Research On Fault Detection Under Stochastic Model For Suspension Systems Of High-speed Railways

The thesis is supported by National Natural Science Foundation of China. With the background of stochastic model, the thesis is to study fault detection technology for the suspension systems of high-speed railways. Regularly, the suspension systems encounter severe environment, track irregularity and random vibration, which would lead to faults easily. In this paper, the work of fault detection is done under the stochastic model taken the actuator and sensor faults into account, which provides effective theoretical way to improve the stability of the vehicle attitude control and ride comfort.Firstly, the dynamic stochastic model and its related parameters is presented for the structure of passive and active suspension systems, which considers the influence of the unknown disturbances caused by the track irregularity and random vibration of the suspension systems. Then the fault types and levels of the suspension systems are described, the types of vehicle sensors and its related measurement objects are also analyzed, which provides a signal acquisition channel for the state monitoring and fault diagnosis.Secondly, with the existences of unknown disturbances and random noise, the sensor faults of the suspension systems is studied in the stochastic environment. The system state and the unknown disturbances are reconstructed, a set of whole dimension state observer is designed to estimate the system state and disturbances simultaneously, and generate the residual signal which is robust to disturbances and sensitive to faults. The random distribution of residual is analyzed and then fault can be detected based on the hypothesis testing. At the same time, the fault amplitude condition is proposed in respect to be detected, and the false alarm rate, miss detection rate in the systems are also discussed.Thirdly, the DC motor and the second stage damper faults are considered in the active suspension systems, and the actuator faults detection is studied in the closed loop system. Compared with the passive suspension systems, the active ones increase the DC motor acted as the actuator. When there are some faults in the motor, the adjustable damper can not be adjusted to control the attitude of vehicle body. Considering the influence of load torque and vibration noise on the motor, the fault detection method based on the observer designing under stochastic model is proposed. The residual is analyzed to hypothesis test and the fault detection is completed with the detectability condition quoted. The damper faults detection in the active suspension system is consider in the closed loop structure. The system state and disturbances are estimated by the cited Kalman filter, then the residual is generated by the To MFIR(Total Measurable Fault Information Residual), which can not only detect the faults in the closed-loop system, but also can distinguish the influence of disturbances and faults in the systems. In the end, the random distribution of residuals analysis for hypothesis testing, and quantitative analysis the false alarm rate, miss detection rate in the systems.Finally, the effectiveness and feasibility of the actuator and sensor faults detection of the suspension system are verified by simulation test.