Early Fault Diagnosis Of Rolling Bearings Based On Enhanced Singular Spectrum Decomposition

Rolling bearing is called "industrial joint",which is widely used in many industrial fields such as aerospace,wind power generation,processing and manufacturing,and its operation status directly affects the safety and service performance of mechanical equipment.In recent years,many experts and scholars have been paying attention to the research of rolling bearing fault diagnosis.Based on this,this paper takes the rolling bearing as the main research object,analyzes the structure of the rolling bearing,the common typical failure forms and causes,and the vibration principle,and then takes the enhanced singular spectrum decomposition as the main research method to solve the problem that the early fault characteristics of the rolling bearing are weak and difficult to find,and the fault characteristics of the roller bearing are seriously disturbed by noise.Based on the enhanced singular spectrum decomposition(ESSD),combined with Gini coefficient and multi-scale mean permutation entropy,the fault diagnosis of rolling bearings is studied.The main contents of this paper are as follows:(1)Early fault detection of rolling bearing based on Gini coefficientTo solve the problem that early fault characteristics of rolling bearings are weak and difficult to find and issue early warning,a method of rolling bearing early fault monitoring based on gini index(GI)is proposed.Firstly,we find the gini index(GI)of the vibration signal of the rolling bearing at different times.Then,the L1 trend-filtering treatment of the gini coefficient curves with large trends but with small fluctuations was performed to obtain a smooth trend line.Finally,whether the trend line crosses the valve line is taken as the standard to judge the healthy running state of rolling bearings,so as to realize the early fault warning of rolling bearing.The effectiveness and generalizability of the proposed method were validated by the analysis of datasets and contrasted by other methods.(2)Early fault feature extraction of rolling bearings based on ESSD-MCKDTo solve the problem that it is difficult to effectively extract weak fault features from rolling bearing vibration signals under strong noise background,an enhanced singular spectrum decomposition(ESSD)method combining maximum correlated kurtosis deconvolution(MCKD)is proposed for rolling bearing early fault feature extraction.based on the singular spectrum decomposition algorithm.the signal decomposition accuracy of the singular spectrum decomposition algorithm and the ability to suppress false components are improved by setting additional termination conditions..First,the early fault signal of rolling bearing is decomposed by ESSD.Then,the optimal component is selected from the value of the correlation coefficient and the value of the kurtosis.Finally,the optimal component is de-noised by maximum correlation kurtosis deconvolution optimized by the honey badger algorithm,and the early fault characteristic frequency is extracted by envelope demodulation analysis of the de-noised signal.The validity of the proposed method is verified by analysis of data sets and comparison with other methods.(3)Rolling Bearing Fault Identification Based on ESSD and HBA-SVMAiming at the problem of difficulty in extracting fault features and low accuracy in rolling bearing fault diagnosis,a pattern recognition method based on enhanced singular spectrum decomposition algorithm and honey badger algorithm optimized support vector machine is proposed.To deal with single type fault,a pattern recognition method based on the combination of enhanced singular spectrum decomposition,maximum correlation kurtosis deconvolution and honey badger algorithm optimized support vector machine is adopted.Firstly,the rolling bearing vibration signal is decomposed by enhanced singular spectrum,and then the correlation coefficient is used to reduce the signal redundancy,and the component with the largest correlation coefficient is selected as the optimal component.Finally,the maximum correlation kurtosis deconvolution optimized by the honey badger algorithm is used to denoise the optimal component,and the optimal component is imported into the support vector machine optimized by the honey badger algorithm to identify the fault type.In order to deal with multiple types of faults,a pattern recognition method based on the combination of enhanced singular spectrum decomposition,multi-scale scale mean permutation entropy(MMPE)and honey badger algorithm optimized support vector machine is proposed.Firstly,the rolling bearing vibration signal is decomposed by enhanced singular spectrum,and then the correlation coefficient is used to reduce the signal redundancy,and the component with the largest correlation coefficient is selected as the optimal component.Finally,mmpe is used to analyze the optimal component to construct the feature data set,which is imported into the support vector machine optimized by the honey badger algorithm to identify the fault type.Through the analysis of fault data sets and engineering cases,the effectiveness of the proposed method is verified,and the fault identification accuracy is high.