k-epsilon source term
 

Hi everybody I am working with k-epsilon model (density-variable version) to compute a simple jet in still air. My problem is that the convergence rate of the iterative solver of the linear system (Coniugate Gradient Method Stabilized) is very poor. I have already tried to improve it using a lower under-relaxation factor and/or CFL number but I have no improvements. I suppose the kind of the source term linearization for k and epsilon is not the best one. In particular the dilatation term is the one which destroy convergence rate. Have you some references where I can find some source term linearizations and/or critical discussion about them? Thanks MAX

Re: k-epsilon source term
 

USe tri diagnol matrix analysis (TDMA) to get your equations converge.it really workes. you can get the detail from the text book " Intoduction to finite difference methods" by Veerstag and Malashekara.

Re: k-epsilon source term
 

I am sorry, but I can not find this book. Can you check exactly the title and authors? I have another question: does this method work also for unstructure grids? I guess you can apply it only for structured problems. Nevertheless I will have look to your book, once I find it. Thanks

Re: k-epsilon source term
 

i have been running k-e model on unstrcutred grids based on bicgstab and yes i also feel that convergence is very slow. for the same mesh fluent converges very fast, even though i make coarsening level to 0, (i am not sure whether fluent is using multigrids in that option too), but if i assume fluent uses gauss siedel only (when i make coarsening to zero), it is quite surprising that fluent solution conerges so fast. i will only come to know if i implement multigrids into my solver and this will take some time.

i am not sure what other people have experience with k-e using other solvers , for example FOAM (i guess) can work with biCGstab.