Deep generation of material properties and representations using deep latent space models
AI-accelerated property prediction, discovery, and design of materials.
The generator can generate enormous data, eliminating the demerits of both experiments and simulations.
With our proposed approach, the expensive simulator can be replaced with a deep generator and a deep property predictor.
The proposed approach can be helpful in accelerating and scaling the design of lightweight impact-resistant structures.
More about the research is available here.
Deep material property estimation using inverse deep surrogate models
Motivation:
Characterization of composite materials is necessary for non-destructive property measurements, and real-time material degradation aspects.
General problem statement:
Can we find out its material properties without destroying it?
Inverse problem statement:
Material properties affect wave propagation behavior.
Inversion: Decode wave behavior to get material properties.
Solution schemes:
Forward problem: Solved with SMM & SFEM.
Inverse problem: Solved with ML & DL.