CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   FLUENT (http://www.cfd-online.com/Forums/fluent/)
-   -   Divergence problem (http://www.cfd-online.com/Forums/fluent/113472-divergence-problem.html)

Smaras February 20, 2013 08:18

Divergence problem
 
"Already posted this yesterday but no response that's why" :(

Hello,

I am getting this problem with the 3D VOF (air jet impinging over water). The simulation is good till 1.4e-2s but after that it gives the following error. The time step size i am using is 1e-6 with 40 time steps. The setting are as follows:

1. Double with SST
2. Surface tension 0.073 n/m
3. inlet velocity 96.672 m/s
4. PISO Scheme with VF(Mod HRIC) Momentum (2nd ord)
5. URF Pressure 0.3 Density 0.5 Body Forces 1 VF 0.5

In 2D for the same thing, i had mass loss problem but i was able to find the solution with the help of few people here.

Would be thankful for your kind reply.

Regards,
Smaras

===>>>>turbulent viscosity limited to viscosity ratio of 1.000000e+05 in 240257 cells
# Divergence detected in AMG solver: x-momentum -> Increasing relaxation sweeps!

Divergence detected in AMG solver: x-momentum
Divergence detected in AMG solver: y-momentum
Divergence detected in AMG solver: z-momentum
Divergence detected in AMG solver: vof-1
Divergence detected in AMG solver: x-momentum
Divergence detected in AMG solver: y-momentum
Divergence detected in AMG solver: z-momentum
Divergence detected in AMG solver: vof-1
Divergence detected in AMG solver: x-momentum
Divergence detected in AMG solver: y-momentum
Divergence detected in AMG solver: z-momentum
Divergence detected in AMG solver: vof-1
Divergence detected in AMG solver: x-momentum
Divergence detected in AMG solver: y-momentum
Divergence detected in AMG solver: z-momentum
Divergence detected in AMG solver: vof-1
Divergence detected in AMG solver: x-momentum
Divergence detected in AMG solver: y-momentum
Divergence detected in AMG solver: z-momentum
Divergence detected in AMG solver: vof-1
Divergence detected in AMG solver: x-momentum
Divergence detected in AMG solver: y-momentum
Divergence detected in AMG solver: z-momentum
Divergence detected in AMG solver: vof-1
turbulent viscosity limited to viscosity ratio of 1.000000e+05 in 251038 cells

Primitive Error at Node 0: floating point exception

Primitive Error at Node 1: floating point exception

Primitive Error at Node 2: floating point exception

Primitive Error at Node 3: floating point exception

Primitive Error at Node 4: floating point exception

Primitive Error at Node 5: floating point exception<<<<====

RodriguezFatz February 20, 2013 08:33

1) Does each time step converge until these errors appear? Can you show a residual plot?
2) Did you make a solution animation? Can you see something curious in the results until the errors?

Smaras February 20, 2013 09:43

Quote:

Originally Posted by RodriguezFatz (Post 408988)
1) Does each time step converge until these errors appear? Can you show a residual plot?
2) Did you make a solution animation? Can you see something curious in the results until the errors?

Thanks Rodriguez,

1. Yes, from very start it was converging, but just before the diverging occurs three or four time steps has no convergence rather turbulent viscosity.

2. Yes, i did only did for the penetration depth.

Please see the attachments. I thinks it's more clear in that.

http://imageshack.us/a/img526/7353/velocity.th.jpg

http://imageshack.us/a/img839/6499/residuals.th.jpg

http://imageshack.us/a/img201/6622/p...ondepth.th.jpg

http://imageshack.us/a/img827/964/fluent1410780.th.jpg

Thanks once again and looking forward for your reply.

Regards,
Azam

shk12345 February 20, 2013 09:47

hi
 
Sometimes divergence problem is due to bad meshing that you have used .
Try to refine your mesh and run the simulation and let me know.
try to decrease the under relaxation factors
also tell what is the courunt number in your simulation.

Regards
shk

RodriguezFatz February 20, 2013 10:19

I think shk is right here. It looks like the fluid firstly moves through a domain of "good mesh" where the solver is able to converge sufficiently, but then enters some bad cells, where the mess starts.
You should a) post a picture of your mesh b) show us the courant number (i think Fluent can show cell courant number) c) in the meantime let the simulation run with a smaller time step.

Edit: before you do c) please try, if you can reduce the residuals right from the start each time step better than now. It looks like you get "continuity" to about 1.0e-3. Can you get better convergence? How? By increasing the number of iterations per time step or only by reducing the time step?

Smaras February 20, 2013 10:58

Quote:

Originally Posted by RodriguezFatz (Post 409044)
I think shk is right here. It looks like the fluid firstly moves through a domain of "good mesh" where the solver is able to converge sufficiently, but then enters some bad cells, where the mess starts.
You should a) post a picture of your mesh b) show us the courant number (i think Fluent can show cell courant number) c) in the meantime let the simulation run with a smaller time step.

Edit: before you do c) please try, if you can reduce the residuals right from the start each time step better than now. It looks like you get "continuity" to about 1.0e-3. Can you get better convergence? How? By increasing the number of iterations per time step or only by reducing the time step?

Thanks Shk. I have modeled it but i think the problem lies in mesh cuz after 1.4e-3 it crashes.
This is my mesh

a) http://imageshack.us/a/img833/6899/meshj.th.png

Now just as shk as said i rechecked in fluent i am getting mesh

http://imageshack.us/a/img685/8620/fluent.th.jpg

http://imageshack.us/a/img708/7589/meshw.th.png

http://imageshack.us/a/img855/1880/fluent1.th.jpg


Rodriguez i have used smaller steps but this is stall point where it crashes. No matter the step size is small or large. I even tried number of iteration but no use.

Now what should i do????


regards,
Smaras

RodriguezFatz February 20, 2013 11:10

1) Post a picture of the Courant number shortly before it crashes.
2) Why do you use a prism layer at the wall? I don't think you expect any flow along that wall, right? As I understand it, a prism layer isn't helpful in such a case.
3) Can you use SIMPLE algorithm for your model? In my experience this is the most robust one.
4) Do you have ICEM for meshing?

Smaras February 20, 2013 11:32

Quote:

Originally Posted by RodriguezFatz (Post 409063)
1) Post a picture of the Courant number shortly before it crashes.
2) Why do you use a prism layer at the wall? I don't think you expect any flow along that wall, right? As I understand it, a prism layer isn't helpful in such a case.
3) Can you use SIMPLE algorithm for your model? In my experience this is the most robust one.
4) Do you have ICEM for meshing?

1) http://imageshack.us/a/img94/8794/courantno.th.jpg

http://imageshack.us/a/img51/6087/courantno1.th.jpg

2,3) Ok i will try that without prism layers. (after your reply)

4) Yes i am using ICEM for meshing.

RodriguezFatz February 21, 2013 03:57

If you have ICEM, you could make an excellent hexa-mesh. Your geometry is really simple and blocking would be straight forward. Don't use tet meshs unless you are not able to do the blocking. They are numerically low-grade.

Smaras February 21, 2013 05:33

Quote:

Originally Posted by RodriguezFatz (Post 409231)
If you have ICEM, you could make an excellent hexa-mesh. Your geometry is really simple and blocking would be straight forward. Don't use tet meshs unless you are not able to do the blocking. They are numerically low-grade.

Thanks
I know Rodriguez but this isn't the real geometry i will be working on more complex geometry, and the requirement for my thesis is tetra meshing. This is just a mock-up or example to make practice which are based on previous research papers.
I am today re-meshing it without the prism layer and plus more refined mesh. Further i have 2 questions:

1. I need to know one more thing i am using densities to get precise result for the jet flow and its impingement. as shown previously. Is it creating the problem????

2. Further after opening opening the geometry in fluent there is seperation line within the air region. and when i saw the velocity profile @ 1.4e-2sec the velocity is reaching the line. Is this meshing problem??????

Regards,
Smaras

RodriguezFatz February 21, 2013 05:40

1) I don't understand the question. Do you mean, if a variable density can cause problems? Then, the answer is yes.
2) What do you mean by "separation line"?

Smaras February 21, 2013 05:47

Quote:

Originally Posted by RodriguezFatz (Post 409258)
1) I don't understand the question. Do you mean, if a variable density can cause problems? Then, the answer is yes.
2) What do you mean by "separation line"?

1.) Ok then that might be one reason of divergence. Should i use prism layers for the region for smooth transition?

2.) http://imageshack.us/a/img28/6784/meshinfluent.th.jpg
The line that starts from top till the water bed.

RodriguezFatz February 21, 2013 05:54

But what kind of line is this? Did you paint it?

Also I am curious why the courant number at the top, next to the needle is so high. I guess the whole top area is a pressure outlet, right? Why would you have such a high courant number there?

Smaras February 21, 2013 06:03

Quote:

Originally Posted by RodriguezFatz (Post 409266)
But what kind of line is this? Did you paint it?

Also I am curious why the courant number at the top, next to the needle is so high. I guess the whole top area is a pressure outlet, right? Why would you have such a high courant number there?

1. Nope i didn't it's the result after i am creating mesh densities.

2. The top area is outlet. And that i don't know. It's appeared just after divergence

http://imageshack.us/a/img571/8189/courantno2.th.jpg

this is the image before divergence


All times are GMT -4. The time now is 12:36.