Cement rotary kiln control: A supervised adaptive model predictive approach

Considering the need of an advanced process control in cement industry, this paper presents an adaptive model predictive algorithm to control a white cement rotary kiln. As any other burning process, the control scenario is to expect the controller to regulate the temperature and the period of baking a fixed quantity of raw material as desired, as well as to have the concentration of the combustion gases under control. To achieve these goals, this work presents a strategy which includes multivariable online identification of the kiln process and a constrained generalized predictive controller. An MLP neural network model derived from real plant data of Saveh cement factory in Iran is used as the kiln process simulator. The control efforts are made taken into account the operating constraints. At last the proposed control strategy is modified so as to gain good disturbance rejection ability.