Towards the automated classification system of worn surfaces

Wear dictates the machine performance, its reliability, and ultimately its lifetime. It is therefore important to know what is the wear status of the machine. The assessment can be done visually which is costly and not always possible (e.g. engines). Therefore, the development of an automated analysis and classification system of worn surfaces is highly desirable. Before such system is developed, surface characterization and classification methods must be first developed and tested. In this paper, we show how the recently developed methods, i.e. directional blanket covering and directional blanket covering curvature, have been adapted to solve this problem. The methods, unlike current standards/methods, measure surface roughness and curvature complexities at individual scales and directions. The methods' ability to differentiate between microscopy images of abrasive and adhesive wear surfaces was evaluated. Minute differences between the two wear modes were detected. This could be of importance in machine condition diagnostic since abrasive and adhesive wear account for most of the machine failures.