Wrong zero crossing function generation

[niklwors]

For this model:

model miniModel
  // PARAMETERS
  parameter Integer nZ = 5;

  // VARIABLES
  Integer lowerZrow;
  Real zIn = 4.3;
  Real[nZ] rowZ = {1,2,3,4,5};
algorithm
  // Finde zIndex
  lowerZrow := 0 "Reset Index";
  for i in 2:nZ loop
    if (i == nZ) then
      lowerZrow := i - 1;
      break;
    end if;
    if zIn < rowZ[i] then
      lowerZrow := i-1;
      break;
    end if;
  end for;
end miniModel;

The omc generate 4 zero crossing functions which is not correct

[lochel]

The debug-dump "relidx" prints:

number of relations : 0
 number of new zc : 4
 number of new zc : 4
 itmp : 4
collectZCAlgsFor result zc : i == 5 index:4
findZeroCrossings1 number of relations : 4
findZeroCrossings1 sample index: 0
findZeroCrossings1 sample index: 0

This output is not very helpful – at least for me.

The debug-dump "dumpindxdae" prints:

Zero Crossings (number of relations: 4)
========================================
2 == 5 with index = 0 in equations [1] and when conditions []
3 == 5 with index = 1 in equations [1] and when conditions []
4 == 5 with index = 2 in equations [1] and when conditions []
5 == 5 with index = 3 in equations [1] and when conditions []


Relations (number of relations: 4)
========================================
2 == 5 with index = 0 in equations [1] and when conditions []
3 == 5 with index = 1 in equations [1] and when conditions []
4 == 5 with index = 2 in equations [1] and when conditions []
5 == 5 with index = 3 in equations [1] and when conditions []

So there are indeed 4 zero crossings. What are the expected zero crossings? At least the simulation result seem to be what Dymola calculates.

[niklwors]

For discrete variables it should not be generate zero crossings.

[lochel]

Well, I can take care of this ticket. Obviously, the check if a relation is discrete doesn't work as it is supposed to. But that is probably not the only issue with zero crossings. I thought it would be good to also involve Willi since he developed the event stuff some time ago...

[lochel]

This seems to be fixed.

[sjoelund.se]

Milestone renamed

[sjoelund.se]

Milestone renamed