| Currently, there are two problems that baffle the 3D direct current (DC) resistivity finite element forward modeling. The first is how to build complex 3D geoelectrical model efficiently and effectively. Most domestic scholars are still build 3D models within a.txt file and by manual input, which is time consuming, error prone and not applicable for complex ones. The second is, though by enforcing mixed boundary conditions we can get relatively high accuracy in an acceptable discretization domain, because the global system matrix is affected by the locations of sources, once the locations have been changed the matrix has to be formed over again, which would make the forward modeling for some survey configurations, such as dipole-dipole array, very time consuming and worthless for inversions. The solutions commonly used for solving the problems mentioned above include taking replacement of the mixed boundary conditions by Dirichlet or Neumann boundary conditions, that is to say to force potentials or there derivatives to be zero on the subsurface boundaries, or assuming the distance between the source and boundary in the mixed boundary conditions to be a constant. All the schemes represented above need the domain of calculation and discretization to be set so large that the effect of the truncated boundaries can be eliminated to some extent, while what is inevitable is that the number of grid nodes must be increasing and so is the computational complexity.The same problems also arise from the 3D controlled source electromagnetic (CSEM) finite element modeling. Due to the large scale of the survey domain, which usually extends from hundreds to thousands meters, the numbers of finite element grids are also huge. Application of Dirichlet boundary conditions has been proved to be an effective and handy method for truncating the infinite domain, but the boundaries must be far away from the source and anomalous bodies, usually tens of thousands meters, thereby bringing so many finite elements in the domain that we are not interested in. Considering in electromagnetic problems, the number of degrees of freedoms for each node is typically more than one. Moreover, the unknowns are all complex numbers. As a result, the global matrices will occupy so many memories and bring difficulties for solving.In this study, fast and visualized pre and post processing for finite element method was achieved by customized development based on the universal 3D modeling software named GiD, while Infinite elements were introduced to form the finite-infinite element coupling method, which can be used to substitute the artificial boundary conditions and reduce the number of finite element nodes. For the customized development on GiD, we only need to program the so-called problem types with simple script language, then models could be build in the graphic interface and output in the format fitting our calculation program. After the calculation process, by using the GiDpost library, the files used for post-processing in GiD could be outputted easily. As to our finite-infinite element coupling method, Astley mapped wave envelope infinite elements were employed to continue the electrical fields to infinity. Meanwhile, a new type of infinite element shape functions was proposed and proved to be the optimal one in both accuracy and time consumption by comparing with several other shape functions during the simulations for 3D DC problems. Finally, the availability and superiority of this coupling algorithm were confirmed by several numerical tests both in 3D DC modeling and 3D CSEM modeling.Overall, by the pre and post processing method proposed in this paper, the 3D models can be built efficiently and applicable to any calculation program, while the finite-infinite element coupling method could derive solutions with high accuracy in a comparatively small descretization domain, which helps to reduce the number of degrees of freedoms and speed up the computation.
|
PDF Full Text Request