Oil and gas pipeline leakage risk assessment method and system based on multi-source data fusion

Abstract

The application discloses a kind of oil and gas pipeline leakage risk assessment method and system based on multi-source data fusion, and relates to oil and gas pipeline leakage assessment technical field. Including: obtaining pipeline multi-source heterogeneous monitoring data and carrying out space-time alignment;Based on the control equation of fluid mechanics, a lower physical model with residual minimization is constructed, and an upper anti-fusion optimization model with maximum feature distribution distance is constructed;The double-layer model is solved by joint iteration, and the optimal data fusion weight matrix and the optimal physical state variable are obtained;Accordingly, the model extreme value condition feature is calculated and extracted, and the holographic leakage criticality index is generated;Combined with dynamic decision strategy, risk assessment result is obtained.The application deeply integrates physical priori law and anti-learning, actively amplifies the abnormal features caused by weak leakage, and realizes high sensitivity, high precision early leakage risk quantitative evaluation and dynamic accurate early warning of oil and gas pipeline.

Description

Technical Field

[0001] This invention relates to the field of oil and gas pipeline leakage assessment technology, and more specifically, to a method and system for assessing oil and gas pipeline leakage risk based on multi-source data fusion. Background Technology

[0002] As vital arteries for long-distance energy transportation, oil and gas pipelines play a crucial strategic role in ensuring energy supply and protecting the ecological environment through their safe and stable operation. With the rapid development of the Internet of Things (IoT) and sensor technologies, modern oil and gas pipelines are equipped with a vast number of monitoring devices that can collect real-time data on internal fluid dynamics such as pressure and flow, as well as external environmental disturbances such as soil stress, vibration, and sound waves. To accurately extract subtle signals characterizing pipeline health from this massive amount of high-frequency monitoring data, the industry has widely adopted multi-source data fusion technology and risk assessment algorithms. These aim to align and map spatiotemporal data from different dimensions by constructing deep mathematical or physical models, thereby achieving a comprehensive understanding of the pipeline's operational status. Based on this, establishing risk assessment models to dynamically quantify the potential leakage criticality of pipelines under complex operating conditions has become a core research direction in the construction of intelligent pipeline operation and maintenance and safety assurance systems.

[0003] Currently, existing pipeline leakage risk assessment models still face significant limitations when dealing with complex operating conditions. On the one hand, mainstream multi-source data fusion methods mostly rely on purely data-driven black-box models or heuristic single-layer weight allocation strategies. The passive response fusion mechanism lacks deep constraints on the physical nature of pipeline fluids, making it easily masked by complex environmental noise when facing early, minor data fluctuations caused by weak leaks. On the other hand, current risk quantitative index systems not only over-rely on empirical thresholds or simple probabilistic statistical models, but also fail to effectively mine the underlying mathematical extreme value characteristics when physical equations conflict with measured data. This makes the existing technology severely lacking in assessment sensitivity in the early stages of minor leak evolution, making it difficult to achieve adaptive identification and high-precision early warning of potential risk states. Summary of the Invention

[0004] The purpose of this invention is to provide a method and system for assessing oil and gas pipeline leakage risk based on multi-source data fusion. It is based on a dual-layer adversarial fusion model driven by physical data, extracts extreme conditions to generate a holographic leakage criticality index, which not only overcomes the shortcomings of traditional methods in lacking physical constraints, but also achieves highly sensitive and accurate quantification and early warning of minute leaks.

[0005] This invention is achieved through the following technical solution:

[0006] A method for assessing oil and gas pipeline leakage risk based on multi-source data fusion, the method comprising the following steps:

[0007] Acquire multi-source heterogeneous monitoring data of oil and gas pipelines during operation;

[0008] Based on the aforementioned multi-source heterogeneous monitoring data, a lower-level physical state estimation model is constructed using fluid dynamics control equations, with the goal of minimizing the state estimation residual.

[0009] Based on the state estimation results output by the lower-layer physical state estimation model, an upper-layer adversarial fusion optimization model is constructed with the goal of maximizing the feature distribution distance.

[0010] The lower-level physical state estimation model is combined with the upper-level adversarial fusion optimization model, and the optimal data fusion weight matrix and the corresponding optimal physical state variables are obtained by joint iterative solution in combination with preset physical and operational boundary constraints.

[0011] Based on the optimal data fusion weight matrix and the optimal physical state variables, the extreme value condition features at the convergence of the model solution are calculated and extracted to generate a holographic leakage criticality index that characterizes the quantitative leakage risk.

[0012] Based on the holographic leakage criticality index and combined with the preset dynamic judgment strategy, the leakage risk assessment result of the oil and gas pipeline is obtained.

[0013] Optionally, acquire multi-source heterogeneous monitoring data of oil and gas pipelines during operation, including:

[0014] Acquire internal fluid state data and external environmental disturbance data of oil and gas pipelines;

[0015] The internal fluid state data and the external environmental disturbance data are projected into a unified manifold space using the Riemannian manifold mapping algorithm;

[0016] Based on the propagation delay of transient fluid waves between nodes in the manifold space, an adjacency matrix with spatiotemporal translation properties is constructed to complete the spatiotemporal manifold alignment of multi-source heterogeneous monitoring data.

[0017] Optionally, a lower-level physical state estimation model is constructed using the fluid dynamics governing equations, with the objective of minimizing the state estimation residual. The objective function of the lower-level physical state estimation model is:

[0018]

[0019] in, The objective function of the lower-level physical state estimation model is... For physical state variables, Let be the physical state estimation matrix for the i-th node; Let be the weight of the i-th node in the data fusion weight matrix; It is a fluid dynamics control mapping operator for embedding partial differential equations of pipeline mass conservation and momentum conservation; For pipeline boundary control variables; The Frobenius norm is used to characterize the residuals of the state estimation. For total variational regularization; It is the first regularization coefficient; This represents the total number of nodes.

[0020] Optionally, an upper-layer adversarial fusion optimization model is constructed with the objective of maximizing the feature distribution distance. The objective function of the upper-layer adversarial fusion optimization model is:

[0021]

[0022] in, The objective function of the upper-level adversarial fusion optimization model is... For data fusion weight matrix; The optimal physical state variables obtained by solving the lower-level physical state estimation model; For based on Constructed current state feature probability distribution; The prior feature probability distribution of historically unleashed healthy samples; Here is the p-th order Wasserstein distance function used to characterize the distance of the feature distribution; This is the second smoothing coefficient.

[0023] Optional, preset physical and operational boundary constraints include:

[0024] Upper-level weight boundary constraints: the sum of the elements of the data fusion weight matrix of all nodes is equal to 1, and the weight value of a single node is greater than or equal to 0.

[0025] The lower-level physical boundary constraint states that the absolute value of the pressure wave propagation speed in the optimal physical state variables is less than or equal to the sound speed propagation limit of the fluid in the pipeline.

[0026] Optionally, a joint iterative solution can be performed by combining preset physical and operational boundary constraints, including:

[0027] The Lagrangian function of the lower-level physical state estimation model is constructed, and the corresponding first-order necessary condition for KKT optimality is derived.

[0028] The first necessary condition for optimality of KKT is used as a complementary relaxation equation constraint and substituted into the upper-level adversarial fusion optimization model to transform the two-level optimization model into a mathematical programming model with balance constraints.

[0029] The mathematical programming model with equilibrium constraints is solved iteratively using a smoothed nonlinear programming algorithm to obtain the optimal data fusion weight matrix and the optimal physical state variables when Nash equilibrium is reached.

[0030] Optionally, the extreme value condition features at the convergence of the extracted model are calculated to generate a holographic leakage criticality index that quantitatively characterizes leakage risk, including:

[0031] Extract the extremum condition features of the mathematical programming model with balance constraints during convergence, wherein the extremum condition features include: the Lagrange multiplier vectors corresponding to the constraints. And the residual gradient vector of the lower-level physical state estimation model with respect to the measured data. ;

[0032] The Lagrange multiplier vector, residual gradient vector, and optimal data are fused into a weight matrix. The holographic leakage criticality index is obtained by performing combined calculations.

[0033] Optionally, the Lagrange multiplier vector, residual gradient vector, and optimal data fusion weight matrix are used. The holographic leakage criticality index is calculated using the following formula:

[0034]

[0035] in, The holographic leakage threshold index. For matrix trace operations; and These are the preset calibration coefficients; This is the Sigmoid nonlinear normalization function.

[0036] Optionally, based on the holographic leakage criticality index and a preset dynamic judgment strategy, the leakage risk assessment result of the oil and gas pipeline is obtained, including:

[0037] The holographic leakage criticality index under multiple consecutive time windows is extracted according to the time series to construct a one-dimensional time series;

[0038] The one-dimensional time series is reconstructed into a high-dimensional phase space trajectory matrix using the time delay embedding theorem;

[0039] Calculate the maximum Lyapunov exponent of the high-dimensional phase space trajectory matrix;

[0040] If the maximum Lyapunov exponent is greater than a preset threshold, it is determined that the state sequence of the oil and gas pipeline has undergone chaotic bifurcation, indicating a risk of leakage. The pipeline node corresponding to the maximum residual gradient vector is then backtracked to generate the leakage source location result.

[0041] A risk assessment system for oil and gas pipeline leaks based on multi-source data fusion includes:

[0042] The data acquisition module is used to acquire multi-source heterogeneous monitoring data of oil and gas pipelines during their operation.

[0043] The lower-level model construction module is used to construct a lower-level physical state estimation model based on the multi-source heterogeneous monitoring data and using fluid dynamics control equations with the goal of minimizing the state estimation residual.

[0044] The upper-layer model construction module is used to construct an upper-layer adversarial fusion optimization model with the goal of maximizing the feature distribution distance based on the state estimation results output by the lower-layer physical state estimation model.

[0045] The model solving module is used to combine the lower-level physical state estimation model with the upper-level adversarial fusion optimization model, and perform joint iterative solving in combination with preset physical and operational boundary constraints to obtain the optimal data fusion weight matrix and the corresponding optimal physical state variables.

[0046] The index calculation module is used to calculate the extreme value condition features when the model converges based on the optimal data fusion weight matrix and the optimal physical state variables, and generate a holographic leakage criticality index that characterizes the quantitative leakage risk.

[0047] The risk assessment module is used to obtain the leakage risk assessment result of the oil and gas pipeline based on the holographic leakage criticality index and a preset dynamic judgment strategy.

[0048] The technical solution of the present invention has at least the following advantages and beneficial effects:

[0049] This invention uses the physical state estimation of the embedded fluid dynamics control equations as a lower-level follower to minimize the state residuals, and the adversarial fusion that maximizes the feature distribution distance as an upper-level leader to actively optimize weights, thereby forcing the model to evolve in the direction that best exposes the characteristics of weak leaks. Based on the two-layer model, this invention not only creatively extracts the KKT Lagrange multipliers and residual gradients of the model in the Nash equilibrium state to generate a brand-new holographic leak criticality index, but also combines the Lyapunov exponent determination in the high-dimensional phase space to achieve accurate quantification starting from the mathematical bottom-level extreme conditions, which greatly improves the early detection sensitivity and positioning accuracy of pipeline micro-leaks. Attached Figure Description

[0050] Figure 1 A flowchart illustrating the oil and gas pipeline leakage risk assessment method based on multi-source data fusion provided by this invention;

[0051] Figure 2This is a schematic diagram illustrating the principle of the oil and gas pipeline leakage risk assessment system based on multi-source data fusion provided by the present invention. Detailed Implementation

[0052] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0053] To address the technical problems of traditional oil and gas pipeline risk assessment models, such as lack of physical constraints, extremely low sensitivity to minor leaks, and difficulty in achieving accurate quantification in the very early stages, this invention proposes an oil and gas pipeline leak risk assessment method based on multi-source data fusion. The implementation details of the oil and gas pipeline leak risk assessment method based on multi-source data fusion in this embodiment are described below. The following content is only for ease of understanding and is not necessary for implementing this solution.

[0054] Example 1

[0055] The oil and gas pipeline leakage risk assessment method based on multi-source data fusion in this embodiment can be applied to electronic devices or cloud server clusters with communication, computing, and data storage capabilities. Its specific process includes:

[0056] Step 110: Obtain multi-source heterogeneous monitoring data of oil and gas pipelines in operation.

[0057] Specifically, multi-source heterogeneous monitoring data refers to diverse information collected in real time by various distributed sensor networks deployed along the entire operating cycle of oil and gas pipelines. Acquiring multi-source heterogeneous monitoring data can be achieved using industrial IoT technology and high-speed data acquisition systems for precise extraction. Pressure transmitters, flow meters, and temperature sensors deployed along the pipeline can continuously record the dynamic physical parameters of the fluid inside the pipeline at millisecond or second-level sampling frequencies. Simultaneously, fiber optic acoustic wave sensing devices, distributed acoustic sensing systems, and stress-strain gauges laid outside the pipeline can accurately capture the micro-vibration waveforms of the surrounding soil and the mechanical deformation state of the pipe wall. These data exhibit significant differences in temporal sampling rate, spatial resolution, and physical dimensions, forming a raw data lake for assessing the health status of the pipeline.

[0058] Step 120: Based on the multi-source heterogeneous monitoring data, a lower-level physical state estimation model is constructed using fluid dynamics control equations with the goal of minimizing the state estimation residual.

[0059] Specifically, the governing equations of fluid dynamics are mathematical expressions that characterize the essential laws governing fluid motion within a pipeline, primarily encompassing the laws of conservation of mass and momentum. Under normal, leak-free conditions, the flow and pressure fluctuations upstream and downstream of the pipeline strictly follow the set of partial differential equations for transient fluid flow. The core idea of ​​constructing the lower-level physical state estimation model is to use these objectively existing physical laws as strict a priori constraints, forcibly regulating the trajectory of data changes. In model construction, the pipeline is discretized into multiple physical nodes, and the state of each node is predicted using fluid control operators. The predicted values ​​are then subtracted from the multi-source heterogeneous monitoring data actually collected by sensors to calculate the state estimation residuals. The goal of the lower-level model is to find a set of optimal physical state variables, given weights, that minimizes the state estimation residuals of the entire pipeline network. This is equivalent to finding a pipeline operating state evolution path that best conforms to the physical laws of fluid dynamics. If a minor leak occurs in the pipeline, the actual physical state of the fluid will deviate from the theoretical governing equations, inevitably leading to the inability to completely eliminate local residuals.

[0060] Step 130: Based on the state estimation results output by the lower-layer physical state estimation model, construct an upper-layer adversarial fusion optimization model with the objective of maximizing the feature distribution distance.

[0061] Specifically, the feature distribution distance is used to measure the statistical difference between the currently estimated probability distribution of state features and the prior feature probability distribution of historical, leak-free healthy samples. This embodiment aims to allocate data fusion weights through an adversarial, proactive optimization mechanism. Traditional weight allocation is often static or heuristic, easily masking weak anomalous signals. The upper-layer adversarial fusion optimization model, acting as the leader in the game mechanism, receives the state estimation results passed up from the lower layer and calculates the distribution distance between the current state features and the healthy state features. Its optimization objective is to adjust the fusion weights of each data node and sensing dimension to maximize this feature distribution distance. In other words, the upper-layer model deliberately seeks out and amplifies those tiny anomalous fluctuations that deviate from the normal healthy state. By maximizing the feature distribution distance, the model can extremely sensitively capture local data anomalies caused by weak leakage and masked by massive environmental noise.

[0062] Step 140: Combine the lower-level physical state estimation model with the upper-level adversarial fusion optimization model, and perform joint iterative solution based on preset physical and operational boundary constraints to obtain the optimal data fusion weight matrix and the corresponding optimal physical state variables.

[0063] Specifically, physical and operational boundary constraints refer to the extreme safety limits and physical limitations that pipelines must meet in actual industrial operations, such as the maximum output pressure of pumping stations and the yield strength of pipeline materials. Combining the upper and lower level models essentially constitutes a two-level Stackelberg game architecture. The upper level model guides the physical estimation direction of the lower level by distributing data fusion weights, while the lower level model strictly follows the physical control equations for state optimization under given weights and feeds back the state results to the upper level. Introducing pre-defined physical and operational boundary constraints during the solution process ensures that the state variables searched in each iteration have real physical meaning. After repeated joint iterative game, when the objectives of the upper and lower level models can no longer be compromised, the system reaches a mathematical Nash equilibrium. At this point, the output optimal data fusion weight matrix represents the feature focusing method most sensitive to the current pipeline anomaly, while the corresponding optimal physical state variables represent the set of physical quantities that best approximate the real fluid state under this focusing method.

[0064] Step 150: Based on the optimal data fusion weight matrix and the optimal physical state variables, calculate and extract the extreme value condition features when the model solution converges, and generate a holographic leakage criticality index that characterizes the quantitative leakage risk.

[0065] Specifically, the extreme condition features at model convergence refer to the gradient and multiplier information at the underlying mathematical programming level when a bi-level optimization problem reaches its optimal solution. When the pipeline is in an absolutely healthy state, the fluid dynamics equations can be perfectly closed, and the model does not need to pay an additional penalty to compensate for the violation of the physical equations. At this time, the extreme condition features are at an extremely low level. However, once a slight leak occurs in the pipeline, the loss of mass will cause an endogenous contradiction in the physical equations. In order to achieve the optimal solution in the game, the optimization solver will generate a huge Lagrange multiplier mutation on the mathematical constraints corresponding to the leak point, accompanied by a sharp reversal of the residual gradient. These purely mathematical underlying extreme condition features are extracted and mathematically transformed and fused with the optimal data fusion weight matrix to generate a holographic leak criticality index. This index completely abandons the traditional scoring method that relies on the fluctuation of surface data, and directly quantifies the risk from the underlying mathematical mutations of the optimization model. Therefore, it has an extremely amazing amplification effect on early leaks.

[0066] Step 160: Based on the holographic leakage criticality index and combined with the preset dynamic judgment strategy, obtain the leakage risk assessment result of the oil and gas pipeline.

[0067] Specifically, the dynamic judgment strategy refers to examining the dynamic evolution of the holographic leakage criticality index over time, rather than relying on manually set fixed empirical dead zones. The continuously generated holographic leakage criticality index is input into the preset dynamic judgment strategy for topological analysis. Once the intrinsic structure of the judgment index sequence undergoes a fundamental change, a risk assessment result indicating pipeline leakage is output, along with the corresponding leak source location information.

[0068] In summary, the oil and gas pipeline leakage risk assessment method based on multi-source data fusion provided in this embodiment acquires multi-source heterogeneous data on pipeline operation and constructs a lower-level physical state model based on fluid mechanics to minimize residuals, ensuring that the assessment process strictly adheres to the essence of fluid physics. Based on the lower-level results, an upper-level adversarial model is constructed to maximize distribution distance, which can significantly amplify the abnormal characteristics of early-stage minor leaks. Combining these two approaches for joint solution with physical constraints forms a two-layer game architecture, overcoming the limitations of passive fitting in traditional single-layer black-box models. Most importantly, the extreme value condition features at model convergence are extracted to generate a holographic leakage criticality index. Starting from the extreme value mutation phenomenon at the bottom layer of mathematical optimization, this achieves a highly sensitive and revolutionary quantitative perception of minor leak risks. Finally, a reliable risk assessment result is output through a dynamic judgment strategy.

[0069] In some optional embodiments, acquiring multi-source heterogeneous monitoring data of oil and gas pipelines in operation includes: acquiring internal fluid state data and external environmental interference data of the oil and gas pipelines; projecting the internal fluid state data and the external environmental interference data onto a unified manifold space using a Riemannian manifold mapping algorithm; and constructing an adjacency matrix with spatiotemporal translation attributes based on the propagation delay of fluid transient waves between nodes in the manifold space to complete the spatiotemporal manifold alignment of the multi-source heterogeneous monitoring data.

[0070] Specifically, internal fluid state data is typically low-frequency time-series data, while external environmental disturbance data is typically high-frequency vibration waveform data. These two types of data cannot be directly compared point-to-point in Euclidean space. This embodiment employs the Riemannian manifold mapping algorithm, treating data with different physical dimensions and sampling frequencies as tangent bundle vectors on the manifold, mapping them to a high-dimensional, unified non-Euclidean manifold space. Within this manifold space, the data is stripped of the constraints of its original physical dimensions, retaining only its geometric and topological features. Furthermore, considering the time required for fluid pressure waves to propagate within the pipeline—that is, a sensing delay exists for downstream sensors after an upstream disturbance—the propagation delay of transient fluid waves is used to determine the spatiotemporal correlation between data, constructing an adjacency matrix with spatiotemporal translation properties. This adjacency matrix guides the time axis alignment of each data node within the manifold space, completely eliminating spatiotemporal misalignment caused by different sensor deployment locations and wave velocity limitations.

[0071] In some optional embodiments, in the construction of a lower-level physical state estimation model based on the multi-source heterogeneous monitoring data and using fluid dynamics control equations to minimize the state estimation residual, the objective function of the lower-level physical state estimation model is:

[0072]

[0073] in, The objective function of the lower-level physical state estimation model is... For physical state variables, Let be the physical state estimation matrix for the i-th node; The data fusion weight matrix is ​​passed to the upper layer; It is a fluid dynamics control mapping operator for embedding partial differential equations of pipeline mass conservation and momentum conservation; For pipeline boundary control variables; The Frobenius norm is used to characterize the residuals of the state estimation. For total variational regularization; It is the first regularization coefficient; This represents the total number of nodes.

[0074] Specifically, The control mapping operator is the physical core of this embodiment. It maps the state of the upstream node i-1 of the pipeline to the boundary control variables of the current pump station or valve. By solving the hydraulic transient equations, the theoretical physical state of downstream node i is derived. The actual physical state to be estimated is... The difference between this theoretically derived value and the actual value represents the residual that violates the laws of physics. It is measured using the squared Frobenius norm and weighted by the upper layers. Weighted amplification is applied. Simultaneously, to prevent drastic, non-physical oscillations in the estimated state, a total variational regularization term is introduced. As a smoothing constraint, the first regularization coefficient. This is used to balance the weight of the physical residual term and the smoothing term. In this embodiment, by minimizing this objective function, the physical state variables are forced to... The system must conform to the pipeline fluid dynamics equations to the greatest extent possible, ensuring that every state estimate does not violate the fundamental physical laws.

[0075] In some optional embodiments, based on the state estimation results output by the lower-layer physical state estimation model, an upper-layer adversarial fusion optimization model is constructed with the objective of maximizing the feature distribution distance. The objective function of the upper-layer adversarial fusion optimization model is:

[0076]

[0077] in, The objective function of the upper-level adversarial fusion optimization model is... For data fusion weight matrix; The optimal physical state variables obtained by solving the lower-level physical state estimation model; For based on Constructed current state feature probability distribution; The prior feature probability distribution of historically unleashed healthy samples; Here is the p-th order Wasserstein distance function used to characterize the distance of the feature distribution; This is the second smoothing coefficient.

[0078] Specifically, It is a baseline probability model established by collecting data over a long period of time during which the pipeline is confirmed to be operating safely and stably without leaks. This is based on the lower-level model's current weights. The latest state is obtained after making efforts to conform to physical laws. The p-order Wasserstein distance function is used to measure the difference between the two distributions because the Wasserstein distance can measure the optimal transmission cost between two non-overlapping distributions, providing smooth and non-zero gradient guidance even when the anomalous distribution changes extremely slowly in the early stages of leakage. The latter part of the objective function is the information entropy regularization term. In this embodiment, by maximizing this objective function, the upper-layer model will actively adjust and search for a set of extreme weights. This intentionally amplifies the deviation between the current state and the healthy state. An information entropy regularization term is added, and a second smoothing coefficient is assigned. This effectively prevents the model from concentrating all weights on a single anomalous noise point, thus balancing the sensitivity against amplification with the robustness of spatial distribution.

[0079] In some optional embodiments, the preset physical and operational boundary constraints include: upper-level weight boundary constraints, where the sum of the elements of the data fusion weight matrix of all nodes is equal to 1, and the weight value of a single node is greater than or equal to 0; and lower-level physical boundary constraints, where the absolute value of the pressure wave propagation velocity in the optimal physical state variables is less than or equal to the sound speed propagation limit of the fluid in the pipeline, and the pipe wall stress satisfies the Von Mises yield criterion.

[0080] Specifically, the solution to the two-layer model must be constrained within a reasonable space. The upper-layer weight boundary constraints define the physical normalization properties of the weight allocation, ensuring that the total energy or information remains constant. The lower-layer physical boundary constraints, on the other hand, are directly based on engineering practice. Based on the physical limitations of fluid transients, they set an absolute limit for pressure variation along the pipe space, i.e., the fluid sound velocity. Exceeding this limit indicates that the result violates physical causality. Simultaneously, the pipe wall stress satisfies the Von Mises yield criterion, ensuring that the internal fluid pressure estimated by the model will not cause plastic deformation of the steel pipe—a fundamental engineering principle. By setting these stringent boundaries, the generation of mathematically feasible but physically absurd spurious solutions during adversarial iterations is completely eliminated.

[0081] In some optional embodiments, the lower-level physical state estimation model is combined with the upper-level adversarial fusion optimization model, and a joint iterative solution is performed using preset physical and operational boundary constraints to obtain the optimal data fusion weight matrix and the corresponding optimal physical state variables. This includes: constructing the Lagrangian function of the lower-level physical state estimation model and deriving the corresponding KKT optimality first-order necessary condition; substituting the KKT optimality first-order necessary condition as a complementary relaxation equality constraint into the upper-level adversarial fusion optimization model to transform the two-level optimization model into a mathematical programming model with balance constraints; and using a smoothed nonlinear programming algorithm to perform a joint iterative solution on the mathematical programming model with balance constraints to obtain the optimal data fusion weight matrix and the optimal physical state variables when a stable solution satisfies a preset convergence criterion.

[0082] Specifically, the bi-level optimization model is mathematically a nondeterministic polynomial hard problem, and direct solutions are prone to getting stuck in local deadlocks. Therefore, the classic Lagrange multiplier method is introduced to transform the lower-level physical control equations and inequality constraints into unconstrained Lagrangian function forms, and then differentiate them to obtain the first-order necessary condition for KKT optimality. This step transforms the lower-level optimization process into a set of complementary relaxation equations with deterministic algebraic relationships. Subsequently, this set of equations is directly implanted into the upper-level model, replacing the original lower-level objective function. This operation flattens the nested bi-level structure into a single-level large-scale mathematical programming model with balance constraints. Since the complementary relaxation conditions contain highly nonlinear absolute values ​​or product zeros, the problem is extremely non-smooth. This embodiment uses a smoothed nonlinear programming algorithm, using exponential or logarithmic approximation functions to perform minor smoothing of non-smooth inflection points, followed by a gradient descent-based joint iterative solution. When the gradient of the objective function converges to the set tolerance range, the system reaches Nash equilibrium. The resulting weight matrix and state variables are the absolute optimal solution that takes into account both the worst-case amplification of anomalies and the most stringent physical constraints.

[0083] In some optional embodiments, based on the optimal data fusion weight matrix and the optimal physical state variables, the extreme value condition features at the convergence of the model are calculated to generate a holographic leakage criticality index characterizing the quantitative leakage risk. This includes: extracting the extreme value condition features at the convergence of the mathematical programming model with balance constraints, wherein the extreme value condition features include the Lagrange multiplier vectors corresponding to the constraints and the residual gradient vectors of the lower-level physical state estimation model with respect to the measured data; and calculating the holographic leakage criticality index by combining the Lagrange multiplier vectors, the residual gradient vectors, and the optimal data fusion weight matrix using the following formula:

[0084]

[0085] in, The holographic leakage threshold index. For matrix trace operations; and These are the preset calibration coefficients; This is the Sigmoid nonlinear normalization function.

[0086] Specifically, in mathematical optimization theory, the Lagrange multiplier vector This represents the sensitivity of the objective function to constraints. When the pipeline is leak-free, the physical equations perfectly match the measured data, requiring no large external force to be applied to the constraints, and the multipliers are extremely small. However, once a leak, even as small as a pinhole, occurs, the measured fluid data will strongly clash with the mass conservation control equations, making the model extremely difficult to solve the constraints at that node, and the corresponding Lagrange multipliers will increase significantly. This will cause a sudden surge. At the same time, in order to balance this contradiction, the residual gradient vector will also fluctuate drastically within its trace space.

[0087] In this embodiment, by calculation and The Hadamard product causes the sensitive weights actively sought by the upper layer to directly amplify the abrupt changes in the multipliers. This is then combined with the inner product trace operation of the residual gradient matrix itself. This process transforms the multidimensional mathematical extremum anomaly into a scalar. After adjusting the weights of the two components using preset calibration coefficients, a sigmoid nonlinear normalization function is used to map them to the interval between zero and one, forming the final holographic leakage criticality index. This index completely breaks away from dependence on superficial physical values ​​such as pressure reduction, directly targeting the underlying mathematical signs of system collapse, thus greatly improving the lead time and sensitivity of warnings.

[0088] In some optional embodiments, obtaining the leakage risk assessment result of the oil and gas pipeline based on the holographic leakage criticality index and a preset dynamic judgment strategy includes: extracting the holographic leakage criticality index under multiple consecutive time windows according to the time series to construct a one-dimensional time series; reconstructing the one-dimensional time series into a high-dimensional phase space trajectory matrix using the time delay embedding theorem; calculating the maximum Lyapunov exponent of the high-dimensional phase space trajectory matrix; if the maximum Lyapunov exponent is greater than a preset threshold, it is determined that the state sequence of the oil and gas pipeline has undergone chaotic bifurcation, indicating a leakage risk, and backtracking to generate the pipeline node corresponding to the maximum residual gradient vector, outputting the leakage source location result.

[0089] Specifically, traditional threshold alarm mechanisms often set an absolute value, making them highly susceptible to false alarms caused by normal start-stop fluctuations in operating conditions. This embodiment introduces chaos theory from nonlinear dynamics. Based on the time-delay embedding theorem, it transforms one-dimensional... The time series is reconstructed according to a certain embedding dimension and time delay, unfolded into a high-dimensional phase space trajectory matrix that reflects the overall dynamic behavior of the pipeline system. The maximum Lyapunov exponent is used to measure the separation rate of adjacent trajectories in the phase space over time. When the pipeline is healthy or subject to normal stationary disturbances, the system dynamics exhibit convergent or periodic trajectories, and the maximum Lyapunov exponent is negative or close to zero. Once a minor leak occurs in the pipeline, the fluid dynamic equilibrium within the pipe is disrupted, and the state evolution produces chaotic bifurcation. At this time, adjacent trajectories will exponentially repel each other and rapidly separate, causing the maximum Lyapunov exponent to turn from negative to positive and sharply exceed a preset threshold. In this embodiment, this is used as the ultimate criterion for determining the occurrence of a deterministic leak. This not only completely eliminates the interference of normal operating condition fluctuations but also allows for tracing back to the geographic coordinate node in the underlying optimization model that generates the maximum residual gradient, achieving accurate location of the leak source with meter-level precision.

[0090] Example 2

[0091] Based on the above embodiments, this embodiment provides a specific field application example of oil and gas pipelines and proposes a complete industrial implementation workflow for an oil and gas pipeline leakage risk assessment method based on multi-source data fusion.

[0092] A major long-distance crude oil transportation artery in western my country traverses hundreds of kilometers of complex hilly and Gobi desert terrain. This crude oil pipeline is characterized by its large diameter, high transport pressure, high fluid viscosity, and dramatic elevation changes along its route. Due to the pipeline's passage through extensive, highly corrosive saline-alkali soil, there is a potential risk of polarization point corrosion perforation in the pipe walls. Traditional detection systems based on mass balance and simple negative pressure waves are extremely insensitive to pinhole leaks of a few hundred liters per hour when dealing with total flow rates of thousands of cubic meters per hour. These leaks are often masked by normal pressure fluctuations from pumping stations along the route, leading to extremely serious soil contamination from crude oil infiltration. Therefore, there is an urgent need to establish a highly sensitive risk assessment model system that integrates the latest physical and mathematical frameworks.

[0093] The specific industrialization process of the oil and gas pipeline leakage risk assessment method based on multi-source data fusion includes:

[0094] High-precision fiber optic Fabry-Perot pressure sensors, Coriolis mass flow meters, and distributed acoustic fiber optic monitoring equipment are deployed on a large scale at various monitoring valve chambers and pump station inlets and outlets along the crude oil pipeline. At a sampling rate of fifty times per second, multi-source heterogeneous monitoring data, including internal pressure, instantaneous flow rate, and micro-vibration frequency around the pipe wall, are acquired 24 / 7 during pipeline operation. This data is then aggregated in real-time to the control center server via an industrial ring network.

[0095] The control center server initiates the Riemannian manifold mapping module, projecting data with different physical dimensions, such as megapascals, cubic meters per hour, and hertz, onto a multidimensional manifold space. Based on the actual sound velocity of crude oil under the given temperature and pressure conditions, it calculates the propagation delay of transient waves in the fluid between stations. An adjacency matrix with spatiotemporal translation properties is established, ensuring that tens of thousands of data points within a 300-kilometer range upstream and downstream are tightly aligned under a unified spatiotemporal reference, forming a high-resolution data snapshot available for computation.

[0096] Based on the aligned data, a lower-level physical state estimation model is constructed using the one-dimensional simplified form of the classic hydraulic Navier-Stokes control equations for pipes, namely the continuity equation and the momentum equation. Sensor-measured pressure and flow rates are substituted into the control equations for derivation. The objective function is set to minimize the Frobenius norm residuals between the derivation and measured states at all computational nodes. Simultaneously, total variational smoothing of the pipe network is incorporated as a physical regularization constraint.

[0097] Extract the current state feature distribution derived from the lower layer and calculate its Wasserstein distance with the feature distribution of historical leak-free healthy samples formed during the past month of safe pipeline operation. Construct an upper-layer adversarial fusion optimization model, the goal of which is to actively find a weight allocation method for the data dimensions that maximizes and widens this Wasserstein distance, thereby making potential anomalies impossible to hide.

[0098] The actual maximum allowable operating pressure of the pipeline, the yield stress limit of the pipeline steel, and the requirement for weight normalization are introduced as physical and operational boundary constraints. Lagrange expansions are performed on the lower-level model to derive the KKT conditions, which are then transformed into complementary relaxation equations. These equations are then substituted into the upper-level model to generate a nonlinear mathematical programming problem with equilibrium constraints.

[0099] The model is solved iteratively using the server-embedded interior-point method and smoothed nonlinear programming algorithm. After hundreds of gradient searches and weight game updates, the model reaches Nash equilibrium convergence. The Lagrange multiplier vectors of each node and the residual gradient vectors of the lower layers are extracted at convergence.

[0100] By performing a Hadamard product between the Lagrange multipliers and the optimal weights obtained from the upper layer, and then combining this with the trace operation of the residual gradient, the holographic leakage criticality index (HLCI) at the second level is calculated and generated. This continuous risk index curve, derived from the underlying mathematical extremes, begins to be plotted in real time on the monitoring screen.

[0101] During the subsequent three days of operation, a very minor flange leak occurred in a section of the pipeline due to geological subsidence. Traditional pressure drop monitoring showed no change in the alarm threshold because the leak amount was less than one ten-thousandth of the total flow. However, the HLCI sequence calculated by the server exhibited abnormal fluctuations. Based on the time-delay embedding theorem, the system reconstructed the one-dimensional HLCI time series of the past ten minutes into a ten-dimensional phase space trajectory matrix and calculated its maximum Lyapunov exponent in real time.

[0102] Calculations revealed that the maximum Lyapunov exponent dramatically reversed from its normal value of -0.5 within seconds and exceeded the preset threshold of +0.2. The system determined that the current state evolution of the oil and gas pipeline had undergone a deterministic chaotic bifurcation and immediately issued a red-level warning for a minor leak.

[0103] The system quickly traced back to the pipeline spatial node identifier corresponding to the maximum residual gradient vector generated within the current computation cycle, accurately marking the weld location 123.5 kilometers from the pumping station on the 3D map of the geographic information system, and outputting a high-precision leak source location result. Inspection personnel promptly arrived at the scene based on the location, successfully preventing an early incident that could have escalated into a large-scale crude oil leak from developing.

[0104] In summary, this embodiment, using a complex scenario of minor leakage in an industrial crude oil pipeline as its implementation background, systematically demonstrates the entire process from high-precision acquisition of multi-source data, manifold spatiotemporal alignment, construction and solution of a two-layer adversarial model, to extracting underlying extreme conditions for chaotic dynamics judgment. Field application verification shows that this method not only successfully captures minor leaks that traditional fixed-threshold models cannot detect, but also, thanks to the underlying mathematical sensitivity of the maximum Lyapunov exponent and Lagrange multipliers, completely eliminates false alarms caused by normal valve regulation and pump shutdowns. This not only provides a revolutionary mathematical quantitative basis for the inherent safety protection of ultra-long oil and gas pipelines, but also greatly improves the safety monitoring technical specifications for the integrated geological engineering of the oil and gas storage and transportation industrial Internet of Things.

[0105] Example 3

[0106] like Figure 2 As shown, another embodiment of this application relates to an oil and gas pipeline leakage risk assessment system based on multi-source data fusion. The implementation details of the oil and gas pipeline leakage risk assessment system based on multi-source data fusion in this embodiment are described in detail below. The following implementation details are provided for ease of understanding and are not necessary for implementing this solution. The oil and gas pipeline leakage risk assessment system based on multi-source data fusion in this embodiment includes a data acquisition module 510, a lower-level model construction module 520, an upper-level model construction module 530, a model solving module 540, an index calculation module 550, and a risk assessment module 560.

[0107] Data acquisition module 510 is used to acquire multi-source heterogeneous monitoring data of oil and gas pipelines in operation.

[0108] The lower-level model construction module 520 is used to construct a lower-level physical state estimation model based on the multi-source heterogeneous monitoring data and using fluid dynamics control equations with the goal of minimizing the state estimation residual.

[0109] The upper-layer model construction module 530 is used to construct an upper-layer adversarial fusion optimization model with the goal of maximizing the feature distribution distance based on the state estimation results output by the lower-layer physical state estimation model.

[0110] The model solving module 540 is used to combine the lower-level physical state estimation model with the upper-level adversarial fusion optimization model, and perform joint iterative solving in combination with preset physical and operational boundary constraints to obtain the optimal data fusion weight matrix and the corresponding optimal physical state variables.

[0111] The index calculation module 550 is used to calculate the extreme value condition features when the model solution converges based on the optimal data fusion weight matrix and the optimal physical state variables, and generate a holographic leakage criticality index that characterizes the quantitative leakage risk.

[0112] The risk assessment module 560 is used to obtain the leakage risk assessment result of the oil and gas pipeline based on the holographic leakage criticality index and a preset dynamic judgment strategy.

[0113] It is worth mentioning that all modules involved in this embodiment are logical modules. In practical applications, a logical unit can be a physical unit, a part of a physical unit, or a combination of multiple physical units. Furthermore, to highlight the innovative aspects of this application, this embodiment does not introduce units that are not closely related to solving the technical problems proposed in this application; however, this does not mean that other units are absent in this embodiment.

[0114] In some optional embodiments, the data acquisition module further includes: a spatiotemporal mapping alignment unit, used to project the internal fluid state data and the external environmental disturbance data into a unified manifold space through the Riemannian manifold mapping algorithm, and to construct an adjacency matrix with spatiotemporal translation properties based on the propagation delay of fluid transient waves.

[0115] In some optional embodiments, the model solving module further includes: an equilibrium solving unit, used to construct the Lagrangian function of the lower-level physical state estimation model, derive the corresponding KKT optimality first-order necessary condition, and substitute it into the upper-level adversarial fusion optimization model as a complementary relaxation equality constraint, and perform joint iterative solving using a smoothed nonlinear programming algorithm.

[0116] In some optional embodiments, the risk assessment module further includes: a phase space reconstruction discrimination unit, used to extract the holographic leakage criticality index under multiple consecutive time windows according to the time series, reconstruct it into a high-dimensional phase space trajectory matrix using the time delay embedding theorem, and determine whether the system state has undergone chaotic bifurcation and locate the leakage source node by calculating its maximum Lyapunov exponent.

[0117] Example 4

[0118] Another embodiment of this application relates to an electronic device, including: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, the instructions being executed by the at least one processor to enable the at least one processor to perform the oil and gas pipeline leakage risk assessment method based on multi-source data fusion in the above embodiments.

[0119] The memory and processor are connected via a bus, which can include any number of interconnecting buses and bridges, connecting various circuits of one or more processors and memories. The bus can also connect various other circuits, such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and will not be described further herein. The bus interface provides an interface between the bus and the transceiver. The transceiver can be a single element or multiple elements, such as multiple receivers and transmitters, providing a unit for communicating with various other devices over a transmission medium. Data processed by the processor is transmitted over the wireless medium via an antenna, which further receives data and transmits it to the processor.

[0120] The processor manages the bus and general processing, and also provides various functions, including timing, peripheral interfaces, voltage regulation, power management, and other control functions. Memory is used to store data used by the processor during operation.

[0121] Example 5

[0122] Another embodiment of this application relates to a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the method embodiments described above.

[0123] That is, those skilled in the art will understand that all or part of the steps in the methods of the above embodiments can be implemented by a program instructing related hardware. This program is stored in a storage medium and includes several instructions to cause a device, such as a microcontroller, chip, or processor, to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, portable hard drive, read-only memory, random access memory, magnetic disk, or optical disk.

[0124] Those skilled in the art will understand that the above embodiments are specific embodiments for implementing this application, and in practical applications, various changes can be made to them in form and detail without departing from the spirit and scope of this application.

[0125] The above are merely preferred embodiments of the present invention and are not intended to limit the present invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A method for assessing the risk of oil and gas pipeline leakage based on multi-source data fusion, characterized in that, The steps of this method include: Acquire multi-source heterogeneous monitoring data of oil and gas pipelines during operation; Based on the aforementioned multi-source heterogeneous monitoring data, a lower-level physical state estimation model is constructed using fluid dynamics control equations, with the goal of minimizing the state estimation residual. Based on the state estimation results output by the lower-layer physical state estimation model, an upper-layer adversarial fusion optimization model is constructed with the goal of maximizing the feature distribution distance. The lower-level physical state estimation model is combined with the upper-level adversarial fusion optimization model, and the optimal data fusion weight matrix and the corresponding optimal physical state variables are obtained by joint iterative solution in combination with preset physical and operational boundary constraints. Based on the optimal data fusion weight matrix and the optimal physical state variables, the extreme value condition features at the convergence of the model solution are calculated and extracted to generate a holographic leakage criticality index that characterizes the quantitative leakage risk. Based on the holographic leakage criticality index and combined with the preset dynamic judgment strategy, the leakage risk assessment result of the oil and gas pipeline is obtained.

2. The method for assessing oil and gas pipeline leakage risk based on multi-source data fusion according to claim 1, characterized in that, Acquire multi-source heterogeneous monitoring data of oil and gas pipelines during operation, including: Acquire internal fluid state data and external environmental disturbance data of oil and gas pipelines; The internal fluid state data and the external environmental disturbance data are projected into a unified manifold space using the Riemannian manifold mapping algorithm; Based on the propagation delay of transient fluid waves between nodes in the manifold space, an adjacency matrix with spatiotemporal translation properties is constructed to complete the spatiotemporal manifold alignment of multi-source heterogeneous monitoring data.

3. The method for assessing oil and gas pipeline leakage risk based on multi-source data fusion according to claim 1, characterized in that, A lower-level physical state estimation model is constructed using the fluid dynamics governing equations, with the objective of minimizing the state estimation residual. The objective function of the lower-level physical state estimation model is: in, The objective function of the lower-level physical state estimation model is... For physical state variables, Let be the physical state estimation matrix for the i-th node; Let be the weight of the i-th node in the data fusion weight matrix; It is a fluid dynamics control mapping operator for embedding partial differential equations of pipeline mass conservation and momentum conservation; For pipeline boundary control variables; The Frobenius norm is used to characterize the residuals of the state estimation. For total variational regularization; It is the first regularization coefficient; This represents the total number of nodes.

4. The method for assessing oil and gas pipeline leakage risk based on multi-source data fusion according to claim 3, characterized in that, A high-level adversarial fusion optimization model is constructed with the objective of maximizing the feature distribution distance. The objective function of the high-level adversarial fusion optimization model is: in, The objective function of the upper-level adversarial fusion optimization model is... For data fusion weight matrix; The optimal physical state variables obtained by solving the lower-level physical state estimation model; For based on Constructed current state feature probability distribution; The prior feature probability distribution of historically unleashed healthy samples; Here is the p-th order Wasserstein distance function used to characterize the distance of the feature distribution; This is the second smoothing coefficient.

5. The method for assessing oil and gas pipeline leakage risk based on multi-source data fusion according to claim 4, characterized in that, The preset physical and operational boundary constraints include: Upper-level weight boundary constraints: the sum of the elements of the data fusion weight matrix of all nodes is equal to 1, and the weight value of a single node is greater than or equal to 0. The lower-level physical boundary constraint states that the absolute value of the pressure wave propagation speed in the optimal physical state variables is less than or equal to the sound speed propagation limit of the fluid in the pipeline.

6. The method for assessing oil and gas pipeline leakage risk based on multi-source data fusion according to claim 5, characterized in that, The solution is obtained through a joint iterative process that combines pre-defined physical and operational boundary constraints, including: The Lagrangian function of the lower-level physical state estimation model is constructed, and the corresponding first-order necessary condition for KKT optimality is derived. The first necessary condition for optimality of KKT is used as a complementary relaxation equation constraint and substituted into the upper-level adversarial fusion optimization model to transform the two-level optimization model into a mathematical programming model with balance constraints. The mathematical programming model with equilibrium constraints is solved iteratively using a smoothed nonlinear programming algorithm to obtain the optimal data fusion weight matrix and the optimal physical state variables when Nash equilibrium is reached.

7. The method for assessing oil and gas pipeline leakage risk based on multi-source data fusion according to claim 6, characterized in that, The extreme value condition characteristics at the convergence of the extracted model are calculated to generate a holographic leakage criticality index that quantitatively characterizes leakage risk, including: Extract the extremum condition features of the mathematical programming model with balance constraints during convergence, wherein the extremum condition features include: the Lagrange multiplier vectors corresponding to the constraints. And the residual gradient vector of the lower-level physical state estimation model with respect to the measured data. ; The Lagrange multiplier vector, residual gradient vector, and optimal data are fused into a weight matrix. The holographic leakage criticality index is obtained by performing combined calculations.

8. The method for assessing oil and gas pipeline leakage risk based on multi-source data fusion according to claim 7, characterized in that, The Lagrange multiplier vector, residual gradient vector, and optimal data are fused into a weight matrix. The holographic leakage criticality index is calculated using the following formula: in, This is the holographic leakage threshold index. For matrix trace operations; and These are the preset calibration coefficients; This is the Sigmoid nonlinear normalization function.

9. The method for assessing oil and gas pipeline leakage risk based on multi-source data fusion according to claim 8, characterized in that, Based on the holographic leakage criticality index and a preset dynamic judgment strategy, the leakage risk assessment result of the oil and gas pipeline is obtained, including: The holographic leakage criticality index under multiple consecutive time windows is extracted according to the time series to construct a one-dimensional time series; The one-dimensional time series is reconstructed into a high-dimensional phase space trajectory matrix using the time delay embedding theorem; Calculate the maximum Lyapunov exponent of the high-dimensional phase space trajectory matrix; If the maximum Lyapunov exponent is greater than a preset threshold, it is determined that the state sequence of the oil and gas pipeline has undergone chaotic bifurcation, indicating a risk of leakage. The pipeline node corresponding to the maximum residual gradient vector is then backtracked to generate the leakage source location result.

10. A risk assessment system for oil and gas pipeline leakage based on multi-source data fusion, characterized in that, include: The data acquisition module is used to acquire multi-source heterogeneous monitoring data of oil and gas pipelines during their operation. The lower-level model construction module is used to construct a lower-level physical state estimation model based on the multi-source heterogeneous monitoring data and using fluid dynamics control equations with the goal of minimizing the state estimation residual. The upper-layer model construction module is used to construct an upper-layer adversarial fusion optimization model with the goal of maximizing the feature distribution distance based on the state estimation results output by the lower-layer physical state estimation model. The model solving module is used to combine the lower-level physical state estimation model with the upper-level adversarial fusion optimization model, and perform joint iterative solving in combination with preset physical and operational boundary constraints to obtain the optimal data fusion weight matrix and the corresponding optimal physical state variables. The index calculation module is used to calculate the extreme value condition features when the model converges based on the optimal data fusion weight matrix and the optimal physical state variables, and generate a holographic leakage criticality index that characterizes the quantitative leakage risk. The risk assessment module is used to obtain the leakage risk assessment result of the oil and gas pipeline based on the holographic leakage criticality index and a preset dynamic judgment strategy.

Images