Research

Our research spans the four thrusts below, from the chemistry of how materials oxidize and break to machine learning for multiscale modeling and prediction. Every thrust pairs a physics-based core with a data-driven layer, working from the governing equations up to fast, reliable prediction. We build physics- and chemistry-based constitutive laws and finite-element frameworks for oxidation, damage, fracture, and multiphysics coupling, grounded in continuum mechanics rather than fit to data. We also characterize materials in our own lab, and use machine learning to predict full fields and design new materials.

Thrust I: Oxidation, from the ground to orbit and re-entry Materials oxidize on very different timescales. Slowly, sitting in heat and oxygen on the ground; aggressively, struck by highly reactive atomic oxygen in low- and very-low-Earth orbit (LEO / VLEO); and almost instantly, in the searing heat of atmospheric re-entry. We connect the microstructural changes that thermo-oxidation drives to the properties that matter, from stiffness to fracture energy, and measure them ourselves with XRD, Raman spectroscopy, and mechanical testing.

Thrust II: The mechanics of breaking We study why and when materials break, and, just as important, how to measure the properties that govern fracture. We use phase-field models to capture crack nucleation and growth, including large-deformation and dissipative response. Good simulations need good inputs, yet for soft materials the methods to accurately measure properties like fracture energy are still not well established, so we develop the experimental techniques to characterize them.

Thrust III: AI for multiscale simulations and material design Full-field simulation is accurate but slow. We build physics-based machine-learning models, using graph and convolutional neural networks, that stay physically consistent while predicting the multiscale response of composites, from the microstructure to the component, orders of magnitude faster than direct simulation. The same models turn prediction into design, including fiber-reinforced composites and shape-memory ceramic systems.

Thrust IV: Infrastructure aging Pavements fail where water and load meet. We model how moisture, especially repeated flooding, couples to the mechanical response of asphalt, from microstructural damage up to the behavior of the full pavement structure, and we use machine learning to predict performance as conditions change.