CFD code for studies on wall roughness

Francesco3097 · January 4, 2021, 14:21

Hi everyone,
I'm new in the world of CFD and I'm starting to study some topics.
I want to make a thesis based on turbulent flows with different kind of roughnesses by modelling them with different boundary conditions. I was beginning to use the code Basilisk but some users suggested to search if there were more appropriate tool for this purpose.

Since I don't have much experience on the field I'm asking you how can I approach the problem and which could be the best code or if any code is anyway a good choice.
Any advice or theoretical paper/method is welcome.
Thank you.

sbaffini · January 5, 2021, 07:31

If you are working on DNS, maybe Nek5000 and/or Incompact3D might be more suited to you (but Gerris was an equally valid choice). Where, by DNS, I also mean that you want to actually describe the roughness explicitly with its geometry and not some model.

If you are, instead, going to use a RANS or LES approach with wall functions then, probably, both Code_Saturne and OpenFOAM would be ok. Investing in OpenFOAM seems, today, a rewarding move.

In any case, you need to give more context to have more detailed advices.

Francesco3097 · January 5, 2021, 09:26

Yes you're right, I didn't right many details.
My aim is to perform DNS simulations on different roughness shapes using channels with dimensions of the order of magnitude of the roughness (1 mm). And then using the results for further studies.

In your opinion one of those you mentioned is more suited? Would it be worth to change code or they're equivalent with Basilisk (Gerry)?
Which are the main differences?

In addition, I'd like to know more about roughness modelling, where can I look for it?
Thanks.

sbaffini · January 5, 2021, 09:35

In theory, all of them are used for DNS studies (so, I don't know why someone would drive you away from Basilisk/Gerris if that is the code you know). In practice, Nek5000 has few advantages: it is a spectral code and is well accepted for DNS studies, besides the fact that it has a more vibrant community (my own perception, might be wrong); also, it can handle complex geometries fairly easily. More generally, if you are getting into DNS, Nek5000 is more commonly used and definitely the tool you want to be able to use.

For what concerns roughness modeling etc., I really have no idea.

sbaffini · January 5, 2021, 09:44

Pros of Nek5000: spectral, scales impressively, has a large DNS community, costantly used in production runs, can handle complex geometries "natively"

Cons of Nek5000: written in old Fortran style, "natively" means you have to build your grid by yourself (but there are tools to help), and it only supports hexa cells without hanging nodes (not sure if something changed on this aspect)

Pros of Gerris (I don't know basilisk): sort of automatic meshing (I guess)
Cons of Gerris: low order, might not scale as well as Nek5000, coded in C

Pros of Incompact3d: coded in a more modern Fortran, still some sort of automated meshing, should be more accurate than Gerris
Cons of Incompact3d: not actual meshing, it's immersed boundary, the less used of the group (personal perception)

Francesco3097 · January 7, 2021, 13:09

And in general we could say that Nek5000 could be easier to get familiar with, in less time? I looked a bit the documentations and Nek's seems to be well structured and there are better and richer explanations.

I received other advices, that, for example, Basilisk could be more suited for complex cases like multiphase system or non-newtonian fluid. In my case I'd like to perform under-resolved DNS in a channel, in which modifying geometries or boundary conditions I can simulate roughness.

sbaffini · January 7, 2021, 14:01

Nek is certainly well documented and getting better. As a personal perception I would say yes, easier to get acquainted to from a user perspective (but code might not be as clear or easy as the other ones).

Gerris is strictly linked to the VOF (I don't know basilisk, but I expect the same) and is thus used in the multiphase field. As a low order code is certainly less suited, with respect to the other ones, for DNS, but that wouldn't honestly stop me if it was the code I knew.

I think that your application might be more easy to be accomplished with Incompact3d, but I would nonetheless invest in Nek5000

sbaffini · January 10, 2021, 09:07

Speaking of the devil, I just saw this tool:

https://github.com/fschuch/xcompact3d_toolbox
https://xcompact3d-toolbox.readthedocs.io/en/latest/

Maybe this is going to be useful for you

Francesco3097 · January 15, 2021, 12:04

Thanks for these informations. Actually I don't know Basilisk, I'm starting to use. So, I'm going to search info also about Nek and Incompact. I'll give a look to that toolbox of Incompact and at the end I'll evaluate which will be the more suited.

January 4, 2021, 14:21	CFD code for studies on wall roughness	#1
Francesco3097 New Member Join Date: Jan 2021 Posts: 4 Rep Power: 5	Hi everyone, I'm new in the world of CFD and I'm starting to study some topics. I want to make a thesis based on turbulent flows with different kind of roughnesses by modelling them with different boundary conditions. I was beginning to use the code Basilisk but some users suggested to search if there were more appropriate tool for this purpose. Since I don't have much experience on the field I'm asking you how can I approach the problem and which could be the best code or if any code is anyway a good choice. Any advice or theoretical paper/method is welcome. Thank you.

January 5, 2021, 09:35		#4
sbaffini Senior Member Paolo Lampitella Join Date: Mar 2009 Location: Italy Posts: 2,152 Blog Entries: 29 Rep Power: 39	In theory, all of them are used for DNS studies (so, I don't know why someone would drive you away from Basilisk/Gerris if that is the code you know). In practice, Nek5000 has few advantages: it is a spectral code and is well accepted for DNS studies, besides the fact that it has a more vibrant community (my own perception, might be wrong); also, it can handle complex geometries fairly easily. More generally, if you are getting into DNS, Nek5000 is more commonly used and definitely the tool you want to be able to use. For what concerns roughness modeling etc., I really have no idea. Francesco3097 likes this.

January 5, 2021, 09:44		#5
sbaffini Senior Member Paolo Lampitella Join Date: Mar 2009 Location: Italy Posts: 2,152 Blog Entries: 29 Rep Power: 39	Pros of Nek5000: spectral, scales impressively, has a large DNS community, costantly used in production runs, can handle complex geometries "natively" Cons of Nek5000: written in old Fortran style, "natively" means you have to build your grid by yourself (but there are tools to help), and it only supports hexa cells without hanging nodes (not sure if something changed on this aspect) Pros of Gerris (I don't know basilisk): sort of automatic meshing (I guess) Cons of Gerris: low order, might not scale as well as Nek5000, coded in C Pros of Incompact3d: coded in a more modern Fortran, still some sort of automated meshing, should be more accurate than Gerris Cons of Incompact3d: not actual meshing, it's immersed boundary, the less used of the group (personal perception) acgnipper and Francesco3097 like this.

January 7, 2021, 14:01		#7
sbaffini Senior Member Paolo Lampitella Join Date: Mar 2009 Location: Italy Posts: 2,152 Blog Entries: 29 Rep Power: 39	Nek is certainly well documented and getting better. As a personal perception I would say yes, easier to get acquainted to from a user perspective (but code might not be as clear or easy as the other ones). Gerris is strictly linked to the VOF (I don't know basilisk, but I expect the same) and is thus used in the multiphase field. As a low order code is certainly less suited, with respect to the other ones, for DNS, but that wouldn't honestly stop me if it was the code I knew. I think that your application might be more easy to be accomplished with Incompact3d, but I would nonetheless invest in Nek5000 Francesco3097 likes this.

Similar Threads
Thread	Thread Starter	Forum	Replies	Last Post
Wall y+, surface roughness and inflation layer CFX	Prak_32	CFX	3	December 12, 2018 05:56
Basic Nozzle-Expander Design	karmavatar	CFX	20	March 20, 2016 08:44
Assigning Wall Roughness in STAR-CD	Sachin Patil	Siemens	0	March 6, 2008 03:05
Multicomponent fluid	Andrea	CFX	2	October 11, 2004 05:12
public CFD Code development	Heinz Wilkening	Main CFD Forum	38	March 5, 1999 11:44

January 5, 2021, 07:31		#2
sbaffini Senior Member Paolo Lampitella Join Date: Mar 2009 Location: Italy Posts: 2,152 Blog Entries: 29 Rep Power: 39	If you are working on DNS, maybe Nek5000 and/or Incompact3D might be more suited to you (but Gerris was an equally valid choice). Where, by DNS, I also mean that you want to actually describe the roughness explicitly with its geometry and not some model. If you are, instead, going to use a RANS or LES approach with wall functions then, probably, both Code_Saturne and OpenFOAM would be ok. Investing in OpenFOAM seems, today, a rewarding move. In any case, you need to give more context to have more detailed advices.

January 5, 2021, 09:26		#3
Francesco3097 New Member Join Date: Jan 2021 Posts: 4 Rep Power: 5	Yes you're right, I didn't right many details. My aim is to perform DNS simulations on different roughness shapes using channels with dimensions of the order of magnitude of the roughness (1 mm). And then using the results for further studies. In your opinion one of those you mentioned is more suited? Would it be worth to change code or they're equivalent with Basilisk (Gerry)? Which are the main differences? In addition, I'd like to know more about roughness modelling, where can I look for it? Thanks.

January 7, 2021, 13:09		#6
Francesco3097 New Member Join Date: Jan 2021 Posts: 4 Rep Power: 5	And in general we could say that Nek5000 could be easier to get familiar with, in less time? I looked a bit the documentations and Nek's seems to be well structured and there are better and richer explanations. I received other advices, that, for example, Basilisk could be more suited for complex cases like multiphase system or non-newtonian fluid. In my case I'd like to perform under-resolved DNS in a channel, in which modifying geometries or boundary conditions I can simulate roughness.

January 10, 2021, 09:07		#8
sbaffini Senior Member Paolo Lampitella Join Date: Mar 2009 Location: Italy Posts: 2,152 Blog Entries: 29 Rep Power: 39	Speaking of the devil, I just saw this tool: https://github.com/fschuch/xcompact3d_toolbox https://xcompact3d-toolbox.readthedocs.io/en/latest/ Maybe this is going to be useful for you

January 15, 2021, 12:04		#9
Francesco3097 New Member Join Date: Jan 2021 Posts: 4 Rep Power: 5	Thanks for these informations. Actually I don't know Basilisk, I'm starting to use. So, I'm going to search info also about Nek and Incompact. I'll give a look to that toolbox of Incompact and at the end I'll evaluate which will be the more suited.