Abstract: The ore properties of Dashan Concentrator of Dexing Copper Mine fluctuate frequently，and manual operation cannot effectively identify the changes of working conditions，resulting in large fluctua⁃ tions in the grinding process and difficult production control. Aiming at the problems that the key parameters in the field grinding process are difficult to measure directly and the traditional simulation system lacks dy⁃ namic real-time data support，a set of digital twin monitoring system for ball mill is developed through mod⁃ eling simulation and digital twin technology. The system can carry out research and analysis on the mecha⁃ nism model of ball mill motion trajectory and carry out dynamic trajectory research. Combined with the actu⁃ al production of grinding process，the self-organizing mapping neural network (SOM) classification method of mixed filling rate inside the cylinder is selected to realize the dynamic parameter classification and recog⁃ nition of mixed filling rate，so as to realize the dynamic calculation of steel ball motion trajectory. Finally， the digital twin monitoring system of ball mill is designed and developed to realize the three-dimensional model modeling and driving of ball mill，process flow simulation and process data monitoring. The industrial application results show that the system can effectively identify the state of mixed filling rate，improve the prediction accuracy of particle size of grinding products，and provide reliable technical support for the intel⁃ ligent operation and optimal control of ball mill.

Key words: ball mill, digital twin, SOM, 3D visualization