[kbaker]

Hi I am trying simulating stratified two phase air water flow in a pipe (its D=98.5 mm and length = 25 m while water level fixed at 8 mm) with CFX from couple of a days ago. I set the case as 2D I divide the inlet into two parts (one for water and the second for air) to set the velocities of the individual phases separately (all details mentioned at the attached report). I used the homogeneous model because the phases never mixed and with steady state analysis type. Also I made a CEL expression to some variables similar to that mentioned in free surface flow over a Bump tutorial (see attachment for ccl file too).
The problem I face currently the water is formulated near the lower inlet in a very small portion of pipe length then disappeared (see first image attached) then re-formulated after nearly 2 meters of length (which full length is 25 m) and still fixed at this level (as I want it) to the outlet (see the other image for vof attached). I am wondered why this discontinuity appeared in water at the results? what happened? Any suggestions please?

[ghorrocks]

What is the mesh size in this region?

[kbaker]

Hi Glenn the mesh is equally spaced over the entire length there is 600 elements distributed over the 25 m length of the pipe (I draw the mesh in Gambit) while for the vertical dimension there is 40 elements for the 100 mm diameter as I said the mesh distributed equally over the length so there is no different in mesh quality at inlet and outlet as I expect.

[ghorrocks]

Your mesh across the diameter is a bit coarse, you are resolving the water in only 4 nodes or so.

But the problem you are seeing is going to be something about your set up. Either it is not converged properly or not set up properly.

[kbaker]

There is 10 nodes for water and 30 nodes for air? The convergence is very well and I post my problem setup at the previous zip file in my first post you can have a look? Even if you need my cfx-pre file I can post it here?

[ghorrocks]

The zip file only contains a small section of CCL. Can you post the full CCL?

[kbaker]

Hi Glenn I check the CCL file on the attached zip file it full and nothing missing with it you can have a look at the Report.html file which contain the problem settings too or you can download my cfx-pre file from the link below I uploaded it:

http://www.4shared.com/file/snuV21Oh/two_inlets.html

[kbaker]

Glenn did you check the cfx-pre file I sent you? You not reply me yet?

[ghorrocks]

I only want to look at your CCL. Please extract the full CCL and post that.

[kbaker]

CEL:
EXPRESSIONS:
DenH = (DenWater - DenRef)
DenRef = 1.225 [kg m^-3]
DenWater = 1000 [kg m^-3]
DownH = 0.008 [m]
DownPres = DenH*g*DownVFWater*(DownH-y)
DownVFAir = step((y-DownH)/1[m])
DownVFWater = 1-DownVFAir
UpH = 0.008 [m]
UpPres = DenH*g*UpVFWater*(UpH-y)
UpVFAir = step((y-UpH)/1[m])
UpVFWater = 1-UpVFAir
END
END

[ghorrocks]

That is just the section specifying the CEL. The full CCL includes the bondary condition setup, convergence parameters, output file specification, physical models, material properties and lots more.

[kbaker]

Sorry here is it

[ghorrocks]

Some comments:

1) Free surface modelling is tricky in steady state. Very hard to get convergence. Try doing it as a transient model. This is especially the case as this is a surface tension model.
2) Are you sure surface tension is significant on these length scales? Surface tension modelling increases the difficulty of the mode a lot.
3) Your convergence tolerance is loose.

[kbaker]

Thanks a lot
1) I will take your advice about shifting to transient analysis.
2) I activate surface tension option because I am interested in generating waves at the interface furthermore I took the values of velocities, liquid level , pipe diameter and pipe length depending on an experimental case mentioned in a paper I want to validate CFX results with it.
3) How my convergence tolerance not specified? I set it as 1E-04?

[ghorrocks]

2) You do not need surface tension to generate waves. I think you will find ST is not having a significant effect on the final results, but will make convergence MUCH harder to achieve.
3) Yes, and 1E-4 is quite loose for a steady state model.

[kbaker]

2) You advice me still working with steady state simulations but without activating surface tension?
3) How much you think the better tolerance value for steady-state modeling?

[ghorrocks]

2) Definitely turn ST off unless you know you need it. Try that first, and if you still have problems try transient.
3) Do a sensitivity check to work out the convergence residual you need for this model for the accuracy you want to achieve.

[kbaker]

Glenn I attempt to change my problem to transient but the following message appeared to me:


In Analysis 'Flow Analysis 1' - Domain 'Default Domain': Transient analyses require that initial conditions are specified unless an Initial Values file is specified at run-time.

I tried several options but failed to fix it? may you tell me with details how to resolve it?

[ghorrocks]

That sounds pretty straight forward to me - you just need to define an initial condition.

[kbaker]

I need to make the whole problem unsteady because I still face convergence difficulties with steady state solution you advice me to go through transient at your last posts if you remember (see the above posts pls)?