Fault Diagnosis Of Rolling Bearing Based On Signal Decomposition And Multi-scale Ent-ropy Using Support Vector Machine

Rolling bearing is an important part of large machinery,which has many advantages such as high transmission efficiency,strong bearing capacity and compact structure.However,due to long-term service in the environment of high speed and heavy load,the rolling bearing is extremely prone to failure.If it is not found in time,it will lead to poor working condition of the equipment or even stop working,seriously affecting production and life.Therefore,in order to avoid or reduce unnecessary loss caused by bearing failure,It is of great significance to study the fault diagnosis method of rolling bearing so that it can efficiently and accurately identify and judge the status of rolling bearing under complex working conditions.This article mainly focuses on the following points:(1)From a theoretical and engineering applications perspective,this paper provides a comprehensive discussion of the structural characteristics,failure characteristics,vibration mechanisms,and fault diagnosis methods of rolling bearings.Furthermore,it systematically analyzes the research status,advances,and future trends of each fault diagnosis method for rolling bearing fault diagnosis,in order to provide effective guidance for this discipline.(2)For the analysis of bearing vibration signals,which are often nonlinear in nature,this paper proposes an adaptive noise-complete ensemble modal decomposition algorithm-based rolling bearing fault diagnosis model.To extract low-dimensional features from the highdimensional vibration data,a combination of CEEMDAN decomposition and time-frequency fusion feature indicators is employed,and the obtained information is then fed into a support vector machine classifier for fault identification.The experimental comparison reveals that this method offers higher accuracy and reliability in identifying and predicting rolling bearing faults.(3)In this paper,an improved rolling bearing fault diagnosis model is proposed for dealing with bearing faults and more complex vibration signals under realistic operating conditions.The IMFs are reconstructed into feature signals according to the correlation coefficient thresholds,and the sample entropy at multiple scales in different states is calculated to form a multiscale entropy feature vector.The low-dimensional features are then input to the SVM to realize the recognition of locomotive wheel-to-bearing fault states.The feature extraction effect of the proposed method is verified by comparison with the recognition results of the previous experiments,demonstrating the effectiveness of the model in the recognition of real locomotive wheel-to-bearing faults,as well as the practical significance of the improvement scheme.(4)Aiming to address the shortcomings of existing feature extraction methods and pattern recognition algorithms in intelligent diagnosis,such as over-reliance on expert experience for tuning,this paper proposes a bearing fault diagnosis method based on a joint parameter optimization algorithm.Building upon a preceding model,the Mean Squared Error(MSE)of extracted feature values and the hyperparameters of the Support Vector Machine(SVM)classifier are adaptively optimized through the use of Particle Swarm Optimization(PSO)and Grey Wolf Optimizer(GWO)algorithms,thereby overcoming errors caused by manual parameter setting and effectively improving the accuracy of real bearing fault identification and classification.