Data-Driven ToMFIR-Based Incipient Fault Detection and Estimation for High-Speed Rail Vehicle Suspension Systems
Yunkai Wu, Yu Su, Peng Shi
Abstract
Fault detection and estimation issues of China railway high-speed (CRH) train suspension systems in early stage are addressed in this article based on data-driven design of total measurable fault information residual (ToMFIR). First, a discrete trailer car model of the CRH train is established. Based on this model, input/output (I/O) data matrices and system data models are constructed step by step using ToMFIR theory through sensor measurements. By utilizing the projection on controller residual, the data-driven form of ToMFIR residual can be further obtained. For the purpose of efficient and accurate incipient fault detection, the Kullback–Leibler divergence (KLD), an indirect method, is employed to evaluate and monitor the slight changes in the ToMFIR residual in matrix form. Finally, a fault amplitude estimation method based on KLD for detecting incipient sensor effectiveness loss is introduced. Simulation results demonstrate that the data-driven detection and estimation scheme proposed offers higher sensitivity to spring faults, damper faults, actuator faults, and sensor faults of CRH train suspension systems in early stage.