Learning Forum

[Dato]

Hello everyone,

I am doing the mesh sensitivity analysis and I found some problems of domain extents in scale panel. I have imported  different cell counts .msh.h5 files under same geometry into Fluent and everytime I check the "scale" panel, the domain extents are not always as same as I built the geometry in CAD. The picture showing ledt is correct domain extents while right is wrong.

It always happens that there exist some error in "Zmin" ranging from e-6 to e-7 as I have read the mesh files so many times.

I want to know that if I create some iso-surfaces while are not aware of this problems, would it affect the results on created iso-surfaces because of the slight distance error? And Why would it happens? How can we avoid this problems when importing mesh into Fluent?

Thanks so much! 

[Federico]

Hello Lee, 

are you using single or double precision in Fluent? 

From your screenshots, it appears that there are some slight differences in each of XYZ directions. I would suggest using double precision and choosing consistent units for your comparison.

If the issue persists, please share your findings here.

[Dato]

Hello Federico,

I use double precision all the time and I also change the unit to the same one everytime I read the mesh. This problems haven't been solved yet. 

Besides, I found that there are also some precision problems when  using iso-surface as shown below. 

As you can see it, the minimum z-coordinate is not computed as same in "iso-surface" and "scale". It's so weird. I'm thinking that there shouldn't exist so large error which  is only in e-3 level. It need to  be know that I use double precision and generated the mesh in FluentMeshing. 

The error is so large, I think, and if I post-processing using the value on this surface, obviously it would cause increase error. I wonder if the precision problem will still exist if  I export the solution data  into other post-processing software like cfd-post, paraview, ensight though I seldom use them.????

[Dato]

Additionally, I think the deviation caused by e-3 distance would result in considerable error in my reserach.