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

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