segregated solver
 

Hello, just wondering does any one know if it is reasonable to take the default values of 10^-3 for convergennce criteria, using the segregated solver for a RNG k-e model, or should I perhaps seek a greater reducvtion of some or all of the residuals. thanks, nick

Re: segregated solver
 

(1). In the last several years, when I started the program, I always tried to set the numbers to 10E-08, regardless of the program I used. (2). You should try to run the code such that you can see clearly the slope-and-the flat shape, with the numbers (residuals) drop below 10E-06. (3). If you are not getting such behavior, then there are a lot of improvement can be made. (4). Try a simple problem first, and see whether you can see such convergence history. It will change your life, if you have not done so.

Re: segregated solver
 

thanks for that John. Are you talking about scaled or normaised residuals. I forgot to to mention that in my last posting.Also have you or anyone else any suggestions on the type of schemes(QUICK, SIMPLE etc.,) that are best for using in an incompresssible flow model with the RNG k-e solver in order to sped up convergence. the model is purely 2D. much appreciated nick

Re: segregated solver
 

(1). Don't worry about the definition. Just set it(all of them) to a very small numbers at the begining of the calculation. The purpose is to avoid terminating the iteration processes. Once you have the ski slope shape visible from your iteration history, you can then set the proper number of iterations for your next calculation. So, it is important to remember that. (2). If you are not getting that particular shape in your convergence history, then you have problems. (3). Most of the time, I use "upwind method" to start the iteration. So, you can select upwind method and make sure that you have the converged solution. If you can't get converged solution with upwind method, you have a big problem. (4). With these two approaches, you can get the feeling about the iteration processes and the number of iterations required to get the converged solution. (5). When you change the method, you will run into additional problems, but at least you have a good starting point.