(406f) Evaluation of Filtered Two Fluid Models Against Data from an Industrial Scale Fluidized Bed Reactor

Fluidized bed reactors are challenging to design and scale up, primarily due to the complex multiphase flow dynamics exhibited by these systems. Gas-particle fluidization is characterized by the formation of particle clusters and/or gas bubbles that occur on small length and time scales. These mesoscale structures have a very large influence on the behaviour of the fluidized bed, but are very computationally expensive to simulate. In response to this challenge, the filtered Two Fluid Model (fTFM) has attracted increasing research efforts in recent years.

The fTFM approach models the effects of unresolved mesoscale flow structures on a coarse computational mesh that is unable to directly resolve these flow structures. For example, the most important effect of mesoscale flow structures is to decrease the effective drag force between gas and particles, and this drag reduction can be reproduced on a coarse grid using a filtered model. Using this method, the fTFM approach can reduce the computational expense of fluidized bed reactor simulations by several orders of magnitude.

Achieving these large computational savings without a large loss in accuracy requires accurate fTFM closures. One challenge with deriving such closures is the limited availability of experimental data on a sufficiently large scale to properly evaluate model performance. This study therefore compares model predictions against experimental pressure drop data from a large industrial reactor (30 m in height and 9 m in width) with a complex gas injection zone.

Initial results show that acceptable comparisons can be achieved with fairly rudimentary fTFM closures, but significant uncertainties exist regarding the scaling of these closures to different particle/fluid properties and very large filter sizes. In this particular case, further uncertainty is introduced by the large reactor region that contains very low particle concentrations where the drag model is highly sensitive to slight changes in particle volume fraction.

Further studies will be completed using more sophisticated anisotropic fTFM closures recently developed. The results from this comparison will give valuable information regarding the main focus areas for further improvement of fTFM closures to maximize accuracy in industrial scale simulations.