威布尔分布, 模糊C均值, 特征提取, 故障诊断

A novel approach to extract fault features and fault diagnosis of rolling bearings based on Weibull distribution model and FCM clustering was proposed. Because the Weibull distribution model parameters such as its scale parameters, shape parameters and its similar function of negative logarithm were more different for different fault types, the running state features of the bearings could be characterized better. The feature vectors representing the rolling bearings running state could be constructed by extracting the three parameters including its scale parameters, shape parameters and its similar function of negative logarithm. FCM clustering was realized by determining the distance degrees of the samples according to the samples membership relative to the clustering center and was more effective. In the test process, the vibration signals of rolling bearings were processed firstly by morphological filtering, and a Weibull distribution model for the denoised signals was set up. Then the extracted feature vectors were transmitted to the classifier of FCM clustering for fault diagnosis and recognition. The results show that the proposed method is more accurate and is an important effective means for fault diagnosis.