Processing time, temperature, and initial chemical composition prediction from materials microstructure by deep network for multiple inputs and fused data

Prediction of the chemical composition and processing history from microstructure morphology can help in material inverse design. In this work, we propose a fused-data deep learning framework that can predict the processing history of a microstructure. We used the Fe-Cr-Co alloys as a model material. The developed framework is able to predict the heat treatment time, temperature, and initial chemical compositions by reading the morphology of Fe distribution and its concentration. The results show that the trained deep neural network has the highest accuracy for chemistry and then time and temperature. We identified two scenarios for inaccurate predictions; 1) There are several paths for an identical microstructure, 2) Microstructures reach steady-state morphologies after a long time of aging. The error analysis shows that the majority of the wrong predictions are indeed not wrong, but the other right answers. We validated the model successfully with an experimental Fe-Cr-Co transmission electron microscopy micrograph.