Um algoritmo colônia de formigas aplicado à inversão de dados geofísicos.

Abstract

Nonlinear inversion of geophysical data uses global and local optimization methods. These methods can be separately used, each with its own advantages and disadvantages. Local scope methods tend to have high convergence speed and usually produce very accurate results, but they are highly sensitive to the initial model, easily getting trapped in local optima closer to the starting point. Global scope methods have the ability to escape from local optima; on the other hand, they usually have high computational cost and low accuracy when compared to local scope methods. In order to overcome these negative attributes and fully exploit the positive ones, we propose a hybrid optimization method developed as a product of the combination of the Ant Colony (global search) and linearized inversion (local search). Then, through (1) the generation of 1-D vertical electrical sounding data from multilayer models; and (2) the generation of 2-D gravimetric data from prism models, a comparative study was carried out between the methods ant colony optimization, linearized inversion and the hybrid method resulting from the combination of these ones in data inversion experiments. Furthermore, among the several variants of the Ant Colony family methods, Ant Colony Optimization for Continuous Domains is the most popular and most robust variant, being the chosen one to be discussed in this study.