Turbulence modeling Puns

As per the discussion topic vote, May's monthly topic is Turbulence modeling.

And, what are the major challenges and successes in the turbulence field now?

Trying to get a benchmark for the drag coefficient for a few simple shapes and comparing them to experimental results. I am running at a Reynolds number of 1000 and 10000. Ran a few cases for cylinder and square using k-epsilon and SST models and different meshes. My averaged drag coefficients seem to be under-predicted.

I saw that drag coefficient being under-predicted seemed to be the case in a few papers that used CFX. Is this a limitation in the RANS models/CFX software? I saw that I can modify terms in the CFX setup menu for the turbulence models. Would modifying those values offer better prediction for the drag?

I'm trying to calculate the Reynolds stress tensor for a 2D incompressible turbulent flow over a flat plate.

When modeling the Reynolds stress tensor with the eddy viscosity nu_t, how do you evaluate the turbulent kinetic energy term? I want to use a zero-equation model (Prandtl mixing length with Cebeci-Smith modifier) which should not involve determination of the turbulent kinetic energy.

Hey there,

this might be a bit of a noob question concerning eddy viscosity RANS models.

A few months ago I heard someone talk about (unfortunately I can only recall from memory) talk about the theoretical differences between two- and one-equation turbulence models. While I am quite familiar with practical differences, e.g., between k-e and k-w SST or/and Spalart-Allmaras, and do understand that a two-equation model is typically better suited for (e.g.) adverse pressure gradients - I do not understand why.

That person said (freely quoted), that one equation models are inherently worse in modelling (the effect) of turbulence due to only one additional transport equation, and that at least two transport equations are required to actually model turbulence. I.e. one equation models are a bit of a botched job, which work in reality but are inherently worse suited.

Now, I'm trying to understand (if it's even correctly remembered by me) why two-transport equations are at least necessary and how that can be shown.

Thanks for your help!

Hello

I want to learn to use the LES turbulent model and compare the results with the RANS model. Is there any reference on which CFD software is suitable for using the LES model, commercial or open-source?

And is there any reference on the minimum and best hardware requirements for performing LES simulations?

Thank you

There are many turbulence models, does anyone one has summarized turbulence model cheat sheet.

I am new to CFD and I have no clue about turbulence models. But I want to implement one in my tool cooling simulation. (Image attached)

I have followed the examples from the LS-Dyna site but I get an error (and I do not know what to correct either due to lack of basics)

What model would be the best option for understanding and implementation with respect to LS-Dyna?

P.S.: If there are good resources regarding this topic, please let me know.

https://preview.redd.it/s1fi1qpft2e71.png?width=547&format=png&auto=webp&s=0a12e0b91bd20f30417617161a333a657726d56a

Hello,

I'm studying extreme waves using OpenFOAM. I have assumed a laminar flow condition, following the other's work on similar test cases. In my models I have identified a very common problem of the MULES VOF scheme - spurious air velocities which lead to small oscillations of the interface. I managed to substantially reduce this issue by using the IsoAdvector scheme, but it's still there.

For the last part of my investigations, I've decided to see if turbulence modeling could improve my numerical predictions. I have a very good agreement with experimental data overall, but there were few issues I was hoping to resolve. I didn't get much of an improvement but I was really surprised to see that all these high-velocity air pockets are gone! The interface also looks much better, no oscillations, perfectly smooth.

So here's my question, why did turbulence modeling remove these flow features? How can this be explained?

BTW I'm using a k-omega turbulence model (Larsen & Fuhrman 2018) designed for this particular application (modelling waves).

I am currently working on a 2D RANS simulation with the k-epsilon turbulence model for wind turbine wake simulation. I implemented the transport equations for k and epsilon but stability for k and epsilon seems to be a problem. Are there any good resources for easy implementation on an evenly spaced grid using finite difference methods?

https://preview.redd.it/840x0k2s6f871.jpg?width=708&format=pjpg&auto=webp&s=be1ea0d478a360a1b16a5439cec5d9b41fc2840e

Hi, I'm wondering if anyone can provide an ELI5 style explanation of the difference between these two approaches, and the way they work, and whether one is necessarily more accurate than the other. If anyone could answer or point me in the right direction for some content that explains these things using simple language, I would greatly appreciate it.

Thanks in advance!

Dear foamers,

I want to modify a turbulent dispersion model and calculate the particle relaxation time. Therefore, I have to include the particle Cd calculated from the drag model.

Could anyone give me a hint on where should I start? I have no experience reading other classes objets in OpenFOAM models and I have such limited knowledge on C++ object oriented language that I don't even know how to google it right, so any materials would also be very helpful. Thank you very much!

I am working on a project in which I am simulating a 2D airfoil in turbulent flow using the SA model. I know that the RANS models tend to breakdown as the airfoil enters the stall regime, and I know that it is because the assumption of homogenous isotropic flow starts to become invalid, but I am not quite sure what that means and how it relates to the turbulence models' inability to resolve smaller length scales.

Also, just to make sure I am not mistaken, RANS turbulence models like SA do not need a highly refined grid because they make assumptions (like homogeneity and isotropy - which, again, I don't think I fully understand) that circumvent the need to resolve anything that's happening on very small length scales, like the transfer of energy from the smallest eddies up to the largest largest eddies down to the smallest. Is that correct? Is that also why it is difficult to capture vortex shedding using these models? Because the larger vortices are a result of what's happening on the smallest scales? I feel like I'm really close to connecting all the dots here but I'm missing the final link...