#3500 closed defect (fixed)
Incorrect numerical output not detected
| Reported by: | Owned by: | Willi Braun | |
|---|---|---|---|
| Priority: | critical | Milestone: | 1.9.4 |
| Component: | Run-time | Version: | |
| Keywords: | solver error, state selection | Cc: |
Description
When the stop time of the attached simulation is increased from 2 to 100, a clearly incorrect result is produced but no error indication is given. When the number of output intervals is increased from 500 to 5000 the correct result is given.
It seems that the solver (dassl) uses the stop time and number of intervals for a first guess at step size. With num. of intervals at 500 the step size is much too large, but its error estimator does not properly catch the error.
Attachments (1)
Change History (7)
by , 10 years ago
| Attachment: | InconsistentNumericalResults.zip added |
|---|
comment:1 by , 10 years ago
| Component: | Unknown → Run-time |
|---|---|
| Owner: | changed from to |
| Status: | new → assigned |
comment:2 by , 10 years ago
| Keywords: | state selection added |
|---|---|
| Priority: | high → critical |
| Status: | assigned → accepted |
Currently, we check the state selection to switch states just at output points and that's the reason the simulation throws wrong results here.
comment:3 by , 10 years ago
| Resolution: | → fixed |
|---|---|
| Status: | accepted → closed |
This is fixed by 212791 where the way how starting values are calculated for the non-linear system have been changed.
comment:4 by , 10 years ago
| Milestone: | Future → 1.9.4 |
|---|
Note:
See TracTickets
for help on using tickets.
Modelica source file and plot of results