Loading...
| Friend's email | |
| Your name | |
| Your email | |
| enter code | |
This page was sent successfuly
- Type of Document: M.Sc. Thesis
- Language: Farsi
- Document No: 54986 (06)
- University: Sharif University of Technology
- Department: Chemical and Petroleum Engineering
- Advisor(s): Masihi, Mohsen; Bozorgmehry Boozarjomehry, Ramin; Blunt, Martin
- Abstract:
- Digital analysis of reservoir rocks prepares a powerful tool in the field of porous media studies and provides appropriate information of single / multiphase flows by performing numerical simulations on porous media images. In order to access the necessary tools to benefit from digital rock analysis and characterization, we need accessing the porous media images, the geometric distribution and the spatial distribution of pores and grains through imaging or reconstruction methods. Porous media reconstruction methods are known as an alternative to imaging techniques. In addition, reconstruction methods have the ability to fix problems and imperfections associated with imaging, such as considering information at different scales of imaging resolution.To reconstruct porous media images, utilization of powerful tools such as deep learning causes fast reconstruction with accurate image features, statistical, dynamical and visual quality. Proper design of the deep network structure makes it possible to take into account the image properties of two different scales, and thus the details obtained in small-scale imaging can be added to the images produced in one scale. The algorithms presented in the present study enable the three-dimensional to three-dimensional, two-dimensional to two-dimensional and two-dimensional to three-dimensional reconstruction of new realizations of porous media images. The artificial intelligence paired algorithm for the multidimensional instance inserts the second-scale visual information (intra-granular pores) into the reconstructed images containing the inter-granular pores. With this tool, we can insert image information obtained from high-resolution two-dimensional images into the desired three-dimensional images. The developed statistical-artificial intelligence combination algorithm also allows two-dimensional to three-dimensional reconstruction of single or multidimensional, homogeneous or heterogeneous samples and thin-section samples. In this structure, conditional learning data is generated by the statistical toolbox and provided to the deep neural network. The results indicate the accuracy of the proposed algorithms. The developed algorithms will be a tool in characterization of hydrocarbon reservoirs and will provide a great help in estimating reservoir characteristics with accuracy beyond well logging
- Keywords:
- Porous Media ; Artificial Intelligence ; Deep Learning ; Pore Network Model ; Statistical Modeling ; Generative Adversarial Networks ; Image Reconstruction ; Multi-Scale Images Reconstruction
Digital Object List
-
محتواي کتاب
- view
Bookmark