Pipeline leakage detection method and system

Abstract

The invention discloses a pipeline leakage detection method and system, and belongs to the technical field of pipeline leakage detection.The method comprises the steps that firstly, edge computing nodes are deployed along a pipeline, pipeline parameters of the pipeline are synchronously and continuously collected, the external environment of the pipeline is monitored through a refrigeration type infrared camera, and a gas cloud image is generated; secondly, establishing a noise feature library and a thermodynamic model based on continuously acquired pipeline parameters; then, based on a noise feature library and a thermodynamic model, extracting abnormal feature information of edge calculation nodes; then, based on the extracted abnormal feature information of the edge calculation node, judging whether the leakage of the pipeline is abnormal or not; and finally, if the leakage is abnormal, performing deep fusion on the extracted abnormal feature information of the edge computing node and the generated gas cloud image so as to immediately position when the leakage is confirmed to occur. According to the method, the technical problems of pressure signal interference caused by temperature, pipeline noise, especially leakage identification inaccuracy caused by noise at the joint are solved.

Description

Technical Field

[0001] This application belongs to the field of pipeline leak detection technology, and specifically relates to a pipeline leak detection method and system. Background Technology

[0002] Current mainstream pipeline leak detection methods mostly rely on pressure and vibration signal analysis, but they have the following problems: First, the pressure of the medium inside the pipeline is easily affected by temperature fluctuations, and existing detection technologies cannot effectively eliminate temperature interference. As a result, the measured pressure values ​​under different temperature environments cannot accurately reflect the actual operating conditions of the pipeline, making it difficult to accurately capture the micro-pressure drop characteristics caused by leaks, which can easily lead to misjudgments or missed detections. Second, due to the special structure of pipeline elbows, tees, crosses, and other connections, strong background noise is generated during operation. Existing noise reduction technologies lack specificity and cannot accurately remove this type of structural noise according to the type of pipe fitting and operating conditions. As a result, the leakage characteristics in the vibration signal are masked by noise, which seriously affects the accuracy of leak identification and makes it difficult to meet the detection needs of complex pipeline systems. Summary of the Invention

[0003] To address the aforementioned issues, this application provides a pipeline leak detection method and system that features minimal interference and high accuracy.

[0004] The purpose of this invention is to provide a pipeline leak detection method, comprising: Edge computing nodes are deployed along the pipeline to synchronously and continuously collect pipeline parameters and monitor the external environment of the pipeline using cooled infrared cameras to generate gas cloud images. Based on continuously collected pipeline parameters, a noise feature library and thermodynamic model are established. Based on a noise feature library and a thermodynamic model, abnormal feature information of edge computing nodes is extracted; Based on the extracted abnormal feature information of edge computing nodes, determine whether the pipeline is leaking abnormally; If an abnormality is detected, the extracted abnormal feature information of the edge computing node will be deeply fused with the generated gas cloud image to locate the leak immediately upon confirmation.

[0005] Furthermore, edge computing nodes are deployed along the pipeline to synchronously and continuously collect pipeline parameters, including... Identify high-risk sections and critical nodes in the pipeline; Based on the identified high-risk pipe sections and key nodes, the deployment locations of edge computing nodes are planned along the pipeline route; On the planned edge computing nodes, an integrated sensor array is installed to measure the pipe parameters, including pipe medium parameters and vibration signals of the pipe wall. The medium parameters include the pressure and temperature inside the medium. At the same time, the infrared camera is calibrated so that its field of view includes the space above the pipe and the space at the pre-set node connection.

[0006] Furthermore, based on continuously collected pipeline parameters, a noise feature library and thermodynamic model are established, including: The continuously collected pipeline parameters are archived to form historical data, which includes pressure, temperature, flow rate, and pipe wall vibration signals above the preset frequency at preset pipeline nodes.

[0007] Furthermore, based on continuously collected pipeline parameters, the establishment of a noise feature library and thermodynamic model also includes... Based on the gas law and pipe thermodynamics theory, a thermodynamic model is established to describe the dynamic relationship between pressure and temperature inside a pipe. The collected historical normal data is input into the established thermodynamic model, and the key parameters in the thermodynamic model are optimized and adjusted to calibrate the thermodynamic model. During the real-time monitoring phase, the current measured pipeline temperature value is input into the calibrated thermodynamic model to correct the real-time monitored pressure value and calculate the expected theoretical pressure value under the current temperature conditions without leakage.

[0008] Furthermore, based on the gas law and pipe thermodynamics theory, a thermodynamic model describing the dynamic relationship between pressure and temperature inside the pipe is established: ; in, To correct for the pressure value at the reference temperature, This is the actual measured pressure value. For the selected reference temperature, It is relative to the reference temperature The corresponding reference pressure, This is the actual measured temperature value. There are two functions, which represent the gas compressibility under reference conditions and the gas compressibility under measured conditions, respectively.

[0009] Furthermore, based on continuously collected pipeline parameters, the establishment of a noise feature library and thermodynamic model also includes... Analyze pipe wall vibration signals above the preset frequency to extract vibration and noise characteristics generated at pipe elbows, tees, and crosses under various specific working conditions; The extracted vibration and noise features are quantified and classified and archived according to the pipe fitting type and operating parameters to establish a noise feature library.

[0010] Furthermore, based on continuously collected pipeline parameters, the establishment of a noise feature library and thermodynamic model also includes... Based on the historical data, various typical operating periods covering different combinations of flow and pressure were selected; Based on the selected typical working conditions, continuous vibration signal sequences were extracted from the connection points of elbows, tees, and crosses. Preprocess the captured continuous vibration signal sequence; Each segment of the preprocessed continuous vibration signal sequence is subjected to in-depth analysis, and each segment of the vibration signal is transformed into a digital feature vector that can fully characterize its properties. The transformed digital feature vectors are forcibly associated and labeled with their corresponding continuous vibration signal sequences to form sample data corresponding to feature vectors and working condition labels. The labeled sample data are stored uniformly to form a noise feature library.

[0011] Furthermore, based on the noise feature library and thermodynamic model, the abnormal feature information of the edge computing nodes is extracted, including: Edge computing nodes align and preprocess the corrected pressure signal and the original pipe wall vibration signal; Continuous wavelet transform is performed on the corrected pressure signal, and the transient micro-pressure drop characteristics are obtained by analyzing the subtle changes in the signal in the time and scale planes. The LSTM model is used to analyze the time series data of the original pipe wall vibration signal, learn the normal pattern and predict the trend, and obtain the vibration signal with abnormal trend. The source location of the original pipe wall vibration signal is determined in real time. When the original pipe wall vibration signal is located at the elbow, tee, or cross connection, the corresponding background noise fingerprint is called from the pre-established noise feature library according to the current operating parameters. Adaptive filtering technology is used to effectively remove this inherent structural noise from the real-time total signal, and the vibration residual signal after removing the background noise is obtained.

[0012] Furthermore, based on the extracted abnormal feature information of the edge computing nodes, it is determined whether the pipeline is leaking. The extracted micro-pressure drop characteristics, abnormal vibration signals, and residual vibration signals after removing background noise are fused and analyzed. When multiple abnormal feature indicators deviate from the normal range in a coordinated manner, the edge computing node initially determines that there is a leakage anomaly and generates a high-confidence alarm event.

[0013] Furthermore, the extracted anomaly feature information of the edge computing nodes is deeply fused with the generated gas cloud image to immediately locate the leak when it is confirmed. It receives alarm events from edge computing nodes and reads the sequence of atmospheric cloud images generated by infrared cameras within the same spatiotemporal range. It then aligns and correlates the abnormal feature information with the atmospheric cloud image sequence on a unified time axis and pipeline coordinate axis. The DS evidence theory algorithm is used to assign confidence levels to each type of anomalous feature information and cloud image sequence; After confirming the leak, the arrival time difference of the sound wave or pressure wave signal recorded by multiple edge computing nodes upstream and downstream of the leak point is called, and the preliminary location range of the leak point is calculated by the time difference positioning method. The calculated preliminary location range is superimposed and compared with the central region with the highest gas cloud concentration in the gas cloud image sequence. The location result is then verified and corrected using visual information. Based on the verified and corrected location results, a comprehensive alarm message is automatically generated and output, which includes the precise coordinates of the leak point, the leak intensity assessment, and the gas cloud diffusion range.

[0014] Another object of the present invention is to provide a pipeline leak detection system, comprising, The acquisition module is used to deploy edge computing nodes along the pipeline to synchronously and continuously acquire pipeline parameters and monitor the external environment of the pipeline through a cooled infrared camera to generate gas cloud images. A module is established to build a noise feature library and a thermodynamic model based on continuously collected pipeline parameters; The extraction module is used to extract abnormal feature information of edge computing nodes based on the noise feature library and thermodynamic model; The judgment module is used to determine whether the pipeline is leaking abnormally based on the extracted abnormal feature information of the edge computing nodes; The location module is used to perform deep fusion of the extracted abnormal feature information of the edge computing nodes with the generated gas cloud image if an abnormality is detected, so as to locate the leak immediately when it is confirmed.

[0015] Compared with the prior art, this application has the following advantages: The above method effectively solves the problem of temperature changes interfering with pressure signals by establishing and calibrating a thermodynamic model, avoiding misjudgments of pressure signals due to temperature fluctuations and significantly improving the accuracy of leak detection in temperature-variable scenarios. Furthermore, by establishing a noise feature library, the interference of noise on anomaly detection is reduced, ensuring accurate leak identification even in noisy critical connection areas.

[0016] Other features and advantages of this application will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the application. The objectives and other advantages of this application may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0017] To more clearly illustrate the technical solutions in the embodiments of this application 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 some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0018] Figure 1 A schematic flowchart of a pipeline leakage detection method according to an embodiment of the present invention is shown; Figure 2 A schematic diagram of a pipeline leak detection system according to an embodiment of the present invention is shown. Detailed Implementation

[0019] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0020] like Figure 1 As shown in the illustration, this invention introduces a pipeline leak detection method. The method includes: first, deploying edge computing nodes along the pipeline to continuously collect pipeline parameters and monitoring the external environment of the pipeline using a cooled infrared camera to generate a gas cloud image; second, establishing a noise feature library and a thermodynamic model based on the continuously collected pipeline parameters; third, extracting abnormal feature information from the edge computing nodes based on the noise feature library and thermodynamic model; fourth, determining whether the pipeline is leaking abnormally based on the extracted abnormal feature information from the edge computing nodes; and finally, if a leak is detected, performing deep fusion of the extracted abnormal feature information from the edge computing nodes with the generated gas cloud image to immediately locate the leak upon confirmation. This method, by establishing and calibrating a thermodynamic model, effectively solves the problem of temperature changes interfering with pressure signals, avoiding misjudgments of pressure signals due to temperature fluctuations, and significantly improving the accuracy of leak detection in temperature-variable scenarios. Furthermore, by establishing a noise feature library, it reduces the interference of noise on anomaly detection, ensuring accurate identification of leaks even in noisy critical connection areas.

[0021] Specifically, edge computing nodes are deployed along the pipeline to synchronously and continuously collect pipeline parameters, including... S11. Identify high-risk pipe sections and key nodes, and calculate the deployment location of nodes along the pipeline route to ensure that the monitoring network can cover the entire line and focus on weak links. Among them, the identification of high-risk pipe sections and key nodes is based on the pipeline's historical leakage records, pipe wall corrosion detection data, geological settlement monitoring reports, media corrosivity analysis, and pipeline design data including pipe material, wall thickness, and pressure rating.

[0022] S12. On the planned edge computing nodes, an integrated sensor array is installed to measure the pipeline parameters, including pipeline medium parameters and vibration signals of the pipe wall. The medium parameters include the internal pressure and temperature of the medium. Simultaneously, a calibrated cooled infrared camera is used to maximize the field of view covering the space above the pipe and critical connections. Critical connections are pre-defined node connections, namely the connection points of elbows, tees, crosses, and other pipe fittings in the pipeline system.

[0023] In this embodiment of the invention, establishing a noise feature library and a thermodynamic model based on continuously collected pipeline parameters includes: S21. Continuously collect and archive a large amount of historical data, including pressure, temperature, flow rate, and pipe wall vibration signals above the preset frequency at preset nodes in the pipeline. The preset frequency is set according to the characteristics of the pipeline system. For common fluid pipelines, the frequency range of interest may be from several hundred hertz (Hz) to several thousand hertz (kHz) to capture acoustic emission or stress wave signals caused by leakage, while excluding low-frequency vibrations such as mains interference. Therefore, the specific preset frequency can be determined based on the sensor performance and pipeline characteristics during project implementation.

[0024] S22. Based on the gas law and pipeline thermodynamics theory, a thermodynamic model describing the dynamic relationship between pressure and temperature within the pipeline is established. Historical normal data from collected historical data is then input into this thermodynamic model. Key parameters in the thermodynamic model are optimized and adjusted to achieve the best fit between the model's predicted output and historical measured pressure values, thus completing the accurate calibration of the model. Key parameters mainly include reference temperature, reference pressure, and model parameters used to calculate the gas compressibility factor function. The optimization objective is to achieve the best fit between the corrected pressure calculated by the model and the measured pressure values ​​under historical normal operating conditions. S23. Conduct in-depth analysis of historical pipe wall vibration signal data, extract the vibration and noise characteristics generated by key connections such as pipe elbows, tees, and crosses under various typical working conditions, and quantify these vibration and noise characteristics using time-domain analysis algorithms in signal processing technology. Classify and archive them according to their corresponding pipe fitting types and working parameters to establish a noise feature library. S24. During the real-time monitoring phase, the system inputs the currently measured pipeline temperature value into the calibrated thermodynamic model to correct the real-time monitored pressure value and calculate the expected theoretical pressure value under the current temperature conditions without leakage. Specifically, inputting the currently measured temperature value into the calibrated thermodynamic model during the real-time monitoring phase aims to correct the simultaneously measured pressure value and calculate the corrected pressure value. This correction is intended to eliminate the influence of temperature fluctuations on the pressure reading, ensuring that the pressure value more accurately reflects pressure changes caused by events such as leaks, rather than pressure changes caused by temperature variations.

[0025] In this embodiment of the invention, a thermodynamic model describing the dynamic relationship between pressure and temperature inside a pipe is established based on the gas state equation and pipe thermodynamics theory: ; in, To correct for the pressure value at the reference temperature, This is the actual measured pressure value. For the selected reference temperature, It is relative to the reference temperature The corresponding reference pressure, This is the actual measured temperature value. These are two functions, representing the gas compressibility under reference conditions and the gas compressibility under measured conditions, respectively. A thermodynamic model is established and calibrated using historical data, accurately describing the dynamic relationship between pressure and temperature within a pipeline based on the gas equation of state and pipeline thermodynamics. First, key parameters of the model are optimized using a large amount of historical normal data to achieve an optimal fit between the model's predicted output and historical measured pressure values. Then, the model is used to correct the measured pressure values ​​during real-time monitoring, effectively solving the problem of temperature variation interfering with the pressure signal. Even under different temperature environments, the corrected pressure signal accurately reflects the true pressure state within the pipeline, providing a precise data foundation for subsequent wavelet transform to capture the micro-pressure drop characteristics of leaks. This avoids misjudgments of pressure signals caused by temperature fluctuations and significantly improves the accuracy of leak detection in temperature-variable scenarios.

[0026] Step S23: Conduct in-depth analysis of historical pipe wall vibration signal data, extract vibration and noise characteristics generated by key connections such as pipe elbows, tees, and crosses under various typical operating conditions, and quantify these vibration and noise characteristics using time-domain analysis algorithms in signal processing technology. Classify and archive these characteristics according to their corresponding pipe fitting types and operating parameters to establish a noise feature library. S231. Retrieve historical data of the pipeline during known normal operation, filter out various typical operating conditions covering different flow and pressure combinations, and extract continuous vibration signal sequences flowing through elbows, tees, and crosses. Preprocess these raw signals (i.e., the extracted continuous vibration signal sequences flowing through elbows, tees, and crosses) to improve signal quality and prepare for subsequent feature extraction. Preprocessing includes detrending, filtering, outlier removal, and signal normalization to improve signal quality.

[0027] S232. For each segment of the preprocessed continuous vibration signal sequence, apply multiple signal processing techniques for in-depth analysis, transforming each segment of the continuous vibration signal sequence into a digital feature vector that can comprehensively characterize its properties. Specifically, the application of multiple signal processing techniques for in-depth analysis includes time-domain analysis, frequency-domain analysis, and time-frequency analysis. That is, calculate the statistical characteristics of the signal through time-domain analysis, obtain the spectrum and envelope spectrum of the signal through frequency-domain analysis, and observe the changes of the signal frequency components over time through time-frequency analysis.

[0028] S232. Forcefully associate and label the obtained digital feature vector with its corresponding metadata to form sample data corresponding to feature vector-working condition label; S233. Establish a noise feature library and store the labeled sample data into the noise feature library.

[0029] By establishing a noise feature database, the noise interference problem at pipe bends, tees, and crosses can be addressed in a targeted manner: first, historical vibration data during normal operation is retrieved, signals from multiple typical operating periods are screened and preprocessed, and then the signals are converted into digital feature vectors and associated with operating condition labels to form classified and archived noise samples; when signals are detected to originate from these connections in real time, the corresponding background noise fingerprint can be retrieved according to the current operating condition, and structural background noise can be effectively stripped away through adaptive filtering technology, making the leakage-related features in the vibration signal clearer, reducing the interference of noise on anomaly judgment, and ensuring accurate identification of leaks even in critical connection parts with complex noise.

[0030] Based on a noise feature library and a thermodynamic model, abnormal feature information of edge computing nodes is extracted, including: S31. The edge computing node aligns and preprocesses the pressure signal after thermodynamic model correction and the original vibration signal to eliminate baseline drift and random pulse interference. The preprocessing includes data alignment, noise reduction filtering, and smoothing to eliminate random pulse interference. S32. Perform continuous wavelet transform on the corrected pressure signal, and capture transient micro-pressure drop characteristics by analyzing subtle changes in the signal in the time and scale planes. S33. Use an LSTM (Long Short-Term Memory) model to analyze vibration signal time series data, learn normal patterns and predict trends, and obtain vibration signals with abnormal trends. First, train the LSTM network using historical normal vibration signal time series data to learn its normal dynamic behavior patterns. Then, in real-time monitoring, use the trained LSTM model to predict vibration signals in the short term. Finally, compare the predicted values ​​with the actual measured values. If the deviation continues to exceed a preset threshold, it is judged as an abnormal trend.

[0031] S34. Real-time determination of the source location of vibration signals. When the source is located at bends, tees, or crosses, the system retrieves the corresponding background noise fingerprint from a pre-established noise feature library based on the current operating parameters. Adaptive filtering technology is then used to effectively remove this inherent structural noise from the real-time total signal. The current operating parameters refer to key physical quantities that characterize the real-time operating status of the pipeline system, mainly including the flow rate and pressure of the medium. The inherent structural noise refers to the digital feature vector of the background noise corresponding to specific operating conditions, pre-stored in the noise feature library. In real-time monitoring, the system retrieves the corresponding fingerprint based on the signal location and current operating conditions, using it as reference noise for adaptive filtering.

[0032] S35. The extracted micro-pressure drop characteristics, abnormal vibration signals, and residual vibration signals after background noise removal are fused and analyzed. When multiple feature indicators deviate from the normal range in a coordinated manner, the edge computing node initially determines that there is a leakage anomaly and generates a high-confidence alarm event. The fusion analysis employs multi-feature collaborative decision-making logic, setting a normal threshold range for each feature indicator. Only when multiple heterogeneous feature indicators simultaneously or in most cases exhibit coordinated abnormal deviations, i.e., all exceed their respective normal ranges, will the edge computing node initially determine that there is a leakage anomaly. The extracted anomaly feature information of the edge computing nodes is deeply fused with the generated gas cloud image to immediately locate the leak when it is confirmed. S41. Receive alarm events from edge computing nodes and simultaneously read the sequence of gas cloud images generated by infrared cameras within the same spatiotemporal range. Align and correlate all information on a unified time axis and pipeline coordinate axis. Specifically, aligning and correlating all information on a unified time axis and pipeline coordinate axis involves performing spatiotemporal matching and correlation analysis between the abnormal feature information extracted by the edge computing nodes and the sequence of gas cloud images captured by infrared cameras within the same time period and pipeline section.

[0033] S42. The Dempster-Shafer Evidence Theory (DS Evidence Theory, a mathematical framework for handling uncertainty reasoning) algorithm is used to assign confidence levels to the identification results of each technological source. Specifically, a confidence level is assigned to the analysis results based on pressure signals, vibration signals, and infrared images, respectively. Then, information is fused using the DS Evidence Theory to make a more reliable leak judgment. S43. After confirming the leak, the arrival time difference of acoustic or pressure wave signals recorded by multiple edge computing nodes upstream and downstream of the leak point is retrieved. The initial location range of the leak point is calculated using the time-of-arrival (TOA) positioning method. This initial location range is then overlaid and compared with the central region of the highest gas cloud concentration in the infrared image. Visual information is used to perform secondary verification and precise correction of the positioning result, reducing the positioning error to the centimeter level. Specifically, threshold segmentation and morphological operations using image processing techniques are employed to locate the region of highest gas cloud concentration. Its geographic coordinates are then spatially registered and optimized with the preliminary results of acoustic TOA positioning to correct the leak point coordinates.

[0034] S44. Automatically generate and output comprehensive alarm information including precise coordinates of the leak point, leak intensity assessment, and gas cloud diffusion range. Among them, the leak intensity is estimated based on the pressure drop rate and continuous flow balance calculation; the gas cloud diffusion range is calculated by processing infrared image sequences, using image segmentation algorithms to identify pixel areas above a specific temperature threshold, and combining camera calibration parameters to convert them into actual physical dimensions.

[0035] like Figure 2 As shown in the illustration, this invention also introduces a pipeline leak detection system capable of performing the above-described method. The system includes an acquisition module, an establishment module, an extraction module, a judgment module, and a location module. The acquisition module deploys edge computing nodes along the pipeline to synchronously and continuously acquire pipeline parameters and monitors the external environment of the pipeline using a cooled infrared camera to generate a gas cloud image. The establishment module establishes a noise feature library and a thermodynamic model based on the continuously acquired pipeline parameters. The extraction module extracts abnormal feature information from the edge computing nodes based on the noise feature library and the thermodynamic model. The judgment module determines whether the pipeline is leaking abnormally based on the extracted abnormal feature information from the edge computing nodes. The location module performs deep fusion of the extracted abnormal feature information from the edge computing nodes with the generated gas cloud image if a leak is detected, enabling immediate location of the leak upon confirmation. This system solves the technical problems of inaccurate leak identification caused by temperature interference with pressure signals, pipeline noise, and especially noise at connection points.

[0036] Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for detecting pipeline leaks, characterized in that, include, Edge computing nodes are deployed along the pipeline to synchronously and continuously collect pipeline parameters and monitor the external environment of the pipeline using cooled infrared cameras to generate gas cloud images. Based on continuously collected pipeline parameters, a noise feature library and thermodynamic model are established. Based on a noise feature library and a thermodynamic model, abnormal feature information of edge computing nodes is extracted; Based on the extracted abnormal feature information of edge computing nodes, determine whether the pipeline is leaking abnormally; If an abnormality is detected, the extracted abnormal feature information of the edge computing node will be deeply fused with the generated gas cloud image to locate the leak immediately upon confirmation.

2. The pipeline leakage detection method according to claim 1, characterized in that, Edge computing nodes are deployed along the pipeline to synchronously and continuously collect pipeline parameters, including... Identify high-risk sections and critical nodes in the pipeline; Based on the identified high-risk pipe sections and key nodes, the deployment locations of edge computing nodes are planned along the pipeline route; On the planned edge computing nodes, an integrated sensor array is installed to measure the pipe parameters, including pipe medium parameters and vibration signals of the pipe wall. The medium parameters include the pressure and temperature inside the medium. At the same time, the infrared camera is calibrated so that its field of view includes the space above the pipe and the space at the pre-set node connection.

3. The pipeline leakage detection method according to claim 2, characterized in that, Based on continuously collected pipeline parameters, a noise feature library and thermodynamic model are established, including: The continuously collected pipeline parameters are archived to form historical data, which includes pressure, temperature, flow rate, and pipe wall vibration signals above the preset frequency at preset pipeline nodes.

4. The pipeline leakage detection method according to claim 3, characterized in that, The establishment of a noise feature library and thermodynamic model based on continuously collected pipeline parameters also includes, Based on the gas law and pipe thermodynamics theory, a thermodynamic model is established to describe the dynamic relationship between pressure and temperature inside a pipe. The collected historical normal data is input into the established thermodynamic model, and the key parameters in the thermodynamic model are optimized and adjusted to calibrate the thermodynamic model. During the real-time monitoring phase, the current measured pipeline temperature value is input into the calibrated thermodynamic model to correct the real-time monitored pressure value and calculate the expected theoretical pressure value under the current temperature conditions without leakage.

5. The pipeline leakage detection method according to claim 4, characterized in that, Based on the gas law and pipe thermodynamics, a thermodynamic model describing the dynamic relationship between pressure and temperature inside a pipe is established: ； in, To correct the pressure value to the reference temperature, This is the actual measured pressure value. For the selected reference temperature, It is relative to the reference temperature The corresponding reference pressure, This is the actual measured temperature value. There are two functions, which represent the gas compressibility under reference conditions and the gas compressibility under measured conditions, respectively.

6. The pipeline leakage detection method according to claim 4, characterized in that, The establishment of a noise feature library and thermodynamic model based on continuously collected pipeline parameters also includes, Analyze pipe wall vibration signals above the preset frequency to extract vibration and noise characteristics generated at pipe elbows, tees, and crosses under various specific working conditions; The extracted vibration and noise features are quantified and classified and archived according to the pipe fitting type and operating parameters to establish a noise feature library.

7. The pipeline leakage detection method according to claim 6, characterized in that, The establishment of a noise feature library and thermodynamic model based on continuously collected pipeline parameters also includes, Based on the historical data, various typical operating periods covering different combinations of flow and pressure were selected; Based on the selected typical working conditions, continuous vibration signal sequences were extracted from the connection points of elbows, tees, and crosses. Preprocess the captured continuous vibration signal sequence; Each segment of the preprocessed continuous vibration signal sequence is subjected to in-depth analysis, and each segment of the vibration signal is transformed into a digital feature vector that can fully characterize its properties. The transformed digital feature vectors are forcibly associated and labeled with their corresponding continuous vibration signal sequences to form sample data corresponding to feature vectors and working condition labels. The labeled sample data are stored uniformly to form a noise feature library.

8. The pipeline leakage detection method according to claim 6, characterized in that, Based on a noise feature library and a thermodynamic model, abnormal feature information of edge computing nodes is extracted, including: Edge computing nodes align and preprocess the corrected pressure signal and the original pipe wall vibration signal; Continuous wavelet transform is performed on the corrected pressure signal, and the transient micro-pressure drop characteristics are obtained by analyzing the subtle changes in the signal in the time and scale planes. The LSTM model is used to analyze the time series data of the original pipe wall vibration signal, learn the normal pattern and predict the trend, and obtain the vibration signal with abnormal trend. The source location of the original pipe wall vibration signal is determined in real time. When the original pipe wall vibration signal is located at the elbow, tee, or cross connection, the corresponding background noise fingerprint is called from the pre-established noise feature library according to the current operating parameters. Adaptive filtering technology is used to effectively remove this inherent structural noise from the real-time total signal, and the vibration residual signal after removing the background noise is obtained.

9. The pipeline leakage detection method according to claim 8, characterized in that, Based on the extracted anomaly feature information of edge computing nodes, it is determined whether the pipeline is leaking. Anomalies include: The extracted micro-pressure drop characteristics, abnormal vibration signals, and residual vibration signals after removing background noise are fused and analyzed. When multiple abnormal feature indicators deviate from the normal range in a coordinated manner, the edge computing node initially determines that there is a leakage anomaly and generates a high-confidence alarm event.

10. The pipeline leakage detection method according to claim 9, characterized in that, The extracted anomaly features from edge computing nodes are deeply fused with the generated gas cloud image to immediately locate the leak when it is confirmed. It receives alarm events from edge computing nodes and reads the sequence of atmospheric cloud images generated by infrared cameras within the same spatiotemporal range. It then aligns and correlates the abnormal feature information with the atmospheric cloud image sequence on a unified time axis and pipeline coordinate axis. The DS evidence theory algorithm is used to assign confidence levels to each type of anomalous feature information and cloud image sequence; After confirming the leak, the arrival time difference of the sound wave or pressure wave signal recorded by multiple edge computing nodes upstream and downstream of the leak point is called, and the preliminary location range of the leak point is calculated by the time difference positioning method. The calculated preliminary location range is superimposed and compared with the central region with the highest gas cloud concentration in the gas cloud image sequence. The location result is then verified and corrected using visual information. Based on the verified and corrected location results, a comprehensive alarm message is automatically generated and output, which includes the precise coordinates of the leak point, the leak intensity assessment, and the gas cloud diffusion range.

11. A pipeline leak detection system, characterized in that, include, The acquisition module is used to deploy edge computing nodes along the pipeline to synchronously and continuously acquire pipeline parameters and monitor the external environment of the pipeline through a cooled infrared camera to generate gas cloud images. A module is established to build a noise feature library and a thermodynamic model based on continuously collected pipeline parameters; The extraction module is used to extract abnormal feature information of edge computing nodes based on the noise feature library and thermodynamic model; The judgment module is used to determine whether the pipeline is leaking abnormally based on the extracted abnormal feature information of the edge computing nodes; The location module is used to perform deep fusion of the extracted abnormal feature information of the edge computing nodes with the generated gas cloud image if an abnormality is detected, so as to locate the leak immediately when it is confirmed.

Images