航空发动机, 磨损监测, 滑油滤分析, 图像分析, 野点检测, 遗传算法

An automatic identification technology based on engine oil filter debris image was investigated. Firstly, characteristic quantities of oil filter images which can reflect wear state were extracted with the maximum entropy method and mathematical morphology method. Then, oil filter images that can reflect normal state were collected, and the normal training sample sets were constructed through image analysis and feature extraction. Finally, the normal domain of oil filter image was obtained through normal training samples using novelty detection, and to identify the severity of the wear state of aero-engine according to the normal domain. Besides, the self-adaptive parameter of novelty detection was obtained by genetic algorithm. Engine oil filter monitoring system (EOFMS) was developed, and the recognition function of wear debris images based on novelty detection was realized. Besides, the experiments using actual aero-engine oil filter debris images were carried out, and the results show the effectiveness of the method.