Search

Assuring compliance of intermediate and final quality attributes in a continuous pharmaceutical manufacturing campaign is of utmost importance. Application of corrective actions might be required in real-time. This work exemplifies the steps needed to identify a linear pulse transfer function for the dynamic behavior of the granule liquid-to-solid ratio (%w/w) at the end of the granulation unit of a commercial ConsiGma(TM)-25 production line. Near-infrared spectroscopy was used to monitor the granule composition in-line. The outcome for both the tracking and regulator problem using either conventional or model predictive control was implemented and evaluated. Dynamic setpoints were correctly followed and an RMSE of 0.25%w/w with respect to the setpoint was obtained when inducing artificial disturbances. Important practical challenges were also tackled. Examples are fouling, computational limitations, and the limited flexibility of the automation software. Applying the proposed advanced process control solution offers an answer to upstream material flow rate deviations. (c) 2018 American Institute of Chemical Engineers AIChE J, 64: 2500-2514, 2018

One major advantage of continuous pharmaceutical manufacturing over traditional batch manufacturing is the possibility of enhanced in-process control, reducing out-of-specification and waste material by appropriate discharge strategies. The decision on material discharge can be based on the measurement of active pharmaceutical ingredient (API) concentration at specific locations in the production line via process analytic technology (PAT), e.g. near-infrared (NIR) spectrometers. The implementation of the PAT instruments is associated with monetary investment and the long term operation requires techniques avoiding sensor drifts. Therefore, our paper proposes a soft sensor approach for predicting the API concentration from the feeder data. In addition, this information can be used to detect sensor drift, or serve as a replacement/supplement of specific PAT equipment. The paper presents the experimental determination of the residence time distribution of selected unit operations in three different continuous processing lines (hot melt extrusion, direct compaction, wet granulation). The mathematical models describing the soft sensor are developed and parameterized. Finally, the suggested soft sensor approach is validated on the three mentioned, different continuous processing lines, demonstrating its versatility.