Less dependence on labels
Train with physical constraints when reference solutions are scarce.
PDCO / Surrogate models
Turn physical laws into reusable prediction.
Physics AI for fields and evolution
Reference simulations can be expensive to repeat. PDCO learns from energy and governing equations to predict material fields and evolution without first generating large sets of solution labels.
From microstructure to stress fields
Learn the response. Reuse the model.
High-fidelity labels can be expensive to generate. PDCO learns from energy and governing equations, turning material configurations and initial states into reusable field predictions.
New material configurations
Predict heterogeneous fields for new inputs within the learned problem family.
Repeated exploration
Reuse a trained operator across related prediction tasks.
See how structure redirects stress.
Changing inclusion shape and arrangement changes the load path. The learned operator maps heterogeneous material distributions to their stress fields.
Follow a disturbance through the material.
Predict propagation, scattering and interference in heterogeneous media, connecting the initial disturbance to its evolving displacement field.
Early propagation
Developing field
Later response
Lower uₓHigher uₓ
Turn an initial field into an evolving state.
An evolution operator follows the changing phase field, supporting repeated studies of how an initial state develops over time.
