Statistical model-based damage detection and localization: subspace-based residuals and damage-to-noise sensitivity ratios

Subspace-Based Damage Detection on Steel Frame Structure Under Changing Excitation

“…For the damage detection method in [2,3] a residual function is constructed based on properties from covariance-driven outputonly subspace-based system identification [13,10].…”

“…In the following, the non-parametric damage detection test [1] based on [2,3] is described. Using measured data (y k ) k=1,...,N , a consistent estimate H p+1,q of the Hankel matrix is obtained from the empirical output covariances…”

“…is defined [2,3]. It is tested if this residual is significantly different from zero or not, using the χ 2 test…”

“…In this paper we consider the subspace-based damage detection approach developed in [2,3,5,6]. With this approach, a data-driven model obtained in the reference state is compared to data from the possibly damaged state using a subspacebased residual function and a χ 2 -test built on it for a hypothesis test, without actually estimating the modal parameters in the tested, possibly damaged states.…”

“…Since this kind of methods process the measurement data in a much more direct way, the original algorithms from [2,3] have the drawback of being influenced by changes in the unknown ambient excitation, which may lead to false alarms. Recently, the original algorithms have been extended and modified to be robust under such changes in the excitation in [6].…”