Research On Fault Diagnosis Of Rolling Bearing Based On SVM And BP Neural Network

Accidents or malfunctions of mechanical equipment will cause huge economic losses and even affect people's life safety.As one of the most widely used parts in mechanical equipment,rolling bearing plays the role of "joint" in mechanical equipment.Rolling bearing usually operates in a very bad environment which leads to its large life dispersion and high failure rate.Once the rolling bearing failure,How to accurately identify the fault parts and types,and formulate the corresponding maintenance strategy to ensure the safe and stable operation of equipment,which is of great significance to improve economic benefits and ensure production safety.Therefore,with rolling bearing fault diagnosis as the background,four types of vibration signal features,including normal state,rolling body fault,inner circle fault and outer circle fault,are extracted,and the state information contained in vibration signal is identified by Machine Learning.In this paper,the structure,failure reason,vibration cause and characteristic frequency of rolling bearing are firstly analyzed,and then the experimental data sources are introduced.Based on the characteristics of early fault signals of rolling bearings,this topic put forward of the vibration signal by the wavelet transform and reduces the interference of the noise,Based on the shortcomings of the wavelet transform,the wavelet packet transform is used to extract the characteristics of the normal state,fault of the rolling body,fault of the inner ring and fault of the outer ring of the rolling bearing.The support vector machine(Support Vector Machine,SVM)and the back propagation(Back Propagation,BP)neural network can classify fault characteristics,however,when the traditional SVM conducts fault classification of feature data,the selection of penalty parameter C and kernel parameter g will affect the classification effect of SVM.In this paper,based on the genetic algorithm(Genetic Algorithm,GA)has the characteristics of global optimization,the particle swarm optimization(Particle Swarm Optimization,PSO)algorithm has the characteristics of rapid convergence,A hybrid SVM algorithm(PSO-GA-SVM)based on PSO and GA was proposed for fault diagnosis of rolling bearings.Based on the defect that BP neural network is trapped in local optimal value,this paper proposes the use of the bat algorithm(Bat Algorithm,BA)to optimize BP neural network Algorithm(BA-BP neural network)for fault diagnosis of rolling bearing.By comparing with the traditional and optimized SVM and BP neural network for fault diagnosis,it is found that the algorithm model of PSO-GA-SVM and BA-BP neural network is better for fault classification of rolling bearings.Therefore,this paper concludes that the combined diagnosis model based on wavelet packet transform and PSO-GA-SVM and BA-BP neural network can effectively perform intelligent fault diagnosis of rolling bearing.