Research Towards an Optimal Method of Modeling and Regulating a Cement Mill Using AI Algorithms

This research explores the optimization of modeling and control strategies for a cement mill using advanced Artificial Intelligence (AI) techniques. Neural Nonlinear AutoRegressive with eXogenous input (NNARX) models are employed to capture the nonlinear, multivariable dynamics of the cement grinding process. The Levenberg-Marquardt algorithm demonstrated superior training performance, achieving high accuracy in predicting key process parameters like inlet chamber fill and main drive current. The study also presents a neural control approach based on inverse modeling, enabling precise input predictions to maintain desired output conditions, stabilizing product quality and minimizing energy usage. Comparisons of neural network configurations suggest that simpler models are effective, with only slight benefits from more complex networks.