fractional effective dose (FED) shows different results in smoke view

[ThijsH24]

I have simulated a tunnel fire of 100MW for 500 seconds, with constant inflow of fresh air (Vin = 1.8 m/s). After 500 seconds this is what the FED-slice in smokeview looks like: Which is extremely low, considering there is a lot of smoke inside the tunnel. When I reproduce the FED calculations myself by using the CO,O2 and CO2 concentrations I get the following result after 500 seconds. FED.pdf Which looks more realistic (goes up way higher than 8 in the fire). In the technical reference guide I see that the units of CO needs to be in ppm while the others are in percentages. So I used that in my calculations. What goes wrong?

calculations all the concentrations are in volume fractions (one concentration value each time step). integral_CO = 2.764*10-5(CO1000000)1.036 #ppm FED_CO = np.trapz(integral_CO,time,axis=0)

integral_O2 = 1/(np.exp(8.13-0.54(20.9-O2100))) #% FED_O2 = np.trapz(integral_O2,time,axis=0)

HV_CO2 = np.exp(0.1903CO2[-1]100+2.0004)/7.1 #%

FED_tot = FED_CO * HV_CO2 + FED_O2 FDS_file: V_1_8.txt

[gforney]

I'll take a look.

rerun your case adding devices for FED as in &DEVC ID='F-00', XYZ=0.00,8.0,2.0, QUANTITY='FED' /

put devices in the region where you are doing your computations and where the slice is displayed

[ThijsH24]

The device data shows similar results. So probably it will have to do something with my calculations.

[ThijsH24]

Is the CO-concentration not in percent? When I use percent instead of ppm, the results correspond better with the DEVC data.

[gforney]

smokeview (and fds) performs an integration to compute the fed values. smokeview uses co, co2 and o2 values in slice files. It looked like to me that the time step integration time interval) you defined for slice file output was 2 s. I suggest you decrease it to say 0.1 s.

the fed computation has been removed in recent versions of smokeview and moved to a standalone program named fds2fed. try installing the nightly bundle at link below. you don't need to rerun your case (except if you reduce the slice time interval) . just type fds2fed casename where casename is the name of your case then view your case in smokeview

https://github.com/firemodels/test_bundles/releases/tag/FDS_TEST

[ThijsH24]

Thanks a lot for your reply. What would be the difference if I reduce my timestep? The integration is over the entire time of the simulation right? Not for each timestep.

[gforney]

Intgral from a to b of f(t) dt can be approximated by splitting the interval a to b into N parts . Compute f(t_i) in each part and sum f(t_i)*deltat where deltat=(b-a)/N and t_i=a+(i+0.5)deltat for i from 0 to N-1. Large N ( small deltat) will approximate the integral better than small N. The fed is an integral which both fds and smokeview approximates similar to a procedure like this. I was just guessing that a reason for poor fed values in your earlier post was that your N was too small ( equivalently deltat=2.0 was too big)

On Tue, Nov 12, 2024, 3:33 AM ThijsH24 @.***> wrote:

Thanks a lot for your reply. What would be the difference if I reduce my timestep? The integration is over the entire time of the simulation right? Not for each timestep.

— Reply to this email directly, view it on GitHub https://github.com/firemodels/fds/issues/13724#issuecomment-2469895270, or unsubscribe https://github.com/notifications/unsubscribe-auth/AC6UCRR5E55RLIDKLHNGHUD2AG4NRAVCNFSM6AAAAABRNWY75SVHI2DSMVQWIX3LMV43OSLTON2WKQ3PNVWWK3TUHMZDINRZHA4TKMRXGA . You are receiving this because you were assigned.Message ID: @.***>

[ThijsH24]

Yes that makes sense. It would definitely make it more accurate. Two remaining questions: What formulas/units do I need to use to calculate the FED myself? How do I install the nightly bundle?