Adaptive reconstruction of digital twins for machining systems: A transfer learning approach

Abstract

• The framework for reconstructing decision model of the machining system is proposed, which can realize the adaptive updating of digital twin model under different working conditions. • Combined with the transfer learning theory, an adaptive reconstruction strategy is proposed, which can meet the requirements under different working conditions and improve reconstruction efficiency. • The experimental drilling platform is built to verify the feasibility of the proposed method. The experimental results show that the performance of the decision model reconstructed by the proposed method has certain advantages, and the prediction error is less than 1.6%. Digital twin technology has been gradually explored and applied in the machining process. A digital twin machining system creates high-fidelity virtual entities of physical entities to observe, analyze, and control the machining process in real-time. However, the current digital twin machining systems lack sufficient adaptability because they are usually customized for specific scenes. Usually, if a decision model is directly reused in a different working condition, the accuracy of the decision model is often poor and difficult to work effectively. Meanwhile, the decision model remodeled from scratch will cause a waste of resources and low modeling efficiency. This paper proposes an adaptive reconstruction method to adjust the decision model in the digital twin machining system to enhance adaptability. The proposed method can ensure the rapid development of the digital twin decision model under new working conditions. Finally, taking the drilling process as an example, this paper establishes the experimental drilling platform and verifies the feasibility of this method in the burr prediction task.