Perm National Research Polytechnic University (Perm, Russia) :

A. V. Klyuev, Cand. Phys.-Math., Associate Prof., Dept. of Computational Mathematics, Mechanics, and Biomechanics, e-mail: kav@gelicon.biz
V. Yu. Stolbov, Dr. Eng., Head of Dept. of Computational Mathematics, Mechanics and Biomechanics, e-mail: valeriy.stolbov@gmail.com

Nosov Magnitogorsk State Technical University (Magnitogorsk, Russia):

N. V. Koptseva, Dr. Eng., Professor, Dept. of Foundry Processes and Materials Science, e-mail: kopceva1948@mail.ru
Yu. Yu. Efimova, Cand. Eng., Associate Prof., Dept. of Materials Processing Technologies, e-mail: jefimova78@mail.ru

Currently, the creation of new functional materials is an urgent task of modern materials science. The selection of the chemical composition and methods of thermal and thermomechanical processing, which lead to the creation of new steels with desired properties, require significant time and financial costs for research and preparation of the production technology for such materials. In this case, the analysis of microstructure images, which makes it possible to predict the spectrum of the mechanical properties of the material, is of particular importance. In order to increase the efficiency and objectivity of the properties identification, it is necessary to use up-to-date mathematical methods of data processing and artificial intelligence algorithms to solve the Iss.s of classification and identification of the material microstructure. The work explores the capabilities of artificial neural networks and deep machine learning to predict the mechanical properties of steels after various deformation and thermal actions based on the neural network analysis of microstructures images. A description of the constructed database of marked-up digital photographs of steel microstructures used in deep training of neural networks is given. It is shown that the rather popular deep neural VGG network with high accuracy solves the task of classifying the microstructures of various steels according to the hardness of the material. The results of processing the initial information by the deep neural ResNet network are presented and a comparison of the results with experimental data is given. The achieved accuracy allows to use the trained network as a core of an intelligent system for the integrated assessment of strength properties of functional and structural materials. The lack of large data arrays of marked-up digital photographs of the microstructures of various materials necessary for the training and testing of convolutional neural networks is one of the obstacles to the widespread use of neural network modeling in the field of material science. Therefore, the results obtained can be considered as an important step in the direction of the development of digital technologies when creating functional metallic materials with a given set of mechanical characteristics.
This work was carried out as a part of the implementation of the federal target program “Research and Development in Priority Directions for the Development of the Russian Science and Technology Complex for 2014–2020”. Unique identifier of the project is RFMEFI58617X0055.

1. Hortigon Fuentes B., Gallardo Fuentes J. М., Muñoz Moreno S., Gumpel P., Strittmatter J. Influence of bar thermal hardening via Tempcore method on their properties after stretching. Chernye Metally. 2019. No. 6. pp. 69–77.
2. Gitman M. B., Klyuev A. V., Stolbov V. Yu., Gitman I. M. Complex Estimation of Strength Properties of Functional Materials on the basis of the Analysis of Grain-Phase Structure Parameters. Strength of Materials. 2017. Vol. 49, Iss. 5. pp. 710–717.
3. Baryshnikov М. P., Chukin М. V., Boyko А. B., Dya H., Nazaybekov А. B. Methods of studying the mechanical characteristics of metals and alloys in metal forming processes taking into account the structure heterogeneity. Vestnik Magnitogorskogo gosudarstvennogo tekhnicheskogo universiteta imeni G. I. Nosova. 2014. Vol. 48. No. 4. pp. 26–31.
4. Chukin M. V., Koptseva N. V., Efimova Y. Y., Chukin D. M., Nikitenko O. A. Influence of the speed of high-carbon steel billet in the patenting unit on its fi nal structure and mechanical properties. Steel in Translation. 2018. Vol. 48, Iss. 4. pp. 224–228.
5. Chukin M. V., Poletskov P. P., Nikitenko O. A., Nabatchikov D. G. Study of microstructure of rolled heavy plates made of low-alloyed pipe steel with increased strength and cold resistance. CIS Iron and Steel Review. 2017. Vol. 13. pp. 28–33.
6. Le Cun Y., Boser B., Denker J. S., Henderson D., Howard R. E. et al. Back Propagation Applied to Handwritten Zip Code Recognition. Neural Computation. 1989. Vol. 1, Iss. 4. pp. 541–551.
7. Szegedy Ch., Liu W., Jia Y., Sermanet P., Reed S. et al. Going deeper with convolutions. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2015. pp. 1–9.
8. He K., Zhang X., Ren Sh., Sun J. Deep Residual Learning for Image Recognition. Proceedings of the IEEE Conference on Computer Vision and Pattern Recognition. 2016. pp. 770–778.
9. Hu Jie, Shen Li, Sun G. Squeeze-and-Excitation Networks. IEEE Transactions on Pattern Analysis and Machine Intelligence. 2017 [electronic resource] Available at: https://arxiv.org/pdf/1709.01507.pdf (accessed: 13.05.2020).
10. De Cost B., Toby F., Holm Elizabeth A. Exploring the microstructure manifold: image texture representations applied to ultrahigh carbon steel microstructures. Acta Materialia. 2017. Vol. 133. pp. 30–40.
11. Azimi S. M., Britz D., Engstler M., Fritz M., Mücklich F. Advanced Steel Microstructural Classification by Deep Learning Methods. Scientiu Reports. 2018. Vol. 8, Iss. 1 [electronic resource] Available at: https://www.ncbi.nlm.nih.gov/pmc/articles/PMC5794926/ (accessed: 13.05.2020).
12. Masci J., Meier U., Ciresan D., Schmidhuber J., Fricout G. Steel defect classification with Max-Pooling Convolutional Neural Networks. Proc. of International Joint Conference on Neural Networks (IJCNN). 2012. pp. 1–6 DOI:10.1109/IJCNN.2012.6252468.
13. Shiyang Zhou, Youping Chen, Dailin Zhang, Jingming Xie, Yunfei Zhou. Classifi cation of surface defects on steel sheet using convolutional neural networks. Materiali in tehnologije. 2017. Vol. 51, Iss. 1. pp. 123–131.
14. Soukup D., Huber-Mörk R. Convolutional Neural Networks for Steel Surface Defect Detection from Photometric Stereo Images. Advances in Visual Computing: conference proceedings ISVC. 2014. Springer. 2014. pp. 668–677.
15. Faghih-Roohi S., Hajizadeh S., Nunez A., Babuska R., De Schutter B. Deep convolutional neural networks for detection of rail surface defects. Proceedings of the 2016 International Joint Conference on Neural Networks (IJCNN 2016). 2016. pp. 2584–2589.
16. Ioffe S., Szegedy C. Batch normalization: Accelerating deep network training by reducing internal covariate shift. 2015 [electronic resource] Available at: https://arxiv.org/pdf/1502.03167.pdf (accessed: 13.05.2020).
17. Klyuev А. V., Stolbov V. Yu., Sharybin S. I. Visualization of complex grain structures of metals and alloys with identification of their parameters. Nauchnaya vizualizatsiya. 2016. Vol. 8. No. 3. pp. 95–101.
18. Klestov R., Klyuev A., Stolbov V. About some approaches to problem of metals and alloys microstructures classification based on neural network technologies. Advances in Engineering Research. 2018. Vol. 157. pp. 292–296.