正交小波变换k-中心点聚类算法在故障诊断中的应用

k-中心点聚类算法(k-medoids cluster algorithm,KCA)是改进的机器学习聚类算法,该方法通过初始聚类中心选取和聚类中心更新,对无标记训练样本的学习揭示数据的内在性质及规律,从而区分出机器的运行状态。提出了一种正交小波变换k-中心点聚类算法(orthogonal wavelet transform k-medoids clustering algorithm,OWTKCA)诊断方法,利用正交小波变换(orthogonal wavelet transformation,OWT)方法提取各细节信号作为训练样本,用KCA方法进行分类。通过滚动轴承的试验数据分类结果显示,该方法相对于没有提取特征值的KCA能有效处理复杂机械振动信号,明显提高了故障数据聚类效果,缩短了聚类时间,提高了智能诊断效率。

Orthogonal wavelet transform KCA in fault diagnosis

k-medoids cluster algorithm(KCA)is an improved machine learning clustering algorithm.This method reveals the inherent properties and laws of the data,through selecting the initial clustering center,updating the clustering center and learning the unmarked training samples,so as to distinguish the running state of the machine.In this paper,an orthogonal wavelet transform k-medoids clustering algorithm(OWTKCA)was proposed for diagnosis,which uses the orthogonal wavelet transform(OWT)method to extract the detailed signals as training samples,and uses the KCA method to classify them.The results of test data classification of rolling bearing show that this method can deal with complex mechanical vibration signals more effectively than KCA without extracting characteristic values,it obviously improves the clustering effect of fault data,shortens the clustering time and improves the efficiency of intelligent diagnosis.