Research On Fault Diagnosis Of Servo Motor Bearing Based On GAN And DBN Model

Servo motors are important industrial power sources that provide driving force for rapidly developing modern industrial production lines under the background of automation,high precision and energy saving.In actual use,the deterioration and failure of key vulnerable parts in the motor are the biggest obstacles affecting the smooth running of the motor.This effect will be passed directly to the entire production line that provides the driving force for the dependent motor.For both the producer and the user,it is desirable that the servo motor can operate in a safe,efficient,and stable state.Therefore,real-time understanding of the operating conditions of the motor is of great significance for the detection and fault diagnosis of the critical vulnerable parts and core components of the servo motor.From the perspective of improving the stability of the servo motor and real-time grasping the operating conditions of the core components,this dissertation establishes the fault diagnosis model based on the fault defects of the key components of the servo motor,and combines the characteristics of the signal such as spectral kurtosis.The extraction method uses a model such as a generation confrontation network to generate a missing signal.Finally,based on the deep confidence network,the signal is identified by pattern recognition to complete the fault diagnosis.The main research contents of the thesis are as follows:(1)The paper first introduces the source of the subject and the background and significance of the research.Then the research status of state monitoring and fault diagnosis technology based on signal and model at home and abroad is comprehensively discussed.After fully discussing the structural characteristics,working principle,fault type and cause of servo motor,the core components and research scheme of fault diagnosis are determined.(2)The paper then expounds the techniques of spectral kurtosis,recursive quantitative analysis,etc.,which are at the forefront of signal feature extraction.Combined with the needs of subsequent follow-up models,the method of feature extraction and selection is selected,and the model is used for later use.Technical support is provided for identification and troubleshooting.(3)Combining with the problems of data loss in the field of fault diagnosis,this paper proposes the generation of missing fault signals based on the generation of confrontation networks.This paper introduces the principle of generating a confrontation network,analyzes the mechanical characteristics combined with the field of fault diagnosis,and introduces the generated confrontation network into the field of fault diagnosis.(4)The paper adopts the deep confidence network,combined with the above generated fault data generated by the anti-network,on the one hand to complete the pattern recognition work of fault diagnosis,and on the other hand,it also proves that the fault data generated by the anti-network generation is represented in the pattern recognition model.Out of performance.Finally,through the various servo motor and bearing fault data obtained by the Case Western Reserve University,the University of Cincinnati and the Kunming University of Science and Technology laboratory,the model performance was comprehensively verified.