Ageing of hydraulic structures such as, pumped-storage power plants, rotating machinery (turbines, rotors, etc.), and locks…etc, compels managers to choose among several scenarios: extending the service life, performing repairs, reinforcing, or even complete replacement. Such structures, that naturally experience variable (cyclic) loading over time due to their operation, are generally assemblies of plates and beams, which present stress concentration zones, particularly around weld seams. The risks of crack initiation and propagation due to fatigue are often located near these areas. Fatigue behaviour is an inherently random phenomenon, characterised by various uncertainties related to loading conditions, mechanical properties, material characteristics, manufacturing quality, geometry, and other variables. Initially, international design codes and standards are irrelevant, focusing on lifespan estimation during the initial design phase, making them inapplicable to scenarios requiring an extension of the operational life. Despite the existing research and studies concerning estimating the fatigue life, the stress–life relation, the probabilistic fatigue curves (P–S–N) of mechanically welded metallic structures (Mikulski and Lassen, 2022; Rocher et al., 2020; Schubnell et al., 2024), and fatigue assessment using stochastic finite element (SFE) models (Larsen et al., 2021; Singh et al., 2020). Only a few studies have addressed the fatigue assessment of lock gates, using SFE models, e.g. (Estes et al., 2004; Ramancha et al., 2022). In light of these considerations, this paper proposes a probabilistic fatigue approach, based on fracture mechanics analysis and combined with SFE modelling at the scale of structural-steel welded assemblies. A comparison of different international design codes and standards (EC3: EN 1993-1-9, 2005; BS 7608, 2014) will reveal scientific challenges, potential gaps, and discrepancies that might influence fatigue life assessment, providing a broader perspective to gauge the proposed approach’s potential in enriching fatigue calculations. This work emphases local stochastic characterisation of fatigue crack propagation in welded assemblies. Influencing factors are modelled as random variables, allowing uncertainties to propagate throughout the FE mechanical model. The post-processing of the generated meta-model, based on an implicit analytical function derived from the Paris law, allows conducting a probabilistic assessment of the remaining useful life (RUL). The resulting cumulative distribution function providing the probability of failure for a given failure criterion, serves as an alert tool for maintenance and risk management, helping operators in planning inspections and interventions to ensure the long-term security and reliability of ageing structures.