(133f) On the Fast Fluidized Bed Simulations Using Recurrence CFD

Fluidized beds are systems with a bed of granular particles initially resting on a perforated bottom plate. When the inlet fluid is passed upwards through the bottom plate, it suspends, or fluidizes, the particles to allow a high effective contacting process. This contributes to the chief advantage of fluidized beds in providing a rigorous mixing and favorable heat and mass transfer characteristics. With these attributes, fluidized beds have occupied a high-ranking position in the chemical processing applications, such as coating, drying, coking of fuels and polymers. The dynamics in fluidized beds includes numerous physical phenomena because of the multiscale interactions between the solid-gas phases. For example, the gas flow interacts with the solid particles by the interstitial gas drag, and the particles among themselves, induce kinetic, collisional and frictional stresses. These (spatial and temporal) small-scale (microscopic) interactions compose to larger (meso-and-macroscopic) interactions which take place between clusters of solid and gas bubbles. The temporal evolution of the (microscopic) particle's collision dynamics lasts about few milliseconds, much shorter than the bubble evolution. While the bubbles in turn, move faster than the gas slugs in the bed and phenomena as heat transfer and chemical conversions can take minutes or hours. Due to this variety in motion scales, the numerical simulations of fluidized beds are limited to a short-term duration in case of industrial-size reactors. Therein, the computational cost needed becomes extremely huge and the long-term processes as chemical species transport (like a passive scalar) is out of an appropriate reach. In order to overcome such shortcomings, an innovative approach of recurrence computational fluid dynamics (rCFD), recently introduced by Lichtenegger & Pirker [1,2], can be applied.

The idea behind of rCFD is extracting the degree of similarity between flow structures at two states, in an unsteady pseudo-periodic flow, and using a chosen global norm,

Therein, φ is an active field. By considering multiple reoccurring states, during a short-term simulation, the recurrence matrix (rMatrix) can be plotted to indicate the recurrence events of the total system in time. Afterwards, this rMatrix will be used to distinguish those recurrence statistics that can be stitched to time-advancing states, and generate a generic flow pattern upon which a passive scalar can be traced.

An infinite time-advancing recurrence path is obtained upon the rMatrix, and it’s corresponding recurrent flow patterns will be used in resolving a transport equation (Eulerian model) or evaluating fluid parcel trajectories (Lagrangian model), for a passive scalar. These (postprocessing) two models were classified as flow-based versions of rCFD [2] where the recurrent velocity patterns were essentially needed therein. However, in the so-called transport-based rCFD [3], the recurrent flow patterns are stored directly to the memory as Lagrangian shift information. Namely, if we want to trace a passive scalar, we just have to shift the scalar information from the corresponding start-cell to the receiving end-cell following inertia-less tracers, which are strictly linked to the fluid's velocity. Finally, we end up to a series of cell-to-cell shifts on the computational grid that can transport the passive scalar with a temporal jump, i.e. recurrence time-step, bigger than the full-CFD time-step, and without resolving any equations. By this way, a huge computational demand required for the full-CFD scalar simulation can be drastically reduced to few seconds using this feasible, cheap and fast passive transportation. It is evident that this procedure of information transport is ill-conservative to the mass scalar inside the domain. In this regard, a global balance between the incoming and outgoing fluxes has to be ensured by applying a proper diffusion on the passive scalar. For further details about the methodology the reader is referred to [3].

In the present work, we aim to apply the transport-based rCFD method for the simulation of chemical species transport, which can be considered as a passive scalars, in fluidized bed systems. To do so, a lab-scale bubbling fluidized bed with a lateral injection of a two species mixture has been numerically resolved on ANSYS/Fluent using Two-Fluid model approach. The preliminary outcomes have shown a feasible perdition of species transport with a good chasing to the actual full-CFD transportation, by consuming a very low computational load using this methodology. This in turn, encourages highly the application of rCFD to be the key ingredient in fast fluidized bed simulations and exploring new frontiers in the plant-scale systems. The obtained results will be presented during the meeting.

References

[1] T. Lichtenegger & S. Pirker. Recurrence CFD – A novel approach to simulate multiphase flows with strongly separated time scales. Chem. Eng. Sci., Vol. 153, pp. 394–410 (2016).

[2] T. Lichtenegger, E. A. J. F. Peters, J. A. M. Kuipers & S. Pirker. A recurrence CFD study of heat transfer in a fluidized bed. Chem. Eng. Sci., Vol. 172, pp. 310–322 (2017).

[3] S. Pirker & T. Lichtenegger. Efficient time-extrapolation of single-and multiphase simulations by transport based recurrence CFD (rCFD). Chem. Eng. Sci., Vol. 188, pp. 65–83 (2018).