Investigation of Hold Time Calculation Methodologies for Total
Flooding Clean Extinguishing Agents
Todd M. Hetrick, Paul E. Rivers PE
3M Fire Protection
St. Paul, MN 55144-1000 USA
phone: 651-737-5463, fax: 651-732-8005, e-mail: [email protected]
phone: 651-733-0029, fax: 651-732-8005, e-mail: [email protected]
Abstract
Total flooding fire suppression systems are designed and installed in accordance with widely
published standards. Annex C of NFPA 2001 and Annex E of ISO 14520 are the principal
guides to verify an enclosure’s integrity. The objective of the test procedures contained
therein is to obtain a measure of the equivalent leakage area, which, when coupled with ‘worst
case’ assumptions, may be used to determine the expected minimum hold time. Previous
studies have shown that the established calculation methods result in hold times that vary
appreciably from that predicted using either standard.
This paper evaluates existing hold time calculation methodologies in comparison to newly
generated bodies of both full and small-scale test data. Testing included commercially
available agents selected to span a wide range of properties including HFC-227ea, HFC-125,
FK-5-1-12, and IG-541. All other system variables were maintained at nearly constant
conditions with the exception of the enclosure’s equivalent leakage area, which is not always
assumed to have equal distribution between upper and lower leaks.
Introduction
Total flooding fire suppression involves the discharge of a clean extinguishing agent that is
typically required to provide protection within the design envelope for a minimum ten
minute period. The ‘Fan Integrity Test’ encompasses the test method and leakage modeling
used to evaluate the total flooding system design against the ‘hold time’ or ‘retention time’
requirement. The hold time may be specified as the period of time required for the clean
agent concentration to drop to a specified threshold (usually 80% of the initial discharge
concentration) at a specified height in the enclosure (often chosen as the point of highest
combustibles or at 75% of the maximum enclosure height) [1]. The Authority Having
Jurisdiction (AHJ) may set the hold time requirement as necessary accounting for both the
cooling of potential ignition hazards and emergency response.
This report documents recent full scale research completed under the auspices of the NFPA
2001 Technical Committee on Gaseous Fire Extinguishing Systems. The second of two
research goals is addressed in this paper; to validate industry-standard hold time prediction
models as they apply to a variety of clean extinguishing agents. A 103 cubic meter
experimental enclosure was used to observe leakage flows through enclosure boundaries.
The upper and lower leakage areas were varied to determine their effect on the hold times
for FK-5-1-12, HFC-125, HFC-227ea and IG-541. Previous studies evaluating the observed
agent leakage after discharge have shown that model predictions are often very inaccurate;
resulting in both overly conservative and optimistic hold time approximations [2, 3, 4].
NFPA 2001 Annex C and ISO 14520.1 Annex E contain enclosure integrity design
standards, which have been chosen for comparative analysis in this paper due to the
prevalent adoption and use around the world. Changes to the 2004 version of NFPA 2001
are currently slated for release later in 2007. The analysis included herein utilizes the 2004
version because modifications to the standard nominally result in the same hold time
predictions provided by the 2006 ISO design standard.
Background & Theoretical Considerations
Following the discharge of a total flooding fire suppression system, a relatively uniform
mixture of clean agent and air will remain inside the design envelope. In comparison to that
of atmospheric air, the agent and air mixture density, especially for halocarbons, will be
greater inside the enclosure than that of the air surrounding it. This density disparity will
exert a positive hydrostatic pressure on the lower enclosure boundaries, forcing the agent-air
mixture through any available leakages. This leakage will create a negative interior-toexterior pressure differential at upper enclosure boundaries. Because the enclosure is a fixed
volume, an equal balance of outside air will necessarily flow into the enclosure’s upper
boundaries as agent leaks out lower boundaries.
Based upon the hydrostatic pressure profile inside the enclosure relative to that surrounding
it the gas flow through leakages may be modeled according to well-established theory on
orifice flow. For all available hold time prediction models this aspect of the model remains
nominally the same. Because the hold time is evaluated as an agent concentration threshold
at a specified height the distribution of agent as a function of enclosure height must be
assumed in order to relate the total quantity of agent remaining to an elevation inside the
enclosure.
Generally, three hold time models are available within NFPA 2001 and ISO 14520. These
differ from one another depending upon the assumption of agent concentration distribution
assumed in each. Schematics of the resulting hydrostatic pressure profile for the sharp
interface, wide interface, and continuously-mixed agent distribution model assumptions are
shown in Figures 1, 2 and 3, respectively. In each, the assumed hydrostatic pressure profile
is shown where the neutral plane, Hnp, denotes that elevation where inside and outside
pressures are equal. The first two models employ a stratified agent distribution assumption.
The sharp interface model assumes that as the column of agent-air mixture leaks out
enclosure boundaries, a sharp interface forms between inflowing fresh air and the agent-air
mixture. The wide interface model assumes that incoming air mixes with the upper edge of
the column of agent-air mixture; resulting in a linear decay from full concentration at the
interface height, Hi, to a 0% concentration at the enclosure’s maximum height, H0. The
continuously mixing model assumes that the enclosure’s volume is homogeneous
throughout the hold time; resulting in infinitely fast mixing between inlet fresh air and the
resident agent-air mixture.
⎛ H + Hi
⎞
ΔP = ( ρ mix − ρ air ) g ⎜ 0
− H np ⎟
2
⎝
⎠
c=0
ΔP = ( ρ mix − ρ air ) g (H i − H np )
c=0
Q i (t )
.
ρ (H ) = ρ air + K
Q i (t )
.
ρ = ρ air
⎛ H − Hi
⎝ H 0 − Hi
(ρ mix − ρ air )⎜⎜
c = ci
⎞
⎟⎟
⎠
H i (t )
H0
c = ci
H0
ρ = ρ mix
H i (t )
ρ = ρ mix
H np (t )
H np (t )
Q o (t )
.
Q o (t )
.
ΔP = ( ρ mix − ρ air ) gH np
ΔP = ( ρ mix − ρ air ) gH np
Figure 1: Sharp Descending Interface
Hydrostatic Pressure Profile Schematic
Figure 2: Wide Descending Interface
Hydrostatic Pressure Profile Schematic
ΔP = ( ρ mix − ρ air ) g (H 0 − H np )
Q i (t )
.
H0
ρ mix = ρ CEA (c(t )) + K
K ρ air (1 − c(t ))
c(t )
H np (t )
Q o (t )
.
ΔP = ( ρ mix − ρ air ) gH np
Figure 3: Continuous Mixing Hydrostatic Pressure Profile Schematic
The continuous mixing model and sharp descending interface model are described elsewhere
[1, 2, 5, 6]. ISO 14520 replaced the sharp interface model in the 2006 publication with the
wide descending interface model, which is detailed in references 4 and 5. The new edition of
NFPA 2001 will include this modification as well.
The sharp descending interface model has been derived in a dimensionless form in
references 3 and 7. It is found that the rate of interface descent is governed by four
parameters; the height between inlet and outlet leakage paths, H 0, the ratio between the
enclosure floor area and the outlet leakage area, A c C o A o, the ratio between inlet and outlet
leakage areas, A , and the ratio between the agent-air mixture density relative to the density
of ambient air, ρ . The final dimensionless form presented in the following will be used for
comparison of tests involving differing agent densities and leakage configurations.
h(t ) ⎡ k 2 (t − t 0 )⎤
= 1−
⎥
τ
H 0 ⎢⎣
⎦
2
(1)
1
Where:
⎡ ⎛
1 ⎞ ⎤2
⎜1 − ⎟ ⎥
⎢
⎜
⎟
⎛C A
⎛ A ⎞⎛ H ⎞
τ = ⎜⎜ c ⎟⎟⎜⎜ 0 ⎟⎟ ; k = ⎢ ⎝ ρ ⎠ ⎥ ; A = ⎜⎜ o o
2
⎢ ⎛
⎝ C i Ai
⎝ C o Ao ⎠⎝ g ⎠
1 2 ⎞⎥
⎢ 2⎜1 + A ⎟ ⎥
⎜
⎟⎥
⎠⎦
⎣⎢ ⎝ ρ
1
2
⎞;
ρ
⎟⎟ ρ = mix
ρ air
⎠
Each model makes the simplifying assumption that leakage areas exist in two locations only;
at the extremes of enclosure height, upper and lower. The distribution of leakage area
between upper and lower locations may be referred to as the lower leakage fraction (LLF) and
quantified as Equation 2. ALL and AUL denote leakage areas for lower and upper leakages,
respectively.
F=
ALL
ALL + AUL
(2)
The combination of the LLF with an evaluation of the total enclosure leakage area includes
all that is necessary to define the leakage configuration in the hold time model. These two
model input variables may be used to quantify the total amount of clean agent in the
enclosure envelope as a function of time. When coupled with an assumption of the agent
distribution profile (as a function of height) the hold time may be evaluated as the time at
which a specified agent concentration exists at a designated height.
One method of evaluating the total amount of open leakage in enclosure boundaries is to
perform a fan integrity test. The test procedure consists of a calibrated fan that infuses or
removes air from the enclosure at a known flow rate. The corresponding increase or
decrease in enclosure pressure relative to atmospheric is then measured for a series of fan
flow rates. The pairs of values are used in a regression to fit Equation 3; and thus, to find
the unknown values of the total leakage area, AT, and the flow exponent, n. In the absence
of multipoint testing such as that prescribed by the 2004 publication of NFPA 2001 the total
leakage area may be found by setting the flow exponent equal to 0.5 and performing the fan
integrity test at a single flow rate. A value of 0.5 indicates turbulent flow and a value of 1.0
would indicate laminar flow [1]. When solved for, the flow exponent, n, represents a leakage
area-weighted average value, which approximately describes the total enclosure leakage flow.
⎛ 2ΔP ⎞
⎟⎟
Q = C ⎜⎜
⎝ ρ ⎠
n
(3)
Q is the fan flow rate, ∆P is the pressure differential from atmospheric, ρ is the density of air
and n is the flow exponent [8]. The constant C is equal to ATKdKU where Kd is an orifice
discharge coefficient, AT is the total leakage area, and KU is a constant based on the value of
the flow exponent, n, and the units being used [3]. Typically, a value of 0.61 is used for Kd
which represents the ratio of the actual flow to the theoretical maximum flow through a
sharp-edged circular orifice.
T
Determination of the LLF is not a straight-forward task. Nonetheless, the selection of a
value for this model input variable requires prudence, as the output hold time is highly
sensitive to it. A variety of methods to ascertain this value are available including estimation,
manual inspection of enclosure leak locations, fan integrity testing with upper or lower
leakages temporarily sealed and fan integrity testing with either an above-ceiling or underfloor pressure neutralization technique. In the present study the hydrostatic pressure profile
was recorded at a series of heights inside the test enclosure. These measurements may be
correlated using the derivation that follows to observe the effective LLF throughout the hold
time.
Gas flow through enclosure leakage pathways is described by Equation 3. Upon
rearrangement, this may be solved for the leakage cross sectional area as shown in Equation
4. Assuming that upper and lower leakages are not equally sized, this equation may be used
to describe each leakage area individually, which is then substituted into the LLF, Equation
2. If it is assumed that the volumetric flow of gas through upper leakages is balanced by that
flowing through lower leaks (QLL = QUL) and the coefficients of discharge and unit
correction constants at each are assumed to be equal then Equation 2 may be modified to
result in Equation 5. Correlated empirical values of the LLF are plotted against time in the
Experimental Results section.
AT =
F=
⎛ ρ LL
⎜⎜
⎝ 2ΔPLL
Q LL
K d KU
QLL
K d KU
⎛ ρ LL
⎜⎜
⎝ 2ΔPLL
n
⎞
⎟⎟
⎠
⎞
QUL
⎟⎟ +
K d KU
⎠
Q
K d KU
n
⎛ ρ UL
⎜⎜
⎝ 2ΔPUL
⎞
⎟⎟
⎠
n
⇒
⎛ ρ ⎞
⎜
⎟
⎝ 2ΔP ⎠
n
⎛ ρ LL
⎜⎜
⎝ ΔPLL
⎛ ρ LL
⎜⎜
⎝ ΔPLL
n
(4)
⎞
⎟⎟
⎠
n
⎛ ρ
⎞
⎟⎟ + ⎜⎜ UL
⎠
⎝ ΔPUL
⎞
⎟⎟
⎠
n
⇒
(5)
1
⎛ ρ ΔP
1 + ⎜⎜ UL LL
⎝ ρ LL ΔPUL
⎞
⎟⎟
⎠
n
Experimental Apparatus & Instrumentation
All testing reported herein was conducted at the Fike Corporation test facility in Blue
Springs, Missouri, USA, in the same enclosure with no significant modifications made
between test sessions. A schematic of the experimental enclosure is found in Figure 4
(following page). Internal dimensions are 4.60 m. (15’-1”) square footprint by 4.88 m. (16’)
in height equaling 103 m3 (3640 ft3) in volume. Construction consists of 5.1 cm. by 20.3 cm.
(2”x8”) wood studs on 40.6 cm. (16”) centers with two interior layers of 15.9 mm. (5/8”)
plywood and one layer of fiberglass sheeting as an interior finish. Intentional leakage area
was supplied in two forms; (1) 84 one inch diameter holes drilled about the upper and lower
enclosure boundaries and (2) a ceiling vent for discharge pressure venting of inert agents 1 .
All drill holes are offset from lower and upper boundaries by 30.5 cm. (1’) and equally
distributed across each wall facing such that a nominal 10 upper and 10 lower holes exist per
wall. A floor drain is located in the room’s center that was closed by means of an existing
ball valve.
For each clean agent tested, a series of controlled leakage area configurations were simulated
by plugging and/or unplugging drill holes. Dense rubber stoppers were used to plug holes
from the inside where they made a reliable seal with the fiberglass sheeting. Each specified
leakage configuration was accomplished in such a way as to produce a symmetrical leakage
pattern. For example, an experiment with 16 open drill holes would be accomplished by
opening a single hole at 1/3 and 2/3 of the wall width on each wall, upper and lower.
1
Experiments 2, 3 and 4 from the IG-541 test set involve relatively ‘tight’ leakage configurations. For these
tests only, there was concern over the potential for over pressurization. Tests 2 & 3 left the ceiling vent
open throughout the discharge and retention time. In Test 4 the ceiling vent was closed within a few
seconds after the clean agent discharge ended. For all other tests this vent remained sealed closed.
1’
5’-8”
Dwyer
Magnehelic®
Indicating
Pressure
Transducers
16’
Retrotec DM-2
Pressure
Transducers
6’-3”
5’-8”
15
1’
”
’-1
15’-1”
Figure 4: Schematic of the Test Enclosure
The test facility in which the experimental
chamber is located has a footprint greater
than five times that of the discharge
enclosure. This helps ensure that the exiting
flow of clean agent vapor will not pool and
serve to equalize or diminish the column
pressure driving flow out lower leakages.
Additionally, the surrounding enclosure
minimized pressure gradients produced by
external wind flow affecting the test
chamber. Ambient pressure probes do
reflect the brief openings of facility doors
throughout the hold time, but do not persist
long enough to significantly affect clean
agent retention times. No HVAC ducting
traverses through the test chamber;
eliminating the potential for bias pressures.
A controlled introduction of bias pressure is
not investigated in this study.
Environmental variables were otherwise
uncontrollable. All testing was completed
within August and September, 2006.
Ambient temperatures and pressures did not
vary drastically, as this period of time is
fairly temperate where Fike’s facility is
located. The one variable of concern
however was the relative humidity. Peak
pressures during halocarbon clean agent
discharges have been shown to have a
strong inverse relationship with the ambient
relative humidity levels. Typical electronic
control rooms are maintained at
approximately 55% RH [9].
For all
halocarbon testing completed the enclosure
was purged with dry compressed air until
the relative humidity level dropped to less
than 40%. Clean agent system cylinders,
valves assemblies, pipe networks, discharge
nozzles and other design parameters were
provided by each system manufacturer. The
manufacturers also specified the discharge
volume concentrations and cylinder
pressures used in testing. For each clean
agent, the selected discharge concentrations
represent the agents’ listed Class A design
concentrations 2 .
Table 1: Gas Sampling Probe Locations
Sampling Probe Elevation
Percent of Encl. Maximum Height
Positive Pressure
Vent (14” x 36”)
Clean Extinguishing Agent Tested
HFC-227ea FK-5-1-12 IG-5-4-1 HFC-125
70.0%
41.1%
41.1%
40.0%
75.0%
50.5%
50.5%
50.0%
80.0%
50.5%
59.9%
60.0%
53.6%
77.6%
65.0%
59.9%
87.0%
70.0%
63.0%
75.0%
70.3%
80.0%
77.6%
85.0%
77.6%
81.8%
87.0%
87.0%
Measured quantities include nozzle and
ambient pressures, gas species vapor
concentrations,
and
enclosure
air
Nozzle pressures are
temperatures 3 .
retained as a means to ensure proper agent
delivery and to diagnose potential problems
in system design. Ambient pressures are
recorded to document (1) the peak pressure
pulses generated during agent discharge and
(2) the hydrostatic pressure profile
2
3
Reference Table 2 for a listing of critical test
configuration parameters.
Nozzle pressures, Dwyer Magnehelic® ambient
pressure transducers, and enclosure ambient
temperature measurements typically recorded for
a 60 second duration during agent discharge.
Retrotec DM2 ambient pressure transducers were
used only for the HFC-125 test series.
throughout the hold time. Inert gas vapor concentrations are correlated indirectly from
measurements of enclosure oxygen content.
Halocarbon agent vapor concentrations are observed by means of a gaseous thermal
conductivity measurement technique. Enclosure air temperature measurements may be used
to further analyze the applicability of neglecting this variable in hold time predictions (as
prescribed by NFPA 2001 and ISO 14520 design standards). Gas sampling probes were
terminated at various elevations in the enclosure; bracketing the range of potential protected
heights. Table 1 lists the nondimensional elevations of the sampling probes used in each
experimental test set. All probes sampled air from a vertical axis offset 61 cm. (2’) to the
north of the central enclosure axis.
Experimental Results
Enclosure Ambient Pressure (Pa)
1000
In the experimental phase, agent
concentrations and differential pressures
were the primary measurements.
Two
500
varieties of differential pressure transducers
were employed. Figure 5 displays the
0
composite output from a pair of Dwyer
Series 605 Magnehelic® transducers that was
HFC-227ea (20 Open 1" Dia. Holes)
-500
obtained during the period of agent
FK-5-1-12 (16 Open 1" Dia. Holes)
IG-541 (84 Open 1" Dia. Holes)
discharge 4 . Data traces are selected from
HFC-125 (16 Open 1" Dia. Holes)
-1000
tests with total controlled leakage areas as
0
10
20
30
40
50
60
Time (s)
similar as possible with exception of the IG541 data trace, which required an over- Figure 5: Ambient Pressure During Discharge
pressure relief vent for leakage scenarios tighter than that shown.
The Dwyer pressure transducers’ output did not indicate the presence of an elevated
hydrostatic pressure following the clean agent discharge. For this and other reasons,
multiple Retrotec DM-2 Series digital pressure gauges were obtained. Although used only
throughout the HFC-125 test series, these instruments obtained high precision
measurements at very low differential pressures. Figure 6 provides an example of the raw
data output acquired throughout the experimental hold time. These data were correlated by
use of equation 4 to experimentally determine the effective lower leakage fraction, F. This
proves to be a powerful tool in investigating the test enclosure’s time-varying leakage
condition. Figure 7 exhibits the results of this correlation 5 .
Because this study seeks to analyze the applicability of the theoretical hold time predictions
as directly as possible, it is necessary to experiment in a test enclosure that represents the
theoretical assumptions as exactly as possible. To this end, it should be noted that the test
enclosure, like all enclosures, has uncontrollable leakages distributed randomly about the
enclosure boundaries and not only at the polar extremes of the experimental envelope’s
elevation. Thus, although the controlled leakage split may be equally balanced between the
4
5
The data traces are a composite of a 2500 Pa. positive pressure and a 1500 Pa. negative pressure transducer.
Figures 6, 7 and 8 are each based upon the data acquired in HFC-125 Test #3 (8/8 open upper/lower holes).
upper and the lower, the uncontrolled leakage may upset this balance and render the hold
time prediction input data slightly invalid. Table 2 shows the quantitative analysis used to
experimentally determine the uncontrolled leakage split.
20
1
Pressure (Pa)
15
DPT @ 1' Above Floor
DPT @ 5-2/3' Above Floor
DPT @ 11-1/3' Above Floor
DPT @ 15' Above Floor
Correlated Lower Leakage Fraction (%)
Retrotec
Retrotec
Retrotec
Retrotec
10
5
0
-5
0
1000
2000
3000
4000 5000
Time (s)
6000
7000
Figure 6: Hydrostatic Pressure Profile
Throughout Clean Agent Hold Time
8000
Direct Correlation of Measured Values
Equally-Weighted Moving Average
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
0
1000
2000
3000
4000 5000
Time (s)
6000
7000
8000
Figure 7: Correlated Effective Lower Leakage
Fraction Throughout Clean Agent Hold Time
Table 2: Empirically-Based Correlation of the Background Lower Leakage Fraction
HFC-125
Test #
1
2
3
4
5
Correlated LLF - Empirical
Average
Steady State Period
LLF
Begins
Ends
1000
1600
0.518
1000
2000
0.506
1000
3000
0.504
1000
3000
0.491
1000
3000
0.311
Background
ELA
59.8
59.8
59.8
59.8
59.8
Calculated LLF - Theoretical
Open 1" Dia. Holes
Controlled Leakage
Upper
Lower
Upper
Lower
41
41
207.8
207.8
16
16
81.1
81.1
8
8
40.5
40.5
4
4
20.3
20.3
24
8
121.6
40.5
LLF
(LLFBG = 0.5)
0.500
0.500
0.500
0.500
0.317
Correlated
LLFBG
0.640
0.524
0.508
0.484
0.477
Interestingly, the correlated background lower leakage fraction generally decreases in
magnitude as the total amount of enclosure leakage is lessened. An average of the last
column equals 0.526. Due to a lack of more complete information, this value is assumed to
be correct and is used where necessary in performing hold time prediction calculations.
Agent concentration measurements were made with a variety of instruments. For each, an
exhaustive effort was made to ensure that the recorded values were interpreted, filtered and
scaled into engineering units according to that prescribed by well-established measurement
theory 6 . An example of the finalized agent concentration data sets is provided in Figure 8.
Measurement uncertainty is likely less than ±5% of full scale, however, an investigation of
measurement error bounds and propagated uncertainty in the calculated quantities is yet to
be completed.
As a first investigation into this data set it is necessary to ascertain the agent distribution
profile within the enclosure. Again, three hold time models are available within NFPA 2001
and ISO 14520; the sharp descending interface, wide descending interface and continuously
mixing model. As previously discussed, the primary difference between these is the
assumption of agent distribution within the design envelope. Figure 9 demonstrates this
idealized assumption as employed in the hold time models.
6
Calibration of gas sampling instrumentation was not available. A relative measurement technique was
implemented in scaling of all agent concentration data. Recorded values were scaled into engineering units
based on a Zero value (a 30 second average of sampled fresh air before discharge) and a Full-Scale value (an
average of ≥ 90 seconds of data acquired after agent discharge had ended and readings had stabilized).
instantaneous snapshots in time, which
progresses with increasing luminosity. In
reality, it appears that the clean agent vapor
takes on a highly-stratified form; thus,
invalidating the applicability of the
continuous mixing model for hold time
predictions. Note that this model may
prove to be the only acceptable prediction
method when mechanical mixing is
provided throughout the hold time.
Comparatively, it can be concluded that the
interface width shares an inverse
relationship with the agent-air mixture
density. It seems that the observed agent
distribution profile lies somewhere between
the sharp and wide interface assumption.
By direct comparison with the empirical
results shown in Figure 10, the effective
agent distribution profile may be observed.
In each figure, the data traces represent
85.0%
75.0%
70.0%
65.0%
60.0%
50.0%
40.0%
8
Agent Volume Concentration (%)
7
6
of Max.
of Max.
of Max.
of Max.
of Max.
of Max.
of Max.
Ht.
Ht.
Ht.
Ht.
Ht.
Ht.
Ht.
(Perco1Ch2)
(Perco2Ch2)
(Perco2Ch3)
(Tuure2Ch2)
(Tuure2Ch3)
(TrPtCh2)
(TrPtCh3)
5
4
3
2
1
0
0
1000
2000
3000
4000
Time (s)
5000
6000
7000
8000
Fraction of Enclosure Maximum Height (%)
Figure 8: Example of Agent Concentration Data
Sharp Interface
Model Assumption
Continuous Mixing
Model A ss umption
Wide Interface
Model Assumption
1
0
0.5
1
0
Fraction of Initial Discharge Concentration Remaining (%)
0.5
1
0.8
0.6
0.4
0.2
0
0
0.5
1
Percent of Enclosure Maximum Height (%)
Figure 9: Idealized Agent Concentration Distribution Profiles Assumptions
HFC-227ea
FK-5-1-12
IG-541
HFC-125
100
80
60
40
Observed at
Observed at
Observed at
Observed at
20
0
0
50
500 s.
910 s.
1320 s.
1730 s.
Observed at
Observed at
Observed at
Observed at
100
0
50
500 s.
910 s.
1320 s.
1730 s.
100 0
Observed at
Observed at
Observed at
Observed at
50
500 s.
910 s.
1320 s.
1730 s.
100 0
Observed at
Observed at
Observed at
Observed at
50
500 s.
910 s.
1320 s.
1730 s.
100
Percent of the Initial Discharge Concentration Remaining in the Enclosure (%)
Figure 10: Examples of the Observed Agent Concentration Distribution Profiles for all Tested Agents
Comparison between tests involving various total amounts of leakage, lower leakage
fractions, and clean extinguishing agents (different agent-air mixture densities) may be
realized through the use of nondimensional analysis. Previously, the sharp descending
interface model’s dimensionless form was presented in Equation 1. Figures 11 through 14
show the experimentally observed interface height, h(t)/H0, as a function of the
dimensionless time, k2(t-t0)/τ, when evaluated at a 20% reduction from the initial discharge
concentration. Reference Table 3 for a listing of test configuration parameters including the
predicted and measured hold times. The values represent the time when 80% of the initial
discharge concentration is remaining at 75% of the maximum enclosure height.
1
1
Theory
Test ID #1
Test ID #2
Test ID #3
Test ID #5
0.9
0.8
0.7
0.8
0.7
0.6
h(t)/H0
h(t)/H0
0.6
0.5
0.4
0.4
0.3
0.2
0.2
0.1
0.1
0
0.1
0.2
0.3
0.4
0.5 0.6
k 2(t-t0)/τ
0.7
0.8
0.9
0
1
Figure 11: Dimensionless HFC-227ea Observed
Descending Interface Compared to Theory
0.1
0.2
0.3
0.4
0.5 0.6
k 2(t-t0)/τ
0.7
0.8
0.9
1
1
Theory
Test ID #11
Test ID #12
Test ID #13
Test ID #14
Test ID #15
0.7
0.6
0.8
0.7
0.6
0.5
0.4
0.5
0.4
0.3
0.3
0.2
0.2
0.1
0.1
0
0.1
0.2
0.3
0.4
0.5 0.6
k 2(t-t0)/τ
0.7
0.8
0.9
Figure 13: Dimensionless IG-541 Observed
Descending Interface Compared to Theory
Theory
Test ID #16
Test ID #17
Test ID #18
Test ID #19
Test ID #20
0.9
h(t)/H0
0.8
0
0
Figure 12: Dimensionless FK-5-1-12 Observed
Descending Interface Compared to Theory
1
0.9
h(t)/H0
0.5
0.3
0
Theory
Test ID #6
Test ID #7
Test ID #8
Test ID #9
Test ID #10
0.9
1
0
0
0.1
0.2
0.3
0.4
0.5 0.6
k 2(t-t0)/τ
0.7
0.8
0.9
1
Figure 14: Dimensionless HFC-125 Observed
Descending Interface Compared to Theory
Figures 11 and 13 exhibit a fairly poor relationship between the predicted sharp interface
descent and that observed in the experimental phase. The HFC-227ea test set was executed
nearly a full year prior to the other completed tests by a different research team using
different instrumentation. Further, insufficient fan integrity test data was supplied for these
tests and all the necessary parameters were retroactively calculated based on the number of
upper and lower holes provided. It is likely that the enclosure’s integrity has changed over
this time period, calling into question the assumed leakage configuration, or that the data
provided for analysis was misrepresentative of actual test conditions. The IG-541 data
indicates that the agent generally drains more readily from the design envelope than the
sharp interface model predicts. In part, this could be due to the vapor density of IG-541
being so close to that of ambient air. A large amount of back diffusion is observed (see
Figure 10 at 1320 seconds), which may have invalidated the sharp interface assumption.
Figures 12 and 14 show an excellent accuracy in the theory’s predicted interface descent
compared to the empirical results. A significant amount of data spread is clearly evident
although it seems well centered about the theoretical prediction. Generally, the accuracy of
the predicted interface decreases as the interface descends further towards the floor.
Table 3: Critical Test Configuration Parameters
Test Run Number:
Test Date:
Number of Open Upper Holes (-):
Number of Open Lower Holes (-):
Additional Open Leakages:
Total Measured ELA (cm2):
Discharge Concentration (Vol. %):
Est. Hold Time at 75% of Ht. (min):
Act. Hold Time at 75% of Ht. (min):
Empirical Error Relative to Est. (%):
Test Run Number:
Test Date:
Number of Open Upper Holes (-):
Number of Open Lower Holes (-):
Additional Open Leakages:
Total Measured ELA (cm2):
Discharge Concentration (Vol. %):
Est. Hold Time at 75% of Ht. (min):
Act. Hold Time at 75% of Ht. (min):
Empirical Error Relative to Est. (%):
0
0
64.0
-
1
10/21/05
27
27
N/A
410.4
7.0
9.4
11.0
16.8%
0
0
11
8/30/06
41
43
-
N/A
43.6
-
575.0
34.2
15.3
14.2
-7.7%
HFC-227ea
3
4
5
11/22/05
11/23/05
11/10/05
8
4
14
14
8
4
N/A
N/A
N/A
231.1
160.8
120.5
7.0
7.0
7.0
16.7
24.0
32.0
18.2
49.0
9.1%
53.1%
IG-541
12
13
14
15
8/30/06
8/31/06
8/31/06
8/31/06
16
16
16
41
16
8
16
43
14" by 36" 14" by 36"
open vent N/A
ceiling vent ceiling vent discharge only
272.0
121.2
272.0
575.6
34.2
34.2
34.2
38.0
32.4
72.7
32.4
14.5
28.4
13.3
-12.5%
-8.8%
2
11/9/05
27
27
N/A
410.4
7.0
9.4
10.4
11.1%
0
0
66.9
0
0
FK-5-1-12
6
7
8
9
10
8/28/06 8/29/06 8/29/06 8/29/06 8/30/06
16
41
8
4
7
16
41
8
4
7
N/A
N/A
N/A
N/A
N/A
295.0
596.5
166.3
123.2
157.9
4.2
4.2
4.2
4.2
4.2
12.0
5.9
21.2
28.7
22.4
12.9
5.6
22.6
34.7
24.7
7.9%
-5.0%
6.2%
21.0% 10.6%
HFC-125
16
17
18
19
20
9/26/06 9/26/06 9/27/06 9/27/06 9/27/06
41
16
8
4
24
41
16
8
4
8
-
N/A
N/A
N/A
N/A
N/A
59.8
-
585.4
8.0
7.6
7.0
-7.6%
288.8
8.0
15.3
16.7
8.7%
152.8
8.0
29.0
27.1
-6.7%
116.5
8.0
38.0
43.1
13.4%
288.8
8.0
20.5
23.7
15.6%
Errors shown in Table 3 are all calculated relative to the theoretically predicted hold time
value in order to provide a consistent basis of comparison between tests. For each test set,
the listed percent errors may be combined by finding the quadratic mean (root mean
squared). The combined empirical error of the HFC-227ea, FK-5-1-12, IG-541, and HFC125 data sets is 29%, 12%, 10%, and 11%, respectively. The HFC-227ea tests combined
error is exceedingly large compared to the other test sets. Reasons for this discrepancy are
still being considered. For all other agents tested, the combined error indicates relatively
good accuracy in the prediction of interface descent. Combined error levels are consistently
just greater than 10%. Considering the safety factors included in total flooding system
design, a hold time prediction error of this magnitude, while not an immediate cause for
alarm, is worthy of consideration and further study.
Summary and Conclusions
This paper documents the findings of a research program designed to experimentally
evaluate the applicability of the widely published hold time prediction models. Twenty
experiments involving a variety of enclosure leakage configurations were performed for four
clean extinguishing agents; FK-5-1-12, HFC-125, HFC-227ea, and IG-541. Experimental
results were modified to a dimensionless form to permit direct comparison between tests.
The inert gas agent, IG-541, drained from the test enclosure much more rapidly than the
theory predicts. Results for the halocarbons FK-5-1-12 and HFC-125 showed relatively
good accuracy in the predictions of agent leakage. Experimental results of HFC-227ea
testing indicate that this agent does not leak from the design envelope nearly as rapidly as
predicted. This is possibly due to uncertainties in the test setup configuration or due to a
lack of instrumentation that may have been subject to large measurement uncertainty. The
quadratic mean of empirical to theoretical hold time errors for all 2006 testing is ~10%.
The cooling affect of a clean agent discharge and resultant temperature change is not
accounted for in the models, which may lead to measurable errors in the predicted hold
time. Further analysis of these transient thermal effects is warranted.
Acknowledgements
This research effort was begun under the auspices of the NFPA 2001 Technical Committee
on Gaseous Fire Extinguishing Systems. 3M Company, Ansul Incorporated and Fike
Corporation provided FK-5-1-12, IG-541 and HFC-125 clean agents, respectively, for
discharge testing. Additionally, Fike Corporation provided a modern test facility and
multiple technicians aiding in making this effort possible. 3M Company and Sevo Systems
provided all halocarbon gas sampling instrumentation. Ansul Incorporated and Fike
Corporation provided all oxygen concentration gas analyzers. Equipment for and execution
of all door fan integrity testing was provided by Retrotec Incorporated.
I would like to extend an equal level of gratitude to the industry specialists who were able to
find the time and means to travel to Fike Corporation and support in the test process.
Contributing members of the research team include Dale Edlebeck of Ansul Inc., Colin
Genge of Retrotec Inc., Richard Niemann of Sevo Systems, Bob Whiteley of Tyco Fire and
Integrated Solutions, and Brad Stilwell, Mark McLelland and John Schaefer of Fike
Corporation.
References
[1] Dewsbury, J., and Whiteley, R. A. “Review of Fan Integrity Testing and Hold Time
Standards.” Fire Technology, Vol. 36, No. 4. November 2000. 249-265.
[2] DiNenno, P. J., and Forssell, E. W. “Evaluation of the Door Fan Pressurization
Leakage Test Method Applied to Halon 1301 Total Flooding Systems.” Journal of
Fire Protection Engineering, Vol. 1, No. 4. 1989. 131-140.
[3] Mowrer, F. “Analysis of Vapor Density Effects on Hold Times for Total Flooding
Clean Extinguishing Agents.” Halon Options Technical Working Conference, 16th
Proceedings. May 16-18, 2006, Albuquerque New Mexico. 2006. pp.1-12.
[4] Dewsbury, J., and Whiteley, R. A. “Extensions to Standard Hold Time Calculations.”
Fire Technology, Vol. 36, No. 4. November 2000. 267-278.
[5] “ISO 14520-1: Gaseous Fire Extinguishing Systems – Physical Properties and System
Design – Part 1: General Requirements, Annex E.” International Standards
Organization. 2006.
[6] “NFPA 2001: Standard on Clean Agent Fire Extinguishing Systems.” National Fire
Protection Association. 2004.
[7] O'Rourke, Sean T. “Analysis of Hold Times for Gaseous Fire Suppression Agents in
Total Flooding Applications.” MS thesis, University of Maryland [College Park].
2005.
[8] Saum, D., Saum, A., Messing, M. and Hupman, J. “Pressurization Air Leakage Testing
for Halon 1301 Enclosures.” Presented at Substitutes and Alternatives to
Chlorofluorocarbons and Halons, Washington, D.C., 1988.
[9]
Genge, C. “Preventing Excessive Enclosures Pressures During Clean Agent
Discharges.” Halon Options Technical Working Conference, 15th Proceedings.
May 24-26, 2005, Albuquerque New Mexico. 2005. 1-16.