A Simulated Annealing Approach to Electrical Resistivity Tomography

In this study, a new approach to Electrical Resistivity Tomography (ERT) data inversion was investigated. ERT is a geophysical method for imaging the subsurface. An electrical current is injected into the surface, and the electrical properties of the subsurface is determined from measurements of the voltage difference. The objective of these measurements are to estimate the unknown resistivity distribution of the subsurface from information gathered on the surface. The most common way to approach this problem is by the local optimization method, least squares. Due to the ill-posedness of inverse problems, least squares method require a prior information of the subsurface. This may not be readily available. A prior information is needed to form an approximated initial model, and employ appropriate regularizations. In this study, we use Simulated Annealing, to estimate the resistivity distribution of a resistor network. Simulated Annealing is a stochastic optimization method for approximating global optimum. This method does not require a prior information or regularizations, and can approximate solutions in a large search spaces. An algorithm was developed and implemented from scratch for this study. We showed that the resistivity distribution of a network, can be approximated using Simulated Annealing. However, problem of local indeterminable resistivity is observed for challenging resistor distributions. Time consumption is one of the main obstacle for Simulated Annealing to be a competitive optimization method. Practice, limitations and improvements of the method is also discussed in this thesis