Research

My research develops probabilistic and computational methods to assess the safety and resilience of aging civil infrastructure subject to deterioration and natural hazards. I combine physics-based simulation with data-driven methods to support infrastructure assessment, maintenance, and design.

1. Resilience of Aging Infrastructure and Networks

Deterioration and hazard-induced damage jointly govern how individual structures and infrastructure networks function and recover. Through the MSCA project DREAMS at UCL, I am developing a probabilistic framework that connects corrosion-induced deterioration, state-dependent fragility under mainshock–aftershock sequences, and the functionality and recovery of bridge networks. Related ongoing work examines post-earthquake traffic flow capacity in bridge networks and resilience assessment of corroded prestressed concrete bridges. Together, these efforts aim to inform inspection, maintenance, and recovery planning.

Keywords: Infrastructure Resilience; Bridge Networks; State-Dependent Fragility; Mainshock–Aftershock Sequences; Decision Support

Current project: DREAMS — MSCA Postdoctoral Fellowship

2. Probabilistic Modeling of Structural Deterioration

Deterioration mechanisms such as strand corrosion and fatigue cracking involve local processes that are not explicitly represented in conventional global structural models. To capture them, I link detailed finite element analyses at the material and component levels to structural-scale predictions through probabilistic constitutive models and surrogates. These frameworks predict the behavior of corroded strands and prestressed concrete girders while propagating uncertainty across scales. Bayesian inference is used to estimate fatigue crack growth parameters and quantify their uncertainty from limited experimental data.

Keywords: Corrosion; Fatigue; Multi-Scale Modeling; Bayesian Inference; Uncertainty Quantification

Selected publications: Corroded strand behavior  (2025); Probabilistic multi-level modeling of PSC girders (2026)

3. Efficient Reliability, Fragility, and Design Optimization

Reliability and fragility analyses based on high-fidelity finite element models are often computationally demanding. To reduce this cost, I develop computational tools such as FERUM-MIDAS, which integrates reliability analysis with MIDAS/Civil, and use sequentially updated surrogates and active-learning Kriging. Applications include bridges under heavy-truck loads, buried pipelines under earthquakes, damage-index-based fragility assessment of reinforced concrete bridges, and bridge performance under flood hazards. Related work extends these methods to reliability-based design optimization of structures subject to fatigue-induced sequential failures.

Keywords: Finite Element Reliability Analysis; Surrogate Modeling; Fragility; System Reliability; Reliability-Based Design Optimization

Selected publications: Sequentially updated surrogate modeling (2024); FERUM-MIDAS platform (2025); Damage-index-based bridge fragility (2025)

4. AI-Accelerated Simulation and Data-Driven Structural Assessment

Deep learning can accelerate computationally demanding simulations and infer structural responses from incomplete measurements. Models trained on physics-based simulations and measurement data support four tasks:

• predicting seismic floor responses of nuclear power plant structures with uncertainty propagation;

• reconstructing unmeasured seismic floor responses to prioritize post-earthquake inspection of critical equipment;

• accelerating stochastic landslide runout simulation with graph neural networks;

• detecting anomalies in long-term bridge monitoring data.

Keywords: Deep Learning; Virtual Sensing; Graph Neural Networks; Structural Health Monitoring; Anomaly Detection

Selected publications: Virtual sensing for inspection prioritization (2025); Seismic response prediction with uncertainty propagation (2026); Graph neural network landslide simulator (2026)

5. Nondestructive Evaluation and Structural Health Monitoring

Assessing existing structures depends on what can be measured, and how reliably. Building on my work at KRISS, this research combines nondestructive testing and monitoring data with physics-based models and probabilistic learning:

• estimating concrete strength from linear and nonlinear ultrasonic measurements;

• reconstructing ultrasonic signals for absolute stress monitoring of mooring cables in submerged floating tunnels;

• building high-fidelity finite element models of corroded rebars from X-ray CT scans;

• detecting anomalies in long-term bridge monitoring data under changing environmental conditions.

Keywords: Nondestructive Testing; Ultrasonic Testing; X-ray CT; Structural Health Monitoring; Anomaly Detection

Selected publications: Ultrasound-based concrete strength estimation (2026); Anomaly detection for cable-stayed bridges (2024)