Automated detection of physical fatigue in transportation maintenance workers through physiological and motion data

Lifting bags of dry concrete mix has caused significant physical fatigue and work-related musculoskeletal disorders (WMSDs) in transportation maintenance workers. Detection of physical fatigue can effectively prevent developing WMSDs. However, there lacks studies exploring the automated detection of physical fatigue in transportation activities. Additionally, previous efforts on physical fatigue detection, primarily focused in the construction sector, may not be directly applicable to transportation activities, due to different activity duration, intensity, and frequency. Also, prior detection methodologies have been examined under laboratory conditions, and their efficacy in real-world environments remains unverified. Therefore, this study explored physical fatigue detection methods by conducting field experiments with 29 transportation maintenance workers lifting bags of dry concrete mix in a highway maintenance facility. Participants' physical fatigue levels were measured using electromyography (EMG), electrodermal activity, and/or heart rate (HR) and motion data with different machine learning algorithms. Results indicated that the Fully Connected Neural Networks algorithm can achieve the highest accuracy (87.84%) with EMG, HR, and motion data for classifying physical fatigue levels in transportation maintenance workers. This study demonstrated the feasibility of automated physical fatigue detection, which can be implemented as a monitoring tool of physical fatigue for enhancing health of transportation practitioners.