An Automated Non–Parametric Healthy Subspace Method for Unsupervised Robust Vibration–Based Damage Detection Under Uncertainty

An automated healthy subspace method for unsupervised robust vibration–based damage detection under uncertainty is postulated. The key idea lies with the approximation of the “healthy subspace” as the union of a number of hyper–spheres, with properly determined centers and radii. The method features full automation capability, aiming at eliminating user intervention, and very good detection performance. Its effectiveness is demonstrated via damage detection for a population of composite beams and comparisons with a Multiple Model based and a Random Coefficient Gaussian Mixture model based method.