As an important part of CRH(China Railway High-speed)trains, the reliability of the high-speed train suspension system is of critical importance to the safety of the entire trains. A data-driven based incipient fault detection scheme is proposed in this paper based on the sensor data of suspension system. Firstly, a suspension system model is established by using Simpack, where the active force of actuator is acted as system input, and the track irregularity is regarded as system disturbance. Secondly, based on the discrete model of suspension system, the data model and input/output data matrices are constructed by using the sensor measurements. Finally, the off-line and on-line detection scheme are proposed based on the residual constructed by data matrices. The simulation results show that the proposed scheme has high sensitivity on incipient actuator and sensor faults of high-speed train suspension systems.