Industrial digital twin empowered soft sensing for key variables in oxidized pellet rotary kilns

The grate-kiln oxidized pellet process plays a crucial role in the ironmaking process, and its efficient operation is of great significance for improving energy efficiency and environmental performance. However, the complexity of the rotary kiln process and the difficulty in directly measuring its internal state limit the monitoring and optimization of the production process. Digital twin technology, as a key enabler of the industrial internet of things in Industry 4.0, can create a virtual copy of a physical entity, enabling interaction between virtual and real systems and real-time monitoring, thus providing an innovative solution for intelligence-driven optimization of smart rotary kilns. Under this background, this study constructs a digital twin system for rotary kilns, which integrates key technologies including a three-dimensional temperature field simulation model, wall thickness monitoring model, and reduced-order model. A multi-physics coupling approach was employed to simulate the kiln's internal temperature field distribution. This simulation accounts for critical processes including pulverized coal combustion and pellet oxidation reactions. To enhance computational efficiency, we constructed a reduced-order temperature field model using random forest algorithms optimized by genetic algorithms. Wall thickness monitoring and caking detection were achieved through laser scanning technology combined with heat transfer principles. Field validation demonstrated the system's effectiveness: temperature prediction errors remained below 1 %, while wall thickness estimation accuracy reached 90 %. These results enable real-time operational guidance for production sites. Additionally, this method offers functions such as temperature monitoring alerts, caking/wall thickness monitoring, and prediction of caking growth trends, which are important for optimizing the production process and ensuring the safe operation of equipment.