Initialization: start value which is an unfixed parameter

[Christian Schubert]

Hi,

while converting a model from Dymola to OpenModelica we encountered the following problem, which I hope is easy to fix.

Consider the following model

model UnfixedParameter
    parameter Real c(fixed = false);
    Real x(start=c,fixed=true);
initial equation 
    c = 1;
equation 
    der(x) = -x;
end UnfixedParameter;

Instead of starting at x(t=0)=1, it starts at 0 which is unexpected.
It's got probably to do with when start values are evaluated.

[Christian Schubert]

Testfile

[Lennart Ochel]

I get the following messages/errors:

Error: Internal error Transformation Module PFPlusExt index Reduction Method Pantelides failed!
Warning: No system for the symbolic initialization was generated. A method using numerical algorithms will be used instead.

- BackendVariable.vararrayNth 1 has NONE!!!

I will fix it.

[Vitalij Ruge]

This issue see very close to #2207

[Lennart Ochel]

The previously mentioned errors just occurred due to my corrupted omc version.
The combination of a fixed state having a start-expression depending on an unfixed parameter is not yet supported. Good ticket ;-)

[Lennart Ochel]

Again, this is a very good ticket. Thanks for contributing it. I included the model from above in the test suite.

Fixed in r18967.

@Volker: Some hpcom tests need to get updated. Can you do that?

[georg.hess@…]

Unfortunately, this doesn't completely fix my problem.
Consider the following model

model Fehler
encapsulated model SubModel
  parameter Real n = 1 annotation(Evaluate = true);
end SubModel;

  parameter Real n_b(fixed = false, start = 2);

  SubModel subModel(n = n_b);

initial equation 
  n_b = 5;

end Fehler;

The value of parameter n is not correctly calculated if n has the annotation Evaluate=true. If the annotation is not present everything works fine.

[Lennart Ochel]

Replying to georg.hess@…:

Unfortunately, this doesn't completely fix my problem.
Consider the following model

model Fehler
encapsulated model SubModel
  parameter Real n = 1 annotation(Evaluate = true);
end SubModel;

  parameter Real n_b(fixed = false, start = 2);

  SubModel subModel(n = n_b);

initial equation 
  n_b = 5;

end Fehler;

The value of parameter n is not correctly calculated if n has the annotation Evaluate=true. If the annotation is not present everything works fine.

n should be initialized with 5. This is the case with and without the Evaluate annotation. The annotation is a Compiler hint to get speedup during symbolic manipulation. (Annotations should never effect the symbolic results of a model.) Hence, your model is working fine and I see no wrong behavior of OpenModelica.