CFD Analysis of the Influence of Particle Characteristics on the Hydrodynamics of a Bubbling Fluidized Bed

Fluidized beds find their applications in various industries due to higher heat and mass transfer rates. A relevant domain of the wide-range of application is biomass-gasification. However, the hydrodynamics of fluidized beds are very complex and are difficult to quantify. Barriers to obtain enough experimental data include very high cost of pilot-scale reactors and limitation of experimental techniques to non-invasive ones due to opaqueness of dense gas-solid regimes. While there have been much investigations into hydrodynamics of two-phase flows, there have been fewer studies on mixing/segregation phenomenon of three phase flows involving binary mixtures of biomass and sand in a fluidized bed, and even more few for heat transfer. Another problem is the scale-up of reactors. When moving to large-scale fluidized bed, a set of parameters e.g. hydrodynamics (particle and bubble size and distribution), heat transfer change drastically, which in turn have strong negative impact on the operation of the fluidized bed. Computational fluid dynamics provides an alternative way to analyze the gas-solid and solid-solid behavior inside the fluidized beds.

For this purpose, 2D and 3D eulerian-eulerian approaches were used to perform simulations on a laboratory-scale bubbling fluidized bed for glass beads bed and validated against the experimental results [1]. The CFD simulations were performed on ANSYS FLUENT 17.2 using finite volume approach for 50 s. Gidaspow’s drag law suitable for dense systems was used to measure gas-solid momentum exchange coefficient. Influence of restitution coefficient e was investigated on the hydrodynamics by varying e from 1 (perfectly elastic collisions) to 0.95, 0.9 and 0.8. Also, the influence of sphericity (ϕ = 0.95, 0.9, 0.85) of particles on the hydrodynamics was investigated. Particle diameter d was replaced in the relation for particle-air exchange coefficient by effective diameter d_eff given as d = ϕd_eff. Preliminary results from both 2D and 3D simulations at e = 0.9 and ϕ =1 agreed in terms of time-averaged pressure drop across the bed, bed expansion and void fraction. Therefore, only 2D simulations were considered for further studies.

At e = 1, there was no bubble formation. As e was reduced to 0.95, bubbles started to form with size slightly increasing for e = 0.9 and 0.85. Also, there were not much variations in granular temperature for e = 0.85, 0.9 and 0.95. So, for further investigations in to the hydrodynamics for bed expansion, void fraction and pressure drop, e = 0.9 was chosen.

Initial simulations considering ideal spherical particles and an initial void fraction of 0.373 showed lesser time-averaged bed expansion and higher pressure drop across the bed, possibly due to the irregular shapes of particles, which allows for more space in between them and hence higher initial volume fraction of air inside the bed. This leads to lower mass of the bed compared to ideally spherical particles for the same bed height. Therefore, bed comprising non-spherical particles have lesser pressure drop and require low minimum fluidization velocity to fluidize the bed. The results revealed the strong dependence of hydrodynamics on the particle shape. Numerical results of time averaged pressure drop and bed expansion matched the experimental results for ϕ = 0.9 and 0.85. Glass beads are particles of uniform density, but biomass is a non-spherical particle as well as porous. Porosity of biomass indicates the voids inside the particle, thereby leading to assumption of wrong initial conditions required for CFD modeling e.g., initial void fraction of the bed and the bed height.

In the final paper we expect to show results concerning heat transfer and mixing/segregation phenomenon in a binary mixture of biomass and sand considering the particle shape and porosity. The objective is to perform numerical study mimicking real conditions, in order to have a proper understanding of the hydrodynamics. Initial results reveal deviations in the numerical and experimental results for bed expansion, pressure drop and solids distribution at various heights from the distributor plate, which initially show some dependence on the bulk density of bed. Further investigations will be carried out to account for the differences in the simulations.