[Markua]

Good Afternoon fellas,

I should start out by saying I'm pretty new to Ansys itself and my knowledge is going to be a bit unfulfilling, but I have put about 50 hours into learning the program so far and have searched up the errors I'm about to ask about with little-to-no luck.

So, I'm trying to generate a mesh for a 2D airfoil for CFX (I will learn Fluent after I'm happy with my knowledge in this program) that will be tested over angles of attack 0-20.

I have done this test once and the data I got was acceptable, but I wanted to increase all of the meshing to about twice as fine and re run the test to determine how much accuracy is sacrificed/gained as mesh decreases/increases in quality. This is where the program starts getting finicky.

I found that the mesh, which was failing very often, would work more often with a less fine inflation layer around the profile. I could somewhat deal with this. I suppose the computer just couldn't work it out, although it did not seem to me to be half as complex as any other mesh I've ever seen.

Now this problem becomes much more annoying when I set my computer to analyze 21 different design points (the angles of attack) and when I come back to the computer, I find that 3-4 points have failed and require a lower intensity inflation, so I need to set all design points to the lowest common denominator which decreases the quality of all of the other design points that would have been able to handle the better inflation. So that's problem number one.

Problem number two, I noticed when I was looking through the down-force generated by each angle, and I saw two or three that didn't quite seem right. Checking these specific point's mesh showed that the inflation layer didn't generate at all. No error, no warning, but no inflation layer. This is dangerous, as I can't check that each angle has properly developed an inflation layer, and ANSYS won't tell me, I'm going to eventually give someone wrong data and waste someones time (read: my own).

So my setup has a lower refinement rectangle, with a high refinement rectangle inside that, and the 2D aerofoil inside that (well, it's all extruded 1mm anyway). The rectangles both have face sizing and a sweep to go the 1mm back, and the wing profile has an edge sizing and an inflation.

I'm going to try include images of the lower quality mesh working, the higher quality mesh failing, and a set of inflation setting that works for 16/21 of the design points.

If anyone can give me more information on this, I'd be very grateful.





e: Can't see the images uploaded on my end. Try these links:
http://imgur.com/pMNV0fY
http://imgur.com/INSuKHQ
http://imgur.com/eLhC63M

[hwet]

The question isnt clear to me, although you wrote a whole essay, summarize please and give more details about the inflation failing and a closeup of the inflation as well.
Have you tried tet mesh instead?

[Markua]

In summary:

I did an analysis of the down-force produced by a 2D wing profile over a series of angles of attack.

One issue was that some angles of attack would be fine with a specific set of inflation settings (say, max thickness of 5 mm, 10 layers), while others AoA's could not handle this and would fail. This was a huge problem as I designed the inflation settings to work with 0 AoA, then set the program to calculate 20 other AoA's for an hour and would come back to find a handful failed. Why would some angles change the amount of inflation I can set, and how can I rectify this, without losing mesh quality on the good design points.

The second error was that some of these angles of attack just did not attempt to produce an inflation at all, without any warning for me. This seems like a pretty big deal, but I just cannot work out what caused this. The mesh follows all other setting (face sizes, etc) but does not apply the inflation. I only found out that the mesh didn't produce the inflation because a couple of angles produced undesired results, so I individually checked the meshing. I wouldn't want to have to check the mesh for 21 or more different angles.

Here is an image of a working inflation setting, where increasing the layer count by 1 or 2 makes it fail: http://imgur.com/rZDnF3Y

[hwet]

One issue was that some angles of attack would be fine with a specific set of inflation settings

Angle of attacks would be fine or the results from those meshes would be fine? Sweep method inflation can be a bit tricky, you have to manually select base face and all.

I can assure you that tet mesh with inflation will give you the same results.

Also to check inflation layer you should see the Y+ values in post processing, just keeping a random inflation thickness is not a good idea.

[Markua]

Angle of attacks would be fine or the results from those meshes would be fine?

Some angles of attack would produce a mesh that works.

I have since removed sweep as an option and have received the same results, but also the exact same error. I will learn more about this Y+ value and how to choose an inflation thickness based on it.

[hwet]

ok so the very first error is because you mesh the whole thing in two parts? So one part is meshed with the size functions you apply and the other is meshed upto its boundary, with the same functions so the software gets confused when you mesh the second part with different size controls.

For the Y+ take a look at this

http://www.computationalfluiddynamics.com.au/tips-tricks-turbulence-part-4-reviewing-how-well-you-have-resolved-the-boundary-layer/

.


The rest of the error msgs i cant see in full so dont know whats going on.

[Markua]

Thankyou for the information on Y+ and determining how much inflation is required, I found it quite informative.

Reading this, I note that the turbulent areas and boundary layers are always going to be varying with different AoA's. Does that mean that my automation of checking each AoA will fail as I am only setting a suitable inflation for a specific angle?

I believe the interface between the two different domains works fine, as I if the mesh fails to produce, and I tell it to only generate for the main domain, it still fails. That shows that the failure is not due to the interface between the two.

Note that the other problem still sometimes rears it's ugly head: sometimes the mesh just does not generate the inflation at all and does not tell me. Is this common?

[hwet]

If it doesnt create inflation, it will tell you why is it doing so. I think at some angles it cannot create a hex mesh and hence inflation fails as well.

Also I think you are using multizone and not actually sweeping? Or are you using hex dominant meshing. What method are you using, doesnt appear that you are sweep meshing.

[Markua]

The only sweep functions I had were to cross the 1mm extrusion of the faces. I have since removed any sweeping.

I have had little-to-no mesh failures after reducing how intense my inflations were, but I still get non-convergent solutions resulting in overflow at seemingly random AoA's. For example, I'll post a working/converging and a failing/diverging solution to see if you can notice where I have gone wrong.

So this picture shows that over AoA's 0-20, only 2,5,6 and 12 failed:
http://imgur.com/dWxM3rB

This is the working/converging data for an AoA of 1:
Mesh: http://imgur.com/SlntYRY
Solver: http://imgur.com/KBz9V7j

This is the failing/diverging data for an AoA of 2:
Mesh: http://imgur.com/75b2ihd
Solver: http://imgur.com/KzEoyQI

I just can't wrap my head around how AoA's 1 and 3 would work but 2 would not.

[Markua]

Originally Posted by hwet

If it doesnt create inflation, it will tell you why is it doing so.

I'd also like to address this. I'm not imagining it.

This image shows how I noticed the error: http://imgur.com/Wt8MRlL
The massive deviations from expected data had to be something going wrong, but no errors are put out

And this image should show the error: http://imgur.com/kNHbUf0
The inflation is not suppressed or anything like that.

Please tell me if the second image is unclear to what I am trying to show.

[hwet]

Your solution diverging is a totally different problem, the mesh could be the reason but there could also be many other reasons.

Inflation is necessary, however, the thickness should be big enough to avoid too many cells in the boundary layer, I am quite sure you can find airfoil simulation tutorials on the ansys website or just on google.

Maybe start with them first.

[Markua]

So my knowledge up to this stage is from doing a few tutorials and a three day seminar to learn ANSYS. I'm now trying to come away from teh tutorials and do it on my own.

Your solution diverging is a totally different problem

While the divergence is a different problem from my mesh failing. I think the underlying question remains, is ANSYS really so picky that a 1 degree change in angle could throw out the meshing so badly that either the meshing would fail or the solver would diverge? And then increasing by a second degree could make everything fine again?

Also, the other problem of the inflation not showing up here and there still has not been solved. I obviously would like to fix that, but I'd be happy just to know whether this has happened to anyone else before.

Thanks, hwet.

[hwet]

No it is not that picky, maybe you are making a mistake somewhere. With the inflation layer does the mesh fail when you have tet mesh, I have said this before try tet mesh.

Maybe you are giving wrong boundary conditions.