A Deep Learning-Based Approach to Modeling the Digital Twin Physical Model of Machine Tools

Intelligent manufacturing is the integration of traditional manufacturing and information technology, and the digital twin is the best way to achieve the integration of physical space and digital space. Currently it is difficult to detect the physical state of the machine tool in real time during machining, and the digital twin physical model combined with physical laws is an effective way to achieve real-time monitoring. Therefore, this paper proposes a modeling architecture for digital twin physical models. First, the physical laws in the real space are extracted into the finite element simulation data, including the working conditions and simulation results. The simulation data is then used to train a deep learning model to achieve encapsulation of the physical laws, and the trained model enables real-time monitoring of the physical state of the machine tool. Finally, a thermodynamic physical model of the five-axis machining center VMC0656e was constructed. Compared to traditional finite element methods, the solution time and the file size generated by the calculation are greatly reduced within acceptable error limits.