搭建油液在线监测实验平台进行磨粒分类识别实验,运用支持向量机和最近邻法相结合的方法对飞机发动机油液中的磨粒进行分类识别;其中基于支持向量机的磨粒分类器的输入为磨粒的主轴长度、纹理相关性、圆度等特征参数,输出为磨粒的分类结果;实验结果表明,基于支持向量机的磨粒分类器的分类准确率高达94%,并且由于最近邻法的使用,分类器的处理速度也提高了30%。
An online oil monitoring system is constructed to conduct the wear debris classification experiment, Support Vector Machine(SVM) combined with the Nearest Neighbor arithmetic was used to classify the wear particles in the aircraft engine's oil ; The inputs of the SVM classifier are the character of the particle: length of the principlc axis, texture relativity, roundness and so on. The output is the result of the classification ; Results of the experiment show that the classification accuracy rate is 94%, and the process speed of the classifier is increased by 30% as using Nearest Neighbor arithmetic.