# [ICEM] Hexa -> small y+ problem

 Register Blogs Members List Search Today's Posts Mark Forums Read

 September 5, 2014, 03:55 #21 Senior Member     Philipp Join Date: Jun 2011 Location: Germany Posts: 1,297 Rep Power: 26 Hey cad, I thought about your post but still don't get your point. I made a small sketch and I can't see here how the quality suffers from refining. sketch_cells.png The original (coarse) cell is the blue one, the green line shows on of the refined cells. For me, it looks like all angles keep constant. __________________ The skeleton ran out of shampoo in the shower.

 September 8, 2014, 07:27 #22 Senior Member     Philipp Join Date: Jun 2011 Location: Germany Posts: 1,297 Rep Power: 26 Ok guys, they changed the algorithm for orthogonal quality in ICEM so it matches with the value in Fluent now. That's why we see different values in 14.5 and 15.0. The new calculation depends strongly on how good the cell center coordinates can be calculated. This calculation however depends on the aspect ratio. So for really large aspect ratios, cell center is inaccurate thus orthogonal quality can be bad. The question is, if "just" the orthogonal quality is bad, or if this inaccurate cell center calculation also influences the algorithm. So I started some tests with a short periodic pipe (L=0.5D) screen.jpg I keep x- and z- direction grid constant, also the total number of gridpoints, just change the y value from 5e-6, 2e-6, 1e-6, 5e-7. Orthogonal quality decreases from >0.75 to >0.45 to >0.25 to >0.1 from large to small y-cell-sizes. k-Omega-SST with low-Re corrections, Fluent 15.0. I made pictures for U, k and omega and plot all four grids in one picture. U looks like this: v_total.jpg A closer look to the wall looks like this: v_wall.jpg and for k: k_wall.jpg and for omega: omega_wall.jpg Omega of course changes at the very first gridpoint, since the boundary condition depends on the y+ value. Maybe this is different for not attached flow, but for the pipe I think it's pretty safe to just not care about these "close to the wall" bad orthogonality. Far, ghost82 and kad like this. __________________ The skeleton ran out of shampoo in the shower.

September 8, 2014, 08:44
#23
Senior Member

Join Date: Feb 2013
Location: Germany
Posts: 200
Rep Power: 23
Quote:
 Originally Posted by RodriguezFatz Hey cad, I thought about your post but still don't get your point. I made a small sketch and I can't see here how the quality suffers from refining. Attachment 33560 The original (coarse) cell is the blue one, the green line shows on of the refined cells. For me, it looks like all angles keep constant.
Okay, I did not made that very clear. I did not mean coarsening in the radial ogrid direction but in the tangential direction. This means you have less nodes on the circumference of the pipe and bad aspect ratio cells on the wall.

 September 8, 2014, 08:58 #24 Super Moderator   Sijal Join Date: Mar 2009 Location: Islamabad Posts: 4,552 Blog Entries: 6 Rep Power: 53 why they consider aspect for orthogonal quality? Since it is mentioned in help you can go up to AR=1000 for single precision and 10,000 for double precision. It means this value has no significance in reality? It also implies that CFX wont care either...

September 9, 2014, 02:35
#25
Senior Member

Philipp
Join Date: Jun 2011
Location: Germany
Posts: 1,297
Rep Power: 26
Quote:
 Originally Posted by Far why they consider aspect for orthogonal quality? Since it is mentioned in help you can go up to AR=1000 for single precision and 10,000 for double precision.
As far as I understood, it is not considered directly, but really large aspect ratios deplete the correctness of the calculation of orthogonal quality (OQ) in Fluent. The question is, if just the OQ is affected, or actually the solver / any other algorithm as well.

The funny thing is, they told me that Fluent reliably supports only aspect rations up to 100 to 250. So basicallly for y+=1 you had z and y+=250.
I asked him why I would use a low-Re RANS model, if my wall grid needs to be good enough for wall modeled LES anyway. Or if that means, that Hexa meshes are not usable for low-Re Fluent RANS.
__________________
The skeleton ran out of shampoo in the shower.

Last edited by RodriguezFatz; September 9, 2014 at 09:20.

September 9, 2014, 15:44
#26
Super Moderator

Sijal
Join Date: Mar 2009
Posts: 4,552
Blog Entries: 6
Rep Power: 53
Quote:
 Originally Posted by RodriguezFatz As far as I understood, it is not considered directly, but really large aspect ratios deplete the correctness of the calculation of orthogonal quality (OQ) in Fluent. The question is, if just the OQ is affected, or actually the solver / any other algorithm as well. The funny thing is, they told me that Fluent reliably supports only aspect rations up to 100 to 250. So basicallly for y+=1 you had z and y+=250. I asked him why I would use a low-Re RANS model, if my wall grid needs to be good enough for wall modeled LES anyway. Or if that means, that Hexa meshes are not usable for low-Re Fluent RANS.
I was told that AR = 1000 is ok with single precision and 10000 with double precision.

Today I just made a mesh around a square and these are the statistics from fluent

Quote:
 Mesh Quality: Orthogonal Quality ranges from 0 to 1, where values close to 0 correspond to low quality. Minimum Orthogonal Quality = 7.07104e-01 Maximum Aspect Ratio = 1.01719e+04
ortho quality = 0.7 and aspect ratio = 10171.9 and results are good

Although i have refined the mesh in the vicinity of wall, but without using ogrid. I guess Ogrid would deteriorate the orthogonal quality.

But what is the orthogonal quality. Is it mix of min angle, determinant 2*2*2 and quality ? If these metrics are good enough, then ortho quality should also be good....

 September 10, 2014, 02:39 #27 Senior Member     Philipp Join Date: Jun 2011 Location: Germany Posts: 1,297 Rep Power: 26 I waver between trying to get a reliable answer from Ansys or just not watching this value ever again... __________________ The skeleton ran out of shampoo in the shower.