what is the proper aspect ratio?
 

Hello every one,
I read somewhere that according to ANSYS documentation maximum aspect ratio is around 18-20. If the maximum aspect ratio is lower than these values mentioned, does that make a problem?
I have an O-grid mesh and my geometry is a simple cylinder where I have two phase flow(a bubble column), when I ask for mesh quality in fluent (since I don't know how to check it in CFX, I check it in fluent but i do my runs and solution in CFX.) it gives these values:
Mesh Quality:
Minimum Orthogonal Quality = 8.39805e-01
(Orthogonal Quality ranges from 0 to 1, where values close to 0 correspond to low quality.)
Maximum Ortho Skew = 1.60195e-01
(Ortho Skew ranges from 0 to 1, where values close to 1 correspond to low quality.)
Maximum Aspect Ratio = 9.29158e+00
do you think it is good or I need to do a mesh refinement?
Thanks for your answer in advance.

That is a nice looking mesh strategy. Nothing wrong with those shapes or aspect ratios. It may be a little coarse, depending on what you are doing, but that is why you do grid dependence study. Start coarse, then reduce, and make sure you are getting similar enough answers.

The mesh quality you require depends on what yo are simulating. If you are modelling a single phase low Reynolds number flow then you can use high aspect ratio elements and big expansion ratioes and it will work fine. If you are simulating surface tension in free surface models you have extremely strict limits on mesh quality - aspect ratioes bigger than 1.2 are bad. (You won't get the laplacian pressure of a spherical drop correct with an aspect ratio bigger than 1.2 - if you don't believe me give it a try:) )

Flows with shock waves are reasonably sensitive to mesh quality, many multiphase models are quite sensitive as well.

So the mesh quality comments in the output file and documentation are just guides. If you want to be really sure you need to check the mesh quality requirements for your model.

Dear ghorrocks
Thanks for your answer.
so this limitations are not specific and I should ignore them and do mesh independence study. right?
In modelling a bubble column, which is 2 phase flow, it is said that too much grids will cause inaccuracy. they usually use less grid numbers in order to eliminate some of the turbulent effects and get an accurate result. In this case, do you think aspect ratio less than 1.2 is still necessary?

Dear evcelica
Thanks for your answer, so you think that this mesh is reasonable?
i have used a coarser mesh before this with aspect ratio 11. and it was not accurate. this is the 2nd mesh which is 48000 cells(about 1.5 times more than the 1st one),
should i care about the aspect ratio, because it is not in the range advised by ANSYS...:o:o

The surface tension free surface multiphase model has some specific aspects of it which make it extremely sensitive to mesh quality. If you are not using the surface tension model your mesh quality requirements will be relaxed - but how far they are relaxed will depend on what you are modelling.

I would suspect your bubble column, assuming you are modelling this with a eularian bubble model (not a free surface model) would not be very strict on mesh quality. But this is just my guess, you should check this.

So you should do a series of models where you deliberately modify the mesh over a range of qualities and see the effect on parameters important to you. Then you will determine how sensitive your case is to mesh quality.

Dear ghorrocks
Yes, I do use Eulerian-Eulerian model. But there is this surface tension that I activate in settings of my CFX simulation (attached picture), is this considered important in mesh quality?

What does Glenn write in his first sentence that you quote? I think his text is more than clear, not?

Sorry...but I did not understand the difference between this surface tension that I activate and set in my solution and the surface tension free surface multiphase model he mentioned...:(:(:( I don't know if he meant this surface tension or not?

When you activate "this surface tension", then you end up in the extremely sensitive mesh dependency.

Dear Gert-Jan
Thank you so much. Now I understand it.:):)

with this answer I suppose that Y+ will be important as well.