Opened 8 years ago
Last modified 5 years ago
#4143 assigned defect
Initialization of systems with dynamic state selection is not handled properly
| Reported by: | Francesco Casella | Owned by: | Karim Adbdelhak |
|---|---|---|---|
| Priority: | blocker | Milestone: | 2.0.0 |
| Component: | Backend | Version: | |
| Keywords: | Cc: |
Description
Lennart Ochel wrote me:
There is no approach in OpenModelica to select properly the initial states for the state sets yet [for the initialization of dynamic state selection]. Right now we choose the states randomly that are used for the initialization phase.
This should be fixed for the 2.0.0 release.
Attachments (1)
Change History (7)
comment:1 by , 8 years ago
comment:2 by , 6 years ago
| Owner: | changed from to |
|---|---|
| Status: | new → assigned |
Karim has some ideas, will discuss them with Willi and Bernhard
comment:3 by , 5 years ago
@Karim, could you post a comment here with an update on the status of this issue?
Thanks!
comment:4 by , 5 years ago
An alpha version for better initialization of models with dynamic state selection is available via the flag "--initialStateSelection", currently only working for C. Is there a model for which it can be tested?
by , 5 years ago
| Attachment: | TestDynamicSelection.mo added |
|---|
comment:5 by , 5 years ago
The attached models are textbook examples - when the point mass is close to the x-axis, one should use y and vy as states, while x and vx should be used when close to the y-axis.
It seems that the current runtime can already select the right initial state in both cases, of course I couldn't easily find a case that fails (Murphy's law).
In the meantime while a proper solution is found for this issue, in case of failure couldn't we just re-try with some other random choice for a certain number of attempts? That would already be much better than an one-shot try