International Construction Specialty Conference of the Canadian Society for Civil Engineering (ICSC) (5th : 2015)

Lessons learned from using bio- and environmental sensing in construction : a field implementation Lee, Wonil; Migliaccio, Giovanni C.; Lin, Ken-Yu; Russo, Francesca Jun 30, 2015

Your browser doesn't seem to have a PDF viewer, please download the PDF to view this item.

Item Metadata

Download

Media
52660-Lee_W_et_al_ICSC15_270_Lessons_Revised_12June2015.pdf [ 313.12kB ]
52660-Lee_W_et_al_ICSC15_270_Lessons_Learned_From_slides.pdf [ 2.46MB ]
Metadata
JSON: 52660-1.0076458.json
JSON-LD: 52660-1.0076458-ld.json
RDF/XML (Pretty): 52660-1.0076458-rdf.xml
RDF/JSON: 52660-1.0076458-rdf.json
Turtle: 52660-1.0076458-turtle.txt
N-Triples: 52660-1.0076458-rdf-ntriples.txt
Original Record: 52660-1.0076458-source.json
Full Text
52660-1.0076458-fulltext.txt
Citation
52660-1.0076458.ris

Full Text

5th International/11th Construction Specialty Conference 5e International/11e Conférence spécialisée sur la construction    Vancouver, British Columbia June 8 to June 10, 2015 / 8 juin au 10 juin 2015   270-1 LESSONS LEARNED FROM USING BIO- AND ENVIRONMENTAL SENSING IN CONSTRUCTION: A FIELD IMPLEMENTATION Wonil Lee1,3, Giovanni C. Migliaccio1, Ken-Yu Lin1 and Francesca Russo2 1 Department of Construction Management, University of Washington, Seattle, WA, USA 2 Department of Civil, Architectural and Environmental Engineering, University of Naples Federico II, Italy 3 wonillee@uw.edu  Abstract: Both physiological status and jobsite environmental stressors influence workforce behavior and performance. Understanding these relationships at the individual worker level is paramount for sustainably managing the construction industry workforce. Astonishing improvements in sensing technology can benefit field research by providing ways to validate occupational performance models based on data that measure workers’ physiological variables and environmental stressors. However, only a few studies have taken advantage of these technological improvements to conduct construction field studies. This paper describes a field monitoring study hosted at a mid-rise, mixed-use building construction site in Seattle, WA. This study was valuable in term of its breadth and period of the observations because it used some of the latest off-the-shelf wearable biosensors to collect 339 hours of workers’ biosignal data from five subjects, during summer and fall, for a total of up to three weeks per subject. This research empirically validated that the heart rate is a good predictor of a worker’s physical strain. Descriptive statistics and a time series plot were used to analyze the heart rate pattern as a predictor of worker’s physical strain level. Correlation analysis was used to analyze the association between the workers’ heart rate and jobsite environmental stressors. Also, analyzing video recordings and questionnaires helped interpreting the analytical results. This paper reports the lessons learned and the challenges of implementing a selected combination of wearable biosensor and environmental sensing technologies. These research findings are preparatory to validating a demand and capability model to be used for predicting construction workers’ performance. 1 INTRODUCTION Identifying the human factors affecting the performance of labor in terms of productivity, quality, and safety is important for sustainably managing workforce in the labor-intensive construction industry. Worker’s low performance is critically related to an unhealthy and overloaded physiological status. In 2010, there were 4,690 fatal occupational accidents in the United States with 802 fatalities being in the construction industry (CPWR 2013). In regard to the workforce employed, the construction industry caused more deaths than any other major industry including transportation, agriculture, retail, and manufacturing. Similarly, the rate of days away from work caused by nonfatal injuries and illnesses is 39% higher for the construction industry than the average in all major industries (CPWR 2013). Accounting for 33% of the total 74,950 non-fatal injuries, bodily reaction and exertion was the largest cause of nonfatal injuries and illness in the U.S. construction industry. Previous studies have identified workers’ accumulated fatigue originating from continuous work activity, and bodily overexertion or repetitive motion as some of the causes of safety accidents and work-related 270-2 musculoskeletal disorders (Everett 1999; Putz-Anderson et al. 1997). Cardiovascular diseases, high blood pressure, and obesity are also major health-risk factors for construction workers (CPWR 2013). Workers’ overexertion and high fatigue are known to be negatively associated with labor productivity and quality of work (Astrand et al. 2003; Bernold and AbouRizk 2010). Overall, increases in physical strain and stress are expected to decrease work quality and productivity (Bernold and AbouRizk 2010; Ringen et al. 1995), and have a negative influence on safe work behavior due to an increasing of distractions and fatigue among workers (Hallowell 2010). Moreover, bodily exertion causes declines in work efficiency, reduces attentiveness, and increases errors (Abdelhamid and Everett 2002). This paper summarizes results and reports lessons learned from a recently-completed field monitoring study that relied on a combination of wearable bio- and environmental sensing technologies. The study aimed at assessing opportunities and challenges in the use of biosensors and tracking devices in a construction site to promote the evaluation and validation of occupational performance models. The study relied on mainstream technology to help collect data related to worker physiological status, activity levels, and jobsite stressors. The adopted technology included a weather station with a wireless data logger, off-the-shelf physiological status monitors (PSM) as the option of biosensors, and global positioning systems (GPS) for tracking workers. 2 DATA COLLECTION AND METHODS 2.1 Data Collection Methods  2.1.1 Biosensor Heart rate (HR) is one of the most important parameters to indirectly measure the physiological demands of workers (Beek and Frings-Dresen 1995; Garet et al. 2005; Takken et al. 2009). Among wearable sensors that measure HR, the Zephyr BioHarness™ 3 was selected for use in our research because it was considered to have high reliability and applicability (Dolezal et al. 2014; Gatti et al. 2011, 2014). Frequent bending and twisting of the waist is often required by construction workers, and for this reason, a sensor mounted on a chest belt, was considered to be more appropriate than a wrist-mounted sensor. Wrist-mounted sensors accurately measure HR of workers when they are resting or performing moderate activities (Terbizan et al. 2002); however, the pulse reading sensors may lose accuracy because of the unexpected fall-out from the wrist during harsh working activities. Also, chest-belt-mounted sensors measure HR similarly to electrocardiogram (ECG) sensors by recording the electrical activity of the muscular tissue of the heart. They are considered to be more reliable than wrist-mounted sensors using photoplethysmogram in physically active states (Schäfer and Vagedes 2013). The selected biosensor provides the functionality to collect HR(bpm), breathing rate (bpm), posture (degrees), activity level (g) and estimated core body temperature (°C). 2.1.2 Environmental Sensor Outdoor temperature is one of the environmental stressors affecting construction workers’ performance. An increase in air humidity also influences sweat evaporation and can increase the heat stress of workers in hot weather. Wind speed also affects the velocity of sweat evaporation, and this may differ depending on the type of clothes worn by the workers (e.g. long-sleeve vs. short-sleeve shirt). If exposed to ultraviolet (UV) rays from the sun for long periods of time, workers may experience dehydration that is one of the variables associated with a worker’s physical strain level. Therefore, the amount of ultraviolet exposure the workers received was measured through UV sensors on a weather station in our research. A wireless weather station (Vantage Pro2TM Plus, Davis Instruments Corp.) was installed in the midsection of the tower crane. The height of installation of the weather station changed from summer to fall data collections to adapt to the building floor level on which the subjects were performing their tasks. 2.1.3 Other Data  Location Tracking: Location tracking was added with the expectation that it could help in performing work sampling to assess how long workers spent in various work areas versus the time they spent travelling 270-3 between working areas. To track the workers’ location, they wore portable global positioning system (GPS) devices throughout the work day. The i-gotU USB GPS Travel & Sports Logger (GT-600) from Mobile Action Technology, Inc. was selected because previous studies found it to be a fairly accurate GPS device for tracking human movement patterns in urban areas (Paz-Soldan et al. 2010; Vazquez-Prokopec et al. 2013). Perceived Fatigue Level and Workers’ Major Tasks Performed: Surveys were administered at the beginning of each break and at the end of the work day to (1) assess each worker’s subjective fatigue level using the Samn‐Perelli Fatigue Checklist (Samn and Perelli 1982), and (2) identify the major tasks each worker had performed between breaks. The Samn-Perelli seven-point fatigue scale (SPS) is a well-established subjective measurement of fatigue. Samn and Perelli (1982) validated the relationship between subjective fatigue levels on a 7-point scale and the performance capabilities of aircraft operators; a higher score indicated the operator felt a higher level of subjective fatigue. The SPS has been used to measure employee fatigue in most transportation industries including aircraft, truck, and rail. Dorrian et al. (2011) used the SPS in rating rail workers’ levels of fatigue to validate a statistical correlation between shift work and fatigue.  Video Recording: Previous studies relied on video recording to capture subjects’ productivity, activity, and behavior. The research participating contractor installed a webcam on an adjacent building, but this camera was only able to capture low-quality videos. Moreover, many critical working areas were located outside of the webcam’s line of sight. After an initial site investigation, the authors deemed it infeasible to install a high resolution site camera in a fixed position, as done in previous laboratory experiments (Cheng et al. 2012, 2013), due to line of sight requirements. Instead, the authors compromised to video recording only small portions of the workday with the plan of using this additional data feed to help interpret the data gather from the wearable devices. Using a digital camcorder (Canon Vixia HF S21), five-minute video observations were recorded three times for each worker: once in the early morning, once in the morning, and once in the afternoon. Video recordings were helpful for interpreting some of the analysis results and identifying the presence of other intervening factors. 2.2 Data Collection Five healthy workers (age range: 27 to 40 years old; height range: 175 to 190 cm; weight range: 84 to 104 kg) were recruited from the selected construction site, a mid-rise mixed-use building project in Seattle, WA. The field observation was approved by the University of Washington Institutional Review Board (IRB). Table 1 includes additional information on these subjects. To guarantee the subjects’ anonymity, specific information such as height, weight, and race are ruled out in the table. Data were collected on workdays usually spanning from 7:00 a.m. to 3:00 p.m.; however, subjects often worked 10-40 minutes of overtime to complete daily assignments or to recover from delayed tasks. Throughout a typical workday, workers started with a short stretch and flex session at 7:00 a.m., and then returned to the trailer to prepare a pre-task plan. At this time, biosensors and GPS devices were provided to subjects to be worn. Workers had two break sessions, one between 10:00 a.m. and 10:15 a.m., and another between 12:00 p.m. and 12:30 p.m. During these two breaks sessions, and again at the end of the workday, short surveys were conducted to assess the workers’ perceived fatigue levels and identify major tasks performed. Data were collected during two different seasons (i.e., hot and cold weather) to increase the variability of environmental stressors. Ideally, the authors would have preferred to collect data in winter (between January and March) that is well representative of the cold weather data to increase the inter-seasonal variability contrasted with the one collected in summer. However, the authors were forced to compromise due to the schedule of the activities supporting the concrete placement that were being performed by the subjects. Because this task was scheduled for completion at the end of October, the data were collected in the following periods: (1) July 29th to August 8th (hot season in Seattle), and (2) October 14th to October 18th (mildly cold season). As workers performed tasks on the ground level in the summer, a weather station was set up at the lowest possible level of the tower crane. As workers worked on the roof in the fall, the weather station was moved and set up at the roof level of the tower crane. While working at the ground level, subjects often performed their activities under a temporary deck; therefore, the weather station data are not fully representative of the environmental stressors on these subjects. 270-4 Table1: Subjects Information Grubbs’ tests were used to detect outliers to be removed from our HR and BR datasets as described in a previous paper (Lee and Migliaccio 2014). Weather data were collected every five minutes, whereas HR and BR data were collected every second. Therefore HR and BR data were calculated as mean values over a time period of five minutes (300 seconds) to perform a correlation analysis.  3 RESULTS 3.1 Heart and Breathing Rates as Predictors of a Worker’s Physical Strain Our data analysis strongly suggests that HR is a more useful parameter in monitoring workers’ physical exertion than BR. For instance, Figure 1 shows HR and BR time series plots for a selected day of the hot season. The plots show clear drops in HR during the break (10:00–10:15 a.m.) and lunch times (12:00–12:30 p.m.), when the workers were resting. The same trend was observed for all five subjects in both seasons, independently from the tasks being performed. On the other hand, the BR data do not provide the same information. This analysis shows how real-time HR data are important to monitor and predict workers’ physical exertion, which is in return associated with workers’ physical strain levels (Bernold and AbouRizk 2010).  Figure1. Comparison of HR and BR Time Series Plots for S.F.1 on July 29th Subject Codes BMI Major Task Study Participation Total Hours of Data Collected    Summer  (Jul. 29 to Aug. 8) Fall (Oct. 14 to Oct. 18)  S.F.1 27.4 Formwork ● ● 120 S.F. 2 25.8 Formwork ● ● 85 F.3 26.9 Formwork  ● 27 S.4 30.4 Concrete Pouring; Cleaning Deck ●  59 S.5 25.1 Layout; Pour Watch ●  48 270-5 3.2 Jobsite Environments  The field weather conditions during the observations are described in Table 2. As expected, higher ambient temperature levels, higher solar radiation levels, and lower relative humidity levels were measured in the summer season than in the fall. The variability of these factors was lower in the fall season than in the summer season. The daily raw data show that the outdoor temperature continuously increased from the beginning to the end of the workday. Conversely, the humidity level decreased throughout the workday. The rainfall data collection failed due to the existence of a hole in the rain collector, which was blocked by wood dust from the construction site. This event suggested that a debris filter should be installed for rain collection in future developments of the research. The amount of solar radiation was generally higher in the summer than in the fall. We did not find differences in wind speed patterns between summer and fall. Table 2: Weather Condition on Jobsite   Summer,Week1 (Jul.29-Aug.2) Summer, Week2 (Aug. 5- Aug. 9) Fall, Week1  (Oct. 14- Oct. 18) Parameter Mean Median Min Max SD Mean Median Min Max SD Mean Median Min Max SD Ambient Temperature (˚F) 61.4 60.9 55.7 70.1 3.5 69.2 69.9 58.8 81.0 5.8 49.9 50.0 42.0 59.4 3.7 Relative Humidity (%) 79.7 81.0 57.0 96.0 9.9 67.2 65.0 37.0 89.0 12.2 89.1 90.0 64.0 98.0 7.4 Wind Speed (mph) 3.9 4.0 1.0 8.0 1.3 4.3 4.0 0.0 11.0 2.1 3.5 4.0 0.0 9.0 2.1 Solar Radiation (W/m2) 344.5 238.0 30.0 923.0 288.3 516.6 643.5 49.0 865.0 312.2 158.4 101.0 0.0 583.0 150.0 3.3 Heart Rate versus Climatic Conditions at the Season Level  Jobsite environmental stressors are expected to affect HR, and so we analyzed our data to this end. To compare the two seasons, we focused our analyses on the subjects who had participated in both summer and fall data collection efforts: S.F.1 and S.F.2. For our analysis, five minute subject-level data were segmented into three sessions per workday depending on the break schedule (i.e., early morning, later morning, afternoon). Moreover, we only included in the analyses those data points corresponding to sessions for which we had data for both the subjects. Finally, we excluded data from August 5th-7th from the data analysis because an unusually heavy workload was performed on these days by subject S.F.2. Therefore, sessions from the following dates were included in the dataset: July 29th-31st, August 1st-2nd, and October 14th-18th. Finally, our dataset included a total of fifteen data points per subject per season (i.e., three sessions multiplied by five days per season). Each data point was represented by a four-dimensional vector that included average, median, minimum and maximum sessional values per subject.  Using this dataset, season-level descriptive statistics were computed (see Table 3). For both the subjects, we found that maximum HR values were higher in fall than in summer. Similarly, we found that average HR values were higher in fall than in summer. To analyze whether this higher average HR in the fall over the summer was statistically significant, a one-sided two-sample t-test was performed. For this analysis, the week 1 summer data were compared with the week 1 data collected in the fall. For the subject S.F.1, the average HR was 115 bpm in the fall season and 109 bpm in the summer season. With 95% confidence, the null hypothesis that the average HR between the two seasons would be the same was rejected (p=0.047, α=0.05 level).  In the case of subject S.F.2, the average HR was 107 bpm in the fall season and 101 bpm in the summer season. With 90% confidence, the null hypothesis was rejected (p=0.061, α= 0.1 level). This result contrasted literature-based expectations that suggested a positive association between HR and outdoor temperature. An analysis of the amount of clothing used by the workers in the two seasons 270-6 can help explain this result. In fact, our subjects wore longer sleeves and thicker clothes during the fall season, which may have restricted their physical movement (see Figure 2). Consequently, this may have caused an increase in their core/skin temperatures, and limited the mobility of their arms and torsos while working, which would have required more worker movement, and therefore increased overall physical exertion. As a result, workers reached higher level of physical strain. This result suggests a significant role of clothing insulation in physical exertion, which should be evaluated in future studies. Table 3: Seasonal Comparison of Heart Rate (bpm) Between Subjects S.F.1 and S.F.2  Subject Codes Summer (Jul. 29 – Aug. 2)  Fall  (Oct. 14- Oct.18)  Two-sample t-test Summer vs. Fall  Mean Median Min Max Mean Median Min Max p-value S.F.1 109.3 110.3 57 156 114.6 118.0 70 174 0.047 S.F.2 101.3 102.4 55 159 107.3 109.4 55 189 0.061   Figure 2: Summer Season Clothing (left) and Fall Season Clothing (right) 3.4 Heart Rate versus Climatic Conditions at Peak  In addition, we analyzed the dataset for peak climatic conditions on August 6th. This day was selected because the daily deviation of the ambient temperature was high and biosignal data from three subjects (S.F.1, S.F.2, and S.4) were collected fully over the eight work hours. S.5 was also present, but the last 3 hours of data for that subject were lost due to malfunctions in the chest belt unit. Subject F.3 did not participate in the summer study because he enrolled only for the fall dates. A Pearson correlation analysis found no correlation between ambient temperatures and HR for subjects S.F.1 and S.F.2, and a positive weak correlation for S.4 (r=0.24, p<0.05) (see Table 4).  Table 4: The Pearson Product-Moment Correlation Coefficient between Ambient Temperature (F˚) and Subject’s Heart Rate (bpm) on August 6th  Subject Codes Pearson-r (r) n p-value (α=0.05) S.F.1 0.11 91 0.320 S.F.2 -0.18 91 0.085 S.4 0.24 91 0.023 Since previous studies have suggested that wet-bulb globe temperature (WBGT) rather than outdoor ambient temperature should be the index used to evaluate workers’ heat stress levels and can be a more useful parameter for predicting workers’ heat strain (OSHA 1999; Rowlinson and Yunyan 2014), we also performed a correlation analysis between WBGT and HR (see Table 5). The outdoor WBGT is estimated 270-7 by the weighted sum of the natural wet-bulb temperature (Tw), the globe temperature (Tg), and the ambient temperature (Ta), as shown in Equation 1. [1] WBGT= 0.7Tw + 0.2Tg + 0.1Ta To estimate the Tw, the dew point temperature, the ambient temperature and atmospheric pressure data collected by the weather station were cross-referenced in the American Society of Heating, Refrigerating, and Air-Conditioning Engineers (ASHRAE) Psychrometric Chart. The Tg calculated by Tonouchi et al. (2001) was estimated from the wind speed and solar radiation as shown in Equation 2 below: [2] Tg= Ta + 0.007×S - 0.208×U  where Tg is the black globe temperature (C˚), Ta is the ambient temperature (C˚), S is a solar ration (W/m2) and U is wind speed (m/s). Since the equation is developed in SI base units, imperial units were converted to SI units to apply this equation. From these processes, the WBGTs were estimated for every five minute period from 7:30 a.m. to 3:00 p.m. on August 8th. Since the WBGT calculation model is empirical, it should be different from the measured WBGT from heat stress monitors incorporating WBGT sensing technology. With this calculated data set, additional Pearson correlation analysis found no statistically significant correlation between WBGT and physical strain measured by HR for subjects S.F.1 and S.F.2, and a positive weak correlation for S.4 (r=0.23, p<0.05), as shown in Table 5, from the uncontrolled field data. Table 5: The Pearson Product-Moment Correlation Coefficient between WBGT (F˚) and Subject’s Heart Rate (bpm) on August 6th  Subject Codes Pearson-r (r) n p-value (α=0.05) S.F.1 0.16 91 0.140 S.F.2 -0.19 91 0.067 S.4 0.23 91 0.027  4 CONCLUSIONS AND LESSONS LEARNED The study narrated in this paper used some of the latest off-the-shelf wearable biosensors to perform an extensive field observation that produced a raw dataset: 339 hours of workers’ biosignal data from five subjects, during two seasons (summer and fall), for a total of up to three weeks per subject. The research empirically validated that HR is a good predictor of a worker’s physical strain. Descriptive statistics and a time-series plot were used to analyze the HR pattern as a predictor of worker’s physical strain levels. Correlation analysis was used to analyze the association between the workers’ HR and jobsite environmental stressors. Whereas no association was observed between the climatic data and HR, the approach initially selected to evaluate environmental stressors may have been flawed. In fact, at the end of the study, the site environmental health and safety (EHS) manager, who performed several daily walk-throughs of the jobsite, informed us that there was a considerable difference in conditions between the location of the weather station in the tower crane and the microenvironment on the deck or in the basement where the workers were working. Ideally, the authors would have preferred to collect data in winter (between January and March) that is well representative of the cold weather data to increase the inter-seasonal variability contrasted with the one collected in summer. However, recent advances in sensor networks may provide a solution to this need for a diffused network of environmental sensors that could be resistant to the harshness of a construction jobsite in the future. The preliminary findings from this extensive field study with five subjects are not sufficient to generalize the statistical outcome to all construction workers in Washington State or the United States because the purpose of this study was to evaluate the use of bio- and environmental sensing technologies for understanding processes affecting construction workers’ performance and health.  270-8 The authors were challenged by various practical issues during this first extensive field study. These issues are shared hereafter for the advantage of other researchers.  Use of GPS Trackers: Whereas the adopted GPS units were found reliable in urban environments by previous studies on tracking general human movements, they did not seem as adequate for tracking construction workers. GPS data lacked reliability due to serious signal disconnections. This issue was attributed to the many barriers (e.g., working under the slab formwork, working in the basement, and using a portable toilet) that obstructed clear GPS satellite communications, as well as artificial errors caused by electrical power lines near the site (see Figure 3). Still, we expect that the same devices may provide satisfactory tracking in open-sky conditions.  Figure 3: Causes of GPS Malfunction Confounding Factors: We observed that some subjects frequently drank energy drinks and smoked during the breaks. These confounding factors may have affected the HR and BR data. Thus, additional survey instruments on the subjects’ dietary and smoking habits should be implemented, and potential compounding variables need to be controlled in the field study.  Fear of Reporting to Supervisors: Based on early conversations during the design of the field study, we perceived construction workers being concerned that the study’s results would be reported to their supervisors at the individual level. This created a sense of resistance for the workers to be monitored through wearing sensors. Therefore, we implemented a careful recruitment process in which a consent form was used to instruct workers about the confidentiality of their data and to assure them that the results of the research would not cause them any disadvantages relative to their supervisors. Still, we adopted a top-down recruitment approach wherein project management helped recruit workers. As a result, a member of the field management team was able to “figure out” who the subjects were and manifested an extreme curiosity in knowing his individual workers’ HR trends. Although the researchers strongly protected the confidentiality of the subjects, this behavior resulted in some residual diffidence by the workers that may have caused some of the dropouts. In the future, we will explore alternative approaches to recruitment that are more bottom-up and use unions and apprenticeship programs as recruitment venues. In addition, we will avoid installing instrumentation in the same trailer used by the field management. Fear of Underperforming against Peers: Insincere reporting probably occurred among subjects in the SPS perceived fatigue survey. For example, a worker was caught looking at other subjects’ survey forms, and another worker asked researchers what levels of fatigue other subjects had reported. These behaviors may indicate attempts to report fatigue levels that did not reflect the actual perceived fatigue. Thus, subjects should complete the surveys away from one another. Trades to be Observed: All the subjects worked for the general contractor on site. As a result, the types of tasks assigned to the subjects presented high variability in scope and target outputs. Some of the inconclusive results may be attributed to this issue. In the future, we plan to perform field observation of subjects belonging to specialty trades as the variability in their scope and target output will be reduced.   Use of Video Recording: While recording short videos of construction activities that could help capture confounding factors, we observed that subjects often became aware of being recorded and changed their 270-9 work behaviors. This was expected and we kept these periods very short. However, positioning several fixed cameras may overcome this issue in the long term.  The cited challenges allowed the authors to outline a roadmap for the successful implementation of bio- and environmental sensors in the construction field. First, it is important to carefully design a study and a managerial strategy to minimize both workers’ resistance to recruitment and insincere reporting of perceived variables. Second, subjects should be recruited from subcontractors involved in the slab framework, roofing, drywall, or masonry trades because these tasks are repetitive and physically intense, require less multitasking, and limit the dimensions of the traveling area, which in return increase the chance of successfully using tracking devices to perform automatic work sampling (Cheng et al. 2013). Third, the use of GPS tracking should be limited to open-sky tasks with minimal barriers presented by built slab framework or roofing. Whereas a need of analyzing indoor activities in established, traditional visual work sampling (Liou and Borcherding 1986) could be applicable to provide an indirect measure of physical exertion and performance. Fourth, cooperation and advice from site superintendents and safety managers are needed to appropriately place environmental sensors for local measurements and cameras that would limit the confounding factors of having a researcher on-site operating these devices. References Abdelhamid, T. S., and Everett, J. G. 2002. Physiological Demands During Construction Work. Journal of Construction Engineering and Management, 128(5): 427-437. Astrand, P., Rodahl, K., Dahl, H. A., and Stromme, S. B. 2003. Textbook of Work Physiology. Human Kinetics, Champaign, IL, USA. Beek, AJ van der, and Frings-Dresen, M. H. W. 1995. Physical Workload of Lorry Drivers: A Comparison of Four Methods of Transport. Ergonomics 38(7): 1508-1520. Bernold, L. E., and AbouRizk, S. M. 2010. Managing Performance in Construction. John Wiley & Sons. CPWR, 2013. The Construction Chart Book-Fifth Edition, The Center for Construction Research and Training (CPWR). Cheng, T., Migliaccio, G. C., Teizer, J., and Gatti,U. C. 2012. Data Fusion of Real-time Location Sensing and Physiological Status Monitoring for Ergonomics Analysis of Construction Workers. Journal of Computing in Civil Engineering 27(3): 320-335. Cheng, T., Teizer, J., Migliaccio, G. C., and Gatti,U. C. 2013. Automated Task-Level Activity Analysis Through Fusion of Real Time Location Sensors and Worker's Thoracic Posture Data. Automation in Construction 29: 24-39. Dorrian, J., Baulk, S. D. Baulk, and Dawson, D. 2011. Work Hours, Workload, Sleep and Fatigue in Australian Rail Industry Employees." Applied Ergonomics 42(2): 202-209. Dolezal, Ba, Boland, D. M., Carney, J., Abrazado , M., Smith D.l., and Cooper C.B. 2014. Validation of Heart Rate Derived from a Physiological Status Monitor-Embedded Compression Shirt against Criterion ECG. Journal of Occupational and Environmental Hygiene 11(12): 833-839. Everett, J. G. 1999. Overexertion Injuries in Construction. Journal of construction engineering and management 125(2): 109-114. Gatti, U. C., Migliaccio, G. C., and Schneider, S. 2011. "Wearable Physiological Status Monitors for Measuring and Evaluating Worker’s Physical Strain: Preliminary Validation." Computing in Civil Engineering (2011): pp. 194-201. Gatti, U. C., Schneider, S., and Migliaccio, G. C. 2014.Physiological Condition Monitoring of Construction Workers. Automation in Construction 44: 227-233. Garet, M., Boudet, G., Montaurier, C., Vermorel, M., Coudert, J., and Chamoux, A. 2005. Estimating Relative Physical Workload Using Heart Rate Monitoring: A Validation by Whole-Body Indirect Calorimetry. European Journal of Applied Physiology, 94(1-2): 46-53. Hallowell, M. R. 2010. Worker Fatigue. Professional Safety, 55(12), 18-26. Retrieved from http://search.proquest.com/docview/763614893?accountid=14784 Lee, W. and Migliaccio, G. 2014. Field Use of Physiological Status Monitoring (PSM) to Identify Construction Workers' Physiologically Acceptable Bounds and Heart Rate Zones. Computing in Civil and Building Engineering (2014): pp. 1037-1044. 270-10 Liou, F. S., and Borcherding, J. D. 1986. Work Sampling Can Predict Unit Rate Productivity. Journal of Construction Engineering and Management 112(1): 90-103. Occupational Safety and Health Administration (OSHA) 1999. OSHA Technical Manual (OTM) Section III: Chapter 4 (TED 01-00-015). Washington D.C., OSHA, Office of Science and Technology Assessment. Accessed January 15, 2015. https://www.osha.gov/dts/osta/otm/otm_iii/otm_iii_4.html.  Putz-Anderson, V., Bernard, B. P., Burt, S. E., Cole, L. L., Fairfield-Estill, C., Fine, L. J., Grant, K. A. et al. 1997.  Musculoskeletal Disorders and Workplace Factors. National Institute for Occupational Safety and Health (NIOSH). Paz-Soldan, V., Stoddard, S.T., Vazquez-Prokopec, G. , Morrison, A. C., Elder, J. P., Kitron, U.,  Kochel ,T. J., and Scott,T. W. 2010. Assessing and Maximizing the Acceptability of GPS Device Use for Studying the Role of Human Movement in Dengue Virus Transmission in Iquitos, Peru. AmJTrop Med Hyg 82 (4): 723-730. Ringen, K., Seegal, J. and England, A. 1995. Safety and Health in the Construction Industry. Annual Review of Public Health 16(1): 165-188. Rowlinson, Steve, and Jia, Y.A. 2014. Application of the Predicted Heat Strain Model in Development of Localized, Threshold-based Heat Stress Management Guidelines for the Construction Industry. Annals of Occupational Hygiene 58(3): 326-339. Samn, S. W., and Perelli, L. P.. 1982. Estimating Aircrew Fatigue: A Technique with Application to Airlift Operations. No. SAM-TR-82-21. SCHOOL OF AEROSPACE MEDICINE BROOKS AFB TX. Schäfer, A., & Vagedes, J. 2013. How Accurate is Pulse Rate Variability as An Estimate of Heart Rate Variability?: A review on Studies Comparing Photoplethysmographic Technology with An Electrocardiogram. International Journal of Cardiology, 166(1): 15-29. Takken, T., Ribbink, A., Heneweer, H., Moolenaar, H. and Wittink, H. 2009. Workload Demand in Police Officers during Mountain Bike Patrols. Ergonomics 52(2): 245-250. Terbizan, D. J., Dolezal, B. A. and Albano, C. 2002. Validity of Seven Commercially Available Heart Rate Monitors. Measurement in Physical Education and Exercise Science 6(4): 243-247. Tonouchi, M., Murayama, K., and Ono, M. 2006. WBGT Forecast for Prevention of Heat Stroke in Japan. In Sixth Symposium on the Urban Environment, American Meteorological Society, JP1(1). Vazquez-Prokopec, G. M., Stoddard, S. T., Paz-Soldan, V. , Morrison, A. C., Elder, J. P., Kochel, T. J., Scott, t. W. and Kitron, U. 2009. Usefulness of Commercially Available GPS Data-Loggers for Tracking Human Movement and Exposure to Dengue Virus. International Journal of Health Geographics 8(1): 68.  5th International/11th Construction Specialty Conference 5e International/11e Conférence spécialisée sur la construction    Vancouver, British Columbia June 8 to June 10, 2015 / 8 juin au 10 juin 2015   270-1 LESSONS LEARNED FROM USING BIO- AND ENVIRONMENTAL SENSING IN CONSTRUCTION: A FIELD IMPLEMENTATION Wonil Lee1,3, Giovanni C. Migliaccio1, Ken-Yu Lin1 and Francesca Russo2 1 Department of Construction Management, University of Washington, Seattle, WA, USA 2 Department of Civil, Architectural and Environmental Engineering, University of Naples Federico II, Italy 3 wonillee@uw.edu  Abstract: Both physiological status and jobsite environmental stressors influence workforce behavior and performance. Understanding these relationships at the individual worker level is paramount for sustainably managing the construction industry workforce. Astonishing improvements in sensing technology can benefit field research by providing ways to validate occupational performance models based on data that measure workers’ physiological variables and environmental stressors. However, only a few studies have taken advantage of these technological improvements to conduct construction field studies. This paper describes a field monitoring study hosted at a mid-rise, mixed-use building construction site in Seattle, WA. This study was valuable in term of its breadth and period of the observations because it used some of the latest off-the-shelf wearable biosensors to collect 339 hours of workers’ biosignal data from five subjects, during summer and fall, for a total of up to three weeks per subject. This research empirically validated that the heart rate is a good predictor of a worker’s physical strain. Descriptive statistics and a time series plot were used to analyze the heart rate pattern as a predictor of worker’s physical strain level. Correlation analysis was used to analyze the association between the workers’ heart rate and jobsite environmental stressors. Also, analyzing video recordings and questionnaires helped interpreting the analytical results. This paper reports the lessons learned and the challenges of implementing a selected combination of wearable biosensor and environmental sensing technologies. These research findings are preparatory to validating a demand and capability model to be used for predicting construction workers’ performance. 1 INTRODUCTION Identifying the human factors affecting the performance of labor in terms of productivity, quality, and safety is important for sustainably managing workforce in the labor-intensive construction industry. Worker’s low performance is critically related to an unhealthy and overloaded physiological status. In 2010, there were 4,690 fatal occupational accidents in the United States with 802 fatalities being in the construction industry (CPWR 2013). In regard to the workforce employed, the construction industry caused more deaths than any other major industry including transportation, agriculture, retail, and manufacturing. Similarly, the rate of days away from work caused by nonfatal injuries and illnesses is 39% higher for the construction industry than the average in all major industries (CPWR 2013). Accounting for 33% of the total 74,950 non-fatal injuries, bodily reaction and exertion was the largest cause of nonfatal injuries and illness in the U.S. construction industry. Previous studies have identified workers’ accumulated fatigue originating from continuous work activity, and bodily overexertion or repetitive motion as some of the causes of safety accidents and work-related 270-2 musculoskeletal disorders (Everett 1999; Putz-Anderson et al. 1997). Cardiovascular diseases, high blood pressure, and obesity are also major health-risk factors for construction workers (CPWR 2013). Workers’ overexertion and high fatigue are known to be negatively associated with labor productivity and quality of work (Astrand et al. 2003; Bernold and AbouRizk 2010). Overall, increases in physical strain and stress are expected to decrease work quality and productivity (Bernold and AbouRizk 2010; Ringen et al. 1995), and have a negative influence on safe work behavior due to an increasing of distractions and fatigue among workers (Hallowell 2010). Moreover, bodily exertion causes declines in work efficiency, reduces attentiveness, and increases errors (Abdelhamid and Everett 2002). This paper summarizes results and reports lessons learned from a recently-completed field monitoring study that relied on a combination of wearable bio- and environmental sensing technologies. The study aimed at assessing opportunities and challenges in the use of biosensors and tracking devices in a construction site to promote the evaluation and validation of occupational performance models. The study relied on mainstream technology to help collect data related to worker physiological status, activity levels, and jobsite stressors. The adopted technology included a weather station with a wireless data logger, off-the-shelf physiological status monitors (PSM) as the option of biosensors, and global positioning systems (GPS) for tracking workers. 2 DATA COLLECTION AND METHODS 2.1 Data Collection Methods  2.1.1 Biosensor Heart rate (HR) is one of the most important parameters to indirectly measure the physiological demands of workers (Beek and Frings-Dresen 1995; Garet et al. 2005; Takken et al. 2009). Among wearable sensors that measure HR, the Zephyr BioHarness™ 3 was selected for use in our research because it was considered to have high reliability and applicability (Dolezal et al. 2014; Gatti et al. 2011, 2014). Frequent bending and twisting of the waist is often required by construction workers, and for this reason, a sensor mounted on a chest belt, was considered to be more appropriate than a wrist-mounted sensor. Wrist-mounted sensors accurately measure HR of workers when they are resting or performing moderate activities (Terbizan et al. 2002); however, the pulse reading sensors may lose accuracy because of the unexpected fall-out from the wrist during harsh working activities. Also, chest-belt-mounted sensors measure HR similarly to electrocardiogram (ECG) sensors by recording the electrical activity of the muscular tissue of the heart. They are considered to be more reliable than wrist-mounted sensors using photoplethysmogram in physically active states (Schäfer and Vagedes 2013). The selected biosensor provides the functionality to collect HR(bpm), breathing rate (bpm), posture (degrees), activity level (g) and estimated core body temperature (°C). 2.1.2 Environmental Sensor Outdoor temperature is one of the environmental stressors affecting construction workers’ performance. An increase in air humidity also influences sweat evaporation and can increase the heat stress of workers in hot weather. Wind speed also affects the velocity of sweat evaporation, and this may differ depending on the type of clothes worn by the workers (e.g. long-sleeve vs. short-sleeve shirt). If exposed to ultraviolet (UV) rays from the sun for long periods of time, workers may experience dehydration that is one of the variables associated with a worker’s physical strain level. Therefore, the amount of ultraviolet exposure the workers received was measured through UV sensors on a weather station in our research. A wireless weather station (Vantage Pro2TM Plus, Davis Instruments Corp.) was installed in the midsection of the tower crane. The height of installation of the weather station changed from summer to fall data collections to adapt to the building floor level on which the subjects were performing their tasks. 2.1.3 Other Data  Location Tracking: Location tracking was added with the expectation that it could help in performing work sampling to assess how long workers spent in various work areas versus the time they spent travelling 270-3 between working areas. To track the workers’ location, they wore portable global positioning system (GPS) devices throughout the work day. The i-gotU USB GPS Travel & Sports Logger (GT-600) from Mobile Action Technology, Inc. was selected because previous studies found it to be a fairly accurate GPS device for tracking human movement patterns in urban areas (Paz-Soldan et al. 2010; Vazquez-Prokopec et al. 2013). Perceived Fatigue Level and Workers’ Major Tasks Performed: Surveys were administered at the beginning of each break and at the end of the work day to (1) assess each worker’s subjective fatigue level using the Samn‐Perelli Fatigue Checklist (Samn and Perelli 1982), and (2) identify the major tasks each worker had performed between breaks. The Samn-Perelli seven-point fatigue scale (SPS) is a well-established subjective measurement of fatigue. Samn and Perelli (1982) validated the relationship between subjective fatigue levels on a 7-point scale and the performance capabilities of aircraft operators; a higher score indicated the operator felt a higher level of subjective fatigue. The SPS has been used to measure employee fatigue in most transportation industries including aircraft, truck, and rail. Dorrian et al. (2011) used the SPS in rating rail workers’ levels of fatigue to validate a statistical correlation between shift work and fatigue.  Video Recording: Previous studies relied on video recording to capture subjects’ productivity, activity, and behavior. The research participating contractor installed a webcam on an adjacent building, but this camera was only able to capture low-quality videos. Moreover, many critical working areas were located outside of the webcam’s line of sight. After an initial site investigation, the authors deemed it infeasible to install a high resolution site camera in a fixed position, as done in previous laboratory experiments (Cheng et al. 2012, 2013), due to line of sight requirements. Instead, the authors compromised to video recording only small portions of the workday with the plan of using this additional data feed to help interpret the data gather from the wearable devices. Using a digital camcorder (Canon Vixia HF S21), five-minute video observations were recorded three times for each worker: once in the early morning, once in the morning, and once in the afternoon. Video recordings were helpful for interpreting some of the analysis results and identifying the presence of other intervening factors. 2.2 Data Collection Five healthy workers (age range: 27 to 40 years old; height range: 175 to 190 cm; weight range: 84 to 104 kg) were recruited from the selected construction site, a mid-rise mixed-use building project in Seattle, WA. The field observation was approved by the University of Washington Institutional Review Board (IRB). Table 1 includes additional information on these subjects. To guarantee the subjects’ anonymity, specific information such as height, weight, and race are ruled out in the table. Data were collected on workdays usually spanning from 7:00 a.m. to 3:00 p.m.; however, subjects often worked 10-40 minutes of overtime to complete daily assignments or to recover from delayed tasks. Throughout a typical workday, workers started with a short stretch and flex session at 7:00 a.m., and then returned to the trailer to prepare a pre-task plan. At this time, biosensors and GPS devices were provided to subjects to be worn. Workers had two break sessions, one between 10:00 a.m. and 10:15 a.m., and another between 12:00 p.m. and 12:30 p.m. During these two breaks sessions, and again at the end of the workday, short surveys were conducted to assess the workers’ perceived fatigue levels and identify major tasks performed. Data were collected during two different seasons (i.e., hot and cold weather) to increase the variability of environmental stressors. Ideally, the authors would have preferred to collect data in winter (between January and March) that is well representative of the cold weather data to increase the inter-seasonal variability contrasted with the one collected in summer. However, the authors were forced to compromise due to the schedule of the activities supporting the concrete placement that were being performed by the subjects. Because this task was scheduled for completion at the end of October, the data were collected in the following periods: (1) July 29th to August 8th (hot season in Seattle), and (2) October 14th to October 18th (mildly cold season). As workers performed tasks on the ground level in the summer, a weather station was set up at the lowest possible level of the tower crane. As workers worked on the roof in the fall, the weather station was moved and set up at the roof level of the tower crane. While working at the ground level, subjects often performed their activities under a temporary deck; therefore, the weather station data are not fully representative of the environmental stressors on these subjects. 270-4 Table1: Subjects Information Grubbs’ tests were used to detect outliers to be removed from our HR and BR datasets as described in a previous paper (Lee and Migliaccio 2014). Weather data were collected every five minutes, whereas HR and BR data were collected every second. Therefore HR and BR data were calculated as mean values over a time period of five minutes (300 seconds) to perform a correlation analysis.  3 RESULTS 3.1 Heart and Breathing Rates as Predictors of a Worker’s Physical Strain Our data analysis strongly suggests that HR is a more useful parameter in monitoring workers’ physical exertion than BR. For instance, Figure 1 shows HR and BR time series plots for a selected day of the hot season. The plots show clear drops in HR during the break (10:00–10:15 a.m.) and lunch times (12:00–12:30 p.m.), when the workers were resting. The same trend was observed for all five subjects in both seasons, independently from the tasks being performed. On the other hand, the BR data do not provide the same information. This analysis shows how real-time HR data are important to monitor and predict workers’ physical exertion, which is in return associated with workers’ physical strain levels (Bernold and AbouRizk 2010).  Figure1. Comparison of HR and BR Time Series Plots for S.F.1 on July 29th Subject Codes BMI Major Task Study Participation Total Hours of Data Collected    Summer  (Jul. 29 to Aug. 8) Fall (Oct. 14 to Oct. 18)  S.F.1 27.4 Formwork ● ● 120 S.F. 2 25.8 Formwork ● ● 85 F.3 26.9 Formwork  ● 27 S.4 30.4 Concrete Pouring; Cleaning Deck ●  59 S.5 25.1 Layout; Pour Watch ●  48 270-5 3.2 Jobsite Environments  The field weather conditions during the observations are described in Table 2. As expected, higher ambient temperature levels, higher solar radiation levels, and lower relative humidity levels were measured in the summer season than in the fall. The variability of these factors was lower in the fall season than in the summer season. The daily raw data show that the outdoor temperature continuously increased from the beginning to the end of the workday. Conversely, the humidity level decreased throughout the workday. The rainfall data collection failed due to the existence of a hole in the rain collector, which was blocked by wood dust from the construction site. This event suggested that a debris filter should be installed for rain collection in future developments of the research. The amount of solar radiation was generally higher in the summer than in the fall. We did not find differences in wind speed patterns between summer and fall. Table 2: Weather Condition on Jobsite   Summer,Week1 (Jul.29-Aug.2) Summer, Week2 (Aug. 5- Aug. 9) Fall, Week1  (Oct. 14- Oct. 18) Parameter Mean Median Min Max SD Mean Median Min Max SD Mean Median Min Max SD Ambient Temperature (˚F) 61.4 60.9 55.7 70.1 3.5 69.2 69.9 58.8 81.0 5.8 49.9 50.0 42.0 59.4 3.7 Relative Humidity (%) 79.7 81.0 57.0 96.0 9.9 67.2 65.0 37.0 89.0 12.2 89.1 90.0 64.0 98.0 7.4 Wind Speed (mph) 3.9 4.0 1.0 8.0 1.3 4.3 4.0 0.0 11.0 2.1 3.5 4.0 0.0 9.0 2.1 Solar Radiation (W/m2) 344.5 238.0 30.0 923.0 288.3 516.6 643.5 49.0 865.0 312.2 158.4 101.0 0.0 583.0 150.0 3.3 Heart Rate versus Climatic Conditions at the Season Level  Jobsite environmental stressors are expected to affect HR, and so we analyzed our data to this end. To compare the two seasons, we focused our analyses on the subjects who had participated in both summer and fall data collection efforts: S.F.1 and S.F.2. For our analysis, five minute subject-level data were segmented into three sessions per workday depending on the break schedule (i.e., early morning, later morning, afternoon). Moreover, we only included in the analyses those data points corresponding to sessions for which we had data for both the subjects. Finally, we excluded data from August 5th-7th from the data analysis because an unusually heavy workload was performed on these days by subject S.F.2. Therefore, sessions from the following dates were included in the dataset: July 29th-31st, August 1st-2nd, and October 14th-18th. Finally, our dataset included a total of fifteen data points per subject per season (i.e., three sessions multiplied by five days per season). Each data point was represented by a four-dimensional vector that included average, median, minimum and maximum sessional values per subject.  Using this dataset, season-level descriptive statistics were computed (see Table 3). For both the subjects, we found that maximum HR values were higher in fall than in summer. Similarly, we found that average HR values were higher in fall than in summer. To analyze whether this higher average HR in the fall over the summer was statistically significant, a one-sided two-sample t-test was performed. For this analysis, the week 1 summer data were compared with the week 1 data collected in the fall. For the subject S.F.1, the average HR was 115 bpm in the fall season and 109 bpm in the summer season. With 95% confidence, the null hypothesis that the average HR between the two seasons would be the same was rejected (p=0.047, α=0.05 level).  In the case of subject S.F.2, the average HR was 107 bpm in the fall season and 101 bpm in the summer season. With 90% confidence, the null hypothesis was rejected (p=0.061, α= 0.1 level). This result contrasted literature-based expectations that suggested a positive association between HR and outdoor temperature. An analysis of the amount of clothing used by the workers in the two seasons 270-6 can help explain this result. In fact, our subjects wore longer sleeves and thicker clothes during the fall season, which may have restricted their physical movement (see Figure 2). Consequently, this may have caused an increase in their core/skin temperatures, and limited the mobility of their arms and torsos while working, which would have required more worker movement, and therefore increased overall physical exertion. As a result, workers reached higher level of physical strain. This result suggests a significant role of clothing insulation in physical exertion, which should be evaluated in future studies. Table 3: Seasonal Comparison of Heart Rate (bpm) Between Subjects S.F.1 and S.F.2  Subject Codes Summer (Jul. 29 – Aug. 2)  Fall  (Oct. 14- Oct.18)  Two-sample t-test Summer vs. Fall  Mean Median Min Max Mean Median Min Max p-value S.F.1 109.3 110.3 57 156 114.6 118.0 70 174 0.047 S.F.2 101.3 102.4 55 159 107.3 109.4 55 189 0.061   Figure 2: Summer Season Clothing (left) and Fall Season Clothing (right) 3.4 Heart Rate versus Climatic Conditions at Peak  In addition, we analyzed the dataset for peak climatic conditions on August 6th. This day was selected because the daily deviation of the ambient temperature was high and biosignal data from three subjects (S.F.1, S.F.2, and S.4) were collected fully over the eight work hours. S.5 was also present, but the last 3 hours of data for that subject were lost due to malfunctions in the chest belt unit. Subject F.3 did not participate in the summer study because he enrolled only for the fall dates. A Pearson correlation analysis found no correlation between ambient temperatures and HR for subjects S.F.1 and S.F.2, and a positive weak correlation for S.4 (r=0.24, p<0.05) (see Table 4).  Table 4: The Pearson Product-Moment Correlation Coefficient between Ambient Temperature (F˚) and Subject’s Heart Rate (bpm) on August 6th  Subject Codes Pearson-r (r) n p-value (α=0.05) S.F.1 0.11 91 0.320 S.F.2 -0.18 91 0.085 S.4 0.24 91 0.023 Since previous studies have suggested that wet-bulb globe temperature (WBGT) rather than outdoor ambient temperature should be the index used to evaluate workers’ heat stress levels and can be a more useful parameter for predicting workers’ heat strain (OSHA 1999; Rowlinson and Yunyan 2014), we also performed a correlation analysis between WBGT and HR (see Table 5). The outdoor WBGT is estimated 270-7 by the weighted sum of the natural wet-bulb temperature (Tw), the globe temperature (Tg), and the ambient temperature (Ta), as shown in Equation 1. [1] WBGT= 0.7Tw + 0.2Tg + 0.1Ta To estimate the Tw, the dew point temperature, the ambient temperature and atmospheric pressure data collected by the weather station were cross-referenced in the American Society of Heating, Refrigerating, and Air-Conditioning Engineers (ASHRAE) Psychrometric Chart. The Tg calculated by Tonouchi et al. (2001) was estimated from the wind speed and solar radiation as shown in Equation 2 below: [2] Tg= Ta + 0.007×S - 0.208×U  where Tg is the black globe temperature (C˚), Ta is the ambient temperature (C˚), S is a solar ration (W/m2) and U is wind speed (m/s). Since the equation is developed in SI base units, imperial units were converted to SI units to apply this equation. From these processes, the WBGTs were estimated for every five minute period from 7:30 a.m. to 3:00 p.m. on August 8th. Since the WBGT calculation model is empirical, it should be different from the measured WBGT from heat stress monitors incorporating WBGT sensing technology. With this calculated data set, additional Pearson correlation analysis found no statistically significant correlation between WBGT and physical strain measured by HR for subjects S.F.1 and S.F.2, and a positive weak correlation for S.4 (r=0.23, p<0.05), as shown in Table 5, from the uncontrolled field data. Table 5: The Pearson Product-Moment Correlation Coefficient between WBGT (F˚) and Subject’s Heart Rate (bpm) on August 6th  Subject Codes Pearson-r (r) n p-value (α=0.05) S.F.1 0.16 91 0.140 S.F.2 -0.19 91 0.067 S.4 0.23 91 0.027  4 CONCLUSIONS AND LESSONS LEARNED The study narrated in this paper used some of the latest off-the-shelf wearable biosensors to perform an extensive field observation that produced a raw dataset: 339 hours of workers’ biosignal data from five subjects, during two seasons (summer and fall), for a total of up to three weeks per subject. The research empirically validated that HR is a good predictor of a worker’s physical strain. Descriptive statistics and a time-series plot were used to analyze the HR pattern as a predictor of worker’s physical strain levels. Correlation analysis was used to analyze the association between the workers’ HR and jobsite environmental stressors. Whereas no association was observed between the climatic data and HR, the approach initially selected to evaluate environmental stressors may have been flawed. In fact, at the end of the study, the site environmental health and safety (EHS) manager, who performed several daily walk-throughs of the jobsite, informed us that there was a considerable difference in conditions between the location of the weather station in the tower crane and the microenvironment on the deck or in the basement where the workers were working. Ideally, the authors would have preferred to collect data in winter (between January and March) that is well representative of the cold weather data to increase the inter-seasonal variability contrasted with the one collected in summer. However, recent advances in sensor networks may provide a solution to this need for a diffused network of environmental sensors that could be resistant to the harshness of a construction jobsite in the future. The preliminary findings from this extensive field study with five subjects are not sufficient to generalize the statistical outcome to all construction workers in Washington State or the United States because the purpose of this study was to evaluate the use of bio- and environmental sensing technologies for understanding processes affecting construction workers’ performance and health.  270-8 The authors were challenged by various practical issues during this first extensive field study. These issues are shared hereafter for the advantage of other researchers.  Use of GPS Trackers: Whereas the adopted GPS units were found reliable in urban environments by previous studies on tracking general human movements, they did not seem as adequate for tracking construction workers. GPS data lacked reliability due to serious signal disconnections. This issue was attributed to the many barriers (e.g., working under the slab formwork, working in the basement, and using a portable toilet) that obstructed clear GPS satellite communications, as well as artificial errors caused by electrical power lines near the site (see Figure 3). Still, we expect that the same devices may provide satisfactory tracking in open-sky conditions.  Figure 3: Causes of GPS Malfunction Confounding Factors: We observed that some subjects frequently drank energy drinks and smoked during the breaks. These confounding factors may have affected the HR and BR data. Thus, additional survey instruments on the subjects’ dietary and smoking habits should be implemented, and potential compounding variables need to be controlled in the field study.  Fear of Reporting to Supervisors: Based on early conversations during the design of the field study, we perceived construction workers being concerned that the study’s results would be reported to their supervisors at the individual level. This created a sense of resistance for the workers to be monitored through wearing sensors. Therefore, we implemented a careful recruitment process in which a consent form was used to instruct workers about the confidentiality of their data and to assure them that the results of the research would not cause them any disadvantages relative to their supervisors. Still, we adopted a top-down recruitment approach wherein project management helped recruit workers. As a result, a member of the field management team was able to “figure out” who the subjects were and manifested an extreme curiosity in knowing his individual workers’ HR trends. Although the researchers strongly protected the confidentiality of the subjects, this behavior resulted in some residual diffidence by the workers that may have caused some of the dropouts. In the future, we will explore alternative approaches to recruitment that are more bottom-up and use unions and apprenticeship programs as recruitment venues. In addition, we will avoid installing instrumentation in the same trailer used by the field management. Fear of Underperforming against Peers: Insincere reporting probably occurred among subjects in the SPS perceived fatigue survey. For example, a worker was caught looking at other subjects’ survey forms, and another worker asked researchers what levels of fatigue other subjects had reported. These behaviors may indicate attempts to report fatigue levels that did not reflect the actual perceived fatigue. Thus, subjects should complete the surveys away from one another. Trades to be Observed: All the subjects worked for the general contractor on site. As a result, the types of tasks assigned to the subjects presented high variability in scope and target outputs. Some of the inconclusive results may be attributed to this issue. In the future, we plan to perform field observation of subjects belonging to specialty trades as the variability in their scope and target output will be reduced.   Use of Video Recording: While recording short videos of construction activities that could help capture confounding factors, we observed that subjects often became aware of being recorded and changed their 270-9 work behaviors. This was expected and we kept these periods very short. However, positioning several fixed cameras may overcome this issue in the long term.  The cited challenges allowed the authors to outline a roadmap for the successful implementation of bio- and environmental sensors in the construction field. First, it is important to carefully design a study and a managerial strategy to minimize both workers’ resistance to recruitment and insincere reporting of perceived variables. Second, subjects should be recruited from subcontractors involved in the slab framework, roofing, drywall, or masonry trades because these tasks are repetitive and physically intense, require less multitasking, and limit the dimensions of the traveling area, which in return increase the chance of successfully using tracking devices to perform automatic work sampling (Cheng et al. 2013). Third, the use of GPS tracking should be limited to open-sky tasks with minimal barriers presented by built slab framework or roofing. Whereas a need of analyzing indoor activities in established, traditional visual work sampling (Liou and Borcherding 1986) could be applicable to provide an indirect measure of physical exertion and performance. Fourth, cooperation and advice from site superintendents and safety managers are needed to appropriately place environmental sensors for local measurements and cameras that would limit the confounding factors of having a researcher on-site operating these devices. References Abdelhamid, T. S., and Everett, J. G. 2002. Physiological Demands During Construction Work. Journal of Construction Engineering and Management, 128(5): 427-437. Astrand, P., Rodahl, K., Dahl, H. A., and Stromme, S. B. 2003. Textbook of Work Physiology. Human Kinetics, Champaign, IL, USA. Beek, AJ van der, and Frings-Dresen, M. H. W. 1995. Physical Workload of Lorry Drivers: A Comparison of Four Methods of Transport. Ergonomics 38(7): 1508-1520. Bernold, L. E., and AbouRizk, S. M. 2010. Managing Performance in Construction. John Wiley & Sons. CPWR, 2013. The Construction Chart Book-Fifth Edition, The Center for Construction Research and Training (CPWR). Cheng, T., Migliaccio, G. C., Teizer, J., and Gatti,U. C. 2012. Data Fusion of Real-time Location Sensing and Physiological Status Monitoring for Ergonomics Analysis of Construction Workers. Journal of Computing in Civil Engineering 27(3): 320-335. Cheng, T., Teizer, J., Migliaccio, G. C., and Gatti,U. C. 2013. Automated Task-Level Activity Analysis Through Fusion of Real Time Location Sensors and Worker's Thoracic Posture Data. Automation in Construction 29: 24-39. Dorrian, J., Baulk, S. D. Baulk, and Dawson, D. 2011. Work Hours, Workload, Sleep and Fatigue in Australian Rail Industry Employees." Applied Ergonomics 42(2): 202-209. Dolezal, Ba, Boland, D. M., Carney, J., Abrazado , M., Smith D.l., and Cooper C.B. 2014. Validation of Heart Rate Derived from a Physiological Status Monitor-Embedded Compression Shirt against Criterion ECG. Journal of Occupational and Environmental Hygiene 11(12): 833-839. Everett, J. G. 1999. Overexertion Injuries in Construction. Journal of construction engineering and management 125(2): 109-114. Gatti, U. C., Migliaccio, G. C., and Schneider, S. 2011. "Wearable Physiological Status Monitors for Measuring and Evaluating Worker’s Physical Strain: Preliminary Validation." Computing in Civil Engineering (2011): pp. 194-201. Gatti, U. C., Schneider, S., and Migliaccio, G. C. 2014.Physiological Condition Monitoring of Construction Workers. Automation in Construction 44: 227-233. Garet, M., Boudet, G., Montaurier, C., Vermorel, M., Coudert, J., and Chamoux, A. 2005. Estimating Relative Physical Workload Using Heart Rate Monitoring: A Validation by Whole-Body Indirect Calorimetry. European Journal of Applied Physiology, 94(1-2): 46-53. Hallowell, M. R. 2010. Worker Fatigue. Professional Safety, 55(12), 18-26. Retrieved from http://search.proquest.com/docview/763614893?accountid=14784 Lee, W. and Migliaccio, G. 2014. Field Use of Physiological Status Monitoring (PSM) to Identify Construction Workers' Physiologically Acceptable Bounds and Heart Rate Zones. Computing in Civil and Building Engineering (2014): pp. 1037-1044. 270-10 Liou, F. S., and Borcherding, J. D. 1986. Work Sampling Can Predict Unit Rate Productivity. Journal of Construction Engineering and Management 112(1): 90-103. Occupational Safety and Health Administration (OSHA) 1999. OSHA Technical Manual (OTM) Section III: Chapter 4 (TED 01-00-015). Washington D.C., OSHA, Office of Science and Technology Assessment. Accessed January 15, 2015. https://www.osha.gov/dts/osta/otm/otm_iii/otm_iii_4.html.  Putz-Anderson, V., Bernard, B. P., Burt, S. E., Cole, L. L., Fairfield-Estill, C., Fine, L. J., Grant, K. A. et al. 1997.  Musculoskeletal Disorders and Workplace Factors. National Institute for Occupational Safety and Health (NIOSH). Paz-Soldan, V., Stoddard, S.T., Vazquez-Prokopec, G. , Morrison, A. C., Elder, J. P., Kitron, U.,  Kochel ,T. J., and Scott,T. W. 2010. Assessing and Maximizing the Acceptability of GPS Device Use for Studying the Role of Human Movement in Dengue Virus Transmission in Iquitos, Peru. AmJTrop Med Hyg 82 (4): 723-730. Ringen, K., Seegal, J. and England, A. 1995. Safety and Health in the Construction Industry. Annual Review of Public Health 16(1): 165-188. Rowlinson, Steve, and Jia, Y.A. 2014. Application of the Predicted Heat Strain Model in Development of Localized, Threshold-based Heat Stress Management Guidelines for the Construction Industry. Annals of Occupational Hygiene 58(3): 326-339. Samn, S. W., and Perelli, L. P.. 1982. Estimating Aircrew Fatigue: A Technique with Application to Airlift Operations. No. SAM-TR-82-21. SCHOOL OF AEROSPACE MEDICINE BROOKS AFB TX. Schäfer, A., & Vagedes, J. 2013. How Accurate is Pulse Rate Variability as An Estimate of Heart Rate Variability?: A review on Studies Comparing Photoplethysmographic Technology with An Electrocardiogram. International Journal of Cardiology, 166(1): 15-29. Takken, T., Ribbink, A., Heneweer, H., Moolenaar, H. and Wittink, H. 2009. Workload Demand in Police Officers during Mountain Bike Patrols. Ergonomics 52(2): 245-250. Terbizan, D. J., Dolezal, B. A. and Albano, C. 2002. Validity of Seven Commercially Available Heart Rate Monitors. Measurement in Physical Education and Exercise Science 6(4): 243-247. Tonouchi, M., Murayama, K., and Ono, M. 2006. WBGT Forecast for Prevention of Heat Stroke in Japan. In Sixth Symposium on the Urban Environment, American Meteorological Society, JP1(1). Vazquez-Prokopec, G. M., Stoddard, S. T., Paz-Soldan, V. , Morrison, A. C., Elder, J. P., Kochel, T. J., Scott, t. W. and Kitron, U. 2009. Usefulness of Commercially Available GPS Data-Loggers for Tracking Human Movement and Exposure to Dengue Virus. International Journal of Health Geographics 8(1): 68.  Lessons	  Learned	  From	  Using	  Bio-­‐	  and	  Environmental	  Sensing	  in	  Construc9on:	  A	  Field	  Implementa9on	  Wonil	  Lee*	  Giovanni	  Migliaccio	  Ken-­‐Yu	  Lin	  University	  of	  Washington	  The	  CSCE	  Interna9onal	  Construc9on	  Specialty	  Conference	  2015	   1	  	  	  Francesca	  Russo	  University	  of	  Naples	  Federico	  II	  Mo9va9on	  •  Industry	  Workforce	  Trends	  – Fatali9es	  and	  Injuries	  – Health	  Issues	  – Presenteeism	  and	  Absenteeism	  – Stagnant	  Labor	  Produc9vity	  Improvement	  – Labor	  Shortage	  The	  CSCE	  Interna9onal	  Construc9on	  Specialty	  Conference	  2015	   2	  Mo9va9on	  •  Factors	  Affec9ng	  Workforce	  Performance	  and	  Health	  The	  CSCE	  Interna9onal	  Construc9on	  Specialty	  Conference	  2015	   3	  Problem	  Statement	  •  Worker’s	  physical	  strain	  affects:	  – Produc9vity	  	  – Quality	  – Safety	  – Health	  The	  CSCE	  Interna9onal	  Construc9on	  Specialty	  Conference	  2015	   4	  Off-­‐the-­‐Shelf	  Sensor	  Technologies	  Wearable	  Biosensors	  Environmental	  Sensors	  Objec9ves	  •  Sharing	  findings	  and	  lessons	  learned	  from:	  	  – Field	  monitoring	  study	  of	  bio-­‐	  and	  environmental	  sensors	  – Par9cipants	  administra9on	  and	  observer	  effect	  issues	  The	  CSCE	  Interna9onal	  Construc9on	  Specialty	  Conference	  2015	   5	  Data	  Collec9on	  Methods	  •  Biosensor	  	  –  Zephyr	  BioHarness™3	  •  Environmental	  Sensor	  	  –  Davis	  Instruments	  Corp.	  Vantage	  Pro2™	  Plus	   The	  CSCE	  Interna9onal	  Construc9on	  Specialty	  Conference	  2015	   6	  Data	  Collec9on	  Methods	  •  Other	  Instruments:	  – GPS	  loca9on	  tracking	  – Perceived	  fa9gue	  level	  – Workers’	  major	  tasks	  performed	  – Video	  recording	  The	  CSCE	  Interna9onal	  Construc9on	  Specialty	  Conference	  2015	   7	  Data	  Collec9on	  •  Five	  healthy	  workers	  •  Mid-­‐rise	  building	  construc9on	  site	  •  SeaXle,	  Washington	  State,	  US	  •  Schedule	  of	  Observa9ons	  – July	  29th	  to	  August	  9th	  (2	  Weeks)	  – October	  14th	  to	  October	  18th	  (1	  Week)	  The	  CSCE	  Interna9onal	  Construc9on	  Specialty	  Conference	  2015	   8	  Data	  Analysis	  The	  CSCE	  Interna9onal	  Construc9on	  Specialty	  Conference	  2015	   9	  3.	  Physical	  Strain	  and	  Environmental	  Job	  Stressor	  Physical	  Strain	  =	  f	  (Average	  HR)	   Correla9on	  Analysis	  2.	  Physiological	  Status	  Physiological	  Acceptable	  Bounds	   Heart	  Rate	  Zones	  1.	  Empirical	  Valida9on	  of	  Previous	  Finding	  Data	  Pre-­‐Processing	   Time	  Series	  Plot	  Data	  Analysis	  •  Subject	  Informa9on	  The	  CSCE	  Interna9onal	  Construc9on	  Specialty	  Conference	  2015	   10	  Subject	  Codes	   BMI	   Major	  Task	   Study	  Par9cipa9on	  Total	  Hours	  of	  Data	  Collected	  	   	   	   Summer	  	   Fall	   	  S.F.1	   27.4	   Formwork	   ●	   ●	   120	  S.F.	  2	   25.8	   Formwork	   ●	   ●	   85	  F.3	   26.9	   Formwork	   	   ●	   27	  S.4	   30.4	   Concrete	  Pouring;	  Cleaning	  Deck	   ●	   	   59	  S.5	   25.1	   Layout;	  Pour	  Watch	   ●	   	   48	  Findings	  The	  CSCE	  Interna9onal	  Construc9on	  Specialty	  Conference	  2015	   11	  Time	  Heart	  Rate	  	  (Beats	  per	  minute)	  Breathing	  Rate	  (Breaths	  per	  minute)	  Findings	  •  Physical	  Strain	  Level	  Measured	  by	  HR	  The	  CSCE	  Interna9onal	  Construc9on	  Specialty	  Conference	  2015	   12	  Findings	  •  Seasonal	  Comparison	  of	  Worker’s	  Physical	  Strain	  Level	  The	  CSCE	  Interna9onal	  Construc9on	  Specialty	  Conference	  2015	   13	  Clothing	  Insula9on	  Summer	   Fall	  The	  CSCE	  Interna9onal	  Construc9on	  Specialty	  Conference	  2015	  Subject	  Codes Pearson-­‐r	  (r) n p-­‐value	  (α=0.05) S.F.1 0.16 91 0.140 S.F.2 -­‐0.19 91 0.067 S.4 0.23 91 0.027 Findings	  •  Wet-­‐Bulb	  Globe	  Temperature	  (WBGT)	  – WBGT=	  0.7Tw	  +	  0.2Tg	  +	  0.1Ta	  •  Tw:	  Natural	  wet-­‐bulb	  temperature	  	  •  Tg	  :	  Globe	  temperature	  •  Ta	  :	  Ambient	  temperature	  •  WBGT	  (F˚)	  and	  Physical	  Strain	  (HR:	  bpm)	  14	  Lessons	  Learned	  •  Use	  of	  GPS	  Trackers	  •  Confounding	  Factors	  •  Fear	  of	  Repor9ng	  to	  Supervisors	  •  Fear	  of	  Underperforming	  against	  Peers	  The	  CSCE	  Interna9onal	  Construc9on	  Specialty	  Conference	  2015	   15	  Causes	  of	  GPS	  Malfunc9on	  Lessons	  Learned	  •  Trades	  to	  be	  Observed	  – Less	  variability	  in	  scope	  and	  target	  outputs	  	  •  Use	  of	  Video	  Recording	  – Posi9oning	  several	  fixed	  cameras	  – Back-­‐up	  via	  tradi9onal	  manual	  work	  sampling	  The	  CSCE	  Interna9onal	  Construc9on	  Specialty	  Conference	  2015	   16	  Industry	  Applica9ons	  	  Use	  of	  Biosensor	  •  S&H	  professionals:	  	  – Monitoring	  for	  workers’	  poten9al	  overexer9on	  	  •  Field	  management:	  	  – Managing	  workers’	  task	  demands	  •  Laborers:	  	  – Self-­‐pacing	  by	  tracking	  physical	  strain	  level	  The	  CSCE	  Interna9onal	  Construc9on	  Specialty	  Conference	  2015	   17	  Special	  Acknowledgements	  •  Skanska	  Innova9on	  Fund	  for	  equipment	  funding,	  and	  Stone	  34	  project	  staff	  and	  labor	  for	  their	  help	  and	  pa9ence	  The	  CSCE	  Interna9onal	  Construc9on	  Specialty	  Conference	  2015	   18	  Q&A	  •  Thank	  you	  for	  your	  aXen9on!	  – Wonil	  Lee,	  PhD	  Candidate,	  Department	  of	  Construc9on	  Management,	  University	  of	  Washington	  – Email:	  wonillee@uw.edu	  The	  CSCE	  Interna9onal	  Construc9on	  Specialty	  Conference	  2015	   19	  

Cite

Citation Scheme:

        

Citations by CSL (citeproc-js)

Usage Statistics

Share

Embed

Customize your widget with the following options, then copy and paste the code below into the HTML of your page to embed this item in your website.
                        
                            <div id="ubcOpenCollectionsWidgetDisplay">
                            <script id="ubcOpenCollectionsWidget"
                            src="{[{embed.src}]}"
                            data-item="{[{embed.item}]}"
                            data-collection="{[{embed.collection}]}"
                            data-metadata="{[{embed.showMetadata}]}"
                            data-width="{[{embed.width}]}"
                            async >
                            </script>
                            </div>
                        
                    
IIIF logo Our image viewer uses the IIIF 2.0 standard. To load this item in other compatible viewers, use this url:
http://iiif.library.ubc.ca/presentation/dsp.52660.1-0076458/manifest

Comment

Related Items