CHARACTERIZATION OF WEAR PARTICLES AND THEIR RELATIONS WITH SLIDING CONDITIONS

Wear particles and worn surfaces generated in pin-on-disk steel slidin g experiments are studied by microscope image analysis and two types o f neural networks. Features of wear particles described by four partic le descriptors depend strongly on sliding conditions. A multilayer neu ral network successfully learns the relations between wear particle fe atures and sliding conditions. If the network is trained with data rep resenting typical features, it also recognizes the particles having si milar features. This suggests that the network can be used as a tool f or condition monitoring in which the network identifies wear particles produced under unknown sliding conditions to predict that conditions. A self-organizing neural network using the competitive learning rule classifies the wear particles based on their features without any supe rvisor data. Particle features are expressed by the position on a two- dimensional feature map. This type of network is useful in finding typ ical particle features, which in turn can be used as supervisor data f or the multi-layer neural network. In another application of the self- organizing network, microscopic images of both wear particles and worn surfaces are automatically classified, and characteristics of each su rface are represented by the distributions of weights on the feature m ap. (C) 1998 Elsevier Science S.A. All rights reserved.