Research On Three-Dimensional Reconstruction Of Digital Porous Media

Coal is one of the main energy sources and is the world's most wide-ranged energy source for exploitation and consumption.With the exploitation and utilization of large quantities of coal,the research on reservoir rock is of great significance.All the physical properties of reservoir rock are closely related to mineral exploration and exploitation,so the comprehensive research of reservoir rock can guarantee the production safety,stability and efficiency.At the same time,it is of great help for subsequent research.At present,the microstructure of coal is mainly obtained by CT imaging,but this method costs too much.By using computer technology,the data of real coal-rock structure is collected,and the threedimensional(3D)structure can be obtained by using the 3D reconstruction algorithm.This paper mainly focuses on the study of the 3D reconstruction algorithm using one single two-dimensional(2D)coal-rock image as the original information.The genetic algorithm and the widely-used multiple-point geostatistics algorithm for 3D reconstruction of porous media are studied.On this basis,the improvement of their respective deficiencies is carried out,and the reconstruction results are further optimized.The main research contents and results are as follows:(1)The coal-rock image detection method based on phase consistency is studied.Aiming at the problem that the pore features obtained by the classical edge detection algorithm are sensitive to the brightness and contrast of the image,the image feature detection method based on frequency domain phase coincidence is introduced.The phase-coincidence operations of three coal-rock pore images with different characteristics are performed respectively.The experimental results show that the frequency-domain phase-consistent algorithm is insensitive to the local brightness and contrast of the image,and it also has a unidirectional correspondence to the pore characteristics,and it is more concise and accurate than the classical airspace edge detection algorithm.Compared with classical edge detection operators,the pore characteristics of coal-rock obtained by this method are stable and can be applied to a variety of image edge detection.(2)The role of pattern density function(PDF)in describing the morphological characteristics of coal-rock media is studied.The morphological aspects of training image(TI)are not well described by the 3D reconstruction results of the coal-rock based on the traditional intelligent optimization algorithm represented by genetic algorithm.To solve the problem,the PDFSIM is proposed by using the template to scan the TI to obtain PDF which is used as the constraint condition of the genetic algorithm.The experimental results show that PDF can constrain the morphological characteristics and physical characteristics of the reconstructed results,and the results are closer to the real structure.In addition,by using different template sizes to scan TI,the comparison experiments with TIs of different porosity are shown,indicating that the algorithm is feasible.(3)A kind of algorithm combining PDFSIM and biggest ball(BB)algorithm is studied.Although the PDFSIM can effectively improve the accuracy of the reconstruction results in morphological characteristics,there is still room for improvement in physical characteristics.Silin's biggest ball algorithm,proposed in 2003,was originally used to study the pore space configuration of sedimentary rocks.The thought of biggest ball can describe the morphological characteristics of TI very well after analysis.Meanwhile,the information contained in the biggest ball is more comprehensive,so that the reconstruction results are subject to multiple constraints.So the BB-PDFSIM algorithm is proposed.Besides,in order to improve the efficiency of reconstruction algorithm,the exchanging mechanism is changed.The random exchanging is changed to selective exchanging.The experimental results show that the BB-PDFSIM can get reconstruction results closer to the real structure,and the efficiency of the algorithm is improved significantly,which is a good solution to the shortage of PDFSIM.