Research On Construction Of Rolling Bearing Fault Data Simulation Platform Based On GAN

Rolling bearing is an indispensable workpiece in electromechanical equipment and is crucial to the electromechanical equipment.Therefore,it is very necessary to study rolling bearings.At present,the management of bearing condition is usually monitored by vibration sensors,which extract the signal characteristics to evaluate,diagnose and predict the bearing status.Nevertheless,with the increasingly continuous progress of machine learning methods,the mining of actual operating data is limited.Consequently,constructing a simulation generator for rolling bearing failure related data has become a solution.Nowadays,data simulation technology is not perfect and exists many disadvantages.The generative adversarial network is a new generation model in recent years and has achieved great success in the field of image generation.Therefore,this paper proposes a method for constructing a rolling bearing data generation platform based on the generative adversarial network.The main research contents include the following three parts:(1)This paper analyzes and studies the structure,fault mechanism,vibration mode,and characteristic signals of rolling bearings to provide a theoretical basis for processing rolling bearing characteristic data.(2)This paper studies the basic principles,network structure and training process of adversarial generative networks,and provides theoretical basis for constructing generative network models.(3)This paper analyzes the time domain analysis of the rolling bearing life cycle monitoring data in practical operation,and divides the data into three periods: normal bearing period,moderate bearing degradation period,and rapid bearing failure period.This paper conducts research on the construction and training of generative network models,comparing the generated data with real data in the time and frequency domains respectively,and verifying the similarity between the generated data and the real data to prove its validity.