Research On Method Of Fault Diagnosis Of Small Sample Gear Based On Support Vector Machine

With the development of modern industry,gear is widely used in modern industry.However,because of the problem that the gear failure causes the machine to not work normally,the intelligent fault diagnosis is very important.At present,the problem of fault intelligent diagnosis is that the number of typical fault samples is not enough.However,the support vector machine(SVM)algorithm can find the classified information in the existing data to the maximum extent on the premise of the limited feature information of the fault.The support vector machine(SVM)is more suitable for the practical problem of gear fault diagnosis from the extension point of fault diagnosis.In this paper,the gear is used as the research object and the support vector machine algorithm is used to identify the gear fault.By optimizing the parameters of support vector machine,the accuracy and generalization of the diagnosis are improved,and the fault diagnosis system is established to realize the intelligent diagnosis of gear fault.The main research work in this paper is as follows:(1)The study of the common fault type mechanism of the gear and the analysis of the fault vibration characteristics.The vibration data of normal,broken,worn,broken,and worn gears in four kinds of loads in foreign university database are applied,and four kinds of working condition data under four loads are analyzed in time-domain and frequency domain to confirm the fault frequency components in gear vibration signal.(2)The study of the feature extraction method which combines the dimensionless parameters and the wavelet packet decomposition and reconstruction of the extraction band energy.This method can effectively distinguish the characteristics of different faults of gears,and serve as the input eigenvectors of support vector machines,which provides effective and correct feature parameters for subsequent support vector machines for gearbox fault pattern recognition.(3)The classification performance of the fault classifier is closely related to the selection of the kernel function parameters of the support vector machine.The selection principle and function type selection,kernel function parameter of important parameters selection of support vector machine,penalty function optimization of fault classification accuracy;is proposed based on the application of grid search algorithm,particle swarm optimization algorithm and genetic algorithm for parameter optimization method of SVM classifier,and the grid search improved algorithm,and improves the searching speed.Analysis of the classification accuracy and classification time by comparison,will determine the grid search algorithm is applied to the support vector machine classifier parameter optimization through feasibility;diagnosis of vibration data of unknown gear,proved that the classification model has good generalization ability and generalization.(4)The gear fault diagnosis system is established by using the support vector machine(SVM)to build the gear fault diagnosis model.The accuracy and effectiveness of the operation of the diagnosis system are verified by the fault signal of the gearbox.Through the above research,by extracting feature data from various states of gear and input to the support vector machine classifier model optimized by the improved grid search algorithm,it can diagnose the gear status very well.Therefore,the application of SVM to gear fault diagnosis and state monitoring is of good practical significance in this paper.