A Risk Prediction Method for EMU Maintenance Operations Based on Dynamic Bayesian Networks

The smooth execution of Electric Multiple Unit (EMU) maintenance operations is fundamental to ensuring the safe and efficient operation of railways. However, numerous risk factors are present during maintenance activities. Without scientific and systematic risk prevention and control, the safety of maintenance personnel may be compromised, potentially affecting the normal operation of trains. Existing research primarily focuses on expert-driven management mechanisms and risk control based on static data analysis. To address this limitation and achieve dynamic risk management, this paper adopts Job Hazard Analysis (JHA) to identify potential risks in Level I maintenance tasks of EMUs. The identified risks are then mapped into a Dynamic Bayesian Network (DBN). By leveraging the diagnostic inference capabilities of DBNs, the proposed method calculates the occurrence probabilities and development trends of risk events throughout the maintenance process. This approach enables the identification of critical risk factors and facilitates real-time risk prediction for maintenance personnel.