A multi-point geostatistical 3D modeling method combined with deep learning

Abstract

The invention discloses a multi-point geological statistics three-dimensional modeling method combined with deep learning. Aiming at the time-consuming problem of the matching process of the traditional multi-point geostatistical algorithm, this method proposes a method combined with deep learning to speed up the reconstruction. Using the method of layer-by-layer reconstruction, the three-dimensional reconstruction is transformed into two-dimensional reconstruction. A deep neural network is designed to learn the mapping relationship from sampling map to pore map, and this relationship is used for the reconstruction of sampling map. The main innovations of the present invention include: the idea of ​​using deep learning to accelerate the matching process of the multi-point geological statistics method and the three-dimensional modeling algorithm of the multi-point geological statistics based on this idea. The datasets used in this method were made, and the performance of the network was measured by quantitative comparison of visual effects and statistical functions. Compared with the time-consuming point-by-point simulation method of the traditional multi-point geological statistical method, the present invention greatly improves the speed under the premise of ensuring accuracy, and has good application value.

Description

technical field [0001] The invention relates to a three-dimensional modeling method of porous media based on two-dimensional images, in particular to a multi-point geological statistics porous media modeling method combined with deep learning, which belongs to the technical field of image processing. Background technique [0002] Porous media such as rock cores, alloys, and ceramics exist widely in nature and man-made environments, and play an important role in practical engineering applications. In practical applications, people need to have an accurate understanding and cognition of their macroscopic properties, and the macroscopic properties are directly determined by their microstructure, so it is very important to study their microstructure. [0003] Taking the core as an example, there are two main ways to obtain the microstructure image of the core. One is to use three-dimensional imaging techniques, such as computed tomography (CT), focused ion beam scanning electro...

AI Technical Summary

Problems solved by technology

However, the matching process of MPS needs to continuously retrieve the patterns of the training images, which is very time-consuming and occupies most of the time of the whole reconstruction process.

Although there is a method based on GPU acceleration at present, the acceleration effect is not very obvious, and it is essentially a point-by-point simulation mechanism.

Images