Heavy equipment machinery operator biosignals

Stress is an inseparable aspect of work, and its constructive and destructive effects on efficiency have been widely studied. This study aims to identify stress in high-risk work environments in the early stages to avoid damage, particularly in scenarios involving heavy-duty robots in factories and power plants. The primary objective is to develop an AI algorithm for stress detection and to evaluate the impact of stress on human factors and system performance in a man-machine loop, specifically for remote maintenance tasks in the ITER fusion power plant, a recognized high-risk environment.As this robot and ITER is under development, we focused on our new developed construction excavator simulator to expose stress for recording data. Investigations on stress identification were conducted on Heavy Equipment Operators (HEOs) using wearable biosensors during simulated construction tasks. Bio-signals such as Electroencephalography (EEG) and Electrocardiography (ECG) were recorded while HEO experts performing various tasks using the excavator simulator. Eight different tasks were exposed to 10 expert HEOs to simulate different levels of stress. Stress was categorized into three levels: low, medium, and high, alongside a baseline rest state. To classify three stress levels and rest, a deep learning algorithm with three configurations of EEG, ECG and a combination of EEG and ECG features were developed. The combined feature space obtained an average accuracy of 94.50% +-0.8. Additionally, power spectral brain maps in different EEG frequency bands were analyzed to identify neurons activation areas during different stress levels. The results indicate that low level of stress enhance mental concentration and positively improve work performance. However, as stress increases, concentration decreased significantly and HEOs continued working with less focus. The proposed deep learning-based approach significantly identified stress levels and the rest state using a limited number of computed features. These findings demonstrate the potential of integrating stress detection into high-risk work environments to improve safety and performance.