Assessment of construction workers' fall-from-height risk using multi-physiological data and virtual reality

Abstract

Efficient identification of at-risk construction workers is crucial for reducing fall-from-height (FFH) accidents. However, current methods of evaluating worker FFH risk rely on manual inspections, which are ineffective because of the complex nature of construction sites. To address this issue, this paper presents a technique for FFH risk assessment using physiological data. A virtual reality experiment with three FFH risk scenarios was conducted, during which four categories of physiological data were recorded. Using the physiological data and machine learning algorithms, FFH risk classification models were developed. Three key findings are as follows. (1) All four physiological metrics showed significant changes in response to varying FFH risk levels (2) EEG was the most effective physiological metric for FFH risk assessment, achieving a test accuracy of 0.924 (3) Combining all four physiological categories provided the highest accuracy of 0.998. The findings demonstrate the feasibility of using physiological signals for effective FFH risk assessment.