Nonlinear mill control

A mill is a mechanical device that grinds mined or processed material into small particles. The process is known to display significant deadtime, and, more notably, severe nonlinear behavior. Over the past 25 years attempts at continuous mill control have met varying degrees of failure, mainly due to model mismatch caused by changes in the mill process gains. This paper describes an on-line control application on a closed-circuit cement mill that uses nonlinear model predictive control technology. The nonlinear gains for the control model are calculated on-line from a neural network model of the process.