A gas delivery valve internal leakage online diagnosis method based on voiceprint feature analysis

Abstract

The present application relates to the technical field of valve internal leakage detection, and particularly relates to a gas delivery valve internal leakage online diagnosis method based on voiceprint feature analysis. The method comprises the following steps: injecting tracer gas into a pipeline, collecting valve node voiceprint signals and pressure response signals and preprocessing; performing cluster analysis on feature data, dividing feature groups according to the similarity of voiceprint spectrum and pressure fluctuation characteristics, and marking group nodes with a distribution range exceeding a benchmark interval as suspected internal leakage areas; extracting suspected area signal evolution trajectories, and determining the risk type as a continuous attenuation type or an intermittent mutation type; matching secondary detection strategies according to the risk type, matching injection methods according to the degree of voiceprint signal attenuation along the pipeline for the continuous attenuation type, and matching injection methods according to the mutation frequency and duration for the intermittent mutation type; and performing secondary detection. The present application realizes accurate positioning of valve internal leakage by screening suspected areas through clustering and dividing risk types in combination with signal evolution, and then matching secondary detection strategies.

Description

Technical Field

[0001] This invention relates to the field of valve internal leakage detection technology, and in particular to an online diagnostic method for internal leakage in gas delivery valves based on acoustic signature analysis. Background Technology

[0002] As a key control component of pipeline systems, gas transmission valves directly affect the safety and efficiency of pipeline operation through their sealing performance. Internal leakage in valves refers to the phenomenon where the medium leaks from the sealing surface even when the valve is closed. Prolonged leakage can lead to media loss, environmental pollution, and even safety accidents. Therefore, timely and accurate detection and diagnosis of internal leakage in valves is of great importance. Currently, commonly used methods for detecting internal leakage in valves include the pressure drop method, flow balance method, ultrasonic detection method, and gas tracer method. The pressure drop method determines leakage by monitoring pressure changes upstream and downstream of the valve, but it is greatly affected by fluctuations in pipeline operating conditions, resulting in limited detection accuracy. The flow balance method determines the leakage amount by comparing the flow differences upstream and downstream, but this method cannot accurately locate the specific position of the leaking valve and requires extremely high accuracy from the flow measurement instruments. The ultrasonic detection method diagnoses leakage by receiving ultrasonic signals generated by the leak, but it is easily affected by external environmental noise and lacks sensitivity to minute leaks. The gas tracer method determines leakage by injecting tracer gas into the pipeline and detecting concentration changes downstream, but this method has a long detection cycle and is difficult to meet the needs of online diagnosis.

[0003] Chinese Patent Publication No. CN121782531A discloses a leak detection method and system based on acoustic signature feature extraction and pattern recognition. The method includes: setting a first acoustic signature sensor that can rotate circumferentially along the pipeline on the outer periphery of the pipeline, and setting a second acoustic signature sensor covering the pipeline within the detection range; synchronously collecting acoustic signature sequences using the first and second acoustic signature sensors, extracting acoustic signature features, determining the leak location along the pipeline path when a preset leak criterion is met, and issuing an alarm signal and triggering a leak scan; performing a leak scan on the pipeline corresponding to the first acoustic signature sensor, and outputting the leak location along the pipeline path and the leak location circumferentially along the pipeline as the leak detection result; by setting a circumferentially rotating acoustic signature sensor on the outer periphery of the pipeline, the location of the leak point can be accurately measured, effectively reducing the error in leak location.

[0004] Therefore, the existing technology has the following problems: Leak location is achieved solely by collecting signals from the circumferential rotation of the acoustic fingerprint sensor, which lacks a preliminary screening step for suspected areas and fails to match differentiated secondary detection strategies according to different types of leak risks, resulting in low detection efficiency. Summary of the Invention

[0005] To address this, the present invention provides an online diagnostic method for internal leakage of gas delivery valves based on acoustic signature analysis. This method overcomes the problems of existing technologies that rely solely on the circumferential rotation of acoustic signature sensors to collect signals for leak location, lack a preliminary screening step for suspected areas, and fail to match differentiated secondary detection strategies according to different leakage risk types, resulting in low detection efficiency.

[0006] To achieve the above objectives, this invention provides an online diagnostic method for internal leakage in gas delivery valves based on acoustic signature analysis, comprising: Step S1: Inject tracer gas into the gas delivery pipeline, collect acoustic signature signals and pressure response signals at each valve node of the gas delivery pipeline, and perform preprocessing to obtain feature data; Step S2: Perform cluster analysis on the feature data. Based on the similarity between the acoustic spectrum features and pressure fluctuation features in the feature data corresponding to each valve node, divide all valve nodes into several feature groups. Determine the feature data distribution range of all valve nodes in each feature group. Mark the valve nodes included in the feature group whose feature data distribution range exceeds the preset benchmark interval as suspected internal leakage areas. Step S3: For each identified suspected internal leakage region, extract the signal evolution trajectory of the feature group corresponding to the suspected internal leakage region on the time axis, and determine the risk type of the suspected internal leakage region based on the signal evolution trajectory. The risk type includes continuous decay type and intermittent mutation type. Step S4: Determine the corresponding secondary detection gas injection strategy according to the risk category; wherein, for the continuous attenuation type, match the corresponding secondary detection gas injection method according to the attenuation degree of the acoustic signal intensity along the transmission distance in the suspected internal leakage area; for the intermittent mutation type, match the corresponding secondary detection gas injection method according to the frequency characteristics and duration characteristics of the acoustic signal mutation event in the suspected internal leakage area. Step S5: Perform secondary detection according to the secondary detection gas injection method to obtain the detection result.

[0007] Furthermore, the cluster analysis in step S2 includes: Step S21: Using the acoustic spectrum features and pressure fluctuation features in the feature data corresponding to each valve node as the basis for clustering, determine the degree of similarity between the features corresponding to any two valve nodes. Step S22: Group valve nodes that meet the preset similarity requirements into the same feature group, traverse all valve nodes, and complete the division of all feature groups; Step S23: Determine the distribution range of feature data for all valve nodes in each feature group; Step S24: The valve nodes included in the feature group whose feature data distribution range exceeds the preset benchmark interval are marked as suspected internal leakage areas.

[0008] Further, in step S3, the risk type of the suspected endoleak region is determined based on the signal evolution trajectory, including: Obtain the acoustic signature amplitude corresponding to the signal evolution trajectory at each acquisition time; The amplitude of the acoustic signature signal is compared point by point along the time axis between adjacent acquisition times to determine the direction and magnitude of the amplitude change over time. In response to the fact that the amplitude change direction between each adjacent acquisition time is decreasing and the amplitude change magnitude between each adjacent acquisition time is less than the preset change magnitude threshold, the risk type of the suspected internal leakage area is determined to be the continuous decay type. In response to the existence of two or more acquisition moments along the time axis, where the amplitude change between the acquisition moment and the previous acquisition moment is greater than or equal to a preset amplitude change threshold, and the direction of amplitude change between the acquisition moment and the subsequent acquisition moment is opposite to the direction of change of the previous moment, the risk type of the suspected internal leakage region is determined to be intermittent mutation type.

[0009] Further, in step S4, for the continuously attenuating type, the corresponding secondary detection gas injection method is matched according to the degree of attenuation of the acoustic signal intensity along the transmission distance in the suspected internal leakage area, including: Obtain the relationship between the acoustic signal intensity at each valve node in the suspected internal leakage area and the position of each valve node along the gas delivery direction; Based on the aforementioned positional correspondence, the attenuation trend of the voiceprint signal intensity along the transmission distance is determined; The injection flow rate and injection pressure of the secondary detection gas are determined based on the attenuation trend.

[0010] Furthermore, after determining the injection flow rate and injection pressure of the secondary detection gas, the following steps are also included: The placement method of the acoustic fingerprint sensor in the suspected internal leakage area is determined based on the injection flow rate and injection pressure. In response to the injection flow rate being greater than the preset injection flow rate and the injection pressure being greater than the preset injection pressure, a sensor array is deployed along the gas delivery direction at a first density to synchronously collect acoustic fingerprint signals at each measuring point, and the leak point is located based on the location of the sudden drop in acoustic fingerprint signal intensity along the path. In response to the injection flow rate being less than or equal to a preset injection flow rate or the injection pressure being less than or equal to a preset injection pressure, sensors are deployed along the gas delivery direction at a second density to collect acoustic fingerprint signals segment by segment. The location of the leak segment is determined based on the differences in the segmented signals, wherein the second density is less than the first density.

[0011] Further, in step S4, for intermittent mutation types, the corresponding secondary detection gas injection method is matched based on the frequency and duration characteristics of the abrupt change in the acoustic signature signal within the suspected endoleak region, including: Acquire the amplitude data of the acoustic signature signal in the suspected internal leakage area at each acquisition time; Traverse the amplitude data along the time axis to identify the acquisition time when a sudden change event occurs. The sudden change event is defined as an amplitude change between the acquisition time and the previous acquisition time that is greater than or equal to a preset sudden change amplitude threshold, and the amplitude change between the acquisition time and the next acquisition time is opposite to the previous change direction. The cumulative number of times the mutation event occurs per unit time is determined as the mutation frequency; The duration from the start of the amplitude jump to the amplitude stabilization time in each mutation event is determined as the duration of each mutation. The injection method and sampling mode of the secondary detection gas are determined based on the frequency range in which the mutation frequency falls and the duration range in which the duration falls.

[0012] Furthermore, the injection method and sampling mode of the secondary detection gas were determined, including: In response to the mutation frequency being greater than or equal to the preset frequency threshold and the duration being less than the preset duration threshold, a high-voltage pulse injection mode and a continuous sampling mode are matched; In response to the mutation frequency being less than the preset frequency threshold and the duration being greater than or equal to the preset duration threshold, a low-pressure continuous injection mode and an interval sampling mode are matched.

[0013] Furthermore, after matching the high-voltage pulse injection method and the continuous sampling mode, it also includes: Simultaneously inject detection gas into both ends of the suspected internal leak area and apply alternating pressure pulses; During the application of alternating pressure pulses, acoustic fingerprint signals of each valve node are acquired in continuous sampling mode during the pulse pressurization and pressure holding periods. The acoustic fingerprint signals of each valve node during the pulse pressurization period are compared one by one with the acoustic fingerprint signals of the corresponding nodes during the pressure holding period; When the difference in acoustic signature between a valve node during pulse pressurization and pressure holding is greater than or equal to a preset difference threshold, the valve node is identified as a transient internal leak point.

[0014] Furthermore, after matching the low-pressure continuous injection method and the interval sampling mode, it also includes: Simultaneously inject detection gas into both ends of the suspected internal leakage area to maintain stable gas pressure within the area; The area is divided into several detection sections along the gas transport direction, and sensors are installed at the nodes of each detection section. Acoustic signals of each valve node are collected at preset time intervals within each detection section, and signal collection of all detection sections is completed in sequence. When abnormal fluctuations occur in the acoustic signature signal collected within a certain detection section, the valve node within that detection section is identified as a suspected internal leakage point.

[0015] Further, in step S5, the detection results are obtained, including: The acoustic fingerprint signals collected by each valve node in the secondary detection are compared one by one with the acoustic fingerprint signals collected by each valve node in the initial detection. When a valve node shows new abnormal components in the acoustic signature during a secondary detection compared to the initial detection, the valve node is marked as an internal leakage confirmed node. When the acoustic signature of a valve node does not show any new abnormal components in the secondary detection compared to the initial detection, the valve node is marked as a node without internal leakage.

[0016] Compared with the prior art, the beneficial effects of the present invention are as follows: This embodiment achieves the beneficial effects of accurate location and efficient online diagnosis of internal leaks in gas delivery valves by using the technical means of initial detection and clustering to screen suspected areas and combining signal evolution trajectory to classify risk types, and then matching differentiated secondary detection strategies for different risk types. This effectively improves the pertinence and efficiency of internal leak detection, reduces unnecessary comprehensive detection workload, and significantly improves the identification ability and diagnostic accuracy of different leak types by distinguishing between two different leak modes, namely continuous decay type and intermittent sudden change type and matching corresponding secondary detection schemes.

[0017] Furthermore, this embodiment combines acoustic signature spectral features with pressure fluctuation features to form a comprehensive feature for cluster analysis, which can comprehensively characterize the comprehensive response characteristics of each valve node in terms of both acoustics and pressure, avoiding one-sided judgments caused by single feature clustering. By comparing the distribution range of feature groups with the preset benchmark interval to identify suspected internal leakage areas, anomaly screening at the group level is achieved, effectively avoiding misjudgments caused by data fluctuations of individual nodes, and improving the accuracy and reliability of suspected internal leakage area identification.

[0018] Furthermore, this embodiment compares the amplitude of the acoustic signature signal at adjacent acquisition times point by point along the time axis to determine the direction and magnitude of the change. Based on the sequence characteristics of both, it distinguishes between two risk types: continuous attenuation and intermittent mutation, thus achieving accurate identification of the physical mechanism of leakage in suspected internal leakage areas. By setting a threshold for the magnitude of the change, it effectively distinguishes between normal signal fluctuations and abnormal mutations, avoiding misjudgments caused by signal fluctuations. Through the dual conditions of monotonically decreasing and stable decrease throughout the continuous attenuation type determination, and the requirement of pulse peak characteristics combined with at least two mutation points in the intermittent mutation type determination, it ensures the rigor of continuous leakage judgment and the effectiveness of eliminating occasional interference, thereby improving the accuracy and reliability of risk type classification.

[0019] Furthermore, this embodiment further matches the corresponding sensor deployment method after determining the injection flow rate and injection pressure, so that the sensor deployment density is adapted to the injection parameters, avoiding the problem of low detection efficiency or insufficient positioning accuracy caused by the mismatch between injection parameters and acquisition method; high-level injection parameters combined with high-density deployment of all nodes achieve accurate positioning, while low-level injection parameters combined with low-density deployment of some nodes achieve segmented screening, which ensures both positioning accuracy and detection efficiency.

[0020] Furthermore, this embodiment maintains stable pressure within the region by synchronously injecting from both ends, allowing the leakage channel to open and close naturally, ensuring that the detection results accurately reflect the valve's sealing performance under actual operating conditions. By dividing the detection area into segments and collecting signals segment by segment, a large suspected area is decomposed into several small detection units, reducing the complexity of a single investigation and avoiding data redundancy and resource consumption caused by comprehensive synchronous collection. By setting a preset time interval less than the average leakage duration, it ensures that at least one acquisition can capture a signal during each leakage opening period, effectively avoiding missed detections due to excessively large sampling intervals. Attached Figure Description

[0021] Figure 1 This is a flowchart of the online diagnostic method for internal leakage of gas delivery valves based on acoustic signature analysis in this embodiment; Figure 2 This is a flowchart illustrating the process of identifying suspected internal leakage areas in the online diagnostic method for internal leakage of gas delivery valves based on acoustic signature analysis in this embodiment. Figure 3 This is a flowchart illustrating the determination of the injection method and sampling mode of the secondary detection gas in the online diagnostic method for internal leakage of gas delivery valves based on acoustic signature analysis in this embodiment. Detailed Implementation

[0022] To make the objectives and advantages of the present invention clearer, the present invention will be further described below with reference to embodiments; it should be understood that the specific embodiments described herein are merely for explaining the present invention and are not intended to limit the present invention.

[0023] Preferred embodiments of the present invention will now be described with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are merely illustrative of the technical principles of the present invention and are not intended to limit the scope of protection of the present invention.

[0024] Please see Figure 1 As shown, it is a flowchart of the online diagnostic method for internal leakage of gas delivery valves based on acoustic signature analysis in this embodiment.

[0025] This embodiment provides an online diagnostic method for internal leakage of gas delivery valves based on acoustic signature analysis, including: Step S1: Inject tracer gas into the gas delivery pipeline, collect acoustic signature signals and pressure response signals at each valve node of the gas delivery pipeline, and perform preprocessing to obtain feature data; Step S2: Perform cluster analysis on the feature data. Based on the similarity between the acoustic spectrum features and pressure fluctuation features in the feature data corresponding to each valve node, divide all valve nodes into several feature groups. Determine the feature data distribution range of all valve nodes in each feature group. Mark the valve nodes included in the feature group whose feature data distribution range exceeds the preset benchmark interval as suspected internal leakage areas. Step S3: For each identified suspected internal leakage region, extract the signal evolution trajectory of the feature group corresponding to the suspected internal leakage region on the time axis, and determine the risk type of the suspected internal leakage region based on the signal evolution trajectory. The risk type includes continuous decay type and intermittent mutation type. Step S4: Determine the corresponding secondary detection gas injection strategy according to the risk category; wherein, for the continuous attenuation type, match the corresponding secondary detection gas injection method according to the attenuation degree of the acoustic signal intensity along the transmission distance in the suspected internal leakage area; for the intermittent mutation type, match the corresponding secondary detection gas injection method according to the frequency characteristics and duration characteristics of the acoustic signal mutation event in the suspected internal leakage area. Step S5: Perform secondary detection according to the secondary detection gas injection method to obtain the detection result.

[0026] This embodiment achieves the beneficial effects of accurate localization and efficient online diagnosis of internal leaks in gas delivery valves by using an initial detection and clustering method to screen suspected areas and combine it with signal evolution trajectories to classify risk types, and then matching differentiated secondary detection strategies for different risk types. This effectively improves the targeting and efficiency of internal leak detection, reduces unnecessary comprehensive detection workload, and significantly improves the ability to identify and diagnose different leak types by distinguishing between two different leak modes: continuous decay type and intermittent sudden change type and matching corresponding secondary detection schemes.

[0027] In this embodiment, a section of the gas delivery pipeline to be tested is first selected, on which several valve nodes are distributed. Before the test begins, a tracer gas source is connected to the inlet of the pipeline system. The tracer gas is selected based on its good tracer properties, and its concentration is preset according to the pipeline volume and the required detection sensitivity. The tracer gas is continuously injected into the pipeline, with the injection pressure maintained within the normal operating pressure range of the pipeline. The injection time is determined by ensuring that the tracer gas can fully diffuse and reach the valve nodes at the end of the pipeline.

[0028] During the tracer gas injection process, acoustic fingerprint sensors and pressure sensors are deployed at each valve node. The acoustic fingerprint sensors are piezoelectric acoustic emission sensors, which are fixed to the outer surface of the valve body or adjacent pipe wall by magnetic attraction or clamps, and are used to collect acoustic fingerprint signals generated by gas flow and potential leakage at the valve node. The pressure sensors are installed at the pressure measurement ports upstream and downstream of each valve node, and are used to synchronously collect the pressure response signals at each valve node, with the sampling frequency synchronized with that of the acoustic fingerprint sensors.

[0029] After data acquisition, the acquired acoustic signature and pressure response signals are preprocessed. First, the raw signals are filtered and denoised. The acoustic signature signal is filtered using a bandpass filter to remove non-target frequency bands, while an adaptive noise cancellation algorithm is used to remove mechanical vibrations from the pipeline and environmental background noise. The pressure response signal is filtered using a low-pass filter to remove high-frequency pressure pulsation interference. After denoising, the filtered acoustic signature and pressure response signals are normalized. A maximum-minimum normalization method is used to map the amplitudes of each signal to a uniform range, eliminating the influence of differences in the dimensions of different sensors. Finally, the normalized acoustic signature and pressure response signals of each valve node are aligned according to the sampling time and combined to generate standardized feature data for that node. The feature data of all valve nodes are then aggregated to form a standardized feature data set.

[0030] Please see Figure 2 As shown, this is a flowchart of the online diagnostic method for internal leakage of gas delivery valves based on acoustic signature analysis in this embodiment, which is used to identify suspected internal leakage areas.

[0031] Specifically, the cluster analysis in step S2 includes: Step S21: Using the acoustic spectrum features and pressure fluctuation features in the feature data corresponding to each valve node as the basis for clustering, determine the degree of similarity between the features corresponding to any two valve nodes. Step S22: Group valve nodes that meet the preset similarity requirements into the same feature group, traverse all valve nodes, and complete the division of all feature groups; Step S23: Determine the distribution range of feature data for all valve nodes in each feature group; Step S24: The valve nodes included in the feature group whose feature data distribution range exceeds the preset benchmark interval are marked as suspected internal leakage areas.

[0032] In this embodiment, the acoustic signature spectrum features and pressure fluctuation features in the feature data corresponding to each valve node are used as the clustering basis to determine the similarity between the features corresponding to any two valve nodes. Specifically, for each valve node, its acoustic signature spectrum features and pressure fluctuation features are combined to form a comprehensive feature, which comprehensively characterizes the node's response characteristics in both acoustic and pressure aspects. For any two valve nodes, the closeness between their comprehensive features is compared. The closer the comprehensive features are, the more similar the acoustic signature-pressure response patterns of the two nodes are, and vice versa.

[0033] Valve nodes whose similarity meets a preset requirement are grouped into the same feature group. The preset requirement is that the proximity between comprehensive features is less than a preset clustering threshold, which is determined using the silhouette coefficient evaluation method. The specific determination process is as follows: A series of candidate clustering thresholds are preset, and all valve nodes are pre-clustered using each candidate threshold to obtain the pre-clustering results corresponding to each candidate threshold; For each pre-clustering result, the average proximity of each valve node to other nodes in its group is calculated as the intra-group cohesion, and the average proximity of the node to other nodes in the group is calculated as the inter-group separation. The difference between the inter-group separation and the intra-group cohesion is divided by the larger of the two to obtain the silhouette coefficient of the node. The closer the coefficient is to the maximum value, the more ideal the clustering effect; The average silhouette coefficient of all nodes under each candidate threshold is taken as the average silhouette coefficient corresponding to that candidate threshold; The average silhouette coefficients corresponding to each candidate threshold are compared, and the candidate threshold corresponding to the highest average silhouette coefficient is selected as the preset clustering threshold.

[0034] After determining the preset clustering threshold, all valve nodes are traversed. For a node that is not currently grouped, the process expands outward from it, grouping all nodes whose comprehensive characteristics are less than the preset clustering threshold into the same group, until no new node meets the grouping criteria. This process is repeated for the remaining ungrouped nodes until all valve nodes are grouped. Thus, valve nodes with similar acoustic-pressure response patterns are grouped into the same feature group, while nodes with significantly different response patterns are grouped into different groups. The feature data distribution range of all valve nodes in each feature group is then determined. For each completed feature group, the comprehensive characteristics of all valve nodes within that group are obtained in each dimension. The maximum and minimum values ​​in each dimension are determined, and the range defined by the maximum and minimum values ​​in each dimension is the feature data distribution range of that feature group.

[0035] Valve nodes included in feature groups whose feature data distribution range exceeds a preset baseline interval are identified as suspected internal leakage areas. The preset baseline interval is determined as follows: First, using the comprehensive features of all valve nodes as samples, the center position and allowable deviation of the normal value range for each feature dimension are determined. The interval defined by the center position plus or minus the allowable deviation is taken as the baseline sub-interval for that feature dimension. The allowable deviation is determined according to the actual detection sensitivity requirements. The smaller the allowable deviation, the higher the detection sensitivity, but the false alarm rate increases accordingly. The larger the allowable deviation, the higher the detection specificity, but minor anomalies may be missed. The baseline sub-intervals of each feature dimension together constitute the preset baseline interval. For each feature group, compare the sub-intervals defined by the maximum and minimum values ​​of the group on each feature dimension with the corresponding baseline sub-interval. If, on any feature dimension, the maximum value of the group is greater than the upper limit of the baseline sub-interval, or the minimum value of the group is less than the lower limit of the baseline sub-interval, then the feature data distribution range of the feature group is determined to exceed the preset baseline interval. If the sub-intervals of the group on all feature dimensions fall completely within the corresponding baseline sub-interval, then the feature data distribution range of the feature group is determined to not exceed the preset baseline interval. For feature groups determined to exceed the preset baseline interval, all valve nodes included in the group are marked as suspected internal leakage areas. For feature groups determined not to exceed the preset baseline interval, the acoustic-pressure response of the valve nodes within the group is determined to be normal, and they are not included in the suspected internal leakage areas.

[0036] This embodiment combines acoustic signature spectral features with pressure fluctuation features to form a comprehensive feature for cluster analysis, which can comprehensively characterize the comprehensive response characteristics of each valve node in terms of both acoustics and pressure, avoiding one-sided judgments caused by single feature clustering. By comparing the distribution range of feature groups with the preset benchmark interval to identify suspected internal leakage areas, anomaly screening at the group level is achieved, effectively avoiding misjudgments caused by data fluctuations of individual nodes, and improving the accuracy and reliability of suspected internal leakage area identification.

[0037] Specifically, in step S3, determining the risk type of the suspected endoleak region based on the signal evolution trajectory includes: Obtain the acoustic signature amplitude corresponding to the signal evolution trajectory at each acquisition time; The amplitude of the acoustic signature signal is compared point by point along the time axis between adjacent acquisition times to determine the direction and magnitude of the amplitude change over time. In response to the fact that the amplitude change direction between each adjacent acquisition time is decreasing and the amplitude change magnitude between each adjacent acquisition time is less than the preset change magnitude threshold, the risk type of the suspected internal leakage area is determined to be the continuous decay type. In response to the existence of two or more acquisition moments along the time axis, where the amplitude change between the acquisition moment and the previous acquisition moment is greater than or equal to a preset amplitude change threshold, and the direction of amplitude change between the acquisition moment and the subsequent acquisition moment is opposite to the direction of change of the previous moment, the risk type of the suspected internal leakage region is determined to be intermittent mutation type.

[0038] In this embodiment, firstly, the acoustic signature signal amplitude of the feature group corresponding to the suspected internal leak area is acquired at each acquisition time during the initial detection period. For each feature group identified as a suspected internal leak area, the acoustic signature signal amplitude records of each valve node within the group are extracted from the feature data set along the entire acquisition time axis, forming the signal evolution trajectory of the suspected internal leak area. This trajectory completely records all information on the change in acoustic signature signal intensity of each valve node in the area as the acquisition time progresses from the start of tracer gas injection to the end of acquisition. Then, the acoustic signature signal amplitudes of adjacent acquisition times are compared point by point along the time axis to determine the direction and magnitude of amplitude change over time. Specifically, for each acquisition time in the signal evolution trajectory, its amplitude is compared with the previous acquisition time and with the next acquisition time to determine whether the amplitude increases or decreases within the adjacent time period; at the same time, the absolute magnitude of the amplitude difference between adjacent acquisition times is calculated as the magnitude of change within that time period. By comparing point by point, a complete sequence of amplitude change direction and magnitude along the entire time axis can be obtained.

[0039] Subsequently, based on the sequence characteristics of the aforementioned change direction and amplitude, the risk type of the suspected internal leakage area is determined. The change direction between all adjacent moments throughout the entire acquisition period is checked. If all changes involve a decrease in amplitude, with no instances of amplitude increase, it indicates that the acoustic signature signal exhibits a monotonically decreasing trend on the time axis. Furthermore, the change amplitude between adjacent moments is examined. If all changes are less than a preset amplitude threshold, meaning there are no instances of abnormally large amplitude decreases during the entire acquisition process, it indicates that the acoustic signature signal is in a uniform and stable decreasing state, without any sudden fluctuations. When both conditions are met, the risk type of the suspected internal leakage area is determined to be a continuously decaying type. The continuously decaying type corresponds to a physical situation where the leakage channel exists continuously and stably, with the tracer gas continuously escaping through a fixed path, and the acoustic signature signal energy continuously and steadily decaying accordingly.

[0040] The direction and amplitude of changes between all adjacent moments within the entire acquisition period are scanned point by point to identify specific acquisition moments that simultaneously possess the following two characteristics: First, the amplitude change between this moment and the previous acquisition moment reaches or exceeds a preset amplitude threshold, indicating an abnormally large amplitude change at this moment; second, the direction of change between this moment and the previous acquisition moment is opposite to the direction of change between this moment and the next acquisition moment. The physical meaning of these two characteristics occurring simultaneously is that the signal first experiences a large surge or drop at a certain moment, and then undergoes a large change of the same magnitude in the opposite direction at the next moment, resulting in a steep pulse peak in the overall signal pattern. When two or more acquisition moments on the entire timeline exhibit the above pulse peak characteristics, it indicates that the acoustic signature signal has undergone multiple sudden jumps against a stable background. In this case, the risk type of the suspected internal leakage area is determined to be intermittent abrupt change. Intermittent abrupt change corresponds to the physical situation where the leakage channel intermittently opens and closes. The leakage does not occur continuously but suddenly opens and quickly recovers under specific triggering conditions, resulting in a pulse-like abrupt change pattern in the acoustic signature signal.

[0041] The preset amplitude threshold is used to distinguish between normal signal fluctuations and abnormal changes that require attention. It is determined by statistically analyzing the distribution of amplitude changes in the acoustic signature signal between adjacent acquisition times during the initial detection period for all valve nodes, taking the median level of all amplitude changes as a benchmark, and then adaptively adjusting this benchmark according to the actual detection sensitivity requirements. The determination of a continuous attenuation type requires that the amplitude change over the entire time axis be less than this threshold, reflecting the stable and continuous nature of the leak; the determination of an intermittent abrupt change type requires the existence of a jump exceeding this threshold, reflecting the sudden nature of the leak.

[0042] This embodiment compares the amplitude of the acoustic signature signal at adjacent acquisition times point by point along the time axis to determine the direction and magnitude of the change. Then, based on the sequence characteristics of the two, it distinguishes between two risk types: continuous attenuation and intermittent mutation, thus achieving accurate identification of the physical mechanism of leakage in suspected internal leakage areas. By setting a threshold for the change amplitude, it effectively distinguishes between normal signal fluctuations and abnormal mutations, avoiding misjudgments caused by signal fluctuations. Through the dual conditions of monotonically decreasing and stable decrease throughout the continuous attenuation type determination, and the requirement of pulse peak characteristics combined with at least two mutation points in the intermittent mutation type determination, it ensures the rigor of continuous leakage judgment and the effectiveness of occasional interference elimination, thereby improving the accuracy and reliability of risk type classification.

[0043] Specifically, in step S4, for the continuously attenuating type, the corresponding secondary detection gas injection method is matched according to the degree of attenuation of the acoustic signal intensity along the transmission distance in the suspected internal leakage area, including: Obtain the relationship between the acoustic signal intensity at each valve node in the suspected internal leakage area and the position of each valve node along the gas delivery direction; Based on the aforementioned positional correspondence, the attenuation trend of the voiceprint signal intensity along the transmission distance is determined; The injection flow rate and injection pressure of the secondary detection gas are determined based on the attenuation trend.

[0044] In this embodiment, firstly, the correlation between the acoustic signal intensity at each valve node within the suspected internal leak area and the position of each valve node along the gas delivery direction is obtained. Specifically, the acoustic signal intensity values ​​of each valve node within the area and their position information in the pipeline are extracted from the initial detection data, and these two values ​​are mapped one-to-one to form a distribution map of the acoustic signal intensity along the delivery distance. Then, based on this distribution map, the change in acoustic signal intensity with position is observed from upstream to downstream along the gas delivery direction to determine the rate of decrease. The rate of decrease reflects the severity of the leak: a more rapid decrease indicates a larger leak; a gentler decrease indicates a smaller leak.

[0045] Finally, the injection flow rate and pressure of the secondary detection gas are determined based on the rate of descent. A rapid descent indicates that the acoustic signal attenuates rapidly between the upstream and downstream ends, resulting in a large escape of the tracer gas over a short distance, indicating a significant leak. The signal characteristics near the leak point are more prominent and easily identifiable from the signal distribution along the leak path. In this case, the injection flow rate and pressure are set to the lower end of the preset standard injection parameter range for the suspected internal leak area to avoid excessive injection causing turbulence and signal distortion near the leak point, thus ensuring positioning accuracy.

[0046] When the decrease is relatively gradual, it indicates that the acoustic signal does not change significantly along the pipeline, the amount of tracer gas escaping is small, the leakage is minor, and the signal characteristics at the leak point are submerged in the normal background signal, making direct identification difficult. In this case, the injection flow rate and injection pressure are set to the high end of the preset standard injection parameter range for the suspected internal leak area. By increasing the injection intensity of the tracer gas, the amount of gas escaping at the leak point is relatively increased, amplifying the acoustic signal response at the leak point and making the leak location stand out in the signal distribution along the pipeline, thus meeting the sensitivity requirements for detecting minute leaks. The standard injection parameter range is a range of acceptable injection flow rates and injection pressures preset based on the pipe diameter, normal operating pressure, and properties of the transported medium. The low and high levels represent the lower and upper halves of this range, respectively. This completes the secondary detection injection parameter matching for continuously attenuating leaks.

[0047] Specifically, after determining the injection flow rate and injection pressure of the secondary detection gas, the process also includes: The placement method of the acoustic fingerprint sensor in the suspected internal leakage area is determined based on the injection flow rate and injection pressure. In response to the injection flow rate being greater than the preset injection flow rate and the injection pressure being greater than the preset injection pressure, a sensor array is deployed along the gas delivery direction at a first density to synchronously collect acoustic fingerprint signals at each measuring point, and the leak point is located based on the location of the sudden drop in acoustic fingerprint signal intensity along the path. In response to the injection flow rate being less than or equal to a preset injection flow rate or the injection pressure being less than or equal to a preset injection pressure, sensors are deployed along the gas delivery direction at a second density to collect acoustic fingerprint signals segment by segment. The location of the leak segment is determined based on the differences in the segmented signals, wherein the second density is less than the first density.

[0048] In this embodiment, after determining the injection flow rate and injection pressure, a corresponding sensor deployment method is further matched according to the determined injection parameters. The determination of the first density and the second density is based on the length of the suspected internal leakage area and the total number of valve nodes in the area. The first density corresponds to the deployment method under high-level injection parameters, specifically: sensors are deployed at each valve node in the suspected internal leakage area to form a sensor array covering all valve nodes, and the spacing between adjacent deployment points is the actual spacing between adjacent valve nodes in the area. The second density corresponds to the deployment method under low-level injection parameters, specifically: sensors are deployed at selected valve nodes along the gas delivery direction in the suspected internal leakage area, with several valve nodes spaced apart between adjacent deployment points, and the spacing between adjacent deployment points is greater than the actual spacing between adjacent valve nodes.

[0049] When high-level injection parameters are used in conjunction with the first-density deployment method, the tracer gas flows faster and diffuses more widely within the pipeline, resulting in a longer signal propagation distance. High-density synchronous acquisition of all valve nodes as measurement points allows for the capture of subtle signal changes along the pipeline. Direct comparison of signal strength at each measurement point identifies the leak point where the signal strength drops abruptly along the transport direction. When low-level injection parameters are used in conjunction with the second-density deployment method, the tracer gas flows more gently, resulting in a relatively gradual signal change and a larger attenuation range. Only a subset of valve nodes needs to be selected as measurement points. The suspected internal leak area is divided into several detection segments along the transport direction, with the endpoints of each segment serving as the valve nodes for sensor placement. Signals are acquired segment by segment, and by comparing the signal differences between the endpoints, the segment with the most significant signal attenuation is identified, and this segment is then determined as the leak location.

[0050] This embodiment further matches the corresponding sensor deployment method after determining the injection flow rate and injection pressure, so that the sensor deployment density is adapted to the injection parameters, avoiding the problem of low detection efficiency or insufficient positioning accuracy caused by the mismatch between injection parameters and acquisition method; high-level injection parameters combined with high-density deployment of all nodes achieve accurate positioning, while low-level injection parameters combined with low-density deployment of some nodes achieve segmented screening, which ensures both positioning accuracy and detection efficiency.

[0051] Specifically, in step S4, for intermittent mutation types, the corresponding secondary detection gas injection method is matched based on the frequency and duration characteristics of the acoustic signature signal mutation events in the suspected endoleak area, including: Acquire the amplitude data of the acoustic signature signal in the suspected internal leakage area at each acquisition time; Traverse the amplitude data along the time axis to identify the acquisition time when a sudden change event occurs. The sudden change event is defined as an amplitude change between the acquisition time and the previous acquisition time that is greater than or equal to a preset sudden change amplitude threshold, and the amplitude change between the acquisition time and the next acquisition time is opposite to the previous change direction. The cumulative number of times the mutation event occurs per unit time is determined as the mutation frequency; The duration from the start of the amplitude jump to the amplitude stabilization time in each mutation event is determined as the duration of each mutation. The injection method and sampling mode of the secondary detection gas are determined based on the frequency range in which the mutation frequency falls and the duration range in which the duration falls.

[0052] In this embodiment, for suspected internal leak areas identified as intermittently abruptly changing, the corresponding secondary detection gas injection method and sampling mode are matched based on the frequency and duration characteristics of the abrupt events. Intermittent abrupt changes correspond to the physical condition of the leak channel opening and closing intermittently; the acoustic signature signal exhibits a pulse-like abrupt change pattern. The frequency of the abrupt changes and the duration of each abrupt change reflect the regularity of the leak triggering and the response characteristics of the leak channel.

[0053] First, the amplitude data of the acoustic signature signal in the suspected internal leakage area is acquired at each acquisition time, that is, the complete amplitude record of this area over the entire acquisition time axis is extracted from the data acquired during the initial detection. Then, all amplitude data is traversed along the time axis, and the acquisition time at which abrupt events occur is identified point by point. The criteria for determining an abrupt event are: the amplitude change between the acquisition time and the previous acquisition time reaches or exceeds a preset abrupt amplitude threshold, and the direction of amplitude change between the acquisition time and the subsequent acquisition time is opposite to the direction of change of the previous time. When both conditions are met simultaneously, it indicates that the signal has undergone a significant jump with a reversed direction at that moment, forming a pulse spike, and this moment is identified as the occurrence of an abrupt event. Next, the cumulative number of abrupt events occurring per unit time is counted as the abrupt frequency of the suspected internal leakage area. Simultaneously, the duration of each identified abrupt event is determined, that is, the time elapsed from the initial moment when the amplitude begins to rise in the abrupt event to the moment when the amplitude returns to a stable state.

[0054] Finally, the corresponding secondary detection gas injection method and sampling mode are matched according to the above-mentioned mutation frequency and duration. The mutation frequency reflects the frequency of leak triggering, and the duration reflects the length of time the leak channel remains open after each opening. Different combinations of the two correspond to different leak behavior modes, requiring different secondary detection schemes. The specific matching method is as follows: when the mutation frequency is high and the duration is short, it indicates that the leak occurs in the form of frequent and brief pulses. In this case, a high-pressure pulse injection method is matched with a continuous sampling mode. The high-frequency pulse injection excites the leak point to generate a response, while continuous sampling is used to capture all mutation signals completely. When the mutation frequency is low and the duration is long, it indicates that the leak occurs in the form of occasional but persistent openness. In this case, a low-pressure continuous injection method is matched with an interval sampling mode. Stable injection maintains the stability of the pipeline state, and interval sampling can meet the detection requirements. Other combinations are matched with a medium-pressure alternating injection method and a segmented sampling mode.

[0055] Please see Figure 3 As shown, it is a flowchart for determining the injection method and sampling mode of the secondary detection gas in the online diagnosis method for internal leakage of gas delivery valve based on acoustic feature analysis in this embodiment.

[0056] Specifically, the injection method and sampling mode of the secondary detection gas are determined, including: In response to the mutation frequency being greater than or equal to the preset frequency threshold and the duration being less than the preset duration threshold, a high-voltage pulse injection mode and a continuous sampling mode are matched; In response to the mutation frequency being less than the preset frequency threshold and the duration being greater than or equal to the preset duration threshold, a low-pressure continuous injection mode and an interval sampling mode are matched.

[0057] In this embodiment, both the preset frequency threshold and the preset duration threshold are determined statistically based on historical acoustic signature signals collected during normal operation of the gas transmission pipeline under known leak-free conditions. Specifically, during normal pipeline operation, a period without any leak records is selected, and acoustic signature signals from each valve node are collected. The number of abrupt events occurring per unit time and the duration of each abrupt event are statistically analyzed within this period. The average level of the abrupt event frequency across all statistical periods is taken as a benchmark value, and the average level of the abrupt event duration is taken as a benchmark value. Then, the two benchmark values ​​are amplified or reduced according to the actual detection sensitivity requirements to obtain the preset frequency threshold and the preset duration threshold.

[0058] When the frequency of sudden changes is greater than or equal to a preset frequency threshold, and the duration is less than a preset duration threshold, the high-pressure pulse injection mode and continuous sampling mode are matched. The high-pressure pulse injection mode involves injecting detection gas into the suspected internal leak area with intermittent high-pressure gas pulses. Each pulse lasts for a certain period and then stops, alternating between pulses. The injection pressure is higher than the normal operating pressure of the pipeline. The continuous sampling mode continuously collects acoustic signature signals from each valve node throughout the entire high-pressure pulse injection process. When the frequency of sudden changes is less than a preset frequency threshold, and the duration is greater than or equal to a preset duration threshold, the low-pressure continuous injection mode and interval sampling mode are matched. The low-pressure continuous injection mode involves continuously injecting detection gas into the suspected internal leak area with a stable low-pressure airflow. The injection pressure is lower than the normal operating pressure of the pipeline. The interval sampling mode collects acoustic signature signals from each valve node at regular time intervals. The collection interval is set according to the duration of this type of leak, ensuring that at least one acquisition can capture a signal during each leak opening period. The specific values ​​of the injection pressure and sampling interval are preset according to the actual operating conditions of the suspected internal leak area and will not be elaborated here.

[0059] Specifically, after matching the high-voltage pulse injection method and the continuous sampling mode, it also includes: Simultaneously inject detection gas into both ends of the suspected internal leak area and apply alternating pressure pulses; During the application of alternating pressure pulses, acoustic fingerprint signals of each valve node are acquired in continuous sampling mode during the pulse pressurization and pressure holding periods. The acoustic fingerprint signals of each valve node during the pulse pressurization period are compared one by one with the acoustic fingerprint signals of the corresponding nodes during the pressure holding period; When the difference in acoustic signature between a valve node during pulse pressurization and pressure holding is greater than or equal to a preset difference threshold, the valve node is identified as a transient internal leak point.

[0060] In this embodiment, detection gas is simultaneously injected into both ends of the suspected internal leak area, and alternating pressure pulses are applied. The alternating pressure pulse is a pressure waveform that alternates between pressurization and pressure holding. During pressurization, high-pressure gas is injected into the pipeline to rapidly increase the pressure, and during pressure holding, the injection is stopped to maintain a stable pressure. This alternating pressurization and pressure holding cycle forms a periodic pressure alternation. Simultaneous injection at both ends allows the alternating pressure pulse to form a symmetrically distributed pressure disturbance within the area, avoiding pressure gradient interference with the leak point signal caused by single-end injection. Throughout the entire process of applying the alternating pressure pulse, acoustic fingerprint signals of each valve node within the area are collected in continuous sampling mode during the pulse pressurization and pressure holding periods, and the acoustic fingerprint signal data of each node during the pressurization and pressure holding phases are recorded separately.

[0061] After data collection, the acoustic fingerprint signals of each valve node during the pulse pressurization period were compared one by one with those during the pressure holding period. For valve nodes without leakage, the acoustic fingerprint signals during pressurization and pressure holding periods may differ slightly due to different pressure conditions, but the difference is within the normal fluctuation range. However, for valve nodes with transient leakage, the high-pressure gas impacts the leakage channel during pressurization, causing it to open passively and generating a significant leakage acoustic fingerprint signal; after the pressure stabilizes during pressure holding, the leakage channel closes immediately, and the acoustic fingerprint signal returns to normal levels. Therefore, the difference in acoustic fingerprint signals between the pressurization and pressure holding periods for this node is significantly larger.

[0062] When the difference in acoustic signature between a valve node during pulse pressurization and pressure holding reaches or exceeds a preset difference threshold, that valve node is identified as a transient internal leak. The preset difference threshold is determined as follows: Under normal operating conditions with no known leakage, acoustic signatures of each valve node are collected before and after pressure fluctuations. The signal difference of each node before and after pressure change is calculated, and the average difference of all nodes is taken as the preset difference threshold to distinguish between normal pressure fluctuation response and abnormal signal difference caused by leakage.

[0063] Specifically, after matching the low-pressure continuous injection method and the interval sampling mode, it also includes: Simultaneously inject detection gas into both ends of the suspected internal leakage area to maintain stable gas pressure within the area; The area is divided into several detection sections along the gas transport direction, and sensors are installed at the nodes of each detection section. Acoustic signals of each valve node are collected at preset time intervals within each detection section, and signal collection of all detection sections is completed in sequence. When abnormal fluctuations occur in the acoustic signature signal collected within a certain detection section, the valve node within that detection section is identified as a suspected internal leakage point.

[0064] In this embodiment, firstly, detection gas is simultaneously injected into both ends of the suspected internal leak area to maintain stable gas pressure within the area. Simultaneous injection from both ends ensures uniform distribution of the detection gas within the area, avoiding interference with the detection results caused by uneven pressure gradients and concentration distribution due to single-end injection. Maintaining stable pressure ensures that the pipeline operating conditions do not experience additional fluctuations, allowing the leak channel to open and close naturally, accurately reflecting the sealing performance of the valve in actual operation. Several detection segments are divided along the gas delivery direction within this area, each covering a portion of the valve nodes. Adjacent detection segments are connected end-to-end, collectively covering the entire suspected internal leak area. The number of detection segments is determined based on the total number of valve nodes included in the suspected internal leak area, with each segment containing at least two valve nodes to ensure sufficient signal reference points within the segment for determining whether abnormal fluctuations occur. Acoustic sensors are deployed at the endpoints of each detection segment.

[0065] After deployment, signal acquisition is performed sequentially from upstream to downstream in each detection section. Within each detection section, acoustic signals from each valve node are acquired at preset time intervals. These preset time intervals are less than the average duration of this type of leak to ensure that at least one acquisition captures a signal during each leak opening period. After completing the acquisition for the current detection section, the process is repeated in the next detection section until signal acquisition is completed for all detection sections.

[0066] After data collection, the acoustic signature signals collected from each detection segment are analyzed to determine if any abnormal fluctuations exist. Abnormal fluctuations refer to significant deviations in the acoustic signature signal amplitude from the normal level within the detection segment. This variation reflects abnormal acoustic signature response caused by leakage of the detected gas as it flows through the segment. The normal level is determined as follows: based on the distribution of acoustic signature signal amplitudes collected under normal operating conditions with no known leaks, the central tendency of all amplitudes is taken as a baseline value. This baseline value is then amplified and adjusted according to the actual detection sensitivity requirements. The adjusted value is used as the boundary between normal and abnormal. If the acoustic signature signal amplitude exceeds this boundary value, it is considered an abnormal fluctuation. When the acoustic signature signal collected in a detection segment exhibits the aforementioned abnormal fluctuations, the valve nodes included in that detection segment are identified as potential internal leak points. If the acoustic signature signal in a detection segment is normal and without abnormal fluctuations, the possibility of leakage in that segment is ruled out, and the next detection segment is checked. If any detection segment is identified as a potential internal leak point during the segmented investigation process, the location of that segment and the valve node information it contains are recorded.

[0067] This embodiment maintains stable pressure within the region by synchronously injecting from both ends, allowing the leakage channel to open and close naturally, ensuring that the detection results accurately reflect the valve's sealing performance under actual operating conditions. By dividing the detection area into segments and collecting signals segment by segment, a large suspected area is decomposed into several small detection units, reducing the complexity of a single investigation and avoiding data redundancy and resource consumption caused by comprehensive synchronous collection. By setting the preset time interval to be less than the average level of the leakage duration, it is ensured that at least one acquisition can capture a signal during each leakage opening period, effectively avoiding missed detections caused by excessively large sampling intervals.

[0068] Specifically, in step S5, the detection result is obtained, including: The acoustic fingerprint signals collected by each valve node in the secondary detection are compared one by one with the acoustic fingerprint signals collected by each valve node in the initial detection. When a valve node shows new abnormal components in the acoustic signature during a secondary detection compared to the initial detection, the valve node is marked as an internal leakage confirmed node. When the acoustic signature of a valve node does not show any new abnormal components in the secondary detection compared to the initial detection, the valve node is marked as a node without internal leakage.

[0069] In this embodiment, the acoustic signature signals collected by each valve node during the secondary inspection are compared one by one with the acoustic signature signals collected by the same valve node during the initial inspection. The initial inspection is a baseline signal collected after the injection of tracer gas, reflecting the acoustic signature response state of each valve node during the initial screening; the secondary inspection is a verification signal collected after matching the corresponding injection strategy for a specific risk type, reflecting the acoustic signature response state of the node under targeted inspection conditions. Comparing the two can effectively eliminate the interference of inherent background noise of the pipeline and operating condition fluctuations on the diagnostic results. The specific comparison method is as follows: the acoustic signature spectrum features of each valve node during the initial inspection are extracted as the baseline feature template for that node, and the acoustic signature spectrum features of the same valve node during the secondary inspection are extracted as the verification features. The verification features are compared item by item with the baseline feature template to determine whether there are any newly added frequency components or abnormally large components in the verification features that are not present in the baseline feature template.

[0070] When the acoustic signature of a valve node shows an additional abnormal component in the secondary detection compared to the initial detection, it indicates that the node did not exhibit this abnormal signal in the initial detection, but it appeared in the secondary detection. This additional abnormal component is caused by the differentiated injection strategy used in the secondary detection stimulating the response of the leakage channel. This valve node is then marked as a node with confirmed internal leakage. Conversely, when the acoustic signature of a valve node shows no additional abnormal component in the secondary detection compared to the initial detection, it indicates that the acoustic response of the node is consistent in both detections, and the secondary detection failed to elicit any additional acoustic response. This valve node is then marked as a node without internal leakage.

[0071] The technical solution of the present invention has been described above with reference to the preferred embodiments shown in the accompanying drawings. However, it will be readily understood by those skilled in the art that the scope of protection of the present invention is obviously not limited to these specific embodiments. Without departing from the principles of the present invention, those skilled in the art can make equivalent changes or substitutions to the relevant technical features, and the technical solutions after these changes or substitutions will all fall within the scope of protection of the present invention.

Claims

1. A method for online diagnosis of internal leakage in gas delivery valves based on acoustic signature analysis, characterized in that, include: Step S1: Inject tracer gas into the gas delivery pipeline, collect acoustic signature signals and pressure response signals at each valve node of the gas delivery pipeline, and perform preprocessing to obtain feature data; Step S2: Perform cluster analysis on the feature data. Based on the similarity between the acoustic spectrum features and pressure fluctuation features in the feature data corresponding to each valve node, divide all valve nodes into several feature groups. Determine the feature data distribution range of all valve nodes in each feature group. Mark the valve nodes included in the feature group whose feature data distribution range exceeds the preset benchmark interval as suspected internal leakage areas. Step S3: For each identified suspected internal leakage region, extract the signal evolution trajectory of the feature group corresponding to the suspected internal leakage region on the time axis, and determine the risk type of the suspected internal leakage region based on the signal evolution trajectory. The risk type includes continuous decay type and intermittent mutation type. Step S4: Determine the corresponding secondary detection gas injection strategy according to the risk category; wherein, for the continuous attenuation type, match the corresponding secondary detection gas injection method according to the attenuation degree of the acoustic signal intensity along the transmission distance in the suspected internal leakage area; for the intermittent mutation type, match the corresponding secondary detection gas injection method according to the frequency characteristics and duration characteristics of the acoustic signal mutation event in the suspected internal leakage area. Step S5: Perform secondary detection according to the secondary detection gas injection method to obtain the detection result.

2. The online diagnostic method for internal leakage of gas delivery valves based on acoustic signature analysis according to claim 1, characterized in that, The cluster analysis in step S2 includes: Step S21: Using the acoustic spectrum features and pressure fluctuation features in the feature data corresponding to each valve node as the basis for clustering, determine the degree of similarity between the features corresponding to any two valve nodes. Step S22: Group valve nodes that meet the preset similarity requirements into the same feature group, traverse all valve nodes, and complete the division of all feature groups; Step S23: Determine the distribution range of feature data for all valve nodes in each feature group; Step S24: The valve nodes included in the feature group whose feature data distribution range exceeds the preset benchmark interval are marked as suspected internal leakage areas.

3. The online diagnostic method for internal leakage of gas delivery valves based on acoustic signature analysis according to claim 1, characterized in that, In step S3, the risk type of the suspected endoleak region is determined based on the signal evolution trajectory, including: Obtain the acoustic signature amplitude corresponding to the signal evolution trajectory at each acquisition time; The amplitude of the acoustic signature signal is compared point by point along the time axis between adjacent acquisition times to determine the direction and magnitude of the amplitude change over time. In response to the fact that the amplitude change direction between each adjacent acquisition time is decreasing and the amplitude change magnitude between each adjacent acquisition time is less than the preset change magnitude threshold, the risk type of the suspected internal leakage area is determined to be the continuous decay type. In response to the existence of two or more acquisition moments along the time axis, where the amplitude change between the acquisition moment and the previous acquisition moment is greater than or equal to a preset amplitude change threshold, and the direction of amplitude change between the acquisition moment and the subsequent acquisition moment is opposite to the direction of change of the previous moment, the risk type of the suspected internal leakage region is determined to be intermittent mutation type.

4. The online diagnostic method for internal leakage of gas delivery valves based on acoustic signature analysis according to claim 1, characterized in that, In step S4, for the continuously attenuating type, the corresponding secondary detection gas injection method is matched according to the degree of attenuation of the acoustic signal intensity along the transmission distance in the suspected internal leakage area, including: Obtain the relationship between the acoustic signal intensity at each valve node in the suspected internal leakage area and the position of each valve node along the gas delivery direction; Based on the aforementioned positional correspondence, the attenuation trend of the voiceprint signal intensity along the transmission distance is determined; The injection flow rate and injection pressure of the secondary detection gas are determined based on the attenuation trend.

5. The online diagnostic method for internal leakage of gas delivery valves based on acoustic signature analysis according to claim 4, characterized in that, After determining the injection flow rate and injection pressure of the secondary detection gas, the following steps are also included: The placement method of the acoustic fingerprint sensor in the suspected internal leakage area is determined based on the injection flow rate and injection pressure. In response to the injection flow rate being greater than the preset injection flow rate and the injection pressure being greater than the preset injection pressure, a sensor array is deployed along the gas delivery direction at a first density to synchronously collect acoustic fingerprint signals at each measuring point, and the leak point is located based on the location of the sudden drop in acoustic fingerprint signal intensity along the path. In response to the injection flow rate being less than or equal to a preset injection flow rate or the injection pressure being less than or equal to a preset injection pressure, sensors are deployed along the gas delivery direction at a second density to collect acoustic fingerprint signals segment by segment. The location of the leak segment is determined based on the differences in the segmented signals, wherein the second density is less than the first density.

6. The online diagnostic method for internal leakage of gas delivery valves based on acoustic signature analysis according to claim 1, characterized in that, In step S4, for intermittent mutation types, the corresponding secondary detection gas injection method is matched based on the frequency and duration characteristics of the acoustic signature signal mutation events in the suspected internal leakage area, including: Acquire the amplitude data of the acoustic signature signal in the suspected internal leakage area at each acquisition time; Traverse the amplitude data along the time axis to identify the acquisition time when a sudden change event occurs. The sudden change event is defined as an amplitude change between the acquisition time and the previous acquisition time that is greater than or equal to a preset sudden change amplitude threshold, and the amplitude change between the acquisition time and the next acquisition time is opposite to the previous change direction. The cumulative number of times the mutation event occurs per unit time is determined as the mutation frequency; The duration from the start of the amplitude jump to the amplitude stabilization time in each mutation event is determined as the duration of each mutation. The injection method and sampling mode of the secondary detection gas are determined based on the frequency range in which the mutation frequency falls and the duration range in which the duration falls.

7. The online diagnostic method for internal leakage of gas delivery valves based on acoustic signature analysis according to claim 6, characterized in that, Determine the injection method and sampling mode of the secondary detection gas, including: In response to the mutation frequency being greater than or equal to the preset frequency threshold and the duration being less than the preset duration threshold, a high-voltage pulse injection mode and a continuous sampling mode are matched; In response to the mutation frequency being less than the preset frequency threshold and the duration being greater than or equal to the preset duration threshold, a low-pressure continuous injection mode and an interval sampling mode are matched.

8. The online diagnostic method for internal leakage of gas delivery valves based on acoustic signature analysis according to claim 7, characterized in that, After matching the high-voltage pulse injection method and continuous sampling mode, it also includes: Simultaneously inject detection gas into both ends of the suspected internal leak area and apply alternating pressure pulses; During the application of alternating pressure pulses, acoustic fingerprint signals of each valve node are acquired in continuous sampling mode during the pulse pressurization and pressure holding periods. The acoustic fingerprint signals of each valve node during the pulse pressurization period are compared one by one with the acoustic fingerprint signals of the corresponding nodes during the pressure holding period; When the difference in acoustic signature between a valve node during pulse pressurization and pressure holding is greater than or equal to a preset difference threshold, the valve node is identified as a transient internal leak point.

9. The online diagnostic method for internal leakage of gas delivery valves based on acoustic signature analysis according to claim 7, characterized in that, After matching the low-pressure continuous injection method and the interval sampling mode, it also includes: Simultaneously inject detection gas into both ends of the suspected internal leakage area to maintain stable gas pressure within the area; The area is divided into several detection sections along the gas transport direction, and sensors are installed at the nodes of each detection section. Acoustic signals of each valve node are collected at preset time intervals within each detection section, and signal collection of all detection sections is completed in sequence. When abnormal fluctuations occur in the acoustic signature signal collected within a certain detection section, the valve node within that detection section is identified as a suspected internal leakage point.

10. The online diagnostic method for internal leakage of gas delivery valves based on acoustic signature analysis according to claim 1, characterized in that, In step S5, the detection results are obtained, including: The acoustic fingerprint signals collected by each valve node in the secondary detection are compared one by one with the acoustic fingerprint signals collected by each valve node in the initial detection. When a valve node shows new abnormal components in the acoustic signature during a secondary detection compared to the initial detection, the valve node is marked as an internal leakage confirmed node. When the acoustic signature of a valve node does not show any new abnormal components in the secondary detection compared to the initial detection, the valve node is marked as a node without internal leakage.

Images