CFD Online Discussion Forums

CFD Online Discussion Forums (https://www.cfd-online.com/Forums/)
-   FloEFD, FloWorks & FloTHERM (https://www.cfd-online.com/Forums/floefd-floworks-flotherm/)
-   -   FloEFD Adaptive meshing (https://www.cfd-online.com/Forums/floefd-floworks-flotherm/90976-floefd-adaptive-meshing.html)

vaneedenp July 27, 2011 05:58

FloEFD Adaptive meshing
 
Hi all,

I am learning FloEFD and have started playing around with adaptive meshing. When I run the simulation and I look at the graph of one of my goals, it seems to converge to a certain value. But when I hit the refine option, the value of the goal drops on the graph and after a few iterations it picks up again to the previous value it was converging to. Everytime I do that the dip is there and it seems to be going to the same values. Shouldn't this dip disappear as the simulation converges? How often does one refine? I tried on every 10 iterations, then 20 and then 40. Same pattern.

Any advice?
Pieter

vaneedenp July 27, 2011 06:55

I found that increasing the total number of cells when the mesh refines each time makes this drop in value less. I will continue increasing number of cells until I do not get a drop anymore when I do a refine. This should tell me if the mesh is refined enough. Does this make sense? Am I correct in my assumption?

vaneedenp July 27, 2011 10:37

Got to a maximum number of cells, and the dip is still there. I am missing something here. Anybody help?

Boris_M August 1, 2011 09:17

Hi vaneedenp,
no, please don't refine too often. Usually I refine maximum 3 times, usually I use 2 times. This drop depends on if you look on absolute or normalized values. It is completely normal to have a jump like that. The strength of the jump can be less for other parameters.
You should also always work from a reasonable good mesh as the adaptive refinement is based on the actual results. So if you have a bad mesh you would refine on bad results which doesn't give a good refinement. These refinements are used to refine areas of stronger gradients as for example on a nozzles, flow separations like in a 90° bent or behind a blunt body or in a shok front to better resolve the physics and parameters over such gradients.

If you want to see if you have a mesh convergence (values don't change anymore with finer mesh) you can do a refinement and if the value converges again towards the same as before you won't get any better results after 10 further refinements.

I hope this helps,
Boris

vaneedenp August 1, 2011 09:44

Hallo Boris,

Thank you for your reply. I try to set up a proper mesh before I start. For the critical areas I do a local initialised mesh and try to get the spacing between cells as good as possible to have enough cells in small openings. When I run the simulation and refine it looks as if it converges to the same value. Which tells me that the mesh is good if I use your criteria.

Thanx for the info!


All times are GMT -4. The time now is 02:04.