Simulation Forums

Find Simulation answers, share expertise, and connect with your peers

Simulation CFD


variation between fine and coarse mesh

110 Views, 9 Replies
02/08/2018 11:51 AM

in the attached cfz files i use region mech in one of them and tried to make the size adjustment very fine for the second one .

big variation between them, especially in the temperature profile.

it is not easy to solve for region meshing spacially for a huge number of nodes, i have 120 G ram but some time fails to solve some thing withs size

is there is any way to combine between reasonable meshing and acceptable results?

 

fine

https://drive.google.com/file/d/1eg30OQV1Y0ld6wI643xeskv75w4KmTUw/view?usp=sharing

 

coarse

 

https://drive.google.com/file/d/1UIXifkMvY9huLnOv10Wpi8WZ_51UDNON/view?usp=sharing


Report Inappropriate Content
Message 1 of 9 ( Views: 110 )

Re: variation between fine and coarse mesh

02/08/2018 1:04 PM

Hi @eng.abass5,

 

120 GB should be plenty for your analysis. Normally you need about 2 GB for each 1 million elements. So are you trying simulations with over 60 M elements? This seems excessive! 

 

Please can you provide an image of the difference between results and I will have a go with your mesh and see if we can find a good compromise. I would imagine the results could be mesh independent after with less than 10 M with optimized settings.

 

All the best,

David


David Short
Technical Support Specialist, Simulation

Report Inappropriate Content
Message 2 of 9 ( Views: 99 )

Re: variation between fine and coarse mesh

02/08/2018 2:20 PM

Hi @eng.abass5,

 

The goal of meshing in CFD is to obtain the best mesh possible without sacrificing reasonable accuracy. The balance between acceptable run time and good results is one for you to figure out, and it really is an art. 5 different CFD experts may take 5 different meshing approaches to the same problem, and all end up with acceptable results. 

 

If you would like more advice on obtaining mesh independent results, check out this webinar. In general, I recommend you start coarse and incrementally make the mesh more dense. This will confirm the mesh is only as dense as it needs to be. Throwing an unrealistic amount of mesh at the domain from the beginning is a bad approach.

 

Thanks,



Matt Bemis

Technical Support Specialist


Report Inappropriate Content
Message 3 of 9 ( Views: 93 )

Re: variation between fine and coarse mesh

02/08/2018 2:40 PM

Maybe this kind of mesh would be a good starting point for the mesh sensitivity study that Matt has suggested.

 

(.cfz attached)

starting mesh.png

 

Then start playing with the shape of the refinement region and the element spacing within it.

 

Definitely make the most of all the really useful webinars and help documentation on meshing!

 


David Short
Technical Support Specialist, Simulation
Report Inappropriate Content
Message 4 of 9 ( Views: 90 )

Re: variation between fine and coarse mesh

02/08/2018 10:45 PM

thanks matt, i watched the webinar it helpful.. but it still mater of time especially when i need to make

many modifications and test the effect of each. 


Report Inappropriate Content
Message 5 of 9 ( Views: 85 )

Re: variation between fine and coarse mesh

02/08/2018 10:48 PM

thank you Davide, let me check the variation in results. 

i will share the results for three of them later 


Report Inappropriate Content
Message 6 of 9 ( Views: 84 )

Re: variation between fine and coarse mesh

02/11/2018 5:42 PM

 

it was found that the fine meshing result is matching the experimental results 

i think it is a matter of time until finding the best possible meshing, no specific way to do it, 

mesh finely, if the number of elements more than what the memory can afford, stop the solver and make it slightly coarser. 

 

could using adaptation help to find the optimum meshing from first solve run?

 

 

543534.PNG


Report Inappropriate Content
Message 7 of 9 ( Views: 68 )

Re: variation between fine and coarse mesh

02/11/2018 7:50 PM

Going from 4.5 to 26m is a huge jump.

What I tend to do is also detailed here. Get a pretty good mesh, one that I think will be accurate and test it out. Then refine by 30% and try again, no more than that or you end up in the position you are with two results and no idea whether there is an optimal one inbetween.

 

Hope that is useful.

 


Report Inappropriate Content
Message 8 of 9 ( Views: 66 )

Re: variation between fine and coarse mesh

02/11/2018 9:26 PM

but i think since the wake size is massive, the growth rate makes the air element size big, right?

the finest meshing i found was when i use the region mesh, first time i i use suggested local mesh size 

and found the number of the element so huge. stoped solver and increased the local mesh size to the limit 

that i think my RAM can deal it. 

now i am running the solver, i think it needs two days to finish 400 iterations, i also activated 

the adaptation( 3 cycles ) this maybe needs more time. 

i know,  doesn't look a practical solution, but what to do with this big wake size :/


Report Inappropriate Content
Message 9 of 9 ( Views: 64 )