Abstract:In controlled-source electromagnetic (CSEM) surveys that utilize electric dipole sources, observations can be greatly influenced by the orientation, length and the shape of the source. However, most of the current 3D CSEM modeling algorithms do not take all these factors into account, and many are only able to deal with very simple source geometries such as a point dipole with a fixed orientation, which may cause significant inaccuracies in the modeling result. A 3D forward modeling code for marine controlled-source electromagnetic (CSEM) surveys has been developed, which permits simulations of complex source geometries.The solution is based on a staggered-grid finite volume (FV) discretization of the second-order Maxwell's equations. The primary/secondary field approach, in which only the secondary fields need to be solved using the FV method, is employed to avoid the source singularity problem. To accurately compute the primary fields over an ID layered background model, the open source software Dipole1D is integrated into the 3D code. Due to Dipole1D's versatility and the fact that the primary field calculation is the only part of the 3D modeling that directly handles the source geometry, the 3D code is capable of simulating arbitrarily orientated finite-length wire sources. The resulting linear system of FV equations is solved by the quasi-minimum residual (QMR) method.To validate the 3D code, we first benchmarked FV solutions against quasi-analytic solutions for a canonical 1D reservoir model. The numerical solutions show excellent agreement with analytic solutions for various transmitter orientations for both a point dipole source and a 300 m wire source. Then we calculated responses of a 3D reservoir model similar to a well-studied one, varying horizontal sizes of the resistor. Reasonable results for the 3D model, which are consistent with published work, can be observed.Further, a suite of numerical experiments on the 3D model were conducted to investigate the influences of source geometries, and the observations include: (1) The full lateral extent of the 3D resistor can be well estimated from the responses normalized by responses of the background model without the resistor, by moving and rotating the transmitter. (2) Electromagnetic fields excited by wire sources with wire lengths ranging from 100 m to 500 m are compared with those by a point dipole, showing that the longer the source, the greater the differences, and the EM field amplitudes of 500 m wire source can be up to more than 4.5 times larger than that of a point dipole. Prominent differences occur not only in the region immediately around the source, but also in four narrow regions at oblique angles to the source. (3) While the 300 m straight wire source is slightly bended in four different ways, either in the horizontal plane or in the vertical plane, one or several components of the EM field change significantly.From results of all the numerical experiments, we can conclude that the presented 3D FV CSEM modeling code is accurate and versatile, capable of modeling relatively complicated source geometries, such as finite-length straight/bent wire sources with arbitrary orientations. In CSEM surveys, source geometries, i.e. the orientation, the length and the shape of the source can have remarkable influence on the observed data, thus accounting for source geometry effects is necessary.

An Electromagnetic Neural Network For Inverse Source Modeling Of Wire-Like Objects

A Spatial SEM-Based Shallow Neural Network for Electromagnetic Inverse Source Modeling

Electromagnetic Machine Learning for Inverse Modeling Using the Spatial Singularity Expansion Method

An End-to-End Neural Network for Complex Electromagnetic Simulations

Application of Artificial Neural Network for Electromagnetic Source Reconstruction

Enhanced Deep Learning Approach Based on the Deep Convolutional Encoder–Decoder Architecture for Electromagnetic Inverse Scattering Problems

A Novel Electromagnetic Interference Source Reconstruction Method Based On Artificial Neural Network

The Spatial Singularity Expansion Method for Electromagnetics.

Machine Learning Target Count Prediction in Electromagnetics Using Neural Networks

A Summary of Artificial Neural Networks on Electromagnetic Interference Diagnosis

Development of Artificial Neural Network for Field Prediction of Unknown EM Source

Neural-network modeling for 3-D substructures based on spatial EM-field coupling in finite-element method

A general forward solver for 3D CSEMs with multitype sources and operating environments

Equivalent Electromagnetic Hybrid Dipole Based on Cascade-Forward Neural Network to Predict Near-Field Magnitude of Complex Environmental Radiation

Neural network EM-Field based modeling for 3d substructure in finite element method

Inverse Design of Electromagnetic Metasurfaces Utilizing Infinite and Separate Latent Space Yielded by a Machine Learning-Based Generative Model

Wide-Band Near-Field Prediction of Unknown EM Source Based on Artificial Neural Network

An Equivalent Dipole Model Hybrid with Artificial Neural Network for Electromagnetic Interference Prediction

Modeling EM Problem with Deep Neural Networks

Three-Dimensional Forward Modeling of the Marine Controlled-Source Electromagnetic Field with Complex Source Geometries

GAN-driven Electromagnetic Imaging of 2-D Dielectric Scatterers