Weathered sandstone micro-image generation method based on conditional generative adversarial network

Abstract

This invention relates to the fields of digital core modeling and artificial intelligence modeling, and particularly to a method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks (GANs). The method includes: acquiring and analyzing sandstone samples with different degrees of weathering to construct a multi-scale database of weathered sandstone; extracting weathering geological knowledge from the multi-scale database to construct a sandstone feature tag set; training the GAN using the sandstone feature tag set to obtain a microscopic image generation model for weathered sandstone; and then using the microscopic image generation model to generate physically consistent microscopic structural images of weathered sandstone. This solves the problems of existing models for modeling the physical and mechanical properties of weathered sandstone, which struggle to reveal the deep correlation between microscopic structure and macroscopic mechanical properties, and the difficulty of effectively integrating the advantages of weathering geological mechanisms and generative modeling in traditional methods, leading to a lack of physical consistency and reasonable weathering response in the reconstruction results.

Description

Technical Field

[0001] This invention relates to the fields of digital core modeling and artificial intelligence modeling, and in particular to a method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks. Background Technology

[0002] Sandstone is widely distributed in the natural environment, and its weathering behavior has a significant impact on geological engineering, geological hazard prediction, and rock mass stability research. During weathering, the mineral composition and microstructure (such as pores and fractures) of sandstone change significantly over time, affecting its physical and mechanical properties. Traditional weathering research methods, such as experimental observation and numerical simulation, suffer from problems such as long data acquisition cycles, high costs, and weak responsiveness to complex environments, making it difficult to reflect the continuous evolution of its mineral composition and microstructure.

[0003] In recent years, Generative Adversarial Networks (GANs) have demonstrated outstanding performance in the image processing domain, compensating for the limitations of limited experimental samples. However, general GAN ​​models have certain limitations in controlling complex conditions (such as the degree of rock weathering) and cannot yet achieve image generation based on control conditions such as the degree of weathering. Therefore, there is an urgent need to develop a GAN model that incorporates the degree of weathering to solve the problem of predicting and reconstructing the evolution of microscopic images of weathered rocks. Summary of the Invention

[0004] This invention provides a method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks, in order to solve the problems of the lack of high-precision, multi-scale, and physically meaningful weathered sandstone databases in existing physical and mechanical property modeling of weathered sandstone, the difficulty in revealing the deep correlation between microstructure and macroscopic mechanical properties, and the difficulty of effectively integrating the advantages of weathering geological mechanisms and generative modeling in traditional modeling methods, resulting in a lack of physical consistency and reasonableness of weathering response in reconstruction results.

[0005] A first aspect of the present invention provides a method for generating microscopic images of weathered sandstone based on a conditional generative adversarial network, comprising the following steps: acquiring and analyzing sandstone samples with different degrees of weathering to construct a multi-scale database of weathered sandstone; extracting weathering geological knowledge from the multi-scale database of weathered sandstone to construct a sandstone feature tag set; training a pre-constructed conditional generative adversarial network using the sandstone feature tag set to obtain a microscopic image generation model of weathered sandstone; and inputting the sandstone feature tags to be predicted into the microscopic image generation model of weathered sandstone to generate a microscopic structure image of weathered sandstone.

[0006] Optionally, the acquisition and analysis of sandstone samples with different degrees of weathering to construct a multi-scale database of weathered sandstone includes:

[0007] Collect sandstone samples with different degrees of weathering; obtain the microscopic parameters of the sandstone samples with different degrees of weathering using a preset diffraction device; obtain the macroscopic physical and mechanical parameters of the sandstone samples with different degrees of weathering through macroscopic physical experiments; process the microscopic parameters and the macroscopic physical and mechanical parameters to obtain a multi-scale database of weathered sandstone with weathering stage labels and structural parameter correspondence.

[0008] Optionally, the step of extracting weathering geological knowledge from the multi-scale database of weathered sandstone to construct a sandstone feature tag set includes:

[0009] Mineral composition data, weathering product data, and microstructure data are extracted from the multi-scale database of weathered sandstone. Quantitative analysis is performed on the mineral composition data to obtain corresponding volume fractions and mass fractions, and mineral composition labels are set for sandstone samples with different weathering degrees based on these volume and mass fractions. Weathering degree labels are set for sandstone samples with different weathering degrees based on a preset weathering grade standard and the weathering product data. Microstructure features are extracted from the microstructure data, and microstructure feature labels are set for sandstone samples with different weathering degrees based on these microstructure features. The mineral crystal orientation distribution in the microstructure data is analyzed, and crystal orientation labels are set for sandstone samples with different weathering degrees based on the mineral crystal orientation distribution. The similarity of local details of sandstone mineral crystals in the microstructure data is calculated to obtain the corresponding structural similarity, and structural similarity labels are set for sandstone samples with different weathering degrees based on the structural similarity. The mineral composition labels, weathering degree labels, microstructure feature labels, crystal orientation labels, and structural similarity labels are preprocessed to construct a sandstone feature label set with real physical properties.

[0010] Optionally, training a pre-constructed conditional generative adversarial network using the sandstone feature label set to obtain a microscopic image generation model for weathered sandstone includes:

[0011] Based on the Adam optimizer, the sandstone feature label set and its corresponding sandstone sample images are input into a pre-constructed conditional generative adversarial network (GAN) to iteratively update the parameters in the GAN until the preset comprehensive loss function converges, thereby obtaining the microscopic image generation model of the weathered sandstone. The preset comprehensive loss function includes an adversarial loss function, a mineral composition loss function, a orientation loss function, a weathering product and porosity ratio loss function, and a structural similarity loss function.

[0012] Optionally, training a pre-constructed conditional generative adversarial network using the sandstone feature label set to obtain a microscopic image generation model for weathered sandstone includes:

[0013] Based on the Adam optimizer, the sandstone feature label set and its corresponding sandstone sample images are input into a pre-constructed conditional generative adversarial network (GAN) to iteratively update the parameters in the GAN until the preset comprehensive loss function converges, thereby obtaining the microscopic image generation model of the weathered sandstone. The preset comprehensive loss function includes an adversarial loss function, a mineral composition loss function, a orientation loss function, a weathering product and porosity ratio loss function, and a structural similarity loss function.

[0014] Optionally, the pre-constructed conditional generative adversarial network includes a generator and a discriminator. The generator uses random noise and the sandstone feature label set as conditional inputs to generate microstructure images of weathered sandstone at the corresponding weathering stage. The discriminator adopts the PatchGAN architecture to distinguish between microsample images of weathered sandstone and microstructure images of weathered sandstone.

[0015] Optionally, the method further includes: comparing the microstructure image of the weathered sandstone with the microstructure sample image of the weathered sandstone corresponding to the feature label of the sandstone to be predicted, so as to quantitatively evaluate the performance of the weathered sandstone microstructure image generation model.

[0016] A second aspect of this invention provides a device for generating microscopic images of weathered sandstone based on a conditional generative adversarial network (GAN), comprising: a database construction module for acquiring and analyzing sandstone samples with different degrees of weathering to construct a multi-scale database of weathered sandstone; a tag set construction module for extracting weathering geological knowledge from the multi-scale database of weathered sandstone to construct a sandstone feature tag set; a training module for training a pre-constructed GAN using the sandstone feature tag set to obtain a microscopic image generation model of weathered sandstone; and a generation module for inputting the sandstone feature tags to be predicted into the microscopic image generation model of weathered sandstone to generate microscopic structural images of weathered sandstone.

[0017] A third aspect of the present invention provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks as described in the above embodiments.

[0018] A fourth aspect of the present invention provides a computer program product that, when executed by a processor, implements the above-described method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks.

[0019] A fifth aspect of the present invention provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks.

[0020] The microscopic image generation method for weathered sandstone based on conditional generative adversarial networks proposed in this invention overcomes the limitations of traditional data-driven methods in weathering mechanism modeling. It integrates weathering mechanism knowledge, physical labels, and generative networks to construct a physically interpretable image prediction model. By building a multi-scale database of weathered sandstone, extracting weathering feature labels, and constructing a physically constrained conditional generative adversarial network model, it achieves the generation of sandstone microscopic images and the prediction of its physical and mechanical properties, possessing realistic physical attributes, structural fidelity, and weathering regularity. It can generate sandstone crystal images with realistic weathering evolution patterns at any weathering stage and supports weathering mechanism analysis and physical and mechanical performance evaluation, exhibiting good scalability and engineering application value. It can be widely applied in fields such as digital modeling and simulation of weathered rocks, possessing good cross-scale scalability and rock-like modeling capabilities.

[0021] Additional aspects and advantages of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0022] The above and / or additional aspects and advantages of the present invention will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein:

[0023] Figure 1 A flowchart illustrating a method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks, provided in an embodiment of the present invention.

[0024] Figure 2 This is a schematic diagram of a multi-scale database for weathered sandstone provided in an embodiment of the present invention;

[0025] Figure 3 This is a schematic diagram of the structure of a conditional generative adversarial network provided in an embodiment of the present invention;

[0026] Figure 4 Comparison of sandstone microscopic images generated by CGAN under different weathering degrees provided in embodiments of the present invention;

[0027] Figure 5 This is a block diagram of a weathered sandstone micro-image generation device based on a conditional generative adversarial network provided in an embodiment of the present invention.

[0028] Figure 6 This is a schematic diagram of the structure of the electronic device provided in an embodiment of the present invention. Detailed Implementation

[0029] Embodiments of the present invention are described in detail below, examples of which are illustrated in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain the present invention, and should not be construed as limiting the present invention.

[0030] The following describes, with reference to the accompanying drawings, a method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks according to embodiments of the present invention.

[0031] Figure 1 This is a schematic flowchart illustrating a method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks, as provided in an embodiment of the present invention.

[0032] like Figure 1 As shown, the method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks includes the following steps:

[0033] In step S101, sandstone samples with different degrees of weathering are acquired and analyzed to construct a multi-scale database of weathered sandstone.

[0034] In some embodiments, sandstone samples with different degrees of weathering are acquired and analyzed to construct a multi-scale database of weathered sandstone, including:

[0035] Collect sandstone samples with different degrees of weathering;

[0036] Microscopic parameters of sandstone samples with different degrees of weathering were obtained using a pre-set diffraction device;

[0037] Macroscopic physical and mechanical parameters of sandstone samples with different degrees of weathering were obtained through macroscopic physical experiments.

[0038] Data processing was performed on microscopic and macroscopic physical and mechanical parameters to obtain a multi-scale database of weathered sandstone with weathering stage labels and corresponding structural parameters.

[0039] like Figure 2As shown, in the actual implementation process, sandstone samples with different weathering degrees are collected. At the microscale, X-ray diffraction (XRD), scanning electron microscopy (SEM), and electron backscatter diffraction (EBSD) can be used to obtain microscopic parameters such as mineral phase information and microstructure information of sandstone. Among them, mineral phase information mainly refers to the types and proportions of mineral components, grain shape, contact relationship, and spatial distribution pattern, while microstructure information mainly includes... At the macroscale, macroscopic experiments such as uniaxial compression, Brazilian splitting, and wave velocity testing can be used to obtain macroscopic physical and mechanical parameters such as elastic modulus, hardness, compressive strength, tensile strength, fracture toughness, and wave velocity of rocks with different weathering degrees. The obtained microscopic and macroscopic physical and mechanical parameters are standardized to establish a multi-scale sandstone database with weathering stage labeling and structural parameter correspondence, providing data support for subsequent weathering geological knowledge extraction, feature label construction, and model training.

[0040] In step S102, weathering geological knowledge is extracted from the multi-scale database of weathered sandstone to construct a sandstone feature tag set.

[0041] In some embodiments, weathering geological knowledge is extracted from a multi-scale database of weathered sandstone to construct a sandstone feature tag set, including:

[0042] Mineral composition data, weathering product data, and microstructure data were extracted from a multi-scale database of weathered sandstone.

[0043] Quantitative analysis of mineral composition data was performed to obtain the corresponding volume fraction and mass fraction, and mineral composition labels for sandstone samples with different degrees of weathering were set according to the volume fraction and mass fraction.

[0044] Based on the preset weathering grade standard, weathering degree labels are set for sandstone samples with different weathering degrees according to the weathering product data;

[0045] Extract microstructural features from the microstructural data, and set microstructural feature labels for sandstone samples with different weathering degrees based on the microstructural features;

[0046] Analyze the mineral crystal orientation distribution in the microstructure data, and set crystal orientation labels for sandstone samples with different weathering degrees based on the mineral crystal orientation distribution;

[0047] The similarity of local details of sandstone mineral crystals in microstructure data is calculated to obtain the corresponding structural similarity, and structural similarity labels are set for sandstone samples with different weathering degrees based on the structural similarity.

[0048] Preprocessing is performed on mineral composition tags, weathering degree tags, microstructure feature tags, crystal orientation tags, and structural similarity tags to construct a sandstone feature tag set with real physical properties.

[0049] In practical application, based on multi-scale weathering research, weathering significantly affects the mineral composition, distribution, and microporosity characteristics of sandstone. Key characteristic mineral composition data, weathering product data, and microstructure data are extracted. It should be noted that mineral composition data refers to the extraction of the main mineral components from the rock sample, combined with nanoindentation testing to obtain the mechanical parameters of the mineral crystals, and the classification of minerals as weather-resistant or weather-prone. Weathering product data refers to identifying weathering products formed during the weathering process (such as kaolinite and hematite), allowing analysis of their formation and distribution patterns at different stages of weathering. Microstructure data refers to the evolution of microstructures such as mineral distribution, mineral morphology, and microporosity during weathering.

[0050] After obtaining mineral composition data, weathering product data, and microstructure data, the mineral composition (including weathering-resistant minerals, easily weathered minerals, and weathering products) of sandstone samples was quantitatively analyzed using techniques such as X-ray diffraction and scanning electron microscopy to obtain their volume fraction and mass fraction. Mineral composition labels for sandstone samples with different weathering degrees were then set based on these volume and mass fractions. According to the weathering grade standards of the International Society for Rock Mechanics, the weathering degree of each sandstone sample was labeled, i.e., weathering degree labels for sandstone samples with different weathering degrees were set based on weathering product data, ensuring a correspondence between the selected weathering characteristics and the physical and mechanical properties of the sandstone. Microstructural features such as porosity and fracture density were extracted using CT scanning technology, and microstructural feature labels for sandstone samples with different weathering degrees were set based on these microstructural features.

[0051] Furthermore, the orientation distribution of mineral crystals was analyzed using microscopic techniques such as electron backscatter diffraction, serving as "orientation" labels to describe the directional characteristics of mineral arrangement in sandstone. Specifically, orientation labels for sandstone samples with different weathering degrees were assigned based on the orientation distribution of mineral crystals. Simultaneously, a structural similarity index was introduced to measure the similarity of local details in sandstone mineral crystals; that is, structural similarity labels for sandstone samples with different weathering degrees were assigned based on structural similarity.

[0052] Preprocessing is performed on mineral composition labels, weathering degree labels, microstructure feature labels, crystal orientation labels, and structural similarity labels. The preprocessing includes classification according to weathering degree, mean-standard deviation normalization, median filtering for noise reduction, and affine transformation for data augmentation. Finally, a sandstone feature label set with real physical properties is constructed using the preprocessed mineral composition labels, weathering degree labels, microstructure feature labels, crystal orientation labels, and structural similarity labels to serve as a representative and diverse high-quality training dataset for subsequent model training.

[0053] It should be noted that the affine transformations used include random rotation, scaling, translation, and mirror flipping to improve the robustness of the dataset and the generalization ability of image generation.

[0054] In step S103, the pre-constructed conditional generative adversarial network is trained using the sandstone feature label set to obtain a microscopic image generation model for weathered sandstone.

[0055] In some embodiments, a pre-constructed conditional generative adversarial network is trained using a sandstone feature tag set to obtain a microscopic image generation model for weathered sandstone, including:

[0056] Based on the Adam optimizer, the sandstone feature label set and its corresponding sandstone sample images are input into a pre-constructed conditional generative adversarial network to iteratively update the parameters in the conditional generative adversarial network until the preset comprehensive loss function converges, thus obtaining a microscopic image generation model of weathered sandstone.

[0057] like Figure 3 As shown, in actual execution, the pre-constructed conditional generative adversarial network includes a generator and a discriminator. The generator employs a convolutional neural network (CNN) structure, containing fully connected layers, upsampling modules, and conditional embedding layers to enhance its ability to model the distribution and variation patterns of minerals. The generator uses random noise and a set of sandstone feature labels as conditional inputs to guide the generator to output microstructure images of weathered sandstone at corresponding weathering stages, which can be expressed as: x' i =G(z,w) j M i ,p k O k SSIM k ), where z is random noise, G is the generator, and x' i For the generated microscopic images of sandstone during the weathering stage, w j M is a weathering degree label. i For mineral composition labels, p k For microstructural feature labels, O kSSIM is a mineral orientation label. k For structural similarity labeling, the discriminator employs the PatchGAN architecture to distinguish between microscopic sample images of weathered sandstone and microscopic structural images of weathered sandstone. It uses the generated sandstone microscopic image x' i and its corresponding real sample x i As input, supplemented by conditional labels w j M i p k O k and SSIM k It can be represented as: D(x) i ,w j M i ,p k O k SSIM k ).

[0058] Furthermore, based on the Adam optimizer, the initial learning rate is set at 1e -4 up to 1e -5 In this process, the sandstone feature label set and its corresponding sandstone sample images are input into a pre-constructed conditional generative adversarial network. The generator generates a fake sample image G(z|y) based on the sandstone feature label set. The discriminator compares the fake sample image G(z|y) with the corresponding sandstone sample image to update the discriminator's parameters based on the comparison results. At the same time, the generator's parameters are updated. The aforementioned modulation process is iteratively executed to achieve explicit control over the weathering stage features and improve the interpretability of the prediction results. This process continues until the trend of the preset comprehensive loss function dynamically decays, that is, the preset comprehensive loss function converges, thus obtaining the weathered sandstone micro-image generation model.

[0059] The preset comprehensive loss function consists of two parts: adversarial loss and physical attribute loss. The adversarial loss function is used to optimize the game between the generator and the discriminator. The physical attribute loss includes image quality loss and physical constraint loss. Image quality loss includes perceptual similarity (lpips) and structural similarity (SSIM) loss. The physical constraint loss is a structural loss term set based on the physical continuity and compatibility rules of mineral distribution in the geological background. Specifically, it includes mineral composition loss function, orientation loss function, weathering product and porosity ratio loss function, and structural similarity loss function.

[0060] The specific expression for the preset comprehensive loss function is as follows:

[0061] L gen =L adv +λ1L mineral-comp +λ2L mineral-shape +λ3L weathering-prod +λ4L porosity +λ5LSSIM (1)

[0062] Where λ1, λ2, λ3, λ4, and λ5 are the weight hyperparameters of each loss term, and L weathering-prod Let L be the directionality loss function. mineral-pro For the proportion of weathering products lost, L porosity This represents the loss due to porosity.

[0063] L adv The adversarial loss function for the Conditional Generative Adversarial Network (CGAN) is expressed as follows:

[0064]

[0065] In the formula, To determine the true data distribution P data The expected value of D(x|y) is the probability output of the discriminator D, given condition y, that it classifies the input sample x as a true sample. To determine the noise distribution P z The expected value, D(G(z|y)), is the probability output of the discriminator D under given condition y, which classifies the sample G(z|y) generated by the generator G as a real sample.

[0066] L mineral-comp The mineral composition loss function ensures that the mineral composition of the generated rock conforms to the weathering process; its specific expression is as follows:

[0067]

[0068] In the formula, M represents the predicted volume fraction of the i-th type of mineral in the generated image. i This represents the volume fraction of the i-th type of mineral in a real geological sample.

[0069] L mineral-shap The mineral morphology loss function uses angular variance and cosine similarity to measure the difference in orientation between the generated mineral crystals and the real samples. The specific expressions are as follows:

[0070]

[0071] In the formula, and O i V represents the orientation distribution of the generated mineral crystal and the real mineral crystal, respectively. θ Variance of control orientation, S cos Similarity of control orientation, λ θ and λ cos These are hyperparameters used to adjust the weights.

[0072] Specifically, to ensure that the overall distribution and local details of the rock mineral structure conform to reality during the generation of the weathered sandstone microscopic image generation model, a structural similarity function L is introduced into the loss function. SSIM It is used to control the local details (such as mineral crystal boundaries) of the generated rock and mineral crystal patterns.

[0073] In step S104, the feature labels of the sandstone to be predicted are input into the weathered sandstone micro-image generation model to generate a weathered sandstone microstructure image.

[0074] In some embodiments, it also includes:

[0075] The microstructure images of weathered sandstone are compared with the microstructure sample images of weathered sandstone corresponding to the feature labels of the sandstone to be predicted, so as to quantitatively evaluate the performance of the weathered sandstone microstructure image generation model.

[0076] like Figure 4 As shown, in actual implementation, arbitrary weathering degree labels and other condition labels are input into the weathered sandstone microstructure image generation model to generate physically consistent weathered sandstone microstructure images. By comparing the physically consistent weathered sandstone microstructure images with actual sample images, the performance of the generated images is quantitatively evaluated using indicators such as Structural Similarity Index (SSIM), Perceptual Similarity Index (LPIPS), and Binary Cross-Entropy (BCE). The physical validity and detail fidelity of the generated images are also analyzed by comparing them with real samples.

[0077] In addition, the evaluation process of the generated results can be supplemented by multi-scale structural similarity (MS-SSIM), which combines quantitative indicators and the interpretation results of domain experts to jointly verify the quality of image generation.

[0078] The following detailed description of the method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks, as proposed in this invention, is provided through a specific embodiment.

[0079] Step one involves collecting sandstone samples from multiple regions, covering different weathering levels from fresh to moderately weathered to strongly weathered. At the microscopic scale, scanning electron microscopy (SEM), electron backscatter diffraction (EBSD), and X-ray diffraction (XRD) are used to analyze parameters such as mineral composition, pore structure, and crystal orientation. At the macroscopic scale, uniaxial compression tests and wave velocity measurements are employed to obtain parameters such as elastic modulus, compressive strength, and permeability. Based on the collected parameters, a multi-scale database of weathered sandstone with weathering stage labels and corresponding structural parameters is constructed to provide data support for model training.

[0080] Step two: Based on the aforementioned multi-scale database of weathered sandstone with weathering stage annotations and corresponding structural parameters, the following tag information is extracted:

[0081] Mineral composition label (M i ): Identify the mass fraction of quartz, feldspar, clay minerals, weathering products, etc.

[0082] Weathering label (w j The samples were graded according to the weathering grade standard (W1–W5);

[0083] Microstructure feature labels (p k ): Based on CT and image analysis, porosity, fracture density, etc. are extracted;

[0084] Orientation Label (O) k ): The orientation distribution of crystals is measured by EBSD, and the variance of orientation angles and orientation deviations are calculated;

[0085] Structural Similarity Labels (SSIM) k ): Used to measure the spatial consistency and edge continuity of image structure.

[0086] All labels are normalized (0–1 interval) during the data preprocessing stage to obtain a sandstone feature label set with real physical properties, and this sandstone feature label set with real physical properties is introduced into the subsequent model as a conditional input.

[0087] Step 3: Obtain the pre-built Conditional Generative Adversarial Network (CGAN) structure and train the CGAN model using a set of sandstone feature labels with real physical properties.

[0088] The Conditional Generative Adversarial Network (CGAN) structure includes a generator G and a discriminator D. The input to the generator G is random noise z and a conditional label (x). i ,w j M i ,p k O k SSIM k The model introduces conditional information into the generative network through a splicing vector injection mechanism to generate mineral crystal images corresponding to the weathering stage. The discriminator D receives real samples or generated images and their corresponding labels, and uses the PatchGAN architecture to discriminate local regions of the image to improve detail fidelity. The model's comprehensive loss function includes adversarial loss, mineral composition loss, orientation loss, weathering product and porosity ratio loss, and structural similarity loss.

[0089] The training strategy employed is as follows: using the Adam optimizer, with an initial learning rate of 1e. -4The exponential decay rate β1 = 0.5 and the exponential decay rate β2 = 0.999. Two-dimensional batch normalization BatchNorm2d is used to batch normalize the sandstone feature label set to suppress gradient vanishing of the sandstone feature label set. Label smoothing real = 0.9 is set to enhance the robustness of the discriminator. Random noise smoothing and data augmentation are performed simultaneously to improve the model's generalization ability. The training loss convergence condition is set to SSIM > 0.85.

[0090] Based on the above strategy, the sandstone feature label set and its corresponding sandstone sample images are input into a pre-constructed conditional generative adversarial network to iteratively update the parameters in the conditional generative adversarial network until the preset comprehensive loss function converges, thus obtaining a microscopic image generation model for weathered sandstone. This model can stably generate microscopic structure images of weathered sandstone under different weathering degrees.

[0091] Step four: The following quantitative indicators are used to evaluate the model performance by comparing the generated microstructure images of weathered sandstone under different weathering degrees with the real images, so as to evaluate the generated images and verify their physical consistency.

[0092] If the mean Structural Similarity Index (SSIM) reaches 0.87, it indicates that the image structures are consistent; if the mean Perceptual Similarity Index (LPIPS) is below 0.12, it indicates that the model has good detail restoration capabilities; if the Binary Cross-Entropy (BCE) is less than 0.2, the reconstruction error is small; if the mineral composition deviation is controlled within ±3%, it conforms to the weathering stage variation law; if the variance difference of orientation angle is less than 0.05, it indicates good crystal orientation consistency. If all of the above are met, it means that the generated microstructure images of weathered sandstone under different weathering degrees have achieved a high level of accuracy in terms of macroscopic trends, detail representation, and label consistency, and have strong engineering application value.

[0093] In summary, the method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks proposed in this invention has the following beneficial effects:

[0094] (1) It breaks through the limitations of traditional data-driven methods in weathering mechanism modeling, integrates weathering mechanism knowledge, physical labels and generative networks, and constructs an image prediction model with physical interpretability;

[0095] (2) By constructing a multi-scale database of weathered sandstone, extracting weathering feature labels, and constructing a conditional generative adversarial network model with physical constraints, the generation of microscopic images of sandstone with real physical properties, structural fidelity and weathering regularity and the prediction of physical and mechanical properties were realized.

[0096] (3) It can generate sandstone crystal images with real weathering evolution laws under any weathering stage, and support weathering mechanism analysis and physical and mechanical property evaluation. It has good scalability and engineering application value.

[0097] (4) It can be widely used in fields such as digital modeling and simulation of weathered rocks, and has good cross-scale scalability and rock-like modeling capabilities.

[0098] Next, referring to the accompanying drawings, a microscopic image generation device for weathered sandstone based on a conditional generative adversarial network, according to an embodiment of the present invention, is described.

[0099] Figure 5 This is a block diagram illustrating a device for generating microscopic images of weathered sandstone based on a conditional generative adversarial network, as provided in an embodiment of the present invention.

[0100] like Figure 5 As shown, the weathered sandstone micro-image generation device 50 based on conditional generative adversarial network includes: a database construction module 501, a label set construction module 502, a training module 503, and a generation module 504.

[0101] The database construction module 501 is used to acquire and analyze sandstone samples with different degrees of weathering to construct a multi-scale database of weathered sandstone. The tag set construction module 502 is used to extract weathering geological knowledge from the multi-scale database of weathered sandstone to construct a sandstone feature tag set. The training module 503 is used to train a pre-constructed conditional generative adversarial network using the sandstone feature tag set to obtain a microscopic image generation model for weathered sandstone. The generation module 504 is used to input the sandstone feature tags to be predicted into the microscopic image generation model for weathered sandstone to generate microscopic structural images of weathered sandstone.

[0102] In some embodiments, the database construction module 501 includes:

[0103] The collection unit is used to collect sandstone samples with different degrees of weathering.

[0104] The first acquisition unit is used to acquire the microscopic parameters of sandstone samples with different degrees of weathering using a preset diffraction device;

[0105] The second acquisition unit is used to obtain the macroscopic physical and mechanical parameters of sandstone samples with different degrees of weathering through macroscopic physical experiments;

[0106] The processing unit is used to process microscopic and macroscopic physical and mechanical parameters to obtain a multi-scale database of weathered sandstone with weathering stage labels and corresponding structural parameters.

[0107] In some embodiments, the tag set construction module 502 includes:

[0108] The extraction unit is used to extract mineral composition data, weathering product data, and microstructure data from the multi-scale database of weathered sandstone;

[0109] The first setting unit is used to perform quantitative analysis on mineral composition data to obtain the corresponding volume fraction and mass fraction, and to set mineral composition labels for sandstone samples with different weathering degrees based on the volume fraction and mass fraction.

[0110] The second setting unit is used to set weathering degree labels for sandstone samples with different weathering degrees based on preset weathering grade standards and weathering product data.

[0111] The third setting unit is used to extract microstructural features from the microstructural data and set microstructural feature labels for sandstone samples with different weathering degrees based on the microstructural features.

[0112] The fourth setting unit is used to analyze the mineral crystal orientation distribution in the microstructure data and set crystal orientation labels for sandstone samples with different weathering degrees based on the mineral crystal orientation distribution.

[0113] The fifth setting unit is used to calculate the similarity of local details of sandstone mineral crystals in microstructure data to obtain the corresponding structural similarity, and set structural similarity labels for sandstone samples with different weathering degrees based on the structural similarity.

[0114] The construction unit is used to preprocess mineral composition labels, weathering degree labels, microstructure feature labels, crystal orientation labels, and structural similarity labels to construct a sandstone feature label set with real physical properties.

[0115] In some embodiments, the training module 503 includes:

[0116] Based on the Adam optimizer, the sandstone feature label set and its corresponding sandstone sample images are input into a pre-constructed conditional generative adversarial network (GAN) to iteratively update the parameters in the GAN until the preset comprehensive loss function converges, thus obtaining a microscopic image generation model for weathered sandstone. The preset comprehensive loss function includes an adversarial loss function, a mineral composition loss function, a orientation loss function, a weathering product and porosity ratio loss function, and a structural similarity loss function.

[0117] In some embodiments, the pre-built conditional generative adversarial network includes a generator and a discriminator. The generator takes random noise and a set of sandstone feature labels as conditional inputs to generate sandstone microscopic images corresponding to the weathering stage. The discriminator adopts the PatchGAN architecture to distinguish between sandstone microscopic sample images and sandstone microscopic generated images.

[0118] In some embodiments, it also includes:

[0119] The microstructure image of weathered sandstone is compared with the actual sample image corresponding to the feature label of the sandstone to be predicted in order to quantitatively evaluate the performance of the microstructure image generation model of weathered sandstone.

[0120] It should be noted that the foregoing explanation of the embodiment of the weathered sandstone micro-image generation method based on conditional generative adversarial networks also applies to the weathered sandstone micro-image generation device based on conditional generative adversarial networks in this embodiment, and will not be repeated here.

[0121] The microscopic image generation device for weathered sandstone based on conditional generative adversarial networks proposed in this invention has the following beneficial effects:

[0122] (1) It breaks through the limitations of traditional data-driven methods in weathering mechanism modeling, integrates weathering mechanism knowledge, physical labels and generative networks, and constructs an image prediction model with physical interpretability;

[0123] (2) By constructing a multi-scale database of weathered sandstone, extracting weathering feature labels, and constructing a conditional generative adversarial network model with physical constraints, the generation of microscopic images of sandstone with real physical properties, structural fidelity and weathering regularity and the prediction of physical and mechanical properties were realized.

[0124] (3) It can generate sandstone crystal images with real weathering evolution laws under any weathering stage, and support weathering mechanism analysis and physical and mechanical property evaluation. It has good scalability and engineering application value.

[0125] (4) It can be widely used in fields such as digital modeling and simulation of weathered rocks, and has good cross-scale scalability and rock-like modeling capabilities.

[0126] Figure 6 This is a schematic diagram of an electronic device provided in an embodiment of the present invention. The electronic device may include:

[0127] The memory 601, the processor 602, and the computer program stored on the memory 601 and capable of running on the processor 602.

[0128] When the processor 602 executes the program, it implements the method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks provided in the above embodiments.

[0129] Furthermore, electronic devices also include:

[0130] Communication interface 603 is used for communication between memory 601 and processor 602.

[0131] The memory 601 is used to store computer programs that can run on the processor 602.

[0132] The memory 601 may include high-speed RAM memory, and may also include non-volatile memory, such as at least one disk storage device.

[0133] If the memory 601, processor 602, and communication interface 603 are implemented independently, then the communication interface 603, memory 601, and processor 602 can be interconnected via a bus to complete communication between them. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus, or an Extended Industry Standard Architecture (EISA) bus, etc. Buses can be categorized as address buses, data buses, control buses, etc. For ease of representation, Figure 6 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0134] Optionally, in a specific implementation, if the memory 601, processor 602, and communication interface 603 are integrated on a single chip, then the memory 601, processor 602, and communication interface 603 can communicate with each other through an internal interface.

[0135] Processor 602 may be a central processing unit (CPU), an application specific integrated circuit (ASIC), or one or more integrated circuits configured to implement embodiments of the present invention.

[0136] This invention also provides a computer program product, which, when executed by a processor, implements the above-described method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks.

[0137] 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 generating microscopic images of weathered sandstone based on conditional generative adversarial networks.

[0138] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of the present invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0139] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this invention, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0140] Any process or method description in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of preferred embodiments of the invention includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as will be understood by those skilled in the art to which embodiments of the invention pertain.

[0141] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as a sequenced list of executable instructions for implementing logical functions, and can be embodied in any computer-readable medium for use by, or in conjunction with, an instruction execution system, apparatus, or device (such as a computer-based system, a processor-included system, or other system that can fetch and execute instructions from, an instruction execution system, apparatus, or device). For the purposes of this specification, "computer-readable medium" can be any means that can contain, store, communicate, propagate, or transmit programs for use by, or in conjunction with, an instruction execution system, apparatus, or device. More specific examples (a non-exhaustive list) of computer-readable media include: an electrical connection having one or more wires (electronic device), a portable computer disk drive (magnetic device), random access memory (RAM), read-only memory (ROM), erasable and editable read-only memory (EPROM or flash memory), fiber optic devices, and portable optical disc read-only memory (CDROM). Alternatively, the computer-readable medium may be paper or other suitable media on which the program can be printed, since the program can be obtained electronically by optically scanning the paper or other medium, followed by editing, interpreting, or otherwise processing as necessary, and then stored in a computer memory.

[0142] It should be understood that various parts of the present invention can be implemented in hardware, software, firmware, or a combination thereof. In the above embodiments, the N steps or methods can be implemented in software or firmware stored in memory and executed by a suitable instruction execution system. If implemented in hardware, as in another embodiment, it can be implemented using any one or a combination of the following techniques known in the art: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (PGAs), field-programmable gate arrays (FPGAs), etc.

[0143] Those skilled in the art will understand that all or part of the steps of the methods in the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0144] Furthermore, the functional units in the various embodiments of the present invention can be integrated into a processing module, or each unit can exist physically separately, or two or more units can be integrated into a module. The integrated module can be implemented in hardware or as a software functional module. If the integrated module is implemented as a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0145] The storage medium mentioned above can be a read-only memory, a disk, or an optical disk, etc. Although embodiments of the present invention have been shown and described above, it is to be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those skilled in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks, characterized in that, Includes the following steps: Acquire and analyze sandstone samples with different degrees of weathering to construct a multi-scale database of weathered sandstone; Weathering geological knowledge is extracted from the aforementioned multi-scale database of weathered sandstone to construct a sandstone feature tag set, specifically including: Mineral composition data, weathering product data, and microstructure data are extracted from the multi-scale database of the weathered sandstone. The mineral composition data are quantitatively analyzed to obtain the corresponding volume fraction and mass fraction, and mineral composition labels are set for the sandstone samples with different weathering degrees based on the volume fraction and mass fraction. Based on a preset weathering grade standard, weathering grade labels are set for sandstone samples with different weathering degrees according to the weathering product data; Extract the microstructure features from the microstructure data, and set microstructure feature labels for sandstone samples with different weathering degrees based on the microstructure features; Analyze the mineral crystal orientation distribution in the microstructure data, and set crystal orientation labels for sandstone samples with different weathering degrees based on the mineral crystal orientation distribution; Calculate the similarity of local details of sandstone mineral crystals in the microstructure data to obtain the corresponding structural similarity, and set structural similarity labels for sandstone samples with different weathering degrees based on the structural similarity; The mineral composition labels, weathering degree labels, microstructure feature labels, crystal orientation labels, and structural similarity labels are preprocessed to construct a sandstone feature label set with real physical properties. The pre-constructed conditional generative adversarial network was trained using the sandstone feature tag set to obtain a microscopic image generation model for weathered sandstone. Input the feature labels of the sandstone to be predicted into the weathered sandstone micro-image generation model to generate a micro-structure image of the weathered sandstone.

2. The method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks according to claim 1, characterized in that, The acquisition and analysis of sandstone samples with different degrees of weathering to construct a multi-scale database of weathered sandstone includes: Collect sandstone samples with different degrees of weathering; The microscopic parameters of the sandstone samples with different degrees of weathering were obtained using a pre-set diffraction device; Macroscopic physical and mechanical parameters of sandstone samples with different degrees of weathering were obtained through macroscopic physical experiments. The microscopic parameters and macroscopic physical and mechanical parameters are processed to obtain a multi-scale database of weathered sandstone with weathering stage labels and corresponding structural parameters.

3. The method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks according to claim 1, characterized in that, The step of training a pre-constructed conditional generative adversarial network using the sandstone feature tag set to obtain a microscopic image generation model for weathered sandstone includes: Based on the Adam optimizer, the sandstone feature label set and its corresponding sandstone sample images are input into a pre-constructed conditional generative adversarial network (GAN) to iteratively update the parameters in the GAN until the preset comprehensive loss function converges, thereby obtaining the microscopic image generation model of the weathered sandstone. The preset comprehensive loss function includes an adversarial loss function, a mineral composition loss function, a orientation loss function, a weathering product and porosity ratio loss function, and a structural similarity loss function.

4. The method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks according to claim 1, characterized in that, The pre-constructed conditional generative adversarial network includes a generator and a discriminator. The generator uses random noise and the sandstone feature label set as conditional inputs to generate microstructure images of weathered sandstone at the corresponding weathering stage. The discriminator adopts the PatchGAN architecture to distinguish between microsample images of weathered sandstone and microstructure images of weathered sandstone.

5. The method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks according to claim 1, characterized in that, Also includes: The microstructure image of the weathered sandstone is compared with the microstructure sample image of the weathered sandstone corresponding to the feature label of the sandstone to be predicted, so as to quantitatively evaluate the performance of the weathered sandstone microstructure image generation model.

6. A device for generating microscopic images of weathered sandstone based on conditional generative adversarial networks, characterized in that, include: The database construction module is used to acquire and analyze sandstone samples with different degrees of weathering in order to build a multi-scale database of weathered sandstone. The tag set construction module is used to extract weathering geological knowledge from the multi-scale database of weathered sandstone to construct a sandstone feature tag set, specifically including: The extraction unit is used to extract mineral composition data, weathering product data, and microstructure data from the multi-scale database of weathered sandstone. The first setting unit is used to perform quantitative analysis on mineral composition data to obtain the corresponding volume fraction and mass fraction, and to set mineral composition labels for sandstone samples with different weathering degrees based on the volume fraction and mass fraction. The second setting unit is used to set weathering degree labels for sandstone samples with different weathering degrees based on preset weathering grade standards and weathering product data. The third setting unit is used to extract microstructural features from the microstructural data and set microstructural feature labels for sandstone samples with different weathering degrees based on the microstructural features. The fourth setting unit is used to analyze the mineral crystal orientation distribution in the microstructure data and set crystal orientation labels for sandstone samples with different weathering degrees based on the mineral crystal orientation distribution. The fifth setting unit is used to calculate the similarity of local details of sandstone mineral crystals in microstructure data to obtain the corresponding structural similarity, and set structural similarity labels for sandstone samples with different weathering degrees based on the structural similarity. The building unit is used to preprocess mineral composition tags, weathering degree tags, microstructure feature tags, crystal orientation tags, and structural similarity tags to construct a sandstone feature tag set with real physical properties; The training module is used to train a pre-constructed conditional generative adversarial network using the sandstone feature label set to obtain a microscopic image generation model of weathered sandstone. The generation module is used to input the feature labels of the sandstone to be predicted into the weathered sandstone micro-image generation model to generate a micro-structure image of the weathered sandstone.

7. An electronic device, characterized in that, include: The memory, the processor, and the computer program stored in the memory and executable on the processor, the processor executing the program to implement the method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks as described in any one of claims 1-5.

8. A computer program product, characterized in that, When the computer program / instruction is executed by the processor, it implements the method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks as described in any one of claims 1-5.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, The program is executed by the processor to implement the method for generating microscopic images of weathered sandstone based on conditional generative adversarial networks as described in any one of claims 1-5.

Images