Title: 3D OcTree Finite-volume method for acoustic wave simulation
André Valente1; Leandro Di Bartolo2; Webe Mansur3
1 IFRJ -- Instituto Federal de Educação, Ciência e Tecnologia do Rio de Janeiro; 2 Observatório Nacional / MCTI, Departamento de Geofísica; 3 COPPE, Universidade Federal do Rio de Janeiro
doi:10.20906/CPS/CILAMCE2015-0203
Resumo
The seismic method is the main tool used for petroleum exploration. It is used to map the structure of subsurface geologic formations in order to infer the existence of possible petroleum reservoirs. It is also used to characterize the lithology, fluid content, fractures, etc. Currently, the huge challenges in exploration and reservoir characterization put the wave equation based algorithms in highlight for processing seismic data. Full waveform inversion (FWI) and reverse time migration (RTM) became important tools. In this context, developing powerful and efficient 3D numerical schemes for seismic modeling is of fundamental importance. The finite-difference method (FDM) is the most popular method used to model seismic waves, mainly because of its robustness: it is applicable in a simple way to complex regions and, at the same time, it is relatively accurate and computationally efficient. Because the grid spacing is fixed in the finite-difference method (FDM), the numerical dispersion usually imposes the use of restrictive small grid spacing (given by the lower velocity) in the whole model. Then, the number of grid points per wave-length is too large in deep layers, resulting in loss of efficiency of the FDM. Although the use of high-order operators allows bigger grid spacing and could make FDM more efficient, this does not solve the problem. In this work, we develop a new 3D explicit scheme based on finite-volume method (FVM) that addresses the aforementioned problem in an efficient way. The algorithm is constructed using a general formalism with sparse matrices and allows both the use of different grid spacing in layers with different velocities and the refinement in regions of interest using OcTree meshes, e.g., in target regions with thin layers containing oil and gas, reflectors and salt flanks. OcTree is easy to be generated using regular grids, avoiding additional complications in meshing. Computational tools for parallel processing on the GPU have been used for the algebraic manipulation of the sparse ma
Palavras-chave: Seismic Modeling; Finite-Volume Method; OcTree Meshes