1、344 2008 ASHRAE ABSTRACTThis paper presents work on the investigation of airentrainment in balcony spill plumes in the under the balconyand the rotating regions using CFD modeling and full-scaleexperiments. Mass flow rates near the balcony area wereexamined to evaluate the applicability of existing
2、balcony spillplume correlations. The results of this study were used todevelop an empirical correlation to calculate air entrainmentrate at the spill edge. This correlation considers the variousfactors affecting air entrainment under the balcony area andthe rotating region. Comparisons between model
3、 predictionsand experimental data indicate that the CFD predictions agreewell with experimental data both of which show a large degreeof air entrainment into the rotating flow.INTRODUCTIONIn the design of atrium smoke exhaust systems, there is anincreasing demand for consideration of the balcony spi
4、llplume that is produced by a fire in a compartment adjacent toan atrium (Lougheed 2000). Low-level compartments adja-cent to atria are typically used as commercial stores, restau-rants or offices and generally have greater fire loads than theatrium. Smoke from a fire in these compartments could eas
5、ilytravel out to the connected large atrium space, threatening theoccupants of the entire building. The balcony spill plumegenerates a greater amount of smoke than an axisymmetricplume for the same size of fire (Milke 2002). This increasedentrainment in the balcony spill plume is due to mixing at th
6、eceiling, under the balcony, at other obstructions such as afascia and entrainment in the rotating region as shown inFigure 1. Figure 1 also depicts the problem and terms used inthis paper, such as fascia, draft curtains, downstand, andbalcony. Possible factors affecting air entrainment of thebalcon
7、y spill plume are the size of the fire compartment, thesize of the door, the depth of fascia, the presence of draftcurtains, and the depth of balcony projection as well as theatrium size.To model this complicated problem of the balcony spillplume, previous studies considered the vertically rising sp
8、illplume as being separate from the fire, so that the source of thespill plume is a wide layer emerging from the spill edge. Basedon this approach, previous studies developed a number ofmethods to estimate the mass flow rate of the balcony spillplume (Ma) (Morgan and Marshall 1975, 1979; Law 1986,19
9、95; Thomas 1987; Thomas et al. 1998; Poreh et al. 1998).Despite the variation of the various methods, there seems to bea general agreement on the following key aspects of the massflow rate of spill plumes:a. The balcony spill plume incorporates the mass flow rateof the vertically moving spill plume
10、and the mass flowrate of the horizontally approaching initial flow at thespill edgeb. The mass flow rate of the plume (Ma) has a linear correla-tion with plume height zc. The slope of the linear correlation depends on QC1/3L2/3as shown in Equation (1), which is the general method formass flow rate o
11、f the spill plume developed in previousstudies.(1)whereMa= the mass flow rate of the balcony spill plume (kgs1)B = the empirical constant of the spill plumeMaBQc1/3L2/3zMs+=CFD Study of the Air Entrainment of Balcony Spill Plumes at the Balcony EdgeYoon J. Ko George Hadjisophocleous, PhD, PEng Gary
12、Lougheed, PhDMember ASHRAE Member ASHRAEYoon J. Ko is a PhD student in the Fire Safety Engineering Program at Carleton University, Ottawa, Ontario, Canada. George Hadjisopho-cleous, FSFPE, is a professor at the Carleton University, Ottawa, Ontario, Canada. Gary Lougheed is a senior research officer
13、for the Fire RiskManagement Program, National Research Council of Canada, Ottawa, Ontario.NY-08-040 (RP-1247)2008, American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc. (www.ashrae.org). Published in ASHRAE Transactions, Volume 114, Part 1. For personal use only. Additional
14、 reproduction, distribution, or transmission in either print or digital form is not permitted without ASHRAEs prior written permission.ASHRAE Transactions 345Qc= convective heat release rate (kW)L = the width of source (m)z = the height of rise (m)Ms= the mass flow rate of the initial approach flow
15、at the spill edge (kgs1)All correlations developed by the previous studies havetried to find the empirical constant B of Equation (1). Anotherimportant parameter that requires further investigation is theinitial mass flow rate (Ms) at the spill edge. The initialapproach flow rate is the focus of thi
16、s study.The Initial Approach FlowA number of calculation methods of the balcony spillplume have been developed based on small-scale experiments(Morgan and Marshall 1975, 1979; Hansell et al. 1993;Marshall et al. 1996; Harrison 2004). A significant discrep-ancy that exists among these methods is the
17、estimation of themass flow rate of the initial approach flow (Ms). The difficultyof quantifying the initial mass flow rate is mainly due to theabsence of sufficient data and lack of studies that focus on theflow under the balcony and in the rotating region. This paperaddresses this issue by focusing
18、 on the flow rates under thebalcony area and the rotation region at the edge of the balcony.Only free balcony spill plumes, which does not adhere orbound back to the wall above the balcony, are considered inthis study. The objectives of this study are described below:a. To determine the limiting hei
19、ght of riseIt has been recognized that the theoretical treatment of thespill plume revealed a problem in the region immediately afterthe rotation of the flow (Thomas 1987). The question raised isthe uncertainty in determining a limiting height of rise, nearthe balcony edge, above which the spill plu
20、me formula has alinear relationship with height. In this study, mass flow ratesnear the balcony area were examined to find the limitingheight. The limiting height should be the point at which theempirical correlation of vertical spill plume is to be defined upand Ms(the initial mass flow rate) shoul
21、d be computed.b. To develop an empirical correlation to quantify the factorsaffecting the air entrainment rate at the spill edge.The existing simple correlations of balcony spill plumeshave been developed primarily based on the common cases ofchannelled flow so that the extent of applicability of sp
22、illplume correlations and methods to calculate Ms are question-able. Therefore, it is necessary to develop a new empiricalcorrelation of Msthat addresses various factors affecting theair entrainment under the balcony area.Full-scale CFD modeling was conducted using the FireDynamics Simulator (FDS) d
23、eveloped by the National Insti-tute for Standards and Technology (NIST). Full-scale experi-ments were also performed to ensure that the CFD modelsused to measure the mass flow rate of the balcony spill plumewere acceptable.CFD MODEL DESCRIPTIONModel Geometry and Boundary ConditionsAs shown in Figure
24、 2, the modeled fire compartment is13.6 by 5.0 m in floor area and 5.0 m high with an openingfacing the atrium area. This compartment has the same dimen-sions as the full-scale test facility at the National ResearchCouncil (NRC) shown in Figure 3. As shown in Figure 2, theceiling of the compartment
25、extends out as a balcony, and thewidth of the balcony is the same as that of the compartment.For some cases, a 1.6 m deep fascia and a 3 m deep channellingdraft curtains were added. Figure 1 The problem of the flow under the balcony area and rotating region.346 ASHRAE TransactionsA free boundary con
26、dition was applied to the computa-tional domain at all boundaries except the floor, and forsimplicity, thermally inert boundary conditions were appliedto the compartment walls. The thermally inert boundarycondition resulted in constant convective heat output due to theconstant heat conduction loss t
27、o wall boundaries. Convectiveheat outputs, mass flow rates, temperatures and velocitieswere calculated by the FDS model. The formulation of theequations and the numerical algorithm are in the Fire Dynam-ics Simulator Technical Reference Guide (McGrattan et al.2002).Computational Domain and Grid Size
28、Domain size and grid convergence tests were conductedby monitoring the mass flow rate at selected planes. The accu-racy of FDS model results, flow velocities and temperatures,depends on the resolution of the numerical grid (McGrattanet al. 2002). The grid convergence tests indicated that the opti-ma
29、l spatial size in this model is 250 mm. This grid size wasapplied to the entire computational domain except in the areaof interest. A finer grid size of 125 mm was used under thebalcony and near the balcony areas in order to capture theboundary layer under the balcony and details of the rotatingflow
30、.Simulation results were found to be very sensitive to thecomputational domain size. A simple coarse grid model wastested with five different computational domain sizes to findthe optimal domain size. These tests indicated that the massflow rate is sensitive to the computational domain size eventhou
31、gh the boundaries were set to be free boundaries. Figure4 shows that the mass flow rate at z=4 m (Height = 9 m)increases as the computational domain size increases. It seemsthat an excessive computational domain size is required inorder to achieve a domain independent solution. Since theprimary inte
32、rest for this study is the air entrainment of theinitial approach flow, air entrainment rates at the rotatingregion, Ms/Mb, were processed and compared. It was foundthat although a larger volume of computational domainentrains more air, the ratio of mass flow rate Ms/Mbconvergedas long as the height
33、 of the computational domain was highenough. The optimal computational domain size was found tobe 23 m (D) 20 m (W) 9.97 m (H).Fire SimulationThe fire source, which was placed at the center of the firecompartment, was defined as a constant heat source, since thegoal of the simulations was to estimat
34、e the mass flow rate atsteady-state conditions.Parameters of Interest and VariationsFor the purpose of investigating variations of air entrain-ment rate under the balcony area and possibly quantifying theeffect of each parameter, the following parameters of interestand their variation were considere
35、d.Fire source (Q): 1, 2, 3, 4 and 5 MW of total heat releaserateWidth of opening (W): 5, 7.5 and 12 mDepth of the fascia (Df): 0 and 1.6 mBreadth of balcony (b): 4.12m fixedDepth of draft curtain(Dc): 0 and 3 mThese variations give four parameter conditions, as inTable 2.To cover the parameters and
36、their variations a number ofsimulations were performed as shown in Table 4.Figure 2 Description of the model.Figure 3 The full-scale experimental facility and the sketchof the instrumentation.Table 1. Computational Domain SizesUsed for the Preliminary RunsSize IDDepth (m)Width (m)Height (m) Ms/MbDom
37、ain QDomain PDomain NDomain CDomain DDomain E18.818.823.027.231.246.2XXXXXX162020283650XXXXXX106101010201.701.521.741.721.751.8ASHRAE Transactions 347DESCRIPTION OF EXPERIMENTAL FACILITYFor CFD validation purposes, full-scale experimentsexamining the under the balcony area, as shown in Table 3,have
38、been conducted by the National Research Council(NRC). Detailed description of the experiments can be foundin Lougheed et al. (2006).Model GeometryThe same dimensions used in the CFD modeling set-updescribed above were used for the experiments. Figure 3shows the fire compartment with dimensions of 5.
39、0 m (D) 13.6 m (W) 5.0 m (H) was built inside the NRCs burn hall.The ceiling of the compartment extends out as a balcony. Thecompartment and balcony were constructed using a steelframe lined with metal sheet. Removable draft curtains andfascia were constructed of steel and installed for selectedcase
40、s. To protect against repeated fire exposure, the interiorwalls and ceiling of the compartment were shielded withnoncombustible ceramic fibre insulation as a lining material(Lougheed et al. 2006).Fire SimulationA rectangular propane gas burner was placed at the centerof the fire compartment, which p
41、roduced the required heatreleased rate.InstrumentationThe temperature rise and velocity were obtained based ondata measured by thermocouples and micromanometers,respectively. The sketch of instrumentation map at the areaunder the balcony is shown in Figure 3.A moveable instrumentation tree containin
42、g thermocou-ples and pitot tubes was built to measure temperatures andvelocities along the centerline of the balcony. This instrumen-tation tree, which could move along a rail underneath thebalcony, collected temperature and velocity data at the desig-nated points. For measurements at the rotating r
43、egion, theinstrumentation tree was rotated upwards so that data could becollected at 30, 60 and 90 to the vertical.RESULTSFlow BehaviorFigure 5 shows velocity vectors on a vertical plane at theend of balcony and a horizontal plane at z=2 m, for Simula-tion 17. The figure demonstrates the flow behavi
44、or under thebalcony area which occurs with simulations without draftcurtains. The flow out from the doorway fully spreads outexceeding the width of the balcony and escapes to the reservoir(the large volume of computational domain representing theatrium) at both sides of the balcony. Mass loss at bot
45、h sides ofthe balcony was observed in all simulations as well as exper-iments for cases without draft curtains. This study is mostTable 2. Parameter Conditions Used in the ModellingParametersConditions FasciaDraft curtains DescriptionCase-YNCase-NNCase-YYCase-NYYNYNNNYYunchannelled flow with fasciau
46、nchannelled flow with no fasciachannelled flow with fasciachannelled flow with no fasciaTable 3. Experiments Conducted at the NRCExperi-ment Test ID Q (MW) W (m) FasciaDraft Curtain1234567T1M5NYT3M5NYT1M5YYT3M5YYT5M5YYT5M5NNT5M5YN13135555555555NNYYYNYYYYYYNNFigure 4 The computational domain size eff
47、ect on the mass flow rate.348 ASHRAE Transactionsinterested in entrainment rate of the wide layer flow roundingthe edge of the balcony, and the mass flow at the sides of thebalcony should be excluded for analysis of entrainment intothe rotating flow since the width of horizontal flows at the edgeof
48、the balcony can not exceed the width of the balcony. Thus,mass flows excluding the mass losses at the sides of thebalcony were used for analysis. Mass flow rates under thebalcony area as well as at the spill plume area were measuredby taking account only the flow through planes of interest asdepicte
49、d in Figure 5. In this way, mass loss at both sides of thebalcony did not affect results because the mass loss was dissi-pated into the large reservoir and was assumed to be part of theambient air. The air entrainment rate was obtained by normal-izing the mass flow rate of the vertical flow to the mass flowrate of the horizontal flow at the end of the balcony.Comparison of CFD Results with Experimental DataTemperature and velocity distributions were comparedalong the centre line at the doorway, at the end of the balconyand at the edge of the balcony. Figure 6 shows a comparison
