Research On Fault Diagnosis And Its Application Based On Time-frequency Analysis Of Non-stationary Signal

The time-frequency analysis method of non-stationary signal is an important research content and direction in the modern signal processing,which has achieved a widespread application.The task of the time-frequency analysis method is describing the variation of the signal frequency spectrum over time,studying and understanding the corresponding relationship of the time-varying spectrum between mathematics and physics,and processing the different signals by constructing an appropriate time-frequency distribution.In this paper,we mainly study the time-frequency analysis method.Then,the manifold learning and Singular Value Decomposition(SVD)methods are applied to extract the fault features.Finally,the Support Vector Machine is applied to classify the fault types.The research works are introduced as follows:The basic theory of the time-frequency analysis method is researched,mainly including the Hilbert-Huang Transform(HHT)and Intrinsic Time-scale Decomposition(ITD)algorithm.For HHT algorithm,we mainly study the instantaneous frequency,intrinsic mode function,Empirical Mode Decomposition(EMD)algorithm and so on.The main existing weaknesses of EMD including mode mixing and end effecting problem are deep analyzed,and the relevant improvement measures are given.For ITD algorithm,the basic principle is studied,and the definition of instantaneous frequency is given.Finally,the advantage of this method is studied.On the basis of Hilbert marginal spectrum analyzing,an intelligent fault diagnosis method based on Hilbert Marginal Spectrum(HMS)-Supervised Locally Linear Embedding(SLLE)and Support Vector Machine(SVM)is proposed in this paper.The HMS is introduced for feature extraction from faulty signals,which is compared with the Fourier spectrum to explain the advantages of HHT method in non-stationary signal time-frequency analysis.Then SLLE is proposed for the dimensionality reduction of high-dimensional fault feature to obtain the fault feature.Finally,the Support Vector Machine is applied to classify the fault type.According to the simulations,the effectiveness of the proposed approach is verified.By studying the non-stationary signal time-frequency analysis method based on ITD,an intelligent fault diagnosis method based on ITD-SVD and Support Vector Machine is proposed in this paper.Firstly,the ITD method is adopted to decompose the signal into several Proper Rotation Components,which is utilized to construct theoriginal feature matrix.Secondly,the singular value decomposition is proposed to obtain the singular value vectors of the feature matrix.Finally,the Support Vector Machine is applied to classify the fault type.The proposed approach can accurately diagnose and identify different fault types under variable conditions.The two fault diagnose methods based on non-stationary signal time-frequency analysis are applied in mechanical vibration signal.The results verify the effectiveness of the fault diagnose methods.According to the compared experiments,the superiority of the proposed approaches is shown.