Erroneous sign of a flow variable

[massimo ceraolo]

In the attached model "asmaFlow" it should be:
aimc.flange.tau=aimc.inertiaRotor.flange_b.tau

instead it is:
aimc.flange.tau=-aimc.inertiaRotor.flange_b.tau

I think this is Critical since it causes a formally valid simulator to be created, but that gives erroneous results.

[Adrian Pop]

Hm, how is that? I get this where there is no equation like you say:

(-aimc.flange.tau) + aimc.inertiaRotor.flange_b.tau + aimc.friction.flange.tau + aimc.strayLoad.flange.tau = 0.0;

[Adrian Pop]

Anyhow you have this equation in PartialBasicMachine:

connect(inertiaRotor.flange_b, flange);

where (if I can still interpret the Modelica Spec rules for connections right):
flange is an outside connector (- sign) and inertiaRotor.flange_b is an inside connector + sign so this will give you a set:
{<aimc.flange.tau, outside>, <aimc.inertialRotor.flange_b.tau, inside>}
So the equation above
(-aimc.flange.tau) + aimc.inertiaRotor.flange_b.tau ... = 0
is correct (as is combined with the other connection set which has only inside connectors with + sign).

I also cc Per as he's our expert on connections. Per, any comments on this?

[Adrian Pop]

Total model of the asmaFlow

[Adrian Pop]

instantiation of asmaFlow + dumpindxdae

[massimo ceraolo]

What is strange is that, despite of the correct equation you mention, ardpo:

(-aimc.flange.tau) + aimc.inertiaRotor.flange_b.tau ... = 0

in the plots I see numbers that indicate that the following equation is actually true:

aimc.flange.tau + aimc.inertiaRotor.flange_b.tau = 0

that is wrong (the other terms summed in the equation are zero or negligible).

The issue can be checked also simulating the model with Dymola, that gives the correct result.
I will try to see if I'm able to produce a simpler example showing the issue (not sure).

I consider this issue serious since OM does simulate the model, but gives wrong result: this behavior tends to undermine the user's trust of OM.

[massimo ceraolo]

I have an additional result.
If we remove from the model the speed sensor "speedSensor" the result becomes correct!
(therefore the quantity inertiaRotor.flange_b.tau changes from +15Nm to -15Nm).
So the presence of a sensor totally alters the result.

[Adrian Pop]

Ok. The front-end seems to generate the correct equations.
Maybe the issue is in the simplification or the back-end.
If you look into trace.txt attached here (the dumpindxdae part)
we can see that aimc.flange.tau is an alias for -(-const.k)

251: aimc.flange.tau:VARIABLE(flow=true )  = -(-const.k)

and aimc.inertiaRotor.flange_b.tau is an alias for -const.k

267: aimc.inertiaRotor.flange_b.tau:VARIABLE(flow=true )  = -const.k

Do you know which of these is wrong?
I guess the back-end does something wrong here, but I don't quite get what it is.

I'll move this from front-end to the back-end

[Adrian Pop]

As far as I can tell the plot should display:

• -const.k for aimc.inertiaRotor.flange_b.tau
• const.k for aimc.flange.tau

Maybe this is a problem on how we plot (or calculate) alias variables from the .mat file.
I mean the simulation goes fine but the plot (or read/write from/to the .mat file) is wrong.

Massimo, could you open and plot the .mat file generated by OpenModelica with Dymola and see if the results in the file are actually correct or not?

[Adrian Pop]

Massimo, what stopTime do you give when you simulate the model?

[Adrian Pop]

Nevermind, you have this in the model:
experiment(StopTime = 10, NumberOfIntervals = 10000)

[Adrian Pop]

In the .mat file the results are the same for both:
aimc.inertiaRotor.flange_b.tau and aimc.flange.tau respectively -15

I think this is a bug in alias removal.
I think RemoveSimpleEquations somehow handles -(-e) a bit wrong (as negated alias).

Willi, can you have a look?

[Adrian Pop]

I think I found the issue. I'll let you know as soon as I test a bit.

[massimo ceraolo]

Sorry for not answering earlier.
I've never looked at OM _res.mat files with Dymola before today, and indeed it seems that something is bad (maybe I make some mistake).

Indeed the plot file that OM creates from asmaFlow seems to be faulty. In OMEdit, for instance, if I try to see
aims.inertiaRotor.flange_a.tau
I don't see anything.

Indeed the file I proposed yesterday, when I opened this ticket, came from a simplification of another one, in which I could see all variables.
That file was faulty also.
When this ticket is solved, I will check what happens with the original file, to see if that one is solved also or something wrong remains.

*
Just in case it can help, I add the mat file that Dymola creates from asmaFlow.mo (for a simulation lasting 1s and having 1000 output points)).

Massimo

[Adrian Pop]

I'm working on a fix for this issue but is more complicated that I though.
I quite know what the problem is now but it will take a while to implement a fix.
I'll continue on it tomorrow and let you know when is fixed.

[Adrian Pop]

model MinimalModel
 Real a;
 Real b = 0.0;
 Real c;
 Real d;
 parameter Real const_k(start = 1.0) = -15.0;
equation
 c = -d;
 d + (-const_k) = b;
 c + a = b;
end MinimalModel;

[Adrian Pop]

I made a minimal model that has this bug.
I now assigned this bug to Willi.

[Adrian Pop]

Added Jens also to CC maybe he has some ideas on what might be wrong.
Me and Willi looked at RemoveSimpleEquations.mo for quite some time and we don't get what is wrong (I think there is something wrong with the negation, Willi thinks is something wrong with the order in which we apply the replacements).

[Jens Frenkel]

• fixed the bug local but svn take some time to load it up.

[Jens Frenkel]

patch for #2429

[Adrian Pop]

Thanks Jens, yes, SVN is a bit slow because Hudson (we start Hudson jobs from SVN post commit hooks) has some issues that I try to fix at the moment. I'll run the testsuite with your patch and see how it goes.

[Adrian Pop]

Fixed in r17836 by Jens.