1、178 2008 ASHRAE ABSTRACT A multi-stage compartment fire model HAICFMRail wasdeveloped for predicting the conditions inside an enclosurewith multiple openings and multiple combustible materialsurfaces. HAICFMRail was successfully validated with gastemperatures and burning rates measured in compartmen
2、t firetest data reported in the literature. HAICFMRail was then usedto predict the conditions that develop inside of a railcarcontaining a fully-developed fire. The railcar modeling resultsindicate that the heat release rate from a railcar is sensitive toinitial ventilation into the railcar, fire pr
3、operties of interiorfinish materials, and window failure. Increasing the number ofdoors initially open from one to two caused the peak heatrelease rate inside the railcar to increase from 13,600 kW to20,500 kW. Material burning rate and duration were alsoobserved to have dramatic affects on the heat
4、 release ratecurves. Window failure provided more air to the fire. If the firewas fuel rich, then the additional air would allow more heatrelease rate inside the railcar which resulted in higher gastemperatures and a larger overall heat release rate. Windowfailure with a fire close to stoichiometric
5、 burning caused thefire to transition into the decay stage. INTRODUCTIONMass transit systems are becoming more numerousthroughout the world. In many of these systems, the railcarsoperate in underground tunnels and stations. In accordancewith NFPA 130 “Standard for Fixed Guideway Transit andPassenger
6、 Rail Systems” 1, these stations and tunnels needemergency ventilation to ensure that passengers and crew cansafely evacuate these areas to a point of safety. A fire involvingthe railcar is one of the worst-case scenarios considered in thedesign of the emergency ventilation system. As a result, theh
7、eat release rate of the railcar fire is an input needed for thedesign of the emergency ventilation system. Data on fully-developed fires inside of actual railcars islimited; however, there have been some large-scale testsconducted on railcars inside tunnels to support tunnel designprojects 2-5. A de
8、tailed description of the tests is provided inRef. 5. The heat release rates 2,3,5and gas temperatures4,5 were measured for an aluminum exterior shell subwayrailcar with internal dimensions of 18.0 m long, 2.8 m high, 3.0m wide and two different intercity railcars both with steel exte-rior shells an
9、d internal dimensions of 26.1 m long, 2.4 m high,2.9 m wide. The subway railcar and intercity IC-traincontained older interior finish materials, while the intercityICE-train contained newer interior finish materials. All rail-cars had glass windows. Initiating fires in the tests were pans of isoprop
10、anollocated at one end of the railcar. Prior to ignition, windows andsometimes doors were opened to allow air to enter into the rail-car. The measured heat release rates and gas temperatures nearthe ceiling are provided in Figure 1. The subway railcar firehad a peak heat release rate of 35 MW, while
11、 the longer inter-city railcars had peak heat release rates of 13-20 MW. The IC-train with older interior finish materials had a lower heatrelease rate than the ICE-train. Figure 1(b) contains gastemperatures 0.020 m below the ceiling inside these three rail-cars and inside a 13.8 m long subway car
12、with a steel exteriorshell. Temperatures during the fully-developed fires rangedfrom 400-900C. Shorter subway railcars were measured tohave uniform temperatures along the length, while longer rail-cars had large temperature changes along the railcar length 5.This could be due to the initiating fire
13、and initial ventilationApplication of a Multistage Compartment Fire Model in Predicting the Heat Release Rate History of Railcar FiresBrian Lattimer, PhD John CutonilliBrian Lattimer was with Hughes Associates Inc. while researching this paper and is now Associate Professor in the department of Mech
14、anicalEngineering at Virginia Tech, Blacksburg, VA. John Cutonilli is an engineer with Hughes Associates, Inc., Baltimore, MD.NY-08-0242008, American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc. (www.ashrae.org). Published in ASHRAE Transactions, Volume 114, Part 1. For per
15、sonal use only. Additional reproduction, distribution, or transmission in either print or digital form is not permitted without ASHRAEs prior written permission.ASHRAE Transactions 179openings being located at one end of the railcar and the failuretimes of the windows during the fire. Windows were h
16、eardbreaking as early as 2 minutes into the test and as late as 42minutes, but in all cases all windows were blown out at the endof the test. In some tests, sounds of shattering windows didcorrespond with abrupt changes in temperatures 5.The focus of this paper is to present an approach formodeling
17、the heat release rates of fully-developed fires insideof railcars. Models were validated against several sets ofcompartment fire data from different studies. Models couldnot be conducted on the railcars used in the large-scale testing5 due to insufficient information the interior finish materialsand
18、 window failure times. The validated models were thenused to provide predictions of railcar heat release rates for asubway railcar. MODELING APPROACHPredicting the heat release rate history of railcars is gener-ally performed using two separate models. A fire growthmodel, HAIFGMRail, is used to pred
19、ict the ignition and flamespread along railcar interior finish materials due to a userselected initiating fire. HAIFGMRail is an improved versionof the fire growth model described in the literature 19,20capable of including multiple interior finish materials on thewalls and ceiling. HAIFGMRail predi
20、cts the initial fire devel-opment and determines whether a fire causes the railcar toreach flashover. If flashover does not occur, the predictionfrom HAIFGMRail is the heat release rate history of the fire.When flashover does occur, all materials inside the railcarignite and burn. For these scenario
21、s, a multi-stage compart-ment fire model, HAICFMRail, is used to determine thehistory of the fire including conditions during the initial devel-opment, fully-developed, and decay stages. HAICFMRailuses the heat release rate from HAIFGMRail as input for theheat release rate up to flashover. Both mode
22、ls provide the totalheat release rate, gas temperatures inside the railcar, and flowsin and out of the railcar. The focus of this paper is on predicting the heat releaserates of a railcar that reaches flashover using HAICFMRail. Inthese simulations, the initial fire growth was representative ofa fir
23、e where flames spread readily over the interior finish caus-ing the railcar to reach flashover in 60 seconds. HAICFMRailused this fire growth curve to initiate burning inside the railcarand predict the heat release rate history of the railcar. Compartment Fire Model (HAICFMRail)The compartment fire
24、model HAICFMRail is used todetermine compartment conditions and burning of materialsduring all stages of a compartment fire (initial development,fully-developed, and decay). Model features include flow inand out of the compartment through one or more vertical open-ings which may have different sill
25、heights (i.e., doors andwindows), predicting window failure time, multiple materialswith different properties burning in the compartment, effectsof compartment thermal environment on material burning,heat release rate inside and outside the compartment, and heatlosses through multiple boundaries wit
26、h different construc-tion. The output from the model includes time varying totalheat release rate, gas temperatures, flow rates in and out of thecompartment, window failure time, remaining mass ofcombustible materials, mass loss rates of combustible materi-als, heat release rate contribution of indi
27、vidual materials, heatrelease rate inside the compartment, and equivalence ratio.The one layer model is a classical method for calculatingcompartment conditions during a fire 11. A one layer modelis typically used to calculate post-flashover fires because theinterface has already moved to a height n
28、ear the floor andconditions are generally uniform during the fully developedburning period. Several researchers have also shown successin using a one layer model to approximate compartment condi-tions 11, 12, 16. The one layer model was chosen because it1a 1bFigure 1 Large-scale railcar test data 5,
29、 (a) heat release rate and (b) gas temperature.180 ASHRAE Transactionsis a simple proven method that can adequately represent arange of fully developed fire conditions. The model is an enhanced version of a previouslypublished model 16. The model itself predicts conditionsinside the compartment base
30、d on a mass and energy balance ofa control volume surrounding the compartment as shown inFigure 2. This approach is the same as proposed in other anal-yses on pre and post-flashover fires 11, 12. Assuming thepost-flashover compartment fire is a quasi-steady stateprocess, the governing equations for
31、mass and energy, respec-tively, are (1)with the mass flow being calculated by:(2)(3)(4)The boundary heat losses were determined using a finitedifference approach capable of calculating heat transferthrough multiple materials composing the boundary, (5)where T is the gas temperature inside the compar
32、tment. Theradiation losses through the vent were calculated from (6)The energy loss due to pyrolyzing the material was deter-mined using the following relation,(7)The specific heat capacity of the incoming air wasassumed to be constant at Cp,air=1.0 kJ/kg-K, making theenthalpy of the incoming air,(8
33、)The specific heat capacity of the fuel from the burningmaterial was assumed to vary with temperature. The enthalpyof the fuel was calculated from(9)The specific heat capacity was taken to be that of propane13, which has a specific heat capacity that is representativeof hydrocarbons and many other f
34、uels. In order to determinethe enthalpy of the outgoing gases, the enthalpy of the hot gasmixture was calculated as follows(10)As seen in the above equation, each enthalpy is scaledusing the mass fraction of the component in the outgoingstream. The mass flow rate of fuel leaving the compartmentwas d
35、etermined by subtracting the fuel burned inside thecompartment from the total fuel pyrolyzed.The mass loss rate of each fuel was predicted from thefollowing equation(11)For charring fuels, the pyrolysis temperature is taken to be75C less than a pseudo-steady-state surface temperaturebased the impose
36、d heat flux and current compartment temper-ature. The heat release rate of the fire was the minimum of theheat release rate of the pyrolyzed fuel and the heat release ratethat the air into the compartment could support, (12)The combustion efficiency, , accounts for how well theair mixes and reacts w
37、ith the fuel. The equivalence ratio of thefire was calculated from(13)momimf+=mi23CdWZnZsill()32Znmin ZsoffitZn,()()32=i2gioi-12mo23CdW ZsoffitZn()32Znmax ZsillZn,()()32=o2gioo-12Qqboundqventqpyrolmihi mfhf moho+ +=qboundAbound j,sgasT4Tw4()hconvTTw()+j 1=n=qventAvent j,gas T4Ti4()j 1=n=qpyrolmfj,hg
38、j,j 1=n=hiCpair,TiTref()=hfCpf,TdTrefT=Figure 2 Energy balance on compartment controlvolume.hoYairCpair,TTref()Yfhf+=mfj,AjsgasT4Tpyrol4()AjhconvTTpyrol()+hgj,-=Qmin mfj,Hc3000mj,j 1=n= mfmair()r=ASHRAE Transactions 181where, (14)The most important feature of the model is the ability ofthe model to
39、predict the burning rate of multiple materials. Anunlimited number of materials can be modeled. The modelallows a user specified burning rate, but the model will alwayslimit the maximum burning rate based on the surface area,mass and compartment conditions. During the early part andlate stages of th
40、e fire (growth and decay stages), the burningrate is based on radiative feedback from the compartment andfrom the fire itself. The burning rate will solely be based onradiative feedback from the compartment if the compartmentreaches fully developed burning.The model also has sophisticated flow rate
41、and heat trans-fer routines. The flow rate routines calculate flows into and outof the compartment based on compartment temperatures. Anunlimited number of vents can be specified and the openingsof each of these vents can change with time. Window failure,which results in additional vent openings in
42、the railcar, waspredicted through a time for the unexposed side of the windowto reach a failure temperature. The heat transfer routines allowa compartment to be divided into an unlimited number ofdifferent heat transfer regions or boundaries. For example, thewalls, ceiling, floor, and windows can al
43、l be specified as differ-ent regions. Each of these regions can be specified with multi-ple materials and each material specified with temperaturedependent properties. This allows a window to be specified asa single material window, while walls, ceilings and the floorspecified with multiple material
44、s (i.e., lining and insulation).HAICFMRail ValidationHAICFMRail has been validated against several differentsets of compartment fire data to demonstrate its ability topredict gas temperatures and mass loss rates of burning mate-rials. Examples of some of the validation performed onHAICFMRail are pro
45、vided in this section. The compartmentconditions and burning behavior of liquid and solid fuels insidea compartment fire were measured in a series of twelvecompartment fire tests conducted by Bullen and Thomas 14.The compartment was 2 m wide, 1 m high, and 1 m deep withboundaries constructed of 12 m
46、m thick asbestos board and 6mm thick ceramic fiber board. Calculations were performedusing 18 mm thick boundaries and ceramic fiber board thermalproperties of k=0.00010 kW/m-C, C=1.0 kJ/kg-C, and=240 kg/m315. Burning was initiated by providing the fuelsurface with a flame heat flux deduced from the
47、free burningrates reported by Bullen and Thomas 14 of the different fuels.Results of the validation are provided in Table 1 using agas layer emissivity of 0.90, pre-flashover efficiency of 0.80,and fully-developed fire combustion efficiency of 0.70. Thesevalues provided the best agreement between th
48、e model and thedata. The gas emissivity is less than 1.0 due to the scale of theexperiment. The model was determined to agree to within10% of the gas temperature data and within 20% of themass loss rate data. In addition to being able to predict thesequantities, HAICFMRail was also able to distingui
49、sh whethera compartment would reach flashover or remain in the pre-flashover stage. HAICFMRail was also used to predict gas temperaturesfor different size fires inside reduced and large-scale compart-ments of tests performed by Bryner et al. 21. In these tests,a natural gas burner was used as the fire source. The reduced-scale c
