Condition Based Maintenance for Railway Turnouts

This article focuses on the specific study of special type A turnout. Today, this type of track apparatus is inspected by visual reconnaissance of the tracks and using specialized measuring equipment to detect irregularities in the rails such as wear or deformation. Both the visual recognition and the measurements made are recorded in a control form that is then evaluated in order to determine the necessary control action.
Thus, this article presents an algorithm based on data analysis that allows us to evolve towards a predictive maintenance model for special track segments.
It comprises the following main technical objectives: Analysis of the potential of data-driven anomaly detection methods, proposing a new approach that incorporates machine learning techniques through statistical pattern recognition. Diagnosis or evaluation of the condition of the track apparatus that allows the fault to be detected, identified, or located. Implementation of a valuable tool that allows the evolution of the maintenance strategy towards predictive maintenance management. Recommendation in terms of maintenance.