Research On Rotating Machinery Fault Feature Extraction And Identification Method Based On Improved RLMD

With the continuous development of modern industry,rotating machineries including steam turbine generator sets present a tendency of upsizing and atomization,and their operating condition is gradually becoming more complicated.Hence,the operative mode of rotating machinery exerts an influence on the entire productive process.In order to ensure the normal operation of rotating machinery and personal safety of staff,it is of great significance to study the fault diagnosis technology of rotating machinery.This paper takes common rotating machinery components as the research object,applies the robust local mean decomposition and its improved methods to the fault diagnosis of rotating machinery,and conducts in-depth research from two aspects: fault extraction and pattern recognition.The main contents are as follows:Firstly,a rolling bearing fault diagnosis method combining robust local mean decomposition(RLMD)and frequency-weighted energy operator(FWEO)is described.The PF component with the largest kurtosis value after robust local mean decomposition is selected,and it is demodulated with frequency-weighted energy operator to obtain energy spectrum.The simulation signal of the bearing and the experimental signal of the Case Western Reserve University in the United States are used for verification.Paralleled with other existing methods,the method has certain merits in anti-noise and anti-interference,and its decomposition effect is better.In addition,in view of the problem of modal aliasing when RLMD processes discontinuous signals,the noise-assisted analysis method is used to improve it,and a time-frequency analysis method based on complete ensemble robust local mean decomposition with adaptive noise(CERLMDAN)and Hilbert transform(HT)is proposed and applied to rotor fault diagnosis.Through the analysis of simulation signals and the signals of turbine rotor rubbing and unbalance faults a 600 MW unit,it is proved that the improved method can extract the rotor fault features more accurately than RLMD,which is beneficial to the fault diagnosis of the rotor.In the end,in order to improve the accuracy of steam turbine rotor fault identification,on a basis of CERLMDAN and Hilbert transform,a fault identification method based on local band energy entropy and gray wolf optimization algorithm optimized support vector machine is explored.Firstly,the Hilbert time-frequency diagram is obtained by CERLMDAN and Hilbert transform.Then,the Hilbert time-frequency diagram is employed to divide the frequency bands,and the energy entropy of the local frequency bands in each frequency band is calculated.In this way,fault features can be extracted,and a feature vector can be constructed.Finally,inputting the feature vector into the support vector machine optimized by the gray wolf optimization algorithm serves the purpose of realizing the rotor fault recognition.Diagnosis and analysis of the fault signals of the measured steam turbine rotors show that this method can effectively distinguish four types of state: normal rotor,rubbing,unbalance,and oil whirl.Compared with other methods,the accuracy rate of the method in this paper is higher.