Wear state identification using dynamic features of wear debris for on-line purpose

Abstract

Wear status identification including wear rate estimation and wear mechanism assessment can be performed using wear debris information. However, although on-line monitoring methods have distinctive advantages over off-line approaches, existing on-line monitoring methods provide limited features of wear particles and have difficulties characterising complex wear states. Most of them determine wear status based on changes in the wear rates, and the wear mechanisms are not taken into consideration. Therefore, comprehensive wear state identification is a bottleneck in real-time machine health monitoring for condition-based maintenance. In order to further advance on-line monitoring technology, this paper, in a case study format, presents a new approach for wear state characterisation using comprehensive wear debris features. For this purpose, wear experiments were carried out on a four-ball rig, and a particle imaging system was employed to capture videos of moving particles to acquire dynamic features. Based on this, wear particles were firstly counted to characterise wear rate. In this stage, a statistical clustering model was established using a mean-shift algorithm to categorise wear debris samples. A trend of wear state evolution was thus obtained. Secondly, the size, shape and colour of wear debris were extracted to identify particles into fatigue, sliding and oxides for wear mechanism analysis. The analysis results of wear mechanisms were related to the trend of the wear state. Correspondingly, a changing chart that contains the wear degree and wear mechanisms was drawn. Therefore, an on-line system has been developed to capture comprehensive particle information to assess the wear severity and mechanisms for in-depth wear analysis and full-life machine condition monitoring.