Application of Synthetically Trained Three-dimensional U-Net to the Detection of Moving Subpixel Objects
Document Type
Article
Publication Date
5-23-2025
Abstract
Hybrid Monte Carlo and molecular dynamics simulations were used to investigate the interaction of light interstitials in multi-element Ni-based alloys. We show that light interstitials such as boron and oxygen fundamentally alter interfacial chemistry by reshaping alloy-element distribution and segregation. Oxygen adsorption drove boron migration from the grain boundary to the free surface, where it co-enriched with Cr, Fe, and Mo and formed BO3 trigonal motifs embedded within mixed-metal oxide networks. Oxygen also promoted M-O-M chain formation, including Nb2O5 clusters at the free surface. In the absence of oxygen, boron segregated to the grain boundary, altering local metal chemistry and underscoring a dynamic, environment-sensitive behavior. Following chlorine exposure, the oxidized surfaces retained strong O-mediated connectivity while forming new Cl-M associations, particularly with Nb and Cr, and exhibited further surface enrichment in Cr, Fe, and Mo. High-temperature MD simulations revealed a dynamic tug-of-war: chlorine exerted upward pull and disrupted weakly anchored sites, while Nb- and BO3-rich oxide motifs resisted deformation. A new stabilization mechanism was identified in which subsurface boron atoms anchored overlying Cr centers, suppressing their mobility and mitigating chlorine-driven displacement. These results demonstrate boron's dual role as a modifier of alloy-element segregation and a stabilizer of oxide networks, and identify Nb as a key element in reinforcing cohesion under halogen attack. More broadly, this study highlights the need to track light interstitial cross-talk and solute migration under reactive conditions, offering atomistic criteria for designing corrosion-resistant surface chemistries in Ni-based superalloys exposed to halogenated or oxidative environments.
Source Publication
Optical Engineering (ISSN 0091-3286 | eISSN 1560-2303)
Recommended Citation
Katherine E. Eismann, Kenneth M. Hopkinson, Bryan J. Steward, and Shannon R. Young "Application of synthetically trained three-dimensional U-Net to the detection of moving subpixel objects," Optical Engineering 64(9), 092204 (23 May 2025). https://doi.org/10.1117/1.OE.64.9.092204
Comments
The "Link to Full Text" on this page opens or downloads the full article at the publisher, SPIE.
This article may have a limited time access period mandated by the publisher or research funding body.
Co-author K. Eismann was an AFIT PhD student at the time of this article.