Underground pipeline leakage point real-time monitoring method and system based on deep learning

Abstract

The invention relates to the technical field of pipeline leakage point detection and flow detection, in particular to an underground pipeline leakage point real-time monitoring method and system based on deep learning. All data acquisition modules are arranged along the underground pipeline, and the data acquisition modules send a real-time pipeline internal pressure value acquired by a pressure sensor, a real-time pipeline internal flow value acquired by a flow sensor and real-time vibration data acquired by a vibration sensor to a server; then constructing a pipeline leakage monitoring model based on a neural network, and performing training; and inputting the received real-time pipeline internal pressure value, the real-time pipeline internal flow value and the real-time vibration data at each sampling moment in the past preset duration into the trained pipeline leakage monitoring model to obtain an output result which is whether the pipeline to be detected leaks or not. According to the scheme, leakage monitoring can be automatically carried out on the to-be-monitored pipeline in real time, manual intervention is not needed, efficiency is higher, and the monitoring effect is more accurate.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of pipeline leak detection flow detection, in particular to a real-time underground pipeline leak detection method and system based on deep learning. BACKGROUND

[0002] With the acceleration of modernization process, the number of underground pipe networks (such as underground water pipes) also increases day by day; a large number of complex underground water pipes need to be positioned in time when they leak to avoid water loss; the environment of underground pipelines is complex and changeable, and is affected by many factors such as soil conditions, surrounding buildings and traffic loads, so the traditional detection method is easily disturbed; therefore, there is an urgent need for an underground pipeline leak detection technical solution with more accurate detection results. SUMMARY

[0003] The main purpose of the present application is to provide a real-time underground pipeline leak detection method and system based on deep learning, which aims to solve the problem of the urgent need for an underground pipeline leak detection technical solution with more accurate detection results.

[0004] The technical scheme provided by the present application is as follows: A real-time underground pipeline leak detection method based on deep learning is applied to a real-time underground pipeline leak detection system based on deep learning; the system comprises a server and a data acquisition module in communication with the server; the number of data acquisition modules is multiple, and each data acquisition module is arranged along the underground pipeline; the data acquisition module comprises a pressure sensor, a flow sensor and a vibration sensor; the method comprises: The data acquisition module sends the real-time pipeline internal pressure value collected by the pressure sensor, the real-time pipeline internal flow value collected by the flow sensor, and the real-time vibration data collected by the vibration sensor to the server, wherein the pressure sensor is arranged inside the pipeline to be detected, the pressure sensor is arranged inside the pipeline to be detected, and the vibration sensor is arranged on the outer wall of the pipeline to be detected; The server acquires historical pipeline data and constructs a pipeline leak detection model based on a neural network, wherein the historical pipeline data comprises historical pipeline internal pressure values, historical pipeline internal flow values, historical vibration data of each sampling time of each underground pipeline in the past preset time length, and whether each underground pipeline leaks; The server trains the pipeline leak detection model based on the historical pipeline data; The server inputs the received real-time pipeline internal pressure value, real-time pipeline internal flow value and real-time vibration data of each sampling time in the past preset time length into the trained pipeline leak detection model to obtain an output result, wherein the output result is that there is a leak or no leak.

[0005] Preferably, the server trains the pipeline leakage monitoring model based on historical pipeline data, comprising: The server takes the historical pipeline internal pressure value, the historical pipeline internal flow value, and the historical vibration data of each underground pipeline at each sampling time within a preset time period in the past as input parameters of the pipeline leakage monitoring model, and trains the pipeline leakage monitoring model by taking whether each underground pipeline has leakage as an output parameter.

[0006] Preferably, the server inputs the received real-time pipeline internal pressure value, real-time pipeline internal flow value, and real-time vibration data at each sampling time within a preset time period in the past into the trained pipeline leakage monitoring model to obtain an output result, and further comprising: When the output result is leakage, the server determines the leakage position point of the pipeline under test based on the real-time pipeline internal pressure value, the real-time pipeline internal flow value, and the real-time vibration data within a preset time period in the past, wherein the sampling periods of the real-time pipeline internal pressure value, the real-time pipeline internal flow value, and the real-time vibration data are consistent.

[0007] Preferably, when the output result is leakage, the server determines the leakage position point of the pipeline under test based on the real-time pipeline internal pressure value, the real-time pipeline internal flow value, and the real-time vibration data within a preset time period in the past, comprising: The server marks the setting position point of the pressure sensor on the pipeline under test as a pressure detection point; The server sorts the pressure detection points according to the water flow direction and determines the unique serial number of each pressure detection point, wherein the serial number of the most upstream pressure detection point is 1, the serial number of the most downstream pressure detection point is N, N is the total number of pressure detection points set on the pipeline under test, and the serial number of the pressure detection point located relatively upstream is smaller than the serial number of the pressure detection point located relatively downstream in two adjacent pressure detection points according to the serial number; The server obtains the real-time pipeline internal pressure change value of each pressure detection point within a preset time period in the past in order of the serial number from small to large: , In the formula, is the jth real-time pipeline internal pressure change value of the pressure detection point with serial number i within a preset time period in the past; is the real-time pipeline internal pressure value collected at the j+1 sampling time within a preset time period in the past of the pressure detection point with serial number i; a real-time pipeline internal pressure value collected at a jth sampling moment in a preset time length in the past for a pressure detection point with a serial number i; i is a positive integer and satisfies: 1≤i≤N; j is a positive integer and satisfies: 1≤j≤M, M is a total number of sampling moments in the preset time length; when the real-time pipeline internal pressure change value is greater than the first preset value, the server marks the pressure detection point with the real-time pipeline internal pressure change value greater than the first preset value as a first target detection point; the server determines a second target detection point, wherein a serial number of the second target detection point is the first target detection point + 1; the server determines whether a real-time pipeline internal pressure change value corresponding to the first target moment in a preset time length in the past of the second target detection point is greater than the first preset value: if yes, the server determines a leakage point of the pipeline under test to be between the first target detection point and the second target detection point.

[0008] Preferably, the server determines a second target detection point, wherein a serial number of the second target detection point is the first target detection point + 1, and further comprises: the server determines a third target detection point, wherein a serial number of the third target detection point is the first target detection point - 1; the server determines whether a real-time pipeline internal pressure change value corresponding to the first target moment in a preset time length in the past of the third target detection point is greater than the first preset value: if yes, the server determines a leakage point of the pipeline under test to be between the first target detection point and the third target detection point.

[0009] Preferably, the real-time vibration data comprises a real-time vibration frequency value and a real-time vibration amplitude value; when the output result is that leakage occurs, the server determines a leakage position point of the pipeline under test based on the real-time pipeline internal pressure value, the real-time pipeline internal flow value and the real-time vibration data in a preset time length in the past, and further comprises: the server marks a setting position point of the vibration sensor on the pipeline under test as a vibration detection point; the server sorts the vibration detection points according to a water flow direction and determines a unique serial number corresponding to each vibration detection point, wherein a serial number of an uppermost upstream vibration detection point is 1, a serial number of a vibration detection point closest to the vibration detection point with the serial number 1 is 2, and a serial number of a lowermost downstream vibration detection point is N, N is a total number of the vibration detection points of the pipeline under test; the server obtains each real-time vibration amplitude change value of each vibration detection point in a preset time length in the past in order of serial numbers from small to large: , In the formula, is the jth real-time vibration amplitude change value of the vibration detection point with the serial number i in the past preset time length; is the (j+1)th sampling time point real-time vibration amplitude value collected by the vibration detection point with the serial number i in the past preset time length; is the jth sampling time point real-time vibration amplitude value collected by the vibration detection point with the serial number i in the past preset time length; When the real-time vibration amplitude change value is greater than the second preset value, the server marks the pressure detection point with the real-time vibration amplitude change value greater than the second preset value as a fourth target detection point, and sets the sampling time point corresponding to the real-time vibration amplitude change value greater than the second preset value as a second target time point; The server determines a fifth target detection point, wherein the serial number of the fifth target detection point is the fourth target detection point+1; The server determines whether the real-time vibration amplitude change value corresponding to the second target time point in the past preset time length of the fifth target detection point is greater than the second preset value: If yes, the server determines the leakage point of the pipeline under test to be between the fourth target detection point and the fifth target detection point.

[0010] Preferably, after the server determines the fifth target detection point, the server further comprises: The server determines a sixth target detection point, wherein the serial number of the sixth target detection point is the fourth target detection point-1; The server determines whether the real-time vibration amplitude change value corresponding to the second target time point in the past preset time length of the sixth target detection point is greater than the second preset value: If yes, the server determines the leakage point of the pipeline under test to be between the fourth target detection point and the sixth target detection point.

[0011] Preferably, after the server determines the fifth target detection point, the server further comprises: The server determines whether the real-time vibration amplitude values corresponding to the second target time point in the past preset time length of the fourth target detection point and all sampling time points after the second target time point are all greater than the second preset value; If yes, the server determines a sixth target detection point, wherein the serial number of the sixth target detection point is the fourth target detection point-1; The server marks the average value of the real-time vibration amplitude values corresponding to the second target time point in the past preset time length of the fifth target detection point and the sampling time points after the second target time point as a first average value; The server marks an average value of the real-time vibration amplitude values corresponding to the second target moment and a sampling moment after the second target moment within a preset time length in the past as a second average value; When the first average value is greater than the second average value, the server determines the leakage point of the pipeline under test to be between the fourth target detection point and the fifth target detection point; When the first average value is less than the second average value, the server determines the leakage point of the pipeline under test to be between the fourth target detection point and the sixth target detection point.

[0012] The application further provides a deep learning-based underground pipeline leakage point real-time monitoring system, which applies the deep learning-based underground pipeline leakage point real-time monitoring method.

[0013] The above technical solution can achieve the following beneficial effects: The deep learning-based underground pipeline leakage point real-time monitoring method can more accurately monitor the underground pipeline leakage point. In specific use, first, arrange the data acquisition modules along the underground pipeline, and send the real-time pipeline internal pressure value collected by the pressure sensor, the real-time pipeline internal flow value collected by the flow sensor, and the real-time vibration data collected by the vibration sensor to the server. The real-time pipeline internal pressure value, the real-time pipeline internal flow value, and the real-time vibration data can reflect the operation of the underground pipeline, thereby reflecting whether the pipeline has a leakage point. Then, construct a pipeline leakage monitoring model based on a neural network, and train the pipeline leakage monitoring model through historical pipeline data. Then, input the received real-time pipeline internal pressure value, real-time pipeline internal flow value, and real-time vibration data of each sampling moment within a preset time length in the past into the trained pipeline leakage monitoring model to obtain an output result, which is whether the pipeline under test has a leakage. The present application can automatically monitor the pipeline under test for leakage in real time without manual intervention, and is more efficient and accurate. BRIEF DESCRIPTION OF DRAWINGS

[0014] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and those skilled in the art can obtain other drawings according to the structures shown in these drawings without creative labor.

[0015] Figure 1A flow step chart of a first embodiment of a deep learning-based underground pipeline leakage point real-time monitoring method according to the present application is provided. DETAILED DESCRIPTION

[0016] It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

[0017] The present application provides a deep learning-based underground pipeline leakage point real-time monitoring method and system.

[0018] As shown in the drawings, Figure 1 In the first embodiment of the deep learning-based underground pipeline leakage point real-time monitoring method according to the present application, the method is applied to a deep learning-based underground pipeline leakage point real-time monitoring system; the system comprises a server and a plurality of data acquisition modules communicatively connected to the server; each of the data acquisition modules is arranged along an underground pipeline; the data acquisition module comprises a pressure sensor, a flow sensor and a vibration sensor; the embodiment comprises the following steps: Step S110: The data acquisition module sends the real-time pipeline internal pressure value collected by the pressure sensor, the real-time pipeline internal flow value collected by the flow sensor, and the real-time vibration data collected by the vibration sensor to the server, wherein the pressure sensor is arranged inside the pipeline to be tested, the pressure sensor is arranged inside the pipeline to be tested, and the vibration sensor is arranged on the outer wall of the pipeline to be tested.

[0019] Specifically, the collection cycle of the pressure sensor collecting the real-time pipeline internal pressure value, the flow sensor collecting the real-time pipeline internal flow value, and the vibration sensor collecting the real-time vibration data is consistent, for example, all 5 seconds.

[0020] Step S120: The server acquires historical pipeline data and constructs a neural network-based pipeline leakage monitoring model, wherein the historical pipeline data comprises historical pipeline internal pressure values, historical pipeline internal flow values, historical vibration data of each underground pipeline at each sampling time in the past predetermined time (for example, 1 hour), and whether each underground pipeline has appeared leakage.

[0021] Specifically, the step constructs a pipeline leakage monitoring model based on a neural network; historical pipeline pressure values, historical pipeline flow values, and historical vibration data can reflect the operation of the underground pipeline, thereby reflecting whether a leakage point of the pipeline occurs, for example: when the pipeline leaks, the pressure sensor adjacent to the leakage point will suddenly reduce the collected pressure in the pipeline, the flow sensor adjacent to the leakage point will suddenly reduce the collected flow in the pipeline, and the vibration sensor adjacent to the leakage point will also change the collected vibration data; the historical pipeline data correspond to a plurality of different underground pipelines, so as to facilitate subsequent model training.

[0022] Step S130: The server trains the pipeline leakage monitoring model based on the historical pipeline data.

[0023] Step S140: The server inputs the received real-time pipeline pressure values, real-time pipeline flow values, and real-time vibration data of each sampling time in the past preset time length into the trained pipeline leakage monitoring model to obtain an output result, wherein the output result is a leakage or no leakage.

[0024] The underground pipeline leakage real-time monitoring method based on deep learning can more accurately monitor the underground pipeline leakage point. Specifically, in use, first, arrange each data acquisition module along the underground pipeline, and send the real-time pipeline pressure values collected by the pressure sensor, the real-time pipeline flow values collected by the flow sensor, and the real-time vibration data collected by the vibration sensor to the server. The real-time pipeline pressure values, real-time pipeline flow values, and real-time vibration data can reflect the operation of the underground pipeline, thereby reflecting whether a leakage point of the pipeline occurs. Then, construct a pipeline leakage monitoring model based on a neural network, and train the pipeline leakage monitoring model through historical pipeline data. Then, input the received real-time pipeline pressure values, real-time pipeline flow values, and real-time vibration data of each sampling time in the past preset time length into the trained pipeline leakage monitoring model to obtain an output result, and the output result is whether the to-be-tested pipeline leaks. The scheme can automatically monitor the to-be-tested pipeline for leakage in real time without manual intervention, is more efficient, and has more accurate monitoring effect.

[0025] In a second embodiment of the underground pipeline leakage real-time monitoring method based on deep learning, based on the first embodiment, step S130 includes the following steps: Step S210: The server trains the pipeline leakage monitoring model by taking the historical pipeline pressure values, historical pipeline flow values, and historical vibration data of each sampling time of each underground pipeline in the past preset time length as input parameters of the pipeline leakage monitoring model, and taking whether each underground pipeline leaks as an output parameter.

[0026] Specifically, this embodiment provides a technical solution for training a pipeline leakage monitoring model based on historical pipeline data.

[0027] In a third embodiment of a method for real-time monitoring of underground pipeline leaks based on deep learning proposed by the present invention, based on the second embodiment, step S140 further includes the following steps: Step S310: When the output result is that a leak occurs, the server determines the leakage location of the pipeline to be tested based on the real-time pipeline pressure value, the real-time pipeline flow value, and the real-time vibration data within a preset time period in the past, wherein the sampling periods of the real-time pipeline pressure value, the real-time pipeline flow value, and the real-time vibration data are consistent.

[0028] Specifically, when the output result of the pipeline leakage monitoring model is leakage, the leakage location of the pipeline to be tested is determined based on the real-time pipeline pressure value, real-time pipeline flow value, and real-time vibration data within the past preset time period.

[0029] In a fourth embodiment of a method for real-time monitoring of underground pipeline leaks based on deep learning proposed by the present invention, based on the third embodiment, step S310 includes the following steps: Step S410: The server marks the location of the pressure sensor on the pipeline to be tested as a pressure detection point.

[0030] Step S420: The server sorts the pressure detection points according to the direction of water flow and determines the unique serial number corresponding to each pressure detection point, wherein the serial number of the upstream pressure detection point is 1, and the serial number of the downstream pressure detection point is N, where N is the total number of pressure detection points set in the pipeline to be tested. Among the two pressure detection points with adjacent serial numbers, the serial number of the relatively upstream pressure detection point is smaller than the serial number of the relatively downstream pressure detection point.

[0031] Specifically, by setting the sequence numbers in this way, the relative upstream and downstream relationships between the pressure detection points can be determined through the sequence numbers of the pressure detection points.

[0032] Step S430: The server obtains the real-time pressure change values ​​of each pressure detection point in the pipeline within the past preset time period in order from the smallest to the largest sequence number: , Where, The jth real-time pipeline pressure change value of the pressure detection point with sequence number i within the past preset time period; The real-time pipeline pressure value collected at the j+1th sampling time within the past preset time period at the pressure detection point with sequence number i; Pi,j is a real-time pipeline internal pressure value collected at a j sampling moment in a preset time length in the past for a pressure detection point with a serial number of i; i is a positive integer and satisfies: 1≤i≤N; j is a positive integer and satisfies: 1≤j≤M, and M is a total number of sampling moments in the preset time length.

[0033] Specifically, the real-time pipeline internal pressure change value can reflect a change condition of the pipeline internal pressure value of each pressure detection point in the past preset time length.

[0034] Step S440: When the real-time pipeline internal pressure change value is greater than a first preset value (0.05 MPa), the server marks the pressure detection point with the real-time pipeline internal pressure change value greater than the first preset value as a first target detection point.

[0035] Specifically, when the real-time pipeline internal pressure change value is greater than the first preset value, it indicates that the pipeline internal pressure value of the pressure detection point has a large numerical mutation (the pipeline internal pressure value decreases), and it can be inferred that a leakage point appears near the pressure detection point, which causes the pipeline internal pressure value of the pressure detection point to suddenly decrease. Therefore, the pressure detection point is marked as the first target detection point, so as to facilitate subsequent determination of the position of the leakage point.

[0036] Step S450: The server sets a sampling moment corresponding to the pipeline internal pressure change value greater than the first preset value as a first target moment, and determines a second target detection point, wherein the serial number of the second target detection point is the first target detection point+1.

[0037] Specifically, the first target moment is a sampling moment corresponding to the large numerical mutation of the pipeline internal pressure value of the first target detection point; and the second target detection point is downstream of the first target detection point and adjacent to the first target detection point.

[0038] Step S460: The server determines whether a real-time pipeline internal pressure change value corresponding to the first target moment in the past preset time length of the second target detection point is greater than the first preset value.

[0039] If yes, step S470 is performed: the server determines that the leakage point of the pipeline under test is between the first target detection point and the second target detection point.

[0040] Specifically, if the real-time pipeline internal pressure change value corresponding to the first target moment of the second target detection point is also greater than the first preset value, it indicates that the pipeline internal pressure value of the second target detection point also has a large mutation, and then it can be determined that the leakage point of the pipeline under test is between the first target detection point and the second target detection point.

[0041] In the fifth embodiment of the underground pipeline leakage point real-time monitoring method based on deep learning, based on the fourth embodiment, the step S450 further comprises the following steps: Step S510: The server determines a third target detection point, wherein the serial number of the third target detection point is first target detection point-1.

[0042] Specifically, the second target detection point is upstream of and adjacent to the first target detection point.

[0043] Step S520: The server determines whether the real-time pipeline internal pressure change value corresponding to the first target time within the past preset time length of the third target detection point is greater than the first preset value.

[0044] If yes, step S530 is performed: the server determines the leakage point of the pipeline under test to be between the first target detection point and the third target detection point.

[0045] Specifically, if the real-time pipeline internal pressure change value corresponding to the first target time of the third target detection point is also greater than the first preset value, it indicates that the pipeline internal pressure value of the third target detection point also has a large mutation, and then it can be determined that the leakage point of the pipeline under test is between the first target detection point and the third target detection point.

[0046] In the sixth embodiment of the underground pipeline leakage point real-time monitoring method based on deep learning, based on the third embodiment, the real-time vibration data comprises real-time vibration frequency value and real-time vibration amplitude value; and step S310 further comprises the following steps: Step S610: The server marks the setting position point of the vibration sensor on the pipeline under test as a vibration detection point.

[0047] Step S620: The server sorts the vibration detection points according to the water flow direction and determines the unique serial number of each vibration detection point, wherein the serial number of the most upstream vibration detection point is 1, the serial number of the vibration detection point closest to the vibration detection point with serial number 1 is 2, and so on, and the serial number of the most downstream vibration detection point is N, N is the total number of vibration detection points of the pipeline under test.

[0048] Specifically, by setting the serial number in this way, the relative upstream and downstream relationship between the vibration detection points can be determined by the serial number of the vibration detection points.

[0049] Step S630: The server obtains each real-time vibration amplitude change value of each vibration detection point within the past preset time length in order of serial number from small to large: , In the formula, a real-time vibration amplitude change value of a vibration detection point with a serial number of i at a jth sampling time in a preset time length in the past; a real-time vibration amplitude value collected at a j+1th sampling time in a preset time length in the past for a vibration detection point with a serial number of i; a real-time vibration amplitude value collected at a jth sampling time in a preset time length in the past for a vibration detection point with a serial number of i.

[0050] Specifically, the real-time vibration amplitude change value can reflect the change of the real-time vibration amplitude value of each vibration detection point in the preset time length in the past. It can be known that if a leakage point appears, the real-time vibration amplitude value collected by the vibration detection point nearby will suddenly increase.

[0051] Step S640: When the real-time vibration amplitude change value is greater than a second preset value (for example, 1 mm), the server marks the pressure detection point with the real-time vibration amplitude change value greater than the second preset value as a fourth target detection point, and sets a sampling time corresponding to the real-time vibration amplitude change value greater than the second preset value as a second target time.

[0052] Specifically, when the real-time vibration amplitude change value is greater than the second preset value, it indicates that the real-time vibration value of the vibration detection point suddenly increases, and it can be inferred that a leakage point appears near the vibration detection point. Therefore, the vibration detection point is marked as the fourth target detection point, so as to facilitate subsequent determination of the position of the leakage point.

[0053] Step S650: The server determines a fifth target detection point, wherein the serial number of the fifth target detection point is the serial number of the fourth target detection point+1.

[0054] Specifically, the second target time is a sampling time corresponding to a sudden increase of a larger value of the real-time vibration amplitude value of the fourth target detection point; the fifth target detection point is downstream of the fourth target detection point and adjacent to the fourth target detection point.

[0055] Step S660: The server determines whether a real-time vibration amplitude change value corresponding to the second target time in a preset time length in the past of the fifth target detection point is greater than a second preset value.

[0056] If yes, step S670 is performed: the server determines that the leakage point of the pipeline under test is between the fourth target detection point and the fifth target detection point.

[0057] Specifically, if the real-time vibration amplitude change value corresponding to the second target time of the fifth target detection point is also greater than the second preset value, it indicates that the real-time vibration amplitude value of the fifth target detection point also suddenly increases by a larger value, and then it can be determined that the leakage point of the pipeline under test is between the fourth target detection point and the fifth target detection point.

[0058] In a seventh embodiment of the underground pipeline leakage point real-time monitoring method based on deep learning, based on the sixth embodiment, after step S650, the following steps are further included: Step S710: The server determines a sixth target detection point, wherein the serial number of the sixth target detection point is fourth target detection point-1.

[0059] Specifically, the sixth target detection point is upstream of and adjacent to the fourth target detection point.

[0060] Step S720: The server determines whether the real-time vibration amplitude change value corresponding to the second target time within the past preset time length of the sixth target detection point is greater than a second preset value.

[0061] If yes, step S730 is performed: the server determines the leakage point of the pipeline under test to be between the fourth target detection point and the sixth target detection point.

[0062] Specifically, if the real-time vibration amplitude change value corresponding to the second target time of the sixth target detection point is also greater than the second preset value, it indicates that the real-time vibration amplitude value of the sixth target detection point also has a large surge, so it can be determined that the leakage point of the pipeline under test is between the fourth target detection point and the sixth target detection point.

[0063] In an eighth embodiment of the underground pipeline leakage point real-time monitoring method based on deep learning, based on the sixth embodiment, after step S650, the following steps are further included: Step S810: The server determines whether the real-time vibration amplitude values corresponding to the second target time within the past preset time length of the fourth target detection point and all sampling times after the second target time are all greater than the second preset value.

[0064] Specifically, when the pipeline under test leaks, the real-time vibration amplitude values collected by the nearby vibration detection points should continue to be at a high level (before the pipeline is repaired), so it can be inferred that the real-time vibration amplitude values corresponding to the second target time within the past preset time length of the fourth target detection point and all sampling times after the second target time should all be greater than the second preset value.

[0065] If yes, step S820 is performed: the server determines a sixth target detection point, wherein the serial number of the sixth target detection point is fourth target detection point-1.

[0066] Specifically, if yes, it is further determined that the pipeline under test has leaked; therefore, the sixth target detection point is further determined in order to subsequently determine the specific position of the leakage point.

[0067] Step S830: The server marks the average value of the real-time vibration amplitude values corresponding to the second target moment within a preset time period in the past and the sampling moment after the second target moment at the fifth target detection point as a first average value.

[0068] Specifically, the first average value here is the average value of the real-time vibration amplitude values collected at the fifth target detection point at the time when the vibration amplitude of the pipeline under test suddenly increases and at all sampling moments after that, which can reflect the vibration duration after the vibration amplitude of the pipeline under test suddenly increases.

[0069] Step S840: The server marks the average value of the real-time vibration amplitude values corresponding to the second target moment within a preset time period in the past and the sampling moment after the second target moment at the sixth target detection point as a second average value.

[0070] Specifically, the first average value here is the average value of the real-time vibration amplitude values collected at the sixth target detection point at the time when the vibration amplitude of the pipeline under test suddenly increases and at all sampling moments after that, which can reflect the vibration duration after the vibration amplitude of the pipeline under test suddenly increases.

[0071] Step S850: When the first average value is greater than the second average value, the server determines the leakage point of the pipeline under test to be between the fourth target detection point and the fifth target detection point.

[0072] Specifically, when the first average value is greater than the second average value, it indicates that the vibration amplitude at the fifth target detection point is greater than that at the sixth target detection point after the leakage occurs, and therefore the leakage point of the pipeline under test is determined to be between the fourth target detection point and the fifth target detection point.

[0073] Step S860: When the first average value is less than the second average value, the server determines the leakage point of the pipeline under test to be between the fourth target detection point and the sixth target detection point.

[0074] Specifically, when the first average value is less than the second average value, it indicates that the vibration amplitude at the fifth target detection point is less than that at the sixth target detection point after the leakage occurs, and therefore the leakage point of the pipeline under test is determined to be between the fourth target detection point and the sixth target detection point.

[0075] The application also provides a deep learning-based real-time underground pipeline leakage point monitoring system, which applies the deep learning-based real-time underground pipeline leakage point monitoring method.

[0076] The above-mentioned serial numbers of the embodiments of the present application are only for description, and do not represent the advantages and disadvantages of the embodiments.

[0077] The embodiments of the present application are described above with reference to the accompanying drawings, but the present application is not limited to the above-mentioned specific embodiments, and the above-mentioned specific embodiments are only illustrative, but not restrictive. Those skilled in the art can make many forms under the inspiration of the present application without departing from the purpose of the present application and the scope protected by the claims, and these all belong to the protection of the present application.

Claims

1. A real-time monitoring method for underground pipeline leakage based on deep learning, characterized in that: A deep learning-based real-time monitoring system for underground pipeline leaks is provided; the system comprises a server and a data acquisition module communicatively connected to the server; the data acquisition modules are multiple and each of the data acquisition modules is arranged along the underground pipeline; The data acquisition module includes a pressure sensor, a flow sensor and a vibration sensor; the method includes: The data acquisition module sends the real-time pipeline pressure value acquired by the pressure sensor, the real-time pipeline flow value acquired by the flow sensor, and the real-time vibration data acquired by the vibration sensor to the server, wherein the pressure sensor is arranged inside the pipeline to be measured, the pressure sensor is arranged inside the pipeline to be measured, and the vibration sensor is arranged on the outer wall of the pipeline to be measured; The server obtains historical pipeline data and constructs a pipeline leakage monitoring model based on a neural network, wherein the historical pipeline data includes historical pipeline pressure values, historical pipeline flow values, historical vibration data, and whether each underground pipeline has a leak at each sampling time within a preset period of time in the past; The server trains a pipeline leakage monitoring model based on historical pipeline data; The server inputs the received real-time pipeline pressure value, real-time pipeline flow value, and real-time vibration data at each sampling moment within a preset time period into the trained pipeline leakage monitoring model to obtain an output result, wherein the output result is whether a leakage occurs or not.

2. The method for real-time monitoring of underground pipeline leakage based on deep learning according to claim 1 is characterized in that: The server trains a pipeline leakage monitoring model based on historical pipeline data, including: The server uses the historical pipeline pressure values, historical pipeline flow values, and historical vibration data of each underground pipeline at each sampling time within a preset period of time in the past as input parameters of the pipeline leakage monitoring model, and uses whether each underground pipeline leaks as an output parameter to train the pipeline leakage monitoring model.

3. The method for real-time monitoring of underground pipeline leakage based on deep learning according to claim 2 is characterized in that: The server inputs the received real-time pipeline pressure value, real-time pipeline flow value, and real-time vibration data at each sampling time within a preset time period into the trained pipeline leakage monitoring model to obtain an output result, and then further includes: When the output result is that a leak occurs, the server determines the leakage location of the pipeline to be tested based on the real-time pipeline pressure value, the real-time pipeline flow value, and the real-time vibration data within a preset time period in the past, wherein the sampling periods of the real-time pipeline pressure value, the real-time pipeline flow value, and the real-time vibration data are consistent.

4. The method for real-time monitoring of underground pipeline leakage based on deep learning according to claim 3 is characterized in that: When the output result indicates leakage, the server determines the leakage location of the pipeline to be tested based on the real-time pipeline pressure value, real-time pipeline flow value, and real-time vibration data within a preset time period, including: The server marks the location of the pressure sensor on the pipeline to be tested as a pressure detection point; The server sorts the pressure detection points according to the direction of water flow and determines a unique serial number corresponding to each pressure detection point, wherein the serial number of the most upstream pressure detection point is 1, and the serial number of the most downstream pressure detection point is N, where N is the total number of pressure detection points set in the pipeline to be tested. Of two pressure detection points with adjacent serial numbers, the serial number of the relatively upstream pressure detection point is smaller than the serial number of the relatively downstream pressure detection point; The server obtains the real-time pressure change values ​​of each pipeline at each pressure detection point within the past preset time period in order of the serial number from small to large: ， Where, The jth real-time pipeline pressure change value of the pressure detection point with sequence number i within the past preset time period; The real-time pipeline pressure value collected at the j+1th sampling time within the past preset time period at the pressure detection point with sequence number i; The real-time pipeline pressure value collected at the j-th sampling time within the past preset time period at the pressure detection point with sequence number i; i is a positive integer and satisfies: 1≤i≤N; j is a positive integer and satisfies: 1≤j≤M, where M is the total number of sampling times within the preset time period; When the real-time pressure change value in the pipeline is greater than a first preset value, the server marks the pressure detection point where the real-time pressure change value in the pipeline is greater than the first preset value as a first target detection point; The server uses the sampling time corresponding to the pressure change value in the pipeline greater than the first preset value as the first target time, and determines the second target detection point, wherein the sequence number of the second target detection point is the first target detection point + 1; The server determines whether a real-time pipeline pressure change value corresponding to a first target moment within a preset time period in the past at a second target detection point is greater than a first preset value: If so, the server determines the leakage point of the pipeline to be tested as between the first target detection point and the second target detection point.

5. The method for real-time monitoring of underground pipeline leakage points based on deep learning according to claim 4 is characterized in that: The server uses the sampling time corresponding to the pressure change value in the pipeline greater than the first preset value as the first target time, and determines the second target detection point, and then further includes: The server determines a third target detection point, wherein the sequence number of the third target detection point is the first target detection point minus 1; The server determines whether a real-time pipeline pressure change value corresponding to a first target moment within a preset time period in the past at a third target detection point is greater than a first preset value: If yes, the server determines the leakage point of the pipeline to be tested as between the first target detection point and the third target detection point.

6. The method for real-time monitoring of underground pipeline leakage based on deep learning according to claim 3 is characterized in that: The real-time vibration data includes a real-time vibration frequency value and a real-time vibration amplitude value; when the output result is leakage, the server determines the leakage location of the pipeline to be tested based on the real-time pipeline pressure value, the real-time pipeline flow value, and the real-time vibration data within a preset time period, and further includes: The server marks the location of the vibration sensor on the pipeline to be tested as a vibration detection point; The server sorts the vibration detection points according to the direction of water flow and determines a unique serial number corresponding to each vibration detection point, wherein the serial number of the most upstream vibration detection point is 1, the serial number of the vibration detection point closest to the vibration detection point with serial number 1 is 2, and so on. The serial number of the most downstream vibration detection point is N, where N is the total number of vibration detection points in the pipeline to be tested; The server obtains the real-time vibration amplitude change values ​​of each vibration detection point in the past preset time period in order from small to large sequence numbers: ， Where, The jth real-time vibration amplitude change value of the vibration detection point with sequence number i within the past preset time length; The real-time vibration amplitude value collected by the vibration detection point with sequence number i at the j+1th sampling time within the past preset time length; The real-time vibration amplitude value collected by the vibration detection point with sequence number i at the j-th sampling moment within the past preset time length; When the real-time vibration amplitude change value is greater than a second preset value, the server marks the pressure detection point where the real-time vibration amplitude change value is greater than the second preset value as a fourth target detection point, and sets the sampling moment corresponding to the real-time vibration amplitude change value greater than the second preset value as the second target moment; The server determines a fifth target detection point, wherein the sequence number of the fifth target detection point is the fourth target detection point+1; The server determines whether the real-time vibration amplitude change value corresponding to the second target moment within the past preset time period of the fifth target detection point is greater than a second preset value: If so, the server determines the leakage point of the pipeline to be tested as between the fourth target detection point and the fifth target detection point.

7. The method for real-time monitoring of underground pipeline leakage based on deep learning according to claim 6 is characterized in that: The server determines a fifth target detection point, and then further includes: The server determines a sixth target detection point, wherein the sequence number of the sixth target detection point is the fourth target detection point - 1; The server determines whether the real-time vibration amplitude change value corresponding to the second target moment within the past preset time period of the sixth target detection point is greater than a second preset value: If so, the server determines the leakage point of the pipeline to be tested as between the fourth target detection point and the sixth target detection point.

8. The method for real-time monitoring of underground pipeline leakage based on deep learning according to claim 6 is characterized in that: The server determines a fifth target detection point, and then further includes: The server determines whether the real-time vibration amplitude values ​​corresponding to the second target moment and all sampling moments after the second target moment of the fourth target detection point within the past preset time period are greater than the second preset value; If yes, the server determines a sixth target detection point, wherein the sequence number of the sixth target detection point is the fourth target detection point - 1; The server marks the average value of the real-time vibration amplitude values ​​corresponding to the second target moment and the sampling moments after the second target moment of the fifth target detection point within the past preset time period as the first average value; The server marks the average value of the real-time vibration amplitude values ​​corresponding to the second target moment and the sampling moments after the second target moment of the sixth target detection point within the past preset time period as a second average value; When the first average value is greater than the second average value, the server determines the leakage point of the pipeline to be tested as between the fourth target detection point and the fifth target detection point; When the first average value is smaller than the second average value, the server determines the leakage point of the pipeline to be tested as between the fourth target detection point and the sixth target detection point.

9. A real-time monitoring system for underground pipeline leakage based on deep learning, characterized in that: A method for real-time monitoring of underground pipeline leaks based on deep learning as described in any one of claims 1 to 8 is applied; the system includes a server and a data acquisition module communicatively connected to the server; there are multiple data acquisition modules, and each data acquisition module is arranged along the underground pipeline; the data acquisition module includes a pressure sensor, a flow sensor and a vibration sensor.

Images