4065 (RP-664)

1997, American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc. (www.ashrae.org). Published in ASHRAE Transactions 1997, Vol 103, Part 2. For personal use only. Additional distribution in either paper or digital form is not permitted without ASHRAEs permission.

Convective Heat Transfer in Building Energy and Thermal Load Calculations

Daniel E. Fisher, Ph.D. Associate MemberASHRAE ABSTRACT

Curtis O. Pedersen, Ph.D. Fe/Iow ASHRAE isothermal and nonisothermalinteriors and with both ceiling and sidewall diffusers. The research showed that actual roomfilm coefficients may be up to 20 times higher than natural convectionbased coefficients. BLAST simulations showed that the error introduced by the natural convectionassumption typically results in an annual calculated coolingload error on the order of 10%. Significantly, the research also showedthat roomswith radial ceiling diffusers are relatively well stirred, evenat ventilative flow rates as lowas 3 ACH. unexpected led to the This fact simplified engineering correlations that are presented in this paper: heat transfer coefficients as a function only of ACH. The correlations can be easily implemented roomthermal load in programs, building energy analysis pro~ams,and other engineeringcalculations. Thepaper also demonstrates importance selecting the the of correct reference temperaturein defining the film coefficient. The uncertainty analysis associated with the experimental method argues in favor of a roominlet reference temperature. The temperatureand flow fields of recirculating cavities and enclosureshaveenjoyedconsiderableattention in the literature during the last 20 years. Hundreds papers present the of results of computational studies, and dozensmorepresent experimental results. Althoughmanyresearchers have studied the buoyancy-drivenenclosure (Bauman al. 1980; Nansteel and et Greif 1981, 1983; Bohnand Anderson 1984, 1986; Allard et al. 1990; Chenet aL 1990) and a significant number have investigated forced convection highventilative flowrates (Spitler et at al. 1991b;Nielsen et al. 1978, 1979; van der Kooiand Bedeke 1983; van der Kooi and Forch 1985; Murakami al. 1987; et Neiswanger al. 1987), relatively few haveinvestigated encloet sure heat transfer at lowventilative flowrates (Neiswanger aL et 1987; Chenet al. 1989; Kapoorand Jaluria 1991;Pavlovic and Penot 1991). Of these, none has developed convective heat transfer correlations for realistic room configurations.

ValMenclosure film coefficients, required by hourly energy and thermal load programs, were experimentally determined for ventilative .flow rates below 12 air changes per hour (ACH).Forty-eight experiments were perfolw~ed a full-scale roomin order to determine fihn coefficients at low ventilative flow rates. The experiments, which were performed over a range of conditions from .3 to 12 ACH,showed that for most roomconfigurations, natural convection film coefficients significantly underpredict the rate of surface convective heat transfer. The new film coefficients were implemented in the BLASTprogram, an hourly heat-balance-based building energy simulation. The significance of the error incurred by using natural convection film coefficients in a ventilated space was estimated by comparing BLAST results obtained with both natural convection and the new.film coefficients. For the case o[ space cooling, where surface-to-air temperaturedifferences are relatively small, errors in calculated space cooling loads were typically on the order o[10%. INTRODUCTION Calculation of roomsurface-to-air heat transmission is dependent an accurate estimate of the film coefficient. Fortyon eight experimentswereperformedin a full-scale room order in to determine the significance of the error incurred by the common practice of using natural convectionheat transfer coefficients in a mechanically ventilated room. The experimental work, which was sponsored by ASHRAE RP-664, was performed over a range of realistic as roomconfigurations. Ventilative flow rates were set between 3 and 12 air changesper hour (ACH), inlet air temperature and setpoints were specified between 10C (50F) and 25C (77F). The experimental room was configured with both

Daniel Fisher a researchengineerandCurtis O.Pedersen a professoremeritus the University Illinois, Urbana. E. is is at of

ASHRAE Transactions:

Research

137

This study used the experimental building enclosure constructedby Spitler et al. (1991a)and designedoriginally for convective heat transfer experimentsat high ventilative flow rates. The facility, which was modified to accommodate research in the mixedand natural convectionflow regimes, is uniquein several respects. First, the enclosure is relatively large--thesize of a smalloffice. Second, interior surfacescan all be temperaturecontrolled. For many the experimentspresented in this paper, all of roomsurfaces were controlled at the same temperature. The isothe~rnal room configurationwas an importantfactor in minimizingthe uncertainty associated with the experimentally determinedfilm coefficients. For this configuration, the radiation component the surface heat transfer was small compared of to the magnitude the convective flux; even a relatively large of uncertaintyin the radiationheat transfer calculationhadan insignificant effect on the uncertaintyassociatedwith the convective heat transfer coefficient. A number of nonisothermal experiments were also performed order to validate the application of correlations to in roomswith nonuniformsurface temperatures. The nonisothermalroom configuredwith three "hot" walls, a "hot" ceiling, was a "hot" floor, and one "cold" wall. THE EXPERIMENTALFACILITY The experimentalfacility, shown Figure 1, consisted of in an office-sized chamber located inside a larger structure. The space betweenthe outer shell and the experimental enclosure was a temperature-controlled "guard space" that maintained a temperaturedifference of -0C (32F) across the walls of the experimentalenclosure. The enclosure, which consisted of 53 individually controlled heated panels, was ventilated through either of two inlets located in the ceiling and side wall. The layout of the heatedpanelsand the air inlets andoutlets is illustrated in Figure 2. Constructionand validation of the facility are discussed by Fisher (1989).

One wall of the experimental room (S. Wall) was constructedwith plate heat exchangers behindthe heated panels. All of the nonisothermal roomexperiments were performed by circulating chilled water through these heat exchangers. The construction and instrumentationof the cold wall are discussed by Fisher (1995) and Mansfield(1993). Eachheated panel was instrumentedwith two surface thermocouples.Local heat transfer coefficients were calculated using the average of the two panel surface temperatures. The calculation of the interior radiant energyexchange basedon was the assumption tire panel surface wasuniform the average that at of the twomeasured temperatures.At the lowairflow rates typical of the mixedconvection regime, temperature differences betweenany two points on a single panel were less than 0.5C (0.9F) (Fisher 1989). The uncertainty introducedby this error wasincorporated the calculationof the surfacefilm coefficient in uncertainty. A chilled-water coil with electric reheat maintainedthe desired roomair inlet temperature. Thermocouple grids located at the roo~n inlets andoutlet measured air enteringand leavingair temperatures. A flow measurement constructed according to box ANSI/ASHRAE Standard 51-1985 (ASHRAE 1985) provided pressure and temperatureinformation required to calculate the air massflow rate to the room. Theroom ventilation rate was calculated fiom the measured p~essuredrop across one or two elliptic flow nozzles of known characteristics. Various nozzle combinations were used to obtain ventilative flow rates rangingfiom.05 to 2 m3/s. Ceiling

N. Wall

l~oor

fW. Wall S. Wall

E. Wall

W = Inlet or OutletOpenings
Figure I Schematicof experimental facility (Fisher 1989). 138 Figure 2 Heated panel layout (Fisher 1989). ASHRAE Transactions: Research

Air speeds and temperatures in the experimentalroomwere measuredby 16 omnidirectional air speed transducers and 16 type-T thermocouples attached to a computer-controlled trolley (Cantillo 1990).Thetrolley, moving horizontally and vertically, typically collected data at 1,000locations in the half-room. Air temperaturemeasurements were particularly important in evaluating possible reference temperatures.A number bulk of air and planar temperatures wereinvestigated as possible referencesfor definingthe heat transfer coefficient. THE EXPERIMENTAL METHOD Eachunique experimentwas defined by a roomventilative flow rate, an inlet air temperature, and roomsurface temperatures. Temperatures flow rates werecontrolled to specified and setpoints until roomair outlet temperatures and time-averaged surface heat fluxes werenearly constant, indicatingthat a quasisteady state had been reached. The computer-controlled trolley then traversed the roomon a specified path logging air speeds and air temperatures; a seconddata-acquisition systemlogged experimentalvariables at 20-second intervals for at least one hour. Thesurface convectiveflux and film coefficient calculations were based on periodic and one-time measurementsat these quasi-steady-stateconditions.Table1 definesthe variables and parametersused in these calculations.

whereqi-pwr is the rate of heat transfer to the panel from the electrical resistance heaters, qi-s represents the radiant exchangebetweenroomsurfaces, and qi-hl is the "back loss" to the guardspace. The electrical powerinput to each panel was calculated fromthe measured voltage, the measured resistance of the heated panel, and the measured area of the panel: V2 (W/m2).
qpwr - R. A i

(2)

The net radiant heat transfer from the ith roomsurface to all the other surfaces in the enclosure is given by Hottel and Sarofim (1967)
qi-~"

1
Ai.~ j=l [Fi] .~

(T~ G" -

2 4. Tj)I(W/m ).

(3)

Finally, the rate of heat transfer fromthe inside surface to the guard space (the "backloss"), qhl, is calculated fromthe measured surface temperaturesand the thermal resistance of the surfaces: qi-h,
[ri- Tg,,~ =

~)(W/m-).

(4)

Thus, the convective flux was explicitly calculated from experimental measurements for each surface in the room. TABLE 1 Since the guardspace was controlled to the inside surface temFilm Coefficient Calculation: perature, for the isothermal roomconfiguration the only sigParametersand Variables nificant termon the right-handside of the equationis qi-pwr" 2) of heatedpaneli (m A Area The convective heat transfer coefficient was calculated i from the rate of convective heat transfer and the temperature z) Grayinterchange factor between panelsi andj (m F 0difference between surface and an arbitrarily selected referthe h Convective transfer coefficientfor paneli (W/(m2/C)) heat i ence temperature. qi-s = net rate of radiation transferfrom heat panel to all i hi - (Ti- Trej)O 2) surfacesin enclosure other (W/m qi - con v (5) Rate heat qi-co,n, 2) of convective transfer frompanel i (W/m Rate electrical resistance transfer2) paneli (W/m of heat to Theselection of the reference temperaturewas arbitrary in the (W/m qi-h! Rateof heat transfer frompaneli to guardspace2) sense that for enclosure heat transfer, a clear and obvious 2. C/W) choice for a temperature reference does not exist. AnimporR Thermal resistance of wall (m t tant part of the investigation wasto examine impactof varthe Stefan-Boltzmann constant (W/(m2/K4)) ious reference temperatures on the proposedcorrelations and Reference temperature (C) Trq on the experimental uncertainty associated with the heat transfer coefficients. The roominlet temperature, the roomoutlet Ti, T./ Average surfacetemperature panels i andj (C) of temperature, and spatially averaged planar and bulk air temGuardspacetemperature (C) peratures were examined possible references. as Linevoltageonpaneli heater(volts) V i SUMMARY OF EXPERIMENTS Of the 48 experimentsperformedduring the course of the investigation, 16 wereperformed radial ceiling diffusers in with the isothermalroomand 20 with radial ceiling diffusers in the nonisothermal room. In addition, several heat gain experiments were performedwith reduced roomvolumein order to estimate the effect of furniture. Thebalanceof the experiments studiedthe free, horizontaljet andare not discussedin this paper.
139

Theconvectiveflux fromeach of the 53 surfaces, qconv, was calculated from spatially and temporally averaged data. An energybalanceat the surface defines the convectiveflux. For a heatedsurface, assuming participation of the enclosureair in no the radiant exchange,the energybalance is
qi-conv = qi-pwr-qi-s-qi-h/

(1)

ASHRAE Transactions:

Research

Theradial ceiling diffuser, used in many commercial appli7but = air outlet temperature, t cations, is of great practical importance. Severalradial diffusers Tint = air inlet temperature, of different design and openarea were used in the ceiling jet = ACvoltage, and experimentsin order to vary the outlet velocity of the jet and Vi = panel electrical resistance. study the effect of diffuser type on surface convection.Table2 Ri shows experiment the code,the ventilative flowrate, the air inlet This preliminaE screening ensured that the facility was temperature, the diffuser openarea for eachof the isothermal and indeed at quasi-steady state and that the data-acquisition sysroomceiling experiments. tem was functioning properly. Calculation of the standard deviation of all measurements over the course of the experiTABLE 2 ment served as an additional check on both instrumentation Summary Isothermal Room of and the data-acquisition system. Finally, uncertainties in both Ceiling Inlet Experiments measured values and calculated results were computed for FlowRate Inlet Temp.DiffuserArea each data set. File 2) (ACH) (m EVALUATION OF THE FLOW e0112933 10.83 0.022 AND TEMPERATUREFIELDS e0905921 3.00 19.98 0.0085 e1010921 e1211921 e1204922 e0828921 e101692l e1202921 e0112932 e1204921 e0619921 e0112931 e0827921 e0624923 e0829921 e0618921 3 ~02 5.95 599 16.72 20.04 9.98 15~04 19.95 14.95 9.91 9.92 20.08 10~02 14.99 25.03 15.03 20.02 !0,0051 0.0051 0.0116 0.0085 0.0116 0.0116 0.022 0.0116 0.022 !0~022 0.0085 !0.022 0.022 0.022 Threemethods wereused to examine validated data sets. the First, raw air speed and air temperaturedata were renderedas color or gray scale mapsin order to visualize the flow field. Second,a large data analysis program, originally written by Spitler (1990),calculatedheat transfer coefficientsbasedonvarious reference temperaturesas well as dimensionlessgroupsand experimentaluncertainties, all of whichwere required for the development CO~Telations. of Finally, special attention was given to the inlet jet. Foreachdiffuser type the ceiling jet wascharacterized by additional air speedmeasurements the ceiling. near The Effect o~ Buoyancyand Coanda Forces on Room Convective Heat Transfer Visualizationof the flow field led to the importantconclusion that over the range of experimental measurements, rate the of convective heat transfer was relatively independentof the densityof the inlet jet. Thegravity force that pulls a dense,cold ceiling jet towardthe floor is opposedby the viscous Coanda force, whichpulls the jet in the opposite direction--toward a low-pressure regionnear the ceiling. For jet inlet temperatures as lowas 10C (50F)and jet inlet velocities as lowas 1.3 m/s (4.2 ft/s), the Coanda effect dorninatedthe flow field and the jet did not detachfromthe ceiling. Theflow and temperaturefields were qualitatively evaluated using air speedsand temperatures measured quasi-steadyat state conditions The measurements were taken at 1,000 locations in the half-roomby the the~mocouples air speedprobes and attachedto the trolley system.Figures3 through5 show single a plane of measured speedsfor a givenset of inlet conditions. air The outline of the "data cube"(measurement region) with the air inlet and outlet is shown illustrate the orientationandlocation to of the measurement plane. The measurement plane itself shows only air speeds abovethe specified threshold velocity~ "Lowspeed"areas in the measurement plane are "transparent." Figures 3a and 3b showair speeds of morethan 20 fpmand 15 fpm for a volumetric flow rate of 9 ACH an inlet air and temperatureof 10C(50F). Since the air speed probes did not penetrate the attached ceiling and wall jets, the "high-speed" ASHRAE Transactions: Research

6.05 9~00 9.05 9.07 11.94 11.94 11.96

Eachdata set wasinitially screenedby calculating an overall room balancefromthe data. It wasrequiredthat the total heat energytransferred to the room duringthe courseof the experair iment, Eair, be within 10% the total electrical energydissiof pated by all room panels, Epwr:
(Eair -Epwr) <- +0.1 (6)

where
Eai r = ~ thtCPair.(Touttt=l n Epwr 53 V~ i

Tint)~

etitne

= Z
t=l

e tune "

2 ~

i=1

tht = massflowrate at timet, Cpair = specificheat of air, etime = timestep, 140

Air

20ft./min. (0.t0 m/s)plane

15ft./min:(0.076 plane m/s)

(a)
Figure 3 Measuredair speeds for 9 ACH,10C: above 20 fpm (a) and above 15 fpm (b).

(b)

(a)
Figure 4 Measuredair speed for 6 ACH,IOC: above 20 fpm (a) and above 15 fpm (b). flow regions represent the turbulent mixingregion rather than the inner flowregionof the .jet. Similar air speed patterns were observed at both higher and lower volumetric flow rates. The flow fields shownin Figure 3 are typical of the Spitler (1990) data and the higher volumetric flow rate experiments performed as part of this investigation. The characteristic flow is also observed at lower volumetric flow rates. Figure 4 shows the same plane of air speed measurements inlet condifor tions of 6 ACH 10C (50 F). Even at this relatively low at volumetric flow rate and low inlet air temperature, the buoyancy of the jet is not sufficient to overcome the Coanda effect. At the lowest volumetric flow rate and lowest air inlet temperature, the effect of buoyancy begins to show. In Figure 5a the ceiling.jet is not visible. This is not because it does not exist, but rather because the jet-affected zone ASHRAE Transactions: Research

(b)

is not penetrated by the air speed probes In contrast, the flow field near the floor, fed by the downward flow of cold air, is relatively active The "still air core" is discernible in Figure 5b but is less well defined than at higher flow rates. The data presented in Figures 3 through 5 indicate that the density differential between the cold, mechanically driven jet and the roomair has a negligible effect on the flow field in comparison to the Coanda effect. The jet remained tenaciously attached over the entire range of conditions such that even the flow along the floor appears to be driven by the attached ceiling and wall jets. Figure 6 shows measured temperatures for points coincident with the air speed measurements shown in Figures 3 through 5. Isotherms at 0.5 C (0.9 F) intervals are shownon the plots. 141

15fi/rnin (0.076m/s)plane

(b)
Figure 5 Measuredair speeds for 6 ACH,10C: above I51~pm(a) and above ]O fpm (b).

(b)
Figure 6 Temperature mapsat 9 ACH(a), 6ACH (b), and 3 ACH IOC inlet for At 9 ACH (Figure 6a), the room appears to be relatively "well stirred" with respect to temperature. There is no sign of stratification, and the temperature in the measurement plane varies by less than 1 C (1.8F). At 6 ACH (Figure 6b), the room still relatively well stirred withrespect is to temperature. The "cold corners" (also clearly visible at 9ACH) the jet leaving the ceiling and reattaching to the show sidewallsbut donot indicate a buoyancy effect. Theunistrut trolley rails mounted the upperright and left corners of the room in tend to exaggerate effect of the uppercorners. the Although room temperatureis still quite uniform the air at 3 ACH (Figure 6c), some temperaturestratification is evident this volumetricflow rate. Buoyant forces are no longer negligible, and one would theoretically expectto see their effect on the rate of convectiveheat transfer----especially on the floor. The momentum the mechanicallydriven jet, however,still domiof nates the flow field, evenat these extreme conditions. Selecting the Reference Temperature Two criteria wereapplied in selecting the reference temperature for room convectiveheat transfer calculations. The first 142

(c)

criterion requiredthat the referencetemperature readily availbe able both in energysimulation programs to building designand ers in general. Thesecondcriterion required that the reference temperature minimizethe experimental uncertainty associated with the calculation of the heat transfer coefficients. Basedon these criteria, the room inlet temperature the best reference was for definingroom convective heat transfer coefficients, Thefizst criterion eliminated but three possibilities: the all roominlet temperature, the roomoutlet ternperature, and the "bulk" (or mean)air temperature Other possible reference temperatures, such as near-wall air temperatures, can only be obtainedin the laborato~3.Of the three possibilities, the room inlet and outlet temperaturesare mostreadily available both in practice and in energy simulation programs. AlthoughSpider used the roomoutlet temperature as the reference the convective transfer coefficient(Spitler et al. for heat 1991b),at lowventilative flow rates the uncertaintyin the film coefficient associated with an outlet reference temperatureis unacceptably high. At low volumetric flow rates, the outlet temperatureapproachesthe room surface temperature. This small AT leads directly to highunce~Xainties the film coefficients. in ASHRAE Transactions: Research

20.0 15.0 10.0 5.0 o.o

2.5 2.0 1.5 1.0 0.5 o.o

-o.5
-lO.O__ - I 0 5 10 15 Air Changes Hour(ACH) per -1.0 I 5 10 Air Changes Hour(ACH) per

(b)
Figure 7 Uncertaintyin wall h for outlet reference (a) and inlet reference (b Figure 7 showsthe calculated uncertainty in film coeffia 1-in. (25.4-mm) with the pan-typediffuser. Thethird case, gap cients using inlet and outlet reference temperatures.Thefigure withan inlet jet velocityof5.11rn/s (16.7ft/s), is for the cylinshowsaverage wall heat transfer coefficients for 15C(59F) drical-type diffuser shown Figure8. Asthe inlet velocity is in inlet air experiments.Theuncertainties shown the figure are in changedby greater than a factor of four with a constant volutypical of room transfer coefficients over the rangeof experheat metric flow rate of 12 ACH, convective flux remains relathe imentaldata tively constant. The variations in the convective flux are Basedon the experimentaluncertainty, the inlet tempera- explainedby slight variations in inlet temperature, volumetric ture is clearly the better choicefor a referencetemperature. The flowrate, and perhapsalso by the jet attachment point. plots also indicate that the inlet referencetemperature corremay late the data better. Theheat transfer coefficients basedon the EXPERIMENTAL RESULTS inlet temperature showedgood sensitivity to ACH over the Development Correlations of entire rangeof conditions. The Effect of the Radial Diffuser on Room Convective Heat Transfer Fora givenvolumetric flowrate, the "openarea" of a ceiling diffuser determinesthe velocity of the .jet at the room inlet. Experiments were designedto answerthe question of whetheror not the rate of convective heat transfer is dependent the inlet on jet velocity The experimental evidencepresentedin the following paragraphs indicates that for a radial diffuser, surface convection independent the inlet jet velocity. is of A numberof tests at 12 and 6 ACH were performedin the isothermal roomwith 20C (68F) inlet air and 30C (86F) surfaces For each test, the velocity of the jet at the inlet was varied for a constant volumetricflow rate by changing open the area of the radial ceiling diffuser. Two types of diffusers were used for the tests, as shown in Figures 8 and 9. The first type was simply a cylindrical cover with evenlyspacedslots or holes on the cylinderwall; the second type wasan adjustable pan-typediffuser. The measuredrate of convective heat transfer showed no sensitivity to the inlet velocity regardlessof the proximity the to inlet jet. Figure 10 showsthe averagerate of convective heat transfer for the three heatedceiling panelsclosest to the ceiling diffuser. Thefirst case, with an approximate inlet jet velocity of 1.19 m/s (3.9 fds), is for the pan-typediffuser witha 2-in~ (51mm)gap between the cover plate and the diffuser pan. The second case, withan inlet jet velocityof 2.37 m/s(7.7 ft/s), is for ASHRAE Transactions: Research Severalconclusionscan be drawnfromthe evaluation of the flow and temperaturefields. First, due to the dominance the of Coanda effect, the flow field is momentum driven and is relatively unaffected by buoyancy.Second,the temperatureof the inlet jet is the best reference the definitionof the surfaceheat for transfer coefficients. Finally, notwithstanding importance the of momentum determining flow field, the surface heat transin the fer is independent the inlet .jet velocity. of Thefact that the film coefficients are independent the .jet of velocity precludes the possibility of correlating surface film coefficients to jet-velocity-based parameterssuch as the outlet Reynolds number: ReUoL v (7)

whereU is the .jet velocity at the diffuser, L is a diffusero based length scale, and v is the kinematicviscosity. Attempts . diffuser

ceilin~~,~

/fixture

Figure 8 Cylindrical-o,pe radial diffuser.

143

Diffuser

F~

Fixture -~ ._. 160,] 140 ,--. 120. 10o. 80= .~ 60. o 20O2enterPanel N. CenterPanel S. CenterPanel Jet Inlet Velocity (m/s) Figure 10 Convective flux for three heated panels near the radial diffuser. must be understood in terms of the roomcontrol volumerather than in terms of the surface boundary layer. The ventilative flow rate is proportional to the total energy delivered to the room a given inlet air temperature. at When normalized to the roomvolume, this paratneter scales the roomconvective heat transfer to its fundamentaldriving potential--the difference between the surface temperature and the air inlet temperature. Preliminary experimental evidenceindicates that the validity of the correlations depends onlyon a basic similarity of the flow field. Thusa single correlation for each surface orientation described the roomconvective heat transfer from 3 ACH 100 ACH a variety of to for diffusers, including the diffusers described in this paper and the commercial diffuser used by Spitlero As shown Figures 12 through14, the volumetricflow rate in dependence is of the form (h-ACH). The figures show room convectiveheat transfer coefficients as a function of volumetric flow rate for an inlet air referencetemperature. Experimental Uncertainty and Rangeof Correlations Theart of uncertaintyanalysisis well established the therin mal sciences; numeroustexts and papers address the topic, including excellent texts by Taylor (1982) and Holman (1989). Of patlicular importanceto this investigation is the fact that Spitler validated both the experimental procedure the perforand mance the facility at highventilative flowrates (Spitler 1990). of Theclassic paper on uncertainties in single-sampleexperiments was published by Kline and McClintock (1953), who showed for the special case of a linear function with indethat pendent variables, each of whichis notxnally distributed, the relation between uncertaintyinterval for the variable and the the uncertaintyinterval for the result is givenby ASHRAE Transactions: Research

Figure 9 Pan-type radial diffuser. at correlating the heat transfer coefficients to the Reynolds number and other velocity-based parameters predictably failed. Momentum dependencewas tested by calculating the jet momentum number,J, which had been previously used to correlate room transfer coefficients at highventilative flowrates. heat Theresults of this test showed that room surface heat transfer was also independent of jet momentum the ceiling inlet for configuration. The jet momentum numberis defined as J (8)

9g~room

where ,h = massflow rate of ventilation air, = jet velocityat the diffuser, p g = densityof air, = acceleration due to gravity, and volumeof room.

For high ventilative flow rates in the enclosure, Spitler correlated the forced convectionheat transfer coefficient to the jet momentum number(Spitler et al. 1991b). Variable diffuser area tests performed the sidewall inlet at high ventilation for rates showed that for the conditions under consideration, the rate of surface heat transfer was sensitive to changesin the inlet momentum. However, the radial ceiling diffuser, the for rate of surface heat transfer is completely insensitive to variations in the jet momentum number. Figure 11, which shows heat transfer coefficients based on an inlet air reference temperature, illustrates the independence surface heat transfer of and the momentum the inlet jet. of The development successful correlations was based on of the observation that althoughthe heat transfer coefficient was independent both the inlet jet velocity andthe inlet jet momenof tum, it wasdependent the jet massflow rate. For incompresson ible flow over the temperaturerange studied, volumetric flow rate also resulted in reasonablecorrelations of the surface heat transfer. Theimplicationof this observation that in developing is whole-room convective heat transfer coefficients, the physics 144

e. -- [(~Rn ,j~2 {OR "~2 Lt,~e

(OR )2"11/2

(9)

where R e R e i = calculated result,R= R(v v2 ..... Vn), l, uncertainty intervalin the result, and =
=

Thecorrelations shown Table3 are applicableto an in isothermal room with a radial ceiling jet between (50F) 10C and25 C(77 F) andan enclosure change (ACH) air rate within the range(3 < ACH 100). Theconvective transfer coef< heat ficient, h, is basedon a referencetemperature measured the in supply duct. air TABLE3 HeatTransferCoefficients for CeilingInlet Configuration

uncertainty intervalin the ith variable.

Thepartialderivative, n, is a measure the sensitivity ~Rl~v of Surface Type Correlation of the result to a singlevariable.Multiplying sensitivityor the 8 = 0.19~ACH "influence" coefficientby the estimated uncertainty the variin Walls h ableprovides estimate the variables an of contribution uncer- Floor to 8 h = 0.13ACH tainty in the result. Themethod influencecoefficientshas of 8 = 0.49ACH Ceiling h traditionallybeenextended nonlinear to relationships firstby orderexpansion the governing of equation evaluation the and of partial derivatives the basevalues(Holman at 1989). Although correlations subjectto restrictions,theyare the are suitable for many applications.Theexperimental HVAC rangeof This method usedto calculate the uncertaintyassociwas bothvolumetric rate andinlet air temperature flow coversthe ated with the experimentally determined transfer coeffiheat normal operating of most range buildings. spiteof the fact that a In cients. Figures15 through17 showthe uncertainty in heat number variables of (such room as aspect ratio, interaction inlet of transfer coefficients for a room configuredwith a ceiling multiple diffusers, furnishings, internalheatsources) and diffuser.Thefilmcoefficients basedona jet inlet reference jets from are temperature. were checked the parametric of experiments, fact that not by set the 12 ACH Ceiling
5--4

Convection

12 ACH Wall Convection

i i I I I ~

0 .4 10 4.0

3 8.010-4 1.210"3 1.6 lff

.4 4~010

.3 8.010 ,4 1,210"3 1.610 J

6 ACH Ceiling Convection

5 4~3~3 t I ~ I

6 ACH Wall Convection

5
I I I 1 I I I

1 I I 0 ~ 10 .4 0.0 2.010 I 4.010 -4 6.010-4 0 5.01~51.510-42.51~3.510-44.51~

J J Figure 11 Convective transfer coefficient as a function of jet momentum heat numbet:

ASHRAE Transactions: Research

145

2O

15

16

30

44

58

72

86

100

LEGEND OFisher [95] VSpider [90]

ACH

2.0

4.6

7.2 ACH

9.8

12.4

15.0

Figure12 Ceiling h vs. ACH a radial ceiling diffuser for with an inlet air reference temperature.

Figure 15 Floor heat transfer coefficients with uncertainty intervals.

16

30

44 ACH

58

72

86

100

0 2.0

~ 4.6

--t 7.2 ACH

t 9.8

~ 12.4

15.0

LEGEND OFisher [95] V Spitlez [90]

Figure 13 Wall h vs. ACH a radial ceiling diffuser with fbr

an inlet 10 , air reference ~ ~ temperature, , , ~ ....................

Figure 16 Wall heat transfer coefficients with uncertainty intervals.

4 6

16

30

44

58

72

86

100

2.0

~ 4.6

7.2 ACH

9.8

12.4

15~0

LEGEND O Fisher [95] V Spitler [90]

Figure 14 Floor h vs, ACH jbr a radial ceiling diffuser with an inlet air referencetemperature~ ceiling heat transfer dominates overall room convection expected is to mitigatethe sensitivity of the conelations these variables~ to Twentyexperiments were performed in the nonisothermal room determinethe applicability of the correlations (derived to fromisothermalroom data) to realistic room surface temperature profiles. Thisset of experiments showed the correlationscan be that applied to nonisothermal roomswith surface temperaturedifferencesof less than 20C F) withoutserious error. Although (68 the room flow fields are significantly different, the maximum devia146

Figure17 Ceiling heat transfer coefficients with uncet;tain~., intervals. tion in the heat transfer coefficient at 6 ACH a 20C with (68F) surface temperaturedifferential was15%,as shown Figure 18. in APPLICATIONS Implementing the Correlations The expressions for roomconvective heat transfer coefficients presented Table3 are suitable for implementation any in in heat balance basedbuilding energyor thermal load program with
ASHRAE Transactions: Research

sufficient detail to supportthe calculations.Asignificant requirementis that the radiation and convectioncoefficients not be combined computationalpurposes. The correlations cannot for be easily incorporated in roomweightingfactors or in surface responsefactors that include the air nodes. Likewise, simplified fenestration modelswith combinedradiation and convection coefficientswill not be able to utilize the results of this study. The new convective heat transfer coefficients were implemented in the BLAST (Building Loads and System Thermodynamics)program. BLAST implementation of timevarying film coefficients was essentially demonstrated in AppendixG of Spitler (1990). The correlations can also implemented computational fluid dynamiccodes, where in in most cases they will represent a significant improvement over typical wall functions. Significance of Results: A Case Study

ventilation rate of 6 ACH a 10C(50F) reference temperwith ature. Thedeadair spaceheat transfer coefficients wereset at the BLAST defaults (natural convection) for all 12 cases. Column three shows the percent difference between columnone and column two The example,thoughvalid for only one building in one location, showsup to an 11% error in the daily calculated load on the system. TABLE 4 BLAST Loads for Natural and Mixed Convection Heat Transfer Coefficients Case Sens. Load Sens. Load Percent (kBtu/h),Natural (kBm/h),Low Error Convection Momentum 3.897E+ 03 2.662E+ 04 5.149E+ 03 3~580E 04 + 5.092E+ 03 3o598E 04 + 10.7% 5.05% 8.5% 4A% 11 6.7%

ABLAST of a 1.4 million ft 2 office wasused for this model exercise. Thefacility, a seven-storybuilding with masonry walls (24% glass) and a built-up asphalt roof with fiberglass and felt insulation, was modeled seven occupiedzones with an interas nal occupancy of 1 person/100 2, a lighting load of 2.0 W/ load ft ft 2, and an equipment load of 0.4 W/ft2. The "deadair spaces" between suspended the ceilings (with recessed lighting) and the next floor (or roof) were modeled separate, "uncontrolled" as zones. "Highinternal loads" were simulated by increasing the occupancy load to 1 person/50 ft 2 and increasing the equipment 2. load to 2.0 W/ft Highsolar loads weresimulatedby increasing the glazing area to 54% the total wall surface area. For the of purposes of this exercise, both buildings were located in the central Midwest (St. Louis, Missouri). Table4 showsthe error in the daily cooling load due to an error in the ceiling convectiveheat transfer coefficient. Column one shows daily coolingload for each case with the currently the used BLAST convective heat transfer coefficients. Column two showsthe daily cooling load for each case calculated with the heat transfer coefficients that are associated with a moderate 1

Single story, low 3.520E+ 03 internal loads Multistory,low 2.534E+ 04 internal loads Singlestory, high internal 4.743E+ 03 loads Multistory,high + internal loads 3o438E 04 Singlestory, highsolar loads 4.57E+ 03 Multistory, high 3.372E+ 04 solar loads

SUMMARY AND CONCLUSIONS Heatbalancebasedcalculation of building energyor thermal loads requireseither explicit or implicit estimation the surface of film coefficients. Simulationstudies, whichwere performedto estimate the impactof film coefficients on overall room heating and cooling loads, indicated that commonly film coeffiused cients canresult in a daily coolingload error on the order of 10%. Equations relating room convective transfer coefficients to heat the air change rate of ceiling-ventilatedrooms weredeveloped from experimental data. Thecorrelations that wereoriginally developed for ventilative flow rotes rangingfrom3 to 12 ACH extended were to coverthe Spitler data set (15 ACH 100ACH). convective to The heat transfer coefficientswere based an inlet jet reference on temperature, which onlyresulted in the lowestexperimental not uncemainty but also significantlyimproved correlations. the Severalobservations to the selection of ACH the correled as lating parameter the convective transfer coefficient. First, for heat it wasobserved that the ceiling jet did not detachfromthe room surfacesoverthe entire rangeof experimental parameters. Thusthe flow field in the roomwas roughly similar from 3 to 100 ACH. Second, wasdemonstrated the rate of convective transit that heat fer wasindependent both the velocity and the momentum the of of inlet jet. The choice of ACH the correlating parameter is as supported the physicsof the control volume, by which requires the overall roomconvectiveheat transfer to be proportional to the ventilativeflowrate at a constant inlet air temperature. 147

%0.8
"=~13.6

0.4.
"~ 0.:2. .~. ~ 0. 15% Error Line

-2" .... ......

Surface Panel Number Figure 18 Deviation o.[ non-isothermal tvom from isothermal room heat transfer coefficients. ASHRAE Transactions: Research

Theexperimental investigationalso left a number questions of unanswered. First, sensitivity of the con:elations the aspectratio to of the room not investigated. Arelated problem was involves the interaction of mtdtiple diffusers in a large room.The effect of furnishings and heat sources in roomsalso requires additional experimental investigation. REFERENCES Allard, F., C. Inard, and J.P. Simoneau. 1990. Experimental study of numerical simulationof natural convectionin a room with heated ceiling or floor. Roomvent Proceedings 90. ASHRAE. 1985. A NSff ASHRAE Standard 51-1985, Laboratory methods testing fans for rating. Atlanta: American of Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc. Bauman, A. Gadgil, R. Kammemd, R. Greif. 1980. BuoyF., and ancy-drivenconvectionin rectangular enclosures: Experimental results and numericalcalculations. ASME 80-Ht-66. New York: American Society of MechanicalEngineers. Bohn, M.S., and R. Anderson.1984. Temperature heat flux and distribution in a natural convectionenclosure flow. SERI Technical Paper 252-2482. Golden, Colo.: Solar Energy Research Institute. Bohn, M.S., and R. Anderson.1986. Temperature heat flux and distribution in a natural convection enclosureflow. Journal of HeatTransfer 108(May):471-475. Cantillo, J. 1990. Air velocity and temperature measurements a in full-scale ventilative coolingresearchfacility. M.S.Thesis, Department Mechanical Industrial Engineering, Uniof and versity of Illinois at Urbana-Champaign. Chen,K.S., A.C.Ku,and C.H.Chou.1990. Investigation of natural convectionin patlially divided rectangular enclosures both withand withoutan opening the partition plate: Meain surement results. Journalof HeatTransfer112(8): 648-652. Chen, Q., C. Meyers,and J. van der Kooi. 1989. Convective heat transfer in roomswith mixedconvection. In Proceedings of Air FlowPatternsbz Ventilated Spaces,February,Liege, pp. 69-82. Fisher, D.E. 1989. Design of an experimentalfacility for the investigation of convective heat transfer in enclosures.M.S. Thesis, Department Mechanical Industrial Engineerof and ing, Universityof Illinois at Urbana-Champaign. Fisher, D.E. 1995. Anexperimentalinvestigation of mixed6onvectiveheat transfer in a rectangularenclosure.Ph.D.Thesis, Department Mechanical Industrial Engineering, Uniof and versity of Illinois at Urbana-Champaign. Holman,J.Po 1989. Experimentalmethods for engineers. New York: McGraw-Hill. Hottel, H.C., and A.F. Sarofim. 1967. Radiative transfer. New York: McGraw-Hill. Kapoor,K., and Y. Jaluria. 1991. Mixed convectiveheat transfer characteristics of a downward turning buoyantceiling jet. ASME-ITID-Vol.163, pp. 9-17. NewYork: American Society of Mechanical Engineers. 148

and F.A. McClintock. 1953. Describinguncertainties in single-sarnple experiments. MechanicalEngineering 57(1):3-8. Mansfield,B.D.1993. Development a cold wall for investigaof tion of convective radiative heat transfer in enclosures. and M.S.Thesis, Department Mechanical Industrial Engiof and neering, Universityof Illinois at Urbana-Champaign. Murakami, S. Kato, and Y. Suyama. S., 1987. Three-dimensional numericalsimulation of turbulent airflow in a ventilated roomby meansof a two-equation model. ASHRAE Transactions 93(2): 621-642. Nansteel, M.W.,and R. Greif. 1981. Natural convectionin undivided and partially divided rectangular enclosures. Jourt~l of Heat Transfer 103: 623-629. Nansteel, M.W.,and R. Greif. 1983. Natural convection heat transfer in complex enclosures at large Prandtl numbers. Journal of Heat Transfer 105: 912-915. Neiswanger, G.A.Johnson, and V.P. Carey. 1987. AnexperiL., mental study of high Rayleighnumber mixedconvection in a rectangularenclosure with restricted inlet and outlet openings. Journalof HeatTransfer109: 446-453. Nielsen,P.V., A. Resfivo, and J. H. Whitelaw. 1978. Thevelocity characteristics of ventilated rooms.Journalof Fluids Engineering 100: 291-298. Nielsen, P.V, A. Restivo, and J.H. Whitelaw.1979. Buoyancyaffected flows in ventilated rooms.Numerical HeatTransfer 2: 115-127. Pavlovic, M.D., and F. Penot. 1991. Experiments the rnixed in convection reNmein an isothermal open cubic cavity. ExperimentalThermaland Fluid Science 4: 648-655. Spitler, J.D. 1990.Anexperimental investigation of air flow and convective heat transfer in enclosureshavinglarge ventilarive flow rates. Ph.D.Thesis, Department Mechanical of and Industrial Engineering,Universityof Illinois at U~bana.Champaign. Spitler, J., C. Pedersen, D. Fisher, P. Menne, J. Cantillo. and 1991a.Designof an Experimental facility for investigation of interior convective transfer underventilative cooling heat conditions. ASHRAE Transactions 97(1). Spitler, J., C. Pedersen, D. Fisher. 1991b. and Interior convective heat transfer in buildingswith large ventilative flowrates. ASHRAE Transactions 97(1). Taylor, J.R. 1982. An introduction to error analysis--The sttMy of uncertainties in physical measurements. Mill Valley, Calif.: UniversityScienceBooks. van der Kooi, J., and K. Bedeke.1983. Improvement cooling of load progams by measurementsin a climate morn with mass. XVlth h~ternational Congressof Refrigeration Proceedings, Tome pp. 71-77. V,
Kline, S.J.,

van der Kooi, J., and E. Forch. 1985. Calculationof the cooling load by meansof a "more-air-points-model." Proceedings of the CLIMA 2000 WorldCongress on Heating, Ventilating andAir-Condiaonh,g 395-40l. 4: ASHRAE Transactions: Research

This paper has been downloaded from the Building and Environmental Thermal Systems Research Group at Oklahoma State University (www.hvac.okstate.edu) The correct citation for the paper is: Fisher, D.E. and C.O. Pedersen. 1997. Convective Heat Transfer in Building Energy and Thermal Load Calculations, ASHRAE Transactions, vol. 103, Pt. 2, pp.137-148 (Winner of the ASHRAE Best Technical Merit Award). Reprinted by permission from ASHRAE Transactions (Vol. #103 Part 2, pp. 137-148). 1997 American Society of Heating, Refrigerating and Air-Conditioning Engineers, Inc.