Study Of Neural Network Self-adapt Control For Vehicle Of Semi-active Suspensions

Suspension is one of the most important parts of vehicles, which has tremendous influence on performance of ride quality and handing stability. The stiffness and damping of traditional passive suspension can not be adjusted, o it difficult to keep optimal performance while the working conditions is variably. The active suspension has the ability of kept the car has optimal ride comfort and handling stability in any circumstances, but it has the disadvantage, such as complex structure, high cost and energy consumption. Semi-active suspension is composed of controllable spring and damper element which can keep the vehicle with preferable ride comfort and handling safety. At the same time, it has the advantage of lower power consumption and easy to realized, so the semi-active suspension has extensive application prospects.Automobile suspension is a complicated nonlinear system, and using traditional control methods to control suspension has a certain limitation. Neural network is the result of simplicity and simulate of cranial nerve network. Neural network has characteristic of learning ability and parallelism, good adaptability and fault tolerance. Neural network is suitable for more extensive system and uncertain environment compared with traditional adaptive systemThe multi-rigid-body system dynamics model of semi-active suspension is established in this paper by using ADAMS software to provide the basis for the constructor of neural network control system. A multi-rigid-body system dynamics model with seven degrees of freedom semi-active suspension is established in ADAMS/View module which has front double wishbone front suspension and towing back suspension. Then the suspension model is imported into Matlab/Simulink by ADAMS/Control module.Body vertical acceleration of vertical is taking as control objective of neural network control system in this paper. Damping force of adjustable damping shock absorber are adjusted by neural network controller to improve the ride quality of vehicle. The structure of neural network controller and neural network identifier are BP neural network. Simulation program was compiled in Matlab to finish simulation study of neural network control system. To verify the performance of neural network control system, simulations were simulated with random road profile at25m/s to demonstrate the effectiveness of the proposed control system in comparison with the passive suspension. The comparison shows that neural network control system improved the performance of the full car suspension system significantly. The simulations also verified that co-simulation is feasible and effective.