Nonlinear identification and Model Predictive Control of a centrifugal fan in a coating machine

This thesis addresses the problem of regulating the inlet airflow in a tablet coating machine by developing and validating advanced optimal control strategies as an alternative to the more traditional approaches commonly adopted in the pharmaceutical industry. To support this objective, a nonlinear model of the centrifugal fan responsible for airflow generation is first identified using real system data acquired through an OPC UA communication interface, with different parameter estimation methods applied according to the specific model structure. The formulation that best reproduces the fan dynamics over a wide range of operating conditions is then selected for controller design, enabling the development of a set of Model Predictive Control (MPC) techniques and their validation in a simulation environment, significantly reducing the effort required for experimental tuning on the real machine. Measurement noise, parametric uncertainties, and other nonidealities were introduced to assess robustness and tracking performance under more realistic scenarios. The proposed controllers were then tested on the physical coating system by overriding the existing regulation loop through the OPC UA protocol. In this final phase, the experimental results were analyzed in detail and compared with the controller currently implemented on the machine, highlighting the performance differences observed in actual operation and the improvements achieved through the integration of a model-based predictive strategy into the airflow regulation process.