Virtual Poster Exhibition

Crystal Plasticity Inspired Deep Learning Surrogate for Instant Structure-property – Prediction of Additive Manufactured Alloys

Crystal Plasticity Inspired Deep Learning Surrogate for Instant Structure-property – Prediction of Additive Manufactured Alloys

Lead Institution: University of Galway

Contact Details

Overview

This poster presents an instant material structure-property prediction tool for a metal additive manufacturing process based on a deep learning model trained and validated with over a thousand crystal plasticity finite element models, as a surrogate for microstructural deformation simulations. This is a tool that will instantly predict the strength of a material from an image of the microstructure.
The project is carried out by Yuhui Tu, a PhD student based at the University of Galway, under the supervision of Dr Noel Harrison (lead supervisor) and Prof Sean Leen, and was funded by the SFI’s I-Form. This work included collaboration with I-Form members at Maynooth University (Dr Caitriona M Ryan, Prof Andrew C Parnell), as well as international collaborators.
Highlights of this poster include:
• Significantly reduced computational cost using deep learning surrogate, from 13 hours to 1.27 seconds compared to full crystal plasticity modelling.
• Feasible application on real EBSD images and varied phase fraction to assist tailored microstructure design for a desired mechanical response
• A viable concept tool for the in-process monitoring capability accounting for inevitable additive manufactured microstructural complexity and heterogeneity.
• User-friendly GUI software developed for straightforward operation, e.g. one-click stress-strain prediction based on microscopy images, and report generation.
The trained code surrogate is capable of predicting mechanical response within milliseconds, making it feasible to be implemented as an in-situ powder bed fusion process quality control tool. More information and codes available at:
GitHub page: https://github.com/I-Form/Deep-learning—Crystal-plasticity-
Journal paper: https://doi.org/10.1016/j.matdes.2021.110345