Probabilistic fatigue reliability assessment of girth-weld surface cracks in API 5L X65 offshore pipelines under cyclic global bending using a Kriging surrogate

Fatigue cracking at girth welds threatens offshore-pipeline integrity and requires uncertainty in flaw size, crack-growth behaviour, loading and weld geometry to be considered. A probabilistic fatigue-reliability framework is developed for a semi-elliptical surface crack in an API 5L X65 pipeline under cyclic global bending. Geometry-consistent stress-intensity-factor histories from a three-dimensional finite-element model provide 60 training and six case-wise held-out simulations. The Paris-law and loading distributions are identified as DNV code-calibrated benchmark inputs, not site-specific X65 measurements. For the fixed Paris exponent, normalisation by CeffΔσeq3 reduces the primary surrogate to a one-dimensional Gaussian process in initial crack depth. The model achieves R2 = 0.997 and MAPE = 4.04% on the test cases, and R2 = 0.996 with MAPE = 3.70% in leave-one-out validation. Using conditionally truncated inputs, the posterior-mean life distribution has a median of 6.84 × 108 cycles and a 95%-reliability design life of 2.12 × 108 cycles. Correlated latent-GP posterior-function propagation gives Pf = 4.147% at Nd = 2 × 108 cycles, with a 95% GP-posterior interval of 3.858–4.444%. Sensitivity analysis identifies the stress-range scale as dominant, showing that site-specific loading characterisation is essential for field application.