A Machine Learning-Driven Pore-Scale Network Model Coupling Reaction Kinetics and Interparticle Transport for Catalytic Process Design.

A pore-scale dual-network model is presented with kinetics (DNMK), enhanced by machine learning (ML), for efficient multiscale modeling of reaction-transport coupled catalytic processes in porous systems. In such systems, apparent catalytic performance arises from the intricate interplay between intrinsic microkinetics and inter-particle transport phenomena. By explicitly resolving these coupled effects, DNMK provides mechanistic insight into how spatial particle arrangements and transport limitations govern overall reactor performance. A key innovation of this work is the integration of ML-based surrogates to accelerate the microkinetic module, effectively bridging the large spatial and temporal scale mismatches between transport and catalytic reactions. This hybrid approach achieves up to a 750-fold computational speed-up while preserving full physical and chemical fidelity. The framework is demonstrated for sorption-enhanced CO2 hydrogenation to methanol, where DNMK identifies optimal catalyst-sorbent configurations that maximize apparent activity and reactor-scale performance. More broadly, DNMK establishes a high-resolution, ML-driven platform for digital catalytic experimentation, enabling predictive, in silico optimization of catalyst scaling, utilization, and process intensification. By allowing rapid, physically consistent evaluation of complex catalytic systems, DNMK reduces reliance on costly experimental trials and opens new pathways for data-driven reactor and process design across diverse chemical engineering applications.