Acceleration strategies for elastic full waveform inversion workflows in 2D and 3D

Full waveform inversion (FWI) is one of the most challenging procedures to obtain quantitative information of the subsurface. For elastic inversions, when both compressional and shear velocities have to be inverted, the algorithmic issue becomes also a computational challenge due to the high cost re...

Descripción completa

Detalles Bibliográficos
Autores: Kormann, Jean, Rodríguez, Juan E., Ferrer, Miguel, Farrés, Albert, Gutiérrez, Natalia|||0000-0001-7921-7322, de la Puente, Josep, Hanzich, Mauricio, Cela, José M.
Tipo de recurso: artículo
Fecha de publicación:2016
País:España
Institución:Universitat Politècnica de Catalunya (UPC)
Repositorio:UPCommons. Portal del coneixement obert de la UPC
Idioma:inglés
OAI Identifier:oai:upcommons.upc.edu:2117/90998
Acceso en línea:https://hdl.handle.net/2117/90998
https://dx.doi.org/10.1007/s10596-016-9593-0
Access Level:acceso abierto
Palabra clave:Physical activity
Mathematical modeling and computation
Seismic data
Elastic waves
Inversion
Elastic
Near offset
Ones elàstiques
Sismologia
Àrees temàtiques de la UPC::Enginyeria electrònica
Descripción
Sumario:Full waveform inversion (FWI) is one of the most challenging procedures to obtain quantitative information of the subsurface. For elastic inversions, when both compressional and shear velocities have to be inverted, the algorithmic issue becomes also a computational challenge due to the high cost related to modelling elastic rather than acoustic waves. This shortcoming has been moderately mitigated by using high-performance computing to accelerate 3D elastic FWI kernels. Nevertheless, there is room in the FWI workflows for obtaining large speedups at the cost of proper grid pre-processing and data decimation techniques. In the present work, we show how by making full use of frequency-adapted grids, composite shot lists and a novel dynamic offset control strategy, we can reduce by several orders of magnitude the compute time while improving the convergence of the method in the studied cases, regardless of the forward and adjoint compute kernels used.