Pipe network fault positioning system based on artificial intelligence

Abstract

The invention discloses a pipe network fault positioning system based on artificial intelligence. The system comprises a data acquisition unit, a preprocessing unit and a diagnostic analysis unit. The data acquisition unit is used for acquiring physical parameter data of the pipeline; the physical parameters comprise pressure, flow, temperature and sound wave signals; the preprocessing unit is used for preprocessing the pipeline physical parameter data to obtain processed data; and the diagnostic analysis unit is used for carrying out diagnostic analysis on the processed data to obtain the position of the pipe network leakage point. The leakage position can be accurately judged, manual excavation is reduced, and technical support is provided for repairing the pipeline in time.

Description

Technical Field

[0001] This invention relates to the field of pipeline network inspection, and more specifically to an artificial intelligence-based pipeline network fault location system. Background Technology

[0002] During the long-term operation of municipal gas and heating pipelines, factors such as complex pipeline structures, large burial depths, long service lives, and variable external environments can all lead to varying degrees of leaks, corrosion, deformation, or blockages. Gas pipeline leaks, in particular, can cause not only energy loss but also major safety accidents such as explosions and poisoning; heating pipeline leaks can result in heat loss, road collapses, and even burns, posing equally serious hazards.

[0003] Currently, traditional methods for locating pipeline faults mainly include manual inspection, excavation along the pipeline, pressure testing, and acoustic leak detection. These methods are not only time-consuming and labor-intensive, but also greatly affected by environmental noise, pipeline overlay structure, and pipe material type. Their detection sensitivity and accuracy are insufficient to meet the needs of modern urban operation and maintenance. Especially for deeply buried underground pipelines or intersecting pipelines in old urban areas, leak points are often highly concealed and have complex, shifting locations, resulting in low location efficiency and frequent false alarms and missed alarms with traditional detection techniques.

[0004] Therefore, accurately locating pipeline faults in the early stages is a core challenge in pipeline maintenance. To solve this problem, an artificial intelligence-based pipeline fault location system is needed, which can accurately determine the location of leaks, reduce manual excavation, and provide technical support for timely pipeline repair. Summary of the Invention

[0005] In view of this, the purpose of this invention is to overcome the deficiencies in the prior art and provide an artificial intelligence-based pipeline fault location system that can accurately determine the location of leaks, reduce manual excavation, and provide technical support for timely pipeline repair.

[0006] The pipeline fault location system based on artificial intelligence of the present invention includes a data acquisition unit, a preprocessing unit, and a diagnostic analysis unit;

[0007] The data acquisition unit is used to acquire pipeline physical parameter data; the physical parameters include pressure, flow rate, temperature, and acoustic signals.

[0008] The preprocessing unit is used to preprocess the pipeline physical parameter data to obtain processed data;

[0009] The diagnostic analysis unit is used to perform diagnostic analysis on the processed data to obtain the location of the pipeline leak point.

[0010] Furthermore, the pipeline physical parameter data undergoes preprocessing, specifically including:

[0011] Noise in the acoustic signal is removed, and time-domain features, frequency-domain features, and time-frequency-domain features are extracted from the acoustic signal to form a feature vector.

[0012] Furthermore, the processed data is analyzed using an AI model; the AI ​​model includes module A, module B, and module C.

[0013] Module A is used to learn the joint distribution of multiple parameters under normal operating conditions. It establishes a normal baseline through a deep autoencoder or a support vector machine. Any state that deviates significantly from the normal baseline will be marked as an abnormal operating condition.

[0014] Module B is used to classify the marked abnormal operating conditions using convolutional neural networks or long short-term memory networks: for thermal pipelines, it identifies energy enhancement in specific frequency bands in the acoustic spectrum to distinguish between leaks, equipment vibrations, or water hammer; for gas pipelines, it analyzes the diffusion pattern of methane concentration sequences and, in conjunction with wind speed and direction data, determines the type of leak and distinguishes it from non-leakage sources.

[0015] Module C is used to calculate the location of pipeline leaks by detecting the time difference between the arrival of sound waves at different sensors.

[0016] Furthermore, significant deviations were identified using the following method:

[0017] By calculating the quantitative deviation between real-time data and a normal baseline, this deviation is compared with a threshold obtained in advance based on historical normal data. If the deviation exceeds the threshold, it is judged as a significant deviation.

[0018] Furthermore, the type of leakage can be determined using the following method:

[0019] An abnormal methane concentration event has been confirmed, and complete time window data from the start of the concentration rise to its fall has been obtained;

[0020] Extract the concentration readings of all methane sensors that triggered alarms during the abnormal event from the database, and time-align the concentration data with wind speed and direction data at the same timestamp.

[0021] For each sensor, a linear fit is performed on the concentration rise phase to calculate its maximum rise slope or the time required to rise from the baseline concentration to the peak concentration.

[0022] If an extremely steep upward slope occurs, and the corresponding time is from a few seconds to a few minutes, or if the concentration curve shows a sharp pulse shape, reaching a peak and then rapidly declining, then the leakage type is instantaneous release.

[0023] If there is a relatively gentle upward slope, and the corresponding time is tens of minutes or longer, or if the concentration does not drop rapidly after the peak but remains at a high level, then the leakage type is a continuous leakage.

[0024] Furthermore, it is distinguished from non-leakage sources, specifically including:

[0025] Obtain the comprehensive feature vector of the non-leakage source; the comprehensive feature vector includes gas composition ratio, reverse trajectory result, concentration sequence morphology, geographical location information, and historical event matching degree;

[0026] By learning the comprehensive feature vectors of non-leaking sources, non-leaking sources can be distinguished.

[0027] Furthermore, the location of the pipeline leak point is calculated, specifically including:

[0028] Identify the pipe section where the pipeline leak occurred, locate the two sensors S1 and S2 placed on that pipe section, with the pipe length between S1 and S2 being L, and the leak point located between S1 and S2.

[0029] Step 1: Identify the starting point of the leakage signal, and timestamp the starting points of the signals detected by S1 and S2, respectively. and ;

[0030] Step 2: Calculate the absolute time difference between the arrival of the signal at the two sensors: ;

[0031] according to The sign of the value determines the relative location of the leak point; assuming... If the signal arrives at S2 first and then at S1, then the leak point is closer to sensor S2.

[0032] Step 3: Determine the speed of sound propagation in the pipe ;

[0033] Step 4: Calculate the distance difference between the leak point and the sensor: ;in, It is the difference in the path from the leak point to S1 and S2;

[0034] Step 5: Let the distance from the leak point to sensor S1 be x meters, then the distance to sensor S2 is Lx meters; the path difference Δd is equal to the absolute value of the difference between these two distances: |2x-L|=Δd;

[0035] Since the leak point has been determined to be closer to S2, the absolute value can be removed. The equation is: 2x-L=Δd; Finally, the position x of the leak point from sensor S1 is calculated: x=(L+Δd) / 2, and the position Lx of the leak point from sensor S2 is calculated.

[0036] Furthermore, the propagation speed is dynamically corrected using the following method:

[0037] For liquid pipelines, the corrected propagation velocity is:

[0038] v=1402.5+5T-0.055T²+0.0003T³+(1.33-0.01T)(P-1);

[0039] Where v is the propagation speed, T is the fluid temperature, and P is the fluid pressure;

[0040] For gas pipelines, the corrected propagation speed is:

[0041] v= ;

[0042] Where v is the propagation speed, T is the absolute temperature of the gas, R is the specific gas constant, and γ is the specific heat ratio of the gas.

[0043] The beneficial effects of this invention are as follows: This invention discloses an artificial intelligence-based pipeline fault location system that acquires key physical parameters such as pressure, flow rate, temperature, and acoustic signals in real time, forming multi-source monitoring data. It performs noise reduction, outlier correction, and feature extraction on the raw data to improve data quality and stability. Based on an artificial intelligence model, it performs fusion analysis on the processed parameters, automatically identifies abnormal patterns, and outputs the specific location of the leak. This invention achieves precise location of pipeline leaks, offering higher detection efficiency and stronger positioning accuracy compared to traditional manual inspections or single-signal analysis methods. It also reduces human judgment errors and improves robustness under complex operating conditions. Attached Figure Description

[0044] The present invention will be further described below with reference to the accompanying drawings and embodiments:

[0045] Figure 1 This is a schematic diagram of the pipeline fault location system of the present invention. Detailed Implementation

[0046] The present invention will be further described below with reference to the accompanying drawings, as shown in the figures:

[0047] This embodiment discloses an artificial intelligence-based pipeline fault location system, including a data acquisition unit, a preprocessing unit, and a diagnostic analysis unit;

[0048] The data acquisition unit is used to acquire pipeline physical parameter data; the physical parameters include pressure, flow rate, temperature, and acoustic signals.

[0049] The preprocessing unit is used to preprocess the pipeline physical parameter data to obtain processed data;

[0050] The diagnostic analysis unit is used to perform diagnostic analysis on the processed data to obtain the location of the pipeline leak point.

[0051] In this embodiment, the data acquisition unit includes pressure, flow, and temperature sensors, as well as fiber optic acoustic sensors or high-performance vibration sensors. The pressure, flow, and temperature sensors upload data at minute-level frequencies, forming the system's basic state perception. At critical pipeline nodes, such as valves, branches, and compensators, distributed fiber optic acoustic sensors or high-performance vibration sensors are used to collect high-frequency sound waves or vibration signals from the pipeline in real time. This is crucial for capturing the hissing sound of minute leaks.

[0052] In this embodiment, by performing local preprocessing on high-frequency, large-volume signals, such as data cleaning and alignment, the burden on the cloud platform is reduced and communication consumption is decreased.

[0053] Preprocessing of pipeline physical parameter data includes:

[0054] Noise, such as vibrations from passing vehicles, is removed from the sound wave signal. Time-domain, frequency-domain, and time-frequency-domain features are then extracted from the sound wave signal to form a feature vector. The time-domain features include the mean and variance; the frequency-domain features can be obtained through FFT transform; and the time-frequency-domain features can be obtained through wavelet transform.

[0055] By preprocessing high-frequency, large-volume acoustic signals locally, the amount of raw data uploaded can be significantly reduced, alleviating the computing and communication pressure on the cloud and improving the overall system response speed. Simultaneously, by removing noise and extracting multi-dimensional features in the time, frequency, and time-frequency domains, more representative feature vectors can be generated, providing high-quality input for subsequent AI model fault identification, thereby further improving the accuracy and stability of leak location.

[0056] In this embodiment, an AI model is used to perform diagnostic analysis on the processed data; the AI ​​model includes module A, module B, and module C.

[0057] Module A is used to learn the joint distribution of multiple parameters under normal operating conditions. It establishes a normal baseline through a deep autoencoder or a support vector machine, and any state that deviates significantly from the normal baseline will be marked as an abnormal operating condition. Among them, multiple parameters include pressure, flow rate, and acoustic spectrum. It aligns pressure, flow rate, temperature, acoustic characteristics, and concentration data from different locations and timestamps and unifies them into a spatiotemporal coordinate system. For example, it performs correlation analysis on the timeline between pressure fluctuations at point A at a certain moment and acoustic events collected at point B.

[0058] Module B is used to classify the marked abnormal operating conditions using convolutional neural networks or long short-term memory networks: for thermal pipelines, it identifies energy enhancement in specific frequency bands in the acoustic spectrum to distinguish between leaks, equipment vibrations, or water hammer; for gas pipelines, it analyzes the diffusion pattern of methane concentration sequences and, combined with wind speed and direction data, determines the type of leak and distinguishes it from non-leakage sources; wherein, the leak types include continuous leaks and instantaneous releases; non-leakage sources include biogas, etc.

[0059] Module C is used to calculate the location of pipeline leaks by detecting the time difference between the arrival of sound waves at different sensors.

[0060] By introducing an AI model consisting of modules A, B, and C, end-to-end intelligent analysis is achieved, encompassing anomaly detection, fault type identification, and leak location. Module A establishes a normal operating condition baseline, improving the reliability of anomaly identification; Module B classifies different anomaly sources based on deep feature analysis, significantly reducing the false alarm rate; and Module C utilizes a multi-sensor time difference method to accurately locate leaks. The synergy of these three modules makes fault identification more comprehensive and location more precise, enhancing the overall intelligence level of the system.

[0061] In this embodiment, the determination of significant deviation does not rely on a single fixed threshold, but is achieved by calculating a quantified deviation score or probability through a deep autoencoder or a type of support vector machine, and comparing this score with a dynamic or static threshold.

[0062] This invention emphasizes deviations from the combined distribution of multiple parameters, rather than just a single parameter exceeding the limit. For example, pressure and flow rate may be within historical ranges individually, but their combined relationship (such as the pressure-flow curve) may be abnormal. This subtle, correlated anomaly is precisely what AI models can capture, but traditional threshold alarms cannot.

[0063] To accommodate slow, legitimate changes in the system, such as seasonal variations in water or gas usage patterns, thresholds can be dynamic. For example, the system can periodically (e.g., monthly) recalculate thresholds using recent normal data, or use a sliding window to update statistics on the normal baseline.

[0064] Significant deviations are determined using the following method:

[0065] By calculating the quantitative deviation between real-time data and the normal baseline, this deviation is compared with a threshold obtained in advance based on historical normal data. If the deviation exceeds the threshold, it is judged as a significant deviation, thereby triggering an abnormal operating condition flag.

[0066] Through the above processing, early and complex anomalies that are difficult to detect by traditional methods can be identified from a multi-dimensional and correlational perspective.

[0067] In this embodiment, the leakage type is determined according to the following method:

[0068] Through the processing of module A, an abnormal methane concentration event was confirmed, and complete time window data from the start of the concentration rise to the fall was obtained;

[0069] Extract the concentration readings of all methane sensors that triggered alarms during the abnormal event from the database, and time-align the concentration data with wind speed and direction data at the same timestamp.

[0070] For each sensor, a linear fit is performed on the concentration rise phase to calculate its maximum rise slope or the time required to rise from the baseline concentration to the peak concentration.

[0071] If an extremely steep upward slope occurs, and the corresponding time is from a few seconds to a few minutes, or if the concentration curve shows a sharp pulse shape that reaches its peak and then drops rapidly, the leakage type is instantaneous release; in this case, the concentration curve shows a sharp pulse shape.

[0072] If a relatively gentle upward slope is observed, and the corresponding time is tens of minutes or longer, or if the concentration does not decrease rapidly after the peak but remains at a high level, then the leak type is a continuous leak. The concentration peak may exhibit a relatively broad plateau shape.

[0073] By analyzing the complete time window of methane concentration anomalies and aligning multi-sensor data with wind speed and direction, it is possible to distinguish between instantaneous releases and continuous leaks based on the rising slope, duration, and curve shape of concentration changes. Through linear fitting and feature extraction, it is possible to accurately identify sharp, pulse-like instantaneous releases and plateau-like continuous leaks, making leak type identification more intuitive and reliable, thereby improving the targeting and safety of subsequent response strategies.

[0074] In this embodiment, interference from non-leakage sources is effectively eliminated through multi-source data fusion and source tracing analysis. Specifically, biogas may exhibit certain periodicity or regularity, such as being related to temperature, air pressure, or human activity, repeatedly showing similar instantaneous release patterns at specific locations. Pipeline leaks, once they occur, will persist unless repaired, will not disappear on their own, and their patterns differ from those of biogas.

[0075] Distinguishing from non-leakage sources specifically includes:

[0076] Obtain the comprehensive feature vector of the non-leakage source; the comprehensive feature vector includes gas composition ratio, reverse trajectory result, concentration sequence morphology, geographical location information, and historical event matching degree;

[0077] The system distinguishes non-leakage sources by learning a comprehensive feature vector. This feature vector is input into an existing classification model and trained on a large dataset of labeled pipeline leaks and methane interference events. This allows the model to learn the complex, nonlinear boundaries that differentiate between the two. Finally, a confidence level is output, for example, 95% confidence for pipeline leaks or 98% confidence for methane interference.

[0078] By constructing a comprehensive feature vector that includes gas composition ratios, reverse trajectories, concentration patterns, geographical locations, and historical event matching, and using a trained classification model for identification, the nonlinear differences between leaks and non-leakage sources such as biogas can be effectively learned. The model can output the confidence level of leaks or disturbances, achieving highly reliable classification judgments, thereby significantly reducing the false alarm rate and improving the accuracy and stability of pipeline fault identification.

[0079] In this embodiment, calculating the location of the pipeline leak point specifically includes:

[0080] Identify the pipe section where the pipeline leak occurred, and locate the two sensors S1 and S2 placed on that pipe section. The length of the pipe between S1 and S2 is L, and the leak point is located between S1 and S2. When the pipeline leaks, the generated sound waves or vibration signals of a specific frequency will propagate along the pipe wall and the medium to both ends, and the sensors S1 and S2 on the pipeline will capture the signal almost simultaneously.

[0081] Step 1: After filtering and denoising the received signal, use the threshold method, cross-correlation method, or wavelet transform modulus maxima method to identify the starting point of the leakage signal, i.e., the moment when the signal first significantly deviates from the background noise. Timestamps are then added to the starting points of the signals detected by S1 and S2, denoted as follows: and ;

[0082] Step 2: Calculate the absolute time difference between the arrival of the signal at the two sensors: ;

[0083] Determine the relative location of the leak point based on the sign of Δt; assuming... If the signal arrives at S2 first and then at S1, then the leak point is closer to sensor S2.

[0084] Step 3: Determine the speed of sound propagation in the pipe Among these parameters, the speed at which sound waves propagate in a specific pipe is a key parameter, which can be determined in advance through experiments or calculated using physical formulas.

[0085] Step 4: Calculate the distance difference between the leak point and the sensor: Where Δd is the path difference from the leak point to S1 and S2;

[0086] Step 5: Let the distance from the leak point to sensor S1 be x meters, then the distance to sensor S2 is Lx meters; the path difference Δd is equal to the absolute value of the difference between these two distances: |2x-L|=Δd;

[0087] Since the leak point has been determined to be closer to S2 (i.e., x > L / 2), the absolute value can be discarded, and the equation is: Finally, the distance x from the leak point to sensor S1 is calculated as x = (L + Δd) / 2, and the distance Lx from the leak point to sensor S2 is calculated as Lx.

[0088] In this embodiment, the propagation speed v of sound waves in the pipe is not constant; it strongly depends on the properties of the fluid medium itself and the physical state of the pipe. Using a fixed sound speed value would introduce a large positioning error. In step three, the propagation speed is dynamically corrected according to the following method:

[0089] For liquid pipelines, the primary influencing factor is fluid temperature. For example, the elastic modulus and density of water change with temperature. The corrected propagation velocity is:

[0090] v=1402.5+5T-0.055T²+0.0003T³+(1.33-0.01T)(P-1);

[0091] Where v is the propagation speed (m / s), T is the fluid temperature (degrees Celsius), and P is the fluid pressure; the data from the temperature sensor T and pressure sensor P of the pipe section where the leak point is located or the adjacent pipe section can be read in real time, thereby calculating a more accurate propagation speed v under the current operating conditions.

[0092] For gas pipelines, the main influencing factors include gas composition, temperature, and pressure; the corrected propagation speed is:

[0093] v= ;

[0094] Where v is the propagation speed (m / s), T is the absolute temperature of the gas (Kelvin), R is the specific gas constant, and γ is the specific heat ratio of the gas (approximately 1.3 for natural gas).

[0095] By integrating data from temperature and pressure sensors and based on a mature physical acoustic model, real-time, dynamic, and high-precision correction of the sound velocity within the pipeline was achieved. This process elevates acoustic positioning from a rough method relying on empirical estimation to a precise quantitative analysis tool, making it one of the core technological guarantees for accurate positioning.

[0096] This invention also includes: weighted fusion of acoustic positioning results and hydraulic model inversion results to improve positioning accuracy. An intelligent weighted fusion module combines the advantages of both to output a more accurate and reliable final positioning result. The workflow of the intelligent weighted fusion module is as follows:

[0097] Under ideal, simple operating conditions, acoustic positioning offers high accuracy and fast response, and the system assigns it a very high weight to achieve rapid and precise positioning. In complex conditions where acoustic signals are interfered with, the hydraulic model provides a reliability supplement at the system level, preventing missed or false alarms due to acoustic method failure. During daily operation, the two methods are cross-validated. If the acoustic positioning results differ significantly from the hydraulic model inversion results, the system triggers a low-confidence warning, prompting manual intervention for verification, thus forming a closed-loop, continuously optimizing diagnostic system.

[0098] The hydraulic model can employ various existing mature pipeline hydraulic and leakage inversion models, possessing the capability to be directly used for pressure-flow response simulation, node state inversion, and leakage location identification; for example, dynamic hydraulic inversion models based on Kalman filtering, such as Extended Kalman Filter (EKF) and Unscented Kalman Filter (UKF). By using the pipeline hydraulic model as the state equation and monitoring pressure or flow rate as the observation, dynamic estimation of leakage points is achieved.

[0099] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A pipeline fault location system based on artificial intelligence, characterized in that: It includes a data acquisition unit, a preprocessing unit, and a diagnostic analysis unit; The data acquisition unit is used to acquire pipeline physical parameter data; the physical parameters include pressure, flow rate, temperature, and acoustic signals. The preprocessing unit is used to preprocess the pipeline physical parameter data to obtain processed data; The diagnostic analysis unit is used to perform diagnostic analysis on the processed data to obtain the location of the pipeline leak point.

2. The pipeline fault location system based on artificial intelligence according to claim 1, characterized in that: Preprocessing of pipeline physical parameter data includes: Noise in the acoustic signal is removed, and time-domain features, frequency-domain features, and time-frequency-domain features are extracted from the acoustic signal to form a feature vector.

3. The pipeline fault location system based on artificial intelligence according to claim 1, characterized in that: The processed data is used for diagnostic analysis using an AI model; the AI ​​model includes module A, module B, and module C. Module A is used to learn the joint distribution of multiple parameters under normal operating conditions. It establishes a normal baseline through a deep autoencoder or a support vector machine. Any state that deviates significantly from the normal baseline will be marked as an abnormal operating condition. Module B is used to classify the marked abnormal operating conditions using convolutional neural networks or long short-term memory networks: for thermal pipelines, it identifies energy enhancement in specific frequency bands in the acoustic spectrum to distinguish between leaks, equipment vibrations, or water hammer; for gas pipelines, it analyzes the diffusion pattern of methane concentration sequences and, in conjunction with wind speed and direction data, determines the type of leak and distinguishes it from non-leakage sources. Module C is used to calculate the location of pipeline leaks by detecting the time difference between the arrival of sound waves at different sensors.

4. The pipeline fault location system based on artificial intelligence according to claim 3, characterized in that: Significant deviations are determined using the following method: By calculating the quantitative deviation between real-time data and a normal baseline, this deviation is compared with a threshold obtained in advance based on historical normal data. If the deviation exceeds the threshold, it is judged as a significant deviation.

5. The pipeline fault location system based on artificial intelligence according to claim 3, characterized in that: Determine the type of leakage using the following method: An abnormal methane concentration event has been confirmed, and complete time window data from the start of the concentration rise to its fall has been obtained; Extract the concentration readings of all methane sensors that triggered alarms during the abnormal event from the database, and time-align the concentration data with wind speed and direction data at the same timestamp. For each sensor, a linear fit is performed on the concentration rise phase to calculate its maximum rise slope or the time required to rise from the baseline concentration to the peak concentration. If an extremely steep upward slope occurs, and the corresponding time is from a few seconds to a few minutes, or if the concentration curve shows a sharp pulse shape, reaching a peak and then rapidly declining, then the leakage type is instantaneous release. If there is a relatively gentle upward slope, and the corresponding time is tens of minutes or longer, or if the concentration does not drop rapidly after the peak but remains at a high level, then the leakage type is a continuous leakage.

6. The pipeline fault location system based on artificial intelligence according to claim 3, characterized in that: Distinguishing from non-leakage sources specifically includes: Obtain the comprehensive feature vector of the non-leakage source; the comprehensive feature vector includes gas composition ratio, reverse trajectory result, concentration sequence morphology, geographical location information, and historical event matching degree; By learning the comprehensive feature vectors of non-leaking sources, non-leaking sources can be distinguished.

7. The pipeline fault location system based on artificial intelligence according to claim 3, characterized in that: Calculating the location of leaks in the pipeline network specifically includes: Identify the pipe section where the pipeline leak occurred, locate the two sensors S1 and S2 placed on that pipe section, with the pipe length between S1 and S2 being L, and the leak point located between S1 and S2. Step 1: Identify the starting point of the leakage signal, and timestamp the starting points of the signals detected by S1 and S2, respectively. and ; Step 2: Calculate the absolute time difference between the arrival of the signal at the two sensors: ; according to The sign of the value determines the relative location of the leak point; assuming... If the signal arrives at S2 first and then at S1, then the leak point is closer to sensor S2. Step 3: Determine the speed of sound propagation in the pipe ; Step 4: Calculate the distance difference between the leak point and the sensor: ;in, It is the difference in the path from the leak point to S1 and S2; Step 5: Let the distance from the leak point to sensor S1 be x meters, then the distance to sensor S2 is Lx meters; the path difference Δd is equal to the absolute value of the difference between these two distances: |2x-L|=Δd; Since the leak point has been determined to be closer to S2, the absolute value can be removed. The equation is: 2x-L=Δd; Finally, the position x of the leak point from sensor S1 is calculated: x=(L+Δd) / 2, and the position Lx of the leak point from sensor S2 is calculated.

8. The pipeline fault location system based on artificial intelligence according to claim 7, characterized in that: The propagation speed is dynamically corrected using the following method: For liquid pipelines, the corrected propagation velocity is: v=1402.5+5T-0.055T²+0.0003T³+(1.33-0.01T)(P-1); Where v is the propagation speed, T is the fluid temperature, and P is the fluid pressure; For gas pipelines, the corrected propagation speed is: v= ； Where v is the propagation speed, T is the absolute temperature of the gas, R is the specific gas constant, and γ is the specific heat ratio of the gas.

Images