Research on turnout system diagnosis based on group decision making

Abstract

Safety has always been the primary consideration in high-speed railway operations. Turnouts are one of the most safety-critical components in the railway system, yet its fault diagnosis still relies on manual inspections, which is time-consuming and can lead to missing reports of turnout malfunctions. This paper proposes an automatic turnout fault diagnosis method based on group decision making. By assembling three individual classification algorithms, including the k-nearest neighbors algorithm, naive Bayes classifier, and deep neural network, this algorithm aims to automate turnout fault diagnosis and reduce the possibility of missing reports. Experiments to compare the performance of the group decision making algorithm and the three individual classifiers based on datasets generated by real turnout systems in simulated fault conditions are carried out. The result shows that the recall, i.e., the sensitivity to turnout fault of this algorithm is superior to the individual classifiers without losing overall accuracy, indicating that the missing report rate can be reduced through the group decision making process.