The increase in size of offshore wind turbines has altered the dynamic response of monopile-based support structures, making them more sensitive to wave-induced response-loading during periods of low aerodynamic damping. Current design standards, such as described in IEC-61400, estimate fatigue damage under these conditions, using long-term hindcasts adjusted to a predefined proportion of non-operational periods. While this provides a standardised basis for fatigue estimation, it risks under-representing the impact of fatigue-critical sea states occurring during low-damping conditions.

This thesis investigates the influence of variance in wave-induced response-loading due to operational variances on the fatigue life of monopile-based offshore wind turbines. In order to achieve this, SCADA availability data of an 8 MW wind farm is used to estimate turbine availability patterns and an existing availability model, based on Markov matrices and originally developed to assess component reliability and turbine availability, is adapted to be SCADA-driven. Together with frequency-domain simulations, this adaptation enables the integration of availability variances into fatigue damage calculations. The framework is applied to an 8 MW turbine and a 14 MW turbine, providing insight into fatigue damage sensitivity to operational variances for turbines of different sizes.

By adapting the availability model, several hyperparameters related to the representation of non-operational periods arose, such as the classification of non-operational periods. Multiple model configurations were evaluated. The available SCADA data were split into training and test sets. The training set was used to estimate the transition probabilities between non-operational and operational states, while the test set was used to evaluate the different hyperparameter configurations. The final model was benchmarked against the test set SCADA availability profiles, which were matched with frequency domain simulations. Here, it slightly overpredicted both the mean fatigue damage and its variance for four sectors, including the driving sector, while for two sectors it substantially overpredicted both metrics.

Finally, the availability model was used to predict fatigue damage over the full design life for an 8 MW and 14 MW turbine. The findings show that incorporating availability variances into fatigue damage calculations results in a relatively small variance in fatigue damage, in contrast to the single-value
estimate provided by the deterministic approach recommended by IEC-61400. For the 14 MW turbine, the results indicate that fatigue damage over the design life could be slightly higher than anticipated by the recommended DNV approach, which applies a 10% non-operational ratio as a conservative basis.

Future work should integrate availability modelling with fatigue limit state evaluation together with variance in metocean datasets, so that both operational and environmental uncertainties are captured. This would allow rare but fatigue-critical sea states to be represented more realistically. To ensure fatigue reliability under site-specific wave climates and long-term climate change effects, design methodologies could either adopt more probabilistic approaches or recalibrate partial safety factors in a site-specific manner.