1. Introduction
Globally, there has been increased concern regarding workers’ safety and health because of many cases of illness and accidents at work [
1,
2,
3,
4,
5,
6,
7,
8,
9,
10,
11,
12,
13,
14,
15,
16,
17]. In the construction industry, which mainly takes place outside and requires high levels of physical activity, the number of casualties is the highest among all industries [
18,
19,
20]. According to a 2018 annual report from the Korea Occupational Safety and Health Agency, both fatal and non-fatal accidents occurred the most in the construction industry (26.6% and 27.1%, respectively). These ranks have been maintained over the last 10 years [
21]. Likewise, in the United States, the mortality rate in the construction industry showed an upward trend from 2011 to 2015; construction is considered to be one of the most hazardous industries [
22].
In addition to accidents on the jobsite, Korean construction workers have high tendencies to suffer from occupational illnesses, such as heat-related illness due to their work environment. The Centers for Disease Control and Prevention (CDC) recognizes heat stress as an element of occupational illness. Certain industry sectors are at higher risk, including construction workers, who are often exposed to extreme heat or work in hot environments [
23]. In fact, according to a heat-related illness surveillance in 2018 conducted by the Korea Centers for Disease Control and Prevention (KCDC), 4526 people suffered from heat-related illness (48 deaths), and 28.1% of these patients were engaged in outdoor work. Moreover, in Korea, an average of 503 reported cases of heat-related illness have occurred annually since 2011 [
24]. Previous studies have found that construction workers in the United States have 13 times the risk of heat-related illness or death compared with other industries, and the increasing frequency of these illnesses presents growing concerns about the occupational safety of workers [
8,
25]. These statistics demonstrate the vulnerability to environmental hazards faced by construction workers and the need to provide effective solutions to reduce incidents.
Traditionally, construction workers’ health conditions have been measured by subjective methods, such as surveys. However, because of the nature of self-reported questionnaires, these measurements may be partially biased. Additionally, surveys typically take place during break times; therefore, they may not reflect a worker’s condition while they are active. Furthermore, stopping workers during their tasks in order to complete a questionnaire hinders their work and workers may respond carelessly [
26]. Therefore, in order to improve the reliability of the test method, sensor-type continuous measurement is encouraged in the construction field, where physical demands vary depending on the time, task type, and work conditions [
5]. Originally, wireless body area sensor networks (WBSNs) were developed for patients recovering from an incident and/or those with chronic diseases that required continuous health monitoring. The Occupational Safety and Health Administration (OSHA) defines heat strain as a body’s physiological response to heat stress; naturally, responses include an increased heart rate and sweating. When these responses function improperly, they lead to an elevated core temperature, which may cause heat-related illness or death [
27]. Moreover, the International Standards Organization (ISO) has indicated that responses to heat strain include a loss of body mass through sweating, and an increase in body core temperature, skin temperature, and heart rate [
28]. For reactions to physiological strain, the heart rate is mainly measured as it is the earliest response [
29].
As sensor technologies have improved, attempts have been made to adopt WBSNs in a variety of other areas, including the construction industry. For example, WBSNs in construction sites can provide effective tools to locate a worker’s position in a large construction site and to track assets in real time. However, the transmitting coverage and accuracy, which can be affected by environmental objects, facilities, the human body, etc., can still be improved [
12,
14,
30]. Although previous results have showed a high accuracy when identifying unsafe acts and conditions, some methods require multiple pieces of equipment. Additionally, in dynamic conditions, like at a construction site, technical obstacles are inevitable [
31,
32,
33]. Recently, based on the fact that wristband-type wearable devices and smart watches can be used to locate one’s position or to check on their physiological status, multiple attempts have been made to replace traditional physiological test methods [
1,
2,
4,
5,
34]. However, most studies related to the construction field have been conducted in lab conditions.
This platform collects a worker’s physiological data through a wearable armband that consists of three sensors. The three sensors of the armband are PPG, temperature, and accelerometer sensors. The PPG sensor measures the volume of blood flow by detecting changes in the intensity of the reflected light. The accelerometer sensor provides the current position of a worker, while the thermometer transmits the current skin temperature of a worker. The hardware consists of a microcontroller (MCU), GPS module, low-power wide-area network (LoRa) module, and power-supply, all of which are embedded into a single circuit board. Data from the MCU form a worker’s current physiological status, which is directly sent to the web and to an smartphone application via the LoRa network for visualization. We used a testbed (the Gyeongbu Expressway Straightening Project, Korea Expressway Corp and Han-ra Corp, Whasung City, Korea) for the performance evaluation.
3. Validation Tests and Results
To validate the developed platform, two types of validation tests were conducted—indoor and outdoor. The indoor test was conducted to compare the transmitted data from the device with the results from the graded exercise test (GXT), which is the most widely used method to study the relationship between exercise and physiological systems [
54]. The measurements were recorded by increasing the slope by 2.5% every minute at a speed of 5.1 km/h, also known as the modified Balke protocol. Three subjects were tested with different devices in case of inferior products. The characteristics of the subjects are described in
Table 7. The results shown in
Table 8 are the average data of the three trials for each subject. The percentage of error between GXT and PWB-300 was found to be less than 1%.
The outdoor experiment was set up in a construction site. The performance evaluation was conducted on the Gyeongbu Expressway Straightening Project, Korea Expressway Corp and Han-ra Corp, Korea. The project was 4.70 km in length, including a 10-lane road in both directions, and was located in Wha-sung City, Korea (
Figure 11). Physiological monitoring data were collected from 18 workers during the summer period of the project (June to August).
The goal of the study was to perform an initial implementation of the system so as to monitor the abnormality under high temperatures for workers in a construction site. Physiological monitoring data were collected for three months. During this period, 18 workers (ranging in age from their 30s to their 50s) were invited to participate. Prior to the test, the workers input their personal information (i.e., age, height, weight, blood type, and normal blood pressure) and the managers input the intensity of the work considering a worker’s daily tasks (
Figure 10b). As shown in
Figure 12, in the test bed, skin temperature measurements were taken using a Polygreen KI-8280 infrared thermometer (KI-8280, Seoul, Korea), and data from the PPG sensor were compared with measurements from a Rossmax automatic heartbeat monitor (BI701, Greencross Medical Science Corp., Yongin, Korea). The comparison results showed a 95.7% accuracy.
In some cases, the risk level in PMS indicated an “attention” level during the break time, while the TCI indicated the “concern” level. Onsite managers conducted interviews, and a lack of sleep and the previous night’s drinking were reported by those workers. The effects and/or correlations between a worker’s heartbeat rate and their sleep or alcohol consumption were not taken into account in this study. The preliminary results showed the reliability of the proposed platform in detecting abnormality based on physiological data via the LoRa network and by acquiring OHS in real time. Considering the nature of the system, it is important to conduct qualitative analyses based on user experiences; in this case, field managers were interviewed. The authors interviewed the field managers who operated the platform during the test period. Interviewees were satisfied with the features of the system, especially with regard to the ability to determine the individual risk level.