Degree

Doctor of Philosophy (PhD)

Department

Construction Management

Document Type

Dissertation

Abstract

Natural disasters, particularly hydrological events such as floods, cause widespread damage, with coastal areas like Louisiana bearing a significant brunt. Vulnerable communities, especially those with low-income populations, often bear a disproportionate share of the consequences during and after these events. To ensure equitable and affordable recovery, it is essential that recovery strategies prioritize these communities based on their specific vulnerability levels. This dissertation proposes a comprehensive framework that addresses these challenges through three interconnected studies: (1) analysis of underlying disaster vulnerability, (2) development of modular-based housing strategies for post-disaster rebuilding, and (3) establishment of digital twin (DT) driven optimization for post-disaster recovery for transportation infrastructure systems.

The first study presents a methodology for identifying and quantifying community vulnerability to floods, addressing existing gaps in understanding the relationship between socio-demographic, infrastructural, and informational factors in disaster impact and recovery. It offers new insights into the primary determinants of community vulnerability and provides a vulnerability index for measuring and comparing risk levels across populations. The methodology has been validated using data from the 2016 Louisiana flood, revealing critical vulnerability factors that significantly affect recovery outcomes. By enabling disaster response teams and policymakers to prioritize recovery efforts for the most at-risk populations, this study advances the field of disaster resilience planning and promotes more informed decision-making.

In addition, the second study examines the potential of modular-based construction approaches to enhance post-disaster housing recovery. Post-disaster recovery efforts hinge significantly on providing permanent housing, which remains one of the most critical yet time-intensive aspects of disaster response. While modular housing presents a viable solution due to its efficiency and scalability, it is often challenged by production delays, overproduction, and misaligned delivery schedules that disrupt on-site reconstruction timelines. To address these challenges, this study proposes a deep reinforcement learning (DRL) framework designed to optimize the allocation of prefabricated modular units across multiple production facilities and construction sites. The model incorporates key constraints, including production capacity, transportation logistics, and social vulnerability prioritization, to ensure an optimized allocation strategy. A case study demonstrates that the DRL-based approach can reduce project delays by up to 46% and lower overall costs compared to conventional optimization methods. Furthermore, the study underscores the importance of aligning construction operations with social equity considerations to ensure that the most vulnerable populations receive prioritized housing support.

Beyond housing, natural disasters often inflict severe damage on critical infrastructure, particularly roadways, which serve as the backbone for overall recovery efforts, including housing reconstruction. However, monitoring and supervising roadway reconstruction manually is labor-intensive, prone to human error, and inefficient. To address these challenges, the third study proposes an innovative DT framework that utilizes audio and other sensor-based data for real-time monitoring of transportation construction projects. By leveraging deep neural network-based sound classification, the DT system enables seamless integration between physical and virtual environments, providing real-time insights into ongoing activities, progress estimation, and early identification of potential bottlenecks. Field experiments validate the system’s capability to enhance the efficiency of a construction site while reducing reliance on manual oversight, positioning it as a transformative solution for managing large-scale infrastructure projects.

By integrating these three studies, this dissertation offers a novel approach to accelerating post-disaster recovery and ultimately improving long-term resilience of infrastructure systems and communities. It highlights the vital role of advanced technologies and data-driven frameworks in strengthening community resilience, optimizing resource allocation, and streamlining construction operations. Together, these contributions lay a robust foundation for future innovations in disaster recovery and sustainable infrastructure development.

Date

5-15-2025

Committee Chair

Lee, Yong-Cheol

Download

Available for download on Thursday, May 06, 2032

Share

COinS