Q-RBSA: High-Resolution 3D EBSD Map Generation Using An Efficient Quaternion Transformer Network

Devendra K. Jangid,Neal R. Brodnik,McLean P. Echlin,Tresa M. Pollock,Samantha H. Daly,B.S. Manjunath

2023-03-20

Abstract:Gathering 3D material microstructural information is time-consuming, expensive, and energy-intensive. Acquisition of 3D data has been accelerated by developments in serial sectioning instrument capabilities; however, for crystallographic information, the electron backscatter diffraction (EBSD) imaging modality remains rate limiting. We propose a physics-based efficient deep learning framework to reduce the time and cost of collecting 3D EBSD maps. Our framework uses a quaternion residual block self-attention network (QRBSA) to generate high-resolution 3D EBSD maps from sparsely sectioned EBSD maps. In QRBSA, quaternion-valued convolution effectively learns local relations in orientation space, while self-attention in the quaternion domain captures long-range correlations. We apply our framework to 3D data collected from commercially relevant titanium alloys, showing both qualitatively and quantitatively that our method can predict missing samples (EBSD information between sparsely sectioned mapping points) as compared to high-resolution ground truth 3D EBSD maps.

Machine Learning,Artificial Intelligence,Computer Vision and Pattern Recognition

What problem does this paper attempt to address?

The paper aims to address the time and cost issues in generating three-dimensional Electron Backscatter Diffraction (3D EBSD) maps. Specifically, the research team proposed a physics-inspired deep learning framework based on an efficient Quaternion Residual Block Self-Attention Network (QRBSA), which is used to generate high-resolution 3D EBSD maps from sparse slices of 3D EBSD mapping. This method utilizes quaternion convolution to effectively learn local relationships in orientation space and captures long-range correlations through self-attention in the quaternion domain. The researchers applied this framework to the 3D data of commercially relevant titanium alloy materials, demonstrating that their method can qualitatively and quantitatively predict missing samples (i.e., missing EBSD information between mapping points of sparse slices) compared to high-resolution real 3D EBSD maps.

The core contribution of the paper is the proposal of the QRBSA network, a deep learning architecture that combines quaternion convolution and self-attention mechanisms, specifically designed to handle the multidimensional nature of EBSD data and the embedded crystallographic information. Quaternion representation is particularly suitable for expressing crystal orientations, and quaternion convolution layers can effectively encode prior orientation information into the convolution kernels, reducing the number of training parameters. Moreover, by introducing self-attention mechanisms, the network can learn global statistical information and long-range correlations between feature maps, thereby improving the super-resolution recovery capability for EBSD mapping.

Experimental results show that the framework can not only visually produce predictions similar to real EBSD maps but also performs excellently on quantitative metrics such as Peak Signal-to-Noise Ratio (PSNR) and Structural Similarity Index Measure (SSIM). Especially when dealing with Ti-6Al-4V alloys that have more complex textures and smaller grain features, despite facing greater challenges, the QRBSA network still demonstrated good performance, indicating that the framework has broad applicability for different types of material structures.

Overall, the study significantly enhances the efficiency and accuracy of 3D EBSD map generation by developing a novel deep learning approach, providing a powerful tool for materials science research, especially in exploring key 3D microstructural information of materials under extreme environments.