Reconstruction of 3D porous media using multiple-point statistics based on a 3D training image

Abstract

To date many methods of constructing porous media have been proposed. Among them, the multiple-point statistics (MPS) method has a unique advantage in reconstructing 3D pore space because it can reproduce pore space of long-range connectivity. The Single Normal Equation Simulation (SNESIM) is one of most commonly used algorithms of MPS. In the SNESIM algorithm, the selection of training image is vital because it contains the basic pore structure patterns. In the previous reconstructions of 3D porous media using SNESIM, a 2D slice was usually employed as the training image. However, it is difficult for a 2D slice to contain complex 3D pore space geometry and topology patterns. In this paper, a 3D training image is used in order to provide more realistic 3D pore structure features. Besides, a multi-grid search template is applied for the purpose of capturing the pore structures of different scales and speeding up the reconstruction process. Two sandstone cores are taken as test examples and the 3D porous media are reconstructed. The two-point correlation function, pore network structure parameters and absolute permeability are applied as the evaluation indexes to validate the accuracy of the reconstructed models. The comparison result shows that the reconstructed models are in good agreement with the real model obtained by X-ray computed tomography scanning in the pore throat geometry and topology and transport property, which justifies the reliability of the proposed method.