Investigating The Permeability Prediction Using Geometric Properties of Micro Computed Tomography Images by Linear Regression Models

Abstract

Challenges on rock absolute permeability prediction from tiny sample when laboratory apparatus is not applicable and without pore network modelling is remarkable. This prediction using the characterization of micro-computed tomography images have been studied in this paper. Twenty series of 2D micro computed tomography rock binary images have been collected, each of them was considered as a 3D binary image. Their geometric measures in 2D and 3D for measuring image properties have been considered using Minkowski functionals and available functions, developing a regression model, absolute permeabilities have been evaluated. Some 2D and 3D geometric properties are considered. The area, the perimeter and the 2D Euler number are 2D binary images properties. The volume, the surface area, the mean breadth also known as integral of the mean curvature, and the 3D Euler Number are 3D binary images properties. Porosity and number of objects also have been considered as parameters of a regression model.To perform linear regression, twenty-four parameters were evaluated and some of them were chosen to be used. An equation is proposed based on the extensive study conducted which can predict rock permeability. This equation has two sets of parameter coefficients, one set predicts high permeability rocks (above two Darcy) and the other for low and medium permeability (less than two Darcy) which can be used for carbonated rock. Average absolute relative error for conducted cases is 0.06.