Sensitivity analysis of subspace-based damage indicators under changes in ambient excitation covariance, severity and location of damage

Abstract

It is common practice in Structural Health Monitoring (SHM) to perform damage detection by detecting changes in subspace-based damage indicators. They are computed by comparing a reference state with the current (possibly damaged) state through a matrix based on the computation of residuals originated from the orthogonality defect of projected subspaces. In the last decades, several scalar indicators have been proposed based on the residual matrix. Still, a multivariate sensitivity analysis was not carried out to estimate the sensitivity of each indicator to the excitation changes. In the current paper, a covariance-based sensitivity analysis applied to a spatial truss structure is carried out to evaluate the sensitivity of eight damage indicators to ambient excitation covariance, damage severity, and damage location. The analysis aims at identifying the ranges of applicability of damage indicators within a simple statistical framework, possibly extendable to more general applications. It is argued whether any equivalent class of indicators can be obtained from the manipulation of the residual matrix. The role of damage location in the probabilistic assessment of damage is discussed: a methodological Bayesian approach, based on the results of the sensitivity analysis, is proposed for the development of a structural reliability analysis driven by damage indicators.