Development of data-driven methods for dynamic risk management in the chemical industry

Large amounts of hazardous substances are handled and stored in chemical facilities, elevating the risk of accidental releases with potentially disastrous consequences. Over the last three decades, safety science has significantly improved and evolved, leading to the formulation of Risk Management (RM) frameworks to manage these risks. Traditional RM methods, however, have limitations due to their static nature, prompting the shift towards Dynamic Risk Management (DRM), which offers a more proactive, dynamic approach to safety, accounting for system changes, safety barriers, and human factors. This paradigm shift raises the need for dynamic and inherently updatable tools to capture the intricate dynamics between risk-influencing factors. In this context, Machine Learning (ML) techniques emerge as valuable tools due to their inherent ability to make predictions under uncertainty and model complex nonlinear relationships between features. However, the potential of these techniques in the context of DRM is still scarcely explored. This Ph.D. study aims to advance DRM by developing ML methods for consequence prediction, frequency evaluation, and safety barrier monitoring. It explores the role of ML in enhancing safety and reliability through various algorithms and data sources, addressing major accident prediction, Risk Based Inspection, Time-To-Failure prediction, and monitoring of alarm systems. Additionally, it examines the integration of traditional risk assessment with ML models and the relationship between human expertise and AI in risk management. Proposed methods have been tested on real-world case studies to demonstrate their potential to leverage digitalization for safety enhancements in industrial settings. The study acknowledges limitations such as data quality, model interpretability, and the importance of human oversight in using AI tools. While the trajectory of progress suggests an increasing adoption of AI tools, domain knowledge and human expertise remain pivotal, ensuring effective oversight of intelligent systems, understanding the limitations of ML models, and contextualizing their predictions.