Multiscale Entropy Theory Based Fault Diagnosis Of Rolling Bearing

Rolling bearings are the most common and vulnerable rotating parts,widely applied in mechanical equipment,and the performance of entire mechanical equipment are obviously affected by its working conditions.Thus,in order to avoid the accidental failure or fault,the real-time monitor on the running status of rolling bearings is essential to increase the working accuracy of mechanical equipment and reduce or avoid the unexpected faults.Meanwhile,the both the working efficiency and the safety of equipment are respectively improved and ensured.Among them,the feature extraction and state identification are particularly emphasized.Multiscale Entropy(MSE),an effective method to measure the complexity of time series,is widely applied to directly extract the model information contained in the original signal.In this,the MSE method is commonly exploited in extracting the fault feature of rolling bearings.As for both the shortcomings of the coarse granulation and the sample entropy,some improved MSE algorithms are proposed here,and then,combined with the classifiers,its validation is to verified with the experimental data of rolling bearing.The detailed contents are described as follows:1.The further study on multiscale entropy theory1)When the scale factor increases,the initial multiscale procedure actually results in the shortening on time series,the fluctuation of the entropy value.As for this problem,the sliding mean theory was correspondingly applied to the coarse graining for the first time,and the fuzzy entropy was taken as the substitution for the sample entropy.The improved multiscale fuzzy entropy(IMFE)was proposed here.The simulation and experimental results validate the stability of entropy value obtained with the method proposed here.2)As for the coarsening insufficiency in the MSE algorithm,combining with the inherent characteristics of the vibration signal,interpolated Multiscale Entropy(InMSE)Algorithm,with the cubic spline function utilized to the coarse graining,was represented.The analytical and experimental results show that the information hidden in signal is sufficiently excavated.3)In order to further excavate the hidden pattern information in the time series,the coarse graining of interpolating multiscale entropy was firstly combined with the initial processing,and meanwhile the composite theory was also exploited,finally,Composite Interpolation Multiscale Fuzzy Entropy(CIMFE),with the fuzzy entropyalgorithm substituting the sample entropy,was proposed here.The validity and superiority of the novel method are,particularly,verified.2.the application of the multiscale entropy algorithm,combining with the learning method,on the fault diagnosis of rolling bearing1)A new fault diagnosis method based on composite multiscale entropy and Laplacian support vector machine was proposed.The simulated results,firstly verify the superiority of the proposed method.Eventually,the experimental analysis suggests the better performance of the method combining with Laplacian support vector machine on accurate identification of faults than support vector machine.2)Aiming at difficultly marking the sample,a new fault diagnosis method of rolling bearing based on fuzzy C-means and interpolated multiscale entropy was proposed.Firstly,compared with the MSE algorithm,the interpolation of multiscale entropy has little dependence on the parameter selection.Secondly,the experimental analytical results suggest the identification superiority of interpolation of multiscale entropy to MSE,combined with FCM.3)Aiming at the small number of fault signal samples in large mechanical equipment,a new fault diagnosis method based on improved multiscale fuzzy entropy and support vector machine was proposed,which is not only suitable for the identification of the same kind of fault,but also improves the fault recognition rate of different fault levels under the smaller sample;4)The multiscale fuzzy entropy of the composite interpolation has less demands on parameters.Based on this,a hybrid fault diagnosis method,based on compound interpolation multiscale fuzzy entropy and Laplacian support vector machine,was proposed.By comparing the multi-scale entropy,multiscale fuzzy entropy,compound multiscale entropy and compound multiscale fuzzy entropy of the simulation signal,the experimental and analytical results suggest that the recognition rate of the proposed method is higher than that of the above algorithm.And then,the Laplacian support vector machine was compared with support vector machine,the results indicate that the recognition rate of the classification method based on Laplacian support vector machine is higher than that of support vector machine.The verification of the above method is based on the simulation signal and test data,which provides a further study on fault diagnosis of rolling bearing.