Novel Tools for Structural Health Monitoring of Offshore Wind Turbine Components Using Ultrasonic Guided Waves

Offshore wind assets increasingly use fibre–matrix composites and bolted or assembled joints. To reduce risk, downtime, and cost, structural health monitoring (SHM) and
non-destructive evaluation must detect, localise, and characterise damage before critical
failure. Ultrasonic guided waves (UGWs) can interrogate large regions from few sensing
locations, but deployment is limited by dispersion, multimodality, anisotropy, attenuation, and the dense measurements needed to characterise full wavefields in space and time. This thesis develops datasets and tools for UGW-based SHM in offshore wind components: (i) bolt-loosening monitoring in assembled structures, (ii) full-field propagation and dispersion analysis in a thick glass-fibre reinforced polymer (GFRP) plate, and (iii) probabilistic reconstruction for sparse sensing.

An experimental study on an aluminium plate shows that bolt loosening alters bolt–
surface contact and stress transfer, producing measurable changes in local guided-wave
features. The change is most notable at the looser conditions and becomes harder to
quantify nearer the initial tight conditions, yet the loosened bolt-hole region can still be
localised.

Full-field undamaged and impact-damaged GFRP wavefields are analysed using
frequency–wavenumber methods to examine modal content and dispersion. A repeatable spectral artefact in the undamaged dataset prevents direct baseline comparability for advanced Fourier-domain procedures, though time-domain baseline subtraction and simpler imaging remain effective. A mode- and direction-aware masking framework is proposed and partially validated on the damaged dataset, enabling separation of outward-propagating components for anisotropy-consistent processing, filtering, and imaging.

Finally, a Gaussian process framework is developed that reconstructs time-domain signals while providing a coherent Fourier-domain representation. Extending Bayesian
Reconstruction of Fourier Pairs with dispersion-informed kernel structure, the method
accurately recovers synthetic two-dimensional dispersive wavefields from sparse observations and outperforms stationary GP baselines.