Understanding the Influence of Environmental and Operational Variability on Wind Turbine Blade Monitoring

For data-driven vibration-based structural health monitoring (VSHM) systems to be considered reliable they must overcome the challenge of mitigating the environmental and operational variability (EOV) on the vibration features. This is particularly important in large and exposed structures such as wind turbine blades (WTB). This work aims to understand the influence of EOV, namely quantifying the influence of input variables on the selected vibration features. Understanding the specific sources of influence can facilitate better prediction of outliers as well as leading to a VSHM system less sensitive to EOV. This study uses an operational wind turbine with an undamaged and incrementally damaged WTB under three operating conditions (idle, 32 and 43 rpm). The approach calculates frequency transformation based features on the vibration responses obtained from an array of accelerometers along the WTB. Subsequently, the features are regressed on environmental and operational parameters (EOPs) via multivariate non-linear regression. The difference between the regression predictions and the actual feature values is used as a new feature. In parallel, to understand the influence of the EOV, inclusive and exclusive sensitivity analyses were conducted. These analyses compared the likelihood of a model based on one or all but one EOP, respectively, against a model using all the EOP. The results showed that the temperature has the largest influence, with respect to the considered EOP, on the regression likelihood. Ultimately, the obtained regression model was used to normalise the effects on the features and enhance damage detection.