CN108488638A - Line leakage system and method based on sound wave suction wave hybrid monitoring - Google Patents

Abstract

The invention provides a pipeline leakage monitoring system and method based on acoustic negative pressure wave mixed monitoring, and relates to the technical field of pipeline risk prediction. The system includes pressure sensors, acoustic wave sensors, lower computer, switch and upper computer located at the beginning and end of the pipeline. The lower computer controls the pressure sensor and the acoustic wave sensor to collect pressure and sound wave data, and sends them to the upper computer through the switch after preprocessing. The computer executes the leakage monitoring program, receives and analyzes the pressure and sound wave data transmitted by the lower computer through the pressure and sound wave data integration storage module, and performs secondary filtering, dimensionless processing and semi-supervision fees on the acquired data through the data processing module Scheer discriminant processing, through the pipeline leakage monitoring module to determine whether the pipeline leaks and perform mixed positioning of pressure signals and acoustic signals. The invention can better shield noise interference, ensure the accuracy of the restored signal after filtering when the signal source changes, and more accurately locate the leakage point.

Description

Pipeline Leakage Monitoring System and Method Based on Acoustic Negative Pressure Wave Hybrid Monitoring

technical field

The invention relates to the technical field of pipeline risk prediction, in particular to a pipeline leakage monitoring system and method based on acoustic negative pressure wave hybrid monitoring.

Background technique

The role of pipeline transportation in economic development is becoming more and more important, such as urban water pipelines, land crude oil pipelines, submarine oil and gas pipelines, etc. Most of the oil transportation is transported in the pipeline in the form of refined oil. With the expansion of the pipeline network year by year, pipeline transportation has become the main mode of land oil and gas transportation. However, the aging, corrosion, sudden natural disasters and man-made damage of the pipeline will all cause the leakage or even rupture of the refined oil pipeline. If it is not discovered and stopped in time, it will not only cause energy waste, economic loss, environmental pollution, but also endanger personal safety. , and even cause catastrophic accidents. Therefore, it is of great significance to carry out real-time online monitoring of oil and gas pipelines, accurately and timely alarm for leakage accidents, and accurately estimate the location of leakage points.

There are many kinds of pipeline leak detection methods today, such as optical fiber leak detection method, acoustic wave detection method, pressure gradient method, negative pressure wave method and so on. Among them, the negative pressure wave method is the most widely used pipeline leakage detection method in the world in recent years. This method has the characteristics of short reaction time and wide range of detectable leakage. The internal pressure changes slowly, and the negative pressure wave method is less sensitive to it, which is prone to false positives. Moreover, due to the complex working condition adjustment of the pipeline transportation system, some common operations such as the start and stop of the main pump, the opening and closing of the valve, and the regulating valve Changes in the opening will cause negative pressure waves, which are highly similar to negative pressure waves caused by leakage, which reduces the detection accuracy of the negative pressure wave method.

Acoustic detection method leak detection has good monitoring accuracy for slow and small flow leakage, but it is easy to confuse the leakage with the abnormal sound wave caused by human or environmental factors, and this method cannot detect simultaneous multi-point leakage .

Contents of the invention

The technical problem to be solved by the present invention is to provide a pipeline leakage monitoring system and method based on acoustic negative pressure wave hybrid monitoring, which can better shield noise interference and ensure that when the signal source changes , The accuracy of the signal is restored after filtering, the false alarm is minimized, the leakage of small flow is more accurately judged, and the location of the leakage point is more accurate.

In order to solve the problems of the technologies described above, the technical solution adopted in the present invention is:

On the one hand, the present invention provides a pipeline leakage monitoring system based on acoustic negative pressure wave hybrid monitoring, including a pressure sensor, an acoustic wave sensor, a lower computer, a switch and an upper computer;

The pressure sensors are arranged one at the head end and the end of the pipeline respectively, and are in contact with the conveying medium for real-time collection of high-precision data reflecting the pressure of the pipeline; two acoustic wave sensors are respectively arranged at the head end and the end of the pipeline, and Placed on both sides of the pressure sensor at the head end and the end, it is used to collect high-speed data reflecting the sound wave of the pipeline in real time; the distance between the two sound wave sensors at the head end or the end is not less than 20Tv, T is the period of the sound wave, and v is the sound wave in the pipeline speed of propagation;

The lower computer, the switch and the upper computer are respectively provided with a group at the head end and the end of the pipeline, and the pressure sensor and the acoustic wave sensor are all connected to the lower computer at the corresponding end, and the lower computer is connected to the The host computer at the corresponding end is connected;

The lower computer is programmed to perform the following steps: control the pressure sensor and the acoustic wave sensor at the corresponding end to collect pressure and acoustic wave data, and preprocess the collected data, including filtering and time stamping according to the GPS collection time; After the data is packaged, it is sent to the upper computer through the switch, and stored in the SD card of the lower computer for backup;

The host computer includes a computer-executable leakage monitoring program, which specifically includes: a pressure and sound wave data integration storage module, a data processing module, and a pipeline leakage monitoring module;

The integrated storage module of pressure and acoustic wave data is used to receive and analyze the pressure and acoustic wave data transmitted by the lower computer, store them according to the data detection time, and combine a set of pressure data and two sets of acoustic wave data at one end of each pipeline into one third-order matrix;

The data processing module is used to obtain data from the pressure and acoustic wave data integration storage module for processing and calculation, including secondary filtering of data, dimensionless processing, and semi-supervised Fisher discriminant processing to obtain data processing and calculation results; , in the secondary filtering process of the data, the sound wave and negative pressure wave signals are decomposed by the improved complete overall empirical mode decomposition ICEEMD, and the sound wave and negative pressure wave signals are filtered out according to the approximate entropy of each decomposed component after calculation The noise in the medium, and remove the acoustic interference signal, to obtain the final filtered signal; in the dimensionless processing of the data after the secondary filtering, the semi-supervised Fisher discriminant is trained through the collected historical pressure and acoustic data, and respectively obtained A database belonging to four working conditions: normal working condition signal, large leakage signal, small leakage signal and working condition adjustment signal;

The pipeline leakage monitoring module is used to use the real-time data of pressure and sound wave to judge whether the pipeline leaks according to the data processing and calculation results of the pressure and sound wave data integration module and the data processing module. When it is judged as a large or small leakage signal, pass The pressure signal and the sound wave signal are mixed to calculate the leakage distance and locate the pipeline leakage point.

On the other hand, the present invention also provides a pipeline leakage monitoring method based on acoustic negative pressure wave hybrid monitoring, which is realized by the above-mentioned monitoring system, and the method includes the following steps:

Step 1: The lower computer at the beginning and end of the pipeline controls the corresponding pressure sensor and acoustic wave sensor to collect pipeline pressure and acoustic wave data, including pressure data and sonic data And filter the collected data to get the filtered pressure signals X ₁ , X ₂ and acoustic wave signals Y ₁₁ , Y ₁₂ , Y ₂₁ , Y ₂₂ ; time stamp;

Step 2: The lower computer packs and sends the filtered and time-stamped pressure and real-time acoustic signals to the upper computer, and stores them in the SD card of the lower computer for backup;

Step 3: The upper computer receives the data sent by the lower computer, determines the required data packet by identifying the header and tail of the data packet, then analyzes the data type and data value in the data packet, and finally stores it in the database of the upper computer ;

Step 4: The upper computer performs secondary filtering on the received data, including the following steps:

Step 4.1: Decompose the sound wave and negative pressure wave signals by using the improved complete overall empirical mode decomposition, namely ICEEMD, to obtain the decomposed eigenmode function components;

Step 4.2: Calculate the approximate entropy of each eigenmode function component, and judge the noise signals in the pressure signals X ₁ , X ₂ and the acoustic wave signals Y ₁₁ , Y ₁₂ , Y ₂₁ , and Y ₂₂ according to the approximate entropy of each component and filter them out ;

Step 4.3: remove the acoustic wave interference from the upstream station and the downstream station in the acoustic wave signal, and obtain the signal after secondary filtering;

Step 5: Carry out standard deviation method dimensionless processing on the signal after secondary filtering, and train semi-supervised Fisher discrimination through the collected historical pressure and acoustic wave data, and obtain signals belonging to normal working conditions, large leak signals, and small leaks respectively The database of the four working conditions of the signal and the working condition adjustment signal;

Step 6: Collect pipeline pressure and acoustic wave data in real time, perform secondary filtering through preprocessing and the secondary filtering method described in step 4, and send it to the semi-supervised Fisher discriminant model obtained in step 5 training in seconds, Comparing the obtained eigenvectors with the established database of four working conditions, and calculating the degree of similarity between the two;

Step 7: Determine the fault type of the real-time data according to the degree of similarity. If it is determined to be a large leak signal or a small leak signal, calculate the position of the leak point by mixing the pressure signal and the acoustic wave signal, and locate the pipeline leak. Otherwise, return to step 6 and continue real-time monitoring.

In the step 4.1, the specific process of decomposing the sound wave and the negative pressure wave signal by using the improved complete overall empirical mode decomposition is as follows:

Step 4.1.1: Add I group of Gaussian white noise to the signal X(t) to generate I new signals, the i-th new signal is ^Xi (t)=X(t)+β _k w ⁱ ; where, X (t) is the original sound wave or negative pressure wave signal; w ⁱ is the i-th group of Gaussian white noise variables, i=1, 2, ..., I; β _k =ε _k std(r _k ), r _k is the k-th Remainder, ε _k = 0.2;

Step 4.1.2: Initialize cycle parameters, let i=1;

Step 4.1.3: Determine all extreme points on the signal X ⁱ (t), including local maximum points and local minimum points; fit the extreme points to obtain the upper envelope and the lower envelope Let X ⁱ (t) satisfy:

Step 4.1.4: Calculate the average value of the upper and lower envelopes of Xi(t), denoted as m(t),

Step 4.1.5: Subtract the average value m(t) from the original ^Xi (t) signal data to obtain the function _hi (t), that is, _hi (t)= ^Xi (t)-m(t);

Step 4.1.6: Set X ⁱ (t) = h _i (t), and judge whether h _i (t) satisfies the two conditions of the intrinsic mode function IMF:

Condition 1: At any point in time, the mean value of the envelope defined by the local maximum and local minimum of the IMF must be zero, that is, for any t, we have

Condition 2: In the entire data sequence, the number of extreme points and the number of zero-crossing points are equal or at most equal to one;

If satisfied, execute step 4.1.7, otherwise, return to step 4.1.3;

Step 4.1.7: Separate _hi (t) from the X(t) signal, which is the IMF _i component, and obtain a function r _i (t) that removes the difference of the high-frequency component, that is, the remainder r _i ( t)=X(t)-IMF _i ;

Step 4.1.8: Judging whether the residual signal r _i (t) is a signal of a monotone function, if so, then X(t) can no longer be decomposed into IMF components, and the decomposition process is completed to obtain each IMF component, which is recorded as IMF ₁ ~ IMFn, n=i, and the final residual signal is r _n (t)=r _nn - ₁ (t)-h _n (t), and the signal X(t) is n eigenmode function components and a residual term and, namely Otherwise, let X(t)=r _i be a new signal, and let i=i+1, return to step 4.1.3, and perform the next decomposition.

The specific process of calculating the approximate entropy of each eigenmode function component in the described step 4.2 is:

Step 4.2.1: Consider the data in the jth eigenmode function as a time series with y points, recorded as {Z _j }={Z _j1 , Z _j2 ,..., Z _jy }; for a given Threshold a and pattern dimension g, Calculate the n*n binary distance matrix B:

in,

Step 4.2.2: Use the elements b _rj of matrix B to calculate the ratios of the number of elements less than a to the total number of distances n-g+1 and n-g+2, respectively recorded as and As shown in the following two formulas:

Step 4.2.3: According to and calculated and As shown in the following two formulas,

Then the approximate entropy A _j of the jth eigenmode function is expressed as:

Step 4.2.4: Arrange the obtained approximate entropy values from large to small, and the corresponding IMF components with approximate entropy values greater than 1 are regarded as noise signals to be filtered out. The IMF components with the approximate entropy of the acoustic wave signal between 0.486 and 0.490 are The coupling signal of the sound wave and the pressure signal is filtered out as an interference signal.

In the step 4.3, the specific process of removing the acoustic interference from the upstream station and the downstream station in the acoustic signal is as follows:

Step 4.3.1: judge the time from each IMF component to each acoustic wave sensor through the second-filtered acoustic wave signal obtained in step 4.2;

Step 4.3.2: The eigenmode function of the time received by the outer acoustic wave sensor at the beginning of the pipe before the time received by the inner acoustic wave sensor and the time received by the outer acoustic wave sensor at the end of the pipe is earlier than that received by the inner acoustic wave sensor The eigenmode function of time is the acoustic wave interference from the upstream station and the downstream station, and these eigenmode functions are removed;

Step 4.3.3: Add the remaining eigenmode functions corresponding to the acoustic signals to obtain the acoustic source signals from which the noise in the station is filtered out, which are the acoustic signals Y ₁ and Y ₂ of the acoustic sensors of the upstream and downstream stations.

The concrete process of described step 5 is:

Step 5.1: The data after the secondary filtering in step 4 is subjected to dimensionless processing by the standard deviation method, and combined into a measurement matrix X _L ; m points whose value is equal to the pressure data value at the previous time point are inserted between each pressure signal , The rounded integer is m, where: f ₁ is the frequency of the sound wave signal, and f ₂ is the frequency of the pressure signal;

Step 5.2: Measurement matrix X _L =[x ₁ , x ₂ , x ₃ ,...x _N ], which includes normal oil delivery signal, small leakage signal, large leakage signal and working condition interference signal, and the kth category contains n _k samples, k=1, 2, 3, 4; n ₁ +n ₂ +n ₃ +n ₄ =N;

Define S _b as the between-class scatter matrix and S _w as the intra-class scatter matrix, as shown in the following two formulas:

Among them, the weight matrix W ^(b) and W ^(w) are defined as:

Among them, C _k represents the set of category number k, C _k = {1, 2, 3, 4};

Define S _t as the global scatter matrix, in,

Step 5.3: Order Where I is the unit diagonal matrix, the semi-supervised Fisher optimization discriminant vector is obtained by the following formula:

Among them, J is the maximum scoring parameter, p is any constant not equal to 0, and P is the load matrix;

The above formula is equivalent to S _rb q=λS _rw q, where λ is the generalized eigenvalue, and q is the corresponding generalized eigenvector;

Arrange the obtained generalized eigenvalues in descending order as λ ₁ ≥λ ₂ ≥…≥λ _N , the corresponding generalized eigenvectors are q ₁ , q ₂ ,…,q _N , and the vectors q ₁ , q ₂ ,…,q _N are Optimize the discriminant vector for semi-supervised Fisher, and its classification ability is also weakened in turn;

Assuming that the prior probability of the sample belonging to each class is equal, it is K is the total number of categories, then the label sample The conditional probability density function of is as follows:

Among them, Q _r =[q ₁ , q ₂ ,...,q _r ] are the first r Fisher discriminant feature vectors, and the space formed by Q _r is r-dimensional semi-supervised Fisher discriminant subspace, is the mean vector of samples of class C _k ;

According to Bayesian rule, The formula for calculating the posterior probability of belonging to the i-th type is:

The half-doctor Fisher discriminant function is defined as

Represents any sample in the test data set The value of the discriminant function belonging to the kth category in the training set;

Predict each test sample in the test set according to the leakage classification criterion shown in the following formula For normal oil delivery signal, large leakage signal, small leakage signal or working condition interference signal:

Step 5.4: Use the normal working condition signal, large leakage signal, small leakage signal and working condition interference signal in the historical data to train the mathematical model obtained in step 5.3 respectively, and obtain the normal working condition signal, large leakage signal, small leakage signal and working condition Corresponding to the interference signal Build a database for pipeline leak detection.

The formula for calculating the degree of similarity in step 6 is as follows:

In the formula, S _k is the similarity, C(x) ^new is the newly collected data, is the test sample of the kth type, is a set of k samples, is a set of all test samples;

If S _i ≥ 0.85, the current data is the kth type of data; otherwise, combine the actual analysis of the site conditions and human diagnosis to determine the type of fault, and store it in the database of pipeline leakage detection.

The specific process of locating the pipeline leakage in the step 7 is:

Step 7.1: Calculate the leak point according to the time difference of the pressure sensors of the upstream station and the downstream station, as shown in the following formula,

In the formula, Z ₁ is the distance from the leakage point to the pressure sensor of the upstream station; l ₁ is the distance between the pressure sensors of the upstream and downstream stations on the pipeline; τ ₁ is the time difference between the negative pressure wave and the upstream and downstream pressure sensors; v ₁ is the negative pressure The propagation velocity of the wave in the pipeline;

Step 7.2: Select the average value of the pressure data 30 seconds before the leakage point and the average value of the pressure data 15 seconds after the leakage point at the upstream station and the downstream station of the leakage point, and calculate the pressure change ratio of the upstream and downstream stations, as shown in the following two formulas respectively,

Among them: δ ₁ , δ ₂ are the pressure change ratios of the upstream station and the downstream station respectively; X ₁₊ , X ₂₊ are the average pressure data values of the upstream station and the downstream station 30 seconds before the leakage occurs; X _1- , X _{2 -} mean values of the pressure data of the upstream station and downstream station 15 seconds after the leak occurred;

Then the total pressure change ratio δ _p when the pipeline leaks is: Among them, μ is a parameter determined by the length of the on-site pipeline and the pressure of the normal delivery pipeline;

Step 7.3: Calculate the leakage point for the time difference between the characteristic value of the sound wave at the upstream station and the downstream station. As shown in the following formula,

In the formula, Z ₂ is the distance from the leak point to the acoustic wave sensor of the upstream station; l ₂ is the distance between the acoustic wave sensors of the upstream and downstream stations on the pipeline; _{τ 2 is} the time difference between the sound wave transmitted to the upstream and downstream acoustic wave sensors; speed of propagation;

Step 7.4: At the upstream station and downstream station of the leakage point, select the average value of the acoustic wave data 5 seconds before the leakage point and the average value of the acoustic wave data 3 seconds after the leakage point, and calculate the acoustic wave change ratio of the upstream and downstream stations, respectively, as shown in the following two formulas,

Among them, δ ₃ , δ ₄ are the acoustic wave change ratios of the upstream station and the downstream station respectively; Y ₁₊ , Y ₂₊ are the mean values of the acoustic wave data of the upstream station and the downstream station 5 seconds before the leakage point; Y _1- , Y _{2 -} mean values of the acoustic wave data of the upstream station and downstream station 3 seconds after the leak occurred;

Then the total acoustic wave change ratio δ _s when the pipeline leaks is:

Step 7.5: According to the final distance from the leak point to the pressure sensor of the upstream station is

Among them, T is the period of the sound wave.

The beneficial effect produced by adopting the above technical solution is that the pipeline leakage monitoring method based on acoustic negative pressure wave hybrid monitoring provided by the present invention has the following advantages compared with the existing system:

(1) the lower computer cost that the present invention uses is lower;

(2) Multiple filtering including separate hardware and software filtering for sound waves and pressure signals, better shielding from noise interference;

(3) In the data preprocessing stage, the signal is decomposed by ICEEMD, and the intrinsic mode function corresponding to the source signal is selected by approximate entropy, which ensures the accuracy of the restored signal after filtering when the signal source changes;

(4) The acoustic interference from the station can be shielded by setting dual acoustic sensors;

(5) The method of Fisher's discriminant analysis is used to comprehensively analyze the pressure and acoustic wave signals, which ensures the accuracy of the characteristic value of the leakage, minimizes false alarms, and has a more accurate judgment on the leakage of small flow;

(6) Comprehensively use sound waves and pressure signals to judge the leak point. In view of the inaccurate positioning of small leaks by the negative pressure wave method and large leaks by the sonic wave method, it is more accurate to set a threshold to locate the leak point.

Description of drawings

Fig. 1 is a schematic structural diagram of a pipeline leakage monitoring system based on acoustic negative pressure wave hybrid monitoring provided by an embodiment of the present invention;

Fig. 2 is the overall algorithm flow chart of the pipeline leakage monitoring method based on the acoustic negative pressure wave hybrid monitoring provided by the embodiment of the present invention;

Fig. 3 is the implementation flowchart of the pipeline leakage monitoring method based on the acoustic negative pressure wave hybrid monitoring provided by the embodiment of the present invention;

FIG. 4 is a flowchart of a secondary filtering method provided by an embodiment of the present invention;

Fig. 5 is the ICEEMD decomposition result diagram of the negative pressure wave under normal operating conditions provided by the embodiment of the present invention;

Fig. 6 is the ICEEMD decomposition result diagram of the negative pressure wave during leakage provided by the embodiment of the present invention;

Fig. 7 is a schematic diagram of the restored signal after filtering the pressure signal provided by the embodiment of the present invention;

Fig. 8 is the ICEEMD decomposition result diagram of the acoustic wave signal during the normal working condition provided by the embodiment of the present invention;

Fig. 9 is the ICEEMD decomposition result diagram of the acoustic wave signal during the leakage provided by the embodiment of the present invention;

FIG. 10 is a schematic diagram of a restored signal after the acoustic wave signal is filtered according to an embodiment of the present invention.

In the figure: 1. The pressure sensor at the beginning of the pipeline; 2. The pressure sensor at the end of the pipeline; 3. The outer acoustic wave sensor at the beginning of the pipeline; 4. The inner acoustic wave sensor at the beginning of the pipeline; 5. The inner acoustic wave sensor at the end of the pipeline; External acoustic wave sensor; 7. Lower computer; 8. Switch; 9. Upper computer; 10. Entry valve; 11. Pump.

Detailed ways

The specific implementation manners of the present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. The following examples are used to illustrate the present invention, but are not intended to limit the scope of the present invention.

A pipeline leakage monitoring system based on acoustic negative pressure wave hybrid monitoring, as shown in FIG. 1 , includes a pressure sensor, an acoustic wave sensor, a lower computer 7 , a switch 8 and an upper computer 9 .

A pressure sensor is installed at the beginning and end of the pipeline, and is in contact with the conveying medium, and is used to collect high-precision data reflecting the pipeline pressure in real time. The pressure sensors at the beginning and end of the pipeline are 1 and 2 in Figure 1, respectively.

In this embodiment, the main parameters of the Rosemount 3051s pressure transmitter are as follows:

(1) The measuring pressure range of the pressure transmitter is 0~8MPa;

(2) The signal resolution is 0.015%, the accuracy is ±0.075%, and the update rate is 50Hz;

(3) The output signal is 4~20mADC (two-wire system), the load capacity is not less than 700Ω, and the power supply is 24VDC;

(4) It has the overload capacity of 1.5 times of the maximum measuring range;

(5) The influence of every 50°F (28°C) change in the ambient temperature is better than: ±(0.025% of the upper limit of the range + 0.125% of the range);

(6) The effect of every 1000psi (6.9MPa) change in static pressure is better than: ±0.1% of the upper limit of the range.

Two acoustic wave sensors are installed at the head end and the end of the pipeline, and placed on both sides of the pressure sensor at the head end and the end, for real-time collection of high-speed data reflecting the sound waves of the pipeline. The two acoustic wave sensors at the head end of the pipeline are respectively 3 and 4 in Fig. 1, and the two acoustic wave sensors at the end of the pipeline are respectively 5 and 6 in Fig. 1, where 3 and 6 are outside, and 4 and 5 are inside. The distance between the two acoustic wave sensors at the head end or the end is not less than 20Tv, T is the period of the sound wave, and v is the propagation speed of the sound wave in the pipeline.

In this embodiment, the acoustic wave sensor adopts CT1000 series, which has the following technical indicators:

(1) Voltage sensitivity: 947.47mv/g;

(2) Measurement frequency range: 0.2～1200Hz;

(3) Range: 5g;

(4) Linearity: ≤1%;

(5) Working temperature: -20～120℃;

(6) Shock resistance: 10g;

(7) Output mode: top M5;

(8) Excitation current: 2～10mA, excitation voltage: 10-24VDC.

The lower computer 7, the switch 8 and the upper computer 9 are respectively provided with a group at the head end and the end of the pipeline. The pressure sensor and the acoustic wave sensor are all connected to the lower computer 7 at the corresponding end. The lower computer 7 is connected to the upper computer at the corresponding end through the network cable and the switch 8 Machine 9 is connected.

The lower computer 7 is programmed to perform the following steps: control the pressure sensor and the acoustic wave sensor at the corresponding end to collect pressure and acoustic wave data, and perform preprocessing on the collected data, including filtering and time stamping according to the GPS collection time; After the data is packaged, it is sent to the upper computer 9 through the switch 8, and is stored in the SD card of the lower computer 7 for backup.

The technical index requirements in the present embodiment are as follows:

(1) Sampling frequency range: 0-1.2KHz;

(2) Drive signal: 3.3V and 5V;

(3) Number of sampling signal channels: at least 4 channels;

(4) Sampling accuracy: 0.001;

(5) Communication method: Ethernet communication;

(6) Power supply mode: 24V DC power supply.

In this embodiment, the sampling frequency of the pressure signal is selected as 50 Hz, and the sampling frequency of the acoustic wave signal is even 1200 Hz.

The upper computer 9 includes a computer-executable leakage monitoring program, specifically including: a pressure and sound wave data integration storage module, a data processing module, and a pipeline leakage monitoring module.

The pressure and sound wave data integration storage module is used to receive and analyze the pressure and sound wave data transmitted by the lower computer 7, store them according to the data detection time, and combine a set of pressure data and two sets of sound wave data at one end of each pipeline into a third-order matrix;

The data processing module is used to obtain data from the pressure and acoustic wave data integration storage module for processing and calculation, including secondary filtering of data, dimensionless processing, and Fisher discriminant processing to obtain data processing and calculation results; among them, in In the secondary filtering process of the data, the improved complete overall empirical mode decomposition ICEEMD is used to decompose the sound wave and negative pressure wave signals, and according to the calculation of the approximate entropy of each decomposed component, the sound wave and negative pressure wave signals are filtered out. noise, and remove the acoustic interference signal to obtain the final filtered signal; in the dimensionless processing of the data after the secondary filtering, the semi-supervised Fisher discriminant is trained through the collected historical pressure and acoustic data, and the normal A database of four working conditions: working condition signal, large leakage signal, small leakage signal and working condition adjustment signal;

The pipeline leakage monitoring module is used to use the real-time data of pressure and sound wave to judge whether the pipeline leaks according to the data processing and calculation results of the pressure and sound wave data integration module and the data processing module. When it is judged as a large or small leakage signal, pass The pressure signal and the sound wave signal are mixed to calculate the leakage distance and locate the pipeline leakage point.

Using the above-mentioned monitoring system to realize the pipeline leakage monitoring method based on acoustic negative pressure wave mixed monitoring, the overall algorithm is shown in Figure 2, and the specific implementation flow chart is shown in Figure 3, and the specific method is as follows.

Step 1: The lower computer at the beginning and end of the pipeline controls the corresponding pressure sensor and acoustic wave sensor to collect pipeline pressure and acoustic wave data, including pressure data and sonic data The collected data is filtered to obtain filtered pressure signals X ₁ , X ₂ and acoustic wave signals Y ₁₁ , Y ₁₂ , Y ₂₁ , and Y ₂₂ .

Due to the long transmission distance of the pipeline, the high-frequency noise generated by the pipeline leakage attenuates quickly in the pipeline, so the high-frequency part of the acoustic signal collected by the lower computer can be regarded as a noise signal, so low-pass filtering is used to filter out the acoustic signal with a frequency greater than 200Hz . Filter out gross errors, high-frequency noise and other interference in the pressure signal of the lower computer.

The lower computer time stamps the filtered pressure and sound wave signals through the time data collected by GPS every second.

Step 2: The lower computer packs and sends the filtered and time-stamped pressure and real-time acoustic signals to the upper computer, and stores them in the SD card of the lower computer for backup.

The data within the last 60 minutes is stored in the SD card. When the upper computer does not receive the data packet of a certain period of time due to network delay and other reasons, it sends a request to the lower computer, and the next second, the lower computer will send the data packet of the current time together with the missing data of the upper computer to the upper computer. .

Step 3: The upper computer receives the data sent by the lower computer, determines the required data packet by identifying the header and tail of the data packet, then analyzes the data type and data value in the data packet, and finally stores it in the database of the upper computer .

According to the analysis of the generation and propagation mechanism of the pipeline leakage signal of acoustic wave and negative pressure wave, it can be known that the acoustic wave signal and negative pressure wave signal will have small fluctuations in the frequency domain during leakage, and the traditional algorithm used in pipeline leakage detection has not Optimizing it will interfere with the subsequent leak location and affect its location accuracy. In this embodiment, ICEEMD is used to decompose the sound wave and the negative pressure wave signal, and the final filtered signal is selected according to the calculation of the approximate entropy of each decomposed component. The specific steps are as follows.

Step 4: The host computer performs secondary filtering on the received data, as shown in Figure 4, including the following steps:

Step 4.1: Use the improved complete overall empirical mode decomposition, namely ICEEMD, to decompose the acoustic wave and negative pressure wave signals, and obtain the decomposed eigenmode function components; the specific process is as follows:

Step 4.1.1: Add I group of Gaussian white noise to the signal X(t) to generate I new signals, the i-th new signal is ^Xi (t)=X(t)+β _k w ⁱ ; where, X (t) is the original sound wave or negative pressure wave signal; w ⁱ is the i-th group of Gaussian white noise variables, i=1, 2, ..., I; β _k =ε _k std(r _k ), r _k is the k-th Remainder, ε _k = 0.2;

In this embodiment, according to the analysis of the approximate entropy of ICEEMD and IMF, data processing is performed on the sound waves and pressure signals collected in the experiment, 200 groups of Gaussian noise with a signal-to-noise ratio of 5 are added, and the maximum number of iterations of EMD is selected as 500 times.

Step 4.1.2: Initialize cycle parameters, let i=1;

Step 4.1.3: Determine all extreme points on the signal X ⁱ (t), including local maximum points and local minimum points; fit the extreme points to obtain the upper envelope and the lower envelope Let X ⁱ (t) satisfy:

Step 4.1.4: Calculate the average value of the upper and lower envelopes of ^Xi (t), denoted as m(t),

Step 4.1.5: Subtract the average value m(t) from the original ^Xi (t) signal data to obtain the function _hi (t), that is, _hi (t)= ^Xi (t)-m(t);

Step 4.1.6: Set X ⁱ (t) = h _i (t), and judge whether h _i (t) satisfies the two conditions of the intrinsic mode function IMF:

Condition 1: At any point in time, the mean value of the envelope defined by the local maximum and local minimum of the IMF must be zero, that is, for any t, we have

Condition 2: In the entire data sequence, the number of extreme points and the number of zero-crossing points are equal or at most equal to one;

If satisfied, execute step 4.1.7, otherwise, return to step 4.1.3;

The first constraint is to ensure the local symmetry of the waveform, and the second constraint is to approximate the definition of the narrow band of the traditional stationary Gaussian process;

Step 4.1.7: Separate _hi (t) from the X(t) signal, which is the IMF _i component, and obtain a function r _i (t) that removes the difference of the high-frequency component, that is, the remainder r _i ( t)=X(t)-IMF _i ;

The first IMF component obtained from ^Xi (t), i.e. h ₁ (t), is denoted as IMF _i1 , i=1, 2, ..., I, and the mean value of all IMF _i1 is obtained to obtain IMF ₁ . The term is r ₁ =X(t)-IMF ₁ ;

Step 4.1.8: Judging whether the residual signal r _i (t) is a signal of a monotone function, if so, then X(t) can no longer be decomposed into IMF components, and the decomposition process is completed to obtain each IMF component, which is recorded as IMF ₁ ~ IMF _n , n=i, and the final residual signal is r _n (t)=rn _-1 (t)-h _n (t), and the signal X(t) is n eigenmode function components and a residual term the sum of Otherwise, let X(t)=r _i be a new signal, and let i=i+1, return to step 4.1.3, and perform the next decomposition.

Step 4.2: Calculate the approximate entropy of each eigenmode function component IMF ₁ to IMF _n , and judge the pressure signals X ₁ , X ₂ and acoustic wave signals Y ₁₁ , Y ₁₂ , Y ₂₁ , and Y ₂₂ according to the approximate entropy of each component. The noise signal is filtered out, the specific process is:

Step 4.2.1: Consider the data in the jth eigenmode function as a time series with y points, recorded as {Z _j }={Z _j1 , Z _j2 ,..., Z _jy }; for a given Threshold a and pattern dimension g, Calculate the n*n binary distance matrix B:

in,

In this embodiment, a is 0.1-0.2, and g is 2.

Step 4.2.2: Use the elements b _rj of matrix B to calculate the ratios of the number of elements less than a to the total number of distances n-g+1 and n-g+2, respectively recorded as and As shown in the following two formulas:

Step 4.2.3: According to and calculated and ( Take the natural logarithm for C _r , and then find the average value of all r), as shown in the following two formulas,

Then the approximate entropy A _j of the jth eigenmode function is expressed as:

Step 4.2.4: Arrange the approximate entropy values obtained in step 4.2.3 from large to small, and the corresponding IMF components with approximate entropy values greater than 1 are regarded as noise signals for filtering, and the approximate entropy of the acoustic wave signal is between 0.486 and 0.490 The IMF component is the coupling signal of the sound wave and the pressure signal, which is filtered out as an interference signal.

Step 4.3: Remove the acoustic wave interference from the upstream station and downstream station in the acoustic wave signal, and obtain the signal after secondary filtering. The specific process is:

Step 4.3.1: judge the time from each IMF component to each acoustic wave sensor through the second-filtered acoustic wave signal obtained in step 4.2;

Step 4.3.2: The eigenmode function of the time received by the outer acoustic wave sensor at the beginning of the pipe before the time received by the inner acoustic wave sensor and the time received by the outer acoustic wave sensor at the end of the pipe is earlier than that received by the inner acoustic wave sensor The eigenmode function of time is the acoustic wave interference from the upstream station and the downstream station, and these eigenmode functions are removed;

Step 4.3.3: Add the remaining eigenmode functions corresponding to the acoustic signals to obtain the acoustic source signals from which the noise in the station is filtered out, which are the acoustic signals Y ₁ and Y ₂ of the acoustic sensors of the upstream and downstream stations.

In this embodiment, the ICEEMD decomposition of the negative pressure wave under normal working conditions is shown in Figure 5, and the approximate entropy corresponding to each IMF component is shown in Table 1. It can be seen from Figure 5 and Table 1 that under normal working conditions, the approximate entropy of IMF1 of the pressure signal is 1.2854, and r has little influence on the signal and the fluctuation range is small, so the pressure signal can be restored by the IMF2-IMF10 signal, and the high High-frequency noise and extremely low-frequency signals that are useless for signal feature extraction.

Table 1 Approximate entropy of IMF function of negative pressure wave under normal working conditions

component number approximate entropy component number approximate entropy component number approximate entropy IMF1 1.2854 IMF5 0.2562 IMF9 0.0052 IMF2 0.8544 IMF6 0.1275 IMF10 0.0037 IMF3 0.6254 IMF7 0.0426 r 6.3882*e ^-4 IMF4 0.5747 IMF8 0.0214 - -

The ICEEMD decomposition of the negative pressure wave during leakage is shown in Figure 6, and the approximate entropy corresponding to each IMF component is shown in Table 2. It can be seen from Figure 6 and Table 2 that the approximate entropy of IMF1 and IMF2 of the pressure signal under leakage conditions is 1.7260 and 1.1563, and these two components can be regarded as high-frequency noise signals. The three components of IMF11, IMF12 and r have little influence on the signal, so the pressure signal can be restored by the IMF3-IMF10 signal, which can remove high-frequency noise and extremely low-frequency signals that are useless for signal feature extraction. The signal after the pressure signal is filtered and restored is shown in Figure 7, where Figure a is the restored signal of the pressure signal under normal working conditions, and Figure b is the restored signal of the pressure signal during leakage.

Table 2 Approximate entropy of IMF function of negative pressure wave during leakage

component number approximate entropy component number approximate entropy component number approximate entropy IMF1 1.7260 IMF6 0.0706 IMF11 0.0004 IMF2 1.1563 IMF7 0.0420 IMF12 7.6097*e ^-4 IMF3 0.5579 IMF8 0.0209 r 2.3579*e ^-4 IMF4 0.2889 IMF9 0.0018 - - IMF5 0.0622 IMF10 0.0010 - -

The ICEEMD decomposition of the acoustic signal under normal working conditions is shown in Figure 8, and the approximate entropy corresponding to each IMF component is shown in Table 3. It can be seen from Fig. 8 and Table 3 that under normal working conditions, the approximate entropy values of all IMF components of the acoustic wave signal are less than 1, and there is no component with a very small approximate entropy value. Therefore, under normal working conditions, the acoustic wave signal can be composed of all IMF components Signal restoration.

Table 3 Approximate entropy of the IMF function of the acoustic signal under normal working conditions

component number approximate entropy component number approximate entropy component number approximate entropy IMF1 0.9229 IMF6 0.2007 IMF11 0.0057 IMF2 0.4960 IMF7 0.0973 IMF12 0.0014 IMF3 0.3220 IMF8 0.0444 r 0.0036 IMF4 0.1938 IMF9 0.0261 - - IMF5 0.2133 IMF10 0.0203 - -

The ICEEMD decomposition of the acoustic signal during leakage is shown in Figure 9, and the approximate entropy corresponding to each IMF component is shown in Table 4. It can be seen from Figure 9 and Table 4 that there is a large amount of noise during leakage, the approximate entropy of IMF1 is 1.2255, and the approximate entropy of IMF2 is 1.0348. It can be seen that IMF1 and IMF2 are noise signals, and the approximate entropy of the residual r is 7.8322*e ^-4 , which has little effect on the signal, so IMF1, IMF2 and the margin r must be filtered out when the signal is restored, so the acoustic signal can be restored by the IMF3-IMF11 component signal. The signal after filtering and restoring the acoustic wave signal is shown in Figure 10, where Figure a is the restored signal of the acoustic wave signal under normal working conditions, and Figure b is the restored signal of the acoustic wave signal during leakage.

Table 4 Approximate entropy of the IMF function of the acoustic wave signal when leaking

component number approximate entropy component number approximate entropy component number approximate entropy IMF1 1.2255 IMF5 0.4863 IMF9 0.0214 IMF2 1.0348 IMF6 0.2553 IMF10 0.0068 IMF3 0.6796 IMF7 0.1212 IMF11 0.0060 IMF4 0.5723 IMF8 0.0480 r 7.8322*e ^-4

Step 5: Carry out standard deviation method dimensionless processing on the signal after secondary filtering, and train semi-supervised Fisher discrimination through the collected historical pressure and acoustic wave data, and obtain signals belonging to normal working conditions, large leak signals, and small leaks respectively The database of the four working conditions of the signal and the working condition adjustment signal;

Based on the nuclear Fisher discriminant analysis and the concept of semi-supervised learning method, by learning and analyzing the database when the historical leakage occurs, it can be judged whether the leakage is really occurred when the host computer monitors the abnormal waveform in the future. The specific process is as follows:

Step 5.1: The data after the secondary filtering in step 4 is subjected to dimensionless processing by the standard deviation method, and combined into a measurement matrix X _L ; in order to ensure the time consistency of the acoustic wave signal and the pressure signal, the pressure signal needs to be expanded, and in each Insert m points whose value is equal to the pressure data value of the previous time point between the pressure signals, The rounded integer is m, where: f ₁ is the frequency of the sound wave signal, and f ₂ is the frequency of the pressure signal;

Step 5.2: Measurement matrix X _L =[x ₁ , x ₂ , x ₃ ,...x _N ], which includes normal oil delivery signal, small leakage signal, large leakage signal and working condition interference signal, and the kth category contains n _k samples, k=1, 2, 3, 4; n ₁ +n ₂ +n ₃ +n ₄ =N;

Define S _b as the between-class scatter matrix and S _w as the intra-class scatter matrix, as shown in the following two formulas:

Among them, the weight matrix W ^(b) and W ^(w) are defined as:

Among them, C _k represents the set of category number k, C _k = {1, 2, 3, 4};

Define S _t as the global scatter matrix, in,

Step 5.3: Order Where I is the unit diagonal matrix, the semi-supervised Fisher optimization discriminant vector is obtained by the following formula:

Among them, J is the maximum scoring parameter, p is any constant not equal to 0, and P is the load matrix;

The above formula is equivalent to S _rb q=λS _rw q, where λ is the generalized eigenvalue, and q is the corresponding generalized eigenvector;

Arrange the obtained generalized eigenvalues in descending order as λ ₁ ≥λ ₂ ≥…≥λ _N , the corresponding generalized eigenvectors are q ₁ , q ₂ ,…,q _N , and the vectors q ₁ , q ₂ ,…,q _N are Optimizing the discriminant vector for the semi-supervised Fisher, its classification ability also weakens successively; In this embodiment, considering the calculation speed problem, only the first four items are used, i.e. N=4;

Usually, the data under normal working conditions can be assumed to satisfy the multivariate Gaussian distribution, and the leakage data can also be considered to satisfy the Gaussian distribution. Assuming that the prior probability of the sample belonging to each class is equal, it is K is the total number of categories, then the label sample The conditional probability density function of is as follows:

Among them, Q _r =[q ₁ , q ₂ ,...,q _r ] are the first r Fisher discriminant feature vectors, and the space formed by Q _r is r-dimensional semi-supervised Fisher discriminant subspace, is the mean vector of samples of class C _k ;

According to Bayesian rule, The formula for calculating the posterior probability of belonging to the i-th type is:

The half-doctor Fisher discriminant function is defined as

Represents any sample in the test data set belongs to the discriminant function value of the kth classification in the training set; then, each sample point in the test set The discriminant value belonging to the kth category can be calculated by the above formula;

Predict each test sample in the test set according to the leakage classification criterion shown in the following formula For normal oil delivery signal, large leakage signal, small leakage signal or working condition interference signal:

Step 5.4: Use the normal working condition signal, large leakage signal, small leakage signal and working condition interference signal in the historical data to train the mathematical model obtained in step 5.3 respectively, and obtain the normal working condition signal, large leakage signal, small leakage signal and working condition Corresponding to the interference signal Build a database for pipeline leak detection.

Step 6: Collect pipeline pressure and acoustic wave data in real time, perform secondary filtering through preprocessing and the secondary filtering method described in step 4, and send it to the semi-supervised Fisher discriminant model obtained in step 5 training in seconds, According to the comparison between the obtained feature vector and the established pipeline leak detection database, the degree of similarity between the two is calculated, as follows:

In the formula, S _k is the similarity, C(x) ^new is the newly collected data, is the test sample of the kth type, is a set of k samples, is a set of all test samples;

If S _i ≥ 0.85, the current data is the kth type of data; otherwise, if the similarity between the current data and any feature vector in the database is less than the threshold 0.85, it is likely to be a new unidentified fault. The actual analysis of on-site conditions and human diagnosis determine the type of fault and store it in the database of pipeline leak detection.

Step 7: Determine the fault type of the real-time data according to the degree of similarity. If it is determined to be a large leak signal or a small leak signal, calculate the position of the leak point by mixing the pressure signal and the acoustic wave signal, and locate the pipeline leak. Otherwise, return to step 6 and continue real-time monitoring.

The specific process of locating pipeline leaks is as follows:

Step 7.1: Calculate the leak point according to the time difference of the pressure sensors of the upstream station and the downstream station, as shown in the following formula,

In the formula, Z ₁ is the distance from the leakage point to the pressure sensor of the upstream station; l ₁ is the distance between the pressure sensors of the upstream and downstream stations on the pipeline; τ ₁ is the time difference between the negative pressure wave and the upstream and downstream pressure sensors; v ₁ is the negative pressure The propagation velocity of the wave in the pipeline;

Step 7.2: Select the average value of the pressure data 30 seconds before the leakage point and the average value of the pressure data 15 seconds after the leakage point at the upstream station and the downstream station of the leakage point, and calculate the pressure change ratio of the upstream and downstream stations, as shown in the following two formulas respectively,

Among them: δ ₁ , δ ₂ are the pressure change ratios of the upstream station and the downstream station respectively; X ₁₊ , X ₂₊ are the average pressure data values of the upstream station and the downstream station 30 seconds before the leakage occurs; X _1- , X _{2 -} mean values of the pressure data of the upstream station and downstream station 15 seconds after the leak occurred;

Then the total pressure change ratio δ _p when the pipeline leaks is: Among them, μ is a parameter determined by the length of the on-site pipeline and the pressure of the normal delivery pipeline;

Step 7.3: Calculate the leakage point for the time difference between the characteristic value of the sound wave at the upstream station and the downstream station. As shown in the following formula,

In the formula, Z ₂ is the distance from the leak point to the acoustic wave sensor of the upstream station; l ₂ is the distance between the acoustic wave sensors of the upstream and downstream stations on the pipeline; τ ₂ is the time difference between the sound wave transmitted to the upstream and downstream acoustic wave sensors; v2 is the time difference of the sound wave in the pipeline transmission speed;

Step 7.4: At the upstream station and downstream station of the leakage point, select the average value of the acoustic wave data 5 seconds before the leakage point and the average value of the acoustic wave data 3 seconds after the leakage point, and calculate the acoustic wave change ratio of the upstream and downstream stations, respectively, as shown in the following two formulas,

Among them, δ ₃ , δ ₄ are the acoustic wave change ratios of the upstream station and the downstream station respectively; Y ₁₊ , Y ₂₊ are the average values of the acoustic wave data of the upstream station and the downstream station 5 seconds before the leakage point; Y _1- , Y _{2 -} mean values of the acoustic wave data of the upstream station and downstream station 3 seconds after the leak occurred;

Then the total acoustic wave change ratio δ _s when the pipeline leaks is:

Step 7.5: In order to shield the problem of poor positioning accuracy of acoustic waves during large leaks and poor positioning accuracy of negative pressure waves during small leaks, the final distance from the leak point to the pressure sensor of the upstream station is

Among them, T is the period of the sound wave.

One uses pressure waves to judge and calculate leak points, and the other uses sound waves to judge and calculate leak points. The results obtained by mixing the two waves are more accurate.

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 them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that: it can still be Modifications are made to the technical solutions described in the foregoing embodiments, or equivalent replacements are made to some or all of the technical features; these modifications or replacements do not make the essence of the corresponding technical solutions depart from the scope defined by the claims of the present invention.

Claims (8)

1. The utility model provides a pipeline leakage monitoring system based on mixed monitoring of sound wave negative pressure wave which characterized in that: the system comprises a pressure sensor, an acoustic wave sensor, a lower computer, a switch and an upper computer;

the pressure sensors are respectively arranged at the head end and the tail end of the pipeline and are in contact with a conveying medium, and the pressure sensors are used for acquiring high-precision data reflecting the pressure of the pipeline in real time; the sound wave sensors are respectively arranged at the head end and the tail end of the pipeline, are arranged at two sides of the pressure sensors at the head end and the tail end and are used for acquiring high-speed data reflecting sound waves of the pipeline in real time; the distance between the two sound wave sensors at the head end or the tail end is not less than 20Tv, T is the period of the sound wave, and v is the propagation speed of the sound wave in the pipeline;

the lower computer, the exchanger and the upper computer are respectively provided with a group at the head end and the tail end of the pipeline, the pressure sensor and the sound wave sensor are both connected with the lower computer at the corresponding end, and the lower computer is connected with the upper computer at the corresponding end through a network cable and the exchanger;

the lower computer is programmed to perform the steps of: controlling a pressure sensor and an acoustic wave sensor at a corresponding end to acquire pressure and acoustic wave data, and preprocessing the acquired data, including filtering and timestamp according to GPS acquisition time; packaging the preprocessed data, sending the data to an upper computer through a switch, and storing the data into an SD card of a lower computer for backup;

the upper computer comprises a leakage monitoring program executable by a computer, and specifically comprises: the device comprises a pressure and sound wave data integration storage module, a data processing module and a pipeline leakage monitoring module;

the pressure and sound wave data integration storage module is used for receiving and analyzing pressure and sound wave data transmitted by the lower computer, storing the pressure and sound wave data according to data detection time, and combining one group of pressure data and two groups of sound wave data at one end of each pipeline into a third-order matrix;

the data processing module is used for acquiring data from the pressure and sound wave data integration storage module for processing and calculating, including secondary filtering, dimensionless processing and semi-supervised Fisher discrimination processing of the data, so as to obtain a data processing and calculating result; in the secondary filtering processing of data, decomposing the sound wave and negative pressure wave signals by adopting an improved complete ensemble empirical mode decomposition ICEEMD, filtering noise in the sound wave and negative pressure wave signals according to the approximate entropy of each decomposed component, and removing sound wave interference signals to obtain final filtered signals; training semi-supervised Fisher discrimination through collected historical pressure and sound wave data in the non-dimensionalization processing of the secondarily filtered data to respectively obtain databases belonging to four working condition conditions of a normal working condition signal, a large leakage signal, a small leakage signal and a working condition adjusting signal;

and the pipeline leakage monitoring module is used for judging whether the pipeline leaks according to the data processing and calculating results of the pressure and sound wave data integration module and the data processing module by utilizing real-time data of pressure and sound waves, and calculating the leakage distance by mixing a pressure signal and a sound wave signal when a large leakage signal or a small leakage signal is judged, so as to position a pipeline leakage point.

2. A pipeline leakage monitoring method based on sound wave negative pressure wave hybrid monitoring is realized by adopting the pipeline leakage monitoring system based on sound wave negative pressure wave hybrid monitoring as claimed in claim 1, and is characterized in that: the method comprises the following steps:

step 1: the head end and tail end lower computers of the pipeline control corresponding pressure sensors and sound wave sensors to acquire pipeline pressure and sound wave data, including pressure dataAnd acoustic dataAnd filtering the collected data to obtain a filtered pressure signal X₁、X₂And acoustic signal Y₁₁、Y₁₂、Y₂₁、Y₂₂(ii) a The lower computer carries out time stamping on the filtered pressure and sound wave signals through time data acquired by a GPS (global positioning system) every second;

step 2: the lower computer packs and sends the filtered and time-stamped pressure and sound wave real-time signals to the upper computer, and simultaneously stores the signals into an SD card of the lower computer for backup;

and step 3: the upper computer receives data sent by the lower computer, determines a required data packet by identifying a packet head and a packet tail of the data packet, analyzes a data type and a data value in the data packet, and finally stores the data type and the data value in a database of the upper computer;

and 4, step 4: the upper computer carries out secondary filtering processing on the received data, and the method comprises the following steps:

step 4.1: decomposing the sound wave and negative pressure wave signals by adopting Improved Complete Ensemble Empirical Mode Decomposition (ICEEMD) to obtain each decomposed intrinsic mode function component;

step 4.2: calculating the approximate entropy of each intrinsic mode function component, and judging the pressure signal X according to the approximate entropy of each component₁、X₂And acoustic signal Y₁₁、Y₁₂、Y₂₁、Y₂₂Filtering out the noise signal;

step 4.3: removing sound wave interference from an upstream station and a downstream station in the sound wave signals to obtain signals after secondary filtering;

and 5: carrying out standard deviation method non-dimensionalization processing on the secondarily filtered signals, training semi-supervised Fisher discrimination through collected historical pressure and sound wave data, and respectively obtaining databases belonging to four working condition conditions of normal working condition signals, large leakage signals, small leakage signals and working condition adjustment signals;

step 6: acquiring pipeline pressure and sound wave data in real time, sending the data into a semi-supervised Fisher discrimination model obtained by training in the step 5 by taking seconds as a unit after secondary filtering is carried out by a preprocessing and secondary filtering method in the step 4, comparing the obtained characteristic vector with the established database of four working condition conditions, and calculating the similarity degree between the characteristic vector and the database;

and 7: and (4) determining the fault type of the real-time data according to the similarity, if the fault type is judged to be a large leakage signal or a small leakage signal, calculating the position of a leakage point by mixing a pressure signal and a sound wave signal, positioning the pipeline leakage, and otherwise, returning to the step 6 to continue real-time monitoring.

3. The pipeline leakage monitoring method based on the acoustic wave negative pressure wave hybrid monitoring is characterized in that: the specific process of decomposing the sound wave and the negative pressure wave signals by adopting the improved complete ensemble empirical mode decomposition in the step 4.1 is as follows:

step 4.1.1: adding I groups of white Gaussian noises to the signal X (t) to generate I new signals, wherein the ith new signal is Xⁱ(t)＝X(t)+β_kwⁱ(ii) a Wherein, X (t) is an acoustic wave or negative pressure wave signal; w is aⁱIs the white Gaussian noise variable of the ith group, I is 1, 2, … and I, β_k＝ε_kstd(r_k)，r_kIs the kth remainder, ε_k＝0.2；

Step 4.1.2: initializing cycle parameters, and enabling i to be 1;

step 4.1.3: determining a signal Xⁱ(t) all extreme points on the surface, including local maximum points and local minimum points; fitting extreme points to obtain an upper envelope lineAnd a lower envelopeLet Xⁱ(t) satisfies:

step 4.1.4: calculating Xⁱ(t) the average of the upper and lower envelopes, denoted m (t),

step 4.1.5: by the original Xⁱ(t) subtracting the mean value m (t) from the signal data to obtain a function h_i(t), i.e. h_i(t)＝Xⁱ(t)-m(t)；

Step 4.1.6: let Xⁱ(t)＝h_i(t), judgment of h_i(t) whether two conditions of the intrinsic mode function IMF are satisfied:

condition 1: at any point in time, the mean of the envelopes defined by the local maxima and minima of the IMF must be zero, i.e. for any t, there is

Condition 2: the number of extreme points and the number of zero-crossing points are equal or differ by no more than one at most in the entire data sequence;

if yes, executing the step 4.1.7, otherwise, returning to the step 4.1.3;

step 4.1.7: h is to be_i(t) is separated from the X (t) signal, i.e. IMF_iComponent, obtaining a function r of the difference value of the removed high frequency component_i(t), i.e. remainder r_i(t)＝X(t)-IMF_i；

Step 4.1.8: determining the residual signal r_i(t) whether it is a signal of monotonic function, if so, X (t) can not be decomposed again to obtain IMF components, and the decomposition process is completed to obtain each IMF component, which is marked as IMF₁～IMF_nN ═ i, and the final residual signal obtained is r_n(t)＝r_n-1(t)-h_n(t), the signal X (t) being the sum of n eigenmode function components and a residual term, i.e.Otherwise, let X (t) r_iAnd (4) returning to the step 4.1.3 for the next decomposition, wherein the new signal is i + 1.

4. The pipeline leakage monitoring method based on the acoustic wave negative pressure wave hybrid monitoring is characterized in that: the specific process of calculating the approximate entropy of each eigenmode function component in step 4.2 is as follows:

step 4.2.1: regarding the data in the jth eigenmode function as a time sequence of y points, which is denoted as { Z_j}＝{Z_j1、Z_j2、…、Z_jy}; for a given threshold a and mode dimension g,calculating a binary distance matrix B of n x n:

wherein,

step 4.2.2: using the elements B of the matrix B_rjRespectively calculating the ratio of the number of elements less than a to the total distance n-g +1 and n-g +2, and respectively recording the ratioAndthe following two formulas are shown:

step 4.2.3: according toAndis calculated to obtainAndas shown in the following two formulas,

the approximate entropy A of the jth eigenmode function_jExpressed as:

step 4.2.4: and arranging the obtained approximate entropy values from large to small, taking the corresponding IMF component with the approximate entropy value larger than 1 as a noise signal for filtering, and taking the IMF component with the approximate entropy of the sound wave signal between 0.486-0.490 as a coupling signal of the sound wave and the pressure signal as an interference signal for filtering.

5. The pipeline leakage monitoring method based on the acoustic wave negative pressure wave hybrid monitoring is characterized in that: the specific process of removing the acoustic wave interference from the upstream station and the downstream station in the acoustic wave signal in the step 4.3 is as follows:

step 4.3.1: judging the time from each IMF component to each acoustic wave sensor through the acoustic wave signal obtained in the step 4.2 after secondary filtering;

step 4.3.2: the intrinsic mode functions received by the outer sound wave sensor at the head end of the pipeline earlier than the time received by the inner sound wave sensor and the intrinsic mode functions received by the outer sound wave sensor at the tail end of the pipeline earlier than the time received by the inner sound wave sensor are sound wave interference from an upstream station and a downstream station, and the intrinsic mode functions are removed;

step 4.3.3: adding the residual intrinsic mode functions corresponding to the sound wave signals to obtain sound source signals with noise from the stations filtered, namely the sound wave signals Y of the sound wave sensors of the upstream station and the downstream station₁、Y₂。

6. The pipeline leakage monitoring method based on the acoustic wave negative pressure wave hybrid monitoring is characterized in that: the specific process of the step 5 is as follows:

step 5.1: carrying out standard deviation method dimensionless processing on the data subjected to the secondary filtering in the step 4, and combining the data into a measurement matrix X_L(ii) a The points where m values are equal to the pressure data value at the previous point in time are inserted between each pressure signal, rounded integers are m, where: f. of₁Is the frequency of the acoustic signal, f₂Is the frequency of the pressure signal;

step 5.2: measurement matrix X_L＝[x₁，x₂，x₃，…x_N]The k category includes n_kSamples, k ═ 1, 2, 3, 4; n is₁+n₂+n₃+n₄＝N；

Definition of S_bIs an inter-class divergence matrix, S_wThe matrix is an intra-class divergence matrix and is respectively shown as the following two formulas:

wherein, the weight matrix W^(b)And W^(w)Are respectively defined as:

wherein, C_kRepresenting a set of class numbers k, C_k＝{1，2，3，4}；

Definition of S_tIn the form of a global divergence matrix, the divergence matrix,wherein,

step 5.3: order toWherein I is a unit diagonal matrix, the semi-supervised Fisher optimization discrimination vector is obtained by the following formula:

wherein J is a maximum scoring parameter, P is an arbitrary constant not equal to 0, and P is a load matrix;

the above formula is equivalent to S_rbq＝λS_rwq, where λ is the generalized eigenvalue and q is the corresponding generalized eigenvector;

arranging the obtained generalized eigenvalues in descending order as lambda₁≥λ₂≥…≥λ_NThe corresponding generalized eigenvector is q₁，q₂，…，q_NVector q₁，q₂，…，q_NNamely, the semi-supervised Fisher optimization discrimination vector is obtained, and the classification capability of the vector is weakened in sequence;

assuming that the prior probabilities of samples belonging to each class are equal, areK is the total number of categories, then the sample is labeledThe conditional probability density function of (a) is shown as:

wherein Q is_r＝[q₁，q₂，…，q_r]Discriminating the feature vector, Q, for the first r Fisher_rThe spanned space is a semi-supervised fisher discrimination subspace of r dimension,is C_kMean vector of class samples;

according to the Bayesian criterion, the method comprises the following steps of,the posterior probability calculation formula belonging to the i-th type is:

the half-Du Fisher discriminant function is defined as

Representing arbitrary samples in a test datasetA discriminant function value belonging to the kth class in the training set;

predicting each test sample in the test set according to a leak classification criterion as shown belowNormal oil transportation signal, large leakage signal, small leakage signal or working condition interference signal:

step 5.4: respectively training the mathematical models obtained in the step 5.3 by using the normal working condition signal, the large leakage signal, the small leakage signal and the working condition interference signal in the historical data to obtain the corresponding mathematical models of the normal working condition signal, the large leakage signal, the small leakage signal and the working condition interference signalAnd establishing a database of pipeline leakage detection.

7. The pipeline leakage monitoring method based on the acoustic wave negative pressure wave hybrid monitoring is characterized in that: the calculation formula of the similarity degree in the step 6 is as follows:

in the formula, S_kFor similarity, C (x)^newIn order for the data to be newly acquired,for the test sample of the k-th type,for a set of samples of the k-class,a set of all test samples;

if S_iIf 0.85, the current data is the kth class data; otherwise, the type of the fault is determined by combining the actual analysis of the field condition and the artificial diagnosis, and the fault is collected into a database for detecting the pipeline leakage.

8. The pipeline leakage monitoring method based on the acoustic wave negative pressure wave hybrid monitoring is characterized in that: the specific process of positioning the pipeline leakage in the step 7 is as follows:

step 7.1: the leak point is calculated from the difference in time between the upstream and downstream station pressure sensors, as shown,

in the formula, Z₁Distance from the leak point to the upstream station pressure sensor; l₁The distance between the pressure sensors of the upstream and downstream stations of the pipeline; tau is₁The time difference of the negative pressure wave transmitted to the upstream and downstream pressure sensors; v. of₁The propagation speed of the negative pressure wave in the pipeline;

step 7.2: the pressure data mean value 30 seconds before the leakage point and the pressure data mean value 15 seconds after the leakage point are respectively selected at the upstream station and the downstream station of the leakage point, the pressure change ratio of the upstream station and the downstream station is calculated, and the pressure change ratio is respectively shown as the following two formulas,

wherein: delta₁、δ₂The pressure change ratio of the upstream station and the downstream station, respectively; x₁₊、X₂₊The pressure data mean values of an upstream station and a downstream station 30 seconds before the leakage occurrence point are respectively obtained; x_1-、X_2-Respectively mean values of pressure data of an upstream station and a downstream station 15 seconds after a leakage occurrence point;

the total pressure change ratio delta at the time of pipeline leakage_pComprises the following steps:wherein mu is a parameter determined by the length of the on-site pipeline and the pressure of the normal conveying pipeline;

step 7.3: calculating the leakage point according to the time difference between the upstream station and the downstream station of the characteristic value of the sound wave, as shown in the following formula,

in the formula, Z₂The distance from the leakage point to the acoustic wave sensor of the upstream station; l₂The distance between the acoustic wave sensors of the upstream and downstream stations of the pipeline; tau is₂The time difference of the sound wave transmitted to the upstream and downstream sound wave sensors; v. of₂The propagation speed of sound waves in the pipeline;

step 7.4: respectively selecting the mean value of sound wave data 5 seconds before the leakage point and the mean value of sound wave data 3 seconds after the leakage point at the upstream station and the downstream station of the leakage point, calculating the sound wave change ratio of the upstream station and the downstream station, respectively as shown in the following two formulas,

wherein, delta₃、δ₄The acoustic wave change ratios of the upstream station and the downstream station, respectively; y is₁₊、Y₂₊Respectively mean values of sound wave data of an upstream station and a downstream station 5 seconds before a leakage occurrence point; y is_1-、Y_2-Respectively mean values of sound wave data of an upstream station and a downstream station 3 seconds after a leakage occurrence point;

the total sound wave change ratio delta at the time of pipeline leakage_sComprises the following steps:

step 7.5: based on the final distance of the leak from the upstream station pressure sensor

Where T is the period of the acoustic wave.