Details of the abstract
Title of paper | Electromagnetic imaging using high-order FE and goal-oriented meshing |
List of authors | Castillo-Reyes, O., Rulff, P., Um E. |
Affiliation(s) | Barcelona Supercomputing Center (BSC), Department of Earth Sciences - Uppsala University, Lawrence Berkeley National Laboratory - Earth and Environmental Sciences |
Summary |
Electromagnetic (EM) modeling routines development has increased in the last decade. Given the recent advances in numerical methods and increased computing power, there are several options to solve the forward EM problem today. However, regardless of the computational meshes used, most of these modeling tools do not support high-order polynomial discretization methods and adaptive meshing. In this work, we present numerical experiments about goal-oriented adaptive mesh refinement and the high-order edge finite element method (HEFEM) for the solution of challenging 3D controlled-source electromagnetic (CSEM) surveys. Each test case presents a particular modeling challenge (e.g., topography, presence of metallic boreholes), suitable for studying the numerical schemes capacities in realistic setups. We use two 3D EM modeling routines to compute synthetic EM responses: elfe3D and PETGEM. We investigate the performance of the goal-oriented meshes and HEFEM discretizations in terms of accuracy and computational cost (e.g., run-time, memory consumption). Our assessment provides relevant information for an in-depth understanding of the pros and cons of employing HEFEM and goal-oriented meshes. The numerical experiments show that high-order polynomial basis functions in conjunction with goal-oriented meshes can obtain synthetic EM responses in agreement with the reference. This conclusion is valid for all the test cases under consideration. However, the code performance depends on the input model: frequency, resistivity, scale variations, mesh quality, source discretization, and computational architecture. Therefore, an aware meshing scheme that considers the polynomial basis order is required to obtain competitive performance ratios. We believe that our numerical experiments will prove useful for the EM community. |
Session Keyword | 2.0 Theory, Modelling and Inversion |
File upload |
2.0_electromagnetic_imaging_u_castillo reyes.pdf
|