Research On Feature Extraction And Diagnosis Of Rolling Bearing Faults Based On Wavelet Packet And Support Vector Machine

Rolling bearing is a necessary key component in rotating machinery.Once a failure occurs,it will affect the normal operation of machinery and equipment.Timely fault diagnosis and troubleshooting of machinery and equipment plays an important role in the normal operation of equipment.Therefore,based on the fault diagnosis of rolling bearing,this paper studies the ball fault,inner ring fault,outer ring fault,three kinds of fault types and vibration signals under normal conditions,which can be used in the efficient operation of machinery and equipment.Safe operation and improving economic benefit are of great significance.Firstly,the main methods of extracting fault information are analyzed in this paper.According to the characteristics of early fault data of rolling bearings,wavelet packet transform(Wavelet Package Transform,WPT)technique is selected to extract 8 energy values after decomposition of the third layer.Eight energy values are taken to represent the characteristics of the original data.The energy eigenvalues after wavelet packet reconstruction need to be normalized as the input of the pattern classifier.Secondly,this paper chooses the pattern recognition model of support vector machine(Support Vector Machine,SVM)to carry on the pattern recognition to the feature data,but its penalty parameter c and kernel parameter g choose different seriously affect the classification effect of SVM.So in this paper,the artificial fish swarm algorithm(Artificial Fish Swarm Algorithm,AFSA),gravity search algorithm(Gravitational Search Algorithm,GSA)optimization support vector machine model is proposed to advance the feature data.In line pattern recognition,the principle of the algorithm is expounded,and two kinds of optimization algorithms are improved.The genetic algorithm(Genetic algorithm,GA)is used to cross-operate the AFSA fish colony mutation to get the IAFSA algorithm,and to add an inertia weight to the GSA to get the IGSA algorithm.Finally,based on the Matlab numerical simulation platform,the optimized SVM model identification method is simulated and realized.Finally,the quasi-optimization results of the unoptimized SVM,and AFSA-SVM,IAFSA-SVM,GSA-SVM,IGSA-SVM models are obtained.By comparison and analysis,their pattern recognition results are 83.33%,9 3.33%,96.67%,96.67% and 98.33%,respectively.It can be seen that the classification effect of the IGSA-SVM algorithm model is better.It has strong convergence ability and fast convergence rate.Therefore,the feature extraction method which based on wavelet packet transform combined with the optimized support vector machine model recognition method can obviously improve the recognition accuracy of rolling bearing fault.