Simulation do not converge

A subforum specific to the development of the OpenFoam-based workbenches ( Cfd https://github.com/qingfengxia/Cfd and CfdOF https://github.com/jaheyns/CfdOF )

Moderator: oliveroxtoby

Post Reply
Sidemountyucatan
Posts: 114
Joined: Wed Apr 17, 2019 2:08 pm

Simulation do not converge

Post by Sidemountyucatan »

About used Freecad -Version:
Version: 0.17
Revisionnumber: 13541 (Git)
Release date: 2018/08/16
Operation system: Windows 10
64 Bit
Branch: releases/Freecad-0-17
Hash: 9948ee4f1570df9216862a79705afb367b2c6ffb

Hi all,

I tried a "real life"-problem (little bit simplified to make the first check run). It is a lense with water flowing around it in a cleaning tray.
The simplified model is made of the lense in the rectangular tray. 1 inlet with Uy and Uz component (flow comes from down below the lense),
1 outlet (static pressure) and 4 walls. The most on top wall would be free surface in reality. Also inlet and outlet would be smaller (pipes) but
for this first approach I changed it.

Acc. to the calc. Reynoldsnumber it should be turbulence flow so I tried RANS. I did several simulations by changing mesh sizes and defined
some refined mesh regions (especially arround the lense). More or less I always got the same unsatisfying results. (the file with the actual
model settings and a print out of its residuals are attached).

In every case the pressure residual was always arround 10^-1 (not good). In the actual case also the residual for the turbulence energy was too
high. (in the actual simulation I called the simulation after 3 hours or more of running).

Therefore I have several questions:

(1) What is my failure by always getting such high pressure residuals ?
(2) Is there any issue with the "Outflow" outlet condition ? (always showed an error when used)
(3) When checking the mesh with Paraview I saw some quite bad cells (skewness) but I think there is
no possibilty to improve the net cellwise, right ?
(4) Is there a hint from anybody what I can do to make the simulation run with meaningful results ?
(5) What is the internal coded condition for the program to stop the simulation ?

for more complex (real) situation simulations:

(5) is there a possibilty to define a free surface ?
(6) any advices how to build the net in case modifying the model in terms of smaller inlet and outlet (pipes in rectangular walls) ?

Thx everybody for help me solving those issues !!!

Best Regards
Lense flow around_Water_cfmesh 10mm_refined_RANS_p outlet.pdf
Linse wasserumspült.FCStd
(27.26 KiB) Downloaded 48 times
Attachments
Lense flow around_Water_cfmesh 10mm_refined_RANS_p outlet.pdf
(22.91 KiB) Downloaded 61 times
User avatar
oliveroxtoby
Posts: 810
Joined: Fri Dec 23, 2016 9:43 am
Location: South Africa

Re: Simulation do not converge

Post by oliveroxtoby »

Sidemountyucatan wrote: Tue Jul 09, 2019 2:06 pm In every case the pressure residual was always arround 10^-1 (not good). In the actual case also the residual for the turbulence energy was too
high. (in the actual simulation I called the simulation after 3 hours or more of running).

Therefore I have several questions:

(1) What is my failure by always getting such high pressure residuals ?
My guess would be that you need to put the inlet and outlet further up/downstream of the body.
(2) Is there any issue with the "Outflow" outlet condition ? (always showed an error when used)
It it only useful in unusual situations (e.g. multiple pressure inlets/outlets). It extrapolates all variables, so you will end up with undefined pressure in this case.
(3) When checking the mesh with Paraview I saw some quite bad cells (skewness) but I think there is
no possibilty to improve the net cellwise, right ?
The OpenFOAM checkMesh utility can be used to print some stats. They look fine.
(5) What is the internal coded condition for the program to stop the simulation ?
The residual tolerance and maximum number of iterations or end time can be accessed on the 'Data' panel when you click on the solver object.
(5) is there a possibilty to define a free surface ?
Yes - there has been some discussion in this forum on how to do it.
Please provide all the information requested in this post before reporting problems with CfdOF.
Sidemountyucatan
Posts: 114
Joined: Wed Apr 17, 2019 2:08 pm

Re: Simulation do not converge

Post by Sidemountyucatan »

Thx oliveroxtoby,

I will try to remodel the case. By the way: I read somewhere that for the outlet pressure condition the value is meant to be the pressure devided
by the density. Is that right ? Altough there is "Pa" in the unit field ?

Regarding the free surface boundary condition:

I found this topic:
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Re: Update on the development of the OpenFOAM CFD workbench (CFDOF)
Diesen Beitrag melden
Zitat
Beitragvon oliveroxtoby » So Okt 01, 2017 10:07 pm
thschrader hat geschrieben: ↑
So Okt 01, 2017 4:18 pm

Here is a little example, that you can do free-surface flow simulation using the
cfd-wb. It is a modified case of the tutorial from
https://www.youtube.com/watch?v=gZ_TqsPwiXY

Design the geometry and the cfd-case in FC, run the simpleFoam-solver to get all necessary files.
Delete all results, overwrite the p/U/alpha.water/fvschemes/fvsolutions/controldict/setFieldsdict with
the files (slightly modified) from the interFoam/dam-break tutorial and activate the bluecfdcore-console.
Set the alpha.water field an run interFoam. The used mesh is very coarse, my goal was to get
the stuff running...

Hi Thomas, thanks for the example. Please note that the CfdOF workbench actually supports free-surface with interFoam (and multiphaseInterFoam) right out of the box (made possible by a contribution from Klaus Sembritzki) - so there should be no need to overwrite dictionaries manually. If you add multiple fluid materials, then provided you choose a transient, incompressible analysis type in the physics task panel, interFoam or multiphaseInterFoam (for more than two fluids) will be used automatically. Field initialisation zones can be used to select the initial fluid volume fractions. At present gravity is restricted to -9.81 m/s^2 in the y direction and there is no surface tension (unless the case files are manually edited, of course).
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

It does not describe how to set up this case using cfdOf. (especially the video describes "just" using "native" OpenFoam). Futhermore, did I understood correctly: surface tensions are not considered at all ?

I didn't found any other useful article describing set up a fee surface boundary in cfdOf. Do you have any discussion in mind facing that topic ?

Thx for feedback.
Sidemountyucatan
Posts: 114
Joined: Wed Apr 17, 2019 2:08 pm

Re: Simulation do not converge

Post by Sidemountyucatan »

Hi Oliveroxtoby,

is it meanningful to define a free suface as a slip surface especially if the movement of the surface water<-> air is not
in interest ?

Than my problem would have been solved quite easy and elegant.

Thx for feedback.
User avatar
oliveroxtoby
Posts: 810
Joined: Fri Dec 23, 2016 9:43 am
Location: South Africa

Re: Simulation do not converge

Post by oliveroxtoby »

Sidemountyucatan wrote: Wed Jul 10, 2019 11:08 am Thx oliveroxtoby,

I will try to remodel the case. By the way: I read somewhere that for the outlet pressure condition the value is meant to be the pressure devided
by the density. Is that right ? Altough there is "Pa" in the unit field ?
No, CfdOF does the conversion for you. You always specify pressure. But the output you view in paraview will be p/rho.
Regarding the free surface boundary condition:

I found this topic:
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Hi Thomas, thanks for the example. Please note that the CfdOF workbench actually supports free-surface with interFoam (and multiphaseInterFoam) right out of the box (made possible by a contribution from Klaus Sembritzki) - so there should be no need to overwrite dictionaries manually. If you add multiple fluid materials, then provided you choose a transient, incompressible analysis type in the physics task panel, interFoam or multiphaseInterFoam (for more than two fluids) will be used automatically. Field initialisation zones can be used to select the initial fluid volume fractions. At present gravity is restricted to -9.81 m/s^2 in the y direction and there is no surface tension (unless the case files are manually edited, of course).
-------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------

It does not describe how to set up this case using cfdOf. (especially the video describes "just" using "native" OpenFoam). Futhermore, did I understood correctly: surface tensions are not considered at all ?
Correct, but you can easily add it by editing the OpenFOAM case. We just haven't created the UI to input the surface tension coefficient (contributions welcomed).
I didn't found any other useful article describing set up a fee surface boundary in cfdOf. Do you have any discussion in mind facing that topic ?

Thx for feedback.
https://forum.freecadweb.org/viewtopic. ... se#p317437
https://forum.freecadweb.org/viewtopic. ... 40#p194045
Please provide all the information requested in this post before reporting problems with CfdOF.
Post Reply