Railway door fault detection improves with cycle-aware autoencoder model
Cycle-Aware Autoencoder with Cross-SignalConsistency for Railway Door Anomaly Detection
Machine LearningArtificial Intelligence
Summary
Railway doors are important for passenger safety, but detecting when they malfunction is hard because problems are rare and varied. The authors propose a method that looks at each door opening and closing cycle as a whole to spot unusual behavior without needing examples of faults beforehand. Their model analyzes multiple types of door signals, checking that they make sense together over time, and this helps catch faults more accurately. Tested on real train data, their approach identifies door issues with high accuracy and low false alarms, showing it could work in real-time on trains.
What this means in practice
- •For railway maintenance teams: Automate detection of abnormal door behavior on trains using onboard sensors to improve safety and reduce manual inspections.
- •For industrial iot developers: Incorporate cycle-level anomaly detection for safety-critical mechanical subsystems where multi-signal consistency is important.
Authors
Ammar Bouketta, Smail Niar, Hamza Ouarnoughi, Eva Mutuzo Brindle
Abstract
Passenger access doors are safety-critical subsystems in railway vehicles, yet detecting abnormal door behavior in real operation is challenging because faults are rare, diverse, and often unlabeled. This paper addresses railway door condition monitoring as a cycle-level unsupervised anomaly detection problem, where each complete opening-dwell-closing cycle is treated as a single monitoring unit. We propose the Temporal Cycle-Aware Attention Autoencoder with Cross-Signal Consistency (TCAA-CS), trained exclusively on nominal cycles. It combines a dual-stream encoder that processes continuous physical measurements (position, current, voltage) and binary logical states (door-closed, door-locked) through separate 1D-CNN branches, an LSTM encoder with temporal attention pooling, and a triple hybrid anomaly score fusing reconstruction error, latent-space deviation, and phase-aware cross-signal consistency. The consistency term helps identify cases where individual signals appear plausible but their inter-signal relationships become physically or logically inconsistent. On real industrial data from a passenger train in commercial service, TCAA-CS achieves 93.8% recall, 97.3% precision, and a 0.5% false-alarm rate, outperforming representative unsupervised baselines. System-level evaluation on an NVIDIA Jetson AGX Xavier supports the feasibility of real-time onboard deployment.