Numerical solution of the elastic wave equation has proven to be a very valuable tool For genphyscists in forward modeling and migration. Among the techniques generally used in modeling wave propagation we conaider in this paper the finite element method, focusing our attention on accuracy and computational requirements. The study has been carried out using several algorithms for time integration; Central Differences (explicit and implicit), Newmark, Wilson and Houholt. The lesulting large sparse system of linear equations is solid using both direct and iterative methods. For the direc. method the matrix is represented using the skyline storage scheme, while for iterative methods the row-wise and the compressed diagonal schemes have been used. The numerical methods have been implemented on an IBM 3090 vector multiprocessor. Performance data for the different implementations are presented and compared with data previously obtained using pseudo-spectral and finite differences methods on the same architecture. From our results it is c ncludcd that Newmark and Explicit Central Differences arc the mast adequate integration methods and Conjugate Gradient with compressed diagonal storage scherc is the most efficient solver for vector computers. © 2018 IOP Publishing Ltd.
Add the full text or supplementary notes for the publication here using Markdown formatting.