Dynamic simulator and model predictive control of a milk pasteurizer

In this study, the design, optimization and dynamic modelling of a milk pasteurization unit have been developed, using the pseudo-component approach for describing milk properties. The fluid has been regarded as a mixture of five major categories, namely water, fats, proteins, carbohydrates, and minerals. Exploiting the optimal pasteurizer configuration, selected based on the total annualized cost, a dynamic model of the process has been also derived. The simulation of the system is then used as a virtual plant to develop a nonlinear model predictive control (NMPC) designed for rejecting the more important disturbances that can enter the system. The predicted trajectories have been calculated with a simplified version of the dynamic model, obtained by neglecting parameters temperature dependence. The NMPC performance has been compared with a PI controller in terms of set-point tracking and disturbance rejection. Similar results have been obtained when using the different control algorithms for the output responses, but the NMPC showed better behaviour of the manipulated variables.