CFD Online Logo CFD Online URL
www.cfd-online.com
[Sponsors]
Home > Forums > General Forums > Main CFD Forum

Can I just use a coarse mesh as my final mesh size?

Register Blogs Members List Search Today's Posts Mark Forums Read

Like Tree5Likes
  • 1 Post By aerosayan
  • 1 Post By FMDenaro
  • 2 Post By FMDenaro
  • 1 Post By FMDenaro

Reply
 
LinkBack Thread Tools Search this Thread Display Modes
Old   January 16, 2021, 04:41
Default Can I just use a coarse mesh as my final mesh size?
  #1
Member
 
Join Date: Feb 2019
Posts: 65
Rep Power: 7
cfdnewb123 is on a distinguished road
I am testing my model, specifically for RANS and it is able produce good result except for anisotropic shear stress (uv) profile where the oscillation is rampant. I can eliminate these oscillation by using a coarser mesh but I am wondering whether is this the right thing to do and does reviewer reject this practice?

However, I also partly have no idea what is the 'appropriate' mesh size. For instance, when I was working on 2D backward test case, my coarsest mesh was >60k cells. However I came across a journal where they use 61 x 41 grid cells and now it make me wonder whether am I over-estimating the required mesh density.

Besides it does make sense to not use such a high mesh density as I am working on RANS and high mesh density will capture unsteady effects which will make prevent RANS from converging. In that case, is it okay to use a coarser mesh for my research and is there a standard number? For instance, for 2D case (e.g. backstep, hills), can I just use mesh number on the order of thousands?
cfdnewb123 is offline   Reply With Quote

Old   January 16, 2021, 05:10
Default
  #2
Senior Member
 
Sayan Bhattacharjee
Join Date: Mar 2020
Posts: 495
Rep Power: 8
aerosayan is on a distinguished road
Quote:
Originally Posted by cfdnewb123 View Post
However I came across a journal where they use 61 x 41 grid cells and now it make me wonder whether am I over-estimating the required mesh density.
Be careful with that. I don't know which paper you selected. But many of the papers that deal with these standard problems, are written in 1970s or 1990s. They could get away with 50 x 50 grids for simple cases, but if your problem is using a different method, or trying to solve for more complex physics like turbulent flow, the coarse mesh may not be acceptable.


If the methods used in the reference paper, and your experiments, are the same, they could be acceptable; but personally, I would like to create a denser mesh for my own sake, and ensuring that there isn't a high difference between the results of the coarse and fine grids.
aero_head likes this.
aerosayan is offline   Reply With Quote

Old   January 16, 2021, 05:26
Default
  #3
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 6,764
Rep Power: 71
FMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura about
Quote:
Originally Posted by cfdnewb123 View Post
I am testing my model, specifically for RANS and it is able produce good result except for anisotropic shear stress (uv) profile where the oscillation is rampant. I can eliminate these oscillation by using a coarser mesh but I am wondering whether is this the right thing to do and does reviewer reject this practice?

However, I also partly have no idea what is the 'appropriate' mesh size. For instance, when I was working on 2D backward test case, my coarsest mesh was >60k cells. However I came across a journal where they use 61 x 41 grid cells and now it make me wonder whether am I over-estimating the required mesh density.

Besides it does make sense to not use such a high mesh density as I am working on RANS and high mesh density will capture unsteady effects which will make prevent RANS from converging. In that case, is it okay to use a coarser mesh for my research and is there a standard number? For instance, for 2D case (e.g. backstep, hills), can I just use mesh number on the order of thousands?



First of all, the quality of the mesh does not depend only on the total number of nodes, you have to specify the details of the flow problem.


However, a reviewer will ask you for a grid refinement assessment until a grid independent solution is reached. RANS produces a solution largely driven by the turbulence model, thus you have to demonstrate that the grid resolution is enough fine to let only the model acts.



And not at all, a high mesh density will not capture unsteady effects. That depends on the proper formulation for solving transient problems.
aero_head likes this.
FMDenaro is offline   Reply With Quote

Old   January 16, 2021, 05:37
Default
  #4
Member
 
Join Date: Feb 2019
Posts: 65
Rep Power: 7
cfdnewb123 is on a distinguished road
Quote:
Originally Posted by aerosayan View Post
Be careful with that. I don't know which paper you selected. But many of the papers that deal with these standard problems, are written in 1970s or 1990s. They could get away with 50 x 50 grids for simple cases, but if your problem is using a different method, or trying to solve for more complex physics like turbulent flow, the coarse mesh may not be acceptable.
Is my case acceptable when I follow their methods (e.g. mesh density, same RANS model) except that I will also be testing against my model which is just a modification of one of the model used?


Quote:
Originally Posted by aerosayan View Post
If the methods used in the reference paper, and your experiments, are the same, they could be acceptable; but personally, I would like to create a denser mesh for my own sake, and ensuring that there isn't a high difference between the results of the coarse and fine grids.
Yes, I will also test my model on three layers of mesh density, with each refinement by 2. But the default mesh I will be using is same as the one used from the journal. However, I am unsure if it is okay to use a default mesh size number in the thousands range for 2D case?
cfdnewb123 is offline   Reply With Quote

Old   January 16, 2021, 05:41
Default
  #5
Member
 
Join Date: Feb 2019
Posts: 65
Rep Power: 7
cfdnewb123 is on a distinguished road
Quote:
Originally Posted by FMDenaro View Post
First of all, the quality of the mesh does not depend only on the total number of nodes, you have to specify the details of the flow problem.


However, a reviewer will ask you for a grid refinement assessment until a grid independent solution is reached. RANS produces a solution largely driven by the turbulence model, thus you have to demonstrate that the grid resolution is enough fine to let only the model acts.



And not at all, a high mesh density will not capture unsteady effects. That depends on the proper formulation for solving transient problems.
Yes, I will test my model on three levels of mesh density, each refinement by 2.

However, may I asked what is the general acceptable number of mesh cells for 2D case? This is because, when I test my model at very high mesh cell number (e.g. >100k), the general flow field (e.g. flow separation) does not vary much compared to low mesh cell number (e.g. >1k), except that the uv profile becomes highly oscillatory at high mesh density. In that case, is it okay to just stick to the thousand range as I do not have to deal with the issue of uv profile oscillation?
cfdnewb123 is offline   Reply With Quote

Old   January 16, 2021, 05:49
Default
  #6
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 6,764
Rep Power: 71
FMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura about
Quote:
Originally Posted by cfdnewb123 View Post
Yes, I will test my model on three levels of mesh density, each refinement by 2.

However, may I asked what is the general acceptable number of mesh cells for 2D case? This is because, when I test my model at very high mesh cell number (e.g. >100k), the general flow field (e.g. flow separation) does not vary much compared to low mesh cell number (e.g. >1k), except that the uv profile becomes highly oscillatory at high mesh density. In that case, is it okay to just stick to the thousand range as I do not have to deal with the issue of uv profile oscillation?



Again, is not the total number of nodes that defines the quality or the good level of resolution.

That depends on the geometry, Reynolds number, type of BCs at the wall.
A reviewer of a relevant journal would not accept only a number of cells. For some 2D flow problems a 100k number could be ridiculous.
aerosayan and aero_head like this.
FMDenaro is offline   Reply With Quote

Old   January 16, 2021, 22:21
Default
  #7
Member
 
Join Date: Feb 2019
Posts: 65
Rep Power: 7
cfdnewb123 is on a distinguished road
Quote:
Originally Posted by FMDenaro View Post
Again, is not the total number of nodes that defines the quality or the good level of resolution.

That depends on the geometry, Reynolds number, type of BCs at the wall.
A reviewer of a relevant journal would not accept only a number of cells. For some 2D flow problems a 100k number could be ridiculous.
Thanks. May I asked whether is it a must to show all the data? Because I found some journals/conferences which only show part of the data, e.g. skin friction coefficient but no velocity profiles even though the paper they are referencing to contain data for velocity profiles too. I know it is a must to report everything but if the papers I am referencing from do not include some of the data, then should I follow their practice?
cfdnewb123 is offline   Reply With Quote

Old   January 16, 2021, 22:48
Default
  #8
Senior Member
 
Kira
Join Date: Nov 2020
Location: Canada
Posts: 435
Rep Power: 8
aero_head is on a distinguished road
Quote:
Originally Posted by cfdnewb123 View Post
Thanks. May I asked whether is it a must to show all the data? Because I found some journals/conferences which only show part of the data, e.g. skin friction coefficient but no velocity profiles even though the paper they are referencing to contain data for velocity profiles too. I know it is a must to report everything but if the papers I am referencing from do not include some of the data, then should I follow their practice?
Hello,

Just as an aside; this would depend on when the paper was published, but have you considered reaching out to the author(s) and asking for the data? Some may give you the velocity profiles, others may not, but the only way to know for sure is to ask.

If anyone ever contacted me about something like that for my work, I would be inclined to assist them.
aero_head is offline   Reply With Quote

Old   January 17, 2021, 05:11
Default
  #9
Senior Member
 
Filippo Maria Denaro
Join Date: Jul 2010
Posts: 6,764
Rep Power: 71
FMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura aboutFMDenaro has a spectacular aura about
Quote:
Originally Posted by cfdnewb123 View Post
Thanks. May I asked whether is it a must to show all the data? Because I found some journals/conferences which only show part of the data, e.g. skin friction coefficient but no velocity profiles even though the paper they are referencing to contain data for velocity profiles too. I know it is a must to report everything but if the papers I am referencing from do not include some of the data, then should I follow their practice?



Submitting an article to a journal means you propose a new methodology or you apply a standard methodology to a new flow problem.
The former case requires you compare your results to the best result you found in literature.
aero_head likes this.
FMDenaro is offline   Reply With Quote

Reply

Tags
fluent, mesh, rans modelling, separate flow, turbulence

Thread Tools Search this Thread
Search this Thread:

Advanced Search
Display Modes

Posting Rules
You may not post new threads
You may not post replies
You may not post attachments
You may not edit your posts

BB code is On
Smilies are On
[IMG] code is On
HTML code is Off
Trackbacks are Off
Pingbacks are On
Refbacks are On


Similar Threads
Thread Thread Starter Forum Replies Last Post
[solidMechanics] Support thread for "Solid Mechanics Solvers added to OpenFOAM Extend" bigphil OpenFOAM CC Toolkits for Fluid-Structure Interaction 686 December 22, 2022 09:10
Maximum number of iterations exceeded chtmultiregionsimpleFoam Moncef OpenFOAM Running, Solving & CFD 28 July 13, 2020 14:26
pimpleFoam: turbulence->correct(); is not executed when using residualControl hfs OpenFOAM Running, Solving & CFD 3 October 29, 2013 08:35
SLTS+rhoPisoFoam: what is rDeltaT??? nileshjrane OpenFOAM Running, Solving & CFD 4 February 25, 2013 04:13
Problems in compiling paraview in Suse 10.3 platform chiven OpenFOAM Installation 3 December 1, 2009 07:21


All times are GMT -4. The time now is 07:33.