Integrating multiscale materials modeling with interpretable automation techniques

Maria Emelianenko
George Mason University

As the materials community is moving towards embracing high throughput calculations and predictions made by ML algorithms, there is a clear need for better integration of physics-based models at multiple scales and for the development of scalable interpretable predictors. This talk will highlight some recent modeling and analytical advances in the context of microstructure evolution and additive manufacturing applications. Specifically, the role of parameter estimation, feature extraction, dynamics identification and dimension reduction techniques will be addressed for a class of problems requiring reliable prediction of optimal processing parameters to yield desired mechanical properties. Numerical experiments and analytical results on newly developed parallelizable CVOD-CUR dimension reduction methodology will be presented.


Back to Workshop II: Scale-Bridging Materials Modeling at Extreme Computational Scales