Turnout Failure Diagnosis System Based on Group Decision Making Strategy

Abstract

The turnout system emerges as the most critical component in railway infrastructure that possesses the function of controlling tracks where its faulty operations should be carefully concerned. The traditional turnout failure diagnosis is conducted manually, field staffs need to monitor thousands of turnout current curves based on expert experience per day, resulting in unstable diagnosis results. Thus, this paper utilized artificial intelligence and the group decision concept to propose a voting algorithm that considered the failure diagnosis results of three machine learning models, aiming to provide a feasible intelligent turnout failure diagnosis system with high accuracy. The training samples were filtered based on an experience-based method and fully cleansed for achieving more stable and reliable classification results. For evaluating the diagnosis performance, Recall and Precision were applied. As a result, the group decision failure diagnosis system indeed revealed high accuracy with low failure case omission.