Converge the numerical result

• 2.1K Views
• Last Post 25 January 2020
• Topic Is Solved
José Mantovani posted this 08 December 2017

Gentlemen, I am trying to converge the value of the numerical result of the lift coefficient  for the experimental result of the airfoil S1223. I've tried it in several ways and am getting values between Cl = 0.95 and Cl = 0.98, but the experimental value is Cl = 2.23. This for conditions of approximately Re = 380000 and an angle of attack alpha= 10 degrees. In my opinion the problem is the mesh, particularly at the trailing edge of the airfoil ... How can I improve the mesh in this region? I think that if a simulation was done in the edX course for a validation case of the NACA0012 airfoil with a relatively simple mesh and cells similar, we can see in below image, in the trailing edge region, I can also obtain a numerical result close to the result experimental, performing knit procedure in the same way ...

Or does anyone know what can be done so that the value approaches the experimental value?

Gentlemans, thanks for help and attention!

peteroznewman posted this 08 December 2017

Mantovani,

If you slice the surface at the trailing edge, you can get good quality elements.

Regards,

Peter

• Liked by
peteroznewman posted this 09 December 2017

Here is that technique applied to S1223.

• Liked by
José Mantovani posted this 09 December 2017

Thank's so much Peter! Very nice!!!

peteroznewman posted this 14 December 2017

Mantovani,

I read Vishal's answer to your other question, and watched the video he recommended and learned how to calculate the first element thickness using a Y+ calculator. For your Reynolds number and the air properties in Fluent, I got a first layer thickness of 5.6e-2 mm. The video also used 60 layers of inflation. You should update your mesh to these values.

Regards,

Peter

• Liked by
José Mantovani posted this 14 December 2017

But Peter, I need update my mesh ou the mesh you sent me by email?

peteroznewman posted this 14 December 2017

Yes, I just watched that video this morning, didn't you watch it last week?

• Liked by
José Mantovani posted this 14 December 2017

I had seen yes, even tried to apply my meshes, the values that I obtained in this calculator of Y +. I run the simulations here and the value was close to the experimental one, it's acceptable, I'll change it and see it gets even closer. Many thanks for the help Peter! Really, thank you so hiper very much! Come here to Brazil for we drink some beer! hahahahaha

Hugs.

peteroznewman posted this 14 December 2017

I have been fighting with inflation and gave up. I am now doing face meshing with biased edge sizing. I have the first layer thickness < 0.05 mm.

I have a nice "inflation" layer manually created after slicing the geometry.

I extended the domain to the size recommended in the video.

I'll send you this file.

Cheers,

Peter

sai193272 posted this 05 October 2018

Dear Peter,

Thanks so much for the advice to partition the domain. I am working currently on a 2D airfoil and have been trying to implement what you did on the S1223 in order to fix the skewness at the trailing edge. However, after splitting the domain (using split face) in design modeller and applying inflation to the airfoil and the new intersections, the mesh fails. If possible, can you detail your procedure slightly more?

Sergio

• Liked by
José Mantovani posted this 05 October 2018

Hello dear sai.

I exactely tested this approach to make a mesh to S1223, whats is sligthy different compared with NACA0012, because the S1223 is a no symmetrical airfoil.

At link below, have a way to how make a mesh to a not symmetric airfoil, in the case Clark-Y, but you can use this approach to do a similar mesh for S1223 airfoi. This PDF is a document of CFD Class of Iowa University.

http://user.engineering.uiowa.edu/~me_160/CFD%20Labs/Lab2/Intermediate%20Lab%202%20Manual.pdf

I hope helped you. Good weekend.

Mantovani.

sai193272 posted this 08 October 2018

Hey,

Thanks very much!

I tried implementing it but I get this weird mesh where the element size becomes larger initially (which it should) but then becomes smaller again towards the inlet. I made sure to assign the same number of divisions along the edges and used the same biases on parallel edges.

as a result, elements become skewed rather than being rectangular as in the example.

Do you know what the issue could be?

Thanks again in advance, much appreciated!

• Liked by
sai193272 posted this 08 October 2018

It seems my settings are being overridden somehow, changing bias for example has no effect at all

• Liked by
José Mantovani posted this 08 October 2018

This is due to the proportion of Edge Sizing. It seeks to correlate a ratio with the size of the domain. Try replicating exactly like the link I submitted. Use even the same domain size and replicate size limitations. See what happens.

I do not think I have the file with the mesh here, I'll take a look. Try to do this and share here if you have succeeded, if not, I search my files and show you.

Mantovani.

• Liked by
sai193272 posted this 08 October 2018

Is there a way to increase the domain size and maintain the same ratio as in the Iowa example? The airfoil used in the example has a cord of roughly 0.300m.  The one I am trying to analyse is about 8 times as large.

José Mantovani posted this 08 October 2018

So, use this to create your domain. Where is one in the Iowa example, in your will be eight. Ok?

In aerodynamic analysis is usually use the chord length around 1 meter. You can also change the chord length of your aerodynamic profile S1223.

• Liked by
sai193272 posted this 08 October 2018

That doesn't work unfortunatley.. The result looks similar to the picture above (with a finer mesh).

Is there perhaps a way to prevent the program from automatically seeking to relate the domain size to edge size?

• Liked by
José Mantovani posted this 08 October 2018

You are having these errors because you are putting different values for number of divisions. Ensure that the values of the region behind the airfoil, the square region of the domain are with similar values for vertical edges with vertical edges and horizontal with horizontal edges. The vertical edges must have the same dimensions as the edges that are directly attached to the C part of the domain. If you have a domain with a height of 50 m and you define division numbers to 47 you will surely get this error. Look for exact values, or in relation to your domain.

For example, if you have a domain of a rectangular channel of 5mx10m and use division numbers of 7x14 you will have a mesh similar to yours.

• Liked by
sai193272 posted this 08 October 2018

Right. I took that into account bu the problem remains.

If you don't mind, can you have a quick look over my mesh file? Maybe there is something obvious that I am missing.

José Mantovani posted this 08 October 2018

I will try to make a mesh over S1223 geometry here, because I lost the files of this simulation. So on I share here.

• Liked by
sai193272 posted this 08 October 2018

Thanks, that would be amazing!

Vinod Kumar posted this 18 January 2019

Hello Peterozenewman, Can you please tell me about this " slice the surface at the trailing edge". I am facing the same problem high skewness near sharp edges/trailing edge.

mjameel posted this 25 January 2020

Hello Peter, can you share with me the Video? I am also dealing with an airfoil with an extended span (thickness). I tried all the necessary methods in the Fluent meshing but am still having a problem with limitation of inflation in relation to Multizone method and face meshing method.

peteroznewman posted this 25 January 2020

The video is in this discussion.