Statistical model-based damage localization: A combined subspace-based and substructuring approach

Abstract

This paper addresses the damage localization problem with a statistical model-based approach applied to vibration-based measurements. Damages are viewed as changes in modal parameters. Damage detection is achieved with a subspace-based residual and a global test, which performs a sensitivity analysis of the residual to the modal parameters, relative to uncertainties in those parameters and noises on the data. Damage localization is achieved by plugging the sensitivities of the modal parameters with respect to structural (finite element model) parameters in this decision framework. For large structures that have thousands of elements, a statistical substructuring method, in which the columns of the latter sensitivity matrix are clustered into different classes, is employed. This paper investigates further the clustering step. Numerical results obtained on the finite element model of a bridge deck with a large number of elements are reported. Copyright © 2007 John Wiley & Sons, Ltd.