FRINK Linked Data Fragments server

Matches in SemOpenAlex for { <https://semopenalex.org/work/W52906850> ?p ?o ?g. }

• W52906850 abstract "In seismology, numerical wave propagation modelling is an essential tool to e.g quantify local seismic risk or to invert multi-source seismic data. Realistic phenomcna are accounted for when three-dimensional (3D) models with complex heterogeneous structures are similated using sophisticated techniques on irregular grids. ln this thesis, I investigate frequency domain finite difference (FDFD) modelling for efficient space discretization optimization algorithms in order to render 3D FDFD modelling feasible. I compare ,FDFD acoustic wave simulations in heterogeneous media by a staggered cross stencil analysis with a grid rotation approach. For two-dimensional (2D) modelling, grid rotation is superior in terms of CPU-time and memory requirements. Therefore, I propose that space discretization by grid rotation shall be investigated for 2D time domain FD modelling. Moreover, 3D discretizations lead to 27-points rotation star, and 19-points staggered cross stencils. I suggest that both shall be examined with respect to numerical accuracy and computer ressources requirements. I propose a multi-level direct and iterative solver combination for realistic 3D FDFD wave simulations. Exact coarse-grid wave solutions are computed by a direct LU matrix factorization. The coarse-grid solution is prolongated on a fine-resolution grid and subsequently used as preconditioner for an iterative solver scheme. I investigate two multi-level approaches based on nested iteration, denoted by Direct-Iterative-Space Solver(DlSS), and a multi-scale wavelet expansion, called Direct-Iterative- Wavelet- Solver (DIWS). DISS and DIWS wave simulations in 2D and 3D heterogeneous media are compared with respect to their CPU-time and memory performance. The DISS approach provides fast and efficient matrix construction algorithms. The convergence of the iterative scheme is strongly dependent on efficient suppression of phaseshift artefacts caused by bilinear interpolation. Either large numbers of iteration steps or V- and W-cycles need to be performed to provide dispersion-free wave simulations. Wavelet-based DIWS preconditioning leads to optimal iterative behaviour at the cost of wavelet projections. Matrix constructions become expensive due to convolution-type computations. Furthermore, non-zero DIWS matrix entries increase and therefore slow down matrix-vector product performances. Nevertheless, the multi-scale wavelet formulation gives fast and stable iterative constructions simultaneously on a series of approximation grids of decreasing resolution, where difficult grid interactions are accounted for naturally. Therefore, DIWS-preconditioning is superior. Moreover, wavelet formulations provide adaptive optimization strategies for efficient implementation schemes. Wave simulation examples are illustrated for several homogeneous and heterogeneous 2D and 3D rnodels. The 3D applications are limited in size by restrictions to sequential computations. Extensions to parallel algorithms on distributed memory computer structures are discussed." @default.
• W52906850 created "2016-06-24" @default.
• W52906850 creator A5027693536 @default.
• W52906850 date "2002-12-17" @default.
• W52906850 modified "2023-09-28" @default.
• W52906850 title "Propagation des ondes : approches espace et ondelette" @default.
• W52906850 hasPublicationYear "2002" @default.
• W52906850 type Work @default.
• W52906850 sameAs 52906850 @default.
• W52906850 citedByCount "0" @default.
• W52906850 crossrefType "dissertation" @default.
• W52906850 hasAuthorship W52906850A5027693536 @default.
• W52906850 hasConcept C11413529 @default.
• W52906850 hasConcept C126255220 @default.
• W52906850 hasConcept C134306372 @default.
• W52906850 hasConcept C154945302 @default.
• W52906850 hasConcept C159694833 @default.
• W52906850 hasConcept C167431342 @default.
• W52906850 hasConcept C181330731 @default.
• W52906850 hasConcept C187691185 @default.
• W52906850 hasConcept C2524010 @default.
• W52906850 hasConcept C2778770139 @default.
• W52906850 hasConcept C28826006 @default.
• W52906850 hasConcept C33923547 @default.
• W52906850 hasConcept C41008148 @default.
• W52906850 hasConcept C459310 @default.
• W52906850 hasConcept C47432892 @default.
• W52906850 hasConcept C73000952 @default.
• W52906850 hasConcept C74050887 @default.
• W52906850 hasConcept C76752949 @default.
• W52906850 hasConceptScore W52906850C11413529 @default.
• W52906850 hasConceptScore W52906850C126255220 @default.
• W52906850 hasConceptScore W52906850C134306372 @default.
• W52906850 hasConceptScore W52906850C154945302 @default.
• W52906850 hasConceptScore W52906850C159694833 @default.
• W52906850 hasConceptScore W52906850C167431342 @default.
• W52906850 hasConceptScore W52906850C181330731 @default.
• W52906850 hasConceptScore W52906850C187691185 @default.
• W52906850 hasConceptScore W52906850C2524010 @default.
• W52906850 hasConceptScore W52906850C2778770139 @default.
• W52906850 hasConceptScore W52906850C28826006 @default.
• W52906850 hasConceptScore W52906850C33923547 @default.
• W52906850 hasConceptScore W52906850C41008148 @default.
• W52906850 hasConceptScore W52906850C459310 @default.
• W52906850 hasConceptScore W52906850C47432892 @default.
• W52906850 hasConceptScore W52906850C73000952 @default.
• W52906850 hasConceptScore W52906850C74050887 @default.
• W52906850 hasConceptScore W52906850C76752949 @default.
• W52906850 hasLocation W529068501 @default.
• W52906850 hasOpenAccess W52906850 @default.
• W52906850 hasPrimaryLocation W529068501 @default.
• W52906850 isParatext "false" @default.
• W52906850 isRetracted "false" @default.
• W52906850 magId "52906850" @default.
• W52906850 workType "dissertation" @default.