Towards predictive modeling of multiphase flow in fluidized beds using coarse-grained CFD-DEM

Fluidized beds are widely used in the energy and chemical process industries due to their excellent heat and mass transfer properties. Their multiphase nature leads to inherently unstable flow behavior and complex dynamics that span multiple time and length scales. On the microscale, particle collisions and particle–gas interactions dominate. On the mesoscale, bubble and cluster formation occur, and on the reactor scale, macroscopic flow structures emerge. These multiscale phenomena pose major challenges for both experimental investigations and numerical modeling. Among available simulation methods, the coupled CFD-DEM offers detailed resolution of particle–fluid interactions by explicitly tracking particle motion while solving the fluid phase on a continuum scale. Although powerful, its computational cost limits applications to laboratory-scale systems with relatively few particles. A promising strategy to overcome this limitation is the CG approach, in which several original particles are represented by larger grains with adjusted properties. This reduces the number of simulated entities significantly but requires suitable scaling laws for contact and drag forces as well as sub-grid closures. While various strategies have been proposed, their predictive accuracy and robustness across different conditions have not been systematically assessed. This thesis addresses this gap by evaluating the CG approach in monodisperse, bidisperse, and large-scale bubbling fluidized beds. Interpolation schemes are analyzed to examine how coupling strategies for CFD-DEM interact with coarse-graining. Different scaling strategies for contact and hydrodynamic forces are benchmarked in mechanically agitated and fluidized systems, providing insight into their respective strengths and limitations. Sub-grid drag models are introduced to improve predictions at higher CG factors. The accuracy of coarse-grained simulations is quantified as a function of the ratio of system diameter to grain diameter, which is a key parameter when conditions are transferred from the laboratory to the technical scale. The methodology is then extended to bidisperse systems to investigate mixing and segregation phenomena, before being applied to a large-scale bubbling fluidized bed reactor. Here, simulations are compared with experimental data to validate the approach under technical conditions. The results demonstrate that the accuracy of coarse-grained simulations depends strongly on the chosen scaling strategy, drag closure, and the underlying ratio of system diameter to grain diameter with no universal approach performing best in all situations. Nevertheless, when combined appropriately, coarse-graining and sub-grid models significantly reduce computational costs while maintaining reliable predictions of key macroscopic quantities. This work establishes a validated framework for coarse-grained CFD-DEM, advancing predictive modeling of inherently multiscale fluidized bed systems and supporting their design, optimization, and scale-up in practice.