Abstract: 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.

Cite: Prokhorov D. , Reshetova G. , Bratchikov D.
Simulated Annealing Algorithm for Model Reconstruction of the Four-Layer Medium with Elliptical Inclusion in the Third Layer
In compilation Computational Science and Its Applications – ICCSA 2024 Workshops Hanoi, Vietnam, July 1–4, 2024, Proceedings. – Springer Nature., 2024. – Т.Part III. – C.334-351. – ISBN 978-3-031-65238-7. DOI: 10.1007/978-3-031-65238-7_23 WOS Scopus OpenAlex

Dates:

Published print:	Jul 26, 2024
Published online:	Jul 26, 2024

Identifiers:

Web of science:	WOS:001294364500023
Scopus:	2-s2.0-85200722318
OpenAlex:	W4401000256

Citing: Пока нет цитирований

Altmetrics: