Offshore jacket-type platforms utilize tubular X-joints, among other configurations, for their high strength-to-weight ratio and bending resistance. However, weld geometry at the brace–chord intersection induces localized stress intensities, quantified as Stress Concentration Factors (SCFs), which strongly affect fatigue performance. Accurate SCF prediction under out-of-plane bending (OPB) is especially critical in joints retrofitted with fibre-reinforced polymer (FRP) composites. While existing models address several geometric and material parameters, the influence of brace inclination (θ) has often been overlooked.

This study develops an artificial neural network (ANN) model that incorporates θ alongside five additional parameters, which are the brace-to-chord diameter ratio (β), chord diameter-to-thickness ratio (γ), brace-to-chord thickness ratio (τ), number of FRP layers (N), and FRP-to-steel stiffness ratio (ξ), to predict SCFs in FRP-reinforced tubular X-joints. Finite element analysis (FEA) of 127 parametric joint configurations was performed in ANSYS Workbench 2024 R1, with SCFs extracted following International Institute of Welding (IIW) guidelines. Results revealed that increases in β, γ, and τ raised SCFs by up to 84.14%, 61.15%, and 58.44%, respectively, whereas steeper brace angles, additional FRP layers, and higher ξ reduced SCFs by up to 38.29%.

The ANN, designed with a 6–8–4–1 architecture and trained in MATLAB, achieved an R² of 0.996. External validation confirmed predictive accuracy within ±10% of FEA values. This study delivers a reliable, angle-inclusive SCF prediction model, supporting more efficient fatigue assessment and retrofitting strategies for offshore tubular joints.