Currently wind turbines are a popular source of renewable energy. Royal HaskoningDHV designs foundations for these turbines. Design optimization of these foundations becomes more and more important due to the growing demand for renewable energy, the development of increasingly large wind turbines and intensification of competition. Design improvements can reduce construction costs, and also reduce the carbon footprint of the tower foundation. Today these foundations are designed using the Eurocode in combination with the results from linear analysis of a 2D model. Expectations are that this yields a conservative model. A 3D non-linear model is expected to provide a more detailed insight into the actual structural behaviour of such a foundation. Specifically, the distribution of concrete and reinforcement stresses and strains can be found, and used to determine whether, and where, reinforcement can be reduced or should be increased. To obtain a non-linear 3Dmodel, first a linear 3D model is developed, starting fromthe existing 2Dmodel. Subsequently the results of the two linearmodels are compared. Once the linear 3Dmodel is deemed to function satisfactorily, it is further developed to incorporate non-linear material properties and reinforcement, taking into account observations made in the linear analysis.
Simplifications made for the 2D model cease to hold. Besides minor details, the pedestal, that was left out of the scope for the 2D model, is introduced in the 3D model. Additionally, the connection between the foundation piles and the structure are placed at the underside, instead of at mid-height. Results of the
2D and 3D linear models are generally similar. However, small differences were observed. In general, these differences can be attributed to these two main modelling differences. Modelling in 3D opened up more possibilities for detailing of the supporting foundation piles. However the effects on the results for different support cases was minor. After comparing the linear analysis results, the next step is a non-linearmodel. The non-linear 3D model includes several non-linear aspects. Reinforcement was incorporated. Non-linear constitutive models were studied and chosen for concrete in tension and compression as well as for the reinforcement steel. Connections between different parts of the structure, such as the anchor cage, were also modelled non-linearly. The results of the non-linear analysis seems to provide an improved insight into the structural behaviour of the foundation structure. More detailed information about concrete strains, reinforcement stresses, crack progression and displacements can be shown, and used to verify, or disprove, assumptions based on the 2D linear analysis. However, problems with non-convergence occur before the expected failure load without a
satisfactory structural explanation. Therefore the model does not provide a plausible structural failure load. Because no ultimate load was found it is difficult to determine whether, and if so, how much the applied reinforcement can be reduced. Nevertheless, the model could be used in a more qualitative way to show and describe where stress concentrations will occur and how the structure will deform. Internal splitting cracks are seen outside of the centre of the structure where only limited reinforcement is present. These splitting cracks may be a cause of model instability. Design alterations can be made to increase the load for which non-convergence occurs. An increased the ultimate load is found when splitting
reinforcement is elongated, or when the amount of applied prestressing load on the anchor cage is altered. Even applying a simply modified constitutive concrete model, to represent steel fibre reinforcement, appears beneficial. However, these alterations may only improved the behaviour of the model, while still no actual
failure mode is found. It cannot be said without a doubt that these modifications will improve the structuralbehaviour of the actual structure. Further research may lead to a solution for the model instability and consequently enabling its use to optimize the reinforcement design.