Fixes for #4424

[casella]

Added small pressure losses to avoid index reduction with fixed pressure boundaries. Changed to explicit steady state initialization to avoid unspecified initial conditions.

Fixes #4424

[casella]

This works in Dymola but not in OpenModelica 😢

[casella]

There is something fishy in the way the pressure loss models work, I'm getting much smaller pressure losse than I expected, which makes the model ill-posed. I'll investigate more.

[casella]

I think I accidentally found a bug in Fluid.Dissipation:

ModelicaStandardLibrary/Modelica/Fluid/Dissipation.mo

Lines 4618 to 4625 in 9e8b7fb

	M_FLOW := IN_var.rho*(if a>0 and b<=0 then
	Modelica.Fluid.Dissipation.Utilities.Functions.General.SmoothPower(
	(1/a)*dp,
	(1/a)*dp_min,
	0.5)
	elseif a>0 and b>0 then
	sign(dp)*(-b/(2*a) + sqrt((b/(2*a))^2 + (1/a)*abs(dp)))
	else b*dp);

I guess the last line (the else case) should be 1/b*dp, not dp. Probably nobody has ever used this option, so it went unnoticed until now. I'm checking with Stefan and I'm going to fix this ASAP.

[casella]

In the meantime, I changed the parametrization of the pressure losses to use a quadratic model and get about 0.1 bar pressure loss. I also switched to FixedInitial to avoid convergence problems. This works both in Dymola and OMC.

@maltelenz could you please check if it also works in WSM?

[maltelenz]

ModelicaTest.Fluid.TestComponents.Pipes.DynamicPipeEnergyConservationCheck2 simulates to completion in WSM.

The new results do not match the previously suggested reference (I assume this is expected).

There are still transients in the beginning, I don't know if this is expected.

[casella]

ModelicaTest.Fluid.TestComponents.Pipes.DynamicPipeEnergyConservationCheck2 simulates to completion in WSM.

Good.

The new results do not match the previously suggested reference (I assume this is expected).

Yes, since I added some pressure losses the pressures are different. But that doesn't matter, since the point of this model is to run the asserts. I'll ask LTX to re-generate the reference results in #4424.

There are still transients in the beginning, I don't know if this is expected.

Yes, it is. I tried steady-state initialization but I got into numerical trouble with OMC, so I though we could keep the same default behaviour we had previously with Dymola (i.e., fix all the states to their initial values), only make it explicit. There is a relaxation transient, then the asserts are only applied at the end. This is also improved by PR #4555.

So, you can approve this without problem. Thanks!

[maltelenz]

Works in System Modeler, and the results seem as expected per previous comment.