Boundary identification in EBSD data with a generalization of fast multiscale clustering.

Area of Science:

• Materials Science
• Crystallography
• Data Analysis

Background:

• Electron backscatter diffraction (EBSD) is crucial for analyzing microstructures.
• Studying cellular and subgrain microstructures presents challenges due to gradual orientation transitions, unlike coarse-grained materials.
• Existing methods struggle to accurately delineate boundaries in microbands and similar structures.

Purpose of the Study:

• To develop an improved segmentation technique for EBSD data with gradual crystallographic orientation transitions.
• To adapt the Fast Multiscale Clustering (FMC) algorithm for EBSD data analysis.
• To enable accurate segmentation of complex microstructural features.

Main Methods:

• Combined Fast Multiscale Clustering (FMC) with quaternion representation of crystallographic orientation.
• Implemented a novel distance function, a modification of Mahalanobis distance, to handle varying boundary magnitudes.
• Developed a new variance update rule to maintain linear runtime for FMC with quaternion data.
• Generalized the FMC algorithm for both 2D and 3D EBSD datasets.

Main Results:

• Successfully segmented EBSD data exhibiting gradual transitions in crystallographic orientation.
• The modified FMC algorithm effectively handled datasets with both high and low magnitude boundaries.
• Demonstrated the generalization of FMC for analyzing 3D EBSD datasets.
• Presented several example segmentations of quaternion EBSD data.

Conclusions:

• The adapted FMC algorithm provides an effective solution for segmenting EBSD data with gradual orientation changes.
• This approach enhances the study of complex microstructures, including cellular and subgrain formations.
• The method is applicable to both 2D and 3D EBSD datasets, offering broad utility in materials science research.