Q-RBSA: high-resolution 3D EBSD map generation using an efficient quaternion transformer network
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 (EB...
| Main Authors: | , , , , , , , |
|---|---|
| Format: | Article |
| Language: | English |
| Published: |
Nature Portfolio
2024-01-01
|
| Series: | npj Computational Materials |
| Online Access: | https://doi.org/10.1038/s41524-024-01209-6 |
| _version_ | 1797273925981306880 |
|---|---|
| author | Devendra K. Jangid Neal R. Brodnik McLean P. Echlin Chandrakanth Gudavalli Connor Levenson Tresa M. Pollock Samantha H. Daly B. S. Manjunath |
| author_facet | Devendra K. Jangid Neal R. Brodnik McLean P. Echlin Chandrakanth Gudavalli Connor Levenson Tresa M. Pollock Samantha H. Daly B. S. Manjunath |
| author_sort | Devendra K. Jangid |
| collection | DOAJ |
| description | 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. |
| first_indexed | 2024-03-07T14:51:06Z |
| format | Article |
| id | doaj.art-f94c3ce970d14265afe8001b9741d3eb |
| institution | Directory Open Access Journal |
| issn | 2057-3960 |
| language | English |
| last_indexed | 2024-03-07T14:51:06Z |
| publishDate | 2024-01-01 |
| publisher | Nature Portfolio |
| record_format | Article |
| series | npj Computational Materials |
| spelling | doaj.art-f94c3ce970d14265afe8001b9741d3eb2024-03-05T19:44:52ZengNature Portfolionpj Computational Materials2057-39602024-01-0110111010.1038/s41524-024-01209-6Q-RBSA: high-resolution 3D EBSD map generation using an efficient quaternion transformer networkDevendra K. Jangid0Neal R. Brodnik1McLean P. Echlin2Chandrakanth Gudavalli3Connor Levenson4Tresa M. Pollock5Samantha H. Daly6B. S. Manjunath7Electrical and Computer Engineering, University California Santa BarbaraMechanical Engineering, University California Santa BarbaraMaterials Department, University California Santa BarbaraElectrical and Computer Engineering, University California Santa BarbaraElectrical and Computer Engineering, University California Santa BarbaraMaterials Department, University California Santa BarbaraMechanical Engineering, University California Santa BarbaraElectrical and Computer Engineering, University California Santa BarbaraAbstract 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.https://doi.org/10.1038/s41524-024-01209-6 |
| spellingShingle | Devendra K. Jangid Neal R. Brodnik McLean P. Echlin Chandrakanth Gudavalli Connor Levenson Tresa M. Pollock Samantha H. Daly B. S. Manjunath Q-RBSA: high-resolution 3D EBSD map generation using an efficient quaternion transformer network npj Computational Materials |
| title | Q-RBSA: high-resolution 3D EBSD map generation using an efficient quaternion transformer network |
| title_full | Q-RBSA: high-resolution 3D EBSD map generation using an efficient quaternion transformer network |
| title_fullStr | Q-RBSA: high-resolution 3D EBSD map generation using an efficient quaternion transformer network |
| title_full_unstemmed | Q-RBSA: high-resolution 3D EBSD map generation using an efficient quaternion transformer network |
| title_short | Q-RBSA: high-resolution 3D EBSD map generation using an efficient quaternion transformer network |
| title_sort | q rbsa high resolution 3d ebsd map generation using an efficient quaternion transformer network |
| url | https://doi.org/10.1038/s41524-024-01209-6 |
| work_keys_str_mv | AT devendrakjangid qrbsahighresolution3debsdmapgenerationusinganefficientquaterniontransformernetwork AT nealrbrodnik qrbsahighresolution3debsdmapgenerationusinganefficientquaterniontransformernetwork AT mcleanpechlin qrbsahighresolution3debsdmapgenerationusinganefficientquaterniontransformernetwork AT chandrakanthgudavalli qrbsahighresolution3debsdmapgenerationusinganefficientquaterniontransformernetwork AT connorlevenson qrbsahighresolution3debsdmapgenerationusinganefficientquaterniontransformernetwork AT tresampollock qrbsahighresolution3debsdmapgenerationusinganefficientquaterniontransformernetwork AT samanthahdaly qrbsahighresolution3debsdmapgenerationusinganefficientquaterniontransformernetwork AT bsmanjunath qrbsahighresolution3debsdmapgenerationusinganefficientquaterniontransformernetwork |