Rolling Bearing Fault Diagnosis Based On Parameter Optimization VMD

Rolling bearing is one of the most widely used mechanical parts in the production equipment.Its running state directly affects the whole production process.When the rolling bearing failure occurs,the light will lead to production equipment failure,production loss,and the heavy will lead to serious consequences such as casualties,so it is of great significance to study the fault diagnosis of rolling bearing.This thesis takes the rolling bearing as the research object,starting from the three parts of fault diagnosis process:signal processing,feature extraction and fault diagnosis.The fault diagnosis technology of rolling bearing has attracted increasing attention in recent years.The main contents of this thesis are as follows:(1)In this thesis,local mean decomposition(LMD)method is compared to verify the advantages of variable mode decomposition(VMD)method in suppressing the end effect and avoiding modal confusion.LMD method belongs to the decomposition method of recursive cyclic screening,and the problem of end effect and mode confusion can not be solved in principle.So VMD method is introduced in this thesis.VMD method transforms the signal decomposition into the optimization problem of the constrained model by introducing the variational model.The simulation results show that VMD can suppress the end effect and avoid modal confusion.(2)Aiming at the problem of irregular parameter selection in VMD method,this thesis studies the VMD method based on genetic mutation particle swarm optimization.The decomposition effect of VMD method is mainly limited by the number of decomposition and the selection of penalty factors.This thesis analyzes the irregularity of the two parameters selection,constructs the fitness function based on the decomposition effect and decomposition efficiency index of VMD method,realizes a VMD method based on genetic mutation particle swarm optimization parameter optimization,and provides the basis for the next step of feature set construction.(3)Aiming at the problem that the fault feature of rolling bearing is weak and difficult to extract,this thesis studies the construction method of rolling bearing feature set based on sample entropy and random forest.When sample entropy is used to measure the complexity of the rolling bearing vibration signal,the entropy obtained is not always related to the complexity of signal.In this thesis,the sample entropy based on the fault mechanism of rolling bearing is proposed,which measures the complexity of vibration signal and is consistent with the result of mechanism analysis.In order to solve the problem that the feature set of single sample entropy is deficient,this thesis adds four evaluation indexes,namely variation index,information entropy,fluctuation index and energy entropy,and establishes the feature set by comprehensive evaluation features.In this thesis,random forest is introduced to rank the importance of features.A method based on sample entropy and random forest is proposed to construct the feature set of rolling bearing,which provides the basis for the next step of diagnosis and recognition.Finally,this thesis uses support vector machine(SVM)as the diagnosis and recognition method,combines the method of parameter optimization VMD and the method of constructing rolling bearing feature set based on sample entropy and random forest,and constructs the rolling bearing fault diagnosis model with parameter optimization VMD as the core.Combined with the data of rolling bearing platform,the simulation experiment is carried out,and the experimental results verify the validity of the diagnosis model.