Date of Award

Winter 2024

Project Type

Dissertation

Program or Major

Civil and Environmental Engineering

Degree Name

Doctor of Philosophy

First Advisor

Erin Santini-Bell

Second Advisor

Masoud Sanayei

Third Advisor

Yashar Eftekhar Azam

Abstract

Finite element models are widely used to simulate the behavior of structures. They can also be utilized for structural condition assessment and health monitoring. However, due to software limitations, simplifying assumptions, and lack of knowledge about the system properties, the predictions of the numerical models may deviate from laboratory or field measurements of the actual structure. This issue may be further exacerbated by the simplifications applied in modeling complex connections, which is a common practice in numerical modeling to save time and effort. While finite element model updating improves the accuracy of numerical models, physics-based surrogate modeling of connections balances simplicity and accuracy in low-fidelity structural simulations. Structural parameters of the connection zones can be estimated by model updating techniques. Modal data have shown significant potential for model updating and elemental parameter estimation. Nevertheless, their capability for parameter estimation of structural connections is yet to be examined.This research develops a methodological framework for proposing low-fidelity physics-based surrogate models of complex connections with structural parameters estimated through finite element model updating within a modal paradigm. Four model updating approaches, including physics-based, Bayesian probabilistic, artificial neural network, and Bayesian neural network, are formulated. The developed methodologies are evaluated for a benchmark laboratory steel structure with bolted joints, the UCF grid, and an in-service steel truss bridge with gussetless connections in Portsmouth, NH, the Memorial Bridge. Parameterization of the connection zones is a significant challenge in employing surrogate physics-based models for complex joints. Moreover, due to the small size of connections, the variations of their structural parameters may not be captured by global modal parameters such as natural frequencies. However, local modal parameters like mode shapes are measured with less accuracy than natural frequencies, and their measurement depends on the instrumentation density and distribution throughout the structure. These challenges are identified and addressed, and practical guidelines are proposed. The results demonstrate the effectiveness of the updated surrogate models of complex connections in improving the reliability of low-fidelity numerical models and addressing uncertainties related to the behavior of structural joints. Efficient strategies for the parameterization and grouping of joints are introduced, novel insights into the semi-rigidity of bolted connections are uncovered, and a model selection protocol is proposed for the low-fidelity representation of gussetless connections. Integrating several model updating approaches and highlighting the synergies of modal-based parameter estimation methods provides a robust and unified scheme for structural joint modeling, identification, and evaluation in an engineer-friendly manner.

Download

Share

COinS