Real-Time Monitoring of Mental Fatigue of Construction Workers Using Enhanced Sequential Learning and Timeliness

Abstract

The demanding nature of construction works exposes workers to prolonged physical labor and high-risk environments, increasing their vulnerability to mental fatigue and consequently posing risks to safety and productivity. Over the years, a number of mental fatigue monitoring models have been created based on the captured vital signs. However, these models are basically based on batch learning algorithms such as machine learning or deep learning, which are time-consuming and costly, and fail to consider the timeliness features embedded in the biological data. Aiming at time-varying vital signs, a novel regularized online sequential extreme learning machine with dynamic forgetting factor (ROSELM-DFF) model was established to realize the real-time monitoring of workers' mental fatigue. A cognitive experiment has been carried out and the results indicate that the proposed OSELM-DFF model outperforms other models in terms of computational efficiency and prediction accuracy, achieving the best Average Accuracy of 96.106%. This study offers an effective solution for proactive management of workers' mental fatigue, which is expected to foster a safer and more productive construction environment.