Research On Fault Signal Analysis And Diagnosis Technology For Rotating Machinery

As the critical components,steam turbine generator rotor system and bearings are widely used in the rotating machineries.Their running states directly affect the performance of the mechanical equipment,even influence on the safety of the whole production lines.In order to ensure the smooth running of the equipments and reduce the occurrence of the accidents,it is of great significance to deeply carry out the research on the fault diagnosis technology for steam turbine generator rotor system and bearings.However,most of the vibration signal of rotating machinery is nonlinear,non-stationary and time-varying signals,so that the state characteristics information of mechanical equipment will not be reflected directly.Therefore,it is always a hot topic of the research on how to apply the correct signal analysis method to extract the feature information of fault signal.Many methods of time-frequency analysis are proposed by researchers.However,the time-frequency analysis methods such as Short Time Fourier Transform(STFT),Wavelet Transform(WT),Wigner-Ville Distribution(WVD)have their own shortcomings,so it is urgent to study new fault diagnosis technology for rotating machinery.In this paper,the theories of the Empirical Mode Decomposition(EMD),Local Mean Decomposition(LMD),Intrinsic Time-scale Decomposition(ITD),Variable Mode Decomposition(VMD)method and its applications in rotating machinery fault diagnosis is studied deeply.The main research work and achievements are as follows:(1)Parameter optimization method of Support Vector Machine(SVM)for Modified Fruit Fly Optimization Algorithm(MFFOA),and the new MFFOA-SVM model is built.Firstly,the theory of support vector machine and fruit fly optimization algorithm are discussed in detail,and the intelligent fruit fly optimization algorithm is Modified.The local search ability of the fruit fly optimization algorithm is gradually enhanced,to improve the early foraging probability of the global optimal solution,to avoid falling into local optimum,to achieve maximum search accuracy in foraging end,and to achieve the balance of global search ability and local search ability.Then,the modified fruit fly optimization algorithm is used to optimize the support vector machine parameters,and a new MFFOA-SVM model is built,which can solve the defects of the traditional parameter selection method and improve the performance of the learning machine by the algorithm of parameter optimization.(2)For modal aliasing problems of the EMD decomposition,the EEMD decomposition algorithm is introduced to eliminate the influence of the modal aliasing.By analyzing the simulation signal,the experiment results show that the EEMD decomposition algorithm can effectively suppress the occurrence of the modal aliasing phenomenon,and improve the accuracy of the EMD algorithm.A bearing fault diagnosis method is proposed.Based on Ensemble Empirical Mode Decomposition(EEMD),Modified Fruit Fly Optimization Algorithm(MFFOA)and Support Vector Machine(SVM).EEMD is used to decompose the fault signals,and to calculate the root mean square value and frequency of the center of gravity,achieving the normalization processing feature vector.In order to improve the classification accuracy rate,a new MFFOA-SVM model is built,and then the feature values are extracted for training and testing,and compared with EMD and MFFOA-SVM diagnosis prediction results,the experiment results show that this method can improve the accuracy of fault feature extraction and can be effectively applied to the fault diagnosis of bearing.(3)A bearing fault diagnosis method is proposed based on LMD ? slice bispectrum and SVM.LMD is used to decompose the fault signals,and to select the product function components with higher kurtosis value to be reconstructed,and then makes slice bispectrum to further reduce the effect of gaussian nosie,and a series of feature vectors are extracted from the 1X,2X and 3X in the bispectrum of the slice.Support vector machine can be used to train and test that eigenvector,and LMD method was proposed to diagnose the bearing with outer-race,inner-race or elements faults.The results indicate that the characteristic frequencies can be extracted effectively using LMD method and be used to make correct judge of the fault type.(4)A rotor fault diagnosis method is proposed based on Modified Intrinsic Time-scale Decomposition(MITD)and MFFOA-SVM,this method has adaptive signal processing.MITD is used to decompose the fault signals,to be demodulated by teager energy operator,and select feature vector of the ISC component.In order to improve the classification accuracy rate,a new MFFOA-SVM model is built,and then the feature values are extracted for training and testing.The experimental results show that this method is effective to extract the rotor fault characteristics,and to implemented rotor fault diagnosis.(5)A rotor fault diagnosis method is proposed based on VMD and MFFOA-SVM.VMD is used to decompose the fault signals,and to calculate the energy,skewness and kurtosis,achieving the normalization processing feature vector.In order to improve the classification accuracy rate,a new MFFOA-SVM model is built,and then the feature values are extracted for training and testing.The experimental results show that,comparing with the EMD,ITD diagnosis methods,the proposed method is able to effectively identify the rotor faults and has obvious advantages.