The Research Of Rotating Machinery Fault Diagnosis Method Based On Deep Support Vector Machine

We can prevent the fault state of the equipment,and improve the reliability and safety of the equipment by using mechanical fault diagnosis technology.Rotating machinery is widely used in industrial production,and fault diagnosis for rotating machinery has always been a key research direction in the field of mechanical failure.With the progress and development of society,large-scale high-speed complex rotating machinery is more widely used,and the reliability and safety requirements of equipment operation are also increasing.Traditional diagnostic methods have been unable to meet the corresponding diagnostic requirements.With the development of methods,fault diagnosis methods have gradually evolved from simple diagnosis,precise diagnosis,and gradually to basic machine learning,intelligent diagnosis methods of artificial intelligence.Among them,a pattern recognition method based on structural risk minimization named support vector machine can learn fault characteristics after fault data training,and realize fault diagnosis.Here,based on the support vector machine as the theoretical basis,a modified deep support vector machine pattern recognition method is proposed around the fault diagnosis of rotating machinery,and related experiments are carried out by design to study the application of deep support vector machine in fault diagnosis of rotating machinery.The research directions and contents of this article mainly include the following aspects:(1)First,an improved deep sparse least squares support vector machine diagnosis method is proposed.This method constructs a multi-layer support vector machine structure.The training samples are trained on the input layer using support vector machines,new samples are generated through data conversion,and the next layer is input for training.Multiple learning is used to improve the classification accuracy.Considering the complexity of algorithm training and the increase in diagnosis time caused by the multilayer structure,the least square support vector machine is selected,and a sparsity theory is used to construct the final training model.Finally,the classification performance test verification and three sets of rotating machinery fault diagnosis experiments are designed respectively,and the diagnosis results of various diagnosis methods are compared.At the same time,the corresponding parameters and structure are discussed in detail.The final results show that the proposed method has good performance.(2)As an improved classification algorithm,deep sparse least squares support vector machine has the same model training as most machine learning and requires enough training samples.However,there are some specific application scenarios,the number of training samples in the target domain is insufficient,and it is difficult to train a diagnostic model with sufficient generalization ability.Therefore,combining sample transfer learning and deep support vector machine theory,a deep transfer least square support vector machine diagnosis is proposed The model uses the auxiliary training set(source domain data)with certain similarity to the target domain sample for reliable identification,and assigns the weight of the source domain and target domain training samples to the target optimization function to make the source domain instance closer to the target domain The sample distribution of,combined with the deep support vector machine idea,carries out multiple feature learning through a multi-layer structure.In order to verify the diagnostic performance of the deep transfer least squares support vector machine,this paper separately designed and conducted the planetary gearbox fault diagnosis experiment,the two-stage helical gearbox fault diagnosis experiment and the rolling bearing public data set verification,and verified the diagnostics of the proposed method.