(149c) Drag Model Implementation for CFD-DEM Modeling of Dense Flow with Non-Spherical Particles in Ansys Rocky

Granular-fluid systems are often characterized by the complex interaction between solid particles and fluids that involve mass, momentum, and energy exchange. Such systems are found in various industrial processes, including sedimentation, filtration, and fluidized bed flows. The use of numerical simulations such as the coupled computational fluid dynamics (CFD) and discrete element method (DEM) model has become increasingly commonplace in modeling and predicting the behavior of granular-fluid systems as it allows quantification of process characteristics that are difficult to measure experimentally. With this approach, the fluid flow is typically treated as a continuous phase using averaged Navier-Stokes equations while the particles are modeled as a discrete phase. The coupling of the two phases relies on an accurate treatment of the drag force. For dilute flows, wherein particles can be assumed to be widely spaced apart and thus there exist no significant particle-particle interactions, empirical or semi-empirical drag laws accounting for the particle aspect ratio, orientation, or roughness are often used. Most granular-fluid systems, however, cannot be considered as dilute. In such dense flows, the effects of particle shape and particle-particle interactions become significant. These interactions may include collisions and the hydrodynamic screening effect that occurs when the drag experienced by each particle in a group of closely packed particles varies as a result of the fluid momentum being ‘blocked’ by the immediately adjacent particles.

This work concerns the selection, implementation, and validation of drag coefficient correlations for dense flows comprising non-spherical particles in Ansys Rocky and Ansys Fluent CFD-DEM coupling. The hybrid Ganser - Di Felice and Hölzer & Sommerfeld - Di Felice drag models were used. These models are formulated using the dense flow drag coefficient correlation presented by Di Felice for spherical particles as the basis, but using the drag coefficient correlations proposed by Ganser and Hölzer & Sommerfeld for application to non-spherical particles.

The drag models were integrated into the numerical model using the Rocky Solver Application Programming Interface (API). The experimental data in Volmari et al. was used to validate the model. As shown in Fig. 1(a), a fluidized bed with sphero-cylindrical particles in a square cross-sectional area with superficial velocities ranging from 0.4 m/s to 2.2 m/s was modeled to match the experimental setup described in the aforementioned reference. The numerical results obtained when using the hybrid drag laws provided a more accurate pressure drop prediction when compared to the result obtained using the original Di Felice drag law model in Ansys Rocky for a wide range of superficial velocities, as shown in Fig. 1(b). In addition, an overprediction of the minimum fluidization velocity results when the particle shape is not taken into account when computing the drag force, as seen in Fig. 1(c).

These results demonstrate that combining the Di Felice drag model and the Ganser and Hölzer & Sommerfeld drag coefficient correlations provides a more accurate estimation of drag forces in dense non-spherical particle flows compared to the standard Di Felice model. Thus, its implementation in the Ansys software framework when modeling granular-fluid systems must be considered.