Tomografia sísmica poço-a-poço utilizando otimização global serial e paralelizada com uso de vínculos estabilizantes

Abstract

Crosswell seismic transit time tomography involves finding a velocity model that minimizes the misfit between calculated and observed transit times. In this sense, this paper presents two studies of this inverse geophysical problem, one using the Particle Swarm Optimization (PSO) global optimization algorithm, analyzing the effectiveness of applying the constraints: Smoothness, Total Variation and Sparsity; the other performing the comparison of the first algorithm with the Genetic Algorithm (GA), in its serial and parallel forms, and for the two experiments using straight ray modeling. For the first experiment, the results of the inversions with the synthetic models were satisfactory. In the first synthetic model, based on graphs of relative cumulative errors, the result of inversions using the Total Variation constraint obtained the best result, followed by the Smoothness constraint. However, in the second synthetic model, the Sparsity constraint (DCT) had the lowest relative cumulative error. In the second experiment, a review of these two optimization techniques is presented. The solution of the crosswell tomography problem using the Smoothness constraint is presented with the objective of illustrating the efficiency of these methodologies in the solution of inverse geophysical problems. The results show that the parallelized versions are more efficient. For each experiment, a scientific article was prepared, totaling two, which were accepted and published in August 2021 by Revista Jr de Iniciação Científica em Ciências Exatas e Engenharia - ISSN 2236-0093, as part of its volume 1, number 22. In addition, the articles were classified in the area of Engineering I, whose qualis in this category is B5.