AI-Driven Predictive Maintenance for Enhanced Reliability in Railway Power Systems: A Technical Review

Reliability in railway power systems is critical for ensuring uninterrupted, safe, and efficient rail operations. Traditional maintenance strategies such as time-based and corrective approaches often fail to anticipate failures, resulting in service disruptions, increased costs, and reduced asset lifespan. This study presents a comprehensive technical review of artificial intelligence-driven predictive maintenance (AI-PdM) as a transformative solution to enhance equipment reliability in railway power systems. The proposed AI-PdM framework leverages advanced techniques such as Long Short-Term Memory Recurrent Neural Networks (LSTM-RNN), Machine Learning (ML), and Autoencoders to analyze data collected from temperature, vibration, partial discharge, and stray current sensors. These models estimate Remaining Useful Life (RUL) and enable early failure detection. Integration with Supervisory Control and Data Acquisition (SCADA), Energy Management Systems (EMS), and Enterprise Asset Management (EAM) platforms ensures actionable maintenance scheduling. Case studies including traction power transformers, gas-insulated switchgear (GIS), and overhead catenary systems demonstrate how AI-PdM can reduce unplanned outages, optimize maintenance resources, and extend asset life. Challenges such as data integration, model interpretability, cybersecurity, and change management are also addressed, along with strategies to overcome them. The findings support AI-PdM as a practical and scalable solution for transitioning from reactive to proactive maintenance in railway power systems. Its implementation promotes improved system reliability, cost efficiency, and safety in the context of growing infrastructure demands.

کلیدواژه ها:

Railway ، Predictive ، Artificial Intelligence ، Maintenance