Automation in Construction 62 (2016) 101–113

Contents lists available at ScienceDirect

Automation in Construction

journal homepage: www.elsevier.com/locate/autcon

Development of an early-warning system for site work in hot and humid

environments: A case study
Wen Yi a,b,c,⁎, Albert P.C. Chan a, Xiangyu Wang b,c,d, Jun Wang b,c
a
Department of Building and Real Estate, The Hong Kong Polytechnic University, Hung Hom, Kowloon, Hong Kong, China
b
School of Built Environment, Curtin University, Australia
c
Australasian Joint Research Centre for BIM, Curtin University, Australia
d
Department of Housing and Interior Design, Kyung Hee University, South Korea

a r t i c l e i n f o a b s t r a c t

Article history: This study presents an early-warning system for working in hot and humid environment. The developed system
Received 5 December 2014 can monitor workers' heat strain level when they have to work under such hostile conditions continuously.
Received in revised form 11 October 2015 Health alert messages with corresponding intervention measures will be prompted to workers to safeguard
Accepted 7 November 2015
their wellbeing. Heat strain is evaluated by a subjective index perception rating of perceived exertion (RPE)
Available online 1 December 2015
and an objective heat strain indicator heart rate. A database containing 550 sets of synchronized work-related,
Keywords:
environmental, and personal data were used to construct the prediction model. Artificial neural networks
Artificial neural networks (ANNs) were applied to forecast the RPE of construction workers. Statistical measures including MAPE, RMSE and R2
Construction industry confirm that the established model is good fitting with high accuracy. The proposed system could be automated
Early-warning system by integrating smart sensor technology, location tracking technology, and information communication tech-
Heat stress nology, which could be in the form of GSM based environmental sensor, smart bracelet, and smart phone appli-
Hong Kong cation, to protect the wellbeing for those who have to work in hot and humid conditions.
Occupational health and safety (OHS) © 2015 Elsevier B.V. All rights reserved.

1. Introduction term) illnesses from their exposure to environmental hazards [4].

Construction is a large, complex, and dynamic sector that generates
employment for millions of people worldwide. However, this sector has
the most fatalities and high incidence of non-fatal occupational injuries
and illnesses on days away from work. Based on the estimates of Inter-
national Labour Organization (ILO), at least 60,000 fatal accidents occur
each year in construction sites around the world, which represent one
fatal accident every 10 min [1]. In addition, the ILO estimates that the
construction workers in industrialized countries are 3 to 4 times greater
than other employees to die from accidents at work [2]. Aside from the
dangers of being the front-liners on a jobsite, workers in the construc-
tion industry also confront potential health hazards (e.g., temperature
extremes, radiation, chemicals, dusts, vibration, and noise) throughout
the building process. Furthermore, about 30% of construction workers
in some European countries suffer from pain and musculoskeletal dis-
orders [3]. The occupational illnesses of construction workers have not
been accurately measured, but an educated guess indicates that con-
struction workers suffer both acute (short-term) and chronic (long-
⁎ Corresponding author at: ZN703, Department of Building and Real Estate, The Hong
Kong Polytechnic University 11 Yuk Choi Road, Hung Hom, Kowloon, Hong Kong
999077, China. Tel.: +852 2766 5794; fax: +852 2766 5131.
E-mail addresses: yiwen96@163.com (W. Yi), albert.chan@polyu.edu.hk (A.P.C. Chan),
xiangyu.wang@curtin.edu.au (X. Wang), jun.wang15@postgrad.curtin.edu.au (J. Wang).
The health and safety of construction workers aroused greater
attention from governments, industry community practitioners,
and the academia.
Heat stress is a well-known occupational hazard in the construction
industry [4], and climate change together with the increased frequency
and intensity of extreme heat events have made risks more severe and
widespread [5-8]. Heat stress causes physiological and psychological
discomforts, deteriorates performance and productivity, increases
incident rates, and even threatens survival [9-11]. Increased ther-
moregulatory, cardiovascular and perceptual strains on the body
promote confusion, irritability, and other emotional stress, which
may cause workers to distract attention from tasks or ignore safety
procedures [12]. Published reports in the United States suggest
that work in construction occupation is associated with an increased
risk of heat-related death and illness [13,14]. In Japan, a total of 47
deaths due to heat stroke were reported as industrial accidents in
hot environments during 2010. The construction industry is found
to be more susceptible to heat stress than other industries, which
is accounted for 64% of all lethal cases [15]. News reports archived
in Hong Kong show alarming incidences of heat stress and verifiable
reported deaths in the construction industry [16]. A recent survey
revealed that 5% of construction workers had suffered from heat stroke
and 23% had experienced symptoms of heat stroke [17]. In Taiwan,
construction workers were identified as the most vulnerable population

http://dx.doi.org/10.1016/j.autcon.2015.11.003
0926-5805/© 2015 Elsevier B.V. All rights reserved.
102 W. Yi et al. / Automation in Construction 62 (2016) 101–113
in which high temperature impacts on health and productivity [18]. The
risk of heat stress experienced by construction workers may even be
higher in the Middle East where ambient air temperatures often reach
45 °C and higher with 90% humidity [19]. Considering the high frequency
of heat-related incidents in the construction industry, understanding how
heat stress results in heat-related illnesses and how it affects construction
workers is important in planning for intervention strategies.
To prevent heat stress, a series of fundamental practice notes and
guidelines (e.g., appropriate work arrangements, shelters at work or
rest places to reduce radiant heat gain, ventilation in indoor working en-
vironment, air-conditioned rest rooms, and provision of drinking water
or sports drinks) have been promulgated [20-23]. Likewise, the issue of
working under hot weather has been a concern of academic researchers.
The limits of human tolerance in terms of physiological parameters
(e.g., core temperature, heart rate, skin temperature) have been evalu-
ated at different levels of heat exposure [24-26]. Upper tolerance limits
in terms of environmental indicators (e.g., temperature, humidity, wet
bulb globe temperature, thermal work limit) have been explored in lit-
erature [27] and adopted by regulatory organizations [28-30]. Recently,
attempts and efforts were made to establish a safety evaluation model for
assessing the risk of heat stress in the workplace. Evaluating safety is not
only an important means for implementing the policy but also provides a
base for establishing a scientific and standardized management of enter-
prises [31]. For example, Ren et al. [32] established an evaluation frame-
work for assessing the hazards of heat stress in the workplace in an
underground mine. Zheng et al. [33] studied safety evaluation and early
warning rating of hot and humid environment. However, no systematic
and in-depth studies have been conducted with respect to safety evalua-
tion and early warning in hot and humid environment for the identifica-
tion, evaluation, control, and management of human behavioral factors.
The performance of work activities befitting safety and accident preven-
tion is a continuing and dynamic process. The key to achieving these ob-
jectives lies with workers' concentration. The cognitive condition of
concentration can be viewed through the concept of mindfulness [34].
Mindful work organization and performance can be achieved through
the improvement of workers' alertness to and awareness of the hazardous
nature of the operations.
Occupational heat strain results from a combination of factors,
which include environmental conditions, work demands, and individu-
al characteristics. Earlier studies by Chan et al. [35,36] established a mul-
tiple linear model (MLR) to predict a worker's heat strain to different
environmental factors, work-related factors, and personal factors.
However, it was challenged that heat stress and heat strain may not
be linearly correlated [37]. Furthermore, the combination of the large
array of personal health and lifestyle factors and their complex interac-
tion effects is far beyond the predictive power of a MLR [38]. More
advanced analytical techniques are required to tackle these complex is-
sues. Artificial neural networks (ANNs), a form of artificial intelligence
technique, is one such approach which provides a high level of flexibility
and competency in nonlinearities and complex behavior. ANN provides
solutions to many complex problems in biology and medicine that are
beyond the computational capacity of classical mathematical and tradi-
tional statistical techniques [39]. The application of ANN in data treat-
ment is high, particularly where systems present nonlinearities and
complex behavior [40].
Innovative devices and technologies [e.g., physiological status mon-
itor (PSM) and Ultra-Wideband (UWB)] are being used to real-time
location tracking and physiological status of construction workers for
enhancing construction safety and productivity [41-47]. This research
aims to develop an early-warning system for construction workers
against hot and humid climates using modern technologies and ANN
as the principal analytical technique. The objectives of this study are
to (1) develop a model to predict a worker's heat strain in hot and
humid environment; and (2) identify proper precautions against the
hazards and risks in hot and humid environment to prevent and reduce
the harmful effects of heat exposure. Since Hong Kong is in a subtropical
climate zone where air temperature typically reaches 34.5 °C on its hot-
test summer days [48] and business environment of the construction in-
dustry in Hong Kong is highly competitive [49], the Hong Kong
construction industry is selected as a prototype for developing a more
focused methodology which if successful could also be applied to
other regions.
2. Design of the early-warning system
2.1. Factors used to develop early-warning system
2.1.1. Heat strain assessment
Heat stress is defined by National Institute for Occupational Safety
and Health as the sum of the heat generated in the body (metabolic
heat) plus the heat gained from the environment minus the heat lost
from the body to the environment [30]. Heat strain describes the overall
physiological and psychological response resulting from heat stress
[50]. International Standard Organization has identified the indicators
for heat strain, including body core temperature, skin temperature,
heart rate and the loss of body mass through sweating [51]. Heart rate
is the earliest response of physiological strain [52], which was used to
assess the strain of construction workers [53]. Previous studies reported
that the normal heart rates for undertaking heavy tasks in the heat were
in the range of 130–160 bpm [24,54]. According to ACGIH, an
individual's heat stress exposure should be discontinued when his/her
heart rate exceeds 180 beats per minute (bpm) minus the age of the
employee sustained for at least 3 min [55].
Apart from the objective physiological parameters, heat strain can be
evaluated by a subjective perception index rating of perceived exertion
(RPE) which indicates the intensity of subjective effort, stress, or dis-
comfort felt during physical activity [56]. RPE provides a useful indica-
tion of the capacity to continue a task [57]. The RPE scales use both
verbal anchors and numbers that have been reported to possess both
categorical and interval properties [58]. The Borg CR10 Scale was used
in this study. The anchors ranged from 0 = “rest” to 10 = “maximal
exertion.”
Previous studies have demonstrated that the RPE is highly correlated
with environmental factors such as temperature, relative humidity, air
velocity, and solar radiation; personal factors such as age, body fat,
and alcohol drinking/smoking habits; and work-related factors such as
work duration and work intensity [59-62]. The wet bulb globe temper-
ature (WBGT) which responds to all four elements of the thermal envi-
ronment (i.e., air temperature, relative humidity, air velocity, and solar
radiation) is adopted as an environmental parameter in this study.
WBGT is so far the most widely used and accepted environmental
index for managing occupational heat stress [63]. The body mass
index (BMI) measures body fat based on height and weight. Alcohol
drinking and smoking habits will be recorded under three categories ac-
cording to the amount of weekly alcohol intake and cigarette consump-
tion. According to the World Health Organization (WHO) [64], one
“standard” drink contains 10 g of pure alcohol [e.g., 250 ml (3/4 can)
of regular beer has 5% alcoholicity, 100 ml (1 small glass) of wine has
12% alcoholicity, 30 ml (1 pub measure) of hard liquor has 40%
alcoholicity]. “Low-risk” drinking levels for men represent no more
than four drinks on any single day, and no more than 14 drinks per
week [65,66]. Thus, alcohol drinking habits are classified into three cat-
egories, namely “none,” “no more than four drinks on any single day and
no more than 14 drinks,” and “more than four drinks on any single day
and no more than 14 drinks.” Based on the study by Chan et al. [67] on
medical examination for construction workers, smoking habits were
classified in three categories, namely “none,” “no more than 35 ciga-
rettes per week,” and “more than 35 cigarettes per week”. Workers of
different trades may be subject to different degrees of physical de-
mands. Energy expenditure, defined as a sum of internal heat produced
and external work, is a common method to quantify the intensity of
physical works in a hot environment [68]. Workload is classified into
 W. Yi et al. / Automation in Construction 62 (2016) 101–113
three categories, namely, light (b841 kJ/h), moderate (841–1297 kJ/h),
and heavy (1297–1681 kJ/h), according to the guidelines by ACGIH
and ISO 7243. Abdelhamid and Everett [69] investigated the physi-
cal demands of construction work and provided reference values
of energy expenditure among different trades. The workload classi-
fication for construction trade is listed in Table 1.
2.1.2. Warning sign
To prevent heat stroke, it is important to consider which workers are
at risk and provide early warning signs before injury occurs. The warn-
ing sign is designed to alert workers of the seriousness of heat stress, in-
dicate potential consequences as well as recommend proper actions to
avoid such happenings [70,71]. The alerting function can be conveyed
using symbols or signs involving various colors and shapes [71]. In this
study, the level of heat strain will be categorized into “mostly safe,”
“caution,” “warning,” and “severe warning,” which are symbolized as
“blue,” “green,” “orange” and “red,” color respectively, to draw the at-
tention of workers.
Physical work in a hot and humid environment not only imposes
considerable heat strain on workers, but also leads to heat illness and
even deaths. The continuum of heat illness includes mild disease
(e.g., heat edema, heat rash, heat cramps, and heat syncope), heat ex-
haustion, and potentially life-threatening heat stroke, which is the
most severe form of heat illness [72]. Heat exhaustion is a moderate
heat illness in which the core temperature falls 38 °C and 40 °C [73].
Heat stroke is characterized by a core temperature of N40 °C accompa-
nied by changes in mental status [74]. Often, heat stroke is accompanied
by acute symptoms, such as extreme fatigue, hot skin or heavy perspira-
tion, nausea, vomiting, diarrhea, and reddened faces [75]. Information
about the possible consequences (symptoms of heat-related illness)
and the appropriate actions to engage in or avoid have been reported
in a series of practice notes and guidelines on working in hot weather
[22,23,27,76]. These recommendations will be captured and presented
in the warning system. Details of the warning system are shown in
Table 2.
2.2. System design conception
The purpose of the early-warning system is to safeguard the well-
being of construction workers while working in hot and humid environ-
ment. The system should alert construction workers of occupational
heat stress, and indicate probable consequences, and subsequently rec-
ommend appropriate behavior to engage in or avoid. Workers' heat
strain level is evaluated by the heart rate and RPE from the perspective
of physiological and psychobiological responses respectively. The early
warning system collects four sets of data: (a) instant WBGT on construc-
tion sites; (b) work duration and activities undertaken by the workers;
(c) personal characteristics of workers (i.e., age, height, weight, alcohol
drinking habit, and smoking habit); and (d) instant heart rate of
workers. An ANN model is developed to predict the RPE of workers to
automate the assessment and early warning for heat strain. Based on
Table 1
Work load classification for construction trade.
Work load Energy expenditure Construction trade
Light b841 kJ/h Electrician
Moderate 841–1297 kJ/h Asbestos workers
Bricklayer
Cement finisher
Drywall installer
Glazier
Pipe fitter
Sheet metal worker
Heavy 1297–1681 kJ/h Carpenter
Ironworker
Laborer
103
workers' RPE computed by the ANN model and the real-time monitor-
ing heart rate, different levels of warning signs will be prompted to rec-
ommend appropriate actions well in advance before the potential
hazard may realize.
The early-warning system can be considered as a highly dynamic
and network operating system, which involved several automated de-
vices (e.g., GSM based environmental sensors, smart phone, smart
bracelet) that collate, monitor, analyze, and disseminate the safety in-
formation. The system could be automated by integrating smart sensor
technology, location tracking technology, and information communica-
tion technology. Fig. 1 shows the system measures the environmental
parameter of the construction sites and monitors the physiological var-
iables of the workers and automatically sends early-warnings in real
time to both frontline workers and supervisory staff.
To activate the early-warning system, workers could input their per-
sonal and work-related information (i.e., age, height, weight, smoking
habit, alcohol drinking habit, and job nature) into the application soft-
ware (app) and set the stop-watch timer before they start their daily
work. Once starting the daily work, smart bracelet wore by the workers
will monitor their heart rate. Smart phone carried by workers will deter-
mine their location and receive the synchronous WBGT data from the
adjacent environmental sensors. Based on the personal, work-related,
environmental data, workers' RPE can be computed by the ANN model
via a smart phone app. Warning alerts will be sent with sound track to
the workers concerned once certain thresholds exceed. For example,
an alert of “red” warning with suggested strategies to rest in the shade
and cool down, report symptoms of heat illnesses if any will be sent
when the predicted RPE of a worker reached 9 or his/her heart rate ex-
ceeds 180 bpm minus the age of the employee sustained for at least
3 min. During the prolonged operation, frontline workers might suffer
from a spectrum of discomforts such as decreased muscle coordination,
nausea, headaches, dizziness, rapid heart rate and breathing. They can
report the heat-related disorders to their supervisors or safety pro-
fessionals via the smart phone. With the wireless communication
technology (e.g., 3G connection), supervisory staff can immediate
acquire workers' level of heat strain, location, personal information
and symptoms of heat-related illness via a PC/smart phone that connects
with a cloud database. Process map for the early-warning system is illus-
trated in Fig. 2.
2.3. System architecture
As illustrated in Fig. 3, the early-warning system is composed of four
layers: (1) sensing layer, (2) transmitting layer, (3) data processing
layer, and (4) display and control layer.
2.3.1. Sensing layer
The sensing layer is the basis of the early-warning system, which
measures the environmental parameters of construction sites, monitors
the physiological conditions of workers, and determines location and
operation duration of workers. WBGT can be easily computed from
the readings of three thermometers on wet bulb temperature, globe
temperature, and dry bulb temperature. The three thermometers are
simple to use and inexpensive [77]. To measure the WBGT at different
locations of a construction site, several WBGT sensors assembled by
sensor module, microcontroller, and GSM module are utilized. The sen-
sor module contained aforementioned three environmental sensors.
The microcontroller converts the analog signals from the sensor to dig-
ital form, processes the raw data, packages the data into a message, and
sends the message to the GSM module, and orchestrates the routing of
messages through the network. The smart bracelet worn on the wrist
can monitor workers' heart rate and interact with their smartphone
via Bluetooth. Commercially smart bracelet with the function of real
time heart rate monitoring is available from many sources. It could con-
stantly track worker' heart rate by calculating changes in blood volume.
104 W. Yi et al. / Automation in Construction 62 (2016) 101–113

Table 2
Warning sign of the early-warning model in hot and humid environment.

RPE Descriptor Heart rate Signs and symptoms Intervention (recommended strategies) Warning sign

0 Rest • < less than 100 bpm • Able to work in the conditions • Take potable (safe for drinking) water close

without any problem to the work area frequently

1 Very, very easy Mostly safe

2 Easy (Blue)

3 Moderate

4 Somewhat hard • 100 –120 bpm • Moist and pale skin • Fluid intake of more than 1 litre per hour

required Caution
5 Hard
• Work–rest cycling or rotation required (Green)
6 – • Avoid working under direct sunlight

• 120 – 160 bpm
7 Very hard • Painful spasm of muscle in the
arms, legs, or abdomen
• Sweaty skin
• Light–headedness
8 –
• Red rash on skin
• Weakness and fatigue
• Take your breaks in the shade if you can
• Rectify ventilations or cool down facilities if
possible.
Warning
• Report symptoms of heat illnesses if any,
(Orange)
avoid doing heavy work
• Avoid working under direct sunlight
• Avoid working alone

• More than 180 bpm • Headache • Should rest in the shade and cool down
9 – minus the age of the • Irritability or confusion • Report symptoms of heat illnesses if any

employee sustained for • Rapid heart rate and breathing Severe

at least 3 min • High body temperature warning (Red)

10 Maximum • R ed, hot, dry skin

• Decreased muscle coordination

The duration of the operation and location of workers can be tracked by
a stop watch timer and GPS chip embedded in the smart phone.
2.3.2. Transmitting layer
The transmitting layer is an important connection that links the
sensing layer and data processing layer. Physiological data collected by
the smart bracelet could be transferred to their smart phone via
Bluetooth. Environmental data sampled from the WBGT sensor could
be transmitted to smart phone via GSM. A set of data in worker' smart
phone including WBGT, heart rate, heat stress assessment, and symp-
toms of heat illness could also be transmitted to supervisor's PC/smart
phone via GSM. The Global System for Mobile Communications (GSM)

WBGT of Location of
construction sites environmental sensors

Smart sensor Location tracking

technology technology

Heart rate of Location of

construction workers construction workers
Early warning system

Information
communication
technology

Bluetooth GSM

Fig. 1. Early-warning system with integration of smart sensor technology, location tracking technology, and information communication technology.
 W. Yi et al. / Automation in Construction 62 (2016) 101–113 105

Sm
mart brac
celett S rt pho
Smar one (Wor
( rker)

I
Issue healtth ale
ert me
essagees witth
c ding in
spond
corres nterve
ention
n

Mo kers’ h
onitorr work heart rate
ack th
Tra cation of wo
he loc s
orkers

Sm b elet
mart brace
…… t rate
Heart
H e

ker 1
Work …… W er n
Worke
ecord the person
Re nal an
nd
Sm
mart phon
ne (W
Worker)
ork-rellated data
wo
eceive
Re e WBG om
GT fro
mental
adjjacent environm Coomputting in
n
the
e app
S rt pho
Smar one ((Worrker)
……
AN
NN m
modell
En
nviron
nmenttal …… Env
vironm
menta
al
sens
sor 1 senso
s rm

PC
C
G n work
GSM netw
arnin
Wa ng sig
gn

Heatt staiin as
ssess
smen
nt

Sma
art ph
hone
e (Wo
orkerr) GSM
M ne
etworrk S rt phone (Sup
Smar pervis
sor)

Data e
abase Su
upervisory staff

Repo
ort sym
mptom
ms of heat--relate
ed illn
nesses
s

Fig. 2. Process map for the early-warning system.

is the most popular and successful mobile cellular communication sys-
tem worldwide [78]. Due to its wide spread coverage, information
transmission among GSM based WBGT sensor, smart phone, and PC
can be achieved in the proposed system.
3. Material and methods
To construct the heat stress model, field studies were conducted on
six construction sites during the summer time in Hong Kong (July to
September in 2010 and 2011) to collect the necessary dataset.

2.3.3. Data processing layer 3.1. Dataset

The data processing layer is the core of the early-warning system
which analyzes the collected data and assesses workers' heat strain.
This layer consists of computer processing algorithms for heat strain as-
sessment (Fig. 4). Average heart rate of workers performing tasks is ob-
tained by statistical calculation. The mathematical model using ANN
was applied to predict the RPE of construction workers. ANNs are a
class of non-linear mathematical models that are characterized by a
complex structure of interconnected computational elements, which
have been proposed as an alternative to standard statistical techniques
for predicting complex phenomena in medicine and biological research
[79]. The heat stress assessment obtained in the data processing layer
indicates the current level of heat strain of construction workers effec-
tively and scientifically.
2.3.4. Display and control layer
The display and control layer is the highest layer in the architec-
ture of the early-warning system. The safety warnings based on the
heat stress assessment predict whether the psychological and psy-
chological status of workers would reach an unacceptable level,
and notify workers and supervisory staff directly and immediately.
Hence, frontline workers could take effective countermeasures as
suggested rather than working to exhaustion, and site supervisor/
foreman and safety professionals can pay more attention to those
workers at risk.
Thirty-nine healthy and experienced construction rebar workers
participated in the field studies. During the field measurement, con-
struction workers performed tasks as per their usual daily work routine
and were asked to verbally report a RPE scoring for every 5 min
(Fig. 5A). A series of physiological parameters (e.g., energy expenditure,
heart rate, oxygen consumption) were continuously monitored and re-
corded during the test via a telemetry system (K4b2, COSMED, Rome,
Italy) for every 5 s (Fig. 5B). Alongside with the measurement of physi-
ological data, a heat stress monitor (QUESTemp°36, Oconomowoc, USA)
was used to measure and record the prevailing environmental data. The
heat stress monitor was used to measure prevailing environmental data
simultaneously at 1 min interval: dry bulb temperature, wet bulb tem-
perature, globe temperature from which the corresponding WBGT
index can be computed (Fig. 5C). The physiological and environmental
data, together with the self-reported RPE were then synchronized every
5 min for further analysis. A total of 550 sets of environmental, work-
related, and personal data together with the corresponding RPEs were
used in this study as dataset.
3.2. Artificial neural networks (ANNs)
ANN is a type of artificial intelligence technology which incorporates
complex data-analysis features into existing applications by an intelli-
gent, human-like application of knowledge [80]. Widely used in diverse
106 W. Yi et al. / Automation in Construction 62 (2016) 101–113

Sensing layer

Smart phone Smart bracelet Environmental sensor

Data collection

Transmitting layer

Bluetooth GSM network

D ta ana
Dat a aly
ysis
s

Da
ata
a prroc
ces
ssing
g la
aye
er

RP
PE pre cted
edic d by
y ANN
A N He
eartt ratte gra
g ph

Da
ata
a diiss
sem
min
nattion
n

D pla
Disp ay and
a dc ntrol lay
con yerr

Early warning

Intervention measures

Supervisory staff ……
Frontline worker

Fig. 3. Architecture of the early-warning system.

fields in science and industry, ANN is an adaptive system that changes
its structure based on external or internal information processed within
the network during the learning stage [37]. Among the ANN ap-
proaches, the backpropagation neural network (BPNN) is the most
classically and generally used training algorithm [81]. BPNN is mul-
tilayer feed-forward neural network that is trained by the error
backpropagation (BP) algorithms. BP learning method is an optimi-
zation procedure based on gradient descent that adjusts weights to
reduce the system error. In the learning phase, input patterns are
presented to the network in a particular sequence. Each training
pattern is propagated forward layer by layer until an output pattern
is computed. The computed output is then compared to a desired or
target output, and an error value is determined. The errors are prop-
agated backwards layer by layer, with corrections being made to the
corresponding layer weights in an iterative manner. The process is
repeated for several times for each pattern in the training set until
the total output error converges to a minimum or until a limit is
reached in the number of completed training iterations.
 W. Yi et al. / Automation in Construction 62 (2016) 101–113 107

Heat strain assessment

0 ≤ RPE ≤ 3 Yes Mostly Safe Yes HR ≤ 100 bpm

No No

4 ≤ RPE ≤ 6 Yes Caution Yes 100 ≤ HR ≤ 130 bpm

No No

7 ≤ RPE ≤ 8 Yes Warning Yes 130 ≤ HR ≤ 160 bpm

No No

9 ≤ RPE ≤ 10 Yes Yes HR≥ 180-Age last

Sever
for 3 min

Fig. 4. Algorithms for heat stress assessment.

A three-layered (namely, an input layer, a hidden layer, and an out-
put layer) BPNN model was developed to establish the relationship be-
tween RPE and heat strain factors. Therefore, the BPNN model contained
seven input neurons (i.e., WBGT, work duration, age, BMI, alcohol-
drinking habit, smoking habit, and job nature) and one output neuron
(RPE). No general rule exists in determining the number of neurons in
the hidden layer. The optimal number of hidden layer neurons was
determined by attempting various numbers of hidden neurons to deter-
mine the optimized network architecture. In terms of the sum-of-
squares error (MSE), the best performance of the BPNN model was
obtained for the configuration characterized by 11 neurons in the hid-
den layer. Therefore, the 7-11-1 neuron configuration was determined
as the optimal network architecture (Fig. 6). Table 3 lists the parameters
of the optimized BPNN. The BPNN model consists of 3-layered was de-
veloped using the Neural Network Toolbox of MATLAB. The data were
randomly divided into three groups as training data set (70%), valida-
tion data set (15%), and testing data set (15%) for predicting RPE.
3.3. Performance evaluation
Correlation coefficient (R2), mean absolute percentage error
(MAPE), and root-mean-square deviation (RMSE) were used to evalu-
ate the performance of the ANN prediction model. The higher R2 values
indicate greater similarities between the measured and predicted

Fig. 5. Field measurements on construction sites (A) construction worker was asked to verbally report a RPE scoring; (B) physiological parameters were continuously monitored and
recorded during the test via a telemetry system (K4b2, COSMED, Rome, Italy); (C) environmental data were measured and recorded by a heat stress monitor (QUESTemp°36,
Oconomowoc, USA).
108 W. Yi et al. / Automation in Construction 62 (2016) 101–113

Input layer Hidden layer Output layer

WBGT

Age

BMI

Alcohol drinking RPE

habit

Smoking habit

Work duration

Job nature

Fig. 6. The ANN architecture used as the prediction model for RPE.

vffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffiffi
values, and the lower MAPE and RMSE values represent more accurate u N
u1X
prediction results. The R2, and MAPE, RMSE values were calculated RMSE ¼ t ðt −s Þ2 ð3Þ
using Eqs. (1)–(3), respectively. N i¼1 i i

XN
2
ðti −si Þ2 where ti is the measured (experimental) values, si is the predicted
R ¼ 1− Xi¼1 ð1Þ
N
ðt −sÞ2 values, N is the total number of data and s is the average of predicted
i¼1 i
values

N  !
1 X 
ti −si   100
4. Results
MAPE ¼ ð2Þ
N i¼1  ti 
An ANN model using Levenberg–Marquart back propagation was
used on the database. This model was trained using 384 randomly se-
Table 3 lected samples; the remaining 166 samples were equally divided for
Optimized backpropagation neural network (BPNN) parameters. the ANN validation and testing process. All performance measure values
confirmed a good-fit and a robust model. As shown in Fig. 7, the values
Error goal 0.02%
obtained through the training, validation, and testing of ANN model are
Learning rate 0.2
identical with measured data, implying the input parameters were
Iterations 50,000
Transfer function of hidden layer Hyperbolic tangent sigmoid strongly correlated with the output parameter. The determinant coeffi-
Number of input neurons 7 cient (R2) between the measured and predicted values is another
Number of hidden neurons 11 important indicator in checking the validity of the prediction models.
Number of output neurons 1 As the R2 values approach 1, prediction accuracy increases, which
Training algorithms Levenberg–Marquardt back propagation
shows a consistent agreement between the measured results and the
 W. Yi et al. / Automation in Construction 62 (2016) 101–113 109

(A) (B)
Training: R=0.98264 Validation: R=0.9525
8 8
Data Data
7 Fit 7 Fit
Y=T Y=T
6

Predicted RPE
Predicted RPE
6
5
5
4
4
3
3
2

2
1
1 2 3 4 5 6 7 8 2 3 4 5 6 7 8
Measured RPE Measured RPE

(C) (D)
Test: R=0.95798 All: R=0.97308
8 8
Data Data
7 Fit 7 Fit
Y=T Y=T
6 6
Predicted RPE
Predicted RPE

5 5

4
4

3
3
2
2
1
1
1 2 3 4 5 6 7 8 1 2 3 4 5 6 7 8
Measured RPE Measured RPE

Fig. 7. Relationship between the measured values and RPE values predicted by the ANN model for training data (A), validation data (B), testing data (C), and all data (D).

prediction results. R2 values in training, validation, and testing data set
for RPE are 0.966, 0.907, and 0.918, respectively. Therefore, the R2 values
obtained in this study imply that the designed model is capable of
explaining at least 90% of the measured data.
The ANN model exhibited a sufficient accuracy level in the predic-
tion of RPE values. Prediction values were determined with very low
percentage errors. MAPE is a dimensionless statistics that provides an
effective way of comparing the residual error for each data point with
respect to the observed or target value. Smaller values of MAPE indicate
better performance of the model and vice versa. MAPE values are con-
sidered as the most important performance criterion in decision making
[82]. In this study, MAPE values were 0.814%, 1.795%, and 1.283% in the
prediction of RPE values for training, validation, and testing data sets, re-
spectively. RMSE statistics compares the observed values to the predict-
ed values and computes the square root of the average residual error. A
lower value of RMSE indicates good prediction performance of the
model. In this study, RMSE values were 0.565, 1.024, and 0.857 in the
prediction of RPE values for training, validation, and testing data sets,
respectively.
5. Validation of the early-warning system
In order to validate the early warning system, a controlled experi-
ment was conducted during the summer time in a construction site in
Hong Kong. A 45 years old sheet metal worker with BMI of 23.3%, who
smokes and drinks occasionally, was invited to participate. An app of
the proposed system was built on iPhone platform (Fig. 8A). Prior to
the test, the worker input the personal information (i.e., age, height,
weight, smoking habit, and alcohol drinking habit) and job nature into
the system (Fig. 8B).
Assisted-GPS (A-GPS) was used to track the locations of the worker
because: (1) A-GPS is by far the most accurate of the three systems
(i.e. A-GPS, WiFi positioning, and Cellular network positioning) on
iPhone; (2) the smart phone has GPS chips that use both American
and Russian satellite systems giving increased accuracy; (3) the experi-
ment was conducted in an outdoor environment thus the antennae
had a clear view of the sky; and (4) the accuracy of A-GPS ranges from
5–8 m which was fairly accurate to distinguish the differences of the en-
vironmental data [83]. The worker was asked to perform tasks in three
different work zones (Fig. 9). In Fig. 10, three clusters were figured out
from the trajectory data, which were aligned with the predefined
work zones. Fig. 11 shows the heart rate and RPE at different levels of
physiological and psychological conditions respectively. It can be seen
in Fig. 10 that at the 100 min an alert of “orange” warning with sug-
gested strategies to take breaks in the shade, rectify ventilations or
cool down facilities was sent. During the prolonged operation, the
worker also reported heat-related disorders of light-headedness, weak-
ness, and fatigue via the system at the 95 min.
The preliminary results revealed the reliability of the proposed
early-warning system in detecting the extent of heat stress that
workers are subject to; together with appropriate intervention strat-
egies to relieve the harmful impacts of heat stress. Nevertheless, the
110 W. Yi et al. / Automation in Construction 62 (2016) 101–113

Fig. 8. Prototype of the proposed early-warning system (A) main interface of the app; and (B) personal information (i.e., age, height, weight, smoking habit, and alcohol drinking habit) and
job nature can be manually input by the worker.

early-warning system shows limitations that should be addressed to 6. Discussion

fully validate the proposed algorithm. For example, the smart brace-
let for monitoring heart rate has suffered a transient malfunction. This study adopted RPE as a yardstick to evaluate the risk of heat
stress that the construction workers suffered. Many reports have been
published concerning RPE during exercise in the heat. RPE is regarded

Fig. 9. Three predefined work zones for construction task execution. Fig. 10. Recorded GPS location data and trajectory information.
 W. Yi et al. / Automation in Construction 62 (2016) 101–113
160
150
140
130
120
HR
110
100
90
80
70
60
50
1 5 9 13 17 22 26 30 34 38 42
Fig. 11. Heart rate (HR) and RPE at different levels of physiological and psychological conditions respectively.
as an essential indicator to investigate the physical strain in hot environ-
ments. Previous studies of exercise performance in the hot environments
demonstrated that the RPE is linearly correlated with heat strain indicator
(e.g., body temperature, heart rate, sweating rate) [84,85], maximal
values of RPE appears when the core temperature reaches approximately
40 °C [84-86]. It has been suggested that RPE is more than an indicator of
exercise exertion, it is also a crucial component of preventing the exercis-
ing athlete from continuing exercise and causing bodily harm, by being
responsible for the decision volitionally to terminate the exercise bout
[87]. The highest RPE usually accompanied by symptoms of heat illness
including high skin/core temperature [84,85], muscle glycogen depletion
[88], or dehydration [89].
In Hong Kong, the incidence of heat stress in the construction indus-
try has been alarming and caused a number of verifiable reported
deaths [90]. However, the number of accidents and injuries caused by
heat-related illnesses in the construction industry of Hong Kong has
not been monitored by the relevant authorities. To foster better assess-
ment of heat-related injuries, this early-warning system could monitor
heat-related incidents and injuries. Incidents of heat-related injuries
can be further reported to the Labour Department and other relevant
government departments in a similar format as other types of occupa-
tional injuries. Detailed information of the victim, such as age, gender,
smoking and drinking habits; and information about the incident,
such as time, location, WBGT at the time of incident, can be recorded
and reported to the relevant departments for proper record and analy-
sis. These statistics can be monitored and disseminated to the public
through various relevant government departments and statutory orga-
nizations, such as the Labour Department, the Department of Census
and Statistics, the Occupational Safety and Health Council, and the Con-
struction Industry Council.
A limitation of the study is the single sample source and the limited
sample size. In this study, participants were mainly steel bar bender and
fixers. Further research work should be done to increase the sample size
and to replicate the experimentation to different trades. The current
study deals with the work intensity from the perspective of physical
demands of construction works in different trades. It is assumed that
construction workers in the same trades will undertake the same work-
loads. However, this may not always be true. For example, the workload
of a rebar worker fixing steel bars to column is higher than that of a
rebar worker measuring the dimensions of re-bars. The likelihood of
111
10
HR (bpm) RPE
9
8
7
6
RPE
5
4
3
2
1
0
46 50 54 58 62 66 70 74 78 82 86 90 94 98 102
Time (min)
underestimating or overestimating workload has been noted and recog-
nized as a potential limitation. Considering the rights of worker in the
design and use of algorithms, the research environment adapts widely
used rules to collect and analyze data of personnel [91]. The application
of developed system should take into account the legal restrictions to
record personal data, especially from humans in countries where per-
sonal rights are potentially not valued as high enough, or the literacy
level of workers/power of unions is not as high enough to understand
the consequences of using tracking/monitoring technology.
7. Conclusions
Construction workers are at a high risk of exposure to heat stress as
they undertake physically demanding tasks. This study has established
an algorithm to develop an early-warning system against hot and
humid climates, which can alert workers to occupational heat stress
and provide intervention measures as well. In this regard, the study con-
tributes to literature by filling the research gap arising from the limited
number of studies in safety evaluation in hot and humid environment.
This study adopted two approaches to improving construction safety
management. One is from the perspective of management, focusing
on human behavior and hazard management. The other approach fo-
cused on how to apply advanced technology in the construction indus-
try. Furthermore, the current study affirms that ANN method can be
applied in predicting the level of heat strain of workers who need to
work in an outdoor environment.
The early-warning system involves (1) collecting timely information
and undertaking risk assessments of heat stress; (2) generating an accu-
rate and timely warning to trigger prompt health and safety interven-
tion; and (3) disseminating heat strain assessment and symptoms of
heat illness to site supervisor/foreman. The established system has
other functions to support risk analysis, warning sign, and response ca-
pability. Supervisory staff and safety professionals may use this system
as a surveillance method to safeguard the health and safety of frontline
workers when working in a hot and humid condition.
Workers in different climates or industries may have different
degrees of susceptibility to heat stress. The current early warning sys-
tem was derived from the field data collected from the Hong Kong
construction industry. Further investigation should be conducted to ex-
amine other climates or other industries to provide a holistic view. The
112 W. Yi et al. / Automation in Construction 62 (2016) 101–113
findings of these studies will be of great significance in enhancing
worker performance and protecting their safety and health, as well as
in reducing the occurrence of heat illnesses.
Acknowledgments
This project is funded by two grants from the Research Grants Coun-
cil of the Hong Kong Special Administrative Region, China (RGC Project
No. PolyU510409 and PolyU510513). The support from the Hong Kong
Polytechnic University's Institute of Textiles and Clothing (ITC) is deeply
appreciated. The research team is also indebted to the technical support
from technicians of the Hong Kong Polytechnic University and the Hong
Kong Institute of Education. In particular, the participation of volunteers
in this experimental study is gratefully acknowledged. This paper forms
part of the research project titled “Experimental research on health and
safety measures for working in hot weather”, from which other deliver-
ables will be produced with different objectives/scopes but sharing
common background and methodology. The authors also wish to ac-
knowledge the contributions of other team members including Dr Mi-
chael Yam, Dr Daniel Chan, Prof Joanne Chung, Prof Esmond Mok, Dr
Geoffrey Shea, Dr Min Wu, Dr Herbert Biggs, Dr Donald Dingsdag, and
Miss Alice Guan.
References
[1] International Labour Organization (ILO), Facts on safety at work, International
Labour Office, Geneva, 2005 (accessed on 3 December 2014 at http://www.ilo.org/
wcmsp5/groups/public/—dgreports/—dcomm/documents/publication/wcms_
067574.pdf).
[2] ILO, The health of workers in selected sectors of the urban economy: challenges and
perspectives, International Labour Office, Geneva, 2013 (accessed on 3 December
2014 at http://www.ilo.org/wcmsp5/groups/public/—ed_dialogue/—sector/docu-
ments/publication/wcms_208090.pdf).
[3] European Agency for Safety and Health at Work, OSH in figures: work-related
musculoskeletal disorders in the EU — facts and figures, Publications Office of the
European Union, Luxembourg, 2010 (accessed on 3 December 2014 at https://
osha.europa.eu/en/publications/reports/TERO09009ENC).
[4] C.D. Reese, J.V. Edison, Handbook of OSHA construction safety and health, second ed.
CRC Press, Taylor & Francis Group, 2006.
[5] J.C. Semenza, J.E. McCullough, W.D. Flanders, M.A. McGeehin, J.R. Lumpkin, Excess
hospital admissions during the July 1995 heat wave in Chicago, Am. J. Prev. Med.
16 (1999) 269–277.
[6] W. Yi, A.P.C. Chan, Effects of temperature on mortality in Hong Kong: a time series
analysis 59 (7) (2015) 927–936.
[7] G.B. Anderson, M.L. Bell, Heat waves in the United States: mortality risk during heat
waves and effect modification by heat wave characteristics in 43 U.S. communities,
Environ. Health Perspect. 119 (2011) 210–218.
[8] K. Knowlton, M. Rotkin-Ellman, G. King, H.G. Margolis, D. Smith, G. Solomon, et al.,
The 2006 California heat wave: impacts on hospitalizations and emergency depart-
ment visits, Environ. Health Perspect. 117 (2009) 61–67.
[9] R.A. Lucas, Y. Epstein, T. Kjellstrom, Excessive occupational heat exposure: a signifi-
cant ergonomic challenge and health risk for current and future workers, Extreme
Physiol. Med. 3 (2014) 14.
[10] W. Yi, A.P.C. Chan, Critical review of labor productivity research in construction
journals, J. Manag. Eng. 30 (2) (2014) 214–225.
[11] A.P.C. Chan, W. Yi, D.P. Wong, M.C.H. Yam, D.W.M. Chan, Determining an optimal re-
covery time for construction workers after working to exhaustion in a hot and
humid environment, Building and Environment 58 (2012) 163–171.
[12] N.A.S. Taylor, Challenges to temperature regulation when working in hot environ-
ments, Ind. Health 44 (2006) 331–344.
[13] D. Bonauto, R. Anderson, E. Rauser, B. Burke, Occupational heat illness in
Washington State, 1995–2005, Am. J. Ind. Med. 50 (12) (2007) 940–950.
[14] D.B. Petitti, S.L. Harlan, G. Chowell-Puente, D. Ruddell, Occupation and environmen-
tal heat-associated deaths in Maricopa County, Arizona: a case-control Study, PLoS
One 8 (5) (2013), e62596.
[15] S. Horie, Prevention of heat stress disorders in the workplace, J. Jpn. Med. Assoc. 141
(2) (2012) 289–293.
[16] Apple Daily, construction worker died from heat stroke in a falling accident (in
Chinese), 2010 (accessed on 19 September 2010 at http://hk.apple.nextmedia.
com/realtime/art_main.php?iss_id=20100918andsec_id=6996647andart_id=
14465015).
[17] Hong Kong Daily News, 28 percent of construction workers in Hong Kong have suf-
fered heat related illness (in Chinese), 2012 (assessed on 19 July 2012 at http://
www.hkdailynews.com.hk/index.php).
[18] R.T. Lin, C.C. Chan, Effects of heat on workers' health and productivity in Taiwan,
Global Health Action, 22009 (10.3402/gha.v2i0.2024).
[19] D. Joubert, J. Thomsen, O. Harrison, Safety in the heat: a comprehensive program for
prevention of heat illness among workers in Abu Dhabi, United Arab Emirates, Am. J.
Public Health 101 (3) (2011) 395–398.
[20] Construction Industry Council, Hong Kong. Guidelines on site safety measures for
working in hot weather, Publication Version 22013 (Available online: http://
www.hkcic.org/eng/info/publication.aspx?langType=1033&id=4458).
[21] Department of Health, HKSAR Government, Preventive measures against heat
stroke and sun burn, 2010 (accessed on 6 July 2010 at http://www.dh.gov.hk/en-
glish/press/2010/100706-3.html).
[22] Health and safety executive and the health departments in England, Scotland and
Wales, keep your top on health risks from working in the sun, 2011 (assessed on
7 November 2014 at http://www.hse.gov.uk/pubns/indg147.pdf).
[23] US Department of Labor, Heat illness prevention training guide, Occupational Safety
and Health Administration2010 (assessed on 13 October 2014 at https://www.osha.
gov/SLTC/heatillness/osha_heattraining_guide_0411.pdf).
[24] S.L. Lu, N. Zhu, Experimental research on physiological index at the heat tolerance
limits in China, Build. Environ. 42 (12) (2007) 4016–4021.
[25] M.N. Sawka, A.J. Young, W.A. Latzka, P.D. Neufer, M.D. Quigley, K.B. Pandolf, Human
tolerance to heat strain during exercise: influence of hydration, J. Appl. Physiol. 73
(1992) 368–375.
[26] S.S. Cheung, T.M. McLellan, Heat acclimation, aerobic fitness, and hydration effects
on tolerance during uncompensable heat stress, J. Appl. Physiol. 84 (1998)
1731–1739.
[27] D.J. Brake, G.P. Bates, Limiting metabolic rate (thermal work limit) as an index of
thermal stress, Appl. Occup. Environ. Hyg. 17 (3) (2002) 176–186.
[28] American conference of government industrial hygienists, thermal stress, Threshold
limit values for chemical substances and physical agents and biological exposures
indices, ACGIH. Cincinnati, OH, 2007.
[29] ISO 7243, hot environments—estimation of the heat stress on working man, based
on the WBGT-index (wet bulb globe temperature), International Standards Organi-
zation, Geneva, Switzerland, 2003.
[30] National Institute for Occupational Safety and Health (NIOSH), Criteria for a recom-
mended standard: occupational exposure to hot environments revised criteria,
NIOSH Publication No. 86-113, Cincinnati, OH, 1986.
[31] J.L. Zhou, Z.H. Bai, Z.Y. Sun, A hybrid approach for safety assessment in high-risk
hydropower-construction-project work systems, Saf. Sci. 64 (2014) 163–172.
[32] Z.G. Ren, G.Z. Liu, C. He, S.W. Wang, Study on the indicators for ranking of thermal
hazard in heat stress workplace of underground mine, J. Saf. Sci. Technol. 5 (3)
(2009) 41–46.
[33] G. Zheng, N. Zhu, Z. Tian, Y. Chen, B. Sun, Application of a trapezoidal fuzzy AHP
method for work safety evaluation and early warning rating of hot and humid envi-
ronments, Saf. Sci. 50 (2012) 228–239.
[34] K.E. Weick, K.M. Sutcliffe, D. Obstfeld, Organizing for high reliability: process of
collective mindfulness, in: R.S. Sutton, B.M. Staw (Eds.), Research in Organizational
Behavior, Jai Press, Stanford 1999, pp. 81–123.
[35] A.P.C. Chan, W. Yi, D.W.M. Chan, D.P. Wong, Using the thermal work limit as an en-
vironmental determinant of heat stress for construction workers, J. Manag. Eng. 29
(4) (2013) 414–423.
[36] A.P.C. Chan, M.C.H. Yam, J.W.Y. Chung, W. Yi, Developing a heat stress model for
construction workers, J. Facil. Manag. 10 (1) (2012) 59–74.
[37] J.B. Malchaire, Predicted sweat rate in fluctuating thermal conditions, Eur. J. Appl.
Physiol. 63 (3–4) (1991) 282–287.
[38] S. Rowlinson, A.Y. Jia, B. Li, C.C. Ju, Management of climatic heat stress risk in con-
struction: a review of practices, methodologies, and future research, Accid. Anal.
Prev. 66 (2014) 187–198.
[39] P.A. Atanassova, B.D. Dimitrov, Recent advances and challenges in the application of
artificial neural network (ANN) in neurological sciences: an overview, in: S.J. Kwon
(Ed.), Artificial Neural Network, Nova Science Publishers, Inc., New York 2011,
pp. 61–100.
[40] A.R. Khataee, M.B. Kasiri, Artificial neural network modeling of water and wastewa-
ter treatment processes, in: S.J. Kwon (Ed.), Artificial Neural Network, Nova Science
Publishers, Inc, New York 2011, pp. 1–60.
[41] G. Migliaccio, U. Gatti, J. Teizer, Remote construction worker location, activity and
safety, Proceedings of the 28th International Symposium on Automation and Robotics
in Construction, ISARC 2011, pp. 650–651.
[42] T. Cheng, G.C. Migliaccio, J. Teizer, U.C. Gatti, Data fusion of real-time location sens-
ing and physiological status monitoring for ergonomics analysis of construction
workers, J. Comput. Civ. Eng. 27 (3) (2013) (0 320-5).
[43] J. Teizer, T. Cheng, Y. Fang, Location tracking and data visualization technology to ad-
vance construction ironworkers' education and training in safety and productivity,
Autom. Constr. 35 (2013) 53–68.
[44] L. Joshua, K. Varghese, Selection of accelerometer location on bricklayers using deci-
sion trees, Comput.-Aided Civ. Infrastruct. Eng. 28 (2013) 372–388.
[45] X. Wang, Using augmented reality to plan virtual construction worksite, Interna-
tional Journal of Advanced Robotics System 4 (4) (2007) 501–512.
[46] X. Wang, M. Truijens, L. Hou, Y. Wang, Y. Zhou, Integrating Augmented Reality with
Building Information Modeling: Onsite construction process controlling for liquefied
natural gas industry, Automation in Construction 40 (2014) 96–105.
[47] J. Wang, X. Zhang, W. Shou, X. Wang, B. Xu, M.J. Kim, P. Wu, A BIM-based approach
for automated tower crane layout planning, Automation in Construction 59 (2015)
168–178.
[48] Hong Kong Observatory, Daily weather summary and radiation level, 2014
(Available online: http://www.weather.gov.hk/wxinfo/dailywx/dailywx.shtml).
[49] B. Yip, S. Rowlinson, Job burnout among construction engineers working within
consulting and contracting organizations, J. Constr. Eng. Manag. 25 (3) (2009)
122–130.
 W. Yi et al. / Automation in Construction 62 (2016) 101–113
[50] Australian Institute of Occupational Hygienists (AIOH), R. DiCorleto, G. Coles, I. Firth,
Heat stress standard & documentation developed for use in the Australian environ-
ment, AIOH, 2003.
[51] ISO 9886, ergonomics — evaluation of thermal strain by physiological measure-
ments, International Standards Organization, Geneva, Switzerland, 2004.
[52] T.E. Bernard, W.L. Kenney, Rationale for a personal monitor for heat strain, Am. Ind.
Hyg. Assoc. J. 55 (1994) 505–514.
[53] D. Schlagbauer, D. Heck, Change in output performance due to prolonged work, ISEC
2013 — 7th International Structural Engineering and Construction Conference: New
Developments in Structural Engineering and Construction 2013. , University of
Hawaii at Manoa, Honolulu; United States, June 2013 18–23.
[54] Z. Tian, N. Zhu, G. Zheng, H. Wei, Experimental study on physiological and psycho-
logical effects of heat acclimatization in extreme hot environments, Build. Environ.
46 (2011) 2033–2041.
[55] ACGIH, TLVs® and BEIs®: threshold limit values for chemical substances and phys-
ical agents and biological exposure indices, American Conference of Governmental
Industrial Hygienists, Cincinnati, OH, 2014 2014.
[56] G.A. Borg, Psychophysical scaling with applications in physical work and the percep-
tion of exertion, Scand. J. Work Environ. Health 16 (1) (1990) 55–58.
[57] M. Garcin, J. Vautier, H. Vandewalle, H. Monod, Ratings of perceived exertion (RPE)
as an index of aerobic endurance during local and general exercises, Ergonomics 41
(1998) 1105–1114.
[58] G. Borg, I. Lindblad, The determination of subjective intensities in verbal descrip-
tions of symptoms, Reports from the Institute of Applied PsychologyStockholm:
Univ Stockholm1976 75.
[59] S. Rowlinson, A.Y. Jia, B. Li, C.C. Ju, Management of climatic heat stress risk in con-
struction: a review of practices, methodologies, and future research, Accid. Anal.
Prev. 66 (2014) 187–198.
[60] N.L. Benowitz, P.I.I.I. Jacob, R.T. Jones, J. Rosenberg, Interindividual variability in the
metabolism and cardiovascular effects of nicotine in man, J. Pharmacol. Exp. Ther.
221 (2) (1982) 368–372.
[61] J.W. Garrett, J. Teizer, Human factors analysis classification system relating to human
error awareness taxonomy in construction safety, J. Constr. Eng. Manag. 135 (8)
(2009) 754–763.
[62] T. Yoda, L.I. Crawshaw, M. Nakamura, K. Saito, A. Konishi, et al., Effects of alcohol on
thermoregulation during mild heat exposure in humans, Alcohol 36 (2006)
195–200.
[63] G.M. Budd, Wet-bulb globe temperature (WBGT)—its history and its limitations, J.
Sci. Med. Sport 11 (2008) 20–32.
[64] WHO, Evidence for the effectiveness and cost–effectiveness of interventions to
reduce alcohol-related harm, WHO Regional Office for Europe, 2010 (accessed on
17 September 2014 at http://www.euro.who.int/document/E92823.pdf).
[65] T.K. Greenfield, Individual risk of alcohol-related disease and problems, in: N.
Heather, T.J. Peters, T. Stockwell (Eds.), International Handbook of Alcohol Depen-
dence and Problems, pp John Wiley and Sons, Ltd., Chichester, England 2001,
pp. 413–437.
[66] H. Wechsler, T.F. Nelson, Binge drinking and the American college student: what's
five drinks? Psychol. Addict. Behav. 15 (2001) 287–291.
[67] A.P.C. Chan, F.K.W. Wong, Y.P. Guo, W. Yi, M. Teng, J. Chung, D. Chow, et al., Prelim-
inary findings from pilot medical examination scheme: for construction workers in
Hong Kong, Aedifice J. Hong Kong's Constr. Ind. 2 (2015) 34–49.
[68] D.P. Wong, J. Chung, A.P.C. Chan, F.K.W. Wong, W. Yi, Comparing the physiological
and perceptual responses of construction workers (bar benders and bar fixers) in
a hot environment, Appl. Ergon. 45 (6) (2014) 1705–1711.
[69] T.S. Abdelhamid, J.G. Everett, Physiological demands during construction work, J.
Constr. Eng. Manag. 128 (5) (2002) 427–437.
113
[70] A. Adams, S. Bochner, L. Bilik, The effectiveness of warning signs in hazardous work
places: cognitive and social determinants, Appl. Ergon. 29 (4) (1998) 247–254.
[71] B. Matthews, R. Andronaco, A. Adams, Warning signs at beaches: do they work? Saf.
Sci. 62 (2014) 312–318.
[72] M. Kjertakov, Y. Epstein, Exertional heat stroke in athletes, Mace J. Med. Sci. 6 (4)
(2013) 473–477.
[73] A.S. Howe, B.P. Boden, Heat-related illness in athletes, Am. J. Sports Med. 35 (8)
(2007) 1384–1395.
[74] A. Bouchama, J. Knochel, Heat stroke, N. Engl. J. Med. 346 (2002) 1978–1988.
[75] E.E. Coris, S.M. Walz, R. Duncanson, A.M. Ramirez, R.G. Roetzheim, Heat illness
symptom index (HISI): a novel instrument for the assessment of heat illness in ath-
letes, South. Med. J. 99 (4) (2006) 340–345.
[76] J.D. Ramsey, F.N. Dukes-Dobos, T.E. Bernard, Evaluation and control of hot working
environments: part I — guidelines for the practitioner, Int. J. Ind. Ergon. 14 (1994)
119–127.
[77] K.C. Parsons, Heat stress standard ISO 7243 and its global application, Ind. Health 44
(3) (2006) 368–379.
[78] W. Gerstacker, R. Schober, R. Meyer, F. Obernosterer, M. Ruder, et al., GSM/EDGE:
a mobile communications system determined to stay, AEU Int. J. Electron. Commun.
65 (8) (2011) 694–700.
[79] J.L. Patel, R.K. Goyal, Applications of artificial neural networks in medical science,
Curr. Clin. Pharmacol. 2 (3) (2007) 217–226.
[80] S.J. Kwon, Artificial neural network, Nova Science Publishers, Inc., New York, 2011.
[81] H.H. Örkcü, H. Bal, Comparing performances of backpropagation and genetic algo-
rithms in the data classification, Expert Syst. Appl. 38 (4) (2011) 3703–3709.
[82] S. Tiryaki, A. Aydin, An artificial neural network model for predicting compression
strength of heat treated woods and comparison with a multiple linear regression
model, Constr. Build. Mater. 62 (2014) 102–108.
[83] P.A. Zandbergen, S.J. Barbeau, Positional accuracy of assisted GPS data from high-
sensitivity GPS-enabled mobile phones, J. Navig. 64 (03) (2011) 381–399.
[84] B. Nielsen, T. Hyldig, F. Bidstrup, et al., Brain activity and fatigue during prolonged
exercise in the heat, Pflugers Arch. 442 (2001) 41–48.
[85] L. Nybo, B. Nielsen, Perceived exertion is associated with an altered brain activity
during exercise with progressive hyperthermia, J. Appl. Physiol. 91 (2001)
2017–2023.
[86] R.J. Robertson, Central signals of perceived exertion during dynamic exercise, Med.
Sci. Sports Exerc. 14 (1982) 390–396.
[87] R. Tucker, The anticipatory regulation of performance: the physiological basis for
pacing strategies and the development of a perception-based model for exercise
performance, Br. J. Sports Med. 43 (2009) 392–400.
[88] J. Baldwin, R.J. Snow, M.J. Gibala, et al., Glycogen availability does not affect the TCA
cycle or TAN pools during prolonged, fatiguing exercise, J. Appl. Physiol. 94 (2003)
2181–2187.
[89] L.B. Baker, K.A. Dougherty, M. Chow, et al., Progressive dehydration causes a
progressive decline in basketball skill performance, Med. Sci. Sports Exerc. 39
(2007) 1114–1123.
[90] Y. Xu, A.P.C. Chan, An investigation of heat related incidents on construction
workers in Hong Kong, The 19th Cross-strait conference on occupational safety
and health, Hong Kong Occupational Safety and Health Association, 3 November
2011.
[91] J. Teizer, Status quo and open challenges in vision-based sensing and tracking of
temporary resources on infrastructure construction sites, Advanced Engineering
InformaticsSpecial Issue Infrastructure Computer Vision, 292015 225–238.