Fuzzy Information and Engineering

ISSN: 1616-8658 (Print) 1616-8666 (Online) Journal homepage: https://www.tandfonline.com/loi/tfie20

Vibration Control of Active Vehicle Suspension

System Using Fuzzy Logic Algorithm

Essam Allam, Hesham Fath Elbab, Magdy Abdel Hady & Shawki Abouel-
Seoud

To cite this article: Essam Allam, Hesham Fath Elbab, Magdy Abdel Hady & Shawki Abouel-
Seoud (2010) Vibration Control of Active Vehicle Suspension System Using Fuzzy Logic
Algorithm, Fuzzy Information and Engineering, 2:4, 361-387, DOI: 10.1007/s12543-010-0056-3

To link to this article: https://doi.org/10.1007/s12543-010-0056-3

Published online: 25 Jan 2019.

Submit your article to this journal

Article views: 108

View related articles

Citing articles: 4 View citing articles

Full Terms & Conditions of access and use can be found at

https://www.tandfonline.com/action/journalInformation?journalCode=tfie20
Fuzzy Inf. Eng. (2010) 4: 361-387
DOI 10.1007/s12543-010-0056-3
ORIGINAL ARTICLE

Vibration Control of Active Vehicle Suspension System Using

Fuzzy Logic Algorithm

Essam Allam · Hesham Fath Elbab · Magdy Abdel Hady · Shawki

Abouel-Seoud

Received: 18 March 2010/ Revised: 30 August 2010/

Accepted: 21 November 2010/
© Springer-Verlag Berlin Heidelberg and Fuzzy Information and Engineering Branch of the Operations
Research Society of China 2010

Abstract Vehicle suspension along with tires and steering linkages is designed for
safe vehicle control and to be free of irritating vibrations. Therefore the suspension
system designs are a compromise between ride softness and handing ability. How-
ever, this work is concerned with a theoretical investigation into the ride behavior of
actively suspended vehicles. It is based on using fuzzy logic control (FLC) to im-
plement a new sort of active suspension system. Comparisons between the behavior
of active suspension system with FLC with those obtained from active systems with
linear-quadratic regulator (LQR), ideal skyhook system and the conventional passive
suspension systems. Results are introduced in such a way to predict the benefits that
could be achieved from a fuzzy logic system over other competing systems. Further-
more, a controller is designed and made by using results of FLC system, theoretical
inputs are used to examine the validity of this controller. Moreover, comparison be-
tween actual outputs from this controller with those obtained theoretically is made to
judge the validity of the controller. The results indicate that the controller has a good
capability in simulation of the theoretical model.

Keywords Body acceleration · Suspension working space · Dynamic tire deflection

· Fuzzy control · Active suspension · Skyhook · Passive suspension

1. Introduction
During the past thirty years a number of linear and non-linear suspension system
models have been proposed to study the effect of suspension design on ride and han-
dling improvements. The models range in complexity from simple quarter vehicle
models to complex multi degree of freedom models. The assumptions made in deriv-
ing the equations of motion of the linear suspension models are: 1) the sprung mass
Essam Allam () · Hesham Fath Elbab · Magdy Abdel Hady · Shawki Abouel-Seoud
Automotive and Tractors Engineering Department, Helwan University, Cairo, Egypt
email: [email protected]
362 Essam Allam · Hesham Fath Elbab · Magdy Abdel Hady · Shawki Abouel-Seoud (2010)

is considered to be a rigid mass as are similarly all unsprung masses, 2) the avail-
able working space is large enough so that the suspension always operates without
contacting the bump stops, 3) the tires assumed to be in constant contact with the
road surface and are modeled as linear springs (sometimes damping elements are in-
cluded), 4) the front and rear suspensions are modeled as linear spring and damper
units with the possible addition of anti-roll bars. The next problem is to combine the
input, described either in the frequency or time domain, with the vehicle model to
obtain the output of interest [1, 2].
The performance of passive suspension systems have been studied by using either
two mass quarter car model or various configurations for a three mass system rep-
resentation consisting of an additional mass attached to the sprung mass and/or the
unsprung mass. It is concluded that it is possible to reach a high degree of ride com-
fort without deterioration in the contact between wheels and road by using a dynamic
absorber of mass equal to the wheel mass, but incorporating an auxiliary mass equal
to the unsprung mass seems to be impractical. Furthermore, the validity of some of
these linear models using actual measured road roughness profiles. The roughness
inputs and the acceleration at the body connection points were measured for different
quality roadways. Using these measurements, they showed that the seven degree of
freedom model predicts more accurately the vehicle acceleration in the low frequency
range (up to 10 Hz). They attributed the discrepancies which appeared over 10 Hz to
excitation from tire unevenness [3-6].
The optimization problem essentially treated the suspension system design as an
optimal control problem either with full state feedback capabilities or limited state
feedback active systems using a gradient search technique or a slow active suspen-
sion. One of the practical difficulties in using the active system is the need to measure
the body and wheel positions relative to the road. This difficulty is due to the fact that
most of the possible methods of measurements are either too expensive or unreliable
on all types of road surfaces. In the most important practical limitation is the need
to measure the road profile height. Therefore, there has been a movement to use the
concept of the limited state feedback active suspension systems. In these systems, the
classical optimal control law based on full state feedback is replaced by one involving
limited state feedback-omitting, in particular, the ground input information and one
which may involve measurement errors were the first to formulate an optimization
problem of the active suspension using a three dimensional vehicle model [7-10].
The performance of quarter car model controlled by a fuzzy inference system
strategies with either two inputs (suspension working space and sprung mass veloc-
ity) and ignored the unsprung mass and tire contact with road or with three inputs
(sprung mass acceleration, sprung mass velocity and suspension working space), the
results were compared with those of the sky hooks model. The investigations made
by setting up a controller based on fuzzy logic control with two inputs (sprung mass
velocity and unsprung mass velocity), while fuzzy rules were made to minimize a de-
fined performance index. Moreover, the theoretical investigation made by setting up
a controller based on neuro-fuzzy logic control with two inputs (sprung mass acceler-
ation error and the rate of error with reference to sky hooks body acceleration). The
results were compared with those of active suspension system with fuzzy logic con-
Fuzzy Inf. Eng. (2010) 4: 361-387 363

troller and also of active suspension system with linear quadratic regulation (LQR)
controller. The neuro-fuzzy active suspension was shown to be the best performing
system without any increase in the system complexity. The results were compared
with those of active suspension system with LQR controller. The analysis show a
great development in performance [11,12].
However, the aim of this work is concerned with a theoretical investigation into
the ride behavior of actively suspended vehicles. It is based on using fuzzy logic
control (FLC) to implement a new sort of active suspension system. Results achieved
from the system with FLC are compared with those obtained from active systems
with linear control theory, ideal skyhook system and the conventional passive sus-
pension system. Results are introduced in such a way to predict the benefits could
be achieved from fuzzy logic system over other competing systems. Furthermore, a
controller is designed and made by using results of FLC system, theoretical input is
used to examine the validity of this controller. Comparison between actual outputs
from this controller with those obtained theoretical is made to judge the validity of
the controller.

2. Nomenclature
A Road input amplitude, m 0.05
BA Body acceleration, m/s2 -
C Coefficient of power spectral density 5
C2 Suspension damping coefficient, Ns/m 1300
C3 Skyhook damping coefficient, Ns/m 20000
DT D Dynamic tire deflection, m -
K1 Tire stiffness, N/m 192000
K2 Suspension stiffness, N/m 20000
M1 Unsprung mass, kg 45
M2 Sprung mass, kg 310
Rc Road roughness coefficient 4 × 10−4
S WS Suspension work space, m -
ts Sampling time, s 0.01
V Vehicle velocity, m/s -
y1 Body acceleration as an output, m/s2 -
y2 Suspension working space as an output, m -
y3 Dynamic tire loading as an output, m -
λ Road wave length, m 10.46
ωc Sprung mass natural frequency, rad/s 7

3. Theoretical Considerations Passive Suspension Systems

This section consists of three parts; the first part introduces simulation of the road
disturbance, while the second part concerns with the modeling of the quarter car pas-
364 Essam Allam · Hesham Fath Elbab · Magdy Abdel Hady · Shawki Abouel-Seoud (2010)

sive suspension system and its analysis. Finally the third part discusses the skyhook
suspension model analysis.
Road Disturbance
Many researches have been made in order to represent the road surface disturbance
is so complex to make a single model to reflect all variation of profile on road with
different velocities. In this work the road input is represented as a sinusoidal sig-
nal. Referring to Ref [3], the road roughness as a sinusoidal signal is represented as
follow:

2πV
x0 = A sin(t), (1)
λ
where A is the amplitude = 0.05m, V is the vehicle velocity = 20m/s, λ is the wave-
length of the road profile = 12.56m.
Passive Suspension System Analysis
A quarter car suspension models are used to represent the passive suspension system.
It is consisted of sprung and unsprung masses connected together by a spring and
damper. The tire is modeled by a linear spring. Using Fig.1, matrices of masses,
forces, stiffness and damping coefficients are as follow:

Fig. 1 Passive suspension model

⎡ ⎤ ⎡ .. ⎤ ⎡ ⎤⎡ ⎤
⎢⎢⎢ M1 0 ⎥⎥⎥ ⎢⎢⎢ Z1 ⎥⎥⎥ ⎢⎢⎢ 1 −1 ⎥⎥⎥ ⎢⎢⎢ f1 ⎥⎥⎥
⎢⎢⎣ ⎥⎥⎦ ⎢⎣⎢ .. ⎥⎥⎦ = ⎢⎢⎣ ⎥⎥ ⎢⎢ ⎥⎥
0 M2 Z2 0 1 ⎦ ⎣ f2 ⎦

or (2)
..
MXDD Z = M f f ,
where MXDD is the matrix of the system masses, M f is the connection matrix, Z is the
vector of the system coordinates, f is the passive suspension and the tire force vector.
The most convenient form to describe the forces x is
Fuzzy Inf. Eng. (2010) 4: 361-387 365

.
f = −Mk d − Mc d, (3)
where Mk and Mc are (n × n) diagonal matrices containing the stiffness and damping
coefficients respectively. In the case of the quarter car suspension model, these two
matrices are as follows:
⎡ ⎤ ⎡ ⎤
⎢⎢ K1 0 ⎥⎥⎥ ⎢ ⎥
Mk = ⎢⎢⎢⎣ ⎥⎥⎦ , Mc = ⎢⎢⎢⎢⎣ 0 0 ⎥⎥⎥⎥⎦ . (4)
0 K2 0 C2
The vector d includes the relative displacement across the connecting elements and
is given by:
⎡ ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤
⎢⎢⎢ Z1 − x0 ⎥⎥⎥ ⎢⎢⎢ 1 0 ⎥⎥⎥ ⎢⎢⎢ Z1 ⎥⎥⎥ ⎢⎢⎢ −1 ⎥⎥⎥
d = ⎢⎣ ⎢ ⎥
⎥=⎢ ⎢ ⎥⎥ ⎢⎢ ⎥⎥ + ⎢⎢ ⎥⎥ x0
Z2 − Z1 ⎦ ⎣ −1 1 ⎦ ⎣ Z2 ⎦ ⎣ 0 ⎦

or (5)
d = M f T Z + MDu x0 .
Substituting by Equation (5) and (3) in Equation (2), the results may be written as:
.. . .
M xDD Z +M xD Z +M x Z = MuD x + Mu x0 , (6)
0

where
M xD = M f Mc M Tf , M x = M f Mk M Tf ,
(7)
MuD = −M f Mc MDu , Mu = −M f Mk MDu .
For the quarter car suspension model used in this work, these matrices are given
as:
⎡ ⎤
⎢⎢⎢ C2 −C2 ⎥⎥⎥
M xD = ⎣ ⎢ ⎥⎦ ,
−C2 C2
⎡ ⎤
⎢⎢⎢⎢ K1 + K2 −K2 ⎥⎥⎥⎥
Mx = ⎣ ⎦, (8)
−K2 K2
⎡ ⎤ ⎡ ⎤
⎢⎢ 0 ⎥⎥ ⎢⎢ K1 ⎥⎥
MuD = ⎢⎢⎣ ⎥⎥⎦ , Mu = ⎢⎢⎣ ⎥⎥⎦ .
0 0
The second order Equation (7) may be transferred into a first order equation by
defining new state variable x. and the new result may be written as:
. .
x = Ax + B2 x0 + B3 x0 , (9)
where A, B2 and B3 are constant matrices which are related to M xDD , M xD , M x , MuD
and Mu by:
⎡ ⎤
⎢⎢ 0 I ⎥⎥⎥
A = ⎢⎢⎣ −1 −1
⎥⎦ ,
−M xDD M x −M xDD M xD
⎡ ⎤ ⎡ ⎤ (10)
⎢⎢ 0 ⎥⎥⎥ ⎢⎢⎢ 0 ⎥⎥⎥
B2 = ⎢⎢⎣ −1
⎥ ,
⎦ 3 ⎣
B = ⎢ −1
⎥⎦ .
M xDD Mu M xDD MuD
366 Essam Allam · Hesham Fath Elbab · Magdy Abdel Hady · Shawki Abouel-Seoud (2010)

For the quarter car suspension model shown in Fig.1, these matrices are given by:
⎡ ⎤
⎢⎢⎢ 0 0 1 0 ⎥⎥⎥
⎢⎢⎢ ⎥⎥
⎢⎢⎢ 0 0 0 1 ⎥⎥⎥⎥
⎢⎢⎢ ⎥⎥
A = ⎢⎢⎢⎢ −(K2 + K1 ) K2 −C2 C2 ⎥⎥⎥⎥⎥ ,
⎢⎢⎢ M1 M1 M! M1 ⎥⎥⎥
⎢⎢⎢ K2 −K2 C2 −C2 ⎥⎥⎥⎥⎦
⎢⎣
M2 M M M (11)
⎡ ⎤ 2 ⎡ 2⎤ 2
⎢⎢⎢ 0 ⎥⎥⎥ ⎢⎢⎢ 0 ⎥⎥⎥
⎢⎢⎢ ⎥
⎢⎢⎢ 0 ⎥⎥⎥⎥⎥ ⎢⎢⎢ ⎥⎥⎥
⎢⎢ 0 ⎥⎥
⎢ ⎥
B2 = ⎢⎢⎢ K1 ⎥⎥⎥ , B3 = ⎢⎢⎢⎢ ⎥⎥⎥⎥ .
⎢ ⎥
⎢⎢⎢ ⎥ ⎢⎢⎢ 0 ⎥⎥⎥
⎢⎢⎣ M1 ⎥⎥⎥⎥⎦ ⎢⎣ ⎥⎦
0 0

The vector of state variables x is related to the vector Z by x1 = Z1 , x2 = Z2 ,

x3 = Z3 , x4 = Z4 .
The output coordinates in the state variable vector x and the road input x0 may be
combined into various outputs variables. The result may be written as:
. .
y = τ xD x +τ x x + τuD x + τu x0 . (12)
0

To indicate this transformation, the following outputs are chosen based on the
quarter car suspension model shown in Fig.1.

1. Body acceleration,
.. .
Z 2 = x4 ; (13)
2. Suspension working space,

S WS = x1 − x2 ; (14)

3. Dynamic tire deflection,

DT D = x1 − x0 . (15)

So

⎡ . ⎤ ⎡ ⎤ .
⎡ . ⎤ ⎡ ⎤ ⎢ x1 ⎥ ⎡ ⎤ ⎢ x1 ⎥ ⎡ ⎤ ⎡ ⎤
⎢⎢⎢ x4 ⎥⎥⎥ ⎢⎢⎢ 0 0 0 1 ⎥⎥⎥ ⎢⎢⎢⎢⎢ . ⎥⎥⎥⎥⎥ ⎢⎢⎢ 0 0 0 0 ⎥⎥⎥ ⎢⎢⎢⎢⎢ ⎥⎥⎥⎥⎥ ⎢⎢⎢ 0 ⎥⎥⎥ ⎢⎢⎢ 0 ⎥⎥⎥
⎢⎢⎢ ⎥⎥⎥ ⎢⎢⎢ ⎥⎥⎥ ⎢⎢⎢ x2 ⎥⎥⎥ ⎢⎢⎢ ⎥⎥⎥ ⎢⎢⎢ x2 ⎥⎥⎥ ⎢⎢⎢ ⎥⎥⎥ . ⎢ ⎥
⎢⎢⎢ x1 − x2 ⎥⎥ = ⎢⎢⎢ 0 0 0 0 ⎥⎥⎥ ⎢⎢⎢ . ⎥⎥ + ⎢⎢⎢ 1 −1 0 0 ⎥⎥⎥ ⎢⎢⎢ ⎥⎥ + ⎢⎢⎢ 0 ⎥⎥⎥ x0 + ⎢⎢⎢⎢⎢ 0 ⎥⎥⎥⎥⎥ x0 .
⎥ ⎥ ⎥ (16)
⎢⎣ ⎥⎦ ⎢⎣ ⎥⎦ ⎢⎢⎢ x3 ⎥⎥⎥ ⎢⎣ ⎥⎦ ⎢⎢⎢ x3 ⎥⎥⎥ ⎢⎣ ⎥⎦ ⎢⎣ ⎥⎦
x1 − x 0 0 0 0 0 ⎢⎣ . ⎥⎦ 1 0 0 0 ⎢⎣ ⎥⎦ 0 −1
x4 x4

In order to carry out the ride analysis in response to road input for the vehicle
models, several possible methods are available to solve the previous set of equations
of motion. One of them is based on using the difference theories. The basic idea is as
follows:
. .
The equation x = Ax + B2 x0 + B3 x0 is written in the form:
Fuzzy Inf. Eng. (2010) 4: 361-387 367

xk+1 = eAT xk + (eAT − I)A−1 B2 x0k + (eAT − I)−1 B3 x0k , (17)

where eAT : Exponentional matrix, T : Sampling time, I: Unit matrix.
Sky-hook Suspension Model
A skyhook suspension model represents a system which is not real and could not be
achieved in practice. Nevertheless, it is used to represent an ideal suspension system.
This system is similar to a conventional passive suspension system (see Fig.2) where
the spring and unsprung masses are connected together by a spring and damper and
the tire is presented by a linear spring. The only difference is that the sprung mass is
hooked to a rigidly via a passive damper element.

Fig. 2 Skyhook suspension model

The equations of motion may be written as:

⎡ ⎤
⎡ ⎤ ⎡ .. ⎤ ⎡ ⎤⎢ f ⎥
⎢⎢⎢ M1 0 ⎥⎥⎥ ⎢⎢⎢ Z1 ⎥⎥⎥ ⎢⎢⎢ 1 −1 0 ⎥⎥⎥ ⎢⎢⎢⎢ 1 ⎥⎥⎥⎥
⎢⎢⎣ ⎥⎥ ⎢⎢ .. ⎥⎥ = ⎢⎢ ⎥⎥ ⎢⎢⎢ f2 ⎥⎥⎥
0 M2 ⎦ ⎣ Z2 ⎦ ⎣ 0 1 −1 ⎦ ⎢⎢⎣ ⎥⎥⎦
fs

or (18)
.. .
MXDD Z = M f f , f = −Mk d − Mc d,
where Mk and Mc are (3×3) matrices containing the stiffness and damping coefficients
respectively. In the case of the quarter car suspension model, these two matrices are
as follows:
⎡ ⎤ ⎡ ⎤
⎢⎢⎢ K1 0 0 ⎥⎥⎥ ⎢⎢⎢ 0 0 0 ⎥⎥⎥
⎢⎢ ⎥⎥ ⎢⎢ ⎥⎥
Mk = ⎢⎢⎢⎢ 0 K2 0 ⎥⎥⎥⎥ , Mc = ⎢⎢⎢⎢ 0 C2 0 ⎥⎥⎥⎥ . (19)
⎣⎢ ⎥⎦ ⎢⎣ ⎥⎦
0 0 0 0 0 Cs
The vector d includes the relative displacement across the connecting elements and
is given by:
368 Essam Allam · Hesham Fath Elbab · Magdy Abdel Hady · Shawki Abouel-Seoud (2010)

⎡ ⎤ ⎡ ⎤ ⎡ ⎤
⎢⎢⎢ Z1 − x0 ⎥⎥⎥ ⎢⎢⎢ 1 0 ⎥⎥⎥ ⎡ ⎤ ⎢⎢⎢ −1 ⎥⎥⎥
⎢⎢⎢ ⎥⎥⎥ ⎢⎢⎢ ⎥⎥⎥ ⎢⎢⎢ Z1 ⎥⎥⎥ ⎢⎢⎢ ⎥⎥⎥
d = ⎢⎢⎢ Z2 − Z1 ⎥⎥⎥ = ⎢⎢⎢ −1 1 ⎥⎥⎥ ⎢⎢⎣ ⎥⎥⎦ + ⎢⎢⎢ 0 ⎥⎥⎥ x0
⎢⎣ ⎥⎦ ⎢⎣ ⎥⎦ Z2 ⎢⎣ ⎥⎦
0 − Z2 0 −1 0

or (20)

d = M f T Z + MDu x0 .

Substituting by Equation (18) and (20) in Equation (17), the results may be written
as
.. . .
M xDD Z +M xD Z +M x Z = MuD x + Mu x0 , (21)
0

⎡ ⎤ ⎡ ⎤
⎢⎢ C2 −C2 ⎥⎥⎥ ⎢ K + K2 −K2 ⎥⎥⎥
M xD = ⎢⎢⎣ ⎥⎦ , M x = ⎢⎢⎢⎣ 1 ⎥⎦ ,
−C2 C2 −K2 K2
⎡ ⎤ ⎡ ⎤ (22)
⎢⎢ 0 ⎥⎥ ⎢⎢ K1 ⎥⎥
MuD = ⎢⎢⎣ ⎥⎥⎦ , Mu = ⎢⎢⎣ ⎥⎥⎦ .
0 0
The second order Equation (21) may be transferred into a first order equation by
defining new state variable x, as explained earlier to be in the form:

. .
x = Ax + B2 x0 + B3 x, (23)
0

where
⎡ ⎤
⎢⎢⎢ 0 0 1 0 ⎥⎥⎥
⎢⎢⎢ ⎥⎥⎥
⎢⎢⎢ 0 0 0 1 ⎥⎥⎥
⎢⎢⎢ ⎥⎥⎥
A = ⎢⎢⎢⎢ −(K 2 + K 1 ) K 2 −C 2 C 2 ⎥⎥⎥ ,
⎢⎢⎢ ⎥⎥⎥
⎢⎢⎢ M1 M1 M! M1 ⎥
⎢⎣ K2 −K2 C2 −(C2 + C s ) ⎥⎥⎥⎥⎦
M2 M2 M2 M2 (24)
⎡ ⎤ ⎡ ⎤
⎢⎢⎢ 0 ⎥⎥⎥ ⎢⎢⎢ 0 ⎥⎥⎥
⎢⎢⎢ ⎥
⎢⎢⎢ 0 ⎥⎥⎥⎥⎥ ⎢⎢⎢ ⎥⎥⎥
⎢⎢ 0 ⎥⎥
⎢ ⎥
B2 = ⎢⎢⎢ K1 ⎥⎥⎥ , B3 = ⎢⎢⎢⎢ ⎥⎥⎥⎥ .
⎢ ⎥
⎢⎢⎢ ⎥ ⎢⎢⎢ 0 ⎥⎥⎥
⎢⎢⎣ M1 ⎥⎥⎥⎥⎦ ⎢⎣ ⎥⎦
0 0

The output coordinates will remain unchanged as in the case of a passive suspen-
sion system, see Equation (16).
Active Suspension System with LQR Controller Analysis
It consists of sprung and unsprung masses connected together by a spring, damper
and actuator. The unsprung mass interacts with the ground and between them there is
a spring. Using Fig.3, matrices of masses and stiffness, damper and forces are used
for analysis without any difficulties to formulate the equations, as follows:
Fuzzy Inf. Eng. (2010) 4: 361-387 369

Fig. 3 Active (LQR) suspension model

⎡ ⎤ ⎡ .. ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤
⎢⎢⎢ M1 0 ⎥⎥⎥ ⎢⎢⎢ Z1 ⎥⎥⎥ ⎢⎢⎢ 1 −1 ⎥⎥⎥ ⎢⎢⎢ f1 ⎥⎥⎥ ⎢⎢⎢ −1 ⎥⎥⎥
⎢⎢⎣ ⎥⎥⎦ ⎢⎢⎣ .. ⎥⎥⎦ = ⎢⎢⎣ ⎥⎥ ⎢⎢ ⎥⎥ + ⎢⎢ ⎥⎥ u
0 M2 Z2 0 1 ⎦ ⎣ f2 ⎦ ⎣ 1 ⎦

or (25)
..
MXDD Z = M f f + M f 2 u,
where M f 2 is the connection matrix.
.
f = −Mk d − Mc d, (26)
where Mk and Mc are (2×2) matrices containing the stiffness and damping coefficients
respectively. These two matrices are as follows:
⎡ ⎤ ⎡ ⎤
⎢⎢⎢ K1 0 ⎥⎥⎥ ⎢⎢⎢ 0 0 ⎥⎥⎥
Mk = ⎢⎣ ⎢ ⎥
⎥ , Mc = ⎢⎣ ⎢ ⎥⎥ . (27)
0 K2 ⎦ 0 C2 ⎦
The vector d includes the relative displacement across the connecting elements and
is given by:
⎡ ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤
⎢⎢⎢ Z1 − x0 ⎥⎥⎥ ⎢⎢⎢ 1 0 ⎥⎥⎥ ⎢⎢⎢ Z1 ⎥⎥⎥ ⎢⎢⎢ −1 ⎥⎥⎥
d = ⎢⎣ ⎢ ⎥
⎥=⎢ ⎢ ⎥⎥ ⎢⎢ ⎥⎥ + ⎢⎢ ⎥⎥ x0
Z2 − Z1 ⎦ ⎣ −1 1 ⎦ ⎣ Z2 ⎦ ⎣ 0 ⎦

or (28)
d = M f Z + MDu x0 .
T

Substituting by Equation (26) and (28) in Equation (25), the results may be written
as
.. . .
M xDD Z +M xD Z +M x Z = MuD x + Mu x0 , (29)
0

where
370 Essam Allam · Hesham Fath Elbab · Magdy Abdel Hady · Shawki Abouel-Seoud (2010)

⎡ ⎤ ⎡ ⎤
⎢⎢⎢ C2 −C2 ⎥⎥⎥ ⎢⎢⎢ K1 + K2 −K2 ⎥⎥⎥
M xD =⎣ ⎢ ⎥
⎦ , Mx = ⎣ ⎢ ⎥⎦ ,
−C2 C2 −K2 K2
⎡ ⎤ ⎡ ⎤ (30)
⎢⎢⎢⎢ 0 ⎥⎥⎥⎥ ⎢⎢ K1 ⎥⎥
MuD = ⎣ ⎦ , Mu = ⎢⎢⎣ ⎥⎥⎦ .
0 0
The second order Equation (29) may be transferred into the first order equation by
defining new state variable x, as explained earlier, to be in the form
.
x = Ax + B2 x0 + Bu u, (31)
where A, B2 and B3 are constant matrices which are related to M xDD , M xD , M x , MuD
and Mu by:
⎡ ⎤
⎢⎢ 0 I ⎥⎥⎥
A = ⎢⎢⎣ −1 −1
⎥⎦ ,
−M xDD M x −M xDD M xD
⎡ ⎤ ⎡ ⎤ (32)
⎢⎢ 0 ⎥⎥⎥ ⎢⎢⎢ 0 ⎥⎥⎥
B2 = ⎢⎢⎣ −1
⎥
⎦ 3 ⎣ , B = ⎢ −1
⎥⎦ ,
M xDD Mu M xDD M f 2
⎡ ⎤
⎢⎢⎢ 0 0 1 0 ⎥⎥⎥
⎢⎢⎢ ⎥⎥
⎢⎢⎢ 0 0 0 1 ⎥⎥⎥⎥
⎢⎢⎢ ⎥⎥
A = ⎢⎢⎢⎢ −(K2 + K1 ) K2 −C2 C2 ⎥⎥⎥⎥⎥ ,
⎢⎢⎢
⎢⎢⎢ M1 M1 M! M1 ⎥⎥⎥⎥
⎢⎣ K −K 2 C 2 −C 2 ⎥
⎥⎥
2 ⎦
M2 M2 ⎡M2 ⎤M2 (33)
⎡ ⎤ ⎢⎢⎢ 0 ⎥⎥⎥
⎢⎢⎢ 0 ⎥⎥⎥ ⎢⎢⎢⎢ ⎥⎥
⎢⎢⎢ ⎥⎥⎥ ⎢⎢⎢ 0 ⎥⎥⎥⎥⎥
⎢⎢⎢ 0 ⎥⎥⎥ ⎢ ⎥
B2 = ⎢⎢⎢⎢⎢ K1 ⎥⎥⎥⎥⎥ , B3 = ⎢⎢⎢⎢⎢ −1 ⎥⎥⎥⎥⎥ .
⎢⎢⎢ ⎥⎥⎥ ⎢⎢⎢ M1 ⎥⎥⎥
⎢⎢⎣ M1 ⎥⎥⎦ ⎢⎢⎢⎢ 1 ⎥⎥⎥⎥
0 ⎣ ⎦
M2
The output coordinates in the state variable vector X and the road input x0 may be
combined into various outputs variables. Therefore the result may be written as:
. .
y = τ xD x +τ x x + τuD x0 + τu x0 . (34)
To indicate this transformation, the following outputs are chosen based on the
active suspension quarter car model shown in Fig.3.
Body acceleration,
.. .
Z 2 = x4 . (35)
Suspension working space,

S WS = x1 − x2 . (36)

Dynamic tire deflection,

DT L = x1 − x0 . (37)
So
Fuzzy Inf. Eng. (2010) 4: 361-387 371

⎡ . ⎤ ⎡ ⎤ .
⎡ . ⎤ ⎡ ⎤ ⎢ x1 ⎥ ⎡ ⎤ ⎢ x1 ⎥ ⎡ ⎤ ⎡ ⎤
⎢⎢⎢ x4 ⎥⎥⎥ ⎢⎢⎢ 0 0 0 1 ⎥⎥⎥ ⎢⎢⎢⎢⎢ . ⎥⎥⎥⎥⎥ ⎢⎢⎢ 0 0 0 0 ⎥⎥⎥ ⎢⎢⎢⎢⎢ ⎥⎥⎥⎥⎥ ⎢⎢⎢ 0 ⎥⎥⎥ ⎢⎢⎢ 0 ⎥⎥⎥
⎢⎢⎢ ⎥⎥⎥ ⎢⎢⎢ ⎥⎥⎥ ⎢⎢⎢ x2 ⎥⎥⎥ ⎢⎢⎢ ⎥⎥⎥ ⎢⎢⎢ x2 ⎥⎥⎥ ⎢⎢⎢ ⎥⎥⎥ . ⎢ ⎥
⎢⎢⎢ x1 − x2 ⎥⎥⎥ = ⎢⎢⎢ 0 0 0 0 ⎥⎥⎥ ⎢⎢⎢ . ⎥⎥⎥ + ⎢⎢⎢ 1 −1 0 0 ⎥⎥⎥ ⎢⎢⎢ ⎥⎥⎥ + ⎢⎢⎢ 0 ⎥⎥⎥ x0 + ⎢⎢⎢⎢⎢ 0 ⎥⎥⎥⎥⎥ x0 . (38)
⎢⎣ ⎥⎦ ⎢⎣ ⎥⎦ ⎢⎢⎢ x3 ⎥⎥⎥ ⎢⎣ ⎥⎦ ⎢⎢⎢ x3 ⎥⎥⎥ ⎢⎣ ⎥⎦ ⎢⎣ ⎥⎦
x1 − x0 0 0 0 0 ⎢⎣ . ⎥⎦ 1 0 0 0 ⎢⎣ ⎥⎦ 0 −1
x4 x4

Active Suspension System with Fuzzy Controller

Fuzzy Logic Control (FLC) is a problem-solving control system methodology that
lends itself to implementation in systems ranging from simple, small, embedded
micro-controllers to large, networked, multi-channel PC or workstation-based data
acquisition and control systems. It can be implemented in hardware, software, or a
combination of both. FLC provides a simple way to arrive at a definite conclusion
based upon vague, ambiguous, imprecise, noisy, or missing input information. FLC
approach to control problems mimics how a person would make decisions, only much
faster.
Active Suspension with Fuzzy Controller Analysis
It consists of sprung and unsprung masses connected together by a spring, damper
and actuator. The unsprung mass interacts with the ground and between them there
is a spring, as seen in the Fig.4, matrices of masses and stiffness, damper and forces
are used for analysis without any difficulties to formulate the equations as follows:

Fig. 4 Active (FLC) suspension model

⎡ ⎤ ⎡ .. ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤
⎢⎢⎢ M1 0 ⎥⎥⎥ ⎢⎢⎢ Z1 ⎥⎥⎥ ⎢⎢⎢ 1 −1 ⎥⎥⎥ ⎢⎢⎢ f1 ⎥⎥⎥ ⎢⎢⎢ −1 ⎥⎥⎥
⎢⎢⎣ ⎥⎥⎦ ⎢⎣⎢ .. ⎥⎥⎦ = ⎢⎢⎣ ⎥⎥ ⎢⎢ ⎥⎥ + ⎢⎢ ⎥⎥ u
0 M2 Z2 0 1 ⎦ ⎣ f2 ⎦ ⎣ 1 ⎦

or (39)
..
MXDD Z = M f f + M f 2 u,
where M f 2 is the connection matrix number two,
.
f = −Mk d − Mc d, (40)
372 Essam Allam · Hesham Fath Elbab · Magdy Abdel Hady · Shawki Abouel-Seoud (2010)

where Mk and Mc are (2×2) matrices containing the stiffness and damping coefficients
respectively. In the case of the quarter car suspension model, these two matrices are
as follows:

⎡ ⎤ ⎡ ⎤
⎢⎢⎢ K1 0 ⎥⎥⎥ ⎢⎢⎢ 0 0 ⎥⎥⎥
Mk = ⎢⎢⎣ ⎥⎥ , Mc = ⎢⎢⎣ ⎥⎥ . (41)
0 K2 ⎦ 0 C2 ⎦

The vector d includes the relative displacement across the connecting elements and
is given by:
⎡ ⎤ ⎡ ⎤⎡ ⎤ ⎡ ⎤
⎢⎢⎢ Z1 − x0 ⎥⎥⎥ ⎢⎢⎢ 1 0 ⎥⎥⎥ ⎢⎢⎢ Z1 ⎥⎥⎥ ⎢⎢⎢ −1 ⎥⎥⎥
d = ⎢⎢⎣ ⎥⎥ = ⎢⎢ ⎥⎥ ⎢⎢ ⎥⎥ + ⎢⎢ ⎥⎥ x0
Z2 − Z1 ⎦ ⎣ −1 1 ⎦ ⎣ Z2 ⎦ ⎣ 0 ⎦

or (42)

d = M f T Z + MDu x0 .

Substituting by Equation (42) and (40) in Equation (39), the results may be written
as

.. . .
M xDD Z +M xD Z +M x Z = MuD x + Mu x0 , (43)
0

where

⎡ ⎤ ⎡ ⎤
⎢⎢ C2 −C2 ⎥⎥⎥ ⎢ K + K2 −K2 ⎥⎥⎥
M xD = ⎢⎢⎣ ⎥⎦ , M x = ⎢⎢⎢⎣ 1 ⎥⎦ ,
−C2 C2 −K2 K2
⎡ ⎤ ⎡ ⎤ (44)
⎢⎢ 0 ⎥⎥ ⎢⎢ K1 ⎥⎥
MuD = ⎢⎢⎣ ⎥⎥⎦ , Mu = ⎢⎢⎣ ⎥⎥⎦ .
0 0

The second order Equation (43) may be transferred into the first order equation by
defining new state variable x, as explained earlier to be in the form

.
x = Ax + B2 x0 + Bu u, (45)

where A, B2 and B3 are constant matrices which are related to M xDD , M xD , M x , MuD
and Mu by:

⎡ ⎤
⎢⎢ 0 I ⎥⎥⎥
A = ⎢⎢⎣ −1 −1
⎥⎦ ,
−M xDD M x −M xDD M xD
⎡ ⎤ ⎡ ⎤ (46)
⎢⎢ 0 ⎥⎥⎥ ⎢⎢⎢ 0 ⎥⎥⎥
B2 = ⎢⎢⎣ −1
⎥ ,
⎦ 3 ⎣
B = ⎢ −1
⎥⎦ .
M xDD Mu M xDD M f 2

For the quarter car suspension model show in Fig.4, these matrices are give by:
Fuzzy Inf. Eng. (2010) 4: 361-387 373

⎡ ⎤
⎢⎢⎢ 0 0 1 0 ⎥⎥⎥
⎢⎢⎢ ⎥⎥
⎢⎢⎢ 0 0 0 1 ⎥⎥⎥⎥
⎢⎢⎢ ⎥
A = ⎢⎢⎢⎢ −(K2 + K1 ) K2 −C2 C2 ⎥⎥⎥⎥ ,
⎢⎢⎢ ⎥⎥
⎢⎢⎢ M M1 M! M1 ⎥⎥⎥
C2 −C2 ⎥⎥⎥⎥⎦
1
⎢⎣ K2 −K2
M2 M2 ⎡M2 ⎤M2 (47)
⎡ ⎤ ⎢⎢⎢ 0 ⎥⎥⎥
⎢⎢⎢ 0 ⎥⎥⎥ ⎢⎢⎢ ⎥
⎢⎢⎢
⎢⎢⎢ 0 ⎥⎥⎥⎥⎥
⎥ ⎢⎢⎢ 0 ⎥⎥⎥⎥⎥
⎢⎢ ⎥⎥
B2 = ⎢⎢⎢⎢⎢ K1 ⎥⎥⎥⎥⎥ , B3 = ⎢⎢⎢⎢⎢ −1 ⎥⎥⎥⎥⎥ .
⎢⎢⎢ ⎥ ⎢⎢⎢ M1 ⎥⎥⎥
⎢⎢⎣ M1 ⎥⎥⎥⎥⎦ ⎢⎢⎢ 1 ⎥⎥⎥
0 ⎢⎣ ⎥⎦
M2
The output coordinates will remain unchanged as in the case of an active suspen-
sion system with LQR, seen in Equation (38).

4. Fuzzy Logic Controller

The fuzzy rule sets usually have several antecedents that are combined by using
fuzzy operators, such as AND, OR and NOT, though again the definitions tend to
vary: AND, in one popular definition, simply uses the minimum weight of all the
antecedents, while OR uses the maximum value. There is also a NOT operator that
subtracts a membership function from 1 to give the “complementary” function.
There are several different ways to define the result of a rule, but one of the most
common and simplest is the “max-min” inference method, in which the output mem-
bership function is given the truth value generated by the premise. Rules can be
solved in parallel in hardware, or sequentially in software. The results of all the rules
that have fired are “defuzzified” to a crisp value by one of several methods. There are
dozens in theory, each with various advantages and drawbacks.
The “centroid” method is very popular, in which the “center of mass” of the re-
sult provides a crisp value. Another approach is a “height” method, which takes the
value of the biggest contributor. The centroid method favors the rule with the output
of greatest area, while the height method obviously favors the rule with the greatest
output value. The example as in Fig.5 demonstrates max-min inferencing and cen-
troid defuzzification for a system with input variables “x”, “y”, and “z” and an output
variable “n”. Notice how each rule provides a result as a truth value of a particular
membership function for the output variable. In centroid defuzzification the values
are OR’d, that is, the maximum value is used and values are not added, and the results
are then combined by using a centroid calculation. Fuzzy control system design is
based on empirical methods, basically a methodical approach to trial-and-error. The
general process is as follows:

• Document the system’s operational specifications and inputs and outputs.

• Document the fuzzy sets for the inputs.

• Document the rule set.

374 Essam Allam · Hesham Fath Elbab · Magdy Abdel Hady · Shawki Abouel-Seoud (2010)

Fig. 5 Centroid defuzzification using MAX-MIN inferencing

• Determine the defuzzification method.

• Run through test suite to validate system, adjust details as required.

• Complete document and release to production.

The design of a fuzzy controller for the active suspension system, where inputs are
a body acceleration error and rate of its change when referred to that of the theoretical
model “skyhook model”. The block diagram of this control system appears as Fig.6.
There are two input variables of body acceleration error and rate of its change, and
a single output variable, the actuating force setting. The suspension bodies oscillate
from upwards to downwards, so the actuating force can be positive or negative. The
fuzzy set mappings are shown in Fig.7. The linguistic terms are defined as follows:
NB: Negative Big.
NM: Negative Medium.
NS: Negative Small.
ZE: Zero.
PS: Positive Small.
PM: Positive Medium.
PB: Positive Big.
Fuzzy Inf. Eng. (2010) 4: 361-387 375

Fig. 6 Fuzzy controller schematic

Fig. 7 The fuzzy logic active suspension

The rule set includes such rules as Table 1.

Rule 13: If Error is NM AND Delta Error is PM, then Actuating Force is Z.
Rule 12: If Error is NM AND Delta Error is PS, then Actuating Force is NS.
376 Essam Allam · Hesham Fath Elbab · Magdy Abdel Hady · Shawki Abouel-Seoud (2010)

Table 1: Fuzzy logic controller rules set.

ΔBAerror

NB NM NS Z PS PM PB
NB NB NB NM NM NS NS Z
NM NB NM NM NS NS Z PS
BAerror NS NM NM NS NS Z PS PS
Z NM NS NS Z PS PS PM
PS NS NS Z PS PS PM PM
PM NS Z PS PS PM PM PB
PB Z PS PS PM PM PB PB

Rule 20: If Error is NS AND Delta Error is PM, then Actuating Force is PS.
Rule 19: If Error is NS AND Delta Error is PS, then Actuating Force is Z.
In practice, the controller accepts the inputs and maps them into their membership
functions and truth values. These mappings are then fed into the rules. If the rule
specifies an AND relationship between the mappings of the two input variables, as
the examples above do, the minimum of the two is used as the combined truth value;
if an OR is specified, the maximum is used. The appropriate output state is selected
and assigned by a membership value at the truth level of the premise. The truth values
are then defuzzified.

5. Controller with Error Input

As mentioned early, the FLC depends on feedback of known measured input to gen-
erate a control output. The input is error in body acceleration. This means simply
that the system requires a direct measurement of the body acceleration via a suitable
accelerometer. The road input is feed to the central processing unit (CPU) of the
controller to evaluate the ideal acceleration. This acceleration is compared with the
measured acceleration signal to evaluate the error. The CPU also evaluates the rate of
error (Δ error). These two inputs are fed to the controller and the output is the control
force u. The graphical user interface (GUI) is used to generate the membership func-
tions for input and output as shown in Fig.8. The input is the body acceleration error
with 7 membership functions as follows:

NB: Negative Big error.

NM: Negative Medium error.
NS: Negative Small error.
Z: Zero error.
PS: Positive Small error.
PM: Positive Medium error.
PB: Positive Big error.
Fuzzy Inf. Eng. (2010) 4: 361-387 377

Fig. 8 Indicates a schematic presentation of the FLC

And the universe of discourse is from −1 to 1 and the error signal is modified
before entering the controller through gain and this gain is used to tune the fuzzy
inference system. The second step is to start up the output and it is a single output
(actuated force) and has 7 membership functions:
NB: Negative Big force.
NM: Negative Medium force.
NS: Negative Small force.
Z: Zero force.
PS: Positive Small force.
PM: Positive Medium force.
PB: Positive Big force.
The universe of discourse is from −1 to 1 and an actuating force signal is modified
before entered the plant through a gain and this gain is used to tune the fuzzy inference
system. The third step is to generate the lookup table however, this is the most difficult
part of the problem due to the need of experience and the great nonlinearity of the
system, and for example it is not easy to define how much force is needed for a big
negative error and a small positive change in error. In order to overcome this problem
the performance is evaluated for each crisp output and used to tune the controller.

Table 2: Summarizes the optimum set of FLC rules.

Body acc. error NB NM NS Z PS PM PB

Actuated force B NM NS Z PS PM PB

For instant, if the error of the body acceleration is NS (Negative Small), the con-
troller output will be NS (Negative Small). On the other hand, if the error is PM
(Positive Medium), the control output will be PM (Positive Medium) in Table 2. To
evaluate the performance parameters of FLC active suspension system, a simulink
model is introduced to combine the control with the system. Fig. 9 shows this combi-
nation. The road input is fed to the skyhook model to generate the ideal acceleration.
378 Essam Allam · Hesham Fath Elbab · Magdy Abdel Hady · Shawki Abouel-Seoud (2010)

This acceleration is then used to determine the error. At time t=0, the error is set
equal to zero. As mentioned above in the previous section, the error is used to evalu-
ate control signal u; this control signal is used to evaluate the system performance of
an FLC active suspension system.

Fig. 9 The fuzzy logic active suspension model with two inputs represented by
simulink

The first controller is based on using two inputs, error and change of error of body
acceleration. Seven membership functions are used to describe each of these inputs.
A centroid method is employed to generate crisp value of the output control signal.
In the second controller type is simplified to use only error of body acceleration is the
single input.
GUI is used to build up computer programs of the generated systems. The results
are generated for the two designed fuzzy logic controllers and include time histories
of body acceleration, suspension working space and dynamic tire deflection.

6. Controller Design

For theoretical study, the fuzzy logic control has the capability of improving further
ride comfort and road holding parameter without any increase in the used working
space. However, there is a need to examine the validity of this controller this experi-
mentally. In principle at least, one needs to build up a test rig including a suspension
system with a fuzzy logic controller, suitable excitation presentation, various needed
sensors and a complete measuring system. This is however, far behind the scope of
this work. Here the ides is simplified to the following steps:

1. Introducing a practical and simple design of controller.

2. Testing the controller through a defined and known input.

3. Evaluating the controller response to this input.

4. Comparing between achieved results with those calculated theoretically using

the same excitation.

5. Evaluating the controller effectiveness.

Fuzzy Inf. Eng. (2010) 4: 361-387 379

In this stage, the designed program is loaded to the interface and then to the micro-
controller using a programmer interface shown in Fig.10 and Fig.11. The design
of the controller is only introduced based on one input i.e., the error in the body
acceleration. This controller uses only the error in body acceleration as input and
generates it in response the required control the signal. A complete set of computer
programs are used using Visual Basic language introduced to drive this controller. A
simulink model is made to evaluate the relationship between the body acceleration
and control force needed.
The controller is tested for various acceleration inputs. Besides, experimental out-
puts are compared with those obtained theoretically using same inputs. Therefore,
good agreements between experimental and theoretical outputs were found.

Fig. 10 Indicates a schematic presentation of the FLC

Fig. 11 The programmer connected to PC by serial

380 Essam Allam · Hesham Fath Elbab · Magdy Abdel Hady · Shawki Abouel-Seoud (2010)

7. Results and Discussion

Fig.12 to Fig.15 show the time histories for the performance parameters of the pas-
sive, skyhook, active with LQR and active with FLC (one and two inputs) suspen-
sion systems respectively. To highlight the benefits that could be achieved from us-
ing fuzzy logic control active suspension system over other competing systems, the
passive suspension is used here as a baseline system, while the skyhook suspension
system is used to represent the ideal system. Under these bases, a direct compari-
son between LQR active suspension system with FLC active suspension system is
obtained.

Fig. 12 Body acceleration of all competing system

Fig. 13 Dynamic tire deflection of all competing system

The root mean square (RMS) of the time histories for the performance parameters
of body acceleration (BA), suspension working space (SWS) and dynamic tire de-
flection (DTD) are collected and presented in Fig.16 to Fig.18 respectively. In terms
of RMS value, the following remarks can be done as follows:
Fuzzy Inf. Eng. (2010) 4: 361-387 381

Fig. 14 Suspension working space of all competing system

One input
382 Essam Allam · Hesham Fath Elbab · Magdy Abdel Hady · Shawki Abouel-Seoud (2010)

Two input

Fig. 15 Performance of the (FCL) active suspension system

1. The FLC active suspension system with signal input approaches closely the
system with two inputs. This is an important finding since it shows that the
fuzzy logic control can still perform well even for the case of using limited
feedback information.

2. The FLC active suspension systems approach the performance level of ideal
skyhook model, this finding is also important, because a part of fuzzy controller
takes input from this ideal system, and tune the active suspension to reduce the
error and hence to perform very closely to this ideal system.

3. Fuzzy logic control shows that significant improvements can be achieved in

comparison with LQR active suspension system. In terms of body accelera-
tion, the improvement appears as a reduction in RMS value by 41.4%; at the
same usage of suspension working space. Also, FLC reduces the dynamic tire
Fuzzy Inf. Eng. (2010) 4: 361-387 383

Fig. 16 Body acceleration of all competing system

Fig. 17 Suspension working space of all competing system

deflection RMS value by 55.5% when comparing with LQR active suspension
system.

4. In comparison with passive suspension system, LQR active suspension system

reduces body acceleration, suspension working space and dynamic tire deflec-
tion by 27.5%, 28.2% and 23.8% respectively. These improvements become
57.5%, 28.2% and 66.2% in case of using FLC system.

In terms of system compiling, FLC appears to be much more easily to realize in

practice. It depends on using an accelerometer to measure the body acceleration. A
micro-controller with a suitable processor would be enough to evaluate the control
384 Essam Allam · Hesham Fath Elbab · Magdy Abdel Hady · Shawki Abouel-Seoud (2010)

Fig. 18 Dynamic tire deflection all competing system of all competing system

signal. On the other hand, the LQR active suspension system needs to have feedback
of body acceleration and a relative displacement transducer to measure body to wheel
displacement as well as a sensor or a state estimator to measure or estimate the road
input. Therefore, it could be stated that the suspension performance level of FLC
active suspension system is accomplished with less difficulty to be applied in practice
when compared with LQR active suspension system.
A simple and practical design of fuzzy logic controller is introduced. This con-
troller uses only the error in body acceleration as input and generates in response the
required control the signal. The results generated are designed to linearize the rela-
tionship between controller input/output responses as shown in Fig.19 and Fig.20.

Fig. 19 BA and actuating force for active suspension system with FCL controller

This approximation is used to evaluate the system performance and compared with
the obtained based on an actual nonlinear form. Results are collected and presented
in Fig.21.
Fuzzy Inf. Eng. (2010) 4: 361-387 385

Fig. 20 BA and actuating force for active suspension designed controller system with
FCL controller

Fig. 21 Performance parameters with FCL and the designed controller

This figure compares directly between linearized system and actual FLC system.
386 Essam Allam · Hesham Fath Elbab · Magdy Abdel Hady · Shawki Abouel-Seoud (2010)

In terms of body acceleration, suspension working space and dynamic tire deflection,
no much difference can be seen between linear and nonlinear systems. Under this
finding, the linear model is used to design the controller.

8. Conclusion

1. Fuzzy logic control shows that significant improvements can be achieved in

comparison with LQR active suspension system at the same usage of suspen-
sion working space: In terms of body acceleration, the reduction in RMS value
by 41.4%; in terms the dynamic tire deflection rms value by 55.5%.

2. The FLC active suspension system with signal input approaches closely the
system with two inputs. Moreover, the FLC active suspension systems ap-
proach the performance level of an ideal skyhook model.

3. An actual fuzzy logic controller is made and used in evaluating experimental

results. The controller uses only error in body acceleration as an input and
generates in response to the required control crisp values.

4. The controller is tested for various acceleration inputs. Experimental outputs

are compared with those obtained theoretically using same inputs. Good agree-
ments between experimental and theoretical outputs were found.

Acknowledgements
I would like to thank the staff and research students within the Advanced Materials
Research Institute, Matria Engineering College, College-Cairo, Egypt.

References
1. Horton D (1986) An introduction to ride analysis in vehicle dynamics. Research Report DAC12,
Dept. of Mechanical Engineering, Leeds University
2. International Standard ISO 2631, (1974), (1978), (1978/Al), (1982) Guide for the evaluation of hu-
man exposure to whole-body vibrations
3. Ryba D (1974 a) Improvements in dynamics characteristics of automobile suspension systems (Part
I, Two Mass System). Vehicle System Dynamics 3: 17-46
4. Ryba D (1974 b) Improvements in dynamics characteristics of automobile suspension systems (Part
II, Three Mass System). Vehicle System Dynamics 3: 55-98
5. Healey A J, Nathman E, Smith C C (1977) An analytical and experimental study of automobile
dynamics with random roadway input. Transaction of the ASME, Journal of Dynamic Systems,
Measurement, and Control, December: 284-292
6. Sharp R S, Hassan S A (1984) The fundamentals of passive automotive suspension system design.
Society of Environmental Engineers Conference on Dynamics in Automotive Engineering, Cranfield
Inst. Tech. : 104-115
7. Abdel Hady M B (1989) The effect of active suspension control on vehicle ride behaviour. Ph.D.
thesis
8. Abdel Hady M B, Crolla D A (1989) Theoretical analysis of active suspension performance using a
four-wheel vehicle model. Proc. Instn. Mech. Engrs 203: 125-135
9. Eldemerdash S M (1995) A practical hydro-pneumatic slow active suspension. Engineering Research
Journal 4: 160-172
Fuzzy Inf. Eng. (2010) 4: 361-387 387

10. Abouel-seoud S A, Abdel-tawaab A (1998) A theoretical study of the hydro-pneumatic semi-active

suspension system with preview. Heavy Vehicle Systems, Int. J. of Vehicle Design, Vol. 5, No. 2
11. Abdel Hady M B (2003) A fuzzy controller for automotive active suspension systems. Society of
Automotive Engineers, Inc. 2003-01-1417.
12. Abdel Hady M B (2003) A neural fuzzy approach for controlling active vehicle suspension systems.
International Mechanical Engineering Congress IMECE2003-42875