Experimental Verification of Substructure Identification for Damage Detection in Shear Buildings

Abstract

Damage detection for civil structures is limited by several factors including poor signal-to-noise ratios, a large number of unknown parameters, and a limited set of measured responses. Global vibration techniques that track modal parameters often remain insensitive to common forms of structural damage. Moreover, large sets of identified parameters make inverse problems ill-conditioned. To overcome some of these limitations, researchers have advanced substructure identification as a methodology to directly detect local stiffness changes using measured responses to improve damage detection and scalability in civil structures. This paper develops a substructure identification estimator that identifies the story stiffness of a shear building. Concurrent with the estimator derivation, identified parameter confidence intervals are developed and identification performance is predicted. Using the developed estimator, experimental testing is performed on a 3.66-m (12-ft) four-story steel structure subject to base excitation. Several structural configurations are tested where the story-level stiffness is decreased by loosening floor-level connections. These changes simulate damage and are mostly detected by substructure identification within computed confidence intervals. The substructure identified parameters are compared against modal measures and found to be more sensitive to damage. Furthermore, the estimator’s performance follows predictions from the error analysis and motivates future work with identification assisted by structural control.