Fluid pipeline leakage positioning method based on multi-point pressure difference regression model
By installing pressure transmitters on pipelines, calculating voltage differences and drop rates, and using machine learning regression models to locate leak points, the problem of insufficient multi-source sensor data fusion in existing technologies is solved, achieving high-precision pipeline leak location.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-03-10
- Publication Date
- 2026-06-12
AI Technical Summary
Existing technologies struggle to efficiently integrate multi-source sensor data in pipeline leak detection, resulting in insufficient leak location accuracy, especially under complex operating conditions where precise prediction is difficult.
A method based on a multi-point pressure difference regression model is adopted. By installing pressure transmitters at the beginning, middle and end of the pipeline, the voltage difference and the drop amplitude are obtained and calculated. A voltage drop amplitude matrix is established and normalized. A machine learning regression model is used to train and locate the leak point.
It significantly improves the accuracy and precision of pipeline leak location, simplifies the model structure, has real-time computing capabilities, adapts to different pipeline topologies and fluid characteristics, and improves the efficiency of pipeline leak monitoring.
Smart Images
Figure CN122196962A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of fluid pipeline leak location technology, and in particular to a fluid pipeline leak location method based on a multi-point pressure difference regression model. Background Technology
[0002] Pipelines are considered one of the safest modes of transporting oil and gas compared to rail and road transport. However, during pipeline operation, oil and gas pipelines may experience malfunctions, including mechanical failures, operational errors, natural disasters, third-party incidents, and corrosion. Improper management of these malfunctions can lead to pipeline leaks or ruptures, causing significant losses and environmental hazards. Therefore, pipeline leak detection is crucial. Timely leak detection can reduce environmental pollution, prevent property damage, and prevent safety accidents. Machine learning regression algorithms have the advantage of efficiently fusing multi-source sensor data and accurately predicting leak locations through nonlinear modeling, significantly improving location accuracy. Compared to traditional physical models and simple statistical methods, machine learning regression algorithms can adapt to complex operating conditions, making them a key area of research in leak location. Summary of the Invention
[0003] This invention proposes a fluid pipeline leak location method based on a multi-point pressure difference regression model, which reduces the difficulty of pipeline leak detection and improves leak location accuracy.
[0004] The present invention adopts the following technical solution.
[0005] A method for locating leaks in fluid pipelines based on a multi-point pressure difference regression model is proposed. This method uses a machine learning regression model to fuse multi-source sensor data and uses nonlinear modeling to predict the leak location. Specifically, it includes the following steps:
[0006] Step 1: Install pressure transmitters at sampling points at the beginning, middle, and end of the pipeline, and obtain the average voltage output of the pressure transmitters after continuous monitoring for several seconds under stable pipeline operation.
[0007] Step 2: Obtain the average voltage output of the pressure transmitter after the pipeline leak has occurred, which is continuously monitored for several seconds.
[0008] Step 3: Calculate the output voltage difference of the pressure transmitter before and after the pipeline leak;
[0009] Step 4: Calculate the voltage drop of each pressure transmitter at different sampling points in the pipeline after a leak occurs, and establish a voltage drop matrix.
[0010] Step 5: Normalize the voltage drop magnitude matrix;
[0011] Step 6: Establish the sample matrix;
[0012] Step 7: Train the initial regression model using the training dataset. Input each sample feature from the test dataset into the initial regression model; use the sample matrix as training data and input it into the machine learning regression model for training.
[0013] Step 8: Use the trained regression model to locate the leak.
[0014] In step one, three sampling points are first set up at the beginning, middle, and end of the pipeline. A pressure transmitter is installed at each sampling point, and the average voltage output of the pressure transmitter at each sampling point is continuously monitored for 10 seconds under stable pipeline operation. , , ,in:
[0015] The average voltage output of the pressure transmitter at the first sampling point is continuously monitored for 10 seconds under stable pipeline operation.
[0016] The average voltage output of the intermediate sampling point pressure transmitter is continuously monitored for 10 seconds under stable pipeline operation.
[0017] The average voltage output of the pressure transmitter at the end sampling point is continuously monitored for 10 seconds under stable pipeline operation.
[0018] In step two, the average voltage output of three pressure transmitters is obtained after continuous monitoring for 10 seconds following the occurrence of a leak at the pipeline leak point. , , ,in:
[0019] The average voltage value is continuously collected for 10 seconds by the pressure transmitter at the beginning of the pipeline after a leak occurs at the nth leak point.
[0020] The average voltage value is the result of continuous 10-second sampling by the intermediate pressure transmitter after a leak occurs at the nth leak point in the pipeline.
[0021] The voltage average value is the result of continuous 10-second data collection by the end-point pressure transmitter after a leak occurs at the nth leak point in the pipeline.
[0022] and the distance from the nth leak point to the beginning of the pipeline. , where n is the leak point identifier, which takes the value (1, 2, 3, 4, ....).
[0023] Step 3: The formula for calculating the voltage difference of the pressure transmitters at the three sampling points after the pipeline leak is as follows:
[0024] ,
[0025] ,
[0026] ,
[0027] Let be the voltage difference at the beginning of the pipeline after a leak occurs at the nth leak point.
[0028] Let be the voltage difference in the middle of the pipeline after a leak occurs at the nth leak point.
[0029] Let be the voltage difference at the end of the pipeline after a leak occurs at the nth leak point.
[0030] In step four, the voltage drop of the pressure transmitters at three sampling points after leakage at different points in the pipeline is calculated, and a voltage drop matrix is established. The formula is as follows:
[0031] ,
[0032] in: , , ,
[0033] , , ,
[0034] , , ,
[0035] ... ,
[0036] , , ,
[0037] This represents the voltage drop at the beginning of the pipeline after a leak occurs at point 1.
[0038] This represents the voltage drop in the middle of the pipeline after a leak occurs at point 1.
[0039] This represents the voltage drop at the end of the pipeline after a leak occurs at point 1.
[0040] This represents the voltage drop at the beginning of the pipeline after a leak occurs at point 2.
[0041] This represents the voltage drop in the middle of the pipeline after a leak occurs at point 2.
[0042] This represents the voltage drop in the middle of the pipeline after a leak occurs at point 2.
[0043] This represents the voltage drop at the beginning of the pipeline after a leak occurs at point 3.
[0044] This represents the voltage drop in the middle of the pipeline after a leak occurs at point 3.
[0045] This represents the voltage drop in the middle of the pipeline after a leak occurs at point 3.
[0046] ... ,
[0047] Let be the voltage drop at the beginning of the pipeline after a leak occurs at the nth leak point.
[0048] This represents the voltage drop in the middle of the pipeline after a leak occurs at the third leak point.
[0049] This represents the voltage drop at the end of the pipeline after a leak occurs at the third leak point.
[0050] In step five, the voltage drop magnitude matrix is row-normalized using the following formula:
[0051] ,
[0052] in:
[0053] ,
[0054] ,
[0055] ,
[0056] ,
[0057] ,
[0058] ,
[0059] ,
[0060] ,
[0061] ,
[0062] ...
[0063] ,
[0064] ,
[0065] ,
[0066] for The maximum value among the three. for The minimum of the three.
[0067] In step six, a sample matrix is established.
[0068] ,
[0069] Each row represents a sample. Let X be the feature, and X be the target value, which corresponds to the distance from the leak point to the beginning. All sample features and their corresponding target values are divided into a training dataset and a test dataset. The training dataset includes multiple sample features and their corresponding target values. The test dataset includes multiple sample features and their corresponding target values.
[0070] In step seven, the initial regression model is trained using a training test dataset formed from the sample matrix. Each sample feature from the test dataset is input into the initial regression model to obtain the test result X for each sample feature in the test dataset. test The test results of each sample feature in the test dataset are compared with the target value of the sample to obtain the localization error. If the localization error is less than or equal to the required error, the initial detection model is determined as the leak localization model.
[0071] The formula for calculating the positioning error is as follows: ,
[0072] The length of the pipeline is the distance from the pressure transmitter at the beginning to the pressure transmitter at the end.
[0073] Step eight specifically involves: acquiring the voltage values collected by three pressure transmitters during pipeline leak monitoring operation, calculating and normalizing the matrix of voltage drop amplitude, inputting the feature to be measured into the leak location model, and thus locating the pipeline leak.
[0074] In step one, the fluid is tap water, and the installed pressure transmitter is directly connected to the water in the pipeline.
[0075] The pipe sidewall is equipped with an M20x0.5 internal threaded interface pipe. Figure 2 The fluid in the pipeline is connected to the pressure transmitter through the interface pipe.
[0076] like Figure 3 As shown, the pipeline is made of 304 stainless steel, with an outer diameter of 60mm and an inner diameter of 56mm. The pipeline transports tap water. In the simulation experiment, the total length of the pipeline for leak location was set to 66m, and the inlet water pressure was 0.083MPa. The first pressure transmitter was located at 0m, the second at 33m, and the third at 66m. The pressure transmitter parameters were: output 0-10V, range 0.15MPa, accuracy 0.2%FS, and power supply 24V.
[0077] In this invention, three pressure transmitters are used to monitor pressure changes caused by leaks in fluid delivery pipelines and output corresponding voltage values. After data processing, these values are used as sample features, and the distance to the leak point is used as the target value of the training set. By training a suitable regression model, leak location can be achieved without human intervention, thus improving the maintenance efficiency of water supply pipelines.
[0078] This invention has the following advantages: The fluid pipeline leakage location method based on the multi-point pressure difference regression model constructs a regression model by fusing voltage data from multiple measurement points, which can accurately capture the pressure gradient characteristics caused by leakage and significantly improve the location accuracy; at the same time, the model structure is simple and efficient, combining real-time computing capabilities and engineering practicality, and can be flexibly optimized for different pipeline topologies and fluid characteristics, maintaining stable performance even under complex working conditions, providing a reliable and efficient solution for pipeline leakage monitoring. Attached Figure Description
[0079] The present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments:
[0080] Appendix Figure 1 This is a schematic diagram of the pressure transmitter installation location and the sample sampling layout corresponding to the leakage point in an embodiment of the present invention;
[0081] Appendix Figure 2 This is a schematic diagram of a pressure transmitter installed at a pipeline in an embodiment of the present invention;
[0082] Appendix Figure 3 This is a schematic diagram of the installation of the pressure transmitter in the pipeline during the experimental simulation of an embodiment of the present invention;
[0083] Appendix Figure 4 This is a cross-sectional schematic diagram of a pressure transmitter installed on a pipeline in an embodiment of the present invention. Detailed Implementation
[0084] This invention is described in detail below with reference to the accompanying drawings, so as to fully reveal the technical features and key points of implementation of this invention. It should be noted that the embodiments described in this section are merely representative examples of typical application scenarios of this invention and are not an exhaustive list of the scope of protection. Any implementation scheme based on the core technical concept of this invention, achieved through equivalent substitution or adaptive adjustment, falls within the protection scope defined by the claims of this invention.
[0085] A method for locating leaks in fluid pipelines based on a multi-point pressure difference regression model is proposed. This method uses a machine learning regression model to fuse multi-source sensor data and uses nonlinear modeling to predict the leak location. Specifically, it includes the following steps:
[0086] Step 1: Install pressure transmitters at sampling points at the beginning, middle, and end of the pipeline, and obtain the average voltage output of the pressure transmitters after continuous monitoring for several seconds under stable pipeline operation.
[0087] Step 2: Obtain the average voltage output of the pressure transmitter after the pipeline leak has occurred, which is continuously monitored for several seconds.
[0088] Step 3: Calculate the output voltage difference of the pressure transmitter before and after the pipeline leak;
[0089] Step 4: Calculate the voltage drop of each pressure transmitter at different sampling points in the pipeline after a leak occurs, and establish a voltage drop matrix.
[0090] Step 5: Normalize the voltage drop magnitude matrix;
[0091] Step 6: Establish the sample matrix;
[0092] Step 7: Train the initial regression model using the training dataset. Input each sample feature from the test dataset into the initial regression model; use the sample matrix as training data and input it into the machine learning regression model for training.
[0093] Step 8: Use the trained regression model to locate the leak.
[0094] In step one, such as Figure 1 As shown, a pressure transmitter is installed at the beginning, middle and end of the fluid transport pipeline. The pressure transmitter outputs a voltage of 0~10V to monitor the pressure changes in the pipeline.
[0095] In step one, three sampling points are first set up at the beginning, middle, and end of the pipeline. A pressure transmitter is installed at each sampling point, and the average voltage output of the pressure transmitter at each sampling point is continuously monitored for 10 seconds under stable pipeline operation. , , ,in:
[0096] The average voltage output of the pressure transmitter at the first sampling point is continuously monitored for 10 seconds under stable pipeline operation.
[0097] The average voltage output of the intermediate sampling point pressure transmitter is continuously monitored for 10 seconds under stable pipeline operation.
[0098] The average voltage output of the pressure transmitter at the end sampling point is continuously monitored for 10 seconds under stable pipeline operation.
[0099] In step two, the average voltage of the first-end pressure transmitter is continuously collected for 10 seconds after a leak occurs at point n in the pipeline. The intermediate pressure transmitter continuously collects the average voltage for 10 seconds after a leak occurs at point n in the pipeline. The end pressure transmitter continuously collects the average voltage for 10 seconds after a leak occurs at point n in the pipeline. And the distance from the nth leak point to the beginning of the pipeline. .
[0100] In step two, the average voltage output of three pressure transmitters is obtained after continuous monitoring for 10 seconds following the occurrence of a leak at the pipeline leak point. , , ,in:
[0101] The average voltage value is continuously collected for 10 seconds by the pressure transmitter at the beginning of the pipeline after a leak occurs at the nth leak point.
[0102] The average voltage value is the result of continuous 10-second sampling by the intermediate pressure transmitter after a leak occurs at the nth leak point in the pipeline.
[0103] The voltage average value is the result of continuous 10-second data collection by the end-point pressure transmitter after a leak occurs at the nth leak point in the pipeline.
[0104] and the distance from the nth leak point to the beginning of the pipeline. , where n is the leak point identifier, which takes the value (1, 2, 3, 4, ....).
[0105] Step 3: The formula for calculating the voltage difference of the pressure transmitters at the three sampling points after the pipeline leak is as follows:
[0106] ,
[0107] ,
[0108] ,
[0109] Let be the voltage difference at the beginning of the pipeline after a leak occurs at the nth leak point.
[0110] Let be the voltage difference in the middle of the pipeline after a leak occurs at the nth leak point.
[0111] Let be the voltage difference at the end of the pipeline after a leak occurs at the nth leak point.
[0112] In step four, the voltage drop of the pressure transmitters at three sampling points after leakage at different points in the pipeline is calculated, and a voltage drop matrix is established. The formula is as follows:
[0113] ,
[0114] in:
[0115] , , ,
[0116] , , ,
[0117] , , ,
[0118] ... ,
[0119] , , ,
[0120] This represents the voltage drop at the beginning of the pipeline after a leak occurs at point 1.
[0121] This represents the voltage drop in the middle of the pipeline after a leak occurs at point 1.
[0122] This represents the voltage drop at the end of the pipeline after a leak occurs at point 1.
[0123] This represents the voltage drop at the beginning of the pipeline after a leak occurs at point 2.
[0124] This represents the voltage drop in the middle of the pipeline after a leak occurs at point 2.
[0125] This represents the voltage drop in the middle of the pipeline after a leak occurs at point 2.
[0126] This represents the voltage drop at the beginning of the pipeline after a leak occurs at point 3.
[0127] This represents the voltage drop in the middle of the pipeline after a leak occurs at point 3.
[0128] This represents the voltage drop in the middle of the pipeline after a leak occurs at point 3.
[0129] ... ,
[0130] Let be the voltage drop at the beginning of the pipeline after a leak occurs at the nth leak point.
[0131] This represents the voltage drop in the middle of the pipeline after a leak occurs at the third leak point.
[0132] This represents the voltage drop at the end of the pipeline after a leak occurs at the third leak point.
[0133] In step five, the voltage drop magnitude matrix is row-normalized using the following formula:
[0134] ,
[0135] in:
[0136] ,
[0137] ,
[0138] ,
[0139] ,
[0140] ,
[0141] ,
[0142] ,
[0143] ,
[0144] ,
[0145] ...
[0146] ,
[0147] ,
[0148] ,
[0149] for The maximum value among the three. for The minimum of the three.
[0150] In step six, a sample matrix is established.
[0151] ,
[0152] Each row represents a sample. Let X be the feature, and X be the target value, which corresponds to the distance from the leak point to the beginning. All sample features and their corresponding target values are divided into a training dataset and a test dataset. The training dataset includes multiple sample features and their corresponding target values. The test dataset includes multiple sample features and their corresponding target values.
[0153] In step seven, the initial regression model is trained using a training test dataset formed from the sample matrix. Each sample feature from the test dataset is input into the initial regression model to obtain the test result X for each sample feature in the test dataset. test The test results of each sample feature in the test dataset are compared with the target value of the sample to obtain the localization error. If the localization error is less than or equal to the required error, the initial detection model is determined as the leak localization model.
[0154] The formula for calculating the positioning error is as follows: ,
[0155] The length of the pipeline is the distance from the pressure transmitter at the beginning to the pressure transmitter at the end.
[0156] Step eight specifically involves: acquiring the voltage values collected by three pressure transmitters during pipeline leak monitoring operation, calculating and normalizing the matrix of voltage drop amplitude, inputting the feature to be measured into the leak location model, and thus locating the pipeline leak.
[0157] In step one, the fluid is tap water, and the installed pressure transmitter is directly connected to the water in the pipeline.
[0158] like Figure 4 As shown, the pipe sidewall is provided with an M20x0.5 internal thread interface pipe. Figure 2 The fluid in the pipeline is connected to the pressure transmitter through the interface pipe.
[0159] like Figure 3 As shown, the pipeline is made of 304 stainless steel, with an outer diameter of 60mm and an inner diameter of 56mm. The pipeline transports tap water. In the simulation experiment, the total length of the pipeline for leak location was set to 66m, and the inlet water pressure was 0.083MPa. The first pressure transmitter was located at 0m, the second at 33m, and the third at 66m. The pressure transmitter parameters were: output 0-10V, range 0.15MPa, accuracy 0.2%FS, and power supply 24V.
[0160] In this example, three pressure transmitters are used to monitor pressure changes caused by leaks in the fluid delivery pipeline and output corresponding voltage values. After data processing, these values are used as sample features, and the distance to the leak point is used as the target value of the training set. By training a suitable regression model, the leak can be located without human intervention, thus improving the maintenance efficiency of the water supply pipeline.
[0161] The above-disclosed embodiment is only one specific embodiment of the present invention. However, the present invention is not limited thereto. Any equivalent substitutions or modifications made by those skilled in the art within the scope of the technology disclosed in the present invention, based on the technical solution and inventive concept of the present invention, should be covered within the protection scope of the present invention.
Claims
1. A method for locating fluid pipeline leaks based on a multi-point pressure difference regression model, characterized in that: The method uses a machine learning regression model to fuse multi-source sensor data and nonlinear modeling to predict the leak location, specifically including the following steps; Step 1: Install pressure transmitters at sampling points at the beginning, middle, and end of the pipeline, and obtain the average voltage output of the pressure transmitters after continuous monitoring for several seconds under stable pipeline operation. Step 2: Obtain the average voltage output of the pressure transmitter after the pipeline leak has occurred, which is continuously monitored for several seconds. Step 3: Calculate the output voltage difference of the pressure transmitter before and after the pipeline leak; Step 4: Calculate the voltage drop of each pressure transmitter at different sampling points in the pipeline after a leak occurs, and establish a voltage drop matrix. Step 5: Normalize the voltage drop magnitude matrix; Step 6: Establish the sample matrix; Step 7: Train the initial regression model using the training dataset, and input the features of each sample from the test dataset into the initial regression model respectively; Step 8: Use the trained regression model to locate the leak.
2. The fluid pipeline leakage location method based on a multi-point pressure difference regression model according to claim 1, characterized in that: In step one, three sampling points are first set up at the beginning, middle, and end of the pipeline. A pressure transmitter is installed at each sampling point, and the average voltage output of the pressure transmitter at each sampling point is continuously monitored for 10 seconds under stable pipeline operation. , , ,in: The pressure transmitter at the first sampling point continuously monitors the average voltage output for 10 seconds under stable pipeline operation. The average voltage output of the intermediate sampling point pressure transmitter is continuously monitored for 10 seconds under stable pipeline operation. The average voltage output of the pressure transmitter at the end sampling point is continuously monitored for 10 seconds under stable pipeline operation. In step two, the average voltage output of three pressure transmitters is obtained after continuous monitoring for 10 seconds following the occurrence of a leak at the pipeline leak point. , , ,in: The average voltage value is continuously collected for 10 seconds by the pressure transmitter at the beginning of the pipeline after a leak occurs at the nth leak point. The average voltage value is the result of continuous 10-second sampling by the intermediate pressure transmitter after a leak occurs at the nth leak point in the pipeline. The voltage average value is the result of continuous 10-second data collection by the end-point pressure transmitter after a leak occurs at the nth leak point in the pipeline. and the distance from the nth leak point to the beginning of the pipeline. , where n is the leak point identifier, which takes the value (1, 2, 3, 4, ....).
3. The fluid pipeline leakage location method based on a multi-point pressure difference regression model according to claim 2, characterized in that: Step 3: The formula for calculating the voltage difference of the pressure transmitters at the three sampling points after the pipeline leak is as follows: ， ， ， Let be the voltage difference at the beginning of the pipeline after a leak occurs at the nth leak point. Let be the voltage difference in the middle of the pipeline after a leak occurs at the nth leak point. Let be the voltage difference at the end of the pipeline after a leak occurs at the nth leak point.
4. The fluid pipeline leakage location method based on a multi-point pressure difference regression model according to claim 3, characterized in that: In step four, the voltage drop of the pressure transmitters at three sampling points after leakage at different points in the pipeline is calculated, and a voltage drop matrix is established. The formula is as follows: ， in: ， ， ， Let be the voltage drop at the beginning of the pipeline after a leak occurs at the nth leak point. This represents the voltage drop in the middle of the pipeline after a leak occurs at the third leak point. This represents the voltage drop at the end of the pipeline after a leak occurs at the third leak point.
5. The fluid pipeline leakage location method based on a multi-point pressure difference regression model according to claim 4, characterized in that: In step five, the voltage drop magnitude matrix is row-normalized using the following formula: ， in: ， ， ， for The maximum value among the three. for The minimum of the three.
6. The fluid pipeline leakage location method based on a multi-point pressure difference regression model according to claim 5, characterized in that: In step six, a sample matrix is established. ， Each row represents a sample. Let X be the feature, and X be the target value, which corresponds to the distance from the leak point to the beginning. All sample features and their corresponding target values are divided into a training dataset and a test dataset. The training dataset includes multiple sample features and their corresponding target values. The test dataset includes multiple sample features and their corresponding target values.
7. The fluid pipeline leakage location method based on a multi-point pressure difference regression model according to claim 6, characterized in that: In step seven, the initial regression model is trained using a training test dataset formed from the sample matrix. Each sample feature from the test dataset is input into the initial regression model to obtain the test result X for each sample feature in the test dataset. test The test results of each sample feature in the test dataset are compared with the target value of the sample to obtain the localization error. If the localization error is less than or equal to the required error, the initial detection model is determined as the leak localization model. The formula for calculating the positioning error is as follows: , The length of the pipeline is the distance from the pressure transmitter at the beginning to the pressure transmitter at the end.
8. The fluid pipeline leakage location method based on a multi-point pressure difference regression model according to claim 7, characterized in that: Step eight specifically involves: acquiring the voltage values collected by three pressure transmitters during pipeline leak monitoring operation, calculating and normalizing the matrix of voltage drop amplitude, inputting the feature to be measured into the leak location model, and thus locating the pipeline leak.
9. The fluid pipeline leakage location method based on a multi-point pressure difference regression model according to claim 2, characterized in that: In step one, the fluid is tap water, and the installed pressure transmitter is directly connected to the water in the pipeline.
10. The fluid pipeline leakage location method based on a multi-point pressure difference regression model according to claim 9, characterized in that: An interface pipe is provided on the side wall of the pipeline, and the fluid in the pipeline is connected to the pressure transmitter through the interface pipe.