[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [TRNSYS-users] Stuck simulation



Dear David,

I finally found the problem using debugging with Compaq Visual Fortran.
The simulation got stuck in an infinite loop in the TYPE753 (TESS, Heating
Coil with by-pass fraction). In the source code there is a convergence
criteria (tolerance : T_CONV = 0.01) that remains unsatisfied in my case.
As there is no condition to stop it it leads to an infinite loop.

The developer writes in the file : "IF THE DATA APPEARS "SPIKY" OR THE
SIMULATIONS ARE TOO SLOW, YOU MAY HAVE TO ADJUST THE CONVERGENCE CRITERIA
ON THE TEMPERATURE (T_CONV)". If I set T_CONV = 1, the simulation runs
without any problem.

For the moment I cannot explain why I got this non-convergence. I need to
read the source code more carefully.

Kind regards,
Vincent


> Vincent,
>   Usually when you get into that "disassembly window" it means that the
> actual crash has occurred inside the TRNExe and not inside the Fortran
> code. Its not any easy thing to figure out. One thing you might try, if
> you haven't already, is to turn on the debugging switch in the Studio
> control cards. Among other things, that will activate two checks at the
> end of each time step. One check will make sure that none of the
> components wrote outside of the OUT() array space that is allocated to
> them. The other check will make sure that none of the outputs of any of
> the components get set to the "NaN" (not a number) condition.
> Kind regards,
>  David
>
>
> Vincent Dolisy wrote:
>> Dear David,
>>
>> I've got a Fortran compiler and I can run the simulation from this.
>> (Compaq Visual Fortran)
>>
>> I have set a breakpoint in the trnsys.for file at the end of the
>> timestep:
>>
>> C    AT THIS POINT THE SIMULATION IS FINISHED AT THE CURRENT TIMESTEP
>>      RETURN
>>
>> In fact I add this breakpoint at the timestep just before the simulation
>> begins to be stuck. Then I run it line by line and hope to find an
>> infinite loop or something like that. When I run it line by line the
>> source code window switches to the "Disassembly window" so that I cannot
>> see any source code lines anymore. In the "Disassembly window" each line
>> is scanned one by one until the cursor stops at the line:
>>
>> 0046F993   call        eax
>>
>> Do you know how I can find the source code line (or file) corresponding
>> to
>> a line in the "Disassembly window"?
>>
>> The way I debug is maybe not good. Do you know any good strategy to
>> debug
>> this kind of problem ?
>>
>> Thanks a lot.
>>
>> Kind regards,
>> Vincent
>>
>>
>>
>>
>>> Vincent,
>>>
>>>
>>>> *TRACE*  UNIT  51  TYPE  56  AT TIME   2.6764166666666665E+03
>>>> ITERATION
>>>>  2  CALL   13105  POST-CONVERGENCE CALL. (can you tell me what it
>>>> means?)
>>>>
>>>>
>>> once TRNSYS converges on a solution at a given time step, the kernel
>>> calls all types once more so that they can do anything that they might
>>> need to do (update summaries, print reports for that time step, make
>>> control decisions that are not supposed to be iterative, etc.)
>>>
>>>> The simulation time step is 1 minute and if I use 5 minutes it does
>>>> not
>>>> hang up at the same hour.
>>>>
>>>> Moreover I have already encountered the problem during the night, that
>>>> is
>>>> to say when the system is turned off (heating coils, cooling coils,
>>>> humidifiers, PID's are switched off).
>>>>
>>>>
>>> all of those suggest that there is something that gets stuck in an
>>> infinite loop. The only way to really track it down is to run it from
>>> the Fortran compiler I am afraid,
>>> Kind regards,
>>>  David
>>>
>>>
>>> --
>>> ***********************************************************************
>>> Thermal Energy System Specialists (TESS), LLC
>>> David BRADLEY                      22 N. Carroll Street - Suite 370
>>> Partner                            Madison, WI 53703
>>>                                    USA
>>> P: +1.608.274.2577
>>> F: +1.608.278.1475
>>> E-mail: bradley@tess-inc.com
>>> Web Pages:  http://www.tess-inc.com    and      http://www.trnsys.com
>>>
>>> ***********************************************************************
>>>
>>> _______________________________________________
>>> TRNSYS-users mailing list
>>> TRNSYS-users@cae.wisc.edu
>>> https://www-old.cae.wisc.edu/mailman/listinfo/trnsys-users
>>>
>>>
>>
>>
>>
>
> --
> ***********************************************************************
> Thermal Energy System Specialists (TESS), LLC
> David BRADLEY                      22 N. Carroll Street - Suite 370
> Partner                            Madison, WI 53703
>                                    USA
> P: +1.608.274.2577
> F: +1.608.278.1475
> E-mail: bradley@tess-inc.com
> Web Pages:  http://www.tess-inc.com    and      http://www.trnsys.com
>
> ***********************************************************************
>
>


-- 
Vincent DOLISY
Ingénieur de recherche

Université de Liège
Département des Sciences et Gestion de l'Environnement
Unité de Surveillance de l'Environnement (Energie)
185, Avenue de Longwy
B-6700 ARLON
Tel : 32-(0)63 230 937
Fax : 32-(0)63 230 800
Mail: Vincent.Dolisy@ulg.ac.be
WEB : http://www.dsge-arlon.ulg.ac.be/energie/