Research On Fault Diagnosis Technology Of Rolling Bearing Based On Machine Learning

As a key part of rotating machinery equipment,rolling bearing runs in the harsh environment of high speed and high load for a long time,which is easy to fail and affect the normal operation of the equipment.Therefore,it is of great significance to study its fault diagnosis technology.Taking rolling bearings as the main research object,this thesis starts from three aspects,namely signal decomposition,feature extraction and fault recognitionand diagnosis,to realize fault diagnosis of rolling bearings.The main research content of this thesis is as follows:(1)Firstly,the basic structure of rolling bearing is introduced,and its failure form,vibration mechanism and failure frequency are expounded.Since the empirical mode decomposition method has mode alipping phenomenon,variational mode decomposition is introduced to further analyze the two decomposition methods.Aiming at the endpoint effect of VMD decomposition,the boundary local feature scaling method is used to improve VMD,and the decomposition effect of EMD,VMD and improved VMD is compared and analyzed by simulation signals.It is proved that the improved VMD method is effective in signal decomposition.(2)Vibration signals collected from the experimental bench of Case Western Reserve University were used to extract bearing fault features from three aspects: time domain,frequency domain and multi-scale perarrangement entropy.Time domain and frequency domain features were respectively extracted from the collected vibration signals,and the vibration signals were decomposed into IMF components through IVMD adaptively.The multi-scale permutation entropy of IMF component was calculated as the fault feature,and the time-domain feature,frequence-domain feature and multi-scale permutation entropy feature were effectively dimensionally reduced by principal component analysis to construct the fault feature data set,on which fault diagnosis was carried out in the next step.(3)In the aspect of fault diagnosis,the support vector machine is used as the fault classification algorithm,and the locust optimization algorithm is introduced to optimize the parameters of SVM for the problem that the penalty factor and kernel parameters of SVM affect the classification accuracy,and the GOA-SVM fault diagnosis model is constructed.The vibration data obtained from the CWRU data set and the rolling bearing fault simulation test platform are used.Compared with unoptimized SVM,PSO-SVM and GA-SVM models.The results show that the GOA-SVM fault diagnosis method adopted in this thesis has higher recognition accuracy,and can realize the effective diagnosis of bearing faults.(4)This thesis developed a rolling bearing fault diagnosis system based on MATLAB APP Designer,and embedded the bearing signal processing method and intelligent fault diagnosis method studied in this thesis into the system.The system includes the login interface and the analysis interface of each link,and uses the intelligent algorithm to realize the running state diagnosis of rolling bearings,which provides the application prospect for realizing the intelligent fault diagnosis of rolling bearings.