The CEE Department is pleased to announce Amir Moayyedi's PhD Defense: "Risk-Based Bridge Life-Cycle Management under Extreme Events and Condition Deterioration: A Novel Approach with Interpretable Reinforcement Learning"
Date: Thursday, November 5th, 2026
Time: 8:40 AM
Location: in-person in EB 315 or online on Zoom: pdx.zoom.us/j/83646884751
Advisor: Dr. David Yang
Abstract: "Transportation agencies increasingly rely on risk-based management to select interventions and allocate limited resources throughout the service life of bridges. Reliable decisions require accurate quantification of extreme-event risk, realistic prediction of condition deterioration, and interpretable life-cycle policies. This dissertation addresses methodological challenges in these three areas through complementary approaches for risk-based bridge life-cycle management.
The first approach uses Gaussian quadrature to approximate the probabilistic risk integral with a small set of objectively derived, weighted hazard scenarios. It preserves the practical risk-table structure used in bridge management systems while avoiding subjective scenario selection. The second approach uses nonhomogeneous Poisson regression to model deterioration directly from National Bridge Element condition data. It estimates quantities transitioning between condition states and reflects the influence of bridge age and structural, traffic, and environmental characteristics. The third approach employs proximal policy optimization with interpretable tree-based policy representations to identify interventions using element-level condition information. These policies account for deterioration, intervention costs, and risk-related consequences over the decision horizon, while providing decision logic that bridge managers can examine and verify.
Together, these contributions provide a methodological foundation for objective hazard risk assessment, element-level deterioration prediction, and interpretable intervention planning. Future research can integrate and validate the three components for multi-element bridges and bridge networks while accounting for inspection uncertainty and shared resource constraints."