Abstract: This paper presents a study on hybrid modeling of automotive suspension system equipped with magnetorheological(MR) dampers. The linear dependences of the damping force of MR damper on the velocity and acceleration are identified using the linear regression method, while the nonlinear dependences are modeled using neural network, which produces a dynamic model of MR damper. The hybrid model of a half-automotive semi-active suspension system of four-degrees-of-freedom with two MR dampers is then established through integrating the model of MR damper with the prior knowledge-based model of the system. The developed system model is compared with that built based on nonlinear hysteresis model of MR damper in terms of the predicted dynamic responses for a SANTANA2000 car suspension system. The results show that the presented hybrid model has the advantages of simple network structure, small computational cost, high accuracy. It is convenient for analysis of dynamic systems.