Prediction method for corrosion rate of natural gas pipeline

Abstract

The invention relates to a method for predicting the corrosion rate of a natural gas pipeline, in particular to the field of natural gas pipeline corrosion prediction.The high-precision dynamic early warning is achieved through the four-step cooperative technology, firstly, visible light and radar data are fused to generate a three-dimensional corrosion base film fully covering the pipeline, and blind areas caused by the complex environment are effectively eliminated; secondly, speculating corrosion probability distribution of a shielding area based on a physical constraint generative adversarial network, and solving a hidden area prediction problem; then, in combination with the stress field and historical data, deducing a space-time diffusion path of corrosion through a graph neural network, and outputting a future corrosion thermodynamic diagram; and finally, dynamically adjusting model parameters according to actually measured data to form a corrosion risk decision matrix of closed-loop optimization, thereby realizing the transformation of pipeline corrosion from passive protection to intelligent prevention in a breakthrough manner.

Description

Technical Field

[0001] This invention relates to the field of natural gas pipeline corrosion prediction, and more specifically, to a method for predicting the corrosion rate of natural gas pipelines. Background Technology

[0002] Natural gas pipelines, as vital energy transmission facilities, face inspection challenges in complex surface environments. In mountainous, swampy, and densely populated urban areas, pipeline surfaces are often obstructed by terrain undulations, vegetation cover, and man-made structures, making some areas undetectable. Furthermore, hidden areas of coating damage and diverse corrosion patterns may exist on the pipeline surface, making it difficult for traditional inspection methods to obtain comprehensive and accurate corrosion data. Especially in obscured or inaccessible areas, traditional manual inspections often fail to provide complete corrosion information, leading to missed detections. Current inspection methods lack the ability to cope with complex geographical conditions and harsh environments, particularly in rainy or snowy weather, making it difficult to obtain clear images and increasing the risk of major pipeline safety accidents.

[0003] Existing inspection technologies primarily rely on visible light and infrared cameras. While these can detect some corrosion areas, in complex terrain, their limited field of view makes it difficult to penetrate obstructions, resulting in incomplete coverage of the pipe surface. Current image stitching technologies suffer from significant errors when processing discontinuous damaged areas, making it difficult for the stitched image to accurately reflect the corrosion status. Furthermore, environmental factors such as severe weather increase image noise, further affecting the accuracy of corrosion feature extraction. These issues limit the effectiveness of existing technologies in complex environments, significantly increasing the missed detection rate and raising the risk of leaks or explosions caused by undetected corrosion. Therefore, there is an urgent need for a novel multi-source obstruction-penetration imaging and 3D reconstruction technology that can overcome the limitations of traditional technologies, achieve millimeter-level precise monitoring of corrosion areas, ensure efficient coverage under various environmental conditions, and minimize missed detections and potential risks. Summary of the Invention

[0004] This invention addresses the technical problems existing in the prior art by providing a method for predicting the corrosion rate of natural gas pipelines. It achieves high-precision dynamic early warning through a four-step collaborative technology, thereby solving the problems mentioned in the background.

[0005] The technical solution of this invention to solve the above-mentioned technical problems is as follows: specifically, it includes the following steps: Step S1: When the visible light image, synthetic aperture radar data and environmental sensing data are received, a cross-modal deformation energy function registration operation is performed on the texture gradient of the visible light image and the polarization scattering matrix of the synthetic aperture radar to generate a three-dimensional corrosion feature base film covering the pipe surface. Step S2: When there is an environmental occlusion area in the three-dimensional corrosion feature base film, the observation area features of the base film are used as input conditions, combined with environmental sensing parameters and pipeline material characteristic parameters, and a physical constraint generative adversarial network is used to generate a corrosion probability distribution map of the occlusion area. Step S3: When the pipeline stress distribution data and historical corrosion data are obtained, the corrosion probability distribution map is used as the initial field. A graph neural network is used to couple the corrosion-stress-diffusion equation to perform a spatiotemporal extrapolation operation and output a heat map of the corrosion propagation path for a future preset time period. Step S4: When new inspection data is input, a multi-objective optimization function is constructed based on the difference between the predicted value and the measured value, and the weight parameters of the generative adversarial network and the graph neural network are dynamically updated to generate a corrosion risk decision matrix. In a preferred embodiment, the visible light image, synthetic aperture radar data, and environmental sensor data in step one are acquired synchronously in the following manner: Visible light images are acquired by an inspection drone equipped with a multispectral gimbal camera on a preset track. Its imaging band covers the 400-900nm visible light and near-infrared spectral range. At the same time, synthetic aperture radar data is acquired by an X-band polarimetric interferometric radar on the same platform, penetrating ground obstructions. The radar incident angle is set to an adjustment range of 25°-50° to adapt to different terrain elevation angles. Environmental sensing data is monitored in real time by a network of sensor nodes buried in the soil around the pipeline. Its sampling parameters include at least three key physical quantities: temperature, conductivity, and pH value.

[0006] In a preferred embodiment, the specific process of the registration operation is as follows: First, feature tensors are constructed. For visible light images, spatial gradient tensors are calculated to enhance the edge features of the image. The Sobel operator is used to perform convolution operations on the image to obtain the gradient information of each pixel. For synthetic aperture radar data, scattering features are extracted and polarimetric scattering entropy is calculated to describe different polarization features. Next, a cross-modal deformation energy function is established, which consists of three parts: feature consistency constraint, thin plate spline constraint, and thermal deformation compensation. In the optimization phase, an improved alternating direction multiplier method algorithm is used to iteratively solve the two-dimensional non-rigid transformation field in order to optimize the matching process of cross-modal deformation. Finally, the optimized transform field was used to generate a three-dimensional corrosion feature base film that combines the real and imaginary parts of the complex scattering features of SAR data with the curvature features of the visible light image.

[0007] In a preferred embodiment, the specific operation of generating the corrosion probability distribution map of the shielded area in step S2 is as follows: A1. First, the observation area features of the base film are used as input. These features include depth information and spatial coordinates in the image. Second, the occluded areas in the image are identified, specifically those areas whose depth features are unreliable. By calculating the gradient of the base film in the depth dimension, if the gradient value is less than a preset threshold, these areas are marked as occluded areas. A2. Integrate environmental sensor parameters and pipe material characteristic parameters; A3. Next, a two-channel network is constructed to generate erosion feature maps. This network consists of three modules: Feature encoder: It fuses the input base film features and environmental sensor data to generate a feature vector containing spatial correlation information; Physical constraint: By processing material property parameters and environmental sensitivity to corrosion rate, it generates a physical constraint output; Corrosion generator: Combines random noise, encoded feature vectors, and the output of the physical constraint to generate the final corrosion feature map; During training, the generator's output is optimized through adversarial training, with the training objective being to minimize two loss functions: Dual-discrimination loss: This loss function is optimized by the difference between the generated image and the real image, encouraging the generator to produce more realistic erosion features; Physical loss: This loss measures the difference between the generated corrosion characteristics and the actual corrosion driving force field, which is related to the iron ion concentration gradient and pH changes; A4. After training and optimization, an accurate corrosion probability distribution map is finally output.

[0008] In a preferred embodiment, the specific steps for performing the spatiotemporal extrapolation operation are as follows: A1. Use the corrosion probability distribution map output in step S2 as the initial field of spatial corrosion depth, and simultaneously load the pipeline stress tensor and the historical corrosion time series library. A2. Establish a graph structure consisting of corrosion feature nodes, stress feature nodes, and diffusion channel edges, where: The corrosion feature nodes are bound to spatial coordinates, carrying the corrosion probability value at the corresponding location and its historical change gradient; The stress characteristic node carries an equivalent stress value, which is calculated based on the horizontal principal stress, vertical stress, and shear stress components. The edge weights of the diffusion channel are determined by a curvature weighting function of the distance between adjacent nodes and the difference in equivalent stress. A3. Input the graph structure into the graph neural network, generate the corrosion depth prediction field for future time periods through iterative solution of the corrosion-stress-diffusion coupling equation, spatially discretize and encode the corrosion depth prediction field at the current time step, and output the corrosion expansion path heat map after superimposing the exponential decay memory effect of the historical prediction field.

[0009] In a preferred embodiment, the corrosion-stress-diffusion coupling equation includes: Stress corrosion acceleration term based on material stress sensitivity coefficient; Nonlocal spatial diffusion integral term containing anisotropic kernel function; Stress-assisted diffusion term driven by hydrogen ion concentration gradient.

[0010] In a preferred embodiment, the specific operation of constructing the multi-objective optimization function in step S4 is as follows: A1. First, new inspection data is received and a corrosion extension path heat map is generated simultaneously. Then, by comparing the current corrosion heat map with historical data, the spatiotemporal alignment difference tensor is calculated and gradient calculation is used to optimize model prediction. A2. Define an innovative third-order optimization objective function, which includes three main sub-terms: Geometric topology loss: measures the changes in the corrosion morphology of a material, using gradient and tensor operations to describe the changes in the material; Corrosion risk functional: Calculates the weighted sum of material corrosion risk and predicts future corrosion risk based on stress depth changes at different time points; Memory entropy: Used to estimate the learning ability of a neural network based on past experience, reflecting the model's memory and adaptive capabilities.

[0011] In a preferred embodiment, the dual-network coupling mechanism specifically comprises: By utilizing two neural networks working together and updating parameters through a gradient projection operator, while introducing a Jacobian response tensor, the model is further optimized by calculating the relationship between network weights and the response of erosion prediction.

[0012] In a preferred embodiment, the triggering condition for the dynamic update is: If the Frobenius norm of the weight update is greater than the product of the set threshold and the failure factor, then the network parameter update is triggered; otherwise, it is not triggered.

[0013] In a preferred embodiment, the generated corrosion risk decision matrix is ​​constructed in the following manner; A1. Decision Matrix Framework: Construct a two-dimensional decision matrix consisting of m rows and n columns of decision units, where each decision unit is bound to a preset spatial location coordinate; A2. Decision Unit Calculation: The value of each decision unit is generated by the core decision function, which is executed in the following order: Calculate the third mixed partial derivatives of the corrosion propagation thermogram with respect to specified spatial coordinates, time variables, and horizontal stress; After multiplying the above partial derivatives by the risk accumulation factor, the result is input into the activation function for normalization. A3. Risk Factor Construction: The risk accumulation factor consists of the product of two items: Stress deep coupling term: After performing a series multiplication operation on the first, second and third gradient magnitudes of the equivalent force field, the hyperbolic tangent function output value is taken; Failure time gain term: a natural exponential function with the critical failure time as the exponent, where the exponent coefficient is a preset material sensitivity constant; A4. Matrix Output: The values ​​of all decision units form a corrosion risk decision matrix, and the row and column dimensions of this matrix are consistent with the pipeline space segmentation.

[0014] The beneficial effects of this invention are as follows: The method for predicting the corrosion rate of natural gas pipelines achieves high-precision dynamic early warning through a four-step collaborative technology. First, it uses the fusion of visible light and radar data to generate a three-dimensional corrosion base film covering the entire pipeline, effectively eliminating blind spots caused by complex environments. Next, it uses a generative adversarial network based on physical constraints to infer the corrosion probability distribution of the obscured area, solving the problem of predicting hidden areas. Then, it combines stress field and historical data to deduce the spatiotemporal diffusion path of corrosion through a graph neural network, outputting a future corrosion heat map. Finally, it dynamically adjusts the model parameters based on measured data to form a closed-loop optimized corrosion risk decision matrix, thereby achieving a breakthrough in transforming pipeline corrosion from passive protection to intelligent prevention. Attached Figure Description

[0015] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation

[0016] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0017] In the description of this application, 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 one or more of the stated features. In the description of this application, "multiple" means two or more, unless otherwise explicitly specified.

[0018] In the description of this application, the term "for example" is used to mean "used as an example, illustration, or description." Any embodiment described as "for example" in this application is not necessarily to be construed as being more preferred or advantageous than other embodiments. The following description is provided to enable any person skilled in the art to make and use the invention. Details are set forth in the following description for purposes of explanation. It should be understood that those skilled in the art will recognize that the invention can be made without using these specific details. In other instances, well-known structures and processes will not be described in detail to avoid obscuring the description of the invention with unnecessary detail. Therefore, the invention is not intended to be limited to the embodiments shown, but is consistent with the broadest scope of the principles and features disclosed in this application.

[0019] Example 1 This embodiment provides, for example Figure 1 The method for predicting the corrosion rate of a natural gas pipeline, as shown, specifically includes the following steps: Step S1: When the visible light image, synthetic aperture radar data and environmental sensing data are received, a cross-modal deformation energy function registration operation is performed on the texture gradient of the visible light image and the polarization scattering matrix of the synthetic aperture radar to generate a three-dimensional corrosion feature base film covering the pipe surface. Step S2: When there is an environmental occlusion area in the three-dimensional corrosion feature base film, the observation area features of the base film are used as input conditions. Combined with environmental sensing parameters and pipeline material characteristic parameters, a physical constraint generative adversarial network is used to generate a corrosion probability distribution map of the occlusion area. Step S3: When the pipeline stress distribution data and historical corrosion data are obtained, the corrosion probability distribution map is used as the initial field. A graph neural network is used to couple the corrosion-stress-diffusion equation to perform a spatiotemporal extrapolation operation and output a heat map of the corrosion propagation path for a future preset time period. Step S4: When new inspection data is input, a multi-objective optimization function is constructed based on the difference between the predicted value and the measured value. The weight parameters of the adversarial network and the graph neural network are dynamically updated to generate a corrosion risk decision matrix.

[0020] In this embodiment, it is particularly important to explain step one, where visible light images, synthetic aperture radar data, and environmental sensing data are collected simultaneously in the following manner: Visible light images are acquired by an inspection drone equipped with a multispectral gimbal camera on a preset track with sub-meter positioning accuracy. Its imaging band covers the 400-900nm visible light and near-infrared spectral range. At the same time, synthetic aperture radar data is acquired by X-band polarimetric interferometry radar on the same platform to penetrate surface obstructions. The radar incident angle is set to an adjustment range of 25°-50° to adapt to different terrain elevation angles. Environmental sensing data is monitored in real time by a network of sensor nodes buried in the soil around the pipeline. The sampling parameters include at least three key physical quantities: temperature, conductivity, and pH value. All acquired data streams are aligned to a unified timestamp and geographic coordinate frame through the BeiDou spatiotemporal reference system. When the data is transmitted to the edge computing node, a fusion preprocessing process is completed, including POS information association, radar echo intensity correction, and environmental parameter outlier filtering, to finally form a three-dimensional spatiotemporal synchronous data packet that meets the reception conditions of step one. The specific procedure for registration is as follows: First, a feature tensor is constructed. For visible light images, the spatial gradient tensor is calculated to enhance the image's edge features. The Sobel operator is used to perform convolution operations on the image to obtain the gradient information of each pixel. The calculation formula is as follows: ; in, Represents the spatial gradient tensor of a visible light image. Indicates the image in color channels The gradient (in red, green, or blue) represents the edge intensity of the image. This represents the convolution operator. This represents the Sobel operator kernel, used to calculate the gradient of an image and emphasize image edges. This represents the squaring operation, used to calculate the intensity of the gradient and enhance edge features in an image; For synthetic aperture radar (SAR) data, different polarization characteristics are described by extracting scattering features and calculating polarization scattering entropy. These scattering features are used to further analyze the physical information in the image, and their calculation formula is as follows: ; in, This represents the scattering characteristics (such as scattering entropy) in synthetic aperture radar data at location. calculate, Represents the polarization scattering matrix of synthetic aperture radar data The tensor representation contains multiple polarization channels (such as HH, HV, VH, VV). This represents the conjugate transpose operation, which transposes the matrix and takes its conjugate. The operator or weight, depending on the polarization decomposition model, is used to map the scattering matrix to certain polarization features (such as scattering entropy). ,in, Represents the scattering matrix of the first The probability distribution of each polarization component (e.g., HH, HV, VH, etc.). This represents a base-3 logarithmic operation, commonly used in polarization entropy calculations to measure the uncertainty or mixing degree of the scattering matrix. This formula calculates the entropy value of the scattering matrix, reflecting the complexity and information content of the scattering characteristics under different polarization states. Next, a cross-modal deformation energy function is established, which consists of three parts: feature consistency constraint, thin plate spline constraint, and thermal deformation compensation. Its expression is as follows: ; in, A function representing the energy of cross-modal deformation. It is a two-dimensional non-rigid transformation field used to describe the deformation when aligning two data sources (visible light image and PolSAR image). This indicates the scattering characteristics of SAR data. Mapping to an optical feature space is achieved using a Wishart likelihood classifier, which converts the polarization features of SAR into features in the visible light image space. This represents the encoding of gradient directions in a visible light image, typically using a histogram of oriented gradients (HOG) to capture edge features of the image. This represents the polarization scattering features (such as scattering entropy) extracted from synthetic aperture radar data, which are then registered with features from a visible light image. This represents the spatial gradient features extracted from a visible light image, used to represent edge and texture information in the image. This indicates that the field will be transformed. Applied to HOG features, it describes the result of performing a non-rigid transformation on the features of an image. This represents the weight of the feature consistency constraint, which is used to balance the matching accuracy of features from different data sources. To represent a thin-plate spline constraint, the transformed field is required. The second-order gradient in space (i.e., the curvature of the deformation) should be as smooth as possible. This constraint helps to avoid excessive discontinuities or sharp deformations during registration. The weights representing the thin-plate spline constraints determine the priority of the transformed field smoothness. This represents the thermal deformation compensation term, and represents the temperature. For deformation field This item is designed to compensate for the thermal expansion and contraction effects of metals or materials caused by temperature changes, thereby ensuring accurate modeling of deformation, especially during the deformation process of materials or metals. This represents the amount of temperature change, indicating the compensating effect of temperature change on the deformation length. The weights of thermal deformation compensation are used to control the contribution of temperature compensation; feature consistency constraints ensure feature alignment between PolSAR and visible light images; thin plate spline constraints are used to smooth the transformation field and avoid excessive deformation; and thermal deformation compensation improves accuracy by compensating for the metal stretching effect caused by temperature gradient. In the optimization phase, an improved Alternating Direction Multiplier Method (ADMM) algorithm is used to iteratively solve the two-dimensional non-rigid transformed field to optimize the matching process of cross-modal deformation. This optimization process continuously updates the transformed field to ultimately achieve accurate data fusion, and its expression is as follows: ; in, Indicates the first In the next iteration, the transformed field The value of represents the transformation field, which is a solution to the objective function and indicates the deformation of image registration in the current iteration. Represents the transformation of image data. Indicates the first The target data in the next iteration This represents the regularization parameter, used to balance the different terms in the optimization objective and help stabilize the optimization process. Representing the Lagrange multipliers, used for handling constraint terms. The Lagrange multipliers are updated at each iteration to ensure that the constraints are satisfied. Indicates the first The transformation field in the next iteration represents the transformation effect of the current iteration step. This represents the changes that occur in the transformed field during the iteration process, and is used to update the Lagrange multipliers; Finally, the optimized transformation field was used to generate a three-dimensional corrosion feature base film, the expression of which is: ; in, This represents the output three-dimensional base film, i.e., in spatial position. The base film generated at this location is established through the fusion of deformation and image features. The real part of the complex scattering feature of the SAR image is represented here. Represents the transformed field sar features inverse transform, Representing the real part of a complex number, SAR images typically contain complex scattering features, reflecting different scattering and polarization information. The imaginary part of the complex scattering characteristics of a SAR image is represented. The imaginary part of a complex number is taken. In SAR data, the imaginary part reflects phase information, which is crucial for deformation modeling. The curvature feature of a visible light image is used to capture changes in edges or textures within the image. Curvature can be calculated using the determinant (det(Hessian)) of the image's Hessian matrix. Representing curvature features, it reveals morphological details of the image, such as curves and surface variations. This base film combines the real and imaginary parts of the complex scattering features of SAR data with the curvature features of visible light images, thus effectively representing the corrosion status. The core innovation of this method lies in the introduction of a thermal deformation Jacobian term to compensate for the deformation caused by the temperature gradient, thereby improving the accuracy of corrosion detection and assessment.

[0021] In this embodiment, the specific operation of step S2, which generates the corrosion probability distribution map of the occluded area, is as follows: A1. First, the characteristics of the observation area of ​​the base film (i.e., the three-dimensional corrosion characteristics of the base film) will be analyzed. As input, this feature includes depth information and spatial coordinates in the image. These features help determine which regions in the image have unreliable depth information or are occluded. Secondly, occluded regions in the image are identified, specifically those regions with unreliable depth features. This is done by calculating the gradient of the base film in the depth dimension. If the gradient value is less than a preset threshold (0.05), these regions are marked as occluded regions. The expression is as follows: ; The expression defines an occlusion region, where It is the gradient of the base film in the depth dimension (representing the degree of depth change). If this gradient is less than a threshold... This means that the depth information for that area is unreliable and it is an occluded area. This represents a preset threshold of 0.05, used to determine the reliability of depth information. If the gradient is less than 0.05, the depth information is considered unreliable. This represents the depth dimension characteristics of the base film, i.e., the data in the depth direction of the three-dimensional etched base film; A2. Parameters from environmental sensors (like temperature, Corrosive medium density, pH value) and characteristic parameters of pipe materials (like Electrochemical potential, Iron ion activity coefficient The corrosion rate constant is fused, where the set of environmental sensor parameters describes the influence of environmental conditions on corrosion, and the set of pipe material characteristic parameters describes the influence of the material's inherent properties on corrosion. These parameters provide the necessary physical constraints for the subsequent generation of corrosion features; A3. Next, a two-channel network is constructed to generate erosion feature maps. This network consists of three modules: Feature encoder: It fuses the input base film features (including depth data with coordinates) and environmental sensor data to generate a feature vector containing spatial correlation information. Its expression is: ; in, This represents the spatial correlation feature vector output by the feature encoder, with dimension 1. That is, a 256-dimensional feature vector. This represents a convolutional gated recurrent unit, a model that combines convolutional neural networks and gated recurrent units for extracting spatiotemporal features. These represent the three-dimensional corrosion characteristic base films. exist Observational data in the direction, This represents an environmental sensing parameter vector, which includes environmental information such as temperature, density of corrosive media, and pH value. Physical constraint: By processing material property parameters and environmental sensitivity to corrosion rates, it generates a physical constraint output, the expression of which is: ; in, This represents the physical constraint information output by the physical constraint controller. This represents a multilayer perceptron, a type of feedforward neural network used to map input features to an output space. This represents a vector of pipe material property parameters, including electrochemical potential, iron ion activity coefficient, and corrosion rate constant. This represents a concatenation operation, which joins two vectors or matrices together. Representation matrix The diagonal matrix here It is the corrosion rate environmental sensitivity matrix, defined as: This represents the effect of environmental sensing parameters (temperature, density, pH) on the corrosion rate constant. Sensitivity; Corrosion generator: Combining random noise, encoded feature vectors, and the output of the physical constraint generator, it generates the final corrosion feature map, expressed as: in, This represents the final generated corrosion probability distribution map in the occluded area. Corrosion prediction on This represents the generator function used to generate the final erosion image. Describe a standard normal distribution The noise vector sampled in the middle is used to generate randomness in the adversarial network. This represents the feature vector output by the feature encoder. This represents the physical constraint information output by the physical constraint controller. This represents a vector concatenation operation; During training, the generator's output is optimized through adversarial training, with the training objective being to minimize two loss functions: Dual-discrimination loss: This loss function is optimized based on the difference between the generated image and the real image, encouraging the generator to produce more realistic erosion features. Its expression is: ; in, This represents the adversarial loss function, used to train the adversarial relationship between the generator and the discriminator. This represents the expected value, i.e., the discriminator's response to the input erosion image. The goal of the discrimination result is to enable the discriminator to better distinguish between real and generated images. This represents the expected value, i.e., the discriminator's assessment of the generated image. The goal of the discrimination result is to enable the generator to produce more realistic images. This represents the image generated by the generator. This represents a noise vector sampled from a standard normal distribution. The output image of the corrosion generator (i.e., the corrosion probability distribution map); Physical loss: This loss measures the difference between the generated corrosion characteristics and the actual corrosion driving force field, ensuring that the generated characteristics conform to physical laws. The corrosion driving force field is related to the iron ion concentration gradient and pH value changes, and its expression is: in, This represents the physics regularization loss, used to incorporate physics knowledge into the generator training process. Image generated by erosion For the input spatial location The gradient represents the rate of spatial change of the image. This represents the corrosion-driving field, located as follows: ,in, Represents the gas constant. Indicates temperature. This represents the iron ion diffusion coefficient, used to describe the ability of iron ions to diffuse in a material (material-dependent). Characteristic of corrosion base film exist The second-order gradient in the spatial dimension represents the spatial variation of corrosion characteristics; A4. After training and optimization, the final output is an accurate erosion probability distribution map. This map represents the erosion probability of each region (including occluded regions) in the image, and has high accuracy and stability. The output format is as follows: ; in, of This represents a location point in the image. Representing a specific region or set, usually referring to the "occupied region," that is, the area in an image or space where erosion occurs, meaning when the location... When located within the occupied area, the generated probability distribution map It depends on the maximum value in the formula; Represents generator Given input noise The generated image This represents the noise input for each Monte Carlo sampling. This indicates the number of Monte Carlo sampling operations, which means ensuring the probabilistic stability of the generated image by sampling the noisy space multiple times. This means using 128 Monte Carlo samplings to ensure the stability of the results. This is achieved by sampling multiple times and averaging (or selecting the maximum value) to reduce noise and random fluctuations, making the generated image more consistent with the actual probability distribution. Indicates the generated corrosion-characteristic base film In position The value, Indicates in In the Monte Carlo sampling, the maximum value is selected as the output to enhance the stability and reliability of the results. Indicates when position Not part of the occupied area At that time, the generated probability distribution map Take another value, which typically represents the probability distribution of the non-corroded area; The specific steps for performing a spacetime extrapolation operation are as follows: A1. Obtain the corrosion probability distribution map output in step S2. As the initial field for spatial corrosion depth, the pipeline stress tensor is also loaded. and loading historical corrosion time series library ,in, Indicates the horizontal stress in the pipe. This indicates the vertical stress in the pipe. This represents the shear stress in the pipeline; these stress values ​​affect the progression and distribution of corrosion in the corrosion model. This indicates a time-series library containing historical corrosion data. Indicates the location and time Corrosion characteristics below, Indicates the initial time. Indicates the time step. Indicates the number of time steps. This represents the total number of time steps; A2. Establish a graph structure consisting of corrosion feature nodes, stress feature nodes, and diffusion channel edges, where: Corrosion feature nodes Bind spatial coordinates, carrying the corrosion probability value of the corresponding location. and its historical change gradient ; Stress characteristic nodes Carrying equivalent stress value This value is calculated based on the horizontal principal stress, vertical stress, and shear stress components. diffusion channel edge The weights are determined by a curvature-weighted function of the distance between adjacent nodes and the difference in equivalent stress, at the edge of the diffusion channel. Weights are defined as: ; in, This represents connecting nodes in a graph model. and The edges are used to describe the interaction between corrosion, stress, and diffusion. This indicates the material-related corrosion correlation scale, used to describe the propagation range of corrosion effects within a material; a larger scale indicates a more significant corrosion impact. This indicates that the corrosion effect can affect a larger area. Represents a node and The Euclidean distance between two nodes measures the spatial distance between them. The position curvature coupling factor measures the curvature difference between locations. This factor affects how stress changes couple with the diffusion process. Represents a node and The difference in equivalent stress between them, which takes into account the influence of stress in different directions, affects the corrosion process. This represents the activation function, used to compress a value into the range [0,1]. Here, it is used to adjust the effect of stress differences on the diffusion channel weights. A3. Input the graph structure into a graph neural network, and generate a corrosion depth prediction field for future time periods through iterative solution of the corrosion-stress-diffusion coupling equation. Spatially discretize and encode the corrosion depth prediction field for the current time step, and after superimposing the exponentially decaying memory effect of the historical prediction field, output a corrosion propagation path heatmap. The formula for the corrosion-stress-diffusion coupling equation is: ; in, Indicates corrosion depth or corrosion state, used to describe the depth of corrosion at a specific location and time. Represents effective stress, used to describe the stress state inside a material during corrosion, and affects the corrosion rate. This represents a constant related to stress corrosion behavior, describing the effect of stress corrosion on the corrosion rate. This represents a nonlocal diffusion kernel function, describing how the corrosion state at a given point is affected by the surrounding region. ,in, This represents the position vector, which is the distance between two points. The covariance matrix represents the anisotropy of diffusion, controlling the behavior of the diffusion process in different directions, and ,in Indicates in The diffusion coefficient in the direction describes the anisotropic diffusion characteristics of a material. The stress-diffusion coupling factor describes the coupling strength between the stress field and the diffusion process, indicating how stress affects diffusion. Representation matrix The determinant is used to normalize the kernel function. Representation matrix The inverse matrix is ​​used to calculate the direction and intensity of diffusion. This represents a spatial location variable, i.e., different locations within the distribution of corrosion states. This represents the entire computational region. The hydrogen diffusion coefficient describes the rate at which hydrogen diffuses through a material. This represents the stress field of hydrogen. The stress effect of hydrogen on materials is usually caused by the accumulation of hydrogen. The gradient operator is used to describe the spatial variation of hydrogen diffusion or stress field. This constant represents hydrogen-assisted diffusion and controls the effect of hydrogen diffusion on the corrosion process. The expression for the spatiotemporal propagation operator of a graph neural network is: ; in, Represents a node In the The feature vector of the layer, Represents a node In the The feature vector of the layer, This represents a gated recurrent unit, a neural network structure used for sequential data. Here, it's used to capture the spatiotemporal relationships between nodes. Represents a node In the set of neighboring nodes, that is, with All directly connected nodes, The transition function represents the edge's characteristics, used to describe the influence of the characteristics of neighboring nodes on the characteristics of the current node. The weight of an edge is used to describe a node. and nodes The strength of the connection between them; This represents the hyperbolic tangent function, used to compress input values ​​to the range [-1, 1]. This represents the weight matrix, used to map the combination of stress gradient and time difference to a suitable space. This represents the product of stress gradient and corrosion depth, used to describe the influence of stress variations within the material on the corrosion depth. ,in It represents the stress gradient, describing how stress changes with location within a material. Indicates the progress of corrosion. This represents the rate of change of stress with respect to corrosion depth, used to describe how corrosion affects stress in a material. Indicates the time step, indicating the time from time The time interval is used for integration calculations; Represents a node and nodes The time difference between them Represents the weight matrix, used for weighting... Perform a linear transformation. This represents the kernel function, used to describe the effect of the distance difference between nodes on the corrosion state. This represents element-wise multiplication; The heatmap representation of corrosion propagation paths is as follows: ; in, The heatmap showing the corrosion propagation path is located at... and time step The value of is used to describe the corrosion propagation state at that location at that moment. Indicates spatial location, This indicates the predicted time step, i.e., the future moment predicted during grinding. This represents the total number of time steps (or the number of historical time steps), referring to the current moment. The previous historical period, Indicates time step Time position The corrosion depth or corrosion state is used as a characteristic quantity to describe the corrosion process of the material at each time step. This represents an exponential decay function, meaning that the influence of historical corrosion conditions gradually weakens over time. It is the decay coefficient, which controls how the influence of corrosion history decays over time. The larger the value, the faster the decay. Indicates the prediction time step The corresponding time, Indicates historical time step The corresponding time, This represents the Sigmoid activation function, used to compress the calculation results to the range of 0 to 1, making it suitable for representing the intensity of the heatmap. Used to smooth the output and map it to the [0,1] interval; The corrosion-stress-diffusion coupling equation includes: Stress corrosion acceleration term based on material stress sensitivity coefficient; Nonlocal spatial diffusion integral term containing anisotropic kernel function; Stress-assisted diffusion term driven by hydrogen ion concentration gradient.

[0022] In this embodiment, the specific steps to construct the multi-objective optimization function, namely step S4, are as follows: A1. First, receive new inspection data. ,in, This represents the new inspection dataset, containing observations of the material and the current time. This indicates the spatial location of each point in the inspection data. Indicates the location The corrosion data values ​​observed above, Indicates the current time and simultaneously generates a heatmap of the corrosion propagation path. This indicates that based on the predicted time The generated heatmap of corrosion propagation path, To determine the location, the spatiotemporal alignment difference tensor is calculated by comparing the current corrosion heatmap with historical data. The formula for calculating the spatiotemporal alignment difference tensor is: in, This represents the spatiotemporal alignment difference tensor, used to describe erosion changes at different points in time. This represents the aligned time, calculated as the current time. Subtract the initial time , The gradient over time relates to the rate of change of time. The gradient of the heatmap representing the corrosion propagation path with respect to location describes the spatial distribution of corrosion variation. This represents the spatial gradient of the observed corrosion data. Indicates curvature sensitivity factor (inheritance step S3) This affects the response of the corrosion model in different spatial regions. This difference tensor reflects the corrosion changes of the material at different time points, and gradient calculations are used to optimize model predictions; A2. Define an innovative third-order optimization objective function, the expression of which is: ; in, This represents the overall optimization objective function, and multi-objective optimization is performed by weighting different sub-objective functions. The weighting coefficients represent the geometric topology loss, controlling its impact on the overall optimization objective. The weighting coefficients represent the corrosion risk functional, controlling the impact of corrosion risk on the overall optimization objective. The weighting coefficients represent the memory entropy and control its impact on the overall optimization objective. This includes three main sub-items: Geometric topological loss: measures the change in corrosion morphology of a material, using gradient and tensor operations to describe the material's changes. Its expression is: ; in, This represents the geometric topology loss function, used to measure the geometric shape changes of the erosion propagation path. These represent constants related to the loss function, used to adjust the shape of the function. This represents a topological tensor used to quantify the difference between predicted and measured geometric features, and ;in This represents the spatiotemporal alignment difference tensor, i.e., the spatial difference between the CPH heatmap and the measured values. Represents the horizontal principal stress gradient. Hessian represents the horizontal principal stress. Represents a tensor product, representing two tensors (e.g., and The product of ) Let represent the norm (i.e., the magnitude) of the cross product of two gradients, and This indicates the gradient of the corrosion prediction thermogram. This represents the measured corrosion depth gradient. This represents the defined area, that is, the spatial area on the material surface. This represents the exponential decay function, used to describe the effects of corrosion over time. This represents the attenuation coefficient, used to control sensitivity to geometric differences; Corrosion risk functional: This function calculates the weighted sum of material corrosion risks, predicting future corrosion risks based on changes in stress depth at different time points. Its expression is as follows: in, This represents a corrosion risk functional used to calculate the corrosion risk at different locations. This represents the set of key points, specifically the coordinates of high-risk locations that require close monitoring. This represents the stress sensitivity function, i.e., the second-order mixed partial derivative of the equivalent stress with respect to time and space. Represents equivalent stress, and ,in Indicates the predicted start time. Indicates the predicted end time. Represents the spatiotemporal evolution rate of stress. The spatiotemporal alignment difference tensor represents the erosion changes at different points in time. Indicates the location The risk indicator function is used to describe the risk under specific conditions; Memory entropy: Used to estimate the learning ability of a neural network from past experience, reflecting the model's memory and adaptive capabilities. Its expression is: in, Memory entropy measures the ability of a neural network to learn and remember past data. The trace operation on a matrix yields the sum of the elements on the diagonal. This represents the weight matrix of a neural network, used to describe the connection strength of each layer in the network. This represents the historical weight matrix of the neural network, used to calculate the network's learning memory. This represents an exponential function that describes the effect of the gradient on memory entropy. It is the memory decay factor. This represents the gradient of the difference field, used to describe the spatial rate of change of the corrosion depth difference. Represents the L2 norm, Represents logarithmic operations, used to describe a measure of the difference in weight distribution; The dual-network coupling mechanism is as follows: Two neural networks work together, with parameters updated via a gradient projection operator. A Jacobi response tensor is introduced, and the model is further optimized by calculating the relationship between network weights and the response to erosion prediction. This process involves dynamic adjustment of the neural network weight matrix to optimize erosion prediction. The gradient projection operator is calculated using the following formula: ; in, Indicates the weight update amount. This represents the learning rate, with values ​​ranging from 1 to 2. , This represents the gradient of the total loss function with respect to the weights. This represents the Hadamard product (element-by-element multiplication). Indicates the adaptive mask strength coefficient. Represents the Jacobian response tensor; The expression for the Jacobian response tensor is: in, This represents the weights of the generative adversarial network. Represents the weights of a graph neural network. Let represent the norm of the erosion propagation rate vector, and ; The triggering condition for dynamic updates is: If the Frobenius norm of the weight update is greater than the product of the set threshold and the failure factor, then a network parameter update is triggered; otherwise, it is not triggered. The expression is as follows: ; in, The Frobenius norm represents the weight update and is used to measure the degree of change in the weight matrix. This indicates a pre-defined threshold used to determine whether an update should be triggered. When the size of the weight update exceeds this threshold, it means that the network weights need to be updated. This represents the failure factor, used to adjust the update triggering conditions. It is calculated by taking corrosion risk into account. To dynamically adjust the triggering conditions, specifically, it represents a probability factor that controls when an update is triggered, and ,in This represents the natural exponential function. This represents the cumulative effect of corrosion risk over time. The expression measures the risk of corrosion in a given area. The rate of change of internal corrosion risk, when integrated, represents the cumulative effect of the risk. This represents the rate of change of the risk function with respect to time, reflecting the sensitivity of corrosion risk to time. The generated corrosion risk decision matrix is ​​constructed in the following way; A1. Decision Matrix Framework: Construct a two-dimensional decision matrix consisting of m rows and n columns of decision units, where each decision unit is bound to preset spatial coordinates, and its expression is: ; A2. Decision Unit Calculation: The value of each decision unit is generated by the core decision function, which is executed in the following order: Calculate the third mixed partial derivatives of the corrosion propagation thermogram with respect to specified spatial coordinates, time variables, and horizontal stress; After multiplying the above partial derivatives by the risk accumulation factor and inputting them into the activation function for normalization, the expression is as follows: ; in, This represents the value of the decision matrix unit (i.e., the position of the pipeline). (corrosion risk level). This represents the third-order mixed partial derivative of the corrosion propagation thermogram with respect to spatial coordinates / time / stress. This indicates activation function normalization. This represents the optimized corrosion prediction heatmap (the spatiotemporal distribution field of corrosion depth output in step S3). Spatial coordinate index: Pipe surface index Each normalized grid position (inherited from step S1) system), Represents time variables (points on the prediction timeline). It represents the horizontal principal stress (the horizontal component of the duct stress tensor). This represents the risk accumulation factor (an enhancement coefficient composed of stress gradient and failure time). Furthermore, if... If , it indicates a safe zone (no intervention required). If it is, it indicates a high-risk area (requiring urgent maintenance); A3. Risk Factor Construction: The risk accumulation factor consists of the product of two items: Stress deep coupling term: After performing a series multiplication operation on the first, second and third gradient magnitudes of the equivalent force field, the hyperbolic tangent function output value is taken; Failure time gain term: A natural exponential function with the critical failure time as the power, where the exponent coefficient is a preset material sensitivity constant, and its expression is: ; in, This represents the hyperbolic tangent function, used for nonlinear adjustment of risk. It smooths the accumulation of stress gradients. This represents the gradient product of effective stress, reflecting the spatial distribution of stress variations. Generally, a larger stress gradient indicates a higher risk of corrosion. This represents an exponential function, indicating the failure time gain. It is related to the failure time. The correlation reflects the cumulative risk effect of the corrosion process over time; A4. Matrix Output: The values ​​of all decision units form a corrosion risk decision matrix, and the row and column dimensions of this matrix are consistent with the pipeline space segmentation.

[0023] It should be noted that the descriptions of each embodiment in the above embodiments have different focuses. For parts that are not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0024] 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.

[0025] 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 computer, 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. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0026] 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.

[0027] 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.

[0028] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0029] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A method for predicting the corrosion rate of natural gas pipelines, characterized in that, Specifically, the steps include the following: Step S1: When the visible light image, synthetic aperture radar data and environmental sensing data are received, a cross-modal deformation energy function registration operation is performed on the texture gradient of the visible light image and the polarization scattering matrix of the synthetic aperture radar to generate a three-dimensional corrosion feature base film covering the pipe surface. Step S2: When there is an environmental occlusion area in the three-dimensional corrosion feature base film, the observation area features of the base film are used as input conditions, combined with environmental sensing parameters and pipeline material characteristic parameters, and a physical constraint generative adversarial network is used to generate a corrosion probability distribution map of the occlusion area. Step S3: When the pipeline stress distribution data and historical corrosion data are obtained, the corrosion probability distribution map is used as the initial field. A graph neural network is used to couple the corrosion-stress-diffusion equation to perform a spatiotemporal extrapolation operation and output a heat map of the corrosion propagation path for a future preset time period. Step S4: When new inspection data is input, a multi-objective optimization function is constructed based on the difference between the predicted value and the measured value. The weight parameters of the generative adversarial network and the graph neural network are dynamically updated to generate a corrosion risk decision matrix.

2. The method for predicting the corrosion rate of a natural gas pipeline according to claim 1, characterized in that: The visible light image, synthetic aperture radar data, and environmental sensor data in step one are acquired simultaneously using the following methods: Visible light images are acquired by an inspection drone equipped with a multispectral gimbal camera on a preset track. Its imaging band covers the 400-900nm visible light and near-infrared spectral range. At the same time, synthetic aperture radar data is acquired by an X-band polarimetric interferometric radar on the same platform, penetrating ground obstructions. The radar incident angle is set to an adjustment range of 25°-50° to adapt to different terrain elevation angles. Environmental sensing data is monitored in real time by a network of sensor nodes buried in the soil around the pipeline. Its sampling parameters include at least three key physical quantities: temperature, conductivity, and pH value.

3. The method for predicting the corrosion rate of a natural gas pipeline according to claim 2, characterized in that: The specific procedure for the registration operation is as follows: First, feature tensors are constructed. For visible light images, spatial gradient tensors are calculated to enhance the edge features of the image. The Sobel operator is used to perform convolution operations on the image to obtain the gradient information of each pixel. For synthetic aperture radar data, scattering features are extracted and polarimetric scattering entropy is calculated to describe different polarization features. Next, a cross-modal deformation energy function is established, which consists of three parts: feature consistency constraint, thin plate spline constraint, and thermal deformation compensation. In the optimization phase, an improved alternating direction multiplier method algorithm is used to iteratively solve the two-dimensional non-rigid transformation field in order to optimize the matching process of cross-modal deformation. Finally, the optimized transform field was used to generate a three-dimensional corrosion feature base film that combines the real and imaginary parts of the complex scattering features of SAR data with the curvature features of the visible light image.

4. The method for predicting the corrosion rate of a natural gas pipeline according to claim 3, characterized in that: In step S2, the specific operation for generating the corrosion probability distribution map of the occluded area is as follows: A1. First, the observation area features of the base film are used as input. These features include depth information and spatial coordinates in the image. Second, the occluded areas in the image are identified, specifically those areas whose depth features are unreliable. By calculating the gradient of the base film in the depth dimension, if the gradient value is less than a preset threshold, these areas are marked as occluded areas. A2. Integrate environmental sensor parameters and pipe material characteristic parameters; A3. Next, a two-channel network is constructed to generate erosion feature maps. This network consists of three modules: Feature encoder: It fuses the input base film features and environmental sensor data to generate a feature vector containing spatial correlation information; Physical constraint: By processing material property parameters and environmental sensitivity to corrosion rate, it generates a physical constraint output; Corrosion generator: Combines random noise, encoded feature vectors, and the output of the physical constraint to generate the final corrosion feature map; During training, the generator's output is optimized through adversarial training, with the training objective being to minimize two loss functions: Dual-discrimination loss: This loss function is optimized by the difference between the generated image and the real image, encouraging the generator to produce more realistic erosion features; Physical loss: This loss measures the difference between the generated corrosion characteristics and the actual corrosion driving force field, which is related to the iron ion concentration gradient and pH changes; A4. After training and optimization, an accurate corrosion probability distribution map is finally output.

5. The method for predicting the corrosion rate of a natural gas pipeline according to claim 4, characterized in that: The specific steps for performing the spatiotemporal deduction operation are as follows: A1. Use the corrosion probability distribution map output in step S2 as the initial field of spatial corrosion depth, and simultaneously load the pipeline stress tensor and the historical corrosion time series library. A2. Establish a graph structure consisting of corrosion feature nodes, stress feature nodes, and diffusion channel edges, where: The corrosion feature nodes are bound to spatial coordinates, carrying the corrosion probability value at the corresponding location and its historical change gradient; The stress characteristic node carries an equivalent stress value, which is calculated based on the horizontal principal stress, vertical stress, and shear stress components. The edge weights of the diffusion channel are determined by a curvature weighting function of the distance between adjacent nodes and the difference in equivalent stress. A3. Input the graph structure into the graph neural network, generate the corrosion depth prediction field for future time periods through iterative solution of the corrosion-stress-diffusion coupling equation, spatially discretize and encode the corrosion depth prediction field at the current time step, and output the corrosion expansion path heat map after superimposing the exponential decay memory effect of the historical prediction field.

6. The method for predicting the corrosion rate of a natural gas pipeline according to claim 5, characterized in that: The corrosion-stress-diffusion coupling equation includes: Stress corrosion acceleration term based on material stress sensitivity coefficient; Nonlocal spatial diffusion integral term containing anisotropic kernel function; Stress-assisted diffusion term driven by hydrogen ion concentration gradient.

7. The method for predicting the corrosion rate of a natural gas pipeline according to claim 6, characterized in that: In step S4, the specific operation of constructing the multi-objective optimization function is as follows: A1. First, new inspection data is received and a corrosion extension path heat map is generated simultaneously. Then, by comparing the current corrosion heat map with historical data, the spatiotemporal alignment difference tensor is calculated and gradient calculation is used to optimize model prediction. A2. Define an innovative third-order optimization objective function, which includes three main sub-terms: Geometric topology loss: measures the changes in the corrosion morphology of a material, using gradient and tensor operations to describe the changes in the material; Corrosion risk functional: Calculates the weighted sum of material corrosion risk and predicts future corrosion risk based on stress depth changes at different time points; Memory entropy: Used to estimate the learning ability of a neural network based on past experience, reflecting the model's memory and adaptive capabilities.

8. The method for predicting the corrosion rate of a natural gas pipeline according to claim 7, characterized in that: The dual-network coupling mechanism is as follows: By utilizing two neural networks working together and updating parameters through a gradient projection operator, while introducing a Jacobian response tensor, the model is further optimized by calculating the relationship between network weights and the response of erosion prediction.

9. The method for predicting the corrosion rate of a natural gas pipeline according to claim 8, characterized in that: The triggering condition for the dynamic update is: If the Frobenius norm of the weight update is greater than the product of the set threshold and the failure factor, then the network parameter update is triggered; otherwise, it is not triggered.

10. The method for predicting the corrosion rate of a natural gas pipeline according to claim 9, characterized in that: The generated corrosion risk decision matrix is ​​constructed in the following way; A1. Decision Matrix Framework: Construct a two-dimensional decision matrix consisting of m rows and n columns of decision units, where each decision unit is bound to a preset spatial location coordinate; A2. Decision Unit Calculation: The value of each decision unit is generated by the core decision function, which is executed in the following order: Calculate the third mixed partial derivatives of the corrosion propagation thermogram with respect to specified spatial coordinates, time variables, and horizontal stress; After multiplying the above partial derivatives by the risk accumulation factor, the result is input into the activation function for normalization. A3. Risk Factor Construction: The risk accumulation factor consists of the product of two items: Stress deep coupling term: After performing a series multiplication operation on the first, second and third gradient magnitudes of the equivalent force field, the hyperbolic tangent function output value is taken; Failure time gain term: a natural exponential function with the critical failure time as the exponent, where the exponent coefficient is a preset material sensitivity constant; A4. Matrix Output: The values ​​of all decision units form a corrosion risk decision matrix, and the row and column dimensions of this matrix are consistent with the pipeline space segmentation.

Images