Modelica.Fluid.Examples.TraceSubstances.RoomCO2WithControls is problematic because of chattering

[casella]

As reported in #4305, Modelica.Fluid.Examples.TraceSubstances.RoomCO2WithControls is problematic: Dymola 2024x fails with it and so does OpenModelica. Apparently there are some reasons to generate chattering.

This should be investigated before the 4.1.0 release.

[maltelenz]

We also see chattering around the same time in System Modeler.

[casella]

I'll have a closer look ASAP

[MatthiasBSchaefer]

This is the commit, causing the chattering according to my tests in Dymola:
930456a

[HansOlsson]

I tried to debug it and preliminarily it seems that the linear part makes sense, but needed to be applied in additional cases.

Specifically the code has:

      if  xi1 < x0 or xi2>x1 then
        // The limits don't make sense, just use linear interpolation
        y := (y1-y0)*(x-x0)/(x1-x0) + y0;
      else

And what happens here is that xi1 and xi2 are practically identical, but x0 and x1 are different.
Adding or abs(xi2-xi1)<100*Modelica.Constants.eps fixes this specific case, but it should be investigated a bit more.

[HansOlsson]

I tried to apply the above and run all the Fluid-tests (without looking at the results) (on maint/4.1.x since I needed MSL-binaries).

In Dymola 2025x (updated) with Advanced.Translation.PlaceDymolaSourceFirst = 2; all fluid examples ran.
I'll make a proper PR of this.

[casella]

It seems that @HansOlsson managed to fix the remaining issues with regfun3 in PR #4522. When that is merged on master we can check if there are any issues popping up with other models and if we have no evidence of that we can back-port it to 4.1.0.

[casella]

As already noted in #3758, some of these issues are popping up because we are using flow models that are meant to handle static head in the trivial cases where the pipe is exactly horizontal. As I understand, this makes the numerical behaviour of regfun3 tricky. As suggested by @sielemann in #3758, we should probably fix Modelica.Fluid.Pipes.BaseClasses.FlowModels.PartialGenericPipeFlow so that it completely skips the part that accounts for static head if max(abs(dheights)) < eps.

Given the quite complex structure of the pipe model, this is not completely trivial, so I'd keep this as an improvement for MSL 4.2.0, see #4525.

[casella]

Reopen until back-ported to maint/4.1.x

[casella]

@Esther-Devakirubai please back-port #4522 to maint/v4.1.x.

We are only closing this ticket for good after we check that there are no regressions on 4.1.0 on the LTX serves

[beutlich]

please back-port #4522 to maint/v4.1.x.

Done by #4533.