| Author |
Message |
    
hnili
Member
Username: hnili
Post Number: 7
Registered: 02-2008
| | Posted on Wednesday, May 26, 2010 - 08:28 am: | |
Hi.
I am running a staged problem (axisymmetrical - 2D), with an errlim of 2e-6. After 8 grids, and in stage 1, I was given the 'Exceeded Regrid Limit' error message. What surprises me is that the error message says: Max err is 1.955e-6, but that's just below my errlim of 2e-6, so why should I be given the message at all?
Advance thanks |
rgnelson
Moderator
Username: rgnelson
Post Number: 1368
Registered: 06-2003
| | Posted on Wednesday, May 26, 2010 - 12:57 pm: | |
Are you using a FRONT or RESOLVE statement?
These are not included in the Max Err report.
|
hnili
Member
Username: hnili
Post Number: 8
Registered: 02-2008
| | Posted on Thursday, May 27, 2010 - 11:55 am: | |
I am using a RESOLVE statement.
Is that why I get this message whereas normally I shouldn't? Are you suggesting I won't get a similar message if I do not use RESOLVE?
Thanks. |
rgnelson
Moderator
Username: rgnelson
Post Number: 1369
Registered: 06-2003
| | Posted on Thursday, May 27, 2010 - 02:39 pm: | |
The error estimate for RESOLVE is separate from the PDE residual reported on the ERR lines in the status window. In version 6, you can see the RESOLVE error estimate on the "log error" plot in the status window.
If the RESOLVE condition is harder to satisfy than the PDE residual, this may be the controlling factor in the regrid process. This does not mean you shouldn't use it if it is meaningful to your model.
You can increase the number of regrids before a warning is issued using the GRIDLIMIT selector.
|
hnili
Member
Username: hnili
Post Number: 9
Registered: 02-2008
| | Posted on Sunday, May 30, 2010 - 11:49 pm: | |
That was valuable information.
Thanks very much. |
rgnelson
Moderator
Username: rgnelson
Post Number: 1371
Registered: 06-2003
| | Posted on Monday, May 31, 2010 - 07:43 pm: | |
PS
ERRLIM=2e-6 is very ambitious requirement, and will probably create a very large mesh and run a long time for no real gain.
You should try your problem with the default value (0.002) and gradually decrease it until the result stops changing. |