ICME 2023: Scientific Workflows for ICME: I (Microstructure)
Program Organizers: Charles Ward, AFRL/RXM; Heather Murdoch, U.S. Army Research Laboratory
Tuesday 8:00 AM
May 23, 2023
Room: Caribbean VI & VII
Location: Caribe Royale
Session Chair: Michael Tonks, University of Florida
8:00 AM Invited
NexusLIMS: A Laboratory Information Management System for Shared-use Electron Microscopy Facilities: June Lau1; 1NIST
Presently, the management and the processing of research data produced by scientific instrumentation is a significant challenge. Electron microscopy researchers work with different make and model of microscopes and detectors, and with these, data produced in various (sometimes proprietary) formats. Users often implement ad hoc strategies to curate and disseminate this data. Often, data and the context under which they were acquired is irrecoverable beyond the few published images in journal articles. To capture microscopy data consistent with the FAIR data principles, we built NexusLIMS, a laboratory information management system managing microscopy output across two NIST campuses. While this example may be microscopy-specific, many of the components and concepts within are extensible to data produced by other instruments. We will show additional data infrastructure components that are co-integrated with NexusLIMS, such as central file servers, schedulers, and compute resources, and how this integration has nucleated an ecosystem for research data.
8:30 AM
Automated Analysis Pipeline to Investigate Bond-wire Corrosion Under Salt-water Exposure
: Jayvic Cristian Jimenez1; Liangyi Huang2; Kristen J. Hernandez1; Harsha Madiraju1; Pawan K. Tripathi1; Alp Sehirlioglu1; Roger H. French1; 1Case Western Reserve University; 2Arizona State University
X-ray computed tomography (XCT) is a powerful tool for studying corrosion of commercial bond-wires. This investigation amasses large amounts of image data. Analysis of a 3D-rendered object can be computationally costly and time-consuming, while performing the task manually is impractical. The cylindrical geometry of commercial bond-wires is challenging to characterize as 3D renders, adding to the computational complexity of the analysis. Developments in computer vision tools, which leverage convolutional neural networks (CNNs) are computationally efficient and fast, making them desirable tools for automated feature extraction. In our work, we demonstrate an automated workflow for transforming a cylindrical object into a 2D representation and performing background denoising that allows for full surface view for further characterization. We integrated semantic segmentation algorithms such as DeepLab into our workflow pipeline allowing for further characterization of the surface features of the bond-wires. SNL is managed and operated by NTESS under DOE NNSA contract DE-NA0003525.
8:50 AM
Microscopy Data Acquisition and Analysis Workflows for Microstructure Quantification: Michael Uchic1; 1Air Force Research Laboratory
Quantitative surface-based microscopy measurements of key and ubiquitous microstructural features (e.g., grains, fibers, internal porosity) is a foundational activity in materials engineering to establish process-to-structure-to-property relationships, as well as to provide ground truth measurements for other purposes (such as advancing nondestructive characterization methods). The need for rapid and unbiased microstructure measurements has spurred the materials community over the past three decades to develop automated microscopy methodologies to provide such information with little-or-no human intervention, which critically rely on robust data acquisition and data post-processing workflows. This contribution will present an overview of such workflows for microstructure quantification in 2D and 3D of structural materials using both reflective light and scanning electron microscope-based techniques, and highlight examples of quantitative measurements of microscale features and defects within mm-to-cm scale volumes. In addition, the contribution will discuss potential areas of improvement to further enhance digital microscopy workflows.
9:10 AM
Image Processing Pipeline for Fluoroelastomer Crystallite Detection in Atomic Force Microscopy Images: Mingjian Lu1; Sameera Venkat1; Jube Augustino1; Jayvic Jimenez1; Pawan Tripathi1; Yinghui Wu1; Roger French1; Laura Bruckman1; 1Case Western Reserve University
Fluoroelastomer crystallization can be easily observed using atomic force microscopy, AFM, to look at surface properties and macro-scale morphologies. In-situ measurements investigating phase-transition kinetics of fluoropolymers, under isothermal heating, generate a large dataset of time-lapsed image sequences. Interpretation of the resulting images is guided by domain-knowledge and image processing is done manually using software, which is time-consuming. In our work, we integrate automated image detection and image segmentation methods, based on convolutional neural networks in our image processing. The resulting pipeline is an end-to-end framework, which aims to automatically classify and analyze images as part of batch processing. The product of this framework can extract individual crystallites and track their growth throughout the course of the image sequence with only a few training data. Then, statistical analysis can then be incorporated opening opportunities to investigate fluoroelastomer crystallization kinetics.
9:30 AM
Towards Interoperability: Digital Representation of a Material Specific Characterization Method: Bernd Bayerlein1; Ghezal Ahmad Zia1; Markus Schilling1; Philipp von Hartrott2; Jörg Waitelonis3; Thomas Hanke2; Birgit Skrotzki1; 1Bundesanstalt für Materialforschung und -prüfung (BAM); 2Fraunhofer-Institut für Werkstoffmechanik (IWM); 3Leibniz-Institut für Informationsinfrastruktur (FIZ)
Certain metallic materials gain better mechanical properties through controlled heat treatments. In age-hardenable aluminum alloys, the strengthening mechanism is based on the controlled formation of nanometer sized precipitates, which hinder dislocation movement. Analysis of the microstructure and especially the precipitates by transmission electron microscopy allows identification of precipitate types and orientations. Dark-field imaging is often used to image the precipitates and quantify their relevant dimensions.The present work aims at the digital representation of this material-specific characterization method. Instead of a time-consuming, manual image analysis, a digital approach is demonstrated. The integration of an exemplary digital workflow for quantitative precipitation analysis into a data pipeline concept is presented. Here ontologies enable linking of contextual information to the resulting output data in a triplestore. Publishing digital workflow and ontologies ensures the reproducibility of the data. The semantic structure enables data sharing and reuse for other applications and purposes, demonstrating interoperability.
9:50 AM Break