A dynamic updating method of digital twin knowledge model based on fused memorizing-forgetting model

Abstract

The customized manufacturing process of complex products (with complex structures and production processes) involving randomness and dynamics, is facing the problem of high data management costs and data waste caused by the accumulation of massive amounts of information. Especially in the digital twin-based workshop, the high-fidelity and interactive operation mechanism produces more massive data, aggravating the difficulty of data management. The combined effect of these complex manufacturing processes and dynamic batch production requirements poses a huge challenge to digital twin data management. To overcome this challenge, this paper proposes an updating method for digital twin knowledge based on a memorizing-forgetting model. Firstly, a multi-level representation model is proposed to fuse product, process flow, and manufacturing data. Secondly, the fused memorizing-forgetting model is proposed for dynamically updating digital twin knowledge. Finally, taking ship block manufacturing as an example, the effectiveness of the proposed method in modeling and fusion analysis is proved by the visual analysis of its resources and process knowledge. Considering the dynamic nature of production, it is believed that the data management method will significantly help improve the refined control of workshop resources and manufacturing processes, as well as the efficient use of massive processing data.