ICME Gap Analysis in Materials Informatics: Databases, Machine Learning, and Data-Driven Design: Session I
Sponsored by: TMS Materials Processing and Manufacturing Division, TMS: Computational Materials Science and Engineering Committee, TMS: Integrated Computational Materials Engineering Committee
Program Organizers: James Saal, Citrine Informatics; Carelyn Campbell, National Institute of Standards and Technology; Raymundo Arroyave, Texas A&M University
Wednesday 8:30 AM
February 26, 2020
Room: 30D
Location: San Diego Convention Ctr
Session Chair: Raymundo Arroyave, Texas A&M; James Saal, Citrine Informatics
8:30 AM Invited
The MGI and ICME: James Warren1; 1National Institute of Standards and Technology
The Materials Genome Initiative (MGI) was, in no small part, built upon the success of ICME. Determining and remedying the gaps in the ICME infrastructure is fully aligned with MGI. In this presentation, I will discuss the new strategic directions within the MGI, and the increased emphasis on informatics as a tool for materials R&D. NIST has focused its support of the MGI on the management, quality, and use of data to accelerate the discovery, design, development, and deployment of new materials. While the program foresaw the day when data-driven materials science would be a rising paradigm for research, the recent demonstrated successes, in a variety of disciplines, of artificial intelligence algorithms has changed the conversation on the value of AI while simultaneously reinvigorating interest in related techniques. This confluence has left the MGI and the application of ICME perfectly poised to capitalize on these remarkable developments.
9:10 AM Invited
Polymer Informatics: Current Status & Critical Next Steps: Rampi Ramprasad1; Chiho Kim; 1Georgia Institute of Technology
The Materials Genome Initiative (MGI) has heralded a sea change in the philosophy of materials design. In an increasing number of applications, the successful deployment of novel materials has benefited from the use of computational, experimental and informatics methodologies. Here, we describe the role played by computational and experimental data generation and capture, polymer fingerprinting, machine-learning based property prediction models, and algorithms for designing polymers meeting target property requirements. These efforts have culminated in the creation of an online Polymer Informatics platform (https://www.polymergenome.org) to guide ongoing and future polymer discovery and design. Challenges that remain will be examined, and systematic steps that may be taken to extend the applicability of such informatics efforts to a wide range of technological domains will be discussed. These include strategies to deal with the data bottleneck, new methods to represent polymer morphology and processing conditions, and the applicability of emerging algorithms for design.
9:50 AM
Artificial Intelligence for Material and Process Design: Marius Stan1; Noah Paulson1; Elise Jennings1; Joseph Libera1; Richard Otis2; Brandon Bocklund3; Aaron Kusne4; James Warren4; Zi-Kui Liu3; Gregory Olson5; 1Argonne National Laboratory; 2Jet Propulsion Laboratory; 3Pennsylvania State University; 4National Institute of Standards and Technology; 5QuesTek Innovations LLC
We discuss examples of data-driven material design of alloys and ceramics for additive manufacturing, battery electrodes and computer memory gates. Data is generated via experiments and computer simulations that operate at various length and time scales, such as density functional theory, ab-initio molecular dynamics, phase field, finite elements, and CALPHAD. The experimental and computational information is analyzed using elements of Artificial Intelligence (AI), especially machine learning and computer vision algorithms, resulting in optimal material property models with credible uncertainty intervals estimated using Bayesian inference. We further employ AI to design experiments and optimize in real time complex synthesis processes, such as flame spray pyrolysis. The presentation includes a discussion of the impact of AI on science and technology in general and on material and process design in particular.
10:10 AM Break
10:30 AM Invited
Gaps, Limitations, and Pitfalls of Materials Informatics: Taylor Sparks1; 1University of Utah
Materials Informatics, or applying data science to address materials research challenges, has become an in important part of computational materials science and the Materials Genome Initiative. However, despite the enormous promise and potential of materials informatics, there still remains significant gaps, limitations, and pitfalls preventing successful implementation. In this talk I will go over several of these challenges and discuss best practices for materials informatics utilization. Specific examples will include the following: How to avoid model overfitting during hyper parameter tuning by incorporating leave-one-cluster-out cross validation. Insight from chemistry-encoded data visualization. Bias and imbalance in training data and fundamental data engineering approaches. Physics informed feature engineering. Regression versus classification when extrapolating using machine learning. Model deployment, repositories, and code availability. Using standardized benchmark data sets when developing features, algorithms, and cross-validation approaches.
11:10 AM Invited
View on Data Ecosystem of Materials: Zi-Kui Liu1; 1Pennsylvania State University
Computational thermodynamics based on CALPHAD (CALculation of PHAse Diagram) method models properties of individual phases and has proven to be foundational for Integrated Computational Materials Engineering and the Materials Genome Initiative and enables computational materials design in terms of chemistry and process (2009, http://dx.doi.org/10.1007/s11669-009-9570-6). To further enhance the predictive power of the CALPHAD method, new tools needed for connecting data repositories and provide feedbacks to create a sustainable ecosystem of materials data are discussed (2018, https://doi.org/10.1007/s11669-018-0654-z; 2019, https://doi.org/10.1557/mrc.2019.59). Furthermore, the synergy of the first-principles calculations for individual configurations and the CALPHAD modeling of individual phases consisting of individual configurations will be presented so that the emergent properties of individual phases that individual configurations do not possess can be predicted and tailored for materials design (2014, https://doi.org/10.1038/srep07043). The role of entropy of individual configurations and the statistics among the configurations on the synergy will also be articulated (2019, https://doi.org/10.1007/s11669-019-00736-w).