Studying liquid injection in fluidised beds through simulations and experiments

Fluidised bed reactors (FBRs) are often used for gas-solid conversion processes where a uniform temperature distribution and solids mobility are desired. They feature low manufacturing costs and can be scaled to the desired production capacity. The vigorous bubble induced solids motion in the bed provides the mechanism for a uniform temperature distribution. One of the methods used to intensify fluidised bed operation in the presence of highly exothermic reactions is to add a liquid phase. The addition of the liquid phase can serve several purposes. For example, the liquid can be a reactant as is the case in Fluid Coking™. In Spray Fluidised Bed Granulation, the liquid acts as a carrier to deposit solid constituents for particle growth, while the liquid injection in Condensed Mode gas-phase polymerisation is used for heat management in the reactor. In the latter case, the latent heat of evaporation is used to absorb the heat liberated due to the highly exothermic reaction.

Although the liquid injection is essential in all of these examples, it also increases the complexity of the operation of the FBR. The injected liquid can affect the solids motion in the reactor and thus influence the temperature uniformity. Additionally, the liquid can induce the formation of clusters of particles through the cohesive forces introduced by the
presence of the liquid. Although the formation of agglomerates might be desired in some processes, the increased cohesion between the particles significantly affects the hydrodynamic behaviour of the bed. Ultimately, the increased cohesion between particles can lead to total defluidisation of the bed. To improve the operating conditions and the efficiency of these processes, it is critical to improve the understanding of the effects of liquid injection.
To model the liquid injection in fluidised beds, using the a Discrete Element Method (DEM), assumptions are required regarding the outcome of the encounter of a particle and droplet. A common assumption is that after collision the droplet forms a thin liquid layer covering the entire particle. For a better understanding of this phenomenon, the complete process of droplet-particle interaction was studied. More specifically, the spreading behaviour of a droplet on a spherical surface was studied using Direct Numerical Simulations.
A coupled Immersed Boundary and Volume of Fluid method was used to represent the gas-liquid-solid interactions. Small viscous droplets are used to reduce interfacial oscillations as well as low impact velocities to avoid droplet fragmentation.
A decrease of spreading area with increasing curvature was observed. To quantify the initial spreading rate of the droplet, a power-law was fitted to the data, following the approach for flat surfaces. A strong contact-angle dependence was found for the pre-factor of the power law, which is expected. The exponent shows a linear decrease with increasing curvature of the surface. In addition, the liquid coverage fraction of the particle is dominated by the volume ratio of the droplet and the particle and can be accurately predicted. Additionally, by using the non-dimensional capillary inertial time, the time needed for 90% of the equilibrium spreading width can be calculated. The required spreading time corresponds to three non-dimensional time units, which can be calculated based of the droplet properties.
This provides a useful time-scale that can be compared with the collision frequency of the particles in the bed.
To study the effects of liquid injection on the bulk behaviour of the bed, fluidisation experiments were performed in a pseudo-2D fluidised bed. Firstly, the measured particle temperature distribution was used to quantify the the temperature uniformity in the bed without liquid injection. For several fluidisation regimes, the degree of temperature
non-uniformity was quantified using Infra-Red Thermography (IRT). The particle temperature distributions were obtained from the whole-field temperature data, which were characterised using the standard deviation, i.e. the width of the distribution, and skewness, i.e. the dominant temperatures in the distribution. Based on the heat loss data and bubble frequencies, the standard deviation and skewness revealed to be good indicators of the temperature uniformity for the studied fluidisation regimes.
Subsequently, the hydrodynamic and thermal behaviour of the bed was studied using liquid injection. The experimental results revealed that injecting smaller droplets has a larger negative impact on the uniformity of the bed compared to larger droplets. In addition, the injection velocity impacts the average temperature of the bed more with increasing droplet size. Besides the temperature distribution, the IRT can be used to study the agglomerates in the bed as they are visible as dark spots in the IRT-images. The experiments reveal that the formation of agglomerates is more pronounced at increased droplet size. This can be counteracted by an increased solids motion either through a higher background velocity or spout injection velocity. Lastly, the presence of agglomerates was not well reflected in the particle temperature distribution and its properties. Therefore, these are unsuitable to quantify the agglomeration behaviour of the bed.
Finally, the effect of liquid injection on a FBR with porous particles was studied to better represent the above mentioned industrial processes. In the case of porous particles, defluidisation occurs eventually at sufficiently high liquid injection rates due to density driven segregation. For particles with a higher porosity, defluidisation occurs at lower liquid
injection rates, as well for particles with a lower pore size. Similar to the agglomeration, the defluidisation due to liquid imbibition can be counteracted by increasing the background fluidisation velocity and the spout velocity.