An intelligent pipe network leakage detection method based on edge computing

Abstract

The application discloses an intelligent pipe network leakage detection method based on edge computing and relates to the technical field of pipe network monitoring and leakage detection.The method comprises the following steps: an edge analysis terminal extracts a pressure stability coefficient, a sound wave correlation coefficient and a flow mutation coefficient from a pretreatment data set, and inputs the pressure stability coefficient, the sound wave correlation coefficient and the flow mutation coefficient into a leakage probability evaluation model to output a leakage risk index; and an edge decision terminal generates a leakage level signal according to the leakage risk index and triggers a positioning instruction.The application extracts the pressure stability coefficient, the sound wave correlation coefficient and the flow mutation coefficient through the edge analysis terminal, inputs the pressure stability coefficient, the sound wave correlation coefficient and the flow mutation coefficient into the leakage probability evaluation model to output the leakage risk index, realizes accurate quantitative evaluation of the leakage risk, and enables the model to be self-adaptive to different working conditions by introducing dynamic weights, thereby improving the adaptability and accuracy of the evaluation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of pipe network monitoring and leakage detection, and particularly relates to an intelligent pipe network leakage detection method based on edge computing. BACKGROUND

[0002] With the rapid advancement of urbanization and the continuous deepening of infrastructure construction, various pipe networks as the core carriers for the transportation of urban energy and resources have become the key link to guarantee the quality of life of residents, the efficiency of industrial production and public safety. Pipe network leakage not only causes a large amount of loss of water resources, energy and other resources, but also may cause ground subsidence, environmental pollution and even explosion and other serious safety accidents. Therefore, real-time, accurate and efficient detection and positioning of pipe network leakage have become an important research direction in the field of intelligent management of urban infrastructure.

[0003] At present, pipe network leakage detection technology has gradually evolved from traditional manual inspection to intelligent monitoring. Early traditional technologies mostly rely on manual listening for leaks, pressure point testing and other methods, which are limited by high labor costs, low detection efficiency and are easily affected by subjective experience. With the development of sensing technology, single sensors such as pressure sensors and sound wave sensors have been applied to pipe network monitoring, and preliminary leakage judgment is achieved by collecting single physical quantities such as pressure fluctuations and sound wave signals. In recent years, some technical solutions have introduced multi-sensor cooperative collection mode and combined cloud computing for centralized data processing, which has improved the automation degree of detection to a certain extent, but the overall still remains in the stage of transition to deep intelligence.

[0004] However, the existing technology still faces many bottlenecks in actual application: first, the pressure, sound wave, flow and other data collected by multi-source sensors often have problems such as time asynchronization and significant noise interference, which leads to insufficient accuracy of key feature extraction and directly affects the reliability of leakage risk assessment; second, the centralized data processing mode relying on cloud computing has defects such as large data transmission delay and weak real-time response capability, which is difficult to meet the needs of early warning and rapid disposal of pipe network leakage; third, the feature weights in the detection model are mostly fixed values, which cannot dynamically adapt to the changes of different pipe network working conditions, resulting in low leakage positioning accuracy, especially insufficient recognition sensitivity to small leaks, and frequent false positives and false negatives.

[0005] Therefore, it is necessary to invent an intelligent pipe network leakage detection method based on edge computing to solve the above problems. SUMMARY

[0006] The purpose of the present application is to provide an intelligent pipe network leakage detection method based on edge computing to solve the problems raised in the background.

[0007] To achieve the above object, the present application provides the following technical scheme: an intelligent pipe network leakage detection method based on edge computing, comprising an edge perception terminal, an edge preprocessing terminal, an edge analysis terminal, an edge decision terminal, an edge positioning terminal and an edge control terminal, specifically comprising the following steps:

[0008] S1, the edge perception terminal collects multi-source physical quantity data in real time at the pipe network node through multiple sensors to form a pipe network operation state data set; the pipe network operation state data set comprises pressure fluctuation data, sound wave feature data and flow anomaly data;

[0009] S2, the edge preprocessing terminal performs time synchronization and adaptive noise filtering processing on the pipe network operation state data set to generate a standardized preprocessing data set;

[0010] S3, the edge analysis terminal extracts features from the standardized preprocessing data set to obtain a pressure stability coefficient, a sound wave correlation coefficient and a flow mutation coefficient, and inputs the pressure stability coefficient, the sound wave correlation coefficient and the flow mutation coefficient into a leakage probability evaluation model to output a leakage risk index;

[0011] The pressure stability coefficient is specifically:

[0012] ,

[0013] Wherein, σ p is the pressure variance, P max is the rated pressure, Δp is the pressure difference between adjacent time points, Δt p is the time interval of pressure sampling, V max is the maximum allowable pressure change rate, e is the natural constant, k1 and k2 are normalization weight factors, k1+k2=1 and k1 and k2∈[0,1];

[0014] The sound wave correlation coefficient is specifically:

[0015] ,

[0016] Wherein, λ is the scale factor, R xy is the sound wave cross-correlation value of adjacent nodes, R0 is the reference correlation threshold, f m is the sound wave main frequency, f max is the maximum measurable frequency, and e is the natural constant;

[0017] The flow mutation coefficient is specifically:

[0018] ,

[0019] Wherein, δ qis the instantaneous flow deviation, γ is a nonlinear scaling factor, β1 and β2 are adjustment parameters, tanh is a hyperbolic tangent function, is the instantaneous flow deviation of the i th sampling point, N is the number of sampling points in the time window, and the calculation formula is , wherein T is a detection period, △t F is a time interval of flow data collection;

[0020] The leakage probability evaluation model is specifically:

[0021] ,

[0022] wherein I is a leakage risk index, S is a pressure stability coefficient, C is a sound wave correlation coefficient, F is a flow mutation coefficient, ω1, ω2 and ω3 are dynamic weights, ω1+ω2+ω3=1 and ω1, ω2 and ω3∈[0, 1], and ε is a smoothing factor;

[0023] The dynamic weights ω1, ω2 and ω3 are dynamically adjusted based on historical data, and specifically:

[0024] ,

[0025] wherein t is a current time period, b is a historical weight attenuation factor, K is a number of similar historical events; φ h，m is a contribution weight of the feature h in the m th historical event, wherein h=1, 2, 3;

[0026] S4, the edge decision terminal generates a leakage level signal according to the leakage risk index, and triggers a positioning instruction;

[0027] S5, the edge positioning terminal receives the positioning instruction, and calculates the three-dimensional spatial coordinates of the leakage point through time delay difference analysis and pressure gradient distribution;

[0028] Preferably, the calculation method of the three-dimensional spatial coordinates of the leakage point comprises the following steps:

[0029] S51, based on the pipeline geometric parameters, the fluid dynamics parameters and the preset flow Q, the theoretical pressure gradient without leakage is calculated through the fluid mechanics equation ;

[0030] S52, the spatial interpolation is performed on the pressure sensor array data, and the measured pressure gradient ∇p at the leakage point X is calculated;

[0031] S53, a target function containing the time delay difference feature and the pressure gradient feature is established:

[0032] ,

[0033] wherein X is the three-dimensional coordinates of the leakage point, Xj Let v be the three-dimensional coordinates of the j-th sensor. s Let M be the velocity of sound in the pipe medium, M be the total number of sensors involved in leak location, and Δt be the velocity of sound in the pipe medium. j Let η be the time delay for the sound wave to reach the j-th sensor, and η be the regularization factor. The theoretical pressure gradient at coordinate X;

[0034] S54. The minimum value of the objective function J(X) is solved iteratively using a nonlinear least squares optimization algorithm. The initial value is set as the coordinates of the center of the pipeline node, and the convergence condition is:

[0035] ,

[0036] Among them, X current X represents the estimated coordinates of the leak point in the current iteration step. previous The coordinates of the leak point are estimated in the previous iteration step, and △X is the preset convergence threshold.

[0037] S55. If the variance of the leak point coordinates calculated by multiple nearby sensors exceeds the threshold σ threshold If the result is positive, the recalibration process is triggered; otherwise, the final positioning result X is output. opt ;

[0038] S6. The edge control terminal closes the corresponding pipe section valve and starts the emergency drainage device according to the three-dimensional spatial coordinates.

[0039] Preferably, the pressure fluctuation data includes the pipe segment pressure variance, the pressure difference between adjacent time points, and the time interval of pressure sampling; the acoustic characteristic data includes the acoustic dominant frequency and the cross-correlation value of acoustic waves at adjacent nodes; and the flow anomaly data includes instantaneous flow deviation.

[0040] Preferably, the leakage level signal is specifically:

[0041] When the leakage risk index I is greater than the threshold θ1, the leakage level signal is a level one leakage.

[0042] When the leakage risk index I is greater than or equal to the threshold θ2 and less than the threshold θ1, the leakage level signal is a level two leakage.

[0043] When the leakage risk index I is less than the threshold θ2, the leakage level signal is a level three leakage.

[0044] The technical effects and advantages of this invention are as follows:

[0045] 1、The present application extracts the pressure stability coefficient, the sound wave correlation coefficient and the flow mutation coefficient through the edge analysis terminal, inputs them into the leakage probability evaluation model, and outputs the leakage risk index, realizes the accurate quantitative evaluation of the leakage risk, and the introduction of the dynamic weight makes the model self-adapt to different working conditions, improves the adaptability and accuracy of the evaluation;

[0046] 2、The present application divides the leakage level into first, second and third levels according to the leakage risk index through the edge decision terminal, and corresponds to different response measures, realizes the hierarchical disposal of the leakage, and improves the pertinence and efficiency of the pipe network management;

[0047] 3、The present application calculates the three-dimensional space coordinates of the leakage point by using the method of time delay difference analysis combined with pressure gradient distribution through the edge positioning terminal, realizes the accurate positioning of the leakage point, and improves the positioning accuracy;

[0048] 4、The present application avoids the transmission delay problem of the centralized processing mode depending on cloud computing by constructing the complete edge computing architecture of edge perception, preprocessing, analysis, decision, positioning and control, realizes the real-time response from data acquisition to control action, can quickly close the corresponding pipe section valve and start the emergency drainage device, and effectively reduces the resource loss and safety risk caused by leakage. BRIEF DESCRIPTION OF DRAWINGS

[0049] Figure 1 It is the device connection schematic diagram of the present application.

[0050] Figure 2 It is the method step flow schematic diagram of the present application. DETAILED DESCRIPTION

[0051] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative labor are within the scope of protection of the present application.

[0052] The present application provides a device connection schematic diagram as shown in Figure 1 , which comprises an edge perception terminal, an edge preprocessing terminal, an edge analysis terminal, an edge decision terminal, an edge positioning terminal and an edge control terminal,

[0053] The present application provides an intelligent pipe network leakage detection method based on edge computing as shown in Figure 2 , which specifically comprises the following steps:

[0054] S1, the edge-aware terminal collects multi-source physical quantity data in real time at the pipe network node through multiple sensors to form a pipe network operation state data set; the pipe network operation state data set includes pressure fluctuation data, sound wave feature data and flow anomaly data;

[0055] Further, in the above technical scheme, the pressure fluctuation data includes pipe segment pressure variance, pressure difference value between adjacent time points and pressure sampling time interval; the sound wave feature data includes sound wave main frequency and adjacent node sound wave cross-correlation value; and the flow anomaly data includes instantaneous flow deviation.

[0056] The pipe segment pressure variance is collected in real time by the pressure sensor from the pressure time sequence data of the pipe network node, such as every Δt p sampling time, and the continuous pressure values are statistically analyzed to calculate the sample variance;

[0057] The pressure difference value between adjacent time points is directly obtained from the pressure sensor from the continuous two sampling values, and the absolute value of the difference between the two is the pressure difference value between adjacent time points;

[0058] The pressure sampling time interval is determined by the hardware sampling frequency of the pressure sensor, that is, the fixed time interval Δt p between two samplings;

[0059] The Δt p is determined by the hardware sampling frequency of the pressure sensor;

[0060] The extraction of the sound wave main frequency is completed by spectrum analysis: the fast Fourier transform is performed on the filtered sound wave signal, and the frequency point corresponding to the maximum amplitude in the energy spectrum is identified as the characteristic main frequency;

[0061] The adjacent node sound wave cross-correlation value is obtained by deploying the acceleration sensor and the hydrophone at the pipe network node, synchronously collecting the vibration acceleration signal and the sound wave pressure signal, and the sampling frequency is not less than 10kHz; based on the IEEE 1588 precision clock protocol, the microsecond level time alignment is realized, and the signal transmission delay error is eliminated; a band-pass filter is used to filter out low-frequency mechanical vibration and environmental noise at 100Hz-1kHz, and the leakage characteristic frequency band is retained; the normalized cross-correlation operation is performed on the adjacent node signals, and the calculation formula is:

[0062] ,

[0063] Where, x(t), y(t) are the synchronized signal sequences, τ is the time delay parameter, and T is the window length; in order to improve real-time performance, a sliding window is used, such as a window length of 5s and a step of 1s to dynamically update R xy , combined with wavelet denoising to suppress transient interference, to ensure the capture sensitivity of early weak signals of leakage;

[0064] The instantaneous flow deviation is obtained by using an electromagnetic flowmeter or an ultrasonic flowmeter at fixed intervals Δt F

[0065] The instantaneous flow value q(t) is collected, and then calculated by the following formula:

[0066] δ q = |q(t) - Q ref |,

[0067] Wherein, Q ref is a preset theoretical flow value, which is dynamically generated by pipeline design flow or real-time flow smoothing value such as moving average;

[0068] S2, the edge preprocessing terminal performs time synchronization and adaptive noise filtering processing on the standardized preprocessing data set to generate a standardized preprocessing data set;

[0069] It should be noted that the time synchronization processing is based on IEEE 1588 precision clock protocol, and the pressure, sound wave and flow data collected by multiple sensors are aligned in microseconds, so that the signal time delay error caused by the difference in deployment position and transmission path of different sensors is eliminated, and the consistency of key parameters such as pressure variance and adjacent node sound wave cross-correlation value in time dimension is ensured, laying a foundation for joint analysis of multi-source data;

[0070] The adaptive noise filtering processing adopts differentiated strategies for different noise characteristics of different data types: for sound wave feature data, a 100Hz-1kHz band-pass filter is used to filter out low-frequency mechanical vibration such as pipeline friction and high-frequency environmental noise such as external sound wave interference, and to retain the sound wave frequency band specific to leakage events; At the same time, wavelet denoising technology is used to suppress transient interference and improve the capture sensitivity of weak sound wave signals in the early stage of leakage; For pressure fluctuation data and flow anomaly data, a sliding window filtering algorithm is used to smooth data fluctuations to eliminate inherent noise of equipment and environmental interference, and to highlight the characteristic changes related to leakage such as pressure mutation and flow anomaly;

[0071] The wavelet denoising technology selects db4 wavelet basis for 3-layer decomposition and uses soft threshold processing;

[0072] The window length in the sliding window filtering algorithm is set to 1 / 5 of the detection period, and the step is 1 / 2 of the window length;

[0073] The standardized preprocessing data set is converted into a unified format such as JSON format through time synchronization and noise filtering, and the data scale is normalized, such as mapping pressure data to the interval [0, 1], normalizing sound wave data according to the maximum measurable frequency, and scaling flow data according to the rated flow;

[0074] S3, the edge analysis terminal extracts features from the preprocessed data set to obtain a pressure stability coefficient, a sound wave correlation coefficient and a flow mutation coefficient, and inputs the pressure stability coefficient, the sound wave correlation coefficient and the flow mutation coefficient into a leakage probability evaluation model to output a leakage risk index;

[0075] Further, in the above technical solution, the pressure stability coefficient is specifically:

[0076] ,

[0077] Where σ p is the pressure variance, P max is the rated pressure, Δp is the pressure difference between adjacent time points, Δt p is the time interval of pressure sampling, V max is the maximum allowable pressure change rate, e is the natural constant, k1 and k2 are normalization weight factors, k1+k2=1 and k1 and k2∈[0,1].

[0078] It should be noted that the rated pressure P max is pre-set by the pipeline design specification, such as the pipeline pressure limit value;

[0079] The maximum allowable pressure change rate V max is a threshold value pre-set based on the fluid dynamics model and the pipeline safety specification, such as the critical change rate to prevent water hammer effect;

[0080] The initial values of the normalization weight factors k1 and k2 are set by expert experience, and can be dynamically adjusted by adaptive algorithms such as gradient descent method during operation;

[0081] Further, in the above technical solution, the sound wave correlation coefficient is specifically:

[0082] ,

[0083] Where λ is the scale factor, R xy is the sound wave cross-correlation value of adjacent nodes, R0 is the reference correlation threshold, f m is the sound wave main frequency, f max is the maximum measurable frequency, and e is the natural constant.

[0084] The setting of the reference correlation threshold R0 is based on historical data calibration, and R xy value is continuously monitored for 30 days under no leakage state, and the 95% quantile is taken as the dynamic threshold reference;

[0085] The maximum measurable frequency f max is directly determined by the sensor hardware specification, such as the upper limit of the frequency response of the hydrophone is 20 kHz;

[0086] The initial value of the scale factor λ follows the physical characteristics of the pipeline: if it is a metal pipeline, the initial value of λ is set to 10; if it is a non-metal pipeline, the initial value of λ is set to 5; and the initial value is optimized online by gradient descent algorithm combined with historical leakage data to adapt to the aging or working condition changes of the pipeline.

[0087] Further, in the above technical solution, the flow mutation coefficient is specifically:

[0088] ,

[0089] Where δ q is the instantaneous flow deviation, γ is a nonlinear scaling factor, β1 and β2 are adjustment parameters, and tanh is a hyperbolic tangent function, is the instantaneous flow deviation of the i-th sampling point, N is the number of sampling points in the time window, and the calculation formula is where T F is the detection period, and △t F is the time interval of flow data acquisition.

[0090] It is known that the initial values of the adjustment parameters β1 and β2 are set to β1=0.6 and β2=0.4, which will be adaptively adjusted with seasonal factors later. The specific logic includes: if it is a summer working condition, increase the β1 weight to improve the sensitivity of the instantaneous flow deviation, such as increasing β1 to 0.7 and decreasing β2 to 0.3; if it is a winter working condition, increase the β2 weight to strengthen the detection ability of the cumulative deviation, such as increasing β2 to 0.5 and decreasing β1 to 0.5.

[0091] The initial value of the nonlinear scaling factor γ is calculated by the formula where α is the hydraulic response coefficient, which is set to α=0.1, and is adaptively adjusted by noise later. Specifically,

[0092] ,

[0093] where σ N is the standard deviation of the flowmeter noise, μ N is the mean of the flowmeter noise, and k N is the sensitivity gain factor, with a default value of 0.3, , increase k N to 0.5; if , reduce k N to 0.1;

[0094] The time interval △t F of flow data acquisition is determined by the sampling frequency of the flowmeter;

[0095] Further, in the above technical solution, the leakage probability evaluation model is specifically:

[0096] ,

[0097] wherein I is a leakage risk index, S is a pressure stability coefficient, C is a sound wave correlation coefficient, F is a flow mutation coefficient, ω1, ω2 and ω3 are dynamic weights, ω1+ω2+ω3=1 and ω1, ω2 and ω3∈[0, 1], and ε is a smoothing factor.

[0098] The value range of the smoothing factor ε is 0.01-0.05, and the default value is set to 0.03;

[0099] Further, in the above technical solution, the dynamic weights ω1, ω2 and ω3 are dynamically adjusted based on historical data, and specifically:

[0100] ,

[0101] wherein t is a current time period, b is a historical weight attenuation factor, K is a number of similar historical events; φ h，m is a contribution weight of feature h in the mth historical event, wherein h=1, 2, 3;

[0102] It should be noted that the default value of the historical weight attenuation factor b is 0.7, which can be adjusted;

[0103] The default values of the dynamic weights ω1, ω2 and ω3 are set to ω1=0.4, ω2=0.4 and ω3=0.3;

[0104] The dynamic adjustment meets the following constraint condition:

[0105] ;

[0106] The selection criteria of the historical events are as follows: if the current leakage level is a first-level leakage, all first-level leakage events in the past 30 days are selected; if the current leakage level is a second-level leakage, all second-level leakage events in the past 60 days are selected; if the current leakage level is a third-level leakage, all third-level leakage events in the past 90 days are selected;

[0107] The quantitative formula of the contribution weight φ h，m of feature h in the mth historical event is:

[0108] ,

[0109] wherein when h=1, Q 1，m is a pressure stability coefficient of the historical event m, when h=2, Q 2，m is a sound wave correlation coefficient of the historical event m, and when h=3, Q3，m is the flow mutation coefficient of the historical event m, μ h and σ h is the mean and standard deviation of the feature k under normal working conditions.

[0110] S4, the edge decision terminal generates a leakage level signal according to the leakage risk index, and triggers a positioning instruction;

[0111] Further, in the above technical scheme, the leakage level signal is specifically:

[0112] When the leakage risk index I is greater than the threshold θ1, the leakage level signal is a first-level leakage;

[0113] When the leakage risk index I is greater than or equal to the threshold θ2 and less than the threshold θ1, the leakage level signal is a second-level leakage;

[0114] When the leakage risk index I is less than the threshold θ2, the leakage level signal is a third-level leakage.

[0115] It should be noted that the threshold θ1 is set in the range of 0.7-0.9, and is set to 0.8 by default, which can be adjusted.

[0116] The threshold θ2 is set in the range of 0.3-0.5, and is set to 0.4 by default, which can be adjusted.

[0117] The corresponding response measure of the first-level leakage is to immediately trigger the positioning instruction and close the valve and start the drainage device;

[0118] The corresponding response measure of the second-level leakage is to trigger the positioning instruction, and after manual confirmation, to delay the closing of the valve and the opening of the drainage device;

[0119] The corresponding response measure of the third-level leakage is only to record the warning, without triggering the positioning instruction and control action.

[0120] S5, the edge positioning terminal receives the positioning instruction, and calculates the three-dimensional spatial coordinates of the leakage point through time delay difference analysis and pressure gradient distribution;

[0121] Further, in the above technical scheme, the calculation method of the three-dimensional spatial coordinates of the leakage point includes the following steps:

[0122] S51, based on the pipeline geometric parameters, fluid dynamics parameters and the preset flow Q, the theoretical pressure gradient without leakage is calculated through the fluid mechanics equation ;

[0123] It should be noted that the pipeline geometric parameters include the pipeline diameter D, the pipeline wall thickness δ and the pipeline length L.

[0124] The fluid dynamics parameters include fluid density p and viscosity m;

[0125] The fluid mechanics equation is Darcy-Weisbach formula;

[0126] The theoretical pressure gradient The calculation formula is:

[0127] ,

[0128] Wherein, v s is the tube sound speed, and g is the acceleration of gravity;

[0129] S52, the spatial interpolation is carried out on the pressure sensor array data, and the measured pressure gradient of the leakage point X is calculated.

[0130] It should be known that the spatial interpolation is inverse distance weighted method or Kriging interpolation;

[0131] The calculation formula of the measured pressure gradient is Wherein, , The partial derivatives of pressure in three-dimensional space are respectively;

[0132] S53, the target function including time delay difference characteristics and pressure gradient characteristics is established:

[0133] ,

[0134] Wherein, X is the three-dimensional coordinates of the leakage point, X j is the three-dimensional coordinates of the jth sensor, v s is the tube sound speed, M is the total number of sensors participating in the leakage positioning, t j is the time delay of the sound wave reaching the jth sensor, and η is the regularization factor, is the theoretical pressure gradient at the coordinates X;

[0135] It should be known that the value of the regularization factor η ranges from 0.2 to 1.0, and the default value is set to 0.5, which can be adjusted. If the sound wave sensor accuracy is high, that is, the error is less than 1ms, η takes a smaller value of 0.2, and the time delay difference positioning is preferred; if the pressure sensor density is high, such as every 10 meters, η takes a larger value of 0.8, and the role of the pressure gradient is strengthened;

[0136] The time delay of the sound wave reaching the jth sensor t j is obtained by calculating the peak position of the cross-correlation function R xy (τ), and the τ value corresponding to the peak is the time difference t j between the sound wave propagating from the leakage point to the sensor j and the reference sensor;

[0137] S54, the minimum value of the objective function J(X) is solved by iteration using a nonlinear least squares optimization algorithm, and the initial value is set as the pipe network node center coordinate, and the convergence condition is:

[0138] ,

[0139] wherein X current is the coordinate estimate value of the leakage point in the current iteration step, X previous is the coordinate estimate value of the leakage point in the last iteration step, and △X is a preset convergence threshold;

[0140] It should be known that the preset convergence threshold △X is specifically △X = η X × pipe network node spacing, wherein η X is a precision coefficient, and the value of η X is set to 0.01.

[0141] S55, if the variance of the leakage point coordinates calculated by multiple adjacent sensors exceeds the threshold σ threshold , a recalibration process is triggered, otherwise the final positioning result X opt is output.

[0142] It should be known that the threshold σ threshold is specifically , wherein η t is a reliability coefficient, and the value of η t is set to 0.1.

[0143] S6, the edge control terminal closes the valve of the corresponding pipe section according to the three-dimensional space coordinates and starts the emergency drainage device.

[0144] Finally, it should be pointed out that: the above only describes the preferred embodiments of the present application and does not limit the present application, although the present application has been described in detail with reference to the foregoing embodiments, for those skilled in the art, the technical solutions recorded in the foregoing embodiments can be modified, or some technical features can be replaced, any modification, equivalent replacement, improvement, etc. within the spirit and principles of the present application should be included in the protection scope of the present application.

Claims

1. An intelligent pipeline leakage detection method based on edge computing, characterized in that, It includes edge sensing terminals, edge preprocessing terminals, edge analysis terminals, edge decision-making terminals, edge positioning terminals, and edge control terminals, and specifically includes the following steps: S1. The edge sensing terminal collects multi-source physical quantity data in real time at pipeline nodes through multiple sensors to form a pipeline operation status dataset; the pipeline operation status dataset includes pressure fluctuation data, acoustic characteristic data and flow anomaly data; S2. The edge preprocessing terminal performs time synchronization and adaptive noise filtering on the pipeline network operation status dataset to generate a standardized preprocessed dataset. S3. The edge analysis terminal extracts features from the standardized preprocessed dataset to obtain the pressure stability coefficient, acoustic correlation coefficient, and flow mutation coefficient. It then inputs the pressure stability coefficient, acoustic correlation coefficient, and flow mutation coefficient into the leakage probability assessment model and outputs the leakage risk index. The pressure stability coefficient is specifically: ， Where, σ p For the pressure variance, P max The pressure is the rated pressure, Δp is the pressure difference between adjacent time points, and Δt is the pressure. p V is the time interval for pressure sampling. max The maximum allowable rate of change of pressure is given by e, which is the natural constant, and k1 and k2 are normalized weighting factors, where k1+k2=1 and k1 and k2∈[0,1]. The acoustic correlation coefficient is specifically as follows: ， Where λ is the scaling factor, R xy The cross-correlation value of acoustic waves between adjacent nodes is given by f, where R0 is the baseline correlation threshold. m f is the dominant frequency of the sound wave. max The maximum measurable frequency is given by e, where e is the natural constant. The flow rate mutation coefficient is specifically: ， Where, δ q γ is the instantaneous flow rate deviation, β1 and β2 are adjustment parameters, and tanh is the hyperbolic tangent function. Let N be the instantaneous flow deviation at the i-th sampling point, and N be the number of sampling points within the time window. The calculation formula is: Where T is the detection period, and Δt F The time interval for collecting traffic data; The leakage probability assessment model is as follows: ， Where I is the leakage risk index, S is the pressure stability coefficient, C is the acoustic correlation coefficient, F is the flow mutation coefficient, ω1, ω2 and ω3 are dynamic weights, ω1+ω2+ω3=1 and ω1, ω2 and ω3∈[0,1], and ε is the smoothing factor; The dynamic weights ω1, ω2, and ω3 are dynamically adjusted based on historical data, specifically as follows: ， Where t is the current time period, b is the historical weight decay factor, and K is the number of similar historical events; φ h，m The contribution weight of feature h in the m-th historical event, where h = 1, 2, 3; S4. The edge decision terminal generates a leakage level signal based on the leakage risk index and triggers a location command. S5. The edge positioning terminal receives the positioning command and calculates the three-dimensional spatial coordinates of the leak point through time delay difference analysis and pressure gradient distribution. The method for calculating the three-dimensional spatial coordinates of the leakage point includes the following steps: S51. Based on pipe geometric parameters, fluid dynamics parameters, and a preset flow rate Q, calculate the theoretical pressure gradient under leak-free conditions using fluid dynamics equations. ; S52. Perform spatial interpolation on the pressure sensor array data to calculate the measured pressure gradient ▽p at the leak point X; S53. Establish an objective function that includes time delay difference characteristics and pressure gradient characteristics: ， Where X represents the three-dimensional coordinates of the leak point, X j Let v be the three-dimensional coordinates of the j-th sensor. s Let M be the velocity of sound in the pipe medium, M be the total number of sensors involved in leak location, and Δt be the velocity of sound in the pipe medium. j Let η be the time delay for the sound wave to reach the j-th sensor, and η be the regularization factor. The theoretical pressure gradient at coordinate X; S54. The minimum value of the objective function J(X) is solved iteratively using a nonlinear least squares optimization algorithm. The initial value is set as the coordinates of the center of the pipeline node, and the convergence condition is: ， Among them, X current X represents the estimated coordinates of the leak point in the current iteration step. previous The coordinates of the leak point are estimated in the previous iteration step, and △X is the preset convergence threshold. S55. If the variance of the leak point coordinates calculated by multiple nearby sensors exceeds the threshold σ threshold If the result is positive, the recalibration process is triggered; otherwise, the final positioning result X is output. opt ; S6. The edge control terminal closes the corresponding pipe section valve and starts the emergency drainage device according to the three-dimensional spatial coordinates.

2. The intelligent pipeline leakage detection method based on edge computing according to claim 1, characterized in that, The pressure fluctuation data includes the pipe section pressure variance, the pressure difference between adjacent time points, and the time interval of pressure sampling; the acoustic characteristic data includes the acoustic dominant frequency and the cross-correlation value of acoustic waves at adjacent nodes; the flow anomaly data includes instantaneous flow deviation.

3. The intelligent pipeline leakage detection method based on edge computing according to claim 1, characterized in that, The leakage level signal is specifically as follows: When the leakage risk index I is greater than the threshold θ1, the leakage level signal is a level one leakage. When the leakage risk index I is greater than or equal to the threshold θ2 and less than the threshold θ1, the leakage level signal is a level two leakage. When the leakage risk index I is less than the threshold θ2, the leakage level signal is a level three leakage.

Images