Document Type : Research Paper
Authors
1 Petroleum Engineering and Geophysics Laboratory, School of Mining Engineering, College of Engineering, University of Tehran, Iran
2 Petroleum Engineering and Geophysics Laboratory, School of Mining Engineering, College of Engineering University of Tehran, Iran
Abstract
Direct Current (DC) Electrical Resistivity Tomography (ERT) inversion is a non-linear, ill-posed problem requiring robust solvers for accurate subsurface resistivity reconstruction. This study systematically compares four second-order unconstrained iterative solvers which are Simple Newton, Newton-CG, Trust-Exact, and Trust-NCG, using synthetic dipole-dipole data with conductive and resistive anomalies in a homogeneous background. Inversion is performed within a finite-element framework with unstructured triangular meshes. Solver performance is assessed through reconstruction accuracy and computational efficiency using error metrics and data misfit. Results show Simple Newton fails due to instability and fixed step sizes. Newton-CG, Trust-Exact, and Trust-NCG converge reliably, with Trust-Exact and Trust-NCG markedly superior in resolving sharp resistivity contrasts, especially the conductive anomaly, owing to adaptive trust-region globalization. Newton-CG and Trust-NCG exhibit minor smoothing, reflecting a trade-off between speed and precision. These findings highlight the superiority of trust-region methods in weakly regularized non-linear geophysical inversion, emphasizing their ability to preserve sharp contrasts.
Keywords
)