A Review: Risk Assessment Methodology for Onboard Hydrogen Storage Cylinders Under Fire Conditions

Abstract Hydrogen, with a wide explosion limit range (4%-75%) and extremely low ignition energy, poses inherent safety risks of leakage and explosion during storage, transportation, and utilization. Type III (aluminum liner) and type IV (polymer liner) composite cylinders are widely adopted in hydrogen fuel cell vehicles (HFCVs) for the storage of compressed gaseous hydrogen due to their lightweight and exceptional mechanical strength characteristics. Since the pressure of the compressed hydrogen stores in the cylinder reaches up to 70MPa, the carbon fiber/epoxy resin composite layers are combustible. Regulations and standards require thermally activated pressure relief devices (TPRDs) to be installed on the hydrogen cylinder to prevent rupture in fire scenarios. However, delayed TPRD activation during vehicular fires may trigger catastrophic explosions, generating fireballs, blast waves, and fragments, particularly in confined environments like tunnels or underground garages. This review evaluates existing risk assessment methodologies for onboard hydrogen storage systems, including the qualitative risk assessment method, quantitative risk analysis (QRA) method, computational fluid dynamics (CFD)-based numerical modeling, and accident consequence analysis based on artificial neural networks and machine learning. While progress has been achieved in scenario-specific assessments (open spaces, underground garages, tunnels), critical limitations persist. Key challenges include insufficient experimental data for validating CFD models, difficulties in modeling multi-physics phenomena, and the absence of standardized risk prediction frameworks. To address these gaps, integrating CFD with large-scale experimental data could enhance model accuracy, while AI-assisted frameworks may improve computational efficiency for multi-scale accident simulations. Future efforts should prioritize scenario-tailored risk assessment standards and user-friendly predictive tools to support the safe large-scale deployment of HFCVs.