Divergence of Continuity Residual

  • 52 Views
  • Last Post 27 August 2019
Joshuakoekt posted this 26 August 2019

Hi everyone!

I am a very very new user of Fluent and am running a simulation of a simple 3D slugging fluidized bed on ANSYS Fluent 2019 R2. I have tried looking around the forum but there does not appear to be any fast solution to my troubles. The following are the "settings" I used in my simulation.

Geometry

  • Cylindrical riser of height 2m and diameter 0.1m.

Meshing

  • Face mesh with 1e-04m

Models

  • Eulerian model

Materials

  • Fluid-air (primary phase) --> properties left very close to default values 
  • Fluid-sand (secondary phase) --> granular
    • diameter: 1.5e-04m
    • density: 2660 kg/m3
    • viscosity: left as air viscosity as I understand the value is unimportant for Two-Fluid Models in Fluent.

Boundary Conditions

  • inlet:
    •    gas phase: 
      •    velocity: 0.04638m/s in (0,1,0) direction
    • particulate phase:
      •    velocity: 0m/s
  • outlet
    •    default
  • wall 
    •    gas phase:
      •    non-slip
    • particulate phase
      •    specified shear: 0.5 coefficient

Initialisation

  • hybrid initialisation
  • Patching: x,z = (-5e-2, 5e2); y = (0,1), solid volume fraction = 0.6.

Autosave every 100 time steps

Calculation

  • time step: 0.0001s
  • maximum iterations per time step = 500
  • convergence criteria: 1e-3

My problem is this: initially, everything was converging very quickly each time step. However, after approx 1000+ time steps the continuity residual started increasing while the other residuals continued to converge reasonably. 

I read from some papers that the grid size used should 3-4 times the diameter of the particles, but when I tried this scale the meshing took an unreasonably long amount of time and I settled for a much coarser mesh size.

May I know how I can resolve the divergent continuity residual issue? 

cvelikonja posted this 27 August 2019

What's your pressure-momentum under-relaxation factors like? Sometimes if these are too low, a simulation may appear to converge, but then will diverge over time. If yours are low and your simulation is converging in the first 1000 time steps, you could also try increasing these controls after the 1000 time steps. This can help your simulation converge after it initial results are created. Especially as you are working with a fluidized bed, it may be a good practice to start with low pressure/momentum/species URFs, and then increase them every couple 1000 iterations. 

Also, what order of magnitude does your continuity residual reach before it starts increasing?

Close