(544g) Incorporating Materials Surrogate Models into Process Models for Adsorption-Based Gas Seperations

Advanced functional materials and novel process systems are two main research aims in transitioning to sustainable energy1. Currently, materials and processes research are largely in parallel, even though they are closely integrated in energy systems and need to be considered simultaneously for optimal performance/economic objective2. There exist research opportunities for developing methodologies and applications for process-materials co-optimization. One potential application area is adsorption-based gas separation technologies. Many novel process development efforts (e.g., pressure swing adsorption, PSA) simply ignore adsorbent materials’ design space3. On the other hand, advanced adsorbent developments (e.g., metal-organic frameworks, MOFs) rarely consider process-level performance4. This work proposes to tackle this research problem by incorporating materials surrogate model with equation-oriented process model, enabling direct process-materials co-optimization.

We develop and illustrate our proposed methodologies with PSA process and MOFs adsorbents. On the first step, we developed an automatic pipeline for learning materials surrogate models from user-input MOFs structure files. The pipeline starts by initiating molecular simulation runs to get adsorption isotherm data points of input structures and then regress those data points to obtain algebraic isotherm model parameters. Meanwhile, a pool of molecular descriptors is computed that describe MOFs design space. The quality of surrogate model depends on the selection of features and data points. We thus build machine learning models to facilitate feature selection and data points filtration. Then we learn an algebraic surrogate model that satisfactorily links the MOFs descriptors (i.e., features) with the isotherm model parameters (i.e., labels). Those materials surrogate models can thus be easily integrated into equation-oriented process models. We illustrate this by constructing a customized fidelity-tunable PSA column unit model via IDAES-PSE framework5 and incorporate the learned surrogate model with the process model. The resulting integrated model have both materials descriptors variables and process decision variables, thus enabling process-materials simultaneous simulation, design and optimization.

[1] Chu, S. and Majumdar, A., 2012. Opportunities and challenges for a sustainable energy future. nature, 488(7411), pp.294-303.

[2] Farmahini, A.H., Krishnamurthy, S., Friedrich, D., Brandani, S. and Sarkisov, L., 2018. From crystal to adsorption column: challenges in multiscale computational screening of materials for adsorption separation processes. Industrial & Engineering Chemistry Research, 57(45), pp.15491-15511.

[3] Biegler, L.T., Jiang, L. and Fox, V.G., 2005. Recent advances in simulation and optimal design of pressure swing adsorption systems. Separation & Purification Reviews, 33(1), pp.1-39.

[4] Daglar, H. and Keskin, S., 2020. Recent advances, opportunities, and challenges in high-throughput computational screening of MOFs for gas separations. Coordination Chemistry Reviews, 422, p.213470.

[5] Miller, D.C., Siirola, J.D., Agarwal, D., Burgard, A.P., Lee, A., Eslick, J.C., Nicholson, B., Laird, C., Biegler, L.T., Bhattacharyya, D. and Sahinidis, N.V., 2018. Next generation multi-scale process systems engineering framework. In Computer Aided Chemical Engineering (Vol. 44, pp. 2209-2214). Elsevier.