A rock fracture information identification method and system based on variational autoencoder

Abstract

The present invention relates to a rock crack information identification method and system based on a variational autoencoder, and belongs to the technical field of petroleum engineering rock mechanics fracturing test. A microresistivity scanning imaging test and a CT scanning test are performed on each of a plurality of rock specimens to obtain a two-dimensional image of the microresistivity distribution for training and a three-dimensional image of the CT scanning for training of the rock specimen, so as to train the initial variational autoencoder and obtain a trained variational autoencoder. In practical applications, a two-dimensional image of the microresistivity distribution of the rock to be tested is first obtained, and a three-dimensional CT scanning image can be reconstructed using the trained variational autoencoder, so that the crack information of the rock to be tested can be accurately determined at a low cost.

Description

Technical Field

[0001] The present invention relates to the technical field of petroleum engineering rock mechanics fracturing tests, and in particular to a rock fracture information recognition method and system based on micro-resistivity scanning imaging technology and deep learning variational autoencoder. Background Art

[0002] China is rich in unconventional oil and gas resources. Vigorously promoting their exploration and development will play a vital role in alleviating the contradiction between energy supply and demand and adjusting the energy structure. However, the physical properties of unconventional oil and gas reservoirs are generally poor (generally permeability is less than 0.1×10 -3 μm 2 , porosity less than 10%), so hydraulic fracturing technology is often used to transform reservoirs to improve oil and gas production efficiency. The quality of fracturing depends largely on the affected area of ​​hydraulic fracturing, that is, the volume of reservoir transformation. Therefore, clarifying the formation mechanism of complex fracture networks during fracturing is of great significance for accurately predicting the volume of reservoir transformation.

[0003] At present, indoor hydraulic fracturing physical model experiments are helpful to understand the laws of crack expansion and the formation mechanism of complex fracture networks. To carry out physical model experiments, it is necessary to monitor the cracks of rock specimens. The main monitoring methods today are acoustic emission technology and CT scanning technology. Both technologies are relatively mature, but they also have their limitations. For example, acoustic emission technology cannot accurately locate cracks inside rocks. Therefore, it is difficult to accurately describe the spatial position, number, direction, length, width and other information of cracks inside rocks. In addition, this technology requires the deployment of a large number of probes to record monitoring data, with a large data capacity and a time-consuming and labor-intensive processing process. Although CT scanning technology can accurately obtain information on cracks inside rocks, the testing cost is expensive.

[0004] Based on this, there is an urgent need for a rock fracture information identification method and system with low cost but the same accuracy as the rock fracture information obtained by CT scanning technology. Summary of the invention

[0005] The purpose of the present invention is to provide a rock fracture information identification method and system based on variational autoencoder, which can accurately obtain rock fracture information with low cost.

[0006] To achieve the above object, the present invention provides the following solutions:

[0007] A rock fracture information recognition method based on a variational autoencoder, the recognition method comprising:

[0008] Obtain multiple rock specimens;

[0009] For each of the rock specimens, a micro-resistivity scanning imaging test is performed on the rock specimen using a micro-resistivity scanning imaging test system to obtain a two-dimensional image of the micro-resistivity distribution of the rock specimen for training; a CT scanning test is performed on the rock specimen to obtain a three-dimensional CT scanning image of the rock specimen for training;

[0010] Taking the training microresistivity distribution two-dimensional images and the training CT scan three-dimensional images of all the rock specimens as input, training the initial variational autoencoder to obtain a trained variational autoencoder;

[0011] The microresistivity scanning imaging test system is used to perform a microresistivity scanning imaging test on the rock to be tested, so as to obtain a two-dimensional image of the microresistivity distribution of the rock to be tested; and the two-dimensional image of the microresistivity distribution is used as input to reconstruct a CT scanning three-dimensional image using the trained variational autoencoder; and the crack information of the rock to be tested is determined according to the CT scanning three-dimensional image; the crack information includes the position, number, direction, length and width of the cracks.

[0012] A rock fracture information recognition system based on a variational autoencoder, the recognition system comprising:

[0013] A training module is used to obtain a plurality of rock specimens; for each of the rock specimens, a microresistivity scanning imaging test is performed on the rock specimen using a microresistivity scanning imaging test system to obtain a two-dimensional image of the microresistivity distribution for training of the rock specimen; a CT scanning test is performed on the rock specimen to obtain a three-dimensional CT scanning image for training of the rock specimen; and an initial variational autoencoder is trained using the two-dimensional images of the microresistivity distribution for training and the three-dimensional CT scanning images for training of all the rock specimens as input to obtain a trained variational autoencoder;

[0014] The identification module is used to use the microresistivity scanning imaging test system to perform a microresistivity scanning imaging test on the rock to be tested to obtain a two-dimensional image of the microresistivity distribution of the rock to be tested; and use the two-dimensional image of the microresistivity distribution as input to reconstruct a CT scanning three-dimensional image using the trained variational autoencoder; and determine the crack information of the rock to be tested according to the CT scanning three-dimensional image; the crack information includes the position, number, direction, length and width of the cracks.

[0015] According to the specific embodiments provided by the present invention, the present invention discloses the following technical effects:

[0016] The present invention is used to provide a rock crack information identification method and system based on a variational autoencoder, and each of a plurality of rock specimens is subjected to a microresistivity scanning imaging test and a CT scanning test to obtain a training microresistivity distribution two-dimensional image and a training CT scanning three-dimensional image of the rock specimen, so as to train the initial variational autoencoder and obtain a trained variational autoencoder. In practical applications, a two-dimensional image of the microresistivity distribution of the rock to be tested is first obtained, and a CT scanning three-dimensional image can be reconstructed using the trained variational autoencoder, so that the crack information of the rock to be tested can be accurately determined at a low cost. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative work.

[0018] Figure 1 A method flow chart of the identification method provided by Embodiment 1 of the present invention;

[0019] Figure 2 The technical route of the identification method provided in Example 1 of the present invention;

[0020] Figure 3 A schematic diagram of the structure of a microresistivity scanning test system provided in Example 1 of the present invention;

[0021] Figure 4 A schematic diagram of the structure of a fixing device provided in Example 1 of the present invention;

[0022] Figure 5 A schematic diagram of the distribution of the electric buckles on the electrode plate provided in Example 1 of the present invention;

[0023] Figure 6 A schematic diagram of the Freeman chain code value provided in Example 1 of the present invention;

[0024] Figure 7 A schematic diagram of the neural network structure of the encoder provided in Example 1 of the present invention;

[0025] Figure 8 A schematic diagram of a decoder neural network structure provided in Example 1 of the present invention;

[0026] Fig. 9 This is a system block diagram of the identification system provided in Example 2 of the present invention.

[0027] Explanation of symbols:

[0028] 1- rock specimen; 2- electrode plate; 3- fixing device; 4- LCR digital bridge tester; 5- electrical imaging processing component. DETAILED DESCRIPTION

[0029] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0030] The purpose of the present invention is to provide a rock crack information identification method and system for obtaining accurate rock crack information, which can accurately characterize the cracks formed in the rock during the fracturing test, has a simple structure, is easy to operate, and has a low implementation cost.

[0031] In order to make the above-mentioned objects, features and advantages of the present invention more obvious and easy to understand, the present invention is further described in detail below with reference to the accompanying drawings and specific embodiments.

[0032] Embodiment 1:

[0033] This embodiment is used to provide a rock fracture information identification method based on variational autoencoder. Figure 1 and Figure 2 As shown, the identification method includes:

[0034] S1: Obtain multiple rock specimens;

[0035] The rock specimen of this embodiment may be a cuboid or a cube, preferably a cube, and the size may be 100 mm×100 mm×100 mm.

[0036] S2: for each of the rock specimens, a microresistivity scanning imaging test is performed on the rock specimen using a microresistivity scanning imaging test system to obtain a two-dimensional image of the microresistivity distribution for training of the rock specimen; a CT scanning test is performed on the rock specimen to obtain a three-dimensional CT scanning image for training of the rock specimen;

[0037] like Figure 3 As shown, the micro resistivity scanning imaging test system used in this embodiment includes a resistivity measuring component and an electrical imaging processing component 5 which are connected in communication. The resistivity measuring component is used to measure the resistivity of two relative surfaces of the rock specimen 1 to obtain resistivity data, and the electrical imaging processing component 5 is used to image the resistivity data to obtain an electrical imaging image, and perform image preprocessing and crack information extraction on the electrical imaging image to obtain crack information.

[0038] Specifically, Figure 4As shown, the resistivity measurement component includes an LCR digital bridge tester 4, a fixing device 3 and two plates 2. The plate 2 is made of insulating material, and its function is to fix the electric buckle and calibrate the position of the electric buckle. A plurality of electric buckles are evenly distributed on the first side of each plate 2, and the electric buckle can be a circular electric buckle with a diameter of 5 mm. As for the design of the plate 2 and the arrangement of the electric buckles, the size of the rock specimen 1 (rock to be tested) and the electric buckle, the insulation problem of the electric buckle and the maximum utilization of the area of ​​the plate 2 should be fully considered when designing the plate 2. Therefore, the first side of the plate 2 can be closely arranged with grooves for placing the electric buckle, that is, a plurality of rows of gapless grooves are arranged on the first side of the plate 2, and a plurality of gapless electric buckles are arranged in each row of grooves, and the electric buckles in two adjacent rows of grooves are staggered. Specifically, as Figure 5 As shown, according to the size of the rock specimen 1 (rock to be tested), the size of the electric buckle and taking into account the insulation problem of the electric buckle, 15 rows of closely spaced grooves are arranged on the first side of each electrode plate 2, and each row of grooves can be arranged with 15 electric buckles. The electric buckles between rows are staggered, which can maximize the utilization of the area of ​​each electrode plate 2 and allow as many electric buckles as possible to be arranged on the first side of each electrode plate 2.

[0039] In the measurement state, that is, when conducting a resistivity test, the specimen is first immersed, and the rock specimen 1 (rock to be tested) is immersed in saturated salt water until it is fully saturated, so that the salt water fully penetrates into the rock specimen 1 (rock to be tested), and the salt water acts as an electrolyte in the resistivity test. Then, the two plates 2 and the rock specimen 1 (rock to be tested) are placed on the fixture 3, and the first side of one plate 2 fits with the first surface of the rock specimen 1 (rock to be tested), and the first side of the other plate 2 fits with the second surface of the rock specimen 1, and the first surface and the second surface are two opposite surfaces of the rock specimen 1. The fixture 3 is used to apply uniform pressure to the two plates 2, so that the electric buckles on the plates 2 fit tightly with the first surface and the second surface, that is, they fit tightly with the two opposite surfaces (opposite surfaces) of the rock specimen 1 (rock to be tested). The LCR digital bridge tester 4 is electrically connected to each electric buckle, and each electric buckle is connected to the LCR digital bridge tester 4 through a wire, ensuring that the LCR digital bridge tester 4 can accurately identify the electrical signal of each electric buckle. The LCR digital bridge tester 4 is used to measure the average resistance of two relative surfaces, and further convert the resistivity of the area where each electric buckle is located, thereby obtaining resistivity data.

[0040] The LCR digital bridge tester 4 is connected to a matching electrical imaging processing component 5, which is used to record the resistivity data of various parts of the rock specimen 1 (rock to be tested) during the test, and then image the resistivity data to obtain an electrical imaging image, and perform image preprocessing and crack information extraction on the electrical imaging image to obtain crack information.

[0041] Specifically, imaging the resistivity data to obtain an electrical imaging image may include: color calibration of the resistivity data to obtain a spatial color value, and then arranging the spatial color value according to the position of the electric buckle to obtain the electrical imaging image. After the resistivity data is color calibrated, a chromatogram image is generated by mapping, thereby obtaining a color matrix of each part of the rock specimen 1 (the rock to be tested), and then according to the position of the electric buckle on each electrode plate 2, the color matrix is ​​drawn on the computer screen according to the azimuth position, and finally a visually intuitive electrical imaging image can be obtained.

[0042] The color calibration process includes: determining the maximum and minimum values ​​of the measured resistivity data, which are denoted as E max and E min ; Determine the maximum and minimum values ​​of the RGB color space, denoted by K max and K min , the RGB color space and resistivity data are transferred through two intermediate parameters M and E offset Establish a connection, that is, determine K j =E i ×M+E offset The intermediate transfer parameters M and E in this formula offset The value of E min The above formula can be transformed into K min , so that E max The above formula can be transformed into K max , K j is the spatial color value corresponding to each resistivity value; E i The resistivity value measured for each electric buckle is then converted into the color value corresponding to the color space using the above formula.

[0043] Image preprocessing and crack information extraction are performed on the electrical imaging image, and the crack information obtained may include:

[0044] (1) graying, image enhancement, image denoising, image segmentation and image filtering are performed on the electro-imaging image to obtain a filtered binary image; image thinning is performed on the filtered binary image to obtain a pre-processed image;

[0045] Among them, grayscale includes the following steps: ① Image grayscale: Perform HSV color change on the electrographic image, that is, convert the electrographic image to the HSV color space, and then extract the image of the brightness channel (i.e., V channel) as the grayscale image. ② Grayscale transformation: Use the grayscale value linear transformation method to stretch the existing grayscale value in the grayscale image to the entire grayscale level of 0 to 255 to obtain the grayscale image.

[0046] The formula used for linear transformation of gray value is:

[0047]

[0048] Among them, g(x,y) is the grayscale value of the grayscale image at the coordinate (x,y) after linear transformation; f(x,y) max is the maximum grayscale value of the grayscale image before linear transformation; f(x,y) min is the minimum grayscale value of the grayscale image before linear transformation; f(x,y) is the grayscale value of the grayscale image at coordinate (x,y) before linear transformation.

[0049] Among them, image enhancement may include: using a histogram equalization enhancement method to evenly distribute the grayscale of the grayscale image throughout the entire grayscale range to obtain an enhanced image, thereby improving the contrast of the image and making the crack information more prominent, showing more details compared to the original linear scale image, and making the entire image softer.

[0050] Among them, image denoising may include: considering that crack information needs to be extracted when the image is segmented later, the image edge information needs to be retained as much as possible during the noise removal process, so the median filtering method is used to reduce the noise of the enhanced image to obtain a denoised image.

[0051] The image segmentation may include: using the Otsu optimal global threshold method to perform segmentation processing on the denoised image, removing the image background information, highlighting the crack information, and obtaining a preliminary binary image.

[0052] The image filtering may include: for the preliminary binary image, using a small area removal algorithm to remove noise points in the background, and using a small hole removal algorithm to remove noise points inside the crack, to obtain a filtered binary image.

[0053] Wherein, image thinning may include: morphological dilation, skeleton extraction and skeleton burr removal. Then, image thinning of the filtered binary image to obtain the preprocessed image may include: firstly, performing morphological dilation processing on the filtered binary image to obtain a dilated image; then, using the Zhang-Suen algorithm (also known as the Zhang-fast parallel algorithm) to extract the crack skeleton in the dilated image to obtain a crack skeleton map; finally, removing the burrs on the crack skeleton in the crack skeleton map to obtain the preprocessed image.

[0054] The specific implementation steps for crack skeleton burr removal can include:

[0055] ① Count the information of nodes and endpoints in the crack skeleton map: Use a sliding window of size 3×3 to traverse each pixel in the crack skeleton map, take the current pixel as the center point of the sliding window, and count the total number of pixels in the sliding window. If the total number is 2, the current pixel is regarded as an endpoint; if the total number is 3, the current pixel is regarded as an ordinary point that is neither an endpoint nor a node; if the total number is greater than or equal to 4, the current pixel is regarded as a node;

[0056] ② Starting from each endpoint, calculate the length to each node to get the length of the branch chain;

[0057] ③ Judge the length of the branch chain, and remove the branch chain that is too short as a burr. If the length of the branch chain is less than a preset length, it is considered that the length of the branch chain is too short.

[0058] (2) Determine the location, number and area of ​​the cracks based on the filtered binary image; determine the direction and length of the cracks based on the preprocessed image; and calculate the width of the cracks based on the area and length to obtain crack information.

[0059] Wherein, determining the position, number and area of ​​cracks according to the filtered binary image may include: the position of the cracks may be directly observed from the filtered binary image, the number and area of ​​the cracks may be obtained by calculating the connected domain of the filtered binary image, the number of cracks may be processed by the filtered binary image using a connected domain algorithm based on the 8-connected mode, the number of connected domains in the image may be counted using a statistical method, and the number of connected domains obtained by counting is the number of cracks; the area of ​​the cracks may be obtained by calculating the area of ​​each connected domain. The specific calculation method is: firstly, the filtered binary image is processed using a connected domain algorithm (cv2.connected Components), wherein the search method is an 8-connected search, different connected domains are marked with different numbers from small to large (starting from 1), the largest number obtained by marking is the number of cracks, and the total number of pixels corresponding to each number is the area of ​​each crack.

[0060] Among them, determining the direction and length of the crack based on the preprocessed image can include: for each crack in the preprocessed image, performing Freeman eight-direction chain code writing on the crack to obtain the Freeman chain code value of each pixel point included in the crack; determining the direction and length of the crack based on the Freeman chain code value of each pixel point, that is, obtaining the direction and length information of each crack by processing the Freeman chain code value.

[0061] When writing the Freeman eight-direction chain code, an eight-direction search method is used. Specifically, a sliding window range of 3×3 pixels centered on the selected point is used for searching. The search range is the 8 neighborhood pixels around the selected point to obtain the Freeman chain code value of each pixel.

[0062] Specifically, the crack direction is obtained as follows: the crack direction information is obtained by the Freeman chain code value, wherein when the Freeman chain code value is an odd number, it represents that the direction is oblique 45°, and when the Freeman chain code value is an even number, it represents that the direction is horizontal or vertical. Figure 6 As shown in the figure, the chain code values ​​1, 2, 3, 4, 5, 6, 7, and 0 represent different directions, that is, the eight directions of the crack direction (0°, 45°, 90°, 135°, 180°, 225°, 270°, 315°, and 360°) can be obtained through the Freeman chain code value. The actual crack direction will produce an error with the calculated value. For example, when the crack direction is between 0 and 45°, the maximum deviation is 22.5°. When the actual width of the crack is 100, the calculated crack width is 100 / cos(22.5°)=108.24, which is very close to the actual width of the crack, so this deviation can be ignored.

[0063] The crack length is obtained as follows: For each crack, the crack length is calculated as follows:

[0064] L=W o N o +W e N e ;

[0065] Where, L is the length of the crack in rock specimen 1 (the rock to be tested); W o is the weight when the Freeman chain code value is an even number; N o is the number of even-numbered Freeman chain code values ​​in the crack; W e is the weight of the Freeman chain code when the value is an odd number; N e is the number of odd Freeman chain code values ​​in the crack.

[0066] The above calculation formula can be used to calculate the length of each crack. The cracks produced by hydraulic fracturing are relatively tortuous. The horizontal and vertical directions (i.e., the chain code value is an even number) and the 45-degree direction (i.e., the chain code value is an odd number) of each crack can be calculated separately, and the sum of the two can be added. Each crack contains many 3×3 pixel areas, and the length of each crack is the sum of the lengths of the cracks in these areas.

[0067] After obtaining the length and area of ​​the crack, the calculation formula for the crack width is:

[0068]

[0069] Among them, b a is the width of the crack; S is the area of ​​the crack; L is the length of the crack.

[0070] After the image preprocessing process of grayscale, image enhancement, image denoising, image segmentation, image filtering and image thinning, a preprocessed image can be obtained, and crack parameters can be extracted in the above preprocessing process to finally determine the number of cracks and the position, direction, length and width of each crack to obtain crack information, which is presented in the form of a cross-sectional two-dimensional image.

[0071] Based on the micro-resistivity scanning imaging test system introduced above, a micro-resistivity scanning imaging test is performed on a rock specimen 1 using the micro-resistivity scanning imaging test system to obtain a two-dimensional image of the micro-resistivity distribution for training of the rock specimen 1, which may include:

[0072] For any two opposite surfaces of rock specimen 1, perform the following operations:

[0073] (1) Before the fracturing test is performed on the rock specimen 1, the rock specimen 1 is first immersed in saturated salt water until it is fully saturated, so that the salt water fully penetrates into the rock. The salt water acts as an electrolyte in the resistivity test. Then, the resistivity of the two relative surfaces is measured using a resistivity measuring component to obtain resistivity data before fracturing. Finally, the resistivity data before fracturing is processed using an electrical imaging processing component 5 to obtain crack information before fracturing.

[0074] (2) After the rock specimen 1 is subjected to the fracturing test, the rock specimen 1 is first immersed in saturated salt water until it is fully saturated, so that the salt water fully penetrates into the rock, and the salt water acts as an electrolyte in the resistivity test. Then, the resistivity of the two relative surfaces is measured using a resistivity measuring component to obtain resistivity data after fracturing. Finally, the resistivity data after fracturing is processed using an electrical imaging processing component 5 to obtain crack information after fracturing.

[0075] (3) Subtract the fracture information before fracturing from the fracture information after fracturing to obtain the fracture information corresponding to the two relative surfaces. The fracture information corresponding to all two relative surfaces of the rock specimen 1 constitutes a two-dimensional image of micro-resistivity distribution for training.

[0076] Taking a cubic rock specimen 1 as an example, this embodiment requires that the rock specimen 1 be tested twice, before and after the fracturing test, and each test requires that three sets of opposite surfaces (two relative surfaces) of the rock specimen 1 be tested. By comparing the two-dimensional images of the microresistivity cross section of the rock specimen 1 before and after the fracturing test, the two-dimensional image crack information formed in the rock specimen 1 during the fracturing test can be obtained. The crack information of each set of opposite surfaces can form a two-dimensional image, and all two-dimensional images constitute a two-dimensional image of microresistivity distribution for training.

[0077] The micro-resistivity scanning imaging test system is used to test and process the rock specimen 1 before and after the fracturing test, respectively, to obtain the crack information of the rock specimen 1 before and after the fracturing test. After comparative analysis, the crack information of the rock specimen 1 after the fracturing test is subtracted from the crack information in the rock before the fracturing test, so as to obtain the crack information generated in the rock specimen 1 during the fracturing test, that is, the two-dimensional image of the micro-resistivity distribution for training.

[0078] In S2, a CT scanning test is performed on the rock specimen 1 to obtain a training CT scanning three-dimensional image of the rock specimen 1, which may include: the CT scanning test can obtain a CT scanning three-dimensional image of the rock specimen 1, the CT scanning test is performed twice, and the CT scanning three-dimensional images of the rock specimen 1 before and after the fracturing test are obtained respectively, and by comparing the information of the two CT scanning three-dimensional images obtained, the accurate crack information generated by the rock specimen 1 during the fracturing test can be obtained, that is, the training CT scanning three-dimensional image is obtained, and the accurate crack information inside the rock specimen 1 (including the position, number, direction, length and width of the cracks) is presented in the form of a three-dimensional image.

[0079] S3: using the training microresistivity distribution two-dimensional images and the training CT scan three-dimensional images of all the rock specimens as input, training the initial variational autoencoder to obtain a trained variational autoencoder;

[0080] like Figure 7 and Figure 8 As shown, the deep learning variational autoencoder is used to construct the mapping relationship between the microresistivity distribution two-dimensional image and the CT scan three-dimensional image. It can be divided into two parts: an encoder neural network structure and a decoder neural network structure. The encoder neural network structure can process the CT scan three-dimensional image to obtain the microresistivity distribution two-dimensional image, and the decoder neural network structure can process the microresistivity distribution two-dimensional image to obtain the CT scan three-dimensional image.

[0081] The variational autoencoder (VAE) based on the deep learning algorithm used in this embodiment essentially constructs a mapping relationship between a two-dimensional image of microresistivity distribution and a three-dimensional image of CT scanning (spatial distribution of rock mass fractures). The variational autoencoder is a generator model based on representation learning. The difference between it and the traditional autoencoder is that the former adds a probability distribution to the latent space, and then learns this distribution so that the output distribution of the decoder is consistent with the observed data.

[0082] During training, the CT scanned 3D image is imported into the encoder neural network structure to obtain a 2D image of the microresistivity distribution, and the 2D image of the microresistivity distribution is introduced into the decoder neural network structure to reconstruct the CT scanned 3D image. The variational autoencoder adds the probability distribution describing the crack information to the latent space, and then learns this distribution so that the output distribution of the decoder is consistent with the observed data. In particular, the latent space output vector is sampled from the distribution learned by the encoder. The input data of the variational autoencoder is sampled from the distribution representing the real crack information (specifically, 2D microresistivity imaging). The encoder and decoder are jointly trained by minimizing the reconstruction error. The measure of the reconstruction error is the KL divergence between the parameter posterior distribution and the true posterior distribution.

[0083] S4: using the microresistivity scanning imaging test system to perform a microresistivity scanning imaging test on the rock to be tested, and obtaining a two-dimensional image of the microresistivity distribution of the rock to be tested; and using the two-dimensional image of the microresistivity distribution as input, and reconstructing a CT scan three-dimensional image using the trained variational autoencoder; determining the crack information of the rock to be tested according to the CT scan three-dimensional image; the crack information includes the position, number, direction, length and width of the cracks.

[0084] In S4, a microresistivity scanning imaging test is performed on the rock to be tested using the microresistivity scanning imaging test system to obtain a two-dimensional image of the microresistivity distribution of the rock to be tested. This is the same process as that in S2, a microresistivity scanning imaging test is performed on the rock specimen 1 using the microresistivity scanning imaging test system to obtain a two-dimensional image of the microresistivity distribution for training of the rock specimen 1. It is only necessary to replace the rock specimen 1 in S2 with the rock to be tested, and no further details are given here.

[0085] The present embodiment provides a method for identifying accurate rock crack information based on microresistivity scanning imaging technology and deep learning variational autoencoder. The method mainly obtains a two-dimensional image of microresistivity distribution and a three-dimensional CT scan image of a rock specimen 1 through a microresistivity scanning imaging test and a CT scanning test. The mapping relationship between the two-dimensional image of microresistivity distribution and the three-dimensional CT scan image is constructed through a deep learning variational autoencoder, which is used to detect the crack information generated by the rock after the fracturing test. A new method for accurately characterizing crack information is provided for rock fracturing tests. Compared with acoustic emission technology, the method is more accurate in crack identification and can accurately describe the spatial position, quantity, direction, length and width information of the cracks. Compared with CT scanning technology, the method has a small budget and is relatively simple to operate.

[0086] Embodiment 2:

[0087] This embodiment is used to provide a rock fracture information identification system based on variational autoencoder. Fig. 9 As shown, the identification system comprises:

[0088] The training module M1 is used to obtain a plurality of rock specimens; for each of the rock specimens, a microresistivity scanning imaging test is performed on the rock specimen using a microresistivity scanning imaging test system to obtain a two-dimensional image of the microresistivity distribution for training of the rock specimen; a CT scanning test is performed on the rock specimen to obtain a three-dimensional CT scanning image for training of the rock specimen; the initial variational autoencoder is trained using the two-dimensional images of the microresistivity distribution for training and the three-dimensional CT scanning images for training of all the rock specimens as input to obtain a trained variational autoencoder;

[0089] The identification module M2 is used to perform a microresistivity scanning imaging test on the rock to be tested using the microresistivity scanning imaging test system to obtain a two-dimensional image of the microresistivity distribution of the rock to be tested; and to reconstruct a CT scanning three-dimensional image using the trained variational autoencoder with the two-dimensional image of the microresistivity distribution as input; and to determine the crack information of the rock to be tested according to the CT scanning three-dimensional image; the crack information includes the position, quantity, direction, length and width of the cracks.

[0090] Each embodiment in this specification focuses on the differences from other embodiments, and the same or similar parts between the embodiments can be referred to each other. For the system disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0091] The principles and implementation methods of the present invention are described in this article using specific examples. The description of the above embodiments is only used to help understand the method and core idea of ​​the present invention. At the same time, for those skilled in the art, according to the idea of ​​the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as limiting the present invention.

Claims

1. A rock fracture information identification method based on variational autoencoder, characterized in that: The identification method comprises: Obtain multiple rock specimens; For each of the rock specimens, a micro-resistivity scanning imaging test is performed on the rock specimen using a micro-resistivity scanning imaging test system to obtain a two-dimensional image of the micro-resistivity distribution of the rock specimen for training; a CT scanning test is performed on the rock specimen to obtain a three-dimensional CT scanning image of the rock specimen for training; Taking the training microresistivity distribution two-dimensional images and the training CT scan three-dimensional images of all the rock specimens as input, training the initial variational autoencoder to obtain a trained variational autoencoder; The microresistivity scanning imaging test system is used to perform a microresistivity scanning imaging test on the rock to be tested, so as to obtain a two-dimensional image of the microresistivity distribution of the rock to be tested; the two-dimensional image of the microresistivity distribution is used as input, and a CT scanning three-dimensional image is reconstructed using the trained variational autoencoder; the crack information of the rock to be tested is determined according to the CT scanning three-dimensional image; the crack information includes the position, number, direction, length and width of the crack; The microresistivity scanning imaging test system comprises a resistivity measuring component and an electrical imaging processing component which are communicatively connected; The resistivity measuring component is used to measure the resistivity of two relative surfaces of the rock specimen to obtain resistivity data; the electrical imaging processing component is used to image the resistivity data to obtain an electrical imaging image, and perform image preprocessing and crack information extraction on the electrical imaging image to obtain crack information; The performing image preprocessing and crack information extraction on the electrical imaging image to obtain crack information specifically includes: Graying, image enhancement, image denoising, image segmentation and image filtering are performed on the electro-imaging image to obtain a filtered binary image; image thinning is performed on the filtered binary image to obtain a pre-processed image; Determine the position, number and area of ​​the cracks according to the filtered binary image; determine the direction and length of the cracks according to the preprocessed image; and calculate the width of the cracks according to the area and the length to obtain crack information; The method of performing a micro-resistivity scanning imaging test on the rock specimen using a micro-resistivity scanning imaging test system to obtain a two-dimensional image of the micro-resistivity distribution of the rock specimen for training specifically includes: For any two opposite surfaces of the rock specimen, perform the following operations: Before conducting a fracturing test on the rock specimen, the rock specimen is immersed in saturated salt water until it is fully saturated, and the resistivity measuring component is used to measure the resistivity of the two relative surfaces to obtain resistivity data before fracturing; the electrical imaging processing component is used to process the resistivity data before fracturing to obtain crack information before fracturing; After the rock specimen is subjected to a fracturing test, the rock specimen is immersed in saturated salt water until it is fully saturated, and the resistivity of the two relative surfaces is measured by the resistivity measuring component to obtain resistivity data after fracturing; the resistivity data after fracturing is processed by the electrical imaging processing component to obtain crack information after fracturing; Subtracting the pre-fracturing fracture information from the post-fracturing fracture information to obtain fracture information corresponding to the two relative surfaces; all fracture information corresponding to the two relative surfaces of the rock specimen constitute the training microresistivity distribution two-dimensional image; The initial variational autoencoder adopts a deep learning variational autoencoder, which is used to construct a mapping relationship between a two-dimensional image of microresistivity distribution and a three-dimensional image scanned by CT, and includes an encoder and a decoder. The encoder processes the three-dimensional image scanned by CT to obtain a two-dimensional image of microresistivity distribution, and the decoder processes the two-dimensional image of microresistivity distribution to obtain a three-dimensional image scanned by CT; The deep learning variational autoencoder is a generator model based on representation learning. During training, the CT scan three-dimensional image is imported into the encoder to obtain a two-dimensional image of microresistivity distribution, and the two-dimensional image of microresistivity distribution is imported into the decoder to reconstruct the CT scan three-dimensional image. The probability distribution describing the crack information is added to the latent space. This distribution is learned so that the output distribution of the decoder is consistent with the observed data. The output vector of the latent space is sampled from the distribution learned by the encoder. The encoder and decoder are jointly trained by minimizing the reconstruction error. The measure of the reconstruction error is the KL divergence between the parameter posterior distribution and the true posterior distribution.

2. The identification method according to claim 1, characterized in that: The resistivity measuring component comprises an LCR digital bridge tester, a fixing device and two plates; a plurality of electric buckles are arranged on the first side of each plate; In the measuring state, the first side surface of one electrode plate is in contact with the first surface of the rock specimen, and the first side surface of the other electrode plate is in contact with the second surface of the rock specimen; the first surface and the second surface are two opposite surfaces of the rock specimen; the fixing device is used to apply pressure to the two electrode plates so that the electric buckles are tightly fitted with the first surface and the second surface; the LCR digital bridge tester is electrically connected to each of the electric buckles, and the LCR digital bridge tester is used to measure and obtain resistivity data.

3. The identification method according to claim 2, characterized in that: A plurality of rows of gapless grooves are arranged on the first side surface of the electrode plate, and a plurality of gapless electric buckles are arranged in each row of the grooves; the electric buckles in two adjacent rows of the grooves are arranged alternately.

4. The identification method according to claim 2, characterized in that: The step of imaging the resistivity data to obtain an electrical imaging image specifically includes: performing color calibration on the resistivity data to obtain spatial color values; and arranging the spatial color values ​​according to the positions of the electrical buckles to obtain an electrical imaging image.

5. The identification method according to claim 1, characterized in that: The performing image thinning on the filtered binary image to obtain the preprocessed image specifically comprises: Performing morphological dilation processing on the filtered binary image to obtain a dilated image; The crack skeleton in the expanded image is extracted using the Zhang-Suen algorithm, and the burrs on the crack skeleton are removed to obtain a preprocessed image.

6. The identification method according to claim 1, characterized in that: Determining the direction and length of the crack according to the preprocessed image specifically includes: For each crack in the preprocessed image, Freeman eight-direction chain code is performed on the crack to obtain the Freeman chain code value of each pixel point included in the crack; the direction and length of the crack are determined according to the Freeman chain code value of each pixel point.

7. A rock fracture information identification system based on variational autoencoder, characterized in that: The identification system comprises: A training module is used to obtain a plurality of rock specimens; for each of the rock specimens, a microresistivity scanning imaging test is performed on the rock specimen using a microresistivity scanning imaging test system to obtain a two-dimensional image of the microresistivity distribution for training of the rock specimen; a CT scanning test is performed on the rock specimen to obtain a three-dimensional CT scanning image for training of the rock specimen; and an initial variational autoencoder is trained using the two-dimensional images of the microresistivity distribution for training and the three-dimensional CT scanning images for training of all the rock specimens as input to obtain a trained variational autoencoder; The identification module is used to perform a microresistivity scanning imaging test on the rock to be tested by using the microresistivity scanning imaging test system to obtain a two-dimensional image of the microresistivity distribution of the rock to be tested; and to reconstruct a CT scanning three-dimensional image using the trained variational autoencoder with the two-dimensional image of the microresistivity distribution as input; and to determine the crack information of the rock to be tested according to the CT scanning three-dimensional image; the crack information includes the position, quantity, direction, length and width of the crack; The microresistivity scanning imaging test system comprises a resistivity measuring component and an electrical imaging processing component which are communicatively connected; The resistivity measuring component is used to measure the resistivity of two relative surfaces of the rock specimen to obtain resistivity data; the electrical imaging processing component is used to image the resistivity data to obtain an electrical imaging image, and perform image preprocessing and crack information extraction on the electrical imaging image to obtain crack information; The performing image preprocessing and crack information extraction on the electrical imaging image to obtain crack information specifically includes: Graying, image enhancement, image denoising, image segmentation and image filtering are performed on the electro-imaging image to obtain a filtered binary image; image thinning is performed on the filtered binary image to obtain a pre-processed image; Determine the position, number and area of ​​the cracks according to the filtered binary image; determine the direction and length of the cracks according to the preprocessed image; and calculate the width of the cracks according to the area and the length to obtain crack information; The method of performing a micro-resistivity scanning imaging test on the rock specimen using a micro-resistivity scanning imaging test system to obtain a two-dimensional image of the micro-resistivity distribution of the rock specimen for training specifically includes: For any two opposite surfaces of the rock specimen, perform the following operations: Before conducting a fracturing test on the rock specimen, the rock specimen is immersed in saturated salt water until it is fully saturated, and the resistivity measuring component is used to measure the resistivity of the two relative surfaces to obtain resistivity data before fracturing; the electrical imaging processing component is used to process the resistivity data before fracturing to obtain crack information before fracturing; After the rock specimen is subjected to a fracturing test, the rock specimen is immersed in saturated salt water until it is fully saturated, and the resistivity of the two relative surfaces is measured by the resistivity measuring component to obtain resistivity data after fracturing; the resistivity data after fracturing is processed by the electrical imaging processing component to obtain crack information after fracturing; Subtracting the pre-fracturing fracture information from the post-fracturing fracture information to obtain fracture information corresponding to the two relative surfaces; all fracture information corresponding to the two relative surfaces of the rock specimen constitute the training microresistivity distribution two-dimensional image; The initial variational autoencoder adopts a deep learning variational autoencoder, which is used to construct a mapping relationship between a two-dimensional image of microresistivity distribution and a three-dimensional image scanned by CT, and includes an encoder and a decoder. The encoder processes the three-dimensional image scanned by CT to obtain a two-dimensional image of microresistivity distribution, and the decoder processes the two-dimensional image of microresistivity distribution to obtain a three-dimensional image scanned by CT; The deep learning variational autoencoder is a generator model based on representation learning. During training, the CT scan three-dimensional image is imported into the encoder to obtain a two-dimensional image of microresistivity distribution, and the two-dimensional image of microresistivity distribution is imported into the decoder to reconstruct the CT scan three-dimensional image. The probability distribution describing the crack information is added to the latent space. This distribution is learned so that the output distribution of the decoder is consistent with the observed data. The output vector of the latent space is sampled from the distribution learned by the encoder. The encoder and decoder are jointly trained by minimizing the reconstruction error. The measure of the reconstruction error is the KL divergence between the parameter posterior distribution and the true posterior distribution.

Images