Study On Rolling Bearing Performance Degradation Assessment And Fault Diagnosis Method Based On Preferred Features

The health status of rolling bearings directly affects the operation of rotating machinery equipment.Therefore,it is necessary to carry out performance degradation assessment research on rolling bearings,construct high-quality degradation indicators,detect early faults and diagnose them in time.This paper carried out the following research work :(1)Aiming at the problem of redundant information in the feature set,a preferred feature construction method based on correlation between features and performance evaluation is proposed.Firstly,the initial feature set containing 66 features is extracted.Secondly,the initial feature set is divided into multiple feature clusters by using the modified hierarchical clustering algorithm with the absolute value of Pearson correlation coefficient as the distance calculation scheme.The performance of each feature is evaluated by three feature evaluation indexes of monotonicity,correlation and robustness,and the optimal features in each cluster are selected to form the optimal feature subset.Through the verification of the whole life data set of rolling bearings,the correlation within the selected feature subset is significantly reduced.(2)Aiming at the problem that the use of Shewhart control chart(6σ method)is limited by data distribution and the alarm threshold setting is affected by interference such as abnormal impact in the normal operation stage of rolling bearing,an early fault detection method of rolling bearing based on modified EWMA control chart is proposed.Firstly,the preferred feature subset is input into the autoencoder to construct the degradation index,which is decomposed by the moving average method to separate the random interference.Secondly,the modified genetic algorithm is used to optimize the relevant variables in the parameter estimation method,and the appropriate value is selected to accurately estimate the mean and standard deviation of the EWMA control chart.Finally,the upper and lower limits of EWMA control chart are determined to detect the early fault of rolling bearing.The modified genetic algorithm is to improve the probability of offspring inheriting excellent genes by selecting some excellent individuals from the population to mate alone,and to accelerate the convergence speed of the algorithm.Through experimental data,it is verified that the degradation index can accurately reflect the performance degradation process of rolling bearings,and shows sensitivity to early faults.The effectiveness of the proposed early fault detection method,compared with envelope spectrum analysis,isolated forest and other methods,can detect early faults earlier,and the false alarm rate is lower,which proves that the method has certain advantages.(3)Aiming at the difficulty of fault feature extraction and intelligent diagnosis of rolling bearings,a fault diagnosis method of rolling bearings based on preferred features is proposed.Firstly,the original vibration signal is decomposed by CEEMD,and five IMF components are selected according to the correlation principle,and the initial feature set is extracted respectively.Secondly,the preferred feature subset is formed by the preferred feature method combining PHC algorithm and feature evaluation index.Finally,the XGBoost model is input,and the grid method is used to optimize the model parameters for fault diagnosis.Experimental results show the effectiveness and feasibility of the proposed method.