Research And Application Of Fault Diagnosis Method For Motor Bearing Based On Multi-Source Information Fusion

As the core component of various mechanical equipment,motors have been widely used in various fields of national production and life.Accidents caused by motor faults often occur,causing serious economic losses and casualties.Rolling bearing faults are very common motor faults.Therefore,it is of great practical significance to study the fault diagnosis technology of motor bearings.Due to the poor ability of the motor bearing to withstand impact,and the operating environment is harsh,the fault mechanism is complex and diverse.The traditional single-sensor fault diagnosis system often cannot obtain complete information of bearing faults.The diagnosis process has great uncertainty and may even be misdiagnosed.Aiming at the above problems,this thesis proposes the use of multi-source information fusion technology for fault diagnosis of motor bearings,using the structure of decision-level fusion,the algorithms of data layer,feature layer and decision layer are studied and analyzed separately,and verified by the analysis of an example.In the data layer,based on the modal aliasing phenomenon in the empirical mode decomposition(EMD)algorithm,the ensemble empirical modal decomposition(EEMD)algorithm is proposed to use,it is verified that EEMD algorithm can effectively suppress modal aliasing through simulation,finally,based on EEMD algorithm,the extraction of motor bearing fault features and the construction of eigenvectors are completed.In the feature layer,based on the information collected by a single sensor,the RBF neural network and probabilistic neural network(PNN)are used to diagnose the motor bearing fault respectively,after comparison and analysis,it can be proved that the PNN bearing fault diagnosis system has obvious advantages in real-time,fault diagnosis accuracy and additional sample capability.In the decision layer,the fusion algorithm is D-S evidence theory.Aiming at the problem that D-S evidential reasoning rule can’t deal with high conflict evidence,based on the idea of correcting evidence source,a new improved algorithm based on evidence source correction is proposed,this algorithm determines the weight of each evidence based on the cosine similarity between two evidences,then weights each evidence,and finally combine them with D-S evidential reasoning rule,through the comparative analysis of classical examples,it is found that the algorithm can correctly combine highly conflicting evidence and the convergence speed is faster,based on the same decision rule,this algorithm can obtain better and more reasonable results.Based on the decision-level fusion structure,a motor bearing fault diagnosis system based on multi-source information fusion technology is constructed,the bearing fault standard data set of Case Western Reserve University motor data center is simulated and analyzed.The data layer is based on the EEMD algorithm to construct the fault eigenvectors;the feature layer uses three parallel probabilistic neural networks to classify the fault modes;in the decision layer,each summation layer output of PNN is fused by improved D-S evidence theory,because it can avoid the problem that it is difficult to obtain the basic probability assignment function.The simulation results show that the system can identify bearing faults quickly and accurately.