CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   Error: Divergence detected in AMG solver: pressure correction (http://www.cfd-online.com/Forums/fluent/117708-error-divergence-detected-amg-solver-pressure-correction.html)

wanna88 May 14, 2013 03:09

Error: Divergence detected in AMG solver: pressure correction
 
Hi All,

I would like to ask some opinions/suggestions on my problem.

Basically my project is the simulation of the multiphase, non -newtonian and turbulent flow of blood vessel inside the aortic disease.

The mesh number that I am using is approximately 600000. I used ICEM CFD for meshing (prism/tetrahedral mesh).
I have run the simulation for steady state (initial condition) for 300 iterations and no error have been detected. Besides, I have reduced the number of relaxation factor (pressure=0.1; momentum = 0.1, turbulent viscosity = 0.5) too for the simulation.

Before starting the transient/unsteady simulation, I checked the case.
There is a message:

" Maximum cell skewness is greater than 0.98. Consider improving the mesh quality before proceeding with your simulation.
Skewness exceeded 0.98 at 71 cells on thread ID = 1.Skewness exceeded 0.98 at 71 cells on thread ID = 1 "

I tried to ignore the message and run the simulation of transient simulation with the fixed time stepping method for 2 cycle:
a) Time step size : 0.02
b) Number of time step :100 (100 x 0.02 = 2 s)
c) Max iteration / time step : 20

BC : Inlet= mass flow inlet; mass flow profile, outlet= outflow

I successfully run above step without any errors with approximately 2550 iterations.
Based on the paper, I need to run 5 cycle.

Then, I changed the time step size to 0.01 and number of time step : 100 (1 s) to run for another 1 cycle

At 2572 iterations, the error come out:
"Error: Divergence detected in AMG solver: pressure correction"

May I know, is the error come out because of :
1) Mesh quality? but I have successfully run the simulation for about 2500 iteration
2) Change the time step size?

Any opinions about this?

Hopefully someone can help me to answer this question. I have been tried to solve this problem for quite sometimes.

Thank you so much.

Naimah

RodriguezFatz May 14, 2013 05:47

I am pretty sure it is the mesh. Show us the geometry and your mesh! Maybe you can improve it. ICEM can do much better...

wanna88 May 14, 2013 07:00

1 Attachment(s)
Hi,

Thank you for your help.
Attached is the picture of the geometry that I have used for the simulation.

Besides, Is that any probability the error came from the max iterations/time step being used?

I only set the max iterations/time step to be 20.

Thank you

wanna88 May 14, 2013 07:07

1 Attachment(s)
Attached is the cross sectional plane of the mesh in ICEM CFD.
When I looked at the left column section of the ICEM CFD, I found out that the volume consist of :
a) Tetrahedral
b) Prism
c) Pyramid

Is this normal?

RodriguezFatz May 14, 2013 07:08

Yes, this is always a possibility.

Is this the whole domain? The grid doesn't look very fine... how comes you have 600000 cells? (or what do you mean by "mesh number"!?)
Can you upload the geometry and mesh files?

Edit: to your second post: Yes, you have prism mesh at the walls, tetra in the middle and pyramids as "connectors".

wanna88 May 14, 2013 07:22

3 Attachment(s)
Mesh Size

Level Cells Faces Nodes Partitions
0 1773753 4116899 663019

Above is the data that I get from Fluent software.
Am I using the wrong word for mesh elements? I am a beginner in CFD, so may I know what is the correct representative for the mesh elements, is that nodes, cells or faces?

Attached is the full geometry image and some part of the mesh.

RodriguezFatz May 14, 2013 07:27

Hi, Fluent stores variables in the cell centers. Therefore, I think the number of cells is the relevant value, because it is directly linked to the actual number of different calculated velocities, pressures, ...

I don't understand the above numbers. How can the number of cells (1.7mio) be larger than the number of Nodes (663019) ? Each cell has several nodes...

wanna88 May 14, 2013 07:31

Hi,

Thank you for the infromation
Actually I do not know about this. I just took the value directly from Fluent.
Mesh> Info> Size
Am I doing the wrong thing?

RodriguezFatz May 14, 2013 07:38

It is hard to emagine how this could work... I use hexa meshes only. Maybe you should start a thread in the ICEM / Meshing forum!? Could you upload the geometry or is it restricted by copyright?

wanna88 May 14, 2013 07:42

Hi,

About the geometry, I cannot upload it since it is a restricted copyright.
Thank you for the suggestion.

I just started to think about the cell and nodes. Thank you for raising this thing.

RodriguezFatz May 14, 2013 07:56

From the few pictures it looks like your prism mesh is really coarse and skewed.
Go to ICEM and let ICEM show you the bad spots. Then improve the mesh until the skewness is better. Also I would try to get a much smaller stable mesh running and after it is stable to refine it. The geometry doesn't look that complicated. For first runs, just use newtonian fluid with some simple turbulence model.

wanna88 May 14, 2013 21:55

Hi,

Thank you for the suggestion.
Is that mean do I need to run the Newtonian fluid in the steady state condition?
What is the simple turbulent flow?
Currently I am using k-omega model.

RodriguezFatz May 15, 2013 02:15

k-omega is simple, yes. Reynolds-stress would be complicated for example...
Do you have unsteady boundary conditions or why do you want to finally run it time dependent?

wanna88 May 15, 2013 02:19

Hi,

I have mass flow inlet profiles (unsteady boundary condition).
So, in the transient simulation, do i need to run the simulation by implementing the profile on the BC or do i need to set up with 1 constant number first?

RodriguezFatz May 15, 2013 03:02

I would go for a constant number first. If the error is caused by the mesh you want to keep the simulations as short as possible for debugging.

niloogh May 23, 2013 19:07

Quote:

Originally Posted by RodriguezFatz (Post 427432)
From the few pictures it looks like your prism mesh is really coarse and skewed.
Go to ICEM and let ICEM show you the bad spots. Then improve the mesh until the skewness is better. Also I would try to get a much smaller stable mesh running and after it is stable to refine it. The geometry doesn't look that complicated. For first runs, just use newtonian fluid with some simple turbulence model.

hi philipp
i had this problem too"Error: divergence detected in AMG solver"
my mesh dos n,t have good quality cause it is complicated in some places
it has some faces with periodic boundary condition.when i use symmetry in stead of periodic i don.t have any problem with fluent but when i use periodic i receive this message;Error: divergence detected in AMG solver
(i,m forced to use periodic)

would you plz help me????

wanna88 May 23, 2013 23:09

Hi,

Do you have any experiences handle multiphase flow case?
I have try to run the simulation in order to detect the cause of error in Fluent I am using the same mesh element for turbulent, single phase case and turbulent, multiphase and non newtonian case.
The simulation can successfully run (both for steady and unsteady/transient) for single phase turbulent flow but it failed for multiphase, non newtonian tubulent flow?

Is this because of my mesh or because of the multiphase flow?
Is there any suggestion for multiphase flow?

Thank you.

Dear niloogh

Hopefully you can get some information regarding the mesh based on this link:
http://www.cfd-online.com/Forums/ans...h-quality.html

Regards,
Naimah

wanna88 May 23, 2013 23:11

Quote:

Originally Posted by RodriguezFatz (Post 427432)
From the few pictures it looks like your prism mesh is really coarse and skewed.
Go to ICEM and let ICEM show you the bad spots. Then improve the mesh until the skewness is better. Also I would try to get a much smaller stable mesh running and after it is stable to refine it. The geometry doesn't look that complicated. For first runs, just use newtonian fluid with some simple turbulence model.

Hi,
May I know how to repair the bad elements at the location shown by ICEM?

RodriguezFatz May 24, 2013 01:51

Quote:

Originally Posted by wanna88 (Post 429661)
Hi,
May I know how to repair the bad elements at the location shown by ICEM?

I only use ICEM hexa / blocking... maybe others can help here.

Amna April 6, 2016 02:57

DDPM issues
 
hi everyone

i m working on gas-solid fluidized bed.Geometry is a simple 2D cylindrical column.During simulation as first step , i kept gas velocity equal to zero & selected surface injections in DDPM to inject solids from bottom of column i.e from gas inlet.

i have to inject 800 particles which will give 0.25 m static bed height (as per literature). but in my case static bed height i just get 0.125 m upon injection. Kindly let me know how could i get the required 0.25 m static bed height?

secondly, just to see the phenomena at 0.125 m static bed height , i gave 1.4 m/s gas velocity, i didn't get a proper fluidized bed rather some random solid fraction from inlet-outlet obtained, which isn't matching to any of fluidization stage.

Kindly tell me how could i get a proper fluidized bed by using this euler-Lagrangian approach?

Hope to get a prompt reply ......


All times are GMT -4. The time now is 16:14.