Dual Adaptive Model Predictive Controller Application to Vertical Roller Mill Process Used in the Cement Industry

Diversified operating conditions, input-output constraints, and parametric variations in the Vertical Roller Mill (VRM) make it to have complicated dynamics and closed-loop instability. Existing traditional controllers are not superlative and may lead to an uneven plant shutdown. Model predictive controller with adaptive models can track these parametric variations and ensure the plant's smooth running, which has been addressed in this paper. Data from the real-time VRM is acquired, and correlation analysis is carried out, which illustrates the use of outlet temperature and differential pressure as the output variables with tensile pressure and booster fan speed as the input variables. The base model for VRM is identified using the selected variables by data-driven system identification methods. The fourth-order state-space model was found to be optimal in capturing the dynamic behavior of VRM. Dual Adaptive Model Predictive Controller (DAMPC) is designed to handle each output variable individually. The use of DAMPC minimizes the complexity involved in the on-line parametric estimation for higher-order models by distributing the control authority to different controllers. The performance of the proposed DAMPC is compared with the existing Proportional Integral (PI) controller and Model Predictive Controller (MPC). Simulation experiments for reference tracking and rejection of slowly varying internal disturbances by considering parametric variations are carried out. Results illustrate DAMPC provides lesser overshoot and faster settling time amidst parametric variations.