ICF10B

simulations were applied on the synthetically generated self-affine surfaces with different fractal dimensions. In the generation of synthetic representative surfaces, typical fractal dimension values obtained through the measurements were used. Finally, the simulation results were utilized to relate the surface roughness and the fluid transport in fractures. MEASUREMENT OF THE FRACTURE SURFACE ROUGHNESS To map the surface roughness, a computer-controlled surface scanning device designed by Develi et al. [1] was used. The device consists of three main parts, each of them capable of moving in three orthogonal directions. The displacement of each part is supplied by three step motors. The scanning is accomplished by a needle that is capable of moving up and down. The needle is fixed at a known distance over the sample and calibrated based on this distance. Once the needle touches the fracture surface, the number of step motor rounds is counted to calculate the traveled distance from the initial point. Then, the needle moves to next pixel by means of other two step motors. When the 55x55 array of pixels (each pixel is 1 mm) have been scanned, the process is terminated. Detailed description of the system can be found in the relevant reference [1]. 3-D representations of natural and synthetic surfaces mapped using this device are shown in Fig. 1. Natural surfaces are the outcrop marble samples cored 55 mm in diameter. Synthetic surfaces were created by Brazilian tests (indirect tensile fractures) on three different marble samples. During the tests, different loading rates between 0.05 and 0.2 kN/s were applied. The fracture surfaces were 55x55 mm. Only the middle 32x32 mm part of the samples was used in fractal analysis. FRACTAL DIMENSION CALCULATIONS Methods Applied to Measure Fractal Dimension The fractal dimension values were calculated using three methods applicable to self-affine fractal sets. Variogram analysis The variogram is defined as the mean squared increment of points: [∑ = + − = n i i h i V x V x n h 1 2 ( ) ( ) 2 1 ( ) γ ] (1) where h is the lag distance (distance between two successive points), γ (h) is variogram at lag distance h, n is the number of pairs at a lag distance h, and V(xi) is the sample values at location xi. Fractal distributions are characterized by a variogram model of the following form: γ γ (h) ho H = 2 (2) where H is called the Hurst exponent. H is related to the fractal dimension by D = 2-H = 2 - β/ 2 where β is the slope of lag distance, h vs. variogram, γ(h) plot in log-log scale and equal to 2H. The 32x32 points taken from the middle portion of 55x55 data set were used for the analysis. The fractal dimension of each profile with 32 points was calculated for both horizontal and vertical directions. Then the 32 fractal dimensions were arithmetically averaged. It was observed that the optimum lag distance yielding the most accurate results is between 4-6 for 32 data points [6] and therefore, maximum length distance was selected to be 5 in the application of Eq. 1. 2

RkJQdWJsaXNoZXIy MjM0NDE=