A Probabilistic Graphical Model for Semi-Autogenous Grinding Processes

In the area of automation and process control, a pressing issue is the uncertainty in the dynamics of the SAG mill ore feeding process and the complexity in identifying the optimal feeding configuration to maintain a stable charge consistently over time. In this paper, we develop a Probabilistic Graphical Model using a Bayesian network that may enhance the decisionmaking process for operating a semi-autogenous grinding (SAG) mill. The Bayesian network is initially constructed based on prior knowledge. Conditional probabilities are calibrated using a subset for training and validated using a test subset. The obtained results demonstrate that the model achieves high accuracy in determining the fresh mineral feed rate and the power consumed by the mill. Thisshowcases the model&s effectiveness in capturing the primary operational standards.