2. Collaboration
Theory & Data Analysis
Forman Williams Michael Gollner
UC San Diego UC San Diego
Small-Scale Testing Ali Rangwala Kris Overholt Cecelia Florit
WPI WPI WPI/University of
Marsielle, France
Jonathan Perricone
Schirmer Engineering
Corporate Sponsor
2
3. Presentation Overview
Current Commodity Classification Limitations
A Fundamental Approach Towards Classification
Upward Turbulent Fire Propagation Theory
Nondimensional Parameters
Experimental Approach
Conclusion and Future Recommendations
3
4. Recent Loss Case Example
2007 – Tupperware storage
warehouse fire*
15,392m2 warehouse burned
24 hours to extinguish fire
Sprinklers met state & local
requirements including NFPA 13
Significant property losses have
Once plastic became involved occurred in the United States as
fire was uncontrollable recent as 2007 as a result of such
shortcomings.
Building was a total loss (Photo: Georgetown Country
Fire, Dept., Hemingway, SC)
*The Problem with Big, NFPA Journal, March/April 2009
4
5. Recent Loss History
2006 – Fire destroys warehouse, vehicles at UPS
facility*
2002 – Storage Warehouse Fire Phoenix, Az.+
2001 – Supermarket Fire in Phoenix, Az.‡
1998 – Warehouse fire in Tempe, Az.‡
1996 – Lowes store fire in Albany, Ga.‡
*Arizona News briefs. Fire destroys warehouse, vehicles at UPS facility. Newspaper. Mar. 28, 2006.
+Duval, R. F. Fire Investiagtion Report - Storage Warehouse - Phoenix, AZ. Quincy, MA : National Fire
Protection Association, 2002.
§Duval, R. F. and Foley, S. N. Fire Investigation: Supermarket Fire, Phoenix Arizona, March 14, 2001. Quncy,
MA : National Fire Protection Association, 2002.
‡Harrington, J. L. Lessons Learned from Understanding Warehouse Fires. Fire Protection Engineering.
Winter, 2006. 5
7. Aspects of Material Flammability
Ignition
Fire Growth
Burning Intensity
Extinction/Suppression
Generation of Smoke & Toxic Compounds
7
8. Warehouse
Fire
Protection
Model
*Zalosh, R. G., Industrial Fire Protection
Engineering. John Wiley and Sons, 2003
8
9. Current Commodity Classification
Current classification methods use ranking scheme (Class I-IV,
Group A-C Plastics) based upon the free-burning heat-release
rate of a given fuel
Intermediate-scale measurements of this parameter are used
as the cornerstone for fire suppression design in modern
storage facilities
The sprinkler industry identifies a database of full-scale fire
tests as validation for this approach
There is a lack of detailed measurements associated with
performance analysis of these tests. Full-scale fire tests are
typically judged in terms of pass/fail with great significance
attached to subjective observations
9
10. Current Commodity Classification
NFPA and FM Global recognize their classification
schemes at the very best can only provide a preliminary
indication of relative flammability*
These results can be misleading and dangerous!
European (CEN) standards classify commodities into 4
groups much like NFPA 13, but there are contradictions
between these categories*
Repeatability cannot even be explored - extensive large-
scale testing is too expensive*
*Zalosh, R. G., Industrial Fire Protection Engineering. John Wiley and Sons, 2003
10
12. Why have a fundamental approach?
Fire scenarios will become more predictable
Scientifically verifiable results from testing
Moves towards guaranteed protection because the
worst case fire conditions and suppression or
extinction requirements can be better estimated
12
13. Our Approach
Current Research Large/Full Scale
Intermediate Modeling
Scale Testing
(Proof of (Proof of concept)
Small Scale Testing concept)
Commodity type
classification
Cone Calorimeter testing Engineering Approach to Commodity
Classification
13
15. Stage 1 – Laminar Case
Boundary layer
B is a function of:
1. Corrugated board
Buoyant Plume
Plume Radiative +
Convective Heat Transfer
Commodity
Excess
Combusting Plume
Flame Radiative +
Pyrolyzate Convective Heat Transfer
Pyrolysis Zone flame XF
mF
(~ 20 to 25 cm Laminar
XP Flame Propagation)
• Flame height <25 cm
Y-axis
• Unrealistic in fire situation Corrugated board
• Study important because
provides physical understanding of the problem 15
16. Stage 2 – Turbulent Case Buoyant Plume
Plume Radiative +
Boundary layer Convective Heat Transfer
B is a function of:
1. Corrugated board Combusting Plume
2. Commodity pyrolysis vapor Flame Radiative +
Convective Heat Transfer
Excess
Pyrolyzate
Commodity
mF
Pyrolysis XP XF
(Turbulent flame height >25 cm)
Zone
• Flame height >25 cm flame
• Realistic fire situation
• Cardboard still intact Y-axis
Corrugated board
16
17. Buoyant Plume
Plume Radiative +
Convective Heat Transfer
Stage 3 – Mixed Case
Combusting Plume
Flame Radiative +
B is a function of: fla Convective Heat Transfer
(from pool and wall fire)
1. Corrugated board me
2. Commodity pyrolysis vapor
3. Commodity Excess
Pyrolyzate
Commodity
XF Boundary layer
Corrugated
Solid/Liquid
board m
F Pool fire
Pyrolysis
Zone
• Flame height >25 cm
• Realistic fire situation
m F Pyrolysis
• Cardboard breaks Zone Y-axis
Commodity leakage
17
18. Theory
•The B-number is a ratio that compares a summation of the various
impetuses (i.e. heat of combustion) for burning to a summation of
the various resistances (i.e. heat of vaporization) to the process.
“Thermodynamic Efficiency”
•It can be described in relation to a mass-loss rate of a commodity
which can be measured in a laboratory
[1]
•Solving for B and using several other well-known heat transfer
relations a formula for estimating an average B-number based on
mass loss is
[2]
[1] Kanury, A. M. An Introduction to Combustion Phenomena. s.l. : Gordon & Breach Science Publishers, Inc, 1977.
[2] M.J. Gollner, T. Hetrick, A. S. Rangwala, J. Perricone, and F. A. Williams. Controlling Parameters Involved in the Burning of
Standard Storage Commodities: A fundamental approach towards fire hazard classification. 6th U.S. National Combustion
Meeting, 2009.
19. Flame Height Theory
•Using theory given by Annamali & Silbulkin [2], the expected flame
height for a vertically oriented material can be found by using a B-
number and constant material and air properties
[1,2]
•A procedure to perform this calculation is found in reference [1]
[1] Overholt, K., Gollner, M.J., Rangwala, A.S. Characterizing flammability of corrugated cardboard using a
cone calorimeter. 6th US National Combustion Meeting of the Combustion Institute. May, 2009.
[2] 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. 19
20. Cone Calorimeter Results
Flame heights measured in the
small-scale cone calorimeter
tests are compared to the
predicted flame heights
calculated by A&S model. The
shaded area represents the
flame heights from the cone
tests. The dashed line shows the
predicted flame heights from
the model using an average B-
number from 2 cardboard tests.
The dark circles represent the
predicted flame heights from
previous literature (0.8)
Overholt, K., Gollner, M.J., Rangwala, A.S. Characterizing flammability of corrugated cardboard using a cone
calorimeter. 6th US National Combustion Meeting of the Combustion Institute. May, 2009. 20
21. Cone Calorimeter Results
Flame heights from FM
experiments are compared to
predicted flame heights
calculated by the A&S model.
The shaded area represents the
flame heights from the FM
tests. The dashed line shows the
predicted flame ehgiths from
the model using an average B
number from the 4 cardboard
tests. The dark circles represent
flame heights using a B-number
from previous literature (0.8).
Overholt, K., Gollner, M.J., Rangwala, A.S. Characterizing flammability of corrugated cardboard using a cone
calorimeter. 6th US National Combustion Meeting of the Combustion Institute. May, 2009. 21
22. Commodity Test Results
Measured vs. Predicted Flame Heights
150 B=1.26 B=0.7 B=1.41 Predicted flame
height from Group-A
Flame Height (cm)
Region with
plastic commodity
Transition
cardboard region
using A&S data.
100 front face
burning
•Blue dotted lines are
predicted flame heights
50 Region with
cardboard & •Red x’s denote measured
PS burning flame heights
0
0 50 100 150 200
Time (s)
22
24. Nondimensionalization
Nondimensionalization is the removal of units from a
mathematical equation by a suitable substitution
of variables
This technique can simplify and parameterize problems
where measured units are involved
Useful in scaling analyses (small-scale to full-scale)
For Example:
Reynolds number is a nondimensional parameter
24
26. Parameters Involved in Study
Fire Propagation
Fire Propagation Index (FPI)
Ignition
Critical Heat Flux (CHF)
Burning Rate
B-number
A nondimensional form will allow useful scaling
analyses of these parameters
26
27. Fire Propagation Index
Fire propagation Index, FPI is proportional to the square
root of the flame spread velocity
FPI is nondimensionalized by the regression velocity, VR of
the material: qfHg
VR
FPI is expressed as a nondimensional parameter of the
form FPI FPI
FPI
*
qfHg / VR
27
28. Calculations of FPI*
VR
Material FPI [(m/s)1/2] FPI*
[m/s.10-5]
Polymethylmethacrylate
31 3.2 5.5
(PMMA)
Polypropyelene (PP) 32 3.7 5.3
Polystyrene (PS) 34 4.2 5.2
Polyethylene (PE) 28 3.2 4.9
Polycarbonate (PC) 14 1.4 3.7
Wood Slab (Doug Fir) 14 4.1 2.2
Polyvinylchloride (PVC) 7 1.5 1.8
* Data for calculations taken from SFPE Handbook of Fire Protection Engineering, Fourth Edition. 28
29. Critical Heat Flux
The Critical Heat Flux is the minimum flux applied to a
material that will cause it to ignite [W/m2]
CHF can be nondimensionalized by means of a maximum
heat-release rate, HRR [W/ m2] of the commodity
CHF
CHF *
HRR
29
30. The B-number
Spalding’s B-number, or Mass Transfer Number
Derived directly from governing equations for combustion
Dimensionless ratio that compares a summation of the
various impetuses (i.e. heat of combustion) for burning to
a summation of the various resistances (i.e. heat of
vaporization to the process. “Thermodynamic Efficiency”
B-number can be found experimentally from burning rate*
h
mf
ln(B 1)
Cg
*Kanury, A. M., Introduction to Combustion Phenomena. Gordon and Breach Science Publishers, New York. 1975 30
31. Nondimensional Parameters
Non-Dimensional FPI used to quantify flame spread
FPI FPI = Fire Propagation Index
FPI* ρ = Flame Density
qfHg /
∆Hg = Heat of gasification
qf‘’ = Feedback flux
Non-Dimensional Flux to quantify heating flux from the
burning commodity
CHF
CHF * CHF = Critical Heat Flux (flux which will cause material to
HRR ignite)
HRR = Average heat-release rate of material
H f YO , Cg (T TB ) B-number to characterize burning rate
B CHF = Critical Heat Flux (flux which will cause
L C l (TB TR ) material to ignite)
HRR = Average heat-release rate of material
31
33. Experimental
Setup- Cone
Tests
Overholt, K., Gollner, M.J., Rangwala, A.S. Characterizing flammability of corrugated cardboard using a cone
calorimeter. 6th US National Combustion Meeting of the Combustion Institute. May, 2009. 33
35. Cone Testing
35
*Cone calorimeter work conducted by K. Overholt, WPI
36. Experimental Setup:
Small-Scale Test
Class III
Commodity
Group-A Plastic
Commodity
Standard Group-A Plastic Commodity
Polystyrene cups in compartmented cardboard carton
36
37. Picture of Experimental Setup
WPI, Summer 2008
TC wires
Heat flux sensors
Back View Front View
37
42. Cone Calorimeter Results
Average B numbers found
Cardboard: B ≈ 1.78
Polystyrene (PS): B ≈ 2.76
Cardboard with PS Backing: B ≈ 4.41
Values are preliminary data
Cardboard values approximately double previously
reported values
When used to predict flame heights, B-number values for
cardboard match test data
42
43. Commodity Test Results
30 s 92 s 100 s 132 s 150 s
Front Face of Cardboard
Plateau PS Cups & Cardboard
Burning
Burning
44. Commodity Test Results
m
f PS cups
burning
(g/m2s)
Measured mass loss 1
rate during the Packing
material Extinction
experiments. After 0.8 (cardboard)
120s the PS cups
started burning and 0.6 Front face of
cardboard
test was terminated. burning
0.4
0.2
0
0 20 40 60 80 100 120 140 160
Time from Ignition [s]
44
45. Commodity Results – Mass Loss Mass Loss Rate, Test 1 Mass Loss Rate, Test 2
0.016 0.016
Mass Loss Rate [kg/m2]
Mass Loss Rate [kg/m2]
0.014 0.014
0.012 0.012
0.01 0.01
0.008 0.008
0.006 0.006
0.004 0.004
0.002 0.002
0 0
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 160 180 200
Time from Ignition, [s] Time from Ignition, [s]
Mass Loss Rate, Test 3 Mass Loss Rate, Test 4
0.018 0.018
0.016 0.016
Mass Loss Rate [kg/m2]
Mass Loss Rate [kg/m2]
0.014 0.014
0.012 0.012
0.01 0.01
0.008 0.008
0.006 0.006
0.004 0.004
0.002 0.002
0 0
0 20 40 60 80 100 120 140 160 180 200 0 20 40 60 80 100 120 140 160 180
45
Time from Ignition, [s] Time from Ignition, [s]
46. Commodity Results – Mass Loss B-number, Test 1 B-number, Test 2
5 5
4.5 4.5
4 4
3.5 3.5
3 3
B
B
2.5 2.5
2
Average B 2
1.5 1.5
1 1
0.5 0.5
0 0
0 20 40 60 80 100 120 140 0 20 40 60 80 100 120 140 160 180 200
Time from Ignition, [s] Time from Ignition, [s]
B-number, Test 3 B-number, Test 4
5 5
4.5 4.5
4 4
3.5 3.5
3 3
B
B
2.5 2.5
2 2
1.5 1.5
1 1
0.5 0.5
0 0
0 20 40 60 80 100 120 140 160 180 200 0 20 40 60 80 100 120 140 160 180
Time from Ignition, [s] Time from Ignition, [s]
46
47. Commodity Test – Measured
Values
Test 1 Test 2 Test 3 Test 4
Average MLR (g/s): 2.97 3.88 2.54 4.15
Average B number: 0.60 0.82 1.33 1.26
Median B number: 0.31 0.51 1.01 0.47
StdDev for B: 0.64 0.87 2.59 1.94
Average B-
number 1.2595
47
49. Warehouse
Fire
Protection
Model
*Zalosh, R. G., Industrial Fire Protection
Engineering. John Wiley and Sons, 2003
49
50. Conclusions
A new method of hazard ranking is introduced in this study based on
nondimensional parameters: B, FPI*, and CHF*
In a warehouse setting, where the burning rate is the dominant fire
hazard, the B-number may appropriately classify the hazard of a
grouped commodity – especially if we can correlate FPI* and CHF*
with B
These parameters can be determined by small-scale laboratory tests
The B-number can be calculated by the Cone Calorimeter and/or
grouped commodity tests
FPI* can be determined using current testing methods by incorporating
parameters already measurable on the NIST LIFT apparatus
CHF* could possibly be determined by testing of a single grouped
warehouse commodity
50
51. Conclusions
These parameters are nondimensional and in preliminary
tests show good correlations to full-scale 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.
51
There has never been a test on a plastic tote!There were plastic totes insidethe warehouse.In-rack has been suggested as a solution… - Increasing suppression for all materials requires a method of almost literally removing all Oxygen. Some warehouses might as well be built underwater.
Mentoin environmental impacts of warehoues fires as well – such as in Sweden….
Discuss how each of these plays a role in
“Both Factory Mutual and NFPA realize that their generic classification schemes are more valuable for providing a preliminary indication of relative flammability than a firm irrefutable determination.” – Zalosh pg. 131i.e. A highly hazardous material may only be moderately hazardous in Europe, and vice versa.
Approach to classify commodities starts with laboratory tests on single commodities, then mid-scale tests on mixed commodities, and eventually full-scale validation. Once full scale validation is accomplished, only laboratory tests will be required in the future.
Mf is the mass loss rate per unit area per unit time, all other variables are constants in this case. They depend on the properties of air and the material.
Xp is the pyrolysis height, and the flame height is found using phi.
This does depend on densityDifferent density papers lightweight heavywheight in 13 and how this is effected.
Conce Calorimeter single sheet cardboard test (identical to single sheet polystyrene & polystyrene backed cardboard tests)
Cone calorimiter tests were conducted on single-cell setups, double-cell setups (as shown above), as well as individual materials.Data will be showed later individually for cardboard, polystyrene, and mixed setups.
One small piece of a big project.To get to a point to accurate equations to calculate in the design phase this is what we have to go through.