14-07-2014, 04:24 PM
Adaptive Neuro Fuzzy Inference Controller for Full Vehicle Nonlinear Active Suspension Systems
Adaptive Neuro Fuzzy Inference Controller.pdf (Size: 776.74 KB / Downloads: 42)
Abstract
The main objective of designed the controller
for a vehicle suspension system is to reduce the discomfort
sensed by passengers which arises from road roughness
and to increase the ride handling associated with the
pitching and rolling movements. This necessitates a very
fast and accurate controller to meet as much control
objectives, as possible. Therefore, this paper deals with an
artificial intelligence Neuro-Fuzzy (NF) technique to
design a robust controller to meet the control objectives.
The advantage of this controller is that it can handle the
nonlinearities faster than other conventional controllers.
The approach of the proposed controller is to minimize
the vibrations on each corner of vehicle by supplying
control forces to suspension system when travelling on
rough road. The other purpose for using the NF controller
for vehicle model is to reduce the body inclinations that
are made during intensive manoeuvres including braking
and cornering. A full vehicle nonlinear active suspension
system is introduced and tested. The robustness of the
proposed controller is being assessed by comparing with
an optimal Fractional Order PIλ
Dμ
(FOPID) controller.
The results show that the intelligent NF controller has
improved the dynamic response measured by decreasing
the cost function
INTRODUCTION
A number of researchers have suggested control methods
for vehicle suspension systems. Some have designed a linear
controller for a quarter or half vehicle [1-9]. In reference [10]
the authors used a robust controller for a full vehicle linear
active suspension system using the mixed parameter
synthesis. A sliding mode technique is designed for a linear
full vehicle active suspension system [11]. A method is
developed for the purpose of sensor fault diagnosis and
accommodation. In reference [12] the authors presented the
development of an integrated control system of active front
steering and normal force control using fuzzy reasoning to
enhance the full vehicle model handling performance. A
fuzzy logic based fast gain scheduling controller is proposed
for control nonlinear suspension systems for quarter car
system [13]. In fact, nonlinearity inherently exists in damper
and spring models [14-16]. Therefore, the nonlinear effect
should be inevitably taken into account to design the
controller for practical active suspension system.
This paper will be developed a novel NF controller for full
vehicle nonlinear active suspension systems. The full vehicle
model will be investigated to take into account the three
motions of the vehicle: vertical movement at centre of
gravity, pitching movement and rolling movement. It is
believed that, this is the first time to use the neuro-fuzzy
method to design the controller for a full vehicle nonlinear
active suspension system.
A neurofuzzy model combines the features of a neural
network and fuzzy logic model. A large class of neuro-fuzzy
approaches utilizes the neural network learning algorithms to
determine parameters of the fuzzy logic system [17]. The
neuro-fuzzy system is more efficient and more powerful than
either neural network or fuzzy logic system [18] which has
been widely used in control systems, pattern recognition,
medicine, expert system, etc. [19].
In this paper optimal FOPID controller will be designed for
full vehicle nonlinear active suspension by using the
Evolutionary Algorithm (EA). The data obtained from the
optimal FOPID controller will be used as reference to design
the NF controller. The learning ability of the neural network
has been used to tune parameters of the membership function
of the Fuzzy Inference Systems (FIS). The performance of the
NF controller has been improved by adding the scaling gains.
The results of the proposed controller will be compared with
THE STRUCTURE AND TRAINING OF ADAPTIVE NEURO FUZZY INFERENCE SYSTEM (ANFIS)
The ANFIS is one of the methods to organize the fuzzy
inference system with given input/output data pairs. The
ANFIS is a combination of a fuzzy logic controller and a
neural network, which makes the controller self tuning and
adaptive. If we compose these two intelligent approaches, it
will be achieve good reasoning in quality and quantity. This
technique gives the fuzzy logic capability to adapt the
membership function parameters that best allow the
associated fuzzy inference system to track the given
input/output data. The data obtain from FOPID controller will
be use to modify the parameters of ANFIS model. In order to
process a fuzzy rule by neural networks, it is necessary to
modify the standard neural network structure accordingly.
Fig. 3 depicts the structure of Neuro-fuzzy inference system
(the type of this model is called Takagi-Sugeno-Kang) [20].
For the simplicity, the following assumptions will be
assumed: (a) the model has two inputs x and y and one
output z, (b) it has just two rules (R1and R2)
MATHEMATICAL MODEL OF THE CONTROLLED SYSTEM
A framework is suggested that the NF controller generates
the suitable command signals (inputs of the hydraulic
actuators) to improve the vehicle performance including
riding comfort and road handling stability. The rigid comfort
can be measured by evaluating the acceleration and
displacement of the sprung mass. The handling stability can
be obtained by minimizing the vertical motion of tires and the
rotational motions of the vehicle body such as rolling and
pitching movements during sharp manoeuvres cornering and
braking.
The full vehicle active suspension physical model is shown in
Fig. 5. This model consists of five parts: the sprung mass (M)
and four unsprung masses mi (where ∈[ ] ,,,i 4321 ). The
sprung mass is assumed as rigid body and has freedom of
motion in vertical, pitch and roll direction. The vertical
displacem
SIMULATIONS AND RESULTS
For the full vehicle nonlinear active suspension system
discussed in Section 4, the numerical values of the hydraulic
actuators and full the vehicle model which are used in this
simulation are given in Table 1. To design the NF controller,
the optimal parameters of FOPID controller should be
obtained first using the EA. The input/output data obtained
from the FOPID controller have been used to design the NF
controller. Figures 6-10 show the changing of the FOPID
controller parameters (proportional constant Kp, derivative
constant Kd, integral constant Ki, integral order λ and
derivative order μ) with respect to optimization steps. Fig. 11
shows the response of the cost function (which is described in
Eq. 10) with respect to the optimization steps. After 225
optimization iteration steps, the optimal values of the FOP
CONCLUSION
A novel Neurofuzzy controller has been successfully
developed for a full vehicle nonlinear active suspension
system. The results have been compared with optimal FOPID
controller and the corresponding system without controller.
From these results, the NF controller has capability of
minimizing the control objectives better than the optimal
FOPID controller. The test of the robustness proves that the
NF controller is still stable and it forces the cost function to
be minimum even significant disturbances occurred. The
results have been confirmed that when the NF controller has
been used, the cost function is still away from zero while
when the optimal FOPID controller is used the cost function
has much bigger values.