Fault Diagnosis And Performance Degradation Assessment Of Rolling Bearing

Machinery and equipment has become more and more important in people’s daily production. It provides a powerful boost for the development of social productivity. However, at the same time, a large of catastrophic accidents caused by mechanical equipment failure and performance degradation ensued. Rolling bearing is one of the most important parts in rotating machinery and vulnerable parts. The research of fault diagnosis and performance evaluation for rolling bearing has become an urgent problem which need to be solved urgently.Therefore, in this paper, the fault diagnosis and performance assessment methods of rolling bearing has been studied. In order to determine the feature selection rule, Correlation and monotonous trend has been studied with the life test data of rolling bearing. Then, with the bearing failure data form Western Reserve University, rolling bearing fault has been diagnosed and fault degree has been identified. At last, three kinds of performance assessment method were used to assess the rolling bearing performance state. The main research content is as follows:(1) The Fuzzy C-Means clustering, the Gaussian Mixture model and Logistic Regression Models are introduced, including the theory of mathematical model and the method of parameter estimate.(2) Based on all the life test data of the rolling bearing, correlation and monotonous trend of conventional characteristic in time domain and frequency domain have been studied, and it is a feature selection rule to solve the question how to select the characteristic to evaluate the performance of the rolling bearing. And in this part, in order to reduce characteristic value dimension, principal component analysis method is introduced.(3) Based on the Gaussian Mixture Model, two different methods, with which to diagnose the fault of rolling bearing and identify the fault degree, have been proposed. Using bearing experiment data from Case Western Reserve University, the accuracy of the two methods is proven.(4) The Gaussian Mixture model, the Fuzzy C-Means clustering and Logistic Regression Models were used to assess the rolling bearing performance state. Using the experimental data of the whole life of rolling bearing, the result of performance evaluation was validated. As the same time, the result of performance evaluation that was made by different methods was compared to give the advantages and disadvantages and scope of application.