Achieving carbon neutrality and reliable scale-up of chemical reactors are among the major challenges in modern chemical engineering. Gas-solid multiphase flows are widely involved in fluidized bed reactors, packed bed catalytic systems, and the thermochemical conversion of biomass and plastics. Because most solid particles in these applications are non-spherical, understanding their complex, multi-scale behavior is difficult. With growing demand for predictive design, process intensification, scale-up, and real-time digital twins, multi-scale simulations that bridge particle and reactor scales are necessary.
The discrete element method-computational fluid dynamics (DEM-CFD) approach is a powerful tool for resolving particle-level collisions, hydrodynamics, heat, and mass transfer. However, extending this to non-spherical particles—specifically using the superquadric equation (SuperDEM)—brings extra challenges due to the highly coupled effects of hydrodynamics, heat, mass transport, and chemical reactions. The researchers systematically review the state of this technology and identify four critical bottlenecks that currently restrict its industrial application: anisotropic contact hydrodynamics and numerical singularities at sharp edges, incomplete heat transfer frameworks, immature non-spherical chemical reaction models, and high computational costs for large-scale particle systems.
To address these challenges, the article proposes several targeted remedies. To resolve contact mechanics issues, they suggest a reduced-space Newton iteration method to circumvent numerical singularities. For heat transfer, they recommend developing generalized shape-factor Nusselt correlations. To model chemical reactions accurately, they propose a shrinking core reaction model based on superquadric coordinates. Additionally, to handle large-scale simulations, the authors point to the synergy of coarse-grained particle models combined with graphics processing unit (GPU) acceleration.
The article also elaborates on the transformative potential of artificial intelligence (AI). The authors highlight that deep learning can construct universal drag and heat transfer correlations from direct numerical simulation (DNS) data, generate interpretable chemical reaction kinetics, and accelerate stiff solvers to cut simulation overhead. Modern AI tools also have the potential to assist in the analysis of convergence and the optimization of stability when developing mathematical algorithms.
The authors note that these proposed pathways deliver clear development roadmaps for the next-generation multi-scale SuperDEM-CFD platform. By overcoming current barriers, this technology will support the predictive design and real-time digital twin construction of gas-solid reactors, ultimately aiding in the transition toward more efficient and sustainable chemical processes.
DOI
10.1007/s11705-026-2705-z