Wind-driven cross ventilation in long buildings

Wind-driven cross ventilation in long buildings

Accepted Manuscript Wind-Driven Cross Ventilation in Long Buildings Chia-Ren Chu , Bo-Fan Chiang PII: S0360-1323(14)00167-X DOI: 10.1016/j.buildenv...

3MB Sizes 2 Downloads 79 Views

Accepted Manuscript Wind-Driven Cross Ventilation in Long Buildings Chia-Ren Chu , Bo-Fan Chiang PII:

S0360-1323(14)00167-X

DOI:

10.1016/j.buildenv.2014.05.017

Reference:

BAE 3713

To appear in:

Building and Environment

Received Date: 5 November 2013 Revised Date:

24 April 2014

Accepted Date: 6 May 2014

Please cite this article as: Chu C-R, Chiang B-F, Wind-Driven Cross Ventilation in Long Buildings, Building and Environment (2014), doi: 10.1016/j.buildenv.2014.05.017. This is a PDF file of an unedited manuscript that has been accepted for publication. As a service to our customers we are providing this early version of the manuscript. The manuscript will undergo copyediting, typesetting, and review of the resulting proof before it is published in its final form. Please note that during the production process errors may be discovered which could affect the content, and all legal disclaimers that apply to the journal pertain.

ACCEPTED MANUSCRIPT

Wind-Driven Cross Ventilation in Long Buildings

1 2 3

Chia-Ren Chu* and Bo-Fan Chiang

4

Department of Civil Engineering, National Central University, Taiwan, R.O.C.

Abstract

6

RI PT

5

The rule of thumb for effective wind-driven cross ventilation suggests that the

8

building length L should be less than five times of ceiling height H. This study uses a

9

Large Eddy Simulation model and wind tunnel experiments to investigate the

10

mechanism behind this rule of thumb. The numerical results reveal that the ventilation

11

rate decreases as the building length increases. This is partly due to the pressure

12

difference between the windward and leeward façades of long buildings (aspect ratio

13

L/H ≥ 2.5) is smaller than that of a short building (L/H = 1.25). The other reason is

14

owing to the internal friction, which can produce a sluggish zone with a low-wind speed

15

inside the building. For buildings with aspect ratio L/H ≥ 5, the ventilation rate will be

16

over-estimated, as much as about 20%, by ventilation models that do not consider the

17

internal resistance. The location of the external openings can also influence the

18

ventilation rate. When the openings are located in the opposite corners of the windward

19

and leeward walls, also due to the internal friction, the ventilation was 15.5% less than

20

that of openings on the centerline of the building. The mitigation effect of internal

21

resistance on the ventilation rate can be quantified by a resistance model.

AC C

EP

TE D

M AN U

SC

7

22 23

Keywords: Wind-driven cross ventilation; Computational Fluid Dynamics; Large Eddy *

corresponding author: Chia-Ren Chu Mailing address: Department of Civil Engineering National Central University 300 Jhong-Da Road Jhong-Li, Taoyuan Taiwan 32001, R.O.C. E-mail address: [email protected] 1

ACCEPTED MANUSCRIPT 24

Simulation; Wind tunnel experiment, Building length.

25

27

1. Introduction Wind-driven natural ventilation is an effective way to maintain a comfortable

RI PT

26

indoor environment in residential and commercial buildings, as well as to reduce the

29

energy consumption required for mechanical ventilation [1, 2]. However, wind-driven

30

ventilation is dependent on the external wind speed, direction, openings and building

31

configurations [2, 3]. Building designers need a simple and accurate method for

32

assessing the effectiveness of wind-driven ventilation. The most widely used method to

33

calculate the ventilation rate, Q, through a building opening, is the orifice equation

34

[4-5]:

2∆P ρ

M AN U

Q = Cd A

(1)

TE D

35

SC

28

where A is the cross-sectional area of the opening, ∆P = Pe − Pi is the difference in the

37

external and internal pressures, ρ is the density of the air and Cd is the discharge

38

coefficient. Typical discharge coefficients given in the literature are in the range of 0.60

39

~ 0.65 for sharp-edged openings [3-5]. Eq. (1) is derived from Bernoulli’s assumption of

40

inviscid, incompressible flows, which has been widely used in network airflow models

41

and multi-zone models [6-8].

AC C

42

EP

36

For a building with only two openings, one on the windward façade and another

43

on the leeward façade, the dimensionless ventilation rate through the openings can be

44

found by using Eq. (1) and the continuity equation Q1 = Q2 [7, 8]:

45

 r 2C 2  Q Q = = Cd1  a 2 dr 2 C p1 − C p2  U H A1 1 + ra Cdr  *

2

1/2

(2)

ACCEPTED MANUSCRIPT where A1 is the opening area of the windward opening, UH is the external wind velocity

47

at the building height and Cp is the pressure coefficient on the external wall. The

48

pressure coefficient is defined as the pressure difference between the wall and the free

49

stream flow divided by the dynamic pressure of the approaching flow at the building

50

height z = H. The ratio of opening areas is defined as ra = A2/A1 and the ratio of

51

discharge coefficients is Cdr = Cd2/Cd1, with subscripts 1 and 2 representing the

52

windward and leeward façades, respectively. Eq. (2) can be used to calculate the flow

53

rate of cross ventilation [8]. However, it does not take into consideration the flow

54

resistance caused by the furniture or the wall friction [9].

M AN U

SC

RI PT

46

Chu et al. [10] used wind tunnel experiments to investigate the wind-driven cross

56

ventilation of partitioned buildings. They found that, due to the extra resistance caused

57

by the internal partition, the ventilation rate of a partitioned building is always lower

58

than that of a single-zone building. Chu & Wang [11] and Chu & Chiang [12] used the

59

energy equation to derive a resistance model to predict the dimensionless ventilation

60

rate Q* when there are obstacles in the buildings:

Q1*

1  C p1 − C p2  =   A1  ζ 1 + ζ i + ζ 2 

1/ 2

(3)

EP

61

TE D

55

where ζ1 and ζ 2 are the resistance factors of the external openings and ζi is the

63

internal resistance factor of the obstacle. The resistance factors of external openings can

64

be calculated as follows:

65

AC C

62

ζ1 =

1 C A12 2 d

ζ2 =

1 C A 22 2 d

(4)

66

The unit of the resistance factor is [m-4] and Eq. (3) is consistent in dimension. This

67

model considers the pressure difference between the windward and leeward openings as

3

ACCEPTED MANUSCRIPT the driving force in overcoming the resistances of cross ventilation. The larger the

69

resistance factor, ζ , the smaller the ventilation rate, Q, will be. This model can be used

70

for buildings with no internal obstacles (ζi = 0). The extra resistance caused by the

71

partition wall and furniture can be expressed in terms of internal resistance factor ζi.

72

RI PT

68

For wind-driven cross ventilation, the rule of thumb [13, 14] is that the building length L should be less than five times the ceiling height H (see Figure 1). For long

74

buildings with length L/H > 5, the wind-driven ventilation will not be effective. But

75

these references do not explain the reason behind this limitation. Is the ventilation rate

76

reduced because of the obstacles in the building or because the wall friction increases as

77

the building gets longer? Or is it because the pressure difference between the windward

78

and leeward façades (driving force) decreases as the building gets longer?

M AN U

79

SC

73

This study used a Large Eddy Simulation (LES) model to investigate the influences of building length on the flow rate of cross ventilation. The simulation results

81

were verified by wind tunnel experiments and then were used to develop a predictive

82

model for the ventilation rate and resistance factor of long buildings. This model can be

83

used to assess the wind-driven cross ventilation in residential and commercial buildings.

85 86

EP AC C

84

TE D

80

2. Numerical Model In recent years, turbulence models have been successfully applied to building

87

ventilation simulations [15-18]. The numerical results can reveal the flow parameters

88

that are difficult to measure in experiments. Kobayashi et al. [19] employed a standard

89

k- model and the Reynolds stress model to study the transported power and power

90

loss of cross ventilation in pitched roof and rectangular buildings. Hu et al. [20] used a

91

Large Eddy Simulation model to study shear-driven and wind-driven cross ventilation. 4

ACCEPTED MANUSCRIPT Tominaga et al. [21] and Ramponi & Blocken [22] discussed the application of

93

computational fluid dynamics (CFD) models to simulate the wind environment around

94

buildings. They found that the computational domain, mesh size, boundary conditions,

95

numerical scheme and convergence criterion could influence the simulation results, and

96

they concluded that the numerical setup should be carefully checked.

97

RI PT

92

The Large Eddy Simulation (LES) model used in this study to investigate the cross ventilation of long buildings was the same as that used in Chu & Chiang [12]. The

99

governing equations are:

SC

98

∂ui =0 ∂ xi

101

∂ui ∂ui u j 1 ∂P ∂ 2 u ∂τij + =− + ν 2i − ∂t ∂x j ρ ∂ xi ∂x j ∂x j

M AN U

100

(6)

(7)

where the subscripts i, j = 1, 2, 3 represent the x, y and z directions, respectively; t

103

represents the time; u and P are the velocity and pressure, the over-bar represents these

104

quantities as spatially averaged values [23], ρ is the air density, ν is the kinematic

105

viscosity of the air and τij ar and τij are the sub-grid scale stresses:

107 108 109

110

111

EP

1 τij = τkk δ ij −2µ t Sij 3

(8)

AC C

106

TE D

102

where µt is the viscosity of the sub-grid scale turbulence, defined as:

µ t = ρ(Cs ∆ s ) 2 2Sij Sij

(9)

where Cs is the Smagorinsky coefficient [24] and Sij is the rate of strain:

1  ∂u ∂u j  Sij =  i +  2  ∂x j ∂x i 

(10)

where ∆s is the mixing length of the sub-grid scales, which can be calculated as:

5

ACCEPTED MANUSCRIPT ∆s = min( κd, Cs V 1/3 )

112 113 114

(11)

where κ (= 0.41) is the von Karman constant, d is the distance to the closest wall and V is the volume of the computational cell. The value of the Smagorinsky coefficient Cs is dependent on the flow types [25]. In this study, Cs = 0.15 was chosen by

116

comparing the simulation results with the experimental data, as described in the next

117

section. The velocity near the ground and solid wall was calculated by the wall function

118

[26]:

SC

u(z) 1  ρu*z  = ln E( ) u* κ  µ 

(12)

M AN U

119

RI PT

115

120

where z is the distance from the wall, u∗ is the shear velocity, E (= 9.793) is a constant

121

and µ is the dynamic viscosity of the air.

122

The no-slip and no-penetration boundary conditions were specified on the ground and on the building walls, and the free-slip boundary condition was specified on the

124

upper and two lateral boundaries. At the outlet boundary, the zero stream-wise gradient

125

condition was used for the velocities and pressure on the cell faces. The governing equations were solved using a finite volume method. The

EP

126

TE D

123

second-order upwind scheme was employed to solve the momentum equations, and the

128

PISO (Pressure Implicit with Splitting of Operators) scheme was adopted for the

129

pressure-velocity coupling. For the unsteady flow calculation, the time derivative terms

130

were discretized using the second-order implicit scheme. The convergence criterion for

131

the continuity and momentum equations was set as 10-5 after several tests.

AC C

127

132 133 134

3. Wind Tunnel Experiments The simulation results of the present LES model were first compared with the 6

ACCEPTED MANUSCRIPT results of two wind tunnel experiments to demonstrate the accuracy of the numerical

136

model. The first experiment was conducted by in a suction, open-type wind tunnel. The

137

total length of the wind tunnel is 30 m long, the test section was 18.5 m long, 3.0 m

138

wide and 2.1 m high. The rectangular building models (height H = 0.6 m and width W =

139

0.3 m) of different lengths L = 0.3 m, 0.6 m, 1.20 m, 1.50 m and 1.80 m were mounted

140

on the centerline of the test section. The blockage ratio of the model was 2.86%. The

141

surfaces of the model were made of smooth acrylic plate without opening.

SC

142

RI PT

135

There were pressure taps (flush to the external wall) on the roof, windward and leeward sides of the building model. All of the pressure taps were connected to a

144

multi-channel high-speed pressure scanner (ZOC33/64PX, Scanivalve Inc.). The

145

measuring range of the pressure sensor was ± 2758 Pa, with a resolution of ±2.2 Pa. The

146

sampling frequency was 100 Hz, the sampling duration was 163.84 sec. The pressure

147

module was placed inside the building model with short pneumatic tubing to the

148

pressure taps.

151

TE D

stream-wise velocity U(z) follow the power law function:

EP

150

The approaching flow was a boundary layer flow and the vertical profile of

U(z)  z  =  Uo  δ 

α

(13)

AC C

149

M AN U

143

152

where Uo is the free stream velocity outside the boundary layer, z is the height from the

153

ground, δ is the thickness of the boundary layer and α is the exponent of the velocity

154

profile. In this experiment, δ = 1.45 m, α = 0.21, Uo = 10 m/s, and the wind speed at

155

the building height UH = 8.30 m/s. The wind direction ( θ = 0o) was normal to the

156

windward façade.

157

Figure 2 compares the simulated and measured pressure coefficients Cp on the 7

ACCEPTED MANUSCRIPT centerlines of building surfaces. The pressure coefficients Cp1 on the windward façades

159

are in the range of 0.55 ~ 0.80, and the profiles of pressure coefficients for building

160

lengths L/H = 0.5 and 1.0 agree very well. This indicated that the windward pressure is

161

independent of the building length. The pressure coefficient Cp1 near the leading edge of

162

building roof decreased to some extent because the air flow accelerated as it pass

163

through the roof. For the leeward pressure coefficients, the value Cp2 ≈ -0.2 of building

164

length L/H = 0.5 is smaller than the leeward pressure coefficients (Cp2 ≈ -0.1) of other

165

building lengths (L/H = 1.0, 2.0, 2.5 and 3.0). Also, the agreement between the

166

measured and predicted results validates the capability of the present LES model to

167

predict the pressure coefficients on building surface.

M AN U

SC

RI PT

158

The second experiment was conducted by Karava et al. [8]. They used particle

169

image velocimetry (PIV) to measure the velocity field inside a single-zone building

170

model, and investigated the influence of the opening configurations on the flow field

171

and ventilation rate of the building. In their model, the building height was H = 0.08 m,

172

the width W = 0.10 m and the length L = 0.10 m. The windward and leeward façades

173

each has one opening at the center of the façades. The areas of the inlet and outlet

174

openings were the same: A1 = A2 = 0.046 m x 0.018 m, and the wall porosity is r1 =

175

A1/AF = 10.35%.

EP

AC C

176

TE D

168

The wind direction was normal to the windward façade ( θ = 0o). The approaching

177

flow was a turbulent boundary layer flow with the boundary layer thickness δ = 0.6 m,

178

the exponent α = 0.15. The wind speed at the building height is UH = 6.97 m/s. The

179

Reynolds number based on the wind speed UH and building height is Re = HUH /ν =

180

37,200. The vertical profile of the time-averaged velocity U(z) and stream-wise

181

turbulence intensity Iu(z) (= σu(z)/U(z)) are shown in Figure 3, where σu is the standard

8

ACCEPTED MANUSCRIPT 182

deviation of the stream-wise velocity. Based on the experimental results of Karava et al.

183

[8], the vertical profile of the turbulence intensity is set at the inlet boundary: z Iu ( z ) = A   δ 

B

(14)

RI PT

184

where the empirical coefficients are A = 6.33, B = -0.29. The spectral synthesizer and

186

the vortex method [27, 28] were used to generate the fluctuating turbulence velocity at

187

the inlet boundary for the same turbulence intensity profile described by Eq. (14).

SC

185

The requirements as suggested by Tominaga et al. [21] and Ramponi & Blocken

189

[22] for the boundary conditions, computational domain and mesh were followed in this

190

study. The computational domain (height 6H = 0.48 m, width 10.5H = 0.84 m and

191

length 22.5H = 1.8 m) and mesh layout are shown in Figure 4. The blockage ratio of the

192

cross section of the building to the computational domain was 2.0%. The wall thickness

193

was set as 2.0 mm.

TE D

M AN U

188

The grid resolution plays an essential role in the accuracy of the LES simulation

195

[25]. In this study, the areas inside and around the building were discretized by a

196

uniform grid. For the area far away from the building, non-uniform grids with a

197

stretching ratio of 1.10 ~ 1.25 were adopted. The grid sensitivity was checked by

198

comparing the simulation results of three different grids, and it was found that the grid

199

235 × 68 × 59 with the smallest grid size of ∆x = 1.0 mm, ∆y = 3.0 mm, ∆z = 3.0 mm

200

had the best agreement with the experimental results of Karava et al. [8]. Therefore, it

201

was used for the rest of the simulation.

AC C

EP

194

202

The simulated stream-wise velocities U(x)/UH along the centerline of the openings

203

are shown in Figure 5 as dash lines. The velocity at the building height is UH = 6 m/s.

204

The solid symbols represent the experimental results of Karava et al. [8], and the dotted

9

ACCEPTED MANUSCRIPT line the simulation results of Ramponi and Blocken [22] by the shear-stress transport

206

(SST) k-ω model (their reference case). Although the simulated velocities deviated from

207

the measured velocities near the inlet and outlet openings, the agreement between the

208

numerical and the experimental results in the region x/L = 0.5 ~ 0.8 was satisfactory.

209

The discrepancy between the simulated and the measured velocities near the inlet and

210

outlet openings was due to the edge reflections and shadows of the laser light sheet [8,

211

21]. By comparing the simulated velocities and the experimental results of Karava et al.

212

[8], it was concluded that the results of the spectral synthesizer were better than those of

213

the vortex method. Therefore, the spectral synthesizer was used for the rest of this study.

214

Figure 6 illustrates the simulated velocity vectors on the mid-plane of the building.

M AN U

SC

RI PT

205

There was a standing vortex upstream of the building. Inside the building, the air flow

216

was deflected downward to the building floor as it passed through the inlet opening. The

217

flow patterns inside the building and on the building roof were very similar to the

218

experimental results of Ohba et al. [29], Karava et al. [8] and the numerical simulation

219

of Hu et al. [20].

The ventilation rate Q into the building can be calculated by two different methods.

EP

220

TE D

215

The first method is to integrate the simulated time-averaged horizontal velocities at the

222

openings. The ventilation rates at the inlet and outlet openings are represented by Q1

223

and Q2, respectively. The second method is to substitute the opening areas, discharge

224

coefficients and the predicted windward and leeward pressure coefficients into Eq. (2)

225

to obtain the ventilation rate Qmodel. The time-averaged velocities and pressure

226

coefficients are computed from the simulation results in between 30 ~ 40 sec.

AC C

221

227

The dimensionless ventilation rates through the inlet and outlet openings,

228

computed by the integration method, were: Q*1 = 0.423 and Q*2 = 0.414. The relative

10

ACCEPTED MANUSCRIPT difference 2.1% indicated that the mass conservation between the inflow and outflow

230

was satisfactory. The ventilation rate predicted by Eq. (2) was Q*model = 0.419, with the

231

simulated external pressure coefficients right above the windward and leeward openings

232

Cp1 = 0.66, Cp2 = -0.17 and discharge coefficients Cd1 = 0.65, Cd2 = 0.65. In other words,

233

the ventilation rates calculated by the integration model were very close to the

234

prediction of Eq. (2). The good agreement between the computed ventilation rates by

235

the two different methods validated the capability of the present LES model to simulate

236

the wind-driven cross ventilation.

SC

RI PT

229

238 239

4. Computational Setup

M AN U

237

The present LES model was used to investigate the wind-driven cross ventilation through a single isolated low-rise building. The external height and width of the

241

building were fixed as height H = 4 m and width W = 10 m (see Figure 7). The length of

242

the building was varied in the range of L = 5.0 ~ 44.0 m (the aspect ratio L/H = 1.25 ~

243

11). The reference case was set as the building length L = 10 m (L/H = 2.50). For all the

244

cases, the windward and leeward façades each had one opening on the centerline of the

245

façades. The cross-sectional areas of inlet and outlet openings were identical A1 = A2 =

246

2 m x 2 m, and the wall porosity was r1 = r2 = 10%. The thickness of the building wall

247

was 0.1 m.

EP

AC C

248

TE D

240

The approaching flow followed Eq. (13) with the value of δ = 400 m and the

249

free-stream wind speed was Uo = 12 m/s. Therefore, the wind speed at the building

250

height was UH = 6.0 m/s. The Reynolds number based on the wind speed UH and

251

building height was Re = HUH /ν = 1.6 × 106. The turbulence intensity profile followed

252

Eq. (14), and the turbulence intensity at the building height was Iu = 23.9%. The wind 11

ACCEPTED MANUSCRIPT 253

direction was normal to the windward façade ( θ = 0°) of the building for all the

254

simulation cases. The computational domain was 24 m (= 6H) in height and 70 m (= 7W) in width.

256

The length of the computational domain varied according to the building length, the

257

distance from the leeward façade to the outlet boundary was 63 m (= 15.75H). The

258

blockage ratio of the building to the computational domain was 2.38%. The

259

computational grid was adjusted from the grid used in the model verification. For all the

260

simulation cases, the grid size inside the building was the same: ∆x/H = 0025, ∆y/H =

261

0.05, ∆z/H = 0.05. The area far away from the building employed a non-uniform grid

262

with the stretching ratio 1.25. For the reference case (building length L/H = 2.5), the

263

grid number inside the building was 87,952 cells and the total grid number was

264

1,064,770 cells (= 245 × 82 × 53 cells).

M AN U

SC

RI PT

255

TE D

265

5. Results and Discussion

267

5.1 Building Length

EP

266

Figure 8(a) shows a comparison of the predicted pressure coefficients on the

269

centerlines of the windward façades for different building lengths. The windward

270

pressure coefficient Cp1 above the inlet opening was in the range of Cp1 = 0.60 ~ 0.75,

271

and independent of the building length. This phenomenon is similar to the results of

272

wind tunnel experiments. Furthermore, the values of windward pressure coefficient are

273

in the same range of experimental results. The pressure coefficient Cp1 right above the

274

windward opening (z/H = 0.6) decreased slightly because the air accelerated as it flow

275

into the opening.

276

AC C

268

On the other hand, Figure 8(b) shows that the leeward pressure coefficient Cp2 ≈ 12

ACCEPTED MANUSCRIPT -0.60 for L/H = 1.25 was much smaller than the leeward pressure Cp2 = -0.15 ~ -0.20 for

278

longer buildings (L/H ≥ 2.5). It also illustrated that the leeward pressure was

279

independent of the building length when L/H ≥ 2.5. Again, this is similar to the

280

experimental results of building length L/H longer than 1.0 (see Figure 2(b)). Altogether,

281

Figures 2 and 8 imply that the pressure difference between the windward and leeward

282

façades (the driving force for cross ventilation) of short buildings was higher than that

283

for long buildings. In other words, the ventilation rate of short buildings will be larger

284

than that of long buildings.

SC

RI PT

277

The contours of time-averaged pressure coefficient Cp on the mid-plane of the

286

building for building lengths L/H = 1.25, 2.5, 5.0 and 8.0 are illustrated in Figure 9. Due

287

to the separated shear layer on the building roof, the pressure at the leeward region of

288

building length L/H = 1.25 was lower than for the other building lengths. For building

289

length L/H ≥ 2.5, because the length of separated shear layer was less the building

290

length and the reattachment of eddies on the building roof, the leeward pressure was

291

much alike. Also the pressure distribution inside the building was comparable.

TE D

It is also noteworthy that, both the experimental and numerical results show that

EP

292

M AN U

285

the aspect ratio L/H of buildings is more influential to the leeward pressures than to the

294

windward pressures. This information is important to those choosing pressure

295

coefficients for ventilation design [30]. It can be explained by the separation bubble at

296

the building roof. However, the critical building length is L/H > 1.0 for the wind tunnel

297

experiment, the critical building length is L/H > 2.5 for the numerical simulation. This

298

difference is believed due to the aspect ratio H/W of the building. The aspect ratio H/W

299

= 0.4 for the wind tunnel experiments, while H/W = 2.0 for the numerical simulation.

300

AC C

293

The time-averaged velocity vectors on the mid-plane of the building for L/H =

13

ACCEPTED MANUSCRIPT 1.25 and 2.5 are compared in Figure 10. The separated shear flows around the leading

302

corner on the building roof look very much alike, but there was reattachment toward the

303

trailing corner of the building for L/H = 2.5, while no reattachment for L/H = 1.25

304

because the separation bubble was longer than the building length. Also, there was a

305

standing vortex inside the building between the ceiling and outlet opening, and the

306

counter-clockwise rotating vortex generated resistance for the air flow passing through

307

the building.

SC

308

RI PT

301

The dimensionless ventilation rates Q*, computed by the integration method, are shown in Figure 11. The values of Q* were normalized by the dimensionless ventilation

310

rate of reference case Q*r = 0.448. When building length L/H 2.5, Eq. (2) predicted the

311

ventilation rates very well. But for longer buildings (L/H ≥ 5.0), the ventilation rate

312

decreased as the building length increased and Eq. (2) over-predicted the ventilation

313

rates Q*. The ventilation rate of L/H = 11 was only 80% of the ventilation rate of L/H =

314

2.5. This suggested that the diminishing effect of the internal friction on the ventilation

315

rate cannot be neglected when the building length L/H ≥ 5.0.

TE D

The extra resistances caused by the friction inside the building are expressed in

EP

316

M AN U

309

terms of the internal resistance factor ζi in Eq. (3). The value of ζi can be computed by

318

substituting the simulated pressure coefficients Cp1 and Cp2 from LES into Eq. (3). The

319

resistance factors of the external openings ζ1 = ζ2 = 0.148 m-4 were computed by setting

320

the discharge coefficient Cd = 0.65 and the opening area A1 = A2 = 4 m2. Figure 12

321

displays a non-linear relation between the dimensionless resistance factor ζi/ζ1 and the

322

building length L/H. The internal resistance was negligible when L/H < 2.5, but

323

increased drastically when L/H > 5. This led to the decline in the ventilation rate of long

324

buildings.

AC C

317

14

ACCEPTED MANUSCRIPT 325

Because the building floors are smooth surfaces, the skin friction estimated by the boundary layer theory [31] is quite small. The energy loss is mainly due to the

327

circulating, turbulent flow in the building. When building length L/H ≥ 5, the internal

328

friction caused by the turbulent flow increased almost linearly with the building length.

329

Figure 12 also infers that the resistance factors ζ1 and ζ2 of the external openings are the

330

dominant parameters for short buildings without any internal obstacle (furniture and

331

partition). However, the resistance factor ξi cannot be neglected for buildings with large

332

aspect ratio L/H.

SC

Figure 13 displays the time-averaged stream-wise velocity along the centerline of

M AN U

333

RI PT

326

the openings at height z/H = 0.25 (z = 1 m). The stream-wise velocities were normalized

335

by the wind speed UH. For building length L/H = 1.25, the internal velocity was the

336

highest due to the large pressure difference across the building, but it decreased as the

337

building length increased. For short buildings, the air flow past thought the windward

338

and leeward openings resembling a strong jet flow, without much energy loss. However,

339

the air flow in long buildings like a diffused jet, velocity decrease as it goes further into

340

the building. For L/H ≥ 5.0, the internal velocity became u/UH < 0.4 in the region near

341

the leeward opening. The velocity rise near the outlet is due to the suction at the leeward

342

façade. The region with a low internal velocity is called the sluggish zone. The length of

343

the sluggish zone was about x/L = 0.65 ~ 0.95 for L/H = 5, but increased to x/L = 0.25 ~

344

0.95 for L/H = 11. Designing cross ventilation in long spaces should pay attention to

345

this phenomenon.

AC C

EP

TE D

334

346 347 348

5.2 Diagonal Openings Besides the building length, the opening configurations can generate additional

15

ACCEPTED MANUSCRIPT resistance for cross ventilation. Karava et al. [8] and Ramponi & Blocken [18] studied

350

the wind-driven cross ventilation of different opening configurations. However, they did

351

not quantify the influences of the location of the openings on the ventilation rate. The

352

purpose of this simulation was to compare the ventilation rates of openings located on

353

the diagonal of the building (see Figure 14) with that of the reference case (openings at

354

the center of the building façades). The size of the building (height H = 4.0 m, width W

355

= 10.0 m and length L = 10.0 m), opening areas (A1 = A2 = 2 m x 2 m), wind direction

356

( θ = 0°) and velocity profile at the inlet boundary were identical to those of the

357

reference case. But the openings were placed in opposite corners of the windward and

358

leeward facades, along the diagonal of the building.

SC

M AN U

359

RI PT

349

Figures 15(a) and 15(b) illustrate the pressure coefficients above the windward and leeward openings, respectively. The location of the pressure profile was on the

361

centerline (y = 0) for the reference case, while the pressure profile was taken at y/W =

362

±0.40 for the diagonal case. The windward pressures were very similar for openings

363

located on the centerline and on the diagonal direction. However, the leeward pressures

364

were slightly different for the reference and diagonal cases.

EP

365

TE D

360

The pressure difference ∆P between the windward and leeward openings was the driving force for wind-driven cross ventilation. As listed in Table 1, the pressure

367

coefficient difference ∆CP = Cp1 - Cp2 above the opening (z/H = 0.7) of the reference

368

and diagonal cases was 0.925 and 0.911, respectively. However, the dimensionless

369

ventilation rate of diagonal case Q* = 0.368, as computed by the integration method,

370

was much lower than that of reference case Q*r = 0.448. This disparity was attributed to

371

the internal resistance as the air flow passed through the diagonal openings. In other

372

words, the ventilation models that did not consider internal resistance, such as Eq. (2),

AC C

366

16

ACCEPTED MANUSCRIPT 373

over-estimated the ventilation rate for openings not on a straight line. The additional resistance for the air flow through diagonal openings could be

375

quantified using the internal resistance factor ζi. Similar to the calculation used for the

376

resistance factors of long buildings, the dimensionless resistance factor of the diagonal

377

openings was obtained as ζi / ζ1 = 0.856. For the same opening size and similar pressure

378

difference, the ventilation rate (calculated by integration method) through the building

379

with diagonal openings was 15.5% less than that of the reference case (openings on the

380

centerline of the building). For long buildings or buildings with furniture inside, the

381

wind-driven cross ventilation will be further reduced.

M AN U

382 383

SC

RI PT

374

6. Conclusions

The rule of thumb for effective wind-driven cross ventilation suggests that the

385

building length should be less than five times the ceiling height. This study used a Large

386

Eddy Simulation model and wind tunnel experiments to investigate the mechanism

387

behind this rule of thumb in a single-zone, low-rise building. The numerical and

388

experimental results both revealed that the windward pressure was independent of the

389

aspect ratio of the building, but the leeward pressures of short buildings were lower than

390

that of long buildings. This is due to the separation bubbles on the building roof can

391

reach the leeward sides of short buildings. In other words, the ventilation rate of a short

392

building is higher owing to the larger pressure difference between the windward and

393

leeward façades.

394

AC C

EP

TE D

384

Furthermore, it was found that the ventilation rate of building length L/H = 11 is

395

20% less than that of building length L/H = 2.5, even when the pressure difference

396

across the building and the opening size were the same. Plus, there was a sluggish zone 17

ACCEPTED MANUSCRIPT with low wind speed inside the building when the building length was L/H ≥ 5.0. It is

398

the internal friction that caused the sluggish zone and the decline in ventilation rate. The

399

conventional ventilation models that do not consider the internal resistance will

400

over-estimate the ventilation rate of long buildings.

RI PT

397

The location of the openings also affected the ventilation process. The simulation

402

results indicated that the ventilation rate for building with openings in opposite corners

403

of the windward and leeward façades was 15.5% lower than that of openings on the

404

centerline of the building, even though the pressure differences between the windward

405

and leeward façades of the building were very close. This was attributed to the

406

additional resistance for air flow to pass through the diagonal openings. The resistance

407

model of Chu and Chiang [12] can be used to quantify the internal friction of different

408

building lengths and opening configurations.

M AN U

SC

401

410

Acknowledgement

TE D

409

The authors thank the Architecture and Building Research Institute, Ministry of

412

The Interior (Grant no. 10162B001), and National Science Council of Taiwan (Grant no.

413

102-2221-E-008-055) for their support of this study.

AC C

414

EP

411

415

References

416 417

[1] Allard F. Natural ventilation in buildings: a design handbook, James and James Ltd., London, England; 1998.

418

[2] Aynsley R. Estimating summer wind driven natural ventilation potential for indoor

419 420 421 422

thermal comfort. J Wind Eng Ind Aerodyn 1999; 83 (1-3):515-25. [3] Linden PF. The fluid mechanics of natural ventilation. Annual Review of Fluid Mechanics 1999; 31:201-38. [4] Etheridge D. Natural ventilation of buildings: Theory and measurement and design. 18

ACCEPTED MANUSCRIPT 423

John Wiley and Sons, Chichester, England, 2012.

424 425

[5] Awbi HB. Ventilation of Buildings. 2nd ed. Taylor and Francis, London, England,

426

[6] Etheridge D, Sandberg M. Building ventilation: Theory and Measurement. John

428 429 430

Wiley and Sons, Chichester, England, 1996. [7] Chu CR, Chiu YH, Chen YJ, Wang YW, Chou CP. Turbulence effects on the

RI PT

427

2003.

discharge coefficient and mean flow rate of wind-driven cross ventilation. Build Environ 2009; 44: 2064-72. doi:10.1016/j.buildenv.2009.02.012.

[8] Karava P, Stathopoulos T, Athienitis AK. Airflow assessment in cross-ventilated

432

buildings with operable façade elements. Build Environ 2011; 46(1): 266-79.

433 434

SC

431

[9] Johnson M-H, Zhai Z, Krarti M. Performance evaluation of network airflow models for natural ventilation. HVAC&R Research 2012; 18(3): 349-365. [10] Chu CR, Chiu YH, Wang YW. An experimental study of wind-driven cross

436 437

ventilation in partitioned buildings. Energy Build 2010; 42(5): 667-73. doi:10.1016/j.enbuild.2009.11.004.

438 439 440

[11] Chu CR, Wang YW. The loss factors of building openings for wind-driven ventilation. Build Environ 2010; 45(10): 2273-79. doi:10.1016/j.buildenv. 2010.04.010.

441 442

[12] Chu CR, Chiang BF. Wind-driven cross ventilation with internal obstacles. Energy Build 2013; 67,:01-09. doi:10.1016/j.enbuild.2013.07.086.

443 444 445

[13] CIBSE. Natural ventilation in non-domestic buildings: Applications Manual 10, AM10: Chartered Institution of Building Services Engineers (CIBSE), London, 1997.

446

[14] Mumovic D, Santamouris M. A Handbook of Sustainable Building Design and

TE D

EP

Engineering: An integrated approach to energy, health and operational performance, Earthscan, England, 2009.

AC C

447 448

M AN U

435

449 450

[15] Evola G, Popov V. Computational analysis of wind driven natural ventilation in buildings. Energy Build 2006; 38:491-501.

451 452 453

[16] Kato S, Murakami S, Mochida A, Akabashi S, Tominaga Y. Velocity-pressure field of cross-ventilation with open windows analyzed by wind tunnel and numerical simulation. J Wind Eng Ind Aerodyn 1992; 41-44: 2575-86.

454 455

[17] Jiang Y, Chen Q. Effect of fluctuating wind direction on cross natural ventilation in buildings from large eddy simulation. Build Environ 2002; 37(4): 379-86.

456 457

[18] Ramponi R, Blocken B. CFD simulation of cross-ventilation flow for different isolated building configurations: validation with wind tunnel measurements and 19

ACCEPTED MANUSCRIPT 458 459

analysis of physical and numerical diffusion effects. J Wind Eng Ind Aerodyn 2012; 104-106: 408-18.

460

[19] Kobayashi T, Sagara K, Yamanaka T, Kotani H, Sandberg M. Power transportation

461 462

inside stream tube of cross-ventilated simple shaped model and pitched roof house. Build Environ 2009; 44(7):1440-51. [20] Hu CH, Ohba M, Yoshie R. CFD modelling of unsteady cross ventilation flows using LES. J Wind Eng Ind Aerodyn 2008; 96(10-11):1692-706.

465 466 467

[21] Tominaga Y, Mochida A, Yoshie R, Kataoka H, Nozu T, Masaru Y, Shirasawa T. AIJ guidelines for practical applications of CFD to pedestrian wind environment around buildings. J Wind Eng Ind Aerodyn 2008; 96:1749-61.

468 469

[22] Ramponi R, Blocken B. CFD simulation of cross-ventilation for a generic isolated building: Impact of computational parameters. Build Environ 2012; 53:34-48.

470

[23] Germano M, Piomelli U, Moin P, Cabot WH. A dynamic subgrid scale eddy

SC

M AN U

471

RI PT

463 464

viscosity model. Physics of Fluids A 1991; 3(7): 1760-65.

472 473

[24] Smagorinsky J. General circulation experiments with the primitive equations. I. The basic experiment. Monthly Weather Review 1963; 91(3): 99-164.

474

[25] Chen Q. Ventilation performance prediction for buildings: A method overview and recent applications. Build Environ 2009;44:848-58. doi:10.1016/ j.buildenv.2008.05.025.

TE D

475 476

[26] Launder BE, Spalding DB. The numerical computation of turbulent flows. Computer Methods in Applied Mechanics and Engineering 1974; 3:269-89.

479

[27] Meroney RN. CFD prediction of dense gas clouds spreading in a mock urban

480 481

environment. In: The Fifth International Symposium on Computational Wind Engineering, 2010. Chapel Hill, North Caroline, USA.

EP

477 478

[28] Huang SH, Li QS, Wu JR. A general inflow turbulence generator for large eddy simulation. J Wind Eng Ind Aerodyn 2010; 98 (10-11): 600-617.

484

[29] Ohba M, Irie K, Kurabuchi T. Study on airflow characteristics inside and outside a

485 486

cross-ventilation model, and ventilation flow rates using wind tunnel experiments. J Wind Eng Ind Aerodyn 2001; 89:1513-24.

487 488 489 490

AC C

482 483

[30] Costola D, Blocken B, Hensen JLM. Overview of pressure coefficient data in building energy simulation and airflow network programs Build Environ 2009; 44:2027-2036. [31] Schlichting H. Boundary-Layer Theory. New York: McGraw-Hill; 1979.

491

20

ACCEPTED MANUSCRIPT

Notation

493

A1, A2

cross-section area of external opening (m2)

494

A2/A1

opening ratio (dimensionless)

495

Aw

cross-section area of building interior (m2)

496

AF

area of the windward facade (m2)

497

Cd

discharge coefficient of opening (dimensionless)

498

Cp = (P-Po)/0.5ñU2

external pressure coefficient (dimensionless)

499

H

height of building (m)

500

L

length of building (m)

501

Po

reference pressure (Pa)

502

Q

ventilation rate (m3/s)

503

Q*

dimensionless ventilation rate

504

r1, r2 = A/AF

external wall porosity (dimensionless)

505

Uo

free stream wind speed (m/s)

506

UH

wind speed at the building height (m/s)

507

W

width of building exterior (m)

508

z

509

ρ

510

g

511

α

512

δ

513

θ

514

ζi

515

æ1 , æ2

516

Subscripts

517

1

windward side

518

2

leeward side

519

i

internal

TE D

M AN U

SC

RI PT

492

elevation (m)

air density (kg/m3)

EP

gravitational acceleration (m/s2) exponent of the velocity profile

AC C

thickness of boundary layer (m) wind direction (degree)

internal resistance factor (m-4) external resistance factor (m-4)

21

ACCEPTED MANUSCRIPT

1

Table Caption:

2

Table 1. Simulation results of diagonal openings.

3 4

RI PT

5 6 7

10

Case

ΔCP

Q*

Reference

0.925

0.448

Diagonal

0.915

ζi (m-4)

ζi/ζ1

0

0

M AN U

9

Table 1. Simulation results of diagonal openings.

SC

8

0.368

0.127

0.856

ΔCP is the difference between the windward and leeward pressure coefficients above the opening (z/H = 0.7).

AC C

EP

TE D

11 12

1

ACCEPTED MANUSCRIPT Figure Caption: Figure 1. Schematic diagram of wind-driven cross ventilation in long buildings. Figure 2. Comparison of simulated and measured pressure coefficient on the centerlines of building surface. (a) windward façades of building height L/H = 0.5 and 1.0. (b) leeward façades of L/H = 0.5, 1.0, 2.0, 2.5 and 3.0. The symbols are

RI PT

the measured pressure coefficients, the lines are the numerical predictions.

Figure 3. Vertical profiles of time-averaged velocity U(z) and stream-wise turbulence intensity Iu(z) (= σu(z)/U(z)) of the approaching flow.

Figure 4. Computational domain and mesh layout for the numerical simulation.

SC

Figure 5. Comparison of predicted stream-wise velocity U(x)/UH along the centerline of boundary.

M AN U

the openings by different schemes to generate velocity fluctuation at the inlet

Figure 6. Time-averaged velocity vectors on the mid-plane of the building. The flow condition followed the experiment of Karava et al. [8]. Figure 7. Schematic diagram of the full-scale building for the numerical simulation. The openings are on the centerlines of windward and leeward walls, opening

TE D

area A1 = A2 = 2 m x 2 m.

Figure 8. Pressure coefficients on the centerlines of building external walls: (a) above windward opening; (b) above leeward opening. Figure 9. Time-averaged pressure distribution on the mid-plane of the building for

EP

different building lengths. (a) L/H = 1.25; (b) L/H = 2.5; (c) L/H = 5.0; (d) L/H = 8.0.

AC C

Figure 10. Time-averaged velocity vectors on the mid-plane of the building for different building length. (a) L/H = 1.25; (b) L/H = 2.5.

Figure 11. Effect of building length on the ventilation rate, Q*r is the dimensionless ventilation rate of reference case (L/H = 2.5).

Figure 12. Dimensionless internal resistance ζi/ζ1 as a function of building length. Figure 13. Time-averaged streamwise velocity along the centerline of the openings at height z/H = 0.25. Figure 14. Schematic diagram of a full-scale building with openings on the diagonal corners of the windward and leeward walls. The opening area A1 = A2 = 2 m x 2 m. 1

ACCEPTED MANUSCRIPT Figure 15. Pressure coefficients on the building external walls: (a) above windward opening (x/L = 0, y/W = -0.4); (b) above leeward opening (x/L = 1.0, y/W =

M AN U

SC

RI PT

0.4).

Wind

L

P1

H

TE D

Q1

EP

A1

P2 Q2 A2

AC C

Figure 1. Schematic diagram of wind-driven cross ventilation in long buildings.

2

ACCEPTED MANUSCRIPT (a) 1.0 LES, L/H=0.5 Exp. L/H=0.5

0.8

Exp. L/H=1.0

0.4 0.2

0.2

(b) 1.0

0.6

0.8

1.0

TE D

AC C

EP

Z/H

0.6

0.2

Cp

LES, L/H = 0.5 Exp. L/H = 0.5 Exp. L/H = 1.0 Exp. L/H = 2.0 Exp. L/H = 2.5 Exp. L/H = 3.0

0.8

0.4

0.4

M AN U

0.0 0.0

SC

Z/H

RI PT

0.6

0.0 -0.8

-0.6

-0.4

-0.2

0.0

Cp

Figure 2. Comparison of simulated and measured pressure coefficient on the centerlines of building surface. (a) windward façades of building height L/H = 0.5 and 1.0. (b) leeward façades of L/H = 0.5, 1.0, 2.0, 2.5 and 3.0. The symbols are the measured pressure coefficients, the lines are the numerical predictions. 3

ACCEPTED MANUSCRIPT

0.48

0

5

10

Iu (%)

15

20

25

Exp., Karava et al. (2011) Inlet Velocity Profile

0.40

RI PT

Exp., Karava et al. (2011)

z (m)

0.32

Inlet turbulence intensity

0.24

SC

0.16

0.00

0

4

M AN U

0.08

8

12

16

20

U (m/s)

TE D

Figure 3. Vertical profiles of time-averaged velocity U(z) and stream-wise turbulence

AC C

EP

intensity Iu(z) (= σu(z)/U(z)) of the approaching flow.

4

ACCEPTED MANUSCRIPT

RI PT

Wind

4.25H L+1.5H

M AN U

SC

6H

TE D

15H

AC C

EP

Figure 4. Computational domain and mesh layout for the numerical simulation.

5

ACCEPTED MANUSCRIPT

1.2 LES, vortex method LES, spectral synthesizer k-ω, Ramponi and Blocken (2012) Exp., Karava et al. (2011)

1.0

RI PT

U(x) / UH

0.8 0.6

0.2 0.0 -0.35

0.5

1.0

M AN U

0.0

SC

0.4

1.5

x/L

Figure 5. Comparison of predicted stream-wise velocity U(x)/UH along the centerline of the openings by different schemes to generate velocity fluctuation at the inlet boundary.

AC C

EP

TE D

The flow condition followed the experiment of Karava et al. [8].

6

M AN U

SC

RI PT

ACCEPTED MANUSCRIPT

Figure 6. Time-averaged velocity vectors on the mid-plane of the building. The flow

AC C

EP

TE D

condition followed the experiment of Karava et al. [8].

7

RI PT

ACCEPTED MANUSCRIPT

A2

4.0 m A1 z

x

SC

2.0 m

2.0 m

10 m

M AN U

y

U

L

TE D

Figure 7. Schematic diagram of the full-scale building for the numerical simulation. The openings are on the centerlines of windward and leeward façades, opening area A1 = A2

AC C

EP

= 2 m x 2 m.

8

ACCEPTED MANUSCRIPT (a) Windward facade

1.0 0.8

RI PT

z/H

0.6 0.4

L/H = 1.25 L/H = 2.5 L/H = 5.0 L/H = 8.0 L/H = 11.0

0.0 0.0

0.4

0.6

0.8

1.0

M AN U

0.2

SC

0.2

Cp

(b) Leeward facade

1.0

TE D

0.8

0.4

AC C

0.2

L/H = 1.25 L/H = 2.5 L/H = 5.0 L/H = 8.0 L/H = 11.0

EP

z/H

0.6

0.0 -0.8

-0.6

-0.4

Cp

-0.2

0.0

0.2

Figure 8. Pressure coefficients on the centerlines of building external walls: (a) above windward opening; (b) above leeward opening.

9

ACCEPTED MANUSCRIPT

z (m)

M AN U

x (m)

SC

RI PT

(a) L/H = 1.25

TE D

AC C

EP

z (m)

(b) L/H = 2.5

x (m)

10

C

p

ACCEPTED MANUSCRIPT

x (m)

M AN U

(d) L/H = 8.0

SC

z (m)

RI PT

(c) L/H = 5.0

TE D

z (m)

EP

x (m)

AC C

Figure 9. Time-averaged pressure distribution on the mid-plane of the building for different building lengths. (a) L/H = 1.25; (b) L/H = 2.5; (c) L/H = 5.0; (d) L/H = 8.0.

11

ACCEPTED MANUSCRIPT

M AN U

SC

RI PT

(a) L/H = 1.25

AC C

EP

TE D

(b) L/H = 2.5

Figure 10. Time-averaged velocity vectors on the mid-plane of the building for different building length. (a) L/H = 1.25; (b) L/H = 2.5.

12

ACCEPTED MANUSCRIPT

150

RI PT

125

75

*

*

Q / Qr (%)

100

SC

50

Integration method Eq. (2)

0

0

2

4

M AN U

25

6

8

10

12

14

L/H

Figure 11. Effect of building length on ventilation rate, Q*r is the dimensionless

AC C

EP

TE D

ventilation rate of reference case (L/H = 2.5).

13

ACCEPTED MANUSCRIPT

1.0

0.8

RI PT

ζi / ζ1

0.6

0.4

0.0

0

2

4

6

8

12

14

TE D

M AN U

L/H

10

SC

0.2

Figure 12. Dimensionless internal resistance ζi/ζ1 as a function of building length.

AC C

EP

Openings on the centerlines of the windward and leeward façades.

14

ACCEPTED MANUSCRIPT

2.0

L/H = 1.25 L/H = 2.50 L/H = 5.00 L/H = 8.00 L/H = 11.0

1.6

RI PT

0.8 0.4 0.0 0.0

0.2

0.4

0.6

0.8

1.0

TE D

M AN U

x/L

SC

u / UH

1.2

Figure 13. Time-averaged streamwise velocity along the centerline of the openings at

AC C

EP

height z/H = 0.25.

15

4.0 m

RI PT

ACCEPTED MANUSCRIPT

2.0 m

A2

2.0 m

z

A1

SC

x

10 m

TE D

U

M AN U

y

10 m

Figure 14. Schematic diagram of a full-scale building with openings on the diagonal

AC C

EP

corners of the windward and leeward façades. The opening area A1 = A2 = 2 m x 2 m.

16

ACCEPTED MANUSCRIPT (a) Windward facade

1.0 0.8

RI PT

z/H

0.6 0.4

Reference case Diagonal case

0.0 0.0

0.2

0.4

0.6

0.8

1.0

M AN U

Cp

SC

0.2

(b) Leeward facade

1.0

TE D

0.8

0.4

AC C

0.2

EP

z/H

0.6

0.0 -0.4

Reference case Diagonal case

-0.3

-0.2

-0.1

0.0

Cp

Figure 15. Pressure coefficients on the building external walls: (a) above windward opening (x/L = 0, y/W = -0.4); (b) above leeward opening (x/L = 1.0, y/W = 0.4).

17

ACCEPTED MANUSCRIPT Title: Wind-driven cross ventilation in long buildings

* This study used numerical model to investigate the ventilation of long buildings.

RI PT

* The ventilation rate decreased as the building length increased. * The rule of thumb for wind-driven cross ventilation is clarified.

AC C

EP

TE D

M AN U

SC

* The location of the building openings also influenced the ventilation rate.