CFD Online Discussion Forums

CFD Online Discussion Forums (
-   FLUENT (
-   -   Is 2-way FSI valve modeling with seperate fluid domains possible? (

Maurosso March 5, 2013 07:16

Is 2-way FSI valve modeling with seperate fluid domains possible?

I have come to a more or less stand still in my work on modeling a hydraulic car shock absorber. The study includes a velocity boundary condition applied in Ansys Mechanical (workbench, not APDL) on a plate, that will in response compress the fluid, and interact with the valves in the damper.

At the moment I am doing preliminary studies, with huge simplifications to the problem, just to get the hang of Mechanical-Fluent system coupling and 2-way fsi modeling.

The problem that I have, is that all the information I've found, all the tutorials and instructions, present a solution for problems where the fluid domain is continuous.
I've made a test of concept for the whole problem by preparing a model visible on Pic1, where the horizontal bar in the middle of fluid domain was to oscillate according to the velocity boundary condition of the top bar. It was more or less successful. The "valve" indeed oscillated, and the fluid structure interface and system coupling worked nicely, I used mesh smoothing for the problem and in got the job done.

But here's the catch. On the final model (no yet ready), the valves will be closed on the first steps, basically creating 2 non-continuous, fluid domains. The problem is schematically shown in Pic2, where the gray bar in the middle would be a fixed solid, and the pink bar would "ease" and open downwards under the rising pressure induced by the movement of the top plate. are my questions: Is it possible to prepare such an study? Where at the first few steps, the fluid domains are separated by a solidbody, but after large enough deformation of the solid, the flow begins between the two domains? Could you provide me with some information on such modeling, or point to towards some literature or tutorials on the subject?

If not possible in any way with fluent-mechanical system coupling-> What methods, tricks, tips, ways, would to suggest to explore to prepare such an analysis.

Martin K.



stumpy March 5, 2013 12:32

You would need to have a very small gap at the right end of the "valve", so the two fluid domains are connected. The challenge will then be to avoid the mesh collapsing or becoming too skewed at the valve opens. There'll be some leakage when the valve is "closed" due to the gap. There's no easy way to deal with that.

Maurosso March 5, 2013 14:41

Well...ok, that might work for this example (even thou slightly, as modeling a valve that you know is no fun...:) ).

You said that there are no easy ways to deal with it, are there difficult ones? Or should I focus on preparing a model of the whole valve assembly with micro and mini leaks? I believe meshing and setting up the dynamic mesh will be a huge pain in the butt...
Are, for example, heart valves modeled this way?
When modeling combustion in a cylinder with intake and outtake valves included, those analisis sort of do it right, and no leakaged was to be spotted, unless they hid it in post-procesing.

Sorry for my abnouxious questions...It's just that, as I've said, I am at a stand still, and don't know whether to change the whole idea behind the subject, do some tinkering with it (as you suggested: micro leaks etc), or dig deeper for other possibilities.

stumpy March 6, 2013 17:17

The difficult way to stop the leaks is use a UDF to identify the cells in the gap and then introduce a porous loss or momentum sink to block the flow. Identifying the cells is the difficult part. The latest version of Fluent has a contact detection model that does exactly this, but you can't use it with System Coupling yet.
I'm not sure what IC engines do in this situation.

Maurosso March 7, 2013 09:56

Thank you for your reply, I will try my best to resolve the subject using the "gaps" solution and dynamic meshing, as I have no experience with UDF's or C programing...thou maybe it's time to dig into that :)

Again, thank you stumpy for your help

All times are GMT -4. The time now is 02:42.