Rock sample damage area prediction method and device

Abstract

This invention discloses a method and apparatus for predicting the damage area of ​​rock samples. The method includes: acquiring a grayscale image of a rock sample scanned by a nuclear magnetic resonance imaging (NMR) device; identifying a pixel value matrix representing the pore structure of the rock from the grayscale image; inputting the pixel value matrix into a rock damage area prediction model and outputting a prediction result; the rock damage area prediction model is obtained by training a BP neural network using a historical pixel value matrix as input and historical measured damage area as output; and determining the prediction result as the damage area of ​​the rock sample. This invention can achieve accurate prediction of rock damage area, avoid the secondary damage problem caused by traditional rock sampling, and realize non-destructive testing of the damage area of ​​rock samples.

Description

Technical Field

[0001] This invention relates to the field of magnetic resonance imaging technology, and in particular to a method and apparatus for predicting the damage area of ​​rock samples. Background Technology

[0002] This section is intended to provide background or context for the embodiments of the invention set forth in the claims. The description herein is not an admission that it is prior art simply because it is included in this section.

[0003] With the accelerating pace of industrialization, traditional energy sources such as oil and natural gas are increasingly unable to meet the needs of economic and social development. Geothermal energy has become an important alternative to traditional oil and natural gas. Hot dry rock reservoirs are typically composed of dense granite, buried at depths of approximately 5-6 kilometers, with temperatures ranging from 150 to 600 degrees Celsius. Geothermal resources mainly include three categories: shallow geothermal resources, hydrothermal geothermal resources, and hot dry rocks. Hot dry rocks (HDRs) are particularly favored by researchers due to their abundant reserves.

[0004] Traditional methods for utilizing hot dry rock resources primarily involve establishing injection wells and production wells. Low-temperature fluids are injected through the injection wells, heated by the hot dry rock formation, and then discharged from the production wells. The heat from the fluids is then used for heating, power generation, and other applications. This process repeats continuously as the reservoir temperature recovers. During this process, the granite matrix in the reservoir is affected by cyclical high temperatures and cooling, leading to deterioration of the rock's mechanical and porosity properties. This can cause wellbore instability and collapse, shortening the well's lifespan.

[0005] Regarding granite damage, patent number CN202110755077.X proposes a rock damage analysis method, device, and equipment based on industrial CT scanning. This solves the problem that existing technologies lack methods for rock damage analysis using industrial CT scanning equipment, achieving improved accuracy and applicability in rock damage analysis. Patent number CN202310842001.X proposes a rock damage quantification device and method during alternating hot and cold processes. This method can achieve real-time monitoring of the evolution of the internal structure of rocks through a combination of optical and acoustic methods during alternating hot and cold rock tests, quantifying rock damage. Patent number CN202111400440.2 proposes a statistical damage calculation method for layered rocks under thermo-mechanical coupling conditions. Applied to triaxial compression tests of layered slate under thermo-mechanical coupling, the proposed new model has stronger comprehensiveness and applicability than existing models, and a higher degree of fitting to the stress-strain curves obtained from the tests.

[0006] Current methods for predicting granite damage to geothermal well walls typically involve breaking up the rock, extracting rock samples, and conducting mechanical experiments. This process inevitably causes secondary damage to the well wall. Furthermore, the complex underground geological environment, influenced by high underground confining pressure, makes it difficult to predict the area of ​​granite damage under different geological conditions and the coupled effects of different operating conditions. Summary of the Invention

[0007] This invention provides a method for predicting the damage area of ​​rock samples, used for non-destructive testing of the damage area of ​​rock samples, achieving accurate prediction of the damage area of ​​rock samples. The method includes:

[0008] Acquire grayscale images of rock samples obtained by scanning with a nuclear magnetic resonance imaging device, and identify pixel value matrices characterizing the pore structure of the rock from the grayscale images;

[0009] The pixel value matrix is ​​input into the rock damage area prediction model, and the prediction result is output. The rock damage area prediction model is obtained by training the BP neural network with the historical pixel value matrix as input and the historical measured damage area as output.

[0010] The predicted results were used to determine the damage area of ​​the rock sample.

[0011] This invention also provides a device for predicting the damage area of ​​a rock sample, used for non-destructive testing of the damage area of ​​a rock sample, thereby achieving accurate prediction of the damage area of ​​the rock sample. The device includes:

[0012] The pixel value matrix recognition module is used to acquire grayscale images obtained by scanning rock samples using nuclear magnetic resonance imaging equipment, and to identify pixel value matrices that characterize the pore structure of rocks from the grayscale images; the position of elements in the pixel value matrix maps the distribution of cracks, and the corresponding pixel value in the pixel value matrix maps the degree of crack development;

[0013] The rock damage area prediction model training module is used to input the pixel value matrix into the rock damage area prediction model and output the prediction result. The rock damage area prediction model is obtained by training the BP neural network with the historical pixel value matrix as the input and the historical measured damage area as the output.

[0014] The rock sample damage area prediction module is used to determine the damage area of ​​the rock sample based on the prediction results.

[0015] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described method for predicting the damage area of ​​rock samples.

[0016] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for predicting the damage area of ​​rock samples.

[0017] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described method for predicting the damage area of ​​rock samples.

[0018] In this embodiment of the invention, a grayscale image of a rock sample obtained by scanning with a nuclear magnetic resonance imaging (NMR) device is acquired, and a pixel value matrix representing the pore structure of the rock is identified from the grayscale image. The pixel value matrix is ​​then input into a rock damage area prediction model, which outputs a prediction result. This model is trained using a historical pixel value matrix as input to a backpropagation (BP) neural network and historical measured damage areas as output. The prediction result is then determined as the damage area of ​​the rock sample. In this process, this embodiment of the invention acquires grayscale images through non-destructive scanning with a nuclear magnetic resonance imaging (NMR) device. It combines the dual mapping feature of mapping the position of elements in the pixel value matrix to crack distribution and the pixel value to crack development degree with the training of the BP neural network, achieving accurate prediction of the rock damage area. This avoids the secondary damage problem caused by traditional rock sampling and enables non-destructive detection of the damage area of ​​rock samples. Attached Figure Description

[0019] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort. In the drawings:

[0020] Figure 1 This is a flowchart of the rock sample damage area prediction method in an embodiment of the present invention;

[0021] Figure 2 This is a flowchart illustrating the identification of the pixel value matrix in an embodiment of the present invention;

[0022] Figure 3 This is a schematic diagram of the pore and fracture distribution corresponding to the magnetic resonance imaging of high-temperature-cooled granite in an embodiment of the present invention;

[0023] Figure 4 This is a schematic diagram of pixel value distribution in high-temperature-cooled granite magnetic resonance imaging in an embodiment of the present invention;

[0024] Figure 5This is a structural diagram of the BP neural network in an embodiment of the present invention;

[0025] Figure 6 This is a flowchart of weight adjustment in an embodiment of the present invention;

[0026] Figure 7 This is a schematic diagram of the rock sample damage area prediction device in an embodiment of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings. Here, the illustrative embodiments of the present invention and their descriptions are used to explain the present invention, but are not intended to limit the present invention.

[0028] The acquisition, storage, use, and processing of data in this application all comply with the relevant provisions of national laws and regulations.

[0029] Figure 1 This is a flowchart of a method for predicting the damage area of ​​rock samples in an embodiment of the present invention. The method includes:

[0030] Step 101: Obtain a grayscale image of the rock sample obtained by scanning with a nuclear magnetic resonance imaging device, and identify the pixel value matrix representing the pore structure of the rock from the grayscale image; the position of the element in the pixel value matrix maps the distribution of cracks, and the corresponding pixel value in the pixel value matrix maps the degree of crack development;

[0031] Step 102: Input the pixel value matrix into the rock damage area prediction model and output the prediction result; wherein, the rock damage area prediction model is obtained by training the BP neural network with the historical pixel value matrix as the input and the historical measured damage area as the output.

[0032] Step 103: The prediction result is determined as the damage area of ​​the rock sample.

[0033] Before elaborating on each step, the following preparatory steps are required:

[0034] The first step, using granite as an example, involves processing the granite into cylindrical specimens with a diameter of 50 mm and a height of 100 mm for magnetic resonance imaging (MRI) experiments. Fine sandpaper is used to polish both ends of all specimens to ensure that both ends are smooth and intact, meeting the preset standards.

[0035] The second step involved dividing the samples into three groups of three cylindrical specimens each. The specimens were placed in a heating furnace and heated to 500°C at a rate of 5°C / min, maintaining this temperature for 150 minutes to ensure complete heating. The heated rocks were then removed and subjected to natural cooling, water cooling, and LN2 cooling, respectively, for 120 minutes each. They were then allowed to reach room temperature, completing one thermal cycle. This heating and cooling process was repeated 1, 2, 4, 8, and 12 times. After each target number of heating and cooling cycles, MRI experiments were performed on the granite specimens. A control group of specimens at room temperature was prepared; these specimens underwent no thermal cycling or cooling treatment, only the same MRI experiments.

[0036] Each step is explained in detail below.

[0037] In step 101, a grayscale image of the rock sample is obtained by scanning it with a nuclear magnetic resonance imaging device, and a pixel value matrix representing the pore structure of the rock is identified from the grayscale image; the position of the element in the pixel value matrix maps the distribution of cracks, and the corresponding pixel value in the pixel value matrix maps the degree of crack development.

[0038] Figure 2 This is a flowchart illustrating the process of identifying a pixel value matrix in one embodiment of the present invention. In one embodiment, a grayscale image obtained by scanning a rock sample using a nuclear magnetic resonance imaging (NMR) device is acquired, and a pixel value matrix characterizing the pore structure of the rock is identified from the grayscale image, including:

[0039] Step 201: Obtain the hydrogen proton distribution from the nuclear magnetic resonance imaging device scanning the water-saturated rock sample; the water-saturated rock sample is obtained by vacuum treatment and water saturation treatment of the rock sample, wherein the water-saturated rock sample is filled with hydrogen proton liquid.

[0040] Step 202: Convert the hydrogen proton distribution into a grayscale image and identify the grayscale value distribution from the grayscale image;

[0041] Step 203: Generate a pixel value matrix representing the pore structure of the rock based on the gray value distribution; wherein the pixel value is positively correlated with the hydrogen proton density.

[0042] Figure 3 This is a schematic diagram of the pore and fracture distribution corresponding to the high-temperature-cooled granite magnetic resonance imaging image in an embodiment of the present invention. Figure 4 This is a schematic diagram of pixel value distribution in high-temperature-cooled granite magnetic resonance imaging according to an embodiment of the present invention. In a specific embodiment, nuclear magnetic resonance imaging was performed on granite samples subjected to different thermal cycling conditions to obtain the distribution of hydrogen protons. The imaging signal was visualized as a grayscale image and then converted into pixel values ​​of a pseudo-color image. Considering that the grayscale value depends on the density of hydrogen protons, the larger the pores and cracks, the more water penetrates into the sample, and the higher the density of hydrogen protons. Figure 4The fitted curves show that the stronger the signal, the higher the pixel value. Analyzing the combinations of different colored pixels, color differences represent different proton densities, thus revealing the distribution of pores and fissures, such as... Figures 3-4 As shown. Observe the changes in pixel cluster color; a darker color indicates the aggregation of pores and the development of cracks.

[0043] In step 102, the pixel value matrix is ​​input into the rock damage area prediction model, and the prediction result is output. The rock damage area prediction model is obtained by training the BP neural network with the historical pixel value matrix as input and the historical measured damage area as output.

[0044] In one embodiment, before using the historical pixel value matrix as input to the BP neural network, the following steps are included:

[0045] Historical grayscale images of rock samples under different thermal cycling conditions were obtained by scanning them with nuclear magnetic resonance imaging equipment; thermal cycling conditions included temperature conditions, cooling methods, number of cycles, and confining pressure conditions.

[0046] Identify historical pixel value matrices representing the pore structure of rocks under current thermal cycling conditions from historical grayscale images.

[0047] In a specific embodiment, a BP neural network algorithm is used to train a rock cross-sectional area damage calculation model. The BP neural network is a multi-layer feedforward network that uses the steepest gradient descent method to backpropagate error information to achieve the learning objective. Its learning process includes two stages: forward propagation of input information and backward adjustment of error information. In the forward propagation stage, the input information is transmitted from the input layer through the hidden layers to the output layer. If the expected output is not obtained in the output layer, the output error will be backpropagated layer by layer until it reaches the input layer. Along the way, the connection weights and thresholds between neurons in each layer are modified to gradually minimize the error. The forward propagation of information and the backward adjustment of error are carried out alternately until the network output error is reduced to a preset error range or the preset number of learning iterations is reached.

[0048] Figure 5 This is a structural diagram of a BP neural network in an embodiment of the present invention. The BP neural network includes an input layer, a hidden layer, and an output layer. The input layer contains m neurons, where m is the dimension of the input parameters. The output layer contains 1 neuron, and the hidden layer of the network contains 20 neurons, where x1, x2, ..., x... m y is the input to the neural network, t is the target output of the output neuron, e is the error of the output neuron, i is the number of nodes in the hidden layer, and j is the number of nodes in the output layer.

[0049] Figure 6This is a flowchart of weight adjustment in an embodiment of the present invention. In one embodiment, the historical pixel value matrix is ​​used as the input of the BP neural network, and the historical measured damage area is used as the output of the BP neural network. The BP neural network is trained, including:

[0050] Step 601: The historical pixel value matrix is ​​used as the input of the BP neural network. The information is processed through the hidden layer of the BP neural network, and the training result after information processing is output through the output layer of the BP neural network.

[0051] Step 602: Using the backpropagation mechanism, adjust the weights of the BP neural network based on the error between the training results of the output layer and the historical measured damage area, until the BP neural network reaches the preset conditions.

[0052] Step 603: Based on the adjusted weights, the trained BP neural network is obtained.

[0053] In one embodiment, the historical pixel value matrix is ​​used as the input to the BP neural network, and the historical measured damage area is used as the output of the BP neural network. The BP neural network is trained by including:

[0054] The BP neural network is trained by using the correlation between the pixel value matrix and the damage area as a constraint, the historical pixel value matrix as the input of the BP neural network, and the historical measured damage area as the output of the BP neural network.

[0055] In one embodiment, the BP neural network is trained using the correlation between the pixel value matrix and the damage area as a constraint, the historical pixel value matrix as the input of the BP neural network, and the historical measured damage area as the output of the BP neural network. The training process includes:

[0056] The correlation between the pixel value matrix and the damaged area is as follows:

[0057]

[0058] Where Δ is the absolute value of the damaged area, i is the index of the pixel value matrix, and x i P represents the number of corresponding pixels, T represents the total number of pixels, and P represents the total number of pixels. i Here, m represents the corresponding pixel value, and m is the maximum pixel value. By mapping the crack distribution to the element positions in the pixel value matrix and the crack development degree to the pixel values, and by using the pixel values ​​from magnetic resonance imaging to define the smooth variation range of the damaged and undamaged areas respectively, the damage variation area of ​​the cross-sectional area of ​​the granite sample can be described, thus determining the method for calculating the absolute value of the damage area.

[0059] In step 103, the prediction result is determined as the damage area of ​​the rock sample.

[0060] In a specific embodiment, the trained BP neural network model is deployed in the cloud. The user inputs the actual parameter values ​​on the client and sends the parameter values ​​to the rock cross-sectional area damage calculation model. Finally, the cloud-based rock cross-sectional area damage calculation model obtains the rock cross-sectional area damage calculation result and sends it to the client for the user to receive.

[0061] Downhole nuclear magnetic resonance was used to measure the distribution of pores of different sizes on the wellbore rock. The results were then connected to the surface system for calculation. Combined with the calculation of the wellbore mechanical state, the strength variation law of the wellbore rock under different thermal cycling conditions was analyzed, and possible wellbore instability or collapse was predicted.

[0062] This invention utilizes MRI technology to obtain the distribution of hydrogen protons in rocks, abstracting the rock sample into two parts: a damaged area and an undamaged area. The damaged area includes cracks and solid parts that have lost their load-bearing capacity, while the undamaged area represents the effective load-bearing region. The distribution of cracks is mapped by the position of elements in the pixel value matrix, and the pixel value maps the degree of crack development. The smooth variation range of the damaged and undamaged areas is defined using the pixel value magnitude of MRI, thus illustrating the damage variation area of ​​the granite sample's cross-sectional area and determining the method for calculating the absolute value of the damage area. This rock damage calculation method is input into the system, and MRI equipment is brought downhole into geothermal drilling projects to achieve non-destructive downhole measurement. Connecting to the surface system, calculations are performed to obtain the wellbore rock damage condition, avoiding damage to the wellbore during rock sampling and obtaining first-hand data from the engineering site. This method can evaluate the changes in wellbore rock strength during geothermal well drilling, predict potential wellbore instability or collapse, and provide an effective reference for improving the development efficiency of hot dry rock and controlling wellbore stability.

[0063] This invention also provides a device for predicting the damage area of ​​rock samples, as described in the following embodiments. Since the principle behind this device is similar to that of the rock sample damage area prediction method, its implementation can be found in the implementation of the rock sample damage area prediction method; repeated details will not be elaborated further.

[0064] Figure 7 This is a schematic diagram of a rock sample damage area prediction device according to an embodiment of the present invention. The device includes:

[0065] The pixel value matrix recognition module 701 is used to acquire grayscale images obtained by scanning rock samples using a nuclear magnetic resonance imaging device, and to identify pixel value matrices representing the pore structure of rocks from the grayscale images; the position of elements in the pixel value matrix maps the distribution of cracks, and the corresponding pixel value in the pixel value matrix maps the degree of crack development;

[0066] The rock damage area prediction model training module 702 is used to input the pixel value matrix into the rock damage area prediction model and output the prediction result; wherein, the rock damage area prediction model is obtained by training the BP neural network with the historical pixel value matrix as the input of the BP neural network and the historical measured damage area as the output of the BP neural network.

[0067] The rock sample damage area prediction module 703 is used to determine the prediction result as the damage area of ​​the rock sample.

[0068] In one embodiment, the pixel value matrix recognition module 701 is specifically used for:

[0069] The distribution of hydrogen protons was obtained by scanning a water-saturated rock sample with a nuclear magnetic resonance imaging device. The water-saturated rock sample was obtained by vacuum treatment and water saturation treatment of the rock sample, in which the water-saturated rock sample was filled with hydrogen proton liquid.

[0070] The hydrogen proton distribution is converted into a grayscale image, and the grayscale value distribution is identified from the grayscale image;

[0071] A pixel value matrix characterizing the pore structure of rocks is generated based on the gray value distribution; the pixel value is positively correlated with the hydrogen proton density.

[0072] In one embodiment, the rock damage area prediction model training module 702 is further configured to:

[0073] Historical grayscale images of rock samples under different thermal cycling conditions were obtained by scanning them with nuclear magnetic resonance imaging equipment; thermal cycling conditions included temperature conditions, cooling methods, number of cycles, and confining pressure conditions.

[0074] Identify historical pixel value matrices representing the pore structure of rocks under current thermal cycling conditions from historical grayscale images.

[0075] In one embodiment, the rock damage area prediction model training module 702 is specifically used for:

[0076] The BP neural network is trained by using the correlation between the pixel value matrix and the damage area as a constraint, the historical pixel value matrix as the input of the BP neural network, and the historical measured damage area as the output of the BP neural network.

[0077] In one embodiment, the rock damage area prediction model training module 702 is specifically used for:

[0078] The correlation between the pixel value matrix and the damaged area is as follows:

[0079]

[0080] Where Δ is the absolute value of the damaged area, i is the index of the pixel value matrix, and x i P represents the number of corresponding pixels, T represents the total number of pixels, and P represents the total number of pixels. i is the corresponding pixel value, and m is the maximum pixel value.

[0081] In one embodiment, the rock damage area prediction model training module 702 is specifically used for:

[0082] The historical pixel value matrix is ​​used as the input of the BP neural network. The information is processed by the hidden layer of the BP neural network, and the training result after information processing is output by the output layer of the BP neural network.

[0083] Using the backpropagation mechanism, the weights of the BP neural network are adjusted based on the error between the training results of the output layer and the historical measured damage area until the BP neural network reaches the preset conditions.

[0084] Based on the adjusted weights, the trained BP neural network is obtained.

[0085] This invention also provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, it implements the above-described method for predicting the damage area of ​​rock samples.

[0086] This invention also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for predicting the damage area of ​​rock samples.

[0087] This invention also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described method for predicting the damage area of ​​rock samples.

[0088] In this embodiment of the invention, a grayscale image of a rock sample obtained by scanning with a nuclear magnetic resonance imaging (NMR) device is acquired, and a pixel value matrix representing the pore structure of the rock is identified from the grayscale image. The pixel value matrix is ​​then input into a rock damage area prediction model, which outputs a prediction result. This model is trained using a historical pixel value matrix as input to a backpropagation (BP) neural network and historical measured damage areas as output. The prediction result is then determined as the damage area of ​​the rock sample. In this process, this embodiment of the invention acquires grayscale images through non-destructive scanning with a nuclear magnetic resonance imaging (NMR) device. It combines the dual mapping feature of mapping the position of elements in the pixel value matrix to crack distribution and the pixel value to crack development degree with the training of the BP neural network, achieving accurate prediction of the rock damage area. This avoids the secondary damage problem caused by traditional rock sampling and enables non-destructive detection of the damage area of ​​rock samples.

[0089] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0090] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0091] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0092] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0093] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for predicting the damage area of ​​a rock sample, characterized in that, include: Acquire grayscale images of rock samples obtained by scanning with a nuclear magnetic resonance imaging device, and identify pixel value matrices characterizing the pore structure of the rock from the grayscale images; The position of each element in the pixel value matrix maps to the distribution of cracks, and the corresponding pixel value in the pixel value matrix maps to the degree of crack development. The pixel value matrix is ​​input into the rock damage area prediction model, and the prediction result is output. The rock damage area prediction model is obtained by training the BP neural network with the historical pixel value matrix as input and the historical measured damage area as output. The predicted results were used to determine the damage area of ​​the rock sample.

2. The method as described in claim 1, characterized in that, Acquire grayscale images of rock samples obtained by scanning with a nuclear magnetic resonance imaging (MRI) device, and identify a pixel value matrix characterizing the pore structure of the rock from the grayscale images, including: The distribution of hydrogen protons was obtained by scanning a water-saturated rock sample with a nuclear magnetic resonance imaging device. The water-saturated rock sample was obtained by vacuum treatment and water saturation treatment of the rock sample, in which the water-saturated rock sample was filled with hydrogen proton liquid. The hydrogen proton distribution is converted into a grayscale image, and the grayscale value distribution is identified from the grayscale image; A pixel value matrix characterizing the pore structure of rocks is generated based on the gray value distribution; the pixel value is positively correlated with the hydrogen proton density.

3. The method as described in claim 1, characterized in that, Before using the historical pixel value matrix as input to the BP neural network, the following steps are included: Historical grayscale images of rock samples under different thermal cycling conditions were obtained by scanning them with nuclear magnetic resonance imaging equipment; thermal cycling conditions included temperature conditions, cooling methods, number of cycles, and confining pressure conditions. Identify historical pixel value matrices representing the pore structure of rocks under current thermal cycling conditions from historical grayscale images.

4. The method as described in claim 1, characterized in that, The BP neural network is trained by using the historical pixel value matrix as input and the historical measured damage area as output, including: The BP neural network is trained by using the correlation between the pixel value matrix and the damage area as a constraint, the historical pixel value matrix as the input of the BP neural network, and the historical measured damage area as the output of the BP neural network.

5. The method as described in claim 4, characterized in that, Using the correlation between pixel value matrix and damage area as constraints, the historical pixel value matrix as input to the BP neural network, and the historical measured damage area as output, the BP neural network is trained, including: The correlation between the pixel value matrix and the damaged area is as follows: Where Δ is the absolute value of the damaged area, i is the index of the pixel value matrix, and x i P represents the number of corresponding pixels, T represents the total number of pixels, and P represents the total number of pixels. i is the corresponding pixel value, and m is the maximum pixel value.

6. The method as described in claim 1, characterized in that, The BP neural network is trained by using the historical pixel value matrix as input and the historical measured damage area as output, including: The historical pixel value matrix is ​​used as the input of the BP neural network. The information is processed by the hidden layer of the BP neural network, and the training result after information processing is output by the output layer of the BP neural network. Using the backpropagation mechanism, the weights of the BP neural network are adjusted based on the error between the training results of the output layer and the historical measured damage area until the BP neural network reaches the preset conditions. Based on the adjusted weights, the trained BP neural network is obtained.

7. A device for predicting the damage area of ​​a rock sample, characterized in that, include: The pixel value matrix recognition module is used to acquire grayscale images obtained by scanning rock samples using nuclear magnetic resonance imaging equipment, and to identify pixel value matrices that characterize the pore structure of rocks from the grayscale images; the position of elements in the pixel value matrix maps the distribution of cracks, and the corresponding pixel value in the pixel value matrix maps the degree of crack development; The rock damage area prediction model training module is used to input the pixel value matrix into the rock damage area prediction model and output the prediction result. The rock damage area prediction model is obtained by training the BP neural network with the historical pixel value matrix as the input and the historical measured damage area as the output. The rock sample damage area prediction module is used to determine the damage area of ​​the rock sample based on the prediction results.

8. The apparatus as claimed in claim 7, characterized in that, The pixel value matrix recognition module is specifically used for: The distribution of hydrogen protons was obtained by scanning a water-saturated rock sample with a nuclear magnetic resonance imaging device. The water-saturated rock sample was obtained by vacuum treatment and water saturation treatment of the rock sample, in which the water-saturated rock sample was filled with hydrogen proton liquid. The hydrogen proton distribution is converted into a grayscale image, and the grayscale value distribution is identified from the grayscale image; A pixel value matrix characterizing the pore structure of rocks is generated based on the gray value distribution; the pixel value is positively correlated with the hydrogen proton density.

9. The apparatus as claimed in claim 7, characterized in that, The rock damage area prediction model training module is also used for: Historical grayscale images of rock samples under different thermal cycling conditions were obtained by scanning them with nuclear magnetic resonance imaging equipment; thermal cycling conditions included temperature conditions, cooling methods, number of cycles, and confining pressure conditions. Identify historical pixel value matrices representing the pore structure of rocks under current thermal cycling conditions from historical grayscale images.

10. The apparatus as claimed in claim 7, characterized in that, The rock damage area prediction model training module is specifically used for: The BP neural network is trained by using the correlation between the pixel value matrix and the damage area as a constraint, the historical pixel value matrix as the input of the BP neural network, and the historical measured damage area as the output of the BP neural network.

11. The apparatus as claimed in claim 10, characterized in that, The rock damage area prediction model training module is specifically used for: The correlation between the pixel value matrix and the damaged area is as follows: Where Δ is the absolute value of the damaged area, i is the index of the pixel value matrix, and x i P represents the number of corresponding pixels, T represents the total number of pixels, and P represents the total number of pixels. i is the corresponding pixel value, and m is the maximum pixel value.

12. The apparatus as claimed in claim 7, characterized in that, The rock damage area prediction model training module is specifically used for: The historical pixel value matrix is ​​used as the input of the BP neural network. The information is processed by the hidden layer of the BP neural network, and the training result after information processing is output by the output layer of the BP neural network. Using the backpropagation mechanism, the weights of the BP neural network are adjusted based on the error between the training results of the output layer and the historical measured damage area until the BP neural network reaches the preset conditions. Based on the adjusted weights, the trained BP neural network is obtained.

13. A computer device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the method of any one of claims 1 to 6.

14. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method of any one of claims 1 to 6.

15. A computer program product, characterized in that, The computer program product includes a computer program that, when executed by a processor, implements the method of any one of claims 1 to 6.

Images