PSI - Issue 43

V.I. Golubev et al. / Procedia Structural Integrity 43 (2023) 29–34 V.I. Golubev et al. / Structural Integrity Procedia 00 (2022) 000 – 000

34 6

fluid filled fracture with the free slip mode. Fig. 3c presents the results of the predictions. One can see that the absolute errors are significant. Improving these estimations requires further experiments with the network architecture and the encoding of the features to be predicted.

a) c) Fig. 3. The horizontal (a) and vertical (b) predicted position of the inclusion center. The real values are depicted with dashed lines. The predicted value of the Coulomb friction coefficient q over the used one is presented at (c). 4. Conclusion Dynamic processes in the heterogeneous fractured elastic medium were considered. Based on the continual model of solid media with a discrete set of slip planes the synthetic seismograms were calculated. For a stable numerical solution an explicit-implicit grid-characteristic method of the second order of approximation was applied. The inverse problem of spatial localization and estimation of fractured region properties was successfully solved. Deep convolutional neural network with UNet architecture was used. This algorithm predicted the location of the inclusion well, even though the signal to noise ratio was relatively small. On the other hand, the prediction of the quantitative values of the mechanical parameters of the inclusion was significantly worse. Improving these estimations will be the question of the further research. Acknowledgements This work was carried out with the financial support of the Russian Science Foundation, project no. 19-71-10060. References Araya-Polo, M., Farris, S., Florez, M., 2019. Deep learning-driven velocity model building workflow. Geophysics 38 (11). Das, V., Pollack, A., Wollner, U., 2019. Convolutional neural network for seismic impedance inversion. Geophysics 84 (6), R869 – R880. Diederik, P.K., Jimmy, B., 2014. Adam: A method for stochastic optimization. arXiv, preprint at https://arxiv.org/abs/1412.6980. Dujardin, J.R. and Sauvin, G., Vanneste, M., 2020. Acoustic Impedance Inversion of High Resolution Marine Seismic Data with Deep Neural Network, NSG2020 4th Applied Shallow Marine Geophysics Conference Proceedings, 1-5. Golubev, V.I., Shevchenko, A.V., Khokhlov, N.I., Nikitin, I.S., 2021a. Numerical investigation of compact grid-characteristic schemes for acoustic problems. Journal of Physics: Conference Series 1902 (1), art. no. 012110. Golubev, V., Shevchenko, A., 2021b. EXPLICIT SIMULATION OF SEISMIC WAVES IN FRACTURED VTI MEDIA. 82nd EAGE Conference and Exhibition 6, 4778 – 4782. Golubev, V.I., Guseva, E.K., Petrov, I.B., 2022a. Application of Quasi-monotonic Schemes in Seismic Arctic Problems. Smart Innovation, Systems and Technologies 274, 289 – 307. Golubev, V., Shevchenko, A., Khokhlov, N., Petrov, I., Malovichko, M., 2022b. Compact Grid-Characteristic Scheme for the Acoustic System with the Piece-Wise Constant Coefficients. International Journal of Applied Mechanics 14 (2), art. no. 2250002. Martin, G.S., Wiley, R., Marfurt, K.J., 2006. Marmousi2: An elastic upgrade for Marmousi. The Leading Edge 25 (2), 156 – 166. Nikitin, I.S., Golubev, V.I., 2022. Higher Order Schemes for Problems of Dynamics of Layered Media with Nonlinear Contact Conditions. Smart Innovation, Systems and Technologies 274 (1), 273 – 287. Park, M.J., Sacchi, M.D., 2020. Automatic velocity analysis using convolutional neural network and transfer learning. Geophysics 85 (1). Ronneberger, O., Fischer, P., Brox, T., 2015. U-Net: Convolutional Networks for Biomedical Image Segmentation. arXiv, preprint at https://arxiv.org/abs/1505.04597. Stankevich, A., Nechepurenko, I., Shevchenko, A., Gremyachikh, L., Ustyuzhanin, A., Vasyukov, A., 2021. Learning velocity model for complex media with deep convolutional neural networks. arXiv, preprint at https://arxiv.org/abs/2110.08626. Yang, F., Ma, J., 2019. Deep-learning inversion: A next-generation seismic velocity model building method. Geophysics 84 (4), R583 – R599. b)