Fractured inclusion localization and characterization based on deep convolutional neural networks
В. И. Голубев, И. С. Никитин, A. V. Vasyukov, A.D. Nikitin
Abstract
Seismic exploration is a standard method of prospecting for hydrocarbon deposits. The major goal is to correctly localize the spatial position of the fractured inclusion and estimate its parameters. In this work we investigate the applicability of machine-learning algorithms for this problem. To generate synthetic seismograms the model of a deformable solid medium containing slip planes with nonlinear slip conditions on them is used. The explicit-implicit scheme is applied for obtaining the numerical solution of a constitutive system of equations. The problem of the seismic wave propagation in an inhomogeneous fractured geological model based on the well-known Marmousi2 model in a two-dimensional case is considered. Deep convolutional neural networks are used to provide a fast solution for the inverse problem of restoring the parameters of the fractured inclusion based on the surface measurements. The neural network architecture follows a segmentation approach and targets primarily a spatial localization of the fractured inclusion. However, the mechanical parameters of the inclusion are also estimated using a single run of the same network.