Research On Bearing Fault Diagnosis Based On Neural Network And Support Vector Machine

As one of the important parts of modern mechanical equipment,the state of bearings plays an important role in the operation of the whole mechanical system.If the bearing fails,it will not only lead to the abnormality of the whole mechanical system,resulting in serious economic losses,and even threaten the safety of people’s lives.Therefore,accurate diagnosis of bearing failure and early formulation of corresponding maintenance strategies to ensure stable operation of equipment are of great significance to ensure production safety and improve industrial production efficiency.In order to improve the accuracy of bearing fault diagnosis and the performance of diagnostic model,this paper investigates the bearing fault diagnosis method based on neural network and support vector machine,proposes a bearing fault diagnostic model with higher accuracy rate,and verifies the validity of the proposed model and the superiority of its performance through simulation experiments.The research content of the paper includes the following aspects:(1)Analyze the general structure of bearings,failure mechanisms,and forms of structural failure of bearings.and introduce four methods of fault feature extraction for bearing raw vibration signals: time domain analysis,frequency domain analysis,power spectrum analysis and wavelet analysis.Describe the experimental data acquisition platform of this study and the pre-processing work of feature extraction,downscaling and normalization on the experimental data set.(2)A bearing fault diagnosis model(BAS-BA-BP)based on the beetle antennae search-bat algorithm(BAS-BA)optimized BP neural network is proposed.The model optimizes the initial weights and thresholds of the BP neural network with the BAS-BA algorithm to overcome the problem of the network easily falling into local optimum brought by the random initial weights and thresholds,in order to improve the convergence ability of the BP neural network and the accuracy of the bearing fault diagnosis.The experimental results show that the proposed model is better than the comparison algorithm model in terms of the precision and accuracy of bearing fault diagnosis,as well as the fitting ability of the model.Meanwhile,the experimental results also show that the BAS-BA-BP model is less effective in the diagnosis of bearing outer ring faults.(3)An improved gray wolf optimization algorithm optimized support vector machine(IGWO-SVM)for bearing fault diagnosis model is proposed.The model first improves the convergence factor in the gray wolf optimization algorithm,and then optimizes the penalty factor c and kernel function parameter g of the support vector machine using the improved gray wolf optimization algorithm,so as to reduce the impact of randomly selected parameters on the classification accuracy of the SVM.The differences between the proposed model and the support vector machine models optimized by genetic algorithm,particle swarm algorithm and the pre-improved gray wolf optimization algorithm respectively are analyzed through experimental comparison,and the results show that the accuracy,fitting ability and fault diagnosis accuracy of the proposed model are better than the comparison models.Meanwhile,the problem that the BAS-BA-BP model does not have a high accuracy rate for the diagnosis of bearing outer ring faults is solved.(4)The comparative analysis of the two proposed models shows that the BAS-BABP model is more suitable for the diagnosis of bearing inner ring faults and rolling shaft faults,and the diagnostic accuracy is 11.11% and 5.56% higher than that of the IGWOSVM model,respectively,while the IGWO-SVM model has a higher diagnostic accuracy for outer ring faults,which is 11.11% higher than that of the BAS-BA-BP model.Therefore,the combination of the two models can provide a more effective solution for bearing fault diagnosis.