Analyss of old Tme Models for Total Floodng Clean Extngushng Agents Todd M. etrck, Al S. Rangwala Department of Fre Protecton Engneerng Worcester Polytechnc Insttute Worcester, MA 69-8 Abstract Ths study documents the expermental results of a research program desgned to evaluate the valdty of the wdely publshed hold tme predcton models found n NFPA, Annex C and ISO 45-, Annex E. The models dscussed n these standards obtan a measure of the equvalent leakage area, whch, when coupled wth worst case assumptons, can be used to determne the mnmum hold tme. Three hold tme predcton theores are adopted from these standards for valdaton; a wde descendng nterface model as mplemented n ISO 45- and two sharp descendng nterface models from the 4 and 8 publcatons of NFPA. The expermental program s comprsed of thrty four tests conducted n a 3 m 3 test enclosure. Seven clean agents are utlzed n the study; selected to nclude both nert gas and chemcal agent types whle spannng a wde range of agent vapor denstes. Ths ncludes FK-5--, FC-5, FC-7ea, FC-3, IG-, IG-54, and IG-55. A seres of holes were drlled through enclosure boundares at upper and lower elevatons whch were opened or closed as a means of regulatng the amount of leakage area for any gven test. Vertcal profles of agent concentraton and ambent pressure are used to evaluate the agent concentraton dstrbuton, rates of agent dranng, and the effectve lower leakage fracton. A nondmensonal hold tme s used to compare expermental results nvolvng dfferng agent types and leakage areas. Results show that expermental values of the hold tme are generally up to 5% longer than the theoretcal hold tme predctons when evaluated as the tme to reduce the agent concentraton to half ts ntal value. When evaluated as a 5% drop n concentraton each model s valdty s sgnfcantly reduced. Under ths condton, expermental hold tme values are up to 5% shorter than the predctons of the sharp descendng nterface models and up to 5% longer than the wde descendng nterface model. Nomenclature A F Enclosure floor area [m ] A Orfce area for gas flowng nto enclosure [m ] A o Orfce area for gas flowng out of enclosure [m ] C o Dscharge coeffcent for the orfce descrbed by A [-] o C U Unt converson constant n the sem-emprcal orfce flow equaton [s n- /m n- ] c Intal clean agent volume concentraton [Vol. %] c f Fnal clean agent volume concentraton [Vol. %] F Lower leakage fracton, Equaton & 4 [-] F Dmensonless rato of outflow and nflow orfce areas [-] g Acceleraton due to gravty [9.8 m/s ]
Enclosure maxmum heght [m] Interface equvalent heght [m] Interface heght [m] Protected heght [m] p Dmensonless heght for the sharp nterface theory of the 4 NFPA publcaton SDI4 [-] Dmensonless heght for the sharp nterface theory of the 8 NFPA publcaton SDI8 [-] Dmensonless heght for the wde nterface theory of publshed n ISO 45- [-] WDI n Orfce flow exponent n the orfce flow equaton [-] t Dmensonless hold tme used when orfce flow exponent s varable [-] t Dmensonless hold tme used when orfce flow exponent s equal to.5 [-] e Greek Symbols β Dmensonless coeffcent used when orfce flow exponent s varable [-] β Dmensonless coeffcent used when orfce flow exponent s equal to.5 [-] ρ Dmensonless densty parameter [-] ρ ag Clean extngushng agent vapor densty at C [kg/m 3 ] ρ ar Densty of ar;. n NFPA &.5 n ISO standards [kg/m 3 ] ρ Agent and ar mxture densty [kg/m 3 ] mx Abbrevatons FK Fluoroketone FC ydrofluorocarbon IG Inert Gas ISO Internatonal Standards Organzaton NFPA Natonal Fre Protecton Assocaton Introducton Total floodng fre suppresson nvolves the dscharge of a clean extngushng agent that s typcally requred to provde protecton wthn the desgn envelope for a mnmum ten mnute perod. The hold tme s defned as the perod of tme requred for a clean agent concentraton to drop to a specfed threshold (usually 8% of the ntal dscharge concentraton) at a specfed heght n the enclosure (often chosen as the pont of hghest combustbles or at 75% of the maxmum enclosure heght) []. The goal of ths study s to valdate ndustry-standard hold tme predcton models as they apply to a varety of clean extngushng agents. A 3 cubc meter expermental enclosure s used to observe leakage flows through enclosure boundares. The upper and lower leakage areas are vared to determne the effect on hold tmes of seven commercally avalable gaseous suppresson agents: FK-5--, FC-5, FC-7ea, FC-3, IG-, IG-54 and IG-55. Prevous studes evaluatng agent leakage rates show that model predctons are often naccurate; resultng n both overly conservatve and optmstc hold tme approxmatons [-5]. The fan ntegrty test encompasses the test method and leakage modelng used to evaluate the total floodng system desgn wth respect to the hold tme or retenton tme
requrement. NFPA, Annex C and ISO 45-, Annex E contan enclosure ntegrty desgn standards, whch are chosen for comparatve analyss n ths paper due to the prevalent adopton and use around the world. Expermental results wll be valdated aganst these standards predctons of hold tme. Modfcatons were made to the hold tme theory n NFPA from the 4 edton to the 8 publcaton. Comparatve analyss n ths paper wll utlze both NFPA publcatons to further help n understandng these changes. Background & Theoretcal Consderatons Total floodng fre protecton systems dscharge a gaseous agent n large quanttes such that a protected enclosure wll be flled to an extngushng or nertng concentraton. In the present study, only the applcaton of clean extngushng agents s assumed. Clean agents are frequently referred to as halon replacement agents due to the enactment of the Montreal Protocol n 989, whch prohbts the contnued use of halogenated (or ozone-depletng) agents n total floodng applcatons. Clean extngushng agents are classfed n NFPA as electrcally nonconductve and ether readly volatle, or gaseous fre extngushants that do not leave a resdue upon evaporaton [6, 7]. A typcal total floodng system ncludes one or many hgh pressure, agent cylnders that are connected to a delvery ppe network through a manfold. The ppe network termnates at one or many dscharge nozzles wthn a desgn envelope. Upon manual or automatc actvaton all cylnders are smultaneously opened and the dscharge duraton s targeted for s for chemcal agents and 6 s for nert gas agents. The dscharge event s turbulent, resultng n a relatvely unform mxture of clean agent and ar nsde the desgn envelope. The agent and ar mxture mass densty nsde the desgn envelope s generally greater nsde the enclosure than the densty of ar surroundng t. Ths densty dsparty exerts a postve hydrostatc pressure on lower enclosure boundares and a negatve nteror-to-exteror pressure dfferental at upper enclosure boundares. These pressure dfferentals drve a convectve flow of agent-ar mxture out lower leakages n enclosure boundares, whch, s balanced by fresh ar flowng n upper leakages. Ths s the only transport method consdered n evaluatng the global rate at whch agent drans from the enclosure. The present study seeks to evaluate three dfferent hold tme models that predct the rate of agent dranng; two sharp descendng nterface models (as theorzed n the 4 and 8 edtons of NFPA ) and the wde descendng nterface model (as employed n the 6 edton of ISO 45-). Extensve dervatons of the hold tme theory contaned n these documents are publshed elsewhere [, 4, 5]. An analytcal form relatng the rate of agent dranng to other confguraton parameters can be acheved by ntroducng the followng assumptons (central to each hold tme model consdered n ths study). () Thermal effects are gnored. The agent-ar mxture resultng after the dscharge event and the ar surroundng the enclosure are both assumed to exst at standard atmospherc temperature ( C). Further, the thermal affects produced durng a real fre ncdent are not consdered. () Speces dffusve transport s ether neglected (sharp nterface) or assumed to mx at an nfntesmal rate n known proportons (wde nterface). (3) The leakage areas n enclosure boundares are assumed to exst at only two locatons: the extremes of upper and lower elevaton. The three theoretcal models governng equatons are presented n the followng. Only the man equatons are dscussed. A detaled dervaton can be found n prevous work reported n lterature [, 8, 9]. 3
The wde descendng nterface model ncorporated n ISO 45- s gven as Equaton. In order to facltate drect comparsons between varous test confguratons a dmensonless form s derved, whch results n WDI [ t ] n = β, () where e e WDI =, A F F o = = A, F ρ ρ = ρ mx, n c = ( p ), CACg o o U t = ( tf t ), ( ) ( ρ ) n c f n A F ar β n F =, + ρ n and c c ρ mx = ρ ag + ρ ar. The dmensonless parameters found to govern the rate of agent dranng nclude the rato of the equvalent heght to the enclosure s maxmum heght, WDI, the rato between the outlet and nlet leakage areas, F, and the rato of the agent-ar mxture densty relatve to the densty of ambent ar, ρ. The equvalent nterface heght,, s gven as a functon of the e enclosure s maxmum heght,, the protected heght,, and the ntal and fnal agent p concentratons, c and c, respectvely. The hold tme s thus evaluated as the tme at whch f a specfed concentraton ( c f ) exsts at a specfed heght ( ). The hold tme n seconds s p gven as the ( tf t ) where the dmensonless hold tme s formulated as t. β s a combned dmensonless coeffcent. The vapor denstes of atmospherc ar and the agentar mxture are ρ and ar ρ, respectvely. mx The 8 edton of NFPA, Annex C espouses a sharp nterface model that uses a varable value of the orfce flow exponent, n, as mplemented n the wde nterface theory [6, 9]. The above theory may be smplfed nto the 8 edton of the sharp nterface theory by settng the agent concentraton at the end of the hold tme, c f, equal to one half = ). By redefnng the dmensonless heght the ntal concentraton, c (resultng n e p parameter as the rato of the actual nterface heght, the ntent of the 8 edton of NFPA, Annex C s met. Equaton shows ths model n dmensonless form as [ t ] n, to the enclosure s maxmum heght SDI = β, () 8 where SDI =. 8 The sharp descendng nterface theory s also publshed n the 4, and pror edtons of NFPA, Annex C [7]. The prmary dsparty between the 4 and 8 edtons of NFPA les n the applcaton of the orfce flow equaton. In the theoretcal models above the orfce flow exponent, n, s a varable model nput parameter. The 4 edton of NFPA assumes that n s equal to.5. Equaton 3 gves the dmensonless governng equaton for the smple, sharp descendng nterface model as ( β ) SDI 4 t =, (3) 4
where SDI =, t ( ) 4 = t t C A o Ao g, F ρ β = ( ). + ρ F Ths can be derved from the sharp nterface model (Equaton ) by ntroducng the assumpton that n equals.5. Thorough dervaton of ths model s publshed elsewhere [4, 5]. Expermental Apparatus & Instrumentaton All testng reported heren was conducted at the Fke Corporaton test faclty n Blue Sprngs, Mssour, USA, n the same enclosure wth no sgnfcant modfcatons made between test sessons. A schematc of the expermental enclosure s shown n Fgure. Internal dmensons are 4.6 m (8.5 n) by 4.6 m (8.75 n) by 4.88 m. (9 n) n heght, whch totals 3.8 m 3 (364 ft 3 ) n volume. Constructon conssts of 5. cm by.3 cm ( n by 8 n) wood studs on 4.6 cm (6 n) centers wth two nteror layers of 5.9 mm (5/8 n) plywood and one layer of fberglass sheetng as an nteror fnsh. Intentonal leakage area s suppled n two forms; () 84 drll holes.5 cm ( n) dameter about the upper and lower enclosure boundares and () a celng vent for dscharge pressure ventng of nert agents. The drll holes are located near the extremes of enclosure elevaton such that assumpton (3) n the hold tme models s met as closely as possble. All drll holes are offset from lower and upper boundares by 3.5 cm ( n) and equally dstrbuted across each wall facng such that a nomnal upper and lower holes exst per wall. A floor dran s located n the room s center that was closed by means of an exstng ball valve. For each clean agent tested, a seres of controlled leakage area confguratons were smulated by pluggng and/or unpluggng drll holes. Dense rubber stoppers were used to plug holes from the nsde where they made a relable seal wth the fberglass sheetng. Each specfed leakage confguraton was accomplshed n such a way as to produce a symmetrcal leakage pattern. For example, an experment wth 6 open drll holes would be accomplshed by openng a sngle hole at /3 and /3 of the wall wdth on each wall, upper and lower. Measured quanttes nclude nozzle and ambent pressures, gas speces vapor concentratons, and enclosure ar temperatures. Nozzle pressures are retaned as a means to ensure proper agent delvery and to dagnose potental problems n system desgn. Ambent pressures are recorded to document () the peak pressure pulses generated durng agent dscharge and () the hydrostatc pressure profle throughout the hold tme. The present study focuses on the later, leavng the topc of room ntegrty for subsequent analyss. Clean agent volume concentratons are used to observe the drop n agent concentraton as a functon of heght and tme. Enclosure ar temperature measurements are used to further analyze the applcablty of neglectng ths varable n hold tme predctons (as prescrbed by NFPA and ISO 45- desgn standards). Envronmental condtons perceved to have an affect on agent dranng were controlled as closely as possble. For all tests conducted the relatve humdty was below 4% and the average ambent temperature before dscharge ranged between and 3 C (7 to 88 F). Generally, bas pressures & wnd affects were suffcently avoded smply due to the test enclosure locaton beng encapsulated n a much larger warehouse. Experments, 3 and 4 from the IG-54 test set nvolve relatvely tght leakage confguratons. In these tests only, concern for the potental to over-pressurze the enclosure was mtgated by usng the vent. 5
Postve Pressure Vent 36 x 9 cm (4 x 36 n) 3 cm ( n).73 m (68 n) 4.88 m (9 n) Pressure Transducers 3 cm ( n) 4.6 m (8.5 n).73 m (68 n) N S 4.6 m (8.75 n) Fgure : Schematc of the test enclosure. Ambent pressure probes and controllable leakage areas are shown. The postve pressure relef vent was allowed to open only for select IG-54 tests. Agent concentraton measurements were made wth a varety of nstruments. For each, an exhaustve effort s made to ensure that the recorded values are nterpreted, fltered and scaled nto engneerng unts accordng to that prescrbed by well-establshed measurement theory. Note that the recorded values are from a relatve measurement technque whch results n the nablty to measure unts of absolute concentraton. Measurement uncertanty s lkely less than ±% of full scale, however, an nvestgaton of measurement error bounds and propagated uncertanty n the calculated quanttes s yet to be completed. Expermental Results and Analyss A total of 34 hold tme tests were conducted. Seven clean agents are utlzed n the study; selected to nclude both nert gas and chemcal agent types whle spannng a wde range of agent vapor denstes. Ths ncludes FK-5--, FC-5, FC-7ea, FC-3, IG-, IG-54, and IG-55. Dscharge concentratons are typcally at the agents lsted Class A desgn concentraton. Leakage confguratons used throughout the test phase result n hold tmes n the range of.8 to 46.3 mnutes wth a medan value of 9. mn 3. Ths suffcently spans the lkely range of system use and allows for the enclosure leakness level to be nvestgated as a potental source of predcton error. Fgure shows the dmensonless theoretcal hold tme predcton plotted wth respect to the dmensonless expermental hold tme where the data seres are grouped by agent type. Proper calbraton of all gas samplng nstrumentaton was not avalable. Recorded values were scaled nto engneerng unts based on a Zero value (a 3 s average of sampled fresh ar before dscharge) and a Full- Scale value (an average of 9 s of data acqured after agent dscharge and readngs had stablzed). The full scale value represents the agent dscharge concentraton, whch was calculated usng the NFPA Total Floodng Tables wth the agent mass as an nput (clean agent retaners were weghed before and after dscharge). Data traces exhbtng suspect behavor are dscarded. 3 old tme calculatons performed accordng to NFPA, 8 Ed. wth the protected heght equal to 85% of the maxmum enclosure heght. 6
Ether axs ranges from to where the nterface can be magned as travelng from the celng of the test enclosure (at dmensonless tme = ) to the floor of the enclosure (at dmensonless tme = ). For each test conducted a seres of agent concentraton measurements are taken across a range of elevatons. Each nstrument provdes a sngle expermental value of the hold tme duraton where probes at upper elevatons result n nondmensonal hold tmes nearer to a value and lower elevatons tend towards values of. Fgure evaluates the hold tme as a 5% reducton n agent concentraton; relatve to the ntal, dscharge concentraton. Theoretcal hold tme predctons are based on the sharp descendng nterface model as publshed n the 8 edton of NFPA, Annex C. Nondmensonal Theoretcal old Tme [-].5.4.3...5.4.3...3.4.5...3.4.5 Nondmensonal Expermental old Tme [-] Exact Correlaton +/- 5% Error +/- 5% Error +/- 75% Error +/- % Error FK-5-- FC-5 FC-7ea FC-3 IG-54 IG-55 Fgure : Valdaton plot of the nondmensonal theoretcal hold tme versus the nondmensonal expermental hold tme wth data seres grouped by clean agent type. Plotted values are calculated as the quantty ( β t ). Expermental hold tmes are evaluated as the tme when the agent concentraton descends to 5% of the ntal dscharge concentraton. Error lnes represent the percent devaton from the theoretcal hold tme predcton. Theoretcal hold tme predctons are based on the sharp descendng nterface model publshed n NFPA, 8 Ed. Fgure shows the degree of correlaton observed for one agent type versus another. Generally, data ponts are equally scattered; ndcatng that the theory works equally well for a range of agent types (an analyss of the mean quadratc error for each agent s data set confrms ths). Only the agent FC-7ea does not conform to ths trend as nearly all data ponts le to the lower-rght of the other agent types. Ths s potentally due to an unrealstcally low vapor densty havng been used for calculatons 4. Data ponts n the lower-left regon of the chart (hghest elevaton n the enclosure) dsplay poorer correlaton. Ths s lkely due to the close proxmty of hgher agent concentraton probes to the turbulent mxng of nflowng fresh ar. Data ponts below the exact correlaton lne represent a conservatve condton where the agent s observed to dran more gradually from the enclosure than predcted by the theory. Conversely, data ponts above ths lne represent an overly optmstc condton where the 4 Fgure analyzes the data accordng to NFPA, 8 Ed. All agent vapor denstes used for hold tme calculatons are adopted from ths standard and not confrmed wth the agent manufacturer. 7
agent s found to dran more rapdly than the theory predcts. It s observed that when evaluated at a 5% concentraton reducton the large majorty of the data s n the conservatve doman of the chart; generally wth an error magntude less than 5% (excludng FC-7ea). Fgure analyzes all expermental hold tme data accordng to 8 sharp descendng nterface model. Fgure 3 analyzes the entre data set of Fgure accordng to the three hold tme models n queston where data seres are grouped by the theory appled n nondmensonalzng the data. Once agan, the hold tme s evaluated as the tme requred to reduce the agent concentraton to one half the ntal value. Nondmensonal Theoretcal old Tme [-].5.4.3...5.4.3.....3.4.5...3.4.5 Nondmensonal Expermental old Tme [-] Exact Correlaton +/- 5% Error +/- 5% Error +/- 75% Error +/- % Error NFPA `4 - Sharp Int. NFPA `8 - Sharp Int. ISO `6 - Wde Int. Fgure 3: Valdaton plot of the nondmensonal theoretcal hold tme versus the nondmensonal expermental hold tme wth data seres grouped theory type. Plotted values are calculated as the quantty ( β t ). Expermental hold tmes are evaluated as the tme when the agent concentraton descends to 5% of the ntal dscharge concentraton. Error lnes represent the percent devaton from the theoretcal hold tme predcton. In Fgure 3 the data tends to le below the lne of exact correlaton. The group of data ponts for any of the three types of appled theory appears to be equally scattered. Although not vsually apparent, dependng on the value of the flow exponent, n, the sharp nterface theory used by the 4 edton of NFPA consstently predcts shorter hold tmes than the 8 verson of ths theory. As explaned n the Background and Theoretcal Consderatons secton, when evaluated at a 5% concentraton reducton the wde descendng nterface theory of ISO 45- collapses nto the sharp nterface theory found n the 8 edton of NFPA. Slght dfferences n the two standards do exst ncludng the assumed ambent temperature and method of evaluatng the enclosure leakness. Due to ths, data ponts for these two theores almost always show perfect correlaton except n select nstances. The 8 publcaton of NFPA states that a mnmum concentraton of 85 percent of the desgn concentraton shall be held at the hghest level of combustbles for a mnmum perod of mnutes or for a tme perod to allow for response by traned personnel (.e. the authorty havng jursdcton may set any tme threshold deemed approprate) [6]. Ths recent modfcaton to NFPA suggests that the hold tme model presented theren wll 8
accurately predct the 5% reducton n agent concentraton as opposed to the 5% concentraton reducton, whch s assumed n Fgures and 3. Fgure 4 dsplays the result of analyzng the entre data set for each of three hold tme theores when the hold tme s assumed to represent a 5% drop n agent concentraton. Varyng the assumed concentraton reducton threshold has an affect on both the expermental and theoretcal hold tmes. Expermental values of the hold tme are found by locatng the moment n tme, for any gven nstrument s recorded data trace, at whch the agent concentraton s found to drop below the specfed concentraton reducton threshold. The obtaned value of the expermental hold tme s then rendered dmensonless accordng to the three theoretcal models, whch results n slghtly varyng values for each. The theoretcal hold tme s a functon of the concentraton reducton (below the ntal dscharge concentraton) only when consderng the wde nterface theory. ISO 45-, Annex E lmts the applcablty of the wde descendng nterface to concentraton reducton thresholds rangng up to 5% (even though the theoretcal assumptons allow for a wder applcablty range). The sharp descendng nterface theores do not mplement the percent concentraton reducton as an nput varable. As such, theoretcal hold tme values for the sharp nterface theores data ponts n Fgures 3 and 4 are dentcal although the expermental values devate sgnfcantly. Nondmensonal Theoretcal old Tme [-].5.4.3...5.4.3.....3.4.5...3.4.5 Nondmensonal Expermental old Tme [-] Exact Correlaton +/- 5% Error +/- 5% Error +/- 75% Error +/- % Error NFPA `4 - Sharp Int. NFPA `8 - Sharp Int. ISO `6 - Wde Int. Fgure 4: Valdaton plot of the nondmensonal theoretcal hold tme versus the nondmensonal expermental hold tme wth data seres grouped theory type. Plotted values are calculated as the quantty ( β t ). Expermental hold tmes are evaluated as the tme when the agent concentraton descends to 5% of the ntal dscharge concentraton. Error lnes represent the percent devaton from the theoretcal hold tme predcton. When the hold tme s evaluated at a 5% reducton n concentraton a clear trend s observed. The data set for each theory type n Fgure 4 are not equally dstrbuted about one another any more. Rather, a dstnct separaton between the sharp nterface theores and the wde nterface theory s found. The sharp nterface theores typcally result n an overly optmstc predcton of the hold tme. The expermental hold tme s usually around than 5% shorter than the theoretcal predcton but may devate by as much as 5%. Conversely, 9
the wde descendng nterface theory provdes overly conservatve hold tme predctons. The data ndcate that actual clean agent retenton tmes evaluated for a 5% concentraton reducton threshold are typcally longer than and up to twce as long n duraton as the wde descendng nterface theory predcts (where data from FC-7ea tests exhbt error of up to 5%). As seen n Fgure 4, the valdty of the hold tme predctons s greatly dmnshed when appled to predct a 5% reducton from the ntal, dscharge concentraton. Although ths may be the ntended applcaton when usng any gven theory to predct the hold tme, t s apparent that the data spread n Fgures and 3 exhbt better correlaton than that resultng n Fgure 4. Fgure 4 demonstrates ths wth a wdely scattered data pattern whch ranges from above 5% error n the overly optmstc regon (above the lne of exact correlaton) to over 5% error n the conservatve drecton. It can be concluded that the valdty of the hold tme models for all agent types s drastcally reduced when the hold tme s not evaluated as a 5% reducton n concentraton. Summary and Conclusons Ths paper documents the fndngs of a research program desgned to expermentally evaluate the applcablty of the wdely publshed hold tme predcton models found n NFPA, Annex C and ISO 45-, Annex E. Thrty four experments nvolvng a varety of enclosure leakage confguratons are presented for seven clean extngushng agents: FK-5--, FC-5, FC-7ea, FC-3, IG-, IG-54 and IG-55. Expermental results are modfed to a dmensonless form to permt drect comparson between tests. Results ndcate that the actual hold tme s longer than the theoretcal hold tme predcton when evaluated as the tme requred for the agent concentraton to drop to 5% of the ntal dscharge concentraton. Under ths condton, expermental hold tmes are typcally up to 5% longer than the theoretcal predcton. The accuracy of theoretcal hold tme predctons dmnshes greatly when the hold tme s evaluated as a 5% reducton n agent concentraton. Theoretcal predctons accordng to the sharp descendng nterface theores are typcally overly optmstc; resultng n expermental hold tmes up to 5% shorter than the predcted value. The wde descendng nterface theory typcally results n overly conservatve hold tme estmates; yeldng expermental values typcally below % but also up to 5% longer than the theoretcal value. The 8 edton of NFPA mandates that the clean agent must be retaned for the specfed hold tme duraton at no less than 5% below the ntal dscharge concentraton. Ths study ndcates that the applcaton of the theory from NFPA or ISO 45- desgn standards to the predcton of hold tme for a 5% reducton n agent concentraton wll nevtably yeld naccurate results. Dependng on the desgn standard of choce the user can expect the actual hold tme to devate from the theoretcal predcton by anywhere from negatve 5% to postve 5%. Recommendaton for Future Work The deal assumptons of spatal agent dstrbuton employed n the sharp and wde nterface models do not match the expermental results well. Ths assumpton should be reevaluated and rensttuted n the theoretcal dervaton to yeld a more robust analytcal soluton. The lkely approach would be to model the nterface wdth dfferently for the varous clean agent types; loosely based upon the tendency of an agent to dffuse n ar.
Nearly all structures are subject to bas pressure whether ntentonal (.e. VAC desgn, smoke control pressurzaton) or not (.e. stack effect). A controlled ntroducton of postve and negatve bas pressure from hgh and low elevatons should be nvestgated. If feasble, the legacy hardware ncorporated for halocarbon gas samplng should be calbrated n order to convert the relatve measurement technque to an absolute one. The varous pressure transducers used smultaneously n testng devate sgnfcantly from one another. More accurate means of montorng the enclosure pressure profle and peak pressures durng agent dscharge should be obtaned for any future testng. The coolng affect of a clean agent dscharge and resultant temperature change s not accounted for n the models, whch may lead to measurable errors n the predcted hold tme. Further analyss of these transent thermal effects s warranted. Acknowledgements Ths research effort was begun under the auspces of the NFPA Techncal Commttee on Gaseous Fre Extngushng Systems. Prototypcal work was conducted by Fke Corporaton on a full-scale and by a jont effort from 3M Company and the Unversty of Maryland on a bench-scale apparatus [4, 5]. Clean agents, system components and desgns were donated by 3M Company, Ansul Incorporated, Chemetron Fre Systems, Fke Corporaton and Kdde Fre Systems. Addtonally, Fke Corporaton provded a modern test faclty and multple techncans adng n makng ths effort possble. 3M Company, Chemetron Fre Systems, DuPont Fluoroproducts, Sevo Systems provded all halocarbon gas samplng nstrumentaton. Ansul Incorporated, Chemetron Fre Systems and Fke Corporaton provded all oxygen concentraton gas analyzers. Equpment for and executon of all door fan ntegrty testng was provded by Retrotec Incorporated. I would lke to acknowledge the ndustry specalsts who were able to fnd the tme and means to travel to Fke Corporaton and support n the test process. Contrbutng members of the research team nclude Mchael Barrera and Paul Rvers of 3M Company, Dale Edlebeck and emng Da of Ansul Inc., Coln Genge of Retrotec Inc., Danel ubert and Erc aberchter of Chemetron Fre Systems, oward ammel of DuPont Fluoroproducts, Brad Stlwell, John Schaefer, Gene ll, Allen Clem, and Mark McLelland of Fke Corporaton, Seth Senkewcz and Jm Marquedant of Kdde Fre Systems, Rchard Nemann of Sevo Systems, and Bob Whteley of Tyco Fre and Integrated Solutons. Contnued collaboraton wth Coln Genge s greatly apprecated. The practcal methods and gudance of John Woycheese provded the groundwork for all data analyss conducted. Dscussons wth faculty and fellow students at Worcester Polytechnc Insttute are gracously acknowledged. References [] J. Dewsbury and R.A. Whteley, Revew of fan ntegrty testng and hold tme standards, Fre Technology, Vol. 36, No. 4, pp. 49-65, Nov.. [] T. etrck, Analyss of hold tme models for total floodng clean extngushng agents, Fre Technology, Publshed onlne, Aprl, 8, http://www.sprngerlnk.com/content/b45854t. [3] P.J. DNenno and E.W. Forssell, Evaluaton of the door fan pressurzaton leakage test method appled to alon 3 total floodng systems, Journal of Fre Protecton Engneerng, Vol., No. 4, pp. 3-4, 989. [4] F. Mowrer, Analyss of vapor densty effects on hold tmes for total floodng clean extngushng agents, n alon Optons Techncal Workng Conference, 6th
Proceedngs, Albuquerque, New Mexco, pp.-, May 6. [5] S.T. O'Rourke, Analyss of hold tmes for gaseous fre suppresson agents n total floodng applcatons Master thess, Unversty of Maryland, College Park MD, 5. [6] NFPA : Standard on clean agent fre extngushng systems, Natonal Fre Protecton Assocaton, Quncy, MA, Annex C, 8. [7] NFPA : Standard on clean agent fre extngushng systems, Natonal Fre Protecton Assocaton, Quncy, MA, Annex C, 4. [8] J. Dewsbury and R.A. Whteley, Extensons to standard hold tme calculatons, Fre Technology, Vol. 36, No. 4, pp. 67-78, Nov.. [9] ISO 45-: Gaseous fre extngushng systems physcal propertes and system desgn part : general requrements, Internatonal Standards Organzaton, Geneva, Swtzerland, Annex E, 6.