grid dependency
 

Hi, I'm running 2D incompressible viscous analysis on some airfoils; I'm using a kwsst model with boundary layer solution (Y+<1), second order interpolation schemes for all the variable and steady segregated implicit solver. Every time I try a new grid I obtain different results. Is there anybody that could suggest me a good grid or some grid convergence criteria? Does anybody notice some better turbolence model? Thanks

Re: grid dependency
 

If you want to obtain a grid independend solution, you have to refine it till results will not change anymore changing the grid.

Hi :)

ap

Re: grid dependency
 

Hi!

Do you use structured mesh? You should obtain beter results (or the same but faster) with structured C or O type mesh...

MATEUS

Re: grid dependency
 

Thaks ap for your answer. Do you mean hanging node adaption based on velocity gradients? I've already done it starting from two different grids (one structured and one hybrid) but when, after some adaptions, both sets of results are stable I noticed they're significantly different. Every time I obtain results that are stable but specific for that particular mesh. Did you try to obtain same results with different grids?

Re: grid dependency
 

Thaks Mateus. I've tryied C structured and hybrid with O structured bounday layer but the results are always different. Moreover I think the non conformal adaption performed by Fluent could couse some mistakes inthe flux balance. A the time I've no idea of how get rid of grid dependency at all.

Re: grid dependency
 

You say:

"...when, after some adaptions, both sets of results are stable I noticed they're significantly different."

Does this means that if you refine the grid, results doesn't change but still are different on different grids?

If so, I'd think to some numerical problem of implementation on a particular kind of grid.

Hi :)

ap

Re: grid dependency
 

I thought Fluent's Grid Adaption only worked with unstructured meshes. I usually do grid refinement by going back to the mesh generator (GAMBIT) and doubling all the mesh parameters (i.e. number of nodes on an edge)

The only problem with this method though is that I have to make a new case file and re-input all the boundary conditions/etc

Dan

Re: grid dependency
 

Yes ap, this is what's happeneing: different stable results for every kind of grid. My be my grid generation method is wrong; is there anybody can show me the grid he uses for an airfoil? I'm not using wall functions, so I'd need y+<1. Thanks

Re: grid dependency
 

You're right: maybe there is a problem during the grid generation.

Unfortunately, airfoils don't belong to my field, so I can't be of great help.

Anyway, I think a grid like the one shown here (from NAG) should work:

http://www.nag.co.uk/simulation/Fast...ml/node141.htm

Just refine it to have y+ close to 1, using y+ adaption in FLUENT.

Hi :)

ap