Research On Rolling Bearing Fault Diagnosis Method Based On Manifold Learning And EEMD

When the rolling bearing fails,the vibration signal is relatively complex,and the traditional power spectrum analysis and time-frequency analysis methods often cannot effectively diagnose the type of failure.In response to the above problems,based on the techniques of ensemble empirical mode decomposition(EEMD),bispectrum analysis and manifold learning,this paper carried out the study of feature extraction and troubleshooting methods and applied them to the fault diagnosis of rolling bearings.The main research contents of this article are as follows:(1)Aiming at the problem that the vibration signal of rotating machinery is often mixed with Gaussian noise and non-Gaussian noise,which cannot be dealt with by traditional methods,a rolling bearing feature extraction method based on cloud model improved EEMD and bispectrum analysis is proposed.First,the bearing signal is EEMD decomposed to obtain a series of intrinsic modal components(IMF)from high to low frequency;secondly,the proposed cloud model parameter method is used to identify the false IMF component and eliminate non-Gaussian noise in the signal;Bispectrum analysis is performed on the filtered real IMF components to remove Gaussian noise,thereby extracting the fault feature information submerged in the noise.Simulation analysis and practical application of rolling bearings verify the effectiveness and superiority of this method.(2)Aiming at the fact that a single-scale entropy value is difficult to fully express the characteristics of rolling bearing failure,a rolling bearing fault diagnosis method is proposed based on the EEMD,dispersion entropy(DE)and local reservation projection(LPP).The method adopts dispersion entropy as the characteristic parameter,the first nine IMFs of the signal are extracted with EEMD and used as the characteristic signals;their dispersion entropies are calculated and used as the characteristic vectors and the sample set is obtained.Then,LPP is used to extract the initially obtained feature set,and finally the K-Neighbor Classification Algorithm(KNN)is used to identify the characteristic extraction method.The results show that the method has higher classification accuracy by using the rolling bearing fault sample for test verification and compared with several other methods.(3)Aiming at the problem of poor real-time performance of EEMD distributed entropy-LPP rolling bearing fault diagnosis method,a rolling bearing fault diagnosis method based on multi-scale distributed entropy-LPP is proposed.This method replaces the EEMD algorithm with a multi-scale spread entropy that takes less time,and directly constructs a high-dimensional feature space.First,calculate the distribution entropy of the bearing signal at 15 scales to construct a high-dimensional feature space,then use LPP to reduce the dimension to obtain a low-dimensional feature set,realize the second extraction of features,and finally input a KNN classifier for classification recognition.The rolling bearing fault diagnosis example proves that it has higher efficiency without reducing the recognition accuracy,and is more suitable for real-time diagnosis of ball bearings.(4)Based on the feature extraction and fault diagnosis methods studied in this paper,as well as conventional time-domain analysis and frequency-domain analysis methods,a set of rotating machinery fault monitoring and diagnosis system was designed using GUI platform of MATLAB to provide online monitoring and Fault diagnosis provides an effective analysis tool and verifies the effectiveness and practicability of the system through practical application.Finally,the work of this paper is summarized,and the future research direction is prospected.