An automated predictive maintenance system has been developed based on the Hough Transform. The Hough Transform, used in image processing to extract geometrical primitives from image data, has been adapted to extract linear trend features from monitored data. The predictive maintenance system uses the extracted trend features to diagnose incipient fault conditions. It then predicts the time for the system to reach a critical fault condition. The transform provides an early warning of a developing fault allowing operating staff to anticipate the types of maintenance required before secondary damage and complete breakdown occurs. It can also deal robustly with noisy data, including impulsive noise.
It is proposed that logistic curves may be useful for the empirical description of wear failure. A multi-point generalisation of the Hough Transform (Flint et al, 1992) was developed for extracting logistic features from developing fault signatures. The new generalisation of the Hough Transform differs from earlier generalisations by incorporating early data fusion to reduce the size of previously large cumbersome arrays.
A brief description is given of the Hough Transform's application to two predictive maintenance case studies. The first case study is the uni-feature problem of providing predictive maintenance decisions for a diesel-engine lubricating oil system. The subject of the second case study is a pneumatic sliding-door system. Here a multivariate generalisation of the Hough Transform is required. The proposed generalisation involves the accumulation of vector rather than scalar votes in an array.
Ballard, D.H.1981. 'Generalizing the Hough Transform to detect arbitrary shapes'. Pattern Recognition.13, 111-122.
2.
Chatfield, C., and Collins, A.J.1992. Introduction to Multivariate Analysis. Chapman and Hall.57-81.
3.
Dolan, S., and Sasena, E.1989. 'Hough in hardware: processors get algorithm-specific '. Electronic Sys Design Mag. February. 51-56.
4.
Draper, N., and Smith, H.1981. Applied Regression Analysis (2nd Edn). John Wiley & Sons, Inc.505-517.
5.
Duda, R.O., and Hart, E.R.1972. 'Use of the Hough Transform to detect lines and curves in pictures'. Communications of theACM, 15(1), 11-15.
6.
Flint, A., Ingleby, M., and Morton, D.1992. 'A new generalisation of the Hough Transform - applied to automatic prognosis of condition from monitored data'. Proc IEEE conf Intelligent Control92, Glasgow. 261-267.
7.
Hough, P.V.C.1962. Method and means for recognizing complex patterns. US Patent 3 069 654.
8.
Li, H., and Lavin, M.A.1986. Fast Hough Transform based on a bintree structure. Proc of the IEEE Computer Vision and Pattern Recognition Conference , San Diego, USA. 640-642.
9.
Low, A.1991. Introductory Computer Vision and Image Processing. McGraw-Hill Book Company. 67-115.
10.
Orchard, G., Hurton, A., and Fry, K.1990. 'Diesel engine condition monitoring, System installation and progress'. British Rail Internal Report, MR VS 65.
11.
Price Waterthouse in association with Aladon.1990. 'Reliability-Centred Maintenance.' Reliability-Centred Maintenance Course Notes for British Rail.
12.
Radon, J.1917. Über die Bestimmung von funktionen durch ihre Integralwerte langs Mannigfaltigkeiten. Berichten Sächs, Akad. Wiss. Leipzig, Math Phys. kl, 69, 12-39. Translation: 1986. 'On the determination of functions from their integral values along certain manifolds'. IEEE Trans on Medical Imaging, MI-5 (4).
13.
Tatari, S.1981. 'Procedures for combination of physically different signal channels for classification tasks'. FhG-Bericht1/2, 19-22.
