An intelligent system for machinery wear debris analysis using evolutionary algorithms

Wear debris analysis is becoming an efficient method for machinery condition monitoring due to recent developments in image analysis techniques. It provides information about not only the wear mode but also the wear mechanism of a machine component. Five types of debris are produced during the operation of a machine: sphere, platelet, longthin, cutting and chunky. A variety of parameters that relate to the identification process of wear debris can affect the performance of image analysis. This paper presents five numerical features to describe the boundary morphology of debris. A ratio-based methodology using genetic algorithms is used for classification. The experimental results indicate that, due to the simplicity of the proposed features, the classification of debris can be carried out quite rapidly and accurately.