(391c) Advanced, Material-Aware Model Predictive Control Strategies for Evaporation Processes in the Pharmaceutical Industries

Pharmaceutical plants are required to operate with very strict product quality specifications and deal with complex and highly integrated processes, varying production targets and incoming material variability. The process design step, in particular, is an important and major factor in determining how well controlled the process can be. This step is time consuming, expensive and resource intensive. Model Based Control (MbC) is an important advanced control technique for complex multivariable control problems. It is able to predict the future behavior of a system, deal with complex dynamics, and to incorporate process constraints (safety, business, equipment, etc.) within a single optimization based problem formulation.

This work illustrates how using model based control approaches greatly improves the time and resource utilization for the development process. Moreover, certain pharmaceutical process design problems are transformed into process control problems. Consider a pharmaceutical process for the manufacture of a newly active pharmaceutical ingredient (API). The traditional approach requires a series of Design of Experiments (DoE) to be performed for the new process to be designed. An alternative MPC approach, by using the process model and being aware of material properties can make the necessary adjustment to adapt to the new API/solvent mixtures where only information regarding the material properties of the new APIs/solvent mixtures are required. Open loop experiments can still be performed and used for parameter estimation. This will eliminate the need of DoE’s as well as the need for the definition of design space since this is an open loop concept that loses relevance in the presence of a controller for the quality attributes.

We develop an advanced multi-parametric model predictive controller (mp-MPC) for a semi-continuous evaporation process designed to work with different APIs/solvent mixtures. For the general design, a high fidelity model of the process was developed that is further used for testing and validation purposes [1]. Possible thermodynamic scenarios and their corresponding effects on the control strategies are then identified and classified to determine what type of controller design works for each scenario. Therefore, for every new API/solvent mixture, the user only needs to feed the corresponding thermodynamic properties, production targets and concentration setpoints to the controller, which will adjust the design and tuning parameters accordingly. To test the performance and limitations of the developed control strategy, it is applied to different mixtures for varying operating targets and process disturbances.

References

1. N. Pistikopoulos, E., et al., PAROC—An integrated framework and software platform for the optimisation and advanced model-based control of process systems. Vol. 136. 2015.