FDS v6.7.1 calculates spurious HRR which is above expected for ventilation limited scenario

[DavidB]

I have come across an issue in which FDS (v6.7.1) predicts a higher global HRR than the input mass loss rate would indicate.

Background: The scenario is a ventilation limited compartment (fans on supply and extract).  Fire is in a vertically stacked set of cable trays. As the fire progresses up through the trays (all forced based on clock timers, fixed spread rates and fixed HRRPUA) the MLR increases and, due to ventilation, some burning does occur away from the fuel, predominantly in the fresh air stream that is entering the compartment via the HVAC vent.

This is all expected and is consistent with the experimental data, but for some reason when the width of the cable trays is altered on one of the sensitivities (I have looked at 3 different tray widths) I see the HRR diverge from the expected behaviour and go higher than the combination of MLR and HoC would indicate is OK.

So at the moment I have two scenarios that seem to give physical / credible results (in line with MLR and HoC data) and one scenario which diverges from the expected and has an excess of HRR. The only difference between the input decks of these scenarios is the width of the tray (obstruction and associated vent) and the HRRPUA of the vent (which I have adjusted to ensure total MLR is not altered in the sensitivity). The MLR values in the output files are almost identical (except for minor differences due to the spread pattern along a narrow vs wider cable tray), I therefore consider the changes to the input decks to be OK.

The mesh resolution is unchanged for all cases, so for the spurious/problematic result the tray width (and vent) is only 4 cell faces wide (the others that seem credible are 5 or 6 cells wide). D*/dx is around 4 at room initial conditions so not a great resolution, can a coarse vent resolution be linked to unphysical excess HRR.

Is there somewhere I should be looking to understand why FDS has overpredicted the HRR? Is there a known issue with this version that I maybe missed which would explain this? 

[dr_jfloyd]

Is the integral over time of the HRR exceeding the integral over time of MLR*HOC, or was the HRR integral lower than MLR*HOC integral for a period of time and now higher than MLR*HOC? If the HRR integral is always less than or equal tl the MLR*HOC integral (accounting for numerical error of course when using the hrr.csv file for integration which likely does not have every time step) then FDS isn't releasing to much HRR and changing the tray width has just impact the flow field in some unanticipated way. If you are getting an HRR integral that is noticeably larger than MLR*HOC, I suggest you trry the current release. 6.7.1 is over four years old at this point.

[DavidB]

Thank you for getting back so promptly.

Yes, the integral HRR (hrr.csv based data) for the spurious result is 175% of the integral of MLR*HOC, i.e. not credible. This compares to an integral percentage 96% for one of the credible scenarios. 

I accept that 6.7.1 is old(er) but i was attempting to maintain solver version consistency with the other simulations that were performed on a study from that time. 

For future reference, is there anything to look for that would indicate that the solver has failed/struggled. I searched for errors in the .out file, there were no obvious differences between the credible and spurious result but i may not be looking for the right keywords.  

Thanks again!

[Kevin]

In the _hrr.csv file, there is a column indicating the mass loss rate of fuel at the burner surface or inlet vent. Multiply by the heat of combustion and see what you get. 

[DavidB]

Thanks Kevin. I think what you are suggesting is what i have done. I have integrated the HRR to get the actual energy release and i have separately integrated the MLR*HoC to get the "input" energy release. I have then divided one integral by the other to get the percentages in my message above, i.e. spurious result "output" is 175 % of MLR*HoC "input"

[DavidB]

Here are the results, to give some context. Y-axis is output HRR. The grey line is the spurious result, the other two lines differ only in the tray (fuel vent) width. It appears (at first) to be in line with the other sensitivities, but then stays high and seems unphysical. For reference all MLR vs time curves are broadly overlapping.

If there is something that I should be looking out for to indicate the model is struggling that would be helpful, but other than that I will take note of dr_jfloyd's  advice to use the latest version. 

 

Thanks!

[Kevin McGrattan]

Also, it might be time to create an issue using a simplified version of the case. This is when the car mechanic tells you that you have to bring the car to the garage.

[jacques Frezabeu]

Hello

I had a similar issue on the 6.6.0 that i solved with the use of glmat pressure solver.

Le jeu. 18 mai 2023, 10:57 PM, Kevin McGrattan <mcgr...@gmail.com> a écrit :

Also, it might be time to create an issue using a simplified version of the case. This is when the car mechanic tells you that you have to bring the car to the garage.

--
You received this message because you are subscribed to the Google Groups "FDS and Smokeview Discussions" group.
To unsubscribe from this group and stop receiving emails from it, send an email to fds-smv+u...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/fds-smv/CAAJimDGKq3%2B4VLa7gaaGU1XTRvWWHLYbrJQ-nNA%3DwQ1cc9CVfA%40mail.gmail.com.

[DavidB]

Thanks both.

Jacques, i will take a look at changing the solver. Did you find anything in the .out file that indicated a problem or was it the spurious result (HRR) that made you change solver?

Kevin, i can probably share the problematic model if it is helpful, but i will perhaps try the glmat solver first and get back once this has run.