Research On Feature Extraction And Fault Classification For Bearing Intelligent Diagnosis

With the rapid development of China’s industrial application field,the practical application of mechanical equipment is also developing toward industrialization and intelligence.As one of the important parts of rotating machinery,rolling bearing damage has a great impact on equipment.How to obtain the fault characteristic information of rolling bearing effectively and determine the fault type in time has become a hot research topic in the fields of fault diagnosis and health monitoring.Therefore,the research on rolling bearing fault diagnosis is mainly carried out from two aspects: fault feature extraction method and intelligent diagnosis model.For fault signal spectrum analysis of the combination of VMD and MOMEDA fault signal processing method,by VMD decomposition after strong correlation and retain more fault frequency components of the IMF component MOMEDA filtering method,and then to the Hilbert envelope demodulation signal after filtering,get the frequency distribution,then based on the theory of bearing parts fault characteristic frequency compared to the actual fault characteristic frequency of the tested by envelope demodulation and to assess the fault type of bearing.The bearing experimental data were used to verify the results.The results show that the method effectively filters the noise components of the original signal,and the obtained envelope spectrum frequency can clearly distinguish the bearing fault types.In order to solve the problems of mode mixing and end effect existing in EMD and the low calculation rate and incompleteness existing in the improved EEMD algorithm.In this paper,an empirical mode decomposition method based on the complete set of adaptive noise is adopted,and finite times of adaptive white noise is added into the original signal decomposition process,and the integrated multi-scale permutation entropy has certain anti-interference and anti-noise ability when dealing with complex random time series.The CEEMDAN multi-scale permutation entropy was used to extract the fault feature information from the complex vibration signals,and the Fisher ratio,correlation coefficient criterion and kurtosis criterion were combined to select the features that could best represent the fault information components.A GG intelligent recognition model was established for intelligent recognition of different fault types of rolling bearings.The feature vector obtained after feature selection is input into the GG clustering recognition model,and the classification effect of other clustering models is compared,which shows the superiority of the GG clustering model.In view of the fuzzy c-means clustering algorithm is sensitive to the initial clustering center of the problem and is related to the local convergence of GA-KFCM intelligent diagnosis model is put forward,by using the genetic optimized nuclear improved FCM algorithm,the feature extraction of different types of fault sample input classification model,compared with other classification model,the results show that GA-KFCM model fault type average diagnostic accuracy was 98.09%.It shows that this method can improve the classification accuracy,speed up the classification rate of FCM algorithm and process the high-dimensional complex data well.