Inversão da forma de onda completa de fontes superficiais

• Autor(a) principal: Silva, Bruno de Souza
• Data de Publicação: 2020
• Tipo de documento: Tese
• Idioma: por
• Título da fonte: Repositório Institucional da UFRJ
• Texto Completo: http://hdl.handle.net/11422/21777
• topic: Modelagem sísmica; Operador de síntese; Família de múltiplas fontes; Inversão sísmica; Engenharia Civil

Full Waveform Inversion is among the most robust technologies for velocity field determination in areas of high geological complexity, such as those which present salt tectonics. The success of this method may be explained by the fact that it is not restricted to inverting primary reflected waves. All events predicted by the differential equation, which governs the wave phenomenon, are employed in its algorithm. This increases the seismic illumination tremendously in relation to the methods which linearize the forward problem. In practice, however, due to its high computational cost, which is mainly associated with the large number of forward and reverse extrapolations, this method is typically restricted to frequencies of a few Hertz (around peak frequency of 10 Hertz), which may be insufficient to guarantee the velocity field resolution required for effective depth migrations. In this work, we propose an algorithm that aims to reduce the computational cost with the goal of improving the feasibility of this method for frequencies up to 30 Hertz so as to consider the final velocity field as an interpretation tool itself. The strategy employed is decreasing the number of forward and reverse extrapolations using the concept of areal shot record technology. Rather than solving the wave equation for all common shot gathers, we solved the wave equation for a small number of areal shot records to increase the seismic illumination on the target zones. Ultimately, we activated passive sources (diffractions or reflections) when the controlled wavefield propagates (or reverse propagates) along the background velocity models. In our numerical experiments, performed on Marmousi-2 velocity model, our method successfully converged to the target model. Further, a substantial reduction was observed in the computational cost compared to the traditional method.