Details of the abstract

Title of paper A parallel adaptive finite-element method for 3-D large-scale controlled-source
electromagnetic forward modelling with hierarchical tetrahedral grids
List of authors "Author, Zhengguang Liu., Co-anthor, Zhengyong Ren., Co-anthor, Hongbo Yao., Co-anthor, Jingtian Tang., Co-anthor, Xushan Lu., Co-anthor, Colin Farquharson."
Affiliation(s) "School of Geosciences and Info-Physics, Central South University., Department of Earth Sciences, Memorial University of Newfoundland."
Summary We have developed a parallel, adaptive, finite-element approach for frequency-domain 3-D CSEM forward modeling. Our algorithm is capable of using
hierarchical tetrahedral grids with an adaptive mesh refinement (AMR) technique, which yields more accurate electromagnetic responses for large-scale complex models. Our algorithm solves the total electric field vector
equation. The geo-electrical model is discretized by unstructured tetrahedral grids which can deal with complex underground geological models with arbitrary surface topography. Unlike previous adaptive finite-element software working on unstructured tetrahedral grids, we implemented a novel mesh refinement technique named the longest edge bisection method to generate hierarchically refined grids. New cells obtained from the refinement method are all part of the original cells in the coarser grid. Therefore, a one-to-one mapping relationship between the newly refined cells and the original cells could be efficiently and precisely obtained such that the conductivity model represented by the refined mesh stays the same as that represented by the original mesh. This means there will be no inconsistency in the conductivity model in the inversion while
transitioning from early-stage coarser meshes to late-stage refined meshes to obtain better recovered models. In addition, we employ a parallel domain-decomposition technique to accelerate the computational speed of our forward modeling algorithm. The flexible generalized minimum residual (FGMRES) iterative solver with an auxiliary-space Maxwell preconditioner is used to solve the final large-scale linear system of equations.
Session Keyword 2.0 Theory, Modelling and Inversion
File upload 2.0_a_parallel_adaptive_finit_liu.pdf
Back