Wind turbine blades fault detection using system identification-based transmissibility analysis

• DOI: 10.1784/insi.2022.64.3.164
• Publication Date: 2022-03-09T05:17:40Z
• Authors: Xuefei Wang; Long Zhang; W P Heath
• Keywords: Overfitting

Wind turbines (WTs) are extensively installed nowadays and the blades integral components within WT systems. Condition monitoring fault diagnosis (CMFD) for is challenging due to fact that they usually suffer from non-stationary time-varying loads load information often unknown or hard collect. This paper proposes system identification-based transmissibility function (TF) methods effectively detect blade defects further help prevent potential economic loss. The novelty proposed only use output response in time domain, which can therefore remove impact of input excitation. Four different models used this work estimate structure parameters, including autoregressive with eXogenous (ARX) model, moving average (ARMAX) error (OE) model non-linear ARX polynomial model. Regularisation then employed address overfitting issues may occur during parameter estimation. effectiveness demonstrated laboratory using three naturally damaged industrial-scale blades.