Modern-day offshore wind turbine foundations are designed by using the so-called Sequential Integrated approach, where two parties are involved: a Wind Turbine Manufacturer and a Foundation Designer. The information exchange between the parties is realized by means of reduced and linearized systems. The accuracy of the final design strongly depends on the quality of these reduced models and their ability to describe the physical structure, i.e. the internal dynamics, wave loading and soil modelling. This thesis investigates the quality of the currently used approach and strives for improving the accuracy and efficiency, by implementing new methods.
The first part of the thesis describes the theoretical background for the reduction methods, including the common methods i.e. Guyan and Craig-Bampton, but also the relatively new Augmented Craig-Bampton. Furthermore, the soil modelling approaches are described, including the non-linear methods, e.g. Winkler spring method. The theoretical basis points out the limitation of the methods and serves as a starting point for further investigations.
The second part investigates the main scope of work i.e. the reduced foundation models, the modal representation of the tower and the non-linear soil modelling. In order to assess the quality of the reduced foundation models, three methods are compared with the non-reduced reference system. The standard load calculation procedure is performed by using a high _delity jacket, modelled in Rambøll's in-house Offshore Structural Analysis Program ROSAP combined with the reference 5 MW NREL turbine that is implemented in the Rambøll's in-house aero-elastic code. The results demonstrate that the Guyan method cannot properly describe the internal brace dynamics while the Craig-Bampton in some cases cannot efficiently describe the influence of the internal wave loading. Therefore, the robust Augmented Craig-Bampton method is proposed, which delivers the same accuracy as the reference solution. Moreover, the Augmented Craig-Bampton reduction method combined with the efficient Direct Expansion recovery-run procedure significantly increases both, the efficiency and accuracy of the analysis.
Secondly, the modal tower representation in an aero-elastic code{ FLEX5 is investigated. The approach implemented in the code, which includes only two tower bending modes, is compared with the reference, non-reduced tower model. A significant error has been found. Therefore, it was concluded that the number of internal modes is not sufficient. A sensitivity study has been performed, which demonstrated that ten modes should be implemented in order to properly describe the tower stiffness.
Finally, the linearized soil representation for an offshore wind turbine has been investigated. The model based on linear soil is compared to a model with non-linear soil representation. The results significantly differ, especially when a high displacement level has been introduced in the soil. Therefore, two solutions have been proposed for mitigating the problem. The most precise and robust solution is to exclude piles from the linear jacket reduced model and implement the non-linear soil representation externally. This approach delivers the same results as the non-reduced jacket with non-linear soil, but demands an additional implementation in the aero-elastic code. In order to overcome this problem, a so-called user-defined linearized soil approach is proposed, which can properly estimate the ultimate pile displacement, but requires geotechnical expertise.