Hi, I'm implementing non-linear eddy viscosity models (for instance the model proposed by Larsson in Two-equation turbulence models for turbine blade heat transfer simulations) and using them in impinging plane jets. However, in the stagnation region (using a grid with 180X120 CV's), these models present oscillations and non-smoothing results in velocity profiles. I want to know if this trend has been observed by anybody else? or can it be an error in my implementation?

That model is usually not much more unstable than a normal linear k-eps model.

Andreas Abdon used the nonlinear model for impingement simulations with some success. You can find references to his papers on the summary of his PhD thesis here:

http://www.lub.lu.se/cgi-bin/show_diss.pl/tec_445.html

Hi,

Can you give the refernce of the paper. I would like to read it.

Thanks

Apurva

Hi, The paper is: Two-Equation Turbulence Models for Turbine Blade Heat Transfer Simulations, J. Larsson, ISABE Paper 97-7163, Proc. 13th ISABE Conf. vol. 2 pp. 1214-1222, Chattanooga, TN, 1997.

You can also download a postscript version of this paper and a few others directly from my homepage here:

http://www.cfd-online.com/Users/jola/

Thanks for your answer Jonas. I've read the papers By Abdon. I have another question: you mention that the NLEVM (Larsson's model) is as stable as a linear k-eps model. I'm using finite volumes, over a staggered grid, with time implicit integration, SIMPLEC method and multigrid. I introduce the nonlinear terms explicitly. However the convergence is seriously affected. Do you use any restriction or relaxation method to improve stability? because Abdon doesn't mention nothing. Thanks.

I have only used this model in a coupled explicit code, and there it didn't lead to any additional stability problems. Running a k-eps model in genereal in this type of code is often a bit tricky though and we experimented with various forms of point-implicit treatment of the turbulence model to improve stability. We ended up treating it fully explicit though, with a set of well-tuned limiters. We saw no big difference between a standard linear k-eps and the no-linear model in terms of stabitlity though.

I will email you Andreas Abdon's current email address so you can contact him about how he implemented it in his code. I don't remember if he used a SIMPLE code, but it is likely.

Thanks for the email address. Could you tell me, how do you treat the possibility of negative values in the square root of your model in the C2 constant?. Because I'm taking an absolute value and I think it can be one problem.