ICF10B

where β is the slope of the log-log plot. The fractal dimension, D, is related to the slope as follows: D=(8+ β ) / 2 (5) Results Results obtained by the above methods are summarized below. The power spectral measurement directly yields fractal dimension for 2-D data set. Whereas other two methods give the fractal dimension of 1-D set and these fractal dimensions were extended to represent 2-D data set by Dsurface = Dprofile + 1. For natural fracture surfaces: • The power spectral method yielded fractal dimensions between 1.70 and 2.25. They were expected to be greater than 2 to be in fractal regime but more than half of the samples yielded values less than 2. • The fractal dimensions obtained through variogram analysis varied between 2.10 and 2.33 for vertical and 2.25 and 2.62 for horizontal profiles (for maximum lag distance of 5). Roughness length measurement fractal dimensions were between 2.03 and 2.52 for vertical and 2.03 and 2.63 for horizontal direction profiles. In general two methods were observed as consistent. Also, the fractal dimensions of the profiles in horizontal direction were generally greater than vertical ones. For synthetic fracture surfaces: • The power spectral method yielded fractal dimensions between 2.00 and 2.50. They were expected to be greater than 2 to be in fractal regime and all the values fell into this interval unlike natural fracture surfaces. The fractal dimensions of upper and lower surfaces were not consistent for any loading rate. • The fractal dimensions obtained through variogram analysis varied between 2.21 and 2.52 for vertical and 2.43 and 2.59 for horizontal profiles (for maximum lag distance of 5). The fractal dimensions of upper and lower surfaces were consistent at any loading rate. As similar to the natural surfaces, dimensions in horizontal direction were generally higher than the dimensions in vertical direction. The difference between the fractal dimensions of horizontal and vertical profiles indicates the anisotropic nature of the fracture surfaces. Variogram analysis and roughness length measurement were, therefore, found as suitable methods to assess this feature of the surfaces. Detailed analysis of the fracture surfaces [4-6] and selection and application criteria of the methods [6,8,9] can be found in previous publications. The results gave an idea of the typical range of the fractal dimensions. Based on these observations, computer generated fracture surfaces were created for invasion percolation simulations. GENERATION OF SYNTHETIC SURFACES Mid-point displacement and successive random addition algorithm introduced by Voss [10] was used to generate 2-D self-affine fractional Brownian motion (fBm) data. The method relies on the displacements of mid points using a value of independent Gaussian random variable. This process produces 2-D selfaffine distributions, which have a fractal dimension D=3-H. Fig. 2 illustrates 3-D representation of fBm surfaces generated using the same random number seed with different fractal dimensions. As can be visually observed, increasing fractal dimension yields more tortuous surfaces. As the fractal dimension decreases and approaches Euclidean dimension of 2, smoother surfaces are obtained. Invasion percolation simulations were performed on these lattices to qualitatively analyze the displacement patterns between two fracture surfaces. INVASION PERCOLATION SIMULATIONS 4

RkJQdWJsaXNoZXIy MjM0NDE=