Connection Graph is not broken correctly

[Christian Schubert]

the models
Modelica.Electrical.QuasiStationary.SinglePhase.Examples.SeriesResonance
Modelica.Electrical.QuasiStationary.SinglePhase.Examples.ParallelResonance
fail because there is one equation missing. Looking into the code (+d=dumpdaelow) one notes that there is no equation for ground.pin.reference.gamma.
However, there are 4 equalityConstraint involving ground.pin.reference which do not count as equations, since they return an empty array.

I drew the connection graph with yEd (see attachment, dashed lines may be broken) and one can see that the "broken connections" (= equality constraints) do not even occur as edges in the connection graph. It seems that the equality constraints are produced from the connection set rather than the connection graph.

[anonymous]

These are the generated equality constraints:

2 : {} = Modelica.Electrical.QuasiStationary.Types.Reference.equalityConstraint(resistor.pin_n.reference, ground.pin.reference)
3 : {} = Modelica.Electrical.QuasiStationary.Types.Reference.equalityConstraint(inductor.pin_n.reference, ground.pin.reference)
4 : {} = Modelica.Electrical.QuasiStationary.Types.Reference.equalityConstraint(capacitor.pin_n.reference, ground.pin.reference)
5 : {} = Modelica.Electrical.QuasiStationary.Types.Reference.equalityConstraint(voltageSensor.pin_n.reference, ground.pin.reference)

[Adrian Pop]

You can run:

omc +locale=C +d=cgraphGraphVizFile Modelica.Electrical.QuasiStationary.SinglePhase.Examples.ParallelResonance.mos

to get Modelica.Electrical.QuasiStationary.SinglePhase.Examples.ParallelResonance.gv
with the connection graph and where it was broken. You can view the graph using GraphViz.
See the file I will attach next.

[Adrian Pop]

obtained via ./omc +locale=C +d=cgraphGraphVizFile Modelica.Electrical.QuasiStationary.SinglePhase.Examples.ParallelResonance.mos

[Adrian Pop]

I did:

loadModel(Modelica, {"3.2.1"}); getErrorString();
instantiateModel(Modelica.Electrical.QuasiStationary.SinglePhase.Examples.ParallelResonance); getErrorString();

and I get a model having equation:

  if false then
  ground.pin.reference.gamma = 0.0;
  end if;

it comes from:

  Connections.potentialRoot(pin.reference, 256);
  if Connections.isRoot(pin.reference) then
    pin.reference.gamma = 0;
  end if;

which is rather correct as Connections.isRoot(ground.pin.reference) is only a potential root, not a root. Also, checkModel reports:

Class Modelica.Electrical.QuasiStationary.SinglePhase.Examples.ParallelResonance has 99 equation(s) and 99 variable(s).
57 of these are trivial equation(s).

so the model is balanced. The problem might be in the back-end someplace.

[Adrian Pop]

Hm, the connection graph obtained via OpenModelica is rather different than your hand done one. Are you sure yours is correct?

[Jens Frenkel]

Modelica Code:

connect(currentSource.pin_n, resistor.pin_p);
connect(currentSource.pin_n, inductor.pin_p);
connect(currentSource.pin_n, capacitor.pin_p);
connect(currentSource.pin_n, voltageSensor.pin_p);
connect(currentSource.pin_p, ground.pin);
connect(resistor.pin_n, ground.pin);
connect(inductor.pin_n, ground.pin);
connect(capacitor.pin_n, ground.pin);
connect(voltageSensor.pin_n, ground.pin);
connect(f.y, currentSource.f);
connect(polarToComplex.y, currentSource.I);
connect(phi.y, polarToComplex.phi);
connect(I.y, polarToComplex.len);
connect(voltageSensor.y, complexToPolar.u);

Dymola Flat Code:

connect(capacitor.pin_n.reference, currentSource.pin_p.reference);
connect(capacitor.pin_n.reference, ground.pin.reference);
connect(capacitor.pin_n.reference, inductor.pin_n.reference);
connect(capacitor.pin_n.reference, resistor.pin_n.reference);
connect(capacitor.pin_n.reference, voltageSensor.pin_n.reference);
connect(capacitor.pin_p.reference, currentSource.pin_n.reference);
connect(capacitor.pin_p.reference, inductor.pin_p.reference);
connect(capacitor.pin_p.reference, resistor.pin_p.reference);
connect(capacitor.pin_p.reference, voltageSensor.pin_p.reference);

[Adrian Pop]

There is no description in the specification on how is best to connect things to get the connection graph. I see that Dymola goes via ground.pin .reference and currentSource.pin_p and connects the capacitor.pin_n to those. Why is done this way, i have no idea.

[Jens Frenkel]

It seems this is not the problem but for the conntection between

ground.pin.reference and currentSource.pin_p.reference is no equation generated

currentSource.pin_p.reference.gamma = currentSource.pin_n.reference.gamma;
currentSource.omega = der(currentSource.pin_p.reference.gamma);

if false then
ground.pin.reference.gamma = 0.0;
end if;
{} = Modelica.Electrical.QuasiStationary.Types.Reference.equalityConstraint(resistor.pin_n.reference, ground.pin.reference);
{} = Modelica.Electrical.QuasiStationary.Types.Reference.equalityConstraint(inductor.pin_n.reference, ground.pin.reference);
{} = Modelica.Electrical.QuasiStationary.Types.Reference.equalityConstraint(capacitor.pin_n.reference, ground.pin.reference);
{} = Modelica.Electrical.QuasiStationary.Types.Reference.equalityConstraint(voltageSensor.pin_n.reference, ground.pin.reference);

[Adrian Pop]

Ok, I get it if I look at the flattened code from OpenModelica.
There is not equation from connect(ground.pin.reference, currentSource.pin_p.reference) even if the connect is not broken it that case. I'll have a look at it.

[Adrian Pop]

Done by hand using GraphViz from equations posted by Jens.

[Jens Frenkel]

in the functino handleOverconstrainedConnections pefore calling the funtion findResultGraph there is the equation

EQUEQUATION(capacitor.pin_n.reference.gamma = ground.pin.reference.gamma)

after calling this funktion it is not there anymore.

[Adrian Pop]

Yes, i'm debugging that now. The equations are removed based on the equation source so it might be something wrong there.

[Adrian Pop]

I found the issue and I'm doing a way better solution by deleting the connections directly from the connections sets before generating the final equations. I'll keep you informed on how it goes.

[Adrian Pop]

Fixed in revision r13778.
These models now simulate but there are some error messages that should not be there.