Hi everybody!

I am having problems when using the mesh deformation functionality of cfx5. I have a 2D circular geometry with a thin membrane inside.

This membrane is displaced (mesh displacement) according to the pressure difference of both sides of the membrane using CEL (using integrated values force() and area()). The shape of the membrane is a nearly quadratic function (which drops to zero where it is connected to the wall. I am also using a step function to guarantee that the displacement at the wall is exactly 0).

See picture of geometry (size approx 3mm x 3mm): http://www.tyberis.com/geometry.png

The fluid comes into motion due to small momentum sources.

If I leave the mesh displacement exactly zero (no mesh displacement) then the pressures value in the domain look okay (within [-0.1, 0.1] Pa). But as soon as there is a small displacement of the membrane (even not noticable to the eye), the pressure gets very high in the whole domain (range [-1000, +1000] Pa)

Does anybody have an idea why the pressure explodes in this case? Does anybody have an idea what I could try? I have tried a lof of different parameters but without success.

Thanks for any help!

Hi,

I don't have the answer, but here aer some suggestions:

1) Make sure you have CFX patched up to the current version (5.7.1 SP1). Some of the recent patches included updates to the moving mesh stuff.

2) Moving mesh can require tighter convergence and/or smaller timesteps than stationary mesh simulations.

3) If the simulation is closed (ie no inlets, outlets or openings) then small variations in volume caused by discretisation can cause spurious pressure fluctuations.

Just some suggestions, they may or may not help.

Regards, Glenn Horrocks

Hi, thanks for your answer.

Unfortunately, evening using an extremely small timestep won't help. Also, I am now using only one plane for the membrane (i.e. both sides are at the same location) to guarantee no change in the total volume when the mesh is displaced. Also, I have added an opening into the side wall. Last but not least: For testing purposes, I have made the mesh displacement static (independent of pressure) with a nearly parabolic profile with a very small displacement in the center of the order of eps.

All this won't help; the relative pressure gets 10e5 even in the very first timestep. If I set the static value mentioned above to exactly zero instead of eps (i.e. no mesh deformation) then I get the correct results for the scenario without mesh deformation.

This is quite odd.

Hi,

Is your fluid incompressible? How much does the mesh move in the first timestep relative to the mesh size? If what you are seeing are pressure waves (ie acoustic waves) then you should use a compressible fluid.

Regards, Glenn