Change to NoInit in ClockedWithDiscretizedContinuousController #4465
Conversation
beutlich
added
example
Modelica/Clocked/Examples/SimpleControlledDrive/ClockedWithDiscretizedContinuousController.mo
Show resolved
Hide resolved
Co-authored-by: Hans Olsson <[email protected]>
There was a problem hiding this comment.
With @HansOlsson's suggestion this looks good to me, also for MSL 4.1.0
|
@Esther-Devakirubai, please back-port this to maint/4.1.x as soon as it is merged in. Thanks! |
| initType=Modelica.Blocks.Types.Init.InitialOutput, | ||
| y_start=0, | ||
| initType=Modelica.Blocks.Types.Init.NoInit, | ||
| x(fixed=true), |
There was a problem hiding this comment.
Isn't x(fixed = true) a violation of this rule from https://specification.modelica.org/master/synchronous-language-elements.html#initialization-of-clocked-partitions?
Attribute
fixedcannot be applied to variables in clocked partitions. The attributefixedis true for variables to whichpreviousis applied, otherwise false.
There was a problem hiding this comment.
Ouch. So, how are initial conditions specified for such clocked partitions?
There was a problem hiding this comment.
That's a valid question, but more work on the specification side is needed to answer it. Until then, I would just find it best to avoid breaking one of the few simple rules we have.
By the way, System Modeler also warns about underdetermined initialization where PI.x was automatically selected to be initialized from start, so it is not only in Dymola that this example will not run as cleanly as we'd ideally want it to.
There was a problem hiding this comment.
I should also mention that System Modeler is lacking a check for
Attribute
fixedcannot be applied to variables in clocked partitions.
so I can't tell how close the rest of the MSL is to complying with the rule.
There was a problem hiding this comment.
After some more thinking I think the specification got a bit messed up when trying to clarify.
Basically I don't see how we can get this to work as desired for MSL with the current specification:
- All examples in MSL should have fully defined initial conditions.
- Clocked components should have neither initial equations, nor fixed; https://specification.modelica.org/master/synchronous-language-elements.html#initialization-of-clocked-partitions
However, going back to 3.5 the rule for initial equations and fixed only applied to "clocked discrete-time partitions", which are now called "discrete-time sub-partitions". That should be restored, and we can see that example in https://specification.modelica.org/master/synchronous-language-elements.html#solver-methods has fixed=true.
Obviously we will later clarify initialization for "discrete-time sub-partitions".
There was a problem hiding this comment.
The "having fully defined initial conditions" is a strong requirement for tool-independent behaviour. If this is not possible with the current specification, I agree at relaxing/changing it. @HansOlsson would the proposed changes to the code OK with your proposed change to the MLS?
There was a problem hiding this comment.
With the proposed change to the specification there is no need to update this model at all, as the restrictions apply equally to initial equations and fixed=true.
|
@casella @HansOlsson Will Backport this as per Casella's comment #4465 (comment) |
No, we first need to clear up if it is needed or not (and if needed exactly how). My current interpretation is that it isn't needed and the specification should be updated instead. |
|
Should not be needed after modelica/ModelicaSpecification#3577 is merged. |
@HansOlssonn, I'm closing this PR accordingly. |
Fixes #4458.
As discussed in today's MAP-Lib monthly meeting.
There appears to be no need to update the reference result.