Graduation Semester and Year
Fall 2025
Language
English
Document Type
Dissertation
Degree Name
Doctor of Philosophy in Civil Engineering
Department
Civil Engineering
First Advisor
Nur Yazdani
Second Advisor
Jessica Eisma
Third Advisor
Shuoyi Wang
Fourth Advisor
Raad Azzawi
Abstract
The United States faces a critical challenge in managing its aging bridge infrastructure, with more than 600,000 structures listed in the National Bridge Inventory (NBI) and many approaching or exceeding their design service life. Agencies such as the Texas Department of Transportation (TxDOT) are responsible for maintaining a vast statewide network under increasing traffic demand, constrained budgets, and evolving environmental conditions. Although bridge inspection programs provide extensive data—ranging from NBI ratings and National Bridge Element (NBE) condition states to detailed defect records—these datasets are not yet fully leveraged to generate consistent maintenance decisions, prioritize structures systematically, or anticipate future deterioration across the network. Current approaches often rely heavily on condition ratings, lack a unified way to link defect observations to repair actions and costs, and seldom validate prioritization outcomes over time, resulting in uncertainty regarding the long-term reliability of agency decision frameworks.
This dissertation addresses these gaps through a multi-stage research effort designed to convert inspection information into actionable maintenance intelligence, develop a predictive and data-driven priority scoring system, and verify its credibility through temporal and real-world validation. The first stage establishes a defect-centered maintenance methodology using detailed bridge inspection reports from TxDOT’s AssetWise system. Defects across CS-1 to CS-4 condition states were analyzed to characterize deterioration patterns and associate them with appropriate repair interventions and unit cost estimates derived from TxDOT bid items. This analysis was then operationalized into a rule-based decision process that automates defect interpretation and supports consistent, replicable maintenance recommendations.
Building upon these insights, the second stage develops a statewide machine learning–based Priority Score (PS) model utilizing approximately 58,000 Texas bridges between 2016 and 2024. Four engineered indices capturing structural deficiency, traffic criticality, age-related deterioration, and scour vulnerability were synthesized into a continuous 0–100 score and grouped into Low, Medium, and High priority categories. Eight machine learning algorithms were trained and evaluated, with XGBoost demonstrating superior predictive performance. The model was then used to forecast bridge priority over a five- to ten-year horizon by incorporating expected deterioration trends and traffic growth, enabling the identification of structures likely to transition into higher-risk conditions in the future.
The final stage strengthens the practical validity of the prioritization framework through comprehensive temporal and case-based evaluation. A multi-year Walk-Forward Validation (2016–2023) assessed the stability and predictive reliability of the Priority Score over time, while a case-based comparison aligned predicted high-priority structures with TxDOT’s major repair and replacement actions documented in AssetWise. The alignment between predicted classifications and real-world program decisions demonstrated that the model maintains strong fidelity to actual agency practices and offers credible support for long-term planning.
Together, these studies deliver an integrated, data-driven framework that transforms raw inspection information into maintainable decision rules, predictive priority scores, and validated forecasts. The outcomes enhance transparency, repeatability, and forward-looking capability for bridge maintenance planning, equipping transportation agencies with a robust toolset to allocate limited resources more effectively and strengthen the safety and resilience of the bridge network.
Keywords
Bridge maintenance, AssetWise inspection data, Defect classification, Machine learning, Priority scoring model, Deterioration forecasting, Walk-Forward Validation, Transportation asset management
Disciplines
Civil Engineering | Structural Engineering
License
This work is licensed under a Creative Commons Attribution 4.0 International License.
Recommended Citation
Zenika, Naima Tahsin, "Predictive Bridge Maintenance Prioritization Using Machine Learning: A Data-Driven Framework for Decision Support" (2025). Civil Engineering Dissertations. 529.
https://mavmatrix.uta.edu/civilengineering_dissertations/529