Hello, I am a new user to CFX and I want to model Benard convection cells with CFX. I went over the free surface tutorial and its a good starting point but is there a another tutorial for free surface modeling with CFX? My model is a liquid in a container with the side walls insulated, heat addition from below and at the top the water is in contact with air. How do I go about setting up boundry conditions for such a model? I created a rectangle with six regions - bottom, two sides (insulation), two sides (symmetry) (trying to model this in 2-D), and top; I applied the symmetry and bottom constant temp boundry. Not sure how to incoporate the side insulation and VF conditions. -FER

Hi,

If the free surface is not important or does not move much then the simulation will be far simpler if you just put a wall with appropriate thermal boundary conditions.

Glenn Horrocks

Glenn, thanks for the info. But the surface interaction is what I want to monitor.

Hi,

Can't the side walls just be adiabatic walls? That's pretty straight forward.

What part of the multiphase setup is confusing you?

Glenn Horrocks

Hi FER,

i would take Glenn's suggestion first since adding free surface to the calculation would appear to be a secondary effect once you get the weak convection started.

The problem you are trying to to is infact very hard to simulate unless you are very careful in controling factors which affect the stability and symmetries of the problem. You want to eliminate any damping, numerical (ie non physical perturbations) and breaking of symmetry inherrent in the code.

For example I would use a very symmetrical structured grid for the problem as an unstructured grid even with all other bc and geometric symmetries will not necessarily lead to a symmetrcal solution.

Even if you have a symmetric problem domain, symmetric boundary conditions, and symmetry in the body forces (the last 2 ammount to symmetries in the RHS of A{x} = b) the code can introduce assymetries which are not physical.

You will also want numerical discretizations which do not introduce unnecessary damping nor assymetries. Crank Nicholson in time and CDS in space are good choices. I know there is an expert flag in CFX to switch to CDS. Never use hi-res or something like that for this kind of problem. Hence my comment about not trying to do the VOF at first because it uses a compressive discretization like hi-res.

There are some good books on Non-linear dynamics and the general issues to simulate such problems. Try Seydel(sp): From Equilibrium to Chaos it has lots of references.

I hope this is of help.

Regards,

Bak_Flow