CN113803647A - Pipeline leakage detection method based on fusion of knowledge characteristics and mixed model - Google Patents
Pipeline leakage detection method based on fusion of knowledge characteristics and mixed model Download PDFInfo
- Publication number
- CN113803647A CN113803647A CN202110979611.5A CN202110979611A CN113803647A CN 113803647 A CN113803647 A CN 113803647A CN 202110979611 A CN202110979611 A CN 202110979611A CN 113803647 A CN113803647 A CN 113803647A
- Authority
- CN
- China
- Prior art keywords
- pipeline
- model
- knowledge
- pressure
- flow
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F17—STORING OR DISTRIBUTING GASES OR LIQUIDS
- F17D—PIPE-LINE SYSTEMS; PIPE-LINES
- F17D5/00—Protection or supervision of installations
- F17D5/02—Preventing, monitoring, or locating loss
- F17D5/06—Preventing, monitoring, or locating loss using electric or acoustic means
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/25—Fusion techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N5/00—Computing arrangements using knowledge-based models
- G06N5/02—Knowledge representation; Symbolic representation
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Evolutionary Computation (AREA)
- General Physics & Mathematics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Molecular Biology (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Acoustics & Sound (AREA)
- Mechanical Engineering (AREA)
- Examining Or Testing Airtightness (AREA)
Abstract
The invention discloses a pipeline leakage detection method based on fusion of knowledge characteristics and a hybrid model, which comprises the following steps: firstly, establishing a pipeline hybrid modeling method fusing a real-time transient mechanism model and a convolutional neural network data driving model to accurately predict the operation condition of a pipeline; secondly, embedding a knowledge characteristic model of the pipeline conveying process into the mixed model aiming at the ubiquitous uncertainty of the process, and improving the robustness of the model and the prediction capability of the pipeline running state; and finally, analyzing the deviation obtained by comparing the prediction output of the mixed model with the actual measurement value, and judging whether the pipeline leaks. The method improves the accuracy of the model, and solves the problems of false alarm, missing report and the like caused by the change of working conditions and small leakage characteristics.
Description
Technical Field
The invention relates to the field of pipeline leakage detection, in particular to a pipeline leakage detection method based on fusion of knowledge characteristics and a mixed model.
Background
Pipelines are an important component of industrial, agricultural, and municipal infrastructure, but pipelines often create a number of pipeline leakage problems due to aging, corrosion, weld defects, third party damage, and the like. Pipeline leak detection therefore takes into account the hydraulics of the pipeline transport, as well as process uncertainty.
In daily life and industrial production, pipeline transportation plays a crucial role. The pipeline transportation has the advantages of large transportation capacity, easily controlled transportation process, low cost, simple construction, no influence of factors such as ground, climate and the like, and strong environmental adaptability. However, the problems of corrosion, aging, connection and the like of the pipeline exist, the influence of various external forces acts on the pipeline, the problem of pipeline leakage generally exists, and resource waste and huge economic loss are caused. Therefore, it is very important to detect the pipeline leakage. To this end, the industry and academia have proposed many methods for leak detection of pipelines. Modern pipeline leakage detection methods are generally based on external or internal measured variables of the pipeline to build a detection model to enable real-time online pipeline leakage detection. External methods are generally based on detecting the characteristics of leaks outside the pipe, such as acoustic based methods, optical fiber based methods. The internal methods are generally based on operating parameters such as pipeline pressure, flow, temperature, etc., such as real-time transient model methods, negative pressure wave methods. From the model establishment perspective, the method based on the mechanism model and the data-driven model is two main methods for detecting the pipeline leakage, and the two methods have the advantages and the disadvantages respectively. In general, methods based on mechanistic models rely on the generation and evaluation of deviations. The performance of these methods is highly dependent on the model parameters and the accuracy of the sensors, and requires a lot of simulation and calibration work. The data-driven approach does not require any specific in-depth knowledge of pipeline hydraulics, and only requires learning from collected historical data, plus some statistical or pattern recognition tools. Among them, the applications based on Artificial Neural Network (ANN) and Support Vector Machine (SVM) are the most.
Disclosure of Invention
The invention provides a pipeline leakage detection method based on fusion of knowledge characteristics and a mixed model.
Firstly, establishing a pipeline hybrid modeling method fusing a real-time transient mechanism model and a convolutional neural network data driving model to accurately predict the operation condition of a pipeline; secondly, embedding a knowledge characteristic model of the pipeline conveying process into the mixed model aiming at the ubiquitous uncertainty of the process, and improving the robustness of the model and the prediction capability of the pipeline running state; and finally, analyzing the deviation obtained by comparing the prediction output of the mixed model with the actual measurement value to judge whether the pipeline leaks. The model obtained by the method is high in precision and good in leakage detection effect.
The purpose of the invention is: due to uncertain operation conditions of the pipeline, inaccurate relevant parameters of the model and poor data matching in the actual process, the prediction of a mechanism model generates deviation, and thus the pipeline leakage is misinformed. The invention establishes a mixed model combining a data driving model fusing knowledge characteristics and a pipeline transient mechanism model, wherein a convolutional neural network fusing knowledge characteristics is used as a pipeline data driving model to correct the pipeline transient model, so that the model precision is improved, and the problems of false alarm, missing report and the like caused by change of working conditions and small leakage characteristics are solved.
The technical solution of the invention is as follows: CNN is used as a data driving model to correct a real-time transient pipeline mechanism model, so that the accuracy of the model is improved, and the problems of misinformation, missing report and the like caused by change of working conditions and small leakage characteristics are solved. Along with the increase of the running time of the pipeline, the resistance structure in the pipeline changes, the output of the mixed model also generates deviation and generates a false alarm phenomenon, so that the pipeline friction knowledge is merged into CNN, and the robustness of the model is enhanced.
The invention effectively combines a mechanism model and a data driving model, and fuses knowledge characteristics on the basis to effectively and accurately predict the running state of the pipeline. Firstly, establishing a simplified Real Time Transient Model (RTTM) mechanism Model based on a hydraulics principle; secondly, extracting implicit characteristic information of pipeline operation data through a Convolutional Neural Network (CNN) data driving model, and supplementing an RTTM mechanism model; in consideration of process uncertainty, the method introduces a learning element of domain knowledge on the basis of a CNN data-driven model, and provides a pipeline leakage detection method integrating data, knowledge and mechanisms.
A pipeline leakage detection method based on fusion of knowledge characteristics and a hybrid model comprises the following steps:
step 4, establishing a data-driven model fusing pipeline knowledge and a mechanism and data-driven hybrid model;
step 5, inputting a convolutional neural network in a data driving model fusing pipeline knowledge into a boundary condition input value of a pipeline fluid real-time transient pipeline mechanism model, wherein the input value is the head end pressure and the tail end flow when the pipeline runs, and the head end flow and the tail end pressure when the pipeline runs can be obtained through the model calculation in the step 2;
step 6, obtaining pipeline operation data with real-time performance through the mixed model based on mechanism and data driving in the step 4; when the pipeline leaks, the head end flow and the tail end pressure of the pipeline are obtained after calculation in the step 5 and are used as model calculation values, the model calculation values are compared with actual measurement values of the sensors, and when the deviation exceeds a given threshold value, the leakage is indicated to occur, and the calculation is represented as:
wherein Q is1Is an actual measured value of the flow at the head end of the pipeline,
calculating a value for head end flow, P2Is an actual measurement of the pressure at the end of the pipeline,
calculating the value, σ, for the end pressure1、σ2Respectively, set flow and pressure thresholds.
In step 1, establishing a pipeline fluid real-time transient pipeline mechanism model, specifically comprising:
establishing a continuity equation and a motion momentum equation for the motion of the pipeline fluid:
wherein, the formula (1) is a flow continuity equation, the formula (2) is a motion momentum equation, and in the formula, P is the average pressure of the section of the pipeline; v is the average flow velocity of the pipeline section; ρ is the fluid average density; f is the hydraulic friction coefficient; ag is the acceleration of gravity; θ is the angle of the fluid to the horizontal axis; d is the inner diameter of the pipe; a is the pressure wave propagation velocity; t is time; x is the along-line distance.
In the step 2, a pipeline fluid real-time transient pipeline mechanism model in the step 1) is optimized by adopting a characteristic line method, and the method specifically comprises the following steps:
solving the nonlinear partial differential equations (1) and (2) in the step (1) by adopting a characteristic line method; the characteristic line method needs to set boundary conditions and initial values of the pipeline, so that the method can be obtained by the following equations (1) and (2) by considering a steady-state model of the pipeline and not considering the influence of time:
the equations (4) and (5) are steady-state models of the pipeline, are ordinary differential equations, and can determine initial values of the real-time transient pipeline mechanism model by solving by adopting a fourth-order Runge Kutta method; the head end and the tail end of the pipeline are used as boundary conditions, and input values of the boundary conditions are set as head end pressure and tail end flow.
based on the pipeline basic principle, the knowledge characteristics of the equivalent friction coefficient in the pipeline are extracted, and the method specifically comprises the following steps:
the pipeline steady state model equations (4) and (5) can be used to obtain:
and (3) carrying out integral solution on equation set formulas (6) and (7) to obtain the relation between the equivalent friction coefficient in the pipeline and the parameter in the pipeline:
wherein Q is1For head end flow of the pipeline, Q2For end flows of pipes, P1For head pressure of the pipeline, P2In order to obtain the pressure at the end of the pipeline,
in order to calculate the equivalent friction coefficient according to the head end flow,
the equivalent friction coefficient is calculated according to the terminal flow.
In step 4, establishing a data-driven model fusing pipeline knowledge and a mixed model based on mechanism and data driving, specifically comprising:
fusing the knowledge characteristics of the equivalent friction coefficient in the pipeline extracted in the step (3) into a one-dimensional convolutional neural network (1DCNN), and obtaining a data driving model of the fused knowledge through a full connection layer after combination;
and (3) correcting the mechanism model by a data-driven model with knowledge fused as shown in formulas (4) and (5) to obtain a mixed model based on the mechanism and the data drive.
Compared with the prior art, the invention has the following advantages:
the CNN is used as a data driving model to correct a real-time transient pipeline mechanism model, so that the accuracy of the model is improved, and the problems of misinformation, missing report and the like caused by change of working conditions and small leakage characteristics are solved. Along with the increase of the running time of the pipeline, the resistance structure in the pipeline changes, the output of the mixed model also generates deviation and generates a false alarm phenomenon, so that the pipeline friction knowledge is merged into CNN, and the robustness of the model is enhanced. The invention integrates the pipeline knowledge into the mixed model, and the precision of the corrected real-time transient pipeline mixed model can not be greatly reduced along with the increase of the pipeline running time and the change of the in-pipe resistance structure. The method improves the accuracy of the model, and solves the problems of false alarm, missing report and the like caused by the change of working conditions and small leakage characteristics.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the principles of the invention and not to limit the invention. In the drawings:
fig. 1 is a design diagram of an experimental pipeline of the present invention, wherein upstream reservoir, downstream reservoir, controller, pressure and flow measurement, and leakpoint, water recovery point, are upstream reservoir, downstream reservoir, controller, and controller.
FIG. 2 is a diagram of a real-time transient pipeline hybrid model incorporating a convolutional neural network.
Fig. 3 is a flow chart of the present invention.
FIG. 4 is a graph of RTTM and CNN-RTTM versus actual pressure when changing the pipeline inlet pressure.
FIG. 5 is a graph of CNN-RTTM and K-CNN-RTTM versus actual pressure for varying pipe inlet pressure and coefficient of friction.
Detailed Description
The following detailed description of the embodiments of the present invention will be provided with reference to the drawings and examples, so that how to apply the technical means to solve the technical problems and achieve the technical effects can be fully understood and implemented.
As shown in fig. 1 to 3, a method for detecting pipeline leakage based on fusion of knowledge characteristics and a hybrid model includes the following steps:
Establishing a continuity equation, a motion momentum equation and an energy equation for the motion of the pipeline fluid:
wherein, the formula (1) is a flow continuity equation, the formula (2) is a motion momentum equation, the formula (3) is an energy equation, and in the formula, P is the average pressure of the section of the pipeline; v is the average flow velocity of the pipeline section; ρ is the fluid average density; f is the hydraulic friction coefficient; ag is the acceleration of gravity; θ is the angle of the fluid to the horizontal axis; d is the inner diameter of the pipe; a is the pressure wave propagation velocity; t is time; x is the along-line distance; cVIs the heat energy of the liquid in the pipeline; k is the ground thermal conductivity; t (r) is a function of temperature; t is the liquid temperature; r is the radial distance from the center of the pipe diameter at a particular location along the pipeline.
When the fluid is a liquid, the effect of temperature on the fluid flow, i.e., the energy change in the pipe, can be ignored, and therefore equation (3) can be ignored.
And 2, solving the nonlinear partial differential equation set in the step 1 by adopting a characteristic line method. The characteristic line method needs to set boundary conditions and initial values of the pipeline, so that a pipeline steady-state model is considered, and the influence of time is not considered. The formula (1) and (2) can be used for obtaining:
the equations (4) and (5) are steady-state models of the pipeline, are ordinary differential equations, and can determine initial values of the real-time transient pipeline mechanism model by solving with a fourth-order Runge Kutta method. The head end and the tail end of the pipeline are used as boundary conditions, input values of the boundary conditions are set to be head end pressure and tail end flow, and the head end flow and the tail end pressure of the pipeline can be obtained through model calculation.
And 3, establishing a mechanism and data driving based hybrid model. The mechanism model is modified by taking the convolutional neural network as a data driving model as shown in formulas (4) and (5). The input of the convolutional neural network is the boundary condition input value of the transient pipeline mechanism model, namely the head end pressure and the tail end flow when the pipeline runs. The input is a time sequence of pressure and flow extracted by the sensor, a one-dimensional convolutional neural network (1DCNN) is adopted to analyze and calculate the time sequence, and the output of the 1DCNN is the difference between a transient pipeline mechanism model and an actual measurement value.
And 4, extracting knowledge characteristics based on the pipeline basic principle.
The steady state model equations (4) and (5) of the pipeline can be obtained
And (3) carrying out integral solution on equation set formulas (6) and (7) to obtain the relation between the equivalent friction coefficient in the pipeline and the parameter in the pipeline:
wherein Q is1For head end flow of the pipeline, Q2For end flows of pipes, P1For head pressure of the pipeline, P2In order to obtain the pressure at the end of the pipeline,
in order to calculate the equivalent friction coefficient according to the head end flow,
the equivalent friction coefficient is calculated according to the terminal flow.
And 5, fusing the pipeline knowledge extracted in the step 4 into the 1DCNN, namely combining the characteristics extracted by the 1DCNN with the pipeline knowledge and then passing through a full connection layer.
And 6, correcting the real-time transient pipeline mechanism model through the model established in the step 5, so as to obtain pipeline operation data with real-time performance. When the pipeline leaks, the head end flow and the tail end pressure of the pipeline are obtained after calculation, the model prediction value is compared with the actual measurement value, and when the deviation exceeds a given threshold value, the leakage is indicated, which can be expressed as:
wherein Q is1For a measurement of the flow at the head end of the pipeline,
calculating a value for head end flow, P2Is a measure of the pressure at the end of the pipe,
calculating the value, σ, for the end pressure1、σ2Respectively, set flow and pressure thresholds.
The normal flow data under different inlet pressure conditions are predicted, the inlet pressure of the pipeline is randomly changed at 30s, and the result is shown in fig. 4, and it can be seen that the model obtained after the convolutional neural network is corrected can better predict the actual condition. If the running time of the pipeline is too long and the internal parameters are changed, the prediction effect of the original model on the actual process is poor, and if the deviation value exceeds a set threshold value, false alarm can occur. Therefore, for the situation, firstly, the knowledge, namely the equivalent friction coefficient, of the pipeline is extracted, the knowledge is fused with the features extracted before the convolutional neural network Flatten layer, and then, an accurate pipeline model can be established for a new working condition through a new full-connection layer. When the pipeline runs for 30s, the inlet pressure and the friction coefficient of the pipeline are randomly changed, and the K-CNN-RTTM has better prediction effect as can be found from the graph shown in figure 5.
The length of the whole section of pipeline is 1600m, the pipeline comprises two elbows, the pipe diameter is 0.05m, the propagation speed of negative pressure waves of the pipeline is 1000m/s, the liquid in the pipeline is water with the temperature of 20 ℃, 4 leakage nodes are arranged in the pipeline, the diameter of each leakage pore is 1.2mm, and the leakage point position controls the leakage of each node by a ball valve. Randomly sampling between 8bar and 25bar as the inlet pressure of the pipeline, and sampling the pressure and the flow at two ends of the pipeline by using sensors at the head end and the tail end, wherein the sampling time is 30s, and the sampling frequency is 33 Hz. Therefore, the information collected by the sensor comprises 1001 points, the pressure signal at the head end of the pipeline and the flow signal at the tail end of the pipeline are connected as a sample after noise is added, and each sample comprises 2002 data points. And adding noise to the tail end pressure signal and the head end flow signal of the pipeline and storing the signals as real measurement data output by the pipeline. Samples 400 were generated at different pressures for normal operation, with 200 sets of tubes having an absolute roughness of 0.025mm and the remaining 200 sets of tubes having an absolute roughness of between 0.025mm and 0.05 mm. Leak data 500 sets are generated, with 100 and 200 sets of data for each leak location.
A real-time transient pipeline mechanism model is constructed through Matlab, head end flow and tail end pressure of a pipeline can be obtained through calculation of head end pressure and tail end flow of the pipeline, the mechanism model is corrected through a convolutional neural network, and 200 groups of pipeline data under different pressure flow conditions are adopted for neural network training. The built convolutional neural network comprises 3 convolutional layers, a pooling layer is arranged behind each convolutional layer, and output is obtained through a Flatten layer and a full-connection layer. After training, the accuracy of the obtained training set is 99.5%, and the accuracy of the obtained test set is 99.2%.
When the leakage changes, the leakage signal is transmitted to the head end and the tail end of the pipeline from the leakage point through the negative pressure wave, the pressure threshold value is selected to be 0.15bar, the flow threshold value is selected to be 0.000025m3And/s, when the difference between the output value of the hybrid model and the measured value of the simulation software exceeds a threshold value, the leakage is considered to occur, and the accuracy rate of the leakage judgment is shown in the table 1.
TABLE 1 leak determination accuracy
Claims (5)
1. A pipeline leakage detection method based on fusion of knowledge characteristics and a hybrid model comprises the following steps:
step 1, establishing a pipeline fluid real-time transient pipeline mechanism model;
step 2, optimizing the pipeline fluid real-time transient pipeline mechanism model in the step 1 by adopting a characteristic line method;
step 3, extracting knowledge characteristics of equivalent friction coefficient in the pipeline based on the pipeline basic principle;
step 4, establishing a data-driven model fusing pipeline knowledge and a mechanism and data-driven hybrid model;
step 5, inputting a convolutional neural network in a data driving model fusing pipeline knowledge into a boundary condition input value of a pipeline fluid real-time transient pipeline mechanism model, wherein the input value is the head end pressure and the tail end flow when the pipeline runs, and the head end flow and the tail end pressure when the pipeline runs can be obtained through the model calculation in the step 2;
step 6, obtaining pipeline operation data with real-time performance through the mixed model based on mechanism and data driving in the step 4; when the pipeline leaks, the head end flow and the tail end pressure of the pipeline are obtained after calculation in the step 5 and are used as model calculation values, the model calculation values are compared with actual measurement values of the sensors, and when the deviation exceeds a given threshold value, the leakage is indicated to occur, and the calculation is represented as:
wherein Q is1Is an actual measured value of the flow at the head end of the pipeline,
calculating a value for head end flow, P2Is an actual measurement of the pressure at the end of the pipeline,
calculating the value, σ, for the end pressure1、σ2Respectively, set flow and pressure thresholds.
2. The pipeline leakage detection method based on the fusion of the knowledge characteristics and the hybrid model according to claim 1, wherein in the step 1, the establishment of the pipeline fluid real-time transient pipeline mechanism model specifically comprises:
establishing a continuity equation and a motion momentum equation for the motion of the pipeline fluid:
wherein, the formula (1) is a flow continuity equation, the formula (2) is a motion momentum equation, and in the formula, P is the average pressure of the section of the pipeline; v is the average flow velocity of the pipeline section; ρ is the fluid average density; f is the hydraulic friction coefficient; ag is the acceleration of gravity; θ is the angle of the fluid to the horizontal axis; d is the inner diameter of the pipe; a is the pressure wave propagation velocity; t is time; x is the along-line distance.
3. The pipeline leak detection method based on fusion of knowledge features and a hybrid model according to claim 2,
in the step 2, a pipeline fluid real-time transient pipeline mechanism model in the step 1) is optimized by adopting a characteristic line method, and the method specifically comprises the following steps:
solving the nonlinear partial differential equations (1) and (2) in the step (1) by adopting a characteristic line method; the characteristic line method needs to set boundary conditions and initial values of the pipeline, so that the method can be obtained by the following equations (1) and (2) by considering a steady-state model of the pipeline and not considering the influence of time:
the equations (4) and (5) are steady-state models of the pipeline, are ordinary differential equations, and can determine initial values of the real-time transient pipeline mechanism model by solving by adopting a fourth-order Runge Kutta method; the head end and the tail end of the pipeline are used as boundary conditions, and input values of the boundary conditions are set as head end pressure and tail end flow.
4. The method for detecting the pipeline leakage based on the fusion of the knowledge characteristics and the hybrid model according to claim 1, wherein in the step 3, the knowledge characteristics of the equivalent friction coefficient in the pipeline are extracted based on the pipeline basic principle, and the method specifically comprises the following steps;
based on the pipeline basic principle, the knowledge characteristics of the equivalent friction coefficient in the pipeline are extracted, and the method specifically comprises the following steps:
the pipeline steady state model equations (4) and (5) can be used to obtain:
and (3) carrying out integral solution on equation set formulas (6) and (7) to obtain the relation between the equivalent friction coefficient in the pipeline and the parameter in the pipeline:
wherein Q is1Is a tubeHead end flow, Q2For end flows of pipes, P1For head pressure of the pipeline, P2In order to obtain the pressure at the end of the pipeline,
in order to calculate the equivalent friction coefficient according to the head end flow,
the equivalent friction coefficient is calculated according to the terminal flow.
5. The method for detecting the pipeline leakage based on the fusion of the knowledge characteristics and the hybrid model according to claim 1, wherein in the step 4, a data-driven model fusing the pipeline knowledge and a hybrid model based on the mechanism and the data drive are established, and specifically the method comprises the following steps:
fusing the knowledge characteristics of the equivalent friction coefficient in the pipeline extracted in the step (3) into a one-dimensional convolutional neural network (1DCNN), and obtaining a data driving model of the fused knowledge through a full connection layer after combination;
and (3) correcting the mechanism model by a data-driven model with knowledge fused as shown in formulas (4) and (5) to obtain a mixed model based on the mechanism and the data drive.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110979611.5A CN113803647B (en) | 2021-08-25 | 2021-08-25 | Pipeline leakage detection method based on fusion of knowledge features and hybrid model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110979611.5A CN113803647B (en) | 2021-08-25 | 2021-08-25 | Pipeline leakage detection method based on fusion of knowledge features and hybrid model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113803647A true CN113803647A (en) | 2021-12-17 |
| CN113803647B CN113803647B (en) | 2023-07-04 |
Family
ID=78941844
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110979611.5A Active CN113803647B (en) | 2021-08-25 | 2021-08-25 | Pipeline leakage detection method based on fusion of knowledge features and hybrid model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113803647B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114636113A (en) * | 2022-02-22 | 2022-06-17 | 中国特种设备检测研究院 | Oil depot pipeline leakage detection method |
Citations (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH1125102A (en) * | 1997-06-30 | 1999-01-29 | Atr Chinou Eizo Tsushin Kenkyusho:Kk | Transmission method with blended knowledge and interest |
| CA2598949A1 (en) * | 2006-09-01 | 2008-03-01 | Areva Np Gmbh | Collector line for leakage monitoring and leakage location |
| CN101592288A (en) * | 2009-05-22 | 2009-12-02 | 重庆大学 | A kind of method for identifying pipeline leakage |
| CN101761780A (en) * | 2010-01-11 | 2010-06-30 | 中国石油大学(华东) | Gas pipeline leakage detecting and positioning device and method thereof |
| US20110167914A1 (en) * | 2008-06-27 | 2011-07-14 | Jeffrey Earle Sutherland | Integrated multi-sensor non-destructive testing |
| CN104750742A (en) * | 2013-12-31 | 2015-07-01 | 南京理工大学常熟研究院有限公司 | Fault diagnosis method and system for heading machine hydraulic system |
| CN105042339A (en) * | 2015-06-03 | 2015-11-11 | 中国石化销售有限公司华东分公司 | Leakage rate estimation system and method for product oil pipelines based on zero dimension |
| CN106015951A (en) * | 2016-07-07 | 2016-10-12 | 北京信息科技大学 | Gas transmission pipeline leakage detection system and method suitable for multiple state changes |
| US20180259921A1 (en) * | 2015-11-26 | 2018-09-13 | Jiangnan University | Data-Difference-Driven Self-Learning Dynamic Optimization Method For Batch Process |
| US20180347763A1 (en) * | 2017-05-31 | 2018-12-06 | Schneider Electric Software, Llc | Machine learning detection of pipeline ruptures |
| US20180365555A1 (en) * | 2016-12-22 | 2018-12-20 | Naveed Aslam | Artificial intelligence based algorithm for predicting pipeline leak and corrosion detection |
| JP2019095292A (en) * | 2017-11-22 | 2019-06-20 | 株式会社日立製作所 | Water leakage detection system and water leakage detection method |
| CN110185939A (en) * | 2019-05-16 | 2019-08-30 | 西北工业大学 | Gas pipe leakage recognition methods based on convolutional neural networks |
| CN110296327A (en) * | 2019-06-19 | 2019-10-01 | 常州大学 | A kind of pipeline leakage detection method based on transient flow frequency response analysis |
| CN111059477A (en) * | 2019-12-23 | 2020-04-24 | 浙江工业大学 | Double-layer framework based reverse source-seeking chemical pipeline leakage detection and positioning method |
| CN111125867A (en) * | 2019-11-04 | 2020-05-08 | 浙江工业大学 | Establishment and calculation method of chemical production pipeline real-time transient model based on chaotic particle swarm |
| US20200210826A1 (en) * | 2018-12-29 | 2020-07-02 | Northeastern University | Intelligent analysis system using magnetic flux leakage data in pipeline inner inspection |
| CN111915579A (en) * | 2020-07-24 | 2020-11-10 | 天津大学 | Liquid leakage detection method based on Gaussian mixture model and convolutional neural network |
| CN112036440A (en) * | 2020-07-31 | 2020-12-04 | 山东科技大学 | Satellite attitude control system fault diagnosis and early warning method based on random forest |
| CN112610903A (en) * | 2020-12-10 | 2021-04-06 | 合肥学院 | Water supply pipe network leakage positioning method based on deep neural network model |
| US20210116076A1 (en) * | 2019-10-22 | 2021-04-22 | Eog Resources, Inc. | Anomaly detection in pipelines and flowlines |
| US20210178697A1 (en) * | 2019-12-17 | 2021-06-17 | Northrop Grumman Systems Corporation | Machine-learning-based additive manufacturing using manufacturing data |
| AU2019383049A1 (en) * | 2018-11-23 | 2021-07-01 | The University Of Adelaide | Method and system to analyse pipeline condition |
| CN113077091A (en) * | 2021-04-09 | 2021-07-06 | 中国石油大学(北京) | Machine learning-based leakage rate probability distribution prediction method and system |
| US20230028276A1 (en) * | 2019-11-26 | 2023-01-26 | Technische Universität Berlin | Forecasting industrial aging processes with machine learning methods |
-
2021
- 2021-08-25 CN CN202110979611.5A patent/CN113803647B/en active Active
Patent Citations (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH1125102A (en) * | 1997-06-30 | 1999-01-29 | Atr Chinou Eizo Tsushin Kenkyusho:Kk | Transmission method with blended knowledge and interest |
| CA2598949A1 (en) * | 2006-09-01 | 2008-03-01 | Areva Np Gmbh | Collector line for leakage monitoring and leakage location |
| US20110167914A1 (en) * | 2008-06-27 | 2011-07-14 | Jeffrey Earle Sutherland | Integrated multi-sensor non-destructive testing |
| CN101592288A (en) * | 2009-05-22 | 2009-12-02 | 重庆大学 | A kind of method for identifying pipeline leakage |
| CN101761780A (en) * | 2010-01-11 | 2010-06-30 | 中国石油大学(华东) | Gas pipeline leakage detecting and positioning device and method thereof |
| CN104750742A (en) * | 2013-12-31 | 2015-07-01 | 南京理工大学常熟研究院有限公司 | Fault diagnosis method and system for heading machine hydraulic system |
| CN105042339A (en) * | 2015-06-03 | 2015-11-11 | 中国石化销售有限公司华东分公司 | Leakage rate estimation system and method for product oil pipelines based on zero dimension |
| US20180259921A1 (en) * | 2015-11-26 | 2018-09-13 | Jiangnan University | Data-Difference-Driven Self-Learning Dynamic Optimization Method For Batch Process |
| CN106015951A (en) * | 2016-07-07 | 2016-10-12 | 北京信息科技大学 | Gas transmission pipeline leakage detection system and method suitable for multiple state changes |
| US20180365555A1 (en) * | 2016-12-22 | 2018-12-20 | Naveed Aslam | Artificial intelligence based algorithm for predicting pipeline leak and corrosion detection |
| US20180347763A1 (en) * | 2017-05-31 | 2018-12-06 | Schneider Electric Software, Llc | Machine learning detection of pipeline ruptures |
| JP2019095292A (en) * | 2017-11-22 | 2019-06-20 | 株式会社日立製作所 | Water leakage detection system and water leakage detection method |
| AU2019383049A1 (en) * | 2018-11-23 | 2021-07-01 | The University Of Adelaide | Method and system to analyse pipeline condition |
| US20200210826A1 (en) * | 2018-12-29 | 2020-07-02 | Northeastern University | Intelligent analysis system using magnetic flux leakage data in pipeline inner inspection |
| CN110185939A (en) * | 2019-05-16 | 2019-08-30 | 西北工业大学 | Gas pipe leakage recognition methods based on convolutional neural networks |
| CN110296327A (en) * | 2019-06-19 | 2019-10-01 | 常州大学 | A kind of pipeline leakage detection method based on transient flow frequency response analysis |
| US20210116076A1 (en) * | 2019-10-22 | 2021-04-22 | Eog Resources, Inc. | Anomaly detection in pipelines and flowlines |
| CN111125867A (en) * | 2019-11-04 | 2020-05-08 | 浙江工业大学 | Establishment and calculation method of chemical production pipeline real-time transient model based on chaotic particle swarm |
| US20230028276A1 (en) * | 2019-11-26 | 2023-01-26 | Technische Universität Berlin | Forecasting industrial aging processes with machine learning methods |
| US20210178697A1 (en) * | 2019-12-17 | 2021-06-17 | Northrop Grumman Systems Corporation | Machine-learning-based additive manufacturing using manufacturing data |
| CN111059477A (en) * | 2019-12-23 | 2020-04-24 | 浙江工业大学 | Double-layer framework based reverse source-seeking chemical pipeline leakage detection and positioning method |
| CN111915579A (en) * | 2020-07-24 | 2020-11-10 | 天津大学 | Liquid leakage detection method based on Gaussian mixture model and convolutional neural network |
| CN112036440A (en) * | 2020-07-31 | 2020-12-04 | 山东科技大学 | Satellite attitude control system fault diagnosis and early warning method based on random forest |
| CN112610903A (en) * | 2020-12-10 | 2021-04-06 | 合肥学院 | Water supply pipe network leakage positioning method based on deep neural network model |
| CN113077091A (en) * | 2021-04-09 | 2021-07-06 | 中国石油大学(北京) | Machine learning-based leakage rate probability distribution prediction method and system |
Non-Patent Citations (5)
| Title |
|---|
| 李平;张园园;: "管道泄漏检测技术研究进展", 管道技术与设备, no. 02 * |
| 李慧: "《基于混合模型的管道泄漏检测方法研究》", 《中国优秀硕士学位论文全文数据库工程科技I辑》 * |
| 李慧: "《基于混合模型的管道泄漏检测方法研究》", 《中国优秀硕士学位论文全文数据库工程科技I辑》, no. 3, 15 March 2016 (2016-03-15), pages 51 - 67 * |
| 王军茹;刘丽华;王巧玲;: "基于模型的输气管道泄漏检测与定位系统", 科技视界, no. 27 * |
| 王占山,张化光,冯健,伦淑娴: "长距离流体输送管道泄漏检测与定位技术的现状与展望", 化工自动化及仪表, no. 05 * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114636113A (en) * | 2022-02-22 | 2022-06-17 | 中国特种设备检测研究院 | Oil depot pipeline leakage detection method |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113803647B (en) | 2023-07-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111425768B (en) | Oil leakage point and oil leakage rate detection method for oil pipeline | |
| CN110245411B (en) | Urban central heating system and leakage point detection method | |
| CN104063588B (en) | Multi-source data fusion-based method for predicting pipeline corrosion defect size | |
| CN105042339A (en) | Leakage rate estimation system and method for product oil pipelines based on zero dimension | |
| CN104180166A (en) | Pipeline leakage detection method based on pipeline pressure data | |
| CN112628613A (en) | Method and system for monitoring pipeline leakage, positioning leakage and calculating leakage amount | |
| CN103491651A (en) | Abnormal/fault positioning detection method based on binomial tree model | |
| CN108506740B (en) | Method and system for determining leakage area of liquid pipeline based on flow velocity meter | |
| CN104930355A (en) | Online nondestructive detection method and device applicable to urban water supply pipeline system | |
| CN108019622A (en) | A kind of computational methods of the pipeline leakage positioning based on pressure differential | |
| CN111102476B (en) | Pipe burst searching method based on three-dimensional underground pipeline | |
| US11982410B2 (en) | Methods and smart gas internet of things (IoT) systems for remote control of ultrasonic metering devices | |
| CN113551744A (en) | Ultrasonic flowmeter performance online monitoring method and system | |
| CN105388884A (en) | Alarm system for detecting leakage fault of heat supply network based on identification algorithm driven by data and method | |
| CN111339701A (en) | Pipeline leakage characteristic Godunov simulation method based on Brunone dynamic friction resistance | |
| CN111664823A (en) | Method for detecting thickness of scale layer of voltage-sharing electrode based on difference of medium heat conduction coefficients | |
| CN113803647B (en) | Pipeline leakage detection method based on fusion of knowledge features and hybrid model | |
| CN116906839B (en) | Safety intelligent monitoring and early warning method for thermodynamic pipeline integrating physical measurement and soft measurement | |
| CN111365623B (en) | Negative pressure wave inflection point identification method based on linear fitting | |
| CN107014451A (en) | The method of ultrasonic flow sensor coefficient is speculated based on generalized regression nerve networks | |
| Ravula et al. | Experimental validation of leak and water-ingression detection in low-pressure gas pipeline using pressure and flow measurements | |
| CN116697276A (en) | Thick oil thermal recovery medium conveying pipeline leakage monitoring method and system based on virtual sensing | |
| JP5727890B2 (en) | Impulse tube clogging diagnosis possibility determination system | |
| CN113063100A (en) | Pipeline leakage detection method and device | |
| CN116398825A (en) | Water supply pipe leakage detection and positioning method, system and device based on probability model |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |