Materials Informatics for Images and Multi-dimensional Datasets: Joint Session: "Materials Informatics for Images and Multi-dimensional Datasets" and "Materials Informatics and Modeling for 21st Century Ceramics Research”
Sponsored by: ACerS Basic Science Division, ACerS Electronics Division
Program Organizers: Amanda Krause, Carnegie Mellon University; Kristen Brosnan, General Electric Research; Alp Sehirlioglu, Case Western Reserve University
Wednesday 8:00 AM
November 4, 2020
Room: Virtual Meeting Room 42
Location: MS&T Virtual
Session Chair: Amanda Krause, Carnegie Mellon University; Ming Tang, Rice University
8:00 AM Invited
Computer Vision and Machine Learning for Microstructural Image Data: Elizabeth Holm1; 1Carnegie Mellon University
Microstructural images encode rich data sets that contain information about the structure, processing, and properties of the parent material. As such, they are amenable to characterization and analysis by data science approaches, including computer vision (CV) and machine learning (ML). In fact, they offer certain advantages compared to natural images, often requiring smaller training data sets and enabling more thorough assessment of results. Because CV and ML methods can be trained to reproduce human visual judgments, they can perform qualitative and quantitative characterization of complex microstructures, including segmentation, measurement, classification, and visual similarity tasks, in an objective, repeatable, and indefatigable manner. In addition, these approaches facilitate new characterization techniques that capitalize on the unique capabilities of computers to capture additional information compared to traditional metrics. Finally, ML can learn to associate microstructural features with materials processing or property metadata, providing physical insight into phenomena such as strength and failure.
8:30 AM Invited
Microstructure Representation for Physically Meaningful Descriptors: Olga Wodo; 1
The holy grail of materials science is to discover the mechanism governing the material properties and describe them in terms of a small set of physically meaningful descriptors. The discovery and exploration of materials and their properties critically depend on the availability of easily computable descriptors. In this talk, we present our framework to compute a library of generic descriptors for micrographs. We describe our microstructure representation that is based on the graph and skeleton and enables microstructure characterization in terms of shape (i.e., morphology), geometry, and connectedness (i.e., topology). We explain how this work lays the foundation for machine learning of microstructure-property relationships and enables information fusion between multiple scales. We showcase our framework using examples from organic electronics.
9:00 AM Invited
Incorporating Materials Physics into Imaging Algorithms for Microscope Image Interpretation: Jeff Simmons1; 1U.S. Air Force Research Laboratory
In microscopy, the signal recorded for each pixel is composed of the true signal plus noise. No matter how much data is collected, there will always be at least twice as many parameters to estimate as there are pixel measurements, Microscopy, then, constitutes ill-posed problems with many equally valid solutions to the governing equations. Machine learning excels at solving ill-posed problems, but using physics allows us to exceed the performance of off-the-shelf machine learning algorithms. This can be in the form of either forward modeling of the true signal or by “regularizing.” Regularizing allows for a rationale for choosing a particular solution among the many valid ones. This presentation gives examples of phase field regularization for polycrystalline SiC, of fluid dynamics analogues to continuous fibers in SiC/SiC composite materials, and of object symmetries, as predicted from the Curie Principle, used for tuning deep neural networks for Ni dendrite core detection.
9:30 AM Invited
Accelerate TEM and Tomography Imaging by Deep-learning Enabled Compressive Sensing and Information Inpainting in High-dimensional Manifold: Huolin Xin1; 1University of California, Irvine
Deep learning schemes have already impacted areas such as cognitive game theory (e.g., computer chess and the game of Go), pattern (e.g., facial or fingerprint) recognition, event forecasting, and bioinformatics. They are beginning to make major inroads within physics, chemistry and materials sciences and hold considerable promise for accelerating the discovery of new theories and materials. In this talk, I will introduce deep convolutional neural networks, and how they can be applied to decoding time-domain compressed TEM video streams to improve time resolution by orders of magnitude and how it can solve the missing-wedge problem in tomographic imaging through information inpainting in a high-dimensional manifold.
10:00 AM Invited
FAIR Digital Object Framework and High Throughput Experiment: Zachary Trautt1; Raymond Plante1; Gretchen Greene1; Jason Hattrick-Simpers1; Brian DeCost1; Aaron Kusne1; Andriy Zakutayev2; 1National Institute of Standards and Technology; 2National Renewable Energy Laboratory
With the increasing use of data-driven methodologies, concerns around data discovery, data access, and data interoperability have come to the forefront. Beginning in August 2019, communities have convened to work towards convergence of three complementary visions: (1) Digital Object Architecture, (2) Linked Data and Semantic Web, (3) FAIR Data Principles. This convergence has established the FAIR Digital Object Framework. This talk will overview these developments, summarize work within the NIST Material Measurement Laboratory to support adoption of the FAIR Digital Object Framework within the materials science and engineering community, and provide practical examples of how researchers can leverage these developments.