(584f) Optimization and Control of Modular Chemical Systems for Continuous Manufacturing

Pharmaceuticals have traditionally been manufactured by using batch processing. Interest in recent years in both academia and industry has shifted towards continuous processing, with the aim to reduce drug costs, production times, waste, and product quality variations, while facilitating faster demand change responses [1,2,3]. Other recent advances in advanced manufacturing involve modular chemical systems, which employ process intensification and the potential for increased flexibility and productivity (e.g., see [4,5] and citations therein). This interest is particularly prominent in the chemicals, pharmaceuticals, biotechnology, and energy industries (e.g., [3,6,7]). This talk focuses on formulation, methodological, and computational advances in the optimization and control of modular chemical systems, with a focus on automation, to provide solutions for continuous-flow pharmaceutical manufacturing. Both first-principles-based and data-driven methodologies are discussed.

Linear MPC (LMPC) methodologies are a promising supervisory control approach for modular systems, since they have already undergone extensive testing and implementation in the chemical industries and government regulatory agencies are more likely to approve an industrial drug production facility that uses such control software. An additional advantage of LMPC formulations such as quadratic dynamic matrix control (QDMC) is the usage of input-output (IO) predictive models. IO models enable low computational cost when solving online optimizations, even for systems otherwise described by first-principles models that have very high state dimension which arise, for example, when partial differential equations are discretized (i.e., such as associated with tubular reactors in a modular system). The first part of this talk will review some methodologies for the construction of linear IO model that result in improved closed-loop performance [8], including some results that go beyond our past publication. Closed-loop simulation results for the control of a plant for the upstream synthesis of an important pharmaceutical using QDMC are presented for various scenarios of disturbances acting on the system. This talk also addresses the control of dynamical operating regions of the modular plant by employing dynamic optimization (offline) and QDMC (online) [9]. This hybrid approach results in closed-loop responses exhibiting good robustness under the existence of model parametric uncertainty for a realistic computational case study [9].

An alternative to first-principles and linear IO models are data-driven nonlinear models such as dynamic artificial neural networks (DANNs). DANNs can be used in a nonlinear model predictive control (NMPC) algorithm that is runnable in real time, which was explored in the 1980s but has received increased attention in recent years. An alternative DANN-based approach is to design approximate MPC strategies in which the control law is learned by data. Such strategies have been shown to give good closed-loop performance in simulations, and guarantees of their theoretical properties have been recently derived [10,11]. This talk discusses another alternative DANN-based approach, which employs the mathematical framework of Lyapunov theory and matrix inequalities to design dynamic state and output feedback controllers and observes for DANN models [12,13]. The methodology is demonstrated on two case studies, for a continuously stirred tank reactor control and pH control, where good closed-loop performance is observed. Given that DANNs have been used in the control of nonlinear dynamical systems in industrial practice for decades, it is conceivable that such approaches could someday become sufficiently accepted that they could be applied in the control of modular pharmaceutical manufacturing.

References

1. S. Mascia, P. L. Heider, H. Zhang, R. Lakerveld, B. Benyahia, P. I. Barton, R. D. Braatz, C. L. Cooney, J. M. B. Evans, T. F. Jamison, K. F. Jensen, A. S. Myerson, and B. L. Trout, “End-to-end continuous manufacturing of pharmaceuticals: Integrated synthesis, purification, and final dosage formation,” Angewandte Chemie International Edition, vol. 52, no. 47, pp. 12359–12363, 2013.

2. I. R. Baxendale, R. D. Braatz, B. K. Hodnett, K. F. Jensen, M. D. Johnson, P. Sharratt, J.-P. Sherlock, and A. J. Florence, “Achieving continuous manufacturing: Technologies and approaches for synthesis, workup and isolation of drug substance,” Journal of Pharmaceutical Sciences, vol. 104, no. 3, pp. 781–791, 2015.

3. A. S. Myerson, M. Krumme, M. Nasr, H. Thomas, and R. D. Braatz, “Control systems engineering in continuous pharmaceutical manufacturing,” Journal of Pharmaceutical Sciences, vol. 104, no. 3, pp. 832–839, 2015.

4. A.-C. Bédard, A. Adamo, K. C. Aroh, M. G. Russell, A. A. Bedermann, J. Torosian, B. Yue, K. F. Jensen, and T. F. Jamison. “Reconfigurable system for automated optimization of diverse chemical reactions,” Science 361(6408):1220–1225, 2018.

5. J. A. Paulson, E. Harinath, L. C. Foguth, and R. D. Braatz. “Control and systems theory for advanced manufacturing.” In Emerging Applications of Control and System Theory, edited by Roberto Tempo, Stephen Yurkovich, and Pradeep Misra, Lecture Notes in Control and Information Sciences, Springer Verlag, Chapter 5, 63–80, 2018.

6. R. Lakerveld, P. L. Heider, K. D. Jensen, R. D. Braatz, K. F. Jensen, A. S. Myerson, and B. L. Trout. “End-to-end continuous manufacturing: Integration of unit operations.” In Continuous Manufacturing of Pharmaceuticals, edited by P. Kleinebudde, J. Khinnast, and J. Rantanen, Wiley, New York, Chapter 13, pages 447–483, 2017.

7. M. S. Hong, K. A. Severson, M. Jiang, A. E. Lu, J. C. Love, and R. D. Braatz. “Challenges and opportunities in biopharmaceutical manufacturing control,” Computers & Chemical Engineering, 110:106–114, 2018.

8. A. Nikolakopoulou, M. von Andrian, and R. D. Braatz. “Plantwide control of a compact modular reconfigurable system for continuous-flow pharmaceutical manufacturing,” in Proc. American Control Conference, pp. 2158–2163, 2019.

9. A. Nikolakopoulou, M. von Andrian, and R. D. Braatz. “Fast model predictive control of startup of a compact modular reconfigurable system for continuous-flow pharmaceutical manufacturing,” in Proc. American Control Conference, pp. 2778–2783, 2020.

10. A. D. Bonzanini, J. A. Paulson, D. B. Graves, and A. Mesbah, “Toward safe dose delivery in plasma medicine using projected neural network-based fast approximate NMPC,” in Proc. IFAC World Congress, pp. 5353–5359, 2020.

11. H. H. Nguyen, T. Zieger, S. C.Wells, A. Nikolakopoulou, R. D. Braatz, and R. Findeisen, “Stability certificates for neural network learning-based controllers using robust control theory,” in Proc. American Control Conference, in press, 2021.

12. A. Nikolakopoulou, M. S. Hong, and R. D. Braatz, “Feedback control of dynamic artificial neural networks using linear matrix inequalities,” in Proc. IEEE Conference on Decision and Control, pp. 2210–2215, 2020.

13. A. Nikolakopoulou, M. S. Hong, and R. D. Braatz, “Output feedback control and estimation of dynamic neural networks using linear matrix inequalities,” in Proc. American Control Conference, in press, 2021.