Statistical evaluation of jacking-installation resistance models for suction foundation of offshore wind turbines in sands

Abstract

The American Petroleum Institute (API) method is widely adopted to estimate the installation resistance of suction foundations, while the estimation accuracy remains unknown. This paper first develops a database containing 114 measured jacking-installation resistance data. Then, the database is used to quantify the model uncertainty of the current API methods with and without consideration of the soil plug effect. The results showed on average the predicted installation resistance is about one third of the true value if by the current API unplugged model, and five times larger if by the current API plugged model. The dispersions in prediction accuracy are medium to extremely high. The accuracy of both API models exhibits statistical dependency to the magnitude of the predicted installation resistance mainly due to the correlation between model bias and the penetration depth. A soil plug factor is introduced to not only unify the current API unplugged and plugged models into one formulation, but also capture the development process of soil plug during the foundation installation. Another two empirical terms are then added to calibrate the unified API model. The calibrated unified API model is accurate on average with low dispersion in prediction accuracy and no hidden statistical dependency.