Research On Application Of Fuzzy Clustering In Rolling Bearing Fault Identification

Rolling bearings have a wide range of applications.As an important part of rotating machinery and equipment,they are also one of the most vulnerable parts.Its health will directly determine whether the entire machinery and equipment can operate normally.Therefore,in the field of mechanical fault monitoring and diagnosis,it is particularly important to improve the effectiveness of rolling bearing fault monitoring.Especially in the rapidly developing modern machinery and equipment,due to the complexity,automation and high speed of the equipment structure,the coordination between mechanical parts is closer,which also increases the possibility of rolling bearing failure.The traditional fault diagnosis method is to use the existing empirical knowledge to perform fault diagnosis on the detected fault signal,but this kind of method has great limitations.Therefore,this paper proposes a fault diagnosis algorithm based on fuzzy clustering analysis to classify rolling bearing faults without prior conditions.In this paper,an improved Markov distance fuzzy clustering algorithm WKSM-FCM(Wavelet K-Smote SM-FCM)based on wavelet transform and unbalanced data processing is proposed.Firstly,the original fault data is denoised by wavelet analysis,and then the fault feature vector is extracted by wavelet packet analysis.As the rolling bearing fault is a small sample classification problem with similar sample characteristics,it is easy to be mistakenly classified into a category.In order to improve the accuracy of fault diagnosis,the data set should be divided effectively through imbalance processing,improve the identification accuracy of the fault signal.Therefore,an improved SM-FCM fuzzy clustering algorithm based on Markov distance is adopted to classify the faults,which can adaptively adjust the geometric distribution of data and solve the problem caused by the inability of Euclidean distance to distinguish the attribute differences between different categories in traditional fuzzy clustering,thus reducing the error rate.The analysis and comparison of simulated signals and measured signals prove the feasibility of this method.Taking the failure data provided by the Fault Simulation Test Bench of Case Western Reserve University as the research object,different types of fault data were collected.Using MATLAB simulation and experimental data for real-time monitoring and analysis,realization of fuzzy clustering in Rolling Bearing Fault Recognition.