AI BASED PREDICTIVE MAINTENANCE SYSTEM FOR RAILWAY ASSET

Abstract

This project presents on AI based predictive maintenance framework for critical railway assets, specifically focusing on track geometry, switching points and rolling stock components. The proposed architecture utilizes a hybrid Deep Learning (DL) model combining Long Short-Term Memory (LSTM) networks and Convolutional Neural Networks (CNNs) to extract spatial-temporal features and forecast Asset Health Index (AHI) degradation curves. Validated on a realworld regional transit dataset, the AI framework demonstrated a 94.2% accuracy in Remaining Useful Life (RUL) estimation and successfully predicted critical track faults up to 14 days in advance. The implementation of this system yields a projected 22% reduction in unplanned downtime and a 15% decrease in overall maintenance expenditure, offering a scalable solution for cognitive asset management in next-generation rail infrastructure. The modern railway networks demand high operational availability, safety, and cost-efficiency, rendering tranditional reactive and timebased scheduled maintenance paradigms inadequate.