Model predictive control for pharmaceutical agitated liquid–liquid-extractions

Increasing efficiency of production processes is of great importance to reduce emissions and save costs at the same time. A resource-intensive production phase is product purification including liquid–liquid-extraction.
A specialized model predictive control (MPC) algorithm was developed for agitated extraction columns and compared against conventional PID approaches. The MPC framework incorporated dependent boundary handling, output error correction, and soft constraints to address pharmaceutical manufacturing requirements. Contrasting conventional control, the algorithm uses all process variables simultaneously to reach quality and optimization targets. Control strategies were evaluated using a non-linear virtual plant model under systematic disturbance scenarios. Results demonstrate that the MPC approach reduced organic waste by 70% while achieving 15% lower impurity overshoots compared to conventional control. The controller maintained robust performance during worst-case disturbance scenarios, ensuring required product quality. A practical approach for extraction processes with impacts on both product quality and operational efficiency is provided.