ROLLING STOCK MAINTENANCE PLANNING. AUTOMATIC OPTIMIZATION BASED ON FAILURE PREDICTION

Abstract

ABSTRACT Nowadays, industry faces the optimisation of all the processes that concerns its activities: costs, quality, timing, production, etc. In this way, maintenance in transportation system, including underground passengers transport, deals with the same requirements. So, maintenance management strategies can be optimised both from technical and economical point of view. In this paper a system for optimising maintenance planning in underground metropolitan transport vehicles is presented. The system is based on failure prediction that uses failure data filed along time. Two different prediction approaches are used. The first one is based on the probability density function estimation associated to the time and mileage between failures. The second one considers the time and mileage between failures as a time series and forecasts values using a Support Vector Machine. The predictions obtained are compared against maintenance planning, which is designed according to coach manufacturer, operator experience and applicable regulations. Depending on the comparison result, the system can propose to modify the maintenance planning in three possible ways: to create new maintenance operations, to delete some of them and to change planned date or mileage of selected maintenance tasks. Preliminary results are shown.