Towards the development of an automated wear particle classification system

Wear particles can be found in any tribological system. They are the direct consequence of wear processes occurring and their morphology reflects the wear mechanisms involved in their generation. This dependence of wear particle morphology on the wear mechanisms has been exploited and widely used in machine condition monitoring. The problem with the methods currently in use is that they require a trained staff, i.e. these methods are time consuming, costly and not always reliable since they rely on the technicians' experience. Recent efforts are therefore directed towards the development of an automated particle classification system. This is not a simple task as there are three major problems which need to be overcome first before the system can find practical industrial applications. The first problem is the establishment of the extensive and detailed database of wear particle morphologies generated under different operating conditions and with different materials used for sliding pairs. The second problem is the development of reliable and accurate description of 3D wear particle morphology, while the final problem is the development of a wear particle classification system. As such system must be robust and reliable this involves the development, selection and testing of appropriate classifiers. It is hoped that the system, when fully operational, will eliminate the need for highly trained staff, and provide more accurate and reliable diagnosis of the machinery health status. In this paper, it is shown how these problems are gradually addressed and what possible solutions are chosen. Also, the recent advances in the development of an automated wear particle classification system for the application in machine condition monitoring are presented. (c) 2006 Elsevier Ltd. All rights reserved.