CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   OpenFOAM Running, Solving & CFD (https://www.cfd-online.com/Forums/openfoam-solving/)
-   -   time step of twophaseeulerfoam (https://www.cfd-online.com/Forums/openfoam-solving/228465-time-step-twophaseeulerfoam.html)

qi.yang@polimi.it July 2, 2020 12:02

time step of twophaseeulerfoam
 
Hi guys,


I am using twophaseeulerfoam to simulate the solid-liquid pipe flow. The total number of the cells are 180000 (unstructure mesh) but I need to set a very very small time step to keep simulation normally runnning. Do you have any suggestion that I can define a larger time step? I saw some people used this solver to simulate the same phonomenon. They used more cells than mine. I cannot imagine how much time they used to finish one simulation if the situation like mine.

Thanks,

rdbisme July 2, 2020 14:40

Quote:

Originally Posted by qi.yang@polimi.it (Post 776700)
Hi guys,


I am using twophaseeulerfoam to simulate the solid-liquid pipe flow. The total number of the cells are 180000 (unstructure mesh) but I need to set a very very small time step to keep simulation normally runnning. Do you have any suggestion that I can define a larger time step? I saw some people used this solver to simulate the same phonomenon. They used more cells than mine. I cannot imagine how much time they used to finishone simulation if the situation like mine.

Thanks,

Hello qi.yang, I think you might get more helpful comments if you give more details about your configuration. You can check my old post for an example of which details might be interesting. :)

Ardali July 2, 2020 20:47

Quote:

Originally Posted by qi.yang@polimi.it (Post 776700)
Hi guys,


I am using twophaseeulerfoam to simulate the solid-liquid pipe flow. The total number of the cells are 180000 (unstructure mesh) but I need to set a very very small time step to keep simulation normally runnning. Do you have any suggestion that I can define a larger time step? I saw some people used this solver to simulate the same phonomenon. They used more cells than mine. I cannot imagine how much time they used to finish one simulation if the situation like mine.

Thanks,




If you do not need energy equation discard it in the solver.

If you need it specify correct boundary conditions in T files.

diameterModel in phaseProperties matters as well.
If comprisibility matters to you a lot, e.g.sonic jet or so on, It will be reflected in p_rgh, therefore it wont be a sudden crash. Anyhow, small courant number in the order of 0.1 may help too.
Good luck
Ardalan

qi.yang@polimi.it July 3, 2020 03:15

Thanks a lot. As you said, I already neglected energy euqation. Even though I set maxCo 0.1, the maximum time step is 2e-5 or 2e-6. The totall cells are 260000. How about your time step used in the simulation?

Quote:

Originally Posted by Ardali (Post 776742)
If you do not need energy equation discard it in the solver.

If you need it specify correct boundary conditions in T files.

diameterModel in phaseProperties matters as well.
If comprisibility matters to you a lot, e.g.sonic jet or so on, It will be reflected in p_rgh, therefore it wont be a sudden crash. Anyhow, small courant number in the order of 0.1 may help too.
Good luck
Ardalan


qi.yang@polimi.it July 3, 2020 03:24

1 Attachment(s)
Thank you. I uploaded my test files and mesh here. Could you please help me to have a look? By the way, the parameters in slurry pipe I used are the following:

pipe diameter 0.1 m (only half)
particle diameter 1e-4 m
Uin=2.5 m/s
alphas=0.2


Quote:

Originally Posted by rdbisme (Post 776716)
Hello qi.yang, I think you might get more helpful comments if you give more details about your configuration. You can check my old post for an example of which details might be interesting. :)


qi.yang@polimi.it July 3, 2020 03:29

1 Attachment(s)
I uploaded the files here without mesh, due to the large size.

qi.yang@polimi.it July 3, 2020 03:38

I saw your previous post and I checked your test case uploaded there. The timestep in your case is 2e-4 s. So you need several days or more to finish one simulation?

Quote:

Originally Posted by rdbisme (Post 776716)
Hello qi.yang, I think you might get more helpful comments if you give more details about your configuration. You can check my old post for an example of which details might be interesting. :)


qi.yang@polimi.it July 6, 2020 10:30

Nobody is curious about this issue?

rdbisme July 6, 2020 11:38

Quote:

Originally Posted by qi.yang@polimi.it (Post 776765)
I saw your previous post and I checked your test case uploaded there. The timestep in your case is 2e-4 s. So you need several days or more to finish one simulation?

Hello,


actually.. no. Running the test case in parallel the test would finish in a matter of hours... Why do you say days? Are you running in parallel?

qi.yang@polimi.it July 7, 2020 08:27

Thanks. I did not try parallel but in the end I will do it. I am curious that there is the way to increase the time step.


Quote:

Originally Posted by rdbisme (Post 777005)
Hello,


actually.. no. Running the test case in parallel the test would finish in a matter of hours... Why do you say days? Are you running in parallel?


rdbisme July 7, 2020 08:32

With a very basic explanation: The time step of a simulation can't be arbitrarily changed. Based on the time and space schemes you are using, there's a so-called CFL condition that restricts the maximum time step you can achieve. This CFL condition takes also in consideration the cell size: that means if you have a very fine mesh, you're generally gonna have a small time step to keep the numerical scheme stable.

What you can do to increase the time step is to compromise on some aspects:
- Use a coarse mesh
- Use low order numerical schemes that have larger stability regions
- Use implicit time schemes
- Increase the Courant number (that will reduce your stability)

qi.yang@polimi.it July 7, 2020 08:42

4 Attachment(s)
Hi guys, I met a new problem that the converency issue between the unstrcture mesh created by workbench and structure mesh generated by blockMesh. Why the iteration of P_rg is 1000, which means unconverged, right? Do you habe any suggestion? Thank u.

qi.yang@polimi.it July 7, 2020 08:52

Thanks a lot. I already considered what you mentioned below however, the time step is the samll one. Now I regard that issue as normal one. Really thank u. Gentilissimo!

Quote:

Originally Posted by rdbisme (Post 777084)
With a very basic explanation: The time step of a simulation can't be arbitrarily changed. Based on the time and space schemes you are using, there's a so-called CFL condition that restricts the maximum time step you can achieve. This CFL condition takes also in consideration the cell size: that means if you have a very fine mesh, you're generally gonna have a small time step to keep the numerical scheme stable.

What you can do to increase the time step is to compromise on some aspects:
- Use a coarse mesh
- Use low order numerical schemes that have larger stability regions
- Use implicit time schemes
- Increase the Courant number (that will reduce your stability)



All times are GMT -4. The time now is 04:51.