Pipe network leakage prediction method and system based on distributed high-speed water pressure detection system

Abstract

The application provides a pipe network leakage prediction method and system based on a distributed high-speed water pressure detection system, and belongs to the technical field of big data collection. The bidirectional long short-term memory neural network is used to realize dynamic cumulative calculation of pipe fatigue damage and quantitative evaluation of residual fatigue life. When the fatigue accumulation value exceeds the safety threshold, the correlation between fatigue damage and micro defect morphology is established through nonlinear space geometric topology mapping. Then, the local transient water pressure value of the crack tip is determined by combining the high-speed water flow velocity sequence and the micro crack geometric parameter, and the wedge tearing effect of water flow on the crack is quantified. On this basis, when the local transient water pressure exceeds the critical state threshold, the tearing propagation rate of pipe wall damage is determined, and then the pipe burst leakage risk prediction result corresponding to the target pipe section is determined according to the damage tearing propagation rate, which provides a basis for pipe network managers to take preventive maintenance measures before the deterioration of small damage.

Description

Technical Field

[0001] This application belongs to the field of big data acquisition technology, specifically relating to a method and system for predicting pipeline leakage based on a distributed high-speed water pressure detection system. Background Technology

[0002] Currently, the safe and stable operation of urban water supply networks is crucial for ensuring the smooth functioning of society. Early detection and prediction of the evolution of leaks in these networks are of decisive significance in preventing major pipe bursts. Current detection methods typically locate leaks retrospectively through acoustic or flow anomalies after obvious leakage has occurred, often overlooking the hidden damage to pipe structures caused by the complex hydraulic impacts the network endures over long-term operation. This makes it difficult for managers to take effective preventative measures before minor damage worsens.

[0003] The above information in the background section of the invention is only used to enhance the understanding of the background of this application, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0004] To address the above issues, this application provides a method and system for predicting pipeline leaks based on a distributed high-speed water pressure detection system, which can provide early warnings before minor damage to the pipeline network worsens.

[0005] To achieve the above objectives, the technical solution adopted in this application is as follows: In a first aspect, embodiments of this application provide a pipeline leakage prediction method based on a distributed high-speed water pressure detection system. The method includes: acquiring the time-series characteristic sequences of the opening and closing states of each sampling valve in the water supply pipeline network; based on the time-series characteristic sequences of the opening and closing states, using a random forest regression model to determine the transient waveform characteristic sequence of non-stationary water hammer impact, and determining the alternating stress time sequence of the inner wall surface of the target pipe section according to the transient waveform characteristic sequence; based on the alternating stress time sequence, using a bidirectional long short-term memory network to determine the accumulated stress characteristic sequence, and determining the fatigue accumulation value and residual fatigue life value of the pipe material in the target pipe section according to the accumulated stress characteristic sequence; if the fatigue accumulation value exceeds a preset fatigue accumulation threshold, then... Nonlinear spatial geometric topological mapping is performed on the residual fatigue life value to determine the microscopic defect morphology characteristics of the inner wall of the target pipe section. The microscopic defect morphology characteristics include crack length and crack depth. The water flow velocity sequence in the water supply network is obtained, and the local transient water pressure value is determined based on the water flow velocity sequence. If the local transient water pressure value exceeds a preset water pressure threshold, the critical fracture toughness parameter of the target pipe section is obtained, and the dynamic stress field at the crack tip is determined based on the current local transient water pressure value and the critical fracture toughness parameter. The damage tear propagation rate of the target pipe section wall is determined based on the dynamic stress field at the crack tip, and the corresponding pipe burst leakage risk prediction result is determined based on the damage tear propagation rate.

[0006] Secondly, embodiments of this application provide a pipeline leakage prediction system based on a distributed high-speed water pressure detection system. This system includes: a first acquisition module, a first determination module, a second determination module, a third determination module, a second acquisition module, a fourth determination module, and a fifth determination module. The first acquisition module is used to acquire the time-series characteristic sequences of the opening and closing states of each sampling valve in the water supply pipeline network. The first determination module is used to determine the transient waveform characteristic sequence of non-stationary water hammer impact using a random forest regression model based on the time-series characteristic sequences of the opening and closing states, and to determine the alternating stress time sequence of the inner wall surface of the target pipe section based on the transient waveform characteristic sequence. The second determination module is used to determine the accumulated stress characteristic sequence using a bidirectional long short-term memory network based on the alternating stress time sequence, and to determine the fatigue accumulation value and residual fatigue life value of the pipe material in the target pipe section based on the accumulated stress characteristic sequence. The third determination module is used to perform nonlinear spatial geometric topology analysis on the residual fatigue life value if the fatigue accumulation value exceeds a preset fatigue accumulation threshold. The system employs a first module to map the microscopic defects on the inner wall of the target pipe section, including crack length and crack depth; a second module to acquire the water flow velocity sequence within the water supply network and determine the local transient water pressure value based on the water flow velocity sequence; a fourth module to determine the critical fracture toughness parameter of the target pipe section if the local transient water pressure value exceeds a preset water pressure threshold, and to determine the dynamic stress field at the crack tip based on the current local transient water pressure value and the critical fracture toughness parameter; and a fifth module to determine the damage tear propagation rate of the target pipe section wall based on the dynamic stress field at the crack tip, and to determine the predicted risk of pipe burst leakage corresponding to the target pipe section based on the damage tear propagation rate.

[0007] This application provides a method and system for predicting pipeline leakage based on a distributed high-speed water pressure detection system. By utilizing a bidirectional long short-term memory neural network, dynamic accumulation calculation of pipe fatigue damage and quantitative assessment of residual fatigue life are achieved. When the accumulated fatigue value exceeds a safety threshold, nonlinear spatial geometric topology mapping technology is used to transform abstract damage values ​​into geometric data of the length and depth of microcracks, establishing a correlation between fatigue damage and micro-defect morphology, and presenting the evolution process of the pipe from fatigue accumulation to microcrack initiation. Subsequently, tensor product calculation is performed by combining the high-speed water flow velocity sequence and microcrack geometric parameters to obtain the local transient water pressure value at the crack tip, quantifying the wedging and tearing effect of the water flow on the crack. Based on this, when the local transient water pressure exceeds the critical state threshold, the tearing propagation rate of the pipe wall damage is determined. Then, based on the tearing propagation rate, the predicted risk of pipe burst leakage corresponding to the target pipe section is determined, providing a basis for pipeline managers to take preventative maintenance measures before minor damage worsens, reducing the probability of large-scale pipe burst accidents, and thus improving the safety and reliability of urban water supply network operation. Attached Figure Description

[0008] To more clearly illustrate the technical solutions in the embodiments of this application or the conventional technology, the drawings used in the description of the embodiments or the conventional technology will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0009] Figure 1 This is a flowchart illustrating a pipeline leakage prediction method based on a distributed high-speed water pressure detection system provided in an exemplary embodiment of this application.

[0010] Figure 2 This is a flowchart illustrating a pipeline leakage prediction method based on a distributed high-speed water pressure detection system, provided in another exemplary embodiment of this application.

[0011] Figure 3 This is a flowchart illustrating a pipeline leakage prediction method based on a distributed high-speed water pressure detection system, provided in another exemplary embodiment of this application.

[0012] Figure 4 This is a flowchart illustrating a pipeline leakage prediction method based on a distributed high-speed water pressure detection system, provided in another exemplary embodiment of this application.

[0013] Figure 5 This is a flowchart illustrating a pipeline leakage prediction method based on a distributed high-speed water pressure detection system, provided in another exemplary embodiment of this application.

[0014] Figure 6 This is a flowchart illustrating a pipeline leakage prediction method based on a distributed high-speed water pressure detection system, provided in another exemplary embodiment of this application.

[0015] Figure 7 This is a flowchart illustrating a pipeline leakage prediction method based on a distributed high-speed water pressure detection system, provided in another exemplary embodiment of this application.

[0016] Figure 8 This is a flowchart illustrating a pipeline leakage prediction method based on a distributed high-speed water pressure detection system, provided in another exemplary embodiment of this application.

[0017] Figure 9 This is a flowchart illustrating a pipeline leakage prediction method based on a distributed high-speed water pressure detection system, provided in another exemplary embodiment of this application. Detailed Implementation

[0018] To enable those skilled in the art to better understand the technical solution, the present application will be described in detail below with reference to the embodiments. The description in this section is only exemplary and explanatory, and should not be used to limit the scope of protection of the present application in any way.

[0019] Currently, the safe and stable operation of urban water supply networks is crucial for ensuring the smooth functioning of society. Early detection and prediction of the evolution of leaks in these networks are of decisive significance in preventing major pipe bursts. Current detection methods typically locate leaks retrospectively through acoustic or flow anomalies after obvious leakage has occurred, often overlooking the hidden damage to pipe structures caused by the complex hydraulic impacts the network endures over long-term operation. This makes it difficult for managers to take effective preventative measures before minor damage worsens.

[0020] Specifically, during routine operation, the emergency opening and closing of valves in a pipe network can cause drastic changes in water flow velocity, generating strong secondary water hammer shock waves within the pipes. As these shock waves propagate through the network, they continuously exert alternating pressures on the pipe walls. Like repeatedly bending a wire, the pipe walls accumulate fatigue under long-term alternating stress, leading to tiny cracks in already vulnerable sections. Once these micro-cracks form, the high-speed water flow within the pipe acts like a wedge, constantly pushing into the cracks. If the local water pressure exceeds the pipe's critical withstand limit, the tiny leaks will be instantly torn apart, rapidly escalating into large-scale pipe bursts. For example, during off-peak water usage at night, the sudden closure of large valves can generate unpredictable pressure fluctuations. These fluctuations wander through the network and repeatedly impact older pipe sections, causing leaks that were initially only pinhole-sized to rapidly expand and rupture under the constant tearing force of the water.

[0021] Therefore, how to accurately assess the degree of fatigue damage to the pipe wall caused by complex pressure fluctuations in the pipeline network, and predict the critical state of tearing and expansion of minor damage under hydraulic impact, is the technical problem that this application needs to solve.

[0022] This application provides a method for predicting pipeline leakage based on a distributed high-speed water pressure detection system, such as... Figure 1 The method shown is a pipeline leakage prediction method based on a distributed high-speed water pressure detection system. This method may include the following steps: Step S110: Obtain the time sequence of opening and closing status of each sampling valve in the water supply network; Step S120: Based on the time sequence characteristics of the opening and closing states, a random forest regression model is used to determine the transient waveform characteristic sequence of non-stationary water hammer impact, and the time sequence of alternating stress on the inner wall of the target pipe section is determined according to the transient waveform characteristic sequence. Step S130: Based on the alternating stress program sequence, a bidirectional long short-term memory network is used to determine the cumulative stress characteristic sequence, and the fatigue accumulation value and residual fatigue life value of the pipe material in the target pipe section are determined according to the cumulative stress characteristic sequence. Step S140: If the fatigue accumulation value exceeds the preset fatigue accumulation threshold, perform nonlinear spatial geometric topological mapping on the residual fatigue life value to determine the micro-defect morphology characteristics of the inner wall of the target pipe section. The micro-defect morphology characteristics include crack length and crack depth. Step S150: Obtain the water flow velocity sequence within the water supply network, and determine the local transient water pressure value based on the water flow velocity sequence; Step S160: If the local transient water pressure value exceeds the preset water pressure threshold, obtain the critical fracture toughness parameter of the target pipe section, and determine the dynamic stress field at the crack tip based on the current local transient water pressure value and the critical fracture toughness parameter. Step S170: Determine the damage and tear propagation rate of the target pipe section wall based on the dynamic stress field at the crack tip, and determine the prediction result of the pipe burst leakage risk corresponding to the target pipe section based on the damage and tear propagation rate.

[0023] According to the pipeline leakage prediction method based on a distributed high-speed water pressure detection system provided in this application, this method can utilize a bidirectional long short-term memory neural network to realize the dynamic accumulation calculation of pipe fatigue damage and the quantitative assessment of residual fatigue life. When the fatigue accumulation value exceeds the safety threshold, the abstract damage value is transformed into the geometric dimensions of the length and depth of microcracks through nonlinear spatial geometric topology mapping technology, establishing the correlation between fatigue damage and micro-defect morphology, and presenting the evolution process of the pipe from fatigue accumulation to microcrack initiation. Subsequently, tensor product calculation is performed by combining the high-speed water flow velocity sequence and the geometric parameters of the microcracks to obtain the local transient water pressure value at the crack tip, quantifying the wedging and tearing effect of the water flow on the crack. On this basis, when the local transient water pressure exceeds the critical state threshold, the tearing propagation rate of the pipe wall damage is determined, and then the pipe burst leakage risk prediction result corresponding to the target pipe section is determined according to the tearing propagation rate. This provides a basis for pipeline managers to take preventive maintenance measures before minor damage deteriorates, reduces the probability of large-scale pipe burst accidents, and thus improves the safety and reliability of urban water supply network operation.

[0024] The following is a detailed description of each step in the pipeline leakage prediction method based on a distributed high-speed water pressure detection system provided in this application: In one embodiment of this application, step S110, which involves obtaining the timing characteristic sequence of the opening and closing states of each sampling valve in the water supply network, further includes the following steps: Figure 2 As shown, the specific content is as follows: Step S210: Obtain the opening feedback signal of each sampling valve in the water supply network; Step S220: Determine the valve opening value based on the opening feedback signal, and determine the opening and closing state timing characteristic sequence based on the valve opening value; Step S230: If there are consecutive missing data in the time series feature sequence of the open and closed states, spline interpolation is used to fill in the missing data.

[0025] Specifically, the rotation angle of the valve plate can be detected in real time by the angular displacement opening sensor built into each sampling valve, converting the physical angle into a standard analog signal. The analog-to-digital converter built into the sensor converts the analog signal into a digital signal; the digital signal is transmitted in real time to the edge computing gateway through a narrowband IoT communication module; the edge gateway performs preliminary verification on the received signal, eliminating abnormal signals that are significantly outside the physical range, and obtaining an effective opening feedback signal. The sampling frequency can be set to 100Hz, a parameter determined based on the time characteristics of the valve opening and closing process and the transient characteristics of water hammer impact. For example, the complete opening and closing process of butterfly valves and gate valves commonly used in water supply networks is usually between 0.5 and 5 seconds. A sampling frequency of 100Hz can provide a time resolution of 0.01 seconds, fully capturing the details of the opening change during the valve opening and closing process; at the same time, this sampling frequency forms a 10:1 ratio with the 1000Hz sampling frequency of the subsequent water pressure sensor, facilitating the time synchronization and alignment of the subsequent valve opening and closing state with the water hammer impact waveform, and avoiding misjudgments of causality due to time deviations.

[0026] Furthermore, the current value of each opening feedback signal can be converted into a corresponding valve opening value. 4mA corresponds to 0% valve opening (fully closed), and 20mA corresponds to 100% valve opening (fully open). Intermediate opening values ​​are calculated using linear interpolation. The formula is: Valve opening value (%) = (Current value - 4mA) / (20mA - 4mA) × 100%. Then, a timestamp accurate to milliseconds and a unique valve identifier are added to each valve opening value. Next, based on the trend of valve opening value changes at adjacent time points, the valve's open / closed status is determined at each time point. Specifically: when the opening value continuously increases and the rate of change exceeds the preset rate of change threshold of 0.5% / 0.01 seconds, it is marked as "open". When the opening value continuously decreases and the rate of change exceeds the preset rate of change threshold of 0.5% / 0.01 seconds, it is marked as "closed". When the opening value remains within the range of 0% ± 0.5%, it is marked as "closed". When the valve opening value remains within 100% ± 0.5%, it is marked as "open". Finally, the timestamp, valve number, valve opening value, and opening / closing status identifier are combined into a temporal feature sequence of opening / closing status. Each sequence element is formatted as [timestamp, valve number, opening value, status identifier]. The preset opening change rate threshold is determined based on the mechanical characteristics of the valve's normal opening and closing. For example, the opening change rate of a water supply network valve during normal opening and closing is typically between 0.1% / 0.01 seconds and 10% / 0.01 seconds. A threshold of 0.5% / 0.01 seconds can effectively distinguish between the valve's static holding state and dynamic opening / closing state, avoiding misjudgments of the status due to sensor measurement noise.

[0027] Furthermore, the open / closed state time-series feature sequence is traversed, and the existence of the sequence element corresponding to each timestamp is checked. If there are three or more consecutive timestamps with missing sequence elements, it is determined that there is continuous missing data. Then, five valid data points before and after the missing data segment are extracted as interpolation control points. Then, the valve opening value corresponding to each missing time point is calculated using cubic spline interpolation. Based on the valve opening value obtained by interpolation, the corresponding open / closed state identifier is determined according to step S220. The interpolated sequence element is inserted into the corresponding position in the open / closed state time-series feature sequence to complete the missing data completion.

[0028] For example, on a DN500 butterfly valve node in a city's water supply network, an angular displacement opening sensor with an accuracy of 0.1% and a narrowband IoT communication module were deployed. The sampling frequency was set to 100Hz, the calibration parameters were 4mA corresponding to 0% opening, 20mA corresponding to 100% opening, the opening change rate threshold was set to 0.5% / 0.01 seconds, and the threshold for determining consecutive missing data was set to 3 sampling points. The angular displacement opening sensor collected the valve's opening change signal in real time. From 10:00:00.000 to 10:00:00.050 on May 10, 2026, the sensor output current signals were 4.0mA, 4.8mA, 5.6mA, 6.4mA, 7.2mA, and 8.0mA, respectively. These signals were transmitted to the edge gateway via narrowband IoT and, after verification, obtained a valid opening feedback signal. Subsequently, the current signal is converted into valve opening values ​​according to the calibration parameters: 4.0mA corresponds to 0%, 4.8mA to 5%, 5.6mA to 10%, 6.4mA to 15%, 7.2mA to 20%, and 8.0mA to 25%. A millisecond-level timestamp and valve number V001 are added to each opening value. Based on the continuous increase of the opening value and the change rate of 5% / 0.01 seconds (exceeding the threshold of 0.5% / 0.01 seconds), the status of all time points is marked as "open". The final generated start / stop state time-series feature sequence segments can be: [2026-05-10 10:00:00.000,V001,0%,On]; [2026-05-10 10:00:00.010,V001,5%,On]; [2026-05-10 10:00:00.020,V001,10%,On]; [2026-05-10 10:00:00.030,V001,15%,On]; [2026-05-10 10:00:00.040,V001,20%,On]; [2026-05-10 10:00:00.050,V001,25%,On]. Next, it was detected that there were four consecutive missing sampling points within the time period from 10:00:00.060 to 10:00:00.090 on 2026-05-10 (exceeding the threshold of three sampling points). Five valid data points before and after the missing segment were extracted as interpolation control points, and the opening values ​​of the missing time points were calculated using cubic spline interpolation, which were 30%, 35%, 40%, and 45%, respectively. Based on the trend of the opening value, the status of these time points was marked as "open". The interpolated sequence elements were inserted into the corresponding positions to complete the missing data and finally obtain the complete open and closed state time series feature sequence.

[0029] In one embodiment of this application, in step S120, a random forest regression model is used to determine the transient waveform feature sequence of non-stationary water hammer impact based on the time sequence feature sequence of the opening and closing states.

[0030] Specifically, the dynamic features of the valve opening and closing process can be extracted by traversing the time-series feature sequences of the opening and closing states. These dynamic features include the maximum opening change rate, average opening change rate, opening and closing duration, initial opening, final opening, number of opening change inflection points, and inflection point time. Then, the static physical features of the valve and pipe section are extracted, including valve type (butterfly valve / gate valve / ball valve), valve nominal diameter, target pipe section diameter, pipe material type, pipe section length, and basic water pressure of the pipe section. Finally, the dynamic and static features are combined into a 12-dimensional input feature vector, with each feature vector corresponding to a complete valve opening and closing process. The constructed input feature vector is then input into a pre-trained random forest regression model. Each decision tree in the model independently splits and calculates the input features, outputting its own predicted water hammer pressure time series. The prediction results of all decision trees are weighted and averaged, with the weights positively correlated with the prediction accuracy of each decision tree on the validation set. Finally, the original sequence of non-stationary water hammer impact transient waveforms, combining the prediction results of all decision trees, is obtained. Next, a timestamp accurate to the millisecond level is added to each data point of the original waveform sequence, with the time start point perfectly aligned with the start time of the valve opening and closing action; baseline correction is performed on the waveform sequence to uniformly adjust the basic water pressure value before the water hammer impact to the actual basic water pressure of the target pipe section; abnormal predicted values ​​in the waveform sequence that exceed the physical reasonable range (less than 0 MPa or greater than 10 MPa) are removed; finally, a non-stationary water hammer impact transient waveform feature sequence containing timestamp, predicted pressure value, and waveform type identifier is generated, with each sequence element in the format of [timestamp, predicted pressure value, waveform type identifier].

[0031] For example, on a DN500 ductile iron pipe section of a city's water supply network, a butterfly valve numbered V001 and a high-speed water pressure sensor numbered P001 are installed. At 10:00:00.000 on May 10, 2026, valve V001 rapidly closes from 100% opening, with an opening / closing duration of 2.0 seconds, ultimately reaching 0% opening. The following input feature vector is extracted from the opening / closing state time-series feature sequence generated in step S110: maximum opening change rate 50% / second, average opening change rate 50% / second, opening / closing duration 2.0 seconds, initial opening 100%, final opening 0%, number of opening change inflection points 0, valve type: butterfly valve, nominal valve diameter: 500mm, target pipe section diameter: 500mm, pipe material type: ductile iron, pipe section length: 120m, basic water pressure: 1.2MPa. Subsequently, a pre-trained random forest regression model (100 decision trees, maximum depth 10 layers) was used for prediction. Each of the 100 decision trees in the model output its own predicted waveform. The 32nd decision tree predicted a pressure peak of 3.38 MPa, and the 76th decision tree predicted a pressure peak of 3.47 MPa. A weighted average of the predictions from all decision trees yielded the original water hammer waveform sequence, with a predicted pressure peak of 3.42 MPa, a pressure rise time of 0.12 seconds, and a pressure decay time constant of 0.8 seconds. Next, a timestamp (time range 10:00:00.000 to 10:00:02.500) was added to the original waveform sequence, baseline correction was performed (adjusting the base water pressure to 1.2 MPa), and outliers were removed to generate a non-stationary water hammer impact transient waveform feature sequence.

[0032] In one embodiment of this application, step S120, when determining the alternating stress on the inner wall of the target pipe section based on the transient waveform characteristic sequence, further includes the following steps: Figure 3 As shown, the specific content is as follows: Step S310: Based on the transient waveform feature sequence, the first-order forward difference algorithm is used to determine the absolute amplitude difference sequence of adjacent sampling points, and the variance and mean absolute amplitude difference of the absolute amplitude difference sequence within the preset time window are calculated. Step S320: Obtain the physical parameters of the inner wall surface of the target pipe section, including geometric parameters and material parameters; Step S330: Determine the circumferential stress amplitude of the inner wall surface of the target pipe section based on the physical parameters and the difference in average absolute amplitude; Step S340: Procedure sequence for determining the alternating stress on the inner wall of the target pipe section based on the circumferential stress amplitude.

[0033] Specifically, the characteristic sequence of the non-stationary water hammer impact transient waveform is traversed, and the absolute value of the pressure difference between two adjacent sampling points is calculated using a first-order forward difference algorithm to generate an absolute amplitude difference sequence. The calculation formula is: D i =|P i+1 -P i |, where D i P represents the absolute amplitude difference at the i-th sampling point. i Let be the predicted pressure value for the i-th sampling point. Then, a sliding window is used, with a window length of 50 sampling points and a step size of 1 sampling point, to traverse the absolute amplitude difference sequence. For each sliding window, the arithmetic mean of all absolute amplitude differences within the window is calculated to obtain the average absolute amplitude difference, and the variance of all absolute amplitude differences within the window is calculated to obtain the pressure fluctuation variance. Finally, an average absolute amplitude difference sequence and a pressure fluctuation variance sequence synchronized with the original waveform sequence are generated. The time window length is determined based on the time-domain characteristics of water hammer impact. For example, the pressure rise time of typical water hammer impact in a water supply network is 0.01-0.1 seconds. A window length of 0.05 seconds can completely cover the main fluctuation process of the water hammer wavefront, while effectively suppressing the influence of high-frequency random noise. A sliding step size of 1 sampling point can ensure that the time resolution of stress calculation is consistent with the original waveform sequence, avoiding the loss of transient characteristics of water hammer impact.

[0034] Furthermore, based on the Lamé stress formula for thick-walled cylinders, a mapping relationship between fluid pressure and circumferential stress of the pipe wall can be established. Then, a dynamic impact amplification factor is introduced to correct the static stress calculation results to reflect the transient dynamic load effect of water hammer impact. Finally, the calculation formula for the circumferential stress amplitude of the inner wall surface of the target pipe section is as follows:

[0035] in, This represents the circumferential stress amplitude of the pipe wall. This is the dynamic impact amplification factor. The outer diameter of the pipe section The inner diameter of the pipe section. This represents the average absolute amplitude difference.

[0036] Furthermore, a corresponding timestamp is added to the circumferential stress amplitude calculated for each sliding window, with the timestamp being the center time point of that sliding window. All timestamped circumferential stress amplitudes are arranged in chronological order to form an initial alternating stress sequence. Linear interpolation is performed on the initial sequence to fill in the data gaps at the beginning and end caused by the sliding window calculation. The final program sequence for generating the alternating stress on the inner wall of the target pipe section is then generated, with each sequence element in the format of [timestamp, circumferential stress amplitude, pressure fluctuation variance].

[0037] For example, in a DN500 ductile iron pipe section of a city's water supply network, the physical parameters of the target pipe section are: outer diameter 500mm, wall thickness 12mm, inner diameter 476mm, elastic modulus 206GPa, Poisson's ratio 0.3, and dynamic impact amplification factor 1.5. The non-stationary water hammer impact transient waveform feature sequence generated in step S120 has a sampling frequency of 1000Hz, a total duration of 2.5 seconds, a base water pressure of 1.2MPa, and a water hammer impact peak value of 3.42MPa. A first-order forward difference algorithm is used to calculate the absolute amplitude difference between adjacent sampling points. For example, if the pressure value at the 100th sampling point is 2.10MPa and the pressure value at the 101st sampling point is 2.13MPa, then the absolute amplitude difference at that point is 0.03MPa; after traversing the entire sequence, an absolute amplitude difference sequence is generated. Subsequently, a sliding window with a length of 50 sampling points and a step size of 1 sampling point was used for statistical calculation. Within the sampling window of sampling points 100-149 at the leading edge of the water hammer wave, the average absolute amplitude difference was calculated to be 0.15 MPa, and the pressure fluctuation variance was 0.02 MPa. 2 Next, the difference between the physical parameters and the average absolute amplitude is substituted into the formula for calculating the circumferential stress amplitude to obtain a circumferential stress amplitude of 28.5 MPa corresponding to the sliding window. Finally, a timestamp of 2026-05-10-10:00:00.125 (corresponding to the center time point of the window) is added to this circumferential stress amplitude. After traversing all sliding windows and performing linear interpolation, a complete sequence of alternating stress on the inner wall of the target pipe section is generated. The sequence time range is from 10:00:00.000 to 10:00:02.450, and the stress values ​​exhibit non-stationary alternating characteristics between 12.5 MPa and 95.6 MPa.

[0038] In one embodiment of this application, in step S130, a bidirectional long short-term memory network is used to determine the cumulative stress characteristic sequence based on the alternating stress time program sequence.

[0039] Specifically, all circumferential stress amplitude data in the program sequence during alternating stress on the inner wall of the target pipe section can be extracted, and the mean and standard deviation of the sequence can be calculated. Z-score normalization is used to normalize all stress amplitude data, mapping the data to a standard normal distribution interval with a mean of 0 and a standard deviation of 1. The normalized stress sequence can be strictly aligned with timestamps to ensure that the time information of each data point is completely consistent with the original sequence. The normalized stress sequence is input into the input layer of a bidirectional long short-term memory neural network. The input layer passes the data to the forward long short-term memory layer, which processes the sequence data in chronological order from earliest to latest. Through the synergistic effect of forget gates, input gates, and output gates, historical stress information that has a significant impact on fatigue damage accumulation is retained, while irrelevant high-frequency noise is filtered out. Simultaneously, the input layer passes the data to the backward long short-term memory layer, which processes the sequence data in chronological order from latest to earliest to determine the impact of future stress changes on current fatigue accumulation. Finally, the output features of the forward and backward layers are concatenated to obtain an intermediate feature sequence containing bidirectional temporal information. Then, the intermediate feature sequence after bidirectional splicing is summed point by point with weights. The weight coefficients are automatically learned by the model during pre-training. The weight magnitude is positively correlated with the contribution of the stress value at the corresponding time to fatigue damage. The weighted summation result is denormalized to restore the stress value with physical meaning. A corresponding timestamp is added to each accumulated stress feature value. The timestamp is completely aligned with the timestamp of the original alternating stress sequence. Finally, all accumulated stress feature values ​​with timestamps are arranged in chronological order to generate a standardized accumulated stress feature sequence. For example, the format of each sequence element is [timestamp, accumulated stress feature value, historical stress extreme value].

[0040] In the above method, by using the forward and backward time-series modeling capabilities of a bidirectional long short-term memory neural network, the long-term dependencies and non-stationary accumulation characteristics in the program sequence under alternating stress can be determined. At the same time, by using a gating mechanism to filter out irrelevant high-frequency noise interference, the final generated cumulative stress feature sequence can accurately reflect the fatigue damage accumulation process of the pipe under water hammer impact.

[0041] In one embodiment of this application, step S130, which determines the cumulative fatigue value and residual fatigue life value of the pipe material in the target pipe section based on the cumulative stress characteristic sequence, further includes the following steps: Figure 4 As shown, the specific content is as follows: Step S410: Based on the accumulated stress characteristic sequence, the rainflow counting algorithm is used to determine the stress amplitude and stress mean in each stress cycle. Step S420: Based on the stress amplitude and the mean stress, the equivalent symmetrical cyclic stress is determined using the Goodman mean correction algorithm; Step S430: Obtain the SN curve of the target pipe section and determine the limit number of cycles for the corresponding equivalent symmetrical cyclic stress from the SN curve; Step S440: Determine the fatigue damage increment generated by a single stress cycle based on the limit cycle number, and accumulate each stress cycle over time to determine the cumulative fatigue value; Step S450: Obtain the average damage accumulation rate of the target pipe section and determine the residual fatigue life value based on the fatigue accumulation value and the average damage accumulation rate.

[0042] Specifically, the rainflow counting method is a standard method for counting fatigue load cycles in engineering. Its principle is to decompose a non-stationary alternating stress time sequence into multiple independent, complete stress cycles. This accurately captures the contribution of stress amplitude and mean to fatigue damage, avoiding the misjudgment of complex stress sequences by simple peak-valley counting methods. The algorithm simulates the process of raindrops flowing and dripping on a roof, traversing the stress-time history to identify all closed stress cycles and extracting the peak value, trough value, stress amplitude (half the difference between the peak and trough values), and stress mean (the average of the peak and trough values) for each cycle. The rainflow counting method requires setting a minimum stress amplitude threshold, which is set to 10% of the pipe's fatigue limit. The rationale is that small stress fluctuations below the pipe's fatigue limit will not cause fatigue damage; filtering such fluctuations significantly reduces computational load and avoids the accumulation of invalid damage. For example, the fatigue limit of Q235 steel pipe is approximately 110 MPa, therefore the minimum stress amplitude threshold is set to 11 MPa.

[0043] Furthermore, stress cycles in actual water supply networks are mostly asymmetric cycles (the mean stress is not zero), while the SN curves of materials are usually obtained through fatigue tests under symmetric cycle conditions (the mean stress is zero). The Goodman mean correction algorithm is a widely used method in engineering for equivalent transformation of asymmetric cyclic stress. It introduces the tensile strength parameter of the material to convert asymmetric cyclic stress into equivalent symmetric cyclic stress, thus enabling direct calculation of fatigue life using standard SN curves. The Goodman correction formula is:

[0044] in, For equivalent symmetrical cyclic stress, The stress amplitude, The average stress value, This refers to the tensile strength of the pipe.

[0045] Furthermore, the SN curve (stress-life curve) describes the relationship between the stress amplitude and the number of cycles required to reach failure under cyclic stress, and is obtained from standard fatigue tests on materials. Significant differences exist in the SN curves of materials with different heat treatment states and under different corrosive environments. Using a pre-established database of pipe SN curves, the corresponding SN curve is selected based on the material, service life, and corrosive environment of the target pipe section. Then, the limiting number of cycles corresponding to the equivalent symmetrical cyclic stress is determined using linear interpolation. The selection of the SN curve must simultaneously consider three factors: material, operating environment, and degree of corrosion. For example, for buried water supply networks, SN curves under corrosive environments must be used, not those under dry laboratory environments. The rationale is that corrosive media such as chloride ions and sulfates in the soil accelerate fatigue crack propagation in the pipe material, significantly reducing its fatigue life. Using SN curves under dry environments would lead to overly optimistic fatigue life predictions.

[0046] Furthermore, the linear damage accumulation theory (Miner's rule) is used to calculate the cumulative fatigue value. Its core assumption is that fatigue damage to a material under different stress levels is linearly additive, and fatigue failure occurs when the total cumulative damage value reaches 1. The increment of fatigue damage generated by a single stress cycle is the ratio of the actual number of cycles to the corresponding limiting number of cycles, i.e.

[0047] in, For the first The actual number of stress cycles corresponding to each stress cycle For the first The limit number of stress cycles corresponds to a given stress cycle. The total cumulative fatigue value is the sum of the damage increments from all stress cycles.

[0048] The Miner's rule is applicable when each stress cycle is independent and the loading sequence has a negligible impact on damage accumulation. However, for special high-risk pipe sections, a nonlinear damage accumulation model can be used for correction.

[0049] Furthermore, the average damage accumulation rate refers to the amount of fatigue damage generated in the pipe per unit time, calculated by statistically analyzing the cumulative fatigue damage of the target pipe section over a past period. The residual fatigue life value refers to the time required for the pipe to reach the fatigue failure threshold (D=1) from its current cumulative fatigue value, calculated using the following formula:

[0050] in, Where D is the residual fatigue life, and D is the current cumulative fatigue value. The average cumulative damage rate is used. The statistical period for the average cumulative damage rate can be set to 30 days. It is important to note that a statistical period that is too short (e.g., 7 days) will result in the rate being overly affected by short-term hydraulic shock fluctuations, leading to unstable prediction results; a statistical period that is too long (e.g., 90 days) will fail to reflect recent changes in the pipeline network's operational status (e.g., increased valve opening and closing frequency, water pressure adjustments, etc.). The system updates the average cumulative damage rate daily to ensure the timeliness of the prediction results.

[0051] For example, taking a Q235B steel pipe section buried in slightly corrosive soil after 10 years of operation as an example, the cumulative stress feature sequence output by the bidirectional long short-term memory neural network has a sampling frequency of 1000Hz, a duration of 1 second, and contains 1000 stress data points with a stress range of 12.5MPa to 95.6MPa. The system calls the rainflow counting algorithm to process this sequence, sets the minimum stress amplitude threshold to 11MPa according to the principle of 10% of the pipe material fatigue limit, filters out small fluctuations with amplitudes less than 11MPa, and identifies one complete stress cycle with a peak value of 145.8MPa and a valley value of 95.2MPa. The calculated stress amplitude of this cycle is 25.3MPa, and the stress mean is 120.5MPa. Combining the tensile strength of Q235B steel pipe ≥375MPa, the actual tensile strength of this pipe section is determined to be 337.5MPa. Substituting into the Goodman mean correction formula, the equivalent symmetrical cyclic stress is calculated to be approximately 39.2MPa. The SN curves of Q235B steel pipes under corresponding corrosive environments were selected from the database. Using linear interpolation, the limiting cycle number corresponding to the equivalent symmetrical cyclic stress was determined to be approximately 2.5 × 10⁻⁶. 5 Based on Miner's linear damage accumulation theory, the fatigue damage increment generated by a single stress cycle is calculated to be 4 × 10⁻⁶. -6 The increment is then added to the historical fatigue accumulation value of 0.456 for this pipe segment over time to obtain the current total fatigue accumulation value of 0.456004. The cumulative fatigue damage of this pipe segment over the past 30 days is then calculated to be 0.045, resulting in an average damage accumulation rate of 0.0015 / day. Finally, these values ​​are substituted into the residual fatigue life calculation formula to obtain a residual fatigue life of approximately 363 days for the target pipe segment, which is then synchronized to the cloud monitoring platform.

[0052] In one embodiment of this application, in step S140, if the fatigue accumulation value exceeds a preset fatigue accumulation threshold, a nonlinear spatial geometric topological mapping is performed on the residual fatigue life value to determine the microscopic defect morphology characteristics of the inner wall of the target pipe segment. The microscopic defect morphology characteristics include crack length and crack depth. The process also includes the following steps: Figure 5 As shown, the specific content is as follows: Step S510: Perform nonlinear spatial geometric topology mapping on the residual fatigue life values ​​to construct a nonlinear spatial geometric coordinate set; Step S520: Determine the topological node matrix between the macroscopic damage state and the microscopic defect morphology based on the nonlinear spatial geometric coordinate set; Step S530: Determine the set of three-dimensional coordinate points of the crack based on the topological node matrix; Step S540: Use a numerical integration algorithm to perform curve fitting on the three-dimensional coordinate point set of the crack to determine the crack length and crack depth.

[0053] Specifically, the fatigue damage evolution of water supply pipeline materials is a multi-scale, highly nonlinear process, ranging from microscopic dislocation movement to macroscopic crack propagation. When the cumulative fatigue value exceeds a preset fatigue accumulation threshold, the pipe material has undergone a qualitative change from having no obvious defects to the initiation of microscopic cracks. The preset fatigue accumulation threshold can be determined based on the soil corrosion environment. For example, buried water supply pipelines are exposed to a corrosive soil environment for extended periods. Corrosive media such as chloride ions and sulfates synergistically interact with fatigue stress, accelerating crack initiation and propagation, reducing the actual fatigue life of the pipe material by 20%-40% compared to dry environments. Therefore: the fatigue accumulation threshold is typically set at 0.850 in dry environments; reduced to 0.800 in weakly corrosive soil environments; further reduced to 0.750 in strongly corrosive soil environments; and for older pipe sections that have been in operation for over 20 years, the threshold can be adjusted to 0.700-0.750 to improve early warning capabilities.

[0054] Furthermore, due to the highly nonlinear mapping relationship between macroscopic fatigue cumulative damage and microscopic crack morphology, it cannot be directly derived using simple linear formulas. However, by using nonlinear spatial geometric topological mapping, the one-dimensional residual fatigue life value and the multi-dimensional cumulative damage state vector can be jointly projected onto a high-dimensional manifold space. This allows manifold learning to capture the intrinsic geometric correlation between macroscopic damage evolution and microscopic defect initiation and propagation. The residual fatigue life value can be fused with the cumulative damage state vector weighted by a time-series forgetting gate to generate a high-dimensional damage feature vector. This vector is then mapped onto the nonlinear space using radial basis functions, constructing a nonlinear spatial geometric coordinate set containing multi-dimensional information such as damage degree, damage rate, and stress history. The time-series forgetting gate weight is set in the range of 0.8 to 0.95, determined through nonlinear regression analysis of the historical damage sequence of the target pipe section over the past 90 days. The rationale is that recent damage data has a greater impact on the current microscopic defect state, while the impact of older damage data gradually weakens; the time-series forgetting gate weight is used to quantify this time decay effect. The nonlinear space dimension can be set to 6 dimensions. The basis for this design is that the micro-crack morphology of water supply network pipes is determined by the degree of damage, stress amplitude, average stress, corrosion rate, service life, and temperature change. Six-dimensional space can fully characterize the interaction of these factors.

[0055] Furthermore, the topology node matrix is ​​a data structure connecting macroscopic damage states and microscopic defect morphologies. Each node corresponds to a coordinate point in the nonlinear space, and the connection weights between nodes represent the correlation strength between macroscopic damage parameters and microscopic defect sizes. By using an equidistant mapping algorithm to find local neighborhood geometric correlations for each point in the nonlinear space geometric coordinate set, and combining this with a damage-defect mapping database pre-established through numerous pipe fatigue tests, the microscopic defect characteristic parameters corresponding to each topology node are determined, ultimately constructing a complete topology node matrix. The number of topology nodes can be set to 1000. With 1000 nodes, the computation time for a single topology mapping can be controlled within 100 milliseconds while ensuring computational accuracy, meeting the response requirements of real-time pipeline network monitoring.

[0056] Furthermore, each node in the topological node matrix only represents a discrete feature point of the microscopic defect and cannot directly describe the continuous morphology of the crack. A spatial coordinate inverse mapping algorithm can be used to convert the high-dimensional feature points in the topological node matrix into actual coordinate points in three-dimensional Euclidean space, generating a discrete three-dimensional coordinate point set for the crack lead and crack surface. This coordinate point set accurately reflects the spatial location, orientation, and geometry of the crack. Then, a cubic spline interpolation algorithm is used to fit the crack's three-dimensional coordinate point set to generate a continuous crack lead curve and crack depth curve. Next, a trapezoidal numerical integration algorithm is used to integrate the arc length of the crack lead curve to calculate the total crack length. The maximum value of the crack depth curve is extracted to obtain the maximum crack depth. The cubic spline interpolation algorithm ensures that the fitted curve is continuous and smooth at all coordinate points, while the trapezoidal numerical integration algorithm is computationally simple and its accuracy meets engineering requirements.

[0057] For example, taking a Q235B steel pipe segment buried in slightly corrosive soil after 10 years of operation as an example, when the fatigue accumulation value of the segment is detected to be 0.856 and greater than the preset fatigue accumulation threshold of 0.800, the temporal forgetting gate weight at the current moment is first calculated to be 0.925. This weight is then multiplied element-wise with the multidimensional cumulative damage state vector [0.45, 0.67, 0.88, 0.32, 0.55, 0.71] to obtain the weighted damage feature vector [0.416, 0.620, 0.814, 0.296, 0.509, 0.657]. Subsequently, the radial basis function is called to map this 6-dimensional feature vector to a preset 6-dimensional nonlinear space, constructing a nonlinear spatial geometric coordinate set containing 6 coordinate dimensions. Next, the nonlinear spatial geometric coordinate set is processed using the equidistant mapping algorithm to determine the 15 nearest neighbor nodes of each coordinate point and construct a local neighborhood graph. At the same time, the pre-established Q235B steel pipe damage-defect mapping database containing 5000 sets of microcrack test data under different fatigue damage states is called. The crack characteristic parameters corresponding to each topological node are calculated through coordinate matching, and finally a 1000×3-dimensional topological node matrix is ​​constructed. Then, a spatial coordinate inverse mapping operation is performed on the topological node matrix. According to the preset requirement of 5 coordinate points per millimeter of crack length and coordinate accuracy of 0.01 millimeters, a crack three-dimensional coordinate point set containing 7 discrete points (0,0,0), (0.25,0.08,0.01), (0.50,0.16,0.02), (0.75,0.24,0.03), (1.00,0.32,0.04), (1.25,0.40,0.05), and (1.25,0.42,0.05) is generated. Finally, the cubic spline interpolation algorithm was used to fit the three-dimensional coordinate points of the crack. The maximum deviation between the fitted curve and the original coordinate points was found to be 0.003 mm, which is less than the preset fitting accuracy threshold of 0.005 mm. Then, the trapezoidal numerical integration algorithm with a step size of 0.01 mm was used to integrate the arc length of the crack leading edge curve, and the crack length was calculated to be 1.25 mm. At the same time, the maximum value of the crack depth curve was extracted, and the crack depth was found to be 0.42 mm.

[0058] In one embodiment of this application, step S150, which involves obtaining a water flow velocity sequence within the water supply network and determining a local transient water pressure value based on the water flow velocity sequence, further includes the following steps: Figure 6 As shown, the specific content is as follows: Step S610: The Navier-Stokes hydrodynamic control equations are used to perform tensor product calculations on the water flow velocity sequence, crack length, and crack depth to construct a three-dimensional fluid-structure coupled stress field model inside the crack of the target pipe section. Step S620: Use the Newton-Raphson iteration method to solve the nonlinear boundary conditions of the three-dimensional fluid-structure interaction stress field model to determine the crack wall friction coefficient and crack opening displacement of the target pipe segment; Step S630: Obtain the preset transient water hammer pressure model, and input the crack wall friction coefficient and crack opening displacement into the preset transient water hammer pressure model to output the local transient water pressure value at the crack of the target pipe section.

[0059] Specifically, based on the microcrack length and depth data generated in step S140, a three-dimensional geometric model of the crack interior is constructed. This model comprises two parts: a fluid domain (the water flow channel inside the crack) and a solid domain (the pipe wall material surrounding the crack). Then, the finite volume method is used to mesh the three-dimensional geometric model. Next, the high-speed water flow velocity sequence collected by a Doppler current profiler is used as the inlet boundary condition of the fluid domain, the base water pressure of the target pipe section is used as the outlet boundary condition of the fluid domain, and the elastic modulus and Poisson's ratio of the pipe wall material are used as the material properties of the solid domain. Finally, tensor product operations are used to couple the Navier-Stokes fluid dynamics governing equations with the solid mechanics equilibrium equations to construct a three-dimensional fluid-structure interaction stress field model inside the crack of the target pipe section. The Navier-Stokes fluid dynamics governing equations are the fundamental partial differential equations describing the motion of viscous incompressible fluids, including the mass conservation equation and the momentum conservation equation, which can accurately simulate the velocity, pressure, and stress distribution of the fluid within complex geometric boundaries. Tensor product operations are used to couple physical quantities of different dimensions, such as fluid velocity field, crack geometry, and pipe wall material properties, to construct a unified mathematical model of a fluid-structure interaction system.

[0060] Furthermore, initial values ​​of pipe wall surface roughness and hydrodynamic viscosity are input into the three-dimensional fluid-structure interaction stress field model. Then, the Newton-Raphson iteration method is used to solve the nonlinear boundary conditions of the model. After each iteration, the calculated values ​​of the crack wall friction coefficient and crack opening displacement are updated. Next, the relative error between two adjacent iterations is calculated. If the relative error is less than a preset iteration convergence threshold, the iteration is considered converged. If the relative error is greater than the preset iteration convergence threshold and the number of iterations has not exceeded the preset maximum number of iterations, the next iteration continues. Finally, when the iteration converges, the final crack wall friction coefficient and crack opening displacement are output. The Newton-Raphson iteration method, by continuously iterating to approximate the solution of the equation system, has the characteristics of fast convergence speed and high computational accuracy, and is widely used in solving nonlinear problems such as fluid-structure interaction. Nonlinear boundary conditions refer to the nonlinear relationship between the boundary conditions at the fluid-structure interaction interface and the fluid velocity, pressure, or solid deformation. For example, the frictional resistance of the crack wall is proportional to the square of the fluid velocity. The coefficient of friction of the crack wall is a dimensionless parameter characterizing the frictional resistance between the water flow and the crack wall, and it is related to the surface roughness of the pipe wall, the water flow velocity, and the crack geometry. Crack opening displacement refers to the perpendicular distance between the two crack walls; it is an important physical quantity for measuring the degree of crack opening and directly affects the flow rate and pressure distribution inside the crack. The iterative convergence threshold can be determined based on the computational accuracy. In this application, it can be set to 1×10⁻⁶. -6 This threshold ensures that the calculation accuracy of the crack wall friction coefficient and crack opening displacement reaches 0.1%, meeting the accuracy requirements of subsequent local transient water pressure calculations. Simultaneously, this threshold controls the number of iterations of the Newton-Raphson method to within 20, ensuring computational efficiency. The maximum number of iterations can also be determined based on the computational accuracy; in this application, it can be set to 50. If convergence is not achieved after 50 iterations, it indicates an anomaly in the model's boundary conditions or parameter settings, requiring parameter recalibration. Setting a maximum number of iterations prevents the program from entering an infinite loop, ensuring system stability.

[0061] Furthermore, a pre-defined transient water hammer pressure model can be established based on water hammer dynamics theory. This model can calculate the transient pressure distribution at any location within the pipe under water hammer impact, according to pipe geometric parameters, fluid physical properties, and boundary conditions. The crack wall friction coefficient and crack opening displacement are used as input parameters to the model, along with physical parameters such as the water flow velocity sequence, water density, dynamic viscosity, and bulk modulus. Next, the fourth-order Runge-Kutta algorithm is used to solve the transient water hammer pressure model using time integration, with a time step set to 0.001 seconds. Finally, the pressure time series at the crack tip is extracted to obtain the local transient water pressure value at the crack tip of the target pipe section. The local transient water pressure value refers to the instantaneous water pressure value at the crack tip that changes with time, and it is the direct driving force leading to crack propagation and tearing.

[0062] It should be noted that crack opening displacement refers to the vertical distance between the two walls of a crack, defining the cross-sectional area of ​​the channel through which water flows within the crack. According to the continuity equation in fluid mechanics, under a constant flow rate, the cross-sectional area of ​​the channel is inversely proportional to the water velocity, which directly determines the magnitude of the dynamic water pressure. Furthermore, crack opening displacement directly affects the flow regime within the crack. When the crack opening displacement is small, the water flow may be laminar; when the crack opening displacement increases to a certain extent, the water flow will transition to turbulent flow, and this change in flow regime will further affect pressure loss and energy distribution. The coefficient of friction of the crack walls is a dimensionless parameter characterizing the frictional resistance between the water flow and the crack walls, directly determining the pressure loss during water flow within the crack. Pressure loss in a pipe is directly proportional to the coefficient of friction and inversely proportional to the pipe diameter. For micro-cracks in water supply networks, due to their extremely small channel size and extremely high relative surface roughness, the impact of frictional resistance on pressure distribution is particularly significant. Therefore, it is necessary to simultaneously introduce the crack wall friction coefficient and crack opening displacement as model input parameters to consider the influence of crack geometry and fluid-wall interaction on local water pressure distribution.

[0063] For example, on a DN500 ductile iron pipe section of a city's water supply network, a microcrack has a length of 1.25 mm and a depth of 0.42 mm. The Doppler velocity profiler collected a high-speed water flow velocity sequence with a sampling frequency of 1000 Hz, and the current water flow velocity is 18.5 m / s. The base water pressure of the target pipe section is 1.2 MPa, the pipe material's elastic modulus is 206 GPa, and Poisson's ratio is 0.3. Based on the crack geometry, a three-dimensional fluid-structure interaction stress field model was constructed. The fluid domain is a rectangular channel with a length of 1.25 mm, a width of 0.42 mm, and a height of 0.05 mm, and the solid domain is the pipe wall material with a thickness of 1 mm surrounding the crack. The finite volume method was used for mesh generation, resulting in a total of 4200 elements. A water flow velocity of 18.5 m / s was used as the inlet boundary condition, and a base water pressure of 1.2 MPa was used as the outlet boundary condition. A complete three-dimensional fluid-structure interaction stress field model was constructed by coupling the Navier-Stokes equations and solid mechanics equilibrium equations through tensor product operations. Subsequently, the initial values ​​for the pipe wall surface roughness (0.05 mm) and the dynamic viscosity of water (1.002 × 10⁻⁶) were set. -3 The Pa·s input is used in the above model. The nonlinear boundary conditions are solved using Newton's iterative method. After 12 iterations, the relative error between two adjacent iterations is 8.7 × 10⁻⁶. -7 Less than the preset convergence threshold of 1×10 -6The iteration converged. The final calculated crack wall friction coefficient was 0.012, and the crack opening displacement was 0.003 mm. Next, a preset transient water hammer pressure model was retrieved, and parameters such as the crack wall friction coefficient (0.012), crack opening displacement (0.003 mm), water flow velocity (18.5 m / s), and water density (1000 kg / m³) were input into the model. Time integration was then performed with a time step of 0.001 seconds. The final calculated local transient water pressure at the crack tip was 4.5 MPa.

[0064] In one embodiment of this application, in step S160, if the local transient water pressure value exceeds a preset water pressure threshold, the critical fracture toughness parameter of the target pipe section is obtained, and the dynamic stress field at the crack tip is determined based on the current local transient water pressure value and the critical fracture toughness parameter. The method also includes the following steps: Figure 7 As shown, the specific content is as follows: Step S710: Obtain the peak value of transient water pressure and the frequency of water pressure pulsation in the local transient water pressure value; Step S720: Based on the transient water pressure peak value, water pressure pulsation frequency and critical fracture toughness parameters, the dynamic stress field at the crack tip is determined using the finite element analysis algorithm.

[0065] Specifically, the local transient water pressure value sequence can be read, the maximum pressure value within each window of the sequence can be calculated, and the maximum value among all maximum pressure values ​​in all windows can be taken as the transient water pressure peak value, and the timestamp corresponding to the peak value can be recorded. Next, a fast Fourier transform is performed on the local transient water pressure value sequence to convert the time-domain signal into a frequency-domain signal, obtaining the amplitude spectrum of the water pressure signal. Finally, the frequency components with amplitudes greater than the preset effective amplitude threshold of the pulsation frequency (i.e., 10% of the transient water pressure peak value) are found in the amplitude spectrum, and the frequency component with the largest amplitude is taken as the water pressure pulsation frequency. The transient water pressure peak value refers to the maximum pressure value in the local transient water pressure value sequence, which is the main factor leading to the maximum stress concentration at the crack tip and directly determines the maximum driving force for crack propagation. The water pressure pulsation frequency refers to the frequency of the periodic fluctuation of the local transient water pressure value over time, reflecting the reflection and superposition effect of the water hammer shock wave inside the crack, and directly affecting the dynamic change characteristics of the stress field at the crack tip.

[0066] Furthermore, based on the geometric dimensions of the microcrack length and depth, a two-dimensional plane strain finite element model of the crack tip is constructed. Then, the mesh at the crack tip is refined; for example, the mesh size in the refined area is set to 0.01 mm, and the mesh size in the unrefined area is set to 0.05 mm, with the total number of elements controlled to within 8000. Next, the transient water pressure peak value and water pressure pulsation frequency are converted into a time-varying water pressure load function and applied to the inner surface of the crack. Simultaneously, the elastic modulus, Poisson's ratio, and critical fracture toughness parameters of the target pipe section are input into the model as material properties. An implicit dynamic finite element algorithm is used to solve the model; for example, the time step is set to 0.001 seconds, and the total computation time is 2.5 seconds. Finally, the stress distribution data of the region surrounding the crack tip is extracted to generate a dynamic stress field at the crack tip containing timestamps, nodal coordinates, and stress components. The dynamic stress field at the crack tip refers to the stress distribution field in the region surrounding the crack tip that changes over time under dynamic hydraulic load. It includes information such as stress magnitude, direction, and distribution characteristics, and is the direct basis for calculating the driving force of crack propagation.

[0067] In one embodiment of this application, step S170, determining the fracture tear propagation rate of the target pipe segment wall based on the dynamic stress field at the crack tip, further includes the following steps: Figure 8 As shown, the specific content is as follows: Step S810: Extract the local transient hydrostatic sequence synchronized with the time difference step from the corresponding time node of the dynamic stress field at the crack tip; Step S820: Determine the transient water pressure gradient value based on the local transient water pressure sequence; Step S830: Obtain the preset dynamic energy release rate integral algorithm model, and synchronously input the transient water pressure gradient value and the dynamic stress field at the crack tip into the preset dynamic energy release rate integral algorithm model to output the energy release rate at the crack tip; Step S840: Determine the micro-tear propagation driving force based on the energy release rate, and input the micro-tear propagation driving force into the fatigue crack propagation law equation to output the failure tear propagation rate.

[0068] Specifically, the dynamic stress field at the crack tip is read, and the timestamp information of all time nodes is extracted. Then, the timestamps are discretized according to a preset time difference step size of 0.001 seconds to generate a time grid. Next, the water pressure value corresponding to each standardized time grid node is extracted from the local transient water pressure value sequence. Finally, the extracted water pressure values ​​are arranged in chronological order to generate a local transient water pressure sequence that is completely synchronized with the dynamic stress field at the crack tip. The format of each sequence element is [timestamp, local transient water pressure value]. The fourth-order central difference algorithm is used to perform gradient differentiation of the local transient water pressure sequence in the time dimension. The transient water pressure gradient value refers to the rate of change of the local transient water pressure value with time, with the unit being megapascals per second. It reflects the drastic dynamic change of water pressure load and is the direct driving force leading to dynamic stress concentration and tear propagation at the crack tip. The system retrieves a pre-defined dynamic energy release rate integral algorithm model from a cloud database. The transient hydraulic pressure gradient sequence and the dynamic stress field at the crack tip are synchronously input into this model to generate a dynamic energy release rate sequence synchronized with the dynamic stress field at the crack tip. The dynamic energy release rate, measured in kilojoules per square meter, refers to the energy released by the system per unit area of ​​crack propagation. It is a key fracture mechanics parameter for measuring the magnitude of the driving force for crack propagation. When the dynamic energy release rate exceeds the material's critical fracture toughness, the crack will rapidly and unstablely propagate. The dynamic energy release rate integral algorithm model is a numerical calculation model based on the J-integral principle. It can accurately calculate the energy release rate under dynamic loads by integrating the stress and strain field around the crack tip.

[0069] Furthermore, the critical fracture toughness parameters of the target pipe section are obtained from the cloud database; then, the difference between the dynamic energy release rate and the critical fracture toughness is calculated to obtain the micro-tear propagation driving force; next, the micro-tear propagation driving force is converted into the corresponding stress intensity factor amplitude; subsequently, the stress intensity factor amplitude is input into the Paris law equation to calculate the crack propagation amount under a single stress cycle; finally, combined with the water pressure pulsation frequency, the single-cycle propagation amount is converted into the propagation rate per unit time, and the failure tear propagation rate of the target pipe section wall is output in millimeters per second. The micro-tear propagation driving force refers to the effective energy that causes micro-tearing and propagation at the crack tip, which is equal to the difference between the dynamic energy release rate and the material's critical fracture toughness; when this difference is positive, the crack will propagate, and the larger the difference, the faster the propagation rate. The fatigue crack propagation law equation is an empirical formula describing the relationship between the fatigue crack propagation rate and the stress intensity factor amplitude. This application adopts Paris law, whose expression is:

[0070] in, This represents the crack propagation rate. denoted as the stress intensity factor amplitude, and C and m as material constants.

[0071] For example, on a DN500 ductile iron pipe section of a city's water supply network, the local transient water pressure peak value is 4.5 MPa, and the water pressure pulsation frequency is 45.5 Hz; the critical fracture toughness of the target pipe section is 55.2 MPa. Timestamps for all time nodes are extracted from the dynamic stress field at the crack tip, and a standardized time grid is generated with a time difference of 0.001 seconds. Water pressure values ​​for corresponding time nodes are extracted from the local transient water pressure value sequence, generating a synchronous local transient water pressure sequence with 100 time nodes. The water pressure values ​​at t=0.125 seconds are 4.5 MPa, t=0.126 seconds are 4.48 MPa, t=0.127 seconds are 4.45 MPa, t=0.123 seconds are 4.42 MPa, and t=0.124 seconds are 4.46 MPa. Subsequently, the transient water pressure gradient at t=0.125 seconds was calculated using a fourth-order central difference algorithm, resulting in a value of -16.7 MPa / s, generating a complete sequence of transient water pressure gradient values, with a maximum gradient value of 150 MPa / s. Next, a preset dynamic energy release rate integral algorithm model was retrieved, and the transient water pressure gradient value sequence and the dynamic stress field at the crack tip were synchronously input into the model, yielding a dynamic energy release rate of 51.2 kJ / m². 2 Finally, the driving force for microscopic tear propagation was calculated to be 51.2 - 55.2 = -4.0 kJ / m. 2 (A negative value indicates that the crack is currently in a stable state and has not propagated unstably). Converting the dynamic energy release rate to a stress intensity factor amplitude of 48.5, and substituting this into the Paris law equation, the crack propagation in a single cycle is calculated to be 3.3 × 10⁻⁶. -7 The tear propagation rate was calculated to be 1.5 × 10 mm, based on the water pressure pulsation frequency of 45.5 Hz. -5 Millimeters per second.

[0072] In one embodiment of this application, step S170, determining the predicted risk of pipe burst leakage corresponding to the target pipe segment based on the rate of tear propagation, further includes the following steps: Figure 9 As shown, the specific content is as follows: Step S910: If the rate of tear propagation exceeds the preset diffusion rate threshold, it is determined that there is a risk of damage to the pipe wall of the current target pipe segment; Step S920: If the rate of damage and tear propagation does not exceed the preset diffusion rate threshold, it is determined that there is no risk of damage to the pipe wall of the current target pipe segment.

[0073] Specifically, the preset diffusion rate threshold refers to the critical crack propagation rate value determined based on the pipe material's mechanical properties, pipeline operating conditions, and industry safety standards. It serves as a quantitative boundary distinguishing between a safe pipe network and a damage risk state. When the pipe wall tear propagation rate exceeds this threshold, the crack enters a rapid propagation stage, posing a risk of pipe rupture and leakage in the short term. For example, in a city's water supply network, there are two pipe sections with different parameters, section A and section B. Section A (ductile iron pipe, operating pressure 1.2 MPa) has a pipe wall tear propagation rate of 1.5 × 10⁻⁶. -5 Millimeters per second, with a preset diffusion rate threshold of 1.0 × 10⁻⁶. -4 Millimeters per second, which shows 1.5 × 10⁻⁶. -5 <1.0×10 -4 Therefore, it is determined that there is no risk of wall damage in pipe section A. The wall tear propagation rate of pipe section B (steel pipe, operating pressure 1.8 MPa) is 2.2 × 10⁻⁶. -4 Millimeters per second. It can be seen that 2.2 × 10⁻⁶ -4 >1.755×10 -4 Therefore, it was determined that the wall of pipe section B currently has a high risk of damage. This can improve the early warning capability and emergency response efficiency of water supply network pipe burst accidents, and effectively reduce the economic losses and social impact caused by pipe burst accidents.

[0074] This application also provides a pipeline leakage prediction system based on a distributed high-speed water pressure detection system. This system may include a first acquisition module, a first determination module, a second determination module, a third determination module, a second acquisition module, a fourth determination module, and a fifth determination module. Specifically, the first acquisition module acquires the opening and closing state time-series characteristic sequences of each sampling valve in the water supply pipeline network; the first determination module uses a random forest regression model to determine the transient waveform characteristic sequence of non-stationary water hammer impact based on the opening and closing state time-series characteristic sequences, and determines the alternating stress time sequence of the inner wall of the target pipe section based on the transient waveform characteristic sequences; the second determination module uses a bidirectional long short-term memory network to determine the accumulated stress characteristic sequence based on the alternating stress time sequence, and determines the fatigue accumulation value and residual fatigue life value of the pipe material in the target pipe section based on the accumulated stress characteristic sequence; the third determination module performs nonlinear spatial determination on the residual fatigue life value if the fatigue accumulation value exceeds a preset fatigue accumulation threshold. The system employs a geometric topology mapping method to determine the microscopic defect morphology features of the inner wall of the target pipe section, including crack length and crack depth. A second acquisition module acquires the water flow velocity sequence within the water supply network and determines the local transient water pressure value based on the sequence. A fourth determination module acquires the critical fracture toughness parameters of the target pipe section material if the local transient water pressure value exceeds a preset water pressure threshold, and determines the dynamic stress field at the crack tip based on the current local transient water pressure value and the critical fracture toughness parameters. A fifth determination module determines the damage and tear propagation rate of the target pipe section wall based on the dynamic stress field at the crack tip, and determines the predicted risk of pipe burst leakage corresponding to the target pipe section based on the damage and tear propagation rate.

[0075] It should be noted that the embodiments of the pipeline leakage prediction system based on the distributed high-speed water pressure detection system provided in this application can be used to execute the processing flow of the embodiments of the pipeline leakage prediction method based on the distributed high-speed water pressure detection system in the above embodiments. Its functions will not be repeated here, but can be referred to the detailed description of the above method embodiments.

[0076] This application also provides an electronic device, which includes one or more processors and memory resources represented by a memory for storing instructions executable by the processor, such as application programs. The application programs stored in the memory may include one or more modules, each corresponding to a set of instructions. Furthermore, the processor is configured to execute instructions to perform the aforementioned method for predicting pipeline leaks based on a distributed high-speed water pressure detection system.

[0077] The electronic device may also include a power supply component configured to perform power management of the electronic device, a wired or wireless network interface configured to connect the electronic device to a network, and an input / output (I / O) interface. The electronic device can be operated based on operating devices stored in memory.

[0078] In one embodiment, a computer device, which may be a server, is also provided. The computer device includes a processor, memory, input / output interfaces (I / O), and a communication interface. The processor, memory, and I / O interfaces are connected via a system bus, and the communication interface is connected to the system bus via the I / O interfaces. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes non-volatile storage media and internal memory. The non-volatile storage media stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device stores data. The I / O interfaces of the computer device are used for exchanging information between the processor and external devices. The communication interface of the computer device is used for communicating with external terminals via a network connection. When the computer program is executed by the processor, it implements a method for predicting pipeline leakage based on a distributed high-speed water pressure detection system.

[0079] In one embodiment, a computer device is provided, which may be a terminal. The computer device includes a processor, memory, input / output interface, communication interface, display unit, and input device. The processor, memory, and input / output interface are connected via a system bus, and the communication interface, display unit, and input device are also connected to the system bus via the input / output interface. The processor of the computer device provides computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The input / output interface of the computer device is used for exchanging information between the processor and external devices. The communication interface of the computer device is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, mobile cellular networks, NFC (Near Field Communication), or other technologies. When the computer program is executed by the processor, it implements a method for predicting pipeline network leakage based on a distributed high-speed water pressure detection system. The display unit of the computer device is used to form a visually visible image and may be a display screen, a projection device, or a virtual reality imaging device. The display screen can be an LCD screen or an e-ink screen. The input device of the computer device can be a touch layer covering the display screen, or buttons, trackballs, or touchpads set on the casing of the computer device, or external keyboards, touchpads, or mice, etc.

[0080] This application also provides a non-transitory computer-readable storage medium, which, when the instructions in the storage medium are executed by the processor of the aforementioned electronic device, enables the aforementioned electronic device to execute a method for predicting pipeline leakage based on a distributed high-speed water pressure detection system.

[0081] This application may take the form of a computer program product implemented on one or more storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing program code. Computer-readable storage media include permanent and non-permanent, removable and non-removable media, and information storage can be implemented by any method or technology. Information may be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to: phase-change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, CD-ROM, digital versatile optical disc (DVD) or other optical storage, magnetic tape, magnetic disk storage or other magnetic storage devices, or any other non-transfer medium that can be used to store information accessible by a computing device.

[0082] It should be noted that although the steps of the pipeline leakage prediction method based on the distributed high-speed water pressure detection system in this application are described in a specific order in the accompanying drawings, this does not require or imply that these steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additional or alternative steps, such as omitting certain steps, combining multiple steps into one step, and / or decomposing one step into multiple steps, should all be considered part of this application.

[0083] It should be understood that this application is not limited to the detailed structure and arrangement of the module of the distributed high-speed water pressure detection system for pipeline leakage prediction proposed in this specification. This application can have other embodiments and can be implemented and executed in various ways. The foregoing variations and modifications fall within the scope of this application. It should be understood that the invention and definition of this application extend to all alternative combinations of two or more individual features mentioned or apparent in the text and / or drawings. All these different combinations constitute multiple alternative aspects of this application. The embodiments described in this specification illustrate the best known mode for implementing this application and will enable those skilled in the art to utilize this application.

[0084] It should be noted that, in this document, the terms "comprising," "including," and any other variations are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Specific examples have been used in this document to illustrate the principles and implementation methods of the technical solutions of this application. The above examples are only for the purpose of helping to understand the methods and core ideas of this application. The above are merely preferred embodiments of this application. It should be pointed out that, due to the limitations of written expression and the objective existence of infinite specific structures, those skilled in the art can make several improvements, modifications, or changes without departing from the principles of this application, and can also combine the above technical features in an appropriate manner; these improvements, modifications, changes, or combinations, or the direct application of the concept and technical solutions of this application to other situations without modification, should all be considered within the scope of protection of this application.

Claims

1. A method for predicting pipeline leakage based on a distributed high-speed water pressure detection system, characterized in that, include: Obtain the time sequence characteristics of the opening and closing status of each sampling valve in the water supply network; Based on the opening and closing state time sequence characteristics, a random forest regression model is used to determine the transient waveform characteristic sequence of non-stationary water hammer impact, and the program sequence for determining the alternating stress on the inner wall of the target pipe section is determined according to the transient waveform characteristic sequence. Based on the alternating stress program sequence, a bidirectional long short-term memory network is used to determine the cumulative stress characteristic sequence, and the fatigue accumulation value and residual fatigue life value of the pipe material in the target pipe section are determined according to the cumulative stress characteristic sequence. If the fatigue accumulation value exceeds the preset fatigue accumulation threshold, then the residual fatigue life value is subjected to nonlinear spatial geometric topological mapping to determine the micro-defect morphology characteristics of the inner wall of the target pipe section. The micro-defect morphology characteristics include crack length and crack depth. Obtain the water flow velocity sequence within the water supply network, and determine the local transient water pressure value based on the water flow velocity sequence; If the local transient water pressure value exceeds the preset water pressure threshold, the critical fracture toughness parameter of the target pipe section is obtained, and the dynamic stress field at the crack tip is determined based on the current local transient water pressure value and the critical fracture toughness parameter. The rate of tear propagation of the target pipe section wall is determined based on the dynamic stress field at the crack tip, and the risk prediction result of pipe burst leakage corresponding to the target pipe section is determined based on the tear propagation rate.

2. The pipeline leakage prediction method based on a distributed high-speed water pressure detection system according to claim 1, characterized in that, The acquisition of the timing feature sequence of the opening and closing states of each sampling valve in the water supply network includes: Obtain the opening feedback signal of each sampling valve in the water supply network; The valve opening value is determined based on the opening feedback signal, and the opening and closing state timing characteristic sequence is determined based on the valve opening value. If the time-series feature sequence of the open / closed state has consecutive missing data, spline interpolation is used to fill in the missing data.

3. The pipeline leakage prediction method based on a distributed high-speed water pressure detection system according to claim 1, characterized in that, The procedure sequence for determining the alternating stress on the inner wall of the target pipe section based on the transient waveform characteristic sequence includes: Based on the transient waveform feature sequence, a first-order forward difference algorithm is used to determine the absolute amplitude difference sequence of adjacent sampling points, and the variance and mean absolute amplitude difference of the absolute amplitude difference sequence within a preset time window are calculated. Obtain the physical parameters of the inner wall surface of the target pipe section, including geometric parameters and material parameters; The circumferential stress amplitude of the inner wall surface of the target pipe section is determined based on the physical parameters and the difference between the average absolute amplitudes. The procedure sequence for determining the alternating stress on the inner wall of the target pipe section based on the circumferential stress amplitude.

4. The pipeline leakage prediction method based on a distributed high-speed water pressure detection system according to claim 1, characterized in that, The step of determining the cumulative fatigue value and residual fatigue life value of the pipe material in the target pipe section based on the accumulated stress characteristic sequence includes: Based on the accumulated stress characteristic sequence, the stress amplitude and average stress in each stress cycle are determined using a rainflow counting algorithm. Based on the stress amplitude and the stress mean, the equivalent symmetrical cyclic stress is determined using the Goodman mean correction algorithm. Obtain the SN curve of the target pipe section and determine the limit number of cycles corresponding to the equivalent symmetrical cyclic stress from the SN curve; The fatigue damage increment generated per stress cycle is determined based on the limit cycle number, and the stress cycles are accumulated over time to determine the cumulative fatigue value. The average damage accumulation rate of the target pipe section is obtained, and the residual fatigue life value is determined based on the fatigue accumulation value and the average damage accumulation rate.

5. The pipeline leakage prediction method based on a distributed high-speed water pressure detection system according to claim 1, characterized in that, The step of performing nonlinear spatial geometric topological mapping on the residual fatigue life value to determine the microscopic defect morphology characteristics of the inner wall of the target pipe segment includes: The residual fatigue life value is subjected to nonlinear spatial geometric topology mapping to construct a nonlinear spatial geometric coordinate set; The topological node matrix between the macroscopic damage state and the microscopic defect morphology is determined based on the nonlinear spatial geometric coordinate set. The set of three-dimensional coordinate points of the crack is determined based on the aforementioned topological node matrix; A numerical integration algorithm is used to perform curve fitting on the three-dimensional coordinate point set of the crack to determine the crack length and the crack depth.

6. The pipeline leakage prediction method based on a distributed high-speed water pressure detection system according to claim 1, characterized in that, The step of determining the local transient water pressure value based on the water flow velocity sequence includes: The Navier-Stokes fluid dynamics control equations are used to perform tensor product operations on the water flow velocity sequence, the crack length, and the crack depth to construct a three-dimensional fluid-structure interaction stress field model inside the crack of the target pipe segment. The nonlinear boundary conditions of the three-dimensional fluid-structure interaction stress field model are solved using the Newton-Raphson iteration method to determine the crack wall friction coefficient and crack opening displacement of the target pipe segment. A preset transient water hammer pressure model is obtained, and the coefficient of friction of the crack wall and the crack opening displacement are input into the preset transient water hammer pressure model to output the local transient water pressure value at the crack of the target pipe section.

7. The pipeline leakage prediction method based on a distributed high-speed water pressure detection system according to claim 1, characterized in that, The dynamic stress field at the crack tip is determined based on the current local transient water pressure value and the critical fracture toughness parameter, including: Obtain the transient water pressure peak value and water pressure pulsation frequency from the local transient water pressure value; Based on the transient water pressure peak value, the water pressure pulsation frequency, and the critical fracture toughness parameter, the dynamic stress field at the crack tip is determined using a finite element analysis algorithm.

8. The pipeline leakage prediction method based on a distributed high-speed water pressure detection system according to claim 1, characterized in that, The step of determining the fracture tear propagation rate of the target pipe segment wall based on the dynamic stress field at the crack tip includes: From the corresponding time nodes of the dynamic stress field at the crack tip, extract the local transient water pressure sequence synchronized with the time difference step size; The transient water pressure gradient value is determined based on the local transient water pressure sequence; A preset dynamic energy release rate integral algorithm model is obtained, and the transient water pressure gradient value and the dynamic stress field at the crack tip are synchronously input into the preset dynamic energy release rate integral algorithm model to output the energy release rate at the crack tip. The micro-tear propagation driving force is determined based on the energy release rate, and the micro-tear propagation driving force is input into the fatigue crack propagation law equation to output the damage tear propagation rate.

9. The pipeline leakage prediction method based on a distributed high-speed water pressure detection system according to claim 1, characterized in that, The step of determining the predicted risk of pipe rupture and leakage corresponding to the target pipe segment based on the tear propagation rate includes: If the rate of tear propagation exceeds a preset diffusion rate threshold, it is determined that the target pipe segment wall is at risk of damage. If the rate of damage and tear propagation does not exceed the preset diffusion rate threshold, then it is determined that there is no risk of damage to the target pipe segment wall.

10. A pipeline leakage prediction system based on a distributed high-speed water pressure detection system, characterized in that, include: The first acquisition module is used to acquire the time sequence of opening and closing status of each sampling valve in the water supply network; The first determining module is used to determine the transient waveform characteristic sequence of non-stationary water hammer impact based on the opening and closing state time sequence using a random forest regression model, and to determine the alternating stress of the inner wall of the target pipe section according to the transient waveform characteristic sequence. The second determining module is used to determine the cumulative stress characteristic sequence based on the alternating stress program sequence using a bidirectional long short-term memory network, and to determine the fatigue accumulation value and residual fatigue life value of the pipe material in the target pipe section according to the cumulative stress characteristic sequence. The third determining module is used to perform nonlinear spatial geometric topological mapping on the residual fatigue life value if the fatigue accumulation value exceeds a preset fatigue accumulation threshold, so as to determine the micro-defect morphology features of the inner wall of the target pipe section. The micro-defect morphology features include crack length and crack depth. The second acquisition module is used to acquire the water flow velocity sequence in the water supply network and determine the local transient water pressure value based on the water flow velocity sequence. The fourth determining module is used to obtain the critical fracture toughness parameter of the target pipe section if the local transient water pressure value exceeds the preset water pressure threshold, and determine the dynamic stress field at the crack tip based on the current local transient water pressure value and the critical fracture toughness parameter. The fifth determining module is used to determine the damage and tear propagation rate of the target pipe section wall based on the dynamic stress field at the crack tip, and to determine the predicted risk of pipe rupture and leakage corresponding to the target pipe section based on the damage and tear propagation rate.

Images