Analysis and Prediction of Air Leakage through Door Assemblies DANIEL GROSS Center for Fre Research National Bureau of Standards Goithersburg, Maryland 20899, USA WILLIAM L. HABERMAN Engineering Consultant Rockville, Maryland, USA assRact A generalized relationship is presented for determining air flow rates through narrow gaps around door edges. The celationship provid Values of leakage rates for steady, Laminar flow through gaps. over wide Fangs of pressure difference and eliminates appronisations associated ith the often inappropriate use of discharge coefficients and exponents in che flow equation Q'= © A(ap)". The analysis covers straight theoush, ‘single bend and double bend gaps of constant thickness, a2 well a Connected gaps of constant thicknesses. Comparison of measured flow rates for installed stairvell door assemblies with those predicted by use of the relationship shows agreement vithin 20 percent. The volumetric How of heated air through simple door gaps haz been calculated by use of the kelationship. The results shov thet the flow race jay increase oF Gecresse vith tonperacure depending on gap size and flow region ntaopUcTroN Prediction of the flow of air through gaps around exterior doors te of considerable importance in deteraining ventilation races and heat Tosses in buildings. Knowledge of air leakage 2 also of importance in estimating the effectiveness of Interior doors as barriers te shoke fos. and in the design of smoke control systeas where stait and elevator shafts, vestibules, and other building spaces may be pressurized or exhausted, Im tho usual approach, use iz made of an equation of the form 210 € A(G9)" for descefbtng flow through single gape or shecugh she uleiple gaps of complete, closed door a Were, Q ie the volume flov tater € te a flav cocttictent, & le the efess-seecional aren of the Bap, Op is the pressure drop across the gap, and-n is an exponent. The constants ¢ snd n are generally not knovn of vary with gap geonetiy and Reynolds nunber.” Te has boon noted that previous assumptions of a~ 0.5 (che square root approximation for thin plate orifices) are not alvays appropriate [1]. “Nevertheless, it is comon practice in smoke control Gesign calculations’ to retain this assunption for all gape ar to utilize a value of n= 0.5 for latge gaps and n= 0.625 for saail openings (2,9) ‘The gape im conventional door assemblies approximate deep (thick- walled) rectangular orifices, i.e. wide rectangular duces, rather than thine walled orifices for which typleal discharge coefficients are often sssuned, Such will be the case of a plain door at floor Level, where there 15 no door stop, or at the aecting edges of swinging doors. in the ore typical case, where there are simple stops, tho flow passage may be Considered a rectangular duct wieh = sharp 90° bend of ‘ell", wheredoors are rabbeted, the flow passage will be in the for of « double 90° bend or "double elit. In addition, there my be Fe. or cavities within the gap az well ar variations in gep thickness, parcloularly in Coubinatfon with 90° bends, Finally, vhere sealing devices (flexible imetal, rubber, or plastic strips, labyrinths, ete) are ingtalled around the edges of Joore or door franes to reduce air flow, the geometry of the Flow passages may be quite complex. Air flow through stealghe-ehrough gape nay be considered equivalent to sir flow in receangelar duces, for “Len theory and experimental data are available [4-11]. Flow in such geps Ls typically laminar but nay also extend into the transition region fat sufficiently high pressure differences Measuring air flow through very saall, well-defined geps involves aeeting up special apparatus including flov devices suitable for very Low flow rates. Only a Linited snount of experimental data has bees published. The study by Hopkins and Hansford [1] provided dats on flow Through three simple gap configurations (straight-through, single bend and double bend). Homa [12] neasured flov through straight-through gaps Of various thicknesses and depths. ‘Thomas and Dick [13] measured flow through short sections of steel and vood window assezbiies, The configuration of these gaps is very similar to double bend gaps vith calarged conter portions, Nearurenents for simple and complex geps have ‘iso been reported by Ishira [14]; also, faired curves releting Ap and Volume Flow rate through straight gaps. and through a varlety of coaplex gaps containing bends, recesses, and labyrinths are summarized in (13) Standard measurenents can be made of flow through door assemblies at several op levels using chanber, a fan or blower, and suitable flow Control devices. Simple Laboratory teat methods have been in comnon use for exterior door and Window assentlies where infiltration iz of concern and there 2 standard pressure difference simulating vind pressure, ef 75 va, is normaly used [16], Nore recently, it has been applied alco co interior doors [17-20] where the principal concern is the control of potential leakage of sacke from building fires. a short review of the Evatlable air leakage measurenente on interior doce agsenblies was published several yeare age [21] If Q 4s plotted sgeinst ap, a single value approximation of che ‘exponent n 4s\sonetines obeatned in certain flow rate regions. Normally. full scale flow measurements of door assenbLies are time-consuming and expensive, and this {s espectelly true If heated aly is used. Hence, it Would be desirable to obtain « generalized relationship suitable for pre Aiccive calculations of aly leakage through door gaps over a vide range Elow conditions. ‘The purpose of this paper 1s to provide a relation ship capable of predicting flow chrough gaps of prescribed geometry and fo make comparisons vith available experimental data on flow through gaps land around selected door asseablies anaysts 1, Flow in Straight-Through Gaps “ine ‘autre One dtstingutshing feature of the flov tn gaps, unlike flow in ordinary pipe or duce installations, is the presence Of Flow development cegton (Fig). Fa — ‘The fluid shich enters the gap is, due to viscosity, reduced to rere velocity Fig 1 Development of Laninar long the entire gap wall. The flutd Flow n'a Gap 170layer at the wall gradually slovs dovm the faster-noving flutd next to Ae, until the effect of viscosity {s felt actors the entire gap. In one model of flow development, it is assumed chat che fluid enters with unt- form velocity. In this case, the inviscid core region of uniform flow gradually disappears and is replaced by the fully developed flow. Zhe Hlow development region often comprises a large portion of the flow through gap. To correlate data from a series of measurements, the use of non- ainensional pataneters based on the applicable set of physical variables becones advantageous. The significance of using nondinensional per. meters in Lieu of dimensional ones ies in the feduetion in the munber of Daranetere required to describe the phenomenon. Ar a consequence, con: Siderably fever experinental measurements are required to ertablich = valid relationship between variables of the gap flow. An appropriate sot of six significant physical variables 48 Q~ £8, pv. %y By), where 1s the volune flow Fate, Op it the pressure difference across" the Bap, Ls the flufa density, v ix the kinematic viscosity of the fluid, x is the depth of gap in the Flow direction, and D, is the hydraulic dianeter For rectangular gaps, D, ~ 28, where a 1s che" thickness of the gap. Dimensional enalysis vat utilized to provide the following set of three pondimensionsl parameters: G, = £(Re, D,/2), shore C, iz the prescure coefficient ap/(1/2)e¥" and Re tz the Reynoids number V'D,/v., Here, V (the average Flow velocity) = G/L a= 29/L Dy for wide rectangula® cops ‘Tho form of the analytic solution for Laminar flow in the entrance region of rectangular ducts [7], where the pressure coefficient slong the duct vas related to the nondimensional paraneter x/0,R0, suggests @ reduction in the nunber of paraneters. “Thus, the Cvo parameters Re and D,/x wore coalesced into one, nasoly Re 0, /x, ylelding a simplified set containing only two paranecers: 6, = ike Be) wo 1c should be emphasized that only a general functional dependency, not a specific functional relationship, (e-g. Linear) is implied here Te results of the analytic solution of Miller and Han [7] are shown in Fig.2, vhere the prossure coefficient G, for Laninar flow in s wide rectangular duct Is presented as s function of the paraneter Re Dy/x Extensive experimental data on developing flov of alr in vide rectangular ducts vere obtained by Beavers et al [10] and cover the range in Re D,/s fron 20 to 2500, These data ate alzo included in Fig.2. Excellent greenent between snalytic and experieental reculcs a foted, For values Of Ro d,/x loss than 10, the analytic solution merges into the straight Line 96°x/0,Re, the solution for fully developed laminar flov in a wide rectangular duct. Ie should be noted chat this sctaight Line always Lies Delov the developing flow curve; hence, the pressure drop in the entrance tegion of a duct will alvays be greater than that in an equal length of duct in fully developed flow. For short sections, the pressure drop in developing flow can be several times ax great az thar for fully developed ‘low, The analycie and experimental results do not contain entrance and leaving losses. Since there {z inviscid flow to the gap entrance and = viscous jet Leaving the gap, a combined entrance and leaving Loss coefficient of 1.3 vill be aerused. A curve for the total G, (the pressure drop in the duct plus entrance and leaving losses) has also boon [neluded in Fig.2, The reasonableness of the assumption will be con- Fired by comparison vith experimental data for straight-through gaps which includes auch losses in its measurements m100 10 Z inalytic Solution aa 10 F557 100 1007-0080 ~~ 100000 Fig 2 Presture Cosffictent for Wide Rectangular Ducts Teo previous studies of afr flow tn gape [1,22] sought to extabtich A simple Correlation equation for the presture coeéfiefent C, using # Linear exponent for the x/DyRe pacaneter over che entire parkmecer range ‘This oversimplification leads to the anomalous result that, at lazge values of the paraneter x/D,Re, the double-bend gap gives a lover pressure drop than that for" the straight-through gap. Ae will be shown, fo single slope is applicable over the entice parancter range. Tn this Paper. we are proposing « different set ef nondinensional parameters hich’ are more appropriate to the problem of gap leakage, where Lt is Usually required te find the fluld leakage rate for a given pressure Gifference. The proposed set constating of three nondiaensionai, para Ipeters ts O/Ly = f[apD,/ev, Dy/x). Combining the products az suggested by the analytic results of Hiller and Han (7), the number in the set is reduced to tho paranaters 00, ae = £1 ap0,2/or?) (D098) @ More (fs = ¥ a/v = y/o to 4 Reynolds number exprasned in terat of ohe olan Flow rete’a. “spitefae! tau vitecan peentare escttietone woten Vetoclty term: The stiaor ef tol Grand Baraner Re BV aise tn Findaaes conworted ios voburs of chetburtaceets' BNSC ast {ebint/pet) (ON and ave thom aa avbelid eave ia Flg-} eopecher vith GePaageeanGnt Gh data om Refevences {ieiin ial for suesighe Saree ap in'caran of the ey parencceres ivan distin saglone ney be noeeaee Re {ou Saloon of the diesioniens presse aifferenes (egies 1." cho Slate Flow rate varie Linear! uh pressure elfferereer tee” he slope Ie'TTat nigh vues of pressure affefence (eplon 3)" the volume ties Men’ cteitatisth the sqttareect af te peeatere steacaree Tie ehe Hope"ts'Ov3r he inceetbate pressure stfferenes (epion 2); he npeeone of the presure alffectnce wastes fron 893 Get For computer calculations, the folloving analytic expressions may be uused to represent the relationship with engineering precision q = 0.01042 sP uPs2s0 NQ = -3,30540.2915 NP-*40,01665 NP-?9-0.0002749 NE a50cuPc10# & (3) NQ = 0,555 Nes reio® where MQ = QD, /Lax and NP = (apb,2/p02)(D,/)? 12Region lee ——— Begton 2a} negton 9 1000 100 n@ = 0, /Loe 10 ‘oo 10 10000 00000 ~~ 1000000 - 9000000 NP = (0p0,2/e¥4] (2,/20® Fig 9 Nondinensional Q vs 4p for Stvatght-Through Gaps 2. Flow dn Gaps with Sharp 90° Bends In addition to the pressure drops due to skin frtetion within the gap and the entrance and Leaving losses, it te sonetines necessary to Include losses due to bends. The data of References [1] and [a] aay be used to derive nondinensional relationships for vide tectangular geps with one and with two sharp 90" bends. ‘These Telations tapether with the appropriate data are shown in Fig.4, ‘The relationships are essentially fdentical with chat for eho steafght-through gap for values of NP up co 4000. The sulttplication factors shown in the inset may be used tn conjunction with Fig.3 to obtain the appropriate values’ of NO 1000 “BINGLE 100 @ : =a male Bend $ 0 Bos 2 07 TPouble Ben are ears ‘oo 1 00 ‘i600 Wobo00 xe Fig 4 Nondisensional Q vs ap for Single and Double-Bend Cape. 17326a 10 x TO0ORv, 7~f0000 ~~ 400000 ~~ 1000000 Fig 5 Boundary Layer Growth in Wide Rectangular Ducts 1. Results for Connected Gaps with Different Thicknesses The analytic solution of Miller and Han (7] vas obtained for a uniform velocity distribution at the duct inlet, "when an enlargenent or contraction in the gap thickness occurs, the domstreaa section of the gap will experience a non-uniform velocity aizeribution at its inlet Since no applicable analytical solutions are available, an approxination to'such duct Flows may be obtained by assuning that the boundary layer thickness at che enlargement or contraction vensins unchanged. The Variation in boundary Layer thickness vith the parazeter Re Dy/x Was detived feom velocity dixtetbution data given in [0] and fro "an analytic expression for the boundary layer thickness piven in [23]. The result is Shown in Fig.5, The approximation iz based on the following consider ation: "As filuserated in Inset A of Fig.5, the flow entering an enlarged fap fron @ narrover gap vill separate fron the solld boundary and nove along a free streanline unetl {t reattaches {teelf to the soltd boundary The Free streaaline separates che main strean from the recirculating region. The boundary layer along a free streamline will grow at a Lesser Fate than if the surface were a solid boundary. The approxination ‘assunes that the growth along the longer free streamline (ac) is equal to thee obtained fron the shorter sold boundary (be) 4n the enlarged gap (n= f). Similarly, ae illustrated in Inset B of Fig.5, the flow ahead of'a gap contraction will form a separated region in the wider gap. The approximation again assures that the boundary layer growth along the free streanline 1s equal to that obtained by the shorter solid boundery 4Consider the flow through a gap consisting of a narrow section Followed by 4 vider section ax shown in Fig-6.. In this case, it Ls not Proper to add C, values for the to sections since each is based on ‘ifferent Fluid’ velocity; instead, corresponding Ap values showid be ‘added. The following procedure is suggested a2 a reasonable approxiatton, fend 15 based on che assumption that the boundary layer thickness encering, the second section is equal to that leaving the flret section: Bix neous) = 6 om ae tag /t) = (D/P PDs Secoraihe Cy Crip GI Calculate aby = C1 (HoV,?) @ ac necOys 4,9, seccenine 26/0, cFte.s) FH @ Ae 26/0, = (26/m,(a,/a,), decermine I Bettys fm") (Fins) AE Re(Dy/"), dteening yy" (Pt8.2) OO ereteaet een team (26/0) ta sty owe the felatton BenuiAe fey ta compote Ge © se Re(Dya/bny"479)), determine Gp Gan Cicdiate ag, ee E+ aleulace apt = s8f cat?) Caleulace Zap - apt + ap, Rey Pe To convert the results to QDy/Lox values, rake (0, 80/2); to convert fo C600,2/004 VCD ADE values, multiply Hey by Ce B/D! é a | Calculations have been made using this technique and have Gesonstrated that the upstream section iz clearly the most critical in terms of pressure drop and flow. Ifa narrow thickness section prececes a larger Section, the flow Le considerably Lesa than the flow in the Yeverse direction. As one example, the flow in a 38 am long gap of 5 sa Unlekness vill be reduced 814 Lf a 2.1m long gap of Imm thickness t= added at the entrance (to provide a total gep length of 40 tm), but will be reduced only 4iv if the’ sane Ima thick gap is placed at the exit. Tt {5 also of Interest co note that for two sections of equal length, whon the second section 1s larger than 3 times the thiekiess of the initial narrow section, the contribution of the larger section {2 very onall and rnay be neglected with Iitele error 2. Results for Door Assenblies Reference [26] reported measurements of six leakage rates for stair Mell door assenblies installed in eight high-rise butléings in Canada To make predictions of the leakage rates, it is necessary to have detailed measurements of gap thicknesses and dinensions. These vero Kindly provides by Tamura [25]. The gaps were meesured at three points slong each vertical edge and at two points at the top and bottom edges of door assenblies using s series of metal shina. The shins vere used in conbination to provide measurenents to the neatest 0.5 mr, providing « Fessonably good average for each door type. Along the vertical sides and the top edges of most door assemblies, the gaps consisted exsentially of one Long section plus one short 90° acction provided for the door stop For andlysis, this vas considered as a s{ngle-dend, constant wideh gap witha path lengeh equal to the sua of the to lege.” All these neasure= 175mente were cosbined £0 compute an average gap th{ekness from vhich a Teakage flow, at « pressure difference of 25°Pa, vas computed from the relationships given in Figures 3 and 4. The gap slong the door bettoa ‘ear considered separately since it iss strsight-chrough gap, ané in wo=t cases vas significantly thicker than the other gaps. fhe cvo partial Flows were combined to yield an estinate of the total flow through che gaps around the door assembly. Sinilar predictions were nade for Comparison with measured Leakage rates for door arseablies tested tn the aboratory [26,27]. Tn casos where the gap consisted of = narrow section plus a considerably larger section, che Latter vas ignored since the bulk Of the pressure drop occurs in che more constricted Section. Table 1 Contains a summary of the predicted end the measured sir leakage rates at pressure difference of 25 Pa. In most cases, agreenent between the predicted and mearured flow raves war within 20 percent Table 1 Measured and Predicted Alr Flow Ratt through Door Assonblics Source Ddestgnacton thiEknels veptn Langen|® feedicted: © */weasure Eh oe CHEE, US 2 cael es Botton? 1716 Me Olah 0.136) 0.151 (ol BMH eggs ils 6 88 Bco2s | aoe : Bottom 320 4B 084 0.023 9.033 | 0.030 3. Effect of Air Temperature on Flow through Gaps In fire sftuations, knowledge of the flow of heated alr or of a mixture of heated gases and smoke will be of interest. Moderately Glevated tenperstures up to at Least 200 of 300°C may be anticipated on door aesenblies uhich separate rooms and corridors or which serve Sneke ov fire bartiers in corridors or stairwells, For such door sesenblies, vhere door deforsation and warpage ave snail (belov 300°C), Flow cates have been extlnated as a function of tenperature. For the Elnplece case of heated ait, Table? Lata values of the computed flov Entre ct selected tenperatutes and pressure differences. These values Sete computed using the relationship shown in Fig 3 for straight-chrough gape and the values for density and kinenatie viscosity of dry air listed En table 2, Te is of Interest to note that due to the decrease in density and the increase in kinenatic viscosity of air with teaperature the flow rate may increase, decrease, of reaain essentially constant a3 Temperature increases, depending on the flow regions involved (zee Table 2) 6Table 2 Effect of Air Temperature through a Straight Gap (Depth=40 ma) Temp Density vigcopte Payal” Tor aE 1030 a oF3%a6 70183 LSE OB TNS S02 ‘Dold. x83 3g B35 UB: 0GTaE B:90UR8, Stes 8.0828 300 Grete 2:95 8080088 BUF BSELES 815283 25 39 2.20 © n.s1 fo.g00358 0.00727 0.0243 0.0507 wo abe 335 | 8:600859 | B:0088? 8.8983 9.0595 Boo Grete 2:98 | 8100083 8:80 8:83 8:8 RS 100 29 1.208 1.52 fo.g0138 9.00628 9.0507 9.1012 yoo b:RG 239 OMT BGS B:088 8.25 380 Grete 4:98 N8:B0dba 888887 B:0e5e | 8:28 1. The experimental data and examples have dealt with air as the fluid fat ambient’ and elevated teaperatures, In actual situations, the fluid ay be a mixture of heated combustion products, e.g. nitrogen, earbon dioxide, vater vapor, hydrocarbons, ete rather than dry aif. Since the analysis applies to any fluld, values of density and kinesaiic viscosity appropriate for the actual mixture should be used. 2. In other situations, the fluid may consist of multiple phases, te gases, condensed vapors, solid particulates. Unfortunately, no analyei~ eal work appears to he currently eva{lable dealing with mulciple-phase Flow in che entrance region of a g3P. 3. The described prodtettve procedure for eatimating the pressure losses for developing flow in connected rectangular gaps (sudden enlargement. and contraction) is sn approximation. Experimental Gata on flow through such ‘gaps over a wide range of flow paransters are needed for confirmation, 4. The pressure difference acting on a door assenbly may vaty with sight and ith cine rather then being uniform and constant. Tn a typical room involved ina fire (or othervire heated), positive prossur relative to the exterior will exise over the upper 1/3 or 2/3 of the door height and negative pressures below. This situation can be treated by considering the vertical gaps to consist of miltiple segments. ‘The rocal Flow can then be obtained by suming up the contributions of the tnd vidual regnence. 5. Gaps between doors and frases may not be of simple geomettic shape and may contain special sealing devices, including flexible gaskets, fiber brushes, ete, Tn such situations, standard laboratory testing or special research studies may be needed to define leakage flows properly conctusroxs The generalized relationship presented in graphical and analytical form in this paper permits prediction of fluid flew through gape of constant thickness over a wide range of pressure difference, with direct application to leakage of air or heated conbustion products’ through gaps in door assenblies, In addition, a procedure is proposed for extending the “predictive capabiIity to the flov through connected geps with Aifferent thicknesses. In the design of pressurized or exhausting smoke control systems, this relationship provides a more accurate means of predicting alr flou through door sesenblies than the sinplified equation QC acap” currencly in use wsREFERENCES Yophing, 0, god aneford, 9, thle Hhow Trough Cracks" Build oneness artnet Rete Bay Teme 2. Heietite ce in SPF. Spon Pabtishers’ Landolt E Klote, JH. and Fothergill. it agyol_ Svstene geal ae se mares is Hite oc SEMIS tage SOE TEED a tn wygensrnte erage seein Push Hebetette BS etch aregte nie Pee fe pe 7, SEE nu. and won, 1.8, *Ereasure Losses for Laninar Flow in, che Entrande degignof Dieta. 5f Rectang ssitgeetal HetanguLsr Ebsas Sect lGne a apple Hoch Wor eabe 1585-87 C570) 8. Faehfbang, Wo and 2ehocon ¥, cSteady iapinas Flow 4n the Innee Heplon of Recenguiae Dicea® Balfetin of the J8NE T1024, "AS. 103 9. Sparred, 2.4.; Mixon, 6.W, apd Shavit, ¢. Experigents, on Laminar iy bce ubsyiesetlan ces 9. RECIRS aes ASHE Vol. BS, Feeds (F968 vo. BEsbe Sduadesiata de ttt Setanta Rosia te Hereer Ratios The. J of Heat Macs ‘Transfer Vel, 13, 689-702, 197 n U5;don, A. te ehalnar How Forced Eatweetion 1s bucpar chapter Vir, Advances In Heat Teanafer, Academic Press, New ese agi i} 2 "i jalgnetlation of Dwel}ings, ang ite Disturbances” Tech. Rpt eae Suedian inst “esting, sng vent! tation (1975 13, FRogage DIA. find Dick, Sene"Uae indiieracton thigugh Gapg Around Sipser enting ie Eras. 1 ee CoeEEtctents of Gaps", Japan Arch. Soc, Rpt. 29 15 2.18, Japan Arch, Soc., 10th Edicion ra lat Method for Rate of Att Leakage cheouth Begrier iiiaoes *eartatn Welise and’ boots? “Asta, PRESHE ghia S 17. {30"58i0/1 spire Taste ; Bvgluatton of Fesfornance, of Saoke Concrol Door Asganolies™. Part i;"vanblone Tempeesture: Test’ 180 (19 18, Bre 105 ""instatietion of saoxe™ and Neclonai Fire Brotection Ageociation, (1385 ro, REAR! Pare ST?Mhechods for deseuring Smoke Penetration chrough Dootgees and shuccer Asseablies; anbiefe Fempersture’ Consitions BUgEEstoncrot Door anes abt est Entrance tie ‘of Straight, Plane Gaps* Jahrbuch der Deutschen LUsftfaheeforschung 121-39 (i941) Fite tia ‘hwy “Cav. PALF Leakage Data for the Design of eRe ee Eee ca ac tne cr a7. esthobe, fu. and deberatt Tr Resul ce of a eee eae eS ests, for Fesforsance of v8