2022 Annual Meeting

(481e) Multi-Objective Catalyst Shape Optimization for Solid-Catalyzed Gas-Phase Reactions with q-Expected Hypervolume Improvement Via Particle-Resolved CFD Simulations

Authors

Kang, W. - Presenter, Seoul National University
Na, J., Carnegie Mellon University
Lee, W., Seoul National University
Despite the widespread use of non-spherical catalysts in industrial and chemical fields, neither the study about packing structures of such particles nor characterization of those have been done that much yet. However, particle size and shape have a significant contribution on reaction properties like pressure drop, liquid holdup, and catalyst effectiveness. Since computational fluid dynamics (CFD) simulations of packed-bed reactors have gained fairly decent status in reaction engineering, investigation and analysis of particle shape could be achieved with a more effective and economical way via simulation.

This presentation describes an integrated workflow for multi-objective optimization (MBO) of optimal catalyst shape in fixed bed reactor from computational generation and meshing of arbitrary-shaped particles in randomly packed beds through to the CFD calculation. The workflow, physics engine based automated packed bed reactor generation, geometry processing with shrink-wrap method for contact regions handling, and particle-resolved CFD simulations for continuum-scale fluid dynamics to simulate solid-catalyzed gas-phase reactions (methanol, dimethyl-ether, Fischer-Tropsch synthesis) in a fixed-bed reactor. The effect of the shape of catalyst particle is investigated with sphere, cylinder, trilobe, quadrilobe, cross-web and Raschig ring. Since the discrete element method is computationally costly to be implemented, a rigid-body model in Blender that showed accurate agreement for porosity compared to DEM in the commercial software STAR-CCM+ is applied. Furthermore, MBO which use Gaussian process as a surrogate model, and q-expected hypervolume improvement (qEHVI) as an acquisition is utilized to reduce the required number of CFD simulations for optimal catalyst shape selection. The resulting Pareto front yields near-optimal particle shapes with significantly reduced simulation time compared to conventional methods reported. We anticipate that this integrated workflow can be applied not only to optimal packed-bed reactor design problems with arbitrary-shaped catalyst particles but also to an expansion of any solid-catalyzed gas-phase reactions.