NUREG/CR-24 31 SAND81-0079 RP BURN MODE ANALYSIS OF HORIZONTAL CABLE TRAY FIRES Fritz R. Krause* Willard H. Schmidt Printed February 1982 Systems Safety Technology Division Sandia National Laboratories Albuquerque, New Mexico 87185 operated by Sandia Corporation for the U. S. Department of Energy Prepared for Division of Engineering Technology GCfioce of Nuclear Regulatory Research U. S. Nuclear Regulatory Commission Washington, DC 20555 Under Memorandum of Understanding DOE 40-550-75 NRC FIN No. A1010 *Now at Los Alamos National Laboratory i
Transcript
NUREG/CR-24 31SAND81-0079
RP
BURN MODE ANALYSIS OF
HORIZONTAL CABLE TRAY FIRES
Fritz R. Krause*Willard H. Schmidt
Printed February 1982
Systems Safety Technology DivisionSandia National Laboratories
Albuquerque, New Mexico 87185operated by
Sandia Corporationfor the
U. S. Department of Energy
Prepared forDivision of Engineering Technology
GCfioce of Nuclear Regulatory ResearchU. S. Nuclear Regulatory Commission
Washington, DC 20555Under Memorandum of Understanding DOE 40-550-75
NRC FIN No. A1010
*Now at Los Alamos National Laboratory
i
ACKNOWLEDGEMENTS
This research was sponsored by the U.S. Nuclear
Regulatory Commission, Office of Nuclear Regulatory
Research. Mr. Ronald Feit was the contract monitor.
Information presented in this report could only
be made available through in-house cooperation of
Sandia's technical staff. Robert Bayette retrieved
and reformatted data tapes. Leo Klamerus provided all
the raw data records. CarlosTellez modified our data
reduction software for retrieving the 17 tray fire
test data.
ii
ABSTRACT
Electrical cables constitute a serious fire hazard fornuclear power plants because the plastic insulation materialis combustible and large quantities of cables are used inthe plants. Nuclear power plant fires often continue toburn in the presence of smoke, whereas building fires usuallyburn in the presence of clear air, since smoke escapes throughwindows and doors before descending to the fuel. Fire growthclassifications (realms) by the National Fire ProtectionAssociation (NFPA) thus may not be completely applicablefor fire hazards analyses of nuclear power plants.
Electrical cable fire tests have been conducted at theSandia Fire Research Facility in Albuquerque, New Mexico, inorder to evaluate cable tray fire safety criteria for theNuclear Regulatory Commission. A burn mode concept wasdeveloped in order to describe and classify the thermodynamicphenomena which occur in the presence of smoke and to comparethe fire growth and recession of different cable types underotherwise unchanged fire test conditions. The importance ofdeep seated fires in cable trays from the standpoint ofpropagation, detection, and suppression is emphasized. Thecable tray fire tests demonstrate that fire recession anddeep seated fires can result from a descending smoke layerand that reignition and secondary fire growth is possibleby readmission of fresh air.
iii/iv
CONTENTS
Page
I. * INTRODUCTION ..................................... 1
II. PHENOMENOLOGICAL DESCRIPTION OF CABLE FIRE
III. BURN MODE ANALYSIS OF FIRE RESULTS ............... 14
VII. APPENDIX A ....................................... 48
VIII. APPENDIX B ............... ..... .......... **** 55
V
LIST OF FIGURES
Figure Title Page
1 Horizontal Open Space Cable Tray Arrangement 7
2 Stacked Cable Trays for Fire Test 8
3 Burn Room Temperature vs. Time 11
4 Cable Temperature vs. Tray Height 13
5 Qualitative Reaction Rate Profile 16
6 Electrical Cable Cross Sections 18
7 Two Tray Cable Fire Test Arrangement 19
8 Temperature Profile, Acceptor Tray, Test 49 23
9 Temperature Profile, Donor Tray, Test 21 27
10 Temperature Profile, Acceptor Tray, Test 21 28
11 Temperature Profile, Donor Tray, Test 20 29
12 Temperature Profile, Acceptor Tray, Test 20 30
13 Temperature Profile, Donor Tray, Test 17 31
14 Temperature Profile, Acceptor Tray, Test 17 32
15 Fire Histories of Two-Tray Stacks, Pre-383Cables 38
16 Fire Histories of Two-Tray Stacks, IEEE-383Cable 39
Al Thermocouple and Calorimeter Placement for 49Single-Tray Tests
A2 Test Data Histories, Donor Tray, Test 20 51
A3 Test Data Histories, Acceptor Tray, Test 20 52
A4 Burn Room Temperature Histories, Test 20 53
A5 Burn Room Temperatures vs. Distance FromCeiling, Test 20 54
Bl Burn Room Geometry, Sandia Fire ResearchFacility 55
vi
LIST OF TABLES
Table 1
Table
Table
2
3
NFPA Classification of Fire GrowthPhenomena
Heat Release Activation Temperatures
Summary of Modal Life Fractions FromCable Fire Test Data
Match of Electrical Cable and BuildingFire Phenoma
Page
2
20
34
40Table 4
vii/viii
I. INTRODUCTION
A fire hazards analysis of Light Water Reactor (LWR)nuclear power plants requires a description of fire phenomenathat can be verified in independent fire tests. A burn modeclassification of fires both for rooms and cables representinitial efforts towards developing such phenomenologicaldescriptions.
The U.S. Nuclear Regulatory Commission has establishe•.guidance on fire hazards analysis for nuclear power plantswhich includes the following analysis tasks:
(a) Simulate fire phenomena from fire introduction,through its development, to propagation intoadjoining spaces;
(b) Confirm or modify principles of industrial fireprevention and control;
(c) Indicate the effect of postulated fires on safety-related plant areas, with and without activationof the automatic suppression system.
i
Heat release rates and room temperature in residential buildingshave been used most widely to quantitatively describe and'simulate fire phenomena. Heat release rates were chosensince they indicate the size of the fire, the rate of firegrowth and the time available for escape or suppression.Room temperature has been used to quantitatively describefire effects, since it indicates which rooms may block theescape of people as well as the heat loading of equipmentand structures in the vicinity of the fire.
Use of such quantitative measures of fire phenomena andfire effects showed that fire phenomena cannot be matched tothe fuel load per unit area or other parameters of the fi.ezone architecture because of the great variability of fires.Fires grow and recede, once flammable gases evolve, in a waythat is beyond the forecast capailities of present determin-istic compartment fire models.•" Measured heat release rato:show a greatyariability even under controlled experimental.conditions. Growth and recession of heat release rates havebeen observed for the same fuel material and fuel load. Thisindicates that chemistry and fuel ai:ea are riot the dominantfactors once flammable gases have avulvad.
-1-
The uncertainty of heat release rates is the main reasonwhy fire protection ýurrently represents an art and not anengineering method.(4 The National Fire Protection Association,NFPA, is trying to develop a probabilistic description of firesfor developing an engineering method of fire hazards analysisthat at least can be applied to residential buildings. Firephenomena categories called "realms" are defined thermodynam-ically to match observed fire test phenomena rather thanad hoc design events which describe failures of architecturalelements. Fire tests are then used to determine the probabilitydistributions of realm lifetime and realm transitions. Thefollowing thef9dynamic definitions are used to identifythese realms.' /
TABLE 1
NFPA Classification of Fire Growth Phenomena
Realm Phenomena Thermodynamic Definition
1 Pre-burning No flames
2 Sustained Ignition (including smoldering) hasburning occurred in the room of origin but
heat release rate does not exceed 2 kW.
3 Vigorous Heat release rate inside the room ofburning origin is between 2 and 50 kW, but the
upper peak room temperature is lessthan 1500C.
4 Interactive Average upper room temperature isburning between 150*C and 400*C; causing
secondary ignitions beyond the roomof origin but with heat release ofless than 2 kW.
5 Remote Average temperature in room ofburning origin is greater than 400*C; causing
secondary ignitions beyond the room oforigin with heat release of less than2 kW.
6 Full room Burning beyond the room of origininvolvement releasing 2 to 50 kW; secondary fires
have reached realm-3 conditions.
-2-
The present NFPA classification of fire growth ispractical for analyzing residential building codes, sincethe code already defines a class of similarly constructedfire zones (rooms, corridors) and since no special hazardsare anticipated in ordinary buildings. However, the archi-tecture and ventilation of LWR nuclear power plant firezones cover a much wider range of conditions than roomsof residential buildings. Nuclear power plant fire zonescontain electrical cable bundles, electrical equipment, andflammable liquid and materials that are not usually found inresidential or commercial buildings. LWR fires also aremore likely to burn in the presence of smoke. The applicationof a building fire hazards analysis to nuclear power plantfire zones is thus uncertain, even if the NFPA should succeedin developing a reliable engineering method of fire hazardsanalysis for buildings.
The uncertainty of eitending NFPA fire hazards analysisto nuclear power plant fire zones can be reduced by develop-ment of methods for simulating both the growth of fires andthe performance of fire protection systems in a reproduciblemanner. For fire prevention studies, ad hoc definitionsof events based on architectural design should be replacedby thermodynamic definitions of fire events or modes whichare common to most building fires and nuclear power plantfires. Such commonality will increase the reproducibilityof building and nuclear power plant fire simulation in twoways:
(a) The data base is vastly increased since the samethermodynamic processes are reflected in a myriadof different architectural designs, ventilationsystems, and fire suppression systems.
(b) The uncertainty of nuclear power plant fire phenomenais reduced to the statistical deviation betweendata segments that reflect the same thermodynamicprocess. These deviations should be much smallerthan the deviation between fire test data segmentsthat describe different thermodynamic processes,i.e., uncertainty is reduced.
In addition to developing thermodynamic definitions offire events or modes for buildings and nuclear power plants,it is possible to develop thermodynamic burn modes, or eventmodes from the analytical standpoint, for electrical cablebundles in nuclear power plants. It is important to be ableto classify and describe the thermodynamic phenomena whichoccur in cable tray fires because such fires represent aserious singular threat to the safety of all nuclear powerplants.
-3-
In a nuclear power plant, literally miles of electricalcables are required to provide electrical power and systemscontrols throughout the facility, and the insulationmaterials in the cables constitute a very large source ofcombustible fuel. The potential hazard of an electricalcable fire is probably best demonstrated by the fire atthe Browns Ferry nuclear plant in 1975 which causedextensive damage to the facility and force16 he plantto be shut down for a period of two years.
At the Sandia Fire Research Facility in Albuquerque,New Mexico, horizontal full scale electrical cable trayfire tests have been conducted in order to observe andevaluate different candidate cable types under realisticconditions. In order to classify the various thermodynamicphenomena observed and measured in these tests, a conceptof electrical cable burn modes has been developed.
The burn mode analysis of cable fires has been veryuseful in revealing the basic processes of fire growth incable trays. The cable tray tests have demonstrated theimportance of air introduced into smoke saturated hot gasenvironments on flame development and spreading as well asthe reignition of deep-seated fires by the readmission offresh air.
-4-
II. PHENOMENOLOGICAL DESCRIPTION OF CABLE FIRE GROWTH
a. General
Horizontal cable tray fire tests at Sandia, and ver-tical cable tray tests at Underwriter's Laboratory, bothshowed that jacket or insulation material may melt or formconsiderable char.
Four volatil ztion reactions were observed in verticalcable tray fires: ''
1. Pyrolysis - "Flaming was uniform over outer sur-face of the cable bundle as well as throughoutthe cable bundle. The cable region involved infire grew steadily for the duration of the test."
2. Smoldering Melt - "The jacket and/or insulationmaterial melted and coalesced into a large mass,and flaming occurred principally on the outersurface of the fused mass. Fire involvement wasvery dependent upon shape and position of the fusedmass within the cable tray."
3. Deep-Seated Combustion - "The jacket and/or in-sulation material formed considerable char, andflaming occurred principally on the outer surfaceof the cable bundle. Flaming was not continuousor uniform but rather occurred as sporadic burstsof fire. After the surface flaming subsided, aglowing cable region slowly progressed along thecables with sporadic flaming issuing from the region.The glowing region propagated for up to 4 hoursbefore extinguishing."
4. Interior Combustion - "Flaming was uniform overthe outer surface as well as throughout the cablebundle. The cable region involved in fire grewsteadily and was continuous. After the surfaceflaming subsided, a glowing region slowly progressedalong the cables with sporadic flaming issuing fromthis region."
Underwriter's Laboratory associated these classifica-tions with particular cable test descriptions. The abovefour labels were chosen by one of the authors (F. R. Krause)to relate vertical cable tray fire phenomena and horizontaltray fire phenomena.
-5-
The above observations, together with those made duringcable fire tests at Sandia, i14ustrate%,th teectriF I cablein trays constitutes a porous solid fuel thatmay eveo6pboth deep-seated fires and interior fuel temperatures in""excess of the flash point. Cellulose materials in buildingsrarely show the above combination of rapid volatilizationand deep-seated fire in the same fuel materials. Somereasonable doubt thus exists that fire growth character-istics of building fires are representative of electricalcable tray fires which can occur in nuclear power plants.
b. Stacked Horizontal Cable Trays
NRC Regulatory Guide 1.75 specifies minimum separationdistances for areas, where the fire damage potential islimited to fixtures oT faults internal to the electricalequipment or circuit.ý8' Minimum physical separation dis-tances are based on open ventilated cable trays, as wellas flame retardant cable insulation and jacket materials.These minimum separation distances are illustrated inFigure 1. They are designed to prevent fire propagationamong cable trays of one safety division and fire spreadbetween safety divisions. Sandfa.verified these separationcriteria in a 17 tray fire test at the Sandia FireResearch Facility that replicated the cable tray arrangementof Figure 1. The following discussion of cable fire growthphenomena is based on temperature records from this test.
Figure 2 shows two stacks of cable trays before the firetest. The seven lower trays are .875 feet apart from eachother and from the floor. They represent one safety divisionaccording to NRC guide 1.75. An eighth tray located fivefeet above tray 7 represents t minimum separation betweenindependent safety divisions-.ta Thermocouples are located alongthe center line of the north stack at the tray centers.
Smoke density prohibited visual flame observations ofthe upper trays. Insulation in the 4 inch diameter cableconduits under the trays turned to ash without flaming,and the conduits all showed electrical shorts above traylevel 3. At 69 minutes into the test the fire departmentextinguished the fire to safeguard the explosion ratedconstruction of the building. Manual discharge of 75gallons of water over a 15 minute period was needed tosuppress smoke production from the cable trays.
Figure 3 shows selected temperature time histories ofthe north stack of trays. The temperature profile of tray4, characterized by a broad peak, denotes the growth andrecession of a surface fire. Similar peaks were observed
-6-
NORTH STACK SOUTH STACK
8N 8S
5 FT. VERTICALSEPARATIONBETWEEN SAFETYDIVISIONS
7N
3 FT. HORIZANTALSEPARATIONBETWEEN SAFETY
IDIVISIONS
ONE TRAY ONLYIN THIS STACK6N
5N
4N
3N
10.5 INCHVERTICALSEPARATION 2NI-BETWEENTRAYS
kM99999dCABLE
CABLE TRAY
1N
IGNITION TRAY
8 INCH HORIZONTAL SEPARATION BETWEEN TRAYS
tntr
j18 FLOOR LEVEL
I8
Figure 1. Horizontal Open Space Cable Tray Arrangement
-7-
I
Figure 2. Stacked Cable Trays for Fire Test
-8-
i i•'
for trays IN, 2N, 3N, and 6N which are not shown in Figure3. These peaks illustrate the propagation of fire throughthe north stack. Films of the fire show that this propaga-tion was caused by an unanticipated "leap frog" phenomena,and not by flames from the tray below.
After a surface fire is sustained in tray number one,a fire ball forms at the bottom of the third tray up thestack. This fire ball grows and subsequently touchesdown on the second tray. The fire ball is then replacedby a surface fire, that starts at the top surface of thetray and not at the flame exposed bottom surface. Thewhole propagation sequence then repeats itself for nexthigher level of the stack. The surface fires grow slightlyin area with higher peak fuel temperatures at each higherlevel. Based on these observations, which show that firesare not propagated from tray to tray due to direct flameexposure, we feel that Regulatory Guide 1.75 separationrequirements are adequate to prohibit such propagation.However, physical separation by itself obviously does notnecessarily inhibit other mechanisms of fire propagation.
Tray 5N shows an extreme temperature rise at 32minutes and this temperature exceeded typical peak valuesof 1500 to 1600°F recorded in all other trays. Peak tem-peratures could not be recorded since the thermocouplesstopped operating around 2300*F. The most likely explana-tion for the sharp temperature rise is a sudden flash of afuel vapor engulfing tray 5.
Flashover is a common means of fire propagation amongphysically separated fuel elements in a room. In the caseof the above cable tray fire, however, flashover occurredtoo late to play such a role. Ceiling tray 8N alreadyhad reached flame temperatures of 1200*F before the flash Noccurred. Oxygen starvation by engulfing fuel vapor prob-ably prohibited surface fire development in tray 5N.
Tray 7N behaved abnormally by not developing a peak -Vtemperature that is characteristic for the growth andrecession of a surface fire. One explanation is that ddescending smoke and/or fuel vapor accumulation preventedia surface fire by oxygen starvation. Even the flashoverof tray SN did not succeed in igniting a surface fire ontray 7N. Tray 7N temperatures, however, rose slowly andsteadily. Water spray halted the temperature rise, butdid not cool the tray. The temperature appeared to riseagain after the water spray stopped. The most likelyexplanation for this is. a deep-seated fire. Observations
-9-
for trays 1N, 2N, 3N, and 6N which are not shown in Figure3. These peaks illustrate the propagation of fire throughthe north stack. Films of the fire show that this propaga-tion was caused by an unanticipated "lleap frog" phenomena,and not by flames from the tray below.
After a surface fire is sustained in tray number one,a fire ball forms at the bottom of the third tray up thestack. This fire ball grows and subsequently touches downon the second tray. The fire ball is then replaced by asurface fire, that starts at the top surface of the trayand not at the flame exposed bottom surface. The wholepropagation sequence then repeats itself for next higherlevel of the stack. The-surface fires grow slightly inarea with higher peak fuel temperatures at each higher level.Based on these observations, which show-that fires are notpropagated from tray to tray due to direct flame exposure,we feel that Regulatory Guide 1.75 separation requirementsare adequate to prohibit such propagation. However, physicalseparation by itself obviously does not necessarily inhibitother mechanisms of fire propagation.
Tray 5N1 shows an extreme temperature rise at 32 minutesand this temperature exceeded typical peak values of 1500 to1600*F recorded in all other trays. Peak temperatures couldnot be recorded since the thermocouples stopped operatingaround 2300*F. The most likely explanation for the sharptemperature rise is a sudden flash of a fuel vapor engulfingtray 5.
Flashover is a common means of fire propagation amongphysically separated fuel elements in a room. In the caseof the above cable tray fire, however, flashover occurredtoo late to play such a role. Ceiling tray 8N alreadyhad reached flame temperatures of 1200*F before the flashoccurred. oxygen starvation by engulfing fuel vapor prob-ably prohibited surface Lire development in tray 5N1.
Tray 7N1 behaved abnormally by not developing a peaktemperature that is characteristic for the growth and re-cession of a surface fire. 'one explanation is that descendingsmoke and/or fuel vapor accumulation prevented a surfacefire by oxygen starvation. Even the flashover of tray 5Ndid not succeed in igniting a surface fire on tray 7N.Tray 7N temperatures, however, rose slowly and steadily.Water spray halted the temperature rise, but did not coolthe tray. The temperature appeared to rise again afterthe water spray stopped. The most likely explanation forthis is a deep-seated fire. Observations indicate that:(1) this fire was not terminated with a single water dischargeand (2) smoke is not necessary to indicate a deep-seatedfire.
-10-
3000
2500
LLq
0 2000wcc
D< 1500w7-W 1000
500-
0 0
wcr.
zLIMIT 0
or
ESTIMATED
~LEGEND= TRAY 5N 6'-4" OFF FLOORA TRAY 4N 5'-2" OFF FLOOR
OVE 0 TRAY 7N 8'-9" OFF FLOOR
o TRAY 8N 14'-1" OFF FLOOR
10 20 30 40 50 60 70 80
TIME, MIN.Figure 3. Burn Room Temperatures vs. Time
A comparison of test data for trays 5N and 7N indicatesthat fuel temperature transients provide an important basisfor distinguishing between deep-seated and surface fires.(See Figure 3). Figure 4 shows three vertical temperaturestratifications along the north stack center line: a) justbefore the hydrocarbon flash (32 minutes), b) after theflashover (35 minutes), and c) when post flashover coolingslowed down (48 minutes). The data shows that the flashwas created by the interaction of multiple fires that isnot present in single tray tests and may not occur in twotray tests. At 32 minutes tray 5N had reached a tempera-ture just below the autoignition threshold. The absenceof thermal transients in this temperature range shows thatfuel vapor was accumulating inside the tray. Burn roomceiling temperatures in excess of 1000*F indicate a layerof burned gas above tray SN that may have helped to confinea hydrocarbon cloud about the tray. The cloud was heatedby the hot trays 4 and 6 from above and below. Flashoveroccurred a short time later, when the temperature reachedthe autoignition threshold.
Conditions between 32 and 35 minutes can be compared tothe NFPA building classifications. The average room temperatureis 700*F (371*C). The heat release of fire may be estimatedcrudely by multiplying the air mass in the room (334 lb)with the specific heat of air (.240 BTU/lb *F) and theceiling temperature rise rate (80*F/min). This gives aheat release of 113 kW. According to the NFPA classifica-tions in Table 1, the fire phenomena should be characterizedby interactive burning, (heat release in excess of 50 kW)that is close to remote burning, i.e., average room tem-peratures greater than 400*C.
The NFPA classification thus indicates that a suddentransition to remote burning could occur. Such a transitiondid occur, however, remote burning of tray 8N was alreadyfully developed some 12 minutes earlier. See Figure 3.
The NFPA classification, therefore, is uncertain forthe type of cable fire that burns in the presence of smoke.Remote burning of cables may occur at temperatures and heatrates which would, in the absence of smoke, only supportinteractive burning. In Chapter V we will describe anextension of the NFPA classification to fires that burnin the presence of smoke.
-12-
3000
2500
0.2000w
1500
w0-W 1000
500
0
w
zLIMIT 0
C
ESTIMATEDLEGEND
FLAýSH TRAY SN 6'-4" OFF FLOOR
OVER A TRAY 4N 5'-2" OFF FLOOR* TRAY 7N 8'-9" OFF FLOOR
o1 TRAY 8N 14'-1" OFF FLOOR
0 10 20 30 40 50 60 70 80
TIME, MIN.Figure 3. Burn Room Temperatures vs. Time
A comparison of test data for trays 5N and 7N indicatesthat fuel temperature transients provide an important basisfor distinguishing between deep-seated and surface fires.(See Figure 3). Figure 4 shows three vertical temperaturestratifications along the north stack center line: a) justbefore the hydrocarbon flash (32 minutes), b) after theflashover (35 minutes), and c) when post flashover coolingslowed down (48 minutes). The data shows that the flashwas created by the interaction of multiple fires that isnot present in single tray tests and may not occur in twotray tests. At 32 minutes tray SN had reached a tempera-ture just below the autoignition threshold. The absenceof thermal transients in this temperature range shows thatfuel vapor was accumulating inside the tray. Burn roomceiling temperatures in excess of 1000'F indicate a layerof burned gas above tray 5N that may have helped to confinea hydrocarbon cloud about the tray. The cloud was heatedby the hot trays 4 and 6 from above and below. Flashoveroccurred a short time later, when the temperature reachedthe autoignition threshold.
Conditions between 32 and 35 minutes can be compared tothe NFPA building classifications. The average room temperatureis 700'F (371C). The heat release of fire may be estimatedcrudely by multiplying the air mass in the room (334 lb)with the specific heat of air (.240 BTU/lb *F) and theceiling temperature rise rate (80*F/min). This gives aheat release of 113 kW. According to the NFPA classifica-tions in Table 1, the fire phenomena should be characterizedby interactive burning, (heat release in excess of 50 kW)that is close to remote burning, i.e., average room tem-peratures greater than 400*C.
The NFPA classification thus indicates that a suddentransition to remote burning could occur. Such a transitiondid occur, however, remote burning of tray 8N was alreadyfully developed some 12 minutes earlier. See Figure 3.
The NFPA classification, therefore, is uncertain forthe type of cable fire that burns in the presence of smoke.Remote burning of cables may occur at temperatures and heatrates which would, in the absence of smoke, only supportinteractive burning. In Chapter V we will describe anextension of the NFPA classification to fires that burnin the presence of smoke.
-12-
3200 -
CABLE TEMPERATURE2800 - vs
TRAY HEIGHT
2400 ~LEGEND
O-- 0 32 MIN. RUN TIMEx 0--0 35 MIN. RUN TIME
.0 2000 -,** A -'-- 48 MIN. RUN TIME
w/| I NOTE: NORTH STACK1600 TRAY 5N HEATED -CENTERLINE
FROM ABOVE I TEMPERATURESwj AND BELOW -
w 1200 AI -
BURNED GASREGION
800 =- - - -AGASEVOLUTION
400 -
00 2 4 6 8 10 12 14 16
CABLE CENTER DISTANCE ABOVE FLOOR (ft)
Figure 4. Cable' Temperature vs. Tray Height
III. BURN MODE ANALYSIS OF FIRE TEST RESULTS
During the course of an electrical cable fire a numberof different thermodynamic phenomena are produced. They canbe observed, measured, recorded, and classified. These phe-nomena are the result of the generation of flammable gasesdue to the initial heating of the cables and combustion ofthe gases. Heat release due to combustion causes a furthertemperature rise and the process continues to accelerate.The availability of oxygen, the cable temperature, and thetype of combustible material in the cables are all fundamentalin determining the limits for fire growth.
The different tray heating phenomena which take placeduring the lifetime of a cable fire can be considered eventsfrom an analytical standpoint, and event tree analysis canbe performed to obtain burning characteristics for differentcable types. The thermodynamic events which denote suddentransitions between cable tray temperatures can be classifiedand these are called burn modes.
Burn modes are studied for a generic classification offire growth phenomena that can be used for comparing buildingfires, electrical equipment fires, and flammable liquid fires.The following classification was restricted to the use oftemperature measurements, since temperature is the only para-meter that has been recorded in full-scale compartment firetests of both building materials and electric equipment.
The purpose of a burn mode classification is to sub-divide a raw data record into segments such that the segmentsof one class reflect a generic chemical process. Recognitionof volatilization and combustion reactions requires a temper-ature signature of individual reactions. Thermogravimetriclaboratory tests use the concept of a weight loss activationtemperature to characterize volatilization of polymers. Inthe case of flammable liquids, this weight loss activationtemperature is simply the boiling point. We have introducedthe concept of a heat release activation temperature tocharacterize not only volatilization reactions but alsocombustion reactions. Using heat release in lieu of weightloss is more in line with the above NFPA definition of firephenomena and permits correlation of reactions with thelocation of individual thermocouples. Such space resolutionis absent in global measurements of fire weight loss.
The introduction of heat release activation temperaturesmakes it possible to divide the burn mode classification ofraw data records into three basic tasks:
1. Retrieval of heat release activation temperatures,
-14-
2. Grouping of raw data records into burn mode dedicateddata segments,
3. Analysis of classified data segments.
A preliminary approach to these basic test data reductiontasks is described in the remainder of this chapter.
III.1 Heat Release Activation Temperatures
Following the qualitative discussion of the 17 tray firetest, we use the rate of fuel temperature rise above the tem-perature of the ambient atmosphere as a qualitative indicatorof local heat release. Heat release activation temperaturesare retrieved by plotting the temperature rise rate againstthe fuel temperature as shown in Figure 5. Experience withthe 17 tray fire test data also showed that temperaturerise rates in excess of 160*F/min indicate flaming combustionnear the tray. The peak level of the temperature rise ratecan therefore be used to separate surface fires above theinstrumented fuel section from other heat release reactions.Any temperature excursion found is then traced back to thepreceding minimum of the temperature rise rate. The associatedfuel temperatures are then identified with a heat releaseactivation temperature according to the following screeningcriteria:
1. Flammable gas evolution
The minimum temperature rise rate indicates areversal of a sharp cooling trend which wascaused by shutting down the burner and whichdropped through the 1600 F/min threshold fromabove.
2. Autoignition
The minimum temperature rise rate indicates areversal of a cooling trend that was independentfrom the burn shut down. The subsequent excursionis steep and reaches peak temperature rise ratesthat are comparable to burner shut down values.
The third type of heat release activation temperatureis associated with the maximum of the fuel temperature.This maximum is interpreted as char oxidation. T. E. Harmathyintroduced the concept of fuel v9talilization by a slowlypropagating char oxidation front " while analyzing over250 full-scale building fires.
-15-
C
E
Ll
ccz01=
LI-
-J
a
600
400
200
0
-200
-400
-600
FLAMMABLE GASEVOLUTION AUTO
I IGNITION
BURNER OUT
LOWER LIMIT OF FUELSVAPOR COMBUSTION
MAXIMUM FUEL TEMP
ACCEPTOR TRAY, TEST 50
0 250 500 750 1000 1250 1500 1750 2000
CABLE CENTER TEMP (OF)
Figure 5. Qualitative Reaction Rate Profile
Flammable gas evolution temperatures, autoignitiontemperatures and char oxidation temperatures were retrievedfrom a series of 21 special effects electrical cable tests.The selection of these tests was based on the conditionthat an electrical short occurred in the burning tray. Thetests differed in material composition and cable design asshown in Figure 6. The tests also differed in using fourdifferent cable arrangements:
1. Single tff•)with three conductor cables in centerof room, duplicating geometry for tray lN ofFigure 1.
2. Two tray stack with three conductor cables in bottomtray and single condVMYr cables in top tray incenter of open room, duplicating tray geometrylN and 2N of Figure 1.
3. Single tray as in (1) above in corner of a metalenclosure.1
4. Two tray s z as in (2) above in corner of metalenclosure.tMI
The single and two-tray arrangements were covered by apermanent barrier on top to crudely simulate the radiationexchange within a tray stack as in the 17 tray fire (seeFigure 7).
Retrieving heat release activation temperatures fromthese special effects tests produced the results given inTable 2. The following chemical classification was used:
These classes represent an increasing amount of flame retardantmaterials. However, each class still includes both singleand 3-conductor cables and both single tray and two traytest configurations. Classes 1 and 2 also include cornertest arrangements.
-17-
-POLYETHYLENE
DIAMETER .450" 0(7 ,,,, FILLER0 ,COPPER
POLYVINYL CHLORIDE
PRE-383 3 CONDUCTOR CABLE
DIAMETEF
FIBER GL
CROSS LINKED POLYETHYLENE
R .155" ( tCO• COPPER
SILICON GLASS
IEEE-383 SINGLE CONDUCTOR CABLE
ASS-_~ ;S LINKED POLYETHYLENE
-SILICON GLASS
DIAMETER .450"
COPPER
CROSS LINKED POLYETHYLENE
IEEE-383 3 CONDUCTOR CABLE
CROSS LINKED POLYETHYLENE
FIBERGLASS0 SILICON GLASS
COPPER 0)-FIRE RETARDANT COATING
CROSS LINKED POLYETHYLENE DIAMETER .450"
IEEE-383 CABLE WITH FIRE RETARDANT COATING
Figure 6. Electrical Cable Cross Sections
-18-
UPPER BARRIER
I'0
BARRIER(REMOVED
AFTERBURNER
SHUTDOWN)
DONOR LV
UNDER DONOR TRAY2
Figure 7. Two Tray Cable Fire Test Arrangement
The results indicate that all cable tray burns includeat least two different combustion reactions which are triggeredat different autoignition temperatures. The difference betweenthe first autoignition temperatures for the different cabletypes may be statistically significant and indicates thatthe low temperature combustion reactions are affected byflame retardant components of the cable surface. The effectof the IEEE qualification is counter-intuitive; it lowersthe autoignition temperature. The high temperature combustionreactions on pre-383 and 383 qualified cables are ignitedat very similar temperatures that exceed most flammable gasflash points. Coating an IEEE-qualified cable raises bothautoignition temperatures but lowers the char oxidationtemperature threshold.
TABLE 2
Heat Release Activation Temperatures
Maximum Flammable CharCont. Use Gas Evolution 1st Auto 2nd Auto Oxidation
Flammability handbooks for plastics(14 ) list the followingweight loss activation temperatures.
Commercial grade polymer
Polyvinylcloride, PVC
High density polyethylene,PE
Pyrolysis inAbsence ofoxygen ('F)
392 to 572
635 to 842
Ignition inPresence of
Flame (*F)
536 to 604
740 to 811
-20-
Comparing these values with Table 2 suggests that flammablegas evolution of pre-383 cables is associated with decompo-sition of the PVC jacket. Flammable gas evolution of IEEE-qualified cables probably reflects volatilization of aplasticizer or fire retardant. First stage autoignitionof both 383 and pre-383 cables probably denotes the combus-tion of polyethelene decomposition products. The secondautoignitg5 1 threshold is characteristic of spontaneousignition of surface heated cellulose materials (930to 1200*F) and plastics (1200"F).
The results of Table 2 support our belief that heatrelease activation temperatures are common to a diverserange of fuel element construction, fuel element arrange-ments, and fire zone architecture. The generic aspect ofheat release activation temperatures is demonstrated by thefollowing observations:
1
(a) A crude chemical classification of the combustiblematerial allowed reproducible heat release actt-vation temperatures with 4 to 9 cable tray burnsin spite of significant variations within each-class of element construction (1 versus 3 conductors),element arrangement (single tray, two trays), andarchitecture (open room, corner of enclosure).
(b) The results of special effects tests (chemicalclass 2) agree closely with full scale replicationof LWR fire zones that used 7 tray stacks.
(c) First autoignition temperatures agree with the hand-book values of polythylene ignition in the absenceof flames, and second autoignition temperaturesagree with handbook values of spontaneous surfaceignitions.
111.2 Burn Mode Classification
The qualitative discussion of cable fire growth indicatedtwo fundamentally different types of combustion reactions:flammable vapor combustion and char oxidation. The ignitionof the first type is controlled by the temperature of thefire zone atmosphere and the ignition of the second by theinterior temperature of the fuel. Fuel vapor reactions canbe subdivided into additional generic subclasses accordingto observed phenomena which occur at simultaneously recordedtemperatures of the ambient atmosphere and interior fueltemperatures.
-21-
Computerized raw data records from Sandia's cable firetests were used for a preliminary burn mode classificationbased on temperature alone. Additional information on thesetests and the test data is given in Appendix A. The burnroom geometry is illustrated in Appendix B.
Thermocouples were cemented to the cable insulation andnot the copper conductor. The measured temperatures are thusuncertain, whenever:
(a) The insulation has burned away from the originalthermocouple junction, or
(b) Heat deformation of cables has caused deviationfrom the vertical separation distance betweenthermocouple locations in the cable bundle.
Consequently, depending/on test conditions, the thermo-couples measured the temperature of the gas near the cable,a char surface, or the temperature of the copper conductor.This uncertainty permits a general classification of testresults, but the data may not be sufficiently definitivefor deterministic modeling of heat transfer near a reactionzone.
The preliminary classifications of the raw data recordsare derived from two thermocouple positions. The firstscreening temperature, called fuel internal temperature, Tf,represents the temperature in the cable tray center betweenthe two ribbon burners. It was calculated by taking thearithmetic mean of thermocouples 2 and 4 (see Figure Al).The second screening temperature, T , called fuel surfacetemperature, represents the measurea output of thermocouple3. This thermocouple was cemented to one cable at the topof the filled tray facing downwards.
Burn modes were identified by subdividing the above twoindependent temperature records into intervals that are de-termined by the heat release activation temperature thresh-olds (see Table 2). This postulation of burn modes is illus-trated graphically in Figure 8. The shaded bars in Figure 8represent the experimental uncertainty of the heat release'activation temperatures. The blank inner spaces are labeledburn modep according to the phenomenological descriptions ofvolatilization and combustion reactions that were discussedin the last section.
-22-
PRE-IEEE-383 CABLE THRESHOLDS
I
U-
0~Z..'
wI-w0U.ccCI,
-JwUI.
1800
1600
1400
1200
1000
800
600
400
200
00 200 400 600 800 1000 1200 1400 1600 1800
FUEL INTERNAL TEMP (OF)
Figure 8. Temperature Profile, AcceDtor
Electrical engineering handbooks list a continuous usetemperature that assures electrical resistance and dielectricproperties of insulatin materials for 20,000 to 100,000hours of operation. Maximum continuous run temperatureratings are:
PVC, flexible, filled: 130 to 150°F
PE, .91 to .925 g/cm3 : 180 to 212*F
PE, .92 to .94 g/cm3 : 220 to 250*F
PE, .941 to .965 g/cm3 : 250OF
PE, cross linked 275°F
These continuous use temperatures thus give the lowesttemperature at which volatilization can occur. The rangebetween continuous use and pyrolysis temperature denotesaccelerated aging. The range between pyrolysis and lowerautoignition temperature denotes pyrolysis of highly volatilefuel components that are characterized by low energy chemicalbonds.
Volatili $4on reaction rates are known to follow theArrhenius law•L" i.e., they grow exponentially with tem-perature. Volatilization reactions, then, are not independentfrom combustion reactions. The coupling of volatilizationand combustion reactions is typical for nonflaming combustion.The temperature range between the first and second autoignitiontemperature thresholds will sustain combustion of flammablegases in the absence of flames if we assume that the secondignition threshold denotes the spontaneous ignition temperatureof gases from a cable tray surface (see discussion followingTable 2). In Figure 8, the area between these two ignitionthreshold tmperatures denotes nonflaming combustion or smoldering.Friedman?1, describes smoldering as follows:
"It may propagate by a 'front' or 'wave' whichinvolves air oxidation generally combined withpyrolysis. It may be self-sustaining, or itmay require assistance from an adjacent energysource. It occurs in bulk porous materialwhich may be in contact with a heat sink aslong as the porous material is thicker thanthe critical value. Flexible polyurethanefoams, even if fire retardant, can smolderin a self-sustained mode."
-24-
Summarizing common observations of smoldering, Friedmanalso notes that smoldering generates an aerosol of condensed,high molecular weight species formed by pyrolysis in or nearthe smoldering zone, which is combustible. However no vis-ible products may emerge until the smoldering zone reachesthe surface because of self-absorption of products withinthe bed.
Self-sustained smoldering can create a local hot zoneinside a porous material, while the surface temperature isstill cool. This is a deep-seA*~_i fire. We hereby proposeto define a deep-seated fire as a subclass of smolderingwith the following conditions (see Figure 8):
Fuel interior temperature is between thefuel vapor and surface autoignitiontemperatures of the f~iel
and /
Fuel surface temperature is below the upperor surface autoignition temperature.
This is a quantitative definition of deep-seated fires thatcan be used to monitor deep-seated fires on line. Figure 8illustrates such monitoring. Realizing that data points are30 seconds apart, we find that a deep-seated fire developedin the acceptor tray of corner test 49 (Pre-383 cable) acouple of minutes after the burner was shut down. Thedeep-seated fire lasted about 1.5 minutes and then starteda surface .fire. We have thus clearly identified thesegment of the test 49 data record, where a deep-seatedfire occurred.
A subdivision of the smoldering zone into self-sustained combustion (deep-seated fire) and externallyheated combustion (henceforth called smoldering) is justi-fied, in our opinion, due to the extreme importance of thedeep-seated fire phenomena relative to fire protection. Adeep-seated fire is very difficult to suppress since firesuppressing agents cannot easily get to the seat of the fire,and it is also difficult to detect since combustion is pri-marily under the cooler surface. The NFPA thus requires(NFPA-12A) or strongly recommends (NFPA 12, NFPA 15) theverification of suppression agent discharge requirements bytest. However, a formal a priori definition of deep-seatedfire so far does not exist and no suppression verificationtest method is presently available. We hope that the conceptof burn modes will help fill this gap.
-25-
Fuel elements can sustain flames, once the fuel tem-perature exceeds the upper autoignition threshold. Wedistinguish two burn modes. Interior gas combustion denotesa fuel with a surface temperature too cool to allow externalsurface fires. External surface fires should occur if bothfuel interior and fuel surface temperature are between theupper (spontaneous) autoignition temperature and the charoxidation temperature. Fire balls are characterized by hotcombustible gases with temperatures in excess of the upperignition threshold but with a fuel surface too cool forignition of surface flames.
To achieve fuel interior temperatures beyond the charoxidation temperature requires intense external heating.Figures 3 and 8 show two such observations. The discus-sion of Figure 3 associated flashover with such extensiveheating. The flashover region is designated in Figure S.
We did not observe excursions of the char oxidationtemperatures. Labeling the remaining area in Figure 8 as"deflagration" is thus speculative. our main motivationwas that temperature of unburned gas in excess of the charoxidation temperature can probably be achieved only bycompression heating of the gas and this is the definitionof a deflagration.
111.3 Results From Test Data
All of the data used to classify and describe the thermo-dynamic phenomena associated with electrical cable fires inthis report came from the thermocouple records of the 21tests mentioned previously. From these data we were ableto determine the thermodynamic burn classifications or modesgiven in the previous section of this report, and their re-spective temperature limits for the three cable types testedas indicated in Table 2. The thermodynamic history of each ofthe cable fire tests is given by temperature profiles-asillustrated in Figures 9 through 14. These are plots'-of thecable bundle internal temperatures (fuel internal temperatures)vs. fuel surface temperatures imposed on a burn mode matrix foreach of the three types of cables. The data points are at 30second time intervals.
-26-
PRE-IEEE-383 CABLE THRESHOLDS
1800
DEFLAGRATION
1600
-- • / FLAMES (TV)
1400FIRE BALL SURFACE FIRE I
a 1200
1000 1iW SMOLDERING>
o0
LLocc 8000:) c r 1 -J
- PYROLYSIS-J 600 -wUCL INTERIOR GAS
400 RATED o COMBUSTION
AGING200 -USET
-TEMPSI 1I
0 200 400 600 800 1000 1200 1400 1600 1800
FUEL INTERNAL TEMP (OF)
Figure 9. Temperature Profile, Donor Tray, Test 21
-27-
PRE-IEEE-383 CABLE THRESHOLDS
1800
DEFLAGRATION1600
1400-1.40 - FIRE BALL
a. 1200 -- FIREw
S1000 0Cw UUJ - SMOLDERING .•i>
800 ZmA
600 -PYROLYSIS•-J 600 -- :
" • INTERIOR GAS400 _CELERATE < COMBUSTION
0w
--AGING oe c
- USE
0 200 400 600 800 1000 1200 1400 1600 1800
FUEL INTERNAL TEMP (°F)
Figure 10. Temperature Profile, Acceptor Tray, Test 21
-28-
IEEE-383 CABLE THRESHOLDS
1800 DEFLAGRATION
1600 --
1400 FIRE BALL FSURFACE FIRE
0 1200CL 10
1000 l"w SMOLDERING 0
800
0 0 PYROLYSIS oIS600 - f
u.400 -ACCELERATED t INTERIOR GAS
CON I
200 USE A" : I
TEMPS
0 200 400 600 800 1000 1200 1400 1600 1800
FUEL INTERNAL TEMP (OF)
Figure 11. Temperature Profile, Donor Tray, Test 20
Figure 13. Temperature Profile, Donor Tray, Test 17
-31-
"C" COATED IEEE-383 CABLE THRESHOLDS
U-%.0
09w
co-JwDI
1800
1600
1400
1200
1000
800
600
400
200
n
DEFLAGRATION
FIRE BALL SURFACE FIRE I
~I
I
0 200 400 600 800 1000 1200 1400 1600 1800
FUEL INTERNAL TEMP (OF)
Figure 14. Temperature Profile , Acceptor Tray, Test 17
-32-
Each of the temperature test profiles is characterizedby a curve which rises and falls as the respective surfaceand internal temperatures increase and cool. Each of theprofiles, therefore, has a heating cycle and a cooling cycle.The burn modes which are activated during a fire test canbe determined readily from the temperature profile plots,and the length of time which each mode is activated canbe determined by counting the data points (which are separatedby 30 second time intervals). A summary of modal life fractionsfrom fire test data for those types of cables tested isgiven in Table 3.
The modal lifetime fraction data given in Table 3 includesresults from all of the 21 special effects electrical cablefire tests. These tests include the four different cablearrangements and the three different chemical classifications(cable types). The data has been separated into thesethree classifications in order to reduce the uncertaintyof the results as indicated by the standard deviations.However, because the number of tests is so low, the datawas not separated according to the different cable trayarrangements.
In any case it is clear that there is a large spreadof data insofar as modal lifetimes for cable types is con-cerned. This indicates that fires of the same fuel con-figurations and chemical composition grown and recede ina non-reproducible manner. Nevertheless, the activationof certain burn modes provides valuable information forunderstanding and controlling critical burn modes thatinitiate fire growth, and this is addressed in the nextsection of this report.
-33-
TABLE 3
Summary of Modal Life Fractions From CableFire Test Data
Pre-IEEE-383 Cables
BURN MODES
I I IDeep Seated I II Surface I nterior Ga•l ITest I Pyrolysis I Smoldering I Fire I Fire Ball II Fire I Combustion I Deflagration I Flashover I Totd]
Cycles C and H refer to cooling and heating cycles, respectivelyTest numbers with D and A pertain to donor and acceptor trays, respectivelyThe symbol a refers to the standard deviation
IV. BURN MODE DESCRIPTION OF FIRE GROWTH
Burn modes describe local reactions in the vicinity ofthe thermocouples. They do not, by themselves, describe firegrowth. Such growth can only be described by the interactionof physically separated burn modes.
We propose to extend the current NFPA classification offire growth stages (realms) from building fires to cable fires,by correlating heat release rate and ceiling temperature signa-tures with burn mode signatures.
A preliminary burn mode classification of fire growthphenomena is given in Table 4. The potential value of sucha classification is illustrated by applying it to selectedtwo-stack-tray fires. The results are given in Figures 15and 16. Both figures clearly show recession followed bygrowth and recession of cable fires. They also demonstratethat the duration of individual fire growth stages (realms)and the sequence of such stages is dominated by plant-specific,special effects such as use of fire retardant coatings andthe proximity of a corner. Clearly, fire growth historyis very fire-zone and fuel-specific, although the underlyingburn mode classification is not.
-37-
NFPAREALMS
5 FLASH OVER
4
3
2
SMOKE DESCENDSTO TOP TRAY
(RUN 49)
OPEN SPACE (RUN 21)
1
BURNER 10 20 30OUT BURN TIME, mrin
40
Figure 15. Fire Histories of Two-Tray Stacks, Pre 383 Cables
TABLE 3 (Continued)
Summary of Modal Life Fractions From CableFire Test Data
IEEE-383 Cables
BURN MODES
I I Ie~ep Seated III surface lInterior Gasl I ITest I Pyrolysis I Smoldering I Fire I Fire Ball II Fire I Combustion I Deflagration I Flashover TotdI
Summary of Modal Life Fractions From CableFire Test Data
"C" Coated IEEE-383 Cables
BURN MODES
ýeatedI I1 Surface linterior CI Fire Ball II Fire I Combustic
H C H C H C
.024 .11 .061 -
-- .12 .078
.2 11 -. .12----------
.49 ---- -----------
~as I~n
0aDef lagration
C H
IIFlashover I Total
C H
1.0
1.0
1.0
1.0
.21 .15 .15 .11 .21 .49 .072 .11 .086 .099
o025 .021.13 .092 .067 .042 .11 0.0 .048 0.0
Cycles C and H refer to cooling and heating cycles, respectivelyTest numbers with D or A suffix pertain to donor and acceptor trays,
The symbol a refers to the standard deviationrespectively
NFPAREALMS
'0
40=ABURNER r_ "-OUT BURN TIME, min
Fire Histories of Two-Tray Stacks, IEEE-38 3 CablesFigure 16.
TABLE 4
Match of Electrical Cable and BuildingFire Phenomena
NFPA Definition Extension to Cable Fires
Realm Name
I Preburning
2 SustainedBurning
3 VigorousBurning
4 InteractiveBurning
5 RemoteBurning
Fuel decomposition in the absenceof a surface fire by any one ofthe following reactions:Pyrolysis, smoldering, deep-seatedfire, interior gas combustion.
Single external reaction (surfacefire or fire ball) with decompo-sition linked to one fuel package.
Single external reaction withmultiple fuel package decomposition.
Multiple external reactions withambient temperature below upperautoignition threshold.Multiple external reaction withambient temperature above upperautoignition threshold.
In spite of the uncertainty of the overall fire growthcycle, some common aspects of fire growth might exist forbuilding and cable tray fires. We hypothesize that thefollowing generic aspects dominate the observed fire growthcycles:
Fire recession is caused by descending smoke.
Fire growth is caused by readmission of freshair to deep-seated fires.
Based on this hypothesis, Figures 15 and 16 provide thefollowing insight into cable fire growth.
Pre-383 cables develop smoke rapidly and the smoke istrapped under the upper tray barrier at the time the burneris shut off. The trapped smoke blocks the oxygen supplyand makes fires on both the upper and lower trays recedeor become deep-seated. The rate of smoke release is thendiminished such that updraft from the hot fuel suffices
-40-
to clear the trapped smoke. The associated readmission offresh air ignites intense surface fires and smoke releaserates are increased enough to repeat the cycle.
In the center of the burn room, smoke descends suffi-ciently fast to prevent development of self-sustained fires(Run 21). Readmission of fresh air does not reignite surfacefires in the absence of deep-seated fires. In the cornertest smoke is trapped and hovers near the acceptor traysurface, and small flames (Run 49) flicker on and off accordingto short repetition of the smoke descent and readmissioncycle until the deep-seated fire is terminated by the overallcooling of the acceptor tray.
IEEE qualified cables behave differently, as shown inFigure 16. The open ladder tray in the center of a room(Run 20) shows the same recession/growth/recession as theequivalent pre-383 cable (Run 21). However, when the stackis placed in a corner (Run 4 ) the use of flame retardantmaterials combined with a heavy smoke release is sufficientto prevent a surface fire" and the burner generates insteada deep-seated fire. This fire reignites as soon as theburner exhaust no longer blocks the updraft of fresh air.The associated surface fire is maintained by the deep-seatedfire until the deep-seated fire is terminated by tray cooling.Cooling periods are very similar for 383 (Run 48) and pre-383(Run 49), as botli deep-seated fires terminate approximately25 minutes aftei burner shutdown.
Coated cable in the open ladder tray provides a dra-matically different fire (Run 17 versus Run 20). Readmis-sion of fresh air at burner shutdown does not immediatelyrekindle the deep-seated donor fire. The deep-seated firehas to burn for another 3 minutes before flames developabove the donor tray. The subsequent donor fire releasessmoke at a lower rate, such that little smoke is trappedtemporarily at the upper barrier. A deep-seated fire developsin the upper tray only after the smoke blanket has descendedfrom the ceiling to the tray. This fire does not ignitesince the smoke blanket prevents the readmission of freshair. The donor fire is terminated 2 minutes later by thestill descending smoke. However, contrary to the uncoatedtray, the deep-seated acceptor fire persists to the endof the run.
The above interpretations of Figures 15 and 16 illus-trate how the observed variety of fire growth and reces-sion can be explained by only two common factors namely,smoke descent and admission of fresh air. The abovedescriptions of possible events were deliberately designed
-41-
to fit the observed facts to the hypothesis. Other explana-tions are conceivable and burn mode analysis of many moretests is needed to confirm or ammend the hypothesis.
The above hypothesis, based on the effects of smoke andfresh air on fire growth, can be used to design and evaluatefire confinement and fire suppression requirements includingthe role of ventilation systems, if substantiated by experi-ments. A fire protection strategy could be to prohibit thedevelopment of deep-seated fires that last one minute orlonger. Fire growth would then be prevented in all electricalequipment configurations that meet Regulatory Guide 1.75requirements for physical separation.
Some additional evidence is already available from post-test observation of char formation and from current unpublishedfire suppression tests. Figures 15 and 16 indicate that therate of smoke development is highest in pre-383 cables,somewhat lower in 383 qualified cables, and considerablylower in coated cables. It is reasonable to expect thatthe diminished rate of smoke release manifests itself ina higher rate of char formation. Pre-383 cables shouldthen show the least amount of char on the burned sectionof cable trays. IEEE-383 cable should show somewhat moreand coated cables should show much more char. This isconfirmed by all post-fire cable tray inspections.
Evidence from Figure 3 and Figure 16 also shows thatdeep-seated fires, which are due to the descending smokeblanket, last much longer than deep-seated fires that aregenerated by temporary fire ball touchdown. The durationof a deep-seated fire is related to the area and the durationof burned gas exposure. This is demonstrated by the Haloncable tray fire suppression tests in which a strong deep-seatedfire was generated by holding thligscending smoke blanketat the top of the acceptor tray. This deep-seated firegrew into the strongest surface fire ever observed as soonas fresh air was readmitted after 10 minutes. A 4 minutesoak with 6% Halon 1301 was not sufficient to terminatethis fire. A ten minute Halon soak was required to preventreignition upon fresh air admission. The suppression testsprovide additional evidence that cable fire propagation canbe started by readmission of air to deep-seated fires andthat deep-seated fires are sustained by a hovering layerof burned gas.
-42-
V. SUMMARY AND CONCLUSIONS
Quantitative temperature records from 21 fire tests ofhorizontal cable trays were reduced to thermodynamicallydefined burn modes in order to develop a physical classi-fication of fire phenomena which meets NRC Regulatory Guide1.170 requirements for fire hazards analysis. This database is neither statistically significant nor extensive enoughto cover the wide range of architecture, ventilation, andfire protection design parameters encountered in LWR plants.The tests do, however, provide important insight as to howa suitable classification of fire phenomena might be developed,especially for electrical cables.
Burn modes describe local volatilization and combustionreactions which have been observed in many porous fuel andflammable liquid fires. A preliminary classification methodwas developed which identifies such reactions using only thetime history of fuel internal and surface temperatures.
The classification of raw data records from 21 specialeffects cable fire tests into data segments, which reflectone burn mode each, confirmed flame aspects of the genericburn mode reactions with independent observations of flamesand post-test inspection of charred surfaces. There wasgood correlation between the phenomena indicated by burn modeanalysis with flame and char observations from tests forwhich TV films were available. These tests covered threedifferent types of fire retardant materials, two differentcable combinations and four different tray stack geometriesas well as a full scale 17 tray replication of an LWR firezone. This partial verification of burn modes encouragesus to think that our preliminary test record classificationsmight indeed give generic descriptions of fire phenomenathat are applicable for a wider range of architectural firezone parameters as well as for a wider range of ventilationand fire suppression system operations. Further verificationof the generic nature of burn modes is warranted.
The following conclusions are tentative, since theyare based on the 21 test data base which is consideredmeager as mentioned above. They are nevertheless given toillustrate the insight which burn mode analysis can providefor confirmation or modification of fire protectionrequirements.
1. The cable fire burn modes reflect volatilizationreactions and oxygen consumption modes, that havebeen observed previously in full scale compartmentfire tests. This commonality should provide a
-43-
technical basis for confirming or modifying prin-ciples of industrial fire prevention and controlfor LWR plants.
2. Duration of burn modes and transitions between burnmodes depend on fuel chemistry, fuel arrangementand smoke descent. A reproducible simulation ofentire fire life cycles is unlikely since manydifferent patterns of fire growth and recessioncan develop from the same fuel configuration oncethe fuel is volatized.
3. Burn mode analysis provides a new physical definitionof deep-seated fires that allows monitoring on-line.The screening method of this paper illustrates theprinciples for such monitoring, and also revealsthat both propagation and reignition of cable firesare frequently preceded by a deep-seated fire inexcess of 1 minute duration. The temperature criteriafor deep-seated fires thus represent a direct indicatorof fire growth potential that has been derived froma cable fire replication test and 21 associatedspecial effects tests.
4. Burn mode analysis of fire confinement and firesuppression verification tests is needed to confirmthat prevention of deep-seated fires will preventfire propagation between fuel elements that meetRegulatory Guide 1.75 separation requirements. Withthis information NRC fire protection requirementscould possibly be verified in special effects testswithout replicating full-scale LWR fire zones and/orprotection system operations.
5. Deep-seated fires were generated in the electricalcable tests by a hovering layer of burned gas. Inhorizontal cable trays such hovering was caused bya descending fire ball and/or by a descending smokeblanket. Consideration should thus be given to in-specting existing porous fuel arrangements for previouslyunknown fire propagation hazards associated withtrapping of burned gas.
6. The use of fire retardant materials (IEEE-383 cablequalifications, cable tray coatings) tend to increasethe duration of deep-seated cable fires. A reasonabledoubt thus exists that industrial experience withIEEE cable qualification and fire retardant coatingsapplies fully to multiple. cable tray arrangements.
-44-
The use of fire retardant materials does significantlyreduce the probability of self-sustained surfacefires, but associated longer-lasting deep-seatedfires might increase the probability of propagatingsurface fires once started.
-45-
VI. REFERENCES
1. NRC Regulatory Guide 1.120: Fire Protection Guidelinesfor Nuclear Power Plants, November 1977.
2. Krause, F. R., Dungan, K. W., DiNeuro, R. D., LWR FireProtection Design Verification by ThermodynamicsLimitations, SAND80-1020, to be published.
3. Coulbert, C. D., Energy Release Criteria for EnclosureFire Hazards Analysis - Part I, Fire Technology, Vol. 13,No. 3, p. 173, 1977.
4. National Fire Protection Association, System Methodologiesand Some Applications, NFPA Symposium, February 1979,University of Maryland. Proceedings in publication.
5. Berlin, G. N., Connelly, E. M., Fahey, R., Russet, D. P.,Swartz, J. A., Fire Safety Systems Analysis for Resi-dential Occupancies, NFPA Final Report on HUD ContractH-2316, NFPA - Boston, July 1978.
6. Pryor, Andrew J., "Browns Ferry Revisited," Fire Journal,May 1977, p. 85.
7. Christian, W. T., private communications based onUnderwriters Laboratories, Inc., File NC555-2, p.4,issued May 8, 1978.
8. Regulatory Guide 1.75, Physical Independence ofElectrical Systems, U.S. Atomic Energy Commission,February 1974.
9. Klamerus, L. J., A Preliminary Report on the FireProtection Research Program, July 6, 1977 Test,SAND 77-1424, October 1977.
10. Harmathy, T. Z., Mechanism of Burning of FullyDeveloped Compartment Fires, Combustion and Flame,Vol. 31 (1978), pp. 265-273.
11. Klamerus, L. J., A Preliminary Report on FireProtection Research Program Fire Retardant CoatingTests, SAND75-0518, March 1978.
12. Klamerus, L. J., A Preliminary Report on the FireProtection Research Program Fire Barrier and FireRetardent Coating Tests,
-46-
13. Klamerus, L. J., Fire Protection Research ProgramCorner Effects Tests, SAND79-0966, Sandia NationalLaboratories, Albuquerque, NM, December 1979.
14. Hilado, C. T., "Flammability Handbook for Plastics"Technomic Publishing Co., Inc., Westport, Connecticut,1974.
15. Friedman, R., Ignition and Burning of Solids in FireStandards and Safety, ASTM STP 614, A. F. Robertson,Ed., American Society for Testing and Materials, 1977,pp. 91-111.
16. Harper, C. A., Handbook of Plastics and Elastomers,Chapter 2, Electrical Design Properties of Plasticsand Elastomers, McGraw-Hill, 1975.
17. Tissot, B. P., Welton, D. H., Petroleum Formationand Occurrence - A New Approach to Oil and GasExploration, Springer, NY, NY, 1978.
18. Klamerus, L. J., private communications.
-47-
VII. APPENDIX A
Notes On Fire Test Data
Thermocouple placements for single tray tests are shownin Figure Al. In the case of two tray tests the acceptortray has the same instrumentation as shown for the donortray. The observed interior cable bundle temperatureswere interpolated to the position of the stack centerline by averaging the corresponding thermocouple readingson each side of the center line.
Cables with Coating C: Tests 2, 17, 35Cables with Coating G: Tests 27, 29
IEEE 383 qualified cables have crosslinked polyethyleneinsulation and jackets. Pre-383 cables had linear poly-ethylene insulation with PVC jackets. Coated cables referto crosslinked polyethylene insulation and jackets plusfire retardant coatings of 1/8 inch (3.2 mm). The bottomtrays were filled with a three conductor cable of 1300 feetlength, and the acceptor trays were filled with one conductorcable of 6800 feet length. The inside dimensions of thecable tray were 12' x 18" x 4". Loaded trays weighedbetween 190 and 220 pounds. The weight of an empty trayis 20 pounds.
-48-
THERMOCOUPLE PLACEMENT
4- EAST
r1 616 r iNO.j INO. NO.3 I1*4 *10
ILJ I
TOP VIEWOF CABLE TRAY
D AND NO. 10 UNDER BARRIER
I
I 8' 14 I
It 12' -I
NO. 2 0 NO. 3 *NO. 4
NO. 0
BURNEREAST
aNO. 1nBURNER
WEST
NO. 0 AND NO. 1 IN BURNER FLAME
NO. 2 AND NO. 4 IN MIDDLE OF CABLEBUNDLE OVER BURNERS
NO. 3 IN TOP LAYER OF CABLESUNDER COATING (IF PRESENT)
CALORIMETER PLACEMENT
BARRIER
NO. 7
NORTH-
8w 9.5'NO. 6 NO. 5
3.715"F7CABLE TRAY
Figure Al. Thermocouple and Calorimeter Placementfor Single-Tray Tests
-49-
Figures A2 and A3 show representative fire test datahistories. Burner shutdown is indicated by a descendingexposure fire temperature. The start of test time has beenarbitrarily defined as the time when the burner temperaturecrosses the 900*F level from above. An electrical shortis considered to exist when the current to ground exceeds100 milliAmpere. Burn room temperature profiles as a functionof distance from the ceiling are illustrated in Figure A4for test 20. The extent of the burned gas layer can beinferred from a two layer linear fit of the burn roomtemperature stratification as shown in Figure A5 for t = 13min, the time of maximum ceiling temperature.
-50-
1500 4500 BURNER
u. -0 CABLE CENTER1250 A CABLE TOP 375
w BARRIERCURRENT (mA)
j,, 1000 Co
ul E- 750 225 E
z z
WJ 500 150 cc1-z00
250 75
0 0
0 4 8 12 16 20 24 28 32 36
TIME, min
Figure A2. Test Data Histories, Donor Tray, Test 20
1500
L-0 -" .
4 5 0
w 1250
4-
" LEGEND '375
0: 10003700 -0 BURNER
0 0 CABLE CENTER 3
I-- CABLE TOP -.
wj 750 -BARRIER
Cl
zEaCURRENT (mA) 2
ccw-
225 E
500
zw z
050 cc
25250
15-
0
75
0 4 8 12 16 20 24 28 32 36
TIME 8 mn
0
Figure A3. Test Data HiSttories, Acceptor Tray, Test 20
LEGENDaI•PI li 1.'
50 UI I ArIA I-IVM tIjILINIU - 13000 1 FEET
" 0 3 FEET40 A 5 FEET 1100
z * 7 FEET 0W N 9 FEET 9030 14 FEET 90090
020 ccca'CC
Clow - 500w 10 0
wu 3000 AMBIENT TEMP-•) 47°F
-10 1 1 100OV 4 8 12 16 20 24 28 32 36
TIME, min
Figure A4. Burn Room Temperature Histories, Test 20
50
ILU-
zu~j
w0
0-wI--
40
30
20
10
LEGEND
o 7.00 MIN. (BURNER OUT)o 8.50 MIN. (ELECTRIC SHORT, DONOR)A 9.50 MIN. (ELECTRIC SHORT, ACCEPTOR)A 11.50 MIN. (MAX ACCEPTOR TEMP)o 13.00 MIN. (MAX CEILING TEMP)* 22.00 MIN. (DECOMPOSITION STOPS)
AMBIENT TEMP = 47* F
0
-100 2 4 6 8' 10 12 14 16 18
DISTANCE FROM CEILING (ft)
Figure A5. Burn Room Temperatures vs. Distance fromCeiling, Test 20
VIII. APPENDIX B
Burn Room Geometry
STEELSTRUCTUREFOR CABLETRAYS
SECTION A- A
DOOROPENING
56'
A
-S
Figure BI. Burn Room Geometry, Sandia Fire Research Facility
-55-
DISTRIBUTION:
U.S. Nuclear Regulatory Commission (369)Distribution Contractor (CDSI)7300 Pearl StreetBethesda, MD 20014
344 copies for RP25 copies for NTIS
Author selected distribution - 11 copies(List available from author)
3652 V. L. Duke4400 A. W. Snyder4440 G. R. Otey4442 W. A. Von Riesemann4442 L. J. Klamerus4442 D. A. Dube4442 L. D. Lambert4443 D. A. Dahigren4444 W. H. Schmidt (5)8144 M. A. pound3141 L. J. Erickson (5)3151 W. L. Garner (3)
