Research On Fault Diagnosis Method Based On MED-VMD And Optimized Support Vector Machine

With the progress of science and technology,the structure and function of mechanical equipment tend to be intelligent,the moving parts of mechanical equipment are increasing day by day,and the mutual coupling and nonlinear coupling between different parts become more and more closely,once a component failure,even a minor fault will affect the operation of the mechanical components,which will have a huge impact on the safe operation of the mechanical system.Therefore,the study of mechanical fault diagnosis technology is of great significance.The mechanical fault diagnosis mainly includes three aspects: acquisition and processing of signal,fault feature extraction and pattern recognition.Based on these three aspects,this paper studies the method of fault diagnosis based on minimum entropy deconvolution and variational mode decomposition and optimized support vector machine(SVM).Firstly,because the empirical mode decomposition method and local mean decomposition method has end effects and mode mixing problem,the variational mode decomposition is introduced into fault feature extraction,VMD method is a kind of non recursive signal decomposition method,the end effects and the mode mixing are much less than EMD and LMD,which can extract the feature information of the signal more effectively.Secondly,aiming at the problem that the fault characteristic information of bearing vibration signal is not obvious by noise interference,a method of feature extraction based on minimum entropy deconvolution and variational mode decomposition is proposed.Minimum entropy deconvolution can reduce the noise interference and enhance signal fault feature information,the fuzzy approximate entropy to quantify the components of VMD,the construction of feature vector.The effectiveness of the feature extraction method is verified by real data analysis.Additionally,according to the traditional parameter optimization algorithm of SVM has low efficiency,the extended particle swarm algorithm is proposed to optimize parameters of SVM,compared with the traditional parameter optimization algorithm,it has higher efficiency and better search ability,extended particle swarm algorithm compared to the standard particle swarm algorithm also has better optimization ability,the fault feature vector for extraction to the analysis,the algorithm can correctly identify the fault.Finally,taking Case Western Reserve University of rolling bearing fault data as the experiment subjects,analyzing from the two aspects of bearing with different degree of damage and different fault locations,the experimental results show that there are great differences between feature vectors of different fault types,verify the effectiveness of this feature extraction method.The SVM based on the extended particle swarm optimization algorithm can correctly identify the different fault types,so the proposed method can achieve a good diagnosis effect.