SVN 3255 and SVN 3263 produce significantly different results

[gforney]

SVN 3255 and SVN 3263 were run on a case involving helium 
release in a box with vents (case file is attached). 
Helium concentrations were significantly different at 3600 s 
after release of helium. Small differences between the results 
for the two runs can be seen as early as 30-40 s after the release 
of helium.

Original issue reported on code.google.com by prasad.kuldeep on 2009-03-17 21:08:57

---

• _Attachment: He_3255.fds_

[gforney]

SVN 3200s are from over a month ago.  Current SVN is 3514.  Run with the latest
version and determine if you feel the results are incorrect and if so how they are
incorrect.  Differences (even large differences) are anticipated as we implement new
features and code changes between releases.  There have been changes to velocity
boundary conditions and bugs fixed during implementation of new approaches for
determining thermophysical properties.

Original issue reported on code.google.com by drjfloyd on 2009-03-17 21:27:11

[gforney]

SVN 3193, 3221, 3255 produce virtually identical results for the case He_3255.fds.
SVN 3263 produces dramatically different results as compared with 3255. Results for

SVN 3294, 3392, 3480 produce results virtually identical to those with svn 3263.

I have just started a case with svn 3514, I expect the results to be similar to 
those from 3263. 

Results from svn 3255 show better comparison with experimental data and the 
comparison is not as good with svn 3263.  Otherwise the results from svn 3263 do not

appear to be incorrect. 

Suspect an issue with the specific heat calculation for non mixture fraction based

scenarios.

Original issue reported on code.google.com by prasad.kuldeep on 2009-03-17 21:45:38

[gforney]

I've got the case running now with source 3490 and with a quick glance SPECIFIC HEAT
and MOLECULAR WEIGHT looks OK.

Try running the old versions with outputs of SPECIFIC HEAT, MOLECULAR WEIGHT, and the
MASS FRACTION of HELIUM.  Specific heat should be a mass weighted sum of the specific
heats of the constituent gasses.  

Original issue reported on code.google.com by drjfloyd on 2009-03-17 22:17:58

[gforney]

He specific heat was added to SVN3479 (issue 677).

Original issue reported on code.google.com by jukka.vaari@vtt.fi on 2009-03-18 13:07:17

[gforney]

I ran 200 s of the case with the latest version of FDS.  At the burner surface SLCFs
show that the He volume fraction is ~0.78, the cp is 2.4 kJ/kg-K, the enthalpy is 820
kJ/kg, and the MW is 9.4 g/mol.  Using the volume fraction and computing by hand cp
and MW I get 2.4, 820, and 9.4.  I do not see any indication that the current version
of FDS is incorrectly computing thermophysical properties for this case.

There were changes made to precision and accuracy with which those properties were
computed between SVN 3255 and SVN 3263.  

Original issue reported on code.google.com by drjfloyd on 2009-03-18 13:56:05

[gforney]

Another quantity to monitor (via a SLCF or DEVC) is TEMPERATURE. The temperature is

related to the density and species mass fractions via the equation of state. The 
temperature should remain close to ambient, but there can be minor deviations. If 
these are large, it can indicate a problem.

Original issue reported on code.google.com by mcgratta on 2009-03-18 14:05:34

[gforney]

TEMPERATURE won't do anything here as ISOTHERMAL=.TRUE. has been set in the input

Original issue reported on code.google.com by drjfloyd on 2009-03-18 14:08:02

[gforney]

I notice that the pressure fields for runs with svn 3255 and 3263 are quite 
different. Images are attached.

Original issue reported on code.google.com by prasad.kuldeep on 2009-03-18 14:18:26

---

• _Attachment: He_3255_0020.png
  _
• _Attachment: He_3263_0020.png
  _

[gforney]

Quite different?  Differences are fractions of Pa.

In the current version, do you feel that there is a bug in the computation and if so
what quantity do you feel is in error?

You have a case that appears to be a not very well resolved plume in a box with an
opening that is also not highly resolved using grids with high aspect ratios.  The
fact that answers have changed between versions is not surprising.  I don't see
anything as of yet to indicate that there is an error in the calculation with the
current version of FDS: molecular weight appears correct, specific heat appears
correct, enthalpy appears correct, flow vectors appear OK, and the velocity at the
opening given the pressure is reasonable.

Original issue reported on code.google.com by drjfloyd on 2009-03-18 14:40:30

[gforney]

Enthalpy of air in version 3255 and 3263 that are computed are quite different. 
Specific heats of air computed by the two versions are similar. 

I have attached enthalpy and specific heat slice files for 3255 and 3263.

Original issue reported on code.google.com by prasad.kuldeep on 2009-03-18 15:43:27

---

• _Attachment: He_3255_Enthalpy.png
  _
• _Attachment: He_3255_Cp.png
  _
• _Attachment: He_3263_Enthalpy.png
  _
• _Attachment: He_3263_Cp.png
  _

[gforney]

Changes were made in 3263 to make the enthalpy reference point consistent for the
gasses.  As stated in comment 9, the specific enthalpy appears to be done correctly
in the current version for your case.  

I note in your inputs that you have changed the default values of some of the
dimensionless parameters.  Was this done to tune FDS to get match the experiment? 
If
that is the case, then the fact that improvements to FDS change the code behavior
does not seem to be a great concern. I expect as we work to improve FDS that there
will be changes in the solutions to prior cases.  I think the changes made in recent
months to the computation of thermophysical properties is improving the consistency
of those values over the various modes of FDS operation.  That is not to say that
there may not still be bugs in that approach, but as of yet I do not see any results
with your inputs that suggest this is the case.  What in the current version do you
feel is incorrect?  

Original issue reported on code.google.com by drjfloyd on 2009-03-18 15:57:36

[gforney]

In version 3263, the enthalpy of the surrounding air is computed to be higher than 
the enthalpy of the helium jet. In version 3255 the enthalpy of the surrounding air

is lower than that for the helium jet. As a result, the net enthalpy flux due to 
species diffusion will be different for the two cases. This will have a direct 
effect on the divergence computed in equations 3.18 and 3.16 of the technical 
reference guide. 

The specific heat for helium is higher than that for air, so the enthalpy computed

should also be higher for helium than for air (irrespective of the reference point).

Original issue reported on code.google.com by prasad.kuldeep on 2009-03-18 17:33:34

[gforney]

Old versions are in general not the concern of the issue tracker, they are old
versions because we fix bugs / make improvements and make new versions which we hope
are improvements without introducing new errors. Our concern is with the current
version and is the current version correct. The current version shows that air plus
helium has a higher enthalpy than air alone.

From "Thermodynamic Properties in SI", WC Reynolds the specific enthalpy of air at
300 K (27 C) is 459.  FDS (the current version) is showing air at 21 C (your calc)
to
be 452. Since cp air is ~ 1.04 kJ/kg-K this is correct.  From the same reference pure
helium at 300 K is 1570 kJ/kg.  At 200 s into the calc with the current version there
is about 1/3 mass fraction of helium just above the burner.  FDS has enthalpy of ~815
kJ/kg and a hand calc gives the same.  Where is the error in the current version? 

You results have changed with a new version of FDS, but unless it can be demonstrated
that there is an error in the latest version (not in 3263 or 3255 we are not at
3514), there is nothing for us to fix.

Original issue reported on code.google.com by drjfloyd on 2009-03-18 17:52:49

[gforney]

Calculations that I have made with svn 3480 indicate that the enthalpy of the helium

jet is smaller than that of the air. This is incorrect, as you have indicated based

on hand calculations and thermophysical property data from WC Reynolds. 

I am making calcs with 3520 to see if the problem still exists.

Kuldeep

Original issue reported on code.google.com by prasad.kuldeep on 2009-03-18 18:22:42

[gforney]

Attached image (run last night with 3490) shows FDS predicts the enthalpy of the
mixture to be higher as indicated in comment 13. Note that 3481 fixed an error in the
MW calc which wouldn't surprise me if it affected enthalpy as well.

Original issue reported on code.google.com by drjfloyd on 2009-03-18 18:28:59

---

• _Attachment: He_3263_0020.png
  _

[gforney]

attached wrong image.  correct image attached.

Original issue reported on code.google.com by drjfloyd on 2009-03-18 18:30:10

---

• _Attachment: He_3255_1000.png
  _

[gforney]

Yes, I see the change that was made in 3481. This might have fixed the bug. All 
my calculations have been with version 3480 and earlier. 
I will run with 3520 and then close the issue, if everything looks good.

Kuldeep

Original issue reported on code.google.com by prasad.kuldeep on 2009-03-18 18:34:26

[gforney]

Enthalpy plots with svn 3520 again indicate that the enthalpy of helium jet is lower

than that of the surround air. see attachment. 
This is consistent with those from r3263. 

Original issue reported on code.google.com by prasad.kuldeep on 2009-03-18 19:26:16

---

• _Attachment: He_3520_0003.png
  _

[gforney]

You graphic cuts off the plot units, are you plotting SPECIFIC ENTHALPY (kJ/kg) or
ENTHALPY (kJ/m^3) SPECIFIC ENTHALPY should be higher (the results I showed in comment
16), but since helium is much less dense ENTHALPY will be lower.

Original issue reported on code.google.com by drjfloyd on 2009-03-18 19:33:27

[gforney]

I notice that the U Guide was not updated to reflect SPECIFIC ENTHALPY and ENTHALPY.
 I'll make that change.  

Original issue reported on code.google.com by drjfloyd on 2009-03-18 19:46:41

[gforney]

Summary:

svn 2087 through 3255 produced similar set of results on the helium test case 
problem. These results compared favorably with the experimental data. 

svn 3263 through 3480 produced results that were significantly lower than the 
experimental data. 

svn 3520 produces results that again compare with experimental data and are similar

to those obtained from svn 2087-3255.

I suspect that a bug was introduced into the software around 3263 and then the bug

got removed around svn 3481. The bug was most likely related to the calculation of

molecular weights. 

Original issue reported on code.google.com by prasad.kuldeep on 2009-03-18 20:44:23

[gforney]

Good sleuthing, everyone. This is the power of version control. 

Original issue reported on code.google.com by mcgratta on 2009-03-18 22:32:24