Deep-Learning-based damage detection of an experimentally tested full-scale wind turbine blade.

Abstract

Reducing the Operation and Maintenance (O&M) costs of Wind Farms can substantially contribute towards adopting environmentally sustainable and cost effective energy sources. O&M costs in Wind Turbines can account up to 50% of the Turbine’s whole life cycle, with the majority of it being the frequent manual inspection needed for Wind Turbine Blades (WTB). Vibration-based structural health monitoring (SHM) approaches can be used for detecting WTB damages in an early state, before they lead to catastrophic failure, while they can be used to perform inspection on an informed basis, reducing O&M costs. In the present study, a data-driven damage detection framework is developed for WTBs. A Finite Element Model (FEM) of a full-scale WTB (the Aventa AV-7 “Light Wind Turbine”, sold under the Leichtwindanlagen© name) is developed in ABAQUS FEA using sparce material properties and dimensions and subsequently modified to match a physical WTB. Damage is modelled in the form of a Trailing Edge (TE) longitudinal crack propagated in a stepwise manner. The FE model is updated based on the geometry altering and material property degrading effects of the TE crack. Numerical simulations are performed to obtain the dynamic responses of the blade, under white noise and chirp excitation. A damage detection framework is then developed using Auto-Regressive, with exogenous excitation (ARX) or without (AR), models and its Nonlinear counterparts NARX and NAR models. In addition, model dimensionality reduction is performed using Principal Component Analysis (PCA) and Deep-Autoencoders (D-AE). The results are presented in terms of Receiver Operating Characteristics (ROC) curves.