Application Of Rolling Bearing Fault Diagnosis Based On Improved Entropy Algorithm

Rolling bearing is a kind of widely used mechanical parts,poor working environment can easily lead to a variety of faults,and then lead to accidents,so it is of great significance to diagnose its faults.The fault feature of rolling bearing vibration signal is weak,so the traditional linear signal processing method is difficult to extract fault information effectively.Such as empirical mode decomposition(EMD)and wavelet transform,all of which have the defects of terminal effect and mode aliasing.In recent years,many nonlinear parameters have been proposed,such as sample entropy(SE),permutation entropy(PE),dispersion entropy(DE)and so on.Multiscale entropy(MSE)reflects the self-similarity and complexity of time series at different scales.SE was calculated over multiple scales so as to formulate a feature vector which is subsequently input into PSOSVM for fault type and severity level identification.MSE is adopted to extract the features of bearing,and then input these fault features to PSO-SVM for pattern recognition.The experimental results show that the feature extraction effectiveness of MSE is better than that of multiscale approximate entropy(MAE)and wavelet packet decomposition.The fault recognition effect of PSO-SVM is better than that of SVM and Grid-SVM.However,SE is greatly affected by abrupt signal when dealing with short time series,while PE can detect the dynamic mutation behavior.Multiscale permutation entropy(MPE)has strong anti-noise ability and high computing power.Composite multiscale weighted permutation entropy(CMWPE),based on the idea of composite coarse graining and weighting,overcomes the shortcomings of MPE and can well distinguish different modes of signals.The results show that the feature extraction effectiveness of CMWPE is better than that of MPE and MSE,and the fault recognition accuracy combined with PSO-SVM is improved by 1.08% and 0.5% respectively.However,PE fails to take into account the effect of signal amplitude difference and is sensitive to noise.Compared with MSE and MPE,MFDE has higher calculation efficiency and better feature extraction performance.However,the intrinsic deficiencies of coarse graining process lead to unstable results and imprecise.To solve these problems,this paper propose an improved multiscale fluctuation dispersion entropy—generalized refined composite multiscale fluctuation dispersion entropy(GRCMFDE).The experimental results show that the feature extraction effectiveness of GRCMFDE is better than that of MFDE,MWPE and MSE,and the fault recognition accuracy combined with PSO-SVM is improved by 0.54%,0.6% and 0.73%respectively.In this paper,the defects of the existing entropy algorithm are improved.The improved entropy algorithm is used to extract the signal characteristics of rolling bearing and artificial intelligence technology is used for fault identification.The experimental analysis shows that the improved entropy algorithm combined with PSO-SVM fault diagnosis model has high accuracy in identifying different fault types and fault degrees of bearings,provides an excellent method for improving the detection accuracy of bearing faults,and has certain engineering application value.