A fuzzy based maintenance decision making methodology for railway infrastructures to solve the problem of a specific rolling contact fatigue defect called squat is presented. Relying on Axle Box Acceleration (ABA) measurements to detect squats, a robust model is used to predict the evolution of detected squats under three different growth scenarios (fast, average, slow). Once a track is candidate for preventive maintenance, to facilitate grinding operations, we propose a fuzzy clustering methodology to group light squats that are good candidates to be treated together because of their proximity. Using a new α% method based on clustering, Pareto fronts are analyzed to show the trade-offs between the number of non-treated squats and coverage percent related to α%. Based on the prediction model, four key performance indicators at the technical level are combined using a fuzzy expert system which estimates the global health condition of each cluster of squats. Estimated cost effectiveness of different maintenance actions are compared. Therefore, infrastructure managers will be able to estimate track condition per cluster, and then easily rank the clusters of light squats according to their importance. Also, the predictive indicators enable infrastructure managers to decide which parts of tracks need to be replaced when severe squats are detected in the track. To illustrate the proposed method, the track Groningen-Assen of the Dutch railway network is used.

This content is only available via PDF.
You do not currently have access to this content.