(121b) Closed-loop control of roller compaction in a dry granulation process

Authors: 
Moreno, M., Purdue University
Gupta, A., Purdue University
Giridhar, A., Purdue University
Reklaitis, G., Purdue University
Nagy, Z. K., Purdue University

Dry granulation is a widely-used process in pharmaceutical manufacturing wherein blended powders are compacted under pressure to granules to be further processed into tablets and other product forms. Roller compaction is the primary unit operation to compress powders to compacts (ribbons), which are then milled into granules. Roller compaction is sufficiently complex to warrant work on control strategies, in order to keep granule density at specified levels and to reduce the occurrence of fines and off-size particles. Control strategies for roller compaction become particularly important in a QbD context for continuous manufacturing.

Process control for roller compaction is generally achieved in practice by simultaneously controlling ribbon density and ribbon thickness for a range of production rates. Manipulated variables are usually the roller rpm, the feed screw rpm (moving powder to the compaction zone), and the pressure exerted by the rolls on the powder (usually hydraulically actuated). Ribbon densities are usually measured online through NIR spectroscopy, or also microwave spectroscopy.

In this work, we present our results from simulation and experiments of different control strategies on roller compaction. We start with simple independent SISO PID loops, treating ribbon thickness and density as independent variables. We show how control loop interaction can be minimized with MIMO schemes, including cascade control and model-predictive control, both linear and non-linear. We compare the performance and complexities of different control schemes, and quantify the tradeoff between implementation ease and process performance. We then offer practical considerations for closed-loop control of roller compaction, describe common process faults seen in roller compaction, and describe  how the control system should interact with fault management systems.