ICME 2023: AI/ML: Properties I
Program Organizers: Charles Ward, AFRL/RXM; Heather Murdoch, U.S. Army Research Laboratory
Monday 1:10 PM
May 22, 2023
Room: Boca I-III
Location: Caribe Royale
Session Chair: Shankarjee Krishnamoorthi, ATI Specialty Materials
1:10 PM Invited
Artificial Intelligence and High-performance Data Mining for Accelerating Materials Discovery and Design: Ankit Agrawal1; 1Northwestern University
The increasing availability of data from the first three paradigms of science (experiments, theory, and simulations), along with advances in artificial intelligence and machine learning (AI/ML) techniques has offered unprecedented opportunities for data-driven science and discovery, which is the fourth paradigm of science. Within the arena of AI/ML, deep learning (DL) has emerged as a game-changing technique in recent years with its ability to effectively work on raw big data, bypassing the (otherwise crucial) manual feature engineering step traditionally required for building accurate ML models, thus enabling numerous real-world applications, such as autonomous driving. In this talk, I will present ongoing AI research in our group with illustrative applications in materials science. In particular, we will discuss approaches to gainfully apply AI/ML/DL on big data as well as small data in the context of materials science. I will also demonstrate some of the materials informatics tools developed in our group.
1:40 PM
Managing Uncertainty in the Strength of Ceramics: Eric Walker1; Jason Sun1; James Chen1; 1University at Buffalo
A known challenge in decision-based manufacturing of ceramics is that there are input uncertainties from data, models and model parameters. In a decision-based manufacturing environment, it is critical to know the probability that a ceramic will have a flexural strength below an acceptable limit. Regarding the flexural strength of silicon carbide (α-SiC), an absolute error bound of ±15% was set as a baseline previously in literature. However, this is inadequate because the uncertainty can shift both in its mean and standard deviation due to manufacturing-caused changes in microstructure and porosity. A Bayesian-based approach is proposed to dynamically and precisely provide a live uncertainty quantification (UQ) in a smart manufacturing environment. Five scenarios with variably perturbed flexural strength data are demonstrated. Across the scenarios the goodness-of-fit to the mean, as measured by R^2, is improved by a range from 0.12-0.61. Also, uncertainty is reduced from ±15% to ±12-±8% depending upon scenario.
2:00 PM
Discovery of Multi-functional Polyimides through High-throughput Screening using Explainable Machine Learning: Ying Li1; 1University of Wisconsin-Madison
Aiming to expedite the discovery of high-performance polyimides, we utilize computational methods of machine learning (ML) and molecular dynamics (MD) simulations. We first build a comprehensive library of more than 8 million hypothetical polyimides based on the polycondensation of existing dianhydride and diamine/diisocyanate molecules. Then we establish multiple ML models for the thermal and mechanical properties of polyimides based on their experimentally reported values, including glass transition temperature, Young’s modulus, and tensile yield strength. The obtained ML models demonstrate excellent predictive performance in identifying the key chemical substructures influencing the thermal and mechanical properties of polyimides. Applying the well-trained ML models, we obtain property predictions of the 8 million hypothetical polyimides. Then, we screen the whole hypothetical dataset and identify three (3) best-performing novel polyimides that have better-combined properties than existing ones through Pareto frontier analysis.
2:20 PM
A Machine Learning-based Virtual Lab to Predict Yield Surfaces from Crystal Plasticity Simulation: Anderson Nascimento1; Sharan Roongta2; Martin Diehl3; Irene Beyerlein1; 1University of California, Santa Barbara; 2Max-Planck-Institut für Eisenforschung; 3Katholieke Universiteit Leuven
At the continuum level, the plastic anisotropy of a wide range of metals and alloys is well described by advanced phenomenological yield surfaces. Relevant difficulties in their usage, however, are associated with the non-trivial parameter identification process and the non-uniqueness of the anisotropy coefficients. Alternative avenues for plastic flow prediction have been studied, and machine learning based approaches have gained notoriety due to their high fitting capabilities. We present a Machine Learning Virtual Lab for yield surface prediction, a framework that integrates crystal plasticity with deep neural network models and provides performance comparable to 3D yield functions. Important features such as well-defined flow vector, convexity, and multi-axial yield prediction are analyzed and compared against benchmark yield criteria.
2:40 PM
Intelligent Design & Manufacturing of High-performance Iron Castings Using AI/ML: Jiten Shah1; 1Product Development and Analysis (PDA) LLC
High performance iron castings design and manufacturing have many variables with uncertainty impacting the casting quality and performance. A framework is developed for building meta models utilizing historical data and generating required missing data under controlled experiments and ICME based predictions. Author will present the meta models developed using AI/ML for predicting properties and porosity for sand cast ductile iron castings. For the design community, models assist predicting properties as a function of section thickness, feature orientation with respect to gravity and type of mold media coupled with processing parameters that will be presented. These meta models are demonstrated for a near real time intelligent processing of the castings in a production environment.
3:00 PM Break