Fault Diagnosis of High-Speed Railway Turnout Based on Convolutional Neural Network

Abstract

Fault diagnosis is critical to ensure the safety and reliable operation of high-speed railway. The traditional fault diagnosis methods for high-speed railway turnout rely on manual features extraction using turnout raw data, but the process is an exhausted work and greatly impacts the final result. Convolutional neural network (CNN), as a typical deep learning model, can automatically learn the representative features from the raw data. This paper investigates an intelligent fault diagnosis method for high-speed railway turnout based on CNN. The turnout current signals in time domain are converted to the 2-D grayscale images, and then the grayscale images are fed into the CNN for turnout fault classification. The proposed method is an automatic fault diagnosis system which eliminates the complex process of handcrafted features. The experimental results show a significant improvement over the state-of-the-art on the real turnout dataset for current curve and prove the effectiveness of the proposed method without manual feature extraction.