| swarup
|
April 3, 2005 01:17 |
Re: Unsteady Convergence
unsteady state convergence should be obtained per time step if you are solving a unsteady state problem for incompressible flow (Ma<0.3). if you want a steady state solution from unsteday state simulation, then do not bother about convergence per time step. the number of iterations per time step (default 20) will be most likely sufficient for convergence once transients have died down. you will always see jumps in residuals at the start of a new time step. they will reduce once convergence is approached. you should try in addition to monitor a primitive variable or another quantity that you hope to see at a strategic point/area/volume of your choice. it is always better to check conservation law confirmations for physically meaningful solutions. residuals if are increasing then you better check their definition. when source terms are increasing scaled un-normalized residuals (default for FLUENT) keep increasing. it may very well happen that they are misleading you while your solution is under control. in my opinion, you should monitor many quantities instead of residuals only. a check of residual definitions will reveal something meaningful. regards, Swarup
|
