Abstract:In order to model the nonlinear MR damper,the neural network models for the direct model and the inverse model of a single MR damper are created respectively. On this basis,two different recurrent neural network (RNN) inverse models for four MR dampers are designed and applied to the control of the semi-active suspension of the full-vehicle model. The simulation results demonstrate that the first RNN inverse model can greatly reduce the vertical acceleration and pitch angular acceleration and roll angular acceleration of the sprung-mass so that the ride comfort and handling of the semi-active suspension can be dramatically improved. However,there is still room for improvement for the second RNN inverse model.