关键词: 车辆工程, 冗余转向系统, 电流传感器, 故障诊断及容错, 数据与机理混合驱动, 硬件在环

Abstract: A fault diagnosis and fault-tolerance strategy of current sensor based on data and mechanism hybrid driven is proposed for redundant steering system with dual-winding motor. The redundant steering system architecture is analyzed, the current sensor fault model and dual-winding permanent magnet synchronous motor(DW-PMSM) model are established. On this basis, the fault diagnosis and fault-tolerance strategy is designed. The coordinate transformation is used to estimate the motor phase current, which is used as the external input of nonlinear autoregressive with eXogenous inputs(NARX) neural network. The single-step and multi-step prediction of DW-PMSM phase current is realized by changing the structure of NARX neural network. Therefore, the fault detection, location and classification strategies of current sensor are designed. The fault severity of the sensor is extracted, and the fault-tolerance control is realized through the compensation of the current sensor or switching. Based on Matlab/Simulink and dSAPCE, a steering system simulation platform and a hardware-in-the-loop test bench are built to test and verify the algorithm. The results show that the proposed algorithm can quickly detect the fault of current sensor and achieve fault-tolerance control under various working conditions, and improve the reliability and safety of redundant steering system.

Key words: vehicle engineering, redundant steering system, current sensor, fault diagnosis and fault-tolerance, data and mechanism hybrid driven, hardware-in-the-loop