(235e) Pharmapy: A Process Synthesis Framework for Hybrid Pharmaceutical Manufacturing Systems

Pharmaceutical processes have been traditionally operated in batch mode, due to the flexibility the batch mode offers in manufacturing of a variety of low volume – high value products. However, there is an increased interest in converting these processes to a continuous operating mode, to benefit from the potential advantages that continuous offers. The advantages include better controllability/safety, more uniform product quality (elimination of batch-to-batch variability), reduced manual labor, reduced intermediate material inventories and material handling and reduced equipment footprint; all of which can drive decreases in capital and operational costs.

While use of continuous operation in pharmaceutical manufacturing can be desirable from the technical-economical point of view, it can be limited by intrinsic process characteristics such as reaction kinetics, thermodynamic properties, transport process limitations, or stability concerns. As a result, the best operating mode could well consist of a combination of batch and continuous unit operations, i.e. a hybrid processing scheme. Hybrid processes pose two important challenges from the design point of view. First, the flowsheet structure space, i.e. the number of feasible process configurations that must be explored, escalate rapidly by virtue of the need to consider batch and continuous alternatives for each of the processing tasks that must be executed. Secondly, the introduction of discontinuities that are inherent to hybrid operation makes the coordination problem of solving and optimizing such flowsheets more difficult compared with purely batch or purely continuous processes [1].

These considerations give rise to the need for a tool capable of simulating and optimizing the candidate set of drug substance (DS) and/or drug product (DP) unit operations that define a particular pharmaceutical process flowsheet. The tool should be capable of automatically generating candidate process configurations in a hierarchical manner so as to allow systematic evaluation via solution of a sequence of optimization problems. Such an integrated tool would support decision-making during the early process design stages, using mathematical models and technical criteria, and would enable the design of processes and products aligned with the Quality-by-Design and Quality-by-Control paradigms [2], [3].

PharmaPy is designed as a two-level platform for the design and optimization of new pharmaceutical processes, or for the retrofitting of existing pharmaceutical facilities. The lower level consists of a series of state-of-the-art numerical tools [4] for the efficient solution and linkage of dynamic DS/DP pharmaceutical batch/semibatch/continuous unit operations, using the modular strategy for equation-solving [5], [6]. It is designed to facilitate the manual, heuristic or algorithmic generation of flowsheets using the Python syntax.

The upper level of the platform involves a design rule-enhanced, algorithmic structure which oversees the automatic generation of process configurations. In particular, mathematical tools such as disjunctive programming are employed to determine optimal flowsheet configurations by representing flowsheet choices as logical disjunctions. Solving such models rigorously results in large-scale MINLPs [7] which may be intractable given the complexity of many pharmaceutical unit operations that are of key interest for our work, i.e. detailed crystallization models. By implementing a design rule-enhanced system, many flowsheet alternatives can be heuristically eliminated leading to a much more reasonable number of flowsheet options which need to be accommodated within the disjunctive framework using, for instance, logic-based outer approximation.

A proof-of-concept case study involving the API synthesis and separation unit operations typically used in small molecule drug substance manufacture will be reported as test of the two-level platform.

References

[1] P. Stechlinski, M. Patrascu, and P. I. Barton, “Nonsmooth differential-algebraic equations in chemical engineering,” Comput. Chem. Eng., vol. 114, pp. 52–68, 2018.

[2] U.S. Department of Health and Human Services, Food and Drug Administration (FDA), and Center for Drug Evaluation and Research (CDER), “Pharmaceutical CGMPs for the 21s Century - A risk-based approach,” Rockville, MD, 2004.

[3] B. Szilagyi and Z. K. Nagy, “Model-based analysis and quality-by-design framework for high aspect ratio crystals in crystallizer-wet mill systems using GPU acceleration enabled optimization,” Comput. Chem. Eng., vol. 126, pp. 421–433, Jul. 2019.

[4] A. C. Hindmarsh et al., “Sundials,” ACM Trans. Math. Softw., vol. 31, no. 3, pp. 363–396, Sep. 2005.

[5] M. Hillestad and T. Hertzberg, “Dynamic simulation of chemical engineering systems by the sequential modular approach,” Comput. Chem. Eng., vol. 10, no. 4, pp. 377–388, Jan. 1986.

[6] V. Kulikov, H. Briesen, R. Grosch, A. Yang, L. Von Wedel, and W. Marquardt, “Modular dynamic simulation for integrated particulate processes by means of tool integration,” Chem. Eng. Sci., vol. 60, no. 7, pp. 2069–2083, 2005.

[7] M. Türkay and I. E. Grossmann, “Logic-based MINLP algorithms for the optimal synthesis of process networks,” Comput. Chem. Eng., vol. 20, no. 8, pp. 959–978, Aug. 1996.