Abstract: In order to solve the problems of control silos and information silos caused by the dispersion of multi-time control subsystems in Taiyuan coal preparation plant, and improve its production efficiency and intelligent management level, this research has developed a digital twin system based on UE5 technology, which builds a high-precision virtual model through UAV aerial photography, laser point cloud scanning and 3D modeling, combined with industrial ring network and 5G technology to achieve real-time data acquisition and transmission; for the heavy-media sorting process, a density dynamic prediction model based on convolution-two-way short- and long-term memory network (Conv-BiLSTM) is proposed, and a fuzzy control rule base is designed to realize intelligent adjustment of sorting density; develop intelligence The inspection system dynamically plans the path based on FMEA.The results show that: after the system is deployed, the sorting density control error is reduced to ±0.05 g/cm³, and the equipment failure rate is reduced by 28%; the intelligent inspection system increases the coverage rate of high-risk equipment to 100%, and the time-consuming of a single inspection is shortened by 37.5%, and the path length is reduced by 32%.Through the dynamic simulation and intelligent regulation of key processes by digital twin technology, the Taiyuan Coal Preparation Plant has achieved production process optimization and reduced operation and maintenance costs by 15%.