Data-driven identification of local structural dynamics in wind turbines under various aerodynamic & marine loads

Abstract

The power generation from wind turbines constitutes an example of highly complex engineering system especially in offshore applications where flow around the tower and nacelle coupled with inflow turbu- lence and rotation of the turbine blades create unpredicted aerodynamic forces which are transmitted into structures like critical joints causing resonance that drastically reduces the design lifetime. Standard approaches that are used in the design to determine stress in structural components typically use geometrically detailed FEM models including the effect of external loads. Loads are provided by aero- hydro-servo-elastic models, which are geometrically simplified to account for global modes and most relevant non-linarites. Even though the component FEM models are detailed, they are considered as linear and quasi- static w.r.t. global system dynamics. Therefore, actual non-negligible effects such as local dynamics or local non-linearities are not considered during the design process. The thesis aims to establish a methodology to study whether these effects are actually present in the real prototypes which are used to validate the design. This is mainly done by modelling the structural response using experimental time-series. For this purpose dynamic correlation models such as ARMAX and TARMAX is used. The local modal parameters (natural frequencies, damping ratios and mode shapes) are identified and structural frequency response function (FRF) is reproduced. The methodology is validated successfully by a synthetic model under different SNR. An industrial appli- cation is done on the Haliade 6MW Wind Turbine Prototype under different meteorological conditions. Similar modal parameters are identified.