Since the early 1970s wear particles have been used as indicators of the health status of industrial machinery. Their quantity, size and morphology was utilized in machine condition monitoring to diagnose and predict the likelihood or the cause of machine failure. In particular, the wear particle morphology was found useful as it contains the vast wealth of information about the wear processes involved in particle formation, and the wear severity. However, the application of wear particle morphology analysis in machine condition monitoring has limitations. This is due to the fact that the process largely depends on the experience of the technicians conducting the analysis. Research efforts are therefore directed towards making the whole wear particle analysis process experts-free, i.e. automated. To achieve that a detailed database of wear particle morphologies, generated under different operating conditions and with different materials for sliding pairs, must be assembled. Next, the reliable and accurate methods allowing for the description of 3-D wear particle morphology must be found. Multiscale and nonstationary characteristics of wear particle surface topographies must be accounted for. Finally, a reliable wear particle classification system must be developed. This classification system must be reliable and robust hence the selection of appropriate classifiers becomes a critical issue. It is hoped that the system, once fully developed, would eliminate the need for experts in wear particle analysis and make the whole analysis process less time consuming, cheaper and more reliable. In this presentation it is shown how the problems leading towards the development of such system are gradually overcome. Also, the recent advances towards the development of expert-free wear particle morphology system for the application in machine condition monitoring are presented.
The Segmentation of Wear Particles Images UsingJ-Segmentation Algorithm
