Data Analysis for Condition-Based Railway Infrastructure Maintenance

Condition assessment is crucial to optimize condition‐based maintenance actions of assets such as railway infrastructure, where a faulty state might have severe consequences. Hence, railways are regularly inspected to detect failure events and prevent the inspected item (e.g. rail) to reach a faulty state with potentially safety critical consequences (e.g. derailment). However, the preventive measures (e.g. condition‐based maintenance) initiated by the inspection results may cause traffic disturbances, especially if the expected time to a faulty state is short. The alarm limits are traditionally safety related and often based on geometrical properties of the inspected item. Maintenance limits would reduce the level of emergency, producing earlier alarms and increasing possibilities of planned preventive rather than acute maintenance. However, selecting these earlier maintenance limits in a systematic way while balancing the risk of undetected safety‐critical faults and false alarms is challenging. Here, we propose a statistically based approach using condition data of linear railway infrastructure assets. The data were obtained from regular inspections done by a railway track measurement wagon. The condition data were analysed by a control chart approach to evaluate the possibility for earlier detection of derailment hazardous faults using both temporal and spatial information. The study indicates that that the proposed approach could be used for condition assessment of tracks. Control charts led to earlier fault warnings compared to the traditional approach, facilitating planned condition‐based maintenance actions and thereby a reduction of track downtime. Copyright 2014 The Authors.