Research On Fault Diagnosis Of Rotating Machinery Based On Singular Spectrum Decomposition

Rotating machinery is an indispensable functional unit of many industrial devices,its operation status directly affects the quality,efficiency and safety of production.The research on the fault diagnosis technology of rotating machinery is of great significance to improve the reliability and safety of equipment operation.Signal decomposition technology is recognized as the most effective tool for fault diagnosis of rotating machinery due to its excellent characteristics in dealing with non-linear and non-stationary signals.In this paper,singular spectrum decomposition(SSD),a new adaptive signal decomposition method,is studied theoretically and applied to fault diagnosis of rotating machinery.Based on the deep analysis of the characteristics of the SSD algorithm,this paper enriches and perfects its theoretical methods,and explores effective solutions for the problem of fault feature extraction and pattern recognition of key components of rotating machinery.The main research contents and innovations of the thesis are as follows:(1)The decomposition characteristics of SSD and its application to rotor fault feature extraction are studied.In the aspect of decomposition characteristics,the anti modal aliasing performance and double harmonic decomposition ability of SSD are analyzed.The results show that SSD can effectively overcome the modal aliasing caused by "abnormal events",and its dual harmonic decomposition ability is better than the empirical mode decomposition(EMD)method.In the aspect of rotor fault feature extraction,a harmonic fault detection method based on the SSD-HT time-frequency analysis is studied.The results show that SSD can effectively separate the characteristic components of the rotor vibration signal,and the SSD-HT time-frequency spectrum can accurately show the instantaneous non-stationary features of each decomposition component,which provides sufficient basis for judging the fault type of rotor.(2)SSD regards the energy ratio of the residual signal and the original signal as the stop condition of the decomposition iteration.It is impossible to predict the optimal energy ratio threshold to determine the reasonable decomposition scale in the process of fault diagnosis.To solve this problem,an optimal singular spectrum decomposition(OSSD)method is proposed.The correlation coefficient is introduced as the supplementary criterion of the iteration stop condition and the component selection criterion of SSD,which can effectively overcome the over decomposition and under decomposition problems caused by improper setting of energy ratio threshold,reduce the false components and improve the analysis stability.(3)How to overcome the interference of the environment noise and the vibration harmonics,and how to realize the separation of the composite fault features are the difficult problems in the impact fault feature extraction of rotating machinery.Hence,a weak impact fault detection method based on enhanced singular spectrum decomposition(ESSD)is proposed.By integrating the differential and integral operators into the SSD analysis,this method improves the detection ability of SSD for weak impact feature component which is not dominant in the signal and the decoupling ability of composite fault impact signal.(4)The application of SSD in fault signature extraction of rotating machinery under variable speed conditions is explored.The SSD-HT time-frequency analysis method is used to extract the fault characteristics of rotor under variable speed conditions.Furthermore,SSD is combined with speed transform(ST)to realize the extraction of the fault features of rolling bearing under variable speed conditions.The research shows that SSD still has good harmonic fault detection and weak impact fault detection function under the condition of slow speed change.(5)With considering the problem that the vibration analysis method based on single-channel signal is easy to omit the key fault feature information in the fault feature extraction,a complex singular spectrum decomposition(CSSD)method is proposed to extend SSD to the complex domain,and a homologous information fusion fault diagnosis scheme is constructed.The experimental analysis shows that the scheme can comprehensively consider the difference of fault characteristics of two-channel quadrature sampling signals,obtain a more comprehensive basis for fault diagnosis and improve the efficiency of fault diagnosis.(6)A fault pattern recognition method based on hierarchical instantaneous energy density dispersion entropy(HIEDDE)and dynamic time warping(DTW)is proposed to determine the fault type and evaluate the fault degree of rotating machinery.HIEDDE combines fault feature enhancement and information evaluation at the same time,which can effectively characterize the characteristics of vibration signals in different states.Using DTW to measure the similarity of feature information can automatically determine the fault mode.Experimental analysis shows that the method can guarantee high analysis accuracy without relying on too many training samples.The research results of this paper provide new ideas for the further research of harmonic fault detection,weak impact fault detection,variable speed time-varying fault feature extraction,homologous information fusion and fault pattern recognition in the process of fault diagnosis of rotating machinery.