Skip to Main Content
 

Global Search Box

 
 
 
 

Files

File List


BrustThesis.pdf (62.75 MB)

ETD Abstract Container

Abstract Header

Applications of Graph Cutting for Probabilistic Characterization of Microstructures in Ferrous Alloys

Brust, Alexander Frederick
http://rave.ohiolink.edu/etdc/view?acc_num=osu1555523646156822

Abstract Details

2019, Doctor of Philosophy, Ohio State University, Materials Science and Engineering.
Processing of martensitic steels requires a thermally driven phase transformation into the austenite phase field, where rapid cooling initiates the diffusionless transformation into martensite. The resultant microstructural constituent is a hard, brittle phase that requires subsequent heat treatment to soften the material for optimized mechanical properties. Although the transformation microstructure has the largest influence on these mechanical properties, the prior austenite microstructure has been shown to significantly affect the final product material in the form of ductile to brittle fracture occurrence, classification of creep and cavitation sites, increasing martensite packet and block sizes resulting in Hall-Petch effects, and temper embrittlement. Therefore, reconstruction of the prior austenite phase field can help optimize both the processing of a sample steel or binary ferrous alloy and predicative examinations on the material. However, analysis of the austenite to martensite transformation is hindered by the large volume of noise associated with the transformation. This can be attributed to the scale of the transformation, which results in a single prior austenite grain producing up to 24 martensitic variants; the plasticity associated with the massive formation of martensite; variations in the orientation relationship across variable compositions and morphologies; errors associated with the EBSD-indexing of the transformation microstructure; and annealing twins forming across the prior austenite microstructure. Due to the inherent noise associated with the transformation, modern reconstruction algorithms using point-to-point or flood-fill algorithms struggle to produce accurate and consistent reconstructions of the austenite microstructure. We therefore propose a probabilistic approach to austenite reconstruction in steels and ferrous alloys based on the graph cutting algorithm. This technique can be applied to a number of inverse problems in materials science, such as image segmentation, microstructure phase and constituent segmentation, atomic cluster identification from atom probe tomography data sets, and the reconstruction of the parent microstructure from the EBSD-indexed post-transformation data set. The chosen algorithm used an energy-minimization technique known as graph cutting to perform the reconstructions. In order to most properly describe the algorithm, information related to the misorientation relationships between martensite variants associated with the same prior austenite grain or twin were utilized. Additionally, an accurate and automated measurement of the orientation relationship for the desired steel data sets was performed through a Bayesian implementation and used conditionally within the reconstruction algorithm. This information was used to perform automated reconstructions on the prior austenite microstructure from transformation martensite. Analysis of a number of steel and binary ferrous alloy data sets with variable orientation relationships was then performed to assess the range of the technique, as well as an analysis on the amount of noise that could be handled within a single data set and the identification of twins across the microstructure. Additionally, segmentations were performed to extract the packet boundaries residing within both a large and small PAG, as well as an analysis on a parent-twin system to identify packets containing shared variants. It was found that the technique can sufficiently produce reasonable, probabilistically defined reconstructions of the prior austenite microstructure, even with the addition of copious amounts of noise. Additionally, it further separated itself from existing techniques through its ability to capture a number of prior austenite twins spanning across the parent grains. Validation was performed through both comparisons to an optical micrograph of the same chemically etched steel and comparison of retained austenite orientations associated with surrounding, reconstructed grains. A comparison to a manual segmentation was also performed, along with crystallographic analysis on the packet segmentations in order to compare the experimentally segmented martensite orientations with their theoretical counterparts.
Stephen Niezgoda, PhD (Advisor)
Eric Payton, PhD (Committee Member)
Antonio Ramirez, PhD (Committee Member)
Yunzhi Wang, PhD (Committee Member)
Nicholas Brunelli, PhD (Committee Member)
276 p.
Engineering; Materials Science
Martensite; Austenite; Steel; Austenite Reconstruction; Crystallography; Misorientation; Graph Cutting; Image Segmentation; Clustering; Microstructure; Transformation; Variant; EBSD Analysis; Packet Segmentation; Sub-block Block Packet Boundary;

Recommended Citations

Citations

  • Brust, A. F. (2019). Applications of Graph Cutting for Probabilistic Characterization of Microstructures in Ferrous Alloys [Doctoral dissertation, Ohio State University]. OhioLINK Electronic Theses and Dissertations Center. http://rave.ohiolink.edu/etdc/view?acc_num=osu1555523646156822

    APA Style (7th edition)

  • Brust, Alexander. Applications of Graph Cutting for Probabilistic Characterization of Microstructures in Ferrous Alloys. 2019. Ohio State University, Doctoral dissertation. OhioLINK Electronic Theses and Dissertations Center, http://rave.ohiolink.edu/etdc/view?acc_num=osu1555523646156822.

    MLA Style (8th edition)

  • Brust, Alexander. "Applications of Graph Cutting for Probabilistic Characterization of Microstructures in Ferrous Alloys." Doctoral dissertation, Ohio State University, 2019. http://rave.ohiolink.edu/etdc/view?acc_num=osu1555523646156822

    Chicago Manual of Style (17th edition)

osu1555523646156822
146
© 2019, all rights reserved.
This open access ETD is published by The Ohio State University and OhioLINK.