Study On Neural Network Self-Adapt Control For Vehicle Of Semi-Active Suspension

Suspension is one of the important part of vehicle, which has tremendous influence on performance of ride quality and handing stability. Passive suspension is difficult to meet requirements for a car which pursuits under variable environment. So, electronic suspension ?is one of the development direction of automotive. Active suspension needs extra force to control it, but this could consume more energy and enhance the weight of vehicle. Semi-active suspension is composed of controllable spring and damper element which consume little energy and are easy to design and manufacture. Because semi-active suspension can improve ride comfort so automobile researchs are enthusiastic are about the research into the semi-active suspension.Vibration of automobile in low frequency band has importment effect on automobile performance and the air spring has high performance in this frequency band, the semi-suspension with controllable spring stiffeness can be used widely to improve the vehicle's ride comfort.Variable rate semi-active suspension is a nonlinear system, but groovy control methods take on states limits when it is applied to the system. So more efficient control polices are required have contain limit to control the semi-suspension more practically and more effectively. In this dissertation nonlinear control method-neural network control method in allusion to variasble rate semi-active suspension are investigated based on its characters to carry on research simulation and experiment.In this paper, according to the ride comfort level a quarter car mathematical model of two-DOF has been set up. Considering nonlinear rigidity of spring element, this dissertation use neural network control strategy to control semi-active suspension. The performance indexes body acceleration of vehicle are determined to show the ride comfort, at the same time are concerned about the value of suspension wheel load and the value of suspension displacement which show manipulate stability.Passive suspension, PID control semi-active suspension and neural network control semi-active suspension are researched by simulation. White noise, sine wave and sawtoothwave are used for excitation. The root mean square value of vehicle body acceleration, suspension displacement and vehicle wheel load are analyzed. And the results are showed in tables and charts. From the results we can draw a conclusion that semi-active suspension is much better than the passive suspension in improving the ride comfort and manipulate stability. The neural network control semi-active strategy is better than PID control strategy, it is a feasible and effective strategy..During the experiment, the direct neural network control strategy are used because it is easy to calculate. From the frequency responses of the body acceleration we can see that neural network control strategy reduces the acceleration of the body and can improve the ride comfort of vehicle.Given the time-variability parameters and the nonlinearities in variable rate semi-active suspension, the new strategy is employed to control the nonlinear characteristics of the suspension system. From the simulation and experiment research, we konw that the semi-active suspension self-adapt neural network control is applicable and effective. It also offers a new idea to the field of research on variable rate semi-active suspension control strategies.