Hi all,

Can anyone tell me if there's a rule of thumb for defining the cell sizes of your mesh. I realise this is dependant on the flow type, etc. Is there a way to calculate a reasonable cell size based on flow conditions, for instance?

Jamcam

Hi,

You should satisfy the CFL (Courant - Friedrich - Lewy) stability condition to be sure to have stable solution and choose apropriate cell size.

However, the cell size will determine size of your time step, so be carefull to be sure to satisfy CFL criteria. It also strongly depends on type of the flow you are trying to simulate - give some hints: is that compressible/incompressible flow? what discretization scheme for convective terms are you using? etc. etc.

Best, Maciej Matykahttp://panoramix.ift.uni.wroc.pl/~maq/eng/

For the multiphase fluid flow, you need a sufficient fine mesh to resolve surface tension force, for example. Finer mesh is better, but it costs a lot, so you need to trade off both.

Junseok

For Viscous flows, visit this web-site

http://geolab.larc.nasa.gov/APPS/YPlus/

Apurva

Hi Jamcam,

I'm afraid, the mesh size is more calibrated estimation, guess and intuition.

If you're doing incomressible steady state, you can easily leave CFL sleeping in the corner. There are, still, some good advices:

(1) boundary conditions -inlets/outlets - you need some cells to push the mass through, so at least 4-8 cells across.

(2) fine mesh in places with steep gradients

(3) watch the Y+

(4) do grid-dependency test.

Cannot dig any other out of my mind now.... There is one thing, I've found - finner mesh on outlet/pressure bounday will help the convergence, believe it or not, its working for me. You are -anyway always limited by the memory and CPU time.

matej

Thanks for all the input, guys!