Divergence detected in AMG solver: pressure coupled
 

After 15 iterations, three errors appear:
# Divergence detected in AMG solver: pressure coupled --> Decreasing coarsening group size!
# Divergence detected in AMG solver: pressure coupled --> Increasing relaxation sweeps!
# Divergence detected in AMG solver: pressure coupled --> Changing to W-cycle!

I'm doing a flow study over a bus.
Boundary conditions:
- inlet: velocity-inlet (direction z)
- outlet: pressure-outlet


Turbulence: K-Epsilon(realizable,non-equillibrium).
Intensity of turbulence: 1% at inlet and 5% at outlet

Solution method:
Pressure-velocity coupling, scheme: coupled. Second order

Is this steady state? Did you try SIMPLE? Just give some more information about your case / mesh.

Problem may as well be transient. Additionally, did you try FMG initialization?

OJ

Hello everyone,

Hope everyone is doing fine. I am modeling a 3D model of underdrain pipe flow. And I am using parallel computing. Details about the model is :

Inlet: pressure inlet with hydrostatic pressure
outlet: pressure outlet
wall and symmetry

Solver: Pressure based
Model: Standard k-e model
Solution Method: coupled
Least square cell based gradient and 2nd order upwind momentum

Time : transient

URF was all 1 except k and e. Those were 0.8

I started with 0.0015 time step. After running 5 second ( almost 35000 iteration) it gave the error " divergence detected in AMG solver: k"

so I reduced all URF to 0.7 . Then it gave the error " divergence detected in AMG solver: epsilon ". I reduced the URF to 0.5~0.6. Then it gave several error after 2.7 sec run time.

# Divergence detected in AMG solver: pressure coupled --> Decreasing coarsening group size!
# Divergence detected in AMG solver: pressure coupled --> Increasing relaxation sweeps!
# Divergence detected in AMG solver: pressure coupled --> Changing to W-cycle!

turbulent viscosity limited to viscosity ratio of 10^5 in 93000 cells.

Please find the following image for error and residuals. I tried with Flow courant number 100, but no improvement.

Before showing error it gives same mass flow rate in outlet for long time, which I assume as converged.

I am not an expert in CFD field. Trying to learn these things for my thesis. I really need to solve this. Hope someone can help me how can I get rid of these error. Thanks in advance

Regards,
Tanjina

First, you should check your mesh. If the mesh is bad or ratio is too high, the solution will be divergent and you will get AMG error.
Second, what us the value of your velocity? Also you should check the smallest grid size, and try to calculate the time step size by using CFL condition( the courant number you defined). You dont need to set the time step equal that value but cant not set to large to compare with that value (implicit method) and you must set time step equal or below that value if you use explicit method,
Last one, reduce URF to 0.1 or 0.2 for all
These are my ideal, hope it help

Hello Le Hoang,

Thanks for your reply. I was desperate for any answer. For judging the quality of mesh, which ratios should I Check? I checked the skewness from mesh metric. 0.999 is the maximum, 4.196 is the minimum and 0.216 is the average.

My model starts with zero velocity. Should I calculate the velocity of fluid using sqrt(2gh)?

Before showing the error, mass flow rate at outlet shows convergence though residuals are above 10.

Please help me to to get rid of these error. Thanks in advance.

Regards,
Tanjina

Hello Le Hoang,

I found different ratio on my Model's mesh. Aspect ratio is 14.3 at almost of the element, which should be below 5 everywhere except boundary ( thats what I found from literature). And skewness is high in different region, about 0.85 for 1.21e4 element and 0.95 for 3.85e3 element. Total element number is 632968. And I identified the region where the skewness is high. And fluent console does not give me any warning regarding mesh.I am trying to solve this though don't know how to change these ration and skewness.

But I re-run the model with this problematic mesh with Courant number 50 ( instead of 100-200 as I used earlier), and the model is running smoothly for run time 15 second ( more than 4000 iteration) and residuals are also below 1e-3. And result I found with 6% error from the experimental. Does it mean that this mesh is not problematic for this specific model, or I am just getting a random good result ?

Any suggestion will be highly appreciated. Thanks in advance for your reply.

Regards,
Tanjina