Research On Bearing Fault Diagnosis Method Based On Optimized VMD And PSO-RF

As the heart part of large rotating machinery equipment,the good running condition of rolling bearing is indispensable to ensure the normal operation of large rotating machinery equipment.Aiming at the problems of modal aliasing in signal processing,limitations and subjective factors of traditional variational modal decomposition methods and low classification accuracy,a rolling bearing fault diagnosis method based on optimized parameter variational modal decomposition(VMD)and optimized random forest(RF)was proposed in this paper.Firstly,the seagull algorithm(SOA)was proposed to optimize VMD parameters in order to solve the mode aliasing problem of EMD and EEMD in signal processing.In this method,the minimum envelope entropy value is taken as the fitness function,and the seagull optimization algorithm is used to find the optimal parameter combination [K,A] of the variational mode decomposition algorithm,and then determine the decomposition number k and penalty parameter value A of the variational mode.Then some intrinsic modal components(IMF)are obtained by variational modal decomposition of bearing vibration signals.In order to prove the feasibility of this method,simulation function and bearing data set of Western Reserve University were used for experimental verification.By observing the spectrum diagram,it can be seen that each IMF component is independent of each other.Therefore,the VMD method optimized by gull algorithm can obviously improve the modal aliasing problem in signal decomposition process.Secondly,in view of the subjectivity and limitations of traditional variational modal decomposition in the decomposition process,the method mentioned above is also used.Comparative analysis was carried out by using the method of center frequency to determine the parameters.First,the original signal was VMD processed,and its parameters were set by using the method of determining one and finding two.The value of K was determined by observing the center frequency values of each IMF under different K values,and then the value of A was determined by using this idea.In order to verify the superiority of gull algorithm to VMD parameter optimization method compared with traditional center frequency method,simulation function and bearing data set of Western Reserve University were also used for experimental verification.The results show that this method is more obvious for signal feature extraction.Finally,aiming at the problem that the accuracy of traditional intelligent diagnosis model is not high in the process of classification,on the basis of expatiating the theories of walking entropy,particle swarm algorithm and random forest,an optimization method based on particle swarm algorithm is proposed to select the number of trees and split feature number of random forest.The bearing data set of Western Reserve University was also used for experimental verification,and each walking entropy in IMF component was used as feature vector for training and testing of random forest.Using particle swarm optimization(PSO)algorithm to the two important parameters of random forest tree number and characteristics of the division are selected,a PSO established for rolling bearing fault classification-RF classification model,so as to realize the fault diagnosis of rolling bearing,through the contrast test accuracy of comparative analysis,the results show that the method has the highest classification accuracy,The superiority of the model in diagnosis process is verified.