Neural Controller Design for Suspension Systems

The main problem of vehicle vibration comes from road roughness. An active suspension systempossesses the ability to reduce acceleration of sprung mass continuously as well as to minimizesuspension deflection, which results in improvement of tire grip with the road surface. Thus, braketraction control and vehicle maneuverability can be improved consider ably .This study developeda new active suspension system for a quarter-car model. The designed system is based on neuralnetwork controller with an input as a regressor and it provided through a lag network thatincludes reference input , system output and control signal system to the previous sate. In thispaper, the system is based on neural network controller that is a regressor input provided througha lag network, including reference input, system output and control signal to previous state. Theneural network outputs are the same control signals applied to the suspension system. Feedbacksystem is taken as the output of the displacement body and is applied to lag network. Roughness ofthe road surface is considered as a reference input. To train, the neural network uses different ideaby introducing a cost function for the system and optimizing it, the best coefficients are selectedfor the neural network.