Aiming at the problems of the lack of perceptual monitoring accuracy of the existing industrial robot intelligent equipment modeling and the low accuracy of modelling based on theoretical parameters, this paper takes the industrial robot milling system as the research object and constructs a digital twin measurement system that measures the robot joint turning angle data in real time with a high-precision scale to avoid the influence of joint turning angle errors such as gear gap and encoder code loss on the accuracy of digital twin modelling. The digital twin drive model was developed based on the MD – H kinematic modelling method, and the L – M algorithm was used to identify and correct the industrial robot modelling parameters to reduce the influence of geometric errors in the digital twin modelling of the robot. The use of the identified robot joint parameters for modelling has improved the accuracy of the twin model for modelling the motion points of the industrial robot from ±1.6905 mm to ±0.3304 mm, increased by 4.12 times, which demonstrates the correctness of the digital twin modelling method and the feasibility of the identification of the modelling parameters.