Machine learning for shock compression with scarce data
Interpretable ML models extract Hugoniot behavior from sparse experimental datasets. Sidestepping the data-poverty problem that defeats classical regression.
Research Area
Energetic systems live and die by what happens in microseconds. CECD's mechanics and shock-physics program develops the constitutive models, multiscale frameworks, and validated codes that capture how materials deform, fail, and respond when struck, from the dislocation to the structure.
Strain rates
Length scales
Frameworks
Open codes
A model that cannot be falsified is not a model. CECD's mechanics program is structured around a simple discipline: every framework is paired with the experiment that would prove it wrong.
Multiply parallel FE-DDD for arbitrary geometries; binary junction formation and strength in HCP crystals.
ML-augmented shock compression models trained on scarce experimental data, with emphasis on solid energetics and metals.
Anisotropic full-Brillouin-zone Boltzmann transport, sub-Casimir thermal conductance, and microscale thermal design.
Asymptotic and computational homogenization for woven-fabric composites, layered media, and porous solids under transient impact.
Composable linear operators for lattice defect analysis, design, and spectral characterization.
Constitutive modeling of charged-vacancy migration and electromechanics in dielectric and piezoelectric crystals.
Selected studies. Peer-reviewed work that captures the program's reach from atomistic defect mechanics to structural impact.
Interpretable ML models extract Hugoniot behavior from sparse experimental datasets. Sidestepping the data-poverty problem that defeats classical regression.
A general-purpose solver that captures real phonon dispersions and band-structure anisotropy, a foundation for microscale thermal design.
A foundational asymptotic-homogenization framework linking lattice statics to continuum field theories. Still in active use across the CECD portfolio.
Twenty-five years of code, frameworks, and validated benchmarks. Many co-developed with sponsor laboratories.
Parallel discrete dislocation dynamics in arbitrary geometries.
3D anisotropic Brillouin-zone transport for thermal design.
Crystal plasticity, dielectric-crystal electromechanics, viscoelastic creep.
Multiscale frameworks bridging lattice statics to nonlinear continuum mechanics.
Inquiries about sponsored research, graduate training, or technology transitions in this area can be directed to the center.