2. Collaboration
Prof. Ali Rangwala
Fire Protection Eng.
WPI
Todd Hetrick
MS Student, WPI
Small Scale Testing
Kris Overholt
M.S. Student, WPI
Small-Scale Testing
Cecilia Florit,
French Exchange Student
Heat Flux Measurements
Jonathan Perricone
Creative FPE Solutions, Inc.
Industry Consultant
Opening picture from Bruce Smith, AP, 6/20/07, Charleston furniture warehouse where 9 firefighters died
2
3. Outline
I. Introduction to Commodity Classification
II. Theory
III. Experimental Setup
IV. Experimental Data and Results
V. Conclusion
VI. Future Work
3
5. Current Commodity Classification
Plastic
Group A-C
Warehouse
Commodity
Class I -IV
Warehouse
commodity
(Carton, packaging,
plastic)
Classify grouped
commodity into
one of seven
hazard groups
(Based on HRR)
Use large-scale test
data to design fire
protection system
(NFPA 13)
5
6. Recent Loss Case Example
2007 – Tupperware storage
warehouse fire1
15,392 m2 warehouse burned for
24 hours – a total loss
Sprinklers met state & local
requirements including NFPA 13
but the fire could not be
extinguished once plastic became
involved
2007 – Furniture warehouse fire kills
9 firefighters in Charleston, SC.2
Warehouse fires pose significant risks
to occupants, local environments,
and responding fire personnel
(Photo: Georgetown Country Fire Dept.
Hemingway, SC)
1The
Problem with Big, NFPA Journal, March/April 2009
Career Fire Fighters Die in Rapid Fire Progression at Commercial Furniture Showroom
– South Carolina, Fire Fatality Investigation Report, NIOSH.
2Nine
6
7. Shortcomings of Current Methodology
Current classification uses ranking scheme (Model is based on
commodity classification: Class I-IV, Group A-C Plastics) according
to the free-burning heat-release rate (HRR)
Full-scale fire tests are preferred, but intermediate tests are more
commonly used to assess Free-burning HRR & classification
Tests are not economically feasible
Free-burning HRR is scale dependent
Fundamental physics is glossed over
Tests have not been shown to be repeatable1
Classification by analogy presents a dangerous, yet common
industry practice. (i.e. plastic totes in S.C. Tupperware facility)
1Golinveaux,
“What We Don’t Know about Storage,” NFPA Presentation
7
9. Our Approach
Current Research
Small Scale Testing
Commodity type
classification
Cone Calorimeter testing
Intermediate
Scale Testing
(Proof of
concept)
Large/Full Scale
Modeling
(Proof of concept)
Engineering Approach to Commodity
Classification
9
10. Our Area of Contribution
Computer Fire
Modeling
Model potential rack setups
& sprinkler interactions
Modeling used to test
warehouse designs costeffectively, based on ranking
Commodity
Classification
Provide input parameters
Add influence of large-scale
Large-Scale Testing
(Verification of both)
Bench-scale tests determine
nondimensional parameters (B)
Rank commodities on
fundamental scale used to
design sprinkler system
11. Material Flammability
Factors controlling flammability and fire hazard
Ignition
Fire Growth
Burning Intensity
Generation of Smoke and Toxic Compounds
Extinction/Suppression
11
12. Commodities Used in Testing
Class II
Class III
Class
IV/Group B
Group A Plastic
Commodities Used in Reality
12
14. Commodity Fire: Stage 1 –
Laminar Case
Boundary layer
B is a function of:
1. Corrugated board
Buoyant Plume
Plume Radiative +
Convective Heat Transfer
Commodity
Combusting Plume
Excess
Pyrolyzate
Pyrolysis Zone
mF
Flame Radiative +
Convective Heat Transfer
XF
flame
XP
• Flame height <25 cm
• Unrealistic in fire situation
Corrugated board
• Study important because
provides physical understanding of the problem
(~ 20 to 25 cm Laminar
Flame Propagation)
Y-axis
14
15. Stage 2 – Turbulent Case
Buoyant Plume
Boundary layer
B is a function of:
1. Corrugated board
2. Commodity pyrolysis vapor
Combusting Plume
Flame Radiative +
Convective Heat Transfer
Excess
Pyrolyzate
Commodity
mF
Pyrolysis
Zone
• Flame height >25 cm
• Realistic fire situation
• Cardboard still intact
Plume Radiative +
Convective Heat Transfer
XP
XF
(Turbulent flame height >25 cm)
flame
Y-axis
Corrugated board
15
16. Stage 3 – Mixed Case
• Flame height >25 cm
• Realistic fire situation
• Cardboard breaks
16
17. Buoyant Plume
Plume Radiative +
Convective Heat Transfer
Stage 3 – Mixed Case
Combusting Plume
B is a function of:
1. Corrugated board
2. Commodity pyrolysis vapor
Excess
3. Commodity
Pyrolyzate
flame
Flame Radiative +
Convective Heat Transfer
(from pool and wall fire)
Commodity
XF
Solid/Liquid
Pool fire
Corrugated
board
m
F
• Flame height >25 cm
• Realistic fire situation
• Polystyrene leaks and
starts pool fire
Boundary layer
Pyrolysis
Zone
mF
Commodity leakage
Pyrolysis
Zone
Y-axis
17
18. The B-number
B
im petuses i.e. heat of com bustion for bu rning
resistances i.e. heat of vaporization to the process
“Thermodynamic Driving Force”
B
(1 )( H c YO , ) / s C p , (Tp T )
χ = Fraction of radiation lost [-]
∆Hc = Heat of combustion [kJ/kg]
YO,∞ = Mass fraction of oxygen in ambient [-]
νs = Oxygen-fuel mass stoichiometric ratio [-]
Cp,∞ = Specific heat of ambient air [kJ/kg-K]
Tp = Pyrolysis temperature of the fuel
Hg Q
B-number
T∞ = Ambient temperature [K]
L = Latent heat of vaporization [kJ/kg]
∆Hc = Heat of gasification [kJ/kg]
Cp,f = Specific heat of the fuel [kJ/kg-K]
Q = L + Cp,f(TB-TR) [kJ/kg]
[1] Kanury, A. M. An Introduction to Combustion Phenomena. s.l. : Gordon & Breach Science Publishers, Inc, 1977.
18
19. Reynolds Analogy
Application to Warehouse Commodity Classification
mF
1
Flow
Condition
hT
ln(1 B )
cg
Material Properties
2
B-number
19
20. Experimentally-Measured B
•Solving for B and using Nu correlation for the heat-transfer coefficient:
''
mf
B exp
0.13[G r Pr]1 / 3
g g
1
•Formula for average B-number based on measured rate of mass loss
•Applies in regimes dominated by convective heat transfer, as found in
many small-scale experiments.
•Effective B-number derived by same formula with radiation included
Kanury, A. M. An Introduction to Combustion Phenomena. Gordon & Breach Science Publishers, Inc, 1977.
20
23. Picture of Experimental Setup
WPI, Summer 2008
TC wires
Heat flux sensors
Back View
Front View
23
24. Measurement of Heat Flux
Thin-Skin Calorimeter
Combined heat flux from calorimeter
(accounting for losses)
q i q c q r q sto q c , st
qi
qc
qr
q c , st
q sto
American Society of Testing and Materials, Standard ASTM E 459-97
24
26. Commodity Test Results
30 s
92 s
Front Face of Cardboard
Burning
Stage I
100 s
132 s
150 s
Plateau
PS Cups & Cardboard
Burning
Stage II
Stage III
26
35. Commodity Test Results
Summary of Stages
Stage I
Stage II
Stage III
Outer layer of commodity is ignited,
producing rapid upward turbulent
flame spread over the front face of a
commodity. B is independent of
polystyrene.
Front layer of corrugated cardboard
has burned to top, exposing inner
region, which burns and then
smolders. Polystyrene does not burn
because of its higher ignition
temperature.
B
1.8
&
m
0.83 g/s
X
0.51 m
f ,a vg
&
q f"
1.2 kW/m2
B
1.4
&
m
X f ,a vg
1.7 g/s
&
q f"
Polystyrene ignites and a rapid increase B
in the burning rate occurs.
&
m
X f ,a vg
&
q f"
0.48m
0.38 kW/m2
1.9
2.2 g/s
0.65 m
2.4 kW/m2
35
37. Conclusions
A new method of hazard ranking is suggested in this study
based on a nondimensional parameter: B
In a warehouse setting, where the burning rate is the
dominant fire hazard, the effective B-number may
appropriately classify the hazard of a grouped commodity
The B-number can be calculated using small-scale tests
Commodity Upward Spread via Mass Loss Rate
Cone Calorimeter Upward Spread via Mass Loss Rate
(Overholt et al.)
Flame Standoff Distance (Rangwala et al.)
1. K.J. Overholt, M.J. Gollner, and A. Rangwala, "Characterizing Flammability of Corrugated Cardboard Using a Cone
Calorimeter," Proceedings of the 6th U.S. National Combustion Meeting, 2009.
2. A.S. Rangwala, S.G. Buckley, and J.L. Torero, "Analysis of the constant B-number assumption while modeling
flame spread," Combustion and Flame, vol. 152, 2008, pp. 401-414.
37
39. Conclusions
This parameter is nondimensional and in preliminary tests
predictions from this parameter show good correlations
to test data
The economic advantage of predicting full-scale
performance with small-scale experiments may be an
impetus for a significant evolution in the field of fire
protection engineering.
39
41. Future Work
Flame height prediction (including influence of radiation)
Study possible correlations between B-number and other
relevant flammability parameters (TRP, FPI, CHF, etc.)
Variation of Fuel/Commodity Volume/Mass Ratios
Incorporate suppression – minimum suppressant (water
spray) can be incorporated in B-number via loss term
41
42. Experimental setup to determine the water application
rate ω g/cm2-s at different external heat fluxes
flame
Lab air supply
commodity
External heat flux
nozzle
Pan for
water collection
excess water
collector
regulator
Pressurized
water supply
z (cm)
water
ω
Water application rate
g/cm2s
Load cell
Load cell
42
43. Acknowledgements
David LeBlanc at Tyco for generous donation of standard
Group A storage commodity and sharing full scale test data
conducted at UL labs by Tyco.
San Diego office of Schirmer Engineering for contributing start
up funding at the beginning of the project.
WPI Lab Manager Randy Harris, Research Assistants: Cecelia
Florit and Todd Hetrick, and helpful discussions with Jose
Torero (University of Edinburgh)
43
48. Flame Spread Theory
Many different theories of upward-spreading flames exist
Annamalai & Sibulkin
X
f
~ A( B t )
2
Saito, Quintiere, Williams
X
f
~ Ae
t
f (T R P, X
f
''f )
/ X p,q
1. Annamalai, K. and Sibulkin, M. Flame spread over combustible surfaces for laminar flow systems. Part I & II:
Excess fuel and heat flux. 1979, Combust. Sci. Tech., vol. 19, pp. 167-183.
2. Saito, J.G. Quintiere, and F.A. Williams, "Upward Turbulent Flame Spread," Fire Safety Science-Proceedings of the
First International Symposium, 1985, pp. 75-86.
48
49. Sibulkin & Kim
Yo ,
YF vs
x f 0.64( r / B )
2 / 3
xp
Convection
Convection +
Radiation
Sibulkin and Kim, Comb. Sci. Tech. vol. 17, 1977
49
50. Variation of Fuel/Commodity
Volume/Mass Ratios
Experimental set up allows systematic control of two
important parameters at small scale:
volume fuel / volume total
commodity weight / packing material weight
50
51. Proposed Experimental Setup
Hood
Detailed
front view in
next slide
Quartz tube
External heat flux
IR lamps
Extension plate
Flow seeding
Aluminum
tube
Flow straightener
(honey comb mesh)
O2 + N2 mixture
Load cell
51
54. burning rate
g/cm2s
Top view of experimental
apparatus
Plastic grid
Volume fraction = Ф
50 x 10 x 5 cm
commodity
Corrugated cardboard
(outer cover)
Zero water application
(free burn)
1.5ω g/cm2s
Critical
mass flux
(control)
5 cm
ω g/cm2s
2ω g/cm2s
3ω g/cm2s
z (cm)
Critical mass
flux (extinction)
Radiant heater
Increasing rate of
water application
Water spray nozzle
ω g/cm2s
0
d
c
b
a
External heat flux, kW/m2
Burning rate vs. external heat flux for various water application rates.
Curves are hypothetical, based on experimental data reported by
Magee and Reitz
Magee, R.S. and R.D. Reitz, Extinguishment of radiation augmented plastic fires by water sprays.
Proc. Combust. Inst. 15: p. 337-347.
54
Notas del editor
Initial funding for experiments was provided by Schirmer Engineering.
The goal of this study is to evaluate a method of characterizing the relative fire hazard of a grouped warehouse commodity. * Commodity classification refers to a ranking of estimated fire challenge to sprinkler protection.
Why are warehouse fires, and particularly upward spread problems being revisited for study? Examples of losses, where “fully-protected” facilities are still burning down despite what should be full protection. There are still glaring lacks of scientific understanding of the fire spread process, let alone suppression process for these large fires. This endangers firefighters, occupants, and local environments. Soot production and toxic runoffs from 2-day long fires are extreme.
Commodity classification in warehouse used to determine requirements of suppression system based on Area Density Curves. Full-scale fire tests are preferred by more than just the sprinkler industry. They give the best approximation available; however, there are several problems:By rushing to full-scale tests exclusively, we are glossing over the fundamental physics into something that is so complex that it defies clear comprehensionFull-scale testing is not always economically feasible or even physically possible (i.e. very tall rack storage arrangements)Free-burning HRR paints an incomplete picture. Several other driving and resisting influences govern combustion such as...Free-burning HRR is scale dependent. Ranking via intermediate scale measurements doesn’t necessarily translate into accurate full-scale assessment. The scale dependence may extend to multiple geometric dimensions (i.e. wood cribs).Are results repeatable? Tyco was unable to reproduce data points on the existing curves according to Golinveaux presentation at NFPA Conference 2007.Are results reproducible? Inconsistencies between NFPA and CENClassification by analogy is common industry practice. No testing performed. Analogies made in non-standard manner and generally very poorly substantiated (i.e. plastic totes in Tupperware facility in S.C.)Suppression and control are vaguely defined. The difference between them is unclear.The distinction between a fire under control and out of control is also unclear.
Goal is to make fire scenarios more predictable with scientifically verifiable results from testing. Distinguish & predict levels of fire control, suppression, and extinction for multiple commodities and layouts.Currently: Working on Cone Calorimeter (see other presentation) and small-scale testing on a standard commodity to determine non-dimensional parameters. Later will predict & assess accuracy of new methods with intermediate (Fire products collector) & large scale testing.
Our contribution is Bench-scale tests to determine nondimensional parameters, (B)Will be used to rank commodities on a fundamental scale used to design a sprinkler system.May also be useful as an input into modeling applications in the future.
Commodity classification in warehouse used to determine requirements of suppression system based on Area Density Curves. Full-scale fire tests are preferred by more than just the sprinkler industry. They give the best approximation available; however, there are several problems:By rushing to full-scale tests exclusively, we are glossing over the fundamental physics into something that is so complex that it defies clear comprehensionFull-scale testing is not always economically feasible or even physically possible (i.e. very tall rack storage arrangements)Free-burning HRR paints an incomplete picture. Several other driving and resisting influences govern combustion such as...Free-burning HRR is scale dependent. Ranking via intermediate scale measurements doesn’t necessarily translate into accurate full-scale assessment. The scale dependence may extend to multiple geometric dimensions (i.e. wood cribs).Are results repeatable? Tyco was unable to reproduce data points on the existing curves according to Golinveaux presentation at NFPA Conference 2007.Are results reproducible? Inconsistencies between NFPA and CENClassification by analogy is common industry practice. No testing performed. Analogies made in non-standard manner and generally very poorly substantiated (i.e. plastic totes in Tupperware facility in S.C.)
Laminar case is not a principal hazard but likely present in initial stages and useful for theoretical analyses.Corrugated board is ignited at the base, heats to a pyrolysis temperature where the board pyrolyzes expelling fuel vapors. Oxygen in the air diffuses towards this region and forms a thin diffusion flame at some standoff distance above the commodity. Not all of the fuel is burned directly in front of the pyrolysis region, some flows upwards beyond the pyrolysis zone into the combusting plume, where it burns above. This is called “excess pyrolyzate.” This excess pyrolyzate burning in the combusting plume heats the rest of the commodity incredibly fast as it burns in front of the surface, developing a very quickly growing upward spreading flame.
The same pyrolysis zone, combusting plume, and buoyant plume exist in this case. Now, the flame is larger and the combusting plume begins to extend above the height of the commodity surface, and the flame becomes very turbulent. The fire grows more rapidly. The remaining products are then propelled above into a buoyant plume.
In the mixed case the same processes still occur but now some leakage of the commodity (in the form of melted plastic) pool in front of the commodity. Now remaining cardboard burns as well as a small pool fire at the base of the commodity. The characteristics of the pool fire as well as the flat plate burning must be taken into account. In our tests, for safety reasons the fire was extinguished before significant commodity leakage occurred. We burnt approximately only 3/4 of the commodity. It would take upwards of 3-5 minutes for this to occur based on observations from tests, so characterizing the earlier region to involved suppression is more important for this study.
In the mixed case the same processes still occur but now some leakage of the commodity (in the form of melted plastic) pool in front of the commodity. Now remaining cardboard burns as well as a small pool fire at the base of the commodity. The characteristics of the pool fire as well as the flat plate burning must be taken into account. In our tests, for safety reasons the fire was extinguished before significant commodity leakage occurred. We burnt approximately only 3/4 of the commodity. It would take upwards of 3-5 minutes for this to occur based on observations from tests, so characterizing the earlier region to involved suppression is more important for this study.
“Universal Meaning”The B number can be thought of as a thermodynamic or mass transfer driving force. It was first introduced by Spalding in 1950 to develop an expression for the burning rate of a liquid fuel droplet in a gas stream. The uncorrected B-number is a property of pyrolyzing material, and it appears in boundary conditions of energy conservation at the fuel surface. The corrected B-number accounts for influences of additional heat-transfer processes. Physically, it relates the heat release of combustion (the numerator) to the energy required to generate fuel gasses (the denominator).In a mass-transfer sense it is the ratio of an impetus for interphase transfer to a resistance opposing that transfer.
Mf’’ is the mass loss rate per unit are of the material, which is related to the heat transfer component (h/Cg) times the thermodynamic component ln(B+1). Because of the log relationship of B, heat transfer plays a larger role in this process. H is assumed to be a constant in this process and is determined by a relation first relating it to the Nusselt number, and a nusselt number correlation which is a function of the cubed root of the Grashof number times Prandtl number. This approach is not exact, but for these small-scale experiments it is acceptable to ignore these small variations in h. Future work we are conducting will investigate the heat transfer coefficient numerically. The resulting formula for the average B number is an exponential function of the mass loss rate of the fuel per area over constants minus 1.
Deemphasize
Representative test
Shaded region denotes overlap between regions 1 & 2. Test timelines have been adjusted so transition between stage I and II lie on the same point on the time scale.
Shaded region denotes overlap between regions 1 & 2. Test timelines have been adjusted so transition between stage I and II lie on the same point on the time scale.
Shaded region denotes overlap between regions 1 & 2. Test timelines have been adjusted so transition between stage I and II lie on the same point on the time scale.
Time averaged mass loss rates used to calculate B-number over entire region. This B-number is an ESTIMATE in the region, it has been shown by Rangwala et al. that the B-number does change during upward spread, but deviations on average are minimal and results over this region are roughly repeatable for multiple tests.
Shaded region denotes overlap between regions 1 & 2. Test timelines have been adjusted so transition between stage I and II lie on the same point on the time scale.
Shaded region denotes overlap between regions 1 & 2. Test timelines have been adjusted so transition between stage I and II lie on the same point on the time scale.
Shaded region denotes overlap between regions 1 & 2. Test timelines have been adjusted so transition between stage I and II lie on the same point on the time scale.