UC San Diego

The American Society of Civil Engineers (ASCE) 2021 report card rated the U.S. infrastructure at a C- grade. Therefore, there is an operational need for structural health monitoring (SHM) and damage prognosis (DP) for large-scale civil infrastructure systems. An effective way of performing SHM/DP of structural systems is by using a hybrid physics-based and data-driven digital twin or cyber model. A potential approach for constructing digital twins of civil structural systems consists of using the Bayesian finite element (FE) model updating framework. The process of calibrating probabilistically, using the Bayesian inference framework, a FE model of a structural system using sensor measurement data collected from the system is termed as Bayesian FE model updating. Most methods of FE model updating for SHM consist of updating linear FE models based on changes in modal parameters identified before and after a potentially damaging event (e.g., an earthquake) using low amplitude vibration data. However, these modal methods only identify damage as loss of effective stiffness and can only be used to detect the existence of damage and localize it. This dissertation focuses on the Bayesian FE model updating framework applied in the time domain. This framework can be used to update linear and nonlinear FE models. In contrast to updated linear FE models, a mechanics-based nonlinear FE model of the system (able to capture the damage states and failure modes of interest) updated using measurement data can provide information about other crucial characteristics of damage such as loss of strength, ductility capacity, and low cycle fatigue life, etc., which are very important to identify for comprehensive damage diagnosis and prognosis. The updated mechanics-based nonlinear FE model can be directly used to detect, localize, classify, and assess the severity of the damage and perform damage prognosis. The Bayesian time-domain FE model updating framework is illustrated using three civil infrastructure testbed structures – a concrete gravity dam, a miter gate, and a bridge column. The framework is further extended to account for model form uncertainty, arguably the most significant source of uncertainty in model calibration, in linear dynamic systems. The extended framework is illustrated and validated on simple structural benchmark problems. Surrogate models can be used as fast emulators of FE models to accelerate the extremely computationally expensive model updating process. Part of this dissertation focuses on evaluating the loss of accuracy and the gain in computational time while performing Bayesian model updating by using surrogate model evaluations compared to using direct FE model evaluations.