CFD Online Discussion Forums

CFD Online Discussion Forums (
-   CFX (
-   -   Free Surface not exactly at mean water level (

FluidH March 12, 2011 05:41

Free Surface not exactly at mean water level

I am simulating a free surface problem and have some problems with the initial results.

The problem appears when I have several sections in the domain.
The initial free surface then get a small disturbance when crossing a section.
Do someone know how to avoid this disturbance ?


ghorrocks March 13, 2011 06:39

What is the graph showing? I guess you are extracting the free surface location somehow - so who are you extracting it?

Also, is this initial conditions or the final result of the simulation?

FluidH March 13, 2011 07:27

The figures shows the isovalue of the water.volume fraction at 0.5, Wich will be the location of the free surface in a 2 phase flow between air and water.
The lowest one shows a closer view at x=3m where I have a section and the free surface get a small jump.

The figures shows the initial result with t=0.0s. So it should be a straight line at z=0 [m].

ghorrocks March 13, 2011 18:56

If the surface location lines up closely with a element edge, then when you define the initial conditions one side of the interface could have the element edge just above the level and one just below. When you graph the free surface level this can appear as a jump of one element edge length.

If you don't like this then use a blurring function to blurr the transition over a few elements. Then small differences in node location do not affect the free surface location.

FluidH March 18, 2011 11:14

Thanks for your answer, but maybe I don't understand this blurring function.
The free surface will in general always be evaluated over some grid cells, as my figure shows.

Do You mean that there is a additional blurring function which smooths the surface?

I use the initial volume fraction smoothing in the solver control, but it gives no changes for the location of the free surface.

ghorrocks March 20, 2011 22:14

No, you have the write the blurring function. Hopefully a CEL expression will do it.

But do you see my point? If you have a line of nodes at z=1m and set the free surface at 1m, then with numerical accuracy the nodes at 0.9999999m will be set as water and the nodes at 1.00000001 will be set as air. This will result in the free surface not being as you intended as it jumps up or down with the mesh numerical accuracy.

Possibly an easier fix is to move the free surface up or down by a small amount. Small enough that it does not matter, but big enough that the mesh numerical accuracy does not affect things.

FluidH March 21, 2011 02:55

Thanks! I get the point. The free surface will always lie between some grid cells, so the free surface accuracy will be +-the smallest element size.
As you say one can move the initial free surface condition a small amount from the exact value to avoid numerical jumps of one grid cell.

I will try to set ZWL= +-0.00001 away from the actually zero water level.


All times are GMT -4. The time now is 19:01.