Inversăo de Sondagem Elétrica Vertical Utilizando Algoritmos Genéticos Combinado com Inversăo Linearizada

Inversion of DC resistivity sounding is a nonlinear problem. Local or global optimization methods are commonly used to solve it. Local methods are fast but, require that the start model be close to the true solution and may be trapped in local minimum. Global methods are robust, but computationally expensive since the space is usually very large. Here we combine the genetic algorithm (AG) with the linearized inversion method, Gauss-Newton (GN), to overcome their limitations and explore the advantages of the two methods. The algorithm was tested with a I-D Schlumberger resistivity sounding data and its performance was compared with pure AG. The joint operation improves the convergence even when using a reduced population of models.