Abstract: To improve the consistence and real-time performance of the digital twin workshop system, a rapid construction method for the digital twin workshop system is proposed based on the parallel computing of logic and model data. Through modeling the physical information, the model and logical information can be constructed in digital twin workshop. Also, the construction principles of digital twin workshop are established, and improving the faithful mapping of twin relationships. The data of static model and of dynamic logic are modeled and analyzed. Through sending the static model data such as vertices, patches of the models into the GPU buffer, the static twin model can be computed based on predefined matrix conversion for position transformation and rotation. Then the parallel computing of logic and model data can be realized and the consistence and real-time performance of the digital twin workshop system can be guaranteed. Finally, a numerical experiment is implemented to verify feasibility and effectiveness of the proposed method. The results show that the proposed method is capable to improve the GPU utilization from 43% to 86%, and to improve the rendering efficiency of the digital twin workshop from 22.6 fps to 387.0 fps.

Key words: parallel computing, digital twin workshop, system structure, system optimization