Redução do custo computacional da modelagem sísmica

Resumo

Seismic modeling is a technique widely applied in geophysical methods, used to simulate the propagation of seismic waves, such as those described by the acoustic or the elastic wave equation. These equations are precise mathematical descriptions of the behavior of waves when passing through different types of geological materials, such as rock layers and underground reservoirs. The simulation is based on geophysical models of the subsurface, developed based on observed data and other geological information. Several aspects, such as density, acoustic pressure, and particle velocity, can be modeled, and the more of these parameters that are modeled, the more accurate the seismic modeling will be. However, increasing the number of modeled parameters can make the computation more complex and time-consuming, which is why seismic modeling has a high computational cost. Therefore, it is necessary to develop approaches that reduce the number of operations performed in seismic modeling. Since, for a given instant of time, the wave propagation does not occur throughout the domain, we can calculate it only in a portion of it. From this, it is proposed to simulate spherical modeling to determine the regions of the grid, where the propagation of the acoustic wave field occurs. In this way, the calculation of the Laplacian determined to solve the acoustic wave equation is restricted to the area corresponding to the sphere, in a specific time interval. To implement the spherical modeling, calculations are performed for the limits of the three-dimensional coordinates, where the limits of the sphere will propagate in the middle of the simulation performed over a predefined period, from the triggering of the source until the reception by the receiver. In the performance analyses comparing the approach that uses spherical modeling with the traditional modeling approach, it is possible to observe a performance gain of 22.87% in reduction of the execution time in the average of the tests that vary the size of time steps. For applications that vary the speed at which the wave propagates, there was a gain in time of 32.28% on average for the tests performed. In applications that vary the position of the seismic source, a gain in time performance of 35.13% was obtained for the average of the tests. It is therefore notable that there was a substantial gain in performance using the applied method. For future work, it is desirable to apply the sphere method to other geophysical methods, such as RTM (Reverse Time Migration) and FWI (Full-waveform Inversion), and consequently perform the analysis of these applications.