Author |
Message |
Yinhui Liu (silver)
New member Username: silver
Post Number: 1 Registered: 07-2004
| Posted on Friday, July 09, 2004 - 03:26 pm: | |
I have a flow model (2D transient) which works well using 3.10 version. But when I try to run it using 4.0.7a, it blows up. I also try to run a 3D transient model, it does not converge too. How can I solver this problem? Thanks. |
Robert G. Nelson (rgnelson)
Moderator Username: rgnelson
Post Number: 181 Registered: 06-2003
| Posted on Friday, July 09, 2004 - 04:55 pm: | |
It's hard to diagnose a problem with as little information as you have provided, but I can make a guess. Version 3 used a RANGE declaration on variables to establish the scale of the variables for error estimation. In version 4 we have changed this to a THRESHOLD definition, which has the same purpose but a slightly different meaning. If you use the word RANGE, version 4 will trap and tell you that you need a THRESHOLD instead. But if you merely state the number without the qualifying name, which is syntactically legal, then version 4 will not warn you. In both version 3 and version 4, supplying too large a value for this scaling number will destroy the meaning of the error measures, and allow the time integration to proceed without timestep control. This can result in a bad solution, or even a crash. The solution is to specify a THRESHOLD compatible with its description in the version 4 documentation: the smallest variation in the variable for which you wish to maintain the specified ERRLIM accuracy. It is better to make it too small than too large. The THRESHOLD specification is not necessary if VALUE boundary conditions already indicate a meaningful scale for the variables. If this guess is not pertinent to your problem, send me the descriptor or post it here, and I will try to determine the cause of the errror. |
|