ARTICLE IN PRESS

0360-1323/$ - se

doi:10.1016/j.bu

�CorrespondE-mail addr

(Y. Zhang), sun

Building and Environment 44 (2009) 34–43

www.elsevier.com/locate/buildenv

Experimental and CFD study of unsteady airborne pollutant transportwithin an aircraft cabin mock-up

Wei Yan, Yuanhui Zhang�, Yigang Sun, Dongning Li

Department of Agricultural and Biological Engineering, University of Illinois at Urbana-Champaign, 332B 1304 Pennsylvania Avenue,

Urbana, IL 61801, USA

Received 15 November 2007; received in revised form 22 January 2008; accepted 23 January 2008

Abstract

It has been documented that diseases can spread within an aircraft cabin from the sneezing, coughing or breathing of a sick passenger.

To understand the spreading mechanism it is very important to quantify the airflow and droplet transmission around a sneezing/

coughing incident. In this project, tracer gas experiments were carried out in a full-scale Boeing 767-300 mock-up to study the global

transport process of contaminated air within the cabin. Computational fluid dynamics (CFD) simulation was also used to provide

additional information for understanding the principle. A steady airflow field was simulated first and then it was compared with the

experimental data. The global airflow patterns were similar to those observed experimentally. This velocity field was adopted as the initial

condition for further unsteady pollutant transport simulation. Experimental and simulated results were compared and discussed to

develop a relationship between concentration and airflow pattern, source location, transport direction, and ventilation rate. Finally, the

overall picture of concentration evolution by both experimental and simulated approaches was discussed.

r 2008 Elsevier Ltd. All rights reserved.

Keywords: Indoor air quality; Computational fluid dynamics; Tracer gas; Aircraft cabin; Ventilation

1. Introduction

Every year there are more than one billion peopletraveling by air throughout the world with more than 600million passengers traveling in the US [1,2]. Complaintsabout cabin environments are found to be very common[3,4]. ‘‘Sick airplane syndrome’’ is a phrase referring tohealth problems in aircraft cabins, which is similar to the‘‘sick building syndrome.’’

Many complaints could be explained by thermal andhumidity discomfort and indoor air pollutant. Accordingto previous research [5], sources of pollutants in the cabinair during normal operations include: (i) ground-level airpollution at the departure or arrival airport, (ii) contami-nants of outdoor origin at cruise altitudes (e.g. strato-spheric O3), (iii) human bioeffluents (e.g. acetone, CO2,ethanol, and odors), (iv) environmental tobacco smoke,

e front matter r 2008 Elsevier Ltd. All rights reserved.

ildenv.2008.01.010

ing author. Tel.: +1217 333 2693; fax: +1 217 244 0323.

esses: weiyan2@uiuc.edu (W. Yan), yzhang1@uiuc.edu

4@uiuc.edu (Y. Sun), dli3@uiuc.edu (D. Li).

(v) chemicals in cleaning supplies, and (vi) spraying pesti-cides on certain international flights. It is believed thatthese pollutants have negative effects on a passenger’s health.The target pollutant in this paper is droplets and

particles from the sneezing and coughing of a sickpassenger. Usually these droplets or nuclei contain micro-organisms such as viruses, which transport within the cabinand propagate disease. It is assumed that the risk ofexposure is associated with flight time and the distancebetween source and receptor. According to previousdata, the exposure risk is high when the traveling time islonger than 8 h and the distance is within two rows aroundthe sick passenger. This could be explained by the factthat little longitudinal airflow was observed in aircraftcabins [6].However, the ‘‘two rows’’ transport rule is not reliable

because the aerodynamic character of a nuclei carryingvirus is very complicated. Some micro-organisms rely oninjection and space encroachment during the releaseprocess, while others follow airflow. For example, duringthe SARS outbreak in Hong Kong, there were 22 people

ARTICLE IN PRESS

Fig. 1. Full-scale aircraft cabin experimental setup: (a) front view of

experimental setup and (b) sampling point locations.

W. Yan et al. / Building and Environment 44 (2009) 34–43 35

infected in a 3 h fight carrying 120 passengers from HongKong to Beijing. Among those infected, only half of theinfected passengers were seated within two rows of thesick person [7] while others were beyond the two-rowlimitation. This indicates that the virus may follow airflowand travel far beyond two neighboring rows. However,currently there is still no reliable experimental data to studythis transport mechanism.

Experiments and computational fluid dynamics (CFD)simulations are the two most effective approaches used toinvestigate indoor air pollutant transportation. Experi-ments are usually done in a full-scale mock-up section of anaircraft cabin. The technologies used to evaluate ventila-tion effectiveness and airflow pattern include tracer gas [8],particle image velocimetry (PIV) [9], stereoscopic particleimage velocimetry (SPIV) [10], volumetric particle trackingvelocimetry (VPTV) [11,12], ultrasonic anemometers [13],hotwire anemometers [14], etc. The expiratory dropletsreleased by passengers span a wide size range. These ‘‘poly-diameter’’ droplets disperse in the cabin via differentmechanisms. Some large droplets behave like airborneparticles while extremely fine droplets behave similar togas. The tracer gas method cannot represent large aerosols,because it cannot mimic the gravity effect, inertial effect,and drag force effect. But it is valid for micro-organismswhen expelled from an infectious person that quicklyevaporates into extremely fine droplet nuclei, which followsthe airflow very well [15]. It has been used to investigate thedispersion of contaminants in indoor environments [16]and risk of exposure in commercial flights [17,18]. VPTV isa unique technology developed to measure 3D velocitywithout intruding into the airflow field. This technologycorrelates particle streaks instead of separated points,which distinguishes it from traditional PIV or SPIV. Therelatively large field of view (up to several meters) makes itsuitable for indoor air quality study.

Compared with experimental methods, CFD can providemore detailed information with lesser expense. The earliestnumerical study for aircraft cabins was conducted byAboosaidi et al. [19]. With the development of IT,capabilities of computers have improved quickly. Increas-ingly realistic cabin geometry and manikins have been usedin the CFD study. Singh et al. [20] used a heated cylinder torepresent a passenger in an aircraft cabin. Lin et al. [21,22]used a more detailed geometry to study airflow andpollutant transport in cabins. Zhang and Chen [23] usedan actual cabin to study novel air distribution systems for acommercial aircraft cabin.

In the current work, both tracer gas experiment andCFD simulation are used to study the airborne pollutanttransport mechanism within an aircraft cabin. The entireunsteady tracer gas dispersion process is monitored bytracer gas experiments to study the accumulation anddischarge of pollutants within cabin. Due to limitations ofthe experiment, only the concentration at the horizontalbreathing zone is recorded. CFD analysis is processedto provide further detailed information, which is not

provided by experimental data. The simulated steadyvelocity field is compared to VPTV experiment data [12].The unsteady pollutant dispersion is simulated using thissteady velocity and zero initial concentration condition.The whole evolution picture is achieved based on theseapproaches.

2. Experiment

A full-scale aircraft cabin mock-up, based on a Boeing767-300, is used for the experiment. An overview of thesystem setup is showed in Fig. 1.Conditioned air is supplied by a long narrow overhead

diffuser and exhausted though the grills installed at thebottom of the sidewalls. As in an actual airplane, the airsupply rate is adjustable. In this experiment, it is set at816m3/h (479.5 cfm), 1052m3/h (618.18 cfm), and 1259m3/h (739.8 cfm), corresponding to 80%, 100%, and 120%of the full ventilation load of a realistic operatingcondition, respectively. These values were recommendedby the Boeing Company.

ARTICLE IN PRESS

Fig. 2. Aircraft cabin geometry and computational grids.

Table 1

Boundary conditions for computational fluid dynamic

Boundary Parameters

Velocity inlet (diffuser) Magnitude: 0.47m/s

Turbulent intensity: 10%

Injection angle: 301

Concentration: 0%(CO2)

Velocity inlet (mouth, 5min) Turbulent intensity: 10%

Injection angle: 01

Concentration: 100%(CO2)

W. Yan et al. / Building and Environment 44 (2009) 34–4336

Air nozzles above each passenger and heat buoyancycan produce local airflows or recirculations, which mayaffect the local pollutant concentration. Previous re-search [12] observed no strong thermal plume from heatsources wrapped in the waist of passengers. For simplifica-tion, heat sources are not considered in the currentexperiment.

Carbon dioxide (CO2) is used as the tracer gas tosimulate airborne pollutants expelled by sick passengers.The gas is released at the breathing zone of a specifiedpassenger at the rate of 4.5 L/min for 5min. The gasconcentration in the mock-up cabin is continuouslymonitored for 1000 s after the tracer gas is released.Fourteen non-dispersive infrared CO2 sensors (Telaire,Santa Barbara, CA) are used to monitor the concentrationhistogram at the breathing zone in front of each manikin(about 1 ft ahead). The accuracy of the sensor is740 ppm+3% of the readings. The sampling frequencyis 0.5Hz. Only two rows could be measured at a time.Therefore, the concentration pattern for the entire mock-up is acquired from three measurements. The backgroundCO2 concentration in the lab is about 300–400 ppm. Thewhole experiment lasted over 2 weeks; therefore it isnecessary to monitor the background concentration beforeeach measurement. In the current study, the ventilationsystem runs for 2 h without releasing CO2 before eachexperiment. The time average outlet concentration is takenas the background value. The typical standard deviation ofthis concentration is less than 2 ppm. This backgroundconcentration is subtracted from the time-dependentconcentration to obtain the net value, which is regardedas the real concentration caused by the pollutant sources.

Before use, all the sensors are calibrated using Enviro-nics gas dilution systems and LI-820 CO2.

3. Computational fluid dynamics

A proper turbulence model should be selected accordingto the airflow situation in the cabin and simulationrequirements. Sorensen and Nielsen [24] have a goodrecommendation on how to control the quality of CFD inan IAQ simulation. The standard k–e model is frequentlyused for indoor air quality simulations [25]. According totheir report, the simulation result agrees well with theexperiment data. Encouraged by these researches, thestandard k–e model was selected for the current research.A commercial CFD program, Gambit and Fluent, wasused as a grid generating tool and a numerical solver.Tecplot was used to handle the post-process work. Thesimulation was conducted on a supercomputing system.

Fig. 2 shows the overall geometry and computationalmesh developed based on the prototype. According to thereport of Topp et al. [26], box-shape manikins are sufficientfor global airflow simulations. Therefore such manikinswere adopted in the current research. The mesh is notuniformly generated. The domain close to the manikins ismeshed with fine unstructured grids, while the outer

domain is meshed with structured coarse grids. Theminimum size of the fine grid is 0.02m. The size of thestructured grid is 0.04m. An interior condition is used atthe interface. The reason for this arrangement is that thegeometry near manikins is complex, and the air velocitygradient there is larger compared to the outer domain;therefore, finer mesh is necessary to achieve good accuracy.The total number of grids is 1.57 million and takes about8 h to achieve convergence (processor speed 1.3GHz,interactive system, eight processors, 16GB memory). Forsteady velocity field simulation, the residue of continuity isbelow 1� 10�5. The residue of velocities is below 1� 10�6.For unsteady species concentration, the residue of species isbelow 1� 10�6. The airflow pattern near the inlet diffuseris extremely complicated. It is difficult to give a perfect inletboundary condition. In the current simulation a simplifiedconstant velocity boundary condition was adopted. Thevelocity magnitude is 0.47m/s, which corresponds tothe 100% ventilation rate recommended by Boeing. Theturbulence intensity is set to be 10%. The injection angleof the inlet is 301, which is defined according to theobservation of experiment. The species transport andmixing model provided by Fluent is selected to simulateCO2 transport. CO2 injection rate is 4.5 L/min, and lastsfor 5min. The complete boundary conditions are listed inTable 1.

ARTICLE IN PRESSW. Yan et al. / Building and Environment 44 (2009) 34–43 37

4. Results and discussion

4.1. Airflow pattern

To begin with, a steady airflow pattern is achieved as theinitial condition for the unsteady pollutant transportsimulation. This airflow pattern is compared with theVPTV experimental data [12].

Fig. 3 compares the simulated airflow pattern withthe corresponding experimental result. Because of theobstacles of seats and manikins, only airflow of the upperdomain of the cabin was obtained in the experiments.It is clearly shown that CFD results agree with experi-ments in the sense that the two big vortices were capturedby the simulation, which is clearly shown by the path-linein Fig. 3(b). The other fact is that CFD predicted thevelocity magnitude as being below 0.3m/s, which isconsistent with the experimental result. However, theplume above the middle passenger is not captured bysimulation. This might be explained by the simplifiedmanikin geometry and its boundary conditions in thesimulation.

The tracer gas could be delivered to every corner ofthe cabin in several minutes. Therefore, the ventilationsystem design has tremendous effect on pollutant trans-port and indoor air quality. The simulated airflow fieldprovides complete information for evaluating the pollutant

y (m

)

-2 -1.5 -1 -0.5 0 0.5 1 1.5 20

0.5

1

1.5

2

-0.31m/s

Vzm/s

x (m)

0.30.250.20.150.10.050-0.05-0.1-0.15-0.2-0.25

Fig. 3. Comparison of experimental and simulated airflow pattern: (a)

VPTV experimental velocity vectors and iso-magnitude contours and (b)

simulated velocity vectors and iso-magnitude contours.

Fig. 4. Path-line descriptions of velocity field in the cabin: (a) path-line in

whole volume and (b) path-line in a horizontal plane at the breathing zone

level.

transport. Fig. 4 shows the path-line of airflow in the wholevolume and a horizontal plane.The ventilation system divides the cabin into two

separate domains. A big swirl is generated on each side.Such self-recirculation could help to control the pollutantwithin the releasing side. As showed in Fig. 4(b), long-itudinal airflow is inevitable because of obstacles in thecabin. This may explain why the pollutant could transportbeyond neighboring rows within a short time.

4.2. Tracer gas concentration and releasing source location

Complex geometries (seats and manikins) and airflowpatterns (swirls and longitudinal velocity) are tied together

ARTICLE IN PRESSW. Yan et al. / Building and Environment 44 (2009) 34–4338

and make pollutant propagation in the aircraft cabin verycomplex. The pollutant source location has tremendouseffect on the pollutant transport. Fig. 5 compares theexperiment results on releasing the tracer gas at differentlocations.

These concentration fields are interpolated based on atotal of 35 sampling points, which are located in front ofeach manikin’s breathing zone. Each intersection in thefigures represents one sampling point.

Obviously, the pollutant transport is not isotropic. It isobserved that pollutants released at locations close to thesidewalls are exhausted out of the cabin faster than thosereleased from the sources distant from the walls. This isproven by examining Fig. 5(c), which is the result ofreleasing CO2 from 2B and 2G concurrently. Seat 2B islocated in the B column and 2G is located in the G column.Compared to 2B, 2G is closer to the exhaust grills.Therefore, the tracer gas released by 2B transport fartherthan those released by 2G, and the concentration of the leftdomain is higher than that of the right domain. Releasingat 3D is the worst case because it is difficult to exhaust thepollutant out.

Figs. 5(a) and (b) are comparable, given that both seatsare near the aisle, and have the same rate of release.

Fig. 5. Experimental iso-concentration contours at the breathing zone level for

2B and 2G, and (d) release at 2B (t ¼ 300 s).

However, the experiment result is quite different. Theoverall concentration field in the 2B case is higher than inthe 4F case. Longitudinal airflow and outlet locationcontributed to this difference. Because the ventilationoutlet is near 5A and 5G, the pollutant released at 4F iseasily exhausted out. This is the reason why the tracer gasreleased from 2B transports farther than that from 4F.Fig. 6 shows the simulated concentration when the

tracer gas-releasing source is located at 2B. A highconcentration region is predicted at 3B from the simula-tion. Fig. 6(b) explains why this peak value occurs. Thepollutant-releasing jet curves because of the airflow andgenerates an upward pollutant flow toward the centerregion of the cabin. However, comparing Figs. 5(a) and6(a), this region is not clearly indicated by the result of theexperiment.One possible reason to explain the difference is that theexperimental sampling points are too coarse to describethis concentration gradient. There are only three samplingpoints (2B, 2C, and 2D) located in this region. In addition,the sensors are installed about 1 ft in front of the releasingsource. Therefore, the peak value is not captured by theCO2 sensor. The interpolated experiment concentrationcontour will underestimate the overall concentration field.

different source locations: (a) release at 2B, (b) release at 4F, (c) release at

ARTICLE IN PRESS

Fig. 6. Simulated iso-concentration contours when released at 2B: (a)

concentration field in the horizontal plane at the breathing level and (b)

concentration field in the vertical plane at the releasing source location.

W. Yan et al. / Building and Environment 44 (2009) 34–43 39

4.3. Pollutant transport in latitudinal and

longitudinal direction

The distance between source and receptor will affectboth the peak pollutant concentration value and the peakvalue delay time. There is a non-linear relationship betweenthe distance and the concentration-changing rate due to thecomplex airflow pattern in the aircraft cabin. Fig. 7compares the concentration histogram along time indifferent latitudinal directions. It is observed from Figs.7(a) and (b) that the concentration on the pollutant sourceside is strongly associated with the distance between thereleasing source and the receptors. The tracer gasconcentration at the B column is the highest, with thepeak value at about 700 ppm. The peak value reduces toabout 600 ppm in columns A and C. However, the effect of

distance is not obvious on the side opposite to the releasingsource. The tracer gas concentration curves in columnsE–G are significantly lower than in A–C, with a peakvalue delay of about 100 s. They overlap with each other(shown in Figs. 7(a)–(d)) and are not easily distinguished.Therefore, the effect of pollutant source location does notimpose any strong effect on the seats in the opposite half ofthe cabin. The reason for this is that the airflow swirlcontrols the released pollutant well. Separate zones arecreated, which is consistent with the prediction of theairflow pattern analysis. Comparing Figs. 7(a)–(d), it isobserved that the concentration gradient decreases in thelongitudinal direction, which can be explained by thedominance of mixing effect in this region.Fig. 8 compares the concentration in the longitudinal

direction. Based on the results shown in Figs. 8(a) and (b),the concentration reduced as the distance from thereleasing source increased. The concentration differencebetween the releasing row and the next row is very small, itis less than 50 ppm in 8(a) and (b). This can be explained bythe fact that a longitudinal airflow exists and contributes tothe tracer gas transport. However, it is not comparablewith the mixing effect at locations distant from thereleasing source. This is proven by the fact that theconcentrations in rows 4 and 5 are significantly reducedcompared to that in row 3. In the opposite domain of thereleasing source, the concentration is very low. Thedistance between the source and the receptor contributeslittle to the concentration difference, which is shown inFigs. 8(c) and (d).

4.4. Tracer gas concentration and ventilation rate

A higher ventilation rate generates higher air velocity inthe aircraft cabin. Under a high ventilation rate, air isrefreshed more effectively and better air quality should beexpected. The airflow swirls produced by the air supplydiffuser are also strengthened. The tracer gas is dilutedand mixes more effectively, which will lead to a lowerconcentration.Fig. 9 compares the concentration history in row 3

when the tracer gas is released at 2B. Clearly, a higherventilation rate results in a lower tracer gas concentration.A significant fact is that increasing the ventilation rate from100% to 120% results in little air quality improvement at3A and 3B, which are located near the pollutant source.However, the air quality improvement at 3C and 3D wassignificant.

4.5. Concentration evolution

The evolutions of tracer gas concentration when releasedat 2B and 2G simultaneously are shown in Figs. 10 and 11.Fig. 10 shows the concentration evolution in the horizontalplane based on the experiment results, while Fig. 11(the evolution in the vertical plane at the second row) isobtained through CFD.

ARTICLE IN PRESS

Fig. 7. Concentration histogram with time at different latitudinal directions: (a) row 2, (b) row 3, (c) row 4, and (d) row 5.

W. Yan et al. / Building and Environment 44 (2009) 34–4340

As discussed in the previous section, Fig. 10 indicates thepollutant source location has little effect in the oppositeregion. The principle is shown in Fig. 11. The pollutantfollows the airflow swirls exactly, and forms a self-recirculation within the releasing source region. In general,the air quality in the center of the aircraft cabin is theworst.

As shown in Fig. 10, the tracer gas concentration inthe right domain is significantly lower than in the leftdomain. 2G is closer to the exhaust grills than 2B, and thereleased tracer gas is exhausted directly out of the cabin. Itis obvious that the concentration in the right bottom corneris much higher than that in the opposite side. Thedownward airflow contributes directly to this exhausting.

5. Conclusion

This paper presents a study on the airborne pollutanttransport within an aircraft cabin mock-up with bothexperiment and CFD approaches. The results show:

1.

airborne pollutants are controlled significantly by theairflow due to the ventilation system, which divides theairflow into two halves. Most pollutants are transportedonly within the releasing half and seldom across themiddle line,

2.

the source location significantly affects the pollutanttransport within an aircraft cabin. Releasing at pointsclose to the outlet will cause less air quality problemsthan releasing in the middle region,

3.

longitudinal airflow was observed in the experiment.This may be due to the complex geometry, furniture,and passengers. This longitudinal airflow contributes tothe pollutant transport in a longitudinal direction,

4.

arbitrarily increasing ventilation rates may not necessa-rily benefit the receptor close to the pollutant source.

This study also demonstrates how CFD simulations canhelp to understand the pollutant transport mechanism.They provide more detailed information, which cannot beobtained by experiments exclusively.Virus transport after coughing or sneezing can be

described in two steps. The first step is the intrudingmulti-phase jet, which will seriously affect the local micro-environment. The second step is the global environmenteffect due to the ventilation system airflow field. The pre-sent research revealed only the transport mechanism ofthe second step. Further studies are required to examinewhat occurs during the first step. Although people havedone some work based on CFD simulation, reliableexperiment data is not currently available and is urgentlyneeded.

ARTICLE IN PRESS

Fig. 8. Concentration histogram with time at different longitudinal directions: (a) column B, (b) column C, (c) column E, and (d) column F.

Fig. 9. Concentration histogram with time at different points and different ventilation rates: (a) sampling at 3A, (b) sampling at 3B, (c) sampling at 3C,

and (d) sampling at 3D.

W. Yan et al. / Building and Environment 44 (2009) 34–43 41

ARTICLE IN PRESS

Fig. 10. Experimental iso-concentration contour evolution at the breathing zone level when released at 2B and 2G: (a) t ¼ 120 s, (b) t ¼ 240 s, (c) t ¼ 300 s,

and (d) t ¼ 420 s.

Fig. 11. Simulated iso-concentration contour evolution in the vertical plane when released at 2B and 2G: (a) t ¼ 120 s, (b) t ¼ 240 s, (c) t ¼ 300 s, and

(d) t ¼ 420 s.

W. Yan et al. / Building and Environment 44 (2009) 34–4342

ARTICLE IN PRESSW. Yan et al. / Building and Environment 44 (2009) 34–43 43

Acknowledgements

The authors sincerely acknowledge the National Instituteof Occupational Safety and Health (NIOSH) and the IllinoisCouncil for Agricultural and Food Research (C-FAR) fortheir support, and the National Center for SupercomputingApplications (NCSA) for providing supercomputing facil-ities for this project.

References

[1] Owe JO. Air transportation. In: Brune D, Gerhardsson G, Crockford

GW, Norback D, editors. Major industries and occupations. Geneva:

International Labour Office; 1997. p. 107–25.

[2] Thomas SH, Robert B, Homer B, Barry DC, Colin D, Len L.

Assessment of technologies deployed to improve aviation security:

first report. National Materials Advisory Board 1999;46.

[3] Lindgren T, Norback D. Health and perception of cabin air quality

among Swedish commercial airline crew. Indoor Air 2005;15

(Suppl. 10):65–72.

[4] Lindgren T, Norback D. Cabin air quality: indoor pollutants and

climate during intercontinental flights with and without tobacco

smoking. Indoor Air 2002;12:263–72.

[5] NRC. The airline cabin environment and the health of passengers

and crew. Washington, DC: National Academy Press; 2002.

[6] Hocking M. Indoor air quality: recommendations relevant to aircraft

passenger cabins. Journal of the American Industrial Hygiene

Association 2002;59:446–54.

[7] Olsen SJ, Chang HL, Cheung TY. Transmission of the severe acute

respiratory syndrome on aircraft. New England Journal of Medicine

2003;349(25):2416–22.

[8] Shaw CY. Application of tracer gas techniques to ventilation and

indoor air quality investigations. Indoor and Building Environment

1993;2:374–82.

[9] Zhu S, Kato S, Yang JH. Study on transport characteristics of saliva

droplets produced by coughing in a calm indoor environment.

Building and Environment 2006;41:1691–702.

[10] Sun Y, Zhang Y. An algorithm of stereoscopic particle image

velocimetry for full-scale room airflow studies. ASHRAE Transac-

tions 2004;110(1):85–90.

[11] Sun Y, Zhang Y, Wang A, Topmiller JL, Bennett J. Experimental

characterization of airflows in aircraft cabins. Part I: experimental

system and measurement procedure. ASHRAE Transactions 2005;

111(2):45–52.

[12] Zhang Y, Sun YG. Experimental characterization of airflows in

aircraft cabins. Report to Centers for Disease Control and Preven-

tion, United States Department of Health and Human Services,

Washington, DC, 2003.

[13] Chou PC, Chiang CM, Chuah YK, Wang WA. Effects of

window position on the air flow distribution in a cross-

ventilated residential bedroom. Indoor and Built Environment 1998;

7:300–7.

[14] Zhou X, Ouyang Q, Lin G, Zhu Y. Impact of dynamic airflow on

human thermal response. Indoor Air 2006;16:348–55.

[15] Seitz T, Mortimer V, Martinez K. A tracer gas evaluation at a

garment manufacturing facility with extensive transmission of

tuberculosis. Applied Occupational Environmental Hygiene 1998;13:

335–42.

[16] Rydock JP, Hermansen O. A tracer method for evaluating

recirculation of pollutant releases in buildings. Journal of the

American Industrial Hygiene Association 2002;63(2):234–8.

[17] Rydock JP. Tracer study of proximity and recirculation effects on

exposure risk in an airliner cabin. Aviation Space Environmental

Medicine 2004;75:168–71.

[18] Wang AJ. Quantifying air distribution, ventilation effectiveness and

airborne pollutant transport in an aircraft cabin mock-up. PhD

thesis, Department of Agricultural and Biological Engineering,

University of Illinois at Urbana-Champaign, 2006.

[19] Aboosaidi F, Warfield M, Choudhury D. Computational fluid

dynamics applications in airplane cabin ventilation system design.

Proceedings of the Society of Automotive Engineers 1991;246:

249–58.

[20] Singh A, Hosni MH, Horstman RH. Numerical simulation of airflow

in an aircraft cabin section. ASHRAE Transactions 2002;108(1):

1005–13.

[21] Lin C, Horstman R, Ahlers M, Sedgwick L, Dunn K, Topmiller J.

Numerical simulation of airflow and airborne pathogen transport in

aircraft cabins—part 1: numerical simulation of the flow field.

ASHRAE Transactions 2005;111(1):755–63.

[22] Lin C, Horstman R, Ahlers M, Sedgwick L, Dunn K, Topmiller J.

Numerical simulation of airflow and airborne pathogen transport in

aircraft cabins—part 2: numerical simulation of airborne pathogen

transport. ASHRAE Transactions 2005;111(1):764–8.

[23] Zhang T, Chen Q. Novel air distribution systems for commercial

aircraft cabins. Building and Environment 2007;42:1675–84.

[24] Sorensen DN, Nielsen PV. Quality control of computational fluid

dynamics in indoor environments. Indoor Air 2003;13:2–17.

[25] Finlayson EU, Gadgil AJ, Thatcher TL, Sextro RG. Pollutant

dispersion in a large indoor space. Part 2: computational fluid

dynamics predictions and comparison with a scale model experiment

for isothermal flow. Indoor Air 2004;14:272–83.

[26] Topp C, Nielsen P, Sorensen D. Application of computer simulated

persons in indoor environmental modeling. ASHRAE Transactions

2002;108(2):1084–9.