Fig. 1. (a) Fractured rock samples and rough fracture surfaces. Fracture sizes are 20 20 cm, (b) transparent model fractures casted from the original rock fracture surfaces,
(c) 3-D view of the scanned surface of model fracture Fr1.
sides. Similarly, the complementary parts of the models used by Auradou et al. (2001) were silicone rubber and transparent epoxy.
As discussed by Wang et al. (1988), perfectly matching fractures composed of mirror image rough surfaces would theoretically not present any aperture inside if they are in the closed form (e.g. closed joints in nature). Aperture will theoretically be zero as the adjacent walls will be in contact overall, the fracture and the height difference of the contacting walls from a reference plane will be zero at any point throughout the fracture. Yet, they will have constant aperture structure if they are opened due to the divergent displacement of the adjacent fracture walls along the normal vector of the average fracture plane (e.g. open joints in nature).
Model fractures used in this study exemplify perfectly matching, tightly closed rough fractures with mirror image adjacent surfaces, which can be considered as common case in the deep earth crust. Note that the term ‘‘fracture’’ refers to ‘‘closed joint with perfectly matching adjacent walls’’. It is important to emphasize that no relative displacement was applied to the fracture walls that might generate sheared fractures. Thus, any void space or aperture structure does not occur within our model fractures. In accordance with this, no visible gaps inside the models were recognized in practice, at least at the macroscopic scale. In the case of injection of a fluid into our model fractures, a mechanical aperture structure was induced by injection pressure, and flow is expected to occur predominantly through this forced-mechanical aperture. As many of the fractures in nature are found to be almost closed, the study of flow through tightly closed rough fractures was indicated by Genabeek and Rothman (1999) as an important issue in solving the transport in channel networks.
The main parts of the experimental set-up are composed of an injection stand, a micro annular gear pump (model mzr 7205), a digital pressure gauge, an electronic pump control unit and software, a high resolution camera, and fluorescent black light source (Fig. 3a). The maximum injection rate the pump is able to supply is 4.667E06 m3/s (280 ml/min). Resolution of the digital pressure gauge is 68.95 Pa (0.01 psi).
Quantification of fracture surface roughness
To measure the fractal dimension of each model, we first mapped the rough surfaces of the polyurethane upper parts of the model fractures using the fully computer-controlled surface scanning device introduced by Develi (2006); Develi et al. (2001). A 190 190 mm2 square portion, selected in the middle of each surface by taking its geometrical centre as a reference point, was scanned and elevations (z) were automatically digitized with a sampling interval of 1 mm in the horizontal (x and y) axes. The measurement resolution in vertical (z) axis was 1/10 mm. Following this procedure, a 2D data set consisting of 190 190 data points was automatically obtained for each surface and simultaneously logged into a computer. A 3D view for one of the digitized surfaces (e.g. for model fracture Fr1) plotted using 2D data set is displayed in Fig. 1c.
Fig. 2. (a) Model fractures manufactured after moulding and casting the fracture surfaces seen in Fig. 1. Two sides were sealed and fluid was injected from one side (inlet) in the direction shown above, (b) variogram fractal dimensions (Dva) of the profiles in the example of the three model fractures, namely Fr1, Fr2 and Fr3.
Once the digitization was completed, the fractal dimension value (Dva) of each surface was calculated using the 2D data sets through the variogram analysis. A summary of the mathematical description of this method can be found in Develi and Babadagli (1998) and Babadagli and Develi (2001). In the calculations, 1D profile data extracted from the 2D data sets were used. In a 2D data set, there are 190 lines and 190 columns through x and y-directions, respectively. Thus, the method was applied to all 190 profiles in y-direction. This direction is also parallel to the injection direction in the flow tests. The fractal dimension value of each profile consisting of 190 data points was computed. All the values calculated for each of the 190 profiles ranged between 1 and 2, as theoretically expected. The arithmetic mean of the 190 fractal dimension values for each model fracture surface was taken as the surface fractal dimension.