Computational Modeling of Interacting and Coalescing Surface Cracks - Application to Offshore Wind Turbine Welded Connections

Offshore wind turbine (OWT) foundations operate in harsh marine environments where cyclic loading and corrosion act together to drive fatigue damage. As monopiles are the predominant foundation type for European OWTs, their fatigue deterioration must be characterized credibly to support optimized inspection and maintenance. However, reliability models that capture interaction and coalescence among multiple surface cracks under corrosion fatigue remain limited. Circumferential welds in monopile foundations can host several surface cracks; accurate determination of stress intensity factors (SIFs), including crack to crack interaction, is necessary to predict crack growth rates and assess fracture capacity. This thesis develops a computational and probabilistic framework to quantify the structural reliability of monopile welds subject to corrosion fatigue, linking high-fidelity fracture mechanics simulations with scalable surrogate modeling.

The approach integrates Paris law crack growth with SIFs computed from three-dimensional finite element analysis (FEA). Multiple interacting surface cracks are modeled with explicit treatment of interaction, coalescence, and transition to failure via through-thickness penetration or fracture toughness exceedance. The framework maintains numerical stability near coalescence and resolves realistic welded geometries to capture local stress concentrations. A multilayer perceptron (MLP) surrogate that computes SIFs along evolving crack fronts is trained on high-fidelity FEA results, enabling large-scale Monte Carlo life cycle reliability analysis. Probabilistic crack growth simulations incorporate uncertainties in initial flaw size distributions, crack growth parameters, and long-term stress ranges representative of OWT service conditions.

The framework is demonstrated on a monopile circumferential weld case with two semi-elliptical surface cracks at the weld toe and explicit weld geometry representation. Surrogate assisted evaluation of SIF histories enables efficient simulation of interaction and coalescence. Results show (i) acceleration of failure due to crack interaction and coalescence, (ii) high sensitivity of reliability to initial defect distributions and long-term stress ranges, and (iii) substantial computational speed-ups from the surrogates while retaining accuracy.

The main outcome is an interaction-aware, surrogate-enabled reliability assessment model that accounts for physical system dependencies via crack interaction and coalescence, supporting inspection planning, decision-making under uncertainty, and potential life-extension strategies for OWT assets.