基于核的动态聚类算法用于机械故障模式分类

Application of kernel to classification of mechanical faults based on dynamic cluster

针对k均值聚类对特征样本划分存在误分类的问题 ,提出用基于核的动态聚类算法对风机不同工作状态进行分类识别 .实验结果表明 ,该方法能有效地识别机器运行的异常状态 ,并能对不同的故障模式进行正确的区分 ,可应用于机械设备运行状态的动态识别 .

Because of false classification of feature samples by k -means clustering analysis, a kernel based dynamic clustering method was presented to distinguish different working modes of an air compressor. Experiment results show that the approach is effective in recognizing abnormal during machine working, and it can fulfill fault classification and can be used in monitoring machine working condition.