Research On Performance Degradation Assessment And Trend Prediction Of Rolling Bearing

Rolling bearing is a key component of the machinery device. It has an important impact to the whole equipment whether its working condition is normal. Rolling bearing always goes through a series of degradation states from normal state to failure. The state of bearing performance degradation should be assessed correctly during operation which has an important significance of preventing rolling bearing failure and improving production efficiency. However, proper evaluation and prediction of rolling bearing degradation state need to solve two key issues. First, the appropriate state assessment index that reflecting the process of bearing performance degradation should be determined. Second, appropriate failure prediction model that predicting reasonably and accurately based on historical data should be established. These two questions are investigated in this research and put forward some effective measures.SOM (Self-Organizing Feature Map) is one of the artificial neural network commonly used in fault diagnosis and status assessment.The spectrum characteristics of HHT envelope spectrum of rolling bearings under different working conditions are investigated in this research. A state identification method based on HHT and Self-Organizing Feature Map neural network is presented. The minimum quantization error (MQE) is used as reliable performance degradation assessment indicator which deriving from SOM network. MQE could take advantage of mutual information from multiple features for assessment of rolling bearing performance degradation process. At last, the effectiveness of the method is verified by rolling bearing inner race fault performance degradation test.Due to rolling bearing performance degradation processes affected by many factors, it is difficult to use a single prediction meet the prediction accuracy. So a new hybrid intelligent trend prediction method based on neural networks optimized by Genetic Algorithm (GA) and support vector machine (SVM) is proposed. Firstly, the minimum quantization error is calculated by self-organizing map network model for performance degradation assessment of bearing. Then, it combined with the advantages that BP networks optimized by GA is capable of processing non-linear information and SVM can deal with small samples. Hybrid intelligent prediction model is realized by weighting average of the two trends predicted results. At last, the effectiveness of the method is verified by rolling bearing performance degradation test. Testing result shows that the prediction performance of this model outperforms any one of the two prediction methods and the trend prediction accuracy of bearing performance degradation is improved. Besides, the method is helpful for rolling bearing preventative maintenance and residual life prediction according to the results analysis.