Data‐Driven Damage Detection on a Temperature‐Varying Turbine Blade Using Multiple Sensor Modalities

This study investigates vibration‐based structural health monitoring (SHM) of a small‐scale wind turbine blade using multiple sensor types to evaluate the performance of an output‐only, data‐driven and semisupervised damage detection framework. The primary objective is to assess how different sensing modalities influence damage detection and localisation and to demonstrate the robustness of the proposed SHM approach across diverse sensor configurations in a wind turbine composite material blade. To this end, experimental accelerometer and strain‐gauge time series are considered in this study. The method integrates multivariate autoregressive (MAR) modelling, principal component analysis (PCA), Mahalanobis distance‐based dissimilarity metrics, and K‐means clustering. This hybrid framework requires no labelled damage data and minimal user input, offering a transparent, interpretable and computationally efficient solution suitable for real‐world applications. The blade was experimentally tested under progressive damage, which was introduced as sequential cracks at three locations, and under varying temperature conditions. Results show that the method reliably detects damage across sensor configurations, highlighting its potential as a practical and cost‐effective tool for continuous SHM. Notably, both sensor types were able to detect damage using only one sensor as input in the model; however, strain gauges proved more effective than accelerometers for damage localisation.