AI-Driven Failure Mode and Effect Analysis (FMEA) of Traction Motor Bearings in Urban Rail Systems: Predictive Maintenance Implementation and Cost-Benefit Evaluation

Urban rail systems demand high availability and safety standards, requiring advanced maintenance strategies for critical components such as traction motor bearings. This paper presents an integrated framework combining Failure Mode and Effect Analysis (FMEA) with artificial intelligence (AI) techniques to enable predictive maintenance (PdM) for Bombardier MJC ۲۵۰-۵ traction motors operating in Shiraz Metro. The methodology employs Random Forest (RF), Long Short-Term Memory (LSTM) networks, and Autoencoder-based anomaly detection to analyze vibration and thermal signatures from electrically insulated roller (NU۱۰۱۸-M۱-F۱-J-۲۰AA-C۴) and ball (۶۲۱۴-M-J۲۰AA-C۴) bearings. FMEA identified bearing outer race spalling as the highest Risk Priority Number (RPN = ۴۳۲). Machine learning models achieved ۹۱.۳% accuracy in fault classification with ۱۸-day advance warning capability. Economic analysis demonstrates a payback period of ۱.۸ years with ۳۴% reduction in life-cycle costs. Results validate the feasibility of transitioning from time-based maintenance to condition-based monitoring, enhancing fleet reliability and operational efficiency for metropolitan transit networks.

کلیدواژه ها:

Predictive Maintenance ، FMEA ، Traction Motor Bearings ، Machine Learning ، Urban Rail ، Condition-Based Monitoring