CFD Online Discussion Forums

CFD Online Discussion Forums (http://www.cfd-online.com/Forums/)
-   Main CFD Forum (http://www.cfd-online.com/Forums/main/)
-   -   general question (http://www.cfd-online.com/Forums/main/395-general-question.html)

jay December 16, 1998 19:24

general question
 
In the event of a simulation, what is one to make of the fact that the residuals oscillate wildly for the first few iterations, and then settle down and fall nicely. Aside from the explanations of why this happens, is a soln from such trend to be trusted as being...good/trustable or..."can't trust it, need to re-do until you obtain a decent trend from the beginning." Your thoughts people?

John C. Chien December 17, 1998 13:09

Re: general question
 
When you say " residuals oscillate wildly ...", I think you are talking about getting steady-state solutions using iterative methods. Let's reverse the process, say, you now have the steady-state solution ( regardless how you define it ), then the (n-1)th solution will be considered as a non-steady-state solution. This (n-1)th solution is not going to create any oscillation for you. So , the reason why you are getting oscillating residuals is mainly because the initial guess is a " wild guess ". The right way to do is to study the problem and come up with a good initial algebraic or analytical flow field. I have used this approach since early 70's, and the advantage is "fast convergency". This is an area where the reward is high. This requires knowledge and experience about the fluid dynamic part of the problem, that is why CFD is not equal to CFD code or just running CFD code. The difference can be several minuites instead of several days of computing time.


All times are GMT -4. The time now is 11:55.