test rig to obtain all the spring and
damper characteristics as well as valve response times. A series of basic reliability tests
were also performed to validate the choice of hydraulic seals and valves. The test rig
consisted of a purpose designed test frame and a 100 kN SCHENCK hydropulse actuator
(see Figures 4.14 to 4.18). The prototype suspension unit was instrumented with four
pressure transducers to determine dynamic system pressures, with actuator force and
actuator displacement also being measured. The switching signal to the valves was
recorded for determining the valve response times. A linear potentiometer was installed
on valve 3 to measure the valve plunger displacement.
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
Figure 4.7 – 4S4 suspension schematic diagram
4.10
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
Figure 4.8 – 4S4 suspension system – exterior view
4.11
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
Figure 4.9 – 4S4 suspension system – cross sectional view
4.12
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
Figure 4.10 - Front suspension layout with 4S4 unit fitted
Figure 4.11 – Rear suspension layout with 4S4 unit fitted
4.13
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
Figure 4.12 – 4S4 Prototype 2
4.14
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.15
Figure 4.13 - 4S4 Prototype 2 (left) compared to Prototype 1 (right)
4.7.1 Gas charging procedure
The spring characteristics are completely dependant on the volume of gas in each
accumulator. It is therefore imperative that the gas charging procedure described below
be strictly observed otherwise the spring characteristics will be in error.
i)
ii)
iii)
iv)
v)
During assembly of the suspension unit, both floating pistons must be
pushed in until they touch the valve block.
Move the piston rod to the maximum extended (rebound) position.
Open all solenoid valves by connecting them to a suitable power supply.
Fill the strut completely with oil. Tilt the strut slowly in different directions
in an attempt to get rid of trapped air. If the unit seems to be full, let it stand
for a few hours and top up frequently with oil. Slowly tilt the unit during
each filling attempt. The unit should take at least 1.6 litres of Aeroshell Fluid
41.
Disconnect power to the valves.
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
Figure 4.14 - 4S4 Prototype 2 on test rig
4.16
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
Figure 4.15 - 4S4 Prototype 2 on test rig
4.17
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
Figure 4.16 - 4S4 Prototype 2 on test rig
4.18
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
Figure 4.17 - 4S4 Prototype 2 on test rig
4.19
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
Figure 4.18 - 4S4 strut mounting to test rig
4.20
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.21
vi)
vii)
viii)
Install the unit in the test rig.
Reconnect the valves and apply power so that all the valves are open.
Remove the M8 cap screws used to bleed off gas from the accumulators
from the accumulator end caps.
ix)
Slowly compress the unit to the maximum compression (bump) position,
noting the force on the load cell whilst doing so.
x)
If the force on the load cell starts increasing rapidly before maximum
compression is reached, investigate the problem before continuing.
xi)
When maximum compression is reached, measure the distance between the
accumulator end caps and the floating piston through the gas bleed hole with
a vernier. This distance should be 29 mm for the small (0.1 litre)
accumulator and 62 mm for the big (0.4 litre) accumulator. The cavities in
the accumulator end caps have been designed to result in the correct gas
volumes in the maximum compressed positions when the floating pistons are
resting against the end caps.
xii)
If any of these two distances are greater than indicated, there is not enough
oil in the strut. If this is the case, remove the highest blanking plug on the
valve block. Extend the strut by about 10 mm. Fill the strut with more oil.
Repeat steps (ix) to (xii) until the strut if filled completely. If there is too
much oil in the strut, the excess can be drained off by removing the highest
blanking plug and compressing the strut fully.
xiii) Once filled with oil, charging the accumulators with Nitrogen gas can begin.
xiv) Close the valves.
xv)
Replace the gas bleed valve on the small (0.1 litre) accumulator.
xvi) Load gas into the small accumulator (about 1 MPa maximum).
xvii) Extend the strut by a distance of 40.8 mm by moving the actuator
downwards.
xviii) Load more gas into the small accumulator until the required static spring
force is reached on the actuator. For all the tests in this chapter, the
accumulator was loaded to 7.8 kN (or 4 MPa). This should be done slowly to
allow the gas to reach equilibrium temperature.
xix) Open all valves.
xx)
Replace the gas bleed valve on the big (0.4 litre) accumulator.
xxi) Load some gas into the big accumulator.
xxii) Extend the strut by a further distance of 119.2 mm by moving the actuator
downwards. This represents a total movement of 160 mm downwards.
xxiii) Load more gas into the big accumulator until the required static spring force
is reached on the actuator. For all the tests in this chapter, the accumulator
was loaded to 7.8 kN (or 4 MPa). This should be done slowly to allow the
gas to reach equilibrium temperature.
xxiv) Check to make sure that there are no gas or oil leaks.
xxv) The unit is now ready for testing or vehicle installation.
4.7.2 Bulk modulus
Normally in hydraulic applications, the oil is assumed to be incompressible. In
hydropneumatic suspension systems, ignoring the compressibility of the oil can result in
significant errors. The compressibility effect is aggravated by the fact that there is always
air present in the oil. Air is entrapped in the oil during filling due to mixing and diffusion.
Air also gets trapped in channels in the valve block, behind seals and o-rings and in the
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.22
valves. One possible way of reducing this problem might be to remove air by using a
vacuum pump. The current method of filling the unit is to pour oil slowly into the strut,
giving enough time for air to escape. The strut is also moved slowly in all directions in an
attempt to remove all the trapped air. This procedure may take several hours before all air
bubbles disappear and even then there is a good possibility of air still trapped in the
system. The total volume of oil required to fill the rear strut using this method was
measured to be 1.6 litres.
The bulk modulus of the fluid is given by:
β=
where:
β
=
∆P =
∆V =
V
=
∆P
⎛ ∆V ⎞
⎟
⎜
⎝ V ⎠
(4.1)
Bulk modulus of the fluid [Pa]
change in pressure of the fluid between two conditions [Pa]
change in volume of the fluid between the same two conditions [m3]
total volume of fluid in the system at atmospheric pressure [m3]
To determine the bulk modulus of the oil in the strut, the accumulators are blocked with
steel spacers so that the accumulator pistons cannot move (no gas in accumulators). The
strut is then compressed slowly whilst the force and relative displacement is measured.
Force and displacement is converted to pressure and volume by using the piston area (see
Figure 4.19). The pressure initially stays almost constant until the spacers in the
accumulators start compressing. At this point the accumulators become solid and only the
oil is compressed. This assumes that the strut itself is incompressible which is a good
assumption in this case.
The value for the bulk modulus measured on the strut is 1.368 GPa as shown in Figure
4.19. This compares favourably with typical values of bulk modulus of 1.4 GPa for
hydraulic oil (Poley, 2005).
4.7.3 Thermal time constant
The thermal time constant is a measure of the heat transfer coefficient between a gas in a
closed container and its surroundings (Els and Grobbelaar, 1993). In the case of the
hydropneumatic suspension system, it is determined experimentally by displacing the
strut with a step input displacement at the highest possible velocity. During the step, the
gas is compressed adiabatically (i.e. there is no time for heat transfer between the gas and
its surroundings). The temperature will rise and then slowly return to the ambient value.
On the other hand, in the case of a rebound step input, the gas will expand. The
temperature will drop first and then rise to the ambient value. The time required for the
temperature to change by 63%, between the initial value (immediately after the step) and
final ambient value, is defined as the thermal time constant (τ).
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.23
Figure 4.19 - Measured bulk modulus
Strictly speaking the thermal time constant is defined in terms of temperature. Measuring
temperature fluctuation accurately at high speed is very difficult. If the ideal gas
assumption is valid, then the time constant can be obtained from pressure or force
measurements. The pressure in the strut is measured vs. time as indicated in Figure 4.20.
The strut must be kept stationary both before and after the high-speed step input. The
thermal time constant was measured with no damper packs in the system (i.e. free flow
dampers). The experimentally determined thermal time constants for Prototype 2 are
shown in Table 4.1 for three different test conditions. The values given for the soft spring
are the combined time constant for both accumulators, while the stiff spring results are for
the small accumulator only. The thermal time constants for compression and rebound
compare well for each test, but the values depend significantly on the displacement of the
step. Els (1993) however indicates that the analyses is fairly insensitive to the value of the
thermal time constant and differences as large as 30% still result in acceptable
predictions.
Table 4.1 – Thermal time constants
File name
TYD1 - Compression
TYD1 – Rebound
TYD2 - Compression
TYD2 – Rebound
TYD3 – Compression
TYD3 - Rebound
Spring
setting
Size of step input
[mm]
Pbegin
[MPa]
Pend
[MPa]
Soft
Soft
Soft
Soft
Stiff
Stiff
25
25
50
50
25
25
5.07
4.30
5.99
4.11
8.83
3.43
4.91
4.45
5.55
4.42
6.57
4.23
∆P
[MPa]
0.16
0.15
0.44
0.31
2.26
0.80
63% point
[MPa]
4.97
4.39
5.71
4.31
7.41
3.93
τ
[s]
10.1
9.9
7.1
7.1
4.8
4.85
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.24
Figure 4.20 - Determination of thermal time constant
4.7.4 Spring characteristics
The two spring characteristics are determined by displacing the actuator slowly with a
triangular wave input displacement with a frequency of 0.001 Hz or a period (duration) of
1000 s. This means that the strut is first compressed from the static position to maximum
compression at a constant speed. The strut is then extended to the maximum rebound
position, again at constant speed and finally returned to the static position at constant
speed. This sequence is repeated for typically three cycles, although the graphs in the rest
of this chapter only show data for typically one cycle. Figure 4.21 displays the soft spring
characteristic measured for one complete compression and rebound cycle lasting 300
seconds. The measured value is compared to the predicted isothermal spring characteristic
calculated for a static gas volume of 0.5 litres. Excellent correlation is observed. The
small hysteresis loop in the measured characteristic can be attributed to heat transfer
between the gas and the surroundings. This effect is well documented by Els (1993).
Other possible contributing factors are seal friction and hysteresis in the test frame.
Figure 4.22 indicates the stiff spring characteristic measured for a complete compression
and rebound cycle. The displacement cycle starts in the static position, compresses the
spring to -62 mm, extends the spring to +75 mm, compresses the spring again to -62 mm
and then returns to the static position. This cycle lasts 1000 seconds. The measured value
is again compared to the predicted isothermal spring characteristic, but in this case there
is a significant discrepancy between measured and predicted results. The hysteresis loop
in the measured characteristic can again be attributed to heat transfer between the gas and
the surroundings, friction and hysteresis in the test frame. Further investigation indicated
that the discrepancy in the stiff spring characteristic could partly be attributed to the
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.25
compressibility of the oil (usually deemed negligible). Figure 4.23 indicates a straight line
corresponding to the bulk modulus of 1.368 GPa determined in paragraph 4.7.2. The
compressibility is significant for the stiff spring characteristics and needs to be taken into
account during spring calculations. The figure also indicates the very good correlation
achieved when the spring characteristic is corrected using the bulk modulus. The
correlation is however achieved with a static gas volume of 0.13 litres and not the 0.1
litres expected. Several sets of tests were performed on Prototype 2 where the damper
configuration was changed. This meant that the unit had to be discharged and recharged
every time. At the beginning of each new test series, the spring characteristics were
measured. Significant variations in actual gas volume were found when measured
characteristics were compared to predicted values. This re-iterates the fact that the oil
filling and gas charging procedures are extremely important and still needs improvement
to limit the errors due to static gas volume discrepancies.
Figure 4.21 - Soft spring characteristic
Figure 4.23 indicates measured isothermal characteristics for both the soft and stiff
springs.
4.7.5 Damping characteristics
The hydraulic damper characteristics of the suspension unit consists of different
components, the most important of which are:
i)
ii)
iii)
Pressure drops over valve block channels and ports
Pressure drops over valves (partially and fully open)
Pressure drops over hydraulic damper packs
These pressure drops are dependent on the flow rate through the various components.
Measuring these characteristics on the prototype is very difficult because of all the
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.26
possible combinations and the fact that it is very difficult to isolate specific components
to determine their individual contributions. In most instances it is impossible to measure
or calculate the flow through a specific component as the flow is often split between the
damper, the bypass valve, and the two accumulators. For these reasons the discussion that
follows does not attempt to give exact values for individual components, but rather to
give a better understanding of all the interactions and the orders of magnitude. This
explains why most of the graphs indicate pressure drops against strut speed and not flow.
By closing off the large accumulator with valve 3 (see Figure 4.7), all the flow is forced
into the small accumulator. The flow into the small accumulator can now be calculated by
multiplying the speed with the piston area. Figure 4.24 indicates four different lines for
the pressure difference (P1–P2) against flow. The data for “Valve block channel only” was
measured on Prototype 2 with no damper packs installed in the unit, i.e. the only flow
resistance was that of the valve block. “Valve block channel and valve” was measured
with solid damper packs in the unit, i.e. with all oil flowing through the valve. Also
indicated is the data measured by De Wet (2000) under steady state conditions on a
hydraulic test bench, and the valve manufacturer’s specification. All these values
correlate exceptionally well, especially if taken into account the variation in test
conditions and hydraulic oil used. Curve fits through the data are indicated in Figure 4.25.
As expected the pressure drop is proportional to the square of the flow rate. Values for
both flow directions are also very similar. These curve fits can be used in the
mathematical model.
Figure 4.26 indicates the effect of a single valve in the V3 position as well as for two
identical valves in parallel. It is clear that the concept of two valves in parallel works very
well as the pressure drop is significantly reduced. The ratio between the two graphs is not
exactly a factor of two due to the fact that the ports and channels connecting the two
valves in parallel are not identical.
The most important damper characteristic, as far as vehicle dynamics is concerned, is the
force velocity relationship of the high damping and low damping characteristics
respectively as measured using a triangular wave displacement input at various
frequencies. Figure 4.27 indicates this relationship for the stiff spring with low damping
(V3 closed and V1 open), stiff spring with high damping (V3 and V1 closed) as well as the
soft spring with low damping (all valves open). The damper packs in the strut were
sourced from standard Land Rover rear dampers. Also indicated on the graph is the
baseline Land Rover Defender 110 rear damper characteristic, correctly scaled for the
new application as explained below. The baseline graph is included as an indication of
what could theoretically be expected in the case of the stiff spring with high damping.
The baseline graph is scaled because the standard piston diameter is 35 mm and the piston
diameter on Prototype 2 is 50 mm. For the same linear velocity, the flow in Prototype 2
will be higher than that on the baseline Land Rover damper by a factor of
(0.050)2/(0.035)2 or 2.04. The force on Prototype 2 will also be higher than the force on
the baseline Land Rover damper for the same pressure difference across the damper pack,
also by a factor of 2.04. The Land Rover damper characteristic can therefore be scaled for
Prototype 2 by multiplying the force by 2.04 and dividing the velocity by 2.04. It can be
seen from Figure 4.27 that there is some discrepancy between the expected and measured
characteristics. In the low speed region the Prototype 2 forces are lower than expected.
This is attributed to leakage past the o-ring seals that mount the damper packs into the
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.27
cavity of Prototype 2. At higher speeds the Prototype 2 damping force is higher. This can
be expected due to the extra flow losses through the valve block ports and channels.
Figure 4.22 - Stiff spring characteristic
Figure 4.23 - Soft and stiff spring characteristics
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.28
Pressure drop over valve and ports
2.0E+06
1.5E+06
Pressure Drop [Pa]
1.0E+06
5.0E+05
0.0E+00
-5.0E+05
-1.0E+06
-1.5E+06
-2.0E+06
-80
-60
-40
-20
0
20
40
60
80
Flow rate [l/min]
Valve block channel only
Valve block channel and valve
Measured by De Wet (2000)
Hydraforce specification
Figure 4.24 - Pressure drop over valve 1
Pressure drop over valve and ports
2.0E+06
2
y = 425.7x - 3074.2x
2
R = 0.9902
1.5E+06
Pressure Drop [Pa]
1.0E+06
2
y = -148.92x + 295.38x
2
R = 0.9977
5.0E+05
2
y = 154.19x + 1213.6x
2
R = 0.9975
0.0E+00
-5.0E+05
-1.0E+06
2
y = -331.83x - 257.96x
2
R = 0.9993
-1.5E+06
-2.0E+06
-80
-60
-40
-20
0
20
40
60
Flow rate [l/min]
Valve block channel only
Valve block channel only
Poly. (Valve block channel and valve)
Valve block channel and valve
Poly. (Valve block channel and valve)
Poly. (Valve block channel only)
Figure 4.25 - Curve fits on pressure drop data
Valve block channel and valve
Poly. (Valve block channel only)
80
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.29
2.5E+06
2.0E+06
Pressure drop (P2 - P3) [Pa]
1.5E+06
1.0E+06
5.0E+05
0.0E+00
-5.0E+05
-1.0E+06
-1.5E+06
-2.0E+06
-0.8
-0.6
-0.4
-0.2
0.0
0.2
0.4
0.6
0.8
Speed [m/s]
Single valve
Two valves in parallel
Figure 4.26 - Pressure drop over valve 3 (single valve vs. 2 valves in parallel)
20
15
Force [kN]
10
5
0
-5
-10
-15
-0.8
-0.6
-0.4
-0.2
0
0.2
0.4
0.6
Speed [m/s]
stiff spring - low damping
stiff spring -high damping
soft spring - low damping
Baseline Land Rover rear damper - scaled
Baseline Land Rover rear damper - unscaled
Figure 4.27 - Damper characteristics for Prototype 2
0.8
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.30
4.7.6 Valve response times
Valve response times are very important for predicting the transient response of the
system to valve switching. A typical trend of pressure drop over the valve vs. time is
shown in Figure 4.28. The solenoid switching signal is indicated on the same graph. To
obtain the valve response time, the initial pressure difference (before switching) and the
final pressure difference (after the transient response has died away) is determined. Two
values (represented by horizontal lines) are calculated representing a 5% change and a
95% change in pressure difference respectively. This is done in order to define the
switching points more precisely as the exact moment where the change occurs is very
difficult to determine. The time from the solenoid switching signal to the 5% change
point is defined as the initial delay. This is the time required for the solenoid to build up
enough force so that the valve plunger starts moving. The time between the 5% and 95%
point is defined as the transient response time of the valve and represents the time
required from the initial plunger movement until the valve is fully open. The total valve
response time is the sum of the initial delay and the transient response time as indicated in
Figure 4.28.
The valve response times were measured for all 4 four valves. The damper orifices were
blocked so that all the flow was channelled through the valves. The valve response time
was measured by closing the respective valve, compressing the strut until the required
pressure difference was obtained, and then opening the valve. This resulted in flow
through the valve until the pressure in the system stabilized. The procedure was repeated
in the opposite direction, e.g. closing the valve and extending the strut before opening the
valve.
Figures 4.29 and 4.30 give the valve response time (initial delay, transient response time
and total response time) as a function of pressure drop across the valve for Prototypes 1
and 2 respectively. Prototypes 1 and 2 were both fitted with the same valves, although
Prototype 1 used 24 Volt solenoids. This was changed to 12 Volt solenoids on Prototype
2 to be compatible with the test vehicle’s electrical system.
The valve response time is to some extent dependant on the system (Janse van
Rensburg, Steyn and Els (2002). All four valves in Prototype 2 are fitted in different
positions in the valve block with the result that the channels to these valves are all
different. Valve response times are also dependent on the pressure difference across the
valve as can be seen in the figure. The valve response time varies from 40 to 100
milliseconds over the pressure range of interest and is acceptable for the current
application.
4.7.7 Friction
The isothermal spring characteristic was determined by slowly compressing the spring
through its operating range whilst recording force, displacement and pressure. Figure 4.31
indicates the spring force against spring displacement for the soft spring on Prototype 1.
Two curves are shown namely the force measured by the load cell, and the force
calculated from the pressure data. The measured force shows unacceptable levels of
hysteresis, while the force calculated from the pressure measurement gives the expected
characteristic.
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.31
24
Total response time
Final value
0
Response time
95% Change
19
Pressure [MPa]
-2
14
-4
Initial delay
9
-6
5% Change
4
-8
Initial value
-10
91.15
91.17
91.19
91.21
91.23
91.25
91.27
91.29
91.31
91.33
-1
91.35
Time [s]
Presure drop across valve
Switching signal
Figure 4.28 - Explanation of valve response time definitions
During initial assembly of the unit, it was found that the main cylinder could be moved
easily by hand, while the accumulator pistons had to be moved using compressed air.
There was no way to move the accumulator pistons by hand. The hysteresis was therefore
attributed to seal friction (stick-slip) in the accumulator seals. After considerable research,
two new accumulator pistons were designed and manufactured using wear rings
combined with a state-of-the-art accumulator seal (Turcon AQ Seal5) with negligible
stick slip. The original design used a fairly basic seal layout with a double o-ring and
back-up ring system.
After testing the more advanced sealing concept in the suspension system, it was found
that the hysteresis had improved only marginally. Careful investigation traced the
problem to the bending moment applied to the main cylinder due to the offset of the
chassis mounting arrangement used on Prototype 1. This results in a high side force
between the main cylinder and the piston, causing unacceptable friction and wear.
Figures 4.32 to 4.35 illustrate that friction in Prototype 2 is very low and should not cause
any serious problems. Friction may however degrade the vibration isolation of the system
for small road inputs.
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.32
100
90
80
Total response time = 0.4212(delta P)2 + 0.2038(delta P) + 41.908
70
Time [ms]
60
50
40
30
20
10
0
-12
-10
-8
-6
-4
-2
0
2
4
6
Pressure difference [MPa]
Initial delay
Transient response time
Total Time
Parabolic curve fit on total time
Figure 4.29 - Valve response time for Prototype 1
140
Response Time [milliseconds]
120
100
80
60
40
20
0
-4
-2
0
2
4
6
8
10
Pressure Difference [MPa]
Valve 1 - init delay
Valve 1 - resp time
Valve 1 - tot time
Valve 2 - init delay
Valve 2 - resp time
Valve 2 - tot time
Valve 3 - init delay
Valve 3 - resp time
Valve 3 - tot time
Valve 4 - init delay
Valve 4 - resp time
Valve 4 - tot time
Figure 4.30 - Valve response time for Prototype 2
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.33
Figure 4.31 - Hysteresis problem on Prototype 1
4.8
Mathematical model
A SIMULINK® model of the suspension unit was developed by Theron (Theron and
Els, 2005). This model takes the deflection rate of the suspension unit as input and
employs simple fluid dynamics theory in an iterative manner to calculate the flow rates
from each accumulator to the cylinder. Iteration takes place until pressure balance in the
parallel branches is established. The model then calculates the pressure in the two
accumulators by solving the energy equation for an ideal gas in an enclosed container
(Els and Grobbelaar, 1993) and time integrating the flow rates to determine the gas
volumes in the two accumulators. The model renders the dynamic force generated by the
suspension unit as output.
Physical tests have been performed on Prototype 2, where the spring characteristics,
damper characteristics and valve dynamics have been measured. These tests were
described in previous paragraphs. Generally, good correlation exists between the results
of the SIMULINK® model and the experimental data measured in the laboratory on the
prototype suspension unit. A number of aspects, where the model or the quantification of
its parameters needs improvement, were identified.
The aim was to develop a mathematical model that can be used in vehicle dynamic
simulations and to investigate suitable control strategies for semi-active switching of the
spring and damper.
4.8.1 Modelling philosophy
In developing a mathematical model, a tension force in the unit is considered positive,
while a compressive force is negative. Any extension of the unit relative to a reference
state is considered as a positive (relative) displacement and compression of the unit as
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
Figure 4.32 - Effect of friction on soft spring at low speeds
Figure 4.33 - Effect of friction on soft spring at high speeds
4.34
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
Figure 4.34 - Effect of friction on stiff spring at low speeds
Figure 4.35 - Effect of friction on stiff spring at high speeds
4.35
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.36
negative displacement. An extensional speed is considered positive and a compression
speed as negative.
For the purposes of vehicle dynamics simulation a mathematical model of this unit is
required that calculates the combined spring-damper force for a certain set of valve
settings and a given state of displacement and speed. One may therefore consider the
force of the suspension unit as the output of the model and the valve settings of the three
valves and the displacement and speed of the unit as the inputs to the model, where the
model calculates the output for given inputs. This calculation is typically performed
within a time step in a simulation run and is repeated for each time step.
The working principle of the suspension unit is discussed in paragraph 4.2. Figure 4.7
indicates the various pressures, dampers and valves in the suspension system. The output
force of the unit is essentially directly related to the pressure P2 in the main strut
cylinder. This pressure depends on the pressures in the two accumulators, the flow
through and corresponding pressure drops over the two dampers with corresponding
channels and the valve switching. The pressure in the accumulators depends on the
volume of oil in the accumulators, which is related to the displacement of the suspension
unit and the state of valve 3. An alternative way of looking at the volume of oil in the
accumulators is to realise that this is determined by the flow history, i.e., these volumes
may be determined by integrating the flow rates in the two main branches of the system.
Using this approach makes the mathematical model independent of the displacement of
the unit as an input. This is indeed the approach that was used in modelling the unit. The
input to the model of the suspension unit is therefore, in addition to the three valve switch
signals, only the extensional speed x& of the unit. From this the volume flow rate q = Ax&
into the main strut cylinder, of cross sectional area A , can directly be calculated. The
flow rates in the two branches are taken as qi , i = 1,2 , for the branch associated with
accumulator i , positive in the direction from the accumulator towards the main strut.
4.8.2 Pressure dependent valve switching
It is assumed that the electric signals with which the various valves are switched changes
instantaneously from low to high values, or vice versa. When this happens, valve and
other dynamics prevent immediate pressure and flow changes. These dynamic effects are
not currently modelled mathematically, but are taken into account empirically. The valve
response time was defined and determined in paragraph 4.7.6 (Figures 4.29 and 4.30).
The parabolic curve indicated in Figure 4.29 (although determined for Prototype 1) was
subsequently employed in the mathematical model with respect to all three valves and for
both prototypes.
Wherever the state of the valve is taken into account in the model, a fraction f i between
zero and one is used, where the subscript i = 1,2,3 indicates the valve number. For
switching on the valve (electric signal going from low to high, valve going from closed to
open) f i = 0 before the electrical signal switches, f i = 0.05 at half the valve response
time after the electrical signal switches, f i = 0.95 at the valve response time and fi = 1
after 1.5 times the valve response time. In between these time points a piecewise cubic
Hermite interpolation is used to calculate the fraction. For switching off the valve the
same type of interpolation is used on the reversed sequence.
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.37
4.8.3 Pressure drop over dampers and valves
Due to the complexity of possible dampers that may be used in the suspension unit, it was
decided to use table look-up techniques to get the pressure drop over the damper for a
given flow rate through the damper. Quite often the pressure-flow characteristics display
significant hysteresis. For now the table look-up procedure employed does not provide
for possible hysteresis. The pressure drop over the damper was measured, with the bypass valve both open and closed, for various positive and negative flow rates in a
practically realistic range. This measured data was used to establish a high damping and
a low damping damper curve, corresponding to the by-pass valve being closed and open,
respectively. These curves are used in the table look-up procedure for both dampers 1
and 2, since they currently are identical and their by-pass valves are also identical. The
fact that the internal passages in the valve block for the two dampers at this time are not
identical is neglected in the model.
For a certain flow rate qi the pressure drop over damper i with i = 1,2 , is calculated as
∆Pdi = f i ∆Pdoi + (1 − f i )∆Pdci , where ∆Pdci is the pressure drop interpolated at qi from the
high damping graph of damper i , while ∆Pdoi is the pressure drop interpolated at qi from
the low damping graph of damper i .
When valve 3 is fully open ( f 3 = 1 ), the pressure drop over valve 3, ∆Pv 3 , is calculated
using an experimentally determined loss factor and the flow q 2 . When the valve is
opening ( 0 < f 3 < 1 , f 3 increasing), a value ∆Pv 3o is calculated in exactly the same way as
∆Pv 3 above, but the actual pressure drop over the valve is taken as
∆Pv 3 = f 3∆Pv 3o + (1 − f 3 )∆Pv 3i , where ∆Pv 3i is the pressure drop over the valve before the
switching started. When the valve is closing ( 0 < f 3 < 1 , f 3 decreasing), on the other
hand, the actual pressure drop over the valve is taken as ∆Pv 3 = f 3∆Pv 3o + (1 − f 3 )∆Pv 3e ,
where ∆Pv 3e is the pressure drop over the valve calculated for the scenario where all
variables are at their current values except q = q1 and q 2 = 0 , i.e., as if valve 3 is fully
closed.
4.8.4 Flow and pressure calculation
The mathematical model is essentially based on the assumption that the hydraulic fluid is
incompressible. In the simulation, however, the compressibility of the fluid is taken into
account as a refining correction in the calculation of the gas volumes in the accumulators.
This correction is based on the various major volumes of fluid in the system, each at its
respective pressure, and the bulk modulus of the hydraulic fluid (see paragraph 4.7.2).
Whenever valve 3 is closed, the system can be modelled as a third order non-linear state
space system; otherwise a fourth order non-linear state space system with an algebraic
constraint is obtained. These two alternative situations will now be considered separately.
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
i)
4.38
Valve 3 closed
When valve 3 is closed, q = q1 and q 2 = 0 , due to the assumed incompressibility of the
hydraulic fluid. Let the volume of gas in accumulator i be V gi . The rate of change in the
gas volume in accumulator 1 is
V&g1 = q1 .
(4.2)
The pressure Paccui in the accumulator i is calculated using the ideal gas law
Paccui = mi RTi / V gi = RTi / vi ,
(4.3)
where: mi is the mass of gas with which the accumulator is charged, R = 296.797 is the
gas constant for Nitrogen, vi = V gi / mi is the specific volume and Ti is the absolute
temperature of the gas in the accumulator. ( Paccui = P1 for accumulator 1 and Paccui = P4
for accumulator 2.) Ti is calculated by solving the following differential equation, as
suggested by Els (1993) and Els and Grobbelaar (1993):
T − Ti
Ti ⎛ ∂Paccui ⎞
⎜
⎟ v&i
−
T&i = i 0
(4.4)
cv ⎜⎝ ∂Ti ⎟⎠ v
τi
where Ti 0 is the initial gas temperature, in this taken as the ambient temperature, τ i is the
thermal time constant of the accumulator and c v is the specific heat at constant volume of
the gas. The thermal time constant is taken at experimentally determined values of
4.8 seconds for both accumulators (see paragraph 4.7.3). Calculating the gas temperature
in this way means that if the gas is suddenly compressed, the model calculates the
pressure rise along an adiabatic compression curve, while the temperature rises.
However, if the gas is subsequently allowed to cool down, the model allows the pressure
to drop to the value indicated by the isothermal compression curve.
From equation (4.3) it follows that
∂Paccui R
=
.
∂Ti
vi
Substituting this in equation (4.4) renders
T −T
Ti R
T&i = i 0 i −
qi = fT i (Ti ,Vgi , qi ) ,
cv Vgi
τi
(4.5)
(4.6)
where the fT i (Ti ,Vgi , qi ) on the right hand side indicates that T&i is a function of the
variables Ti , qi and V gi . Equation (4.6) is non-linear due to the appearance of the
product of these variables.
Since q 2 = 0 , there is no change in the gas volume in accumulator 2. The pressure in this
accumulator may however still change, as the gas temperature may change. The third
order system is thus defined by the three differential equations, equation (4.2) and
equation (4.6) for i = 1,2 . Within a simulation time step, in addition to these three
differential equations, various other variables are calculated (for example, the
accumulator pressures with equation (4.3)). There are, however, no algebraic equations
that need to be solved simultaneously with the three differential equations, and the
solution is therefore fairly straightforward.
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.39
Once q1 for the current time step has been calculated, the pressure P2 in the main strut
cylinder is calculated by calculating ∆Pd 1 as described in section 4.8.3 above, and then
P2 = P1 − ∆Pd 1 . With P2 known the output of the model is simply calculated by
multiplying this pressure with the negative of the main strut cross sectional area.
ii)
Valve 3 open, opening or closing.
When valve 3 is partially or fully opened, the flow rate q 2 is no longer zero. Due to the
assumed incompressibility of the hydraulic fluid, q = q1 + q 2 . The rate of change in gas
volume in accumulator 1 is still given by equation (4.2), while the rate of change in the
gas volume in accumulator 2 is
V&g 2 = q2
(4.7)
= q − q1
In this case, however, an additional algebraic equation needs to be solved simultaneously
with the differential equations. This equation may be considered as a constraint that
needs to be satisfied, namely that the pressure P2 in the main strut cylinder calculated
along two different paths must be the same. Let P21 be the pressure in the main strut
cylinder, calculated along the branch connecting this to accumulator 1 as outlined in
section 4.8.4(i) above (which for a given flow rate q1 is also valid in this case). P21 is
therefore a function of the flow rate q1 and the pressure P1 . The pressure P1 , by
equation (4.3), is a function of T1 and V g1 . Therefore P21 = P21 (q1 , T1 , V g1 ) . In a similar
way the pressure P22 in the main strut cylinder, calculated along the branch connecting
this to accumulator 2, may be calculated, by first calculating ∆Pd 2 and ∆Pv 3 at flow rate
q 2 = q − q1 , as described in section 4.8.3 above. Then, P3 = P4 − ∆Pd 2 and P22 = P3 − ∆Pv 3 .
The pressure P4 is a function of T2 and V g 2 , therefore, P22 = P22 (q, q1 , T2 ,V g 2 ) . The
algebraic constraint may then be written as:
0 = P21 (q1 , T1 , V g1 ) − P22 (q, q1 , T2 ,V g 2 ) .
(4.8)
Also, whereas equation (4.6) is still valid for accumulator 1, for accumulator 2 the flow
rate q 2 needs to be substituted with q − q1 , so that the system dynamics may be
summarized in the following non-linear state space representation:
⎡1
⎢0
⎢
⎢0
⎢
⎢0
⎢⎣0
f T 1 (T1 ,V g1 , q1 )
⎤
0 0 0 0⎤ ⎡ T&1 ⎤ ⎡
⎥
⎢ & ⎥ ⎢
⎥
f T 2 (T2 ,V g 2 , q, q1 )
1 0 0 0⎥ ⎢ T2 ⎥ ⎢
⎥
⎥ ,
q1
0 1 0 0⎥ ⎢V&g1 ⎥ = ⎢
⎥
⎥⎢ & ⎥ ⎢
q − q1
0 0 1 0⎥ ⎢V g 2 ⎥ ⎢
⎥
0 0 0 0⎥⎦ ⎢⎣ q&1 ⎥⎦ ⎢⎣ P21 (q1 , T1 ,V g1 ) − P22 (q, q1 , T2 ,V g 2 )⎥⎦
(4.9)
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.40
with input variable q and state variables T1 , T2 , V g1 , V g 2 and q1 . The flow rate q1 is not
truly a state variable, but it is convenient to consider it as such in order to write the four
differential equations and the algebraic constraint in a single equation as above.
The matrix on the left of equation (4.9) is often called a mass matrix. This equation is an
example of a so-called differential-algebraic equation, as the mass matrix is singular.
This singularity is clearly caused by the algebraic constraint.
4.8.5 Implementation in SIMULINK®
As mentioned above, the aim of this research was to develop a mathematical model of the
suspension unit, to be used in vehicle dynamic simulations. It was decided earlier to use
the ADAMS® program for the vehicle dynamics simulation. A very convenient way to
interface a mathematical model like that of the suspension unit as described above with an
ADAMS model of a larger system (in this case the vehicle and its suspension system
components other than the suspension units) is to implement the mathematical model in
the SIMULINK environment. ADAMS can be linked to MATLAB® SIMULINK subprograms. For this reason the mathematical model was implemented in SIMULINK.
MATLAB provides a solution scheme for differential-algebraic equations and as a
consequence SIMULINK has the ability to model algebraic constraints. Solution of the
differential-algebraic equation, equation (4.9), using this functionality has been
unsuccessful thus far. The mathematical model was however implemented successfully
in SIMULINK by, within each time step, first calculating the valve fractions f i , i = 1,2,3 ,
based on the pressure drops over the valves at the end of the previous time step and then
enforcing the algebraic constraint using a Newton-Raphson type iteration to find the
values of q1 , q 2 , P2 and P3 . After these values have been calculated, T1 , T2 , Vg1 and
Vg 2 are calculated by solving the four first order differential equations contained in
equation (4.9). Lastly P1 and P4 are calculated using equation (4.3). During the
Newton-Raphson type iteration the values of P1 and P4 at the end of the previous time
step are used. This iteration is performed in a MATLAB s-function that is called by the
SIMULINK program. Once P2 is calculated, the output force of the suspension unit for
the current time step may be calculated as F = − P2 A and the program may move on to
the next time step. It should be noted that the friction between the piston and the cylinder
walls and the piston rod and its bushing is neglected in the calculation of F .
4.8.6 Validation of the mathematical model
The model of the suspension unit has been validated by comparing its predicted force
output with forces measured on the Prototype 2 unit in a SCHENCK Hydropulse
hydrodynamic testing machine under displacement control.
During testing on the hydrodynamic testing machine, the displacement feedback signal
and resulting force as measured with a load cell were recorded. In addition to these two
signals, the signals from the four pressure transducers measuring pressures P1 to P4 and
the electric command signals for switching the valves were also recorded. All these
signals were filtered to prevent aliasing, digitised and stored on disc.
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.41
Comparing the load cell force and the pressure P2 measurements clearly showed that the
error made in the model by neglecting the friction on the sliding parts of the unit and
taking the output force of the unit as − P2 A , is not insignificant but generally quite small.
There is a second reason, other than friction, for the difference between the load cell force
and − P2 A , especially in situations of oscillation at high frequency. During vehicle
simulation, the inertial properties of the piston and piston rod should be combined with
those of the unsprung mass, so that the associated dynamic effects are taken into account
by the ADAMS model, rather than the SIMULINK model. The output of the SIMULINK
model should therefore be the suspension unit output force before the inertial effect of the
piston and piston rod has been taken into account. The load cell, however, measures the
suspension unit net output force after accelerating this mass. It is therefore prudent, in the
comparison of the mathematical model with the measured results, to compare the output
of the model in terms of measured and calculated − P2 A values. In the discussion that
follows all reference to measured force should be understood to mean force calculated
from the measured pressure P2 and thus neglects friction.
Since the mathematical model does not accept a displacement time history as input, but
rather the extensional speed time history, the measured displacement signal first had to be
differentiated with respect to time. It was always possible to bring the displacement
signal back to its initial value at the end of a test run. The differentiation was therefore
performed by transforming the whole displacement time history of a test run to the
frequency domain using a Fast Fourier Transform (FFT), then multiplying the resultant
double sided complex spectrum with jω , setting all values corresponding to frequencies
above a chosen low pass filter cut-off frequency and below the negative of this cut-off
frequency to zero and lastly back transforming the signal to the time domain using the
inverse FFT. (In this j = − 1 and ω is the circular frequency.) This procedure not only
performs the differentiation but also realizes a low pass filter with very sharp cut-off
properties and no magnitude and phase distortion below the cut-off frequency. During
vehicle simulation this differentiation of the displacement is not required, since the
ADAMS model directly calculates the required speeds.
To first test the spring properties without the influence of the dampers the suspension unit
was cycled through a triangular wave displacement at low speed, as indicated in
Figure 4.36. This figure also shows the output force of the suspension unit, as calculated
from the measured pressure P2, for the case of stiff spring and low damping properties.
Figure 4.37 shows the comparison between the measured and SIMULINK calculated time
histories for this case, for the pressure in the active accumulator, P1, and the main strut,
P2. Even though the nominal gas volume of accumulator 1 at the static wheel load was
designed to be 0.1 litres, during this simulation it was adjusted to 0.135 litres, in order to
obtain what was considered an acceptable correlation between the measured and
calculated results. This adjustment is to some extent justified due to the fact that the
volume calculation during design did not take into account some small cavities and screw
thread inside the accumulator, and it was also determined that it is rather difficult to fill
the suspension unit with oil without trapping small pockets of air inside the unit. The
volume of accumulator 1 could have been adjusted to an even higher value, to get an even
closer correlation between the measured and calculated results at the peak at 100 seconds
in Figure 4.37, but there also was evidence that valve 3 was prone to leak at a high
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.42
pressure differential, which may have caused a reduced pressure during the measurement.
The force-displacement graph obtained during the test that produced Figure 4.37 is shown
in Figure 4.38, once again comparing the measured and calculated results. This test was
repeated with a high damping setting and essentially the same results were obtained, as
expected, since the very slow speed renders very small damping.
Figure 4.36 – Measured input and output: stiff spring and low damping at low speed
Next a similar test was conducted but at considerably higher speeds, to generate a
significant damping effect. The input displacement and output force for a stiff spring and
low damping setting is shown in Figure 4.39. The comparison between the measured and
calculated time histories for this case, for P1 and P2, are shown in Figure 4.40 and the
force-displacement graph obtained during this test in Figure 4.41. The correlation
between measurement and calculation displayed in Figure 4.40 is generally good, except
at the high-pressure peaks. The calculated force-displacement graph shows an interesting
figure eight shape, which was not observed in the measurement nor any other simulation
result. When evaluating the force-displacement graphs generated by the simulation, one
needs to bear in mind that the model does not yet provide for hysteresis in the damping
properties. This may account for the strange curve calculated and displayed in
Figure 4.41.
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.43
Figure 4.37 – Comparison between measured and calculated values of P1 and P2: stiff
spring and low damping at low speed
Figure 4.38 – Comparison between measured and calculated force-displacement curve:
stiff spring and low damping at low speed
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.44
Figure 4.39 – Measured input and output: stiff spring and low damping at high speed
A third stiff spring low damping test was performed at a slightly lower speed but a higher
displacement stroke, as indicated in Figure 4.42. The comparison between the measured
and calculated time histories for this case, for P1 and P2, are shown in Figure 4.43 and the
force-displacement graph obtained during this test in Figure 4.44. The simulation indeed
indicated significantly higher pressures to accompany the higher displacement input, but
the measured pressures failed to reach the high values as expected. The clear kink in the
measured force-displacement graph in Figure 4.44 near –50 mm displacement is seen as a
clear indication of leakage, at high differential pressure, through valve 3.
Next the same kind of test as shown in Figure 4.39 was performed, only now with high
damping (i.e., high stiffness and high damping, triangular displacement excitation at high
speed). The input displacement and measured force time histories are shown in
Figure 4.45. It is clear that the output force is clipped at about zero Newton, and the
reason for this is that the pressure cannot drop very far below zero Pascal (atmospheric
pressure) because at lower pressures the oil starts to boil preventing further pressure drop.
In any case, the pressure cannot drop below zero absolute, which would correspond to a
positive output force of merely 196 N. The time histories of the pressures P1 and P2 are
shown in Figure 4.46. The SIMULINK model has been constructed such that pressure P2
will only drop to zero. It is seen that while P2 is dropping, the model follows the
measurement quite well into the saturation at zero. The model, however, recovers from
this more quickly that the actual physical unit. This causes the calculated pressure to start
rising significantly earlier on the compression stroke than the measured pressure. After a
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.45
Figure 4.40 – Comparison between measured and calculated values of P1 and P2: stiff
spring and low damping at high speed
Figure 4.41 – Comparison between measured and calculated force-displacement curve:
stiff spring and low damping at high speed
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.46
Figure 4.42 – Measured input and output: stiff spring and low damping at high speed,
larger displacement stroke
Figure 4.43 – Comparison between measured and calculated values of P1 and P2: stiff
spring and low damping at high speed, larger displacement stroke
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.47
Figure 4.44 – Comparison between measured and calculated force-displacement curve:
stiff spring and low damping at high speed, larger displacement stroke
delay, the calculation and the measurement meet up again with good correlation until this
is repeated in the next cycle. One possible explanation for this delay is that in the
physical unit some boiling of the oil at low pressure occurs, a phenomenon that is not
provided for in the SIMULINK model. Oil vapour caused by boiling and suspended in
the oil is expected to cause a delay in pressure rise on compression. In this case, during
the low-pressure part of the P1 cycle, the correlation between simulation and
measurement is not as good as observed in the results discussed earlier. This may be
related to the suspected boiling of the oil. The poor correlation in both the P1 and P2
results is not of serious concern, as the situation where the suspension unit is subjected to
a prescribed high speed rebound that can cause P2 to drop to zero, even though easy to
create on a test bench, is highly unlikely with the unit installed in a vehicle, even under
rough road conditions. There is simply no downwards pull on the wheel available to
cause such a condition. The force-displacement graph generated for this test is shown in
Figure 4.47.
Whereas all the results discussed above pertain to stiff spring scenarios, with valve 3
closed, the more complicated part of the model corresponds to the soft spring scenario.
Figure 4.48 shows input displacement and output force measured for a soft spring and
low damping case, at low speed. Figure 4.49 shows the comparison of the measured and
calculated time histories of the two accumulator pressures P1 and P4, while Figure 4.50
shows the same for the pressures P2 and P3. The force-displacement curve is shown in
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.48
Figure 4.51. In this case the gas volumes of accumulator 1 and 2 during the simulation
were taken as 0.135 and 0.4 litres, respectively. Correlation is generally acceptable.
Figure 4.45 – Measured input and output: stiff spring and high damping at high speed
Next the above test was repeated at high speed, the displacement input and measured
force output shown in Figure 4.52. The comparison of the measured and calculated time
histories of P1 and P4 for this case is shown in Figure 4.53 and that of P2 and P3 in
Figure 4.54, with the force-displacement curve in Figure 4.55. Once again the correlation
is generally acceptable.
Lastly, a test was performed on the suspension unit wherein it was compressed some
distance in the stiff spring mode, then kept at this displacement for a while, after which
valve 3 was opened and the pressures in the system allowed to equalize. Valve 3 was
then closed again and the unit was then further compressed. This was repeated twice after
which the unit was extended in a similar stepwise manner. This procedure, referred to
herein as the incremental compression test, is well illustrated in Figure 4.56, which shows
the time histories of the input displacement, the measured output force and the switch
signal for valve 3. The switch signal is not plotted against a specific scale; it merely
indicates when the valve is open (high) or closed (low). This whole test was conducted
with a low damping setting. The measured and calculated time histories of the pressure in
the two accumulators are shown in Figure 4.57, while the time histories of P2 and P3 are
shown in Figure 4.58. In this case the gas volumes of accumulator 1 and 2 during the
simulation were taken as 0.111 and 0.4 litres, respectively. The change in the volume of
accumulator 1 may be justified by the fact that the suspension unit was emptied of both
gas and oil, and then refilled, between this test and the test described earlier. With these
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.49
Figure 4.46 - Comparison between measured and calculated values of P1 and P2: stiff
spring and high damping at high speed
Figure 4.47 – Comparison between measured and calculated force-displacement curve:
stiff spring and high damping at high speed
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.50
Figure 4.48 - Measured input and output: soft spring and low damping at low speed
Figure 4.49 – Comparison between measured and calculated values of P1 and P4: soft
spring and damping at low speed
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.51
Figure 4.50 – Comparison between measured and calculated values of P2 and P3: soft
spring and low damping at low speed
Figure 4.51 - Comparison between measured and calculated force-displacement curve:
soft spring and low damping at low speed
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.52
Figure 4.52 - Measured input and output: soft spring and low damping at high speed
Figure 4.53 – Comparison between measured and calculated values of P1 and P4: soft
spring and low damping at high speed
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.53
Figure 4.54 – Comparison between measured and calculated values of P2 and P3: soft
spring and low damping at high speed
Figure 4.55 – Comparison between measured and calculated force-displacement curve:
soft spring and low damping at high speed
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.54
settings, the only correlation that does not seem good is that between the measured and
calculated time histories of P4. It should however be realized that while valve 3 is closed,
P3 and P4 should practically be identical, as there is no flow through damper 2 or its bypass valve. If the measured values of P4 and P3 from Figures 4.49 and 4.50 are compared,
during the first second, when valve 3 is indeed closed, it is seen that the P4 pressure
transducer reads a pressure slightly higher that the P3 transducer, by the same amount as
the difference in the measured and calculated P4 values in Figure 4.57. If based on this
observation it is assumed that an offset was present in the P4 measurement, the correlation
between the measurement and the simulation result may be considered as very good. The
measured and calculated force-displacement graphs for this test are shown in Figure 4.59.
This figure also shows a very good correlation between measurement and simulation.
It is also worth noting that the slow drop in pressure P1 right after achieving the local
peaks at the end of the compression strokes in Figure 4.57, just before valve 3 is opened,
is predicted quite well by the model. Since the displacement input does not vary in this
period, it is evident that the cooling of gas in accumulator 1 causes this pressure drop.
This effect is captured adequately in the model by the use of equation (4.4).
Since the displacement of the suspension unit is not taken as an input in the mathematical
model, it is necessary to check that the displacement of the unit that would be mandated
by the solution of the differential equations like equation (4.2) does in fact correspond to
the actual displacement experienced by the unit. During all the tests described in this
section this was in fact checked and the correlation was exceptionally good. At this time
it is proposed that a similar check should be incorporated in an implementation of the
SIMULINK model within an ADAMS simulation of vehicle dynamics.
To date no measurement was done to specifically validate the way that the pressure
dependent valve switching was implemented in the mathematical model.
4.9
Conclusion
A prototype four-state semi-active hydropneumatic spring-damper system (4S4) has been
designed, manufactured, characterised on a test rig and modelled mathematically.
The design meets all the initial specifications and can be fitted to the proposed test
vehicle without major modifications to the test vehicle.
The manufactured prototypes (Prototypes 1 and 2) have been extensively tested and
characterised on a SCHENCK hydropulse actuator. Although several problems have been
identified on Prototype 1, these have been adressed and eliminated on Prototype 2.
Prototype 2 meets all the dynamic requirements.
A mathematical model of the suspension unit was developed and implemented in
SIMULINK. Agreement between the model predictions and the measurements was
generally good. Some aspects where the model or the quantifying of its parameters need
improvement were identified. In particular, the tests to date clearly identified the need for
an accurate method of quantifying the mass of gas loaded into the two accumulators.
Further work will be done on testing the model within simulations of a full vehicle
equipped with these suspension units.
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.55
Figure 4.56 – Measured input and output, and valve 3 switch signal: incremental
compression test with low damping
Figure 4.57 – Comparison between measured and calculated values of P1 and P4:
incremental compression test with low damping
THE FOUR-STATE SEMI-ACTIVE SUSPENSION SYSTEM (4s4)
4.56
Figure 4.58 – Comparison between measured and calculated values of P2 and P3:
incremental compression test with low damping
Figure 4.59 – Comparison between measured and calculated force-displacement curve:
incremental compression test
Chapter
5
THE RIDE COMFORT VS. HANDLING DECISION
This chapter describes and analyses various methodologies that can be used to make the
decision whether the suspension should be set to “ride comfort mode” or “handling
mode”. This is referred to as the “ride comfort vs. handling decision”. It does not
attempt to discuss or investigate possible control strategies for ride comfort and/or
handling respectively. It rather assumes that these characteristics and control methods are
known, i.e. that a set of “optimal” suspension characteristics and/or control laws exist for
both ride comfort and handling. These two sets of conditions are in conflict as described
in chapter 2. The importance of the ride comfort vs. handling decision cannot be
overemphasized, as it is a safety critical decision. If the suspension system for example
switches to the “ride comfort” mode during a severe handling or accident avoidance
manoeuvre, the consequences might be severe and loss of control or rollover might result.
For the purposes of this study, the 4S4 will be switched to the soft spring and low
damping characteristics when ride comfort is required and will switch to high damping
and the stiff spring when handling is required. The effects of ride height on ride comfort
and handling is excluded from the analysis. All the analyses will be made with the
suspension set to the same ride height as the baseline suspension system. Effects caused
by acceleration (e.g. squat) or braking (e.g. dive) are neglected at present. Figure 5.1
indicates where the ride comfort vs. handling decision fits into the study.
The aim of the present chapter is to find a strategy that uses parameters that can be
measured directly, or otherwise easily calculated from direct measurements. This
excludes the use of state estimators, integrators and artificial intelligence techniques such
as neural networks.
No literature was found that is directly applicable to the ride comfort vs. handling
decision as applied to off-road vehicles or controllable springs, although some of the
concepts proposed by different authors are worth exploring and will be discussed now.
5.1
Literature
Stone and Cebon (2002) investigate semi-active roll control of a heavy vehicle. They
make use of a system where an anti-rollbar is connected to the vehicle body with
hydraulic cylinders providing switchable roll stiffness. The anti-rollbar can either be
“free” (i.e. transmit little force) or “locked” (i.e. provides high roll stiffness). The low roll
stiffness is intended for use when the lateral forces on the vehicle are small, thus
providing good ride comfort. When large lateral forces are present, the system is switched
THE RIDE COMFORT VS. HANDLING DECISION
5.2
to higher roll stiffness to improve handling. The vertical bounce stiffness of the
suspension system is therefore unaffected. Although only a preliminary analysis is
performed, the authors differentiate between a case where the lateral acceleration builds
up slowly (general driving) and a case where rapid increases in lateral acceleration takes
place (avoidance manoeuvres). For a rapid increase in lateral acceleration, using a lateral
acceleration threshold as control input seems reasonable if control system delays are
small. For the case of a slowly increasing lateral acceleration, a more sophisticated
control strategy (e.g. one that uses steering inputs) might be beneficial.
Assume ride comfort
control strategy is
known
Conflicting
requirements
Assume handling
control strategy is
known
Investigate how the decision between
“ride” and “handling” can be made
based on easily measurable
parameters
Vehicle implementation
Figure 5.1 - The ride comfort vs. handling decision
Jost (2002a) describes the Continental Teve’s four-corner air suspension with
continuously variable semi-active damper control fitted to the Volkswagen Phaeton. The
system adjusts damping force on each wheel within 10 to 15 milliseconds and
automatically adjusts vehicle height. The system uses wheel acceleration sensors on the
shock absorbers as well as body movement sensors (two at the front and one at the rear).
Other inputs include data from the engine management, brake and electronic stability
control systems. The system can recognise when the driver is steering into a curve. The
driver can select between four fixed damper settings ranging from soft to sporty or firm.
The control system will however temporarily override these settings when handling
manoeuvres are encountered.
Nell (1993) and Nell and Steyn (1998) develop a general strategy for the control of twostate semi-active dampers in an off-road vehicle suspension system. Nell focussed on a
THE RIDE COMFORT VS. HANDLING DECISION
5.3
full vehicle model taking all degrees of freedom into account instead of looking at each
wheel separately. He defines suspension control as a “decision making” problem. The
damper is switched to the high damping state whenever handling is required and
controlled using a “minimum product” strategy whenever ride comfort takes preference.
Roll movement over rough terrain is caused by suspension forces whereas lateral
acceleration causes roll movement during handling manoeuvres on smooth roads. The
ride comfort vs. handling strategy needs to differentiate between these two conditions.
Nell measures and compares lateral and vertical acceleration on the centre of the (rigid)
front axle. If this lateral acceleration is greater than the vertical acceleration, the handling
mode (all dampers switched to high damping) is selected; otherwise the “minimum
product” strategy is used to improve ride comfort.
Nell and Steyn (2003) apply the same basic idea to another off-road vehicle, in this case
using measurements from two solid-state gyroscopes and two accelerometers as inputs to
the control system. The control strategy is said to be a derivative of the method proposed
by Nell and Steyn (1998). It switches the two-state semi-active dampers to the conditions
that will provide the highest accelerations opposing the motion of the sprung mass, or the
lowest acceleration in the same direction. The relative damper velocities and absolute
sprung mass velocity are no longer required. Handling is improved over the baseline
vehicle by changing the “on” characteristic of the semi-active damper.
Darling and Hickson (1998) investigate the effect of an active anti-rollbar on the
handling of a vehicle. They aim for a “flat” ride e.g. no body roll. They state that,
although steering angle and vehicle speed can be used as control inputs, this relationship
can vary significantly due to differences in tyre-road friction. They therefore make use of
a lateral accelerometer mounted on the vehicle body in front of the centre of mass to
measure a combination of lateral and yaw acceleration. A simple PID controller was
implemented and the gains were optimised by a process of trial and error vehicle tests.
An electronic modulated air suspension system, as fitted to the 1986 Toyota Soarer is
described by Hirose et. al. (1988). The system is said to control spring rate, damping
force and height with a response time of 70 milliseconds. There are three control steps
namely: i) detection of vehicle travelling conditions, ii) classification of travelling
conditions into one of several preset patterns, iii) adjust suspension parameters according
to selected pattern. Sensors include three height sensors (left front, right front and left
rear), steering angle sensor, throttle position, stop lamp switch and mode select switch.
Vehicle height is detected in 16 steps between maximum and minimum height. Vehicle
height is lowered if the vehicle speed exceeds 90 km/h and only increased again once
vehicle speed drops to below 60 km/h resulting in a hysteresis of 30 km/h. On rough
roads, height is increased above 40 km/h and only decreased again below 25 km/h to
eliminate bump stop contact. Rough road conditions are detected by the left front wheel
displacement using an observation duration of 0.5 seconds (half the sprung mass natural
period). If the displacement measured during the observation duration exceeds a reference
value four times in succession, ride height is increased. The detection period for changing
ride height is 20 seconds to eliminate frequent ride height changes due to cornering for
example. Spring and damper rates are changed simultaneously on all four wheels. Control
of spring and damper settings are performed by either predictive control (see Table 5.1)
or tracking control (see Table 5.2)
THE RIDE COMFORT VS. HANDLING DECISION
5.4
Table 5.1 – Predictive control as implemented by Hirose et. al. (1988)
Situation
Anti-dive
Anti-roll
Anti-squat
Antibump
Sensor
Speed sensor
Stop lamp switch
Speed sensor
Steering sensor
Speed sensor
Throttle position sensor
Speed sensor
Height sensor
Purpose
Suspension is changed to harder setting to restrict attitude
change before the attitude change begins
Irregularity of roads is detected by vertical movement of the
front wheels and suspension is changed softer before the rear
wheels pass through the detected irregularity to reduce shock
Table 5.2 - Tracking control as implemented by Hirose et. al. (1988)
Situation
Response to speed
Sensor
Speed sensor
Response to rough
road
Speed sensor
Height sensor
Purpose
Suspension is set harder to improve travelling stability at
high speed cruising. Since speed change is gradual, tracking
control has satisfactory effect
Suspension is set harder to restrict pitching and bouncing on
rough road
A very similar system, fitted by Mitsubishi, is described by Mizuguchi et. al. (1984). The
suspension consists of air springs used in conjunction with coil springs and semi-active
dampers. Sensors for vehicle speed, steering wheel angular speed, sprung mass
acceleration (lateral, longitudinal and vertical), throttle speed and suspension stroke is
used. Apart from ride height control, the suspension system can be switched from soft to
hard quickly whenever any of the conditions in Table 5.3 are satisfied. The hard setting
increases spring stiffness by approximately 50% and damping by 150%. The soft state is
restored after 2 seconds in the hard state.
Table 5.3 – Strategy used by Mizuguchi et. al. (1984)
Case
1
Item
Vehicle speed
Sensor
Vehicle speed
2
Steering speed
Steering wheel angular
velocity
3
Sprung mass acceleration
Acceleration sensor
4
Throttle speed
Throttle position sensor
5
Front suspension displacement
Displacement sensor
Conditions
Soft to hard above 120 km/h
Hard to soft below 110 km/h
10 km/h hysteresis
Longitudinal acceleration: over 0.3g
Lateral acceleration: over 0.5g
Vertical acceleration: over 1g
Throttle wire moving speed:
*over 0.25 m/s when accelerating
*over 0.5 m/s when decelerating
(with vehicle speed ≥ 3 km/h)
Highest and lowest positions
THE RIDE COMFORT VS. HANDLING DECISION
5.5
Wallentowitz and Holdmann (1997) propose a frequency based control algorithm that
generates high damping only when the vehicle is excited in the vicinity of the natural
frequencies. They also propose a strategy where the vertical movement of each wheel is
controlled individually, but with an overlying controller for roll and pitch movements.
The spring must be switched to the stiff mode during braking and cornering to reduce roll
and pitch angles. The soft spring is said to be only beneficial for frequencies lower than 5
Hz. No simulation or test results are given for the proposed controller.
Armstrong Patents Company Limited of York developed a practical intelligent damping
system described by Hine and Pearce (1988). The system consists of two or three state
adaptive dampers combined with an auxiliary air spring to provide ride height control.
Measurements indicate that reducing the damper setting below standard greatly improves
vibration isolation at frequencies above 2 Hz, while higher than normal settings reduces
the amount of motion around 1 Hz (roll etc.) The aim of the control strategy is to keep the
damper in the lowest setting as long as possible and only switch to higher levels when
required. Switching typically takes place in 12 milliseconds. The system has sensors for
suspension movement, steering wheel angular velocity, vehicle speed, brake application
and wheel or body acceleration. The control strategy is separated into a number of
components namely:
i)
Ride: Ride control is initiated by the relative suspension displacement and vehicle
speed using displacement maps. If the displacement exceeds a pre-programmed
limit for the specific vehicle speed, higher damping will be selected. For returning
to the lower damper setting, the valves have been designed to delay until the
pressure is below a preset limit to reduce hydraulic noise.
ii) Handling (including roll control): Handling is detected based on steering wheel
speed and vehicle speed. Roll information, obtained from the suspension
displacement sensors, is then used to switch back to the lower damper settings
shortly after the vehicle returns to the level position.
iii) Acceleration: Information from the vehicle speed sensor is used to determine the
acceleration. Damper rate is increased when acceleration exceeds a pre-defined
level.
iv) Deceleration (Dive): Information from the brake and speed sensors is used to
immediately switch the dampers to the hard setting. The system returns to the softer
setting once the longitudinal acceleration drops to below a preset level.
v) Ride frequency control and vehicle levelling: Levelling compensates for changes in
payload and aerodynamics.
The system was implemented on a 1986 model GM Corvette (5.7 litre) as well as a Ford
Granada 2.8 Ghia. The improvements in ride comfort and handling is however not
quantified.
An active suspension control approach that consists of an inner loop that rejects terrain
disturbances, an outer loop that stabilises heave, pitch and roll response and an input
decoupling transformation that blends the inner and outer control loops is proposed by
Ikenaga et. al. (2000). The ride control loop isolates the car body from uneven terrain
while the attitude control loop maintains load levelling and load distribution during
handling manoeuvres. Skyhook damping (the term used to describe the feedback of
absolute sprung mass heave, pitch and roll velocities) improves heave, pitch and roll
accelerations at all frequencies below the wheel frequency.
THE RIDE COMFORT VS. HANDLING DECISION
5.6
Truscott (1994) develops a composite controller for a high bandwidth (35 Hz) fully
active suspension system. Only simulation results for a linear quarter car suspension
system are presented. The proposed composite controller consists of two controllers
operating together, but over different frequency ranges. The first controller cancels out
low frequency dynamic loads experienced during cornering and braking, keeping the
vehicle level. This controller operates at frequencies below 5 Hz. The second controller
isolates the car from high frequency terrain induced vibration and operates at frequencies
above 5 Hz. The vibration controller is fully adaptive and auto-tunes the system according
to varying payload, tyre stiffness and varying road frequency spectrum. The 5 Hz
frequency was chosen to be between the sprung mass and wheel-hop frequencies.
Trent and Greene (2002) propose a model-based genetic algorithm predictor to estimate
the potential for rollover. The tyre deflection that will result in vehicle rollover
approximately 50 time steps in future is calculated assuming all other operating
conditions such as vehicle speed remain constant. Advanced rollover warning of 400
milliseconds may be possible, giving enough time for an intelligent suspension system or
stability control (differential braking) system to react and decrease the rollover
propensity.
Active roll control, as developed by TRW, is discussed by Böcker and Neuking (2001).
The system uses hydraulic cylinders fitted to the anti-rollbars. Figure 5.2 indicates the
functioning of the control system schematically. Sensors include steering angle, lateral
acceleration, hydraulic system pressure and vehicle speed.
Figure 5.2 – TRW’s active roll control system according to Böcker and Neuking (2001)
THE RIDE COMFORT VS. HANDLING DECISION
5.7
Hamilton (1985) defines many general aspects for the theoretical operation of
controllable suspension systems. No simulation or test results are given although
prototype hardware was available. The proposed system consists of the following
components:
i)
Ideal damping device: must instantly provide the damping force required by the
computer independent of suspension position or velocity.
ii) Ideal energy storage device: must be capable of changing it’s energy storage
capacity to a value demanded by the computer.
iii) Ideal computer controller: must have all the necessary inputs to calculate all
required parameters.
The ideal theory of operation also consists of many aspects namely:
i)
Optimised ride comfort: requires a very soft spring and virtually no damping force.
Some force is required to control the kinetic energy of the wheel.
ii) Cornering: The centrifugal force on the vehicle’s centre of gravity causes a torque
on the sprung mass, about the roll centre, that can be counteracted by the damping
device by applying equal forces in the opposite direction.
iii) Ideal pitch control: Attempt to maintain a level ride during acceleration or braking
by looking at the change in pitch height.
iv) Ideal level ride control: This is required to compensate for substantial variations in
the loading condition of the vehicle. Height control can also be used to decrease the
frontal area of the vehicle during high speed driving or to increase ride height over
rough terrain.
v) Ideal roll control: e.g. on a mountain road. Can be based on the difference in height
between the sprung mass and the road surface on the left and right hand side of the
car.
vi) Ideal natural frequency control: Observe the two primary natural frequencies
(sprung and unsprung mass) using a Discrete Fourier Transform (DFT) and control
each one of them separately.
vii) Ideal high amplitude or high velocity control: Road inputs that exceed the dynamic
range of the suspension require forces to be applied to the sprung mass to move it
up and over the obstacle. The magnitude of these forces should be such that the
suspension movement limits are never exceeded.
Not all forces acting on the sprung mass can be totally eliminated. The system should aim
to minimise them while optimally controlling vehicle movements within the suspension
working space. There are many counteracting forces that are required simultaneously and
these must be superimposed to control all the dynamics simultaneously.
The concept is to apply these forces when required and keep them as small as possible to
ensure good ride comfort. The author states that only relative suspension movements need
to be measured as all the other required parameters can easily be calculated from these.
The Mercedes-Benz Active Body Control (ABC) is described by Birch (1999). The
system was introduced on the CL Coupe and adapts both spring and damper
characteristics to prevailing conditions. A hydraulic system (called a “plunger”) acts on
the coil spring to change the preload on the spring. The stiffness remains unchanged. The
hydraulic system acts up to a frequency of 5 Hz thereby improving vehicle response to
long wavelength road inputs as well as during braking and cornering. Anti-rollbars are not
THE RIDE COMFORT VS. HANDLING DECISION
5.8
necessary and the system is also self-levelling. The driver can select sport and comfort
settings.
A detailed description of the Citroën Hydractive I, II and III is beyond the scope of this
text, but is described in substantial detail by Nastasić and Jahn (2005). The control
principles employed are however relevant to the ride vs. handling decision. The basic idea
is to map different vehicle parameters against vehicle speed. Figure 5.3 indicates the
steering wheel angle threshold as a function of vehicle speed. The suspension is switched
to the handling mode whenever the measured steering wheel angle exceeds the threshold
at a certain speed. The driver can select one of two different threshold levels by selecting
a “normal” or “sport” mode with a switch. The steering wheel velocity threshold (Figure
5.4) exhibits a similar trend and operates on the same principle. At low vehicle speeds,
large steering wheel angles and velocities are allowed e.g. during parking manoeuvres. As
the vehicle speed increases the threshold levels become smaller, resulting in faster
reaction times.
Body dive and squat (Figure 5.5) is determined by measuring the relative displacement of
the front and rear suspension respectively. Threshold levels also decrease as vehicle speed
increases. The accelerator pedal press and release rate is also used (Figures 5.6 and 5.7) to
reduce squat and pitch during hard acceleration. Slow release of the accelerator pedal
indicates that the driver desires to reduce speed gradually whilst a sudden release will
often be followed by hard braking to quickly reduce speed. Dive and squat effects are
further ignored for the present study.
Steering angle vs. Speed
200
180
160
Steering wheel angle
140
120
100
80
60
40
20
0
34-39
40-49
50-59
60-68
69-78
79-89
90-99
100-119
Vehicle Speed [km/h]
Normal mode
Sport mode
Figure 5.3 – Steering wheel angle vs. vehicle speed
120-139
140-158
159-179
>179
THE RIDE COMFORT VS. HANDLING DECISION
5.9
Steering wheel speed : Hydractive II
600
Steering wheel speed [°/s]
500
400
300
200
100
0
24-29
30-39
40-49
50-59
60-68
69-78
79-89
90-99
100-119
120-139
140-158
>158
Vehicle Speed [km/h]
Normal
Sport
Figure 5.4 – Steering wheel rotation speed vs. vehicle speed
Hydractive II - Body movement thresholds
90
80
70
Displacement [mm]
60
50
40
30
20
10
0
10-33
34-39
40-49
50-59
60-68
69-78
79-89
90-99 100-109 110-119 120-129 130-139 140-149 150-158 159-179
Vehicle speed [km/h]
Dive [mm]
Figure 5.5 – Dive and squat vs. vehicle speed
Squat [mm]
>179
THE RIDE COMFORT VS. HANDLING DECISION
5.10
Hydractive II - Pedal press rate
8
7
Press rate [steps/25 ms]
6
5
4
3
2
1
0
<14
15-49
50-99
100-134
135-199
>199
Vehicle speed [km/h]
Normal
Sport
Figure 5.6 – Accelerator pedal press rate vs. vehicle speed
Hydractive II - Pedal release rate
12
Release rate [steps/25 ms]
10
8
6
4
2
0
<19
20-78
79-168
>168
Vehicle speed [km/h]
Normal
Sport
Figure 5.7 - Accelerator pedal release rate vs. vehicle speed
Interactive vehicle dynamics control on the 2000 Ford Focus is discussed by Broge
(1999). Although not controlling the suspension system of the car, the general concept
may be applicable to the current study. Several parameters measured on the vehicle are
THE RIDE COMFORT VS. HANDLING DECISION
5.11
compared to a dynamic handling map stored in the on-board computer. When any vehicle
parameter deviates from the stored map, corrective action is taken by reducing engine
power and braking appropriate wheels. Sensors include individual wheel speeds, steering
wheel movement, yaw rate sensors and lateral accelerometers.
5.2
Suggested concepts for making the “ride comfort vs. handling decision”
From the literature discussed in paragraph 5.1, several concepts have been identified to
assist in making the “ride comfort vs. handling decision”. These concepts are listed in
Table 5.4 and will be investigated further in paragraph 5.5. It is important to note that the
majority of the applications discussed so far are related to road vehicles. Substantial
differences might be required for off-road driving.
Table 5.4 – Suggested concepts for assisting with the “ride vs. handling” decision
Concept no.
1
5.3
Measurement parameters
Frequency analysis of acceleration
2
Lateral acceleration vs. vertical acceleration
3
Steering angle vs. speed
4
5
Pitch and roll velocity / acceleration
Height, throttle position, brake application,
mode select switch
6
Lateral acceleration
Reference
Wallentowitz and Holdman (1997)
Truscott (1994)
Nell (1993)
Nell and Steyn (1998)
Hirose et. al. (1988)
Hine and Pearce (1988)
Nastasíc and Jahn (2005)
Broge (1999)
Nell and Steyn (2003)
Hirose et. al. (1988)
Hine and Pearce (1988)
Nastasíc and Jahn (2005)
Broge (1999)
Stone and Cebon (2002)
Daling and Hickson (1998)
Easily measurable parameters
At the outset of this study, the decision was made to try and find a strategy that uses
parameters that can be measured directly, or otherwise easily calculated from direct
measurements. This therefore excludes the use of state estimators, integrators and
artificial intelligence techniques such as neural networks. This decision was made based
on several factors namely:
•
This is the first study focussed on the “ride comfort vs. handling decision” for offroad vehicles
•
No previous concepts or algorithms seem to exist
•
Attempting to keep it as simple as possible and only as complicated as necessary
•
The current focus is on a more fundamental understanding of the issues involved
•
It is important to keep the cost of the sensors and control system within the project
budget
•
The “controller” to be used was a personal computer based system with an analog to
digital converter card and a digital input-output card fitted. This excluded the use of
digital signal processing (DSP) cards.
The parameters identified to be easily and directly measurable are listed in Table 5.5
while Table 5.6 lists parameters that can be easily calculated from the directly measured
parameters. Displacement measurements can be differentiated with respect to time to give
THE RIDE COMFORT VS. HANDLING DECISION
5.12
velocities. Although differentiation tends to add high-frequency noise, we are primarily
interested in the low-frequency content of the velocity and good results can be achieved
using simple mathematics and low-pass filters. Integration is also possible in theory, but
creates many obstacles in practice due to the effect of drift. Small offsets in the zero
reading of a sensor (e.g. accelerometer) can cause the integrated value to quickly drift to
the limits. Because we are primarily interested in the low-frequency content, it is very
difficult to control drift by for example high-pass filtering. The signal offsets are often
influenced by effects such as change in temperature or attitude changes of the vehicle
body due to varying load and road conditions. Absolute body movements can presently
only be calculated by integrating acceleration signals twice. Absolute body movements
are thus not easily measured directly, or calculated, and are therefore excluded at present.
Table 5.5 – Directly measurable parameters
No
1
2
3
4
Parameter
Vehicle speed
Relative displacement
Angular velocity (roll,
yaw, pitch)
Relative displacement
5
6
7
8
Acceleration
Kingpin steer angle
Wheel speed
Driveshaft speed
Position
Roof
Every suspension position
Vehicle body
Equipment
Global positioning system (GPS)
Rope displacement transducer
Solid state gyroscope
Steering arm between axle
and body
Vehicle body
Kingpin
Any wheel
Gearbox output rear
Rope displacement transducer
Solid state accelerometer ±4g range
Potensiometer
Optical speed sensor
Optical speed sensor
Table 5.6 – Parameters that can be easily calculated from measurements
No
1
2
3
4
5.4
Parameter
Relative suspension velocities
Relative angles between vehicle body and suspension components
Relative angular velocities
Angular accelerations
Experimental work on baseline vehicle
A test sequence, consisting of six different test routes and manoeuvres, was devised to be
representative of the Land Rover Defender 110 vehicle’s typical application profile. Tests
were performed at representative speeds. For city and highway driving, tests were
performed in and around the city of Pretoria. All other tests were performed at the
Gerotek Vehicle Test Facility West of Pretoria. The legal speed limit was adhered to on
all public roads. For off-road driving, the speed was determined by the driver’s judgement
of ride comfort while on the mountain pass the speed was limited either by vehicle
performance on the steep uphill slopes or by handling around the corners. For the
handling and rollover tests the speed constraint was the vehicle’s handling combined with
the driver’s ability. The chosen test routes are summarised in Table 5.7 with the plan
layouts of the test routes and tracks indicated in Figures 5.8 to 5.12.
It was also postulated that an objective vehicle parameter (e.g. lateral acceleration) might
be correlated with an objective human physiologic parameter (e.g. heart rate or blood
pressure). A series of tests were performed where heart rate and blood pressure was
measured for both driver and passengers in attempt to obtain a correlation between
vehicle parameters and change in heart rate. A total of 85 test subjects were used for the
THE RIDE COMFORT VS. HANDLING DECISION
5.13
physiological measurements. Although very interesting trends were noticed, no
correlation could be found between the measured physiological parameters and vehicle
parameters.
Table 5.7 – Chosen tests and test routes
Test
1
Driving
conditions
City driving
2
Highway
driving
3
4
Off-road
Mountain
pass
Handling
Rollover
5
6
5.5
Test route
Driver
Start: corner Dely & High (point 2)
End: corner Rigel &Buffelsdrift (point 3)
Start: Fountains circle (point 4)
End: corner Lynnwood & Kiepersol (point
5)
Top 800m of Gerotek Rough Track
Gerotek Ride & Handling Track - clockwise
Normal
Duration
[s]
704
Figure
Normal
783
5.8
Normal
Experienced
166
268
5.10
5.11
ISO 3888 Severe double lane change test
Fishhook rollover simulation test
Experienced
Driving
robot
13.4
7.1
5.12
5.9
5.8
Evaluation of concepts
The concepts identified in paragraph 5.2, and summarised in Table 5.4, will now be
implemented on the test data measured on the baseline vehicle and evaluated. Although
this approach is not strictly correct since the vehicle dynamics will change when the
suspension settings change, this method is expected to illustrate trends and provide a first
order evaluation of feasibility.
Only the first three concepts listed in Table 5.4 were investigated in more detail. Concept
4 is a “ride comfort” strategy while concept 5 focuses on the effect of longitudinal forces
(e.g. due to acceleration or braking) on the vehicle and therefore not ride comfort or
handling. Concept 6 is a “handling” strategy and ignores ride comfort.
At this point, the controllable suspension system is assumed to function as a two-state
system that can be switched between a “ride comfort mode” and a “handling mode”.
Several requirements were set for evaluating the feasibility of a control strategy. These
requirements are:
•
Switching should not be too frequent, e.g. the strategy should not “hunt” between
the “ride comfort” and “handling” settings
•
The strategy should work for all six chosen tests without manual driver intervention.
•
The strategy should rather err towards the “handling” mode.
•
For both the handling test and the rollover test, the system should switch to the
“handling mode” as quickly as possible, and remain in the “handling mode” for the
duration of the manoeuvre. Ideally the “handling mode” should be selected before
the start of the manoeuvre, but this is not possible without some kind of preview.
•
During off-road driving, the system should remain in “ride comfort” mode for most
of the time.
THE RIDE COMFORT VS. HANDLING DECISION
Figure 5.8 – City and highway driving route
Figure 5.9 – Fishhook test
5.14
THE RIDE COMFORT VS. HANDLING DECISION
Figure 5.10 – Gerotek rough track top 800 m
Figure 5.11 – Gerotek Ride and handling track
Figure 5.12 – Double lane change test
5.15
THE RIDE COMFORT VS. HANDLING DECISION
5.16
5.5.1 Frequency domain analysis
The first concept implemented is frequency domain analysis proposed by Wallentowitz
and Holdman (1997) as well as Truscott (1994). To determine whether this concept is
feasible, FFT magnitudes were calculated for measurements over the six predefined tests.
Each measurement was divided into bins of 1024 data points each with no overlapping.
The FFT magnitudes were calculated for each bin of 1024 points, and then averaged for
all the bins of the specific measurement at each frequency. Figure 5.13 indicates the
average FFT magnitudes for the left rear and right rear vertical accelerations measured on
the vehicle body. All six superimposed graphs indicate the same two peaks at 2 Hz (Body
natural frequency) and 12 Hz (wheel hop frequency), although the magnitudes differ for
the different terrains.
Figure 5.14 indicates that only very low frequencies can be detected by looking at the
lateral acceleration. Trends are similar for all six tests. Roll, yaw and pitch velocities are
indicated in Figure 5.15. The off-road track causes significant activity around 2 Hz that is
absent in the other tests. Yaw velocity is restricted to low frequencies while the pitch
natural frequency can be seen to be around 2 Hz for all six terrains. Relative suspension
displacements (Figure 5.16) indicates the body natural frequency around 2 Hz. This is
only really noticeable on the off-road test. The FFT magnitudes of the steering
displacement and kingpin steering angle are indicated in Figure 5.17. All activity takes
place at very low frequencies.
FFT magnitudes of relative suspension velocities are indicated in Figure 5.18. The
relative velocity was obtained by differentiating the relative displacement in the time
domain and then calculating the FFT. Again the frequency at 2 Hz is prominent with
activity from about 1 Hz to 12 Hz. Trends do however look the same for all terrains.
The FFT magnitudes of the steering velocities, calculated by first differentiating the
steering displacements in the time domain, are indicated in Figure 5.19. Figure 5.19(b)
indicates the FFT magnitude of the steering velocity at the kingpin. Figure 5.19(a)
indicates the steering velocity calculated from the measured displacement between the
vehicle body and the steering link going to the wheels. The steering velocity clearly
indicates activity around 8 to 10 Hz when driving off-road and through the mountain pass
that is not present on the kingpin steering velocity. This is attributed to bump and roll
steer as well as the kinematic effects resulting from the Panhard rod.
Although the frequency domain analysis provides valuable insight into the various
excitation and natural frequencies, it is concluded that “ride vs. handling” cannot be
detected from the frequency analysis. The same frequencies are excited regardless of
terrain types, manoeuvres and speeds.
5.5.2 Lateral vs. vertical acceleration
The next strategy that was investigated is the one proposed by Nell (1993) and Nell and
Steyn (1998). They compare lateral and vertical acceleration as measured on the rigid
front axle of a heavy off-road military vehicle. The semi-active dampers on their test
vehicle is switched to “hard” when the lateral acceleration is higher than the vertical
THE RIDE COMFORT VS. HANDLING DECISION
Figure 5.13 – FFT magnitude of vertical body acceleration (left and right rear)
Figure 5.14 – FFT magnitude of body lateral acceleration (left front and left rear)
5.17
THE RIDE COMFORT VS. HANDLING DECISION
Figure 5.15 – FFT magnitudes of body roll, yaw and pitch velocity
Figure 5.16 – FFT magnitude of relative suspension displacement (all four wheels)
5.18
THE RIDE COMFORT VS. HANDLING DECISION
Figure 5.17 – FFT magnitude of steering displacement and kingpin steering angle
Figure 5.18 – FFT magnitude of relative suspension velocity (all four wheels)
5.19
THE RIDE COMFORT VS. HANDLING DECISION
5.20
Figure 5.19 – FFT magnitude of steering velocity
acceleration and to the “ride comfort” mode when vertical acceleration is higher than
lateral acceleration.
Figure 5.20 indicates the result of this analysis when applied to our measurements during
city driving. Figure 5.20(a) indicates vertical and lateral accelerations measured on the
left rear of the vehicle while Figure 5.20(b) indicates the suspension switching. A value of
“0” in Figure 5.20(b) indicates “ride” mode while a value of “1” indicates “handling”
mode. Switching seems spurious and random, e.g. between 530 and 600 seconds the
vehicle is stationary, but the suspension switches all the time due to the background noise
on the acceleration signals. The idling engine causes some of this noise.
Figure 5.21 indicates switching during the rollover test. It can be seen that the switching
only works in one direction. The absolute values of the lateral and vertical acceleration
should therefore be compared to enable correct switching. Two fundamental problems
exist with the strategy as proposed by Nell namely:
•
The absolute values of the accelerations should be compared
•
Provision should be made for some type of dead band to prevent spurious switching
when the measured values are close to zero
Nell intended the strategy to be used for accelerations on the front axle of the vehicle. The
vertical acceleration on the axle is however significantly higher (can be 15 to 25g peak)
than the lateral acceleration (about 1g peak). The strategy will therefore favour ride
comfort, especially on rough roads. The strategy will also emphasize wheel hop
frequency and not body motion due to the measuring position on the axle.
THE RIDE COMFORT VS. HANDLING DECISION
Figure 5.20 - Strategy proposed by Nell (1993) as applied to city driving
Figure 5.21 - Strategy proposed by Nell (1993) as applied to the rollover test
5.21
THE RIDE COMFORT VS. HANDLING DECISION
5.22
5.5.3 Lateral vs. vertical acceleration - modified
The strategy proposed in paragraph 5.5.2 is now modified in order to eliminate its
drawbacks. The absolute values of the lateral and vertical accelerations on the vehicle
body are used. A dead band is introduced to prevent spurious switching due to
accelerometer noise and drift. The “ride” mode is always selected if the absolute lateral
acceleration is less than 0.1 g. An upper limit is also included that forces switching to the
“handling” mode when lateral acceleration exceeds 0.3g (see Stone and Cebon, 2002 and
Darling and Hickson, 1998). This however results in negligible improvement during
highway driving (Figure 5.22) although the ride mode is at least selected during periods
when the vehicle is stationary (e.g. between 650 and 700 seconds). A significant
improvement is however noticed for the rollover test (Figure 5.23) where the handling
mode is selected during most of the manoeuvre. The switching to the ride mode at 3.2
seconds is however problematic as this happens at a critical point in the test. The method
is however an improvement on the previous case.
5.5.4 Steering angle vs. speed
The use of a speed dependant steering angle threshold has been applied frequently. Some
examples are discussed by Hirose et. al. (1988), Hine and Pearce (1988), Nastasíc and
Jahn (2005) as well as Broge (1999). The steering angle vs. speed threshold used by
Citroën was indicated in Figure 5.3.
The envelope for the Land Rover was determined by plotting steering angle against
vehicle speed for all the tests. The results are indicated in Figure 5.24. The circles in
Figure 5.24 indicate the measured data points obtained for city driving and the solid lines
indicate the limiting values determined for different terrains. These curves represent the
values of steering angle that is achieved during normal driving.
The strategy itself is very easy to implement and is an “input driven” strategy i.e. it will
react on driver input and not vehicle reaction to driver input as is the case with lateral
acceleration etc. It should therefore also give an early warning before the vehicle reaction
can be detected.
The results of this strategy as implemented on test data is shown in Figures 5.25 to 5.30.
The threshold value used for all the analyses indicated in Figures 5.25 to 5.30 is 50% of
the steering angle limit for city driving (solid red line in Figure 5.24) at any given vehicle
speed. During city driving (Figure 5.25), off-road driving (Figure 5.27) and on the
mountain pass (Figure 5.28) the strategy works well due to the large steering angles
involved. During the highway tests (Figure 5.26) switching seems to occur too often due
to the fact that the steering threshold is very low and measurement noise and sensor drift
has a significant effect on the results.
For the handling test (Figure 5.29) and rollover test (Figure 5.30) results are not entirely
satisfactory, as the suspension will be switched to the “ride” mode whenever the steering
goes through the zero position. This is dangerous as the effect takes place during critical
parts of the manoeuvre.
THE RIDE COMFORT VS. HANDLING DECISION
Figure 5.22 – Modified lateral vs. longitudinal acceleration for highway driving
Figure 5.23 - Modified lateral vs. longitudinal acceleration for rollover test
5.23
THE RIDE COMFORT VS. HANDLING DECISION
5.24
Figure 5.24 – Steering limits vs. vehicle speed measured during three tests
Spurious switching also sometimes occurs due to noise. This can be seen for example in
the first 20 seconds of Figure 5.25 where the vehicle is stationary, but the steering wheel
is turned. This problem can however be solved by using a dead band instead of a single
limit.
The biggest difficulty when applying this strategy to the off-road vehicle is that the
threshold values differ considerably depending on the terrain. If the terrain can be
somehow “identified”, and the threshold values adapted accordingly, then the
performance of the strategy can be improved. Performance for the handling and rollover
tests are however only expected to improve marginally because the system will still
switch to “ride mode” when the steering angle goes through the zero position.
5.5.5 Disadvantages of proposed concepts
All the concepts investigated up to this point suffer from the same disadvantages namely:
i) Switching occurs too frequently
ii) All strategies don’t work properly for all conditions
iii) Strategies that work well for on-road driving fail during off-road tests and vice
versa
Unnecessary switching could be eliminated by applying a dead band, low pass filtering or
delayed switching. It is imperative that absolute values be used.
THE RIDE COMFORT VS. HANDLING DECISION
Figure 5.25 – Steer angle vs. speed implemented for city driving
Figure 5.26 – Steer angle vs. speed implemented for highway driving
5.25
THE RIDE COMFORT VS. HANDLING DECISION
Figure 5.27 – Steer angle vs. speed implemented for off-road driving
Figure 5.28 – Steer angle vs. speed implemented for mountain pass driving
5.26
THE RIDE COMFORT VS. HANDLING DECISION
Figure 5.29 – Steer angle vs. speed implemented for handling test
Figure 5.30 – Steer angle vs. speed implemented for rollover
5.27
THE RIDE COMFORT VS. HANDLING DECISION
5.6
5.28
Novel strategies proposed
To overcome the problems mentioned in paragraph 5.5.5, two additional strategies are
proposed namely relative roll angle and running RMS (RRMS).
These proposed strategies will be discussed in paragraphs 5.6.1 and 5.6.2.
5.6.1 “Relative roll angle” calculated from suspension deflection
Roll angle was identified as a good measure of handling in paragraph 2.1.2.5. Absolute
body roll angle is however very difficult to measure directly. The first proposal is to use
the relative body roll angle between the vehicle body and axle, calculated using the
relative suspension deflection of the left and right suspension systems. Because the body
roll angle is small (< 5°), the roll angle is proportional to the difference between left and
right relative displacements, divided by the distance between the left hand and right hand
displacement measuring points. The difference between left and right displacements is
therefore directly compared to a threshold value, without calculating the actual body roll
angle. If this difference exceeds the threshold, the suspension system is switched to the
handling mode. Results of this concept, applied with a threshold value of 20 mm, are
indicated in Figures 5.31 to 5.36. The left front and right front relative suspension
displacements were used in these calculations, but the same concept could be applied to
the displacements measured for the rear axle.
The strategy works well for city driving (Figure 5.31), highway driving (Figure 5.32) and
mountain pass driving (Figure 5.34). It switches to “handling” mode too frequently during
off-road driving (Figure 5.33). During the handling test (Figure 5.35), the “ride” mode is
selected for most of the manoeuvre. “Handling” mode is selected only at the most critical
part of the test where the vehicle returns to the initial lane (between 70 and 100 meters in
Figure 5.12, corresponding to between 8 and 11 seconds in Figure 5.35). This switch to
“handling” mode at this critical point of the test might have disastrous effects. Although
behaviour during the handling test can be improved by reducing the switching threshold
of 20 mm, “ride” mode will still be selected whenever the relative roll angle crosses
through the zero position. A reduction of the threshold will also result in more unwanted
switching during off-road driving. During the fishhook rollover test (Figure 5.36),
dangerous switching to the “ride comfort” setting occurs where the relative roll angle
crosses the zero position. This is however also the place where the roll velocity (and
therefore kinetic energy due to body roll) is maximum. Switching to “ride” mode under
these conditions is highly undesirable.
5.6.2 Running RMS vertical acceleration vs. lateral acceleration
The second proposal is to use the running RMS (RRMS) value of lateral acceleration
compared to the running RMS of vertical acceleration. This concept will result in an
average absolute value of the required parameters and should therefore reduce spurious
switching and noise.
The running RMS (RRMS) is calculated determining the RMS value of the last N number
of points. The strategy includes hysteresis and will always select the “ride comfort” mode
if the RRMS lateral acceleration is less than 0.05g. It also always selects “handling” mode
when the RRMS lateral acceleration is greater than 0.3g. Between these two limits,
handling mode is selected only when the RRMS lateral acceleration exceeds the RRMS
THE RIDE COMFORT VS. HANDLING DECISION
5.29
vertical acceleration. A running RMS of 1 second (or 100 previous data points) has been
used for this analysis and seems to successfully remove noise without affecting response
time detrimentally.
Figure 5.31 – Relative roll angle strategy for city driving
Figure 5.32 – Relative roll angle strategy for highway driving
THE RIDE COMFORT VS. HANDLING DECISION
Figure 5.33 – Relative roll angle strategy for off-road driving
Figure 5.34 – Relative roll angle strategy for mountain pass driving
5.30
THE RIDE COMFORT VS. HANDLING DECISION
Figure 5.35 – Relative roll angle strategy for handling
Figure 5.36 – Relative roll angle strategy for rollover
5.31
THE RIDE COMFORT VS. HANDLING DECISION
5.32
RRMS strategy results are indicated in Figures 5.37 to 5.42. This strategy works well for
all conditions except for the double lane change where the “ride comfort” mode is
selected about halfway through the test (see Figure 5.41). This is however the point where
the vehicle is in the second lane before it starts turning back into the first lane. This
should not result in serious problems, as long as the switching back to “handling” mode
happens quickly enough.
Figure 5.37 – RRMS strategy for city driving
Figure 5.38– RRMS strategy for highway driving
THE RIDE COMFORT VS. HANDLING DECISION
Figure 5.39– RRMS strategy for off-road driving
Figure 5.40– RRMS strategy for mountain pass
5.33
THE RIDE COMFORT VS. HANDLING DECISION
Figure 5.41– RRMS strategy for handling test
Figure 5.42– RRMS strategy for rollover test
5.34
THE RIDE COMFORT VS. HANDLING DECISION
5.35
For the analyses discussed above, a 100-point or 1 second RRMS was used. The number
of points in the RRMS is expected to influence the response time, threshold levels and
rejection of noise for short duration events. Figure 5.43 indicates the effect of the number
of points in the RRMS on both the RRMS value and the resultant switching of the system
for the handling test. The ideal behaviour would be if the system switches to “handling”
mode immediately upon starting the test (i.e. at 3.7 seconds), and then remains in
“handling” mode for the duration of the test. Figure 5.43(a) indicates the RRMS of the
lateral acceleration for number of points from one to 500. The one point RRMS
corresponds to the absolute value of the measured acceleration, while the 500 point
RRMS corresponds to a five second RRMS. An increase in the number of points results
in more “smoothing”. The RRMS magnitude also decreases with an increase in the
number of points. This means that the threshold levels should be decreased as the number
of points is increased.
The corresponding switching according to the RRMS strategy is indicated in Figure
5.43(b). The y-axis has no units but just indicates the switching pattern for the eight
different analyses. For the one point RRMS, switching occurs quickly after the start of the
test (at 3.7 seconds). The switching delay as a function of the RRMS duration is indicated
in Figure 5.44. As the RRMS duration increases, the switching delay increases
accordingly. A one point RRMS does however result in many switchovers between “ride”
and “handling” mode. As the RRMS duration is increased, the number of switchovers
decreases. RRMS durations of 2 seconds and higher result in the system staying in
“handling” mode for the duration of the test. The percentage time spent in the “handling”
mode is indicated in Figure 5.45 as a function of the RRMS duration. As the RRMS
duration increases above 2 seconds, the initial delay results in a reduction of time spent in
the “handling” mode. The choice of RRMS duration is therefore a trade-off between
response time and switching behaviour. Values between one and two seconds seem to be
a reasonable starting point.
5.7
Conclusion
It is concluded that, of all the proposed strategies, only the running RMS (RRMS) appears
to work for all the test conditions. Vehicle tests must be performed to validate the
strategy.
A combination of strategies may also result in improvements, e.g. the steering angle can
be used to determine the switching point from “ride comfort” to “handling”, but switching
back to “ride comfort” may then be based on the running RMS, or simply delayed by a
fixed time to eliminate spurious switching.
If the terrain or driving conditions could be successfully identified, using for example
artificial intelligence techniques (self organising maps, neural networks etc.), other
concepts (e.g. steering angle vs. vehicle speed) may be successfully implemented by
adapting thresholds according to operating conditions.
THE RIDE COMFORT VS. HANDLING DECISION
5.36
Figure 5.43 – Effect of number of points in the RRMS on switching for handling test
Figure 5.44 – Effect of number of points in the RRMS on the switching delay for
handling test
THE RIDE COMFORT VS. HANDLING DECISION
5.37
Figure 5.45 – Effect of number of points in the RRMS on time spent in “handling” mode
for handling test
Chapter
6
The integration of the 4S4 suspension hardware, associated hydraulics and electronics on
the test vehicle is discussed in this chapter. Ride comfort and handling test results,
performed on the vehicle with the 4S4 system fitted, are quantified, discussed and
compared to baseline values obtained from testing of the baseline vehicle. Results are
interpreted to determine whether the system works as intended and if the proposed “ride
comfort vs. handling” decision strategy performs as predicted.
6.1
Installation of 4S4 hardware on test vehicle
Mounting of the new suspension system to the test vehicle required relatively minor
modifications to the chassis and axle mounting points. Mudguards on the inside had to be
cut to make provision for the units. The struts are mounted on the same centerline as the
baseline suspension system. One notable change is the absence of any rubber elements in
the mounting arrangement compared to the baseline suspension system, where the
dampers were mounted to the chassis and axles with rubber bushes. The original rubber
bump stops and axle-locating links were not modified. This results in exactly the same
suspension travel and suspension kinematics as the baseline suspension system.
The prototype 4S4 units, as fitted to the right hand side of the test vehicle, are illustrated
in Figures 6.1 to 6.5. Purpose-made top and bottom mounting brackets can be seen in
Figure 6.1. The required wiring to the solenoid valves, as well as the hydraulic pipe for
height adjustment is visible in Figure 6.5.
The pressure transducers, used to measure strut pressure, can be seen on top of the
aluminium valve blocks in the figures.
Ride height adjustment capability was also incorporated on the test vehicle. The
requirement for the ride height adjustability is that the system should be able to raise or
lower the vehicle body up to the maximum or minimum elevation in 30 seconds. The
minimum required oil flow for all four struts was calculated to be 1.57 l/min. The pump
used has a volumetric displacement of 1.0 cm³/rev and delivers 3.0 litres per minute at a
motor speed of 3000 r.p.m. The required oil reservoir should hold sufficient oil to
guarantee functionality during lowering or raising of the vehicle. In order to have a
sufficient reserve, a reservoir with a usable capacity of 5.9 litres was selected. The
hydraulic power pack consists of a 12 Volt direct current (DC) electric motor, hydraulic
gear pump and oil reservoir supplied by SPX Stone (Anon, 2005c). The assembled DC
power pack is shown in Figure 6.6. The power pack is driven from a supplementary 12
Volt battery that is connected in parallel to the vehicle’s 12 Volt battery.
VEHICLE IMPLEMENTATION
6.2
Top mounting bracket
Bottom mounting bracket
Figure 6.1 - Right rear suspension fitted to chassis – front view
VEHICLE IMPLEMENTATION
Figure 6.2 - Right rear suspension fitted to chassis – inside view
Figure 6.3 - Right front and right rear suspension fitted to chassis
6.3
VEHICLE IMPLEMENTATION
Figure 6.4 - Right rear suspension fitted to test vehicle – side view
6.4
VEHICLE IMPLEMENTATION
Figure 6.5 - Right front suspension fitted to test vehicle – side view
6.5
VEHICLE IMPLEMENTATION
6.6
Figure 6.6 - Assembled hydraulic power pack
A control manifold (Figure 6.7) is used to regulate the oil flow from the power pack to the
individual struts, or to let the oil flow back to the oil reservoir.
Figure 6.8 indicates the hydraulic pump and associated reservoir and valves used for
height adjustment, mounted in the load area of the vehicle. The solid-state relays used to
switch the solenoid valves are also shown.
6.2
Control electronics
The 4S4 control system controls ride height as well as the different spring and damper
settings by means of solenoid valves. For this purpose it is necessary for the controller to
process analog signals, from sensors measuring the vehicle’s current operating conditions,
to switch the solenoid valves and hydraulic power pack.
The control unit is based on a Coremodule 420 computer (PC-104 form factor) from
AMPRO. Analog inputs are measured with a Diamond Systems MM-16-AT 16-bit analog
to digital convertor card. The digital outputs, controlling the solid-state relays, are
provided by a Diamond Systems Onyx-MM-DIO card. A schematic diagram of the
control unit is provided in Figure 6.9.
VEHICLE IMPLEMENTATION
6.7
Figure 6.7 - Control manifold for ride height adjustment
Left rear 4S4 unit
Right rear 4S4 unit
Pump
Solid state relay box
Height control
valve block
Reservoir
Figure 6.8 – Piping, wiring and electronics
VEHICLE IMPLEMENTATION
6.8
The four relative strut displacements (one for each 4S4 strut) as well as lateral and vertical
accelerations are digitised by the analog to digital converter card. The ride height
adjustment algorithms use the relative strut displacements while the “ride vs. handling”
decision uses only the vertical and lateral body accelerations. After computing the
required settings for all the valves, the valves are switched via the digital output card and
solid state relays.
The control algorithm used for the “ride vs. handling decision” is the running RMS
(RRMS) strategy proposed in chapter 5. The control loop runs at 100 Hz and employs a
100-point (or 1 second) RRMS. Both lateral and vertical accelerations are measured using
a single Crossbow CXL04LP3 tri-axial accelerometer with built-in signal conditioning.
There are several issues that require special attention including zero positions, signal drift
and noise. Initially the aim was to mount the accelerometers on the test vehicle in the
vicinity of the center of mass. This mounting position resulted in high noise content from
presumably the engine or drivetrain vibration. Although the mean signal was zero, the
RMS resulted in an unacceptably high value. The accelerometer was subsequently moved
to a position under the rear seat, where the engine vibration levels were significantly
reduced. As an additional precaution, these accelerations were filtered with a 6th order
analog low-pass Butterworth filter, with a 50 Hz cut-off frequency, to prevent aliasing
and to filter out engine related vibration. The software also recorded measurements before
each test in order to obtain the zero values on all sensors.
Relative strut displacements are measured using ICS-100 In-Cylinder Sensors from Penny
& Giles. The linear potentiometer positioning sensors are mounted inside the struts,
surrounded by the hydraulic oil. They offer low hysteresis, low electrical noise, stable
output under temperature extremes and good dither vibration performance. No signal
conditioning is necessary and the sensors only require a stable supply voltage to operate
reliably.
All the valves are normally closed i.e. in the event of power failure (e.g. due to a flat
battery, cable breaking or control computer that reboots), the 4S4 system will revert to the
“handling” mode (i.e. stiff spring and high damping with no height adjustment). This adds
a failsafe capability to the system. Due to the required reverse logic, the switching signals
indicated in the rest of this chapter have the opposite meaning to those in Chapter 5, i.e. a
logic “1” now means “ride” mode (all valves open) and a logic “0” indicates handling
mode (all valves closed).
6.3
Steady state handling
The steady state handling characteristics of the vehicle were tested using a constant radius
test. In this test, the vehicle was driven around a circle of 25-meter radius starting at crawl
speed and gradually increasing speed until the vehicle reached it’s handling limit (based
on either sliding out or impending rollover). Data is represented as a graph of steering
link displacement against lateral acceleration. A zero slope on this graph indicates neutral
steer. A positive slope (steering angle increases as lateral acceleration increases) indicates
understeer while a negative slope (steering angle decreases as lateral acceleration
increases) indicates oversteer.
VEHICLE IMPLEMENTATION
6.9
Figure 6.9 - Control computer schematic
The effect of front:rear roll-stiffness balance was determined experimentally by
performing preliminary tests without any control applied, but just switching the valves
manually.
Measured characteristics are indicated in Figure 6.10 for the “handling” (all springs hard)
mode, front suspension hard (rear soft) and rear suspension hard (front soft). All three
settings steer neutrally up to 0.3 g after which oversteer develops for “all springs hard”
and “rear springs hard”. In the case where the front suspension is hard, the vehicle steers
neutrally up to 0.4 g and thereafter understeers. This indicates that switching the front
suspension to the hard setting can induce understeer. The opposite scenario is probably
also valid (e.g. switching the rear to hard will result in oversteer) although this is not as
evident from the data in Figure 6.10. A possible handling strategy then is to switch the
front suspension to hard when oversteer is detected and vice versa to counter understeer.
Figures 6.11 to 6.14 indicate the relative roll angle between the body and axle against
lateral acceleration for different combinations of spring stiffness and ride height. It is
clear that stiffening the suspension, as well as lowering the ride height, considerably
reduces the body roll angle.
VEHICLE IMPLEMENTATION
Figure 6.10 – Constant radius test results
Figure 6.11 – Relative roll angle front – effect of ride height
6.10
VEHICLE IMPLEMENTATION
Figure 6.12 – Relative roll angle front – effect of stiffness
Figure 6.13 – Relative roll angle rear – effect of ride height
6.11
VEHICLE IMPLEMENTATION
6.12
Figure 6.14 – Relative roll angle rear – effect of stiffness
It is concluded that the hard suspension setting results in a considerable decrease in body
roll. Further improvements might be obtained by switching the roll stiffness balance
between front and rear to counter over- or understeer.
6.4
Dynamic handling
In order to evaluate the dynamic handling characteristics of the vehicle, the ISO 3888
double lane change test was performed. The vehicle body roll angle is used as a measure
of handling.
Handling test results through the ISO 3888 double lane change at a vehicle speed of 58
km/h is indicated in Figure 6.15. At first valve selection was performed manually without
any control applied. The vehicle was driven in a specific gear against the diesel engine’s
governor in an attempt to keep the vehicle speed as constant as possible, and to ensure the
same speed for different test runs. Test speeds did however vary slightly e.g. between 57
and 61 km/h, 70 and 75 km/h and 82 to 84 km/h respectively for the three gear ratios used
for testing. The “ride” setting (soft spring and low damping), “handling” setting (stiff
spring and high damping) and baseline vehicle is compared to each other at the same
vehicle speed. It is observed that the “handling” setting results in significant
improvements in roll angle (between 61 and 78 %) compared to the baseline vehicle. The
“ride” setting is, however, very soft and results in unsatisfactory handling as expected.
Roll angle was determined in two ways. The top graph indicates the body roll angle
obtained by integrating the roll velocity measurement and correcting for drift. The bottom
graph indicates the relative roll angle between the vehicle body and the axle, calculated
from the measured relative displacement on the left and right hand struts. The values for
all four peaks, based on the relative roll angle between the body and the axle, are
VEHICLE IMPLEMENTATION
6.13
summarised in table 6.1. The “handling mode” results in significant improvements in roll
angle, compared to the baseline vehicle.
Table 6.1 – Comparison between baseline and 4S4 relative roll angles through double
lane change at 57 to 61 km/h
Peak
1
2
3
4
Baseline roll
angle [°]
1.6
-2.1
-2.3
1.8
“Handling mode”
roll angle [°]
0.6
-0.8
-0.9
0.4
“Ride mode” roll
angle [°]
3.0
-4.5
-3.7
4.0
Improvement of “Handling
mode” over baseline [%]
63
62
61
78
Figure 6.16 illustrates the effect of a 50 mm reduction in ride height on the body roll
angle at 58 km/h. There is a slight improvement in roll angle for the “handling mode”.
The major advantage is, however, seen in the “ride comfort mode” where the body roll
angle is reduced substantially to the same levels as for the baseline suspension system.
Note that the vehicle speed for the soft suspension with lowered ride height is marginally
lower than for the other three test runs.
With these large differences between the “handling mode” and the “ride comfort” mode,
it is imperative to investigate whether the RRMS control strategy will switch the 4S4
system to “handling mode” for the duration of the manoeuvre. Figure 6.17 indicates
results for RRMS control at 61 km/h. After a delay of 0.8 seconds, the system switches to
“handling mode”. It does however switch back to “ride mode” between 3.1 and 4.2
seconds. Figures 6.18 and 6.19 indicate that at speeds in the region of 75 km/h, the system
switches back to “ride mode” in some of the tests (Figure 6.19) but remains in the
“handling mode” for others (Figure 6.18). At higher speeds (above 80 km/h) the system
stays in “handling mode” as indicated in Figures 6.20 and 6.21. The switching between
settings at the lower speeds is not regarded as a problem as the vehicle is still far from the
handling limits. When approaching the handling limits at higher speeds, the RRMS
control functions correctly by switching to “handling mode” and remaining in “handling
mode” until the manoeuvre is completed. The initial switching delay is also reduced from
0.8 seconds at 61 km/h to 0.5 seconds at 75 and 0.4 seconds at 84 km/h. The vehicle will
therefore travel 13.5 m at 61 km/h and 9.2 m at 83 km/h before the 4S4 system switches
from the “ride mode” to the “handling mode”. In actual fact the valve response time of
between 0.04 and 0.09 seconds (see paragraph 4.7.6 in Chapter 4) should be added to this
initial switching delay of the control strategy.
The comparison between roll angle for the “handling mode” and RRMS control is
indicated in Figures 6.22 and 6.23. Both figures indicate that the RRMS control does not
perform as well as the “handling mode” with a definite offset noticeable in the data. This
is attributed to the delay from the start of the test until the RRMS strategy selects the
handling mode. This switching delay results in an initial roll angle on the soft suspension.
Once switching takes place, the large accumulator, and the oil in it, is isolated from the
rest of the system. The portion of oil removed, results in a differential change in ride
height between left and right and therefore an initial body roll angle. After switching
takes place, the resulting roll angle corresponds to the “handling mode”, with an offset
equal to the initial roll on the soft suspension. This offset in body roll angle is eliminated
when the system switches back to “ride comfort mode”. The roll angles at 70 and 82 km/h
however still compare favourably with the baseline roll angle at 57 km/h.
VEHICLE IMPLEMENTATION
Figure 6.15 – Body roll with 4S4 settings compared to baseline at 58 km/h
Figure 6.16 - effect of ride height on body roll at 58 km/h
6.14
VEHICLE IMPLEMENTATION
Figure 6.17 - RRMS control at 61 km/h
Figure 6.18 - RRMS control at 74 km/h
6.15
VEHICLE IMPLEMENTATION
Figure 6.19 - RRMS control at 75 km/h
Figure 6.20 - RRMS control at 83 km/h
6.16
VEHICLE IMPLEMENTATION
Figure 6.21 - RRMS control at 84 km/h
Figure 6.22 - RRMS control compared to “handling mode” at 70 km/h
6.17
VEHICLE IMPLEMENTATION
6.18
As a final comparison, Figures 6.24 and 6.25 indicate the roll angle for the handling mode
at three different speeds (Figure 6.24) and the corresponding roll angle for the RRMS
control mode (Figure 6.25). The peak-to-peak roll angles of the RRMS strategy at 73 and
83 km/h are significantly lower than at 60 km/h, primarily due to the fact that the strategy
does not switch between “handling” and “ride” modes during the manoeuvre, as it tends
to do at 60 km/h. This is favourable as it will improve ride comfort at lower speeds but, at
the onset of a handling manoeuvre, switch to handling as the vehicle speed increases,
improving high-speed vehicle stability.
6.5 Ride comfort
For the evaluation of ride comfort, the vehicle is driven over the Belgian paving (see
Figure 2.21 in Chapter 2) at five speeds. The vertical accelerations, measured at three
positions on the vehicle body and weighed according to the BS6841 standard, is used as a
measure of ride comfort.
In order to test if the RRMS control strategy performs correctly, the vehicle was driven
over the Belgian paving track at different speeds in both the “ride comfort” mode (all
soft) and the RRMS control mode. Figure 6.26 indicates that the strategy indeed switches
to the soft setting on the Belgian paving. At 4.8 seconds the driver changes direction to
avoid the very rough test track following the Belgian paving. During this manoeuvre the
RRMS control strategy switches the suspension to “handling” mode. Figure 6.27
confirms that there is no significant difference in the ride comfort, at the three measuring
positions and five speeds, when the “ride mode” is compared to the RRMS control. The
data points for “handling mode” are only indicated for the lowest speed of 17 km/h.
The RRMS strategy performs correctly for driving in a straight line over a rough road.
The “ride comfort mode” results in an improvement in ride comfort, of between 50 and
80%, compared to the “handling mode”. A significant improvement in ride comfort with
respect to the baseline values is however not experienced due to the following reasons:
i)
ii)
iii)
The current 4S4 hardware has the same damper setting front and rear while
on the baseline vehicle, front damping is considerably lower than rear
damping.
The low damping characteristic on the current 4S4 hardware has more or less
the same characteristics as the rear dampers on the baseline vehicle due to
pressure drops in the bypass valves and valve block channels. Significant
improvements in ride comfort are only expected for damper characteristics
less than 50% of the baseline values.
The baseline vehicle’s rear dampers are installed at an angle while the 4S4
dampers are vertical, thus exerting greater damper force even though the
force-velocity characteristics are similar.
Refinement of the 4S4 damper settings for the low damping characteristic is necessary
before ride comfort improvements will be noticed. This will mean enlarging the diameter
of the existing ports and channels, and fitting valves with a lower pressure drop or higher
capacity.
VEHICLE IMPLEMENTATION
Figure 6.23 - RRMS control compared to “handling mode” at 82 km/h
Figure 6.24 - Body roll for “handling mode” at different speeds
6.19
VEHICLE IMPLEMENTATION
Figure 6.25 - Body roll for RRMS control at different speeds
Figure 6.26 - RRMS control over Belgian paving at 74 km/h
6.20
VEHICLE IMPLEMENTATION
6.21
6.6 Mountain pass driving
Performance of the RRMS strategy during mountain pass driving is shown in Figure 6.28.
The RRMS control switches to “handling mode” whenever the RRMS lateral acceleration
exceeds the vertical acceleration. Subjectively the vehicle feels very stable. The
subjective improvement in ride comfort is considerable compared to “handling” mode.
6.7
City and highway driving
Results for city driving and highway driving are indicated in Figures 6.29 and 6.30
respectively. Switching to “handling mode” occurs rarely and only when cornering or
changing lanes. Again subjectively the system performs as expected with a very
noticeable improvement in ride comfort compared to “handling” mode, but also inspiring
confidence when performing handling manoeuvres.
Figure 6.27 - Ride comfort of RRMS control compared to “ride mode”
6.8
Conclusions
The 4S4 suspension system performs according to expectations. Ride comfort in the
“ride” setting, is a 50 to 80 % improvement over the “handling” setting. Body roll angle
in the “handling” setting, is a 61 to 78 % improvement over the baseline vehicle and a 47
to 90 % improvement over the “ride comfort” setting.
The RRMS control strategy performs well under most circumstances, the only drawback
being the time taken to switch to “handling” mode during the double lane change
manoeuvre. Switching between “ride comfort mode” and “handling mode” occurs
seamlessly, without the driver noticing the switching. The low damper characteristic is
not sufficiently low enough to improve the ride comfort compared to the baseline vehicle.
VEHICLE IMPLEMENTATION
6.22
The differences between “ride comfort mode” and “handling mode” are significant,
illustrating that the principle works according to expectation.
Figure 6.28 - RRMS control during mountain pass driving
Figure 6.29 – City driving
VEHICLE IMPLEMENTATION
Figure 6.30 – Highway driving
6.23
Chapter
7
7.1. Conclusions
Controllable suspension systems have been implemented successfully in top-end
passenger cars and are regarded by industry specialists as the development trend of the
future. Basic systems employ a “mode switch” where the driver manually selects a
suspension setting e.g. “comfort” or “sport”. More advanced systems react quicker and
use some form of control to determine suspension settings.
Application of controllable suspension systems to vehicles that require good off-road
capability (high ground clearance, large suspension travel and soft springs), but also good
handling and stability on smooth roads at high speeds (low centre of gravity and stiff
springs) are rare. Military wheeled vehicles, Sports utility vehicles (SUV’s) and
Crossover utility vehicles (CUV’s) all fall within this category. This thesis attempts to fill
this gap.
For off-road vehicles, a “mode switch” where the driver manually selects a suspension
setting e.g. “off-road” or “on-road” can be used, but if the design in any case offers “ride
comfort” and “handling” settings, automatic switching may just as well be employed to
get the best possible benefit from the system. This also relieves the driver from making
this decision. Furthermore, good handling is often required during off-road driving and
good ride comfort is desirable when driving on bad roads. A successful “ride comfort” vs.
“handling” decision can automatically select the required suspension settings according to
the prevailing driving conditions. An important point worth noting is that current
production systems still employ compromised characteristics, i.e. the “low” and “high”
characteristics are often not optimised for ride comfort and handling respectively. The
“low” setting is merely biased towards ride comfort but still results in acceptable
handling. The “high” setting is biased towards handling, but still gives tolerable ride
comfort. The suspension settings used in the present study are at the limits of the design
space, i.e. the “low” setting gives the best possible ride comfort, but with unacceptable
handling. The opposite holds for the “high” setting, i.e. best possible handling with
intolerable ride comfort. This configuration results in large improvements in both ride
comfort and handling respectively, but its successful application in vehicles rely on the
“ride comfort vs. handling decision”
CONCLUSIONS AND RECOMMENDATIONS
7.1.1
7.2
The ride comfort vs. handling compromise
Although no clear-cut answer is available for a metric that quantifies vehicle handling, the
body roll angle was used in this research as an indication of handling.
The following hypotheses were made:
i)
ii)
iii)
iv)
v)
vi)
Ride comfort and handling have opposing requirements in terms of spring and
damper characteristics.
Suspension requirements for off-road use differ substantially from
requirements for high-speed on-road use.
A set of passive spring and damper characteristics, called the “ride comfort
characteristic” can be obtained that results in excellent ride comfort over
prescribed off-road terrains at prescribed speeds. Additional improvements
may be possible by using “control”, but is not considered for the purposes of
this research.
A set of passive spring and damper characteristics, called the “handling
characteristic”, can be obtained that results in excellent handling for
prescribed high-speed maneuvers on good roads. Additional improvements
may be possible by the use of “control” but is not considered for the purposes
of this research.
Advanced suspension system hardware that can switch between the passive
“ride comfort” and “handling” spring and damper characteristics, can be
feasibly implemented. Response time must be rapid enough to enable control
of the sprung mass natural frequencies.
A robust decision can be made whether “ride comfort” or “handling” is
required for the prevailing conditions.
A validated, non-linear full vehicle model was used to investigate the “optimal”
characteristics for both ride comfort and handling. The conflicts between these
requirements were investigated and analysed using simulation. The following conclusions
are made based on the evidence presented:
i)
A passive suspension system is a compromise between ride comfort and
handling, as the respective requirements for ride comfort and handling are at
opposite ends of the design space.
ii)
To eliminate the “ride comfort vs. handling” compromise the following is
required:
a. At least two discrete spring characteristics are required namely:
• A stiff spring for excellent handling.
• A soft spring for excellent ride comfort.
b. At least two discrete damper characteristics are required namely:
• High damping for excellent handling.
• Low damping for excellent ride comfort.
c. The capability to rapidly switch between the two spring and two damper
characteristics.
d. A control strategy that can switch between “ride comfort” mode and
“handling” mode in a safe and predictable way.
CONCLUSIONS AND RECOMMENDATIONS
7.3
7.1.2 Possible solutions to the ride comfort vs. handling compromise
The solution proposed to solve the “ride comfort vs. handling” compromise, is to use a
twin accumulator hydropneumatic spring (two-state) combined with a two-state (on-off)
semi-active hydraulic damper. Although more than two spring and/or damper
characteristics can be incorporated, two is considered sufficient based on the simulation
results presented. The pre-requisite is however that a successful ride comfort vs. handling
decision-making strategy can be developed that will switch automatically between the
“ride comfort” and “handling” modes. This switching must be safe and quick enough to
prevent accidents, but not disturbing to the driver.
Preliminary investigation indicates that further improvements in ride comfort using
control techniques are unlikely, especially when the spring and damper characteristics
have been determined by optimising for ride comfort.
The proposed solution to the “ride comfort vs. handling” compromise is the 4 State Semiactive Suspension System or 4S4.
7.1.3 The four-state semi-active suspension system (4S4)
A possible solution was formulated and investigated in greater detail in Chapter 4 where
the design, manufacturing, testing and mathematical modelling of the proposed prototype
four-state semi-active hydropneumatic spring-damper system (4S4) system was described.
The design meets all the initial design specifications and can be fitted to the proposed test
vehicle with minor modifications to the test vehicle. The manufactured prototypes have
been extensively tested and characterised. Although several problems were identified on
the first prototype, these have been addressed and eliminated on the second prototype.
Prototype 2 meets all the dynamic requirements, except that the low damping
characteristic is too high to achieve the maximum ride comfort benefit.
A mathematical model of the suspension unit was developed and implemented in
SIMULINK. Agreement between the model predictions and the measurements was
generally good. Some aspects where the model or the quantification of its parameters
needs improvement were identified. In particular, the tests to date clearly identified the
need for a better method of quantifying the mass of gas loaded into the accumulators.
7.1.4 The ride comfort vs. handling decision
The crucial “ride comfort” vs. “handling” decision was investigated in chapter 5.
Numerous tests were performed for different driving conditions and the data thoroughly
analysed. Based on this analysis, different decision-making ideas were investigated. It is
concluded that of all the proposed strategies, only the running RMS (RRMS) strategy
appeared to work for all the test conditions.
A combination of strategies may also result in improvements, e.g. the steering angle can
be used to determine the switching point from “ride comfort” to “handling”, but switching
back to “ride comfort” might then be based on the running RMS, or simply delayed by a
fixed time to eliminate spurious switching.
CONCLUSIONS AND RECOMMENDATIONS
7.4
7.1.5 Vehicle implementation
The implementation of the proposed hardware and decision-making strategy in the
vehicle, as well as final test results is discussed in Chapter 6.
The 4S4 suspension system performs according to expectations. Switching between “ride
comfort mode” and “handling mode” occurs seamlessly without the driver being aware of
the switching. Ride comfort with the “ride” setting is 50 to 80 % better than with the
“handling” setting. The “ride comfort mode” does not present an improvement in ride
comfort compared to the baseline vehicle, because the low damping characteristic on the
4S4 prototypes is too high. Body roll angle on the “handling” setting is improved by 61 to
78 % compared to the baseline vehicle and 47 to 90 % compared to the “ride comfort”
setting.
The RRMS control strategy performs well under most circumstances, with the only
drawback being the time taken to switch to “handling” mode during the double lane
change manoeuvre.
7.1.6 Final comments
The proposed solution successfully eliminates the “ride comfort vs. handling”
compromise when designing vehicles for both on- and off-road use. The 4S4 suspension
system can be successfully implemented in hardware form, as this research has proven.
The “ride comfort vs. handling” decision can be made using easily measurable parameters
from freely available sensors.
7.2
Recommendations
Several recommendations to improve the system, and aspects that warrant further
investigation have been identified.
7.2.1 The ride comfort vs. handling compromise
The handling study, presented in chapter 2, should be expanded to include more vehicles
(especially off-road vehicles) and more drivers. This should enable better limits to be
obtained.
For the present study, suspension characteristics for optimal ride comfort were obtained
by simulating the vehicle driving over the Belgian paving at a speed of 60 km/h. Optimal
characteristics for handling were obtained by performing a double lane change at 60 km/h
on a smooth level road. The issue of combined ride comfort and handling was briefly
investigated by performing the double lane change over the Belgian paving.
Before a final verdict can be reached with respect to the optimal suspension
characteristics for ride comfort and handling respectively, it is necessary to investigate the
effects of the following aspects in greater detail:
i. Different terrain roughnesses
ii. Different vehicle speeds
iii. Different handling manoeuvres
CONCLUSIONS AND RECOMMENDATIONS
7.5
iv. Combined ride comfort and handling over a rough terrain e.g. performing the
double lane change manoeuvre over the Belgian paving
v. More design variables such as the low speed and high speed damping
characteristics, different compression and rebound characteristics, as well as the
transition point between the low- and high speed characteristic.
vi. Effect of ride height
vii. Different vehicle loading conditions
viii. Improving vehicle handling compared to passive “handling” setting by applying
control.
7.2.2 Possible solutions to the ride comfort vs. handling compromise
The effect of ride height, on the ride comfort and handling of the vehicle, should be
investigated in more detail. Limited test results discussed in chapter 6 indicate that
handling, with the soft suspension, can be considerably improved by lowering the ride
height. A control strategy to change ride height, while the vehicle is moving, should be
investigated.
7.2.3 The four-state semi-active suspension system (4S4)
The current 4S4 system can be improved in several ways namely:
i.
The “off “ characteristic for the damper is currently too high and compares to the
baseline damper value. This characteristic should be lowered significantly to
between 20% and 50% of the baseline value before substantial improvements in
ride comfort will be realized. This should be achievable by enlarging the ports and
channels in the valve block or replacing the valve with a valve of larger flow
capacity.
ii.
The low-speed “on” characteristic for the damper needs to be increased.
iii. The gas charging procedure needs to be improved to ensure that the correct mass
of gas is initially charged into the unit.
iv.
Weight and cost should be reduced before the system can be commercially viable.
v.
The 4S4 simulation model should be further verified to determine if the transient
response, during valve opening and closing, is correctly simulated.
7.2.4 The ride comfort vs. handling decision
For further improvement of the “ride comfort vs. handling” decision, the use of artificial
intelligence techniques (self organising maps, neural networks, fuzzy logic etc.) to
identify the terrain and operating conditions is suggested. If the terrain or driving
conditions can be successfully identified, then other concepts such as the steering angle
vs. vehicle speed limit values can be implemented and thresholds adapted according to
operating conditions. Possible reduction of the delay time caused by the length of the
RRMS calculation, using additional information, should be investigated.
The SIMULINK model, comprising four of these units, should be incorporated into the
ADAMS vehicle dynamics model of the sport utility vehicle in question. This will enable
investigation of control strategies using simulation instead of vehicle testing.
The possibility of controlling vehicle over- and understeer by altering the front:rear roll
stiffness balance should be investigated.
CONCLUSIONS AND RECOMMENDATIONS
7.6
The capability of the 4S4 system to reduce rollover propensity has not been investigated.
Suspension characteristics required to prevent rollover, and the effect of ride height and
control, must be determined. Early rollover warning systems might be beneficial in this
application because ride height and suspension characteristics can be adapted to operating
conditions. It might for example be possible to reduce ride height rapidly by dumping oil
in the reservoir and prevent rollover. Reduction in centre of gravity height of up to 150
mm may be achieved in this manner.
7.2.5 Vehicle implementation
Concerning implementation of the 4S4 system on a vehicle, the measuring position for the
two accelerometers needs to be investigated. If the lateral accelerometer is mounted at the
front, it might react earlier during a handling manoeuvre. The installation of a steering
angle sensor should also be investigated.
7.2.6 Additional possibilities
Additional improvements may be possible using integrated chassis control, where the
ABS braking system and automatic stability control is linked to the 4S4 suspension
control. Not only can sensors be shared, but additional information can be used e.g. the
system pressure in the 4S4 gives vertical wheel load (not true when bump or rebound
stops are in contact). This could be used as input to the brake or stability control system to
determine which wheels should be braked. This early warning could improve the
performance of the other systems.
Installation of a higher capacity hydraulic pump could facilitate slow-active control, such
as active body control or active anti-rollbars, without need for additional suspension
hardware. The suspension system can then be used as an actuator or force generator
instead of the current application as an adaptive element.
Reduced rollover propensity might require a third set of spring and damper characteristics
or a different combination e.g. soft springs with high damping. The effect of front:rear
stiffness balance has been indicated, but not used in the control yet. Switching spring and
damper characteristics individually for each wheel might also have possible benefits in
other driving scenarios that were not investigated.
Many other driving scenarios (other than the six investigated) should be investigated to
ensure that the switching strategy works under all conditions, or otherwise adapt the
strategy accordingly.
Bibliography / References
A
Abd El-Tawwab, A.M. and Crolla, D.A., 1996, An Experimental and Theoretical
Study of a Switchable Damper, SAE Technical Paper 960937, Society of Automotive
Engineers, Warrendale, 1996.
Abd El-Tawwab, 1997, Twin-Accumulator Suspension System, SAE Technical Paper
970384, Reprinted from Steering and Suspension Technology, SP-1223, Society of
Automotive Engineers, Warrendale, 1997, pp. 257-264.
Alanoly, J. and Sankar, S., 1987, A New Concept in Semi-Active Vibration Isolation,
Transactions of the ASME, Volume 109, June 1987, pp. 242-247.
Alexander, D., 2003, Cadillac SRX, Automotive Engineering International, November
2003, pp. 58-61.
Alexander, D., 2004a, Performance Air Suspension, Automotive Engineering
International, May 2004, p. 32.
Alexander, D., 2004b, Global Viewpoints – North America: Chassis Integration
Keeps Rubber on the Road, Automotive Engineering International, May 2004, p. 32.
Anon., 1998, High performance hydraulic cartridge valves and manifold systems,
HydraForce catalog 1998/99.
Anon, 2002, Getting Started Using ADAMS/View, Version 12, Mechanical Dynamics.
Anon, 2004, ZF Sachs Goes Mainstream with Active Damping, Automotive
Engineering International, December 2004, p. 44.
Anon,
2005a,
Active
suspension
Iltis,
http://www.suffield.drdcrddc.gc.ca/ResearchTech/Products/MilEng_Products/RD95010/index_e.html, accessed on
11 August 2005.
BIBLIOGRAPHY / REFERENCES
BR.2
Anon, 2005b, The Bose Suspension System- Resolving the conflict between Comfort
and control, http://www.bose.com/controller?event=VIEW_STATIC_PAGE_EVENT
&url=/learning/project_sound/suspension_challenge.jsp.
Anon, 2005c, www.stonehydraulics.com/PickAPackframe.html , accessed on 14 May
2005.
Anon, 2005d, Busak and Shamban Seal Catalogues, http://www.busakshamban.com/
accessed on 21 September 2005.
B
Besinger, F.H., Cebon, D. and Cole, D.J., 1991, An Experimental Investigation Into
the use of Semi-active Dampers on Heavy Lorries, Proceedings of the 12th IAVSD
Symposium, Lyoun, France, 26-30 August 1991.
Birch, S., 1998, Mercedes and the “Moose Test”, Global Viewpoints, Automotive
Engineering International, April 1998, pp. 11-13.
Birch, S., Yamaguchi, J. and Demmler, A., 1990, Tech Briefs – Concepts: Citroen’s
Activa 2, Automotive Engineering, Vol. 98, No. 12, pp. 55-56.
Birch, S., 1999, Actively Suspended Mercedes, Automotive Engineering International,
May 1999, pp. 38-40.
Birch, S., 2001a, Global Viewpoints – Europe embraces the AT-factor: Land Rover
introducing new technologies through Range Rover, Automotive Engineering
International, June 2001, pp. 58-60.
Birch, S., 2002a, Global vehicles: 2002 Paris Mondial De L’Automobile, Tech
highlights, Automotive Engineering International, November 2002, pp. 22-24.
Birch, S., 2002b, Global vehicles: 2002 Paris Mondial De L’Automobile, Tech
highlights, Automotive Engineering International, November 2002, p. 26.
Birch, S., 2002c, Global vehicles: Audi A8 and RS6, Automotive Engineering
International, September 2002, pp. 18-24.
Birch, S., 2003a, Global Viewpoints – Europe: Chassis systems integration
Automotive Engineering International, June 2003, pp. 58-62.
Birch, S., 2003b, Global vehicles: Frankfurt Motor Show concepts 2003, Automotive
Engineering International, November 2003, pp. 8-22.
Birch, S., 2003c, Automotive Manufacturing: Aluminium and the XJ, Automotive
Engineering International, April 2003, pp. 97-100.
Birch, S., 2003d, Global vehicles: Geneva Motor Show technical highlights,
Automotive Engineering International, May 2003, pp. 10-23.
BIBLIOGRAPHY / REFERENCES
BR.3
Birch, S., 2004a, Global vehicles: Audi Makes A6 Sportier, Automotive Engineering
International, July 2004, p. 14.
Birch, S., 2004b, Global vehicles: Mercedes-Benz CVT for new A-Class, Automotive
Engineering International, September 2004, pp. 13-16.
Böcker, M. and Neuking, R., 2001, Development of TRW’s Active Roll Control, 16th
European Mechanical Dynamics User’s Conference, 14-15 November 2001,
Berchtesgaden, Germany,
http://www.mscsoftware.com/support/library/conf/adams/euro/2001/proceedings/papers_
pdf/Paper_6.pdf accessed on 18 May 2005 at 14:50.
British Standards Institution, 1987, British Standard Guide to Measurement and
Evaluation of Human Exposure to Whole Body Mechanical Vibration and Repeated
Shock, BS 6841, 1987.
Broge, J.L, 1999, Interactive Vehicle Dynamics, Automotive Engineering International,
December 1999, pp. 47.
Buchholz, K., 2003a, Another smart truck for the U.S. Army, Automotive Engineering
International, April 2003, pp. 14-15.
Buchholz, K., 2003b, Sachs levels at the curbside, Automotive Engineering
International, December 2003, pp. 37-38.
Buchholz, K., 2003c, Global Viewpoints – North America: Body and chassis
developments, Automotive Engineering International, May 2003, pp. 65-70.
Buckner, G.D., Schuetze, K.T. and Beno, J.H., 2000, Active vehicle suspension control
using intelligent feedback linearization, Proceedings of the American Control
Conference, Chicago, Illinois, June 2002, pp. 4014-4018.
C
Carney, D., 2003a, Ferrari 360 takes up challenge, Automotive Engineering
International, October 2003, p.12.
Carney, D., 2003b, Maserati Coupe and Spyder evolve, Automotive Engineering
International, October 2003, p.14.
Carney, D., 2004b, New Vehicle Technology Highlights: Grand Ride for Grand
Cherokee, Automotive Engineering International, November 2004, pp. 54-60.
Cebon, D., (1999), Handbook of Vehicle-Road Interaction, 629.231CEBON, ISBN
9026515545, Swets and Zeitlinger.
BIBLIOGRAPHY / REFERENCES
BR.4
Choi, S.B., Lee, H.K. and Chang, E.G., 2001, Field results of a semi-active ER
suspension system associated with skyhook controller, Mechatronics Vol. 11, pp. 345353.
Chou, J.-H., Chen, S.-H. and Lee, F.-Z., 1998, Grey-Fuzzy Control for Active
Suspension Design, International Journal of Vehicle Design, Volume 19, Number 1,
1998, pp. 65-77.
Cooper, H.W. and Goldfrank, J.C., 1967, B-W-R constants and new correlations,
Hydrocarbon Processing, Vol. 46, No. 12, December 1967, pp. 141-146.
Crolla, D.A., Chen, D.C., Whitehead, J.P. and Alstead, C.J., 1998, Vehicle Handling
Assessment Using a Combined Subjective-Objective Approach, SAE Technical Paper
No. 980226.
Crolla, D.A. and Abdel-Hady, M.B.A., 1991, Semi-Active Suspension Control for a
Full Vehicle Model, SAE Technical Paper 911904, Society of Automotive Engineers,
Warrendale, 1991.
D
Dahlberg, E., 2000, A Method Determining the Dynamic Roll over Threshold of
Commercial Vehicles, SAE paper 2000-01-3492.
Darling, J. and Hickson, H.R., 1998, An experimental study of a prototype active antiroll suspension system, Vehicle System Dynamics, 29 (1998), pp. 309-329.
Data, S. and Frigero, F., 2002, Objective evaluation of handling quality, Proceedings of
the Institution of Mechanical Engineers, Vol. 216, Part D, Journal of Automobile
Engineering, pp. 297-305.
Decker, H., Schramm, W. and Kallenbach, R., 1988, A practical approach towards
advanced semi-active suspension systems, IMechE, 1988, C430/88.
De Wet, G.J., 2000, Semi-aktiewe voertuigdemper: Modellering en eksperimentele
bevestiging van solenoïde klep, “Semi-active vehicle damper: modelling and
experimental verification of solenoid valve”, Unpublished final year project,
Department of Mechanical and Aeronautical Engineering, University of Pretoria.
E
Eberle, W.R. and Steele, M.M., 1975, Investigation of Fluidically Controlled
Suspension Systems for Tracked Vehicles – Final Report, Technical Report No.
12072, TACOM Mobility Systems Laboratory, US Army Tank Automotive Command,
Warren, Michigan, September 1975.
BIBLIOGRAPHY / REFERENCES
BR.5
ElBeheiry, E.M., Karnopp, D.C., Elaraby, M.E. and Abdelraaouf, A.M., 1995a,
Advanced Ground Vehicle Suspension Systems - A Classified Bibliography, Vehicle
System Dynamics, Volume 24, Number 3, April 1995, pp. 231-258, Swets and Zeitlinger.
El Gindy, M. and Mikulcik, E.C., 1993, Sensitivity of a Vehicle’s Yaw Rate Response:
Application to a Three-axle Truck, International Journal of Vehicle Design, Vol. 14,
no. 4, pp. 325-352.
EL Gindy, M. and Ilosvai, L., 1983, Computer simulation study on a vehicle’s
directional response in some severe manoeuvres. Part 2: Steering and braking
manoeuvres, International Journal of Vehicle Dynamics, Vol. 4, No. 5, pp. 501-510.
El Gindy, M. and Mikulcik, E.C., 1993, Sensitivity of a vehicle’s yaw rate response:
application to a three-axle truck, International Journal of Vehicle Design, Vol. 14, No.
4, pp. 325-352.
Els, P.S., 1993, Die Hitteprobleem op Hidropneumatiese Veer-en-Demperstelsels
“The overheating problem on hydropneumatic spring-damper systems”,
Unpublished M.Eng Dissertation, Department of Mechanical and Aeronautical
Engineering, University of Pretoria, South Africa.
Els, P.S. and Grobbelaar, B., 1993, Investigation of the Time- and Temperature
Dependency of Hydropneumatic Suspension Systems, SAE Technical Paper Series no.
930265, Published in Vehicle Suspension and Steering Systems, SAE Special Publication
SP-256, 1993, pp. 55-65.
Els, P.S. and Grobbelaar, B., 1999, Heat Transfer Effects on Hydropneumatic
Suspension Systems, Journal of Terramechanics, Vol. 36, pp. 197-205.
Els, P.S. and Holman, T.J., 1999, Semi-Active Rotary Damper for a Heavy Off-Road
Wheeled Vehicle, Journal of Terramechanics, Volume 36, 1999, pp. 51-60.
Els, P.S. and Van Niekerk, J.L., 1999, Dynamic Modelling of an Off-Road Vehicle for
the Design of a Semi-Active, Hydropneumatic Spring-Damper System, Proceedings
of the 16th International Association for Vehicle System Dynamics (IAVSD)
Symposium: Dynamics of Vehicles on roads and Tracks, Pretoria, South Africa, August
30 to September 3, 1999.
Els, P.S., and Uys, P.E., 2003, Investigation Of The Applicability Of The Dynamic-Q
Optimisation Algorithm To Vehicle Suspension Design, Mathematical and Computer
Modeling, Vol. 37, pp. 1029-1046.
Els, P.S., 2005, The Applicability of Ride Comfort Standards to Off-Road Vehicles,
Journal of Terramechanics, Vol. 42, pp. 47-64.
Els, P.S., Uys, P.E., Snyman, J.A. and Thoresson, M.J., 2003, Obtaining Vehicle Spring
and Damper Characteristics for Improved Ride Comfort and Handling, Using
Mathematical Optimisation, 18th IAVSD Symposium, Dynamics of Vehicles on Roads
and Tracks, Extensive Summaries, IAVSD 2003, August 24-30, Kanagawa Institute of
Technology, Japan.
BIBLIOGRAPHY / REFERENCES
BR.6
Esmailzadeh, E., 1979, Servovalve-controlled Pneumatic Suspensions, Journal of
Mechanical Engineering Science, Vol. 21, No. 1.
F
Fodor, M. and Redfield, R.C., 1996, Experimental Verification of Resistance Control,
Semi-Active Damping, Vehicle System Dynamics, Volume 26, Number 2, August 1996,
pp. 143-159, Swets and Zeitlinger.
Forkenbrock, G.J. and Garrot, W. R., 2001, Light Vehicle Dynamic Roll over
Propensity Phases IV, V and VI, NHTSA Power Point Presentation, NHTSA 2001-010128, http://www-nrd.nhtsa.dot.gov/pdf/nrd-01/SAE/SAE2002/RGarrott_rollover.pdf,
accessed on 18 May 2005, 16:40.
G
Garrot, W.R., Howe, J.G. and Forkenbrock, G., 2001, Results from NHTSA’s
experimental examination of selected manoeuvres that may induce on road
untripped light vehicle rollover, NHTSA 2001-01-0131.
Gehm, R., 2003, Chrysler Pacifica, Automotive Engineering International, October
2003, p. 62-64.
Gehm, R., 2004, Tech Briefs – ZF Sachs goes mainstream with active damping,
Automotive Engineering International, June 2004, p. 20-22.
Ghazi Zadeh, A., Fahim, A. and El-Gindy, M., 1997, Neural Network and Fuzzy Logic
Applications to Vehicle Systems: Literature Survey, International Journal of Vehicle
Design, Volume 18, Number 2, 1997, ISSN 0143-3369.
Giliomee, C.L. and Els, P.S., 1998, Semi-Active Hydropneumatic Spring and Damper
System, Journal of Terramechanics, Volume 35, 1998, pp. 109-117.
Giliomee, C.L., Els, P.S. and Van Niekerk, J.L., 2005, Anelastic Model of a Twin
Accumulator Hydro-pneumatic Suspension System, R&D Journal, South African
Institution of Mechanical Engineering, Vol. 21, No. 2, July 2005.
Gillespie, T.D., 1992, Fundamentals of Vehicle Dynamics, Society of Automotive
Engineers, Inc., Warrendale, PA.
H
Hall, B.B. and Gill, K.F., 1987, Performance Evaluation of Motor Vehicle Active
Suspensions Systems, Proceedings of the Institution of Mechanical Engineers, Volume
201, Number D2, IMechE, 1987.
BIBLIOGRAPHY / REFERENCES
BR.7
Hamilton, J.M., 1985, Computer-Optimized Adaptive Suspension Technology
(COAST), IEEE Transactions on Industrial Electronics, Volume IE-32, No 4, November
1985, pp. 355-363.
Harada, H., 1997, Stability criteria of a driver-vehicle system and objective
evaluation of vehicle handling performance, International Journal of vehicle Design,
Vol. 18, No. 6., pp. 597-615.
Harty, D., 2003, Branding vehicle dynamics, Automotive Engineering International,
July 2003, pp. 53-60.
Harty, D., 2005, A review of dynamic intervention technologies and a method to
choose between them, Vehicle Dynamics Expo 2005, Open Technology Forum, 31 May
- 2 June 2005, Stuttgart Messe, Stuttgart, Germany.
Hashiyama, T., Furuhashi, T. and Uchikawa, Y. 1995, A Study on Finding Fuzzy Rules
for Semi-Active Suspension Controllers with Genetic Algorithm, In Proc. Second
IEEE Conference on Evolutionary Computation (EC-IEEE'
95), volume 1, pages 279-282.
Perth
Hedrick, J.K., Rajamani, R. and Yi, K., 1994, Observer Design for Electronic
Suspension Applications, Vehicle System Dynamics, Volume 23, Number 6, September
1994.
Hedrick, J.K. and Wormley, D.N., 1975, Active Suspensions for Ground Transport
Vehicles - A State of the Art Review, Mechanics of Transportation Suspension Systems,
ASME AMD, Volume 15, 1975, pp. 21-40.
Hennecke, D. and Zieglmeier, F.J., 1988, Frequency Dependent Variable Suspension
Damping - Theoretical Background and Practical Success, IMechE, 1988, C431/88,
pp. 101-111.
Hine, P.J. and Pearce, P.T., 1988, A Practical Intelligent Damping System, IMechE,
1988, C436/88, pp. 141-147.
Hirose, M., Matsushige, S., Buma, S. and Kamiya, K., 1988, Toyota Electronic
Modulated Suspension System for the 1986 Soarer, IEEE Transactions on Industrial
Electronics, Volume 35, Number 2, May 1988.
Hohl, G.H., 1984, Ride Comfort of Off-Road Vehicles, In Proceedings of the 8th
International Conference of the ISTVS, Vol. I of III, Cambridge, England, August 5-11,
1984.
Holdmann, P. and Holle, M., 1999, Possibilities to improve the ride and handling
performance of delivery trucks by modern mechatronic systems, JSAE Review, Vol.
20, pp. 505-510.
Holscher, R. and Huang, Z., 1991, Das komfortorientierte semiaktive
dampfungssystem, Aktive Fahrwerkstechnik, Fortschritte der Fahrzeugtechnik 10,
Vieweg and Sohn Verlaggesellschaft, Braunschweig.
BIBLIOGRAPHY / REFERENCES
BR.8
Horiuchi, S., Yuhara, N. and Takeda, H., 1989, Identification of driver/vehicle
multiloop properties for handling quality evaluation, 11th IAVSD Symposium, 21-25
Aug 1989, Supplement to Vehicle System Dynamics, Vol. 18.
Hrovat, D., 1997, Survey of Advanced Suspension Developments and Related
Optimal Control Applications Automatica, Vol 33, No. 10, pp. 1781-1817, Elsevier
Science.
Hrovat D. and Margolis, D.L., 1981, An Experimental Comparison Between SemiActive and Passive Suspensions for Air-Cushion Vehicles, International Journal of
Vehicle Design, Volume 2, Number 3, 1981, pp. 308-321.
I
Ikenaga, S., Lewis, F.L., Campos, J. and Davis, L., 2000, Active Suspension Control of
Ground Vehicle Based on Full-Vehicle Model, Proceedings of the American Control
Conference, Chicago, Illinois, June 2000, pp. 4019-4024.
International Standards Organisation, 1982, International Standard ISO 4138: Road
vehicles – Steady state circular test procedure, ISO 7401:1988(E).
International Standards Organisation, 1988, International Standard ISO 7401: Road
vehicles – Lateral transient response test methods, ISO 7401:1988(E).
International Standards Organisation, 1995, International Standard ISO 8608:
Mechanical vibration – Road surface profiles – Reporting of measured data, ISO
8608:1995(E).
International Standards Organisation, 1997, Mechanical Vibration and Shock Evaluation of Human Exposure to Whole-Body Vibration, Part 1: General
Requirements, ISO 2631-1, Second Edition, The International Organisation for
Standardisation, 15 July 1997.
International Standards Organisation, 1999, International Standard ISO 3888-1:
Passenger cars – Test track for a severe lane-change manoeuvre – Part 1: Double
lane-change, ISO 3888-1:1999(E).
International Standards Organisation, 2002, International Standard ISO 3888-2:
Passenger cars – Test track for a severe lane-change manoeuvre – Part 2: Obstacle
avoidance, ISO 3888-2:2002(E).
Ivers, D.E. and Miller, L.R., 1989, Experimental Comparison of Passive, Semi-Active
On/Off, and Semi-Active Continuous Suspensions, SAE Technical Paper 892484,
Society of Automotive Engineers, Warrendale, 1989. (Reprinted from “Advanced Truck
Suspensions”, SP-802).
BIBLIOGRAPHY / REFERENCES
BR.9
J
Janse van Rensburg, N., Steyn, J.L. and Els, P.S., 2002, Time delay in a semi-active
damper: modeling the bypass valve, Journal of Terramechanics, Volume 39, 2002, pp.
35-45.
Jolly, M.R. and Miller, L.R., 1989, The Control of Semi-Active Dampers Using
Relative Feedback Systems, SAE Technical Paper 892483, Society of Automotive
Engineers, Warrendale, 1989. (Reprinted from “Advanced Truck Suspensions”, SP-802)
Jost, K., 2002a, Continental gives Phaeton a lift, Automotive Engineering International,
November 2002, p. 49.
Jost, K., 2002b, Top technologies of the year: Delphi improves Cadillac’s ride,
Automotive Engineering International, December 2002, p. 40.
Jost, K., 2004, Segment firsts for Opel Astra, Automotive Engineering International,
January 2004, p. 12.
Jost, K., 2005, Audi Allroad Quattro, Automotive Engineering International, February
2005, p. 28-30.
K
Karnopp, D., 1968, Applications of Random Process Theory to the Design and
Testing of Ground Vehicles, Transportation Research, Vol. 2, pp. 269-278, Pergamon
Press.
Karnopp, D.C., Crosby, M.J. and Harwood, R.A., 1973, Vibration Control using Semiactive Force Generators, ASME paper 73-DET-122.
Karnopp, D., 1983, Active Damping in Road Vehicle Suspension Systems, Vehicle
System Dynamics, Volume 12, 1983, pp. 291-316.
Karnopp, D., 1990, Design Principles for Vibration Control Systems Using SemiActive Dampers, Transactions of the ASME, Volume 112, September 1990, pp. 448455.
Karnopp, D. and Margolis, D., 1984, Adaptive Suspension Concepts for Road
Vehicles, Vehicle System Dynamics, Volume 13, 1984, pp. 145-160.
Karnopp, D., Crosby, M.J. and Harwood, R.A., 1974, Vibration Control Using
Semiactive Force Generators, ASME Journal of Engineering for Industry, Vol. 98, pp.
914-918.
Kelly, K., 2001, Spyder of a different stripe – Maserati’s ragtop lives up to its Italian
heritage, Ward’s Autoworld, December 2001, pp. 65-66.
BIBLIOGRAPHY / REFERENCES
BR.10
Kizu, R., Saito, R., Matsumura, S. and Yokoya, Y., 1989, Technical Note: Suspension
Technology capable of reconciling handling stability and ride comfort, International
Journal of Vehicle Design, Vol. 10, No. 4, pp. 497-501.
Kim, H-J. and Park, Y-P., 2004, Investigation of robust roll motion control
considering varying speed and actuator dynamics, Mechatronics, Vol. 14, pp. 35-54.
Kornhuaser, A.A., 1994, Dynamic modelling of gas springs, Transactions of the ASME,
Vol. 116, September 1994. pp. 414-418.
Kojima, H., Nakano, J., Nakayama, H., Kawashima, N. and Fujimoto, H., 1991,
Development of Toyota Electronic Modulated Suspension - Two Concepts for SemiActive Suspension Control, SAE Technical Paper 911900, Society of Automotive
Engineers, Warrendale, 1991. (Reprinted from “Car Suspension Systems and Vehicle
Dynamics”, SP-878).
Krasnicki, E.J., 1981, The Experimental Performance of an “On-Off” Active
Damper, Shock and Vibration Bulletin, Number 50, May 1981, pp. 125-131.
L
Lieh, J., 1996, Development of Active Suspensions Using Velocity Feedback, SAE
Technical Paper 960935, Society of Automotive Engineers, Warrendale, 1995. (Reprinted
from “Suspension and Steering Technology”, SP-1136).
Lizell, M., 1988, Semi-Active Damping, IMechE, 1988, C429/88, pp. 83-91.
Lord
Corporation,
2005,
Magneto-Rheological
(MR)
Technology,
http://www.lord.com/defaultt.aspx?tabid=762&pid=3 accessed on 23 May 2005.
M
Margolis, D.L., 1982a, The Response of Active and Semi-Active Suspensions to
Realistic Feedback Signals, Vehicle System Dynamics, Volume 11, Number 5-6,
December 1982.
Margolis, D.L., 1982b, Semi-Active Heave and Pitch Control for Ground Vehicles,
Vehicle System Dynamics, Volume 11, Number 1, February 1982.
Masato, A., 1989, Handling characteristics of four-wheel active steering vehicles over
full manoeuvring range of lateral and longitudinal accelerations, 11th IAVSD
Symposium, 21-25 Aug 1989, Supplement to Vehicle System Dynamics, Vol. 18.
Mayne, E., 2002, Land Rover Innovation Goes From Paper to Practice,
www.wardsauto.com, accessed on 08 Jan 2002 at 07:53.
Miller, L.R., 1988a, The Effect of Hardware Limitations on an On/Off Semi-Active
Suspension, IMechE Paper number C442/88, 1988, pp. 199-206.
BIBLIOGRAPHY / REFERENCES
BR.11
Miller, L.R., 1988b, Tuning Passive, Semi-Active and Fully Active Suspension
Systems, Proceedings of the 27th Conference on Decision and Control, Austin, Texas, 79 December 1988.
Miller, L.R. and Nobles, C.M., 1988, The Design and Development of a Semi-Active
Suspension for a Military Tank, SAE Technical Paper 881133, Society of Automotive
Engineers, Warrendale, 1988.
Misselhorn, W.E., Theron, N.J. and Els, P.S., 2006, Investigation of Hardware-in-theLoop for use in suspension development, Vehicle System Dynamics, Vol. 44, No.1,
January 2006, pp. 65-81.
Mizuguchi, M., Chikamari, S., Suda, T. and Kobayashi, K., 1984, Electronic Controlled
Suspension (ECS), SAE Technical Paper 845051, Society of Automotive Engineers,
Warrendale, 1984.
Murphy, R.W., 1984, Further Development in Ride Quality, In Proceedings of the 8th
International Conference of the ISTVS, Vol I of III, Cambridge, England, August 5-11,
1984.
N
Nastasi , Ž. and Jahn, G.D., 2005, The Citroën Technical Guide,
http://www.club_xm.com/files/citroen%20guide.pdf, Accessed on 29 April 2005.
National Highway Traffic Safety Administration, 2000, Roll over prevention Docket
No. NHTSA-2000-6859 RIN 2127-AC64,
www.nhtsa.dot.gov/cars/rules/rulings/Roll over/Chapt03.html. Accessed September 2002.
Nell, S., 1993, ‘n Algemene Strategie vir die Beheer van Semi-Aktiewe Dempers in ‘n
Voertuigsuspensiestelsel, “A general strategy for the control of semi-active dampers
in a vehicle suspension system”, Unpublished PhD Thesis, Department of Mechanical
and Aeronautical Engineering, Faculty of Engineering, University of Pretoria, November
1993.
Nell, S. and Steyn, J.L., 1994, Experimental Evaluation of an Unsophysticated Two
State Semi-Active Damper, Journal of Terramechanics, Volume 31, Number 4, pp. 227238, 1994, Elsevier Science Ltd.
Nell, S. and Steyn J.L., 1998, An alternative control strategy for semi-active dampers
on off-road vehicles. Journal of Terramechanics, Vol. 35, 1998, pp 25-40.
Nell, S. and Steyn J.L., 2003, Development and experimental evaluation of
translational semi-active dampers on a high mobility off-road vehicle. Journal of
Terramechanics, Vol. 40, pp. 25-32.
BIBLIOGRAPHY / REFERENCES
BR.12
O
Ouellette,
J.,
2005,
Smart
Fluids
Move
into
the
Marketplace,
http://www.aip.org/tip/INPHFA/vol-9/iss-6/p14.html , accessed on 23 May 2005.
P
Palmeri, P.S., Moschetti, A. and Gortan, L., 1995, H-Infinity Control for Lancia
Thema Full Active Suspension System, SAE Technical Paper 950583, Society of
Automotive Engineers, Warrendale, 1995. (Reprinted from “New Developments in
Vehicle Dynamics, Simulation, and Suspension Systems”, SP-1074).
Petek, N.K., Romstadt, D.J., Lizell, M.B. and Weyenberg, T.R., 1995, Demonstration of
an Automotive Semi-Active Suspension Using Electrorheological Fluid, SAE
Technical Paper 950586, Society of Automotive Engineers, Warrendale, 1995. (Reprinted
from “New Developments in Vehicle Dynamics, Simulation, and Suspension Systems”,
SP-1074).
Pinkos, A., Shtarkman, E. and Fitzgerald, T., 1993, An Actively Damped Passenger Car
Suspension System with Low Voltage Electro-Rheological Magnetic Fluid, SAE
Technical Paper 930268, Society of Automotive Engineers, Warrendale, 1993. (Reprinted
from Special Publication SP-952), pp. 87-93).
Poley, R., 2005, DSP Control of Electro-hydraulic Servo Actuators, Texas Instruments
Application Report, SPRAA76 – January 2005 from: www.eetchina.com/ARTICLES/
2005MAR/ PDF/2005MAR21_DSP_CTRLD_ANONLINE35.PDF, accessed on 11
January 2006.
Pollard, M.G., 1983, Active Suspensions Enhance Ride Quality, Railway Gazette
International, November 1983.
Ponticel, P., 2002, New magnetorheological fluids from Lord, Automotive Engineering
International, August 2002, p. 13.
Poyser, J., 1987, Development of a Computer Controlled Suspension System,
International Journal of Vehicle Design, Volume 8, Number 1, 1987, pp. 74-86.
Pradko, F. and Lee, R.A., 1966, Vibration Comfort Criteria, SAE Technical Paper
660139, Society of Automotive Engineers, Warrendale.
R
Rajamani, R. and Hedrick, J.K., 1991, Semi-Active Suspensions - A Comparison
Between Theory and Experiments, The Dynamics of Vehicles on Roads and on Tracks,
Proceedings of the 12th IAVSD-Symposium held in Lyon, France, 26-30 August 1991,
Supplement to Vehicle System Dynamics, Volume 20, Swets & Zeitlinger.
BIBLIOGRAPHY / REFERENCES
BR.13
Rakheja, S. and Sankar, S., 1985, Vibration and Shock Isolation Performance of a
Semi-Active “On-Off” Damper, Transactions of the ASME, Volume 107, October
1985, pp. 398-403.
Reichardt, W., 1991, Correlation Analysis of Open/Closed Loop Data for Objective
Assessment of Handling Characteristics of Cars, SAE Technical Paper No. 910238.
S
Salemka, R.M. and Beck, R.R., 1975, Feasibility Analysis and Evaluation of an
Adaptive Tracked Vehicle Suspension and Control System, TACOM, Technical
Report Number 11893(LL-146), June 1975.
Sharp, R.S. and Crolla, D.A., 1987, Road Vehicle Suspension System Design - A
Review, Vehicle System Dynamics, Volume 16, number 3, 1987, Swets and Zeitlinger,
pp. 167-192.
Sharp, R.S. and Hassan, S.A., 1987, Performance and Design Considerations for
Dissipative Semi-Active Suspension Systems for Automobiles, Proceedings of the
IMechE, Volume 201, Number D2, 1987, pp. 149-153.
Sharp, R.S. and Pan, D., 1991, On active control for automobiles, 12th IAVSD
Symposium, Aug 26-30, 1991, Supplement to Vehicle System Dynamics, Vol. 20.
Silani, E., Savaresi, S.M. and Bittanti, S., 2003, Semi-active Suspensions: an Optimal
Control Strategy for a Quarter-car Model, Dipartimento di Elettronica e Informazione,
Polotechnico di Milano.
Simon, D.E., 2001, An Investigation of the Effectiveness of Skyhook Suspensions for
Controlling Roll Dynamics of Sport Utility Vehicles Using Magneto-Rheological
Dampers, PhD Dissertation. Virginia Polytechnic Institute and State University.
Soliman, A.M.A., Abd El-Tawwab, A.M. and Crolla, D.A., 1996a, Adaptive Control
Strategies for a Switchable Damper Suspension System, SAE Technical Paper 960939,
Society of Automotive Engineers, Warrendale, 1996. (Reprinted from “Suspension and
Steering Technology”, SP-1136.
Soliman, A.M.A. and Crolla, D.A., 1996b, Preview Control for a Semi-Active
Suspension System, International Journal of Vehicle Design, Volume 17, Number 4,
1996.
Speckhart, F.A. and Harrison, E., 1968, The Design of a Shock Absorber to Improve
Ride Comfort by Reducing Jerk, SAE Technical Paper 680472, Society of Automotive
Engineers, Warrendale, 1968.
Starkey, J.M., 1993, The effects of vehicle design parameters on handling frequency
response characteristics, International Journal of Vehicle Design, Vol. 14, No. 5/6, pp.
497-510.
BIBLIOGRAPHY / REFERENCES
BR.14
Stone, E. and Cebon, D., 2002, A preliminary investigation of semi-active roll control,
http://www.cvdc.org/recent_papers/StoneCebon_avec02.pdf, accessed on 19 May 2005,
07:45.
T
Temple, N.L. and Hoogterp, F.B., 1992, Semi-Active Suspension: A Mobility
Enhancement for Combat Vehicles, Proceedings of the ISTVS/FISITA 92, Seminar on
Off-road Vehicles, Institution of Mechanical Engineers, London, 9-11 June 1992.
Theron, N.J. and Els, P.S., 2005, Modelling of a Semi-active Hydropneumatic Springdamper Unit, Accepted for publication in International Journal of Vehicle Design
(IJVD), Inderscience Publishers, 3 March 2005.
Thoresson, M.J., 2003, Mathematical optimisation of the suspension system of an offroad vehicle for ride comfort and handling, Unpublished M.Eng Thesis, University of
Pretoria, Pretoria, South Africa.
Tomizuka, M. and Hedrick, J.K., 1995, Advanced Control Methods for Automotive
Applications, Vehicle System Dynamics, Volume 24, 1995, pp. 449-468, Swets and
Zeitlinger.
Trent, V. and Greene, M., 2002, A Genetic Algorithms Predictor for Vehicular
Rollover, 0-7803-7474-6/02/$17.00, IEEE, 2002
Truscott, A.J., 1994, Composite Active Suspension for Automotive Vehicles,
Computing and Control Engineering Journal, June 1994, pp. 149-154.
Tseng, H.E. and Hedrick, J.K., 1994, Semi-Active Control Laws - Optimal and Suboptimal, Vehicle System Dynamics, Volume 23, Number 7, October 1994, pp. 545-569,
Swets and Zeitlinger.
U
Uffelman, F., 1983, Automotive Stability and Handling Dynamics in Cornering and
Braking Manoeuvres, Vehicle System Dynamics, Vol. 12, pp. 203-223.
Uys, P.E., Els, P.S. and Thoresson, M.J., 2006, Criteria for Handling Measurement,
Journal of Terramechanics, Vol. 43, pp. 43-67.
Uys, P.E., Els, P.S., Thoresson, M.J., Voigt, K.G. and Combrinck, W.C., 2005,
Experimental determination of moments of inertia for an off-road vehicle in a
regular engineering laboratory, Accepted for publication in the International Journal of
Mechanical Engineering Education.
BIBLIOGRAPHY / REFERENCES
BR.15
V
Vlk. F., 1985, Handling performance of truck-trailer vehicles: A state-of-the-artsurvey, International Journal of Vehicle Design, Vol. 6, No. 3, pp. 323-361.
Voigt, K.G., 2006, Semi-active spring and damper control for ride comfort, Draft
copy of Masters degree thesis at University of Pretoria submitted to study leaders, Prof.
N.J. Theron and Mr. P.S. Els, for review.
W
Wallentowitz, H. and Holdman, P., 1997, Hardware and Software Demands on
Adjustable Shock Absorbers for Trucks and Passenger Cars, Internet http://www.ika.rwth-aachen.de/vortrag/ph-hdt accessed on 26 August 1997.
Weeks, D.A., Bresie, D.A., Beno, J.H. and Guenin, A.M., 1999, The Design of an
Electromagnetic Linear Actuator for an Active Suspension, SAE Technical paper
1999-01-0730.
Weissler, P., 2003, Continuously Controlled Chassis from Volvo, Automotive
Engineering International, August 2003, pp. 10-13 .
Williams, R.A., 1994, Electronically Controlled Automotive Suspensions, Computing
and Control Engineering Journal, June 1994, pp. 143-148.
Wright, P., 2001, Formula 1 Technology, Society of Automotive Engineers, pp. 325335.
Y
Yoshimura, T., Nakaminami, K. and Hino, J., 1997, A Semi-Active Suspension with
Dynamic Absorbers of Ground Vehicles Using Fuzzy Reasoning, International Journal
of Vehicle Design, Volume 18, Number 1, 1997.
Youn, I., 1991, Optimal Preview Control Design of Active and Semi-Active
Suspension Systems Including Jerk, SAE Technical Paper 960936, Society of
Automotive Engineers, Warrendale, 1991. (Reprinted from “Suspension and Steering
Technology”, SP-1136).
Appendix
A
HANDLING CRITERIA
APPENDIX A: HANDLING CRITERIA
Figure A-1 - Performance related to driver A – Volkswagen Golf 4 GTI on ride and
handling track
Figure A-2 - Performance related to driver B – Volkswagen Golf 4 GTI on ride and
handling track
A.2
APPENDIX A: HANDLING CRITERIA
A.3
Figure A-3 - Roll angle histograms for Drivers A and B – Volkswagen Golf 4 GTI on
ride and handling track
Figure A-4 - Lateral acceleration histogram for Drivers A and B – Volkswagen Golf 4
GTI on ride and handling track
APPENDIX A: HANDLING CRITERIA
A.4
Figure A-5 - Lateral acceleration, yaw rate and roll angle performance of a Ford Courier
on a dynamic handling track
Figure A-6 - Lateral acceleration, yaw rate and roll angle performance of a Ford Courier
on a ride and handling track
APPENDIX A: HANDLING CRITERIA
A.5
Figure A-7 – Lateral acceleration histogram for a Ford Courier on the dynamic handling
track
Figure A-8 - Roll angle histograms for a Ford Courier on a dynamic handling track
APPENDIX A: HANDLING CRITERIA
A.6
Figure A-9 - Lateral acceleration histogram of a Ford Courier on the ride and handling
track
Figure A-10 - Roll angle histogram of a Ford Courier on a ride and handling track
APPENDIX A: HANDLING CRITERIA
A.7
Figure A-11 - Lateral acceleration, yaw rate and roll angle performance of a VW Golf 4
GTI on a dynamic handling track
Figure A-12 - Lateral acceleration, yaw rate and roll angle performance of a VW Golf4
GTI on a ride and handling track
APPENDIX A: HANDLING CRITERIA
A.8
Figure A-13 - Lateral acceleration histogram for a VW Golf 4 GTI on a dynamic
handling track
Figure A-14 - Roll angle histogram for a VW Golf 4 GTI on a dynamic handling track
APPENDIX A: HANDLING CRITERIA
A.9
Figure A-15 - Lateral acceleration histogram for a VW Golf 4 GTI on a ride and handling
track
Figure A-16 - Roll angle histogram for a VW Golf 4 GTI on a ride and handling track
APPENDIX A: HANDLING CRITERIA
A.10
Figure A-17 - Lateral acceleration and yaw rate performance of a Land Rover Defender
110 on the ride and handling track (roll angle data not available)
Figure A-18 – Lateral acceleration histogram for a Land Rover Defender 110 on the ride
and handling track
Was this manual useful for you? yes no
Thank you for your participation!

* Your assessment is very important for improving the work of artificial intelligence, which forms the content of this project

Download PDF

advertisement