key: cord-0720012-iiwcyywc
authors: Han, Mengtao; Ooka, Ryozo; Kikumoto, Hideki; Oh, Wonseok; Bu, Yunchen; Hu, Shuyuan
title: Measurements of exhaled airflow velocity through human coughs using particle image velocimetry
date: 2021-06-09
journal: Build Environ
DOI: 10.1016/j.buildenv.2021.108020
sha: 54be2200948721cb8ca2f4fe828bacc93e812e65
doc_id: 720012
cord_uid: iiwcyywc

The sudden outbreak of coronavirus (COVID-19) has infected over 100 million people and led to over two million deaths (data in January 2021), posing a significant threat to global human health. As a potential carrier of the novel coronavirus, the exhaled airflow of infected individuals through coughs is significant in virus transmission. The research of detailed airflow characteristics and velocity distributions is insufficient because most previous studies utilize particle image velocimetry (PIV) with low frequency. This study measured the airflow velocity of human coughs in a chamber using PIV with high frequency (interval: 1/2986 s) to provide a detailed validation database for droplet propagation CFD simulation. Sixty cough cases for ten young healthy nonsmoking volunteers (five males and five females) were analyzed. Ensemble-average operations were conducted to eliminate individual variations. Vertical and horizontal velocity distributions were measured around the mouth area. Overall cough characteristics such as cough duration time (CDT), peak velocity time (PVT), maximum velocities, and cough spread angle were obtained. The CDT of the cough airflow was 520–560 m s, while PVT was 20 m s. The male/female averaged maximum velocities were 15.2/13.1 m/s. The average vertical/horizontal cough spread angle was 15.3°/13.3° for males and 15.6°/14.2° for females. In addition, the spatial and temporal distributions of ensemble-averaged velocity profiles were obtained in the vertical and horizontal directions. The experimental data can provide a detailed validation database the basis for further study on the influence of cough airflow on virus transmission using computational fluid dynamic simulations.

One is that the detailed distribution of airflow velocity and its temporal and spatial variations need 113 to be further investigated, which is very important to the deep understanding of mechanisms of 114 cough airflow and droplets. Among the above studies, only VanSciver et al. [32] reported the 115 velocity profiles, but the profiles were aligned according to peak velocity position and lost the 116 spatial distribution information. It is challenging to utilize the limited data to provide a validation 117 database and boundary conditions for cough CFD simulation. The other one is that the frequency 118 of PIV was low due to the limitation of equipment, leading to the missing of several airflow 119 characteristics, which relate to a very small time scale. It is essential to employ a high PIV 120 frequency to capture some detailed cough airflow features because several important 121 characteristics were at the time scale of millisecond level. Mahajan et al. [23] reported that the 122 cough air velocity would peak after only a few milliseconds. The temporal and spatial resolution 123 will also be rough when utilizing low frequency, and fine turbulence structure caused by large 124 initial velocity near the mouth cannot be captured. The previous studies' PIV interval was too large 125 Therefore, this study aims to measure the initial velocity of cough airflow using a higher PIV 133 frequency and obtain detailed velocity distribution data. This study is the first and essential part, 134 which provides the detailed validation database and probable boundary conditions for the 135 subsequent CFD simulation of droplet propagation in the building environment. To simulate a 136 complete unsteady cough airflow process, the time series data of initial magnitude and directions 137 of cough airflow exhaled from the mouth are necessary. Hence, the essential data for boundary 138 conditions of CFD simulations include the temporal and spatial distribution of initial velocity, 139 cough spread angle, and cough duration time. Meanwhile, to validate whether the simulation has 140 accurately described the cough airflow features, accurate cough characteristic data (e.g., maximum 141 velocity, peak velocity times, peak velocity, and cough duration time ) are also required from the 142 experiment in addition to the velocity distribution. These data have not been reported in detail and 143 thus is the objective of this experiment. A higher frequency PIV is utilized to capture the 144 characteristics of a small time scale (e.g., peak velocity time and cough duration time) and 145 small-scale airflow turbulence (this will affect the accuracy of turbulent velocity) more accurately. 146

Repeated cough cases from ten healthy nonsmoking volunteers were measured, and ensemble 147 average operations were carried out to eliminate individual variations and obtain general 148 representative cough airflow results. Overall cough airflow characteristics, including peak velocity 149 time, cough duration time, maximum velocity, and cough spread angle, were measured. Spatial To measure the exhaled airflow velocity through human coughs, the equipment was installed in a 156 clean darkroom at the Institute of Industrial Science, University of Tokyo. The room temperature 157 and humidity were constant (24 °C, relative humidity: 40-45%) to imitate the coughing situation 158 in the general and representative indoor environment. The main equipment consisted of a chamber 159

and PIV system. The chamber was constructed using transparent acrylic boards and had 160 dimensions of 0.8 m × 0.5 m × 0.5 m (Fig. 1) . The left panel of the chamber was 161 constructed using opaque boards to prevent subject harm by the laser. A circular opening (diameter: 162 0.05 m) existed in the middle of the panel. The chamber contained stage fog (particle diameter: 1-163 10 μm) and was semi-enclosed (right side remains empty) to assure the air pressure balance during 164 the coughing. During the measurements, the subjects were instructed to sit on an adjustable-height 165 chair in front of the chamber. They could adjust their height to the correct position and place their 166 mouths on the opening. The cough airflow passed through the opening and entered the chamber. 167

The coordinate origin is defined as the position of the mouth center in the middle of the opening. 168

The x-, y-, and z-directions are defined as the streamwise, spanwise, and vertical directions, 169 respectively. 170 171 

For the PIV measurements, stage fog particles were nebulized into the chamber using an oil 174 droplet generator. The parameters of the laser are listed in Table 1 

The laser lens produced a thin laser sheet, and thus the PIV could only detect the two-dimensional 181 velocity on the sheet. Therefore, vertical and horizontal distribution measurements (Fig. 2) were 182 carried out successively to obtain the three-dimensional spatial distribution of the velocity. 

Ten young, healthy nonsmoking adult subjects, including five males and five females, were 204 selected for the experiment (Table 2) . Each subject repeated coughing three times for the vertical 205 plane measurement and three times for the horizontal plane measurement, with a sufficient rest 206 period between coughs. In total, 60 cough cases were measured in the vertical and horizontal 207 measurements. The experiments were conducted from October to December 2020. The 208 experimental procedures were approved by the University of Tokyo Ethic Committee (Approval 209 number: 20-193) and all subjects have provided written informed contents. Fig. 3 shows the 210 experimental scene and a raw image acquired by PIV. Previous researches reported that several 211 cough characteristics vary including peak flow rate of cough airflow, and PVT, due to the 212 physiological differences caused by sex (e.g., larynx size and the cough power) [24, 25] . Therefore, 213

we studied the cough characteristics and velocity distribution of males and females, respectively. 214 215 Firstly, comprehensive insights into the airflow characteristics of the coughing process should be 221 checked. According to our preliminary analysis, the measurement error will be larger at positions 222 farther away from the lens due to the refraction of light and perspective (see Fig. A 

A). In addition, velocities near the opening (e.g., 0 ⁄ = 0.5) were significantly smaller than that 224 at 0 ⁄ = 1.0, which was probably a measurement error due to the insufficient particles near the 225 opening mentioned in Section 2.2, although the supplement tube partly compensated for this 226 phenomenon. Therefore, the position of 0 ⁄ = 2.5 was selected as a representative position 227 because it is located in the middle of the camera's field of view, which is less affected by particle 228 rarefaction and perspective; the accuracy is high. Fig. 4 shows the maximum velocity variation Table 3 . 249 11.8 m/s, which was smaller for females (approximately 10.3 m/s) because the female cough is 251 usually weaker than that of males. The CDT was approximately 500 ms for both males and 252 females. PV is the peak velocity at 0 ⁄ = 2.5, not the maximum velocity for the whole cough. 253 F-tests for PV, PVT, CDT, and maximum velocity between males and females were conducted to 258 validate whether the cough airflows have differences by sex, using the hypothesis of "data 259 between males and females was the same". The results are shown in Table 3 . The p-values varied 260 in the range of 0.16-0.4, implying that the hypothesis was difficult to reject. However, the p-values 261

were also not large enough to prove that the velocity between males and females was the same. In 262 addition, PV and maximum velocity also show differences by sex. Therefore, we consider that the 263 cough airflows varied by sex to a certain extent, and the results will be analyzed by separating 264 males and females in this study. 265 266 The averaged velocity curves for males and females were generally identical, which indicates that 325 the male and female airflow characteristics were similar. The averaged velocity partially collapsed; 326 for example, the ensemble-averaged PV at 0 ⁄ = 2.5 was smaller than one, because the PVs of 327 different cases occurred at different heights owing to the individual differences. 

In addition to the spatial distribution, we also analyzed the velocity variation with time. As stated 336

in Section 3.1, velocity at 0 ⁄ = 0.5 were underestimated due to the insufficient particles near 337 the opening. In addition, the velocity at 0 ⁄ = 3.5 were also small and it was difficult to 338 distinguish meaningful profiles, which probably due to the natural decay of the airflow. Therefore, 339 this section analyzes the vertical and horizontal temporal distributions of velocities at the positions 340 of 0 ⁄ = 1.0 2.0 2.5 and 3.0, as shown in Fig. 9 and Fig. 10 . Same as in the last section, the 341 velocity in the vertical distribution is 〈√ 2 + 2 〉 , while the velocity in the horizontal 342 distribution is 〈 √ 2 + 2 〉.

At all positions, the velocity initially increased rapidly, and then decreased slowly after it peaked. 345

This tendency well agrees with that shown in Fig. 6 . In particular, at 

The horizontal SD curves varied more sharply than the vertical curves. This probably implies that 360 the measured velocity variation in each case is more violent in the horizontal plane, probably due 361 to measurement errors in the horizontal measurements. One possible reason is the particles' 362 vertical settlement, which led to the uneven distribution of particles or attachment to the bottom 363 board and caused measurement errors. 364 not be neglected (e.g., heads' sway during coughing). Fig. 11 also indicates the decaying tendency 378 of the velocity magnitude with the distance, indicating that the airflow energy was damped with 379 the increase in travel distance due to air friction. 380 381 

The initial cough airflow from the mouth behaves as a jet flow. Therefore, the airflow out of the 383 mouth is usually not completely parallel, but there is a spread angle at the edge of the mouth. The 384 vertical and horizontal cough spread angles are discussed in this section.

The key task to obtain the spread angle is to determine the upper/lower and left/right boundaries of 387 the cough airflow. The previous study [25] determines the boundary by directly checking the raw 388 particle image and drawing the boundary line. This method is intuitive, simple, but obviously with 389 large errors. This study introduced a more quantitative method to determine the airflow boundaries. 390

Considering the jet characteristics of coughing flow, the cough airflow boundary was defined as 391 the position where the velocity decayed to 1% of the maximum value at the airflow center (i.e., 392 velocity is reduced by 99%) in every vertical or horizontal direction. This definition referenced the 393 velocity boundary layer in fluid dynamics. The ordinary least squares fitting line of all boundary 394 points was regarded as the airflow boundary's edge (Fig. 12) . 395 respectively. The difference between averaged vertical and horizontal spread angles was less than 405 0.3°. This is very small; therefore, the initial cough airflow exhaled from the mouth was promising 406 to be modeled as a cone, which should be confirmed in the future. 407 408 Table 4 shows the averaged and compared with Gupta et al. [25] , in which a large 409 discrepancy was found. In addition to the differences in the methods of determining the airflow 410 boundary mentioned above, several experimental conditions between the current study and provide boundary conditions and an accuracy validation basis for CFD validation to model cough 425 airflows for the next step (Table 5) . With these data, it is capable of simulating the droplets 426 propagation via cough airflow using CFD. The simulation can help determine the propagation 427 features of droplets (e.g., propagation distances, propagation routes, and concentration 428 distributions), which is essential for the design and research of building space sensitive to social 429 distance hospitals and sanatorium. Furthermore, the simulation is invaluable for evaluating the 430 infection risk in the population in the building environment, which will be further discussed in the 431 next step.

However, some inadequacies of this experiment should be noticed, which limits the utilization of 434 these data to some extent. Firstly, the experimental subjects were healthy young subjects. 435

Therefore, it should be prudent if utilizing these data in modeling the cough airflow from infected 436 subjects or elderly people, because the cough characteristics may change to the difference in the 437 physical condition. In addition, as we reported, the velocity horizontal distribution was noised, and 438 its accuracy may not as high as the vertical distribution, which we should also pay attention to. 439 440 Table 5 Usage of experimental data for CFD boundary condition and simulation validation

Available data of this experiment CFD boundary condition Airflow outlet geometry Mouth width (Table 2) Velocity duration time CDT (Table 3) Velocity magnitude Temporal velocity distribution around mouth (Table 3) , Maximum velocity (Table 3) Temporal velocity variation PVT (Table 3) , Maximum velocity variation with time ( Fig.6) Airflow velocity distribution Spatial velocity distribution (Fig.8) ,

Temporal velocity distribution (Fig.9, 10) 

This study utilized a PIV with the interval of 1/2986 s system to measure the overall 443 characteristics and initial velocity of the exhaled airflow of coughs by five males and five females. 444

The experiment for each person was repeated six times to measure the vertical and horizontal 445 velocity distributions. The measurement results were ensemble-averaged according to sex.

The maximum coughing velocity for females was weaker than that for males, but their PVT or 448

CDT was similar, implying that their cough duration time was almost the same. In addition, the 449 non-dimensional velocity distribution for males and females agreed well in both vertical and 450 horizontal directions, indicating that their cough airflows own similar characteristics. The 451 temporal variations of maximum velocity at 0 ⁄ = 2.5 can be defined as a combination of 452 gamma-probability-distribution functions. The cough spread angles were analyzed by determining 453 the cough airflow boundaries as the position where the velocity decayed to 1% of the maximum 454 value at the airflow center in every vertical or horizontal direction. The vertical and horizontal 455 spread angles were similar, implying that the initial cough airflow exhaled from the mouth was 456 promising to be modeled as a cone. This should be further studied. The cough spread angles of 457 males and females were also almost the same.

However, several limitations of this study should be noticed. The first is the situation of the 460 subjects (young and healthy) we mentioned obviously, which may limit these data in modeling the 461 cough airflows from the infected subjects (e.g., COVID-19 infected patients) or elderly people. In 462 addition, more objects with different physical conditions should be measured to explore the 463 relationship between airflow characteristics and physical indexes (e.g., ages, weight, and height). 464

Another point is that the study of cough airflow cannot replace the research of pathological 465

organizations. This is because pathogenic organisms exhaled from a cough may act as a different The PIV accuracy was analyzed using an air spray and hotwire anemometer. The nozzle of the air 473 spray was arranged at the opening to imitate the human mouth and continuously provided a stable 474 jet flow. The 10-s-time-averaged velocity magnitude was measured using a hotwire anemometer at 475 several positions along the jet flow and was regarded as the benchmark data. The chamber was 476 then filled with fog and PIV was used to measure the velocity field of the jet flow again. Fig. A-1  477 shows that PIV significantly under-estimated the velocity, because the jet flow blew the fog 478 particles away, so that the measurement error occurred because of the insufficient concentration of 479 particles. After the addition of the circular particle supplement tube (Fig. A-2 ) at the opening, the 480 accuracy of the PIV significantly improved and was close to that of the hotwire anemometer. 481 482 Appendix B

485 Table A shows the raw PV, PVT, CDT and maximum cough velocity values of 60 cough cases. In 486 the Case names, the first alphabet indicates the subjects, in which "A-E" represents five males, 487 "a-e" represents five females. The second alphabet shows the case was measured in the vertical (V) 488 or horizontal (H) directions. The third number means the experiment case number. Therefore, 489 B-H2 represents the second horizontal measurement of the male subject B. 490 491 Table A Variation range 

WHO Department of Communicable Disease Surveillance and Response, WHO guidelines for 499 the global surveillance of severe acute respiratory syndrome ( SARS )

World Health Organisation, Pandemic (H1N1) 2009 -update 112

World Health Organization, Case definition for reporting to WHO Middle East respiratory 503 syndrome coronavirus Interim case definition

Study on the initial 507 velocity distribution of exhaled air from coughing and speaking

The role of particle size in aerosolised 510 pathogen transmission: A review

Measurement of Sanitary Ventilation

Chronic Cough: The Spectrum and Frequency of Causes

Key Components of the Diagnostic Evaluation, and Outcome of Specific Therapy

Is mis-swallowing or smoking a cause of 518 respiratory symptoms in patients with gastroesophageal reflux disease?

Effects of anti-reflux surgery on chronic cough and asthma in 521 patients with gastro-oesophageal reflux disease

Short-range airborne route dominates exposure of 526 respiratory infection during close contact

Evaporation and dispersion of respiratory droplets from coughing

Thermal effect of human body on cough droplets evaporation and 531 dispersion in an enclosed space

Numerical investigation of indoor particulate contaminant transport using 534 the Eulerian-Eulerian and Eulerian-Lagrangian two-phase flow models

Distribution of droplet aerosols generated 537 by mouth coughing and nose breathing in an air-conditioned room

Characterization of expiration air jets 541 and droplet size distributions immediately at the mouth opening

How far droplets can move in indoor 544 environments -revisiting the Wells evaporation-falling curve

Modality of human expired aerosol size 548 distributions

Enhanced spread of expiratory droplets by turbulence in a cough jet

Numerical Simulation of Coughed Droplets in 552 Conference Room

Establishment and 556 clinical applications of a portable system for capturing influenza viruses released through 557 coughing

Viable 560 influenza a virus in airborne particles from human coughs

Relationship between expired lung volume, 563 peak flow rate and peak velocity time during a voluntary cough manoeuvre

Relationship of peak flow rate and peak 566 velocity time during voluntary coughing

Flow dynamics and characterization of a cough

Cough peak flow rate

Flow Rate Estimation Based on GA-BP Method

Evaluation of cough using digital particle image velocimetry

Study on transport characteristics of saliva droplets produced by 580 coughing in a calm indoor environment

Investigation into aireborne transport characteristics of airflow due 583 to coughing in a stagnant indoor environment

On Particle Image Velocimetry (PIV) 585 measurements in the breathing zone of a thermal breathing manikin

Particle image velocimetry of human cough

Preliminary prediction of flow and particulate 590 concentration produced from normal human cough dispersion

Numerical study of the transport of droplets or particles generated by 593 respiratory system indoors

The motion of respiratory droplets 596 produced by coughing

Experimental 598 investigation of far-field human cough airflows from healthy and influenza-infected subjects

Short-range bioaerosol deposition and recovery of viable 601 viruses and bacteria on surfaces from a cough and implications for respiratory disease 602 transmission

The snot-spattered experiments that show how far sneezes really spread

The authors would like to thank all the students in the Ooka's lab and Kikumoto's lab of the 494 Institute of Industrial Science, University of Toky, who have made selfless and important 495 contributions to this experiment. 496 497