Just wondering if the future development would include more turbulence models (like large eddy simulation, detached eddy simulation, etc...) other than the ones we see (k-e, RNG, eddy viscoscity, mixing length)?
hey max.. we have done a fair amount of R&D on which turb models to offer. What sort of simulations would you use LES/DES models for? As you know, they are fairly sophisticated models that are pretty computationally expensive..
Derrek, with LES/DES models, we would like to accurately (as much as possbile) model unsteady vortex shedding of vertical axis wind turbine blades in 2D. 3D for horizontal axis wind turbine would be nice too, if possible.
How good is the eddy viscosity model we have? Comparable to LES/DES models in some ways? I agree with you that these are sophisticated and computationally expensive models.
You could attain vortex shedding with RSM, (not that it's implemented) - this model would presumably be less work for the developers than LES DES / hybrid. I'd also like to see RSM implemented.
You are correct that implementing the RSM would probably be more feasible since it is an RANS based model. I am not a developer, but that is my understanding.
The LES/DES would be a very new direction. These turbulance models also require much more mesh and longer runs times, not something that works well for most designers and our tradiational message of trying to be upfront.
What would be a good compromise turbulance model that would provide reasonable error and solve time when you are looking for vortex shedding and accurate lift/drag numbers at low and high reynolds numbers?
As vortex shedding is a boundary layer / transition phenomena, it'd require turbulence models valid all the way to the wall: SST k-omega, or a low-RE RSM model for example. This'd also require work on the mesher, with meshes which can resolve the boundary layers (a low y+ setting, instead of the standard wall treatment).
Although I agree that this increase in mesh requirement goes against the upfront philosophy, I feel the whole user experience, CAD integration and general speed of pre/post processing would offset the extra computational time in terms of whether it is perceived 'up front' enough, as it were. It would expand the software's capabilities massively.
Same goes for LES / DES / Hybrid RANS LES; but I'm not the one who's got to code it!
I m an amateur in the world of CFD. i want to study the flow over cylinder (mainly vorticity nd vortex shedding). I tried its analysis on Autodesk Simulation CFD 2013, the computational results were right, but vortex were not visiible when i used trace points to visualize the flow. Also what all assumptions should i make so that vortex shedding is visble & also if you can recommend me some text regarding the topic, it would be great.
I was wondering what proportion of current SimCFD users will really leverage LES/DES on their designs. RSM seems about the right high-end choice.
I was wondering if it would be possible to include the Realizable k-e model. This one has been (traditionally) more accurate than RNG, because of the realizability it induces in k and eps equations. This would reduce the diffusion in raw k-e model, and at the same time can deal with high mean strain rates as well, where k-e may have problems. Can be a good alternative to RNG for non-swirl cases.
Additionally, it would be nice to have a non-linear k-eps model as well. This would be instrumental in simulating cross-flows by abolising the isotropic turbulence in two equation models, yet being lean on computational resources as compared to RSM. Though, validation still remains the issue for such models.