My research develops data-driven and physics-informed methods for modeling, understanding, and controlling dynamical systems. I am especially interested in reduced-order models that remain reliable over long times, preserve important physical structure, and can be used for prediction, interpretation, and control.
01
Reliable reduced-order models
Learning compact dynamical systems that preserve the important behavior of fluid flows.
High-fidelity fluid simulations are often too large and expensive to use directly for design, prediction, or control. My work develops simpler models that retain the essential dynamics while remaining reliable beyond short training windows.
This research combines modal decomposition, sparse regression, operator inference, and physics-informed constraints to learn low-dimensional dynamical systems from high-dimensional flow data. A central goal is to build reduced-order models that are accurate, interpretable, and stable enough for downstream analysis and control.
02
Physics-informed and bounded model learning
Constraining learned models so that they respect qualitative physical structure.
Data-driven models can make accurate short-term predictions but then drift, diverge, or behave unphysically. I work on methods that build physical structure into learned models so their long-time behavior remains meaningful.
This work focuses on enforcing boundedness, energy consistency, and trapping-region structure in learned dynamical systems. These ideas connect data-driven model discovery with convex optimization, semidefinite programming, and stability analysis.
03
Dynamics, phase, and coherent structure
Using modal, phase-based, and time-frequency tools to understand unsteady flows.
Many fluid flows are organized by recurring structures and rhythms. I use phase, frequency, and modal analysis to identify how these structures evolve and how they respond to forcing or actuation.
This work includes POD-based analysis, wavelet/time-frequency methods, phase identification, and coherent-structure interpretation for separated aerodynamic flows. The goal is to connect high-dimensional flow fields with interpretable dynamical mechanisms.
04
Optimization and flow control
Finding perturbations and actuation strategies that strongly influence nonlinear flows.
Instead of asking only how a flow evolves, I also ask how it can be changed. My control work studies which disturbances or actuator inputs have the largest effect on the future flow state.
This research uses optimization-based methods for finite-amplitude perturbations, sparse nonlinear optimal perturbations, actuator placement, and active flow control. These methods help identify sensitive regions, influential structures, and effective control strategies.
05
Robustness, uncertainty, and emerging directions
Extending model-learning and control methods toward robust prediction and decision-making.
Real fluid systems are uncertain, noisy, and imperfectly modeled. A growing direction of my work is to understand how reduced models and controllers behave under uncertainty and how to make them more robust.
Current and emerging interests include uncertainty quantification, PDE lifting, nonlinear balanced truncation, stochastic operator-inference ideas, and robust nonlinear control.