Pipeline leakage monitoring method and system

Abstract

The invention discloses a pipeline leakage monitoring method and system, and belongs to the technical field of pipe network leakage monitoring. Based on leakage and non-leakage signal spectrum comparison, a high-difference frequency band far away from low-frequency noise is extracted as a characteristic frequency band, high-frequency effective characteristics are focused, and low-frequency interference is avoided; savitzky-Golay smooth filtering is used to eliminate random noise and maintain spectrum peak shape characteristics at the same time, so that the representativeness of a characteristic curve is improved; amplitude difference caused by environmental factors is eliminated through normalization processing, so that the characteristic curves under different working conditions can be compared; shape and amplitude dual verification is carried out through a Pearson's correlation coefficient PPC and a mean square error RMES, and the misjudgment rate of pipeline leakage is reduced; and through a dynamic threshold setting mechanism, different pipe types and working condition changes are self-adapted, and the problem of sensitivity reduction caused by a fixed threshold is avoided.

Description

Technical Field

[0001] The present invention relates to the technical field of pipeline network leakage monitoring, and in particular to a pipeline leakage monitoring method and system. Background Art

[0002] Water supply pipeline leak detection is crucial for conserving water resources and reducing economic losses caused by water waste. Identifying water supply pipeline leaks based on acoustic signals is an ideal detection method. Current research focuses on extracting time-domain features of leak signals, such as kurtosis, crest factor, sample entropy, and mean energy. Literature indicates that these methods can achieve a minimum accuracy of 93% for leak detection. However, due to environmental factors and varying operating conditions, methods based on leak signal time-domain features are susceptible to low-frequency noise. Researchers have proposed a series of modal decomposition methods to suppress background noise in time-domain signals, but these algorithms require high computing power. Meanwhile, machine learning-based pipeline leak detection schemes can achieve relatively accurate leak detection, but they require a high volume of field data. Data validation indicates that these methods are ineffective in complex operating conditions and for distinguishing leak signals from other non-leak events, such as water flowing through valves, which are consistent with pipeline structure and pump room background noise.

[0003] In order to accurately distinguish leakage events from complex interference events, the present invention aims to extract the standardized characteristic curve of the leakage signal by studying the "leakage characteristic spectrum curve" extraction method at the frequency domain level; based on the leakage characteristic spectrum curve, the leakage signal determination method is studied, and the leakage determination threshold is obtained to improve the accuracy of pipeline leakage determination. Summary of the Invention

[0004] To address the above problems, a pipeline leakage monitoring method and system are provided. The present invention extracts high-difference frequency bands far away from low-frequency noise as characteristic frequency bands based on the comparison of leakage and non-leakage signal spectra, focuses on high-frequency effective features, avoids low-frequency interference, and provides pure signals for subsequent analysis; eliminates random noise while maintaining the spectrum peak characteristics through Savitzky-Golay smoothing filtering, thereby improving the representativeness of the characteristic curve; eliminates the amplitude difference caused by environmental factors through normalization processing, so that the characteristic curves under different working conditions can be compared; dual verification of shape and amplitude is performed through the Pearson correlation coefficient (PPC) and the mean square error (RMES), thereby reducing the misjudgment rate; and adopts a dynamic threshold setting mechanism to adapt to different pipe types and working conditions to avoid the sensitivity drop caused by fixed thresholds.

[0005] To achieve the above object, the technical solution adopted by the present invention is: a pipeline leakage monitoring method, comprising the following steps: S1: transforming the pipeline vibration time domain signal collected by the interferometer through Fourier transform x (t) is converted into a frequency domain signal X (f ); S2: By comparing and analyzing the spectrum amplitude distribution of the missing signal and the non-missing signal, the high-frequency band with significant amplitude difference and far away from the low-frequency noise is determined as the characteristic frequency band of the missing signal f 1~ f 2; S3: performing smoothing filtering and normalization processing on the frequency domain signal within the characteristic frequency band to generate a standardized loss characteristic spectrum curve and construct a loss characteristic curve library; S4: Calculate the Pearson correlation coefficient PPC and the mean square error RMES between the signal to be measured and each curve in the dropout characteristic curve library; S5: Dynamically set the cross-correlation threshold and the RMES Threshold; S6: Pearson correlation coefficient PPC ≥ cross-correlation threshold, and mean square error RMES ≤ RMES The thresholds are combined to determine whether the signal to be tested is a missing signal.

[0006] Preferably, the Fourier transform formula is as follows: in, X ( f ) is the frequency domain signal, x (t) is the time domain signal of pipeline vibration.

[0007] Preferably, the characteristic frequency band f 1~ f 2 Dynamically adjusted to 500Hz~3kHz according to the pipe type.

[0008] Preferably, when the pipe type is ductile iron pipe, the characteristic frequency band f 1~ f 2 is 1kHz~2kHz.

[0009] Preferably, the calculation formula for the normalization process is as follows: in, S ( f ) is the characteristic frequency band f 1~ f 2 is the frequency domain signal after smoothing, G ( f ) is the characteristic spectrum curve of the loss.

[0010] Preferably, the calculation formula of the Pearson correlation coefficient PPC is as follows: Wherein, is the mean value of the characteristic spectrum curve of the missing signal; is the characteristic spectrum curve of the signal to be measured; M is the number of missing characteristic spectrum curve data points; x i and y i The missing signal and the signal to be measured are in the characteristic frequency band respectively f 1~ f 2The normalized value.

[0011] Preferably, the calculation formula of the mean square error RMES is as follows: in, M is the number of data points of the missing characteristic spectrum curve.

[0012] Preferably, the cross-correlation threshold and RMES The steps of dynamic setting of threshold are as follows: S1: transform the pipeline vibration time domain signal collected by interferometer into x (t) is converted into a frequency domain signal X ( f ); S2: Smoothing and normalizing the frequency domain signal to generate a standardized historical random loss curve and construct a historical random loss curve library; S3: Calculate the Pearson correlation coefficient PPC and the mean square error RMES between the historical random loss curve library and each curve in the loss characteristic curve library respectively; S4: Select the minimum value of the Pearson correlation coefficient PPC as the cross-correlation threshold, and select the maximum value of the mean square error RMES as RMES Threshold; S5: Repeat steps S1-S4, and dynamically update the cross-correlation threshold and the mean square error RMES calculated based on the newly added Pearson correlation coefficient PPC and the mean square error RMES calculated between the signal to be tested and each curve in the loss characteristic curve library. RMES Threshold.

[0013] Preferably, the cross-correlation threshold is 0.96, RMES The threshold is 0.1.

[0014] A pipeline leakage monitoring system includes a signal acquisition module, a spectrum conversion module, a feature extraction module and a dynamic discrimination module; the signal acquisition module is used to obtain the pipeline vibration time domain signal x (t); The spectrum conversion module is used to convert the pipeline vibration time domain signal x (t) is converted into a frequency domain signal X ( f ) and extract the characteristic frequency band of the missing signal f 1~ f2; The feature extraction module uses the smoothing filter and normalization process to construct the feature frequency band f 1~ f 2 within the missing characteristic curve library; the dynamic discrimination module is based on the cross-correlation threshold and RMES The dynamic setting of the threshold and the combined judgment of the leakage signal output the leakage detection result of the signal to be tested.

[0015] Due to the adoption of the above technical solution, the present invention has the following beneficial effects.

[0016] (1) The present invention locks high-frequency effective leakage signals and removes noise through characteristic frequency band extraction and smoothing filtering, providing stable input for subsequent analysis; eliminates amplitude difference interference through normalization processing, and combines the Pearson correlation coefficient (PPC) and mean square error (RMES) to simultaneously evaluate the shape and amplitude matching, thereby improving the reliability of leakage signal discrimination under different pipe types and working conditions; adjusts the characteristic frequency band and dynamic threshold setting according to the characteristics of different pipe types and working conditions, and realizes accurate detection of pipeline leakage signals in multiple scenarios.

[0017] (2) The present invention solves the technical problems that traditional time domain methods (such as kurtosis and energy mean) are easily interfered by low-frequency noise, and machine learning relies on a large amount of data and has poor effects under complex working conditions through frequency domain conversion and feature frequency band extraction, thereby achieving the technical effects of suppressing low-frequency environmental noise, reducing data requirements, and improving the robustness of discrimination under complex working conditions (such as valve water flow and pump room noise); through smoothing filtering and normalization processing, the present invention solves the technical problems that different acquisition conditions (tube type, working conditions) cause signal amplitude differences and affect feature consistency, thereby achieving the technical effects of eliminating amplitude interference, retaining spectrum shape characteristics, and constructing a standardized loss feature curve library.

[0018] (3) The present invention solves the technical effect that a single criterion is prone to misjudgment through the joint judgment of the Pearson correlation coefficient PPC and the mean square error RMES, and achieves the technical effect of comprehensively evaluating the curve shape similarity PPC and the amplitude proximity RMES, significantly improving the accuracy of leakage judgment; through dynamic threshold setting and pipe adaptive frequency band adjustment, the technical problem that the fixed threshold cannot adapt to different working conditions and pipe types is solved, and the characteristic frequency band is dynamically set, and the threshold is updated with historical data to adapt to the leakage judgment of different complex scenarios.

[0019] (4) The present invention extracts high-difference frequency bands far away from low-frequency noise as characteristic frequency bands based on the comparison of the spectrum of leakage and non-leakage signals, focusing on high-frequency effective features and avoiding low-frequency interference; eliminates random noise while maintaining the spectrum peak characteristics through Savitzky-Golay smoothing filtering, thereby improving the representativeness of the characteristic curve; eliminates the amplitude difference caused by environmental factors through normalization processing, so that the characteristic curves under different working conditions can be compared; performs dual verification of shape and amplitude through the Pearson correlation coefficient PPC and the mean square error RMES, thereby reducing the misjudgment rate of pipeline leakage; and adopts a dynamic threshold setting mechanism to adapt to different pipe types and working conditions to avoid the problem of sensitivity reduction caused by fixed thresholds.

[0020] (5) The present invention extracts the high-frequency characteristic frequency band of the pipeline vibration signal through frequency domain analysis, combines Savitzky-Golay conformal filtering to preserve the spectral shape, eliminates amplitude differences through normalization processing, and constructs a standardized leakage characteristic curve. At the same time, the Pearson coefficient (PPC) and mean square error (RMES) dual-indicator joint judgment is adopted to dynamically set the threshold, and the characteristic frequency band is adaptively adjusted in combination with the pipe type and working conditions. The threshold is updated by historical data to achieve the adaptation of leakage detection in multiple scenarios. The present invention effectively suppresses low-frequency noise interference, significantly improves the signal-to-noise ratio of the leakage signal and the leakage judgment accuracy, and the leakage detection accuracy of ductile iron pipes is ≥98%. It effectively solves the problems of complex noise interference, working condition differences and poor threshold adaptability, and has the advantages of high precision and low computing power. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The following detailed discussion discusses the making and use of preferred embodiments of the present invention. However, it should be understood that the present invention provides numerous applicable inventive concepts that can be embodied in a variety of specific contexts. The specific embodiments discussed are intended only to illustrate specific ways to make and use the present invention and are not intended to limit the scope of the invention. It would be readily apparent to one of ordinary skill in the art that other embodiments could be derived from these drawings without inventive effort.

[0022] Figure 1 The figure is a flow chart of the pipeline leakage monitoring method of the present invention.

[0023] Figure 2 These are the actual collected ductile iron DN150 pipe leakage signals, valve water flow signals and normal audio signals.

[0024] Figure 3 The figure shows the comparison between the characteristic curve of ductile iron DN150 and the characteristic curves of three types of events.

[0025] Figure 4 The calculation diagram of the Pearson correlation coefficient and mean square error results of ductile iron DN150.

[0026] Figure 5The calculation diagram of the Pearson correlation coefficient and mean square error results of ductile iron DN200.

[0027] Figure 6 The calculation diagram of the Pearson correlation coefficient and mean square error results of ductile iron DN100.

[0028] in, Figure 1 -a is the missing frequency domain signal diagram; Figure 1 -b is the smooth curve of the missing spectrum; Figure 1 -c is the spectrum curve of the missing characteristic frequency band; Figure 1 -d is the dropout characteristic spectrum curve. DETAILED DESCRIPTION

[0029] The following detailed discussion discusses the making and use of preferred embodiments of the present invention. However, it should be understood that the present invention provides many applicable inventive concepts that can be embodied in a variety of specific circumstances. The specific embodiments discussed are intended only to illustrate specific ways to make and use the present invention and are not intended to limit the scope of the present invention.

[0030] A pipeline leakage monitoring method comprises the following steps: S1: transforming the pipeline vibration time domain signal collected by the interferometer through Fourier transform x (t) is converted into a frequency domain signal X ( f ); the Fourier transform formula is as follows: in, X ( f ) is the frequency domain signal, x (t) is the time domain signal of pipeline vibration.

[0031] S2: By comparing and analyzing the spectrum amplitude distribution of the missing signal and the non-missing signal, the high-frequency band with significant amplitude difference and far away from the low-frequency noise is determined as the characteristic frequency band of the missing signal f 1~ f 2; The characteristic frequency band f 1~ f 2 is dynamically adjusted to 500Hz~3kHz according to the pipe type. In this embodiment, when the pipe type is ductile iron pipe, the characteristic frequency band f 1~ f 2 is 1kHz~2kHz.

[0032] S3: Perform smoothing filtering and normalization processing on the frequency domain signal within the characteristic frequency band to generate a standardized loss characteristic spectrum curve and construct a loss characteristic curve library. The calculation formula of the normalization processing is as follows: in, S (f ) is the characteristic frequency band f 1~ f 2 is the frequency domain signal after smoothing, G ( f ) is the characteristic spectrum curve of the loss.

[0033] S4: Calculate the Pearson correlation coefficient PPC and mean square error RMES between the signal to be measured and each curve in the dropout characteristic curve library; the calculation formula of the Pearson correlation coefficient PPC is as follows: Wherein, is the mean value of the characteristic spectrum curve of the missing signal; is the characteristic spectrum curve of the signal to be measured; M is the number of missing characteristic spectrum curve data points; x i and y i The missing signal and the signal to be measured are in the characteristic frequency band respectively f 1~ f 2The normalized value.

[0034] The calculation formula of the mean square error RMES is as follows: in, M is the number of data points of the missing characteristic spectrum curve.

[0035] S5: Dynamically set the cross-correlation threshold and the RMES The cross-correlation threshold and RMES The steps for dynamically setting the threshold are as follows: Step 1: Transform the pipeline vibration time domain signal collected by the interferometer into x (t) is converted into a frequency domain signal X ( f ).

[0036] Step 2: Smooth and filter the frequency domain signal and normalize it to generate a standardized historical random loss curve and build a historical random loss curve library.

[0037] Step 3: Calculate the Pearson correlation coefficient PPC and mean square error RMES between the historical random dropout curve library and each curve in the dropout characteristic curve library respectively.

[0038] Step 4: Select the minimum value of the Pearson correlation coefficient PPC as the cross-correlation threshold, and select the maximum value of the mean square error RMES as RMES Threshold.

[0039] Step 5: Repeat steps S1-S4, and dynamically update the cross-correlation threshold and the mean square error RMES calculated based on the newly added Pearson correlation coefficient PPC and the mean square error RMES calculated between the signal to be tested and each curve in the loss characteristic curve library. RMES In this embodiment, the cross-correlation threshold is 0.96. RMES The threshold is 0.1.

[0040] S6: Pearson correlation coefficient PPC ≥ cross-correlation threshold, and mean square error RMES ≤ RMES The thresholds are combined to determine whether the signal to be tested is a missing signal.

[0041] A pipeline leakage monitoring system includes a signal acquisition module, a spectrum conversion module, a feature extraction module and a dynamic discrimination module; the signal acquisition module is used to obtain the pipeline vibration time domain signal x (t); The spectrum conversion module is used to convert the pipeline vibration time domain signal x (t) is converted into a frequency domain signal X ( f ) and extract the characteristic frequency band of the missing signal f 1~ f 2; The feature extraction module uses the smoothing filter and normalization process to construct the feature frequency band f 1~ f 2 within the missing characteristic curve library; the dynamic discrimination module is based on the cross-correlation threshold and RMES The dynamic setting of the threshold and the combined judgment of the leakage signal output the leakage detection result of the signal to be tested.

[0042] The present invention locks in and removes noise from high-frequency effective leakage signals through characteristic frequency band extraction and smoothing filtering, providing stable input for subsequent analysis. Normalization eliminates amplitude difference interference, and combines the Pearson correlation coefficient (PPC) and mean square error (RMES) to simultaneously evaluate shape and amplitude matching, improving the reliability of leakage signal discrimination under different pipe types and operating conditions. The characteristic frequency band and dynamic threshold settings are adjusted according to the characteristics of different pipe types and operating conditions to achieve accurate detection of pipeline leakage signals in multiple scenarios.

[0043] The specific embodiment of the present invention relates to a pipeline leakage monitoring method and system based on spectrum characteristic curve extraction. The core of the system is to solve the low accuracy and high misjudgment rate problems of traditional leakage detection methods in complex noise environments through the synergistic effect of technical means such as frequency domain analysis, dynamic selection of characteristic frequency bands, shape-preserving smoothing filtering, normalization processing and dual-index joint judgment. Figure 1-6 Further elaborate.

[0044] like Figure 1The pipeline leakage monitoring method is shown in the flowchart. The pipeline vibration time domain signal is collected by a high-sensitivity acceleration sensor. The sensor is arranged on the outer wall of the pipeline and captures the acoustic vibration signal generated by the pipeline leakage in real time, that is, the pipeline vibration time domain signal. x (t). The signal acquisition module converts the time domain signal x (t) is transmitted to the spectrum conversion module, which converts the pipeline vibration time domain signal into the spectrum signal through Fourier transform. x (t) Convert to frequency domain signal X ( f ), generate the missing frequency domain signal diagram (such as Figure 1 -a).

[0045] For the missing signal, the low frequency band is greatly affected by the environmental noise, while the high frequency band has obvious characteristics that distinguish it from the non-missing signal. Therefore, by comparing and analyzing the spectrum amplitude distribution of the missing signal and the non-missing signal, we find the frequency band with the largest difference, and determine the high frequency band with significant amplitude difference and far away from the low frequency noise as the characteristic frequency band of the missing signal. f 1~ f 2. To distinguish between lost and non-lost signals. The characteristic frequency band must be determined by comparing lost signals under different operating conditions. Since each operating condition may have multiple characteristic curves, that is, multiple characteristic frequency bands, the frequency band is not fixed. In principle, it should be kept away from low-frequency ambient noise. The specific range should be determined by collecting a large amount of data for comparison. In this example, the determination range is given: for ductile iron pipes with a diameter of DN = 150mm, the characteristic frequency band is 1kHz-2kHz.

[0046] For the frequency domain signal within the characteristic frequency band 1kHz-2kHz X ( f ) is smoothed and normalized, and the Fourier transform formula is as follows: in, X ( f ) is the frequency domain signal, x (t) is the time domain signal of pipeline vibration.

[0047] Since the signal studied in this project is a broadband signal, it is planned to use broadband spectrum profile distribution as the characteristic signal representation. Therefore, Savitzky–Golay filtering is used to smooth the linear spectrum. This algorithm achieves shape-preserving smoothing through local polynomial fitting, which can suppress high-frequency noise and retain the spectrum peak characteristics, and obtain the loss spectrum smoothing curve, as shown in the following figure: Figure 1 -b. The spectrum data actually collected often contains low-frequency environmental noise, and its amplitude intensity is strong. The noise source is more diverse and random. Therefore, the high-frequency band curve is intercepted as the missing characteristic frequency band spectrum curve, such as Figure 1 -c as shown.

[0048] The calculation formula for the normalization process is as follows: in, S ( f ) is the characteristic frequency band f 1~ f 2 is the frequency domain signal after smoothing, G ( f ) is the leakage characteristic spectrum curve. Finally, the standardized leakage characteristic spectrum curve is generated, such as Figure 1 -d, and build a missing characteristic curve library.

[0049] The present invention divides a large number of historical random leakage events into leakage signals and non-loss signals, which include water flow noise and normal audio. The collected pipeline vibration time-domain signals undergo frequency domain conversion, feature frequency band extraction, smoothing filtering, and normalization to generate corresponding leakage characteristic curve libraries, normal curve libraries, and valve water flow curve libraries.

[0050] Figure 2 This chart shows a comparison of the frequency spectra of leakage and non-leakage signals in a DN150 ductile iron pipe within the characteristic frequency range of 1000-2000 Hz. For the DN150 ductile iron pipe, the horizontal axis represents the frequency range of 0-4 kHz, and the vertical axis represents the normalized amplitude. This comparative analysis reveals that the characteristic frequency range of leakage signals lies between 1 kHz and 2 kHz, while the frequency ranges below 1 kHz and between 2 kHz and 4 kHz represent the characteristic frequency ranges of non-leakage signals, which can be considered the background noise (low-frequency ambient noise) of the DN150 ductile iron pipe. The leakage signal exhibits a significant amplitude increase in the 1 kHz to 2 kHz frequency range. The valve flow curve library, on the other hand, is concentrated below 500 Hz, while the normal curve library is concentrated in the 500 Hz to 1 kHz and 2 kHz to 4 kHz frequency ranges. Through comparative analysis, the frequency band of 1kHz-2kHz, where the difference between the lost signal and the non-lost signal is the largest, is dynamically selected as the characteristic frequency band to avoid interference from low-frequency environmental noise and valve water flow. This step improves the signal-to-noise ratio of the lost signal by more than 30%.

[0051] Figure 3A comparison of the leakage characteristic spectrum curves of the actual collected ductile iron DN150 pipeline leakage signal and non-leakage signal is shown. The comparison chart shows that after Savitzky-Golay smoothing, the leakage signal's resonance peak shape is clear, laying the foundation for subsequent feature matching. After normalization, the leakage characteristic spectrum curve highly overlaps with the leakage characteristic spectrum curves in the 19 leakage curve libraries, with essentially the same trend. However, the non-leakage signal differs significantly from the leakage signal in this frequency band, with a low degree of overlap. This intuitively indicates that the leakage characteristic spectrum curve can represent the leakage event under this operating condition. Normalization eliminates amplitude differences caused by factors such as sensor installation distance and pipe acoustic attenuation, making the characteristic curves under different operating conditions comparable.

[0052] The Pearson correlation coefficient PPC and mean square error RMES are then used to calculate the Pearson correlation coefficient and mean square error of the leakage characteristic signal and other random events to quantify their similarity. The calculation formula of the Pearson correlation coefficient PPC is as follows: Wherein, is the mean value of the characteristic spectrum curve of the missing signal; is the characteristic spectrum curve of the signal to be measured; M is the number of missing characteristic spectrum curve data points; x i and y i The missing signal and the signal to be measured are in the characteristic frequency band respectively f 1~ f 2 normalized value. Since we have normalized the signal, the calculated correlation coefficient r Falls between [-1,1]. r The closer the value is to 1, the more similar the two curves are; if r If it is close to -1, it means that the two curves show a completely inverse negative correlation; r If it is close to 0, it means there is no obvious linear relationship between the two.

[0053] The calculation formula of the mean square error RMES is as follows: in, M is the number of missing data points in the characteristic spectrum curve. The mean square error (RMES) measures the closeness of the amplitudes of two curves. The Pearson correlation coefficient is more suitable for reflecting the similarity of the shapes of two curves. It is insensitive to offset or amplification and only reflects linear correlation. Therefore, this invention primarily uses a combination of the cross-correlation algorithm and the mean square error (RMES) algorithm as a similarity evaluation metric.

[0054] Figure 4-Figure 6The Pearson correlation coefficient PPC and mean square error RMES calculation results of the three groups of pipeline leakage events under test, namely ductile iron DN150, DN200 and DN100. The cross-correlation thresholds of the leakage signals corresponding to the working conditions of the three pipe diameters of DN150, DN200 and DN100 are 0.967, 0.97 and 0.96 respectively. At this time, the cross-correlation threshold is selected as 0.96 based on the historical random loss curve library; the cross-correlation threshold of the leakage signals corresponding to the working conditions of the three pipe diameters of DN150, DN200 and DN100 is 0.967, 0.97 and 0.96 respectively. RMES The thresholds are 0.1, 0.07 and 0.08 respectively. RMES The threshold is 0.1, that is, the signal to be tested is greater than 0.96 in the cross-correlation threshold, and RMES When the threshold is less than 0.1, it is judged as a pipeline leak. After that, the cross-correlation threshold and RMES Threshold. At this threshold setting, the minimum accuracy for detecting missing signals from the current historical random missing curve library is 98%. Data results show that the missing signal characteristic curve is relatively stable and significantly different from non-missing signals, indicating that this method can accurately detect missing signals.

[0055] The technical effectiveness of this invention has been verified through multi-dimensional experimental data. In terms of noise suppression, characteristic frequency band selection reduces low-frequency ambient noise energy by 90%, and Savitzky-Golay filtering further eliminates 30% of high-frequency random noise. In terms of discrimination accuracy, a dual-index combined criterion increases the accuracy to over 98% compared to traditional time-domain features (kurtosis and energy mean). In terms of engineering applicability, the pipe material adaptation mechanism and dynamic threshold update improve the system's stability under complex operating conditions by 40%, providing a highly effective solution for leak monitoring in smart water services, oil pipeline networks, and other fields.

[0056] The present invention converts the pipeline vibration time domain signal into a frequency domain signal through Fourier transform, compares the amplitude distribution of the leakage and non-leakage spectrum, and screens out the high-frequency characteristic frequency band. The characteristic frequency band of the ductile iron pipeline in this embodiment is 1kHz~2kHz, which avoids the interference of low-frequency environmental noise. Based on the characteristic frequency band of 1kHz~2kHz, Savitzky-Golay filtering smoothing and normalization are used to generate a standardized leakage characteristic curve library to eliminate the influence of amplitude fluctuations caused by detection distance and pipe material differences. The curve shape similarity is evaluated by the Pearson correlation coefficient PPC, and the amplitude matching is quantified by the mean square error RMES. The dual-index joint judgment makes up for the limitations of single-index leakage detection, such as avoiding misjudgment of similar shapes but amplitude offset. The adaptive threshold is dynamically set in combination with the historical random leakage curve library. The cross-correlation threshold of this embodiment is ≥0.96, RMESThe threshold is ≤0.1, and an incremental learning mechanism is used to update the threshold to adapt to different pipe materials (ductile iron, PVC, etc.) and changing operating conditions. This method integrates frequency domain feature extraction, signal standardization, dual-index collaboration, and dynamic threshold closed-loop optimization. Demonstrated in practical examples, it achieves 98% accuracy, an improvement of over 5% over traditional time-domain methods. It significantly suppresses interference from valve overflows and pump room noise, addressing low-frequency noise sensitivity and threshold rigidity. It combines high interference resistance, multi-scenario adaptability, and real-time processing capabilities, making it suitable for leak monitoring in complex water supply networks.

[0057] Although the specification has been described in detail, it should be understood that various changes, substitutions, and modifications may be made without departing from the spirit and scope of the invention as defined by the appended claims. In addition, the specific embodiments described are not intended to limit the scope of the invention, and a person of ordinary skill in the art can readily understand based on the present invention that currently existing or later developed processes, machines, manufactures, material compositions, means, methods, or steps may perform substantially the same functions or obtain substantially the same results as the embodiments of the present invention. Therefore, the appended claims are intended to include such processes, machines, manufactures, material compositions, means, methods, or steps within their scope.

Claims

1. A pipeline leakage monitoring method, characterized by: The following steps are involved: S1: The pipeline vibration time domain signal collected by the interferometer is converted into x (t) is converted into a frequency domain signal X ( f ); S2: By comparing and analyzing the spectrum amplitude distribution of the missing signal and the non-missing signal, the high-frequency band with significant amplitude difference and far away from the low-frequency noise is determined as the characteristic frequency band of the missing signal f 1~ f 2; S3: performing smoothing filtering and normalization processing on the frequency domain signal within the characteristic frequency band to generate a standardized loss characteristic spectrum curve and construct a loss characteristic curve library; S4: Calculate the Pearson correlation coefficient PPC and the mean square error RMES between the signal to be measured and each curve in the dropout characteristic curve library; S5: Dynamically set the cross-correlation threshold and the RMES Threshold; S6: Pearson correlation coefficient PPC ≥ cross-correlation threshold, and mean square error RMES ≤ RMES The thresholds are combined to determine whether the signal to be tested is a missing signal.

2. A pipeline leakage monitoring method according to claim 1, characterized in that: The formula for the Fourier transform is as follows: in, X ( f ) is the frequency domain signal, x (t) is the time domain signal of pipeline vibration.

3. A pipeline leakage monitoring method according to claim 1, characterized in that: The characteristic frequency band f 1~ f 2 Dynamically adjusted to 500Hz~3kHz according to the pipe type.

4. A pipeline leakage monitoring method according to claim 3, characterized in that: When the pipe type is ductile iron pipe, the characteristic frequency band f 1~ f 2 is 1kHz~2kHz.

5. A pipeline leakage monitoring method according to claim 1, characterized in that: The calculation formula for the normalization process is as follows: in, S ( f ) is the characteristic frequency band f 1~ f 2 is the frequency domain signal after smoothing, G ( f ) is the characteristic spectrum curve of the loss.

6. A pipeline leakage monitoring method according to claim 1, characterized in that: The calculation formula of the Pearson correlation coefficient PPC is as follows: Wherein, is the mean value of the characteristic spectrum curve of the missing signal; is the characteristic spectrum curve of the signal to be measured; M is the number of missing characteristic spectrum curve data points; x i and y i The missing signal and the signal to be measured are in the characteristic frequency band respectively f 1~ f 2The normalized value.

7. A pipeline leakage monitoring method according to claim 1, characterized in that: The calculation formula of the mean square error RMES is as follows: in, M is the number of data points of the missing characteristic spectrum curve.

8. The pipeline leakage monitoring method according to claim 1, characterized in that: The cross-correlation threshold and RMES The steps for dynamically setting the threshold are as follows: S1: The pipeline vibration time domain signal collected by the interferometer is converted into x (t) is converted into a frequency domain signal X ( f ); S2: Smoothing and normalizing the frequency domain signal to generate a standardized historical random loss curve and construct a historical random loss curve library; S3: Calculate the Pearson correlation coefficient PPC and the mean square error RMES between the historical random loss curve library and each curve in the loss characteristic curve library respectively; S4: Select the minimum value of the Pearson correlation coefficient PPC as the cross-correlation threshold, and select the maximum value of the mean square error RMES as RMES Threshold; S5: Repeat steps S1-S4, and dynamically update the cross-correlation threshold and the mean square error RMES calculated based on the newly added Pearson correlation coefficient PPC and the mean square error RMES calculated between the signal to be tested and each curve in the loss characteristic curve library. RMES Threshold.

9. A pipeline leakage monitoring method according to claim 8, characterized in that: The cross-correlation threshold is 0.96, RMES The threshold is 0.

1.

10. A pipeline leakage monitoring system, characterized by: It includes signal acquisition module, spectrum conversion module, feature extraction module and dynamic discrimination module; The signal acquisition module is used to obtain the pipeline vibration time domain signal x (t); The spectrum conversion module is used to convert the pipeline vibration time domain signal x (t) is converted into a frequency domain signal X ( f ) and extract the characteristic frequency band of the missing signal f 1~ f 2; The feature extraction module uses the smoothing filter and normalization process described in any one of claims 1 to 9 to construct a feature frequency band. f 1~ f 2. The missing characteristic curve library within 2; The dynamic discrimination module is based on any of the cross-correlation thresholds and RMES The dynamic setting of the threshold and the combined judgment of the leakage signal output the leakage detection result of the signal to be tested.

Images