| description abstract | The construction industry is one of the most hazardous sectors worldwide, with extremely high rates of occupational deaths and injuries. Worker fatigue, caused by undertaking physically demanding tasks in awkward working postures over prolonged daily durations, has been recognized as the main cause of these accidents. Additionally, fatigue can lead to reduced work efficiency and increased absenteeism, undermining labor productivity. This study aims to develop an accurate and reliable model to estimate the fatigue levels of construction workers. Field studies were conducted with 156 construction workers at four construction sites in mainland China. A series of physiological, personal, work-related, and environmental factors were measured and monitored to establish an interpretable machine learning model for assessing fatigue levels. The developed interpretable machine learning model exhibited good fitting with high accuracy, evidenced by the random forest model attaining an R2 value of 0.9953 through the 10-fold cross-validation method. Furthermore, this model could transparently reveal the mechanisms underlying the prediction of worker fatigue. Work duration, work session (i.e., morning session, afternoon session), environmental parameters (i.e., air temperature, humidity, wind velocity, and radiation), and worker age were identified as key factors affecting the fatigue of construction workers. The developed fatigue model can prevent excessive fatigue among construction workers, and the model interpretation results may benefit the industry by making solid guidelines and practice notes to alleviate worker fatigue. To prevent excessive fatigue among construction workers, this study developed an interpretable machine learning model to assess the fatigue levels of workers during work. The developed model not only accurately evaluates the fatigue status of workers, but also reveals key fatigue-influencing factors. This dual capability facilitates the practicality and acceptability of the model in real construction settings. By assessing real-time fatigue levels, the model can prevent workers from reaching exhaustion thresholds, thereby mitigating the risk of hazardous accidents. The identified key fatigue-influencing factors can support the formulation of more targeted management measures, focusing on aspects such as work duration, work session, environmental conditions, and worker age, to effectively alleviate fatigue. In future research, the fatigue model could be considered integrated into smartphone applications to issue timely fatigue warnings, enhancing fatigue management on construction sites. The fatigue assessments provided by the model can also be considered to offer decision-making insights for optimizing work assignments. Overall, the development of this fatigue model contributes to fostering sustainable work practices among construction workers, thereby enhancing their occupational health and safety. | |
