A stray current corrosion risk dynamic prediction device and method based on digital twinning

Abstract

The application discloses a kind of stray current corrosion risk dynamic prediction device and method based on digital twinning, belong to underground metal structure corrosion protection technical field.The device includes edge monitoring node, digital twinning model construction module, stray current diffusion simulation module, LSTM (Long Short-Term Memory) time sequence prediction module, interference source intelligent classification and positioning module, three-dimensional visual display module, adaptive protection suggestion generation module and closed-loop feedback optimization module.The method is through multidimensional data acquisition, digital twinning modeling, stray current diffusion simulation, interference source classification and positioning, long time sequence risk prediction and closed-loop iteration optimization, realize the full-factor perception of underground pipe network stray current corrosion risk, high-precision simulation, long time sequence prediction, interference source accurate tracing and adaptive protection decision.The present application solves the problem of monitoring lag, low prediction accuracy, interference source positioning difficulty and passive protection in the existing stray current protection, the prediction accuracy is improved by more than 30% compared with traditional method, interference source positioning error ≤5%, can be applied to corrosion protection of urban underground pipe network, oil and gas long-distance pipeline, rail transit metal structure and power grounding network.

Description

Technical Field

[0001] This invention relates to the field of corrosion protection technology for underground metal structures, specifically to a device and method for dynamic prediction of stray current corrosion risk based on digital twins. Background Technology

[0002] Stray current refers to current flowing outside the designed circuit and is widely present in scenarios such as rail transportation, DC power transmission, electrolysis industry, and DC grounding. When stray current leaks into the underground soil, it can cause electrolytic corrosion to facilities such as buried metal pipelines, steel structures, and power grounding grids. It is characterized by rapid corrosion rate, strong concealment, wide range of damage, and great difficulty in tracing its source.

[0003] Existing stray current corrosion protection technologies have the following main drawbacks: (1) The monitoring methods are mainly single-point, static and delayed, which cannot achieve synchronous dynamic sensing of the entire pipeline network, all time period and multiple physical quantities, and it is difficult to capture transient stray current change signals. (2) Corrosion prediction relies on empirical formulas and short-term statistics, lacks long-term dynamic prediction capabilities, has low prediction accuracy and delayed early warning, and cannot achieve early protection. (3) Stray current interference sources cannot be automatically classified and accurately located, relying on manual investigation, which is inefficient, has large errors, and is difficult to suppress corrosion from the source; (4) The lack of three-dimensional visualization means that real-time mapping between physical entities and digital space makes it impossible for maintenance personnel to intuitively grasp the spatial distribution of current diffusion, potential distribution and corrosion risk. (5) Protective measures rely on human experience to formulate, which is passive, undifferentiated, and lacks closed-loop optimization, making it difficult to quantify and continuously improve the protective effect.

[0004] Existing technologies cannot meet the high-precision, intelligent, proactive, and visual requirements for stray current corrosion protection in urban underground spaces and long-distance energy pipelines. Summary of the Invention

[0005] The technical problem to be solved by this invention is that existing stray current corrosion protection technologies have technical defects such as monitoring lag, low prediction accuracy, difficulty in tracing the source of interference, lack of three-dimensional visualization, and passive protection measures.

[0006] To address the aforementioned technical problems, this invention provides a device and method for dynamic prediction of stray current corrosion risk based on digital twins, comprising an edge monitoring node, a digital twin model construction module, a stray current diffusion simulation module, an LSTM time series prediction module, an intelligent classification and localization module for interference sources, a three-dimensional visualization module, an adaptive protection suggestion generation module, and a closed-loop feedback optimization module.

[0007] The edge monitoring node is used for multi-dimensional data synchronous collection; The digital twin model building module is used for pipeline network 3D modeling, multi-source data fusion and model accuracy calibration, realizing real-time mapping between physical entities and digital models; The stray current diffusion simulation module is used to simulate current distribution, corrosion rate, risk area and surface potential based on electromagnetic field and electrochemical principles. The LSTM time-series prediction module is used for long-term time-series prediction and graded early warning of corrosion risk. The intelligent classification and localization module for interference sources is used to extract waveform features, identify the type of interference source, and perform three-dimensional localization. The three-dimensional visualization module is used for the three-dimensional dynamic presentation of pipeline network models, current cloud maps, corrosion thermal maps, and the location of interference sources; The adaptive protection suggestion generation module is used to generate graded protection schemes based on risk levels. The closed-loop feedback optimization module is used to collect protection effects, evaluate execution effectiveness, and iteratively correct model parameters.

[0008] Furthermore, the edge monitoring node is a distributed acquisition terminal that integrates a current acquisition unit, a soil environment acquisition unit, a structural status acquisition unit, a surface potential acquisition unit, and a data transmission module.

[0009] Furthermore, the stray current diffusion simulation module, based on the principle of electromagnetic field and electrochemical coupling, completes numerical simulations of current distribution, corrosion rate, risk area and surface potential.

[0010] Furthermore, the LSTM time series prediction module provides graded early warning based on high, medium, and low risk levels.

[0011] Furthermore, the interference source intelligent classification and localization module uses the time difference method to complete the three-dimensional localization of the interference source.

[0012] Furthermore, the adaptive protection suggestion generation module generates a graded protection scheme that includes inspection, parameter adjustment, insulation modification, and forced drainage.

[0013] Furthermore, the closed-loop feedback optimization module evaluates the execution effectiveness and iteratively corrects the model parameters of the digital twin model construction module and the LSTM time series prediction module based on the collected protection effect data.

[0014] This invention also provides a method for dynamic prediction of stray current corrosion risk based on digital twins, comprising the following steps: S1: Real-time collection of multi-dimensional data; S2: Digital Twin Model Construction and Calibration; S3: Simulation of stray current diffusion and corrosion process; S4: Intelligent classification and three-dimensional localization of interference sources; S5: Intelligent classification of interference sources and time-difference 3D localization based on SVM; S6: Comprehensive 3D visualization display; S7: Automatic generation and push of graded protection recommendations; S8: Execution result feedback and model iterative optimization.

[0015] Furthermore, the classification decision function of the SVM is: ;in, Represents the space consisting of samples. Represents the first sample space One sample, Representing the Lagrange multipliers, which can be obtained by solving. Represents the known first The classification label for each sample: 1 for a positive sample and -1 for a negative sample. Represents the Gaussian kernel function (where and (representing the function's independent variable) This represents the trainable kernel bandwidth. This indicates a trainable bias. It is a symbolic function. It is a sample The prediction category.

[0016] Furthermore, the mathematical model of the LSTM is as follows: in, Indicates time The characteristic variables are obtained from the sensors. express The hidden state at any given moment. express The hidden state at any given moment. , , and Let represent the trainable biases of the input gate, forget gate, output gate, and intermediate states, respectively. , 、· and Let represent the trainable weights of the input gate, forget gate, output gate, and intermediate states with respect to the feature quantity, respectively. , , and Let represent the trainable weights of the input gate, forget gate, output gate, and intermediate states with respect to the hidden state, respectively. , , and These represent the output values ​​of the input gate, forget gate, output gate, and intermediate states, respectively. and Representing time respectively and The storage unit outputs the value. and These represent the trainable weights and biases of the model output, respectively. Indicates at time Predicted risk indicator values, Represents the sigmoid activation function (where (representing the function's independent variable) Represents the hyperbolic tangent function. It is a natural constant, symbol This represents the Hadamard product.

[0017] Furthermore, the interference sources include subways, electrified railways, DC grounding electrodes, high-voltage DC lines, electrolysis devices, DC charging piles, and industrial DC equipment.

[0018] Furthermore, the tiered early warning system includes: a 15-day advance warning for high-risk situations, a 10-day advance warning for medium-risk situations, and a 7-day advance warning for low-risk situations.

[0019] The embodiments of the present invention have the following beneficial effects: (1) The prediction accuracy is greatly improved: the use of digital twin dynamic mapping and LSTM long time series prediction improves the accuracy by more than 30% compared with the traditional static prediction method, and achieves early warning of 7 to 15 days in advance; (2) Accurate source tracing of interference sources: Supports intelligent classification of 7 typical interference sources, with a three-dimensional positioning error of ≤5%, and quickly locks the source of stray current leakage; (3) Three-dimensional visualization: The three-dimensional visualization displays the current distribution, potential changes, corrosion status and interference source location, allowing maintenance personnel to intuitively grasp the safety status of the pipeline network; (4) Intelligent adaptive protection decision-making: Differentiated protection plans are automatically generated according to risk level, without relying on human experience, and the protection measures are more targeted and timely; (5) Closed-loop iterative continuous optimization: A closed-loop system of "monitoring-simulation-prediction-location-protection-feedback" is formed, and the model and algorithm are continuously iterated, and the protection accuracy is continuously improved; (6) Wide range of applicable scenarios: covering stray current corrosion protection scenarios such as urban underground pipe networks, long-distance oil and gas pipelines, rail transit metal structures, power grounding grids and industrial buried metal facilities. Attached Figure Description

[0020] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0021] in: Figure 1 This is an overall structural block diagram of the stray current corrosion risk dynamic prediction device based on digital twins as described in an embodiment of the present invention.

[0022] Figure 2 This is a schematic diagram of the digital twin model of stray current in underground pipe networks according to an embodiment of the present invention.

[0023] Figure 3 This is a cross-sectional view along the AA direction of the digital twin model of stray current in underground pipe network according to an embodiment of the present invention.

[0024] Figure 4 This is a schematic diagram of the overall process of the stray current corrosion risk dynamic prediction method based on digital twins as described in an embodiment of the present invention.

[0025] Explanation of key component symbols: 1. Physical entity of underground metal pipe network; 2. Stray current inflow point; 3. Stray current outflow point; 4. Stray current conduction path in soil; 5. High corrosion risk area; 6. Medium corrosion risk area; 7. Low corrosion risk area. Detailed Implementation

[0026] To facilitate understanding of the present invention, a more complete description will be given below with reference to the accompanying drawings. Preferred embodiments of the invention are shown in the drawings. However, the invention can be implemented in many different forms and is not limited to the embodiments described herein. Rather, these embodiments are provided to provide a thorough and complete understanding of the disclosure of the invention.

[0027] This embodiment is applied to the stray current corrosion protection scenario of buried gas pipelines along urban rail transit lines, such as... Figure 2 As shown, the entire system consists of edge monitoring nodes, a digital Li Sheng model construction module, a stray current diffusion simulation module, an L5T timing prediction module, an interference source intelligent classification and localization module, a 3D visualization module, an adaptive protection suggestion generation module, and a closed-loop feedback optimization module. Figure 1 As shown.

[0028] An edge monitoring node is deployed every 50 meters above the gas pipeline along the rail transit line, and three surface potential acquisition terminals are arranged in an equilateral triangle on the ground surface to form a distributed monitoring network. Figure 1Edge monitoring nodes.

[0029] Three-dimensional laser scanning was used to obtain the spatial coordinates of the pipeline, and a three-dimensional digital twin model of the gas pipeline was constructed using BIM technology, with a modeling error of ±2.8cm. Model calibration was completed by integrating soil resistivity, pipeline material, and burial depth parameters, resulting in an overall error ≤3%. Figure 1 The digital Li Sheng model building module.

[0030] Edge nodes collect stray current, soil temperature, humidity, resistivity, pipeline vibration, and surface potential waveform data in real time, with a transmission delay of 35ms. The stray current diffusion simulation module performs simulations based on Maxwell's equations, identifying areas where current concentrates into the pipeline wall as high-corrosion-risk points. Figure 2 and Figure 3 As shown, it constitutes a stray current diffusion simulation module and a three-dimensional visualization display module.

[0031] The LSTM prediction module identified the problem as high-risk and issued a warning 15 days in advance. The interference source classification and localization module extracted waveform features and identified it as an interference source for the subway traction system using the SVM algorithm. Time difference method was used for localization, with a localization error of 3.1%. Figure 1 The L5T timing prediction module and the intelligent classification and localization module for interference sources.

[0032] The 3D visualization module highlights the locations of corrosion hotspots and subway interference sources. The system automatically generates forced drainage modification and cathodic protection potential optimization schemes, and pushes them to the operation and maintenance platform, forming a complete system. Figure 1 The adaptive protection suggestion generation module.

[0033] After the protection scheme was implemented, monitoring data showed that stray current density decreased by 85%, and corrosion risk was reduced to a low level. The results were fed back to the digital twin model, and simulation parameters were iteratively corrected to form... Figure 1 The closed-loop feedback optimization module.

[0034] like Figure 4 As shown, this embodiment is applied to stray current corrosion protection for long-distance oil and gas pipelines, and is performed according to the following steps: S1: Data Acquisition. Real-time data acquisition is achieved through edge monitoring nodes, including stray current in the pipeline, soil temperature (23℃), humidity (58%), resistivity (120Ω·m), pipeline vibration (0.03mm / s), and transient waveform data of the ground surface potential.

[0035] S2: Model Construction. Based on pipeline CAD drawings and laser scanning data, a three-dimensional digital twin model of the long-distance pipeline is constructed. Five years of historical corrosion data and real-time monitoring data are integrated, with the model updated once per minute and accuracy calibrated to ±2.5cm.

[0036] S3: Simulation Analysis. The three-dimensional diffusion path of stray current was simulated using the finite element method, and the corrosion rate was calculated using Faraday's law. Three pipe sections were identified as high-risk corrosion areas, and two as medium-risk areas.

[0037] S4: Interference Source Processing. Eleven features of the ground potential waveform were extracted, and the SVM algorithm identified it as a DC grounding electrode interference source. A set of location equations was constructed based on the time difference data from four monitoring terminals to solve for the interference source coordinates. The location error was 2.9%, and the assessed impact range was a 5km section of the pipeline.

[0038] S5: Risk Prediction. Input simulation data and real-time data into the LSTM model. High-risk areas will receive an early warning 15 days in advance, and medium-risk areas will receive an early warning 10 days in advance. The model will output a dynamic curve showing the change in risk level.

[0039] S6: Visualization. The platform displays a 3D model of the long-distance pipeline, overlaid with current density cloud maps, corrosion heat maps, and DC grounding electrode interference source markers. High-risk areas are highlighted in red, and medium-risk areas are marked in yellow.

[0040] S7: Protection Generation. For high-risk areas, solutions for interference source shielding and isolation, and pipeline insulation repair are proposed, with a 7-day deadline for completion. For medium-risk areas, suggestions for adjusting cathodic protection current density are provided.

[0041] S8: Iterative Optimization. After the protective measures were implemented, monitoring data for 30 consecutive days showed a 90% decrease in corrosion rate, reducing the risk level to low risk. The protective effect data was fed back to the system to optimize the electric field parameters of the digital twin model and the LSTM prediction weights.

[0042] To verify the technical effectiveness of this invention, a third-party testing organization was commissioned to conduct performance tests. In terms of prediction accuracy, the accuracy of corrosion risk prediction is 32% higher than that of traditional manual monitoring methods, and the accuracy rate of early warning in 7 to 15 days is ≥95%.

[0043] In terms of positioning accuracy, the classification accuracy of the seven types of interference sources was 86%, and the average error of three-dimensional positioning was 3.8%.

[0044] In terms of model accuracy, the modeling error of the digital twin model is ±2.6cm, and the overall simulation error is 4.2%.

[0045] In terms of synchronization performance, the GNSS synchronization time difference of the ground potential acquisition unit is 0.6μs, and the sampling rate is 1GS / s.

Claims

1. A stray current corrosion risk dynamic prediction device based on digital twin, characterized in that, include: The module includes an edge monitoring node, a digital twin model construction module, a stray current diffusion simulation module, an LSTM (Long Short-Term Memory) timing prediction module, an intelligent classification and localization module for interference sources, a 3D visualization module, an adaptive protection suggestion generation module, and a closed-loop feedback optimization module. The edge monitoring node is used for multi-dimensional data synchronous collection; The digital twin model building module is used for pipeline network 3D modeling, multi-source data fusion and model accuracy calibration, realizing real-time mapping between physical entities and digital models; The stray current diffusion simulation module is used to simulate current distribution, corrosion rate, risk area and surface potential based on electromagnetic field and electrochemical principles. The LSTM time-series prediction module is used for long-term time-series prediction and graded early warning of corrosion risk. The intelligent classification and localization module for interference sources is used to extract waveform features, identify the type of interference source, and perform three-dimensional localization. The three-dimensional visualization module is used for the three-dimensional dynamic presentation of pipeline network models, current cloud maps, corrosion thermal maps, and the location of interference sources; The adaptive protection suggestion generation module is used to generate graded protection schemes based on risk levels. The closed-loop feedback optimization module is used to collect protection effects, evaluate execution effectiveness, and iteratively correct model parameters.

2. The stray current corrosion risk dynamic prediction device based on digital twin according to claim 1, characterized in that, The edge monitoring node is a distributed acquisition terminal that integrates a current acquisition unit, a soil environment acquisition unit, a structural status acquisition unit, a surface potential acquisition unit, and a data transmission module.

3. The stray current corrosion risk dynamic prediction device based on digital twin according to claim 1, characterized in that, The stray current diffusion simulation module is based on the principle of electromagnetic field and electrochemical coupling to complete numerical simulations of current distribution, corrosion rate, risk area and surface potential.

4. The stray current corrosion risk dynamic prediction device based on digital twin according to claim 1, characterized in that, The LSTM time-series prediction module outputs a corrosion level index and provides graded early warnings based on high, medium, and low risk levels.

5. The stray current corrosion risk dynamic prediction device based on digital twin according to claim 1, characterized in that, The interference source intelligent classification and localization module uses the time difference method to complete the three-dimensional localization of the interference source.

6. The stray current corrosion risk dynamic prediction device based on digital twin according to claim 1, characterized in that, The adaptive protection suggestion generation module generates a graded protection scheme that includes inspection, parameter adjustment, insulation modification, and forced drainage.

7. The stray current corrosion risk dynamic prediction device based on digital twin according to claim 1, characterized in that, The closed-loop feedback optimization module evaluates the effectiveness of the protection based on the collected data and iteratively corrects the model parameters of the digital twin model construction module and the LSTM time series prediction module.

8. A method for dynamic prediction of stray current corrosion risk based on digital twins, characterized in that, Includes the following steps: S8.1: Real-time collection of multi-dimensional data; S8.2: Digital Twin Model Construction and Calibration; S8.3: Simulation of stray current diffusion and corrosion process; S8.4: Intelligent classification of interference sources and time-difference 3D localization based on SVM (Support Vector Machines); S8.5: LSTM Dynamic Risk Prediction and Graded Early Warning; S8.6: Comprehensive 3D visualization display; S8.7: Automatic generation and push of graded protection recommendations; S8.8: Execution result feedback and model iterative optimization.

9. The method according to claim 8, characterized in that, The sources of interference include subways, electrified railways, DC grounding electrodes, high-voltage DC lines, electrolysis devices, DC charging piles, and industrial DC equipment.

10. The method according to claim 8, characterized in that, The classification decision function of the SVM is: in, Represents the space consisting of samples. Represents the first sample space One sample, Representing the Lagrange multipliers, which can be obtained by solving. Represents the known first The classification label for each sample: 1 for a positive sample and -1 for a negative sample. Represents the Gaussian kernel function (where and (representing the function's independent variable) This represents the trainable kernel bandwidth. This indicates a trainable bias. It is a symbolic function. It is a sample The prediction category.

11. The method according to claim 8, characterized in that, The mathematical model of the LSTM is as follows: in, Indicates time The characteristic variables are obtained from sensors. express The hidden state at any given moment. express The hidden state at any given moment. , , ,and Let represent the trainable biases of the input gate, forget gate, output gate, and intermediate states, respectively. , , ,and Let represent the trainable weights of the input gate, forget gate, output gate, and intermediate states with respect to the feature quantity, respectively. , , and Let represent the trainable weights of the input gate, forget gate, output gate, and intermediate states with respect to the hidden state, respectively. , , and These represent the output values ​​of the input gate, forget gate, output gate, and intermediate states, respectively. and Representing time respectively and The storage unit outputs the value. and These represent the trainable weights and biases of the model output, respectively. Indicates at time Predicted risk indicator values, Represents the sigmoid activation function (where (representing the function's independent variable) Represents the hyperbolic tangent function. It is a natural constant, symbol This represents the Hadamard product.

12. The method according to claim 8, characterized in that, The tiered early warning system includes: 15-day advance warning for high-risk areas, 10-day advance warning for medium-risk areas, and 7-day advance warning for low-risk areas.

Images