A Fault Detection Method for Railway Point Machine Operations Based On Stacked Autoencoders

Abstract

Fault detection of point machine operations is discussed in this paper, which is critical for ensuring the safety of a running train. A fault detection method is proposed based on stacked autoencoders (SAE), which can be easily trained and has great expressive power. The method only requires normal samples to train the SAE model, and integrates feature extraction and fault detection into one step. The proposed method is evaluated by using the historical field data collected from a real high-speed railway. Experimental results show the effectiveness and merits of the SAE based detection method.