Machine learning assisted microtextured regions segmentation in a near-α titanium alloy

Developing automated, reliable, and advanced segmentation methodologies for identifying microtextured regions (MTRs) in complex titanium microstructures is imperative due to their substantial influence on dwell-fatigue performance. This paper introduces an innovative machine learning-assisted approach that combines the k-means and region-growing algorithms to automatically segment MTRs in a near-α titanium alloy, utilizing microstructural data acquired through electron backscatter diffraction (EBSD). The segmentation processes target MTRs at the center and edge of a billet, where the MTRs at the center are prominently visible while those at the edge are less distinct. The accuracy and precision of this methodology are comprehensively validated.