Simcenter STAR-CCM+ Example of Simulating Baffled fluidized bed

Introduction
Fluidized beds are frequently employed in the process industry for applications such as gas-solids mixing combined with heat/mass transfer.  Fluid catalytic cracking (FCC) units are very common in the Oil and Gas industry. Efficient performance of the fluidized bed is dependent on the flow hydrodynamics and consequent bubble behavior. Internals are often placed inside the bubbling fluidized bed for achieving better performance by increasing the residence time and contact area between gas and solids.

Geometry :
We consider here a 2d geometry having gas spargers at the bottom of the bed and inlets at the top for catalyst particles entering the domain. This catalyst comes from the dipleg of cyclones. It has also 2 baffles placed 45 degree to horizontal plane as shown in the figure below. Dimensions of the column is 1m*0.28m.
                                                                  

Mesh
Trimmer mesh model has been used to generate volume mesh; polyhedral mesh can be used as well. Total cell count is ~31000. General guideline on cell size for particulate flows is that the mesh size should be smaller or equal to a value ten times the particle diameter (but is not possible always), especially for fine sized Geldart A particles (90um). So we can have a mesh size as close as possible to the recommendation based on the resources and accuracy we need to achieve.
In the current simulation, we have maintained a cell size of around 3mm, as shown below.
                                                                          
                                                                           
Boundary Conditions
We have considered a gas velocity of 1m/s (just above the minimum fluidization velocity for the conditions used) for the particles to act like fluid. Catalyst particles of size 90um fall at the rate of 0.5m/s from the dipleg and volume fraction at the catalyst inlet equals 0.55. Whole bed is initially filled with catalyst up to 0.75 m height with the volume fraction of catalyst being 0.3.
We have used phase-specific boundary conditions: top boundary allows the air to flow out but is impermeable for the particulate phase, while bottom boundary allows the particles to exit the domain but is impermeable to air phase. Slip condition has been assigned at the walls for the interactaction of the particulate phase with wall boundary.

Below image shows the initial particle fraction and BC's of top and bottom faces.


Physics
We have chosen the Granular flow approach to model the particulate flow assuming air as the continuous phase and particles as the discrete phase with collisions between particles modeled using the Granular temperature model. Please find below snapshots of the  physics model selection:


                  Fig:1                                                             Fig:2                                                               Fig3

• Once we select the Granular Pressure model under EMP, Granular temperature and Granular temperature transport model automatically get selected as shown in Fig:1
  1. Under Granular temperature transport, we have two options for particulate phase: kinetic viscosity and granular diffusion coefficient. Gidaspow or Syamlal models can be selected based on our requirement.
• Sub models for Air phase and Particle phase is shown in Fig:2
  1. Initial conditions for velocity and volume fraction can be set under sub models of the Eulerian phase.
• Once the above model are selected, we have to assign the continuous and dispersed phases under phase interactions as shown in Fig:3. Particles must be chosen as the dispersed phase.
  1. There are a variety of drag models available for different scenarios: for the current simulation we have selected the Gidaspow model as it is widely used for fluidized beds

                                                                             

Solver settings
It's important to have the right combination of settings (such as URF's, convection schemes and timestep) for the different solvers; the actual values may have to be adjusted based on the volume fraction of particle phases. For the current particle fraction (0.3), a timestep of 5e-4 sec is sufficient to resolve particle-particle collisions. However, for particulate fraction values of ~ 0.6 i(which is close to the packing limit), we have to start with a lower timestep (~1e-4s) and slowly ramp up to higher values depending on the monitoring criteria.
We recommend to use a reasonable lower urf combination initially along with 1^st order schemes (for volume fraction and velocity), slowly ramp the values and finally change to 2^nd order schemes once we get a stable solution. This is a general practice to be followed for granular simulations.

Results
Following is the animation of volume fraction of particles and velocity of air; influence of placement of baffles on the fluidization pattern is clearly discernible. Total run time is around 1.8 hr to complete a total simulation time of 2 sec on a 10 core machine.
https://videos.mentor-cdn.com/support/videos/5400/435079b3-8aa3-4d5e-9e5d-82a303cc31ad-en-US-video.mp4
See also :
Rotary Fluidized Bed Simulation
When the number of discrete particles in a mixing problem gets excessively large, do I use the Granular Model or DEM?
Simulating Physics > Multiphase Flow > Using the Multiphase Segregated Flow Model > Modeling Phase Interactions in Multiphase Segregated Flow > Modeling Drag Force > Drag Coefficient Reference