A pipeline leakage monitoring method and system based on industrial voiceprint AI recognition

Abstract

The present invention relates to the field of leakage monitoring technology, and discloses a pipeline leakage monitoring method and system based on industrial voiceprint AI recognition, the method comprising: laying a microphone array along the pipeline to be monitored; collecting a voiceprint signal set, performing noise reduction processing, and obtaining each voiceprint amplitude; comparing with the amplitude standard range to determine whether the voiceprint amplitude is abnormal, when it is determined to be abnormal, extracting all abnormal voiceprint amplitudes and recording the abnormal signal position, establishing a voiceprint fitting curve, and determining the number of abnormal leakage points; when it is determined that the number of abnormal leakage points is not unique, extracting a fluctuation curve, performing segmented processing, determining the simulated peak value of the fluctuation curve according to the first data set and the second data set, and determining the abnormal value according to the fusion curve; comparing the abnormal value with the preset abnormal value, and performing abnormal warning according to the comparison result. The present application reduces the cross-interference of signals and the risk of false alarms and missed alarms through segmented processing, simulated peak values, and curve fusion.

Description

Technical Field

[0001] The present invention relates to the field of leakage monitoring technology, and in particular to a pipeline leakage monitoring method and system based on industrial voiceprint AI recognition. Background Art

[0002] With the continuous advancement of industrialization, pipelines, as important infrastructure, are widely used in industries such as oil, natural gas, chemicals, and water treatment. The safe operation of pipelines is directly related to production efficiency, environmental protection, and personnel safety. In order to ensure the stability and safety of pipelines, pipeline leakage monitoring has become a key technical means. In recent years, acoustic monitoring technology has gradually been applied to pipeline leakage detection. The sound signals generated by pipeline leakage are used for monitoring and analysis to identify the leakage point.

[0003] However, the current environment where pipelines are located is often accompanied by a large number of background noise sources, such as equipment operation, pipeline fluid flow noise, wind noise, etc. Noise will interfere with the collection and analysis of voiceprint signals, resulting in false alarms or missed reports. In practical applications, there may be multiple leakage points on the pipeline, especially on long-distance pipelines. Current voiceprint recognition technology is usually difficult to handle the superposition of signals generated by multiple leakage points at the same time. Since the signals between leakage points may overlap and interfere, a single method of early warning based on voiceprint signals cannot fully reflect abnormal leakage. This affects the early warning capability of the monitoring system.

[0004] Therefore, it is necessary to design a pipeline leakage monitoring method and system based on industrial voiceprint AI recognition to solve the problems existing in current technology. Summary of the invention

[0005] In view of this, the present invention proposes a pipeline leakage monitoring method and system based on industrial voiceprint AI recognition, aiming to solve the problems of large monitoring error and low reliability of monitoring results in the current pipeline voiceprint leakage monitoring technology.

[0006] In one aspect, the present invention proposes a pipeline leakage monitoring method based on industrial voiceprint AI recognition, comprising:

[0007] A microphone array is arranged along the pipeline to be monitored, wherein the microphone array includes a plurality of voiceprint collection devices, and the distances between the voiceprint collection devices are equal;

[0008] Collecting a voiceprint signal set from the microphone array, performing noise reduction processing on the voiceprint signal set, and obtaining the voiceprint amplitude collected by each voiceprint collection device according to the noise-reduced voiceprint signal set;

[0009] Comparing the voiceprint amplitude with the amplitude standard range to determine whether the voiceprint amplitude is abnormal, when it is determined that the voiceprint amplitude is abnormal, extracting all abnormal voiceprint amplitudes and recording abnormal signal positions, establishing a voiceprint fitting curve according to the abnormal voiceprint amplitudes and abnormal signal positions, and determining the number of abnormal leakage points according to the voiceprint fitting curve;

[0010] When it is determined that the number of abnormal leakage points is not unique, extract the fluctuation curve between the leftmost peak and the rightmost peak in the voiceprint fitting curve, segment the fluctuation curve, establish a first data set and a second data set, determine the simulated peak of the fluctuation curve according to the first data set and the second data set, merge the simulated peak with the voiceprint fitting curve after removing the fluctuation curve, and determine the abnormal value according to the merged curve;

[0011] The abnormal value is compared with the preset abnormal value, and an abnormal warning is issued according to the comparison result.

[0012] Furthermore, when performing noise reduction processing on the voiceprint signal set, it includes:

[0013] Converting each voiceprint signal in the voiceprint signal set into a two-dimensional matrix by short-time Fourier transform;

[0014] In the silent interval, the noise signal is collected to obtain the power spectrum of the noise;

[0015] De-noising the voiceprint signal by spectral subtraction according to the power spectrum of the noise;

[0016] The de-noised signal is processed by inverse Fourier transform to obtain the de-noised voiceprint signal set.

[0017] Furthermore, when comparing the voiceprint amplitude with the amplitude standard range to determine whether the voiceprint amplitude is abnormal, it includes:

[0018] When the voiceprint amplitude is greater than the maximum value of the amplitude standard range, determining that the voiceprint amplitude is abnormal;

[0019] When the voiceprint amplitude is less than or equal to the maximum value of the amplitude standard range, it is determined that the voiceprint amplitude is not abnormal.

[0020] Furthermore, obtaining the amplitude standard range includes:

[0021] Based on the soundprint collection device, the soundprint amplitude of the pipeline to be monitored when it runs at a rated flow rate is obtained, and recorded as a standard amplitude;

[0022] Collecting the soundprint amplitude in the non-abnormal record of the pipeline to be monitored, and extracting a non-abnormal soundprint amplitude data set from the soundprint signal in the non-abnormal record;

[0023] The standard amplitude is compared with a non-abnormal voiceprint amplitude data set, and the amplitude standard range is generated according to the numerical value relationship, wherein the amplitude standard range includes a left boundary value and a right boundary value.

[0024] Furthermore, when judging the number of abnormal leakage points according to the voiceprint fitting curve, it includes:

[0025] When the soundprint fitting curve has only one peak, it is determined that the number of leakage points in the pipeline to be monitored is unique;

[0026] When there are two or more peaks in the soundprint fitting curve, it is determined that the number of abnormal leakage points in the pipeline to be monitored is not unique.

[0027] Further, when it is determined that the number of leakage points in the pipeline to be monitored is unique, it includes:

[0028] Taking the peak value in the voiceprint fitting curve as the origin, recording the distance between each voiceprint amplitude and the peak value in the voiceprint fitting curve, and recording the difference between the voiceprint amplitude in the voiceprint fitting curve and the maximum value of the amplitude standard range as Si;

[0029] Calculate the weight of each voiceprint amplitude in the voiceprint fitting curve:

[0030] ;

[0031] Among them, Qi represents the weight of the i-th voiceprint amplitude in the voiceprint fitting curve, and w is the standard deviation;

[0032] Calculate outliers:

[0033] ;

[0034] Wherein, n represents the total number of data after removing the peak value in the voiceprint fitting curve.

[0035] Further, when it is determined that the number of abnormal leakage points is not unique, the fluctuation curve is segmented to establish the first data set and the second data set, including:

[0036] Extracting a line segment showing an upward trend in the fluctuation curve, and listing each voiceprint amplitude corresponding to the upward trend line segment into the first data set;

[0037] Extracting a line segment showing a downward trend in the fluctuation curve, and listing each voiceprint amplitude corresponding to the line segment showing a downward trend into the second data set;

[0038] Each peak value in the fluctuation curve is included in the first data set, and each valley value is included in the second data set.

[0039] Further, determining the simulated peak value of the fluctuation curve according to the first data set and the second data set includes:

[0040]

[0041] Among them, A is the simulated peak value, a1 is the calculation coefficient corresponding to the first data set, n is the number of voiceprint amplitudes in the first data set, fi is the i-th voiceprint amplitude in the first data set, E1 is the average value of the voiceprint amplitude in the first data set, a2 is the calculation coefficient corresponding to the second data set, m is the number of voiceprint amplitudes in the second data set, fj is the j-th voiceprint amplitude in the second data set, and E2 is the average value of the voiceprint amplitude in the second data set.

[0042] Furthermore, the abnormal value is compared with a preset abnormal value, and an abnormal warning is issued according to the comparison result, including:

[0043] The abnormal value is compared with a first preset abnormal value and a second preset abnormal value respectively, and an abnormal warning is issued according to the comparison result; the first preset abnormal value is smaller than the second preset abnormal value;

[0044] When the abnormal value is less than or equal to the first preset abnormal value, a third-level abnormal warning is issued; when the abnormal value is greater than the first preset abnormal value and less than or equal to the second preset abnormal value, a second-level abnormal warning is issued; when the abnormal value is greater than the second preset abnormal value, a first-level abnormal warning is issued; the third-level abnormal warning is smaller than the second-level abnormal warning, and the second-level abnormal warning is smaller than the first-level abnormal warning.

[0045] Compared with the prior art, the beneficial effect of the present invention is that by performing noise reduction processing on the voiceprint signal set collected by the microphone array, the interference of background noise, such as equipment operation noise, pipeline fluid flow noise and environmental noise, is removed, so that the leakage signal can be extracted and analyzed more clearly and accurately. Noise reduction processing improves the quality of the signal. By comparing the voiceprint amplitude with the amplitude standard range and judging the number of leakage points by fitting curve, the signals of multiple leakage points can be identified and distinguished, avoiding the signal overlap and interference problems that may be caused by the traditional method when multiple leakage points occur at the same time. The location and characteristics of each leakage point are analyzed to improve the accuracy and reliability of leakage monitoring. When the signals of multiple leakage points are difficult to distinguish, by extracting the fluctuation curve in the fitting curve, segmented processing and simulating the peak value, and then merging with the fitting curve after removing the fluctuation curve, the cross interference of the signal is reduced, and by comparing the fused abnormal value with the preset abnormal value, an abnormal warning is realized, leakage problems are found and an early warning is issued, and the pipeline safety and timely repair are ensured. The risk of false alarms and missed alarms is reduced, and the emergency response speed of pipeline monitoring is improved.

[0046] On the other hand, the present application also provides a pipeline leakage monitoring system based on industrial voiceprint AI recognition, which is used to apply the above-mentioned pipeline leakage monitoring method based on industrial voiceprint AI recognition, including:

[0047] A microphone array, wherein the microphone array is arranged along the pipeline to be monitored, and the microphone array includes a plurality of voiceprint collection devices, and the distances between the voiceprint collection devices are equal;

[0048] a collecting unit configured to collect a voiceprint signal set from the microphone array, perform noise reduction processing on the voiceprint signal set, and obtain the voiceprint amplitude collected by each voiceprint collecting device according to the noise-reduced voiceprint signal set;

[0049] a judgment unit configured to compare the voiceprint amplitude with a standard amplitude range to judge whether the voiceprint amplitude is abnormal; when the voiceprint amplitude is judged to be abnormal, extract all abnormal voiceprint amplitudes and record abnormal signal positions; establish a voiceprint fitting curve according to the abnormal voiceprint amplitudes and abnormal signal positions; and judge the number of abnormal leakage points according to the voiceprint fitting curve;

[0050] A processing unit is configured to, when it is determined that the number of abnormal leakage points is not unique, extract a fluctuation curve between the leftmost peak and the rightmost peak in the voiceprint fitting curve, perform segmentation processing on the fluctuation curve, establish a first data set and a second data set, determine a simulated peak value of the fluctuation curve according to the first data set and the second data set, fuse the simulated peak value with the voiceprint fitting curve after removing the fluctuation curve, and determine an abnormal value according to the fused curve;

[0051] The early warning unit is configured to compare the abnormal value with a preset abnormal value and issue an abnormal early warning according to the comparison result.

[0052] It can be understood that the above-mentioned pipeline leakage monitoring method and system based on industrial voiceprint AI recognition have the same beneficial effects and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] Various other advantages and benefits will become apparent to those of ordinary skill in the art by reading the detailed description of the preferred embodiments below. The accompanying drawings are only for the purpose of illustrating the preferred embodiments and are not to be considered as limiting the present invention. Moreover, the same reference symbols are used throughout the accompanying drawings to represent the same components. In the accompanying drawings:

[0054] Figure 1 A flow chart of a pipeline leakage monitoring method based on industrial voiceprint AI recognition provided by an embodiment of the present invention;

[0055] Figure 2 A structural block diagram of a pipeline leakage monitoring system based on industrial voiceprint AI recognition provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0056] Exemplary embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although exemplary embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided in order to enable a more thorough understanding of the present disclosure and to fully convey the scope of the present disclosure to those skilled in the art. It should be noted that, in the absence of conflict, the embodiments of the present invention and the features described in the embodiments can be combined with each other. The present invention will be described in detail below with reference to the accompanying drawings and in combination with the embodiments.

[0057] In some embodiments of the present application, see Figure 1 As shown, a pipeline leakage monitoring method based on industrial voiceprint AI recognition includes:

[0058] S100: Arrange a microphone array along the pipeline to be monitored, wherein the microphone array includes a plurality of voiceprint collection devices, and the distances between the voiceprint collection devices are equal.

[0059] S200: Collecting a voiceprint signal set from a microphone array, performing noise reduction processing on the voiceprint signal set, and obtaining the voiceprint amplitude collected by each voiceprint collection device according to the noise-reduced voiceprint signal set.

[0060] S300: Compare the voiceprint amplitude with the amplitude standard range to determine whether the voiceprint amplitude is abnormal. When the voiceprint amplitude is determined to be abnormal, extract all abnormal voiceprint amplitudes and record the abnormal signal positions. Establish a voiceprint fitting curve based on the abnormal voiceprint amplitudes and abnormal signal positions, and determine the number of abnormal leakage points based on the voiceprint fitting curve.

[0061] S400: When it is determined that the number of abnormal leakage points is not unique, extract the fluctuation curve between the leftmost peak and the rightmost peak in the voiceprint fitting curve, segment the fluctuation curve, establish a first data set and a second data set, determine the simulated peak of the fluctuation curve based on the first data set and the second data set, merge the simulated peak with the voiceprint fitting curve after removing the fluctuation curve, and determine the abnormal value based on the merged curve.

[0062] S500: Compare the abnormal value with the preset abnormal value, and issue an abnormal warning according to the comparison result.

[0063] Specifically, in S100, a microphone array is arranged along the pipeline to be monitored, and the spacing between each voiceprint collection device is equal. A wider pipeline area is covered to ensure that sound signals from different leakage points are captured. The voiceprint signal set collected in S200 usually contains a large amount of background noise, and the voiceprint signal set is subjected to noise reduction processing. The influence of background noise is reduced to extract the pipeline leakage soundprint. The noise-reduced signal is used to calculate the voiceprint amplitude of each microphone array collection device and provide a basis for subsequent abnormality detection. In S300, the voiceprint amplitude is compared with the standard amplitude range to determine whether there is an abnormality. When the amplitude exceeds the standard range, it is regarded as an abnormal signal, and the position of the abnormal signal is recorded. By extracting these abnormal amplitudes and establishing a voiceprint fitting curve based on their positions, it is further analyzed and determined whether there is a leakage point. In S400, when there are multiple abnormal signals, the voiceprint fitting curve shows multiple peaks, which makes it difficult to directly determine the leakage situation of the leakage point. By extracting the fluctuation curve between the leftmost and rightmost peaks in the fitting curve and performing segmented processing on the fluctuation curve, two data sets (the first data set and the second data set) are established, and the fluctuation curve is subjected to simulated peak processing. The pipeline abnormal value is determined by combining the simulated peak value and the soundprint fitting curve after removing the fluctuation. In S500, the abnormal value is compared with the preset abnormal threshold. If the abnormal value exceeds the set threshold range, an early warning is triggered, an alarm is processed for the leakage, and the pipeline maintenance personnel are notified in time to intervene and handle it.

[0064] It is understandable that through noise reduction processing and accurate establishment of voiceprint fitting curves, multiple leakage points can be identified and located, avoiding the common false alarms and missed reports in traditional methods. Combined with the analysis of the algorithm, the intelligence level of pipeline leakage monitoring is improved, the need for manual intervention is reduced, and the speed and accuracy of leakage response are improved. Especially in complex industrial environments, it can effectively deal with the interference of background noise and the signal overlap of multiple leakage points.

[0065] In some embodiments of the present application, when performing noise reduction processing on a voiceprint signal set, the process includes:

[0066] Each voiceprint signal in the voiceprint signal set is converted into a two-dimensional matrix through short-time Fourier transform.

[0067] In the silent interval, the noise signal is collected to obtain the power spectrum of the noise.

[0068] The voiceprint signal is denoised by spectral subtraction according to the power spectrum of the noise.

[0069] The denoised signal is processed by inverse Fourier transform to obtain a denoised voiceprint signal set.

[0070] Specifically, the voiceprint signal set collected in this solution is the voiceprint signal within a certain period of time, and the time interval setting can be set according to the collection frequency of the voiceprint collection device. Short-time Fourier transform is performed on each frame of the voiceprint signal in the voiceprint signal set. STFT (short-time Fourier transform) converts the voiceprint signal from the time domain to the frequency domain, and divides the signal into several small windows according to time. The signal in each small window is regarded as a stationary signal. A two-dimensional matrix is ​​obtained, in which one dimension represents time and the other dimension represents frequency. In the process of collecting pipeline leakage voiceprint signals, there will be a "silent interval", that is, a period when no leakage occurs. In these silent intervals, only background noise can be collected. The power spectrum of the noise is calculated through the noise signals in these silent intervals. The power spectrum reflects the energy distribution of noise at different frequencies. Using the noise power spectrum collected in the silent interval, the spectral subtraction method eliminates the noise component by subtracting the power spectrum of the noise from the spectrum of the signal. The background noise is removed in a targeted manner to retain the effective leakage voiceprint signal. Spectral subtraction reduces the impact of noise on signal quality by subtracting the noise spectrum in the frequency domain. The denoised signal is converted back to the time domain through an inverse Fourier transform (inverse STFT). At this point, the resulting signal will no longer contain the original background noise.

[0071] It is understandable that by performing time-frequency domain analysis on the signal through STFT, the spectrum of the signal can be accurately processed, retaining the effective signal while removing the interference of background noise. By adopting the method of noise collection in the silent zone, the noise reduction strategy can be dynamically adjusted so that the processed signal can more accurately reflect the true condition of the pipeline. The application of inverse Fourier transform ensures that the noise-reduced signal can be restored to the time domain for subsequent amplitude analysis and leakage detection. This noise reduction method is particularly suitable for scenarios where background noise changes frequently occur in industrial environments, which improves the accuracy of pipeline leakage monitoring, reduces false alarms and missed alarms, and provides reliable protection for the safe operation of pipelines.

[0072] In some embodiments of the present application, when comparing the voiceprint amplitude with the amplitude standard range to determine whether the voiceprint amplitude is abnormal, it includes: when the voiceprint amplitude is greater than the maximum value of the amplitude standard range, it is determined that the voiceprint amplitude is abnormal. When the voiceprint amplitude is less than or equal to the maximum value of the amplitude standard range, it is determined that the voiceprint amplitude is not abnormal.

[0073] In some embodiments of the present application, when obtaining the amplitude standard range, it includes: obtaining the soundprint amplitude of the pipeline to be monitored when it runs at the rated flow rate based on the soundprint collection device, and recording it as the standard amplitude. Collecting the soundprint amplitude in the non-abnormal record of the pipeline to be monitored, and extracting the non-abnormal soundprint amplitude data set from the soundprint signal in the non-abnormal record. Comparing the standard amplitude with the non-abnormal soundprint amplitude data set, and generating the amplitude standard range according to the numerical value relationship, wherein the amplitude standard range includes a left boundary value and a right boundary value.

[0074] Specifically, when the pipeline is running at the rated flow rate, the amplitude value recorded by the soundprint collection device is used as the "standard amplitude". This standard value represents the soundprint amplitude of the pipeline during normal operation and serves as the basis for determining amplitude anomalies. In order to further establish the amplitude standard range, the soundprint amplitude data from multiple "non-abnormal" records are collected. Non-abnormal records refer to signals collected from the pipeline when no leakage or other abnormal events occur, reflecting the typical amplitude changes of the pipeline under normal conditions. The standard amplitude is compared with the non-abnormal soundprint amplitude data set, and the amplitude standard range is generated based on their numerical relationship. The amplitude standard range contains two boundary values: the left boundary value and the right boundary value, which represent the minimum and maximum allowable ranges of the amplitude, respectively. In the actual signal acquisition process, if the soundprint amplitude exceeds the right boundary of the standard range, it is judged as abnormal.

[0075] It is understandable that the standard amplitude range generated based on the standard amplitude of the pipeline at the rated flow rate, combined with the non-abnormal records actually collected, provides a clear boundary for monitoring. Avoid interference from background noise or environmental factors to ensure that abnormal signals will trigger alarms. The abnormal amplitude judgment method adopts an upper limit judgment, which ensures high sensitivity to leaks or other abnormal conditions, while avoiding misjudgments caused by small fluctuations. It improves the stability of operation in complex industrial environments and improves the accuracy and reliability of monitoring.

[0076] In some embodiments of the present application, when judging the number of abnormal leakage points according to the soundprint fitting curve, it includes: when the soundprint fitting curve has one and only one peak, it is judged that the number of leakage points in the pipeline to be monitored is unique. When there are two or more peaks in the soundprint fitting curve, it is judged that the number of abnormal leakage points in the pipeline to be monitored is not unique.

[0077] In some embodiments of the present application, when it is determined that the number of leakage points in the pipeline to be monitored is unique, it includes: taking the peak value in the soundprint fitting curve as the origin, recording the distance between each soundprint amplitude and the peak value in the soundprint fitting curve, and recording the difference between the soundprint amplitude in the soundprint fitting curve and the maximum value of the amplitude standard range as Si.

[0078] Calculate the weight of each voiceprint amplitude in the voiceprint fitting curve:

[0079] ;

[0080] Among them, Qi represents the weight of the i-th voiceprint amplitude in the voiceprint fitting curve, and w is the standard deviation.

[0081] Calculate outliers:

[0082] ;

[0083] Wherein, n represents the total number of data after removing the peak value in the voiceprint fitting curve.

[0084] Specifically, when the voiceprint fitting curve has one and only one peak, it means there is only one leakage point, so the number of leakage points in the pipeline can be determined to be unique. If there are two or more peaks in the fitting curve, it means there are multiple leakage points, and the number of leakage points is determined to be non-unique. In the case of a unique number of leakage points, the peak value in the fitting curve is taken as the origin, and the distance between the peak value and other voiceprint amplitudes is recorded. This distance is the distance of the voiceprint collection device. For each voiceprint amplitude in the fitting curve, the weight is calculated using a formula based on the relative position of each voiceprint amplitude to the peak value of the fitting curve. The amplitude with a larger weight means that the point is more affected by the leakage point.

[0085] It is understandable that judging whether the number of leakage points is unique by the number of peaks in the fitting curve avoids misjudgment caused by signal overlap. Combined with the calculation of the difference between the amplitude and the standard range, and the evaluation of the weight, the abnormal points in the fitting curve are identified to determine the degree of abnormality of the leakage point. The introduction of weight calculation makes the identification of abnormal values ​​more accurate, which helps to provide more timely and targeted leakage warnings, thereby improving the safety and operation efficiency of the pipeline.

[0086] In some embodiments of the present application, when it is determined that the number of abnormal leakage points is not unique, the fluctuation curve is segmented to establish the first data set and the second data set, including: extracting the line segments with an upward trend in the fluctuation curve, and listing the voiceprint amplitudes corresponding to the upward trend line segments in the first data set. Extracting the line segments with a downward trend in the fluctuation curve, and listing the voiceprint amplitudes corresponding to the downward trend line segments in the second data set. Wherein each peak value in the fluctuation curve is included in the first data set, and each valley value is included in the second data set.

[0087] In some embodiments of the present application, determining a simulated peak value of a fluctuation curve according to the first data set and the second data set includes:

[0088]

[0089] Where A is the simulated peak value, a1 is the calculation coefficient corresponding to the first data set, n is the number of voiceprint amplitudes in the first data set, fi is the i-th voiceprint amplitude in the first data set, E1 is the average voiceprint amplitude in the first data set, a2 is the calculation coefficient corresponding to the second data set, m is the number of voiceprint amplitudes in the second data set, fj is the j-th voiceprint amplitude in the second data set, and E2 is the average voiceprint amplitude in the second data set. The calculation coefficient a1 and the calculation coefficient a2 have a value range of (1,1.5).

[0090] It can be understood that in this embodiment, after obtaining the simulated peak, it is fused with the voiceprint fitting curve after removing the fluctuation curve, that is, the simulated peak is used as the peak of the voiceprint fitting curve, and the horizontal coordinate of the simulated peak is located at the center of the break of the voiceprint fitting curve. Then, the curve is completed according to the slope of the peak and the break of the voiceprint fitting curve to obtain the fused curve. The fused curve has a peak, and the process of calculating the outlier is the same as the single peak.

[0091] It is understandable that by analyzing the fluctuation curve in sections and establishing two data sets, we can effectively deal with the complex situation of signal overlap and interference from multiple leakage points. By extracting the rising and falling trends of the fluctuation curve, we can capture the enhancement and attenuation process of the leakage signal respectively, so as to identify each leakage point more accurately. In particular, for the case where there are multiple peaks in the fluctuation curve, the first data set and the second data set are combined with weighted calculation to evaluate the overall leakage situation, avoiding misjudgment caused by signal overlap. The calculation method of the simulated peak value effectively smoothes the data and improves the accuracy and stability of monitoring.

[0092] In some embodiments of the present application, when comparing an abnormal value with a preset abnormal value and issuing an abnormal warning according to the comparison result, it includes: comparing the abnormal value with a preset first abnormal value and a preset second abnormal value respectively, and issuing an abnormal warning according to the comparison result. The first preset abnormal value is smaller than the second preset abnormal value.

[0093] Specifically, when the abnormal value is less than or equal to the first preset abnormal value, a third-level abnormal warning is issued. When the abnormal value is greater than the first preset abnormal value and less than or equal to the second preset abnormal value, a second-level abnormal warning is issued. When the abnormal value is greater than the second preset abnormal value, a first-level abnormal warning is issued. The third-level abnormal warning is less than the second-level abnormal warning, and the second-level abnormal warning is less than the first-level abnormal warning.

[0094] It is understandable that through the multi-level early warning mechanism, the response strategy can be flexibly adjusted according to the degree of abnormality, so as to achieve hierarchical management. By setting different thresholds, the risks of minor, medium and serious leakage can be effectively distinguished, avoiding the interference of too frequent or too minor early warnings, making monitoring more efficient. The multi-level early warning mechanism can not only provide real-time risk assessment, but also help operators make more appropriate responses. For example, for minor abnormalities, only the third-level early warning is triggered to remind the monitoring personnel to continue to observe and avoid unnecessary overreaction; for serious abnormalities, the first-level early warning will immediately trigger an emergency response to ensure the safe operation of the pipeline. Through the precise early warning mechanism, false alarms and missed alarms are minimized, and the reliability and accuracy of detection are improved.

[0095] In the above embodiment, the noise reduction processing is performed on the voiceprint signal set collected by the microphone array to remove the interference of background noise, such as equipment operation noise, pipeline fluid flow noise and environmental noise, so that the leakage signal can be extracted and analyzed more clearly and accurately. Noise reduction processing improves the quality of the signal. By comparing the voiceprint amplitude with the amplitude standard range and judging the number of leakage points by fitting curve, the signals of multiple leakage points can be identified and distinguished, avoiding the signal overlap and interference problems that may occur when multiple leakage points occur at the same time in the traditional method. The location and characteristics of each leakage point are analyzed to improve the accuracy and reliability of leakage monitoring. When the signals of multiple leakage points are difficult to distinguish, by extracting the fluctuation curve in the fitting curve, segmented processing and simulating the peak value, and then merging with the fitting curve after removing the fluctuation curve, the cross interference of the signal is reduced, and by comparing the fused abnormal value with the preset abnormal value, an abnormal warning is realized, leakage problems are found and an early warning is issued to ensure pipeline safety and timely repair. The risk of false alarms and missed alarms is reduced, and the emergency response speed of pipeline monitoring is improved.

[0096] In another preferred embodiment based on the above embodiment, refer to Figure 2 As shown, this embodiment provides a pipeline leakage monitoring system based on industrial voiceprint AI recognition, which is used to apply the above-mentioned pipeline leakage monitoring method based on industrial voiceprint AI recognition, including:

[0097] A microphone array is arranged along the pipeline to be monitored. The microphone array includes a plurality of voiceprint collection devices, and the distances between the voiceprint collection devices are equal;

[0098] A collection unit is configured to collect a voiceprint signal set from the microphone array, perform noise reduction processing on the voiceprint signal set, and obtain the voiceprint amplitude collected by each voiceprint collection device according to the noise-reduced voiceprint signal set;

[0099] a judgment unit configured to compare the voiceprint amplitude with the amplitude standard range to judge whether the voiceprint amplitude is abnormal; when the voiceprint amplitude is judged to be abnormal, extract all abnormal voiceprint amplitudes and record the abnormal signal positions; establish a voiceprint fitting curve according to the abnormal voiceprint amplitudes and the abnormal signal positions; and judge the number of abnormal leakage points according to the voiceprint fitting curve;

[0100] The processing unit is configured to, when it is determined that the number of abnormal leakage points is not unique, extract a fluctuation curve between the leftmost peak and the rightmost peak in the voiceprint fitting curve, perform segmentation processing on the fluctuation curve, establish a first data set and a second data set, determine a simulated peak value of the fluctuation curve according to the first data set and the second data set, fuse the simulated peak value with the voiceprint fitting curve after removing the fluctuation curve, and determine an abnormal value according to the fused curve;

[0101] The early warning unit is configured to compare the abnormal value with the preset abnormal value and issue an abnormal early warning according to the comparison result.

[0102] It is understandable that by performing noise reduction processing on the voiceprint signal set collected by the microphone array, the interference of background noise, such as equipment operation noise, pipeline fluid flow noise and environmental noise, is removed, so that the leakage signal can be extracted and analyzed more clearly and accurately. Noise reduction processing improves the quality of the signal. By comparing the voiceprint amplitude with the amplitude standard range and judging the number of leakage points by fitting curves, the signals of multiple leakage points can be identified and distinguished, avoiding the signal overlap and interference problems that may occur when multiple leakage points occur at the same time in traditional methods. The location and characteristics of each leakage point are analyzed to improve the accuracy and reliability of leakage monitoring. When the signals of multiple leakage points are difficult to distinguish, by extracting the fluctuation curve in the fitting curve, segmented processing and simulating the peak value, and then merging it with the fitting curve after removing the fluctuation curve, the cross-interference of the signal is reduced. By comparing the fused abnormal value with the preset abnormal value, an abnormal warning is realized, leakage problems are discovered and warnings are issued to ensure pipeline safety and timely repair. The risk of false alarms and missed alarms is reduced, and the emergency response speed of pipeline monitoring is improved.

[0103] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program codes.

[0104] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems) and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0105] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0106] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0107] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, ordinary technicians in the relevant field should understand that the specific implementation methods of the present invention can still be modified or replaced by equivalents. Any modification or equivalent replacement that does not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A pipeline leakage monitoring method based on industrial voiceprint AI recognition, characterized in that: include: A microphone array is arranged along the pipeline to be monitored, wherein the microphone array includes a plurality of voiceprint collection devices, and the distances between the voiceprint collection devices are equal; Collecting a voiceprint signal set from the microphone array, performing noise reduction processing on the voiceprint signal set, and obtaining the voiceprint amplitude collected by each voiceprint collection device according to the noise-reduced voiceprint signal set; Comparing the voiceprint amplitude with the amplitude standard range to determine whether the voiceprint amplitude is abnormal, when it is determined that the voiceprint amplitude is abnormal, extracting all abnormal voiceprint amplitudes and recording abnormal signal positions, establishing a voiceprint fitting curve according to the abnormal voiceprint amplitudes and abnormal signal positions, and determining the number of abnormal leakage points according to the voiceprint fitting curve; When it is determined that the number of abnormal leakage points is not unique, extract the fluctuation curve between the leftmost peak and the rightmost peak in the voiceprint fitting curve, segment the fluctuation curve, establish a first data set and a second data set, determine the simulated peak of the fluctuation curve according to the first data set and the second data set, merge the simulated peak with the voiceprint fitting curve after removing the fluctuation curve, and determine the abnormal value according to the merged curve; Compare the abnormal value with the preset abnormal value, and issue an abnormal warning according to the comparison result; When judging the number of abnormal leakage points according to the voiceprint fitting curve, it includes: When the soundprint fitting curve has only one peak, it is determined that the number of leakage points in the pipeline to be monitored is unique; When there are two or more peaks in the soundprint fitting curve, it is determined that the number of abnormal leakage points in the pipeline to be monitored is not unique; When it is determined that the number of leakage points in the pipeline to be monitored is unique, it includes: Taking the peak value in the voiceprint fitting curve as the origin, recording the distance between each voiceprint amplitude and the peak value in the voiceprint fitting curve, and recording the difference between the voiceprint amplitude in the voiceprint fitting curve and the maximum value of the amplitude standard range as Si; Calculate the weight of each voiceprint amplitude in the voiceprint fitting curve: ; Among them, Qi represents the weight of the i-th voiceprint amplitude in the voiceprint fitting curve, and w is the standard deviation; Calculate outliers: ; Wherein, n represents the total number of data after removing the peak value in the voiceprint fitting curve; When it is determined that the number of abnormal leakage points is not unique, the fluctuation curve is segmented to establish the first data set and the second data set, including: Extracting a line segment showing an upward trend in the fluctuation curve, and listing each voiceprint amplitude corresponding to the upward trend line segment into the first data set; Extracting a line segment showing a downward trend in the fluctuation curve, and listing each voiceprint amplitude corresponding to the line segment showing a downward trend into the second data set; Wherein each peak value in the fluctuation curve is included in the first data set, and each valley value is included in the second data set; Determining the simulated peak value of the fluctuation curve according to the first data set and the second data set includes: ; Among them, A is the simulated peak value, a1 is the calculation coefficient corresponding to the first data set, n is the number of voiceprint amplitudes in the first data set, fi is the i-th voiceprint amplitude in the first data set, E1 is the average value of the voiceprint amplitude in the first data set, a2 is the calculation coefficient corresponding to the second data set, m is the number of voiceprint amplitudes in the second data set, fj is the j-th voiceprint amplitude in the second data set, and E2 is the average value of the voiceprint amplitude in the second data set.

2. The pipeline leakage monitoring method based on industrial voiceprint AI recognition according to claim 1 is characterized in that: When performing noise reduction processing on the voiceprint signal set, it includes: Converting each voiceprint signal in the voiceprint signal set into a two-dimensional matrix by short-time Fourier transform; In the silent interval, the noise signal is collected to obtain the power spectrum of the noise; De-noising the voiceprint signal by spectral subtraction according to the power spectrum of the noise; The de-noised signal is processed by inverse Fourier transform to obtain the de-noised voiceprint signal set.

3. The pipeline leakage monitoring method based on industrial voiceprint AI recognition according to claim 1 is characterized in that: When comparing the voiceprint amplitude with the amplitude standard range to determine whether the voiceprint amplitude is abnormal, it includes: When the voiceprint amplitude is greater than the maximum value of the amplitude standard range, determining that the voiceprint amplitude is abnormal; When the voiceprint amplitude is less than or equal to the maximum value of the amplitude standard range, it is determined that the voiceprint amplitude is not abnormal.

4. The pipeline leakage monitoring method based on industrial voiceprint AI recognition according to claim 3 is characterized in that: When obtaining the amplitude standard range, it includes: Based on the soundprint collection device, the soundprint amplitude of the pipeline to be monitored when it runs at a rated flow rate is obtained, and recorded as a standard amplitude; Collecting the soundprint amplitude in the non-abnormal record of the pipeline to be monitored, and extracting a non-abnormal soundprint amplitude data set from the soundprint signal in the non-abnormal record; The standard amplitude is compared with a non-abnormal voiceprint amplitude data set, and the amplitude standard range is generated according to the numerical value relationship, wherein the amplitude standard range includes a left boundary value and a right boundary value.

5. The pipeline leakage monitoring method based on industrial voiceprint AI recognition according to claim 1 is characterized in that: The abnormal value is compared with the preset abnormal value, and an abnormal warning is issued according to the comparison result, including: The abnormal value is compared with a first preset abnormal value and a second preset abnormal value respectively, and an abnormal warning is issued according to the comparison result; the first preset abnormal value is smaller than the second preset abnormal value; When the abnormal value is less than or equal to the first preset abnormal value, a third-level abnormal warning is issued; when the abnormal value is greater than the first preset abnormal value and less than or equal to the second preset abnormal value, a second-level abnormal warning is issued; when the abnormal value is greater than the second preset abnormal value, a first-level abnormal warning is issued; the third-level abnormal warning is smaller than the second-level abnormal warning, and the second-level abnormal warning is smaller than the first-level abnormal warning.

6. A pipeline leakage monitoring system based on industrial voiceprint AI recognition, used to apply the pipeline leakage monitoring method based on industrial voiceprint AI recognition as described in any one of claims 1 to 5, characterized in that: include: A microphone array, wherein the microphone array is arranged along the pipeline to be monitored, and the microphone array includes a plurality of voiceprint collection devices, and the distances between the voiceprint collection devices are equal; a collecting unit configured to collect a voiceprint signal set from the microphone array, perform noise reduction processing on the voiceprint signal set, and obtain the voiceprint amplitude collected by each voiceprint collecting device according to the noise-reduced voiceprint signal set; a judgment unit configured to compare the voiceprint amplitude with a standard amplitude range to judge whether the voiceprint amplitude is abnormal; when the voiceprint amplitude is judged to be abnormal, extract all abnormal voiceprint amplitudes and record abnormal signal positions; establish a voiceprint fitting curve according to the abnormal voiceprint amplitudes and abnormal signal positions; and judge the number of abnormal leakage points according to the voiceprint fitting curve; A processing unit is configured to, when it is determined that the number of abnormal leakage points is not unique, extract a fluctuation curve between the leftmost peak and the rightmost peak in the voiceprint fitting curve, perform segmentation processing on the fluctuation curve, establish a first data set and a second data set, determine a simulated peak value of the fluctuation curve according to the first data set and the second data set, fuse the simulated peak value with the voiceprint fitting curve after removing the fluctuation curve, and determine an abnormal value according to the fused curve; The early warning unit is configured to compare the abnormal value with a preset abnormal value and issue an abnormal early warning according to the comparison result.

Images