Influence of Modeling Uncertainties in Accelerometer-based Strain Estimation for Wind Turbine Support Structures

Abstract

One challenge in structural health monitoring (SHM) is to make reliable statements about the condition of the structure using only a few sensors, which is desirable from both technical and economic perspectives, as wind turbines often already have one acceleration sensor installed at the nacelle. Virtual sensing methods are used for this purpose, enabling the estimation of the structural response at non-instrumented locations using a structural model and measurement data. Knowledge of the structural response of wind turbines is desirable, as it provides the foundation for lifetime extension or load-optimised operation. The method applied here uses only one DC-capable acceleration sensor to estimate the full-field strain across the entire support structures of wind turbines. It utilises the tilt error present in the measured acceleration of tower structures to estimate the displacement in the frequency range below 0.05~Hz. The strains are subsequently estimated from the displacements using a transfer function based on a multi-band modal decomposition and expansion. For this contribution, the proposed method is applied to measurement data from an operating offshore wind turbine. The focus lies on the influence of uncertainties in the structural model on the accuracy of the strain estimation. The investigated modelling errors include parameters that are usually known to the operators with high levels of uncertainty, such as modelling of the scour protection, the soil stiffness, and the rotor. The influence of the selected parameters on the accuracy of strain estimation is compared using the damage-equivalent strains. The results showcase which parameters have a significant influence on the accuracy of strain estimation and indicate whether measures should be taken to reduce the uncertainty in the modelling to increase the strain estimation accuracy.