AU2020103292A4 - A novel approach for evaluation of the flow paths into underground excavations using three-dimensional discrete fracture network (3D DFN) model - Google Patents

Abstract

The invention discloses a novel approach for detecting the flow paths into underground excavations using three-dimensional discrete fracture network (3D DFN) model. In the procedure, DFN model is generated using Monte Carlo simulation, according to the distribution parameters of the fracture properties obtained from field measurements. Then, the intersections between the fractures as well as between fractures and boundaries are determined, considering their relative spatial positions. After determination of all the intersections within DFN model, to specify the flow network around the excavation, the fractures that have no connection with any boundaries are considered as non-conductive fractures and are eliminated from the DFN model. In the traditional approach, in order to determine the flow paths, dead-end fractures are only removed from fracture networks. However, some of the loop paths arise in the fracture network that the traditional approaches identify them as flow path while water is stationary within them. The algorithm presented in the procedure accurately detects such paths and consequently reduce model uncertainty. 001I rII Im I I ------- CI r'I t.0 + .' d + I+'+ - - - - - - - - - - - - - - - - - - CDC I II I I I I S1Z I I I W > WD 4I 0 0 ) M 0)I I cc I Iu 0 f I IuI Ia I I I I I I I-~ I flcu 00 ct I I ct - - - - - - - - - - - -00 aPWNE

Description

001I rII Im I I---

CI r'I

t.0

+ .' d +

I+'+

- - - - - - - - - - - - - - - - - -

CDC I II I I I I S1Z I I I W > WD 4I

0

) M I 0)I cc I Iu 0 f I IuI Ia I I I I I I I-~ I flcu

00 ct I I ct

- -- - -- - -- - - - 00 aPWNE

Editorial Note 2020103292 There is 4 pages of Description only.

A novel approach for evaluation of the flow paths into underground excavations using three dimensional discrete fracture network (3D DFN) model

BACKGROUND

When an underground space is excavated within a groundwater aquifer, one of the significant problems is water movement toward excavation due to the presence of a low-pressure area in aquifer caused by the excavation. Generally, the basic approaches for fluid flow modeling in rock masses are consist of continuum and discrete that each one has its own specific applications. Since the fractures have a major role in the rock mass hydraulic conductivity, the use of the discrete approach is necessary to obtain accurate estimations of fluid flow. One of the discrete models that satisfactorily express the impact of fractures on flow in a rock mass is discrete fracture network (DFN) model. In order to evaluate the flow in fractured rocks, some DFN models were developed in two-dimensional (2D) space. However, because of the three-dimensional (3D) nature of fractures in real conditions, the 3D models provide a more appropriate prediction of the water flow through fractured rock than the 2D models. A novel approach has been designed to evaluate the flow paths into underground excavations using 3D DFN model. The presented approach is a robust method to assess the fracture connectivity and distinguishes the non-conductive fractures to detect the flow paths in the rock mass surrounding an underground excavation. This procedure potentially could be used to evaluate the flow paths around the injection wells in solution mining or in-situ recovery in unconventional mining methods.

2. BRIEF DESCRIPTION OF THE DRAWINGS

Illustrative embodiments of the present invention are described in detail below with reference to the attached drawing figures, which are incorporated by reference herein and wherein:

Figure 1 illustrates the procedure for identification of inflow paths around underground excavations using 3D DFN model;

Figure 2 schematically illustrates a complete description of the orientation of an elliptical fracture in 3D space;

Figure 3 illustrates the workflow process for determining the intersections between the fractures as well as between fractures and boundaries in a DFN model;

Figure 4 illustrates the workflow process for specifying the flow network around an excavation by eliminating the non-conductive fractures from DFN model;

Figure 5 schematically illustrates an example of the process of tracking a loop path extended from a fracture.

3. DETAILED DESCRIPTION

As shown in Figure 1, workflow process 100 describes steps for specifying the inflow paths to an underground excavation (steps 110-124), which are schematically illustrated in this figure. One of the most important steps is to generate DFN model 118 based on the field measurements 110. To generate a DFN model 118, the fracture properties such as location, size and orientation are considered as random variables with respective distribution functions, which can be obtained from field measurements 110. The fracture properties are reproduced using Monte Carlo simulation based on their probability distribution parameters 112 resulting from the field measurements 110.

To reproduce the fracture locations (step 114), the homogeneous Poisson model has been applied. To calculate the coordinate of fractures centre, the interest domain 126 first is subdivided into sub domains, Ai (128), which are disjoint and then below procedure is followed:

1) In the domain of interest 126, the number of points falling inside Ai(128), N(A), follows a Poisson distribution with mean value of pi=.v(A).Where v(Ai) is the volumetric measure of Ai and L is the fracture density: 'in P (Ni = n) = e-_"( )

To simulate a Poisson random variable with mean pi, generate a sequence of independent random variables uniformly distributed in [0, 1], U, U2..., stopping when the below condition is reached: k

I-IUj < e-P

The variable n=k then has the desired Poisson distribution.

2) Considering N(A1 )= n, for each of the events in sub-domain Ai, generate three values using uniform distributions and they are used to determine the coordinates of the point simulated inside the sub-domain.

In order to reproduce the fracture size, an array of radius value (r) is first generated using Monte Carlo simulation according to a standard normal distribution (normal distribution with a mean of zero and a standard deviation of 1). Then the variable R follows a lognormal distribution with mean pL and standard deviation aL:

R = exp (PL +rcL)

The fractures are simulated as ellipsoid with major radii of R, pL is the mean of the log (r) and UL is the standard deviation log(r).

By assuming independence between sub-sets, each sub-set is modelled separately with a similar procedure. For a complete description of the orientation of an elliptical fracture 210 in 3D space, three angles consisting dip angle 212, dip direction angle 214 and rotation angle 216 is required as shown in Figure 2. The dip direction angle 214 and the rotation angle 218 are defined as deviation of projected dip direction 216 from the North and the angle between the major axis of fracture 210 and the dip direction 216 in the fracture plane 220, respectively. In a sub-set, orientation of the fractures is as values that scattered around a certain value, and so each sub-set has a particular concentration in orientation. Thus, Fisher distribution considered as a suitable distribution to express fracture orientation. Based on the mean and concentration of orientation, two sets of angles stochastically are generated for the fractures using Fisher distribution. Accordingly, two angles named spin and deviation angle are generated by shifting away from the mean dip and dip direction of the fracture sub-set, respectively. The spin angle 0 is generated as a random number between 0 and 1 which is multiplied by 360 so as to be between 0° and 360. To generate the deviation angle RF, a random number R u is first generated from a uniform distribution between 0 and 1. Then RF is calculated from a Fisher distribution as follows:

In (1 - Ru)+ RF =aCOS[( K |

where K is Fisher constant which obtained from field measurements. By adding the spin angle and angle ofRFto the mean dip and dip direction of the fracture sub-set, dip angle 212 and dip direction angle 214 are reproduced for the fractures. In addition, rotation angle 216 is reproduced from a Fisher distribution by the Monte Carlo simulation.

By assigning the size and orientation to the fractures (step 116) that their locations were reproduced using step 114, a three-dimensional DFN model 118 is generated.

Figure 3 shows the workflow process 300 to detect the intersections between the fractures as well as between fractures and boundaries (i.e., the surface of excavation 134 and the surfaces surrounding the domain 126), corresponding to the step 120. Based on the part 310, to find the intersections between the fractures, it is considered around each fracture a sphere with a diameter equal to major radii of the fracture. Then, for each fracture, the possibility of intersection is evaluated only for the fractures whose centre from the centre of the fracture is less than the summation of the radius of spheres related to both fractures. To investigate the connection between the fractures and the boundaries, the possibility of intersection for the fractures whose centre from the boundary is less than their major radii is evaluated using the part 312.

In the traditional approach, to determine the flow paths, dead-end fractures are only removed from fracture networks. Nevertheless, some of the loop paths arise in the fracture network 118 that the traditional algorithms identify them as flow path while water is stationary within them. By the algorithm 400 shown in Figure 4, in addition to the dead-end fractures (Part 410), fractures within such paths are accurately detected and removed (Part 412). The algorithm 400 works as following steps:

a) Remove all the dead-end fractures.

b) Identify the paths that are extended from each fracture.

c) Track each of the paths to evaluate connection with boundaries.

d) If the path does not lead to any boundaries, all the fractures forming the path are removed.

e) If no path extended from the fracture connect to any boundary, the main fracture is removed from the network flow.

Figure 5 shows a schematic of the process of tracking a loop path 510 extended from the fracture 512. To track the path extended from the fracture 512, fracture that is immediately connected to the fracture is considered as the path starter 514 and by advancing through the path, the fractures miss the previous their connections. Since the last remaining fracture of the path has no intersection with any boundary, the fractures forming the path are considered as non-conductive and are removed from the fracture network 118.

Eliminating the non-conductive fractures from the fracture network 118 (step 122) results in a flow network 130 by which water flows into the excavation 134. By linking the connectivity nodes (i.e., the centre of the intersections and centre of fractures) in the flow network 130 (step 124), inflow paths 132 to the excavation 134 can be specified.

Editorial Note 2020103292 There is 1 page of Claims only.

Claims (3)

Claims WHAT IS CLAIMED IS:

1. A DFN-based novel approach to specify the flow paths into underground excavations that comprises:

DFN generation based on the fracture properties includes fracture location, fracture size and fracture orientation, which are reproduced according to their distribution parameter that obtained from field measurements;

Detecting the intersections within DFN model;

Eliminating the non-conductive fractures;

Specifying the inflow paths by linking the connectivity nodes of conductive fractures.

2. The DFN-based procedure of Claim 1, wherein the intersections for a fracture are evaluated only for the fractures and the boundaries that are in a certain distance from the fracture. This takes a lower computational burden than the traditional approaches.

3. The DFN-based procedure of Claim 1, wherein the loop paths at which water is stationary and have no effect on the flow through the fracture network are detected and eliminated from the DFN model. However, traditional approaches are unable to detect such paths. This novel approach improves DFN-based fluid flow modelling and reduces the model uncertainties.

FIG. 1

FIG. 2

FIG. 3

FIG. 4

FIG. 5