Coupling dynamic modelling and Bayesian networks for quantitative risk assessment of hydrogen systems

Project overview

The increasing adoption of hydrogen as a sustainable energy carrier presents significant safety challenges throughout the entire hydrogen value chain, from hydrogen production to storage and utilisation. This thesis aims to develop an integrated model-based framework for quantitative risk assessment (QRA) of hydrogen systems, building upon the toolchain approach proposed by Rogovchenko-Buffoni et al. [4] for functional safety analysis. The research will focus on a specific hydrogen system configuration by considering the dynamic modelling activities of FBK for H2 production, storage or utilisation. By leveraging Modelica language multi-domain modelling capabilities, the framework integrates physics-based system models with functional safety requirements expressed through service-oriented components. The methodology extends beyond traditional Failure Mode and Effects Analysis (FMEA) by incorporating probabilistic risk assessment through automatically generated Bayesian Networks. The proposed approach enables early-stage identification of hazardous scenarios, quantification of failure propagation probabilities, and generation of risk priority metrics specific to hydrogen safety concerns such as leakage, combustion, and material compatibility issues. This model-based integration eliminates the semantic gap between system design and safety verification, ensuring compliance with emerging hydrogen safety standards. The thesis demonstrates how functional requirements for hydrogen systems can be formalised within the modelling environment itself, allowing both dynamic verification during simulation and static analysis for dependency mapping. The resulting Bayesian Networks provide insights into component criticality and enable ”what-if” scenarios for risk mitigation strategies. Furthermore, the framework generates comprehensive FMEA tables that support design optimisation throughout the development lifecycle.