As more and more robots are used in industrial production, real-time state monitoring during robot running process is attracting widespread attention. Aiming at the problem that conventional Finite Element Analysis costs long calculating time, making it difficult to meet the demand of real-time monitoring, the grasping robot is taken as the research object in this paper to construct its digital twin model for simulating the real motion state of the robot itself, and to realize a real-time monitoring in stress state of mechanical grasper based on Deep Learning method. The 3D model of the grasping robot is firstly constructed and then configured in Unity 3D environment. The running data of the robot is collected and transmitted to drive the digital twin model moving as same as the robot itself. In order to realize the real-time monitoring of mechanical grasper’s stress state, a quick generating method for images based on Deep Learning is studied in this research. Finally, high-speed loading of image data and animation generation of mechanical grasper’s stress state are realized in Unity 3D.