Statistical Discrimination of Steady State Shift Damage Localization Metrics

When inspecting a linear structure subjected to spatially invariant, repeatable excitation, the recently proposed steady state shift damage localization (S3DL) method offers damage localization by mapping a postulated damage pattern to the damage-induced change in the steady state response. This postulated damage pattern is – depending on the type of damage one seeks to interrogate for – cast as a stiffness or mass perturbation in a theoretical model. The damage is hereby localized when the overlap between the experimental and analytical subspaces, under ideal conditions, is complete. Obviously, such perfect mapping will never occur under realistic, experimental conditions, where operational, environmental and modelling variabilities in addition to measurement noise will be present. This will generally lead to a reduced localization resolution, which potentially complicates a deterministic distinction between healthy and damaged areas. The present paper examines the application of different well-established statistical evaluation schemes to enhance robustness in the discrimination and, as such, a more unambiguous localization. The application studies are conducted in an experimental context with a cantilevered residential-sized wind turbine blade, which is exposed to a harmonic input and with the output taken as accelerations captured along the blade edges. Damage is manifested as a stiffness change and the damage localization interrogation will be carried out accordingly.