Professional Documents
Culture Documents
Ole Nick
Ole Nick
C
B
presumably this detector would alarm first to the fire. According to the new changes to
the code in NFPA 72 (2007 edition), a detector spaced according to smooth ceiling
spacing could be located sufficiently far from the fire to be in the second beam pocket,
shown as Detector B in Figure 1. A detector on the bottom of the first beam (Detector C)
could also satisfy the new requirements. However, the Detector B location is the worstcase scenario, because it is farthest from the fire and is located behind the barrier created
by the beamed ceiling. For comparison, a model was also run with the same inputs, but
with a smooth ceiling configuration (i.e., the beams no longer present). It should be
noted that the second pocket, where detector B was located, was only modeled as a
partial pocket in the previous analysis in some of the scenarios depending on the spacing
of the pockets. As can be seen in Figure 1, the partial second beam pocket extends to the
edges of the computational domain and therefore is open on two of the four sides,
allowing smoke to simply exit the domain with no buildup.
In the previous work, the Geiman and Gottuk optical density thresholds [4] were
used as the primary means to determine when the smoke detectors activated. A
secondary means of determining time to detector activation, a critical velocity along with
a temperature correlation as surrogate conditions to confirm activation, was also used.
The critical velocity used was 0.13 m/s +- 0.07 m/s, and the temperature correlation used
was a 4C temperature rise for ionization detectors and a 13C temperature rise for
photoelectric detectors. In the initial analysis, it was established that a 60 second
difference between activation of the detectors with beams present when compared to a
smooth ceiling as the threshold of a significant activation time difference. No
justification for the use of a 60 second threshold was provided in the report.
The initial work examined smoke detector activation times for both a beamed and
a smooth ceiling, based on the maximum spacing of detectors allowed for a smooth
ceiling configuration (21). According to the report, the detector activation times for the
case with beams and the case without beams were within 60 seconds of each other.
Based on this activation time difference of less than 60 seconds, it was concluded that the
presence of ceiling beams made no significant difference in detector activation time.
However, as reported in the report, the detector placed in the second beam pocket
according to the smooth ceiling spacing for a 12 on center beam spacing and 2 deep
beams (Detector B) never reached the alarm threshold. The report explained the failure
of the detector to alarm by the lack of a reservoir effect. According to the report, the
reservoir effect occurs when a beam pocket causes the smoke to accumulate within that
pocket before spilling over into the next pocket. As can be seen in Figure 1, if the
smooth ceiling spacing is used, the detector nearest to the fire becomes Detector B, which
is in the second beam pocket. Since in the model for this scenario the second pocket was
only a partial pocket, smoke flowed out of the domain instead of accumulating as it
would if the pocket were complete on all sides. According to the initial report, the
detector in this pocket, Detector B, failed to alarm because the model domain was not
extended sufficiently to create a completely enclosed pocket around Detector B, not
because the smooth ceiling detector spacing was inappropriate.
Based on the modeling, it was determined that the ceiling beams will not
significantly impact the time to smoke detector activation if the flat ceiling detector
spacing is used. The work was presented to the NFPA 72 committee responsible for this
section of the code dealing with the placement of smoke detectors, and the work was
accepted, resulting in the change in NFPA 72, 2007 edition. The reasoning for the
change in the code is that while the beams may block some of the ceiling jet, the pockets
will allow for greater temperature rise and buildup of smoke, and create greater smoke
velocity due to the smoke spilling out of each pocket into adjacent pockets. The trapping
of smoke and heat in the pockets was termed the reservoir effect.
Comparison Models
In order to evaluate the present day accuracy and reproducibility of the initial
work, the authors of this paper performed a series of FDS modeling runs using the new,
validated, Smoke Detector Activation Algorithm (SDAA), as well as the Geiman and
Gottuk and temperature correlations. The ultimate goal of this work was to determine
independently, based on modeling and analysis, whether the conclusions of the initial
work are justified. For this analysis, a geometry where the ceiling was 12 high and the
beams were 1 or 2 deep and spaced 12 on center was used. According to the new
changes in NFPA 72, these geometries would qualify for smooth ceiling spacing of
smoke detectors. Smoke detectors were included in the first beam pocket (Detector A)
according to the previous NFPA 72 requirements, and then included in the second beam
pocket according to smooth ceiling spacing (Detector B) allowed under the current NFPA
72 code. Smoke detectors utilizing realistic lag time coefficients, termed Cleary I1,
Cleary I2, and Heskestad L=1.8m detectors, were implemented in this model because
they represent real detectors that would be expected to be found installed in buildings,
based on testing done by Cleary [3] and by Heskestad [6] .
Two different heat release rates were examined: a fixed 100 kW heat release rate
and a variable heat release rate based on a medium t-squared fire. According to the initial
report, this 100 kW fire should be detectable for a 12 ceiling, with or without beams. In
both cases, the fuel was Methyl Methacrylate (MMA), a fuel that has a relatively large
soot yield of 0.022 kg/kg fuel as defined in the FDS material database, when compared to
other ordinary combustibles such as wood, paper, or cloth. This high soot yield will lead
to earlier detector activations than would be observed with other ordinary combustibles.
Therefore, the use of this high soot yield fuel will tend to reduce the time to detector
activations in both the smooth and beamed ceiling cases.
In order to test the observation in the initial report that the partial pocket was
responsible for the detector failing to alarm in the second pocket (Detector B) for this
particular scenario, for the current simulations, the domain was extended to ensure that
this pocket was fully enclosed, allowing for the so-called reservoir effect to occur. A
depiction of the extended domain as used in this modeling is shown as Figure 2.
C
B
Figure 2. 12 ceiling, 2 beam depth, 12 beam spacing geometry. Note that the
furthest pocket is surrounded on all sides, allowing buildup and the reservoir
effect.
For this analysis, simulations were run with the 12 ceiling and 12 beam spacing.
Two different iterations were performed, one with the beams 2 deep, and the other with
the beams 1 deep. The 2 beam depth is larger than the estimated ceiling jet depth (10%
of 12 or 1.2 ft.), while the 1 beam depth should not be sufficient to fully block the
ceiling jet, based on its estimated depth.
Model Results
The first significant result from the current modeling is that when the second
beam pocket is completed and the 100 kW heat release rate is used for this particular
scenario, the detector in the second pocket never reaches the alarm threshold. This result
is in stark contrast with the assertion in the initial report, which states that this 100 kW
fire should result in alarm of the detector in the second beam pocket if the complete
pocket is modeled. This result can be clearly seen in Table 1 where detection times for
the smooth ceiling spacing are given for three cases: no beams, 1 beams, and 2 beams.
Detector
Cleary 1 Activation
Time (s)
17
Cleary 2 Activation
Time (s)
17
Heskestad L=1.8m
activation time (s)
16
Baseline 15 spacing
1 ft beams Detector
9
8
6
(A)
1 ft beams Detector
> 600
> 600
> 600
(B)
2 ft beams Detector
8
8
7
(A)
2 ft beams Detector
> 600
> 600
> 600
(B)
Table 1. 100 kW fire, 12 ceiling, 12 beam spacing results, 4%/ft sensitive
detectors.
As can be seen from the results, the smoke detectors at an x and y distance of 15
from the plume centerline in the baseline smooth ceiling case activated in 16-17 seconds
after ignition. However, in both the 1 beam case and the 2 beam case, the detectors in
the second beam pocket (Detector B) at the same distance from the fire as the smooth
ceiling case failed to alarm within 600 seconds. Thus, even when a complete second
pocket geometry is included to create the reservoir effect, the detector at the 15
spacing does not activate when either 1 or 2 beams are present.
These results contradict the findings of the initial report, since the current
modeling shows that the beams result in a substantially longer time to detector alarm than
the baseline smooth ceiling case. The differences shown in Table 1 between the smooth
ceiling case and the beamed ceiling case, when the smooth ceiling spacing is used as
allowed in the revised version of NFPA 72 (2007 edition), are clearly outside the 60
seconds criterion that the initial work deemed to be significant. Therefore, the smooth
ceiling spacing does not provide adequate safety for beamed ceilings in this
configuration. However, as can be seen from Table 1, if a detector were included in
every beam pocket (Detector A), as previously required by NFPA 72 (2002 edition),
these detectors would alarm earlier than the detectors placed according to the smooth
ceiling spacing.
To further investigate the analysis of the initial work, the temperature threshold
that was used as a surrogate for smoke detector activation was examined through the
current models. Figure 3 shows plots of temperature rise vs. time for the three cases
investigated with the 15 detector spacing. As can be seen in the figure, the top curve is
for the baseline case of a smooth ceiling, the middle curve shows the temperature rise vs.
time for the 1 beam case, and the bottom curve gives the temperature rise history for the
2 beam case. Also shown in the figure are the lower activation bound of 4C
temperature rise and the upper activation bound of 13C temperature rise that was used
for the temperature correlation. When the change in temperature at the smoke detector
location reaches these thresholds, activation by smoke particles is assumed to have
occurred.
20
Baseline - on ceiling
1ftbeams - on ceiling
2ftbeams - on ceiling
Upper activation bound
Lower activation bound
18
16
14
12
10
8
6
4
2
0
0
100
200
300
400
500
600
Time (s)
0.16
Baseline - on ceiling
1ftbeams - on ceiling
2ftbeams - on ceiling
Upper Ionization Bound
Lower Ionization Bound
Upper Photoelectric Bound
Lower Photoelectric Bound
0.14
0.12
0.10
0.08
0.06
0.04
0.02
0.00
0
100
200
300
400
500
600
Time (s)
baseline case, along with the velocity. Figure 6 shows a similar plot for the case with 2
deep beams. All of the results above are for detectors that activate at 13%/m (4%/ft.)
obscuration, which is an allowable threshold per UL 217.
No beams (baseline) 15'
0.80
30.0
0.70
25.0
20.0
0.50
0.40
15.0
Activation Threshold
0.30
Baseline Exterior Smoke - 15'
Baseline Heskestad 1.8m L
Baseline Cleary I1
Baseline Cleary I2
Baseline Exterior Velocity
10.0
5.0
Velocity (m/s)
0.60
0.20
0.10
0.0
0.00
0
10
20
30
40
50
60
Time (s)
30.0
2-7a Exterior Smoke - in pocket 15'
2-7a In pocket Heskestad 1.8m L
2-7a In pocket Cleary I1
2-7a In pocket Cleary I2
2-7a Exterior Velocity
0.70
0.60
20.0
0.50
0.40
15.0
Activation Threshold
0.30
Velocity (m/s)
25.0
10.0
0.20
5.0
0.10
0.0
0
50
100
150
200
250
300
350
400
450
0.00
500
Time (s)
Figure 6. 100 kW 12 ceiling, 2 beam depth results at the 15 spacing detector, both
inside and outside the detectors.
As can be seen from Figures 5 and 6 above, the smoke obscuration at the smooth
spacing ceiling detector (Detector B) in the baseline case reaches adequate smoke
obscurations both inside and outside the detector. Conversely, in the 2 beams case, the
smoke external and internal to Detector B does not reach adequate thresholds (13%/m or
4%/ft). This lack of adequate smoke at detector location B in the beamed ceiling cases is
likely because the beams act to divert the flow of smoke away from detector location B to
other adjacent pockets, so there is never a chance for these detectors to alarm to this fire.
As discussed previously, a fire that ramps up instantaneously to 100 kW but then
does not continue to grow may not provide the appropriate test for detector spacing.
Therefore, the above analysis was repeated with a realistic medium t-squared fire. For
these t-squared fire models, every other variable was left unchanged from the 100 kW
case, except for the fire size and growth rate. Similar to the 100 kW cases, Table 2 shows
the activation time results using the SDAA for the medium t-squared growth fire.
Figures 7 and 8 show the activation times from the temperature correlation and Geiman
and Gottuks optical density method, respectively, using the smooth ceiling detector
spacing location for both the baseline ceiling and the beamed ceilings.
Cleary 1 Activation
Time (s)
71
Detector
Cleary 2 Activation
Time (s)
73
Heskestad L=1.8m
activation time (s)
75
49
43
131
126
48
58
165
166
Baseline 15 spacing
1 ft beams Detector
43
(A)
1 ft beams Detector
131
(B)
2 ft beams Detector
48
(A)
2 ft beams Detector
165
(B)
Table 2. 12 ceiling, 12 beam spacing results for
sensitive detectors.
20
Baseline - on ceiling
1ftbeams - on ceiling
2ftbeams - on ceiling
Upper activation bound
Lower activation bound
18
16
14
12
10
8
6
4
2
0
0
100
200
300
400
500
600
Time (s)
0.20
Baseline - on ceiling
1ftbeams - on ceiling
2ftbeams - on ceiling
Upper Ionization Bound
Lower Ionization Bound
Upper Photoelectric Bound
Lower Photoelectric Bound
0.18
0.16
0.14
0.12
0.10
0.08
0.06
0.04
0.02
0.00
0
100
200
300
400
500
600
Time (s)
200 kW. Finally, if 2 beams are included, the fire is not detected until over 90 seconds
after the smooth ceiling case, and the fire size is over 5 times larger than if no beams
were present, at over 300 kW.
As a comparison, in Table 4 below, the activation times if a detector is placed in
every beam pocket, as was previously required by NFPA 72, are displayed.
Scenario
For a more realistic fire scenario such as a medium t-squared fire, the authors
determined that a detector in every pocket will detect the fire in less time than the smooth
ceiling configuration. Furthermore, if smooth ceiling spacing is used and the beams are
in place, the fire at the time of detection will be at least (depending on single- or multiblock) 3 times larger if the beams are 1 in depth and will be more than 5 times larger if
2 beams are in place. The fire in the 1 beam case would be ~200 kW and in the 2
beam case would be ~300 kW, both of which are far larger than the 100 kW fire that was
assumed should be detectable. As a result of this peer-review modeling and analysis
utilizing the new SDAA in FDS, the new requirements in NFPA 72 (2007) that allow
smooth ceiling spacing for some beamed ceilings should be reevaluated.
References
1. Beyler, C. and DiNenno, P., Letters to the Editor, Fire Technology, v. 27, n. 2,
1991.
2. Cholin, J.M. and Marrion, C., Performance Metrics for Fire Detection, Fire
Protection Engineering, n.11, 2001.
3. Cleary, T., Chernovsky, A., Grosshandler, W., and Anderson, M., Particulate Entry
Lag in Spot-Type Smoke Detectors, Fire Safety Science - Proceedings of the Sixth
International Symposium, 1999.
4. Geiman, J.A. and Gottuk, D.T, Alarm Thresholds for Smoke Detector Modeling,
Proceedings of the Seventh International Symposium on Fire Safety Science, 2003.
5. Gottuk, D.T., Hill, S.A., Schemel, C.F., Strehlen, B.D., Rose-Phersson, S.L., Shaffer,
R.E.., Tatem, P.A. and Williams, F.W., Identification of Fire Signatures for
Shipboard Multi-criteria Fire Detection Systems, Naval Research Laboratory
Memorandum Report 6180-99-8386, 1999.
6. Heskestad, G., Generalized Characterization of Smoke Entry and Response for
Products of Combustion Detectors, Proceedings of the Fire Detection for Life Safety
Symposium, 1975.
7. Luck, H. and Sievert, U., Does an Over-All Modeling Make Any Sense in
Automatic Fire Detection? AUBE 99 Proceedings of the 11th International
Conference on Automatic Fire Detection, 1999.
8. McGrattan, K. B. and G. P. Forney. Fire Dynamics Simulator (Version 4), Users
Guide. NIST Special Publication 1019, National Institute of Standards and
Technology September, 2005
9. Mowrer, F.W. and Friedman, J., Experimental Investigation of Heat and Smoke
Detector Response, Proceedings of the Fire Suppression and Detection Research
Application Symposium, 1998.
10. OConnor, D.J., Cui, E., Klaus, M. J., Lee, C.H., Su, C., Sun, Z., He, M., Jiang, Y.,
Vythoulkas, J., Al-Farra, T. Smoke Detector Performance for Level Ceilings with
Deep Beams and Deep Beam Pocket Configurations Research Project. NFPA Fire
Protection Research Foundation Report, April, 2006.
11. Roby, R.J., Olenick, S.M., Zhang, W., Carpenter, D.J., Klassen, M.S., and Torero,
J.L. Smoke Detector Algorithm for Large Eddy Simulation Modeling. NIST GCR
07-911, 2007.
12. Schifiliti, R.P. and Pucci, W.E., Fire Detection Modeling State of the Art, Fire
Detection Institute, 1996.
13. Schifiliti, R.P., Fire Detection Modeling The ResearchApplication Gap, AUBE
01 Proceedings of the 12th International Conference on Automatic Fire Detection,
2001.
14. Wakelin, A.J., An Investigation of Correlations for Multi-Signal Fire Detectors,
Masters Thesis, Worcester Polytechnic Institute, Department of Fire Protection
Engineering, 1997.