A fault diagnosis and measurement method for temperature measurement circuits in electric traction drive system

Real-time and accurate measurement of operating temperature in electric traction drive system (ETDS) is essential to ensure the stable operation of the train. To eliminate the impact on the temperature measurement circuit (TMC) faults which include sensor and signal process circuit, this paper proposes a data-driven based fault diagnosis and fault-tolerant measurement framework. Firstly, according to the cooling and temperature measurement mechanism in ETDS, the TMCs of four water temperature sensors and two oil temperature sensors in ETDS are considered. An extreme learning machine (ELM) based estimation method for temperature measurement of each TMC is proposed by learning the system redundant relationship without additional hardware. Secondly, an anomaly detection strategy is proposed by studying the residual law between the estimated and measured value. Thirdly, a fault isolation and fault-tolerant measurement decision making method based on sliding window probability is proposed. Finally, field tests are performed and the results show that, compared with the traditional method, the proposed method has better measurement and diagnostic performance, and it is also valuable for practical application.