Deep Forest-based Fault Diagnosis for Railway Turnout Systems in The Case of Limited Fault Data

Railway turnout system is one of the most important equipment in railway infrastructures. It is also one of the most vulnerable assets that are likely to be affected by hostile weather and mechanical wear. Effective fault diagnosis can play an important role in reducing railway disasters caused by turnout failures. First, this paper analyzes the structure of fault types and power curves of turnouts. Then a new diagnosis method for railway turnouts is presented based on deep forest. The method fully considers the characteristics of fault data, strong sequence, and limited amount of failure data. Finally, this paper conducts an extensive evaluation by using real-world data from the Changsha Railway Station in China. The experimental results show that the proposed method achieves better performance compared to state-of-the-art methods in diagnostic accuracy under limited field data, with a fault diagnosis accuracy of 97%.