Research On Weak Fault Signal Enhanced Recognition And Diagnosis For Health Condition Monitoring Of Rolling Bearing

Rolling bearing is an important part of mechanical transmission system,is one of the main causes of rotating machinery failure,its failure will affect the transmission accuracy of the system,serious bearing failure will even lead to catastrophic accidents,resulting in serious loss of life and economic property.Therefore,it is of great significance to monitor,discover and diagnose bearing faults as early as possible for the operation and maintenance of rotating machinery and to prevent and avoid accidents.The vibration signal is an important carrier of the running state information of rolling bearings.The vibration signal of early fault of rolling bearings is weak.How to effectively monitor,discover weak fault of rolling bearings by using vibration signal processing technology,and conduct pattern recognition of different fault characteristics under varying working conditions is the key problem to be solved in the research of early weak fault diagnosis of rolling bearings.In view of this,this dissertation takes rolling bearings as the research object,and through analyzing the fault forms,causes and fault signal response characteristics of rolling bearings,carries out theoretical research and practical verification exploration on strengthening,identification and diagnosis of weak fault signals of rolling bearings.The main research work can be summarized as follows:(1)Aiming at the problems of pulse characteristics of weak fault signals of rolling bearings,difficulty in extracting weak fault features and lack of applicability research of different filtering methods,The case verification study shows that the deterministic signal separation and random signal separation method based on Intrinsic Time-scale Decomposition(ITD)filter is suitable for weak fault signal separation where the absolute value of signal amplitude is very small.The separation method of deterministic signal and random signal based on Multipoint Optimal Minimum Entroy Deconvolution Adjusted(MOMEDA)filter is more suitable for the separation of fault signal under strong noise or strong interference.(2)Aiming at the problem that weak fault signals of rolling bearings are easy to be annihilated in noise signals and difficult to extract effective features,a weak fault feature extraction method of rolling bearings based on "ITD filtering + adaptive bistable stochastic resonance" was proposed to deal with weak fault signals with relatively small amplitude."MOMEDA filter +adaptive bistable stochastic resonance" is used to deal with the weak fault signals with low signal-to-noise ratio,so as to realize the weak fault signal processing of rolling bearings under stable working conditions/variable working conditions.On this basis,in order to further improve the accuracy of feature extraction,Cuckoo Search(CS)algorithm was introduced to achieve the optimal matching of adaptive bistable stochastic resonance scale factor and step size,thus effectively improving the output signal-to-noise ratio of weak fault signals.Simulation signal analysis and case verification results show that this method can improve the original weak fault signal signal-to-noise ratio of rolling bearing by about 3 times,and greatly improve the feature extraction effect of fault signals.(3)On the basis of the above work,this dissertation takes advantage of the characteristics of fast convergence speed and strong global search ability of the Grey Wolf Optimization(GWO)algorithm,and further proposes the feature extraction method of the combination of MOMEDA and GWO algorithm optimization adaptive stochastic resonance to enhance the weak fault feature signal,and proves that the method is more efficient,reliable and accurate through the example.(4)In view of the problem that the stochastic resonance theory of bistable system is only effective in strengthening the frequency components near the optimal resonance frequency and is strongly dependent on the amplitude-frequency characteristics of noise,this dissertation takes advantage of the weak dependence of the adaptive Wood-Saxon stochastic resonance on noise characteristics.Furthermore,the MOMEDA-CS-Adaptive Wood-Saxon stochastic resonance weak fault signal feature extraction method is proposed.The example verification results show that the proposed method has a significant improvement over the bistable stochastic resonance feature extraction effect,with an average improvement of about 56%.(5)In view of the different fault states of rolling bearings under different working conditions,it is difficult to achieve effective screening only through eigenvalue extraction.A CS optimized enhanced Multiscale Dispersion Entropy(MDE)-Multi-Label K-Nearest Neighbor(ML-KNN)fault pattern recognition and diagnosis method is proposed.The MDE eigenvalue was used as the fault pattern recognition criterion,and the CS algorithm was used to achieve the optimal matching of the scale factor,delay time and embedding dimension of MDE,which effectively improved the accuracy of the extraction of the eigenvalue.In view of the criticism that supervised learning relies heavily on training sample data,Stacked Auto Encoder(SAE)+ Support Vector Machines(SVM)pattern recognition classifier was constructed,which could fully excavate the feature information contained in unlabeled collected data samples and improve the weak fault diagnosis accuracy of rolling bearings.Furthermore,an unsupervised learning model SAE+SVM based on CS optimization enhanced MDE was proposed for weak fault diagnosis of rolling bearings.The experimental results show that the proposed method can effectively classify and identify different fault states of rolling bearings under different working conditions.(6)In order to further verify the engineering practicability of the proposed method of weak fault diagnosis of rolling bearings based on adaptive stochastic resonance and MDE,the method was applied to the early fault diagnosis of rolling bearings at the driving end of dewatering and salt pumps in a chemical plant of an enterprise,and the early weak fault diagnosis and recognition of healthy rolling bearings were realized.