First World Congress on Artificial Intelligence in Materials and Manufacturing (AIM 2022): Machine Learning/Deep Learning in Materials and Manufacturing I
Program Organizers: Taylor Sparks, University of Utah; Michael Dawson-Haggerty, Kerfed, Inc.; Elizabeth Holm, University of Michigan; Jin Kocsis, Purdue University; Adam Kopper, Mercury Marine; Benji Maruyama; James Warren, National Institute of Standards and Technology
Monday 9:30 AM
April 4, 2022
Room: Riverboat
Location: Omni William Penn Hotel
Session Chair: Aagam Shah, University of Illinois at Urbana-Champaign
9:30 AM Break
10:00 AM Invited
Machine Learning Regression Models of Crystalline Materials Properties: Comparing Approaches and Results in Prediction Intervals Determination: Francesca Tavazza1; Kamal Choudhary1; Brian DeCost1; 1National Institute of Standards and Technology
Uncertainty quantification in AI-based predictions of material properties is of immense importance for the success and reliability of AI applications in material science. While confidence intervals are commonly reported for machine learning (ML) models, prediction intervals, i.e. the evaluation of the uncertainty on each prediction, are seldomly available. In this work we compare 3 different approaches to obtain such individual uncertainty, testing them on 12 ML-physical properties. Specifically, we investigated using the Quantile loss function, machine learning the prediction intervals directly, and using Gaussian Processes. We identify each approach’s advantages and disadvantages and compare their results. All data for training and testing were taken from the publicly available JARVIS-DFT database, and the codes developed for computing the prediction intervals are available through JARVIS-tools.
10:30 AM
Accelerating Phase-Field Predictions Via History-Dependent Neural Networks Learning Microstructural Evolution in a Latent Space: Remi Dingreville1; Chongze Hu1; Shawn Martin1; 1Sandia National Laboratories
The phase-field method is a powerful and versatile computational approach for modeling the evolution of the microstructure and properties of a wide variety of physical, chemical and biological systems. However, existing high-fidelity phase-field models are inherently computationally expensive, requiring high-performance computing resources and sophisticated numerical integration schemes to achieve useful degree of accuracy. In this talk I will discuss advanced in developing computationally inexpensive and accurate, data-driven surrogate model that directly learns the microstructural evolution of targeted systems by combining phase-field and history-dependent machine learning techniques. I discuss the advantages/disadvantages of combining various techniques to integrate low-dimensional description of the microstructure, obtained directly from phase-field simulations, with history-dependent deep neural network. Lasty, I will give examples on the performance and accuracy of the established machine-learning accelerated framework to predict the non-linear microstructure evolution as compared to high-fidelity phase-field.
10:50 AM
Semi-supervised Dynamic Sampling for 3D Electron Backscatter Diffraction: Zachary Varley1; Gregory Rohrer1; Marc De Graef1; 1Carnegie Mellon University
Electron backscatter diffraction (EBSD) is a popular microstructure analysis technique due to the direct, relatively high-resolution, measurement of phase and orientation. Three-dimensional EBSD (3D-EBSD) extends this microstructure analysis to a volume of material, measured in 2D sections, with a commensurate penalty in data acquisition time. The present work proposes a dynamic sampling algorithm to leverage the organized granular structure of orientation data present in microstructures, both within individual serial sections, and across consecutive ones. By using online machine learning to avoid redundant measurements, and then infilling unmeasured data during post-processing, significant theoretical savings are realized without offline training. User parameters allow tuning of the tradeoff between reconstruction accuracy and sampling speed. By coupling this sampling approach with real-time Kikuchi pattern indexing the authors aim to create a robust 3D-EBSD sampling technique which improves the Pareto frontier in the tradeoff between scan time and sample volume.
11:10 AM
A Graph Based Workflow for Extracting Grain-Scale Toughness from Meso-Scale Experiments.: Stylianos Tsopanidis1; Shmuel Osovski1; 1Technion Israel Institute of Technology/Faculty of Mechanical Engineering
Extracting the micro-scale material toughness is very challenging. This information is only accessible through delicate experiments, where the overall sampled volume and the experimental complexity limit the statistical assessment of the results and thus their validity. We introduce a novel machine learning computational framework that aims to compute the micro-scale material toughness, after a short training process on a limited meso-scale experimental dataset. This framework relies on the ability of a graph neural network to perform high accuracy predictions of the micro-scale material toughness, utilizing a limited size dataset. The merit of the proposed framework arises from the capacity to enhance its performance in different material systems with limited additional training on data obtained from experiments that do not require complex measurements. We demonstrate the algorithm’s high efficiency in predicting the crack growth resistance in micro-scale level, using a crack path trajectory limited to 200-300 grains for the network’s training.