For offshore wind turbines, the support structure contributes with a significant part to the Levelized Cost Of Energy (LCOE). LCOE is the total cost to build and operate a whole offshore wind turbine structure over its lifetime divided by the total energy output of the wind turbine over that lifetime. That total expected cost of operation & maintenance may be lowered by application of reliability- and risk-based maintenance strategies and updating of the reliability based on e.g. inspections performed during the design lifetime. Updating the reliability (or alternatively the failure probability) of a welded joint can theoretically be done using Bayesian updating. However, for tubular joints in offshore wind turbine substructure when considering a two dimensional crack growth and a failure criterion combined brittle fracture and material strength, the updating is quite complex due to the wind turbine loading obtained during operation. This paper considers how the reliability (or the probability of failure) of welded steel details can be updated in the case where the fatigue failure is modelled by a fracture mechanics approach and a Failure Assessment Diagram (FAD) is used to define a limit state equation. Besides, it is shown how the probability of failure of fatigue critical joints can be updated for various inspection scenarios. A two dimensional bi-linear model is considered for the crack growth. Calculation of the crack depth and the crack length are coupled. The stress intensity factor is calculated following the sophisticated procedure in BS 7910:2005. The initial crack size, the yield and ultimate strengths of steel, the fracture toughness, the stress intensity factor, and the stress-ranges are considered as uncertain and modelled by random variables. The probability of detection is used to account for uncertainty in crack inspections. By using Monte Carlo simulations, stress-range histories are generated randomly and together with the calculated stress intensity factors used to check the limit state condition using the FAD approach. The probability of failure updating is done for three inspection scenarios: No crack detected; Crack detected and repaired; Crack detected but not repaired. The paper presents illustrative results for offshore wind applications for the three scenarios, including comparisons of application of the FAD approach for modelling the limit state equation with a ‘conventional’ approach where it is assumed that failure happens when crack-depth reaches the plate thickness or the stress intensity factor is equal to or larger than the fracture toughness of the material. Finally, it is discussed how application of the updating procedure can be used for inspection planning for offshore wind turbine support structures, and thus also for reducing the required safety factors at the design stage.