Analytical scaling of DEM particles for efficient packed-bed simulations

Abstract

Paper presented to the 10th International Conference on Heat Transfer, Fluid Mechanics and Thermodynamics, Florida, 14-16 July 2014.

Packed and Fluidized beds are commonly found in industries such as chemical processing and refining. A major advantage to these configurations lies in the large solid-surface area exposed to the flow, allowing for rapid interaction between the solid and fluid phases. While these types of flow configurations have been heavily studied over the years, computational software and hardware are only recently becoming advanced enough to allow realistic simulations of industry relevant configurations. Recent developments have allowed for the coupling of Discrete Element Modeling approaches, where conservation equations are typically solved on a particle-by-particle basis, with traditional continuum-fluid dynamics simulations. Nonetheless, when fluid-particle interaction is important, such as in packed bed analysis, modeling of the individual particles may still be computationally prohibitive except for simple applications. Methods to improve computational costs include grain coarsening or parcel-based approaches, where particle sizes may be scaled up or groups of particles are treated statistically. The present study develops and validates an analytical approach for the scaling of the Coefficient of Drag equations in a simplified packed bed simulation with scaled-up particles, using CD-adapco’s STAR-CCM+ software. Pressure drop predictions are compared against the accepted Ergun Correlation for the high density cylindrical packed bed. Container contact forces and packed bed height are also monitored as the flow rate is increased toward fluidization. It is shown that by properly scaling the Coefficient of Drag, while doubling the particle diameter (effectively reducing the total number of simulated particles by 8), a more than 15X speed-up in simulation time is achieved. This speed-up is achieved with an increase in error of only 8% maximum for the cases studied. Additionally, similar physical behavior is observed between the cases. This analytical approach proves to be a robust method of reducing computational expense without sacrificing accuracy, effectively making industrial scale simulations feasible.