Consistent Digital Twin-Driven Fault Diagnosis of Planetary Gearboxes
In view of the limited nature of real data and the data dependency of deep models,it has become a trend to utilize digital twin technology to simulate data to train models,and to ensure the consistency of real and virtual data is still a key technical bottleneck that needs to be broken through for digital twin-driven health monitoring and intelligent operation and maintenance.In this paper,a digital twin virtual-real data consistency evaluation method is designed and combined with 1D convolutional neural network of migration learning for fault diagnosis.First,the simulation data of the digital twin model and the measured data are e-valuated multidimensionally by frequency domain similarity,waveform similarity and cosine similarity,and it is determined that the simulation data driven by multi-sensor real-time data fusion has high consistency.In order to apply the simulation data more efficiently and downgrade its differences with the measured data,a fault diagnosis strategy with simulation pre-training and fine-tuning of the measured data is proposed.This method provides theoretical support for the iterative evolution of digital twins and a new technical path for health monitoring and intelligent operation and maintenance of complex equipment.
digital twinplanetary gearboxmultidimensional evaluationfault diagnosis
万方数据
维普
