Validation of a model-based dual-band modal decomposition and expansion approach for fatigue monitoring of offshore wind turbine foundations

Abstract

The fatigue limit state strongly drives the lifetime of offshore wind turbine support structures. Despite steady advancements in their design, the accumulation of fatigue damage over the lifetime of the structure remains uncertain; one way to improve certainty is to directly monitor structural loads at fatigue-critical locations, e.g., near or below the mudline. However, directly monitoring the loads at these locations is often not possible, since these locations do not allow for easy installation or retrofitting of sensors. To this end, model-based virtual sensing techniques can be employed to estimate the structural response at unmeasured locations based on vibration response data and a finite element model which can accurately describe the modal and operational deflection shapes of a specific turbine [1,2,3]. For this contribution, a modal decomposition and expansion (MDE) algorithm [4] will be used to predict strain time series at fatigue-critical locations, based on acceleration and strain data obtained from the tower of an operational OWT. Emphasis will be put on sparse data situations in which a realistic amount of sensor data is used as input. Fibre-optical strain gauges close to and below the mudline level (Fig 1) will be used for direct validation of the extrapolation capabilities. The required finite element model is built using an in-house developed and verified integrated FE model class [5,6], in combination with a unique database which contains all available geotechnical and structural data of offshore wind turbines (owimetadatabase) [7,8]. This combination allows us to generate fleet-wide, turbine-specific, finite elements models for several wind farms. The results will be presented in terms of time series (e.g., Fig. 2) and fatigue damage-related metrics. Preliminary results show that strain time series estimated with the MDE method show good agreement with the validation data.