An artificial neural network methodology for damage detection: Demonstration on an operating wind turbine blade

• ANN-based methodology for damage detection in wind turbine blades. Minimising the number of outliers robust detection. Learning relationships between sensitive features and novelty indices. Mitigating environmental operational variabilities This study presents a novel artificial neural network (ANN) based within vibration-based structural health monitoring framework The establishes nonlinear selected (DSF) influenced by (EOVs) their corresponding indices computed Mahalanobis distance (MD). ANN regression model is trained validated on reference state (i.e., healthy structure). used to predict MD new observations. prediction error calculated predicted as index Firstly, an 2D feature set generated illustrate how limitations solely using MD-based can be overcome proposed methodology. Secondly, implemented data obtained from in-operation with different artificially induced scenarios one its Finally, performance evaluated metrics accuracy, F1-score Matthews correlation coefficient. results demonstrate advantages improving detectability all despite influence EOVs both simulated real data.