Bedrock discrete fracture network model generation method based on multi-source drilling data

Abstract

The invention discloses a bedrock discrete fracture network model generation method based on multi-source drilling data. The bedrock discrete fracture network model generation method comprises the following steps: step 1, preprocessing multi-source data; step 2, recognizing a drilling fracture image, and clustering trace line segments recognized in a drilling image area into continuous fracture trace lines; step 3, fracture parameter statistics; step 3.1, calculating and determining geometric parameters of the fracture; step 3.2, performing statistics on a rock mass fracture network distribution model and parameters; 4, generating semi-determined fracture networks in batches; 5, multi-source data iterative analysis is carried out, and an optimization model is selected. Modeling is carried out through multi-source data cross comparison without depending on a single data source, and the accuracy of the model is improved. Automatic analysis is achieved through a digital method, manual operation is reduced, dependence on engineering experience is reduced, and large-scale application in the engineering scale can be achieved; and meanwhile, the determined crack of the drilling area and the uncertain crack of the drilling adjacent area are considered, so that the model uncertainty of grouting engineering modeling is greatly reduced.

Description

Technical Field

[0001] The present invention relates to the technical field of digital reconstruction of fractured rock masses in grouting engineering, and in particular to a method for generating a bedrock discrete fracture network model based on multi-source drilling data. Background Art

[0002] As my country's large-scale water conservancy and hydropower projects develop in areas with complex geological conditions, the scale of grouting projects continues to expand. In grouting projects, fractures of various scales and the resulting fracture networks dominate the primary mechanical and seepage properties of the bedrock. The essence of bedrock grouting is to alter the bedrock's geological conditions by injecting grout and allowing the grout to seep through the bedrock. Therefore, thoroughly examining and analyzing changes in bedrock geology before and after grouting is a key component of grouting project design and construction, as well as the evaluation of grouting project effectiveness. In recent years, with the advancement of computer technology and the widespread adoption of digital modeling techniques, it has become possible to achieve high-precision modeling of grouting bedrock fracture networks through multi-source data fusion.

[0003] Traditional grouting engineering data mainly includes construction data such as drilling, inclination measurement, groundwater records, water pressure, grouting, and post-grouting inspection hole water pressure. Among them, water pressure and grouting data have been automatically collected, while drilling, inclination measurement, and groundwater records must be manually entered. In recent years, with the development of sensor technology, image analysis technology, and inversion analysis methods, the introduction of multi-source data in grouting bedrock modeling has gradually become a hot research direction. In the process of integrating multi-source data, the difficulty and key to improving modeling accuracy are to fully utilize various types of data through algorithms and cross-compare the data. Due to the large amount of drilling footage in grouting projects, it is difficult to manually handle the workload of data analysis. Therefore, the basis for rapid modeling of grouting bedrock based on multi-source drilling data is to use a unified coordinate system and model component coding, digitize the basic construction data, and automatically import it into the database.

[0004] As mentioned above, the goal of grouting bedrock modeling is to obtain the spatial geometric configuration and seepage characteristics of the fracture network in the bedrock. Traditional discrete fracture network modeling mainly uses the method of borehole geostatistics to obtain the dominant grouping of fractures, and then analyzes the distribution law of fracture occurrence, size, and density under each dominant group. Based on the mathematical statistical distribution model of each dominant group, a discrete fracture network is generated. For details, see "Generation and visualization technology of three-dimensional network of random structural surface of rock mass" [J], Journal of Rock Mechanics and Engineering. 2007, 26(12): 2564-2569. This method only uses the data of exposed traces and is usually suitable for analyzing the geological sketch modeling of tunnel side arches and tunnel faces, and is more dependent on the experience of geological engineers. For other data during the drilling construction process, such as drilling data and water pressure data, numerical simulation inversion analysis can be used to cross-check with the random modeling results, thereby reducing the uncertainty of the discrete fracture network model. For details, see "Simulation-based investigation on the accuracy of discrete fracture network (DFN) representation" [J], Computers and Geotechnics. 2020, 121: 103487. These methods are widely applicable to all underground projects, and the data source boreholes are all defaulted to geological exploration boreholes. In grouting projects, the main source of geological data is the construction boreholes, which are themselves located within the construction area. The surface traces revealed by the borehole camera are the determined fractures in the constructed model, and its water pressure data also reflects the actual connectivity and permeability properties of the local modeling area. Therefore, the use of semi-deterministic models for discrete fracture network modeling in grouting projects has significant advantages. Summary of the Invention

[0005] The technical solution adopted by the present invention to solve the above technical problems is to propose a multi-source data cross-comparison modeling method to reduce the uncertainty of discrete fracture networks. To solve the above technical problems, the technical solution adopted by the present invention is:

[0006] A method for generating a bedrock discrete fracture network model based on multi-source borehole data includes the following steps:

[0007] Step 1: Preprocess multi-source data and calculate the DPI (Drilling Process Index). Based on the change of the DPI in the depth direction of the borehole, select the sudden change area, determine the area where the borehole may have cracks, and crop the borehole image at this location;

[0008] Step 2: drilling crack image recognition, clustering the trace segments identified in the drilling image area into continuous crack traces;

[0009] Step 3, crack parameter statistics;

[0010] Step 3.1: Calculate and determine the geometric parameters of the crack;

[0011] Step 3.2: Statistical rock mass fracture network distribution model and parameters;

[0012] Step 4: Batch generate semi-deterministic fracture networks;

[0013] Step 5: Iteratively analyze multi-source data and select the optimal model.

[0014] In some specific embodiments, step 1 includes: obtaining drilling parameters such as drilling speed, drill bit speed, and drilling pressure during drilling through digital drilling monitoring, and calculating a drilling process index (DPI), as shown in formula (1):

[0015]

[0016] Where V is the real-time drilling speed [m / min], V ref is the benchmark drilling speed [m / min]; N is the real-time drill bit speed [rpm], N ref is the rated drilling speed of the drill bit [rpm]; P is the real-time drilling pressure [MPa], P ref is the design maximum pressure [MPa]; the value range of α, β, γ is 0~1.

[0017] In some specific embodiments, step 2 includes: identifying the cropped defect area image based on a deep learning algorithm, exporting the defect area into a binary image, and distinguishing between pores and cracks; extracting the skeleton and contour of the crack traces revealed by the drilling image, and obtaining the location and occurrence geometric characteristics of the drilling cracks through burr removal and segmented clustering methods; the skeleton of the crack trace segments revealed by the drilling image can be described as a parameter set F i =[θ i ,φ i ,d i ], respectively representing the inclination and dip of the crack, and the midpoint depth of the trace segment; all trace segments contained in this image are matched for similarity, and their similarity coefficient can be given by formula (2):

[0018]

[0019] Among them, σ θ and σ φ Represent the tolerance difference of inclination and dip, ω θ 、ω φ With ω d Represent the similarity coefficients of dip, inclination and center position respectively; through this step, the trace segments identified in the borehole image area can be clustered into continuous fracture traces.

[0020] In some specific embodiments, the deep learning algorithm in step 2 adopts DeepLabv3+.

[0021] In some specific embodiments, step 3 includes: using the morphological characteristics of each fracture trace clustered in step 2 to obtain the geometric parameters of the fractures revealed by the borehole, and then obtaining the fracture network spatial distribution model and parameters of the formation where the borehole is located through mathematical statistics.

[0022] In some specific embodiments, step 3.1 includes, using the fracture traces revealed by the borehole image in step 2, first, calculating the occurrence of the fracture surface; assuming that the borehole image is developed along the north direction, the fracture traces can be fitted into a trigonometric function, The inclination of the crack is φ, and its inclination angle, θ, can be obtained from formula (3):

[0023]

[0024] Second, the crack trace length is calculated based on the product of the number of pixels in the skeleton and the size of a single pixel;

[0025] Third, the normal direction of the crack skeleton is defined as the width direction of the crack, based on which the average crack width can be calculated;

[0026] Fourth, the position and spacing of the traces can be measured by the line survey method. The position is the depth coordinate of the intersection of the trace and the survey line, and the spacing is the distance between two adjacent cracks in the survey line.

[0027] Fifth, the density of cracks is obtained by calculating the spacing and number of cracks.

[0028] In some specific embodiments, step 3.2 includes: grouping the borehole fractures according to the occurrence of the fractures obtained in step 3.1, and establishing a distribution histogram between the statistical distribution parameters and the fracture frequency for each group of fractures, fitting the probability distribution form obeyed by the parameters according to the drawn histogram, and testing the goodness of fit; in terms of fracture grouping, performing stereographic projection on the occurrence of the fractures to form a pole diagram, clustering the pole diagram using the K-Means clustering algorithm, obtaining the dominant group and the occurrence corresponding to the dominant group, and the number of fractures in each dominant group; performing parameter statistics on the fractures of each group according to the grouping, and fitting the fracture occurrence distribution using the Fisher distribution model, as shown in formula (4):

[0029]

[0030] Where k is the concentration parameter, is the dominant normal vector of the fracture group, is the crack normal vector; the crack spacing and size are fitted using a negative exponential distribution, and the average crack width is fitted using a lognormal distribution; the fitted crack distribution statistical model and its parameters need to be verified using the corresponding fitting test method, the Fisher distribution needs to use a spherical test, and the lognormal distribution needs to perform a QQ test.

[0031] In some specific embodiments, step 4 includes generating a determined fracture network in the area to be modeled based on the determined fractures counted in step 3.1, and generating a random fracture network in the inter-pore area using the Monte Carlo method based on the parameter distribution obtained in step 3.2.

[0032] In some specific embodiments, step 4 specifically includes:

[0033] Step 4.1: Generate and determine the fracture network; some fractures in the area to be modeled have been revealed by drilling. According to step 3.1, their locations and occurrences are basically determined. The fracture sizes are determined by the cross-hole fracture clustering algorithm.

[0034] Step 4.2: Generate a random fracture network between holes; obtain the number of fractures in each group based on the density distribution model and parameters of the grouped fractures obtained in step 3.2, and then generate random numbers for the size, position, occurrence, and width of each fracture based on other parameter distribution models and parameters to establish a random fracture network model; delete the fractures that intersect with the boreholes in the random model; and fuse the deterministic fracture network generated in step 4.1 to form a semi-deterministic fracture network model.

[0035] In some specific embodiments, step 5 includes, in combination with the water permeability of each borehole section during the construction process, using the fracture network method to quickly numerically calculate the model water permeability of the borehole section; comparing the numerical calculation results with the actual water pressure data, and selecting a discrete fracture model with an error of less than 20% with the actual water pressure test data as the preferred model.

[0036] The beneficial effects of adopting the above technical solution are: cross-comparison of multi-source data (including drilling data, borehole camera data, grouting hole water pressure data, etc.) is used for modeling without relying on a single data source, thereby improving model accuracy; and automatic analysis through digital methods reduces manual operation and dependence on engineering experience, which can be applied on a large scale at the engineering scale; at the same time, the determined cracks in the drilling area and the uncertain cracks in the adjacent area of ​​the drilling area are taken into account, greatly reducing the model uncertainty of grouting engineering modeling. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0038] Figure 1 It is the overall flow chart of the present invention;

[0039] Figure 2 In one embodiment of the present invention, borehole image recognition and fracture clustering based on while-drilling data are performed;

[0040] Figure 3 This is a schematic diagram of statistical results of crack determination by drilling in one specific embodiment of the present invention;

[0041] Figure 4a It is a fracture advantage grouping and parameter statistics based on geo-statistical methods in one embodiment of the present invention, and a fracture grouping diagram of the occurrence distribution extreme coordinates;

[0042] Figure 4b In one embodiment of the present invention, the fracture advantage grouping and parameter statistics based on the group fracture size distribution fitting are performed;

[0043] Figure 4c In one embodiment of the present invention, the fracture advantage grouping and parameter statistics based on the group fracture aperture distribution fitting are performed;

[0044] Figure 5a A semi-deterministic discrete fracture network for semi-deterministic modeling of a fracture network in a borehole and its adjacent area according to one embodiment of the present invention;

[0045] Figure 5b It is a semi-deterministic fracture pipe network model for semi-deterministic modeling of a fracture network in a borehole and its adjacent area in one of the specific embodiments of the present invention. DETAILED DESCRIPTION

[0046] The following is a clear and complete description of the technical solutions in the embodiments of the present invention, in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts are within the scope of protection of the present invention.

[0047] In the following description, many specific details are set forth to facilitate a full understanding of the present invention. However, the present invention may also be implemented in other ways different from those described herein. Those skilled in the art may make similar generalizations without violating the connotation of the present invention. Therefore, the present invention is not limited to the specific embodiments disclosed below.

[0048] A method for generating bedrock discrete fracture network models based on multi-source borehole data, referring to Figure 1 , including the following steps:

[0049] Step 1: Multi-source data preprocessing. Through digital drilling monitoring, drilling speed, drill bit speed, drilling pressure and other drilling parameters are obtained during drilling, and the drilling process index (DPI) is calculated as shown in formula (1):

[0050]

[0051] Where V is the real-time drilling speed [m / min], V ref is the benchmark drilling speed [m / min]; N is the real-time drill bit speed [rpm], N ref is the rated drilling speed of the drill bit [rpm]; P is the real-time drilling pressure [MPa], P ref is the maximum design pressure [MPa]; α, β, and γ range from 0 to 1 and are determined by field tests. Based on the change in the drilling process index along the borehole depth, a sudden change region is selected to identify the area where cracks may exist in the borehole, and the borehole image at this location is cropped.

[0052] Step 2: Drilling crack image recognition. Based on deep learning algorithms (e.g., DeepLabv3+), the cropped defect area image is identified, the defect area is exported as a binary image, and the pores and cracks are distinguished. For the crack traces revealed by the drilling image, their skeletons and contours are extracted, and the key geometric features of the drilling cracks, namely the position and occurrence, are obtained through methods such as burr removal and segmented clustering. It should be noted that the skeleton of the crack trace segments revealed by the drilling image can be described as the parameter set F i =[θ i ,φ i ,d i ], respectively representing the inclination and dip of the crack, and the midpoint depth of the trace segment. All trace segments contained in this image are matched for similarity, and their similarity coefficient can be given by formula (2):

[0053]

[0054] Among them, σ θ and σ φ Represent the tolerance difference of inclination and dip, ω θ 、ω φ and ωd represent the similarity coefficients of dip, inclination, and center position, respectively. Through this step, the trace segments identified in the borehole image area can be clustered into continuous fracture traces.

[0055] Step 3: Fracture parameter statistics. Using the morphological characteristics of each fracture trace clustered in step 2, the geometric parameters of the fractures revealed by the borehole are obtained. Then, through mathematical statistics, the spatial distribution model and parameters of the fracture network in the formation where the borehole is located can be obtained.

[0056] Step 3.1: Calculate and determine the geometric parameters of the fracture. Using the fracture traces (in the form of pixels) revealed by the borehole image in step 2, first, the occurrence of the fracture surface can be calculated. Assuming that the borehole image is developed along the north direction, the fracture traces can be fitted into a trigonometric function. The inclination of the crack is φ, and its inclination angle, θ, can be obtained from formula (3):

[0057]

[0058] Second, the length of the crack trace is calculated based on the product of the number of pixels in the skeleton and the size of a single pixel. Third, the normal direction of the crack skeleton is defined as the width direction of the crack, based on which the average width of the crack can be calculated. Fourth, the position and spacing of the trace can be measured by the line survey method. The position is the depth coordinate at the intersection of the trace and the survey line, and the spacing is the spacing between two adjacent cracks in the survey line. Fifth, the density of the cracks can be obtained by calculating the spacing and number of cracks. In summary, step 3.1 can automatically analyze the position, occurrence, average width, and trace length of all determined cracks that intersect with the borehole through the borehole crack trace revealed in step 2; at the same time, the crack spacing and density of the current borehole can also be obtained.

[0059] Step 3.2: Statistical rock mass fracture network distribution model and parameters. According to the borehole fracture occurrence obtained in step 3.1, group them, and establish a distribution histogram between the statistical distribution parameters and the fracture frequency for each group of fractures. Fit the probability distribution form obeyed by the parameters based on the drawn histogram and test the goodness of fit. In terms of fracture grouping, the occurrence of the fractures is projected stereographically to form a pole diagram. The pole diagram is clustered using the K-Means clustering algorithm to obtain the dominant group and the occurrence corresponding to the dominant group, as well as the number of fractures in each dominant group. Parameter statistics are performed on the fractures of each group according to the grouping. Generally, the Fisher distribution model is used to fit the fracture occurrence distribution, as shown in formula (4):

[0060]

[0061] Where k is the concentration parameter, is the dominant normal vector of the fracture group, is the fracture normal vector. Similarly, the fracture spacing and size are fitted using a negative exponential distribution, and the average fracture width is fitted using a lognormal distribution. It is important to note that the fitted fracture distribution statistical model and its parameters must be verified using the corresponding fitting test methods, such as the spherical test for the Fisher distribution and the QQ test for the lognormal distribution.

[0062] Step 4: Batch generation of semi-deterministic fracture networks. Based on the determined fractures counted in step 3.1, generate a deterministic fracture network for the area to be modeled. Using the parameter distribution obtained in step 3.2, use the Monte Carlo method to generate a random fracture network for the inter-pore area.

[0063] Step 4.1: Generate and determine the fracture network. Some fractures in the area to be modeled have been revealed by drilling, as detailed in Step 3.1. Their locations and occurrences are essentially determined, and the fracture sizes are determined by the cross-hole fracture clustering algorithm.

[0064] Step 4.2: Generate a random inter-hole fracture network. Based on the density distribution model and parameters for the grouped fractures obtained in Step 3.2, determine the number of fractures in each group. Then, based on the distribution models and parameters for other parameters, generate random numbers for the size, location, occurrence, and width of each fracture to establish a random fracture network model. Fractures that intersect with boreholes in the random model are deleted. This is combined with the deterministic fracture network generated in Step 4.1 to form a semi-deterministic fracture network model.

[0065] Step 5: Iterative Analysis of Multi-Source Data. The semi-deterministic fracture network model generated in Step 4 still has a certain degree of randomness, so theoretically, an infinite number of fracture networks can be generated in batches. This step combines the permeability of individual borehole segments during construction and utilizes the fracture network method to rapidly numerically calculate the model permeability of each borehole segment. The numerical calculation results are compared with actual water pressure data, and the discrete fracture model with an error of less than 20% from the actual water pressure test data is selected as the preferred model.

[0066] One specific embodiment is described in detail below.

[0067] Step 1: Multi-source data preprocessing

[0068] This embodiment is based on the drilling data of a curtain grouting project in a reservoir construction project in Northwest my country. First, digital drilling monitoring is performed to obtain the drilling parameters of the grouting hole. The drilling parameters are expanded and analyzed along the drilling depth direction. For example, the drilling speed V = 0.8m / min (the reference speed V) at a drilling depth of 2.5m is ref =1.2m / min), speed N = 110rpm (rated speed N ref =150rpm), pressure P = 12MPa (design pressure P ref =15MPa). Taking weight coefficients α=0.5, β=0.3, γ=0.2, we can calculate DPI=0.5×(0.8 / 1.2)+0.3×(110 / 150)+0.2×(12 / 15)=0.71. The change trend of DPI value along the depth direction is analyzed, as shown in the following figure: Figure 2 As shown in (a)-(b), this method automatically marks and intercepts the borehole camera images in the area where the DPI value suddenly changes.

[0069] Step 2: Drilling crack image recognition

[0070] Using DeepLabv3+ network Figure 2 (b) Perform semantic segmentation and output Figure 2 (d) shows the binary image. The suspected crack area is extracted by threshold segmentation. Figure 2 (c) shows that there are multiple obvious traces in this section. After morphological processing to remove burrs, three crack skeletons and contours are extracted, as shown in Figure 2. Figure 2 (d)-(e). The identified crack trace segments are clustered to obtain complete crack traces, and the traces of different cracks are marked with different colors, such as Figure 2 (f) shown.

[0071] Step 3: Fracture parameter statistics

[0072] Step 3.1: Determine fracture parameter statistics. Based on the fracture trace data obtained in step 2, the occurrence, location, and average aperture of each fracture revealed by the borehole can be calculated. Figure 3 As shown in the figure, the parameters of the clustered fractures were extracted: 12 fractures were identified in the 5m hole section, and the fracture line density was 2.4 fractures / meter. The position, occurrence and average aperture of each fracture could be calculated.

[0073] Step 3.2: Statistical analysis of the rock mass fracture network distribution model and parameters. Based on the fracture data from all borehole sections within the same depth range within the modeling area, over 200 confirmed fracture data points were obtained. As shown in Figure 4(a), the fracture occurrences were projected onto a stereographic projection network and subjected to K-Means clustering to obtain three groups of dominant occurrences, based on which the confirmed fractures were grouped. For each group of fractures, the size and aperture distribution model and corresponding distribution model parameters were statistically analyzed, as shown in Figures 4(b)-(c).

[0074] Step 4: Semi-deterministic fracture network generation

[0075] Step 4.1: Generate a definitive fracture network. Based on the definitive fracture data obtained in Step 3.1, generate definitive fractures that intersect the borehole, as shown in Figure 5(a). The definitive fractures revealed by the borehole are marked as red planes. To facilitate subsequent calculations, adjacent parallel fracture planes are integrated into the same fracture plane.

[0076] Step 4.2: Generate random interpore fractures. Based on the random fracture network distribution model and parameters obtained in Step 3.2, 36 random fractures were generated within the modeling area. The random fracture generation process employed an iterative approach, whereby any generated fractures that intersected a borehole were deleted and regenerated. Figure 5(a) shows the retained interpore fracture network (blue plane) after deleting the random fractures that intersected the borehole. The final model contained 7 fixed fractures and 36 random fractures.

[0077] Step 5: Iterative analysis of multi-source data

[0078] It is important to note that the semi-deterministic fracture generation method in step 4 can generate any number of fracture network models. Five groups of semi-deterministic fracture networks are selected each time, as shown in Figure 5(b). The theoretical permeability value for that well section is quickly calculated using the fracture network model and compared with the measured value of 6.37Lu. The model with an error of less than 20% is selected as the regional fracture network model for that well section.

[0079] After applying this method and the aforementioned embodiment in a curtain grouting project at a hydropower station, the accuracy of grouting volume prediction exceeded 65%, and the accuracy of identifying abnormally permeable areas was significantly improved. This overcomes the dilemma of traditional construction, which only models the entire system but lacks a model of the fracture network in the local construction area. The correlation coefficient between the permeability distribution calculated by the model and the actual water pressure test results reached 0.91, verifying the reliability of this method.

[0080] The various embodiments in this specification are described in a progressive manner. Similar parts between the various embodiments can be referred to in conjunction with each other. Each embodiment focuses on the differences between the other embodiments. In particular, the system embodiments are generally similar to the method embodiments, so the description is relatively simple. For relevant parts, refer to the description of the method embodiments.

[0081] The foregoing is merely an embodiment of the present application and is not intended to limit the present application. For those skilled in the art, the present application may have various changes and variations. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application should all be included within the scope of the claims of the present application.

Claims

1. A method for generating a bedrock discrete fracture network model based on multi-source drilling data, characterized in that: The following steps are involved: Step 1: Preprocess multi-source data and calculate the DPI (Drilling Process Index). Based on the change of the DPI in the depth direction of the borehole, select the sudden change area, determine the area where the borehole may have cracks, and crop the borehole image at this location; Step 2: drilling crack image recognition, clustering the trace segments identified in the drilling image area into continuous crack traces; Step 3, crack parameter statistics; Step 3.1: Calculate and determine the geometric parameters of the crack; Step 3.2: Statistical rock mass fracture network distribution model and parameters; Step 4: Batch generate semi-deterministic fracture networks; Step 5: Iteratively analyze multi-source data and select the optimal model.

2. The method for generating a bedrock discrete fracture network model based on multi-source drilling data according to claim 1, characterized in that: The step 1 includes: obtaining drilling parameters such as drilling speed, drill bit speed, drilling pressure, etc. during drilling through digital drilling monitoring, and calculating the drilling process index (DPI), as shown in formula (1): Where V is the real-time drilling speed [m / min], V ref is the benchmark drilling speed [m / min]; N is the real-time drill bit speed [rpm], N ref is the rated drilling speed of the drill bit [rpm]; P is the real-time drilling pressure [MPa], P ref is the design maximum pressure [MPa]; the value range of α, β, γ is 0~1.

3. The method for generating a bedrock discrete fracture network model based on multi-source drilling data according to claim 1, characterized in that: The step 2 includes: based on the deep learning algorithm, identifying the cropped defect area image, exporting the defect area into a binary image, and distinguishing between dissolved pores and cracks; extracting the skeleton and contour of the crack traces revealed by the drilling image, and obtaining the location and occurrence geometric characteristics of the drilling cracks through burr removal and segmented clustering methods; the skeleton of the crack trace segments revealed by the drilling image can be described as the parameter set F i =[θ i ,φ i ,d i ], respectively representing the inclination and dip of the crack, and the midpoint depth of the trace segment; all trace segments contained in this image are matched for similarity, and their similarity coefficient can be given by formula (2): Among them, σ θ and σ φ Represent the tolerance difference of inclination and dip, ω θ 、ω φ With ω d Represent the similarity coefficients of dip, inclination and center position respectively; through this step, the trace segments identified in the borehole image area can be clustered into continuous fracture traces.

4. The method for generating a bedrock discrete fracture network model based on multi-source drilling data according to claim 3, characterized in that: The deep learning algorithm in step 2 adopts DeepLabv3+.

5. The method for generating a bedrock discrete fracture network model based on multi-source drilling data according to claim 1, wherein: The step 3 includes: using the morphological characteristics of each fracture trace after clustering in step 2 to obtain the geometric parameters of the fractures revealed by the borehole, and then obtaining the fracture network spatial distribution model and parameters of the formation where the borehole is located through mathematical statistics.

6. The method for generating a bedrock discrete fracture network model based on multi-source drilling data according to claim 5, characterized in that: Step 3.1 includes, first, calculating the occurrence of the fracture surface based on the fracture trace revealed by the borehole image in step 2; assuming that the borehole image is developed along the north direction, the fracture trace can be fitted into a trigonometric function, The inclination of the crack is φ, and its inclination angle, θ, can be obtained from formula (3): Second, the crack trace length is calculated based on the product of the number of pixels in the skeleton and the size of a single pixel; Third, the normal direction of the crack skeleton is defined as the width direction of the crack, based on which the average crack width can be calculated; Fourth, the position and spacing of the traces can be measured by the line survey method. The position is the depth coordinate of the intersection of the trace and the survey line, and the spacing is the distance between two adjacent cracks in the survey line. Fifth, the density of cracks is obtained by calculating the spacing and number of cracks.

7. The method for generating a bedrock discrete fracture network model based on multi-source drilling data according to claim 6, characterized in that: Step 3.2 includes: grouping the borehole fractures according to the occurrence of the fractures obtained in step 3.1, and establishing a distribution histogram between the statistical distribution parameters and the fracture frequency for each group of fractures; fitting the probability distribution form obeyed by the parameters based on the drawn histogram and testing the goodness of fit; in terms of fracture grouping, the occurrence of the fractures is projected stereographically to form a pole diagram, and the pole diagram is clustered using the K-Means clustering algorithm to obtain the dominant group and the occurrence corresponding to the dominant group, as well as the number of fractures in each dominant group; performing parameter statistics on the fractures in each group according to the grouping, and fitting the fracture occurrence distribution using the Fisher distribution model, as shown in formula (4): Where k is the concentration parameter, is the dominant normal vector of the fracture group, is the crack normal vector; the crack spacing and size are fitted using a negative exponential distribution, and the average crack width is fitted using a lognormal distribution; the fitted crack distribution statistical model and its parameters need to be verified using the corresponding fitting test method, the Fisher distribution needs to use a spherical test, and the lognormal distribution needs to perform a QQ test.

8. The method for generating a bedrock discrete fracture network model based on multi-source drilling data according to claim 1, wherein: Step 4 includes generating a determined fracture network in the area to be modeled based on the determined fractures counted in step 3.1, and generating a random fracture network in the inter-pore area using the Monte Carlo method based on the parameter distribution obtained in step 3.

2.

9. The method for generating a bedrock discrete fracture network model based on multi-source drilling data according to claim 8, characterized in that: Step 4 specifically includes: Step 4.1: Generate and determine the fracture network; some fractures in the area to be modeled have been revealed by drilling. According to step 3.1, their locations and occurrences are basically determined. The fracture sizes are determined by the cross-hole fracture clustering algorithm. Step 4.2: Generate a random fracture network between holes; obtain the number of fractures in each group based on the density distribution model and parameters of the grouped fractures obtained in step 3.2, and then generate random numbers for the size, position, occurrence, and width of each fracture based on other parameter distribution models and parameters to establish a random fracture network model; delete the fractures that intersect with the boreholes in the random model; and fuse the deterministic fracture network generated in step 4.1 to form a semi-deterministic fracture network model.

10. The method for generating a bedrock discrete fracture network model based on multi-source drilling data according to claim 1, wherein: Step 5 includes, combining the water permeability of each borehole section during the construction process, using the fractured pipe network method to quickly numerically calculate the model water permeability of the borehole section; comparing the numerical calculation results with the actual water pressure data, and selecting the discrete fracture model with an error of less than 20% with the actual water pressure test data as the preferred model.

Images