Extracting Grain Orientations from EBSD Patterns of Polycrystalline Materials Using Convolutional Neural Networks.

Area of Science:

• Materials Science
• Computational Materials Science
• Crystallography

Background:

• Electron Backscatter Diffraction (EBSD) is crucial for analyzing material microstructures.
• Accurate indexing of EBSD patterns is essential for reliable crystallographic orientation determination.
• Current dictionary-based indexing methods have limitations in speed and accuracy.

Purpose of the Study:

• To introduce a novel deep learning approach for indexing EBSD patterns.
• To enhance the accuracy of crystal orientation prediction from EBSD data.
• To compare the performance of the deep learning method against traditional dictionary-based indexing.

Main Methods:

• A deep convolutional neural network (CNN) architecture was designed and implemented.
• A differentiable approximation of the disorientation function was developed for model optimization.
• Stochastic gradient descent was used to train the deep learning model on simulated EBSD data.
• The model was evaluated using experimental EBSD patterns of polycrystalline nickel.

Main Results:

• The deep learning model achieved a mean disorientation error of 0.548°.
• Dictionary-based indexing resulted in a mean disorientation error of 0.652°.
• The deep learning approach demonstrated superior accuracy in crystal orientation determination.

Conclusions:

• Deep learning offers a more accurate method for indexing EBSD patterns compared to dictionary-based approaches.
• The developed CNN architecture effectively predicts crystal orientation from EBSD data.
• This advancement has significant implications for materials characterization and analysis.