Published: 2023

Publication Name: Integrating Physics-Based and Data-Driven Models for Community Resilience Assessment Under Wind Storms

Publication URL: https://www.proquest.com/docview/3073246108?pq-origsite=gscholar&fromopenview=true&sourcetype=Dissertations%20&%20Theses

Abstract:

TIDC Project C19.2020

 

Under strong wind storms, such as hurricanes, tornadoes, or Nor’easters, damages to civil infrastructures could lead to substantial social and economic losses. With a changing climate and aging infrastructure, the resilience of structures and communities subject to coastal hazards is paramount. To forecast the impacts of storms on infrastructure systems and communities and quantify the effects of mitigation or adaptation investments, improved predictive models are needed with greater accuracy and computational efficiency. Moreover, such models should be more holistic, incorporating the effects of multi-hazards, social and economic conditions, and interactions between different structural systems and the natural and built environments. To address these research challenges, a general framework for the resilience assessment of coastal communities and critical infrastructures, including transportation and power systems and residential buildings, is constructed. The model integrates physics-based structural analysis with data-driven and statistical modeling to generate models of infrastructure and community vulnerability to disasters in various scenarios. First, a hybrid model for predicting weather-related power outages is developed by integrating physics-based fragility curves with machine learning models to improve accuracy for extreme events and assess the benefits of grid hardening. Second, a physics-informed deep learning model to predict the response of transmission towers under wind loads is presented and is shown to outperform purely physicsbased or data-driven models in terms of computational efficiency and accuracy, respectively. Third, a probabilistic risk assessment framework is generated to predict large woody debris accumulation and scour around bridge piers and assess the effects of different parameters on the bridge vulnerability. Fourth, a multi-objective optimization framework is established to inform ideal mitigation strategies to reduce hurricane damage to residential homes while considering social and economic factors. Finally, a community-level resilience assessment model is built to predict post-storm damages and residents’ access to critical services considering the impacts of damaged trees. The applicability of the established models to predict infrastructure risk and inform effective decision-making to reduce future damages is demonstrated through case studies focused on the New England region.