Vous êtes sur la page 1sur 70

# SIMULATION ANALYSIS OF AN ACTIVE

## VEHICLE SUSPENSION SYSTEM

ABSTRACT

Traditionally automotive suspension designs have been a compromise between three conflicting
criteria of road holding, load carrying and passenger comfort. The suspension system must
support the vehicle, provide directional control during handling maneuvers and provide effective
isolation of passenger payload from road disturbances. Good ride comfort requires a soft
suspension whereas insensitivity to applied load requires stiff suspension. Good handling
requires a suspension setting somewhere between the two. The objective of the project is to
develop an active control law for the suspension system , for this only a quarter model of the
vehicle suspension system with two degree of freedom is conceived and then effect of various
types of input (Step, impulse) on the passive quarter model is evaluated followed by application
of various control algorithms like PID control, Active force control using Crude approximation
and Iterative learning method, Fuzzy logic is applied on the quarter model and the results are
tabulated. The effect of inputs on different control laws is found and the results are compared
which gives the relative significance of various control algorithms. The software utilized is
MATLAB version 7.6.

CHAPTER 1
INTRODUCTION
Traditionally, automotive suspension designs have been a compromise between three conflicting
criteria of road holding, load carrying and passenger comfort. The suspension system must
support the vehicle, provide directional control during handling maneuvers and provide effective
isolation of passenger payload from road disturbances. Good ride comfort requires a soft
suspension whereas insensitivity to applied load requires stiff suspension. Good handling
requires a suspension setting somewhere between the two.
Due to these conflicting demands, suspension design has had to be something of a compromise,
largely determined by the type of use for which the vehicle was designed. Active suspensions are
considered to be a way of increasing the freedom one has to specify independently the
characteristics of load carrying, handling and ride quality
A passive suspension system has the ability to store energy via a spring and to dissipate it via a
damper. Its parameters are generally fixed, being chosen to achieve a certain level of
An active suspension system has the ability to store, dissipate and to introduce energy to the
system. It may vary its parameters depending upon operating conditions and can have knowledge
other than the strut deflection the passive system is limited to.
The following chapters deal with the step by step procedure of developing a mathematical model
for a quarter model of vehicle suspension systems followed by application of various control
laws on the quarter car model.

CHAPTER-2
LITERATURE REVIEW
K.Rajeswari, P.Lakshmis paper [5] dealt with about an intelligent method of Active Force Control
scheme applied to the vehicle suspension system of a quarter car models with actuator dynamics.
The proposed controller structure consists of three loops. Outer loop controller is used for the
target force computation to reject the road disturbances. Inner force tracking loop is used to keep
the actual force close to the target force. Active Force Control (AFC) feedback loop is used for
the estimation of force due to disturbance and it results in robustness of the proposed control
scheme. The performance of suspension system with and without AFC scheme is analyzed.
Simulation results of intelligent active force control based vehicle suspension system exhibits
better vibration isolation of vehicle body and high robustness than those obtained by the Fuzzy
Logic Controller (FLC) without AFC.
Zafer Bingul, Oguzhan Karahans paper [6] dealt with a 2 DOF planar robot controlled by
Fuzzy Logic Controller tuned with Particle Swarm Optimization. For a given trajectory, the
parameters of Mamdani-type-Fuzzy Logic Controller (the centers and the widths of the Gaussian
membership functions in inputs and output) were optimized by the particle swarm optimization
with three different cost functions. In order to compare the optimized Fuzzy Logic Controller
with different controller, the PID controller was also tuned with particle swarm optimization.
In order to test the robustness of the tuned controllers, the model parameters and the given
trajectory were changed and the white noise was added to the system. The simulation results
show that Fuzzy Logic Controller tuned by particle swarm optimization is better and more robust
than the PID tuned by particle swarm optimization for robot trajectory control.
C. Alexandru and P. Alexandrus paper [14] dealt with virtual model of the active suspension
system used for improving the dynamic behavior of a motor vehicle. The study is focused on the
design of the control system, the purpose being to minimize the effect of the road disturbances
(which are considered as perturbations for the control system). The analysis is performed for a
quarter-car model, which corresponds to the suspension system of the front wheel, by using the
DFC (Design for Control) software solution EASY5 (Engineering Analysis Systems) of MSC
Software. The controller, which is a PID based device, is designed through a parametric
optimization with the Matrix Algebra Tool (MAT), considering the gain factors as design
variables, while the design objective is to minimize the overshoot of the indicial response.
Jan Jantzens paper [8] dealt with Design of a fuzzy controller requires more design decisions
than usual, for example regarding rule base, inference engine, defuzzification, and data pre- and
post processing. This tutorial paper identifies and describes the design choices related to single-
loop fuzzy control, based on an international standard which is underway. The paper contains
also a design approach, which uses a PID controller as a starting point. A design engineer can
view the paper as an introduction to fuzzy controller design.
M. M. M. Salem and Ayman A. Alys paper [16] dealt with an active suspension system have
been proposed to improve the ride comfort. A quarter-car 2 degree-of-freedom (DOF) system is
designed and constructed on the basis of the concept of a four-wheel independent suspension to
simulate the actions of an active vehicle suspension system. The purpose of a suspension system
is to support the vehicle body and increase ride comfort. The aim of the work described in the
paper was to illustrate the application of fuzzy logic technique to the control of a continuously
damping automotive suspension system. The ride comfort is improved by means of the reduction
of the body acceleration caused by the car body when road disturbances from smooth road and
real road roughness. The paper describes also the model and controller used in the study and
discusses the vehicle response results obtained from a range of road input simulations. In the
conclusion, a comparison of active suspension fuzzy control and Proportional Integration
derivative (PID) control is shown using MATLAB simulations.

CHAPTER 3
THEORETICAL BACKGROUND
3.1 Introduction
This chapter deals with the background information of the various types of suspension system,
their mathematical modeling, and various techniques available to achieve the required response.
In this the major thrust is given on 3 types of control mechanisms for suspension systems namely
passive, semi-active, active suspension systems are described in detail. PID control, fuzzy
control and active force control (AFC) strategies applying crude approximation, iterative
learning method (ILM) are also discussed.
3.2 Suspension System:
Suspension system is a system that supports a load from above and isolates the occupants of a
vehicle from the road disturbances. Every vehicle moving on the randomly profiled road is
exposed to vibrations which are harmful both for the passengers in terms of comfort and for the
durability of the vehicle itself. Therefore the main task of a vehicle suspension is to ensure ride
comfort and road holding for a variety of road conditions and vehicle maneuvers. This in turn
would directly contribute to the safety. Shock absorption in automobiles is performed by
suspension system that carries the weight of the vehicle while attempting to reduce or eliminate
vibrations. This may be induced by a variety of sources, such as road surface irregularities,
aerodynamics forces, vibrations of the engine and driveline, and non-uniformity of the tire/wheel
assembly. Usually, road surface irregularities, ranging from potholes to random variations of the
surface elevation profile, acts as a major source that excites the vibration of the vehicle body
through the tire/wheel assembly and the suspension system (Wong, 1998).
In general, a good suspension should provide a comfortable ride and good handling within
a reasonable range of deflection. Moreover, these criteria subjectively depend on the purpose of
the vehicle. Sports cars usually have stiff, hard suspensions with poor ride quality while luxury
sedans have softer suspensions but with poor road handling capabilities. A suspension system
with proper cushioning needs to be soft against road disturbances and hard against load
disturbances. A heavily damped suspension will yield good vehicle handling, but also transfers
much of the road input to the vehicle body. When the vehicle is traveling at low speed on a rough
road or at high speed in straight line, this will be perceived as a harsh ride. The vehicle operators
may find the harsh ride objectionable, or it may physically damage vehicle. Where as a lightly
damped Suspension will yield a more comfortable ride, but would significantly reduce the
stability of the vehicle at turns, lane change maneuvers, or during negotiating an exit ramp.
Therefore, a suspension design is an art of compromise between these two goals. A good design
of a passive suspension can work up to some extent with respect to optimized riding comfort and
road holding ability, but cannot eliminate this compromise.
3.3 Types of Suspension System:
Generally there are three types of the suspension system. They are:
a) Passive suspension
b) Semi-active suspension
c) Active suspension
3.3.1 Passive suspension system
Passive suspension system is the conventional suspension system. However it is still to be found
on majority of production car. It consists two elements namely dampers and springs. The
function of the dampers in this passive suspension is to dissipate the energy and the springs is to
store the energy. If a load exerted to the spring, it will compress until the force produced by the
compression is equal to the load force. When the load is disturbed by an external force, it will
oscillate around its original position for a period of time. Dampers will absorb this oscillation so
that it would only bounce for a short period of time. Damping coefficient and spring stiffness for
this type of suspension system are fixed so that this is the major weakness as parameters for ride
comfort and good handling vary with different road surfaces, vehicle speed and disturbances.

Figure3.1: Block Diagram of Passive Suspension System
3.3.2 Semi-active Suspension
The element in the semi-active suspension system is same with passive suspension system and it
uses the same application of the active suspension system where external energy is needed in the
system. The difference is the damping coefficient can be controlled. The fully active suspension
is modified so that the actuator is only capable of dissipating power rather than supplying it as
well. The actuator then becomes a continuously variable damper which is theoretically capable
of tracking force demand signal independently of instantaneous velocity across it [3]. This
suspension system exhibits high performance while having low system cost, light system weight
and low energy consumption.

Figure 3.2: Block Diagram of Semi-Active Suspension System

3.3.3 Active suspension system
The concept of active suspension system was introduced as early as 1958.The difference
compare to conventional suspension is active suspension system able to inject energy into
vehicle dynamic system via actuators rather than dissipate energy. Active suspension can make
use of more degrees of freedom in assigning transfer functions and thus improve performance.
An active suspension system in which a force actuator is placed parallel to passive system. In
active suspension systems, sensors are used to measure the accelerations of sprung mass and
unsprung mass and the analog signals from the sensors are sent to a controller. The controller is
designed to take necessary actions to improve the performance abilities already set. The
controller amplifies the signals which are fed to the actuator to generate the required forces to
form closed loop system (active suspension system). The performance of this system is then
compared with that of the open loop system (passive suspension system. It should be noted, that
an active suspension system requires external power to function, and that there is also a
considerable penalty in complexity, reliability, cost and weight.

Figure 3.3: Block Diagram of an Active Suspension System
3.4 Mathematical Model of a Quarter car model
Quarter car model are used to derive the mathematical model of the active suspension system.
The quarter car model is popularly used in suspension analysis and design because it is simple to
analyze but yet able to capture many important characteristics of the full model. It is also
realistic enough to validate the suspension simulations
Figure (3.3) shows a quarter car vehicle passive suspension systems. Single wheel and axle is
connected to the quarter portion of the car body (sprung mass) through a passive spring and
damper. The tire (un-sprung mass) is assumed to have only the spring feature and is in contact
with the road terrain at the other end. The road terrain serves as an external disturbance input to
the system.

Figure 3.4: Quarter Car Vehicle Passive Suspension
The equations of motion for the passive system are based on Newtonian
mechanics and given as:
m

3.1
m

3.2
Where

And

## : Sprung mass and un-sprung mass respectively

: Damping coefficient

And

And

## : Displacement of sprung mass and unsprung mass respectively

: Deflection of suspension

: Deflection of tire

And

And

## : Acceleration of sprung mass and un-sprung mass respectively

Active suspension system for a quarter car models can be constructed by adding an actuator
parallel to spring and damper. Figure shows a schematic of a quarter car vehicle active
suspension system.

Figure 3.5: Quarter Car Vehicle Active Suspension
The equations of motion for an active system are as follows:
m

fa 3.3
m

fa 3.4
Where
f
a
: actuator force
Some assumptions are made in the process of modeling the active suspension system.
The assumptions are:
i) The behavior of the vehicle can be represented accurately by a quarter car model.
ii) The suspension spring stiffness and tire stiffness are linear in their operation ranges and tyre
does not leave the ground. The displacements of both the body and tyre can be measured from
the static equilibrium point.
iii) The actuator is assumed to be linear with a constant gain.
3.5 Implementation of various control laws
3.5.1Passive suspension system (NO CONTROL LAW)
Passive suspension system is an open loop control system, so we cant use any controller to this
system. Passive suspension system consists of an energy dissipating element, which is the
damper, and an energy-storing element, which is the spring and mass.
3.5.2 PID Control (ProportionalIntegralDerivative)
A proportionalintegralderivative controller (PID controller) is a generic control loop feedback
mechanism (controller). PID is the most commonly used feedback controller. A PID controller
calculates an "error" value as the difference between a measured process variable and a desired
set point. The controller attempts to minimize the error by adjusting the process control inputs.
A simple PID controller applied to a vehicle suspension system can be illustrated as shown in
Figure (3.6)
The effects of the P, I and D parameters to the system are as follows:
a) Proportional (P) action
This parameter provides a contribution which depends on the instantaneous value of the control
error. A proportional controller can control unstable plant but it provides limited performance
and non zero steady-state errors. This later limitation is due to the fact that its frequency response
is bounded for all frequencies.
b) Integral (I) action
Integral parameter gives a controller output that is proportional to the accumulated error, which
implies that it is a slow reaction mode. This characteristic is also evident in its low-pass
frequency response. The integral mode plays a fundamental role in achieving perfect plant
inversion at zero frequency. This forces the steady-state error to zero in the presence of a step
reference and disturbance.
c) Derivative (D) action
Derivative action acts on the rate of change of the control error. Consequently, it is a fast mode
which ultimately disappears in the presence of constant errors. It sometimes referred to as a
predictive mode because of its dependence on the error trend. The main limitation of the
derivative mode is its tendency to yield large control signals in response to high-frequency
control errors, such as errors induced by set-point changes or measurement noise.
3.6 Mathematical model of PID controller
The PID control scheme is named after its three correcting terms, whose sum constitutes the
manipulated variable (MV). The proportional, integral, and derivative terms are summed to
calculate the output of the PID controller. Defining u (t) as the controller output, the final form of
the PID algorithm is:
mt k

etk

et ut k

et

Figure 3.6: Block Diagram of PID
Where

=
Proportional gain, a tuning parameter

=
Integral gain, a tuning parameter

=
Derivative gain, a tuning parameter
= error (output input)

= Derivative error
t = Time or instantaneous time (the present)
3.7 Active Force Control (AFC)
Active force control strategy applied to dynamic system was proposed in the early 80s by Hewit
[8]. High robustness system can be achieved such that the system remains stable and effective
even in the presence of known or unknown disturbances, uncertainties and varied operating
conditions. AFC loop compensates the disturbance force obtained from the error between the
ideal and actual force vector. AFC has a fast decoupling property and it can be applied to various
nd
law of motion, sum of Forces (F) acting on the body is
the product of the mass (m) and the acceleration (a) of the body in the direction of applied forces.

For the dynamic system shown above the equation of motion is,

Where F
a
(constant (0-1)) is the actuator force, Q is the disturbance force; m is the mass, a
acceleration of the body. The estimated value of the disturbance force can be formulated as

The subscript denotes a measured or computed value. The error provides an adjustment signal to
the actuation system which equals -Q and it can be used to decouple the actual disturbance force
Q and hence the system will be stable even under variable external force. The actuator force and
body acceleration can be accurately measured by means of suitable transducers. Mass m can be
estimated by using intelligent methods. The disturbance force (Q) is fed forward through a
transfer function such that controller output would cancel the disturbance Q. Estimation of mass
which is needed in the AFC feed forward loop is essential for a successful AFC strategy. An
appropriate estimation of the estimated mass of the body, M was then multiplied with the
acceleration of the body, a yielding the estimated force.
If equation can be fulfilled, it is expected that very robust system can be achieved. Thus, it is the
main aim of the study to apply the AFC method to control a suspension effectively. Figure shows
a schematic of the AFC strategy applied to a dynamic system.(previously that the estimated
mass, M in Figure can be determined by a number of methods).Such as crude approximation
method, neural network, fuzzy logic, iterative learning and genetic algorithms. In this project,
Crude approximation (CA) and iterative learning method (ILM) were used.

Figure3.7: Block Diagram of AFC Strategy
3.8 AFC-Iterative Learning Method
Iterative learning method (ILM) is one of the popular methods in estimating the next value. It has
been applied to control a number of dynamic systems. As the number of iteration increases, the
track error converges to near zero data and the dynamic system is then said to operate
effectively. In this project, the proposed iterative learning algorithm takes the following form;

ATE

TE

3.6

Figure 3.8: Graphical Representation of the ILM Algorithm.
Where

## = Error value/current root of sum squared position track error

A, B = Learning parameter
3.9 Difference between Crude-Approximation and Iterative Learning Method
The essence of the AFC-CA is to determine the estimated force F.
Iterative learning method (ILM) is used to estimate the next value. It has applied to
control a number of dynamic systems. If the number of iteration increases, the track error
converges to near zero data and the dynamic system is then said to operate effectively.
3.10 Fuzzy Logic
3.10.1 Introduction
Fuzzy logic is a form of multi-valued logic derived from fuzzy set theory to deal with reasoning
that is approximate rather than accurate. In contrast with "crisp logic", where binary sets have
binary logic, fuzzy logic variables may have a truth value that ranges between 0 and 1 and is not
constrained to the two truth values of classic propositional logic. Furthermore, when linguistic
variables are used, these degrees may be managed by specific functions.
The human brain interprets imprecise and incomplete sensory information provided by
perceptive organs. Fuzzy set theory provides a systematic calculus to deal with such information
linguistically, and it performs numerical computation by using linguistic labels stipulated by
membership functions. A fuzzy inference system (FIS) when selected properly can effectively
model human expertise in a specific application. In this chapter the basic terminology of fuzzy
logic is discussed, giving a detailed explanation of all the aspects involved.
A classic set is a crisp set with a crisp boundary. For example, a classical set A of real numbers
greater than 6 can be expressed as

Where there is a clear, unambiguous boundary 6 such that if x is greater than this number, then x
belongs to this set A; or otherwise does not belong to this set. Although classical sets are suitable
for various applications they do not reflect the mature of human concepts and thoughts, which
tend to be abstract and imprecise.
In contrast to a classical set, a fuzzy set, as the name implies, is a set without a crisp boundary.
That is, the transition from belongs to a set to do not belong to a set is gradual, and this
smooth transition is characterized by membership functions that give fuzzy sets flexibility in
modeling commonly used linguistic expressions, such as the water is hot or the temperature is
high. The fuzziness does not come from the randomness of the constituent members of the set,
but from the uncertainties and imprecise nature of abstract thoughts and concepts. In logic, fuzzy
concepts are often regarded as concepts which in their application are neither completely true nor
completely false, or which are partly true and partly false.
3.11 Fuzzy Sets and its Terminology
Let U be a collection of objects denoted generically by {u}, which could be discrete or
continuous. U is called the universe of discourse and u represents the generic element of U.
a) Fuzzy set
A fuzzy set F in a universe of discourse U is characterized by a membership function

which
takes value in the interval [0,1] namely,

## [0,1].A fuzzy set may be viewed as a

generalization of the concept of an ordinary set whose membership function only takes two
values{0,1}. Thus a fuzzy set F in U may be represented as a set of ordered pairs of a generic
element u and its grade of membership function:
F = {(

3.7
When U is continuous, a fuzzy set F can be written concisely as
F =

3.8
When U is discrete, a fuzzy set F is represented as:
F =

3.9
b) Support, Crossover Point, Fuzzy singleton
The support of a fuzzy set F is the crisp set of all points u in U such that

## (u) > 0. In particular,

the element u in U at which

## is calleu the crossover point and a fuzzy set whose support

is a single point in U with

## =1.0 is referred to as fuzzy singleton.

c) Set Theoretic operations
Let A and B be two fuzzy sets in U with membership function

and

## respectively. The set

theoretic operations of union, intersection and complement for fuzzy sets are defined via their
membership functions.

d) Union
The membership function

## of the union A B is point wise defined for all by

(u) = max {

(u),

(u)} 3.10
e) Intersection
The membership function

## of the intersection A B is point wise defined for all by

(u) = min {

(u),

(u)} 3.11
f) Complement
The membership function

## of the complement of a fuzzy set A is point wise defined for all

by

(u) 3.12

g) Fuzzy Relation
An n-array fuzzy relation is a fuzzy set in

and is expressed as

={(

..

3.13

h) Sup-Star Comparison
If R and S are fuzzy relations U and V respectively, the composition of R and S is a
fuzzy relation denoted by R and defined by
R s

(u, v)*

3.14
Where * could be any operator in the class of triangular norms, namely, minimum, algebraic
product, bounded product, or drastic product.
i) Fuzzy Number
A fuzzy number F in a continuous universe U, e.g.., a real line, is a fuzzy set F in U which is
normal and convex, i.e.,

(Normal)

min

(Convex)
The use of fuzzy sets provides a basis for a systematic way for the manipulation of vague and
imprecise concepts. In particular, employ fuzzy sets to represent linguistic variables. A linguistic
variable can be regarded either as a variable whose value is a fuzzy number or as a variable
whose values are defined in linguistic terms.
j) Linguistic variables
A linguistic variable is characterized by a quintuple (x, T(x), U, G, M) in which x is the name of
variable; T(x) is the term set of x, that is the set of names of linguistic values of x with each
value being a fuzzy number defined on U; G is a syntactic rule for generating the names of
values of x; and M is a semantic rule for associating with each value its meaning.
k) Fuzzy Logic and Approximate Reasoning
In Fuzzy logic and approximate reasoning, there are two important fuzzy implication inference
rules named the generalized modus ponens (GMP) and the generalized modus tollens (GMT).
Premise 1: x is

GMP
Premise 1: y is

GMT
The fuzzy implication inference is based on the compositional rule of inference for approximate
reasoning suggested by Zadeh in 1973.
l) Sup-Star Compositional Rule of Inference
If R is a fuzzy relation in , and x is fuzzy set in U, then the Sup-Star Compositional rule
of inference asserts that the fuzzy set y in V induced by x is given by
y = x 3.15
x is the step-star components of x and R, if the star represents the minimum operator, then
definition reduces to Zadebs compositional rule of interference.

Figure 3.9: Frame work of Fuzzy Logic Controller (FLC)
3.12 Membership Functions (MF)
A fuzzy set is completely parameterized by its MF. A crisp or well defined set of values are
converted into fuzzifier set using membership functions. Since most fuzzy sets have a universe
of discourse X consisting of the real line R, it would be impractical to list all the pairs defining a
membership function. So a MF is expressed with the help of a mathematical formula. A MF can
be parameterized according to the complexity required. These also could be one dimensional or
multi dimensional. Here are a few classes of parameterized MFs of one dimension that is MFs
with a single input.
a) Triangular Membership Function
A triangular MF is specified by three parameters {a, b, c} as follows

By using min and max, an alternative expression for the preceding equation

3.16
The parameters {a, b, c} (with a<b<c) determine the x coordinates of the three corners of the
underlying triangular MF. The triangular membership functions as shown in below fig 3.3(a).

Figure 3.9 (a): Triangular membership function
b) Trapezoidal Membership Function
A trapezoidal MF is specified by four parameters {a, b, c} as follows

An alternative concise expression using min and max is

3.17
The parameter {a, b, c, d} (with a < b < c < d) determine the x coordinates of the four corners of
the underlying trapezoidal MF. The trapezoidal membership functions as shown in below fig
3.3(b).

Figure 3.9 (b): Trapezoidal membership function
c) Gaussian Membership Function
A Gaussian MF is specified by two parameters {c, }

3.18
A Gaussian MF is determined completely by c and ; c represents the MFs center and
determines the MFs width. The Gaussian membership functions as shown in below fig 3.3(c).

Figure 3.9 (c): Gaussian membership function
d) Generalized Bell Membership Function
A generalized bell MF (or bell MF) is specified by three parameters {a, b, c}

3.19
The parameter b is positive. It is also called as the Cauchy MF.
e) Sigmoidal Membership Function
A sigmoidal MF is defined by

3.20
Where a controls the slope at the crossover point x = c. Sigmoidal functions are widely used as
the activation function of artificial neural networks.
3.13 Fuzzy Inference Systems (FIS)/Fuzzy Logic Controller (FLC)
There are three main fuzzy logic inference systems (fuzzy logic approximates): Mamdani type,
Sugeno type, and Tsukamoto type. Of these Mamdani fuzzy inference system is used.
As shown in Figure 3.4, the Mamdani type of fuzzy logic controller contains four main parts,
two of which perform transformations.

Figure 3.10: Fuzzy logic inference systems.
The four parts are
Fuzzifier (transformation 1);
Knowledge base;
Inference engine (fuzzy reasoning , decision-making logic);
Defuzzifier (transformation 2).
3.13.1 Fuzzifier
The fuzzifier performs measurements of the input variables (input signals, real variables), scale
mapping and fuzzification (transformation 1). Thus all the monitored signals are scaled, and
fuzzification means that the measured signals (crisp input quantities which have numerical
values) are transformed into fuzzy quantities (which are also referred to as linguistic variables in
the literature). This transformation is performed using membership functions. In a conventional
fuzzy logic controller, the number of membership functions and the shapes of these are initially
determined by the user. A membership function has a value between 0 and 1, and it indicates the
degree of belongingness of a quantity to a fuzzy set. If it is absolutely certain that the quantity
belongs to the fuzzy set, then its value is 1, but if it is absolutely certain that it does not belong to
this set then its value is 0.
The membership functions can take many forms including triangular, Gaussian, bell shaped,
trapezoidal, etc. The knowledge base consists of the data base and the linguistic control rule
base. The data base provides the information which is used to define the linguistic control rules
and the fuzzy data manipulation in the fuzzy logic controller.
3.13.2 Fuzzication Operator
A fuzzification operator has the effect of transforming crisp data into fuzzy sets. Symbolically
X= fuzzifier (

) (3.21)
Where

is a crisp input value from a process; x is a fuzzy set; and fuzzier represents a
fuzzification operator.
3.13.3 Fuzzy conditional statements and fuzzy control rules
The rule base defines (expert rules) specifies the control goal actions by means of a set of
linguistic rules. In other words, the rule base contains rules such as would be provided by an
expert. The FLC looks at the input signals and by using the expert rules determines the
appropriate output signals (control actions). The rule base contains a set of ifthen rules. The
expert knowledge is usually of the form
IF (set of conditions are satisfied), THEN (set of consequences can be inferred).
Since the antecedents and the consequents of these IF-THEN rules are associated with fuzzy
concepts (linguistic terms), they are often called fuzzy conditional statements. In our
terminology, a fuzzy control rule is a fuzzy conditional statement in which the antecedent is a
condition and its application domain and the consequent is a control action for the system under
control. Basically, Fuzzy control rules provide a convenient way for expressing control policy
and domain knowledge. Furthermore, several linguistic variables might be involved in the
antecedents and the conclusions of these rules. When this is the case of two-input-single-output
(MISO) fuzzy systems, fuzzy control rules have the form:
R
1
: if x is A
1
and y is B
1
then z is C
1
,
R
2
: if x is A
2
and y is B
2
then z is C
2
,
.. .
.. .
R
n
: if x is A
n
and y is B
n
then z is C
n
,
Where x, y, and z are linguistic variables representing two process state variables and one control
variables; Ai, Bi, and C
i
are linguistic values of the linguistic variables x, y, and z in the
universes of discourse U, V and W, respectively, with i=1, 2,n; and an implicit sentence
connective also links the rules into a rule set or, equivalently, a rule base.
A fuzzy control rule, such as if(x is A
i
and y is B
i
) then (Z is C
i
), is implemented by a fuzzy
implication (fuzzy relation) R
i
and is defined as follows:

R1
= [

(u) and

(v)]

(w) 3.22
Where

and

is a fuzzy set

in U V;
R
i
(A
i
and B
i
) C
i
is a fuzzy implication (relation) in U V W; and notes a fuzzy
implication function.
The main methods of developing the rule base are:
Using the experience and knowledge of an expert for the application and the control
goals;
Modeling the control action of the operator;
Modeling the process;
Using a self organized fuzzy controller;
Using an artificial controller;
Using artificial neural networks.
When the initial rules are obtained by using expert physical considerations, these can be formed
by considering that the three main objectives to be achieved by the FLC are
Removal of an significant errors in the process output by suitable adjustments of the
control output;
Ensuring a smooth control action near the reference value;
Preventing the process output exceeding user specified values.
3.13.4 Sentence Connective Operators
An FLC consists of a set of fuzzy control rules which are related by the dual concepts of fuzzy
implication and the sup-star compositional rule of inference. These fuzzy control rules are
combined by using the sentence connectives and also. Since each fuzzy control rule is
represented by a fuzzy relation, the overall behavior of a fuzzy system is characterized by these
fuzzy relations. In other words, a fuzzy system can be characterized by a single fuzzy relation
which is the combination of the fuzzy relations in the rule set. The combination in question
Involves the sentence connective also,
Symbolically
R= also (

- - -,

,- -,

)
Also represents a sentence connective.

3.13.5 Compositional Operator
To infer the output z from the given process states x, y and the fuzzy relation R, the sup-star
compositional rule of inference is applied
3.23
Where is the sup-star composition?
3.13.6 Defuzzication Operator
The output of the inference process so far is a fuzzy set, specifying a possibility distribution of
control action. In the on-line control, a non-fuzzy (crisp) control action is usually required.
Consequently, one must defuzzifier the fuzzy control action (output) inferred from the fuzzy
control algorithm, namely:
Z
0
= defuzzier (z), 3.24
Where Z
0
is the non-fuzzy control output and defuzzier is the defuzzification operator.
There are many defuzzification techniques some of which are discussed here.
a) Centroid of Area
z

3.25

z Is the aggregated output MF and z is the output quantity. This is the most widely adopted
defuzzification strategy. For discrete values, above equation can be put in the form
z

3.26
Where

zare the k=1, 2 n sampled values of the aggregated output membership function.
b) Mean-Max Method (Middle of Maxima MOM Method)
In this defuzzification technique, the average output value

3.27
Is obtained, where z

## is the last value, where the output overall

membership function,

z, is maximum.

c) First of Maxima (FOM) Method
When this defuzzification technique, the first value of the overall output membership function
with maximum membership

## z degree is taken. It should be noted that this is equal to z

use
in the MOM defuzzification method.
d) Last of Maxima (LOM) Method
When this defuzzification technique, the last value of the overall output membership function
with maximum membership

## z degree is taken. It should be noted that this is equal to z

used
in the MOM defuzzification method.
In general, the defuzzification operations are time consuming and they are not easily subject to
rigorous mathematical analysis.

Figure 3.11: Structure of the FLC.

3.14 Design Parameters of the FLC
The principal design parameters for an FLC are the following:
l) Fuzzification strategies and the interpretation of a fuzzification operator (fuzzifier),
2) Data base:
a) Discretization/normalization of universes of discourse,
b) Fuzzy partition of the input and output spaces,
c) Completeness,
d) Choice of the membership function of a primary fuzzy set;

3) Rule base:
a) Choice of process state (input) variables and control (output) variables of fuzzy
Control rules
b) Source and derivation of fuzzy control rules,
c) Types of fuzzy control rules,
d) Consistency, interactivity, completeness of fuzzy control rules;
4) Decision making logic:
a) Definition of a fuzzy implication,
b) Interpretation of the sentence connective and,
c) Interpretation of the sentence connective also,
d) Definitions of a compositional operator,
e) Inference mechanism;
5) Defuzzification strategies and the interpretation of a defuzzification operator
(Defuzzifier).
3.14.2 Fuzzification Strategies
Fuzzification is related to the vagueness and imprecision in a natural language. It is a subjective
valuation which transforms a measurement into a valuation of a subjective value, and hence it
could be defined as a mapping from an observed input space to fuzzy sets in certain input
universes of discourse. Fuzzification plays an important role in dealing with uncertain
information which might be objective or subjective in nature
In fuzzy control applications, the observed data are usually crisp. Since the data manipulation in
an FLC is based on fuzzy set theory, fuzzification is necessary during an earlier stage.
Experience with the design of an FLC suggests the following principal ways of dealing with
fuzzification.
1. A fuzzification operator conceptually converts a crisp value into a fuzzy singleton
within a certain universe of discourse. Basically, a fuzzy singleton is a precise value and
hence no fuzziness is introduced by fuzzification in this case. This strategy has been
widely used in fuzzy control applications since it is natural and easy to implement. It
interprets an input

## as a fuzzy set A with the membership function,

equal to zero
except at the point

at which

equals one.
2. Observed data are disturbed by random noise. In this case, a fuzzification operator should
convert the probabilistic data into fuzzy numbers, i.e., fuzzy (possibilistic) data. In this
way, computational efficiency is enhanced since fuzzy numbers are much easier to
manipulate than random variables. The isosceles triangle was chosen to be the
fuzzification function. The vertex of this triangle corresponds to the mean value of a data
set, while the base is twice the standard deviation of the data set. In this way, triangular
fuzzy number which is convenient to manipulate. In this connection, it should be noted
that Dubois and Prade defined a objective transformation which transforms a probability
measure into a possibility measure by using the concept of the degree of necessity.
Basically, the necessity of an event, E, is the added probability of elementary events in E
over the probability assigned to the most frequent elementary event outside of E. Based
on the method of Dubois and Prade, the histogram of the measured data may be used to
estimate the membership function for the transformation of probability into possibility .
3. In large scale systems and other applications, some observations relating to the behavior
of such systems are precise, while others are measurable only in a statistical sense, and
some, referred to as hybrids, require both probabilistic and possibilistic modes of
characterization. The strategy of fuzzification in this case is to use the concept of hybrid
numbers, which involve both uncertainty (fuzzy numbers) and randomness (random
numbers). The use of hybrid number arithmetic in the design of an FLC suggests a
promising direction that is in need of further exploration
The inference engine (reasoning mechanism) is the kernel of FLC and has the capability both of
simulating human decision making based on fuzzy concepts and inferring fuzzy control actions
by using fuzzy implications and fuzzy logic rules of inference. In other words, once all the
monitored input variables are transformed into their respective linguistic variables (by
transformation 1), the inference engine evaluates the set of if then rules (given in the rule base)
and thus a result is obtained which is again a linguistic value for the linguistic variable. The
linguistic result has to be then transformed into a crisp output value of the FLC and this is why
there is a second transformation in the FLC. The second transformation is performed by the
defuzzifier which performs scale mapping as well as defuzzification. The defuzzifier yields a non
fuzzy, crisp control action from the inferred fuzzy control action by using the consequent
membership functions of the rules.
Fuzzy Logic offers several unique features that make it a particularly good choice for many
control problems.
It is inherently robust since it does not require precise, noise-free inputs and can be
programmed to fail safely if a feedback sensor quits or is destroyed. The output control is
a smooth control function despite a wide range of input variations.
Since the Fuzzy Logic controller processes user-defined rules governing the target
control system, it can be modified and tweaked easily to improve or drastically alter
system performance. New sensors can easily be incorporated into the system simply by
generating appropriate governing rules.
Fuzzy Logic is not limited to a few feedback inputs and one or two control outputs, nor is
it necessary to measure or compute rate-of-change parameters in order for it to be
implemented. Any sensor data that provides some indication of a system's actions and
reactions is sufficient. This allows the sensors to be inexpensive and imprecise thus
keeping the overall system cost and complexity low.
Fuzzy Logic can control nonlinear systems that would be difficult or impossible to model
mathematically. This opens doors for control systems that would normally be deemed
unfeasible for automation.
Simulink defines signals as the outputs of dynamic systems represented by blocks in a
Simulink diagram and by the diagram itself.
The lines in a block diagram represent mathematical relationships among the signals
defined by the block diagram. For example, a line connecting the output of block A to the
input of block B indicates that the signal output by B depends on the signal output by A.

1

The Gain block multiplies the input by a constant value
(gain). The input and the gain can each be a scalar, vector,
or matrix
2

The Scope block displays its input with respect to
simulation time. The scope block can have multiple axes
(one per port);all axes have a common time with range
with independent Y-axis
3

on the Stewart platform's motion with two control theory
techniques, the s-plane and the frequency response, both
based on the Laplace transform.
4

The Integrator block outputs the integral of its input at the
current time step
5

The Sum blocks perform addition or subtraction on its
input .this blocks can add or subtract scalar, vector or
matrix inputs. It can also collapse the elements of a single
input vector
6

The add block is an implementation of the sum block.
7

The Derivative block approximates the derivative of its
input by computing
8

The Discrete State-Space block implements the system
described
9

A subsystem block represents a subsystem of the system
that contains it .the subsystem blocks can represents a
virtual subsystem or A true (atomic) subsystem
10

The Band-Limited White Noise block generates normally
distributed random numbers that are suitable for use in
continuous or hybrid systems.
11

The sin function operates element-wise on arrays. The
function's domains and ranges include complex values.
All angles are in radians. The sine wave block provides a
sinusoid. The block can operate in either time based or
sample-based mode.
12

This block provides a step between two definable levels at
a specified time. If the simulation is less than the step time
parameter value ,the blocks output is the initial value
parameter value .for simulation time greater than or equal
to the step time ,the output is the final value parameter
value

13

The Fuzzy Logic Controller block automatically generates
a hierarchical block diagram representation of your FIS.
This automatic model generation ability is called the
Fuzzy Wizard. The block diagram representation only
uses built-in Simulink blocks and, therefore, allows for
efficient code generation.
14

The mux blocks combine its inputs into a signal output.
An input can be a scalar, vector, or matrix signal.
Depending on its inputs, the output of a mux blocks is a
vector or a composite signal, i.e., signal containing both
matrix and vector elements.

15

The Discrete Impulse block generates an impulse (the
value 1) at output sample D+1, where D is specified by
the Delay parameter (D 0). All output samples preceding
and following sample D+1 is zero.

CHAPTER 4
PROBLEM STATEMENT
4.1 Mathematical Modeling of Suspension System
The objective of this project is to find the best possible control law for the suspension of a
vehicle in presence of disturbances. The analysis is based on the quarter model of the suspension.
First only a passive (open loop) control law is developed taking 2 degree of freedom for the
quarter model of a vehicle then followed by application of various control model laws like PID,
AFC-CA, AFC-ILM and Fuzzy Logic. The software used is Simulink available in matlab7.6. A
number of assumptions that are made throughout this project are also described appropriately.
The relative importance of the proposed control algorithms is to be found. The specifications of
the different parameters used for the suspension system are given in the table 4.1
4.2 Quarter car suspension system

Figure 4.1: Quarter car vehicle passive suspension
The equations of motion for the passive suspension system are based on Newtonian mechanics
and given as
m

4.1

m

4.2
The Quarter car of the suspension system is shown below with the various specifications

Table 4.1 Table of parameters for quarter car suspension mode

4.3 Disturbance Models
There are two types of disturbances introduced to the vehicle suspension system in this study.
They are the step input, and impulse disturbances.
Figures (4.2.1) and (4.2.2) shows the disturbances i.e. step input, and impulse respectively.

Figure 4.2.1: Step input Figure 4.2.2: Impulse

Parameters Value
Sprung mass(m
s
) 170kg
Un-Sprung mass(m
s
) 25kg
Spring stiffness(k
s
) 10520 N/M
Tire stiffness(k
t
) 86240 N/M
Damper coefficient(b
s
) 1130Ns/M
CHAPTER 5
SOLUTION METHODOLOGY
The various steps involved in finding relevant solution are
1. To find the mathematical model of the quarter car
2. To develop passive (open loop) control model using simulink
3. To develop PID control model using Simulink
4. To develop AFM-CA control model using Simulink
5. To develop AFM-ILM control model using Simulink
6. To develop Fuzzy control model using Simulink
7. Comparison and discussion of results
8. Final conclusions
5.1 Mathematical Model of Quarter Car (passive suspension system)
Quarter car model is used to derive the mathematical model of the active suspension system. The
quarter car model is popularly used in suspension analysis and design because it is simple to
analyze but yet able to capture many important characteristics of the full model. It is also
realistic enough to validate the suspension simulations
Figure 5.1 shows a quarter car vehicle passive suspension system. Single wheel and axle is
connected to the quarter portion of the car body (sprung mass) through a passive spring and
damper. The tire (un-sprung mass) is assumed to have only the spring feature and is in contact
with the road terrain at the other end. The road terrain serves as an external disturbance input to
the system.
The equations of motion for the passive suspension system are based on Newtonian mechanics
and given as
m

5.1

m

5.2
The dynamic nature of the quarter model is given by the above differential equations.

Figure5.1: Quarter car vehicle passive suspension
Applying Laplace Transform with boundary conditions result in
m

s k

s z

s B

sz

s z

s 5.3
m

s k

s s k

s z

s B

s z

s k

s 5.4
m

sz

s k

sz

s
k

s m

sz

s k

sz

s k

s
k

s m

s

m

sm

s k

5.5

sk

s 5.6
Transfer Function : Laplace Transform of output / Laplace Transform of Input

5.7

Where
Z
s
(s) = Sprung mass displacement
Z
r
Z
u
(s) = unsprung mass displacement
The above equation is transfer function of a passive suspension system. It gives the relationship
between sprung mass (Z
s
r
).
When the specified values are substituted the transfer function results

5.8
In order to dampen the vibrations developed in the system , the best possible method is to find
methods to control sprung mass displacement as it is heaviest component, reducing the vibrations
of this component is equal to reduction of vibrations of the whole system.
Therefore the equations for which control laws are developed are for equation 5.8

The transient specifications like (% overshoot (M
p
)), Rise time (T
r
), peak time (T
p
), settling- time
(T
s
) are defined for only 2
nd
order systems, as the denominator of G(s) is of 4
th
order response.
It is required to check whether the response of G(s) follows 2
nd
order response or not.
To validate the 2
nd
order response of the system is represented by

The above equation 5.8 is converted to equation.5.9
us

5.9
First to check the stability of system represented by equation 5.8
Applying pole zero plot in matlab results in the figure 5.2
- 25 - 20 - 15 - 10 - 5 0
- 60
- 40
- 20
0
20
40
60
Pole- Zer o Map
Real A x is
Im
a
g
in
a
r
y

A
x
is

Figure 5.2: Zeros and poles schematic plot of passive suspension system
Zeros of G(s) : z
1
= - 23.1 z
2
= - 23.1
Poles of G(s) : p
1
= (-23.1297+55.49i)
p
2
= (-23.1297-55.49i)
p
3
= (-2.7938+7.158i)
p
4
= (-2.7938-7.158i)
From the fig.5.2 it is evident as all the zeros and the poles of G(s) are placed to the left side of
the J axis, the system is said to be stable.
Applying perfect Partial fraction to the denominator of Transfer-function G(s) results in
G(s) =

5.10
The partial fraction expansion of this G(s) involving (A, B, C, D) four terms,
A = (-292.45+95.51i)
B = (-292.45-95.51i)
C = (292.45+86.16i)
D = (292.45-86.16i)
us

5.11

Combining the first two terms and last two terms on right-hand side of the equation 5.11 we get
us

5.12
us

5.13
Equation (5.13) is in the form of

,
So we can write in the time domain as by applying inverse laplace
transforms
ft L

us 5.14
ft

5.15
The inverse Laplace transform of F (t) gives

5.16
From the above equation (5.16) it is found that part of solution decays faster than the other part
and it can be neglected. Therefore it results in a 2
nd
order system. Modified transfer function of
the quarter car model is given by equation 5.16
us

5.17
5.1.1 Developing model of Passive Suspension System
The suspension systems Simulink model is started basically for understanding the passive
suspension system of a quarter car models. The dynamical system is separated into two systems,
as this suspension system involves two degrees of freedom. This passive suspension model was
modeled in Simulink form as shown in Figure 5.3. This model was built based on the equations
(5.1) and (5.2). This is an open loop system with no feedback element for appropriate adjustment
of parameters. Simulink model for unit Step type of Input .The range of the input and all the
parameters are given in table 4.1.
Unit Step response:

Figure 5.3: Simulink Model of Passive Suspension System for Step

Figure 5.4: Passive Suspension Response to Step Input Disturbance
Figure 5.4 shows the response of passive suspension system to the unit step input. Response
shown is not initially stable and needs some time to settle down .This causes the sprung mass
displacement to occur for a long period of time. The transient and steady state response values
for the system are given in table 5.1. The damping ratio of the system for step response is 0.547.

Table 5.1: Step Input Transient Responses of a Passive (Open-Loop) Suspension System
Type of
input
Rise time
(t
r
) in Sec
Peak time
(t
s
) in Sec
% overshoot
(Mp)
Settling time
(t
s
) in Sec
(e
ss
)
Step 0.1390 0.2780 44.60 1.980 0.0068

Impulse input to the System:

Figure 5.9: Simulink Model of a passive Suspension System for impulse
Impulse response of the System

Figure 5.6: Passive Suspension Response of Impulse Input Disturbance
Figure 5.6 shows the response of passive suspension system to the impulse input. It takes some
amount of time to respond and also as the settling time is about 3 seconds. The transient and
steady state response values for the system are given in table 5.2.
Table 5.2: Impulse Transient Responses of a Passive (Open-Loop) Suspension System
Type of input Peak time(t
p
) in Sec

Settling time (t
s
) in Sec

Impulse 0.2781 2.995

5.2 Mathematical Model of Quarter Car Active Suspension System
Active suspension system for a quarter car models can be constructed by adding an actuator
parallel to spring and damper. Figure 5.7 shows a schematic of a quarter car vehicle active
suspension system. The equations of motion for the active suspension system are based on
Newtonian mechanics and given as
m

fa 5.18

m

fa 5.19

Figure 5.7: Quarter car vehicle active suspension system
For an active suspension system the PID control equation is,

mt k

etk

et ut k

et

5.20
PID gain used are as follows; K
p
= 10000, K
i
= 8 and K
d
= 6, the values are choose by a method
called manual tuning.

5.2.1 PID tuning
Tuning a control loop is the adjustment of its control parameters (gain/proportional band, integral
gain/reset, derivative gain/rate) to the optimum values for the desired control response.
5.2.2 Manual tuning
If the system must remain online, one tuning method is to first set K
i
and K
d
values to zero.
Increase the K
p
until the output of the loop oscillates, then the K
p
should be set to approximately
half of that value for a "quarter amplitude decay" type response. Then increase K
i
until any offset
is correct in sufficient time for the process. However, too much K
i
will cause instability. Finally,
increase K
d
, if required, until the loop is acceptably quick to reach its reference after a load
disturbance. However, too much K
d
will cause excessive response and overshoot. A fast PID
loop tuning usually overshoots slightly to reach the set point more quickly; however, some
systems cannot accept overshoot, in which an over-damped closed-loop system is required,
which will require a K
p
setting significantly less than half that of the K
p
setting causing
oscillation.
Table 5.3: Effects of Increasing a Parameter Independently

5.2.3 Transfer function of the an active suspension system with PID
To get transfer function of an active suspension system we are multiplying open loop and PID
transfer functions (as both open loop and pid control transfer functions are in serial),
parameters
Rise time
(t
r
) Sec
Overshoot
%(Mp)
Settling
time(t
s
) Sec
K
p
Decrease Increase Small change Decrease

K
i
Decrease Increase Increase
Decrease
significantly
K
d

Minor
decrease
Minor
decrease
Minor
decrease
No effect in
theory
Improve if
K
d
small
us us

5.21
Transfer function of an active suspension system with PID
u

5.22
The above equation 5.22 is converted to equation.5.23
u

5.23
First to check the stability of system represented by equation 5.22
Applying pole zero plot in matlab results in the figure 5.8
- 25 - 20 - 15 - 10 - 5 0
- 60
- 40
- 20
0
20
40
60
Pole-Zer o Map
Real A x is
Im
a
g
in
a
r
y

A
x
is

Figure 5.8: Zeros and poles schematic plot of an active suspension system

Zeros of G(s) : Z
1
= -1667; Z
2
= - 0.0008; Z
3
= - 22.6; Z
4
= - 22.6
Poles of G(s) : P
1
= 0; P
2
= - 2.794; P
3
= - 2.794; P
4
= - 23.13; P
5
= - 23.13
From the figure.5.8 it is evident as all the zeros and the poles of G
1
(s) are placed to the left side
of the J axis, the system is said to be stable.
5.2.4 Developing a Simulink Model of an Active Suspension System
Active suspension system requires an actuator force to provide a better ride and handling than
the passive suspension system. The actuator force, F
a
is an additional input to the suspension
system model. The Simulink model Fig.5.9 was built based on the equation 5.17 and 5.18.The
actuator force is controlled by the PID controller which involves a feedback loop.
Unit step response:

Figure 5.9: Simulink model of active suspension system using step input

Figure 5.10: active suspension response to step input disturbance
Figure 5.10 shows the response given by an active suspension to the step input. PID controller is
used in this suspension and it is a closed loop system. PID is tuned optimizely so that the
response for the step input disturbance is good compared to passive suspension system. PID
gains used are as follows; K
p
= 1000, K
i
= 8 and K
d
= 6.The transient and steady state response
values for the system with PID tuning are given in table 5.4. The damping ratio for step response
is 0.527.
Table 5.4: Step Input Transient Responses of an Active Suspension System
Type of
input
Rise time
(t
r
)in Sec
Peak time
(t
s
)in Sec
% overshoot
(Mp)
Settling time
(t
s
)in Sec
error
(e
ss
)
Step(1sec) 0.1500 0.3000 33.40 1.850 0.01

Figure 5.11: Simulink Model of an Active Suspension System to Impulse input

Figure 5.12: Active Suspension Response to Impulse Disturbance
Figure 5.12 show the response given by active suspension to the impulse input. It takes some
amount of time to respond and also as the settling time (t
s
) is about 2.842 seconds. The transient
and steady state response values for the system with PID tuning are given in table 5.5.
Table 5.5: Impulse Input Transient Responses for an Active Suspension System
Type of input Peak time(t
p
)in Sec Settling time (t
s
)in Sec
Impulse 0.300 2.842

5.3 Active Suspension System Model with AFC-CA Strategy
Instead of using only PID controller, active suspension system in Simulink model was further
develop by introduced active force control with crude approximation (AFC-CA) in the system.
This model is shown in Figure (5.13).The AFC-CA control Simulink blocks includes the
estimated mass gain, parameter 1/Ka gain and the percentage of AFC application gain. The input
to the AFC control is the sprung mass acceleration and the output is summed with the PID
controller output before multiply with the actuator gain which finally results the generated
actuator force. Crude approximation method is used to estimate the estimated mass in the AFC.
Here estimated mass is E
m
(100).
To get transfer function of an active suspension system with (AFC-CA) we are multiplying
active force (F= f
a
- m*a) with active suspension system transfer functions,
Where
F = Active Force (0.9937)
F
a
= Actuator Force 1m/sec
m = Estimated Mass of 100kg
a = Acceleration (0.001629m/sec
2
)
G
2
(s) = G
1
(s)*F
Transfer function of an active suspension system with (AFC-CA)
u

5.24
The above equation 5.24 is converted to equation.5.25
u

5.25
First to check the stability of system represented by equation 5.24
Applying pole zero plot in matlab results in the figure 5.13
The zeros and poles of the active suspension system are plotted in figure 5.13 based on equation
5.24.
- 180 - 160 - 140 - 120 - 100 - 80 - 60 - 40 - 20 0
- 60
- 40
- 20
0
20
40
60
Pole- Zer o Map
Real A xis
I
m
a
g
i
n
a
r
y

A
x
i
s

Figure 5.13: Zeros and Poles Schematic Plot of an Active (AFC-CA)

Zeros of G(s) : z
1
= -166.4; z
2
= - 22.59+57.96i ; z
3
= - 22.59-57.96i;
Z4
= -22.6

Poles of G(s) : P
1
= 0; P
2
= - 2.794; P
3
= - 2.794; P
4
= - 23.13; P
5
= - 23.13
In the above equation 5.24 as all the zeros and the poles of G
2
(s) are placed to the left side of the
J axis, the system is said to be stable.

5.3.1Devloping and Simulink Model of an Active (AFC-CA) Suspension System
Unit step response:

Figure 5.14: Simulink model of an active (AFC-CA) suspension system to step-input

Figure 5.15: AFC-CA suspension system response to step input disturbance
Figure 5.15 shows the response given by active suspension with AFC strategy and crude
approximation method to the step input. Response, under step input is much better compare to
PID controller. AFC gives a good result although disturbance is change. Estimated mass used in
this simulation is 100 kg. The transient and steady state response values for the system are given
in table 5.6. Estimated mass is E
m
100 kg. The damping ratio for step response is 0.521.
Table 5.6: Step Input Transient Responses of an Active (AFC-CA) Suspension System
Type of
input
Rise time
(t
r
) in Sec
Peak time
(t
s
)in Sec
% overshoot
(Mp)
Settling time
(t
s
) in Sec
(e
ss
)
Step 0.2480 0.4960 30.70 2.445 0.0096

Impulse input Reponse:
Figure 5.16: Simulink model of (AFC-CA) suspension system with impulse-input

Figure 5.17: AFC-CA suspension response to impulse disturbance
Figure 5.17 shows the response given by active suspension with AFC strategy and crude
approximation method to the impulse input. It takes some amount of time to respond and also as
the settling time (t
s
) is about 3.686 seconds. The transient and steady state response values for the
system with PID tuning are given in table 5.7.
Table 5.7: Impulse Input Transient Responses of an Active (AFC-CA) Suspension System
Type of input Peak time(t
p
) in Sec Settling time (t
s
) in Sec
Impulse 0.5 3.686

5.4 Active Suspension System Model with AFC-ILM
To estimate the estimated mass for AFC, systematic method such as intelligent method is
appropriate to use rather than trial and error. One of the intelligent methods is iterative learning
method (ILM).This type of method applied with AFC can be modeled as shown in Figure 5.19.
Equation 5.25 is a transfer function of (AFC-ILM). The estimated mass in ILM method is U
k+1
(12.6) and the initial mass (M
I
) is (10)
To get transfer function of an active suspension system with (AFC-ILM) we are multiplying
estimated mass with active suspension system transfer functions
G
3
(s) = G
2
(s) * U
k+1

Transfer function of an active suspension system with (AFC-ILM) is

5.26
The above equation 5.26 is converted to equation.5.27

5.27
First to check the stability of system represented by equation 5.26
Applying pole zero plot in matlab results in the figure 5.18
The zeros and poles of the active suspension system are plotted in figure 5.18 based on equation
5.26.
Pole- Zero Map
Real A x is
I
m
a
g
i
n
a
r
y

A
x
i
s
- 180 - 160 - 140 - 120 - 100 - 80 - 60 - 40 - 20 0
- 60
- 40
- 20
0
20
40
60

Figure 5.18: Zeros and poles schematic plot of an active (AFC-ILM)
Zeros of G(s) : z
1
= - 166.4; z
2
= - 22.59 + 57.96i ; z
3
= - 22.59 - 57.96i; z
4
= - 0.0008

Poles of G(s) : p
1
= 0; p
2
= - 2.794; p
3
= - 2.794; p
4
= - 23.13; p
5
= - 23.13
In the above equation 5.25 as all the zeros and the poles of G(s) are placed to the left side of the
J axis, the system to be said stable.
5.4.1 Developing the simulink of an active (AFC-ILM) suspension system
Unit step response:

Figure 5.19: Simulink model of AFC-ILM suspension system with step-input

Figure 5.20: Subsystem of iterative learning method

Figure 5.21: AFC-ILM suspension response to step input disturbance
Figure 5.21 shows the response of active suspension with AFC strategy and iterative learning
method to the step input. The response of (AFC-ILM) is good as AFC-CA suspension system.
Value of learning parameter A is set to 4 and B = 5.Initial condition used is 10. The transient and
steady state response values for the system are given in table 5.8. Estimated mass is E
m
100 kg.
The damping ratio for step response is 0.514.

Table 5.8: Step Input Transient Responses of an Active (AFC-ILM) Suspension System
Type of
input
Rise time
(t
r
)Sec
Peak time
(t
s
)Sec
% overshoot
(Mp)
Settling time
(t
s
)Sec
(e
ss
)
Step 0.1647 0.3295 27.90 1.955 0.0023

Impulse input to system

Figure 5.22: Simulink model of AFC-ILM suspension system with impulse-input

Figure 5.23: AFC-ILM suspension response to impulse input disturbance
Figure 5.23 shows the response of active suspension with AFC strategy and iterative learning
method to the impulse input. It takes some amount of time to respond and also as the settling
time (t
s
) is about 2.804 seconds is good as AFC-CA suspension system. The transient and steady
state response values for the system are given in table 5.9.
Table 5.9: Impulse Input Transient Responses of (AFC-ILM) Suspension System
Type Of Input Peak Time(t
p
)in Sec Settling Time (t
s
)in Sec
Impulse 0.3295 2.804

5.5 Design of a Fuzzy Controller
It is necessary to formulate fuzzy rules, before simulating for comparison purpose. The input
linguistic variables chosen for the fuzzy controller are sprung mass velocity and the suspension
velocity (relative velocity of sprung mass to un-sprung mass).The output of the controller is the
damping coefficient of the variable damper. The universe of discourse for both the input
variables the sprung mass velocity and suspension velocity was divided in to three sections with
the following linguistic variables. Positive (p), zero (z) and negative (n). The universe of
discourse for the output variable, damping coefficient of the damper, was divided in to three
sections with the following linguistic variables, small (s), medium (m) and large.
Trapezoidal membership functions are used for the linguistic variables because they produce
smoother control action due to flatness at the top of the trapezoidal shape. The objective of
control is contained in the fuzzy rule base in the form of the linguistic variables using the fuzzy
conditional statement. It is composed of the antecedent (IF, AND clause) and the consequent
(THEN-clause). For example, one of the control rules can be stated as If the relative velocity is
negative and the sprung mass velocity is positive THEN the damping coefficient is small. Using
these linguistic variables, a set of fuzzy rules was developed. The fuzzy rule base consisted of 9
rules. The fuzzy reasoning inference procedure used was max-min. The defuzzification
procedure employed was bisector.
Figure 5.24 shows the membership function with body velocity as input, ranging between
[-1.5 1.5], the output is a value (degree of membership) which lies in the range [0, 1].

Figure 5.24: Membership function/input/body velocity
Figure 5.25 shows the membership function with relative velocity as input, ranging between
[-1.5, 1.5], the output is a value (degree of membership) which lies in the range [0, 1].

Figure 5.25: Membership function/input/mass velocity

Figure 5.26 illustrates the output membership function, damping coefficient with damper range
[0, 2250].The output is a numerical value (degree of membership) between 0 and 1.

Figure 5.26: Membership function with damping coefficient
The fuzzy base rule consists of nine rules. These rules are shown in below table (5.10) for the
set. The results of the simulation will be discussed in chapter 6, Results and Discussion
Table 5.10: Fuzzy Base Rules
Relative
Velocity/Body
Velocity
Negative Zero Positive
Negative Large Medium Small
Zero Medium Small Medium
Positive Small Medium Large

Figure 5.27 shows how fuzzy logic surface viewer.

Figure 5.27: fuzzy logic surface viewer
It is found that the set of fuzzy rules, when applied to the active suspension system performed
well, following an assumption.
5.6 Active Suspension System Using Fuzzy Controller
In active suspension system, for vary the coefficient of damper in active suspension system,
fuzzy logic controller was employed. In simulink block system, fuzzy blocks exists, simply
called fuzzy logic controller .Figure shows the simulink arrangement of active suspension system
with fuzzy logic controller toolbox. All the simulations are done using MATLAB/SIMULINK.
Simulink model is a combination of various simulink blocks, which are connected by means of
connectors (arrows) according to the process (mathematical relations in the present case).
Simulink models may also contain subsystems, in which certain process can be represented within
a single block.

5.6.1 Developing the Simulink Model of Fuzzy Active Suspension System
Unit step response:

Figure 5.28: Simulink Model of an Active Suspension with Step Input

Figure 5.29: Active Suspension System Response to Step Disturbance

Figure 5.29 shows the response given by active suspension to the step input. Fuzzy controller is
used in this suspension system and it is a close loop system. The response for the step input
disturbance is good compared to PID active suspension system. The transient and steady state
response values for the system with fuzzy controller are given in table 5.10. Estimated mass is
E
m
100 kg. The damping ratio for step response is 0.489.
Table 5.10: Step Input Transient Responses of an Active Suspension System.
Type
of
input
Rise time
(t
r
)in Sec
Peak time
(t
s
)in Sec
% overshoot
(Mp)
Settling time
(t
s
)in Sec
(e
ss
)
step 0.2269 0.4528 20.20 1.885 0.0005

Figure 5.30: Impulse Input Simulink Model of an Active Suspension System

Figure 5.31: Active suspension system response to impulse disturbance
Figure 5.31shows the response given by active suspension to the impulse input. It takes some
amount of time to respond and also as the settling time (t
s
) is about 1.985 seconds is good as
active PID suspension system. Fuzzy controller is used in this suspension system. It is a close
loop system.The transient and steady state response values for the system are given in table 5.11.
Table 5.11: Impulse Input Transient Responses of an Active Suspension System

5.7 Active suspension system with fuzzy logic controller using AFC-CA
Instead of using only fuzzy controller, active suspension system in Simulink model was further
develop by introduced active force control with crude approximation (AFC-CA) in the system.
This model is shown in Figure. The AFC-CA control Simulink blocks includes the estimated
mass gain, parameter 1/Ka gain and the percentage of AFC application gain. The input to the
AFC control is the sprung mass acceleration and the output is summed with the fuzzy controller
output before multiply with the actuator gain which finally results the generated actuator force.
Crude approximation method is used to estimate the estimated mass in the AFC. Here estimated
mass is E
m
100kg.
Type of input Peak time (t
p
)in Sec Settling time (t
s
)in Sec
Impulse 0.4538 1.985
5.7.1 Developing the Simulink Model of Fuzzy Active (AFC-CA) Suspension System
Unit step response:

Figure 5.32: Step input Simulink model of an active suspension (AFC-CA) system

Figure 5.33: Active suspension system (AFC-CA) response to step disturbance
Figure 5.33 shows the response given by active suspension with AFC strategy and crude
approximation method to the unit step input. Fuzzy controller is used in this suspension system
and it is a close loop system. The response for the step input disturbance is good compared to
active (AFC-CA) PID suspension system. The transient and steady state response values for the
system with fuzzy controller are given in table 5.12. Estimated mass is E
m
100 kg. The damping
ratio for step response is 0.520
Table 5.12: Step Input Transient Responses of an Active (AFC-CA) Suspension System
Type of
input
Rise time
(t
r
)in Sec
Peak time
(t
s
)in Sec
% overshoot
(Mp)
Settling time
(t
s
)in Sec
(e
ss
)
step 0.3024 0.6049 30.06 4.987 0.00016

Figure 5.34: Active Suspension (AFC-CA) System Simulink Model Using Impulse

Figure 5.35: Active (AFC-CA) suspension system response to impulse disturbance
Figure 5.35 shows the response given by active suspension (AFC-CA) to the impulse input
respectively. Fuzzy controller is used in this suspension and it is a close loop system. From the
response graph one can deduce the specifications of transient responses of a given input.
Table 5.13: Impulse Transient Responses of fuzzy Active (AFC-CA) Suspension System
Type of input Peak time (t
p
)in Sec Settling time (t
s
)in Sec
Impulse 0.5123 3.665

CHAPTER 6
RESULTS AND DISCUSSIONS
Introduction
This chapter presents the simulated results of a quarter car suspension responses results that are
described in the previous chapter. The response of sprung mass with respect to applied load is
simulated. Comparative results of various control laws for quarter car model suspension system
for two types of inputs (impulse and step) are given below.
Table 6.1: Comparison Table of response of different type of Suspension Systems for step input:
Type of
system
Rise Time
(t
r
)in Sec
Peak Time
(t
p
)in Sec
% Peak over
shoot (M
p
)
Settling
Time
(t
s
)in Sec
error (e
ss
)

Table6.2: Comparison Table of response of different type of Suspension Systems for step input

Type of system
Peak time (T
p
)in Sec Settling time (T
s
)in Sec
Passive
0.2781 2.995
Active(PID)
0.3000 2.842
Active(AFC-CA)
0.5 3.686
Active(AFC-ILM)
0.3295 2.804
Active(FLC)
0.4538 1.985
Active (AFC-CA)(FLC)
0.5123 3.665

Passive 0.1390 0.2780 44.60 1.980 0.0068
PID control 0.1500 0.3000 33.40 1.856 0.0100
AFC-CA 0.2480 0.4960 30.70 2.445
0.0096
AFC-ILM 0.1647 0.3295 27.90 1.955 0.0023
Active
(FLC)
0.2269 0.4528 20.20 1.885
0.0005
FLC with
AFC-CA
0.3024 0.6049 30.06 4.987
0.00016

CHAPTER-7
CONCLUSIONS
The implementation of an active suspension system, using different control (PID, AFC,
FUZZY) laws to the vehicle suspension system has been successfully done in simulation.
The simulation demonstrates that AFC-CA and AFC-ILM, fuzzy controller give better
performance compared to PID controller.
The most important thing in AFC is to estimate the initial mass. If the approximation is
done accurately for initial mass, AFC will give a better performance.
Crude approximation method is easier than iterative learning. In this the initial mass
value can change directly.
Iterative learning method is more intelligent to estimate the initial mass value as it will
iterate and decrease the error until it get the right value. But the problem in this method is
to tune the learning parameter.
The % overshoot and settling time are best indicators of transient response (speed) of the
system which show fuzzy controller is the best control law for step response.
The best possible % overshoot of 20.2 is obtained by active fuzzy logic controller for
step response which is about 54.7 % better than the uncompensated system.
The system to settle down fast is obtained by active fuzzy logic controller for step
response which is about 5.05 % better than the uncompensated system.
Steady state error is reduced from a value of 0.0068 to 0.00016 (Fuzzy control) with the
implementation of active control laws, which is 97.6 % improvement.
The damping ratio) for uncompensated system is 0.547, which is reduced to a
minimum value of 0.489 (Fuzzy control) with the use of active control laws.
The best possible settling time (t
s
) for impulse response is 1.985sec. It is attained in
active fuzzy logic controller.
The second best possible impulse response settling time (t
s
) is obtained 2.804sec in
iterative learning method.
The system to settle down fast is obtained by active fuzzy logic controller for impulse
response which is about 33.72 % better than the uncompensated system.
FUTURE SCOPE
The above work is only simulation work, in order to check the validity of the results
experimentation has to be done.
The values for active force control law are randomly taken, the accurate approximate
initial mass of AFC give better understanding performance values for various inputs.
The analysis is done for quarter model, it can be extended to full car model and the
degree of freedom for the system

REFERENCES
1. M. Senthil Kumar, S. Vijayarangan, Analytical and experimental studies on active
suspension system of light passenger vehicle to improve ride comfort,PSG College of Technology,
Coimbatore 641 004, India, Proceedings.
of Passenger Car Suspension for Ride and Road Holding.
3. Andrew J.Barr, The Fuzzy Logic Control of an Active Suspension System, B.S.M.E. Grove
City College (1994), MS Thesis, Youngstown State University (1996).
4. K.Rajeswari , P.Lakshmi, Simulation of Suspension System with Intelligent Active Force
Control(2010 IEEE)( 978-0-7695-4201-0/10)
5. A Fuzzy Logic Controller tuned with PSO for 2 DOF robot trajectory control, Zafer
Bingul, Oguzhan Karahan. Department of Mechatronics Engineering, Kocaeli
University, Kocaeli, Turkey Expert Systems with Applications 38 (2011) 10171031
6. Technical University of Denmark, Department of Automation, Bldg 326, DK-2800
Lyngby, DENMARK.Tech. Report no 98-E 864 (design), 19 Aug 1998.
7. Mailah.M & Priyandoko.G, Simulation of a suspension system using adaptive
fuzzy Active Force Controller , (2007).
8. Maziah Mohamad, Musa Mailah, Abdul Halim Muhaimin,Vibration Control
of Mechanical Suspension System Using Active Force Control.
9. F. Chevrie F. Guly Fuzzy logic ECT 191 first issued, December 1998 Cahier Technique
Schneider no 191.
10. M.A.Salim1, A.Noordin2, M.Z.Akop3 Implementation of Active Force Control to
Reduce Vibration Displacement in Active Mass Damper during Seismic Activity
Canadian Journal on Science and Engineering Mathematics Vol. 2, No. 1 January 2011.
11. Reasoning with uncertainty-Fuzzy Reasoning Version 2 CSE IIT, Kharagpur.
12. Mailah.M & Priyandoko.G, Simulation of a suspension system using adaptive
fuzzy Active Force Controller , (2007). ISSN (1725-4529).
13. C. Alexandru, and P. Alexandru Control Strategy for an Active Suspension System
World Academy of Science, Engineering and Technology 79 2011 Product Design and
Robotics Department from the Transilvania University of Brasov.
14. Simulation and experimental analysis of an active Vehicle suspension system, Mohd
Rizal bin ahmad manap University technology Malaysia. Faculty of Mechanical
Engineering.
15. M. M. M. Salem and Ayman A. Aly Fuzzy Control of a Quarter-Car Suspension System,
M. M. M. Salem is with Auto. Department, Faculty of Engineering, El-Minia University,
Egypt. Ayman A. Aly is with Mechatronics Section, Faculty of Engineering, Assuit
University, Egypt ,World Academy of Science, Engineering and Technology 53 2009