Is this situation a redundant initial condition?

[huihui11123]

model Test
  Real x;
initial equation 
  pre(x) = time - 10;
equation 
  when initial() then
    x = time + 5;
  elsewhen time>1 then
    x = 1;
  end when;
end Test;

This model can run in Dymola, but if the elsewhen branch is removed, Dymola will report redundant initial condition errors.

model Test
  Real x;
initial equation 
  pre(x) = time - 10;
equation 
  when initial() then
    x = time + 5;
  end when;
end Test;

In my opinion, the initial equations of these two models are the same; both are pre(x) = time - 10 and x = time + 5.
I don't know why the results in Dymola are different. Also, does this situation count as redundant initial conditions?
I have seen many such cases in the third-party library Buildings.

[henrikt-ma]

Sounds like an issue with Dymola. In System Modeler you get the results you expect.

[huihui11123]

Sounds like an issue with Dymola. In System Modeler you get the results you expect.

I know that during the initialization phase, for Integer and Boolean, x and pre(x) can have different values, but I'm not sure if this is also the case for discrete Real, that is, whether pre(x) = time - 10 and x = time + 5 should be treated as inconsistent initial conditions.

[henrikt-ma]

See https://specification.modelica.org/master/equations.html#initialization-initial-equation-and-initial-algorithm.

Since x is assigned in a when-clause, it is a discrete-time Real, and hence the rule about pre(vc) = vc does not apply.

I suspect the problem with Dymola is that they fail to see that x is discrete-time if the when-equation consists entirely of one branch triggered by initial().

Perhaps this is enough to make it work in Dymola?

  when {initial(), false} then
    x = time + 5;
  end when;

[HansOlsson]

I'll try to verify that it is indeed an issue in Dymola.

However, it is sort of redundant: you give an initialization for pre(x) and then compute x, which means that the value for pre(x) is ignored.

[henrikt-ma]

However, it is sort of redundant: you give an initialization for pre(x) and then compute x, which means that the value for pre(x) is ignored.

Right. In System Modeler, there is no way to access the value of pre(x) during initialization. To access it, one would need to modify the model. For instance by adding this equation:

  when initial() then
    Modelica.Utilities.Streams.print("pre(x): " + String(pre(x)));
  end when;

[HansOlsson]

As far as I can see is a tool-issue; so no change needed specification. (But thankful for the report.)