Research On Suspension System Based On Genetic Algorithm and Neural Network Control
Research On Suspension System Based On Genetic Algorithm and Neural Network Control
Research On Suspension System Based On Genetic Algorithm and Neural Network Control
Open Access
INTRODUCTION
Suspension is the term given to the system of springs,
shock absorbers and linkages that connects a vehicle to its
wheels. Suspension systems can not only contribute to the
car's handling and braking for good active safety and driving
pleasure, but also keep vehicle occupants comfortable and
reasonably well isolated from road noise, bumps, and
vibrations. The suspension also protects the vehicle itself and
any cargo or luggage from damage and wear. The ride
quality of a vehicle is significantly influenced by its suspension system, the road surface roughness, and the speed of
vehicle. A vehicle designer can do little to improve road
surface roughness, so designing a good suspension system
with good vibration performance under different road conditions become s a prevailing philosophy in the automobile
industry. Passive suspension systems use conventional dampers to absorb vibration energy, the dampers and stiffness
coefficients are constant. The active suspension system use
extra power to provide a response-dependent damper, which
is capable of producing an improved ride comfort. Over the
years, both passive and active suspension systems have been
proposed to optimize a vehicles ride quality. O. GD used
genetic algorithm to obtain the optimal set and suspension
design (O. GD. IJIE 2007) [1], Sun L, Cai XM, Yang J. got
the minimum dynamic pavement load through the genetic
algorithm (Sun L, Cai XM, Yang J. JSV 2007) [2]. A
simplified algorithm for the evaluation of a small car
suspension model can be found in the papers of T.G.
Chondros (T.G. Chondros, S. Michalitsis, S. Panteliou and
A.D. Dimarogonas 1994) [3] and (T.G. Chondros, P.A.
Belokas, K. Vamvakeros and A.D. Dimarogonas 1997) [4],
Furthermore, a more detailed model for heavy vehicles
suspension systems are given in the paper (Chondros T. G.,
Michalos G, Michaelides P, Fainekos E 2007) [5]. The stateof-the-art review on neural networks in automotive applications can be found in the papers (J.T. Papadimitropoulos,
T.G. Chondros, S.D. Panteliou, B. Carlsson, S. Kalogirou
and A.D. Dimarogonas 1999) [6] and (S. Kalogirou, T.G.
Chondros, A.D. Dimarogonas 2000) [7].
A variety of research projects and publications deal with
different types of active suspension systems have been
discussed (Yeh, E.C. and Tsao, Y.J. 1994) [8]. Different
vehicle dynamic models have been adopted according to
different study purposes during research. A two degree-of freedom quarter body of vehicle suspension system model
had been widely applied in vehicle suspension control
research, it can indicate the vehicle body vertical movement,
but not include the pitching movement of the vehicle body.
Although a seven degree-of -freedom whole-body of vehicle
suspension system model can describe not only the vertical
movement of the four wheels and the body center of gravity,
but also the pitching and lateral movement, the model is too
complicated to be widely applied. . In our research, a fivedegree-of -freedom half body of vehicle suspension system
model is established, which can describe both the vertical
movement and the pitching movement of the body, whats
more, it can demonstrate the effect of the passenger, which
makes it to be a relatively ideal model for suspension
dynamic description.
From the flow chart before using ADAMS, we can see
that the mechanical engineer and the control engineer use
different software, build repeated model for one concept
design, moreover, different design validation and tests are
made. Once there is something wrong, no matter what kind
of fault appears, both the mechanical system and the control
system will have to be designed again. Fig. (1) shows the
flow chart before using ADAMS, and Fig. (2) shows the
flow chart after using ADAMS.
From Fig. (2), the flow chart after using ADAMS, we can
see that the mechanical engineer and the control engineer
share the same virtual model, thus the mechanical design can
2009 Bentham Open
73
correspond with the control design. Furthermore, more complex model, such as the nonlinear model can be obtained,
and design time is decreased and the design reliability is
improved.
(1)
(2)
ms Z1 + f ks + f cs = 0
(3)
..
m Z 2 + f kf + f kr + f cf + f cr f ks f cs = 0
(4)
..
J + b( f kr + f cr ) + e( f ks + f cs ) a( f kf + f cf ) = 0
(5)
..
m f Z 4 + f ktf f kf f cf = 0
(6)
..
mr Z6 + f ktr f kr f cr = 0
(7)
are the seat spring force and seat damping force, respectively;
f kf = K f (Z3 Z 4 ) and f kr = K r (Z5 Z6 ) are the front suspension
spring force and rear suspension spring force, , respec.
tively; fcf = C f (Z3 Z4 ) and f cr = Cr (Z5 Z6 ) are the front suspension damping force and rear suspension damping force,
respectively; where the coefficients C f and C r come from
.. 2
.. 2
J = 1 z2 + 2 + 3 z1 + 4
= [ Z1 Z2 Z3 Z4 Z5 Z6 Z1 Z2 Z3 Z4 Z5 Z6 Z01 Z02 ]
+ 5
y f = z3 a z4
yr = z2 + bq - z6 ; z1 is
the
y 2f + yr2
(8)
And
75
The state
1
.
J
E p (t)
wij (t)
+ wij (t)
(11)
min f w t +1 f w t
( ) ( ( ))
Fd =
s.t. 0
(12)
SIMULATION RESULTS
Crossover
This is the exchange of chromosomes characteristics
among randomly selected pairs from the parent root. The
crossover rate has proper value, the excessive large value
will damage the excellent mode, and the excessive small
value will decrease the velocity of new chromosome. In the
present work, 0.5 is the considered probability rate of the
crossover.
Mutation
y=
(2)
(1)
2
1 .
1+ ex
(9)
1
d y (t ) 2
kp
2 k kp
(10)
77
Table 1. Performance Analysis
Body Acceleration
(m/ s2)
Pitching Angular
Acceleration (rad/s2)
Before control
0. 4620
0.1745
After control
0. 2524
0.1226
Performance
melioration
45. 4%
29. 7%
Passenger
Acceleration (m/ s2)
Before control
0. 2109
0.0089
After control
0.1150
0.0058
Performance
melioration
45. 5%
34. 8%
Before control
After control
82. 8
102. 5
Performance
melioration
61. 4
89. 9
Before control
25. 8%
12. 3%
ASSOCIATED SIMULATION
First, using the software ADAMS, the mechanical model
of five DOF suspension system is established (Chen Y 2007)
[14], as shown in Fig. (11). The ADAMS model is imported
into the Matlab/Simulink model, to accomplish the associated simulation with Matlab/Simulink (Chen LP 2007)
[15]. Figs. (12 and 13) indicate the vertical acceleration and
pitching angular acceleration response of the sprung mass of
the constituted mechanical model. The solid line is the curve
of matlab simulation, the dash line is the curve of combination of matlab and Adams. From the results, we can see
that although the variation trend of the vertical response and
the pitching acceleration response of the established mathematical model and the mechanical are in consistence, there is
difference between the solid line and the dash line. The
CONCLUSIONS
[3]
[2]
[5]
[6]
[7]
[8]
[10]
[11]
[12]
[13]
O. GD, Optimal seat and suspension design for a quarter car with
driver model using genetic algorithms, International Journal of
Industrial Ergonomics, vol. 37, pp. 327-332, 2007.
L. Sun, X. M. Cai, and J. Yang, Genetic algorithm-based optimum
vehicle suspension design using minimum dynamic pavement load
as a design criterion, Journal of Sound and Vibration, vol. 301,
pp.18-27, 2007.
[4]
[9]
ACKNOWLDGEMENTS
[1]
[14]
[15]
79
Note: This service is not intended for secure transactions such as banking, social media, email, or purchasing. Use at your own risk. We assume no liability whatsoever for broken pages.
Alternative Proxies: