search

ReactionMechanismGenerator / RMG-Java

The Java version of RMG: Reaction Mechanism Generator
http://rmg.sourceforge.net/
MIT License
29 stars 36 forks source link

Dassl AUTO completes simulation although RMG thinks not converged. #114

Closed rwest closed 13 years ago

rwest commented 14 years ago

Running with DASSL AUTO but no pruning. My understanding is that dassl should interrupt the simulation if the critical flux is exceeded. At the end of the simulation, the java code checks for validity.

My simulations are running to the end, but then the java is deciding that at least one of them is invalid. This suggests to me that the java flux evaluation disagrees with the dassl flux evaluation. I have not dug further into this however. I suppose it's possible that it's a rounding error.

If anyone is editing this section of the code (eg. to put in the new expansion / pruning stuff) perhaps put in some checks so we can be sure the fluxes are consistent?


For reaction system: 1 out of 12
At this time: 0.5SEC
Conversion of Mfmt is:6.917966566266642E-8
For reaction system: 2 out of 12
At this time: 0.5SEC
Conversion of Mfmt is:2.175440393381578E-7
For reaction system: 3 out of 12
At this time: 0.5SEC
Conversion of Mfmt is:8.520506442621922E-7
For reaction system: 4 out of 12
At this time: 0.5SEC
Conversion of Mfmt is:1.0
For reaction system: 5 out of 12
At this time: 0.5SEC
Conversion of Mfmt is:1.0
For reaction system: 6 out of 12
At this time: 0.5SEC
Conversion of Mfmt is:1.0
For reaction system: 7 out of 12
At this time: 0.5SEC
Conversion of Mfmt is:1.0
For reaction system: 8 out of 12
At this time: 0.5SEC
Conversion of Mfmt is:1.0
For reaction system: 9 out of 12
At this time: 0.5SEC
Conversion of Mfmt is:1.0
For reaction system: 10 out of 12
At this time: 0.5SEC
Conversion of Mfmt is:1.0
For reaction system: 11 out of 12
At this time: 0.5SEC
Conversion of Mfmt is:1.0
For reaction system: 12 out of 12
At this time: 0.5SEC
Conversion of Mfmt is:1.0
Running Time is: 1120.3586 minutes.
The model core has 7825 reactions and 108 species.
The model edge has 10269 reactions and 9621 species.
After garbage collection:
Memory used: 2147549184
Free memory: 1594369416
Exceeded largest permitted flux for convergence (tolerance=0.001): 1.9017714656526474E-21
PDep Pruning DEBUG:
The number of species in the model's edge, after pruning: 9621

Enlarging reaction model
Time: 0.5SEC
Rmin: 1.9017714656526474E-21
Unreacted species CO2 has highest flux: 2.187000453270849E-21
Network 8728 has highest leak flux: 2.48553716355881E-21
rwest commented 14 years ago 
#  This java code was compiled by ant at:                            #
#    2010-10-16T13:32:52
#  The git repository was on the branch:                             #
#    master
#  And at the commit with the hash:                                  #
#    c9baf9f0f6c538d67acadad95f27c434cfe2ada6
#                                                                    #
#  For details visit:                                                #
#    http://github.com/GreenGroup/RMG-Java/tree/c9baf9         #
#  To see changes since then visit:                              #
#    http://github.com/GreenGroup/RMG-Java/compare/c9baf9...master
gmagoon commented 14 years ago

OK, I added a debugging line to test the hypothesis that this is due to exceeding the threshold on the final ODE step. If it does exceed on the final step, it should still print the message.

Try rerunning with the following version: http://github.com/GreenGroup/RMG-Java/commit/0ea230862087cae757ff1ced596db0e56333ba85

gmagoon commented 14 years ago

RHW reran and results suggested multiple instances of a discrepancy between dasslAUTO kleak and Java kleak, with the dasslAUTO kleak being on the order of 10% lower. It may be a moot point with the recent changes on the PDepRMEmod branch. We will reassess with the recent changes.

gmagoon commented 14 years ago

See Issue #41. I think the fix for that issue should address this issue.

gmagoon commented 13 years ago

This should be fixed now...see comment above.

gmagoon commented 13 years ago

Correction: I've reopened as I don't think the issue has been addressed on the master branch; this should be fixed on PDepRMEMod branch ; we can close once issue #41 is addressed (i.e. we merge PDepRMEMods into master)

gmagoon commented 13 years ago

closed with closure of #41 cf. https://github.com/GreenGroup/RMG-Java/commit/61872e87522a9d8ae703cf1d19846b2b76d31317