Data Driven Bridge Deck Deterioration Modeling and Maintenance Intervention Scheduling

<p>Majority of over 617,000 bridges across the United States are significantly impacted by environmental factors, aging materials, and heavy traffic loads. According to the 2021 Infrastructure Report, nearly 231,000 bridges require repair or preservation work, with 46,154 (7.5%) being classified as structurally deficient, thereby posing risks for mor than 178 million daily trips across these bridges. Despite recent improvements, 42% of U.S. bridges are at least 50 years old, underscoring an urgent need for increased investment to meet repair demands. The nation’s backlog of bridge repairs is estimated at $125 billion, with a 58% increase in annual spending required to address deteriorating conditions adequately.</p> <p>Traditional bridge deterioration models required by bridge management systems often suffer from deterministic assumptions, inspector subjectivity, and data encoding errors, which can lead to inconsistent or biased assessments. These challenges are compounded by the high levels of censoring in bridge inspection data, particularly in the National Bridge Inventory (NBI), where observations (i.e., inspections) are often limited to specific time windows rather than the full lifecycle of a structure. These limitations hinder accurate deterioration modeling and typically result in reactive maintenance strategies, where interventions are delayed until significant deterioration occurs. This research introduces an innovative framework to enhance bridge deterioration modeling by addressing the challenges posed by high censoring in the National Bridge Inventory (NBI) inspection data. Additionally, Artificial Intelligence (AI) methods are explored and compared to traditional models used by Departments of Transportation (DOTs). The study also investigates the effects of various maintenance interventions across each NBI rating for optimized maintenance scheduling. Leveraging extensive bridge datasets, including historical inspection records and environmental factors, this data-driven approach facilitates proactive maintenance planning, reducing costs and improving the safety and longevity of infrastructure assets.</p> <p>This research also explores the integration of large language models (LLMs) technology to develop AI assistant bot to interpret and derive insights from unstructured bridge inspection reports. The AI bot can automatically extract key deterioration indicators, assign condition ratings, and generate maintenance recommendations, thereby reducing subjective bias and standardizing inspection data. This approach not only minimizes reliance on manual data processing but also enables more consistent and scalable bridge assessments among various inspection teams.</p> <p>By shifting from reactive to predictive maintenance, this research aims to advance bridge management practices, reduce costs, and enhance infrastructure safety and longevity. Ultimately, this study contributes to the evolving field of AI applications in infrastructure, demonstrating how artificial intelligence and LLMs can transform the maintenance and management of critical assets on a national scale.</p>

Name thesis:degree_name

Doctor of Philosophy (Ph.D.)

Level thesis:degree_level

Dissertation

Discipline thesis:degree_discipline

Civil Engineering

Year dc:date.available

2025

Author dc:creator

• Kumar, Deepak

Contributors dc:contributor

• Anil K. Agrawal

• Bridge Deterioration Modeling
• National Bridge Inventory (NBI)
• Weibull Analysis
• AFT-Weibull Analysis
• Censoring
• Survival Analysis
• Artificial Intelligence
• Large Language Modeling
• Bridge Inspection
• Maintenance Scheduling
• Civil Engineering
• Operational Research
• Risk Analysis
• Transportation Engineering

Repository record dc:identifier

https://academicworks.cuny.edu/cc_etds_theses/1281

OAI identifier oai:identifier

oai:academicworks.cuny.edu:cc_etds_theses-2289