Digital and AI-Driven Approaches to Hydrogen Asset Integrity: Challenges and Opportunities in the UK Gas Industry

Hydrogen blending and emerging dedicated hydrogen networks in the United Kingdom (UK) introduce integrity threats to legacy gas assets that differ from conventional natural-gas service: hydrogen-assisted cracking (HAC) and embrittlement, altered fatigue crack growth (FCG) kinetics under cyclic pressure loading, increased leakage propensity through joints and polymeric seals, and uncertainties in material compatibility. Limited hydrogen service field data compound these risks, heterogeneous inspection datasets (ILI, NDT), and the safety-critical decision-making context. Artificial intelligence (AI) can improve integrity performance through automated defect detection, anomaly detection in operations data, probabilistic remaining-life estimation, and decision support within risk-based integrity (RBI) frameworks. However, many published models are trained under methane-service distributions and can fail under hydrogen-driven domain shift. This review critically synthesises hydrogen-relevant degradation mechanisms for pipelines and above-ground installations, evaluates AI methods for integrity tasks (data quality control, detection, diagnosis, prognosis, uncertainty quantification, and digital twins), and assesses UK-specific drivers and constraints (HyDeploy, H21, FutureGrid, evolving IGEM guidance, and transmission-level blending policy). The review argues that high-trust deployment will require physics-informed AI, explicit uncertainty quantification, and auditable assurance practices aligned with UK safety cases. A research and implementation agenda is proposed for AI-enabled hydrogen integrity assurance across the UK gas value chain.