CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   CFX (http://www.cfd-online.com/Forums/cfx/)
-   -   How should we define a multiphase interface (http://www.cfd-online.com/Forums/cfx/90245-how-should-we-define-multiphase-interface.html)

zcesd47 July 5, 2011 09:24

How should we define a multiphase interface
 
1 Attachment(s)
Hi, all

I am trying to define a inviscid water-air interface with a cos profile as the intial condition. The purpose is to model the sloshing behavior of water in a tank excited by sinusoidal external excitation. As the sloshing will eventually reach to the steady state (sloshing wave will has the same amplitude during the time history), we can defined a wave with a initial cos profile and start simulation.


I defined the cos interface with CEL and vof mehod, as following

height:0.03[m]*cos(2*pi[rad m^-1]/1.7680*x)+0.198[m]
water volum fraction: step((height-z)/1[m])

However, the interface profile defined as above is not smooth (like sawtooth) and there are some undefined mesh elements,
:(

Can anybody give me some suggestion about how to define a smooth free surface profile, please, thank you very much.

ghorrocks July 5, 2011 19:29

There should be any undefined elements. What elements are undefined?

You need to put a blurring function on the interface. This can be done with a linear blend over a few elements or more complex functions - but it will take a bit of thinking to develop the equation.

Alternately you can just start with the jagged interface and run it for a while to smooth out the jaggedness.

zcesd47 July 6, 2011 08:10

Quote:

Originally Posted by ghorrocks (Post 314856)
There should be any undefined elements. What elements are undefined?

You need to put a blurring function on the interface. This can be done with a linear blend over a few elements or more complex functions - but it will take a bit of thinking to develop the equation.

Alternately you can just start with the jagged interface and run it for a while to smooth out the jaggedness.

Hi, Horrocks, thank you so much for you suggestions and help.

There are some cells in grey color (in the graph) when I check the water.volume fraction in post-processer, are those undefined cell? For other cases I did there are more 'grey cells'.

I will try to define to define the volume fraction of the cells on the interface as a hypobolic function,sorry, I am new to CFD, could you suggest me some name of the function may be useful, eg. function that may be applied for selecting cells on the interface. Great appriciated.

My purpose is to model the sloshing behavior in a fluid tank when it reach the steady state (constant wave crest level). So to check the linear wave behavior I define a initial water surface as above, with small amplitude, and set initial velocity to 0, pressure to hydrostatic pressure, but the wave amplitude still has some decay with 0 viscousity, so I thought the reason maybe the jaggedness. Is it possible to extract the volume fraction of the whole mesh from solution data after jaggedness are smoothed out, then put into mesh in CFX-pre and start another run?Thank you very much.


Sincerely

Dijun

ghorrocks July 6, 2011 19:20

No elements should be undefined. Have a close look at them and check they are OK before doing anything else.

The model will always have some decay due to numerical diffusion. You can reduce this with fine meshes and second order differencing schemes, but you cannot eliminate it.

Yes, you can use a previous result as the initial condition for a new run.

zcesd47 July 7, 2011 09:00

Hi Horrocks,

Thank you so much for your kind help. Could you give me some more help? Could you tell me the way of checking the volume fraction of the cells of the mesh,please.

I used the default second-order backward Euler scheme. But the wave amplitude actually increased by some reason in the first two periods, and decrease, and then increase again for two periods. As the water viscousity was set to zero, I expect the wave height should be constant, so I guess it may because of the initial condition. Do you feel it is wrong to set the initial velocity in all direction to zero? I am greatly appreciated for your help.

Sincerely

Dijun Pan

ghorrocks July 7, 2011 19:11

Don't forget about the spatial differencing scheme. If you need to minimise diffusion then I recommend you use pure second order differencing (ie hybrid with 1 blend factor).

zcesd47 July 8, 2011 10:00

Thank you very much, really appreciated for your help.


All times are GMT -4. The time now is 08:26.