[ternik]

Hi,

when simulating flow past cylinder one will (most likely) end up with the non-orthogonal mesh. How many "NonOrthogonalCorrectors" should be used? In OF-User Guide it is written that 20 "NonOrthogonalCorrectors" should be used for the most non-orthogonal meshes... Non-orthogonality of my mesh (after running checkMesh) is:

• Max: 44.7371
• Average: 16.8382

Thanks for your advice,
Primoz.

[gricci]

Maybe someone more expert will correct me, but i believe there is no fixed rule...
up now i've often set the number of nonOrthogonalCorrectors comparing the velocity profiles on a potential flow. Most of the time, I've noticed that 2 correctors are sufficient.
hope it helps

[sega]

I can confirm gricci's experience.
I have played around with the number of non-orthogonal corrector loops and found that 2 are sufficient.
If I used more than 2 corrector loops the computational time was severely increased without improving the solution.

[ternik]

Originally Posted by sega

I can confirm gricci's experience.
I have played around with the number of non-orthogonal corrector loops and found that 2 are sufficient.
If I used more than 2 corrector loops the computational time was severely increased without improving the solution.

Hi Sega (& Gricci),

thank you for your reply. I will give it try with with 2 (o.k. maybe 3) non-orthogonal corrector loops and... Non-orthogonal corrector loop(s) corrects the solution of pressure filed and velocity filed as well?

With greetings from Slovenia,
Primoz.

[alberto]

I agree with 2 or 3. If you need more, it might be the case of reconsidering the mesh

[hjasak]

On this mesh you do not need non-orthogonal correctors at all. There's usually nothing to worry about until the max non-orthogonality angle is approx 70 deg.

Hrv

[ternik]

Originally Posted by hjasak

On this mesh you do not need non-orthogonal correctors at all. There's usually nothing to worry about until the max non-orthogonality angle is approx 70 deg.

Hrv

Thanks for this - but using non-orthogonal correctors (for example 2 or 3) will result in "more accurate" results or...

Regards,
Primoz.

[hjasak]

Not really "more accurate". If you are running steady, you are doing iterations that will correct non-orthogonality (among other things), and in transient, you are running multiple PISO/SIMPLE correctors which will do the same thing.

Non-orthogonal correctorsare here to save you if your code is blowing up because the mesh is so non-orthogonal that the first solution is driving the velocity to be stupid. If your velocity is OK, you just keep doing "normal" correctors, without special need for non-orthogonal ones.

I use them on bad meshes (some people call them "industrial") when the solver is giving me trouble. Usually, 1 is enough, and I never used more than 3.

Hope this helps,

Hrvoje

[ternik]

Originally Posted by hjasak

Not really "more accurate". If you are running steady, you are doing iterations that will correct non-orthogonality (among other things), and in transient, you are running multiple PISO/SIMPLE correctors which will do the same thing.

Non-orthogonal correctorsare here to save you if your code is blowing up because the mesh is so non-orthogonal that the first solution is driving the velocity to be stupid. If your velocity is OK, you just keep doing "normal" correctors, without special need for non-orthogonal ones.

I use them on bad meshes (some people call them "industrial") when the solver is giving me trouble. Usually, 1 is enough, and I never used more than 3.

Hope this helps,

Hrvoje

Yep, this helps. Thanks again for really helpful information on this subject.

Currently, my mesh is "academic"

With best wishes,
Primoz.