Study On Three-dimensional Reconstruction Algorithm Of Complex Pore Structure Of Rock

Geomaterial is a kind of porous media material,which contains pore structures with different sizes and different properties.These pore structures can be represented by porosity,size,and geometry.In addition,the degree of pore connectivity,the distribution of pores,and some statistically based methods can be used to characterize the pore structure.If an effective model of pore structure of geomaterials can be formed,the macroscopic properties of geomaterials can be predicted by analyzing the characteristics of its internal pore structure,such as the distribution of stress and strain,stress conditions,seepage conditions,and so on.There are some effective three-dimensional reconstruction methods have been developed for pore structures of geomaterials,among which the numerical reconstruction method based on the simulated annealing algorithm is widely used.This method combines statistics-based correlation functions to describe the internal pore structure of natural rock and then uses computer methods to generate different reconstruction models with similar statistical information.The reconstruction method based on the simulated annealing algorithm can introduce more statistical correlation functions under a reasonable calculation time when reconstructing the pore model,showing a flexible superiority.In addition,the convergence speed of the algorithm can be accelerated by developing a statistical correlation function with a small amount of calculation.When the statistical correlation function can fully contain the pore distribution information,this method can be used to reproduce the real structure with similar statistical infomation.However,when high-order statistical correlation functions are added to reflect the connectivity and other properties of the rock pore structure,the amount time of reconstruction computation will increase significantly.In particular,there are also problems with efficiency in the face of larger-scale reconstructions,which take too long and lead to limited application scope.In addition,using this method for reconstruction requires more primitive core images used as a reference model,and obtaining these two-dimensional CT images of the actual pore structure is not only economically expensive,but also requires a large amount of CT scan and calculation time.In particular,when developing a very detailed 3D pore structure model,CT images of such rocks at different resolutions may be needed,and even images acquired by different imaging techniques may be needed,which increases costs and limits the scope of application of this reconstruction model.In many cases,there are needs for quick reconstruction models of natural rocks using only a small number of CT images.For example,at a site of a mining project,it is necessary to quickly reconstruct a corresponding digital model then analyze the model and make decisions.This has become a bottleneck problem in the reconstruction of three-dimensional models of complex structures of rocks.Although the ability of the simulated annealing reconstruction method to reconstruct pore structures has been demonstrated,few studies have used this method to reconstruct the grain structure of granular geomaterials.Examples of granular geomaterials include glutenite or conglomerate,sand,earth-rock mixtures,etc.Its macroscopic physical and mechanical properties are basically determined by its irregular particle structure,which consists of multi-scale,geometrically irregular,randomly distributed particles.If the reconstructed method can be used to reconstruct the granular structure of the geomaterial,it not only facilitates the quantitative description and prediction of the mechanical behavior of the geomaterial,but also can significantly reduce the cost of laboratory experiments and field experiments.According to the analysis of the current reconstruction algorithm for complex pore structures of rocks,this paper focuses on the following aspects:(1)Study on the 3D reconstruction efficiency of pore structures of geomaterials.From the above analysis,it can be seen that in the simulated annealing reconstruction method,the factors influencing the calculation efficiency mainly include the following three points:The first factor is the count of correlation functions used in reconstruction process.In order to pursue a higher accuracy of the reconstruction model,more and more statistical correlation functions are developed and used in the reconstruction process.In particular,some of the high-order statistical correlation functions show good theoretical results in the two-dimensional reconstruction experiments,but when they are applied to the three-dimensional reconstruction process,they will greatly reduce the efficiency of reconstruction.It is difficult to accept in engineering applications.The second factor is the size of the reconstructed model.The three-dimensional reconstruction model of geomaterials is mainly used to study how the internal microstructure of geomaterials affects its macroscopic physical and mechanical properties.At first studies require only a small three-dimensional reconstruction model of geotechnical materials,but as the research progressed further,the size of the reconstructed model required was also growing.In this case,the application of the reconstruction method will face the problems of slow convergence and time-consuming.Moreover,the efficiency of a particular step in the reconstruction algorithm.The simulated annealing reconstruction method includes several specific implementation steps,such as the extraction of statistical information of the reference model,the initialization of the reconstruction model,the selection of interchange points during the reconstruction process,the judgment of whether the interchange is accepted,and the implementation of correlation functions.Traditional reconstruction algorithms consider more of the common processing,but in some specific applications these common processes may waste computing resources.In the two-dimensional reconstruction of theoretical verification,the impact of this waste is not obvious.However,this waste of time is hard to accept when performing three-dimensional reconstruction.For these efficiency problems,this paper considers the introduction of parallel computing into the three-dimensional reconstruction algorithm of geomaterials,and uses parallel computing to accelerate the computational efficiency of the reconstruction algorithm.In addition,this paper also proposes an improved interchange acceptance criterion,which can improve the computational efficiency when using some specific statistical correlation functions for 3D reconstruction.(2)Study on using a minimum number of two-dimensional images of real rock to conduct reconstitution.In the reconstruction algorithm,the reference model is established by using a CT device or other device to obtain cross-sectional images of the rock,and then processing the images to obtain statistical information of the pore structure.However,this requires many high-resolution two-dimensional CT images,which are costly and computationally expensive,and it has a lower reconstruction efficiency for large-scale models.The bayesian information criteria are used herein to determine the minimum number of two-dimensional images required to ensure the accuracy of the expected reconstruction.Then use this minimum number of 2D images to build a reference model and perform reconstruction.Finally,by comparing the statistical correlation function of the pore structure of the reconstructed model with that of the prototype sandstone,and also the geometric and topological characteristics and the mechanical properties,the accuracy and effectiveness of the proposed method are verified.(3)Study on 3D numerical reconstruction method of irregular structure of granular rock.Granular geomaterials consist of a large number of multi-scale,geometrically irregular and randomly distributed particles that strongly influence their physical and/or mechanical properties.Accurately characterizing the complex structure of granular geomaterials remains a challenge.In this paper,a new computerized method is presented to numerically represent three-dimensional irregular particle structures based on natural conglomerate samples.In the proposed method,a two-point probability function along the coordinate axis direction and a diagonal direction and a two-point cluster function along the coordinate axis direction are used.To verify the accuracy of the proposed algorithm,the gravel characteristics and mechanical properties of the reconstructed three-dimensional grain structure were analyzed and compared with prototype conglomerate samples.