Structural health monitoring (SHM) is defined as the capability to monitor the performance behavior of civil infrastructure systems as well as to detect, localize, and quantify damage in these systems. SHM technologies contribute to enhance the resilience of civil infrastructures, which are vulnerable to structural aging, degradation, and deterioration and to extreme events due to natural and man-made hazards. Given the limited financial resources available to renovate or replace them, it is crucial to implement SHM methodologies, which can help detect safety threats at an early stage, evaluate the operational risk of the infrastructure after a catastrophic event, and prioritize the urgency of the repair/retrofit or replacement of these structures.

This research focuses on the development of a novel framework for nonlinear structural system identification. This framework consists of updating mechanics-based nonlinear finite element (FE) structural models using Bayesian inference methods. Recognizing structural damage as the manifestation of structural material nonlinearity, the developed framework provides a new methodology for post-disaster SHM and DID of real-world civil structures.

This research is subdivided in two parts. The first part investigates the accuracy of state-of-the-art nonlinear FE modeling in predicting the cyclic and dynamic inelastic response behavior of reinforced concrete structural components and systems. Sources of inaccuracy and uncertainty in the FE modeling and simulation approach are investigated by comparing the FE-predicted structural response with high-fidelity experimental results. In the second part of this research, two frameworks for nonlinear FE model updating are proposed, developed, and validated using numerically simulated data. In the proposed frameworks, different Bayesian estimation methods are utilized to update the nonlinear FE model of a civil structure using the recorded input excitation and response of the structure during a damage-inducing earthquake event. The initial frameworks are then extended to output-only nonlinear structural system and damage identification methods. This extension not only overcomes the shortcomings of the initial frameworks in handling unmeasured or noisy input measurements, but also paves the way to a general approach to account for model uncertainties. Finally, a new information-theoretic approach is developed for the purposes of nonlinear FE model identifiability, experimental design, and optimal sensor placement.