Experimental study on the flow characteristics of air curtains at building entrances

Building and Environment 105 (2016) 225e235

Contents lists available at ScienceDirect

Building and Environment journal homepage: www.elsevier.com/locate/buildenv

Experimental study on the flow characteristics of air curtains at building entrances Sherif Goubran, Dahai Qi, Wael F. Saleh, Liangzhu (Leon) Wang*, Radu Zmeureanu Centre for Zero Energy Building Studies, Department of Building, Civil and Environmental Engineering, Concordia University, 1455 de Maisonneuve Blvd. West, Montreal, Quebec, H3G 1M8, Canada

a r t i c l e i n f o

a b s t r a c t

Article history: Received 21 April 2016 Received in revised form 26 May 2016 Accepted 27 May 2016 Available online 2 June 2016

Air infiltration through building entrances is one of the main sources of energy loss in modern buildings. Previous studies have shown that air curtains, when used at building entrances, can reduce infiltrationrelated energy loss significantly. A recent computational fluid dynamics (CFD) study proposed a new empirical model to capture air curtain door infiltration/exfiltration characteristics under varying operation conditions and pressure differences. Extending the recent CFD study, this paper presents an experimental study to verify and further investigate the flow characteristics of building entrances equipped with air curtains. A small scale chamber of 2.44 m
2.44 m
1.3 m (L
W
H) was constructed and used for the measurements of infiltration/exfiltration and pressure differentials, which were then used for developing the empirical model across the operating air curtain. A 2-D particle image velocimetry (PIV) system with helium filled soap bubbles as seeds was used to visualize the airflow fields captured at the doorway. Both the PIV and the measurement-based correlations were also compared to CFD simulations. The flow/pressure measurements confirmed that, for the tested pressure difference range, air curtains can significantly reduce infiltration. The PIV results confirmed the existence of multiple flow characteristics subject to pressure differences across the air curtain. The experimental results also validated the CFD modeling methods for air curtain, and verified that the empirical model of air curtain from the literature is valid in estimating infiltration through building entrances equipped with air curtains. © 2016 Elsevier Ltd. All rights reserved.

Keywords: Air curtain Experimental validation Particle image velocimetry Air infiltration Helium filled soap bubble

1. Introduction Air infiltration is the uncontrolled inward leakage of outdoor air into buildings through cracks in the envelope or through large openings such as doors [1]. Air leakage, movement of air in or out of buildings, is mainly caused by pressure difference across the various building enclosure elements. These differences in pressure can be caused by many factors such as wind, stack effect, and/or HVAC system operations [1]. In modern well constructed and insulated buildings, it is estimated that air infiltration can be responsible for up to 25% of the building heating loads and it can mainly be due to entrance doors and their use [2]. Based on the available research and standards [3,4], many building energy codes1 require vestibule

* Corresponding author. E-mail addresses: [email protected], [email protected] (L. Wang). 1 Such as the American Society of Heating, Refrigerating and Air-Conditioning Engineers (ASHRAE) Standard 90.1 e Energy Standard for Buildings Except LowRise Residential Buildings, as well as International Energy Conservation Code [1]. http://dx.doi.org/10.1016/j.buildenv.2016.05.037 0360-1323/© 2016 Elsevier Ltd. All rights reserved.

doors in commercial buildings to reduce the infiltration through entrance doors. Another solution, which has been in use for more than 50 years and offers cost and space savings for building owners, is the use of air curtains [5e7]. Air curtain units, typically made of a fan and casement with a jet outlet, are used as to create an air barriers in various applications: they have become a standard in cold storages and food cabinets, they are used in buildings to block smoke in the event of fires, as well as to control dust in the mining industry [5e8]. For entrance doorways, the units are most commonly mounted above the doorway/opening and are designed to supply one (or more) jet(s) of air that are engineered to reach the floor at a particular velocity and position in order to seal openings aerodynamically [9]. Air curtains are defined by the American Society of Heating, Refrigeration and Air-conditioning Engineers (ASHRAE) as a continuous broad stream of air circulated across a doorway of a conditioned space which reduces penetration of unconditioned air into conditioned spaces by forcing an air stream over the entire entrance [10]. The air stream layer moves with a velocity and angle such that any air that tries to penetrate the

226

S. Goubran et al. / Building and Environment 105 (2016) 225e235

Nomenclatures and abbreviations

hair

Efficiency factor for infiltration reduction (air curtain) Area Width of the air curtain nozzle (air outlet width) Automatic door coefficient (for specific door size) Airflow coefficient Discharge coefficient section Average discharge coefficient for each door operation section CDave Average flow coefficient for a full door operation cycle Dd & DDave Discharge modifier (air curtains) Dm Deflection modulus of air curtain jet Fu Air curtain infiltration usage correction factor H Door height ƞ Efficiency factor for heat transfer reduction (air curtain) Ph People per hour (door usage) Q Volume flow rate Rp Pressure factor Th Usage per hour (doors) u0 Air curtain discharge speed V0 Air curtain volume supply rate a0 Air curtain discharge angle

A b0 C CA Cd CD,

curtain is entrained [10]. Besides restricting air infiltration into buildings, air curtain doors (single doors equipped with air curtain units) can also provide building users with un-obstructed entranceways, blocking outdoor pollutants, respirable dust, insects, and moisture [11]. Manufacturers of the units claim that air curtain doors can reduce the energy losses from entrance doors by 70% compared to building entrance doors without air curtains [8,11,12]. Hayes [13] was able to develop theoretical models that describe the airflow and jet of vertically downwards blowing air curtains for isothermal and non-isothermal cases. Hayes [13] proposed the deflection modulus of air curtains to be used in the design and selection of the units. The deflection modulus, Dm, which is indicative of the deflection of the air curtain, is the ratio of the air curtain jet momentum to transverse forces due to the stack effect on the jet [13]. Hayes and Stoecker then developed, based on the “deflection modulus”, widely used design charts for air curtains that provide the minimum jet outlets momentum needed for the jet to reach the floor [14]. In their work, Hayes and Stoecker defined the operation condition where the air curtain jet reaches the floor as the “optimum condition” and other operation conditions where the air curtain jet does not reach the floor as “break-through condition” [14]. They also proposed evaluating the performance of air curtains by calculating the air infiltration efficiency factor, hair , which evaluates the ability of the air curtain to limit the air infiltration in reference to the unprotected opening. This measure has been widely used by researchers in the industry up till this day [14]. Pappas and Tassou have extended this concept to measure the efficiency of air curtains in reducing heat losses by proposing the heat transfer efficiency factor, ƞ [15]. However, the work of Hayes and Stoecker was limited to fully sealed rooms under constant and steady conditions of pressure and temperature differences; the

2 Such as the Building Services Research and Information Association (BSRIA) Application Guide 2/97 Air Curtains e Commercial Applications for the design of air curtains [5].

DP DPlc DPuc

Pressure difference across the air curtain door Lower critical pressure difference Upper critical pressure difference

Abbreviations AC Air curtain Air curtain door Double swing doors opening out with air curtain unit AMCA Air Movement and Control Association (International) Door mid-plane Vertical plane 30.5 cm away from the left side of the chamber door edge (middle of the door opening) Door side-plane Vertical plane 8.5 cm away from the left side of the chamber door edge e in the experimental chamber Fully open door Door open at 90 HFSB Helium filled soap bubbles PIV Particle Image Velocimetry RMS Root mean square values (PIV velocity fields) Single door Double swing doors opening out (without vestibule or air curtain) Vestibule door Vestibule with double swing doors opening out (2 set of double swing doors)

models developed were only applicable for air curtains operating under optimum conditions [14]. This resulted in other design guides2 to conclude that air curtains were not suitable for leaky buildings under windy conditions [5]. However, other research has shown that even when operating in a breakthrough condition, air curtains can still provide better protection than single or vestibule doors [16]. What is most important to note is that most of the studies dealing with air curtains infiltration have focused on single or steady condition analysis (usually with the door fully open) to evaluate the performance of air curtain doors, neglecting the variable conditions, varying usage and the door opening cycles that entrance doors experience during their normal operation. In contrast, when it comes to doors not equipped with air curtains (i.e. single doors and vestibule doors), the model developed by Yuill [3] is widely used in predicting air infiltration through doors. Yuill’s model (presented in Eq. (1)) uses the well known orifice equation and presents average discharge coefficients for the door, CDave, based on the usage frequency and considers the full door opening cycle [3]: during one operation cycle of a single door, the door opens from the shut-off position (0 ) to the fully open position (90 ) (i.e. the opening section a), stays fully open (i.e. the section b), shuts off (i.e. the section c), and finally closes completely (i.e. the section d).

Q ¼ CDave ATh

sffiffiffiffiffiffiffiffiffi 2DP

(1)

r

where CDave is the overall average discharge coefficient,

CDave ¼

CD a a þ CD b b þ CD c c þ CD aþbþcþd

dd

CD, section (a,b,c,d) is the average discharge coefficient for each door operation section,

S. Goubran et al. / Building and Environment 105 (2016) 225e235

a, b, c & d are the time period (in seconds) for each door operation section, Th is the door usage time per hour (h/h), A is the full door size (m2), DP is the pressure difference across the door (Pa), and r is the air density (kg/m3).

[18]. Based on the CFD simulation setting used their study, more than 330 simulations were conducted in order to develop air curtain infiltration characteristics curves for pressure differences across the door ranging from 30 Pa to 50 Pa. The results of the numerical study of Wang et al. [18] confirmed and extended the concepts proposed by Hayes’ study [13] as it identified 3 main operational conditions for air curtains (seen in Fig. 1).

In a recent study of air curtain by Wang and Zhong [18,19], the methods proposed by Yuill [3] were extended to include the analysis of double swing doors equipped with air curtains [16,18]. In the study, extensive CFD simulations were conducted using ANSYS Fluent 14.0 [20] for the 20 m
24 m
10 m (L
W
H) building within a full domain of 50 m
24 m
10 m (L
W
H). A double swing automatic door of 2 m
2.4 m (W
H) was modeled at the center of the building’s width. The door opening angles were varied from 10 to 90 (fully open). The air curtain was modeled following the methods similar to those proposed by Belleghem et al. [7] with a supply slot of 0.08 m
2 m (b0 ¼ 0.08 m), an indoor air intake, and an air supply of 15 m/s at 20 outwards. The door was then simulated with different pressure conditions that were set as the boundaries of the CFD model (pressure inlet and outlet), the stack effect was considered by setting different indoor and outdoor temperatures and the door opening angle was also varied. The actual pressure difference across the door (in order to consider the effect of temperature) was calculated at the mid-section of the model. The CFD cases assumed steady state flow and uniform supply velocity across the jet based on the standard k  ε model which was found to be widely used in other sources of literature

Outdoor

Po

 Optimum condition: where the air curtain flow reaches the flow and seals the door  Inflow breakthrough: where the air curtain flow is curved inward and does not reach the floor  Outflow breakthrough: where the air curtain flow is curved outwards and does not reach the floor The results indicated that there exist two critical pressure differences across the door as shown in Fig. 1-D: the upper critical pressure, DPuc, higher than which the air curtain operates at the inflow breakthrough condition and the lower critical pressure, DPlc, lower than which the air curtain operates in the outflow breakthrough condition [18]. Wang et al. [18] also developed an empirical model (Eq. (2)) based on that proposed by Yuill [3] by adding to the infiltration rate equation a discharge modifier, Dd, angle (Pa0.5). This indicated that the airflow through a door with an air curtain is determined by both the pressure difference and the air curtain jet. The model which was developed in the study was used to calculate the infiltration rates through the air curtain door [16].

Outdoor

Indoor

Indoor

Po

Pi

(A). Optimum condition

227

Outdoor

Pi

Indoor

Pi

Po

(B). Inflow breakthrough

(C). Outflow breakthrough

when ∆Poi > ∆Puc

when ∆Poi < ∆Plc

when ∆Plc < ∆Poi <∆Puc

Q (m3/s)

5

Air Curtain Door

4

3

Single Door

2

1

-40

-30

-20

∆Plc

0

-10

Outflow Breakthrough

Inflow Breakthrough

∆Puc 0

--1

10

20

Optimum Condition

30

40

50

∆Poi (Pa)

--2 --3 -4

(D). An illustration of the typical Q - ∆P curve for an air curtain compared to a single door. Fig. 1. (A)e(C): the three operation conditions of air curtains [18] and (D): a typical QeDP curve for an air curtain compared to a single door.

228

S. Goubran et al. / Building and Environment 105 (2016) 225e235

e

Q ¼ ð  1Þ CDave ATh

sffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi 2jDPoi j

r

þ DDave Th

sffiffiffi 2

r

(2)

where

DDave ¼

DD a a þ DD b b þ DD c c þ DD aþbþcþd

dd

and e ¼ 0 when DPoi  0 and e ¼ 1 when DPoi < 0. Or in shorter term:

pffiffiffiffiffiffiffi Q ¼ C DP þ D

(3)

The correlation and calculation of the coefficients at each door opening angle was completed using Eq. (4).

Qangle qffiffiffi ¼ ð1Þe CD; A 2r

angle

qffiffiffiffiffiffiffiffiffiffiffiffi jDPoi j þ DD;angle

(4)

Based on numerical simulations, the study by Wang et al. [18] was able to present a method to quantify the amount of infiltration through air curtain doors under variable weather conditions and door usage frequencies using primarily numerical simulations. However, an extensive literature review indicated a lack of field or experimental measurements that could be used to validate the findings of Wang et al. [18], e.g. the existence of the operation conditions as shown in Fig. 1 and the correlation as defined by Eq. (4) for an air curtain. There is also a lack of experimental verification of their CFD method modeling air curtain, for example, the air curtain in their study was modeled as a perfectly uniform jet (i.e. it was assumed that the air curtain is supplying air at a perfectly uniform speed and angle throughout the width of the door) which is not necessarily the case for real units [16]. To answer for these research gaps, this paper presents the findings of an experimental study conducted for a small scale air curtain door in a test chamber with varying pressure differences across the door. CFD simulations using the same method as Wang et al. [18] were also conducted for the test chamber. The study compares the air curtain correlations (i.e. defined by Eq. (4)) based on the experimental volume flow rates and pressure measurements to the ones calculated from CFD simulations. In addition, particle image velocimetry (PIV) is used to capture the flow fields at the door which is also compared to the results of the CFD simulations. The paper aims to confirm the existence of the three air curtain flow conditions (Fig. 1) that are proposed in previous studies [14,16,18]. Additionally, the experimental results are used to confirm that air curtain door’s infiltration characteristics can be modeled by Eq. (4)

proposed by Wang et al. [18]. The comparison between the numerical and experimental results aims to also validate various CFD related modeling methods of air curtains. The paper specifically aims to validate the method of modeling the air curtain supply in CFD and the effect of the supply non-uniformity on the air curtain performance. The experimental study also helps to confirm the suitability of using the standard k-3 model (found to be repeatedly used in literature) when modeling air curtain doors along with its associated turbulence parameters (turbulence intensities and boundary conditions). The experimental study is also used to help validate other CFD related settings that are proposed by Wang et al. [18] such as grid resolutions, high gradient flow regions, discretization schemes and their accuracy (1st/2nd order), wall functions and under-relaxation factors. 2. Methodology 2.1. Experimental setup The experimental setup consists mainly of an airtight transparent/translucent chamber with an exhaust fan, a door equipped with an air curtain, pressure sensors located inside of the enclosure, flow gauge at the exit of the fan, and a PIV system to capture the airflow field near the door. The whole structure of the test chamber is a cube with a side length of 2.44 m (hereafter the CUBE for Concordia University Building Environment chamber) made of aluminum framing and Lexan panels (Fig. 2-A). The chamber has a double swing single door, 0.61 m
0.71 m (W
H), with an adjustable opening angle (for this study the doors are fully open at 90 ) equipped with an above door air curtain unit (Fig. 2-B). The configuration was designed so that the door and the duct fan were the only openings in the chamber. Thus, based on the law of conservation of mass, the volume flow rate measured at the fan are equal to that passing through the door. It is important to note that the CUBE is divided internally into two vertical spaces by a horizontal ceiling. The actual tests only used the lower section of 2.44 m
2.44 m
1.3 m (L
W
H), which is the same as that used by Yuill [3] in his investigations on automatic doors infiltration which represents a 1/3 scale of a real building. In order to calibrate the apparatus, single door measurements were compared to Yuill’s theoretical model [3]. A commercial air curtain unit was made available from an air curtain manufacturer (seen in Fig. 2-B). The air curtain is a 120 V center motor unit (motor placed in the center) with centrifugal fans on each side (total 2 fans). The unit’s supply slot is 6.35 cm
61 cm (depth
width  same width as the door) equipped with three discharge direction vanes (which are set to 20 outwards for this study). To control the supply speed, the unit includes a VFD

Fig. 2. (A) Picture the CUBE in the testing space (left), (B) a picture of the above door air curtain placed inside the CUBE with one of the HFSB nozzles (middle) and (C) HFSB system € control fitted in the CUBE (right)epictures provided by MEKANIC.

S. Goubran et al. / Building and Environment 105 (2016) 225e235

controller with six configurations. The supply speeds at the surface of the unit for different power configurations were measured and averaged based on two sets of measurements completed at 45 points across the supply slot. For this study, two power configurations were selected: (1) the maximum supply of the unit at an average of 13.75 m/s (the average nominal speed) and (2) the medium speed configuration with an average of 9.1 m/s. Preliminary tests were conducted to check the speed variance at the supply slot surface. The average variance of the supply speed was ±40% based on the standard deviation air supply speed data. At both power configurations used, the highest supply speed was recorded approximately 0.085 m away from the edges and the lowest speed was at the center of the unit (0.305 m from the edge). The CUBE’s interior pressure is varied by a fan system with controllable and measurable flow rates [22] (duct blaster fan in Fig. 2-A). Note that when the fan was used to pressurize the chamber (i.e. to supply air into the enclosure), an air diffuser was added to the outlet to limit the direct flow of air to the door. The interior pressure in the room was measured at the center plane between the door and the fan. The reference pressure was measured at a protected point 1.2 m away from the door opening following the same location criteria used by Wang et al. [18]. In order to obtain a close-to-average pressure at this plane, sensors (tube endings) at four different points (nodes) throughout this plane were placed and equipped with adequate protection to limit fluctuations in the readings. Based on the equipment and setup used, the flow rate achieved from the duct fan (airflow through the door/fan) was of a maximum of approximately 0.4 m3/s (infiltration) and a minimum of approximately 0.3 m3/s (0.3 m3/s of exfiltration). For the experiments, the volume flow rate through the fan is measured and plotted as a function of pressure differences across the door. All flow and pressure measurements were conducted based on 10 s time average while allowing adequate time for flow stabilization in the chamber before the measurements. In order to ensure the repeatability of the measurements, all readings were repeated twice and averaged (Table 1). In regards to calculated error for the averaged measurements, the pressure readings had a random instrumental error of ±1.3% or ±0.21 Pa (whichever is larger) and the volume flow rate readings had random instrumental error and a leakage systematic error (based on the air tightness tests of the chamber) of ±9% or ±0.02 m3/s (whichever is larger). To visualize the airflow fields at the doorway, the Solo PIV 120XT model by New Wave Research with its controller [23] was used for this study. The laser is a Nd:YAG dual laser head system with a pulse energy of 120 mJ at 532 nm equipped with a telescope lens. The CCD cameraeDANTEC Dynamics [24]eused in the experiment is a thermo-electrically cooled 14 bit camera with a 2 M resolution (1200
1600 pixels). The camera is equipped with a 60 mm lens (2.8/32, by Nikon). For such a large scale application, one of the suitable seeding methods identified in the literature was Helium Filled Soap bubbles (HFSB). A specially designed HFSB system (Fig. 2-C) was used in this study which was developed in-house based on the commercial HFSB systems available [25] and the concepts presented by Bosbach et al. and Scarano et al. [26,27]. The

Table 1 Experimental data acquired for this study. Condition

Measurements

PIV measurements

AC Speed 1: Avg. 9.1 m/s AC Speed 2 Avg. 13.75 m/s Single Door Total

15 16 10 82 (41 conditions
2)

12 (incl. 3 repeated) 12 (incl. 3 repeated) 24

229

HFSB nozzles used consist of 3D printed plastic polymer bodies each fitted with 5 needle stainless steel tubes connected to a central control station able to operate up to four nozzles simultaneously. During the experiments, all four nozzles were used and each was capable of producing about 80e150 HFSB/s between 0.8 and 1.6 mm. Two of the nozzles were placed inside the air curtain unit casement directly at the supply slot, while the two others were placed on the outside facing the door opening from the opposing sides (Fig. 3). Considering that the air curtain velocity exceeding 10 m/s, the laser light sheets were adjusted to coincide (maximum 1e2 mm apart) at the doorway to capture the seeds within the same plane. The measurement plane were selected (position of PIV measurement planes presented in Fig. 3); (a) at 8.5 cm from the door edge and coinciding with the maximum supply speed of the unit (i.e. door side-plane), and (b) at 30.5 cm from the edge which is the middle of the door opening and coinciding with the lowest supply speed of the unit (i.e. door mid-plane). The PIV data acquisition was completed using the software package of the PIV system, FlowManager 4.71 [28]. Based on the size of the seeds, the airflow conditions of the air curtain and the models available in the literature [29e32], the optimum delay time, Dt, was calculated to be 800 s. Based on testing, 31 frame pairs (62 images) for each PIV case was identified as the optimum number (i.e. the number that allowed for the flow to be correctly captured). Different interrogation window sizes were tested: 8
8, 16
16, 32
32, 64
64 and 128
128 pixels. It was found that 64
64 pixels were the optimum size to represent the flow with maximum number of vectors: the measured flow field did not show major difference for the bigger window size of 128
128. Therefore, the interrogation area was selected to be 64
64 pixels and the correlation was completed using the adaptive correlation method with an overlap of 50% (central differencing method with 1.2/2 min peak for validation and an acceptance factor of 0.1). The data were then averaged with a 3
3 averaging area and the root mean square values of velocity (RMS values) of the velocity vectors were then generated. Again, to ensure the repeatability of the data measured, some selected cases were captured twice and compared (Table 1). The average standard deviation reported for all the cases’ RMS values is approximately ±20%. The overall plan view of the experimental setup is presented in Fig. 3-A, which clarifies the configuration of all equipment as well as the location of the laser planes used in reference to the test chamber. In addition, Fig. 3-A also presents the pressure nodes located in mid-plane of the chamber. The field of view of the CCD camera, the location of the air curtain and the HFSB nozzles (Pl.1 and Pl.2) are clearly presented in Fig. 3-B. It is important to note that the placement of the nozzles inside the air curtain was necessary to achieve the needed seeding density. 2.2. Numerical simulations A CFD model of the CUBE and it surrounding (Fig. 4) was created on ANSYS Fluent 14.0 [20]. The CFD model used the similar boundary conditions as those available in the real laboratory. The air curtain’s supply was modeled using the FLUENT’s User Defined Function (UDF) based on the measured speed profile of the unit with an outward angle. In addition, the duct blaster fan was modeled as a velocity outlet (with a diffuser modeled as a surface inside the chamber): the pressure in the chamber was controlled and decided by the airflow through the fan which was input in the simulation software. The pressures were evaluated at the same points illustrated in Fig. 3-A. The CFD cases, as proposed by Wang et al. [16,18], assumed steady state flow based on the standard k  ε model. For this study, all cases are isothermal to match the experimental conditions in the laboratory, i.e. the thermal condition in

230

S. Goubran et al. / Building and Environment 105 (2016) 225e235

Fig. 3. Illustration of the (A) final experimental setup plan view showing equipment and measurement positions (top) and (B) a section view with the field of view and seeding (nozzle) positions highlighted (bottom).

Fig. 4. CFD model of the experimental chamber and its surroundings.

S. Goubran et al. / Building and Environment 105 (2016) 225e235

231

Table 2 CFD simulations completed for the experimental validation study.

Air curtain (AC) cases

Air curtain supply models

Average supply speed

Simulations

Model using measured AC supply profile

AC AC AC AC

10 13 4 4 9 40

Model assuming uniform supply speed Single door cases Total number of simulations

speed speed speed speed

1: 2: 1: 2:

9.1 m/s 13.75 m/s 9.1 m/s 13.75 m//s

NA

the lab was well maintained all the time so the measured temperature difference between the interior and exterior of the chamber was minimal. Simulations were conducted for the two experimental supply speeds (namely 9.10 m/s and 13.75 m/s average supply speed) as well as for the single door (air curtain off). The details of the simulations conducted are presented in Table 2. It is also worth mentioning that, the CFD model used for the experiments assumes the same CFD parameters used by Wang and Zhong [18] in regards to grid resolutions and wall functions (and other boundary conditions). Grid-independent results were confirmed for two grid resolutions (not shown here). Since the model proposed by Wang et al. [16,18] assumed a uniform supply speed for the air curtain, a number of simulations were thus conducted using the uniform supply speed rather than the actual supply profile. This step was crucial in the validation process of the modeling method of air curtains.

3. Results and discussion For the flow and pressure data, the experimental measurements and the CFD results are both plotted on the same “Pressure vs. Volume Flow Rate” graph. A direct comparison is used to evaluate the cases. Some numerical comparisons are conducted between the correlated CFD data and the experimental results to calculate the differences in the flow rates at the door. For airflow fields, the velocity field obtained from the PIV values are compared to the CFD data at the two measurement planes (namely at the door sideplane and the mid-plane). The CFD results were exported to

match the field of view captured by PIV.

3.1. Air curtain infiltration/exfiltration characteristics In Fig. 5, the infiltration rates from the experiments (i.e. “Experiment”) are compared the CFD simulations (i.e. “CFD” in the figures) for air curtains with the different supply jet velocities (average jet supply jet speed of 9.1 m/s and 13.75 m/s with 20 outward angle). To better show the trends, correlated CFD results based on Eq. (4) (i.e. “Simulation Correlation” in the figure) are plotted. Fig. 5 also compares infiltration rates through the single door as measured in the experiments and calculated by CFD simulations. In addition, the single door and vestibule door infiltration curves obtained, and extrapolated in the negative pressure range, based on Yuill’s study are presented [3]. The results shown in Fig. 5 indicate an agreement between the experimental measurements and CFD simulations for both average air curtain supply speeds of 9.1 m/s and 13.75 m/s under different pressure differences. Fig. 5 also indicate that both the experiment results and CFD simulations data agree well with the empirical curve developed by Yuill [3] for the pressure difference range DP ¼ 0.35e1.25 Pa with an average difference of 3.48%. The experimental data (Fig. 5) confirm the existence of two flow conditions of air curtain: the optimum and the inflow break-through conditions. For both supply speed tests, the zones of the optimum operation condition fall in the negative flow rate zone (i.e. the net flow through the door is negative). When the flows, Q, are close to zero and become positive (inflows), it indicates the change to inflow breakthrough

1.6 Q (m3/s)

Air Curtain - 13.75 m/s (Experiment)

1.4

Air Curtain - 9.1 m/s (Experiment)

1.2

Single Door (Experiment)

1

Air Curtain - 13.75 m/s (CFD)

0.8

Air Curtain - 9.1 m/s (CFD)

0.6

Single Door (CFD)

0.4

Single Door (Yuill, 1996)

0.2

Vestibule (Yuill, 1996)

0 -2

0 -0.2

-0.4

2

4

6

8

10

12

14

16

18 ΔP (Pa)

20

Air Curtain -9.1 m/s (Simulation Correlation) Air Curtain -13.75 m/s (Simulation Correlation)

Fig. 5. Comparison of the measured net airflow rate, Q, versus the pressure difference, DP, among the experimental and CFD simulation data for the air curtain with different jet velocities (average jet supply speed of 9.1 m/s and 13.75 m/s at 20 outwards), and for the single door and vestibule from the previous study (air curtain off).

232 Table 3 CD, 90 and DD,

S. Goubran et al. / Building and Environment 105 (2016) 225e235

90

calculated for the tested cases based on the CFD simulations (the door angle at 90 indicates a fully open door).

Air curtain average supply speed (m/s)

9.1 (DPuc ¼ 5.3 Pa) 13.75 (DPuc ¼ 11.5 Pa)

CD,

DD,

90

90

Optimum condition

Inflow breakthrough

Optimum condition

Inflow breakthrough

0.160 0.268

0.858 0.684

0.478 0.999

2.033 2.417

conditions. This confirms the air curtains operation conditions proposed in previous studies [14,16,18]. Note that the present study is limited to depressurization or slight pressurization because the flow from the blower door fan may potentially interface with air curtain jets complicating airflow inside the chamber when pressurizing the chamber at high pressures. This is similar to Yuill’s experiments, in which many of the blower door tests were depressurization instead of pressurization. When supplying air at 13.75 m/s, the tested air curtain is able to maintain the optimum condition under higher pressure differences than that at 9.1 m/s. As shown in Fig. 5, the upper critical pressure, DPuc, was around 5.3 Pa for the 9.1 m/s case whereas it was significantly higher at around 11.5 Pa for the 13.75 m/s case. Both the experimental and CFD data indicate that, for the same outward supply angle, the higher jet supply speed (in this research 13.75 m/s average supply speed compared to 9.1 m/s) ensure lower infiltration rates under high pressure difference conditions. On the other hand, a higher supply speed may cause more exfiltration under low pressure difference conditions. Compared to the single door, and the vestibule door (for comparison purpose, the empirical model developed by Yuill was used here), the measured data and the correlated curves for the air curtain show a significant reduction of air infiltration through the door for both supply speeds. However, the results indicate a trend that when operating under low pressure difference conditions, air curtains may result in higher exfiltration rates than single/vestibule doors. The correlated CD, 90 and the DD, 90 in Eq. (4) were presented in Table 3. The flow modifier, which represents the outward flow of

the air curtain, for the 9.1 m/s supply speed case is lower than at that for the 13.75 m/s case. For the 9.1 m/s supply case the flow coefficient is lower in the optimum condition range and higher in the inflow breakthrough range when compared to the 13.75 m/s case. For the 9.1 m/s average supply speed case, the average difference between the experimental data and the correlated simulation curve was calculated to be 15.07% and for the 13.75 m/s average supply case the average difference was calculated to be 20.67%. This average differences, even though high, can be considered acceptable considering that the flow measurements equipment systematic and random error can be as high as 9% or higher in low volume flow rate measurements. This confirms that the correlation proposed by Wang et al. [18] in Eq. (4) is valid for quantifying flow characteristics of air curtains. In addition, the agreement between the CFD and experimental data gathered for the case validate the modeling methods used and the choice of CFD-related settings. Note that each test in Fig. 5 was repeated at least once to check the repeatability of the results and an uncertainty analysis based on the repeated tests was also conducted as shown by the error bars in the figure. 3.2. Air curtain PIV measurements The PIV Root Mean Square (RMS) values are compared to the inplane velocity flow fields extracted from the CFD simulations for the air curtain supply of 9.1 m/s 20 outwards in Fig. 6 and 13.75 m/ s 20 outwards in Fig. 7. As observed in Fig. 6 when the air curtain is

Fig. 6. Comparison between experimental and CFD simulation flow visualization at door side-plane for the air curtains with an average supply speed of 9.1 m/s at 20 outwards at different pressure conditionsemoving from an almost outflow breakthrough to optimum then inflow breakthrough conditions (from left to right).

S. Goubran et al. / Building and Environment 105 (2016) 225e235

233

Fig. 7. Comparison between experimental and CFD simulation flow visualization at door side-plane for the air curtains with an average supply speed of 13.75 m/s at 20 outwards at different pressure conditionsemoving from an almost outflow breakthrough to optimum then inflow breakthrough conditions (from left to right).

operating at the lowest pressure difference (0 Pa in the experiments), the air curtain is bent outwards experiencing an almost outflow breakthrough condition. In contrast, at the highest pressure difference (8.5 Pa in the experiments) the air curtain jet is completely curved inwards and experiencing an obvious inflow breakthrough with outdoor air seeping underneath the jet. The rest of PIV results show a transition of airflow patterns between these two extremes at variable pressure differences. Similar trends were observed for the 13.75 m/s supply case in Fig. 7 but at higher pressure differences than Fig. 6: e.g. clear inflow breakthroughs occur when DP is over 11.5 Pa when compared to that over 5.3 Pa for the supply 9.1 m/s. It is also apparent that the air curtain jet is able to resist curving at much higher pressure differences for higher supply speed. The comparison of CFD to the PIV results in both figures that the overall flow patterns captured by PIV are similar to the calculated ones by CFD at respective pressure differences. The velocity magnitudes also compare better between the measurements and simulations at upper portion of the air curtain jets than that near the floor: the jet velocities near the floor are generally higher in the simulations than those from the measurements, or in other words, the PIV measurements seem to indicate higher jet diffusing than the simulations. These differences can be caused by several reasons. In the CFD simulations, the non-uniform distribution of the supply velocity magnitudes was measured and modeled whereas the variations of supply angles were not. Using a simple technique of smoke tracer, the present study found that the air curtain motor located in the center of the unit created a three-dimensional and turbulent flow, which is hard to be captured and quantified by the existing instruments available to this study including the anemometer and the 2D PIV system. Even for the 2D PIV system, a large average standard deviation was reported, ± 20%, due to a number of factors including the transient and three-dimensional nature of the air curtain flow. Another possible source of uncertainties is the HFSB seeding method and the locations of the

nozzles, which need further investigations. Due to the limitations, the comparison between the CFD and PIV velocity flow fields was completed mainly qualitatively in terms of the jets’ curvature, direction and general velocity profile. Above all these limitations, however, it is clear that the captured PIV data show similar flow conditions as obtained from the CFD simulations at corresponding pressure differences. In addition, Figs. 6 and 7 confirm the existence of the three flow conditions of air curtains: optimum condition, inflow breakthrough, and outflow breakthrough as defined in the previous studies [16,18,19]. The PIV results also confirm that air curtains at higher air supply speed are able to resist air infiltration better. The air curtain with a supply speed of 13.75 m/s at 20 outwards maintains its optimum condition up to around 11 Pa, while the 9.1 m/s supply at the same angle maintains the optimum condition only up to around 5 Pa. 3.3. Air curtain supply uniformity The previous numerical studies by Wang et al. [14,16,18] assume uniform air supply for air curtains. The present study investigates the effects of the assumption of uniformity on air curtain flow characteristics. As seen in Fig. 8, for the two supply speeds tested, the simplification of the air curtain assuming a uniform air supply (disregarding the real supply profile) can still capture the air curtain unit performance. The simplified uniform model used in previous numerical studies [16,18] is thus considered to be accurate in capturing the infiltration characteristics of air curtain doors. These findings imply that the tested air curtain’s overall uniformity level (the overall variation of the supply speed, angle and flow condition) seem within the acceptable level (i.e. not significantly effecting the performance of the unit overall). 4. Conclusion A specially designed experimental test chamber equipped with

S. Goubran et al. / Building and Environment 105 (2016) 225e235

Q (m³/s)

234

0.6

Air Curtain - 13.75 m/s (Experiment)

0.4

Air Curtain - 9.1 m/s (Experiment)

0.2

Air Curtain - 13.75 m/s (Simulation Correlation)

ΔP (Pa)

0 -2

0

2

4

6

8

10

12

14

16

18

20

Air Curtain - 9.1 m/s (Simulation Correlation)

-0.2

Air Curtain Uniform 13.75 m/s (Simulations)

-0.4

Air Curtain Uniform Supply 9.1 m/s (Simulation)

Fig. 8. The CFD results for the air curtain with uniform supply compared to experimental tests results and the correlated air curtain simulations that consider the tested unit’s supply profile.

an air curtain was constructed. The chamber was used to conduct an experimental study to investigate the flow characteristics of air curtain doors. A CFD simulation study was also conducted for the chamber and its surroundings based on the methods proposed by Wang et al. [18]. Flow and pressure measurements as well as PIV flow visualization data were obtained at the door from the experimental chamber and the CFD simulations. The flow and pressure measurements showed that the infiltration rates measured through the air curtain door are significantly lower than the measured infiltrations through the single door. It also showed that the infiltration rates measured through the air curtain door are significantly lower the theoretically calculated infiltrations through a vestibule door of the same size by Yuill [3]. The infiltration characteristic curves developed for the door at the two air curtain supply speeds confirmed the existence of two flow conditions of air curtains: the optimum and the inflow breakthrough conditions. Additionally, the experimental results indicated that the higher the air curtain supply speed, the better the air curtain performs (i.e. it is able to continue to operate in the optimum condition at higher pressure differences). The data presented indicate that the experimental infiltration rates measured through the chamber’s air curtain door conform well to the data obtained from numerical CFD simulations within the pressure difference range of 2 Pa to 20 Pa. The analysis of the experimental and numerical data confirmed the validity of the empirical model developed by Wang et al. [16,18] to estimate the air curtain performance in the optimum and inflow breakthrough conditions. A twodimensional PIV system with a specially designed and fabricated seeding helium filled soap bubble (HFSB) apparatus was utilized to visualize the airflow patterns near the door under varying pressure conditions. The captured PIV data confirmed the existence of the three airflow conditions of air curtains from the previous studies [18]: namely, inflow breakthrough, outflow breakthrough and optimum conditions. The air curtain jets’ curvature, direction and general velocity profile calculated by CFD show a general agreement with the PIV measurements. However, noticeable differences of airflow magnitudes were observed, especially near the floor, due to the 3-D and turbulent features of the flow and the limitations of the PIV system. In summary, it is concluded that the air curtain’s theoretical and numerical modeling method used in the previous studies [16,18] are valid and are able to capture the infiltration/exfiltration

performance of various air curtain doors. It also showed that the effect of the air curtain supply uniformity on the performance should be further investigated for future research. It is important to note that the current tests were mostly focused on the inflow breakthrough and optimum conditions of the air curtains operation. The current experimental setup and the equipment allowed for reaching a minimum pressure difference of 0 Pa (with the air curtain average supply of 9.1 m/s). However, testing the more extreme exfiltration breakthrough conditions needs further pressurization which can be the focus of future studies. In addition, this study focused on specific air curtain supply velocities for specific sizes of doors. For other cases, the determination of the flow discharge coefficients and modifiers still need either experiments or numerical simulations. Therefore, further research is needed to develop general and possibly universal models that can be used to estimate the infiltration/exfiltration performance of air curtain doors in terms of any given door sizes and air curtain supply velocities. More work should also be conducted to relate the reduction of infiltrations from air curtains to their effectiveness, energy performance, and occupant comfort. Acknowledgement The authors acknowledge the financial and technical support from the Air Management and Control Association (AMCA) (Grant number: Air Curtain Project - Phase III), Mars Air Systems, Berner International, Powered Aire and InterCode Incorporated. The au€ thors would also like to thank MEKANIC team for the construction of the experimental chamber as well as providing some of the professional images for the experimental setup used in this paper. References [1] ASHRAE, ASHRAE Standard 90.1-2010: Energy Standard for Buildings except Low-rise Residential Buildings, 2010. [2] S.J. Emmerich, A.K. Persily, Energy impacts of infiltration and ventilation in U.S. Office buildings using multizone airflow simulation, in: IAQ and Energy 98 Conference, 1998, pp. 191e203. [3] G.K. Yuill, Impact of High Use Automatic Doors on Infiltration, 1996. Pennsylvania. [4] ASHRAE, ASHRAE Handbookefundamentals, American Society of Heating, Refrigerating and Air Conditioning Engineers, Inc., Atlanta, 2009. [5] F. Alamdari, Air Curtains Commercial Applications e Application Guide 2/97, BSRIA, Berkshire, 1997.

S. Goubran et al. / Building and Environment 105 (2016) 225e235 [6] Y.G. Chen, Parametric evaluation of refrigerated air curtains for thermal insulation, Int. J. Therm. Sci. 48 (10) (2009) 1988e1996. [7] M. Van Belleghem, G. Verhaeghe, C. T’Joen, H. Huisseune, P. De Jaeger, M. De Paepe, Heat transfer through vertically downward-blowing single-jet air curtains for cold rooms, Heat. Transf. Eng. 33 (June 2013) (2012) 1196e1206. [8] D. Johnson, P. Thomas, L. Kordecki, Berner International Corporation, Air Curtains : a Proven Alternative to Vestibule Design, Berner Int. Corp., 2008. [9] J.J. Costa, L. a. Oliveira, M.C.G. Silva, Energy savings by aerodynamic sealing with a downward-blowing plane air curtain e a numerical approach, Energy Build. 38 (10) (2006) 1182e1193. [10] ASHRAE, ASHRAE HandbookeHVAC Applications, American Society of Heating, Refrigerating and Air Conditioning Engineers, Atlanta, 2007. [11] S. Indac, Determination of the Climate Separation Efficiency of Biddle Air Curtains Doorflow HP-200 and Indac S200 under Laboratory Conditions, 2011. Netherlands. [12] I.B.E. Cremers, Air Curtains : Faster Is Not Always Better!, 2012. [13] F.C. Hayes, Heat Transfer Characteristics of the Air Curtain: a Plane Jet Subjected to Transverse Pressure and Temperature Gradients, University of Illinois, 1968. [14] F.C. Hayes, W.F. Stoecker, Design data for air curtains, ASHRAE Trans. (1969) 168e180. [15] T.C. Pappas, S.A. Tassou, Numerical investigations into the performance of doorway vertical air curtains in air-conditioned spaces, ASHRAE Trans. 109 (1) (2003) 7. [16] L. (Leon) Wang, Investigation of the Impact of Building Entrance Air Curtain on Whole Building Energy Use, 2013. [18] L. (Leon) Wang, Z. Zhong, An approach to determine infiltration characteristics of building entrance equipped with air curtains, Energy Build. 75 (Jun. 2014) 312e320.

235

[19] L. (Leon) Wang, Z. Zhong, Whole building annual energy analysis of air curtain performance in commercial building, in: eSim, 2014. [20] ANSYS, ANSYS FLUENT User’s Guide e Release 14.0, 2011. Canonsburg, PA. [22] The Energy Conservatory, Minneapolis Duct Blaster ®, The Energy Conservatory, Minneapolis, 2015. [23] New Wave Research, Solo PIV e Nd: YAG Laser Systems, 2004. [24] Laser Optical CCD and sCMOS Cameras | Dantec Dynamics.” [Online]. Available: http://www.dantecdynamics.com/ccd-and-scmos-cameras. [Accessed 01.12.15]. [25] Sage action Inc, SA/TM BUBBLE GENERATOR MODEL 5 CONSOLE, 2002. ITHACA. [26] F. Scarano, S. Ghaemi, G. Caridi, J. Bosbach, U. Dierksheide, A. Sciacchitano, On the use of helium filled soap bubbles for large-scale Tomographic PIV wind tunnel experiments, in: 17th International Symposium on Applications of Laser Techniques to Fluid Mechanics, 2014, pp. 7e10. [27] J. Bosbach, M. Kühn, C. Wagner, M. Raffel, C. Resagk, R. Puits, A. Thess, in: Large Scale Particle Image Velocimetry of Natural and Mixed Convection 2. Measurement of Convective Air Flow on Large Scales, 2006, pp. 26e29. [28] Dantec Dynamics, Nova Instruments, DynamicStudio Imaging Platform, 2012. [29] X. Cao, J. Liu, N. Jiang, Q. Chen, Particle image velocimetry measurement of indoor airflow field: a review of the technologies and applications, Energy Build. 69 (2014) 367e380. [30] A. Li, E. Qin, B. Xin, G. Wang, J. Wang, Experimental analysis on the air distribution of powerhouse of Hohhot hydropower station with 2D-PIV, Energy Convers. Manag. 51 (1) (2010) 33e41. €nen, Particle Image [31] G. Cao, M. Sivukari, J. Kurnitski, M. Ruponen, O. Seppa Velocimetry (PIV) application in the measurement of indoor air distribution by an active chilled beam, Build. Environ. 45 (9) (2010) 1932e1940. [32] D.P. Hart, PIV error correction, Exp. Fluids 29 (1) (2000) 13e22.