Research On Transfering And Updating Method And Application Of Milling Digital Twin Model

Digital twin has the obvious advantages of virtual real interaction and accurate mapping of physical space.Using digital twin technology to build the digital twin model of milling processing unit is of great significance to realize product quality monitoring and production optimization in the processing process.The actual milling conditions are complex and changeable.Building a digital twin model based on a fixed milling scene usually leads to the decline of its decision-making performance or even wrong decisionmaking because it is difficult to adapt to the changes of working conditions,which affects the whole production and processing process.In view of the above problems,the research on the updating and adjustment method of milling digital twin model has important application value for maintaining the decision-making level of model under variable working conditions.This topic proposes the transfering and updating method of milling digital twin model under changing working conditions,and carries out application research and case verification combined with the prediction of milling surface roughness of aerospace structural parts:(1)A transferable digital twin modeling method for milling is proposed.The milling conditions are described in a unified parametric way.The characteristics of the traditional digital twin model and the transferable digital twin model for milling in decision-making modeling and decision-making model updating and adjustment are analyzed.The definition of the transferable digital twin model for milling is given,and the modeling method of the transferable digital twin model for milling is put forward.(2)The transfering strategy of milling transferable digital twin model is given,which mainly includes active variable condition transfering strategy and passive variable condition transfering strategy.The active variable condition transfering strategy mainly realizes the matching of the source model to be reused based on the Maximum Mean Discrepancy method and the transfer and update of the decision-making model based on the Deep Domain Confusion method.The passive variable condition transfering strategy is mainly based on the Adjusted R-squared method to realize the regular evaluation of the accuracy of the source model and the transfer and update of the decision-making model based on the fine-tuning method.(3)The decision-making algorithm of surface roughness of transferable digital twin in milling is realized.Based on the original milling conditions designed and deployed,the wavelet threshold denoising preprocessing is carried out for the milling vibration signal,and the time and frequency domain features are extracted.The time domain features of the electrical signal of the milling machine tool are extracted,and the Pearson correlation coefficient method is used for feature screening;The one-dimensional convolution neural network is used to build the decision-making model of the original working condition of milling surface roughness.(4)Case and verification,A digital twin surface roughness decision-making prototype system for milling is designed,which realizes the functions of on-line monitoring of workpiece surface roughness,visualization of system model evaluation results and interactive model update and prediction.Three cases of changing working conditions caused by the change of workpiece material,machining characteristics and tool performance state are designed and deployed.The milling experiments are carried out respectively,the vibration signals and machine tool electrical signals in the milling process are collected and processed,and the effect of model transfer and update is verified.The results show that the method proposed in this paper is effective to improve the decision-making performance of digital twin model under changing working conditions.