Geotechnical design of offshore wind monopiles under cyclic loading

Abstract

For supporting Offshore Wind Turbines (OWT), monopiles are currently the most common foundations. The role of a foundation is to transfer safely the loading to the ground. The wind and the wave loads are considered cyclic because they repetitively apply on the OWT. The North Sea is sand dominated in many areas. During cyclic loading, permanent strains develop in the surrounding soil while the soil stiffness and strength are irreversibly affected. Through time, the accumulation of strains can lead to the soil failure. Thus, assessing the behaviour and stability of monopiles under cyclic loading is essential. To model the response of monopiles under lateral loading, the traditional design procedure is the use of p-y curves that express the lateral soil resistance in function of the pile deflection. The p-y curves are nowadays recommended to be calibrated on FE models. The Stiffness Degradation Method (SDM) of Achmus et al. (2009) is a numerical strategy that assesses the behaviour of a monopile under cyclic loading. The method estimates the cyclically degraded stiffness based on the results of a static analysis. The soil stiffness is degraded based on a semi-empirical power law that accounts for the number of loading cycles, the stresses in the soil after the static analysis and two model parameters calibrated on cyclic triaxial tests. The SDM was successfully implemented in PLAXIS 3D via a practical routine coded in Python and the use of soil clusters around the pile. The soil stiffness is degraded by updating the soil material within the clusters. The study model was verified by comparing results with the published reference system of Kuo (2008) for two piles with embedded length to pile diameter ratios of 2.7 and 5.3. The results indicate that the study model provides a stiffer pile-soil response than the reference model because the soil stiffness is overestimated. The degraded stiffness overestimation is attributed to the initial stiffness mismatch and the use of soil clusters. The impact on the short pile is greater than on the long pile because the short pile opposes less resistance to the loading and is thus more affected by the stiffness difference. The study model was validated against three 1-g pile tests for homogeneous uniform and multi-layered dense sand. The numerical results are in agreement with the test data. In the absence of cyclic triaxial tests, the two model parameters were directly calibrated on the pile head displacement of the experiment. The two model parameters have a significant impact on the stiffness degradation. Thus, model parameters from literature were classified from the highest to the smallest estimation of pile lateral displacement. The results of codified and published approaches (DNV-GL-0126; Duhrkop, 2009; Garnier, 2013) were compared with the results of the study model. The study model and the method of Garnier (2013) are in agreement. They both account for the loading amplitude, the number of cycles and the pile geometry. The codified procedure and the method of Duhrkop (2009) estimate higher lateral displacement compared to the study model. Finally, the 1D model was successfully calibrated with the highest displacement estimate of the study model. With this procedure, the 1D model accounts for the number of cycles, the pile geometry and the loading amplitude. The study model provides a less conservative approach for determining the pile lateral displacement under cyclic loading. The calibration of the 1D model on the pile deflection curves of the study model is a promising procedure which will require further research.