Additional Funding Sources
The project described was supported by the National Science Foundation via the Research Experience for Undergraduates Site: Materials for Society (Award Nos. DMR 1658076 and 1950305) and by Boise State University.
Abstract
In this work we developed a microstructure database for a model alloy, i.e., Iron Chromium Alloy (FeCr), to inform machine learning algorithms to predict the materials parameters based on their microstructure evolution morphology. In particular, in this work we have focused on Cr mobility and interfacial energy. We have used the phase field model to generate the FeCr microstructure database. FeCr undergoes spinodal decomposition at a temperature of 500 °C at a concentration of 45% Chromium. We have explored a wide range of parameters for mobility, from 0.5e-26 - 10e-26, and kappa (gradient energy coefficient), from 0.5e-16 - 10e-16. We have modeled the spinodal decomposition through the Cahn-Hilliard equation via MOOSE (Multiphysics Object-Oriented Simulation Environment) framework. In this work, we have generated 400 microstructures. We plan to use these images to train a Convolutional Neural Network toward predicting mobility and kappa just by reading microstructures.
Materials Parameters Prediction from Microstructure by Machine Learning: Database Generation
In this work we developed a microstructure database for a model alloy, i.e., Iron Chromium Alloy (FeCr), to inform machine learning algorithms to predict the materials parameters based on their microstructure evolution morphology. In particular, in this work we have focused on Cr mobility and interfacial energy. We have used the phase field model to generate the FeCr microstructure database. FeCr undergoes spinodal decomposition at a temperature of 500 °C at a concentration of 45% Chromium. We have explored a wide range of parameters for mobility, from 0.5e-26 - 10e-26, and kappa (gradient energy coefficient), from 0.5e-16 - 10e-16. We have modeled the spinodal decomposition through the Cahn-Hilliard equation via MOOSE (Multiphysics Object-Oriented Simulation Environment) framework. In this work, we have generated 400 microstructures. We plan to use these images to train a Convolutional Neural Network toward predicting mobility and kappa just by reading microstructures.