Fault Diagnosis of Railway Turnout Based on Random Forests

The turnout is a key component of the railway infrastructure systems and is considered as a critical issue about the train operation safety. Therefore, the fault diagnosis research of the turnout is important. However, the existing methods of the fault diagnosis for the railway turnout have the problems such as low efficiency, inability to meet timeliness, and insufficient accuracy. To solve these problems, this paper presents a fault diagnosis method based on random forests. The random forests algorithm builds many CART decision tree classifiers, and introduces two random procedures: i.e., random samples and random features, to enhance the diversity of each decision tree classifier. The final classification result is obtained by majority voting method, which improves the execution speed and classification accuracy. In this paper, a case study is also presented by using the electric power data of the S700K switch machine, and the random forests classification model is constructed. The result shows that the random forests algorithm can accurately and quickly give the diagnosis results for the status of the railway turnout.