Field Experiment on Wearable EEG/PPG-based Monitoring of Mental Fatigue in Drillers: Impact of Prolonged Working Hours and Night Shifts

Abstract

Mental fatigue among oil and gas drillers during prolonged equipment operations or night shifts often leads to lapses in attention, resulting in equipment-related accidents. Hence, it is crucial to develop efficient objective detection technologies for monitoring and identifying fatigue characteristics. During this research, a total of eight oil and gas drillers were under constant observation for a duration of 20 days, utilizing wearable EEG and PPG devices. Our objective is to enhance health and safety management for oil and gas drillers through the application of physiological measurement techniques and machine learning algorithms. EEG spectral analysis is frequently employed for monitoring mental fatigue, and EEG functional connectivity analysis can provide insights into the interactions between various regions of the brain. PPG monitoring revealed that heart rates decreased over time with longer hours and night shifts, though they were slightly higher during night shifts compared to day shifts, indicating a relationship between heart rate and mental fatigue. EEG spectral analysis revealed significant changes in α-band and frequency band ratios as mental fatigue increased. Functional connectivity analysis revealed a significant improvement in functional connectivity between the left frontal lobe and various other brain regions when experiencing fatigue. The machine learning model demonstrated its effectiveness in detecting fatigue conditions with an impressive overall accuracy rate of 93.46%. This field experiment validates the extraction of fatigue characteristics in real-world drilling scenarios using wearable EEG and PPG devices and establishes a mental fatigue monitoring model, providing valuable insights for future applications in other high-demand work environments.

Keywords: Mental Fatigue, wearable EEG, wearable PPG, Spectral Analysis, Functional Connectivity, Machine Learning