Research On Data-driven Collaborative Diagnosis Method of Traction Converter Sensors and IGBTs Composite Faults

The composite fault diagnosis of sensors and IGBT power devices in rail transit electric traction systems faces challenges, such as fault characteristics are highly coupled and difficult to recognize online. In this paper, we propose a data-driven collaborative diagnosis method based on the fusion of multi-source timing signals and improved random vector function link network (RVFL) to achieve the simultaneous diagnosis of sensor and IGBT faults. Experimental results obtained from the RT-LAB hardware-in-the-loop simulation platform demonstrate that the proposed method achieves over 97% accuracy in single-fault diagnosis with an average response time below 60 ms. For composite fault diagnosis, the system maintains 96% diagnostic accuracy while responding within 76 ms on average. This result shows that the method can effectively solve the problem of compound fault feature coupling and the difficulty of real-time detection, which provides a new idea for the fault diagnosis of rail transit traction system.