Research On The Characteristic Analysis And Automatic Identification Technology For The Wear Debris

As one of the most important methods for mechanical fault diagnosis and condition monitoring, Ferrography technique has only been operated and analyzed by the experts or workers who have rich experience in ferrography analysis since this technique came out in 1970's. The development and spread of ferrograhpy has been restricted because of its lower automation. Therefore, the automatic analysis of ferrography has been always the highlight in academic researches.Based on the analysis of the former theoretical research work, the digital image processing was employed in Ferrography analysis in this thesis.. The methods of image filtering, image thresholding segmentation, image binarization were used for the pretreatment of the ferrographic images, on the basis of preparation of ferrogram slide, acquisition of ferrographic image and the digital image processing techniques. Based on the analysis of the forming mechanism of the debris and debris feature parameters, the parametric mathematics was suggested to describe the debris feature parameters. In addition, the characterization parameter of debris surface texture was also studied. How to obtain the characteristic parameters of debris region area, eqvivalent circle area, debris region circumference, circularity, aspect ratio, direction index was investigated. Based on the above the basis studying, the serial application of three layers Bp neural networks in debris pattern recognition was used, and a method which ascertains the training times of Bp neural networks was also suggested. Finally, a WDAIS (Wear Debris Analysis and Identity System) was developed with VC++6.0 integrated exploitation environment. And by the examination of the actual input debris, the WDAIS was proved to be able to distinguish several kinds of representative debris and was very practical.