Signal identification and positioning method based on point-type high-sensitivity optical acoustic vibration detector

Abstract

The invention discloses a signal identification and positioning method based on a point-type high-sensitivity optical acoustic vibration detector, and belongs to the technical field of signal detection, and the method specifically comprises the steps: collecting signal data of a potential leakage region, and dividing the signal data into a reference signal and a to-be-detected signal; analyzing the signal through a frequency band energy detection method, calculating frequency band energy, and calculating a frequency band energy transcendental value of the signal to be detected relative to the reference signal; according to an override result, judging event types, and forming a multi-channel multi-batch event matrix; performing anomaly detection on the matrix by using an independent channel and continuous multi-channel anomaly detection method, recording positioning information, and outputting an early warning result; the method improves the accuracy and timeliness of leakage detection, and is suitable for various potential leakage scenes.

Description

Technical Field

[0001] The invention belongs to the technical field of signal detection, in particular to a signal recognition and positioning method based on a point-type high-sensitivity optical acoustic vibration detector. Background Art

[0002] Traditional leak detection methods, such as listening, observation, and regional metering, have limitations such as low efficiency, poor accuracy, and strong reliance on personnel experience. Especially in large-scale high-power laser devices, there are strict requirements for the sealing status of optical transmission pipelines and target chambers. Traditional detection methods cannot meet their efficient and accurate detection needs. Although modern pipeline leak point measurement instruments use high-tech means such as sonar technology, radar technology, infrared thermal imaging technology, and big data analysis, these methods are often complex and costly, and still have limitations in certain application scenarios.

[0003] For example, a Chinese patent application with authorization announcement number CN116973697B discloses a leakage current identification system and method, including: an insulator string, a junction box, a circulator 1, a circulator 2, a photodetector 1, a photodetector 2, an optical amplifier, an acousto-optic modulator, an arbitrary waveform generator, a coupler 1, a coupler 2, a laser source and a data processing module, wherein the insulator string is connected to the junction box, and the junction box divides two lines to connect the circulator 1 and the circulator 2 respectively; the beneficial effects of this technical solution are: the entire system is passive on the tower side and uses optical fiber; winding optical fiber around the metal part of the insulator can improve the insulation performance to a certain extent; the system can locate which specific insulator or insulators have flashovers, and can also distinguish between pollution flashover, ice flashover and whether there is obvious discharge and heating according to the FBG monitoring situation.

[0004] The above existing technologies all have the following problems: Although the above patents can locate which specific insulator or insulators flashover occurs, they may be limited by the optical fiber layout and signal transmission characteristics, affecting the monitoring range and accuracy, and may be affected by environmental conditions such as temperature, humidity, and pollution on the optical fiber transmission performance, resulting in instability of the monitoring results. At the same time, there are limitations in real-time performance and response speed. Summary of the invention

[0005] In view of the shortcomings of the prior art, the present invention proposes a signal recognition and positioning method based on a point-type high-sensitivity optical acoustic vibration detector, which collects signal data of potential leakage areas and divides them into reference signals and signals to be detected; analyzes signals through a frequency band energy detection method, calculates frequency band energy, and calculates the frequency band energy exceedance ratio of the signal to be detected relative to the reference signal; distinguishes event categories based on exceedance results to form a multi-channel and multi-batch event matrix; uses independent channels and continuous multi-channel anomaly detection methods to perform anomaly detection on the matrix, records positioning information, and outputs warning results, thereby improving the accuracy and timeliness of leakage detection and being suitable for various potential leakage scenarios.

[0006] To achieve the above object, the present invention provides the following technical solutions:

[0007] The signal recognition and positioning method based on the point-type high-sensitivity optical acoustic vibration detector includes:

[0008] Step S1: using a point-type high-sensitivity optical acoustic vibration detector to collect signals from a potential leakage area, obtaining signal data of the potential leakage area, and dividing the signal data of the potential leakage area into a reference signal and a signal to be detected;

[0009] Step S2: Analyze the reference signal and the signal to be detected by using a frequency band energy detection method, and calculate the frequency band energy of the reference signal and the signal to be detected;

[0010] Step S3: Calculate the exceedance ratio of the signal to be detected relative to the reference signal in each frequency band according to the frequency band energy of the reference signal and the signal to be detected, combined with the preset threshold discrimination parameter, and record the exceedance result of the signal to be detected and the reference signal;

[0011] Step S4: discriminating the event category of the signal to be detected according to the exceeding result, and dividing the signal to be detected data according to the discriminating result to form a multi-channel multi-batch event matrix;

[0012] Step S5: Based on the multi-channel multi-batch event matrix, anomaly detection is performed using an independent channel anomaly detection method and a continuous multi-channel anomaly detection method, the positioning information of the anomaly detection is recorded, and an early warning result is output according to the positioning information of the anomaly detection.

[0013] Specifically, the specific steps of step S2 include:

[0014] S2.1: Obtain reference signal and signal to be detected, and perform preprocessing;

[0015] S2.2: According to the analysis requirements, the frequency range of the reference signal and the signal to be detected is divided into N1 frequency bands, and the formula is used The energy of the reference signal and the signal to be detected in each frequency band is calculated, where E p represents the energy in frequency band p, wk represents the weight factor of the kth sampling point, α k represents the noise suppression factor of the kth sampling point, S(kΔf,R f ) indicates that the frequency resolution R f Under this condition, the amplitude of the signal at the frequency kΔf, R f represents the frequency resolution, Δf represents the sampling frequency interval, and n p represents the number of sampling points in the frequency band p, |S(kΔf,R f )| represents the S(kΔf,R f ) takes the absolute value;

[0016] S2.3: Record the energy value in each frequency band to form a frequency band energy distribution diagram.

[0017] Specifically, the specific steps of step S3 include:

[0018] S3.1: Set the threshold discrimination parameter h1 to obtain the energy values ​​of the reference signal and the signal to be detected in each frequency band and

[0019] S3.2: Calculate the excess ratio of the energy of the signal to be detected to the energy of the reference signal

[0020] If β>h1, it is considered that the signal to be detected exceeds the reference signal in this frequency band;

[0021] If β≤h1, it is considered that the signal to be detected does not exceed the reference signal in this frequency band;

[0022] S3.3: Record the exceedance result, which includes the exceedance frequency band and the exceedance ratio.

[0023] Specifically, the specific steps of step S4 include:

[0024] S4.1: Get the exceedance result and set the event discrimination threshold h low and h high ;

[0025] S4.2: Set the exceedance ratio β on each exceedance frequency band p The event discrimination threshold h is set low and h high Make comparisons;

[0026] If β p ≤h low , the exceedance ratio is divided into a low interval, and the event category L corresponding to the signal to be detected in the frequency band is judged to be vibration or normal state, and L = 0;

[0027] If hlow <β p <h high , the exceedance ratio is divided into the middle interval, and the event category corresponding to the signal to be detected in the frequency band is judged to be noise or abnormal state, and L = 1;

[0028] If β p ≥h high , the exceedance ratio is divided into a high interval, and the event category corresponding to the signal to be detected in the frequency band is judged to be an impact or fault state, and L = 2;

[0029] S4.3: traverse the exceedance ratio judgment results on all frequency bands, for each frequency band, determine the event category state corresponding to the signal to be detected on the frequency band, and record the event category state on each frequency band;

[0030] S4.4: Organize the sorted signal event category states according to sampling batches and frequency bands to form a multi-channel multi-batch event matrix.

[0031] Specifically, the specific steps of S4.4 include:

[0032] S4.41: Obtain the event category status of the signal to be detected, and determine the number of rows and columns of the event matrix according to the number of sampling batches Z and the number of frequency bands N1;

[0033] S4.42: Create an empty event matrix in, represents the set of real numbers;

[0034] S4.43: traverse each sampling batch and each frequency band, and fill the corresponding event category state value into the corresponding position of the matrix;

[0035] S4.44: Verify the filled event matrix and output the multi-channel multi-batch event matrix M e ′.

[0036] Specifically, the specific steps of step S5 include:

[0037] S5.1: Get the multi-channel multi-batch event matrix M in S4.44 e ′, using the independent channel anomaly detection method core1 = (m i -μ)×σ performs anomaly detection on the event matrix of each frequency band, calculates the anomaly score of each matrix element, and identifies abnormal events in a single channel based on the anomaly score. At the same time, the anomaly detection results of independent channels are recorded, where core1 represents the anomaly score of each matrix element, m i Indicates M e ′, μ represents the matrix element m iThe mean of i The variance of

[0038] S5.2: Use a continuous multi-channel anomaly detection method to perform anomaly detection on the event matrix of multiple frequency bands, identify anomaly patterns or trends among multiple channels, and record the multi-channel anomaly detection results;

[0039] S5.3-1: Based on the multi-channel anomaly detection results, the Spearman rank correlation coefficient is used to calculate the correlation of events between different channels, and the cross-correlation function is used according to the size of the correlation coefficient and the signal propagation speed. Calculate the time delay of different channels Among them, CCF uv represents the cross-correlation value between channel u and channel v, max(CCF uv ) indicates the CCF uv Take the maximum value, represents max(CCF uv ) corresponding to the delay value, represents the weight function, X1(w) represents the result of Fourier transform of the event value of channel u at time t, represents the complex conjugate of X2(w), X2(w) represents the result of Fourier transform of the event value of channel v at time t, W1(w) and W2(w) represent the window signals added to X1(w) and X2(w), τ represents the time delay value, j represents the imaginary unit, w represents the angular frequency, w1 and w2 represent the upper and lower limits of the frequency range of the integration, e -jwτ represents the phase factor;

[0040] S5.4-1: Determine the sequence of events between different channels based on the calculated delay values;

[0041] If τ uv ≥0, it means that the event occurs on channel u before on channel v;

[0042] If τ uv <0, it means that the event occurs on channel v after on channel u;

[0043] S5.5-1: Calculate the actual delay value τ between the events on channel u and channel v uv Synchronization delay value B uv The difference between G uv =τ uv -B uv , and set the difference threshold h2;

[0044] If G uv ≤h2, the channels are considered to be synchronized;

[0045] If Guv >h2, it is considered that there is a synchronization problem;

[0046] S5.6-1: Record the results of the correlation analysis in S5.3-1 and the results of the delay calculation in S5.5-1, and display them in a chart.

[0047] Specifically, the specific steps of step S5 include:

[0048] S5.1: Get the multi-channel multi-batch event matrix M in S4.44 e ′, using the independent channel anomaly detection method core1 = (m i -μ)×σ performs anomaly detection on the event matrix of each frequency band, calculates the anomaly score of each matrix element, and identifies abnormal events in a single channel based on the anomaly score. At the same time, the anomaly detection results of independent channels are recorded, where core1 represents the anomaly score of each matrix element, m i Indicates M e ′, μ represents the matrix element m i The mean of i The variance of

[0049] S5.2: Use a continuous multi-channel anomaly detection method to perform anomaly detection on the event matrix of multiple frequency bands, identify anomaly patterns or trends among multiple channels, and record the multi-channel anomaly detection results;

[0050] S5.3-2: Based on the multi-channel anomaly detection results, select N sensors that simultaneously detect abnormal events of the signal, use the Spearman rank correlation coefficient to calculate the correlation of the detection signals between different sensors, and use the cross-correlation function according to the size of the correlation coefficient i∈[1,N-1],N≥2Calculate the delay between adjacent sensors in, Represents the cross-correlation value of detection signals between different sensors, express The corresponding delay value is Represents the vibration signal x measured by the i-th sensor i The result after Fourier transformation is: express The complex conjugate of Represents the vibration signal x measured by the i+1th sensor i+1 The result after Fourier transformation is: represents the weight function, W i (w) and W i+1 (w) indicates adding and Window signal on;

[0051] S5.4-2: Based on the delay value between adjacent sensors Find the position q,q∈i where the delay value changes from positive to negative or from negative to positive, and obtain the positive and negative intersection of the delay;

[0052] S5.5-2: If Then select the qth and q+1th sensors as positioning sensors, where: represents the time delay between the qth and q+1th sensors;

[0053] like Then the qth and q+2th sensors are selected as positioning sensors;

[0054] like Then the first and second sensors are selected as positioning sensors;

[0055] like Then the N-1th and Nth sensors are selected as positioning sensors;

[0056] S5.6-2: Record the sensor selection results and display them in a chart.

[0057] Specifically, the specific steps of step S5 also include:

[0058] S5.7: according to the calculation results of the correlation and delay of events between different channels, use the built-in optimization rule to find the best two channels and obtain the optimal channel combination, wherein the optimization rule is set in such a way that the channel combination with the highest correlation and the smallest delay is selected as the optimal channel combination;

[0059] S5.8: Based on the optimal channel combination, the abnormal event is located and estimated, and the specific location information of the abnormal event is calculated. Based on the location estimation, the built-in optimization rule is applied again to optimize the parameters of the location estimation to obtain the optimized multi-channel anomaly detection result;

[0060] S5.9: Integrate the location information of abnormal events and abnormal patterns identified in the independent channel abnormal detection results in S5.1 and the optimized multi-channel abnormal detection results in S5.5-1, and generate a record table containing the location information of all abnormal events and abnormal patterns;

[0061] S5.10: Generate an early warning report or early warning signal based on the record table information, and output the early warning result to the monitoring system, operator or decision-making department.

[0062] Specifically, the continuous multi-channel anomaly detection method in S5.2 is achieved by loading a pre-trained deep learning model and using the pre-trained deep learning model to perform anomaly detection on event matrices of multiple frequency bands; each channel in the multi-channel is connected to N sensors.

[0063] An electronic device comprises a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of a signal recognition and positioning method based on a point-type high-sensitivity optical acoustic vibration detector are implemented.

[0064] Compared with the prior art, the present invention has the following beneficial effects:

[0065] 1. The present invention proposes a signal recognition and positioning method based on a point-type high-sensitivity optical acoustic vibration detector. The point-type high-sensitivity optical acoustic vibration detector is used to perform refined signal collection and analysis on the potential leakage area, and the frequency band energy of the signal is calculated using the frequency band energy detection method. In combination with a preset threshold, the exceedance ratio is calculated, thereby achieving accurate recognition and classification of leakage signals, which not only improves the sensitivity of leakage detection, but also can effectively distinguish different types of leakage events.

[0066] 2. The present invention proposes a signal recognition and positioning method based on a point-type high-sensitivity optical acoustic vibration detector, constructs a multi-channel and multi-batch event matrix, and adopts an independent channel and continuous multi-channel anomaly detection method, which can realize all-round and multi-level monitoring and analysis of leakage signals, not only improving the accuracy of anomaly detection, but also quickly locating the source of leakage, timely outputting early warning results, and effectively reducing the risk of safety accidents caused by leakage. BRIEF DESCRIPTION OF THE DRAWINGS

[0067] Figure 1 It is a schematic diagram of a signal recognition and positioning method based on a point-type high-sensitivity optical acoustic vibration detector of the present invention;

[0068] Figure 2 It is a flow chart of the leakage warning algorithm in the signal recognition and positioning method based on the point-type high-sensitivity optical acoustic vibration detector of the present invention;

[0069] Figure 3 The present invention is a schematic diagram of the arrangement of point-type high-sensitivity optical acoustic vibration detectors in the signal recognition and positioning method based on point-type high-sensitivity optical acoustic vibration detectors. DETAILED DESCRIPTION

[0070] Example 1

[0071] See also Figure 1 The present invention provides an embodiment: a signal recognition and positioning method based on a point-type high-sensitivity optical acoustic vibration detector, comprising the following steps:

[0072] Step S1: using a point-type high-sensitivity optical acoustic vibration detector to collect signals from a potential leakage area, obtaining signal data of the potential leakage area, and dividing the signal data of the potential leakage area into a reference signal and a signal to be detected;

[0073] Furthermore, the specific steps of step S1 include:

[0074] (1) Equipment preparation: Ensure that the point-type high-sensitivity optical acoustic vibration detector is functioning normally, has sufficient power, and has been calibrated according to the instructions for use. At the same time, check whether the probe of the detector is clean and undamaged to ensure that the acoustic vibration signal can be accurately collected;

[0075] (2) On-site layout: Place the detector in the potential leakage area to ensure that the probe can fully receive the acoustic vibration signal from the leakage source, and adjust the angle and height of the detector according to the on-site conditions to obtain the best signal collection effect;

[0076] (3) Signal acquisition: Start the detector and begin collecting signals from potential leakage areas;

[0077] (4) Data recording: The collected signal data of the potential leakage area is stored in the detector's built-in storage device, or transmitted to a remote server in real time for storage and analysis to ensure that the data records are complete and accurate;

[0078] (5) The collected data is divided into reference signals and signals to be detected. The reference signal is usually the background signal in the absence of leakage, and the signal to be detected is the signal to be detected, which contains possible leakage information.

[0079] Step S2: Analyze the reference signal and the signal to be detected by using a frequency band energy detection method, and calculate the frequency band energy of the reference signal and the signal to be detected;

[0080] Step S3: Calculate the exceedance ratio of the signal to be detected relative to the reference signal in each frequency band according to the frequency band energy of the reference signal and the signal to be detected, combined with the preset threshold discrimination parameter, and record the exceedance result of the signal to be detected and the reference signal;

[0081] Step S4: discriminating the event category of the signal to be detected according to the exceeding result, and dividing the signal to be detected data according to the discriminating result to form a multi-channel multi-batch event matrix;

[0082] Each element of the multi-channel multi-batch event matrix represents a specific frequency band and the event status on the frequency band.

[0083] Step S5: Based on the multi-band multi-batch event matrix, anomaly detection is performed using an independent channel anomaly detection method and a continuous multi-channel anomaly detection method, the positioning information of the anomaly detection is recorded, and an early warning result is output according to the positioning information of the anomaly detection.

[0084] Example 2

[0085] See also Figure 2 and Figure 3 ,exist Figure 3 In the process, point-type high-sensitivity optical acoustic vibration detectors are arranged at intervals on the pipeline to monitor the vibration signals of the pipeline, such as Figure 3 As shown by the arrow in , when a pipeline leaks, the vibration caused by the leakage is transmitted along the pipeline to one or more nearby highly sensitive optical acoustic vibration detectors. By analyzing the vibration signal received by the detector, it is determined whether a leak has occurred. The specific steps of step S2 in this embodiment include:

[0086] S2.1: Obtain reference signal and signal to be detected, and perform preprocessing;

[0087] S2.2: According to the analysis requirements, the frequency range of the reference signal and the signal to be detected is divided into N1 frequency bands, and the formula is used The energy of the reference signal and the signal to be detected in each frequency band is calculated, where E p represents the energy in frequency band p, w k represents the weight factor of the kth sampling point, α k represents the noise suppression factor of the kth sampling point, S(kΔf,R f ) indicates that the frequency resolution R f Under this condition, the amplitude of the signal at the frequency kΔf, R f represents the frequency resolution, Δf represents the sampling frequency interval, and n p represents the number of sampling points in the frequency band p, |S(kΔf,R f )| represents the S(kΔf,R f ) takes the absolute value;

[0088] It should be noted that the weight factor w is introduced in the present invention. k and the noise suppression factor α k , making the energy calculation more flexible and able to be optimized for specific needs, while taking into account the frequency resolution R f The impact on signal amplitude, thereby improving the accuracy and practicality of frequency band energy calculation.

[0089] S2.3: Record the energy value in each frequency band to form a frequency band energy distribution diagram.

[0090] The specific steps of step S3 include:

[0091] S3.1: Set the threshold discrimination parameter h1 to obtain the energy values ​​of the reference signal and the signal to be detected in each frequency band and

[0092] S3.2: Calculate the excess ratio of the energy of the signal to be detected to the energy of the reference signal

[0093] If β>h1, it is considered that the signal to be detected exceeds the reference signal in this frequency band;

[0094] If β≤h1, it is considered that the signal to be detected does not exceed the reference signal in this frequency band;

[0095] S3.3: Record the exceedance result, which includes the exceedance frequency band and the exceedance ratio.

[0096] The specific steps of step S4 include:

[0097] S4.1: Get the exceedance result and set the event discrimination threshold h low and h high ;

[0098] S4.2: Set the exceedance ratio β on each exceedance frequency band p The event discrimination threshold h is set low and h high Make comparisons;

[0099] If β p ≤h low , the exceedance ratio is divided into a low interval, and the event category L corresponding to the signal to be detected in the frequency band is judged to be vibration or normal state, and L = 0;

[0100] If h low <β p <h high , the exceedance ratio is divided into the middle interval, and the event category corresponding to the signal to be detected in the frequency band is judged to be noise or abnormal state, and L = 1;

[0101] If β p ≥h high , the exceedance ratio is divided into a high interval, and the event category corresponding to the signal to be detected in the frequency band is judged to be an impact or fault state, and L = 2;

[0102] S4.3: traverse the exceedance ratio judgment results on all frequency bands, for each frequency band, determine the event category state corresponding to the signal to be detected on the frequency band, and record the event category state on each frequency band;

[0103] S4.4: Organize the sorted signal event category states according to sampling batches and frequency bands to form a multi-channel multi-batch event matrix.

[0104] The specific steps of S4.4 include:

[0105] S4.41: Obtain the event category status of the signal to be detected, and determine the number of rows and columns of the event matrix according to the number of sampling batches Z and the number of frequency bands N1;

[0106] S4.42: Create an empty event matrix in, represents the set of real numbers;

[0107] S4.43: traverse each sampling batch and each frequency band, and fill the corresponding event category state value into the corresponding position of the matrix;

[0108] S4.44: Verify the filled event matrix and output the multi-channel multi-batch event matrix M e ′.

[0109] The specific steps of step S5 include:

[0110] S5.1: Get the multi-channel multi-batch event matrix M in S4.44 e ′, using the independent channel anomaly detection method core1 = (m i -μ)×σ performs anomaly detection on the event matrix of each frequency band, calculates the anomaly score of each matrix element, and identifies abnormal events in a single channel based on the anomaly score. At the same time, the anomaly detection results of independent channels are recorded, where core1 represents the anomaly score of each matrix element, m i Indicates M e ′, μ represents the matrix element m i The mean of i The variance of

[0111] The larger the core1 score is, the more abnormal the matrix element is.

[0112] S5.2: Use a continuous multi-channel anomaly detection method to perform anomaly detection on the event matrix of multiple frequency bands, identify anomaly patterns or trends among multiple channels, and record the multi-channel anomaly detection results;

[0113] Furthermore, the specific steps of S5.2 include:

[0114] S5.21: Obtain event matrices of multiple frequency bands, which contain signal data collected from each channel;

[0115] S5.22: Load the pre-trained deep learning model, input the multiple frequency band event matrices to be detected into the trained deep learning model, and the pre-trained deep learning model will output the anomaly score or label of each channel in each frequency band;

[0116] S5.23: Based on anomaly scores or labels, identify anomaly patterns or trends among multiple channels, including analyzing the distribution of anomaly signals in time and space, as well as their correlation and mutual influence;

[0117] S5.24: Record the multi-channel anomaly detection results, including the location, intensity, duration and other information of the abnormal signal.

[0118] S5.3-1: Based on the multi-channel anomaly detection results, the Spearman rank correlation coefficient is used to calculate the correlation of events between different channels, and the cross-correlation function is used according to the size of the correlation coefficient and the signal propagation speed. Calculate the time delay of different channels Among them, CCF uv represents the cross-correlation value between channel u and channel v, max(CCF uv ) indicates the CCF uv Take the maximum value, represents max(CCF uv ) corresponding to the delay value, represents the weight function, X1(w) represents the result of Fourier transform of the event value of channel u at time t, represents the complex conjugate of X2(w), X2(w) represents the result of Fourier transform of the event value of channel v at time t, W1(w) and W2(w) represent the window signals added to X1(w) and X2(w), τ represents the time delay value, j represents the imaginary unit, w represents the angular frequency, w1 and w2 represent the upper and lower limits of the frequency range of the integration, e -jwτ represents the phase factor, and the Spearman rank correlation coefficient is the prior art content in this field, which is not the inventive solution of the present application and will not be described in detail here;

[0119] In the present invention, the cross-correlation function is used to further estimate the time delay between channels after determining the correlation between them.

[0120] S5.4-1: Determine the sequence of events between different channels based on the calculated delay values;

[0121] If τ uv ≥0, it means that the event occurs on channel u before on channel v;

[0122] If τ uv <0, it means that the event occurs on channel v after on channel u;

[0123] S5.5-1: Calculate the actual delay value τ between the events on channel u and channel v uv Synchronization delay value B uv The difference between G uv=τ uv -B uv , and set the difference threshold h2;

[0124] If G uv ≤h2, the channels are considered to be synchronized;

[0125] If G uv >h2, it is considered that there is a synchronization problem;

[0126] S5.6-1: Record the results of the correlation analysis in S5.3-1 and the results of the delay calculation in S5.5-1, and display them in a chart, as shown in Table 1, assuming that the number of sensors N=4.

[0127] Table 1 Multi-channel multi-batch event matrix record table

[0128] Sensor 1 Sensor 2 Sensor 3 Sensor 4 Band 1 no yes yes yes Band 2 no no yes no Band 3 no no no no

[0129] As can be seen from Table 1, it records whether each detector in each frequency band has detected any abnormal event. For example, in frequency band 1 in Table 1, sensor 2, sensor 3, and sensor 4 all detected abnormal events, which is a continuous multi-channel abnormality; in frequency band 2 in Table 1, only sensor 3 detected an abnormal event, which is an independent channel abnormality; in frequency band 3 in Table 1, sensors 1-sensor 4 did not detect any abnormal events, indicating that the channel is normal.

[0130] Similarly, the specific steps of step S5 include:

[0131] S5.1: Get the multi-channel multi-batch event matrix M in S4.44 e ′, using the independent channel anomaly detection method core1 = (m i -μ)×σ performs anomaly detection on the event matrix of each frequency band, calculates the anomaly score of each matrix element, and identifies abnormal events in a single channel based on the anomaly score. At the same time, the anomaly detection results of independent channels are recorded, where core1 represents the anomaly score of each matrix element, m i Indicates M e ′, μ represents the matrix element m i The mean of i The variance of

[0132] S5.2: Use a continuous multi-channel anomaly detection method to perform anomaly detection on the event matrix of multiple frequency bands, identify anomaly patterns or trends among multiple channels, and record the multi-channel anomaly detection results;

[0133] S5.3-2: Based on the multi-channel anomaly detection results, select N sensors that simultaneously detect the abnormal events of the signal. Assuming that the number of sensors is N = 5, the location deployment of the sensors is shown in Table 2. The Spearman rank correlation coefficient is used to calculate the correlation of the detection signals between different sensors, and the cross-correlation function is used according to the size of the correlation coefficient. i∈[1,N-1],N≥2Calculate the delay between adjacent sensors in, Represents the cross-correlation value of detection signals between different sensors, express The corresponding delay value is Represents the vibration signal x measured by the i-th sensor i The result after Fourier transformation is: Represents the vibration signal x measured by the i+1th sensor i+1 The result after Fourier transformation is: express The complex conjugate of represents the weight function, W i (w) and W i+1 (w) indicates adding and Window signal on;

[0134] Table 2 Sensor location diagram

[0135] Sensor No. Sensor location (m) 1 0 2 200 3 400 4 600 5 800

[0136] As can be seen from Table 2, the sensor takes every 200 meters as the position interval.

[0137] S5.4-2: Based on the delay value between adjacent sensors Find the position q,q∈i where the delay value changes from positive to negative or from negative to positive, and obtain the positive and negative intersection of the delay;

[0138] S5.5-2: If Then select the qth and q+1th sensors as positioning sensors, where: represents the time delay between the qth and q+1th sensors;

[0139] like Then the qth and q+2th sensors are selected as positioning sensors;

[0140] like Then the first and second sensors are selected as positioning sensors;

[0141] like Then the N-1th and Nth sensors are selected as positioning sensors;

[0142] S5.6-2: Record the sensor selection results and display them in a chart, as shown in Table 3, assuming that the number of sensors is N=5.

[0143] Table 3 Sensor delay diagram

[0144]

[0145]

[0146] As can be seen from Table 3, when the number of sensors is N=5, the delay values ​​of all sensors are positive when the distance is 0-100 meters, that is, Therefore, the first and second sensors are selected as cross-correlation sensors; at 90-110 meters, because the 1,2 delay is zero, that is, Therefore, q = 1, so the first and second sensors are selected as cross-correlation sensors; when the distance is 100 to 300 meters, because the latter sensor covers the former, q = 1, and Therefore, the first and third sensors are selected as cross-correlation sensors; at 290-310 meters, because the delays 2 and 3 are zero, the positive and negative intersection positions of the delays are reselected, so q = 2, and Therefore, the second and third sensors are selected as cross-correlation sensors; at 300-500 meters, similarly, q = 2, and Therefore, the second and fourth sensors are selected as cross-correlation sensors; at 490-510 meters, because the delays of 3 and 4 are zero, similarly, q = 3, so Therefore, the third and fourth sensors are selected as cross-correlation sensors; at 690-710 meters, because the delay of 4 and 5 is zero, similarly, q = 4, so Therefore, the 4th and 5th sensors are selected as cross-correlation sensors; when the distance is 700-800 meters, the delay values ​​of all sensors are negative, that is, Therefore, the 4th and 5th sensors are selected as cross-correlation sensors.

[0147] The specific steps of step S5 also include:

[0148] S5.7: Based on the calculation results of the correlation and delay of events between different channels, use the built-in optimization rule to find the best two channels and obtain the optimal channel combination;

[0149] Furthermore, the specific steps of S5.7 include:

[0150] S5.71: Collect event data from different channels, ensure the accuracy and completeness of the data, and pre-process the data, such as denoising and filtering, to improve data quality;

[0151] S5.72: Obtain the time delay of events between different channels from S5.6 to obtain a delay matrix, which represents the delay between each channel;

[0152] S5.73: According to the actual application requirements, the optimization rule is set. In the present invention, the channel combination with the highest correlation and the smallest delay is selected as the optimal channel combination;

[0153] S5.74: Use a built-in optimization algorithm based on a genetic algorithm to search for an optimal solution in the channel combination space, and evaluate the comprehensive performance of each channel combination according to the set optimization rule, wherein the genetic algorithm is a prior art content in this field and is not an inventive solution of the present application, and is not described in detail here;

[0154] S5.75: From the searched channel combinations, select a combination with the best comprehensive performance as the optimal channel combination.

[0155] S5.8: Based on the optimal channel combination, the abnormal event is located and estimated, and the specific location information of the abnormal event is calculated. Based on the location estimation, the built-in optimization rule is applied again to optimize the parameters of the location estimation to obtain the optimized multi-channel anomaly detection result;

[0156] Furthermore, the specific steps of S5.8 include:

[0157] S5.81: Based on the signals received by the optimal channel combination, the abnormal event is preliminarily located using the time difference positioning method, and the accuracy of the preliminary positioning result is analyzed to determine the positioning parameters that need to be optimized, such as time difference, angle, and distance. The time difference positioning method is a prior art content in this field and is not an inventive solution of the present application, and is not described in detail here;

[0158] S5.82: Prepare resources required for optimization, such as computing resources and model parameters;

[0159] S5.83: Taking the preliminary positioning result as the initial value and the positioning error as the optimization target, the positioning parameters are optimized using a genetic algorithm, wherein the genetic algorithm is a prior art in this field and is not an inventive solution of the present application, and is not described in detail here;

[0160] S5.84: Recalculate the specific location information of the abnormal event based on the optimized positioning parameters, and evaluate the accuracy of the optimized positioning results to ensure that they meet actual application requirements;

[0161] S5.85: Fuse the optimized positioning result with the information of other channels to form a multi-channel anomaly detection result, and output the detection result.

[0162] S5.9: Integrate the location information of abnormal events and abnormal patterns identified in the independent channel abnormal detection results in S5.1 and the optimized multi-channel abnormal detection results in S5.5-1, and generate a record table containing the location information of all abnormal events and abnormal patterns;

[0163] S5.10: Generate an early warning report or early warning signal based on the record table information, and output the early warning result to the monitoring system, operator or decision-making department.

[0164] Furthermore, the specific steps of S5.10 include:

[0165] (1) Read monitoring data from the record table;

[0166] (2) Matching monitoring data with preset warning rules;

[0167] (3) Determine whether there is an abnormal situation. If an abnormal situation is found, generate warning information according to the preset warning level and type;

[0168] (4) Generate warning reports or warning signals based on the warning information. The warning reports should include the warning time, warning level, warning type, abnormal data, possible impacts and recommended response measures; the warning signals should be presented in an intuitive and easy-to-understand manner, such as color coding and sound prompts;

[0169] (5) Output warning reports or warning signals to monitoring systems, operators and decision-making departments.

[0170] The continuous multi-channel anomaly detection method in S5.2 is achieved by loading a pre-trained deep learning model and using the pre-trained deep learning model to perform anomaly detection on event matrices of multiple frequency bands; each channel in the multi-channel is connected to N sensors.

[0171] Example 3

[0172] An electronic device includes a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the steps of a signal recognition and positioning method based on a point-type high-sensitivity optical acoustic vibration detector are implemented. For details, please refer to the above method embodiment, which will not be repeated here.

[0173] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation modes, which are merely illustrative rather than restrictive. Under the guidance of the present invention, ordinary technicians in the field may also change, modify, replace and modify the above-mentioned embodiments without departing from the purpose and scope of protection of the present invention, and all of these are within the protection of the present invention.

Claims

1. A signal recognition and positioning method based on a point-type high-sensitivity optical acoustic vibration detector, characterized in that: include: Step S1: using a point-type high-sensitivity optical acoustic vibration detector to collect signals from a potential leakage area, obtaining signal data of the potential leakage area, and dividing the signal data of the potential leakage area into a reference signal and a signal to be detected; Step S2: Analyze the reference signal and the signal to be detected by using a frequency band energy detection method, and calculate the frequency band energy of the reference signal and the signal to be detected; Step S3: Calculate the exceedance ratio of the signal to be detected relative to the reference signal in each frequency band according to the frequency band energy of the reference signal and the signal to be detected, combined with the preset threshold discrimination parameter, and record the exceedance result of the signal to be detected and the reference signal; Step S4: discriminating the event category of the signal to be detected according to the exceeding result, and dividing the signal to be detected data according to the discriminating result to form a multi-channel multi-batch event matrix; Step S5: Based on the multi-channel multi-batch event matrix, anomaly detection is performed using an independent channel anomaly detection method and a continuous multi-channel anomaly detection method, the positioning information of the anomaly detection is recorded, and an early warning result is output according to the positioning information of the anomaly detection.

2. The signal recognition and positioning method based on a point-type high-sensitivity optical acoustic vibration detector according to claim 1, characterized in that: The specific steps of step S2 include: S2.1: Obtain reference signal and signal to be detected, and perform preprocessing; S2.2: According to the analysis requirements, the frequency range of the reference signal and the signal to be detected is divided into N1 frequency bands, and the formula is used The energy of the reference signal and the signal to be detected in each frequency band is calculated, where E p represents the energy in frequency band p, w k represents the weight factor of the kth sampling point, α k represents the noise suppression factor of the kth sampling point, S(kΔf,R f ) indicates that the frequency resolution R f Under this condition, the amplitude of the signal at the frequency kΔf, R f represents the frequency resolution, Δf represents the sampling frequency interval, and n p represents the number of sampling points in the frequency band p, |S(kΔf,R f )| represents the S(kΔf,R f ) takes the absolute value; S2.3: Record the energy value in each frequency band to form a frequency band energy distribution diagram.

3. The signal recognition and positioning method based on a point-type high-sensitivity optical acoustic vibration detector according to claim 2, characterized in that: The specific steps of step S3 include: S3.1: Set the threshold discrimination parameter h1 to obtain the energy values ​​of the reference signal and the signal to be detected in each frequency band and S3.2: Calculate the excess ratio of the energy of the signal to be detected to the energy of the reference signal If β>h1, it is considered that the signal to be detected exceeds the reference signal in this frequency band; If β≤h1, it is considered that the signal to be detected does not exceed the reference signal in this frequency band; S3.3: Record the exceedance result, which includes the exceedance frequency band and the exceedance ratio.

4. The signal recognition and positioning method based on a point-type high-sensitivity optical acoustic vibration detector according to claim 3 is characterized in that: The specific steps of step S4 include: S4.1: Get the exceedance result and set the event discrimination threshold h low and h high ; S4.2: Set the exceedance ratio β on each exceedance frequency band p The event discrimination threshold h is set low and h high Make comparisons; If β p ≤h low , the exceedance ratio is divided into a low interval, and the event category L corresponding to the signal to be detected in the frequency band is judged to be vibration or normal state, and L = 0; If h low <β p <h high , the exceedance ratio is divided into the middle interval, and the event category corresponding to the signal to be detected in the frequency band is judged to be noise or abnormal state, and L = 1; If β p ≥h high , the exceedance ratio is divided into a high interval, and the event category corresponding to the signal to be detected in the frequency band is judged to be an impact or fault state, and L = 2; S4.3: traverse the exceedance ratio judgment results on all frequency bands, for each frequency band, determine the event category state corresponding to the signal to be detected on the frequency band, and record the event category state on each frequency band; S4.4: Organize the sorted signal event category states according to sampling batches and frequency bands to form a multi-channel multi-batch event matrix.

5. The signal recognition and positioning method based on a point-type high-sensitivity optical acoustic vibration detector according to claim 4, characterized in that: The specific steps of S4.4 include: S4.41: Obtain the event category status of the signal to be detected, and determine the number of rows and columns of the event matrix according to the number of sampling batches Z and the number of frequency bands N1; S4.42: Create an empty event matrix in, represents the set of real numbers; S4.43: traverse each sampling batch and each frequency band, and fill the corresponding event category state value into the corresponding position of the matrix; S4.44: Verify the filled event matrix and output the multi-channel multi-batch event matrix M e ′.

6. The signal recognition and positioning method based on a point-type high-sensitivity optical acoustic vibration detector according to claim 5, characterized in that: The specific steps of step S5 include: S5.1: Get the multi-channel multi-batch event matrix M in S4.44 e ′, using the independent channel anomaly detection method core1 = (m i -μ)×σ performs anomaly detection on the event matrix of each frequency band, calculates the anomaly score of each matrix element, and identifies abnormal events in a single channel based on the anomaly score. At the same time, the anomaly detection results of independent channels are recorded, where core1 represents the anomaly score of each matrix element, m i Indicates M e ′, μ represents the matrix element m i The mean of i The variance of S5.2: Use a continuous multi-channel anomaly detection method to perform anomaly detection on the event matrix of multiple frequency bands, identify anomaly patterns or trends among multiple channels, and record the multi-channel anomaly detection results; S5.3-1: Based on the multi-channel anomaly detection results, the Spearman rank correlation coefficient is used to calculate the correlation of events between different channels, and the cross-correlation function is used according to the size of the correlation coefficient and the signal propagation speed. Calculate the time delay of different channels Among them, CCF uv Indicates the cross-correlation value between channel u and channel v, max(CCF uv ) indicates the CCF uv Take the maximum value, represents max(CCF uv ) corresponding to the delay value, represents the weight function, X1(w) represents the result of Fourier transform of the event value of channel u at time t, represents the complex conjugate of X2(w), X2(w) represents the result of Fourier transform of the event value of channel v at time t, W1(w) and W2(w) represent the window signals added to X1(w) and X2(w), τ represents the time delay value, j represents the imaginary unit, w represents the angular frequency, w1 and w2 represent the upper and lower limits of the frequency range of the integration, e -jwτ represents the phase factor; S5.4-1: Determine the sequence of events between different channels based on the calculated delay values; If τ uv ≥0, it means that the event occurs on channel u before on channel v; If τ uv <0, it means that the event occurs on channel v after on channel u; S5.5-1: Calculate the actual delay value τ between the events on channel u and channel v uv Synchronization delay value B uv The difference between G uv =τ uv -B uv , and set the difference threshold h2; If G uv ≤h2, the channels are considered to be synchronized; If G uv >h2, it is considered that there is a synchronization problem; S5.6-1: Record the results of the correlation analysis in S5.3-1 and the results of the delay calculation in S5.5-1, and display them in a chart.

7. The signal recognition and positioning method based on a point-type high-sensitivity optical acoustic vibration detector according to claim 5, characterized in that: The specific steps of step S5 include: S5.1: Get the multi-channel multi-batch event matrix M in S4.44 e ′, using the independent channel anomaly detection method core1 = (m i -μ)×σ performs anomaly detection on the event matrix of each frequency band, calculates the anomaly score of each matrix element, and identifies abnormal events in a single channel based on the anomaly score. At the same time, the anomaly detection results of independent channels are recorded, where core1 represents the anomaly score of each matrix element, m i Indicates M e ′, μ represents the matrix element m i The mean of i The variance of S5.2: Use a continuous multi-channel anomaly detection method to perform anomaly detection on the event matrix of multiple frequency bands, identify anomaly patterns or trends among multiple channels, and record the multi-channel anomaly detection results; S5.3-2: Based on the multi-channel anomaly detection results, select N sensors that simultaneously detect abnormal events of the signal, use the Spearman rank correlation coefficient to calculate the correlation of the detection signals between different sensors, and use the cross-correlation function according to the size of the correlation coefficient i∈[1,N-1],N≥2Calculate the delay between adjacent sensors in, Represents the cross-correlation value of detection signals between different sensors, express The corresponding delay value is Represents the vibration signal x measured by the i-th sensor i The result after Fourier transformation is: express The complex conjugate of Represents the vibration signal x measured by the i+1th sensor i+1 The result after Fourier transformation is: represents the weight function, W i (w) and W i+1 (w) indicates adding and The window signal on the top; τ represents the delay value, j represents the imaginary unit, w represents the angular frequency, w1 and w2 represent the upper and lower limits of the frequency range of the integration, e -jwτ represents the phase factor; S5.4-2: Based on the delay value between adjacent sensors Find the position q,q∈i where the delay value changes from positive to negative or from negative to positive, and obtain the positive and negative intersection of the delay; S5.5-2: If Then select the qth and q+1th sensors as positioning sensors, where: represents the time delay between the qth and q+1th sensors; like Then the qth and q+2th sensors are selected as positioning sensors; like Then the first and second sensors are selected as positioning sensors; like Then the N-1th and Nth sensors are selected as positioning sensors; S5.6-2: Record the sensor selection results and display them in a chart.

8. The signal recognition and positioning method based on a point-type high-sensitivity optical acoustic vibration detector according to claim 6, characterized in that: The specific steps of step S5 also include: S5.7: according to the calculation results of the correlation and delay of events between different channels, use the built-in optimization rule to find the best two channels and obtain the optimal channel combination, wherein the optimization rule is set in such a way that the channel combination with the highest correlation and the smallest delay is selected as the optimal channel combination; S5.8: Based on the optimal channel combination, the abnormal event is estimated and the specific location information of the abnormal event is calculated. Based on the location estimation, the built-in optimization rule is applied again to optimize the parameters of the location estimation to obtain the optimized multi-channel anomaly detection result; S5.9: Integrate the location information of abnormal events and abnormal patterns identified in the independent channel abnormal detection results in S5.1 and the optimized multi-channel abnormal detection results in S5.5-1, and generate a record table containing the location information of all abnormal events and abnormal patterns; S5.10: Generate an early warning report or early warning signal based on the record table information, and output the early warning result to the monitoring system, operator or decision-making department.

9. The signal recognition and positioning method based on a point-type high-sensitivity optical acoustic vibration detector according to claim 8, characterized in that: The continuous multi-channel anomaly detection method in S5.2 is achieved by loading a pre-trained deep learning model and using the pre-trained deep learning model to perform anomaly detection on event matrices of multiple frequency bands; each channel in the multi-channel is connected to N sensors.

10. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, wherein: When the processor executes the computer program, the steps of the signal recognition and positioning method based on a point-type high-sensitivity optical acoustic vibration detector according to any one of claims 1 to 9 are implemented.

Images