<title>Comparison of neural network and cluster analysis techniques for the reliability of high-speed machinery</title>

Abstract

This paper presents a comparison of three differing methods applied to the analysis of control data from a high speed machinery application. The source data and the pre-processing applied to improve the suitability of the data to the analysis techniques is discussed. The methods compared are cluster analysis, multi-layer perceptron neural networks and self organizing feature maps. The aim of the work is to determine the merits of the techniques in separating normal running operation from faulty operation. The methodology used with each technique is explained and results are computed so as to give the fairest comparison of their respective abilities. Additionally, the ways in which such techniques would be integrated into a final system for the analysis, diagnosis, and control of a high speed machine to give improved reliability are discussed.© (1995) COPYRIGHT SPIE--The International Society for Optical Engineering. Downloading of the abstract is permitted for personal use only.