Data Science and Analytics for Materials Imaging and Quantification: Poster Session
Sponsored by: TMS Structural Materials Division, TMS Materials Processing and Manufacturing Division, TMS: Advanced Characterization, Testing, and Simulation Committee, TMS: Integrated Computational Materials Engineering Committee
Program Organizers: Emine Gulsoy, Northwestern University; Charudatta Phatak, Argonne National Laboratory; Stephan Wagner-Conrad, Carl Zeiss Microscopy; Marcus Hanwell, Brookhaven National Laboratory; David Rowenhorst, Naval Research Laboratory; Tiberiu Stan, Asml
Monday 5:30 PM
March 15, 2021
Room: RM 16
Location: TMS2021 Virtual
High Dimensional Analysis of Abnormal Grain Growth under Dynamic Annealing Conditions: Matthew Higgins1; Jiwoong Kang1; Ning Lu1; He Liu2; Robert Suter2; Ashwin Shahani1; 1University of Michigan; 2Carnegie Mellon University
Under specific annealing conditions, a few grains in polycrystalline structures may significantly outgrow other grains. This process is known as abnormal grain growth (AGG). Unfortunately, the phenomenological aspects of AGG remain a mystery. Synchrotron-based characterization methods such as high energy diffraction microscopy (HEDM) are eroding long-standing barriers to understand the temporal evolution of 3D microstructures. Here, we imaged via HEDM a Cu-17Al-11.4Mn alloy undergoing a cyclical, non-isothermal heat treatment. From the reconstructions, we measured the microstructural characteristics to understand the potential for a given grain to grow abnormally, and the complex interplay between precipitation, dissolution, and grain growth. We quantified a set of 14 independent microstructural features potentially contributing to the growth dynamics, thus necessitating dimensionality reduction. By applying principal component analysis and outlier/novelty detection methods, we discovered the key features that set abnormal grains apart, enabling them to eventually consume the microstructure.
Quantitative EBSD Image Analysis and Prediction via Deep Learning: Yi Han1; Joey Griffiths1; Yunhui Zhu1; Hang Yu1; 1Virginia Tech
In this work, we demonstrate a deep learning based approach to quantitatively analyze and characterize the variation of microstructure from a large dataset of material imaging. Metal samples processed via the Additive Manufacturing (AM) technique known as the additive friction stir deposition (AFSD) are used to validate our approach. The microstructure images obtained from electron backscatter diffraction (EBSD) are processed through a deep neural network called VGG16 to generate high-dimensional features, then a set of low-dimensional principal microstructure descriptors are extracted to represent the key differences among the analyzed microstructures, allowing for quantitative comparison between existing microstructures as well as prediction of new microstructure within the domain spanned by the principal descriptors. This allows us to explore the physical insights into the implicitly expressed microstructure descriptors by mapping the regenerated microstructures as a function of principal component values.
Understanding Powder Morphology and Its Effect on Flowability Through Machine Learning in Additive Manufacturing: Srujana Rao Yarasi1; Andrew Kitahara1; Anthony Rollett1; Elizabeth Holm1; 1Carnegie Mellon University
The use of computer vision and machine learning tools in the additive manufacturing domain have enabled the quantitative investigation of qualitative factors like powder morphology, which affect the flowability in powder bed fusion processes. Flowability is measured through rheological experiments conducted with the FT4 rheometer and the Granudrum. Convolutional Neural Networks (CNN) are used to generate feature descriptors of the powder feedstock, from SEM images, that describe not just the particle size distribution but also the sphericity, surface defects, and other morphological features of the powder particles. These descriptors are then correlated to their respective flowability properties for numerous powder systems to evaluate powder performance in an AM powder bed fusion machine. This framework is intended to be a powder qualification system that can differentiate between powder systems and serve as a method to indicate the usability of recycled powder lots, for instance.