I have a query in my mind which is related to the k-epsilon RNG model. When I start working on it, I faced an unusual phenomena about its instability as compared to k-epsilon standard and realizable. Here I quoted convergence problems as a unstable. Perhaps I tried to decode the program which used in fluent but I am still not sure about what the parameters which cause this are. I have two questions:
1. Is it sensitive model in all cases or in some specific cases?
2. Is there a unique way to use this model (unique way means there is particular solver and solution parameter combination or different way of initialization) to overcome this kind of problem and get the stable converged solution?
{While I was decoding the program, I went across fluent programing guide in which I found this suggestion: When you are using the RNG k- epsilon model, an approach that might help you achieve better stable convergence is to obtain a solution with the standard k- epsilon model before switching to the RNG model. Due to the additional non-linearities in the RNG model which cause for instability. I tried this but still I face the same problem}
Could anyone who already worked with this model or experienced this problem explain this?
Because of some more nonlinearity than the k-e model you may find some instability for a few cases,which is easily cured by a little more relaxation for k and e equations.The advise of using first standard k-e and then use RNG is very good and always works.I am rather certain you have done some other mistake and it is not the RNG instability.At the end it always works,either with some more relaxation or continuing from the k-e solution
I got the solution yes its depend on relaxation factors and you were right that there were other mistake which i figured it out after decoding the fluent code thts in the both case when starting iteration the solver has to be put in to sequence mode otherwise that unstable phenomena comes.
thank you for your guidance.
Hi Shrish, great that you could find the tech errors. But there is a general problem with the physics of all these models. We have meanwhile about 50 of such semi-empirical or phenomenological closures and this was the background that I started to think about the turbulence problem 25 years ago. Notice that e.g. k is a conserved quantity but eps is something else. The attached paper might help if you are interested to understand turbulence and why the semi-empirical closures cannot do the job. E.g. they dont predict the universal constants of Karman and Kolmogorov, to name only the most prominent cases. Helmut PS: If you have questions you're welcome.
k -ε does predict κ of Karman and decay -1.18.
Please send me e-mail to send you proof
My coordinates are [email protected].
But I am afraid that all papers only calculate Karman and decay etc. only from other (emoirical) constants. But of course I am curious, and thanks in advance!
Actually this is one way to determine the empirical constants,so I agree with you about the constants.I will send you mail tomorrow.
Shrish and other young guys are so great in asking fresh questions, I love that. All our science systems need renovation from the bottom up...
I only may wish you and Shrish happy Easter days and thank you both!
Wish you happy Easter days both of you sir !
i am so grateful for your valuable suggestion which fix it up my doubts.
regards
shrish shukla
- Badges
- Science topic