Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Research › peer-review
Simulated Annealing Algorithm for Model Reconstruction of the Four-Layer Medium with Elliptical Inclusion in the Third Layer. / Prokhorov, Dmitry; Reshetova, Galina; Bratchikov, Denis.
Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics). Springer Science and Business Media Deutschland GmbH, 2024. p. 334-351 23 (Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics); Vol. 14817 LNCS).Research output: Chapter in Book/Report/Conference proceeding › Conference contribution › Research › peer-review
}
TY - GEN
T1 - Simulated Annealing Algorithm for Model Reconstruction of the Four-Layer Medium with Elliptical Inclusion in the Third Layer
AU - Prokhorov, Dmitry
AU - Reshetova, Galina
AU - Bratchikov, Denis
N1 - Conference code: 24
PY - 2024
Y1 - 2024
N2 - The paper presents an approach for reconstructing the properties of two-dimensional viscoelastic medium with defined geometry using the simulated annealing algorithm. The inverse problem solution requires a lot of computational resources because the direct seismic modeling is performed at each iteration. The staggered grid finite-difference scheme is implemented using CUDA technology to speed up the solution of the direct problem by parallelization. The choice of the simulated annealing method for solving inverse problem is due to the method’s ability to avoid local minima of the target functional. However, the simulated annealing method needs a good coverage of the model space by realizations of random probing vectors. It leads to enormous computation time in the case of a four-layer medium with elliptical inclusion in the third layer, which has 37 parameters. Therefore, the sequential reconstruction of model parameters, where the simulated annealing algorithm searches for the parameters in 1 or 2-D subspace, is introduced. Nevertheless, the attenuation properties of medium were not reconstructed by simulated annealing. For their study, the deep convolutional neural network is used.
AB - The paper presents an approach for reconstructing the properties of two-dimensional viscoelastic medium with defined geometry using the simulated annealing algorithm. The inverse problem solution requires a lot of computational resources because the direct seismic modeling is performed at each iteration. The staggered grid finite-difference scheme is implemented using CUDA technology to speed up the solution of the direct problem by parallelization. The choice of the simulated annealing method for solving inverse problem is due to the method’s ability to avoid local minima of the target functional. However, the simulated annealing method needs a good coverage of the model space by realizations of random probing vectors. It leads to enormous computation time in the case of a four-layer medium with elliptical inclusion in the third layer, which has 37 parameters. Therefore, the sequential reconstruction of model parameters, where the simulated annealing algorithm searches for the parameters in 1 or 2-D subspace, is introduced. Nevertheless, the attenuation properties of medium were not reconstructed by simulated annealing. For their study, the deep convolutional neural network is used.
KW - seismic modeling
KW - simulated annealing
KW - viscoelastic medium
UR - https://www.scopus.com/record/display.uri?eid=2-s2.0-85200722318&origin=inward&txGid=9891399a30457b117d69f958715630d6
UR - https://www.mendeley.com/catalogue/bcf9a249-3df8-33f4-88ca-b29453b777b9/
U2 - 10.1007/978-3-031-65238-7_23
DO - 10.1007/978-3-031-65238-7_23
M3 - Conference contribution
SN - 9783031652370
T3 - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
SP - 334
EP - 351
BT - Lecture Notes in Computer Science (including subseries Lecture Notes in Artificial Intelligence and Lecture Notes in Bioinformatics)
PB - Springer Science and Business Media Deutschland GmbH
T2 - 24th International Conference on Computational Science and Its Applications
Y2 - 1 July 2024 through 4 July 2024
ER -
ID: 60494695