HI. I am currently working on implementation of explicit algebraic stress model(ex, Gatski & Speziale, 1993) in my code( a 2D-FVM code, AUSM with Muscl). I can calculate the 2-D compressible mixing layer with standard k-epsilon model with linear stress-strain relation. But, when I using the ASM, the wiggles of Reynols stress in the center region of the mixing layer are shown. and they grow for further iteration. So, I cann't obtain the converged solution.I think that this is due to the numerical error in the higher-order term in Reynolds stress. If there is anyone out there who actually done this kind of coding work, I will be most grateful if he/she shares some of his/her experiences with me.

I think you will need to limit the time scale term appearing in the non-linear part of the stresses. This has a tendency to go nuts and take the solution with it.

Srini

You must probably have already checked for any typo or other stuff in your implementation, correct?

Since this type of model is completely nonlinear, one interesting thing to do to guarantee robustness is to apply realizability criterions for the Reynolds-stress terms.

Get this document form internet to help you on the realizability criterions you should apply to your code: http://cfd.me.umist.ac.uk/mcjihsg2/T...R/appendAB.pdf This would be the same approach as clipping the k equation so as k do not getnegative, and so on.

This is usually recommended because, in the middle of the simulation, when you do not have a physically meaningful solution, this could lead the turbulence terms to a nonphysical state as well, and all that will blow on your face.

Hope that helps you.

PS.: I'm currently trying to implement a RSM from people from UMIST. Huge amount of work... =)