CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   ANSYS Meshing & Geometry (http://www.cfd-online.com/Forums/ansys-meshing/)
-   -   [ANSYS Meshing] Inflation layer / Element quality (http://www.cfd-online.com/Forums/ansys-meshing/112163-inflation-layer-element-quality.html)

Spookz January 23, 2013 04:43

Inflation layer / Element quality
 
Hi,

I am meshing a relatively simple geometry (tool) and I have add inflation layer to get the temperature boundary layer. (I want to simulate heat transfer)
The problem is that the first layer of the prisms has to be 0,04 mm (16 layers) to get the y+ at about 1. As a consequence the mesh quality near the inflation layer and in the layer is getting bad (max. skewness near 1 and min. orthogonality near 0).
Is there a way to get better quality?
Do the prisms even need to have the same quality as tetras?

Thanks for answer!

PSYMN January 24, 2013 07:55

It depends on the metric and the solver, but generally speaking skewness and orthogonality are important for inflation.

However, having a thin layer does not always lead to issues like this. Is there some weird kink in the geometry that the prisms are having trouble getting around? Is this just a few bad cells in one location?

A screen shot of a cross section thru the bad area (zoomed in) would help

Spookz January 24, 2013 10:34

Hi Simon,


The bad quality elements are in many regions of the tool (for 0,04 mm).
I postet a link to the picture, so that you can see it.

Actually there are almost only square angles, that the prisms have to get around. Excpet a 1 mm thick and slightly bend plate.

The minimum height of the first prism layer i can define is 0,3 mm.
When i get lower the element quality decreases rapidly.

For 0,2 mm there are only a few (maybe 30) elements that have slightly bad quality in different regions. Does that matter concerning convergence in the simulation?

http://imageshack.us/photo/my-images/854/meshe.jpg/

http://imageshack.us/photo/my-images...rismlayer.jpg/

http://imageshack.us/photo/my-images...delements.jpg/

http://imageshack.us/photo/my-images...elements2.jpg/

Solver is CFX

Peter_J February 2, 2013 14:01

It was always my concern, how to put together two rules of thumb:

1. Satisfy the Y+ criterion (impact on the first inflation layer thickness)
2. Satisfy the skewness and orthogonal quality metrics

For high values of the Re number and big obstacles, the required first layer thickness is really small, in order of magnitude of 1E-03 where Y+ is about 150.
If You do so, the other edges of the inflation grid should also be alike, in order to maintain satisfactory mesh metrics. Of course this leads to great number of cells, and significant rise of computation time and power.

Spookz February 4, 2013 05:03

Thanks for the answer Peter.

My problem is that in my case its about heat transfer and I have read, that
y+ should be smaller or equal 1. But for y+=1 and my Reynolds No. (~550000) the first layer has to be 0.04 mm thick. This leads to a bad element quality near the inflation layer in the transition region to the tetras.

Is there a problem, if there are just few elements of bad quality or does even this lead to bad results or no convergence?

Peter_J February 4, 2013 13:23

Unfortunate it can lead to convergence trouble. There are two solutions that might help:
1. Improving the mesh quality in inflation layer by local sizing functions
2. Switching to double precision

PSYMN February 6, 2013 14:53

I talked to someone here who mentioned that they had this problem at GM and the solution was to turn off the prism smoothing option... Maybe give that a try.

Spookz February 7, 2013 04:24

I am not using the global inflation layer option.
So unfortunately there isn't any smoothing option.

strobel February 7, 2013 11:00

i'm getting this kind of problem as well. This is one of reasons i started to use IcemCfd.

PSYMN February 7, 2013 15:55

If you are in ANSYS Meshing and you switch prism from "pre" inflation to "post" inflation, then you start to use the ICEM CFD prism meshing tool in ANSYS Meshing... Maybe that will work for you.

Spookz February 7, 2013 16:16

How can I use that Prism Tool from ICEM in Ansys Meshing.
Aren't these two different meshing tools?
Or do you mean I should just import a mesh without inflation in ICEM and addthe Prisms there?

PSYMN February 7, 2013 17:34

ICEM CFD Technology is used under the hood in ANSYS Meshing.

If you use "Pre" inflation, you are using TGrid Prism. If you switch to "Post" inflation, you are using the ICEM CFD prism algorithm.

Far February 7, 2013 17:38

Why pre-prism is set to T-grid and post-prism to ICEM?

PSYMN February 7, 2013 17:42

In the background it calls the TGrid or ICEM CFD methods...

Pre Prism means that Prism runs (using the TGrid algorithm) before the tetra volume is filled.

Post Prism means that prism runs (using the ICEM CFD algorithm) after the tetra volume is filled.

The naming is silly since you don't really see the steps, but that is what happens.

Best regards,

Simon

Far February 7, 2013 17:53

Quote:

Originally Posted by PSYMN (Post 406622)
In the background it calls the TGrid or ICEM CFD methods...

Pre Prism means that Prism runs (using the TGrid algorithm) before the tetra volume is filled.

Post Prism means that prism runs (using the ICEM CFD algorithm) after the tetra volume is filled.

The naming is silly since you don't really see the steps, but that is what happens.

Best regards,

Simon

Thats what I am interested to know. Why Tgrid algorithm is used when we want pre-inflation and ICEM prism algorithm is used when we want post inflation.

In other words Tgrid is strong in pre-inflation and ICEM is strong in post inflation? right?

PSYMN February 7, 2013 19:05

Tgrid and ICEM CFD just work differently.

TGrid can not do post inflation at all. It was just designed for Pre-inflation.

ICEM CFD can do pre-inflation, but it is not our normal mode or best practice.

Why were both options included in ANSYS Meshing? Because they each have advantages and each have users who prefer them.


All times are GMT -4. The time now is 21:35.