Note: This discussion is about an older version of the COMSOL Multiphysics® software. The information provided may be out of date.

Discussion Closed This discussion was created more than 6 months ago and has been closed. To start a new discussion with a link back to this one, click here.

Large Eddy Simulation

Rouaissi Ridha
hello
I want to use Large Eddy Simulation on comsol but I can't found it.
Can you help how to do this.

thanks a lot

11 Replies Last Post Nov 15, 2014, 9:22 PM EST
Posted: 4 years ago Apr 25, 2013, 8:13 AM EDT
Hi

I'm not sure there are an "Large" Eddy current, but try out the MEF physics and the related model library cases

--
Good luck
Ivar
Hi I'm not sure there are an "Large" Eddy current, but try out the MEF physics and the related model library cases -- Good luck Ivar

Posted: 4 years ago Apr 25, 2013, 9:55 AM EDT
Hi,

I think the question is about fluid dynamics. Large eddy simulation is somkething
like "direct numerical simulation" where the plain Navier-Stokes equations are solved
in turbulent regime and the eddies are resolved in space and time. In contrast
to ke-model.

Usually done in dedicated solvers, so I dont know if this can be done in Comsol.

Regards

Jens
Hi, I think the question is about fluid dynamics. Large eddy simulation is somkething like "direct numerical simulation" where the plain Navier-Stokes equations are solved in turbulent regime and the eddies are resolved in space and time. In contrast to ke-model. Usually done in dedicated solvers, so I dont know if this can be done in Comsol. Regards Jens

Posted: 4 years ago Apr 26, 2013, 1:52 AM EDT
Hi

sorry I misunderstood as there are different Eddy currents around ;)
For that one I'm not competent enough, have you tried to ask COMSOL directly ?

--
Good luck
Ivar
Hi sorry I misunderstood as there are different Eddy currents around ;) For that one I'm not competent enough, have you tried to ask COMSOL directly ? -- Good luck Ivar

Posted: 4 years ago May 15, 2013, 11:22 PM EDT
COMSOL does not have an LES study type at this time. To approximate this type of simulation in COMSOL, one can use a combination of NS equations on a fine mesh, a transient solver with small time steps (small CFL number), superimposed small random inlet variation (and perhaps asymmetry in the mesh) and time-averaging of results.

--
Steven Conrad, MD PhD
LSU Health
COMSOL does not have an LES study type at this time. To approximate this type of simulation in COMSOL, one can use a combination of NS equations on a fine mesh, a transient solver with small time steps (small CFL number), superimposed small random inlet variation (and perhaps asymmetry in the mesh) and time-averaging of results. -- Steven Conrad, MD PhD LSU Health

Posted: 4 years ago May 28, 2013, 11:44 AM EDT
Dear Steven,

Do you have any example model to demonstrate this ? If yes, can you share it with us ? Thanks a lot. I am trying to run some simulation which requires LES turbulence model, but COMSOL does not have it.

Best,

Xinhui,.
Dear Steven, Do you have any example model to demonstrate this ? If yes, can you share it with us ? Thanks a lot. I am trying to run some simulation which requires LES turbulence model, but COMSOL does not have it. Best, Xinhui,.

Posted: 3 years ago Sep 17, 2014, 11:23 PM EDT
Hi
Dear Steve Conrad, you say :"COMSOL does not have an LES study type at this time". I want to know what's your meaning about "LES study type"? you know comsol using the RANS . So, can we create a LES model using it‘s equations interface or through physical Model Builder?

--
Regards

Meng Jie
Hi Dear Steve Conrad, you say :"COMSOL does not have an LES study type at this time". I want to know what's your meaning about "LES study type"? you know comsol using the RANS . So, can we create a LES model using it‘s equations interface or through physical Model Builder? -- Regards Meng Jie

Posted: 3 years ago Sep 21, 2014, 8:57 PM EDT
Meng,

COMSOL can perform standard Navier-Stokes (NS) as well as several RANS-based turbulence models (i.e. you can select a study type for each of these). The RANS-based study types are quite different from large eddy simulation (LES), which solves for large eddies instead of averaging them, and accounts for small eddies implicitly. However COMSOL does not (yet) have a separate study type for LES. Hopefully in a future release they will.

Direct numerical simulation is another approach to turbulence by directly solving the standard NS equations on a spatial scale that can resolve even the small eddies, but is usually too computationally intensive due to the need for such a superfine mesh.

I hope this answers your question.

--
Steven Conrad, MD PhD
LSU Health
Meng, COMSOL can perform standard Navier-Stokes (NS) as well as several RANS-based turbulence models (i.e. you can select a study type for each of these). The RANS-based study types are quite different from large eddy simulation (LES), which solves for large eddies instead of averaging them, and accounts for small eddies implicitly. However COMSOL does not (yet) have a separate study type for LES. Hopefully in a future release they will. Direct numerical simulation is another approach to turbulence by directly solving the standard NS equations on a spatial scale that can resolve even the small eddies, but is usually too computationally intensive due to the need for such a superfine mesh. I hope this answers your question. -- Steven Conrad, MD PhD LSU Health

Posted: 3 years ago Nov 4, 2014, 4:34 AM EST
Hi Steve,

Thanks for your answer. Recently comsol released version 5.0. But there is still no progress. Looking forward to LES's early accession.

--
Regards

Meng Jie
Hi Steve, Thanks for your answer. Recently comsol released version 5.0. But there is still no progress. Looking forward to LES's early accession. -- Regards Meng Jie

Posted: 3 years ago Nov 4, 2014, 7:11 PM EST
Yes, I was at the COMSOL conference in Boston for its release. I did speak with COMSOL representatives about LES in a future release, and it is being considered. If others are reading this post and are interested in LES, please submit your support for LES to COMSOL. LES is an important tool for flows in biomedicine (and many other fields), which aren't suitable for simulation with NS or RANS models.

--
Steven Conrad, MD PhD
LSU Health
Yes, I was at the COMSOL conference in Boston for its release. I did speak with COMSOL representatives about LES in a future release, and it is being considered. If others are reading this post and are interested in LES, please submit your support for LES to COMSOL. LES is an important tool for flows in biomedicine (and many other fields), which aren't suitable for simulation with NS or RANS models. -- Steven Conrad, MD PhD LSU Health

Stanimir Kolev
Posted: 3 years ago Nov 14, 2014, 5:20 AM EST
Hi,
In this paper:
www.comsol.com/paper/possibility-of-implicit-les-for-two-dimensional-incompressible-lid-driven-cavity-14684

the authors claim that they do some kind of Large eddy simulation in Comsol based on the Navier-Stokes equations and the stabilization terms added in the equations. I am not sure how reliable is this (I am quite skeptical) but it might be useful.

Regards,
Stan
Hi, In this paper: http://www.comsol.com/paper/possibility-of-implicit-les-for-two-dimensional-incompressible-lid-driven-cavity-14684 the authors claim that they do some kind of Large eddy simulation in Comsol based on the Navier-Stokes equations and the stabilization terms added in the equations. I am not sure how reliable is this (I am quite skeptical) but it might be useful. Regards, Stan

Posted: 3 years ago Nov 15, 2014, 9:22 PM EST
I am still somewhat a novice at LES, but here is my understanding. LES is usually considered to be implicitly a 3D approach, so I am not familiar with its use in 2D as demonstrated in the paper.. These authors do bring up something that I failed to mention in my first reply, in that the very small eddies that are not fully resolved need to be numerically filtered, since COMSOL does not have a formal method of modeling sub-grid scales. This is done in COMSOL with the conservative diffusion methods as described in the paper.

The element size needs to be small enough to resolve the eddies of interest. The Kolmogorov microscales are the smallest scales in turbulent flow, and modeling these scales is the goal of Direct Numerical Simulation. These scales are on the order of 10^5 to 10^7 times smaller than the largest eddies. It takes on the order of Re^(9/4) elements to model these in 3D, and hence is too computationally expensive for anything complex. In LES, larger scales are used, hence the need for filtering out the smaller scales. These intermediate Taylor microscales require on the order of Re^(3/2) elements for 3D. This is far more manageable, which is what makes LES a practical alternative to DNS.

The time interval for the time-dependent simulation (delt) should be manually controlled such that delt < h/u, where h is the element size and u is the velocity in the region of that element.

When steady-state results are desired, i.e. when a time-dependent simulation is run long enough to develop a steady-state, it was indicated to me (by someone with much more knowledge about LES) that there should be some small variation on the inflow boundary, and then average the results over time. I have created this by adding a small statistical variation (with average variation of 0) to the inlet flow profile.

The above is to the best of my understanding of LES, and I invite any corrections to my comments.

--
Steven Conrad, MD PhD
LSU Health
I am still somewhat a novice at LES, but here is my understanding. LES is usually considered to be implicitly a 3D approach, so I am not familiar with its use in 2D as demonstrated in the paper.. These authors do bring up something that I failed to mention in my first reply, in that the very small eddies that are not fully resolved need to be numerically filtered, since COMSOL does not have a formal method of modeling sub-grid scales. This is done in COMSOL with the conservative diffusion methods as described in the paper. The element size needs to be small enough to resolve the eddies of interest. The Kolmogorov microscales are the smallest scales in turbulent flow, and modeling these scales is the goal of Direct Numerical Simulation. These scales are on the order of 10^5 to 10^7 times smaller than the largest eddies. It takes on the order of Re^(9/4) elements to model these in 3D, and hence is too computationally expensive for anything complex. In LES, larger scales are used, hence the need for filtering out the smaller scales. These intermediate Taylor microscales require on the order of Re^(3/2) elements for 3D. This is far more manageable, which is what makes LES a practical alternative to DNS. The time interval for the time-dependent simulation (delt) should be manually controlled such that delt < h/u, where h is the element size and u is the velocity in the region of that element. When steady-state results are desired, i.e. when a time-dependent simulation is run long enough to develop a steady-state, it was indicated to me (by someone with much more knowledge about LES) that there should be some small variation on the inflow boundary, and then average the results over time. I have created this by adding a small statistical variation (with average variation of 0) to the inlet flow profile. The above is to the best of my understanding of LES, and I invite any corrections to my comments. -- Steven Conrad, MD PhD LSU Health

Note that while COMSOL employees may participate in the discussion forum, COMSOL® software users who are on-subscription should submit their questions via the Support Center for a more comprehensive response from the Technical Support team.