Probabilistic fatigue life prediction for welded joints with undercuts considering crack initiation variability and multi-crack interaction

Weld-toe undercuts, as common welding imperfections, generate severe localized stress concentrations along the weld toe. The resulting stochastic crack initiation and subsequent multi-crack interactions markedly degrade the fatigue resistance of welded joints. Conventional deterministic methods fail to adequately address these uncertainties, limiting their applicability in safety–critical assessments. This study proposes a probabilistic prediction framework integrating a fracture mechanics-based crack growth model, fatigue testing, and Monte Carlo simulation to predict the fatigue life of welded joints with undercuts. The model explicitly accounts for weld-toe amplification, notch effects, multi-crack interaction, and crack closure to quantify the crack driving force at weld-toe undercut tips, and to accurately simulates these multiple cracks interaction, coalescence, and propagation. Fatigue tests on cruciform-welded (CW) and butt-welded (BW) Q420qFNH steel joints validate this model and provide statistical inputs for simulations. Monte-Carlo simulations then yield reliability-based fatigue life by accounting for the variability of crack initiation at weld-toe notches. Results demonstrate that fatigue strength is predominantly controlled by undercut depth (secondary by curvature radius), with maximum allowable depths of 0.7 mm and 0.5 mm for CW and BW joints to comply with IIW requirements at 95 % reliability. The framework advances defect-tolerance design through reliability-based acceptance criteria, offering a practical tool for welding quality control.