Machine Learning-Based Prediction of Grain Size from Colored Microstructure
Source abstract
Abstract
We constructed a convolutional neural network to estimate average grain size from microstructure images. In the previous study from our research group, the network was trained using GB-type images in which the grain matrix and grain boundary were represented in white and black, respectively. The model well estimated the same GB-type images, but did not properly predict CL-type images where grain boundaries were defined by color contrast between grains. In the present study, the convolutional neural network was trained using CL-type microstructure images, and evaluated the average grain size, for comparison with the previous GB-type model. The relationship between the microstructure image and the average grain size was determined using regression rather than classification. Then, the results were compared with the previous ones. Finally, the proposed approach was used for actual microstructural image analysis. Mid-layer images were extracted to examine how the network recognizes the characteristics of microstructures, such as grain color and grain boundary. Like the previous GB-type model, the present CL-type model seems to estimate the average grain size from the curvature of the grain boundary through edge detection of the grain boundaries. However, the GB- and CL-type models only properly predicted the grain size from the same kind of images as the training data, because the definitions of the grain boundaries of the two models were different.
Source: crossref · Quality: baseline
Research context
OpenAlex topics: Machine Learning in Materials Science · Mineral Processing and Grinding · X-ray Diffraction in Crystallography
SciVX discovery tags are terminology-based. OpenAlex topics are source-provided classifications.
References (27)
Show source-provided references
1. Saunders · 2018 · 1751
2. 10.2172/86139710.2172/861397 ↗
3. 10.1016/j.proeng.2014.09.14110.1016/j.proeng.2014.09.141 ↗
4. 2022
5. 10.1016/j.pmatsci.2018.01.00510.1016/j.pmatsci.2018.01.005 ↗
6. Chollet · 2018
7. 10.1016/j.commatsci.2018.03.00410.1016/j.commatsci.2018.03.004 ↗
8. 10.3365/kjmm.2019.57.3.18410.3365/kjmm.2019.57.3.184 ↗
9. 10.1016/j.actamat.2017.09.00410.1016/j.actamat.2017.09.004 ↗
10. 10.3390/cryst1109106010.3390/cryst11091060 ↗
11. 10.3390/met1105075610.3390/met11050756 ↗
12. 10.3365/kjmm.2021.59.6.43010.3365/kjmm.2021.59.6.430 ↗
13. 10.48550/ARXIV.2005.0963410.48550/ARXIV.2005.09634 ↗
14. 10.3390/met1110154710.3390/met11101547 ↗
15. 10.3390/met1010138110.3390/met10101381 ↗
16. 10.1016/j.compind.2005.05.01210.1016/j.compind.2005.05.012 ↗
17. 10.3390/ma1519695410.3390/ma15196954 ↗
18. Kim · 2006 · 061605
19. Kamachali · 2012 · 2719
20. 10.1016/j.commatsci.2010.09.02310.1016/j.commatsci.2010.09.023 ↗
21. LeCun · 2004 · 97
22. Wasserman · 1993
23. Saito · 2016
24. 2021
25. 2022
26. 10.1109/tpami.2018.287884910.1109/tpami.2018.2878849 ↗
27. 10.1002/cta.21010.1002/cta.210 ↗