Research

My research develops degradation-aware learning and optimization methods for smart and connected manufacturing systems: how to model degradation, predict failures remotely, and turn those predictions into proactive maintenance and operating decisions.

Methodology: degradation modeling and prognostics; machine learning and AI (deep learning, LLMs, and generative models); digital twins; reinforcement learning; optimization and data-driven decision-making.

Applications: smart manufacturing; logistics robotics; fusion facilities and nuclear energy systems.

🩺 Degradation Modeling and Remote Prognostics

Predicting remaining life from sensor and event data, including under unknown failure modes and with limited communication or sensing budgets.

• Dynamic Sensor Selection for Remote Prognostics, IISE Transactions (2026), a feature article. [code]
• Degradation Modeling and Prognostic Analysis Under Unknown Failure Modes, IEEE TASE (2025). [code]
• A Bayesian Spike-and-Slab Sensor Selection Approach for High-Dimensional Prognostics, IEEE TASE (2025).
• A degradation-aware cross-attention framework for multimodal prognostics with discrete event logs and continuous sensor signals (under revision).

🛠 Proactive Maintenance and Decision-Making

Using learning and optimization to decide when to maintain and how to allocate work across a fleet of degrading assets.

• Fleet maintenance and workload allocation under unknown degradation dynamics (in preparation).
• Event-based control for proactive degradation management (in preparation).
• Reinforcement learning for degradation control and maintenance.

⚛️ AI for Nuclear Energy and Advanced Systems

Extracting operational knowledge from unstructured records with large language models, and fleet-level monitoring and maintenance of advanced microreactors and multi-vehicle systems.

🧠 Foundational Machine Learning and Optimization

• Instance Selection via Voronoi Neighbors for Binary Classification Tasks, IEEE TKDE (2024). Best Paper Finalist, IISE 2023. [code]
• Comparative Study on the Performance of Categorical Variable Encoders (under review). [code]
• 3D Dynamic Heterogeneous Robotic Palletization Problem, European Journal of Operational Research (2024). [code]

🎤 Invited Presentations

• Maintenance and Workload Allocation under Unknown Degradation Dynamics: A Learning and Optimization Approach, INFORMS Annual Conference, San Francisco, CA, 11/2026
• Degradation Control and Maintenance via Reinforcement Learning, INFORMS Annual Conference, Atlanta, GA, 10/2025
• Dynamic Sensor Selection for Remote Prognostics, INFORMS Annual Conference, Seattle, WA, 10/2024
• Degradation Modeling and Prognostic Analysis Under Unknown Failure Modes, INFORMS Annual Conference, Phoenix, AZ, 10/2023
• Instance Selection via Voronoi Neighbors for Binary Classification Tasks, IISE Annual Conference (ISERC), New Orleans, LA, 05/2023