Interior Convective Heat Transfer in Buildings with Large Ventilative Flow Rates, ASHRAE Transactions

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:

Spitler, J., C. Pedersen, D. Fisher. 1991. Interior Convective Heat Transfer in Buildings with Large Ventilative Flow Rates, ASHRAE Transactions. 97(1): 505-515.

NY-91-5-2 (RP-529) INTERIOR CONVECTIVE HEAT TRANSFER IN BUILDINGS WiTH LARGE VENTiLATIVE FLOW RATES

J.D. Spi.tler, Ph.D., P.E. C.O. Pedersen, Ph.D. D.E. Fisher Associate MemberASHRAE Fellow ASHRAE Associate MemberASHRAE

ABSTRACT transfer is extremely important; film coefficients should be considerably higher than those normally used. This paper presents the results obtained from a new This paper describes the results of an experimental experimentalfacility designedto investigate convective heat investigation of convective heat transfer in enclosures under transfer. It is described in a companionpaper. A large conditions that represent a wide range of ventilative condi- number of experiments were performed with varying inlet tions for buildings. The lowest volumetric flow rate locations and sizes, inlet temperatures, and flow rates. The examined, 15 ach, is at the high end of what might be rootn outlet temperature was identified as the most suitable considered typical for an occupied office building. Flow reference temperature for the calculation of film coeffi- rates above 15 ach might be encountered when certain cients. Film coefficients were successfully correlated to the ventilative cooling strategies are implemented.Under these jet momentumnumber, J. The correlations form the basis of conditions, the jet and surrounding flow are generally a new convective heat transfer model. This model has been turbulent. The experimentalenclosure used in this investiga- tested in the Building Loads Analysis and System Thermody- tion has approximateinterior dimensionsof 15 ft by 9 ft by namics (BLAST) program. The new model enabled BLAST 9 ft (4.6 m by 2.7 m by 2.7 m) and is considered full scale. to accurately predict the experimental results. The walls, floor, and ceiling are covered by individually ht addition, an office building with a night purging controlled heated panels that were maintained at 86°F strategy was simulated, using both the current model, based (30°C) for all experiments. Heat fluxes from each panel can on natural convection, amt the new model. The two models be determined using an energy balance. The room is well predicted significantly different ventilative cooling rates. instrumented, with more than 100 surface thermocouples, additional thermocouplesfor resolving all energy flows, and INTRODUCTION airspeed sensors and thermocouples that measure air speed and air temperature at 896 points. The ventilation system is An important componentof building energy analysis is capable of providing more than 100 ach, but it can be the prediction of interior convective heat transfer. The throttled downto provide as few as 2 ach. The experimental model currently used is knownas the well-stirred model facility is described in detail in a companionpaper (Spitler and assumes that the zone air has uniform temperature and et al. 1991). that a convective heat transfer coefficient, or film coeffi- The primary objective of this work is to create an cient, can be selected to predict the convective heat trans- experimental background for the development of improved fer. The coefficients given by the 1.989 ASHRAEHand- interior convective heat transfer models for use in building bookwFundamentals and commonly used by building simulation programs. These models will primarily be in the energy analysis programs appear to be based on research form of correlations for film coefficients. With this type of performed in the 1930s (Wilkes and Peterson 1938). model, the important questions to be answered are: Furthermore, this research was based on experimental measurementof heat transfer from fiat plates in a free 1. Whatis an appropriate reference temperature? environment under conditions of natural convection. The 2. Whatparameters influence the film coefficient for applicability of the well-stirred model whencoupled with the conditions of interest? .... these film coefficients has been questioned by numerous researchers (Gadgil 1980; van der Kooi and Forth 1985; Hourly building energy analysis programs simulate Lebrun and Ngendakumana1987; Chen and van der Kooi 8,760 hours for an annual simulation. This leads to some a988). constraints on the complexity of a model that can be The applicability of the natural-convection-based film effectively utilized. Furthermore, since there are many coefficients is particularly questionable under ventilative unpredictable simulation parameters (such as occupants) that cooling conditions, such as night purging. Night purging influence the roomair flow and hence, the convective heat involves ventilating the building at night to precool the transfer, there is a point beyondwhich increased accuracy building structure. Night purging may also involve very maybe a waste of effort. high flow rates, perhaps as high as 100 air changes per hour (ach). The coupling betweenthe building mass and the EXPERIMENTAL RESULTS ventilation air described by the convective heat transfer model is the dominant effect in the simulation of night A total of 44 experimental tests were performed. This purging. In this case, accurate modelingof convective heat included one set of 30 parametric tests, in whichvolumetric

Jeffrey D. Spider is an Assistant Professor, Schoolof Mechanicaland AerospaceEngineering, Oklahoma State University, Stillwater; Curtis O. Pedersenis a Professor and Daniel E. Fisher is a ResearchEngineer, Departmentof Mechanicaland Industrial Engineering, Universityof Illinois, Urbana.

ASHRAETransactions: Symposia 505 5000

4000¯ ~ 21"C

3000"

2000 ¯

1000’ East Wall Ceiling 0 10 30 50 7O 9O 10 30 ~ 70 90 II0 ACH Figure 2a Ventilative cooling rate for sidewall inlet, Figure 1 Ventilative cooling rate for sidewall and three inlet temperatures ceiling inlets--21 °C 8000- flow rate and inlet temperature were varied for each inlet location; 7 tests with reduced inlet area to examine the ----’O---, 16"12 / effects of jet momentum;3 tests with the sidewall inlet diverted awayfrom the north wall; and 4 tests with furni- 6000 ture in the room. The variables and levels for the parametric tests are shownin Table 1. The interior surfaces of the roomwere maintainted at 86°F (30°C) for all tests. Due to the voluminous amount of data obtained, the 4000’ scope of results that cart be presented here is limited. This paper will focus on the subset of results directly related to choosing the best reference temperature and the development of con’elations for convective heat transfer. Other 2000’ experimental results that cannot be adequately covered here include airflow contour plots, temperature profiles, flow visualization, distribution of convective heat flux aroundthe room, and correlations for individual panels and walls. 0 Thesetopics are included in the final project report (Peder- 10 30 50 70 90 110 sen et al. 1990) and in individual theses (Fisher 1989; ACll Menne1989; Spitler 1990). Figure 2b Ventilative cooling rate for ceiling inlet, three Overviewof Results inlet temperatures

Figures 1 through 5 represent the total heat transfer area; second, the ceiling inlet jet impinges on a larger from the roomfor various conditions and flow rates. Figure surface area than the sidewall inlet jet. 1 shows the ventilative cooling rate for each of the two Figures 2a and 2b show the variation in ventilative inlets with varying flow rates but a constant inlet tempera- cooling rate with inlet temperature for each inlet configura- ture of 70°F (21°C). tion. As expected, lower inlet temperatures result in higher For both constant area inlet configurations, with a ventilative cooling rates. Again,all ventilative cooling rates constant inlet temperature, the ventilative cooling rate is are approximately linear with respect to volumetric flow approximately linear with respect to volumetric flow rate. rate. The ventilative cooling rate curve for the ceiling inlet Figure 3 shows the effect of changing the jet momen- configuration is both greater than and has a greater slope tum on the ventilative cooling rates. The jet momentumwas than the cooling rate curve for the sidewall inlet configura- changed by reducing the inlet area. The percentages given tion. This is due to two factors: first, the ceiling inlet has in the legend refer to the percentage of area open. For a higher htlet flow velocities due to its smaller effective inlet constant volumetric flow rate, the 33 %opening has inlet velocities three times as high as the 100 % opening. Like- wise, the 67 %opening has inlet velocities 1,5 times as high TABLE 1 as the 100%opening. Variables and Levels for Parametric Tests As expected, for constant volumetric flow rate, the ventilative cooling rate increases as the inlet area decreases. Variables Levels (For constant volumetric flow rate, inlet jet velocity, 0VolttmcwicFlow Rate (ACI , 15.30.50,70.100 momentum,and energy are all inversely proportional to inlet area.) Also note the "convergence" of the three Inlet Loc~on Ceiling, Sidewall curves at 15 aeh. This implies that at low flow rates, the

506 ASHRAETransactions: Symposia 400O

30OO

10 30 50 70 90 110 10 30 50 70 90 110 ACH ACH Figure Ventilative cooling rate for different size Figure 4 Ventilative cooling rate for diverted and undi- sidewall inlets--21 °C verted sidewall inlets--21 °C effects of jet velocity, momentum,and energy become small comparedwith the effects of buoyancy. Visualization 40o0 of the flow showedthe inlet jet "pouring" into the room and dissipating along the floor at low flow rates (Menne 1989). 30o0 Figure 4 shows the effects of diverting the sidewall inlet jet awayfrom the north wall and toward the center of the room. As shown,diversion of the jet has little effect on the total ventilative cooling rate. 2000 Figure 5 shows the effects of different furniture configurations on the ventilative cooling rate. Thetests with furniture were intended to be a first look at the effects of furniture on ventilative cooling. The surface temperatures ~ OFHCE of the furniture were not measuredand hence the radiative and convective components of the heat flux could not be separated. The conference room furniture configuration 10 30 50 70 90 110 consists of a table and six chairs. This configuration offers little flow resistance. The office furniture configuration ACH consists of two cabinets (5 ft [1.5 m] wide, 3 ft [0.9 m] Figure 5 Ventilative cooling rate for different furniture deep, and 6 ft [1.8 m] high) placed against the center of the configurations (21 °C sidewall inlet) north and south walls. This configuration significantly changes the flow in the room. The effect of the table and chairs in the conference Four different reference temperatures producing four roomfurniture configuration is relatively insignificant. The different heat transfer coefficients were investigated: effect of the, two cabinets in the office furniture configuration is muchlarger. This is probably due to increased local 1. room outlet or return air temperature (HOUT), flow velocities in the room. Note that local flow velocities 2. bulk air temperature (HBULK), are increased because the volumetric flow rate is held 3. air temperature adjacent to a surface (HAIRT),and constant, despite any increased pressure drop across the 4. local air temperature as a function of height roomthat might be caused by the addition of furniture. In (HADJZT). a room ventilated by open windows or a mechanical ventilation system without flow control, the flow rate might The return air temperature is measured by six thermocou- be decreased due to the addition of furniture. pies in the outlet duct. The bulk air temperature, air temperature adjacent to each panel, and local air tempera- Choosinga Reference Temperature ture as a function of panel height are all determined by the local airspeed measurementsystem described in a compan- The first step in correlating convective heat flux is to ion paper (Spitler et al. 1991). The bulk air temperature choose one or more reference temperatures so that film simply the average of the 896 air temperatures measured. coefficients maybe determined as follows: The points are distributed evenly throughout the roomand

II should be a good representation of the true bulk air temper- qc (1) ature. Air t.emperature adjacent to each panel is determined h= by find.ing the measurementpoints closest to each panel and averaging those measurements. Air temperature as a where function of height is determined by averaging the air temperature of each horizontal plane in the measurement convective heat flux (Btu/h-ft 2 or~) W/m grid. The average temperature of the horizontal planes surface temperature (°F or °C) adjacent to each panel is then used as the reference temper- reference temperature (°F or °C). ature.

ASHRAETransactions: Symposia 507 12 programs based on the well-stirred model require no structural changes in order to implementcorrelations using return temperature as the reference temperature. 10 Calculation of Characteristic Parameters

There are several parameters of interest that describe | conditions in the room. These are volumetric flow rate, inlet velocity, Archimedesnumber, jet momentum,and jet energy. Measurementof the volumetric flow rate is discussed in a companionpaper (Spitler et al. 1991). /nlet velocity, Archimedes number, jet momentum, and jet energy are discussed in the following sections. Inlet Velocity For the sidewall inlet, the inlet velocity is the averagevelocity over the face of the inlet, that is,

I~" (2) Inlet Area 0.0 0.2 0.4 0.6 0.8 1.0 For the ceiling inlet, an effective inlet area is estimated Volumetric Flow Rate (m**3/s) from experimental measurements. (With the complex geometry of a radial diffuser, it is somewhatdifficult to Figure 6 Comparisonof vertical wall film coefficients determine the actual inlet area, which is dependent on for different reference temperatures(five tests whereit is measured.The effective inlet area is determined with 21 °C sidewall inleO by measuring the velocity profile near the diffuser. A characteristic velocity is determinedand the effective inlet Figure 6 shows the average film coefficients for the area can be determined by dividing the volumetric flow rate vertical walls with the sidewall inlet configuration. The film by the characteristic velocity.) This effective inlet area is coefficients based on the air temperatures in the room used to estimate the inlet velocity: (HBULK,HAIRT, and HADJZT)all have approximately the same value. This to be expected whenthe heat transfer surface is the vertical walls of the room. In essence, the = (3) reference temperature is just the bulk air temperature averaged in slightly different ways. These coefficients follow the film coefficients based on roomoutlet temperature, with the exception of the 15-ach case. For reasons of consistency with previously published Further insight can be gained by looking at Figure 7, literature, the inlet velocity is referred to as Uo. which shows the outlet and bulk air temperatures. The bulk Archimedes Number The Archimedes number is air temperature diverges from the roomoutlet temperature usually considered to represent the ratio of the buoyant as the flow rate decreases. As the flow rate decreases, the forces to the dynamicforce for an inlet jet. (It also maybe stratification in the roomincreases. The stratified layer in considered to represent the ratio of buoyant force.,~ to the roomhas little effect on the outlet temperature but does momentumflux, momentumdissipation, or inertial forces.) increase the bulk air temperature. The high bulk air Note that the Archimedes number does not take viscous temperature results in a small temperature difference effects into account. The Archimedesnumber can be cast in between the surfaces and the reference. The large film coefficients shownin Figure 6 are a result of this small temperature difference. The linear relationship between outlet-temperature- based film coefficients and the volumetric flow rate indi- cates that the outlet temperature is the most desirable reference temperature. In addition, careful analysis of the experimental uncertainty associated with each reference temperature reveals that using the outlet temperature will [] result in the lowest error in calculated film coefficients, particularly at low flow rates. Experimental tmcertamty, whichis a function of the experimentalfacility, is discussed in a companionpaper (Spitler et al. 1991). Since the outlet reference temperature has no significant disadvantages comparedwith the other reference temperatures, all future analysis in this paper will use the outlet temperatureas the reference temperature. 21 BLASTand other building energy analysis programs 0.0 0.2 0.4 0.6 0.8 use the zone air temperature, not the outlet temperature, as 1.0 the reference temperature. However, because BLASTis Volumetric Flow Rate (m**3/s) based on the well-stirred zone model,it assumesthat all air in the enclosure, including the return air, is at the zone air Figure 7 Comparisonof reference temperatures (five temperature. The result of this fortuitous assumptionis that tests with 21 °C sidewall inlet)

508 ASHRAETransactions: Symposia several different forms. A general form of the Archimedes 1)" U, number is J = (6)

Ar = f~g L~To (4) This is equivalent to dividing the momentumflux by where pgV, o~,. The authors do not give any justification for this methodof nondimensionalization other than it is possible. /3 = coefficient of thermal expansion The effect of room volume on J could not be examined in g = gravitational acceleration the experimentalfacility used for this investigation. There- cL = a characteristic length fore, the limits on roomsize for whichheat transfer coeffi- oAT = temperature difference betweenthe roominlet cients correlated to the jet momentumnumber will be and outlet accurate is not known.In the second paper, somecorrela- U = a fluid velocity. tions are madefor air velocity at the floor of a livestock building in the form of Different researchers have used different values of the characteristic length and fluid velocity. For this study, the (7) characteristic length was chosen to be the maximum Floor velocity = C1 + 6"2 ¯ jcz possible throw of the jet and the fluid velocity was chosen to be the inlet velocity. The reasoning behind these choices where C1, C2, and C3 are empirically defined constants. is discussed in depth by Spitler (1990). Regardless of which form of the Archimedes number Jet Energy The jet energy flux into the roomis calcu- is used, it seemsto have one main application to predict the lated as trajectory of the inlet jet. In the scenarios presented by Nielsen et al. (1979) and Randall and Battams (1979), =,a. (8) inlet jet that is introducedat the top of a roomin a horizontal direction can either travel along the ceiling and then Correlation of Results drop in the roomor just drop into the room, depending on the Archimedesnumber. (There is also the possibility of unstable flows for intermediate values of the Archimedes The large data set generated by the experimental effort number.) allows for the formulation and study of numerouscorrela- Althoughit seemsunlikely that convective heat transfer tions. However,only those correlations that maybe useful would correlate directly to the Archimedesnumber, it may in current building simulation programs are of primary be .possible to use the Archimedesnumber to predict flow interest. These include correlations of the convective heat regimes and/or vertical temperature gradients. Also, transfer coefficient (h) whereh is a function of one or more comparisons of different room flows based on the Archime- parameters (Tq). Only those parameters will be considered des numberwill most likely be valid only whenroom inlet that can be determinedor derived from user input to current building simulation programs. For example, since inlet jet and outlet geometries are the same. In other words, we do conditions, such as jet velocity, jet momentum,and jet not expect the airflow patterns to be the same when the energy, can be determined by most current building simula- Archimedesnumber is the same if, in one case, the inlet is tion programs, correlations based on these parameters will a radial diffuser in the ceiling and, in another case, it is a sidewall inlet. be investigated. Onthe other hand, local air velocity, which For the sidewall inlet in the experimental room, Lc is presently can only be calculated by computational fluid approximately 15 ft (4.6 m). Somevalues of the Archime- dynamics (CFD) programs, will not be considered as des numberencountered in the experiment are." possible parameter. Correlations of the convective heat transfer coefficient as a function of one or moreparameters 7.9 15 ach, 61°F (16°C) inlet, 100%inlet area (h = f(Tq)) for the following enclosure surfaces and air inlet 0.13 50 ach, 70°F (21°C) inlet, 100%inlet area configurations will be investigated: 0.03 100 ach, 79°F (26°C) inlet, 100%inlet area 1. ceiling with ceiling inlet, 4.1 15 ach, 70°F (21°C) inlet, 100%inlet area 2. vertical wails with ceiling inlet, 1.6 15 ach, 70°F (21°C) inlet, 67%inlet area 3. floor with ceiling inlet, 0.6 15 ach, 70°F (21°C) inlet, 33%inlet area 4. ceiling with sidewall inlet, 5. vertical walls with sidewall inlet, and Jet MomentumThe jet momentumflux into the room 6. floor with sidewall inlet. is calculated as Procedures for Correlating Experimental Resalts A = ¯ Up (5) numberof relationships between variables were examined, both graphically and with linear and nonlinear regression techniques, prior to the formulation of the successful where correlations presented in this paper. The procedures discussed in the followingparagraphs present the methodol- rh = mass flow rate (lbm/s or kg/s) ogy that was ultimately used to correlate the data. Up = velocity of the supply air jet (ft/s or m/s). First, film coefficients for ceilings, walls, and floors for both inlet configurati¢ms were plotted against various The jet momentumhas been nondimensionalized for use parameters, including volumetric flow rate, jet momentum, in analyzing ventilationpatterns (Barber et al. 1982; Ogilvie jet energy, bulk air velocity, etc. The most obvious rela- and Barber 1988). The first paper defined the jet momen- tionship that cameout of these plots was that, for all but tum number: two cases, the film coefficients were linearly related to bulk

ASHRAETransactions: Symposia 509 TABLE 2a TABLE 2b Correlationsfor h (Btu/h’ft2" o F) Correlationsfor h (W/m2. o C)

Surface Inlet Correlation Limits Surface Inlet Correlation Linfits Ceiling Ceiling h=2.0+ 36.9 jo.5 0.001 < I < 0.03 Ceiling Ceiling h=l 1.4 + 209.7 j0.5 0.001 < l < 0.03

Verfi~ Ceiling h=0.7 + 14~3 0.001< l < 0,03 Vertical Ceiling ...... h=4.2+ 81.3 jO.5 .... 0.001 < J < 0103 Wails Waits Floor Ceiling h=0.6+ 8.2410.5 0.001 < J < 0.03 Floor Ceiling h=3.5+ 46.8 jo.5 0.001 < J < 0.03

Ceding Side Wall h=Ool+ 10.5 0.002< J < 0.011 and Ceiling Side Wall h=0.6+ 59.4 jo.5 0.002< J < 0.011 and Ar <0.3 Ar <0.3 Vertical Side Wall h=0.3+ 16.3jO.5 0.1302< J < 0.011 Vertical I Side Wall h=l.6 + 92.7 j0o5 0.002 < I < 0.011 Walls Walls l~.oor Side Wall h=0.6 + 7.74 j05 0.002< J < 0.011 and Floor Side Wall h=3.z+,~.o jo.5 0.002 < J < 0.011 and Ar <0.3 air velocity. These two cases--the floor and ceiling with where sidewall inlet--showed increased scatter in the film coefficients as the bulk air velocity becamesmall. This was due _.C = an ernpirically determined constant to the effects of stratification at larger Archimedesnum- V = bulk air velocity (ft/s or m/s). bers. For the floor at low flow rates, the film coefficient increases independently of the bulk air velocity as the Using the dimensionless form of jet momentum(the jet Archimedesnumber increases. For the ceiling, the opposite momentumnumber J-) as the independent variable, linear is true. As the Archimedes number increases, the film correlations were developed using the square root of J in coefficient decreases independentlyof the bulk air velocity the following form: as natural convectionand the resulting stratification retard the heat transfer. h = C1 + 6"2 "J°’S (10) The above results indicate that in order to successfully correlate the data over the entire range of conditions, the where C1 and C2 are empirically determined constants. correlation must include an Archimedes number or other These correlations are presented below. such parameter in addition to a bulk air velocity parameter, Correlations This section contains correlations for the For Ar < 0.3, however, the data correlated well to a single ceiling inlet and sidewall inlet configurations that are parameter, namely, the bulk air velocity. Therefore, it was intended for use in building simulation programs. Table 2 decided to use only the data from tests below Ar = 0.3 and gives the correlation for each surface and inlet configura- develop single-parameter correlations. The correlations tion, along with the limits of the correlation as boundedby presented in this paper are, therefore, valid only for Ar < the experiments performed. Table 3 compares film coeffi- 0.3. For the parametric experiments performed, the tests cients calculated with the correlations with the standard that had Ar < 0.3 were: values from the 1989 ASHRAEHatMbook--Fundamentals. (The "Thermal and Water Vapor Transmission Data" 1. 15 ach, inlet temperatures--16°C, 21°C, 26°C chapter of Fundamentalsgives a table of surface conduc- 2. 30 ach, inlet temperatures--16°C, 21°C tances. These surface conductances include both radiation 2. 50 ach, inlet temperature--16°C. and convection. The radiation componentmust be subtract- ed to yield a convection coetficient.) Table 4 gives several Since current building simulation programs have no statistical parameters for each correlation, including the wayto determine bulk air velocity in a zone, the relation- numberof data points, the multiple correlation coefficient, ships between bulk air velocity and flow rate, jet momen- arid the standard error of the estimate. The standard error tum, and jet energy were investigated. Fromthis investiga- of the estimate is the uncertainty at about a 70%confidence tion it was determined that bulk air velocity correlates level. Plots of the correlations vs. the data are shownin extremely well to jet momentumin the form: Figures 8 through 10. All three correlations for the ceiling inlet have excellent ~ = C ¯ (Jot Momentum)°’s (9) multiple-regression coefficients. Whenused within the

TABLE 3a TABLE 3b SampleValues of h (Btu/h.ft2’°F) Sample Values of h (W/m2.°C) from Correlations and ASHRAE from Correlations and ASHRAE

_ l__._nle~t ~ : ASHRAE : ., 15 ACH Surface Inlet ASHRAE 15 ACH 100 A~’-- Ceilin~ 0.17 3.0 8.3 Ceiling Cetting 0.95 16.8 47.4 VertlcaIWallsCeiling 0.54 1.1 3.2 VerticalWalls Ceiling 3.08 6.3 18.1 Ceiling ..... 0:,~ I ...... 0.8 ,2.0 Ceilinff 4.04 ...... 4.7 11.5 SideWall 0.17 0.2 1.0 Ceiling Side Wall 0.95 41,: 5.8., ...... VerticalWalls SideWall 0.54 0.5 1.7 VerdcalWalls SideWall 3.08 2.8 9.7 Floor SideWall 0.71 0.7 1.2 Floor Side Wall 4.04 3.8 7.0

510 ASHRAETransactions: Symposia 50 15

~ 30

10 20 30 5 10 15 Jx 1000 J× 1000 Figure 8a h (W/mz. °C) vs. J for the ceiling with ceiling Figure 8b h (W/me. °C) vs. J for ceiling with sidewall inlet inlet, Ar < 0.3

20 15

0 10 20 30 5 10 15 J x 1000 Jx 1000 Figure 9a h (Wlrn2" °C) vs. J for averageof vertical walls Figure 9b h (W/me" °C) vs. J for averageof vertical walls with ceiling inlet with sidewall inlet

15 15

I I 0 0 0 10 20 30 0 5 10 15 Jx 1000 Jx 1000 e. Figure lOa h (W/m °C) vs. J for floor with ceiling inlet Figure lob h (W/me. °C) vs. Jforfloor with sidewall inlet, Ar < 0.3 range given, they should give good results. Obviously, as J approacheszero, at somepoint the correlations will cease to give accurate results. In particular, the constant determined for the ceiling is too high for J = 0. This would TABLE 4 mean natural convection, for which we would expect a Statistical Parametersof Correlations value approximatelyequal to 0.2 Btu/h-ft2. °F (1 W/m~. °C). One can imagine that at some point the negative buoyancy Surface Inlet N Multiple R StandardError StandardError would becomemore dominant than the ceiling effect and (s~ n~,, , ~/,.2~, that the inlet jet would separate from the ceiling. At thst 15 0~99 0.28 1.6 point, we might expect an abrupt change in the film VerticalWalls Ceiling 15 0.99 0AI 0.6 coefficient. Identifying this point is an obvious topic for Floor C.ei/ing 15 0.99 0.05 0:3 future research. Ceiling SideWall 12 0.89 0~6,, H,,,O. , The correlations developed for the sidewall inlet have Vea~ealWalls SideWall 22 0~96 0.14 0.8 reasonably high multiple-regression coefficients, though not Floor,,, SideWall 12 0.93 0.07

ASHRAETransactions: Symposia 511 TABLE 5a TABLE 5b Correlationsfor h (Btu/h.ft2.°F) with jo~4 Correlations for h (W/m2.°C) with jo.,

~n]et Correlation Surface Inlet Correladon Ceiling Ceil~g h=l+3 + 28+3jO.4 O.~l

Ceging h--0.5 + 11.0 j0A O.~l

Ce~ng h=0.5 + 6°3 30.4 0,~l

Ceil~g Side Wall h=-0.07 + 7°6 j0A O,~2

Side Wall h=O~4+ 5.7 jO.4 0.~2

If this were the case, then h wouldbe correlated to J in the A simulation of the experimental room was performed following form: with BLASTand used for model verification. The unique feature of the simulation is that all walls are artificially h = a + b -J°’4 (14) maintained at 30°C. The inlet air was treated as ventilation air and was held at 21°C. A comparison of the ventilative cooling rate for the Certainly, using this form of the equation will result in ceiling inlet configuration is shown in Figure 11a. The slightly different correlations. Toinvestigate this possibility, legend "CL 21 NF" corresponds to experimental results the correlations presented in Table 2 were reevaluated in for the 21°C ceiling inlet. The legend "BLAST"refers to the tbrm given by Equation 14. This resulted in the correla- simulation results using the current convective heat transfer tions shownin Table 5. coefficients. The legend "Ceiling Model"refers to simula- Additional information regarding these correlations is tion results using the new version of BLASTwith the new given by Spitler (1990). "[~ae correlations for the ceiling coefficients for the ceiling inlet configuration. inlet configuration have slightly higher correlation coeffi- The new model generates accurate simulation results, cients than the correlations using 3°’ . In addition, the while the old model is clearly inadequate. Even at 15 ach,

512 ASHRAETransactions: Symposia 6000

300o

30OO 20oo

2000 lOOO 1000 ----o,-- BLAST ¯ "--’It-- SideWall Model

~ 20 40 60 ~0 120 0 20 40 60 80 100 120 ACH Figure lla Comparison of new and old models with Figure lib Comparison of new atul oM models with experimentaldata for the ceiling inlet config- experimentaldata for the sidewall inlet configuration uration which represents the upper end of what might be considered temperature reached 55°F (12.8°C). The building was typical for standard HVACsystems, the old model signifi- simulated with a Sacramento,California, weather file. Note cantly underpredicts the total ventilative cooling rate. that this example does not include the effects of moisture Anotherfeature of interest is the shape of the curves. absorption and storage on loads. The purpose of the The BLASTcurve, which uses constant film coefficients, example is not to prove or disprove the efficacy of night- shows a slight curvature, which might be represented by time ventilative cooling; rather, it is to showthe importance volumetric fl~w rate raised to a fractional power. The of accurate convection modeling whenattempting to study experimental results and the new model (which uses film the problem. coefficients proportional to .P’~, yield approximatelylinear Of particular interest is howthe night purging strategy relationships. reduces the cooling load during the day. Figure 12 presents Figure llb shows an analogous comparison for the monthly cooling loads for the example building with no sidewall inlet configuration, Once again, the new model night purging, night purging modeled with the standard shows superior agreement with the experimental results. convection coefficents, and night purging modeledwith the However, the difference at low flow rates is not very newly developed convection coefficients. The monthly significant. cooling loads are the amount of cooling the system must provide to the room. Application of the NewModel The first obvious trend to be noted is that either convection model predicts significant savings in cooling One application of interest is :the analysis of night energy required when night purging is implemented. The purging systems for buildings in locations with suitable new convection model predicts about 50 % savings on an climates. Night purging refers to the introduction of large annual basis comparedwith the no-night-purging scenario. amountsof outside air at night to precool the building. In Secondly, the new convection model predicts signifi- a few cases, this alone may ameliorate the need for me- cantly lower cooling loads than the standard convection chanical air conditioning. As an example, a room the size of the experimental roomwas simulated with massive walls, floor, and ceiling. The four room walls were exposed to the outdoor environment and constructed of 4 in. (100 ram) of face brick, air space, and 4 in. (100 ram) of concrete block. The Standard Model ceiling and floor were not exposed to the outdoor environ- New Model ment and served only as thermal mass. They consisted of 8 No Purging in. (200 ram) of heavyweightconcrete. During the day, the room had 850 Btu/h (2.5 kW) internal electric equipment load and was maintained at approximately 70°F (21. l’C). At night, the room had 125 0 Wof internal gains and 100 ach of ventilation air when 2 3 4 5 6 7 8 9 10 11 12 required by the ventilation strategy. Theventilation strategy Month involved bringing in 100 ach whenever the outdoor temper- Figure 12 Comparisonof cooling loads with atwI without ature was less than the zone temperature until the zone night purging for different convection models

ASHRAETransactions: Symposia 513 model, particularly ih the cooler monthswhen there is more Topics for further research include the following: potential for night purging. During January, the new convection model predicts 35 %less cooling required due to 1. Investigation of the mixed convection problem (Ar increased cooling at night. During August, the new convec- > 0.3) for the sidewall inlet configuration. tion modelpredicts 13 % less cooling required. 2. Investigation of the lower limits of applicability of The differences between the new model and the the correlations developed for the ceiling inlet standard model are certainly significant. The standard configuration. modelsystematically underpredicts convective heat transfer 3. Investigation of applicable scaling for variation in and, therefore, ventilative cooling effects at high flow rates. roomsize (J was used here, but its applicability to much larger, or smaller, rooms has not been investigated. This will require another facility or CONCLUSIONS AND RECOMMENDATIONS modificationof the existing facility). 4. Development of a better implementation of the The following conclusions can be drawn from this correlations into BLAST(current implementation work: has parameters such as inlet area "hard-wired"). Then, investigation of the interaction between 1. A unique experimental facility has been developed convective heat transfer and the rest of the simula- that has offered the opportunity to investigate tion in a more thorough manner. manyfacets of interior convective heat transfer in 5. Investigation of the difference betweenreturn air buildings. The experiments that have been per- temperature and zone air temperature--in many formed to date are a small subset of what poten- cases, the actual zone temperature may be of interest even though it need not be determined for tially maybe investigated with the facility. These experiments investigated flow regimes and surfaces the calculation of convectiveheat transfer. that have not previously been investigated. Previ- 6. Further investigation of the effects of furniture on ous convective heat transfer experimentation in convective heat transfer. buildings has been limited to flow rates of less ACKNOWLEDGMENTS than 20 ach, and heat transfer from floors and ceilings was not investigated. The experiments described here investigated convective heat transfer This project was funded primarily by the American from all roomsurfaces, with flow rates up to 100 Society of Heating, Refrigerating, and Air-Conditioning ach. In addition, the effects of jet momentumwere Engineers (ASHRAE)through RP-529, "The Assessment investigated. and Modification of Standard Hourly Energy Calculation 2. The return air temperature was identified as the Methods for Predicting Performance of Ventilative Cool- best choice for a reference temperature. This is an ing." Its support is sincerely appreciated. In addition, important conclusion since virtually every load and several members of the monitoring committee provided energy calculation algorithm in use today makes useful advice during the course of the project. Theyinclude that assumption. The concept of a well-stirred Fred Bauman, Chip Barnaby, and John Mitchell. zone, whichleads to that model in the algorithms, Additional funding was provided by the University of did not accurately describe an average room Illinois Research Board and the Mechanical and Industrial temperature, but the outlet temperature seems to Engineering Department. Equipment loans were provided integrate the effects of the roomheat transfer when by the U.S. Army Construction Engineering Research used to determine the heat transfer coefficients. Laboratory, the University of Illinois Small HomesCotmeil, 3. Correlations suitable for use in building energy and the University of California, Berkeley. analysis programs that account for the effects of high flow rates and varying jet momentumfluxes REFERENCES for two different inlet configurations have been developed. These correlations are based on the jet Barber, E.M., S. Sokhansanj, W.P. Lampman, and LR. momentumnumber, J, and do not require the use Ogilvie. 1982. "Stability of airflow patterns in ventilat- of CFDprograms to determine room airflows. The ed airspaces." Paper No. 82-4551. St. Joseph, MI: jet momenturn number has not previously been AmericanSociety of Agricultural Engineers. used to correlate heat transfer in buildings. Chen, Q., and J. van der Kooi. 1988. "ACCURACYwA 4. Conventionalheat transfer coefficients published in program for combined problems of energy analysis, the ASHRAEhandbooks are not correct for high indoor airflow and air quality." ASHRAETransactions, ventilative flow rate applications. However,the Vol. 94, Part 2. measuredheat transfer coefficients for the sidewall Fisher, D.E. 1989. "Design of an experimental facility for inlet configuration do extrapolate to near published the investigation of convective heat transfer in enclo- values at low flow rates. The heat transfer coeffi- sures." M.S. thesis, Department of Mechanical and cients obtained using ceiling diffusers are signifi- Industrial Engineering,University of Illinois at Urbana- cantly different and do not approach the published Champaign. values at low flow rates. Gadgil, A. 1980. "On convective heat transfer in building Whenthe new convective heat tranSfer models energy analysis." Ph.D. dissertation, University of were installed in BLAST,an annual simulation of California, Berkeley. a sample building with night purging showed Lebrun, J., and P. Ngendakumana.1987. "Air circulation significant differences in predicted daytimecooling induced by heating emitters and corresponding heat ex- load. The standard coefficients systematically changes along the walls: Test-room results and model- underpredicted convective heat transfer and venti- ling." Proceedings of ROOMVENT87, Vol. 3, Stock- lative cooling effects at high flow rates. holm.

514 ASHRAETransactions: Symposia ling." Proceedings of ROOMVENT87, Vol. 3, Stock- tion of interior convective heat transfer." ASHRAE holm. Transactions, Vol. 97, Part 1. Menne,P.F. 1989. "Flow visualization in a full-scale ven- van der Kooi, J., and E. Forth. 1985. "Calculation of the tilative cooling experiment." M.S. thesis, Department cooling load by means of a ’more-air-points-model."’ of Mechanical and Industrial Engineering, University Proceedings of the CLIMA 2000 Worm Congress on of Illinois at Urbana-Champaign. Heating, Ventilating and Air-Conditioning, Vol. 4, pp. Nielsen, P.V., A. Restivo, and J.H. Whitelaw. 1979. 395-401. "Buoyancy-affected flows in ventilated rooms." Wilkes, G., and C. Peterson. 1938. "Radiation and convec- Numerical Heat Transfer, Vol. 2, pp. 115-127. tion from surfaces in various positions." Transactions, Ogilvie, J.R., and E.M. Barber. 1988. Jet momentum ASHVE,Vol. 44, pp. 513-520. number:An index of air velocity at floor level. BuiMing systems: Roomair and air contaminant distribution. DISCUSSION Atlanta: AmericanSociety of Heating, Refrigerating, and Air-Conditioning Engineers, Inc. Leon Glicksman, Professor, Massachusetts Institute of Pedersen, C., J. Spitler, D. Fisher, P. Menne, and J. Technology, Cambridge: The experimental measurements Cantillo. 1990. "Standard hourly energy calculation are very extensive. However, the use of a general cor- methods for predicting performance of ventilative relation, as proposed, should be done with caution since it cooling." Final Report of ASHRAERP-529. will vary with roomgeometry, diffuser design, etc. Randall, J.M., and V.A. Battams. 1979. "Stability criterion for airflow patterns in livestock buildings." Journal of J.D. Spifler: The correlations presented in the paper are Agricultural Engineering Research, Vol. 24, pp. 361- based on a limited numberof experiments. The experimen- 374. tal conditions are clearly specified in the paper, and the Spitler, J. 1990. "Anexperimental investigation of air flow correlations should certainly not be used if those conditions and convective heat transfer in enclosures having large do not apply. The authors certainly look forward to refine- ventilative flow rates." Ph.D. thesis, Department of ment of the correlations based on additional experimentation Mechanical and Industrial Engineering, University of and future research in computational fluid dynamics. Never- Illinois at Urbana-Champaign. theless, at the current time, these correlations represent the Spitler, 1., C. Pedersen, D. Fisher, P. Menne, and J. best model available for modelingconvective heat transfer Cantillo. 1991. "Anexperimental facility for investiga- under high ventilative flow rates.

ASHRAETransactions: Symposia 515