Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 571 Registered: 06-2003
| Posted on Thursday, March 30, 2006 - 04:36 pm: | |
FlexPDE is an adaptive system. The timestep and mesh sizes are modified to try to maintain the accuracy requested (or defaulted) by the user. In order to avoid infinite cutting of timesteps in case of trouble, there is a cutoff that stops the run if the timestep drops below 1e-9 times the initial timestep. Try specifying an initial timestep small enough to resolve initial transients. A common cause of trouble in time-dependent problems is the use of discontinuous initial conditions. A discontinuity is not representable at any scale by polynomial patches in time or space. In theory, therefore, it requires infinite subdivision in both time and space. FlexPDE includes an initial smoothing pass that "rounds the corners" of such initial discontinuities, but it is not always effective. Try applying initial conditions (including boundary conditions at time=0) that are self-consistent, and then ramp boundary values over a "reasonable" time. Another way to address this is to convert value boundary conditions to fluxes, which are "softer" numerically. Another cause of trouble, of course, is if the equation system is simply wrong. Inconsistent equations that have no solution will cause the timestep to grind to zero. Check your equations and parameter dimensions. |