Methods for quality check of LES

Project Support
1

Hi everybody,

I am doing a LES for a small landscape architecture project. The aim is to show, how two different shaped walls provide protection from wind. Here is a link: https://www.simscale.com/workbench/?pid=1929720385875739693&mi=spec%3A7cb9aa7a-6165-4b57-9646-fc992f7caa7e%2Cservice%3AMESHING%2Cstrategy%3A7&ps=geometryPrimitives%2F0&sh=6

I also shared the link with support, I am taking about run2.

My question is: How can I measure the quality of my simulation result?
I checked the convergence plots and they look stable. Are there any other ways?

Thank you
Sophie

1 Like
2

Hi Sophie, thank you for using the forum! In order to confirm the reliability you can:

  1. Check the Residuals
  2. Check the Convergence Plots for Boundary Conditions
  3. Check the Result Control Items

Check this page for more details: How to Check Convergence of a CFD Simulation? | SimScale

Best regards,
Fillia

Best regards,
Fillia

3

Hi Filia!

Thank you for your reply! I only have convergence plots for residuals, not for boundary conditions. Where would I find them?

Thank you again!

4

I would like to enhace Filia’s answer.

While the residual plot is related to the convergence of the numerical study, and represents an indicator of the quality of how well the numerical solution is being resolved it does not necessarily indicates the quality of the solution (how close is your result to the real-life scenario), specially for LES based approaches. Remember that LES solves a portion of the turbulent spectra, and models the rest through a SGS (sub-grid scale model). Therefore, many quality approaches evaluate how much turbulence you are resolving (Turbulent kinetic energy ratio, or resolved length scale vs von Karman length scale). I found this paper for architecture purposes which might be useful for you: http://www.urbanphysics.net/2013_CAF_PG_BB_GJFvH_QLES__Preprint.pdf

Generally, it is accepted (although discussed in some articles) that if you resolve 80% of the total turbulent kinetic energy, the grid resolution is adequate and therefore your solution quality complies with LES standards (and in this way it is possible to assess your results should relate well with the real-life scenario). The question would be how can you evaluate that in SimScale, and maybe Filia can give us some insight, as I am not sure you can position probes in some sections on the domain to sample velocity data (and thus to obtain resolved turbulent kinetic energy statistics).

3 Likes
5

Really interesting! Placing probes under the Result control prior to the run is possible!

1 Like
6

Tsite’s answer would allow you to obtain both the fluctuating velocity values and some sort of “k-equation” related field values. From the fluctuating velocity values, it is easy to calculate the “resolved k”. However, the modelled “k-portion” of the turbulence may be a bit more difficult to interpret. First, you need to determine which specific LES model are you using and how that specific model relates the SGS-like variable to the sub-grid scale turbulent kinetic energy. This video is very helpful, especially in chapter 4: [CFD] Large Eddy Simulation (LES) 2: Turbulent Kinetic Energy - YouTube. I must say I am not expert in LES as I am focused in DES (hybrid LES-RANS), and there, the modelled portion of the turbulence is obtained directed from the k-equation.

1 Like