A pattern recognition and data analysis method for maintenance management

Abstract

This article presents a pattern recognition method based on grouping by linear relationship a set of faults. The majority of faults can be detected, but only a few experiments can be identified. The algorithm called Principal Component Analysis (PCA) is employed together with the statistical parameters of the signals for detecting and identifying the faults. PCA technique is utilised for modifying dataset reducing the coordinate system, which must be correlated, by linear transformation, into a smaller set of uncorrelated variables called ‘principal components’. The signals analysed were the current and force signals in normal-to-reverse and reverse-to-normal directions of the system.

Keywords:

• reliability centred maintenance
• remote condition monitoring
• maintenance management
• principal component analysis
• railway infrastructure

Acknowledgements

The author would like to thank to M.A. McHutchon from Sheffield University (UK) and Felix Schmid (from Birmingham University) for his helpful comments, suggestions and background.

Notes

Note

1. Some authors refer to the combination of switches and crossing as a turnout, i.e., the mechanism allowing trains to change tracks and access sidings. For this article, the author adopts the term ‘points’ to describe the complete arrangement.