Signal processing for remote condition monitoring of railway points

The dependability of railway points (turnouts or switches) is a key part of any railway system; the Potters Bar crash (10 May 2002) in the UK, which led to seven fatalities is a key example of a failure of this subsystem. Present maintenance of points involves overly frequent inspection by maintenance staff. A remote condition monitoring approach would lead to more efficient inspection routines and directed anticipatory maintenance trips. To assist the creation of a suitable fault-detection algorithm, the authors analysed existing force and current data for the 'as commissioned' case of a turnout and for situations with different fault conditions. Signal processing of this data revealed several different methods that can be used to distinguish between fully functioning points, and different fault conditions of the points. Specifically, clustering of statistical parameters and their application to wavelet levels and coefficients, provided clear discrimination of most critical faults. This demonstrates a good first step towards a condition monitoring-based maintenance regime for points that is both safer for passengers and maintenance personnel and that has the potential to be more effective and economical.