No abstract

The FWI is formulated as a nonlinear optimization problem that traditionally uses local (derivative-based) minimization to find the scalar field of properties that best represents the field seismic data. This problem has a high computational cost and accuracy limited to local minima, in addition to suffering from a slow convergence rate (Cycle Skipping). Therefore, we developed a two-phase hybrid optimization algorithm based on DFO algorithms. The first use global minimization and clustering technique. The second use local minimization. In phase 1 we adopted the modified PSO and K-means algorithms and in phase 2, we adopted the ANMS. We call the hybrid algorithm of the PSO-Kmeans-ANMS. Where K-means is responsible for dividing swarms of particles into 2 clusters at every instant. This strategy aims to automatically balance the mechanisms of exploration and exploitation of the parameter search space by the hybrid algorithm, allowing one to find more precise solutions and consequently improving its convergence. The PSO-Kmeans-ANMS algorithm was validated on the set of 12 benchmark functions and applied to the FWI 1D problem. We compared PSO-Kmeans-ANMS with classic PSO, modified PSO, and ANMS algorithms. The metrics used were are the average execution time and the success rate (an error of ± 4% of the optimal solution). In all validation experiments and the FWI application, the PSO-Kmeans-ANMS performed well in terms of robustness and computational efficiency. In the case of FWI, there was a significant reduction in computational cost, thus presenting a relevant result.

No abstract

No abstract

RESUMONa exploração sísmica, investiga-se as características de subsuperfície usando técnicas de inversão completa, da forma da onda (Full Waveform Inversion -FWI), a qual foi abordada como um problema de otimização não linear. A técnica FWI, tradicionalmente, usa métodos matemáticos baseados em derivadas e, portanto, falha quando a função objetiva é não diferenciável. Ademais, isso acarreta um alto custo computacional e uma precisão limitada a mínimos locais. Portanto, neste trabalho foi adotada uma metodologia sem derivadas, Derivative-Free Optimization (DFO), para encontrar o mínimo global. Neste tipo de abordagem, utiliza-se as técnicas Salto Aleatório (RJT), Busca Aleatória Controlada (CRS) e Simplex Adaptativo de Nelder-Mead (ANMS). Desenvolveu-se um algoritmo, FWI-DFO, que resolve numericamente a equação da onda acústica 2D pelo Método das Diferenças Finitas (FDM) e se utiliza de um método híbrido RJT-CRS-ANMS como técnica de otimização para a inversão sísmica. A estratégia é balancear automaticamente as buscas globais e locais iterativamente pelo CRS e ANMS, respectivamente. Aplicou-se a metodologia em cinco modelos reais de subsuperfície. Os resultados mostraram uma concordância significativa com os modelos reais. O tempo computacional apresentou valores razoáveis e a função objetivo mostrou ser bastante sensível a pequenas alterações nos parâmetros do modelo para os casos aqui analisados. Em síntese, a metodologia, FWI-DFO, empregada mostrou-se bastante promissora na inversão sísmica. DERIVATIVE-FREE OPTIMIZATION HYBRID STRATEGY FOR FULL WAVEFORM INVERSION ABSTRACTIn the seismic exploration, the subsurface characteristics have been investigated using Full Waveform Inversion (FWI) techniques which was approached as a nonlinear optimization problem. The FWI technique traditionally uses mathematical methods based on derivatives and therefore fails when an objective function is non differentiable. In addition, this entails a high computational cost and a precision limited to local minimum. Therefore, in this work, a Derivative-Free Optimization (DFO) methodology was adopted to find the global minimum. In this type of approach, the Random Jump (RJT), Controlled Random Search (CRS) and Adaptive Nelder-Mead Simplex (ANMS) techniques were used. A FWI-DFO algorithm which numerically solves the 2D acoustic wave equation by the Finite Differences Method (FDM) and uses a hybrid method RJT-CRS-ANMS as the optimization technique for the seismic inversion was developed. The strategy is to automatically balance global and local searches iteratively by CRS and ANMS, respectively. The methodology to five real subsurface models was applied. The results showed a significative agreement with the real models. The computational time presented reasonable values and the objective function showed to be very sensitive to small changes in the model parameters for the cases analyzed here. In summary, the FWI-DFO methodology proved to be very promising in the seismic inversion. KEYWORDS: