[sega]

Now that I did the "blocked" mesh I have a liquid phase fraction which is exactly as much as described in Case 10: http://test.interface.free.fr/Case10.pdf

The initial gamma-field is sharp if I take a look at the cell-gamma.

Unfortunately the pressure inside the bubble is still too low. It reaches a level of 0.943 N/m².
As described in the test-setup above it should be 1.04 N/m²

Now we should consider the next two options suggested:
2: No circular state. As I'm still not able to export a contour plot or similar to another post-processing tool, I can't get quantitative results of how the actual shape differs from a circle

3: Surface Tension and snGrad(). In Case10 there is a table listing the numerical error with different degrees of discretization.
In these terms (radius / gridspace) my grid has a value of 36.16 which is far more than the largest value (20) in Case10.
Concerning this the error of my model should not be that big!

So, there is still some thinking to do ...

What about these parasitic currents?
This whole setup is mainly about this.
Is there influenc that strong?

BTW: How can I compute the magnitude of these currents INSIDE the bubble?

[ngj]

Hi Sebastian,

Could you please upload your test case, then I will try to have a look at it in the beginning of next week.
If it is to big, do not hesitate sending it to my email.
Until then, enjoy yourself

- Niels

[sega]

Hey.

I send the files to your account.
For all the other people who want to toy around with this problem: Here are the files attached:

[sega]

Hey.

I send the files to your account.
For all the other people who want to toy around with this problem: Here are the files attached:

http://therealsega.th.funpic.de/open...ionSend.tar.gz

[ngj]

Hi all

Previously, I uploaded a small movie with a stick which turns into 3 spheres. Now I have added gravity just for the fun of it, and it looks so amazing that I wanted to share it with you.

Especially look at the symmetry which remains even though the flow is very complex. The movie shows the contour surface of gamma = 0.5 and all sides are walls.

www.student.dtu.dk/~s001581/OF_upload/cubeStickGravity.mpg

Have fun

Niels

[ngj]

Hi Sebastian

I have now looked at your setup, and I cannot say anything but: "I have run out of ideas.".

I have run your model and tried different things, and finally I tried to run it with double the discretization and the result is given below as "cell pd".



Suddenly large pressure gradients at the interface occur. Could it be that there is missing a term in the pressure or momentum equations? Otherwise these very large oscillations in the 'convergence' of the result is quite strange.

Please anyone who might have an idea, come forth

Best regards,

Niels

[ngj]

Hi

The present scheme used for div(phirb,gamma) is interfaceCompression. I have made a small and fast test, and it looks like the pressure increases if another scheme is chosen.

I used vanLeer, but wouldn't know if that is a clever choice.

Best regads,

Niels

[sega]

Well, ok. I think I will have a talk about this divScheme with the guys at the university.

But first I have to know what div(phirb,gamma) stands for ...

[sega]

Is it possible that we are comparing dynamic with static pressure?

[sega]

Two things:

1. There seems to be a tool called interFoamPressure to calculate the static pressure instead of the dynamic.
Maybe this can settle the error?

2. How can I get a value for the magnitude of these currents inside the bubble?

Thanks so far. S.

[sega]

Somehow I could make interFoamPressure work and ran it over my results.

So now the pressure (p) is 0.85 N/m² ... 0.1 N/m² less than pd
Wrong direction as it should be 1.04 N/m² ...

Any ideas?

[ngj]

Hi Sebastian

Sorry for not replying, but I have been in the lab all week ... quite fun some times to see what reality actually looks like

div(phirb,gamma) is the convective term of the transport equation for gamma, where phirb is a surface normal flux. Changing from interface conpression to vanLeer gives a changes in pd from 0.944 to 0.957, though still not enough to reach 1.04.

I have not been able to find interFoamPressure, so could you please direct me to the source code?

There is a tool in paraFoam where you can see parts of the domain based on a threshold on one of the parameters ... thus using gamma > 0.5 would give you the interior of the bubble. You can find it in the toolbar next to the isosurface (the the left hand side).

Enjoy,

Niels

[sega]

Hey Niels.

I stumbled over interFoamPressure in this post:
http://www.cfd-online.com/OpenFOAM_D...es/1/5619.html
The link (svn checkout) is in the third post ...

Greetings - S.

[ngj]

Hi Sebastian

At first glance I see that the UEqn does not look like the one in interFoam, thus it is not surprising that the results deviate. E.g. in the one fvc:snGrad(gamma) is used and in the other fvc:grad(gamma) is used. The deviation seems odd and it is not the only one!

- Niels

[sega]

Hey.

I'm still playing with this problem.

I did the following:
Defining a spherical gamma-field with this setup:



It's a quarter of a circel (due to symmetrical reasons).

I'm just starting the problem to have a look at the pressure inside the bubble.

Unfortunately the simulation is running at (adjustable) timesteps of 5e-6 by now.
This will take until eternity.

May it be that there is something wrong with the setup, because the solver is "stuck" like this?

[sega]

Of course the file was too big.

Now it's here:
http://therealsega.th.funpic.de/open...10aSend.tar.gz

[ngj]

Hi Sebastian

Your velocity field looks something like this:



thus somethings definitely wrong at the symmetry condition.

1. What happens is you first try to make it a wall? Just to see if that removes the large velocities, it should

2. You have mapped gamma. Could you also map U and pd to get a initial estimate of the fields which are more or less in equilibrium in your symmetric case?

- Niels

[sega]

Hey.

Mapping (funkySetFields) the pd-field brought some even more disturbing results.

What about U? I think it should be zero all over the geometry, thus I leave it uniform (0 0 0).

I think I will have to do something about the right and top faces, concerning the pressure.
Setting it zero may not be the right approach...

[sega]

So, now I made some changes to the boundary conditions:

left & bottom: symmetryPlane for all variable (U,pd,gamma)

top: 0 for U, totalPressure for pd

right: 0 for U, zeroGradient for pd

I initialized the gamma and pd fields using funkySetFields so the values of a circle and the pressure inside the bubble due to surface tension.




The setting of pd with funkySetFields is destorying my boundary conditions totalPressure every time, making the solver demand a pdRefCell and Value. So I insered the totalPressure BC by hand to the nonuniform-pd-file.

The actual setup is here:
http://therealsega.th.funpic.de/open...ase10_a.tar.gz

The solver is stuck with timesteps of 10e-6 from the first 0.1 seconds. For this timestep there is "good" looking gamma-field, but a strange pressure-field:



So, I think there is some problem with the pressure boundary conditions.

Any more ideas?

[sega]

Meanwhile I had a talk with my academic advisor at the faculty I'm writing my student research project.

Obviously there is another student who is doing the same test-case like me.
He is investigating the square liquid volume under zero gravity conditions and looks after the final shape and pressure inside the bubble.
He is doing this with Fluent!

As a matter of fact he has the same problems!!!

On the same sized grid with equal amount of cells he is getting an error with the pressure jump of about 20%.

A guy from the faculty staff used the interFoam-solver with a (self-build) level-set-method and got even worse numbers.

So I think the problem I'm experiencing is not that trivial.
It even looks like my error of about 9% in the pressure jump is a pretty good value.