AIVC 11899

CH-99-1-3

Predicting the Position of the Smoke Layer Interface Height Using NFPA 928 Calculation Methods and a CFO Fire Model

William N. Brooks, P.E.

ABSTRACT

NFPA Standard 92B presents computational methods for determining the position of a smoke layer in a large-volume space. Although NFPA 92B is a guide to smoke management

design, the methods have been adopted, with certain modifi­cations, by model building codes and are mandated for use in atriums and large-volume spaces. This paper makes use of a recently developed CFD fire model to assess the NFPA 92B calculation methods. A total of 13 simulated tests were conducted. Results suggest that the NFPA 92B Equation 9 method may not predict the fastest filling of an enclosure within the range of aspect ratios provided in NFPA 92B.

INTRODUCTION

Prediction of smoke layer interface heights in large­

volume spaces is essential to certain building fire-protection and life-safety strategies. Predicted positions of the smoke

layer interface height can be applied to egress design or to

property conservation.

Figure 1 depicts the development of the fire plume, the formation of the smoke layer as the fire plume radiates

outward to the surrounding walls, and the smoke layer descent. Figure 2 illustrates a cross-sectional view of an

atrium with a descending smoke layer. The terms used are

defined as follows:

H z ZcR

Zt

= highest point of smoke accumulation

= height of smoke layer interface

= critical (design) height of smoke layer

interface

= limiting elevation (defined as the height of the

luminous flame)

A- PLUME RISING TO CEILING

- -___,{_ - -

·-

L.: -, � -- -, \ - ' 'tt - / \- . \-

8- SMOKE LA YER FORMS

\::.. I -1

C - SMOKE LAYER DESCENT

Figure 1 Smoke filling of atrium space: (a) plume rises to ceiling; (b) smoke layer forms; (c) smoke layer descends.

William N. Brooks is fire protection department manager, Peter F. Loftus Division, Eichleay Engineers, Inc., Pittsburgh, Pa.

THIS PREPRINT IS FOR DISCUSSION PURPOSES ONLY, FOR INCLUSION IN ASHRAE TRANSACTIONS 1999, V. 105, Pt. 1. Not to be reprinted in whole or In part without written permission of the American Society of Heating, Refrigerating and Air-Conditioning Engineera, Inc., 1791 Tullie Circle, NE, Atlanta, GA 30329. Opinions, findings, conclusions, or recommendations expressed in this paper are those of the auttior(s) and do not necessat11y reflect the views of ASH RAE. Written questions and comments regarding this paper should be received at ASHRAE no later than February 13, 1999.

H

A:ARJ.A OF ATRIUM

,, . ,

Figure 2 H-Highest point of 'smoke accumulation; 1.; ·

: z_:_Height of s�oke layer' interface; Zcr-.. n Ctitical (design) height of smoke·· layer '\

'· i t� ·, interface;J ·zrL7miting elevation (height"of : u, '< : llJmirlbus flame)•

· '' · c ' 1 ;;, :· · . t� ' t '. ···. (;_.;It. •. If.:,-.� i, '

A I ' .I .. '; , "i• ' ' '• ' ' ltoriZontal cross-sectional area' of atrium

Aspect ratio = A!H2 /·. :. ·:::: ·;_. 7 � b :>=?- t · - ii ' --- '.� t .,,;t,.1M

· f..TP'PA. <:st�ndarcl 101 (Lifej?Jei ·:rcoae) ·1aifrl'\Iir1996 BOCA National Build1��g Coile (Jio't� {ij96)\1ke .. th'e NFPA.. S_ta1,1dard 92B metliodolog;y in � mannex; that link� p�e position of the smoke layer at a particular tf me to the requirement for a smoke management system (see BOCA Code, Section 922.2, and NPP.A l 01, .Sectiop. 6-2A1; ). The llnJform·Building Code (ICBO 199?)J;l9es not �til��e-�tj.rne c<;msideration in its large-volume desi� criteria. I.nstead � sms>�e exhau�t systel!l mu t be capable of1milntainfog an interface height 3".1 m (10 ft) above the high'e t walking sllrface'in the"'sin6k'e ione (see ICBO Code, Seotibn 905;5.2. l). Therefore,itbe time-of smoke layer descent is a critical factor in NFPA l@l!:and the BOCA Codei design but is not relevant to ICBQ Code design. , A �tevi�,�s

.P.�per.,�y t?e �u�or,@roo��:!1��7) showed th t the two desWJ .meth9ds m the BOCA COde produce · . : ·1 v r · ; r i� • •·· •· results that vary signi ficantly. In fact, ari exampTe iri the ·19'93 Comment a',: (B'dcA i �9�) ��s us�cfto ill��trateiffi·ka sm��e manageme�t s;.�i��-·i� re9�i;�d ·v:h in 1the c"aic'Uiati�n �ethod

� ' f . ..... ,: • . ) • . f t • .. � . 1ft.J,:1. for regular spaces is used, while a smoke management sy. teJP is not required when the calculation for irregular ·spac�· 'is used. . :::;.d

A large eddy simulation, fornp\itaifoti'a! fluid° <iyniirrt'ics (LES. CFD) model:(McGra.ttan �t,al.�1199�) is usell.'in an attempt to reconcile:the differenc.:e_s among a.number qf c_alc.:u-lation methods. " . ; .', ,,

-

2

METHODOLOGY

Four calculation methods are used to predict the position of the moke llyer interface in a total of 13�imulated test cases usi,� steaQ.Y,;:�tate heat release rate fires.

• --Method t:__._:_NFPA 92B, Equation 9 (calculation method for regular spaces described in the 1996 BOCA Code) Method 2-NFPA 92B, Equation A-26 Method 3-NEPA.92B, Equations 14 and 15, adjusted for s�oke layer temperature

• Method 4--Large eddy simulation CFD model

Table t ·is a: 'US! of the individual test cases-. With the J . • exception of Test-No. 10, the selected aspect ratio are consis-tent with NFPA.92B and the 199(iJ;�oCA Co4.e�J:Iea.t rel_(!ase rates include the range of fire sizes 'that are mandated for use by the BOCA Code (2110, 4640, kW) and tlie ICBO Code (5275 kW):-Selected areas represent th� author's·opinion as to o} where the design methods will be most widely utilized.

Test No. IO.represents a more· slender aspect ratio that is. not contemplated by_ NpP� 92B or the 19,96 _BOCA Code. It h�.�.�ee� inclu�� for sensitivitr P�Tos��· , ,, All of the cases assume ,a fl.� hotjzontal ceiling and. a uniform square hoi;izontal cross-section'al area y.-ith vertical sideS: r J. f - • ,rr .. . !. l• �·

; .. rt. .,. It,..,

CALCULATION METHODS J ·:·t '.�- 1 ... �·C;.i.HI ···, .• . r ' "'

·,�

... · ..

Methods 2 and 3 :can be defined as zone models, which· predict. well·defiried srnpke laytm positions !atl any' given time and·distirict-differentiatinn,between the smoke lay.en.and the clear<allr b�low. Smoke layer interface positions are assun;ied

• ,, • . ,· • 1 ; r.-.. �·i TABLE 1a' J '-' ' - · "

' ' , :-.::i .... �.: ··�selec,t��fT�lC�ses i· ,,,, · ""'

,'_':nn -: . . ·.- ' .t�leh� 'J:e�t �oi\<!ii�o�s Sf U�I�]) • . , , JJJ'. - ) ' • ·) ' . • . .. � ..<i l _j - • • '1 . • ' I

,, !1 rr: - ,,.,r "iHeatRelcase "Test No;, Aspect Ratio r!Area'(m2): Height (m),,: Rate (kW)

I: ''1 'l :,, '.l• '1 '10''..i' 11. ·-1'9'303 Y; '''·3'0'.5 ' "' 2110

." . .'2':"" •".00.9,','. , " 9#Q�" 1q1:� · .. •I 2110 I,

,-3, .· .. .. o.�lr' , or:r?��o . . . ,lOl .. 6,,, 464Q ,,0.

···�· . . . t ""i:; :'0;9 ;; ... '1 ..• • 921 J '. - � 3f.'1 ... . . ' • • 1 I ' • I . :• -� • 9i7:y ''') ·''3i.'i · � � '-' I , 0,9 I , ---•. , ., , , 1 •

2t10

' i 464� �' \

4640 ·-\_ . '

, ;. S>l.. •· ,, :: :14:.:r;1), · <' <4688'.•-· ·1'>l8:3·0.i'. ,, ·_4400: ,, ·!:_ . 9' '·' · r •Ii 14"'"J u• · "" 9391,,i!· i · 25'�9 JJ, '"· ·4400 ur·

10 0.5 929 43. l r· ·J' 4640

14 'l � �· 1'.le?OO.�J fiS�3':1 1,�7:)8 ., ' 11

12 0.9 �.672 43,1 9693 0.9 . .. :,r167:Z. .43_:1 . � r ! 544 Je .. 13

CH-99-1-3

Test No.

1 2 3 4 5 6 7

'i:.·8 .·.r

, . 9 !' 10

11 12,, L�·

TABLE 1b Sel�cted Test Conditions .1._

(Selected Test Conditions [IP Uni�]) I

H�il't Release Aspect Ratio Area (ft2) Height (ft) Rate (Btu/sec)

10 100,000 100 2000 0.9 100,000 333 2000 0.9 ,100,000 }3.p 4400 10 10,000 31.5 ;2000

J 0.9 10,000 .105,, 2000 0.9 10,000 105 4400 14 10,000 27 , 4400

. . .)4 .�p.900 60 .. .. � '·,4170 ,.!J ;:: � ... 14 L 100,000 85 [ ;-·,!· 4170

JI ·o.5 10�000 141.5 4400 14 �80,000 141.5 14,900

. .. ........ 0.9, ·- 18,000 141.5 9,200 ( ··O�-��t{· 18,000 i·Ml,� · 520 .

. • · I "'./l�·;. , , , ' • ('.,,. to be level and to fill atnum: volumes much like an mverted versio_h of wate� �l�n'g a '.b'ocket. '. ·

. . ') 1-,; , • I ) "'• j 1 ( 0 j ll; '. Method 1 is also a zone model, with the excep'tion th�t it does not define the position of the smoke layer interface'. Instead, the results of this calculation ffeeJ�pc;I are,tq . be viewed as the first indication of smoke, or �- oegin.i}ing of the transi­ti(m between clean air andceontaminated all:. (NIWA 92B, Figures Hhrough4.and:accompanyingidemnitions). ltdoes not assess the .quality :of:the:environment at·the fransi.tion point; only that this is the:point where· the transition begins to occut.

Method 4 is the LES CfP: m<:>QeL, relying on a much more precise calculation t�?l�.J�������o:e�i�JiJ� t!te characteristics of the atrium enviJ-.on ,qien.t: �i&" inedi0:�1 ��1� to d�e the posi-cl101 J ;:{:!!' . 11 � c .. .,, ..: 9 , �. J 'F..,.h. , ·• ti.Q.n \YQI}!"�. tern �r��e �l!!IE_ge_ 9c,cur§_ as� a res��� of p_�u�e in.ter{l.c;won wi�h the ambient air.

1 ITher�.is norsta'n'dilra to define .the -position of the �moke fay�r inteffa�: KD.1iriiber Ofpintii7f a.-aa�fiiiis op16:°\ITiB-a , .. r .. ,..ft".'� • &;'\ I .

;1

numQtr of metlro are·· proposed ·(ehow 1995; Soderbom, ( :;:t , . ' ' I' I' • 1-992; He et aL- 19�8) Experimeiital .. prograrri.s use .combina-tio.lLQf 1yj�t,1.al ...9.P.s�atjo_ns, Uf�W.Q..�oup!�. tree�, �d ·}�ght �bscutation mbasurement equipment to record the p.osition of the sl}ioke lay"fr"at�y �veri time (Yamanii.and Tan'afca 1985). ;nrer�fo;e,' me' re-sutt'S:-oftbtn:aicµlations will �ary �n-that they may be predicting.different-pneriome'na Th�t is, Method l is' p:r�dkti,ng '.'.first in_�lJ�aJj.gl_l _of�smqk�,·� M;etho_ds �.i111d 13 111".f! predi'et:ing si.Jllilar· smoke la�r' positions,, and Method 4 is much· like visuaJ o6�rvations t WOWcfbe made in an ej<.per-imemal��ro-gt:nn;·: -�-· '· - �- · · '

. ' ·

; . : .... _ ' . -

� ; ,i'

Methot( 1-NFPA 9_�!3. _E�u�tjor:t� _

The -posi�on of the smoi(J l�yer-intert'�c�, Z, is predicted at any time mi�ng Jbe''following eq��tjon�_.i_ ·

·· 1

CH-99-1'.a

ZIH = 1.11 - 0.28 ln [(tQ113!H413)/(A/H2)] (1)

where Z = height from the floor to the smoke interface (m),

= time for the interface to extend to Z (s). H = height from floor to flat ceiling (m), Q = steady-state heaf release rate (kW), A = horizontal cross-sectional area of the space being

filled (m2).

Equation 1 is a correlation, representing a "curve fit" of a limited number of fire tests {NFPA 1995).

Method 1 does not readily lend itself to a hand-calculated solutio.n for t, given a particular value of Z/H. It is for this reason that it would most often be used in a "pass-fail" manner, as presented by the BOCA Code to determine if the smoke layer interface has reached the•critical height (design objective level) in a given period of time. (The 1996 BOCA National Building Code and the 1997 NFPALife Safety Code define this period.of time as 1200:1-'<!Conds,) ?.:1 ·:'-''

The usefulness of, this method is presruited by the 1996 BOCA National Building Code as a sin.gle�pqint "test" to determine if the smokie layer interfac�-.has reached a given point (Z/H) in a defined time period. Ho,wever, with some manipulation, the use of the method can be expanded to graph-ically plotth;e,P.�ti,t\O�\?r{ the ]ayer at 81}� �iven ?me. '"

Method 2-NFPA 928, Equation A.;26

. · The. poi;jtion.of the. s�oke 1ay1er int�rf.:!!c.cr, ?i �s J?redi,cted \ �·}

• . ,\

• t ... ·,f • • '1,{... �... .... ' . - • . , · ,_:.. . • 1 • at.an}'..llme usu�� ijiefollovieng equation:. . . . . . ,, .· • • • • • ' ! ,• t ' ' ''\l . ·=

. . .. ·�

Zffl ,:,,:tf +[2kv(tQll3tfl'V·3Y/g(AfW)]}�312 (,.• "- (2) '{'�..!·:�:Ji':� '1:1:�:: • . �,i ;l� ' .: ;:.i , l . J ·;

whereJ,; . _ >".· .r.:.:).:J · . . 2·'. , ;::.Ji: .. , ;,T ·1J · •

z " . ;', 'k::·height from theTflfuor to.the smoke iliterface (m), ·

r : ·' ;: ;::: �e for the ihierfabe to extend to' Z'(s), '' .. � .. ,co

i{ve ·� height frounio6; tbflat ceilirig (m), ·1' .r 11' ·• �: ' J ' ' L < , ' • :.. : i � o O • " 1 � ' ' : j Q, =;= �tef,lCIY,�,state he�t re�ease ra�e (k\Y), : ri A., c ·= -ihorizontal cross-sectional atea of tbe spac¢. being

..... '1filled/in2);.'-'.0" • ' ; • ;jl ' •

� :::: entrainment oonstant (use 0.064 m413t[s-kW113]).- •

. · ;;1 �tho�"g\E<J'ilati°'Jh ' 1 i� �as� o? a.��b�t �f. a ��be� of. ffi:�. �ts, Equ�ti�n �:· acco�ding tb,�A 92B; is a theoreticfill!'. based equ'Ationrfor : inoke fillin� "aen.ved usi.I)g the law� of c?�seNhtio·� �·of'hiass .an.d" 'en·erg'j t?, detenninf th� additional vol�� bbing supplied to' the up#er la)'.'erf,. (NFPA 1995, Appen-·u� ··., ·· :J.I - ·1· ;�· ·· 11· ' - · ,. , . ,,,. aiX" i\-3-6,L..2). • • I I• • • • •• ,•••

� � .. '': a ' ..; !, •I j/ • 'Ji

Method a ....... NFPA 928, Equations 14 and 15, ��jµsted for �woke :Layf!r Jempera�ure

,,

Basic Calculation; Method. '!fhis method relies on an integrative apptdach that utilizes the following NFPA 92B plume mass flow rate correlations: . '·

3

Lm = 0.071Q/3z513 + 0.0018Qc (NFPA 92B, Eq. 14) (3) :1 i, J'

Vl:'.?ere ''li'' • . , '· ·'

m = mass flow rate in plume at height Z (kg/s )·and· :-Qc = convective portion of heat release rate (kW), where

and

(4) .I ,.

m = 0.032�?5z (NFPA 92B, Eq. 15). (5)

Equations 3 and 5 define a mass rate of smoke:]Jroduction. Equation 3 is to be.used above the limiting elevation, calcu­lated as follows:

z1 = 0.l66Q/15 (NFPA 92B, Eq. 13). jA (6)

Equation 5 is to be used at or below the limiting elevatiolj: Calculating 'tile layer position is not straightforward as is

the case with Methoos 1 and 2. Method 3 requires the follow-ing iterative calculation procedure: ·

1. CalcUlate smoke mass production (m) using Equation 3 based on the ceiling height (H'),

2. Convert smoke mas.s production (m) to smoke volume (V) using the folloWing;equation:

, -. ., ... .:.. . . ll = m/p - (7) ·�

where v p

3.

4.

5.

6.

= volumetric rate of smoke production (m3/s) and = density o('rnoke0(Rg7rri3f .:- '. '1' . ; · in.;:. ·;•,

·� l-1 ,. -:.: . , .. . ,..

Distribute the'silloke v<lilinie uit'lfoimly'u'nder the ceiling surface area (A) by dividing the smoke volume produced in the first time period (1 second) by the area of the atrium. This value becomes the incremental depth of the smoke layer. Subtract the incremental smoke layer depth ( dZ) from the ceiling height (H). The new height becomes the smoke layer interface height Z. Calculate smoke mass production based on the smoke layer height (Z). Repeat steps 2 through 5 until the smoke layer interface height descends to the limiting flame height (z1). At this point, calculate the smoke mass production (m) using Equa­tion 5.

Smoke LayertTemperature Correction '. NFPA 92B provides no guidance regarding the ti�e otthis

calculation method. The 1996 BOCA Code Commentary. iilustrat� the''iterative na"iure of the calculation but fai1s 'it;;· adjust the volume of smoke production based on a smoke density correction. As reportc;:d previously(Brooks 1997), tl'fe• BOCA Code biregular ceiling calculation method is.based on smoke density at21°C (70°F). Therefore, the resulting c<iku­lations .will represent the·, slowest filling· of the volume (i.e.:, !

smoke with1 the greatest density win occupy the smallest' volum� per kg). , ' ' ,•r •:Jf

In order to determine the 'sensitivity of the calcultRioni: method tQ .. smoke density, the' smoke layer temperature is assumed·fa1be:uniform and equal to the average temperature: of the plume centerline. . . 1,

Adjusting the layer tempera:rure will result in ·a decrease' in smoke density and will increas'e -th� volume ·elf sm()i(e produced. The increased smoke volume will mean faster fill-ingiof the atrium;r , � ,··-·

For eachtest case, an average plume temperature is calcu-lated using the folldwing method:-;.:�·· · ·,_

fl I • �1 <" 'tt) • • • 0 l • ,.;(1 1. Calculate the plume centerlirie temperature at ilie ceiling of

the atriuni''.inctat the fire's liin1ting elivilion (Eq\iitfio�: 6)�' ' ;L ,, • i· ., (!•,-. .... �-- . • I

T CL= TA+ 25(Q/13/(Z-Zo)513) (Klote 1994) (8) • /'' . 1C . I' -

whe�e, _ . ;-1 "· . .,, ·: ·'· • .. • ·,

T cl '=· �plume1cepferline temperaturei(l'�y, . . : •''' . ·1 :'•

-:r

For the atrium heights and heat release rates considered in this analysis, the virtual origin Will be assumed to be at floor level.

,fi ·'-''' 2. Calculate average plume/centerline temperature:

where ' .

..

TAVE = average pilll'Jl centerline temperature (°C),

TH = plume cei;it€rurie temperature at height H,

(9)

T1 = plume,,C�nterline temperature at limiting elevation. .,.� , � -,r.u

3. Calculate smoke density, using the average plume temper-The calculation method results in instantaneous distribu- -� ,i;ltute: 1•

tion of smoke under the entire ceiling surface, without regard_,:.. . .-- ··· / , ,..,, .. · . , . •

to transport time from the fire to the.ceiling.or wit1toUi·regard .,.. ,..,..-v . P = �lRT . to the time required for th"e smok�ovement horizontally under the ceiling surface. The method also fails to �ccount for . any horizontal entrainment that may occur as the smoke·:, moves ,radially outward from ,.the ppint,, where. �e plume centerline impinges on'"the ceiling.

·

4

wl,l.en( p = absolute pressure (101 ;325 Pa), R = ,g�. con�tant (J/kg K),

1F"\ . '. > - . . • ,\ - . '· - __)\'i� -�

T = ap�olµt1< temperawre.(K)F 27,3 .+.J'AvE·

(10)

CH-99•1-3

4. The value for Equation 10 is inserted in "Equation 7 and , , ,remaiJ;is constant for the entire calculation period. ' )_ ··'

Method 4-Large Eddy Simulation CFO Modef tt . : "

The large eddy simulation CFO model (McGrattan et al. 1997) has been developed in conjunction with National Fire Protection Research .Foundation at the National Institute of Stap9ards and Technology. The model is seen as a valuable tool to redµi.:e the overall costs offire1experiments by devel­oping computational methods· to predict fire phenomena.' In addition to predicting and evaluating experimental results, the mode� �s able to assist in pliµming experimental pro'grams.

Fgr:;this analysis, the LES CFO model is,li>eing.used in plaq� of 13 fire experiments. While the validation ofthe model . is still in process, its use appeared to present .an interesting opportunity �o simµla,te fireJ:mnditions in order, to compare the rt<�µl� from the thre� ,zone models to the LES CFD model.

The LES CFO model predicted the position of the smoke layer interface in each of the 13 cases. Since there is no· math-· ematically defineci methocl to determine the smo� fayer inter­face position, the LES CFD: modeler visually observed the:, s��J-�t�� q1!ffig ,��-the atrium �nclosures. In, a,.�1vn

.se, the i

observatiop pt the !Jmoke layer mterfacc descent u1ang the LES CFD° m0Ci1e1 is

.somewhat similar to observing" an actual

test. :J• . I 1 • I Figure 3 illustrates the differences among observed data

in a fire experiment (Yamana and Tanaka 1985). The test data represent the fillingooh,24 m (18.7.ift),by,30.m.(98.4 ft) by 26.3 m (86.3 ft) high enclosure. T�e aspeHt �atio f.t,!/H2),is 1.04. "

No mecha}}��al V�l�tjl.��on, i usrd q�ring.fhe}e,st. �he fire

has a steaay-state heat release rate re�wc�d t�1}?e_ 1300 kW· (1232 Btu/s). Note the difference among tlie photometer thermocouple, and visual' observatioi\s; of 'the' shloke layer'

1---· ..

1.00

,)' 1')·' •

0.80 .,. ,

"•[;";_ •;.';::_ 0.70

0.60

i � \ " ., ... ,,,,!'ii 0.501 ['" '" ( , • ;.J • _....,.._ • � I •·-•

;�:�,·�•'Ji .. �u. o.49* rrt�h . �

0,20 ME'T D 2 INFPA 928, EQ A·2

I ·, \1 l � • �. O, 10 ETHOO 1 (NFPA 928, ea

0.00

il\terface position:as it descends. Also note the linear nature·of the descent as the interface descends below the 7 m (23 ft) height, corresponding to ZIH = 0.26. The dashed line reprf.:: sents the:<Yamanatranaka predicted filling. Figure 3 most closely represents Test 5 presented in this dise.ussion.

CALCULATION RESULTS I

Figure 4 illustrates predicted volume filling for each of the 13 tests. In this figure, the Method 3 calculations were performed without layer temperature c,o,�ection. Note the

25 0 A-1 o� 0 THERMOCOUPLE "' t. PHOTOMETER a EYE

-e .. -.,1 ...... ::C • I l,1 (.!) LI.I 10 ::c

i ' 0

;J'.:1 ' Figure3

[]() 0 Cl

I J!O I �a I

---CALCULATED BY EO(ol

�fl:' '"TEMPERATURE . �o j \ .: ,

A •

\a '

I• ' Q,0 •

� ... .::•· .•l'· l•JJ, a o-:--�----! 0

5 . 0

TIME cmrn)

10./ I

, J

10

Yamana/I'anaJca, .. cxperimental results. Aspect Ratio = 1.04, H ·,;,, 26.3 m (86.3 ft); Heat

\<,r:'j, _ • :ii �.{f.e.�f!04�/?-a,�� =}�q<J .. k,W (1232, Bt'!(s).:

•' , . . . ••'

j I I 0

la �··

•: , ... ··; '.

(,j I , -. . ==

"!1 •. : ,. · . . H

. .. ·/

l ,;•

• :,.!'

'--' �1

' )

,,oill! !o, .. ,1� •; 30 �� 40 50 60 90 'ioo 110 120 NORMALIZED '.ffMF: .!'. I.: ' ' ., ' i'

Figure 4 Calculation methods 1, '2, ''iifiii 3.· (Cdlculation method 3 riot adjusted'for smoke lqyer temperature; Meth�J''3·:.

CH-99+3

calculaiions id'in.tifiea bjiest n'Umher.) " \ ,, r· ' '

5

clustering of the Method 3 curves around the Method 2 curve� :.: .. longer predicts the fastest filling of the atriums. Tests 4 and 7 Also note the position of the Method 1 curve. For these calcu-;r,,, lations, the Method 1 curve represents the fastest predicted. filling of the atrium volume.

now fall closer to the origin than the Method 1 curve. The Method 2 curve now represents nearly the slowest filling rates.

Figure 6 -compares the results from the LES CFD--simu-Figure 5 also illustrates the predicted filling i'at�s, but in

this instance, �moke densities have been adjusted based on the, calculated average plume temperature. Note the shift of the Method 3 curves toward the origin. Note :�so that Method 1 n�

• l lations to . Methods' 1 and 2. The LES CPD results · are 1 trendlines derived from the raw data observed from the LES CPD simulations (see Table 2 for LES CFD �Qserved data). The LES CFD simulations s_�()W the general dJstribution of the

\.<'

Figures

1.00

--0.90

o.80

0.70

0.60 � � � � 0.50

i 0.40

0.30

0.20 METHOO

0.10

METHO

0.00 0

,)_

__ 1 . ,

=

""

1,: ' I

• 10--·-·- 20- ----30 ---- - -40- . . -50- - -. 60-- - - - --70- 80------ 90---- 100- - ,110 J20 - - - --·---1 �_'� i .i� . ,-,

NORMA�IZED TIME ) I, • ·• . ' I

Calculation methods ·1,1;:2.iiaiirJ i· (.Cafoulatfon ·m.ethod J.;�aajuste(f1"(ir ,s'mqke layer ,t�mperature; Method· 3 calculations identified by:t:est nuinber.r " - '� . �;- -

. --- . "r1 -- ---,�- -

-

--- · - . I

1.00

0.90

0.80

0.70

0.60

0.60

0.40

0.30

0.20

0.10

0.00 0 10 20

. - -- -�--'- ·---- -- -- + ___ ...., . ----

'.· :1

30

- ·�-

40 60

.

60

-·--- - · -- - · -.' ::c;

Jt1 I •

,. - -- ·-·--·--.

•l I

'·

I! •• r

·' ·"""--·- � i. , . ·

L-t �:·

• ,•.,II

• .. =.i , .

trL_

i." , r.,.

( .. :.

r r I ;_ �. • .... ! . --�,,,·- ·-- .. ··-· . - ---- -

.. �. ,:' ,·�:·1Jd;.·:';,i.--{ •1,,· -:. '

• i"- .. :-.- .. _ .. - � .. ·�

-' J • • ·.·:·

' 70

: .. -===-===-�-=----: . --- - .. --;- . '. ,-, •'· j.\ ; II

, • :x: l :::::; '•·I.� �; ·,;L_, ..--.--• .... � _ 90 ". 100 I '10, r •- '12.0 • '•' _L · · ---

. '

NORMALIZED TIME '•

Figure 6 Calculation methods 1 and 2. (Compared to large ed.<;J,y simµJatioh C.fQ prMi.fted c),1,rve�;_ LeS.- CF!) q.�rves identified by test number.)

·

6 CH-99-1-3

Test 1

., Height (Z) Time m ft sec

30.5 100 0 '25 ,,82 •:: 200 20 66 280 15 49 500 10 33 1000 7.5 25 1750

Test2 Height (Z) Time

m ft sec 101.6 333 0

95 312 75 85 279 110 75 246 130 65 213 350 SS 180 sso so 164 600 45 148 850 40 131 l®O 35 115 r200 30 98 1400 25 J 821:,. • ''1700 '

20 66 1900

Test3 Height (Z) Time

m ft sec 101.6 333 0

95 312 75 85 279 90 75 246 100 65 213 250 55 180 375 so 164 450 4S 148 600 40 131 750 3S llS 800 30 98 900-·,_, 2S 82 1100 20 66 1300' lS 49 lSOO 10,.,. _.; 33: 1. 180Q

CH-99-1-3

TABLE2. LES CFO Experimental Observations

Test4 Height'(Z) ' Time

.m ft .. sec 9.6 31 0. ,17 . 23 35 I

6 20 70 s 16 90 3 10 160 2 7 300

Tests Height (Z) Time

m ft sec 32.1 105 0

24 79 30 16 52 80 9 30 220 7 23 280 6 20 290 5 16 310 � .. 10 ,-· 4"oo 2· 7

Test6 Heighf (Z) · ·: ',"

m ft 32.l 105 24 79 16 52 9 30 7 23 6 20 s 16 3 10 2 7

Test7 Height (Z)

m 8.2 7 ·: _

6 .. 5

4 _,. �. '

2

ft 27

___ .. 23 �-·

., -, -

20 16 13 10 7

450

..

Time'· sec 0

30 60

150 200 215 22S 250 310

Time sec 0

·- 20 .- 35

45 63 90

125

'

. ,.

Tests �eight (Z),

m ft 18.3 60

Mi 52 14 46 12 39 9 30 7 23 5 16 3 10 2 7

Test 9 Height (Z)

m ft

Time sec 0 ]S so

12S 250 400 500 700 900

Time.•

2S.9 85 0

20 (;_-66 150

1.

. . 13 43 ' 350 "''

.-·

8 ,, .. /)2.6

Test 10

800 ·1800 "'

"' '· . .. Height (Z) ·' • · '·Tbne

:··rt''·"-'· .. sec'· .. m

43.1 141 0 31 102 25 22 72 50 10 33 135 8 26 150

Test 11 Height (Z) Time

ft ,• m sec ·1

43.1 141 o· 30 98 250 22 72/ 600' I

10 33 /1450, � - .. ,� 13. 1825

..

,

.... �· _.

.....

,. .'I . . , .

' ·,··

Test 12 !·,

Height (Z) Time m ft sec

43.1 141 0 30 98 45 22 72 75 10 33 150 2 7 300

Test 13

Height (Z) Time

m ft sec 43.1 141 0 30 98 65 22 72 150 11 36 375 2 7 900

7

simulated filling curves. Note the similarity to the Figure 5 curves in that Method 1 does not predict the fastest filling rate for five of the tests. In fact, the LES CFD results for Test 7 predict filling times that are 50% faster than the Method 1 predictions.

Figures 7 through 10 compare all of the calculation meth­ods by separating the test results into the four selected aspect ratios. On each of the four figures, the LES CFD predictions are shown as individual points. These points are the raw data points reported by the LES CFD observer during the observa­tions of the test simulations.

Figure 7. Methods 1 , 2, and 3 approximate the LES CFD calculation from ZIH = 1 .0 to ZIH = O.i Method 1 predicts faster filling than the LES CFD observed points; but Methods 2 and 3 predict slower filling rates. Due to ' the- asymptotic nature of the Methods 2 and 3 curves, the predicted smoke layer interface arrival times begin to diverge frmn the LES CFD calculations below a value ofZJH = 0.3. Note the div�.,., --gence of the Method 2 calculation fi:gµt the LES-t::FD obser-vations as ZIH decreases. · •

. � Figure 8. The Method 1 calculation again predicts faster

filling than the LES CFD· observed points. However; .�. the height of the atrium i ncreases, the Method 1 curve begins to

Please refer to Figure 3, which illustrates the results from a fire experiment in an enclosure very similar in aspect ratio to the 0.9 aspect ratios assessed in this analysis. It would appear that the Figure 3 test most closely approximates Test 5. Note that the Test 5 LES CFD observations are closely approxi­mated by the Method 2 and Method 3 calculations, but the Method 1 calculations predict significantly shorter filling times. The LES CFD observations also· reflect the linear nature of the smoke layer interface descent observed in the Yamana/ Tanaka experiments. ,., .. .

Figure 9. Only two cases represent aspect ratios of 10. In both cases, Method 1 predicted filling times with reasonable accuracy. Method 2 does not appear to be reliable for these cases. Method 3 compares favorably to Method 1 and the LES CFD in Test 4 �µt does not predict Test-J. with a similar degree of accuracy,"'

..... Fi�� 10. All of the LES CFD· observations depict substantially, faster filling than the Method 1 calculations. Method .2 'appears to be unsatisfactory and fails to correlate w}.l:h the LES CFD resttlts"for each case. Method 3 predicts the Test 7 filling reasonaoly adequately, but it overpredicts filling

r.;;· tjmes for the other cases.

overpredict fi��ng times by a wide mar�. For ne�ly aH of the \ . , ANALYSIS ,1 , , . , , . , . A/f/2 = 0.9 caSes, the Method . . 3 predibtions prbdUce· curves ...1 '

- • • 1 '

with the proper characteristic shape. Below Z!H = 0.3, the Although the LES CFD mo\fol may ridt'be'fully validated Method 3 curve's asymptotic shape begins to differ from the based on the number of full-scale fire tests, its predicted results LES CFD observed points. Method 2 represents midpoint have been used as a comparison to zone model calculation values for these six tests. Note the linear nature of the LES methods provided in NFPA 92B. CFD observed data as ZIH approaches 0.25 to 0.2. Until this It appears that the currently ava.i:lable calculation methods point, Method 2 and Method 3 produce results that are reason- may not be useful over the full ranf,e of aspect ratio defined ably close to the LES CFD observations. in NFPA 92B. The reasons for tliis are not known, but the "'

8

1 .00

0.90

o_ao

0.70

0.60 !i:' � � 0� 50 � g

0.40

0. 30

0.20

0.10

0.00 " . 0 1 0 20

. .. . . - · """ ·

30 . ,40 50 ' 1'"'so;. ,_-·, '70 NORMALIZED TIME

_.,,,

: I i

l ' /' / /

J '

'lf /f

i I ·' '

,· I x TEST to fO.s. 43.1. 46401 I

80 '> 90 '."; · . _10,0, ' • , 1 1!>',. '. o J 20_ \\'\I\

": . '

Figure 7 Summary of results. Aspect ratio = 0.5. (LES CFD data represented by individual points.)

CH-99-1 -3

J !

··.

1 .

. � : .

I C•t

' "

1 .00

J .. . 0.90

0.80

, 0.70 ,, 1

0.60 "E �

I 0.50

ti · 0.40

1 \;:0.30

METHOD 2 o.2p

. · .

' . , .

. • I ' . ,

I •

. ,,

, . ' .

·,

x TEST 2 10.9, 101 .6, 2 1 1 0) , t TEST :}:10.9; ,1()1 .6, 4640)

1 , ' I ' - TEST 5 10.9, 32. 1 , 2 1 1 0) ' · c TEST 6 10.9, 32. 1, 46401

I . .. :rE�T.12:10.9, 43. 1 , 9693)

:;r.' .ln 'TJ ,r; - TEST 1 3 10.9, 43. 1 , 544)

' J , I • 2

ll ·�·

. ' ·

· 1 , 0 o.oo +-��--+��� ...... ,��-+�(�; ���_,,,.._ ��,� •• �-�,_,_� ..... ......,,,_,_�;-..-...,��-c;.-.� ...... -+'c,...,�r; ...... ...,.::-1'.' ... • ")

·.

0 10 20 30 40 50 60 , 7Q , , .:\,p,�; ; I .90 :

,1 00 1 1 0 120 I '.• . ., , [ ' NORMALIZED TIME. . 1 l " ! ,, .f ; �H J t .

'....! ' t' - ' I �·- '_1 J' .1 � I . t �I .J

Figure 8 Summary of results. Aspect ratio frid�viduall�. i1de1'f:.�ified.) ;

= 0.9;-(f!BSfocFD ��tg' iep.f��e!l��d b>' )��ividuaf,points; Method_) 1�urv;s

·� T ,, i:) ··-x: 1 A t '

' 1.:.>. , , 1 .00

.. · i' 1J ,,: .. ; Oi90 n

0.80

0.70

0.60

� � 0.50

i 0.40

0.30

0.20

0.10 :�···

�-'.I : ' ; ' J -.

' �_, I , ;· i ;_ . . . �.' ', t �' f �

; _,,,., _� � r

:.I' '

: '

r!! q r.1.ff1 ... Jt•

l...! ·<[J t0 l . , . ,

60 70 80

NORMALIZED TIME •

�1 '

,. ·1 t J •

.. --:. :..1 ., .i ., £ r J , ,,; , !

r , .' J j

T•1 ! :

-�-�. 1 _: ;_; "J . C ' ;f�C"l1

c TEST 1 110, 30.5, 2 1 1 0) + TEST 4 110. 9.6, 21 lOI

I

,1

· n

90 100 1 1 0

. . ... . .

j, '. •:

i . � '

120

·,,

• J

: /'!�' :) {

, .

. . .

Figure 9 Summary of results. Aspect.ratio = JO. (LES CFD data represented by individual points; Method 3 curves individually identified.)

c . . . .. . . ,

CH-99-1 -3

. . , i· .

9

1.00

·1:

0.90 I '

I · 1 j I ) :�

o.eo . ;' • I

"t .. 1� ·� · .. • I ... 0_10

0.60

� � 0.50 ;l; �

0.40

0.30

0.20

0.1 0

O.IJI! 80

, . ·

( I I

, .

• TEST 7 114, 8.2, 46401 •TESt 8 1 14, 1 8.3, 44001

I'...

6 TEST 9 114, 25.9, 4400) �TEST 1 1 114, 43.1 . 157181

90 100 1 1 0

• I t .

I';. ·

120

� - - • - • / ',-"'"f� •ti1 . . . ; 'j . i!'" •

Figure 10 Summary of results. Aspect ratio = 14. (LES CFD points reP.resented by indi�i41ftfl ppints,�. Method 3 curves ---- -indfviaually identified.) --- · - " · 1• . . · · .. . , , 1·

_.

_-, ,-.;

diff ���ces_ -�e signifi�.!ht for. as�b.Cha�s .represent�g the sho�� _slr1:1£�Un�s (�pei(t�tio =:. 14).

cases, Method. I overpredids fillilig 'times by a significant margin. In other cases, Meth11>d 1 unde!predicts fiHirtg".tiines

.:. : _J;Ji�ttiod 1.:§eems_J() projlij¢� a sroo�i��er fii_un&;�ufv�:: ¢at exhibits. th� li_{lear nature of the r_ate of filling at ZJ # ':.! 0:3 'and bdow. However, i! appeats tha� the _selection _9f �n Loitial term " of 1 . 1 1 limits the usefulness of this method. Iii some

by a s�giiificarit margiii . . J ;1 . : -· . . i 'i ' -· n

(-lJ � i ' • ·&.J r ;!] ')),, .... I� '· : ' • ; .•' ' � ti _ . : J '- /,f . .1 1!llJ

· ; rtfi� � p co���es, �e�ra,n:,9�t� P.9i�ts fr<;>.xn :.the .LES CFD model with a number of curves representing mq4ifica-tiqn� t!J . th� initial tqrm o� the Equati_qn_ . L Foi;. refe:rence .

Ii · i

. I

. ' , · .. �. ; ( , ,,, . , . f .L • '" . . , i .. ,I . ·). , -·-

c: .

;--... . : :1 . .

".t, !} t : '. ' :j I I i • - :; • l..- • - I ---�j�.� +\il�\1-\Wr-����=����=��---:'.,-,:-. -.;--�-n-::-.,--:.rl1n .�·�� : ;���.�o�;�n,okl . • n ." · , ;; ' '- r i < o.ati'�M\\l'Wr�--T----------------:-·-:-;�·-·_·_•_:;_ .. ·;,-----l: . �l.TEst·3r1o:e'·,-:io1 �(r. .. 6�0)� r: : .. t / ; :) , �

; \ ·i: . JV· .,; � :· .h+T£S1�'4 n'd!-9.6. 211,)'i ·� �-: r:J. 1 � . :i� ,�(.1: • .., r .•. 1;1t: .:

o. 70 +--lffilM''r-'i-'1r..,_,'<----"--'---_;,,;_------"'J ,'--'--'-' ·_•__,Jc..:!i:.:::;_·-'I: "' ' ;,.;rEST,& f0.'9/32.1-. 2:1;101 ' .'' O:i !- 'IL -,,_, . •:

.. . :� ( . ·:.rJ; ,,,,; . �P!JE�f_6 f9!"9,·:��At4:640� - .'J!.':\:.t· .- · ."'!"'!'

0.60 +----1+1'1--T-'l-'<-..,_..-'...-:..._�---��-�-------j flEST_�l1.f;;�.2::�64Jl}, , :, ,..., . ;::;· "·

'i' • TEST 8 114, 18.3, 44001 � u .:-� 1r:�; . ... u·J�!-rEST·a ,.,�: 2s.s: '*ioop( J .\ �;!1 _·"q .:: ;, � 0,50 t---t�l'illll'r-"r'-6"'"'""'�.-�°'------;:-o-;-�; •''"°; .... , ..-. � 1: xtESTJ1 0 .(.0-.'S:.4�1r,<!t640I:

i :.i.ic r:..- ;1 ;1off��l1Jl; \\�d�· 1_.,15z� 8J ,1 ::i 0.40 +---T<�"""''<-'1:--'�-"<:----'"'c----'""'='-�:::-'--';ho::::----,-""""-:-::-t

.... .... . . ,

·) -� • ;. �·i �­·0.20+---__;����,--:�.±...::::.....,__;;:,..,,_,.-.....:::"-<:;;-<.:..,,,...::::..i;��;....:::=-..-=::::,..;,..CIU:=...,ii!i:!:::=---:r:�

o.oo+-���-----1-�...-J�...,p;"--l>-.-..�;.;..-,.,.....;::>...,...-'.!"•....,:::._,,----J,...,�t-"=,....,.,.__....,.,...,-=ii--,..,.,.,"""'!r----:r-';",1' 0 1 0

• : - ! NORMALIZED TIME . .

"'i' ·'

;.

... r-

��. < • . . r · : · • ... ti • .J J-Jot '1� , .J1.,.:. • · , ' --�; i ;!l.d i"' Figure 11 Me_thod 1 curves with q_<Jj1Jstedfirst terrµ_y�. LES CFD obser�fiH°.'!k (J,.E� . .Clf� �P.!a_points,.,)1,� t'i G1' • . , . .

1 0 CH-99-'1 -3

purposes, the curve noted " 1 . 1 " approximates the Equation 1

curve. The additional curves represent Equation 1 initial term values from 0.85 to 1 .35. Note that each of the LES CFD test observations appear to follow one of the adjusted Equation 1 curves. From this observation, it would appear that the adjust­ment of the initial term of Equation 1 for a particular combi­nation of aspect ratio, height, and fire size could result in a calculation method that would reflect the observations from the LES CFD experiments.

Table 3 is a presentation of the test cases, separated on the basis of fire size, aspect ratio, and area. Each of the 1 3 test cases is represented by an entry. Although a number of addi­tional LES CFD tests would be_required to provide entries for each configuration and fire size, it appears that some general observations can be made. 1 .

2. 3.

As the aspect ratio increases, the initial term decreases. As the area increases, the initial term increases. , · •• ,,��- ·

As the fire size increases, the initial term decreases.

. --

CONCLUSIONS

The following conclusions are based on the observed differences betweeh thi'

LEs CFO observations and the other calculation methods. 1 . Metlwd 2 (NFPA .92�,- Equatien A-26) is not«a· reliable:j .

': r pg:diptor of sino,iq! layer positionoiexceptfor a narrow range of atrium sizes. The results from thi!!;equatiqp.,wiJJutµ.<e!y underpredict filling times by a wide margin using. the Joni'biilations ofhea-rre1ei: e ratei, heiglits��an<l arefil'lli this

·h:..·�tudy. t• �::..... J·J u 1 ) "" r.t7 · ' J. (f :� • :K) t""': ! ;-/' '

2.-·:i ;·'Methbd' 11 {NFPA'1-�h1l E�u�ti'oA':9) d�fh�(�rodu6�-

"conservative" results (that is, results that can be relied on to predict filling times -that are faster than would be produced in actual experimental conditions). in all of the cases that are foreseeablf 1yii�� ;t?�- ��PfcaPle range of

_

aspect ratios. Method 1 ap�,to, predi9� pmch faster fill­ing times for certain 0.9 aspect ratips anq-much slower fill­ing times for an aspect ratio .0fd4FAn!;adjusJ1netltof tlie initial term is suggested in'orderto:make'the method a reli­able indicator of smoke layer p6sl.dort. ,,.

: �· � .�: . · ; j ; ,: r2:; ;- • , 3. Method 3 produces reason!:lble results f� �pect ratios up to

10. However, at aspect ratios, greater :than 10, Method 3 is -likely to underpredict the <time of arrival of the smoke layer interface by 100%: A oorreetion°fadoi:' (or method) for Method 3 is beyond the scope �oflhis artruysis. · ' -- - - --- - .. . -·- :: ; . -

4. High aspect ratio spaces do nqt appear to fill in accordance with previously accepted : zone model assumptiohs. µe­reasons for this behavior are not known. ·. -

., RECOMMENDATIONS . .

- :.:. f . ·; ( � j . 1 . A standard definition of smok�_ layer interface should be(Ji .

'

I

developed. This definition should be expressed in mathe� ., matical terms in oi:der to evaluate _ _( or. r�valuate) experi- . . . mental data prevrously

" couet:fel

and �to' eVah.i'ate future

( '

CH-99-1 -3

TABLE 3 Adjusted Initial Term Values for Method 1 Calculations

Area

2 600 0 9300 4 700 1 700 93 0

_,_/ - Area

2 600 0,, .-· -·· 93 @

4700 1 700 93 0

•I'

1 700 93 0

Area

2 600 0 9300 4 700 r

1 700 93 0

Area ·'

2 600 0 9300

- 4700 . 1 700

. _, _.9 �0

Heat Release Rate

Aspect Ratio

0.5 0.9 10

1.13 "

,,. .. Heat Release Rate ' Aspect Ratio

/ 0.5 0.9 10 -

.

1.3 4 1.1

·. I �"\, �� , f-. 1.22 \ ' ) ', 1

n 0; Heaq_�1elease Ra,t,e -.

,,

1.18 L2

Heat Rel�,� �ate

· · Aspect-Ratio

0.5 ' 0.9 10

-·

�

- �'{ ·,aAl.12

,_

� Heat Release Rate .

Aspect Ratio

0.5 0.9 10

t ', � . \.� }� •I'

544 kW

14

2110 kW

14

' "!: ' \i_,,··. �

, I"! ,4500 .. kW_

' '· ·i .

: 0.9 7

0.8 7

9700 kW

14

15700 kW

14

0.98

1 1

experime11tal data. The definition would be particularly useful in CFD simulations where mathematic functions are utilized to determine the characteristics of discrete cells in an enclosure.

2. Additional LES CFD simulations would be useful for aspect ratios between 0.9 and 10. The data will be useful to further validate present Wlle models or can be used to develop improved zone models that can be used for the next several years.

ACKNOWLEDGMENTS

The author would like to thank Mr. Kevin B. McGrattan for his assistance in the preparation of this paper. Mr. McGrat­tan spent many hours working with definitions, assumptions, and procedures in order to produce the data from the 1 3 exper­iments. Without his help, this effort would not have been possible.

REFERENCES

BOCA. 1996. 1996 BOCA National Building Code. Country Club Hills, Ill.: Building Officials and Code Administra­tors International, Inc.

BOCA. 1993. The BOCA National Building Code/1993, Commentary Volume 1. Country Club Hills, Ill.: Build­ing Officials and Code Administrators International, Inc.

Brooks, W.N. 1997. Comparison of "regular" and "irregular" methods of calculating smoke layer interface heights in 1996 BOCA National Building Code. ASHRAE Trans­actions 103(2): 554-568.

1 2

Chow, W.K. 1995 . A comparison of the use of fire zone and field models for simulating atrium smoke filling pro­cesses. Fire Safety Journal 25: 337-353.

He, Y., P. Fernando, and A. Luo. 1998. Determination of interface height from measured parameter profile in enclosure fire experiment. Fire Safety Journal, 3 1(1):

19-38.

ICBO. 1997. Uniform building code. Whittier, Calif.: Inter­national Conference of Building Officials.

Klote, J.H. 1994. Equation 10, Method of predicting smoke movement in atria with application to smoke manage­ment. NISTIR 5516, Building and Fire Research Labo­ratory, National Insliluli: of Standards and Technology, Gaithersburg, Md.

McGrattan, K.B., H.R. Baum, and R.G Rehm. 1998. Large eddy simulations of smoke movement. Fire Safety Jour­nal, 30: 161 -178.

McGrattan, K.B., H.R. Baum, and R.G Rehm. 1997. Large eddy simulations of fire experiments. ASME Heat Transfer Conference Proceedings.

NFPA. 1995. NFPA Standard 92B, Guide for smoke manage­ment systems in atria, covered malls, and large areas. Quincy, Mass.: National Fire Protection Association.

NFPA. 1997. NFPA Standard 101, Life Safety Code. Quincy, Mass. : National Fire Protection Association.

SOderbom, J. 1992. Smoke spread experiments in large rooms. Experimental results and numerical simulations. Swedish National Testing and Research Institute Fire Technology SP Report 1992:52.

Yamana, T., and T, Tanaka. 1985. Smoke control in large scale spaces, Part 2: Smoke control experiments in a large scale space. Fire Science and Technology 5(1):41-

54.

CH-99-1 -3