(341n) Simultaneous Optimization of Membrane Design, Process Design, and Process Operation

Commercially viable water treatment process plants are a prerequisite for the sustainable supply of safe drinking water. However, for a given separation task, process efficiency highly depends on the interplay of several key factors: (1) the selected membrane, (2) the process design, and (3) the operating point. However, in the conventional development of membrane processes, decision making is still decoupled [1,2]. This has inherent shortcomings as process optimization is usually limited by fixing the critical component of the plant – the membrane.

We propose a novel methodology that simultaneously performs the optimization of membrane synthesis protocol, process design, and process operation [3-5]. The method builds on our developed modeling approach [3-5] and our developed optimization method [6].

A mechanistic model mapping the membrane synthesis protocol to its process performance has not been available in the previous literature. We develop a hybrid mechanistic/data-driven membrane model [3-5]. The model uses an extensive experimental data set for novel ion separation membranes synthesized with a layer-by-layer (LbL) fabrication methodology. It combines artificial neural networks with physical mass transport models. The hybrid model predicts ion retention and water flux values based on membrane synthesis protocols [3,4]. This membrane model is integrated into superstructure process models including cost correlations [5].

The optimization of superstructure process models with artificial neural networks embedded results in large-scale mixed-integer nonlinear problems that are difficult to solve to global optimality. Recently, we found that a reduced-space formulation is favorable for optimization problems with artificial neural networks embedded [6] and process flowsheeting [7,8]. We solve the proposed optimization problems to guaranteed global optimality using our open-source solver MAiNGO [9] and machine-learning modeling library MeLOn [10].

A multi-objective optimization approach reveals the inherent trade-off between the conflicting objectives minimal annual operation costs and minimal permeate impurity. The presented method enables the tailoring of membrane fabrication to specific separation tasks. The results outperform conventional design approaches where membranes are selected among a set of commercial membranes. In future work, the methodology has the potential to accelerate the development of new membrane materials through active learning.

[1] Ulbricht, M. (2020). Design and synthesis of organic polymers for molecular separation membranes. Current Opinion in Chemical Engineering, 28, 60-65.

[2] Nunes, S. P., Culfaz-Emecen, P. Z., Ramon, G. Z., Visser, T., Koops, G. H., Jin, W., & Ulbricht, M. (2019). Thinking the future of membranes: Perspectives for advanced and new membrane materials and manufacturing processes. Journal of Membrane Science, 117761.

[3] Rall, D., Menne, D., Schweidtmann, A. M., Kamp, J., von Kolzenberg, L., Mitsos, A., & Wessling, M. (2019). Rational design of ion separation membranes. Journal of membrane science, 569, 209-219.

[4] Rall, D., Schweidtmann, A. M., Aumeier, B. M., Kamp, J., Karwe, J., Ostendorf, K., Mitsos, A. & Wessling, M. (2020). Simultaneous rational design of ion separation membranes and processes. Journal of Membrane Science, 600, 117860.

[5] Rall, D., Schweidtmann, A. M., Kruse, M., Evdochenko, L., Mitsos, A., & Wessling, M. (2020). Multi-scale membrane process optimization with high-fidelity ion transport models through machine learning. Submitted to Journal of Membrane Science.

[6] Schweidtmann, A. M., & Mitsos, A. (2019). Deterministic global optimization with artificial neural networks embedded. Journal of Optimization Theory and Applications, 180(3), 925-948.

[7] Bongartz, D., & Mitsos, A. (2017). Deterministic global optimization of process flowsheets in a reduced space using McCormick relaxations. Journal of Global Optimization, 69(4), 761-796.

[8] Schweidtmann, A. M., Huster, W. R., Lüthje, J. T., & Mitsos, A. (2019). Deterministic global process optimization: Accurate (single-species) properties via artificial neural networks. Computers & Chemical Engineering, 121, 67-74.

[9] Bongartz, D., Najman, J., Sass, S., & Mitsos, A. (2018). MAiNGO: McCormick based Algorithm for mixed integer Nonlinear Global Optimization. In Technical Report. Process Systems Engineering (AVT. SVT), RWTH Aachen University. https://git.rwth-aachen.de/avt.svt/public/maingo

[10] Schweidtmann, A. M., Netze, L., & Mitsos, A. (2020). MeLOn - Machine Learning Models for Optimization. https://git.rwth-aachen.de/avt.svt/public/MeLOn