Advancing Worker Safety Through AI-Powered and IoT Health Monitoring in Hazardous Petroleum Environments

Abstract Ensuring the health and safety of personnel in hazardous petroleum environments remains a critical challenge, particularly in upstream and offshore operations where workers face extreme conditions such as high temperature, toxic gases, and mechanical stress. This study presents an Artificial Intelligence (AI)-powered and Internet of Things (IoT)-integrated health monitoring framework designed for real-time detection, prediction, and prevention of occupational health risks. The system combines physiological data (Inter-Beat Interval (IBI), heart rate, electrodermal activity, blood volume pulse) with environmental air quality parameters (particulate matter, carbon monoxide, ammonia, nitrogen dioxide, and noise) to assess worker wellbeing under operational stress. Using the open-access dataset [1], several machine learning models—Logistic Regression, Decision Tree, CatBoost, and XGBoost—were trained for multi-class wellbeing classification. Among them, the XGBoost model achieved the highest performance on validation and test splits (validation accuracy 0.9648, test accuracy 0.9628), with strong training accuracy (0.9870), demonstrating robust predictive capability on multimodal data. The framework offers a scalable foundation for digital safety infrastructure in petroleum operations, enabling proactive monitoring, early intervention, and predictive decision-making in Health, Safety, and Environment (HSE) management. Results underscore the potential of integrating AI and IoT to move from reactive to predictive safety strategies across oilfield operations.