Industrial-Scale Simulation Technology

Paul Kieckhefen, M.Sc.

Motivation

The design of novel apparatuses can benefit from numerical simulations. These allow for greater insight into the process, rapid evaluation of design variants and enable scale-up after lab-scale vavalidation, reducing the need for costly experiments.

Methodology

The state of the art in granular flow simulation is the Computational Fluid Dynamics-Discrete Element Method (CFD-DEM) which allows for the accurate description of all granular flow phenomena. The fluid phase is described using the Navier-Stokes equations while the granular phase is described using discrete particles obeying the Newtonian equations of motion, with momentum exchange modelled by a drag closure.

The degree of accuracy and applicability requires the resolution of small-scale phenomena like particle collisions and clustering which result in small timesteps and high computational demand. To allow for the treatment of large-scale systems, coarse-graining techniques are applied that attempt to reproduce the bulk behavior by tracking representative computational parcels instead of all individual particles. The Lagrangian nature of CFD-DEM preserves particle identities and enables tracking of physical or chemical processes occurring within the particle, thereby predicting product quality.

Based on these highly-resolved simulations, time extrapolation can be performed using Recurrence CFD (rCFD). This novel approach captures pseudo-periodic flow patterns like bubbling or spouting to allow for the long-term description of transport processes within the system and its constituents.

Objectives

  • Implementation of Recurrence CFD in OpenFOAM.
  • Development and implementation of a new communication scheme for scaling the CFD-DEM software CFDEMcoupling to hundreds of cores.
  • Modelling of heat and mass transfer within pilot scale fluidized and spouted bed granulators.
  • Investigation of intra-particle transport processes over the course of granulation. 

Results

An implementation of the rCFD method was developed and applied to two promising target cases: spray coating and residence time behavior prediction. The particle dynamics is not influenced by the deposition of a thin coating layer in the first case and by the motion of particle through the system in the second one.

For spray coating, a batch spouted bed equipped both with and without stabilizing draft plates was first simulated using state-of the art CFD-DEM and subsequently extrapolated in rCFD. Here, spray parcels were injected and droplet deposition was modelled using a filter correlation, stripping mass from the parcels. The particle surface percentage coated is estimated using a statistical approach that obeys the asymptotic behavior of increasing overlapping droplet impacts. The surface coverage distribution after 1 h of coating is shown in Fig. 1. The unstabilized system shows much more homogenous coating due to inhibited mixing in the system with draft plates. The present rCFD implementation achieved a speedup of 2100x over pure CFD-DEM, without which the study at hand would not have been feasible.

Selected Publications