An experimental model identification of magneto-rheological damper with neural network and its application in seat suspension system

A model identification method of a magneto-rheological (MR) damper with back propagation neural network (BPNN) and Radial Basis Function neural network (RBFNN) is discussed in this study. A vibration testing system is constructed to obtain the experimental data of the MR damper, such as the displacement and the velocity of the piston rod, the input current applied to the coil and the MR damping force. Forward model based BPNN and inverse model based RBFNN of the MR damper is trained with 70% of the experimental data. And the remainder data are used for testing. It is shown by the comparison of the prediction and the target that both forward model and inverse model can demonstrate a good training effect. In order to verify the accuracy of the models of the MR damper and the effectiveness of skyhook control strategy, skyhook control strategy and passive control strategy are applied to the seat suspension system with both forward model and inverse model, respectively. The simulation results show that the trained models of the MR damper is very successful. Meanwhile, skyhook control strategy can achieve better acceleration attenuation for the seat suspension system and improve ride comfort.