No promises

Do I need to change that too for U and other fields?

Idk you could strategize. I never selectively did it so I don't really know. It might even be redundant to set such low residuals. This is just something I found that keeps my transient residuals well behaved.

Thank you 🙂

If it still keeps exploding, it could be the model. You could try a different variant (just for diagnosis) unless there is a specific reason for choosing RNG.

I will run the case

Yes, if it keep blowing up I will switch to Realizeable

@User: The simulation run and completed successfuly!

Again, Thank you very much!

Glad I could help. Just for the sake of closure, we used a parallel compliant preconditioner and made tighter tolerance on the algebraic solvers. Am I correct to say these fixed your issues with 2nd order interpolation of turbulence? Also, what does your comparison with the other variations suggest, are they comparable? Or is there a definite winner?

Yes, you're right, I have also used tighter tolerance for the other equations (U, k, epsilon)

I have run the case using the RNG variant only because I have already used other turbulence models (Spalart-Allmaras, SST k-Omega, etc.)

To sum up, here are the settings I used in

fvSolution

file (before and after):

from t = 0 to t = 2.4s the simulation was run only with 1st order schemes of k and espilon

I could run the case with other variants of k-epsilon but unfortunately due to some technical limitation it's not possible right now

I have unlimited storage but the quota limits the maximum number of files to 1M which could be easily reached after 1-2 hours when running open with 2nd order schemes

It's alright. BTW, you could use purgeWrite if you don't want to save all intermediate files..

Yes, I have used from t=0 to t=2.4

But for t>2.4 up to t=2.55 I need to save the results every 1 degree of rotation (corresponding to saving every 1e-4 seconds)

I know about other solution such as using a much larger writeInterval in the

controlDict

file and using writeObject function object to save only some specific fields

Another approach is to hook a bash script to the solver using systemCall function object to remove the intermediate files. But this could slow down the simulation.

This looks like a good approach

Another question, I forgot to ask about the k-epsilon model implementation in OpenFOAM

I know that the model itself is a High-Re model

but the

epsilonWallFunction

has low-Re corrections

Does that mean the model can be used with resolved viscous sublayer?

I have never used it but I think that's the idea: to integrate all the way upto the wall, with the use of low Re corrections and fine mesh. I am not sure whether "resolve" is the right word to use though.

Hi I am facing an issue when comparing two different models the RNG and BSLRSM predicting vortices performances. There are two vortices involved in a high swirling flow and somehow the RNG model is showing a better trend than the BSLRSM model. ( My investigation is aiming at the RANS applicability thus eliminating the LES and DNS option) . Would anyone mind to give a second thought?

Do you mean RNG kEpsilon?