Model predictive control for continuous pharmaceutical feeding blending units

This paper outlines several control strategies for a feeding blending unit in a continuous pharmaceutical manufacturing setup, with the goal of damping feeder fluctuations by adjusting the blender speed. An advanced model predictive control concept and two standard concepts, PI- and model-based feed-forward control, are described and compared with respect to their performance in the simulation. The concepts which showed satisfactory behaviour in the simulation are further tested on the real system. The paper also provides a detailed explanation of each implementation step in a real system, i.e., measurement device setup, signal processing, parameter identification and controller design. The properties of control concepts are compared and analysed, particularly in view of further work on the complete direct compaction line.