Using Multivariable Predictive Control to Optimize the ASARCO Mission Mill

At the ASARCO Mission Mill, a multivariable, predictive controller was used to optimize parallel grinding circuits, improving throughput and product size. Using a dynamic model of the grinding circuit developed from on-line testing the software manipulates 4 variables to control 14 quality and constraint variables in each circuit. Traditionally, model-based control in grinding is difficult due to ore changes. This is addressed via a novel method to predict mill overload and automatic model switching.