Damage Detection in Hysteretic Structures using Measured Seismic Response

Damage in structures is often correlated with a loss of structural stiffness. However, using dynamic response measurements from structures subjected to earthquakes could show significant decreases in stiffness as a result of yielding that is not necessarily an indicator of permanent damage. The use of hysteretic models in system identification may allow for distinctions between permanent losses in structural stiffness and temporary decreases due to nonlinear yielding response. While yielding parameters cannot be identified using small-amplitude vibration data, such as ambient vibrations or weak earthquakes, the information concerning the behavior of the structure in the linear elastic range can serve as useful prior information for Bayesian model updating of hysteretic models. In this study, a simulated structure is subjected to small-amplitude motion and the resulting updated model is then used for detection of damage during a larger event, which produces nonlinear structural response. Stochastic simulation methods are used to perform the updating, because it is inherently an ill-conditioned inverse problem.