Measuring Effects of Working in High Temperatures on Construction Workers' Hazard Identification Using Physiological Measurements


Student thesis: Doctoral Thesis

View graph of relations



Awarding Institution
Award date11 Sept 2023


Poor hazard recognition levels in construction workplaces have been identified as a principal contributor to poor safety performance in the construction industry. Understanding why workers fail to identify hazards is effective in enhancing their identification abilities. Both personal and situational factors can influence workers’ hazard recognition ability. However, previous studies focused on personal issues such as experience and personality, lacking studies on situational factors, such as hot working conditions which is quite frequently in construction workplaces. Working under high temperatures might impair workers’ cognitions and thus leading to failed hazard identifications. Understanding such effects would be useful to provide insights for enhancing workers’ safety under hot workplaces. This study hence proposed to explore the effects of high temperatures on workers’ hazard identification, and to seek methods to detect risky workers.

Experimental studies were applied to achieve the research objectives. Firstly, the effects on workers’ cognitive states were examined. Twenty construction workers were invited to perform a hazard identification task respectively under a normothermic (25°C) and a hyperthermic (34°C) condition. Their attention ability, mental workload, mental stress, and mental fatigue were recorded by an electroencephalogram (EEG) device. The effects of high temperatures on these cognitions and the correlations between subjects’ behavioral performances and cognitions were analyzed. Meanwhile, the combined effects of high temperatures with noise and fatigue were also examined because workplace noise and physical fatigue are frequent in job sites and might amplify subjects’ thermal discomfort. There were three noise levels (60 dBA, 85 dBA, and 100 dBA workplace noise exposure) and three fatigue conditions (low, medium, and high). Secondly, the effects on the adopted visual search patterns were investigated because the search patterns might impact the identification performances. Twenty workers were required to search for hazards in 15 panoramic virtual construction sites. An eye tracker recorded their eye-movements to indicate their search patterns (including attention distribution, scanpaths, and search engagement). The experimental conditions (i.e., thermal, noise, and fatigue exposure) were the same as the EEG experiments. Last, mental states found to be impaired under high temperatures in the first research stage (mental workload, mental fatigue, and mental stress) were selected to investigate their monitoring methods for detecting risky workers. Directly measuring brain activities to know workers’ cognitive states are not applicable in job sites, so the feasibility of using electrocardiogram (ECG) and galvanic skin response (GSR) to replace EEG was examined. Thirty construction workers were invited to perform hazard identification tasks, with their mental states being manipulated. Their biological signals were collected and used as training data for developing prediction models. Supervised machine learning algorithms were applied, including Support Vector Machines (SVM), Random Forest (RF), KNearest Neighbor (KNN), and Linear Discriminant Analysis (LDA).

The results show that workers had poorer hazard identification accuracy under high temperatures. The heat stress impaired workers’ cognitions, with a significant increase in mental workload and mental fatigue. The missed hazards were correlated with worsening cognition. Besides, workers allocated less attention to hazardous areas in hyperthermic conditions, which might lead to ignorance of hazards. What is worse, the combined effects of high temperatures with noise/fatigue were more significant than the effects of high temperatures alone: the attention ability became significantly impacted when the combined noise level reached 100 dBA or fatigue level above medium; the changes in scanpaths increased with noise or fatigue levels; workers become less engaged in the hazard identification task when they suffer noise or medium/high fatigue levels in hot conditions, ending the searching more quickly and devoting fewer fixations within the construction scenes. As for predicting workers’ cognitive states, the accuracy of frequency-domain of heart rate variability (HRV) features extracted from ECG data can be up to 91.67% for mental fatigue, 96.67% for mental stress, and 86.67%~91.67% for differentiating mental workload levels. The prediction performances are even slightly better than EEG in classifying the levels of mental workload. GSR signals have a relatively poorer prediction performance. It obtains an accuracy of 80.00% in predicting mental fatigue; however, the accuracy is too low in mental stress (60.00%) and mental workload (no more than 70.00%). It indicates that collecting ECG signals is promising in monitoring workers’ mental states at job sites.

This study firstly examined the effects of hot workplaces on construction workers’ hazard identification. The findings could help understand why construction workers fail to detect environmental hazards when suffering heat stress, which could provide insights for the industry to enhance workers’ hazard identification abilities. Besides, revealing the combined effects when there are workplace noise and physical fatigue in hot workplaces could provide insights for implementing effective safety management measures in hot working conditions. In addition, this study found that the HRV features extracted from ECG data achieved satisfying prediction performances in monitoring workers’ cognitive states. It provides a promising approach for monitoring workers’ safety and detecting risky subjects in job sites.

    Research areas

  • physiological measurements, hazard identification, cognitive states, visual search patterns, construction safety