Two-dimensional frequency-domain visco-elastic full waveform inversion: Parallel algorithms, optimization and performance

摘要: Full waveform inversion (FWI) is an appealing seismic data-fitting procedure for the derivation of high-resolution quantitative models subsurface at various scales. modelling and visco-elastic waves from multiple sources allow recovering different physical parameters, although they remain computationally challenging tasks. An efficient massively parallel, frequency-domain FWI algorithm implemented here on large-scale distributed-memory platforms imaging two-dimensional media. The resolution elastodynamic equations, as forward problem inversion, performed in frequency domain unstructured triangular meshes, using a low-order finite element discontinuous Galerkin method. linear system resulting discretization solved with parallel direct solver. inverse problem, which presented non-linear local optimization quasi-Newton method, this allows reliable estimation classes parameters. Two levels parallelism are algorithm, based message passing interfaces multi-threading, optimal use computational time core-memory resources available modern multi-core platforms. realistic targets scales, ranging near-surface geotechnic applications to crustal-scale exploration.