Performance assessment of driveability predictions for different pile geometries and soil conditions

As wind turbines increase in size, the foundation loads increase as does the size of the required foundations. Consequently, the installation of the pile foundations (monopiles and jacket piles) becomes a key component of design and the accurate calculation of the soil resistance to driving (SRD) becomes ever more critical to assess the installation approach, hammer selection, and pile driving fatigue. Various well-established SRD methods are typically adopted for pile installation assessments. These methods all have an empirical or semi-empirical basis and as such are only reliable within the database against which they were calibrated. This paper provides an assessment of the performance of various key SRD methods for as-built piles from several offshore windfarm (OWF) sites, covering a range of pile geometries and soil types. From comparison of the measured installation data to the method predictions it is shown that no single method is considered suitable across a range of pile geometries and soil types. The use of machine learning (ML) techniques is subsequently explored using the same OWF pile installation database and its performance compared to the industry standard methods. The potential advantages and the drawbacks of ML applications for pile driveability predictions are herein explored and discussed.