Research On Data-Model Combined Driven Digital Twin Of Life-Cycle Rolling Bearing

Rolling bearings are widely used in various rotating machinery such as aero-engine,high-speed train,precision machine tool and so on because of its excellent performance like low friction,high precision.As one of the most important components in mechanical equipment,the performance and life of bearing can determine the performance and life of the whole equipment.In the running of bearings,it is necessary to recognize the running state of bearings,estimate the evolution trend of bearings and predict the remaining service life of bearings to prevent the occurrence of faults in advance and avoid the occurrence of major accidents.Therefore,in order to effectively guide fault diagnosis and life prediction of bearing,it is significant to carry out the research on the fault mechanism of rolling bearing,and build its fault dynamics model and life-cycle model.However,in order to obtain high fault diagnosis and life prediction accuracy,which usually require much samples,but the cost of rolling bearing life-cycle experiment is huge.In addition,at present,most of the fault dynamic model of rolling bearings are simple to construct,which are only suitable for simple working conditions such as low speed and low temperature.Moreover,the effect of fault impact is only characterized by a simple displacement excitation function,which makes the generated simulation signal have different from the measured signal.At the same time,the life-cycle dynamic model and research of rolling bearings are also lacking at present.Therefore,in view of the above problems,this paper carries out the research on data-model combined driven digital twin of life-cycle rolling bearing.Firstly,a compound displacement excitation function by considering the timevarying impact and the fluctuation of defect surface,which can better simulate the excitation from the real fault.At the same time,the high-speed effect,dynamic radial clearance and transfer path effect are introduced into the traditional rolling bearing fault dynamic model,and the dynamic model of high-precision rolling bearing is constructed.Using the bearing fault dynamic model,the dynamic responses under different fault sizes and speeds are calculated by Runge Kutta method and Newton Raphson method,and their time-domain and frequency-domain characteristics are analyzed.The results prove the necessity of introducing high-speed effect and dynamic radial clearance,as well as the superiority of composite displacement excitation function.A comparative experiment is carried out on the rolling bearing fault test-bed.The results show that the model is superior to the existing rolling bearing fault dynamics model.In addition,in order to make the model construction more convenient,the rolling bearing fault dynamics modeling software is designed and developed.Then,the construction method of digital twin model of life-cycle rolling bearing is proposed.That is,based on the bearing fault dynamic model,combined with the measured signals,the evolution size of bearing defects is estimated,and the evolution law of bearing defects in the whole life cycle is revealed by BP neural network.The excitation of evolutionary defects is introduced into the dynamic model,and the whole life dynamic model of bearing is established in the virtual space.In addition,in order to make the generated simulation signal closer to the measured signal,a cyclegan mapping network with smooth cyclic consistency loss is constructed to build the mapping relationship between virtual space and physical space.Finally,based on the XJTU-SY rolling bearing accelerated life test data set,the corresponding digital twin model of life-cycle rolling bearing is constructed.The simulation signal of the digital twin model is compared with the measured signal in time domain and frequency domain to verify the effectiveness of the model.