CN107435817A - A kind of 2 leak detection accurate positioning methods of pressure pipeline - Google Patents

Abstract

The invention discloses a precise positioning method for two-point leakage detection of a pressure pipeline, which uses an acoustic emission leakage detection system and a correlator leakage detection system to collect pipeline leakage signals for the same object at the same time and in the same environment. On the one hand, the particle swarm optimization algorithm based on the idea of simulated annealing is used to perform blind source separation on the leak source signal detected by the acoustic emission leak detection system. At the same time, the memory is embedded to construct and apply the memory simulated annealing particle swarm blind separation method to eliminate multi-point leakage in the pipeline. Due to the resulting dispersion characteristics, a more accurate leak source signal is separated, and the separation time is greatly reduced, so as to determine the time when the leak source signal reaches the upstream and downstream sensors; at the same time, the propagation speed of the leak sound wave in the pipeline is calculated by using the correlator detection data ; Finally, the position of the leakage source is calculated according to the cross-correlation positioning algorithm. Realize the precise location of pressure pipeline leakage, and has the advantages of low cost and convenient use.

Description

A two-point leak detection and precise positioning method for pressure pipelines

technical field

The invention belongs to the technical field of pipeline leakage location detection and relates to a precise location method for pressure pipeline two-point leakage detection.

Background technique

Pipeline transportation is welcomed and valued by people for its advantages of continuous transportation, convenient transportation and low transportation cost. However, due to the natural aging of equipment, climate and environment, and man-made damage, pipeline leakage occurs from time to time. It not only causes a waste of resources, but also pollutes the environment and even threatens people's lives and property. Therefore, it has good economic value and social significance to find an effective pipeline leakage detection method and find out the hidden dangers of the pipeline.

For pipeline detection and positioning, many technical methods have been developed at home and abroad. In actual working conditions, when a pipe section leaks, it is usually a multi-point leak. Therefore, people also began to study the problem of pipeline multi-point leakage location. Verde firstly stated in the article "Multi-leak detection and isolation in fluid pipelines" (Control Engineering Practice, 2001, Vol. 9, No. 6, pp. 673-682) The detection data of the flow sensors and pressure sensors at both ends of the water pipe section can locate the two-point leakage of the pipeline, but the two-point leakage cannot be located in real time; Lei Yang et al. , 2014, No. 09, pp. 109-112) proposed to compare the modulus maximum value of the pressure signal with the signal in the pipeline fault model library, and detect and locate the multi-point leakage of the pipeline through the similarity, but This method requires the original pipeline fault model database; Zhang Chong used optical fiber The sensor detects and locates the multi-point leakage source of the pipeline, but the positioning is not accurate enough and the cost is high. There are also small-scale robots, infrared imaging, and SCADA-based systems to detect multi-point leak sources in pipelines, but these methods are either expensive or the system design is too complicated and not universal.

In the field of non-destructive testing, acoustic emission technology can continuously detect pipeline leaks, and does not require high real-time diagnostics. It can detect pipeline leaks just after they occur, and can detect small pipeline leaks in time, which greatly improves Improve the convenience and correctness of diagnosis. However, the traditional acoustic emission location detection cannot meet the requirements for precise location of multi-point leaks. For example, there are two leak points in the pipeline under test. Complex working conditions and environmental noise make it difficult to identify and extract pipeline leakage signals, resulting in low accuracy in locating the leakage source.

Therefore, the present invention uses the particle swarm algorithm based on the simulated annealing idea for blind source separation, and uses the acoustic emission leak detection system and the correlator leak detection system to collect pipeline leakage signals for the same object (pipeline) at the same time and in the same environment; Process the leak signal detected by the acoustic emission leak detection system, which can eliminate the dispersion characteristics between the signals caused by the multi-point leak of the pipeline, accurately separate the pipeline leak signal, and embed the memory at the same time, construct and apply the memory simulated annealing particle swarm blind separation method , so that the repeated optimal value search is avoided in the annealing process, and the blind source separation time is greatly reduced; and the leakage acoustic signal is collected by a correlator, which solves the problem of pipeline leakage acoustic wave propagation velocity and extracts accurate propagation velocity. This method is applied to the location of multi-point leakage sources in pipelines, and then the precise location of multi-point leakage sources can be realized.

Contents of the invention

In order to solve the deficiencies of the prior art in detecting and locating the two-point leakage source of the pressure pipeline, the present invention proposes a method for precise location of the two-point leakage of the pressure pipeline, so as to realize the separation and precise positioning of the two-point leakage source. This method is applied to the location of pipeline multi-point leakage sources, and then the precise location of leakage multi-point sources can be realized.

The technical solution adopted by the present invention to solve the technical problem is: a precise positioning method for two-point leakage detection of pressure pipelines. Based on the positioning formula (1) calculated by waveform cross-correlation time difference, it can be known that only the acoustic emission signal of the leakage source needs to be determined to reach the upstream and downstream The time difference Δt of the acoustic emission signal and the precise propagation speed v of the acoustic emission signal in the pipeline can determine the position of the leak point, so as to realize the separation and precise positioning of the two leak sources.

In the formula: l is the location of the leakage source of the detected pipeline, that is, the distance from the leakage point to the upstream acoustic emission sensor (m); L is the distance between two acoustic emission sensors (m).

The leakage location method of the present invention specifically includes the following steps,

S1: Build a detection system;

Install two acoustic emission sensors on the upstream and downstream of the pipeline to be tested, and connect the acoustic emission sensors to the acoustic emission instrument to build an acoustic emission leak detection system; at the same time, install the two correlator sensors on the same upstream and downstream of the pipeline to be tested position, and connect the two sensors with the correlator to build a correlator leak detection system;

S2: Determine the time difference Δt between the leakage source signal reaching the upstream and downstream acoustic emission sensors;

S2.1: Acquire the original signal of the leakage source of the pipeline through the acoustic emission leakage detection system;

S2.2: Filter and screen the original signals of the upstream and downstream leakage sources of the pipeline collected by the acoustic emission leakage detection system, and extract the mixed positioning signal data with relatively high effective value voltage RMS value and average signal level ASL value and relatively concentrated peak value as Coarse positioning signal data: According to effective value voltage (RMS), average signal level (ASL), energy and other parameters, the original signal of the detected pipeline leakage source is filtered and extracted to obtain mixed rough positioning signal data.

The application of the acoustic emission leak detection system can detect the location of the pipeline leak and obtain a rough location signal for leak detection. Appropriate methods must be used to process the coarse location signal data of pipeline leaks. For multi-point leakage, the coarse positioning signal source must be separated first, and multi-technical comprehensive processing such as noise reduction must be performed to obtain a more accurate time difference Δt, thereby obtaining a more accurate positioning.

The wavelet denoising technology has been widely recognized and applied, and it is simple and convenient to use, and the denoising effect is better. Therefore, in step S2.3, the wavelet denoising technology is used to denoise the rough positioning signal data to obtain the observation signal.

The specific process of wavelet denoising is as follows:

S2.3.1: Wavelet decomposition of the signal. First of all, it is necessary to select the appropriate wavelet base for different signals, and determine the level to be decomposed, and then carry out the decomposition calculation.

S2.3.2: Threshold quantization of high-frequency coefficients of wavelet decomposition. It is necessary to select an appropriate threshold to quantify the high-frequency coefficients at each decomposition scale. Here, a soft threshold is selected to quantify it. Specifically, the soft threshold denoising method in the wavelet threshold denoising in Matlab is used for processing.

S2.3.3: Wavelet reconstruction. One-dimensional wavelet reconstruction is carried out according to the high-frequency coefficients of each layer and the low-frequency coefficients of the bottom layer decomposed by wavelet.

Specifically, it includes importing two coarse positioning signals obtained from the upstream and downstream of the leakage pipeline from the acoustic emission leak detection system into the wavelet analysis module in the MATLAB toolbox, and performing wavelet decomposition, threshold quantization and Wavelet reconstruction to derive the signal after denoising, which is the observation signal I ₁ and I ₂ , where I ₁ represents the signal obtained by the upstream sensor, and I ₂ represents the signal obtained by the downstream sensor;

S2.4: Mix the observed signals I ₁ and I ₂ obtained after noise reduction with the randomly generated matrices in Matlab to form new observed signals L ₁ and L ₂ ; since the number of observed signals is greater than or equal to the number of source signals, This step is used to change the original underdetermined blind source separation problem into a positive definite blind source separation problem.

S2.5: Import the new observation signals L ₁ and L ₂ into Matlab for whitening processing, and then use the blind source separation method of memory simulated annealing particle swarm to perform blind source separation to obtain the separated upstream and downstream leakage source location signals S ₁ , S ₂ ; the blind source separation can be simplified and the blind source separation algorithm can be improved through the whitening process, making the blind source separation process more convenient.

In this algorithm, the maximum likelihood function is used as the objective function. Since the maximum likelihood method can be asymptotically effective when the number of samples is large, an optimal solution can be obtained, so the maximum likelihood algorithm is used to optimize the results; the blind source separation of the particle swarm algorithm based on the simulated annealing idea can improve And get rid of local extremum points, the ability of local optimum, and high separation precision and high stability.

The specific steps of the memory simulated annealing particle swarm blind source separation are as follows:

Since the basic model of blind source separation is:

L(t)=A(t)s(t) (2)

Among them, L(t) is the observation signal matrix, A(t) is the mixing matrix, and s(t) is the source signal matrix.

The model for solving the source signal s(t) is:

y(t)=W(t)L(t) (3)

Among them, y(t) is the output signal matrix after separation, and W(t) is the unmixing matrix.

S2.5.1: Initial parameter setting: Set the number of particles in the particle swarm to n, and initialize each particle, the weight is w, the cognitive factor and social learning factor are c ₁ and c ₂ respectively, and a certain number of The unmixing matrix W(t) of is used as the initial particle, at the same time, the initial velocity of each particle is randomly generated, and the individual extremum and the global extremum are initialized, given the starting temperature T, the termination temperature T ₀ and the simulated annealing speed λ ;

S2.5.2: Separate the signal according to the position of the particle, perform centering and whitening operations on y(t), and use the maximum likelihood estimation function as the objective function to calculate the fitness value of each particle;

S2.5.3: Take the fitness value of each particle as the individual optimal extremum p _i of the particle, and select the optimal value among the individual extremums as the global extremum p _g ;

S2.5.4: Judging whether the termination condition is satisfied, if it is satisfied, the calculation is terminated, otherwise continue;

S2.5.5: Compare the fitness value of each particle with the individual extremum p _i and the global extremum p _g , and take the optimal value to update the individual extremum p _i and the global extremum p _g of each particle;

S2.5.6: Update the velocity and position of the particle, and limit the range of the given maximum velocity and maximum position respectively, and calculate the fitness value of each updated particle;

S2.5.7: Calculate the variable ΔE of the fitness value caused by the two particle positions before and after, if ΔE<0, accept the new position; if exp(-ΔE/T)<δ, δ∈(0,1) random number, also accept the new position, otherwise reject and return to step S2.5.2;

S2.5.8: Embed and set the initial position and adaptive value of the memory variable, which is the optimal position and adaptive value of the first cycle;

S2.5.9: Compare the new position and the adaptation value with the stored position and adaptation value in the memory, if the new position and the adaptation value are the same as the position and the adaptation value in the memory, return to step S2.5.2, otherwise record it into the memory;

S2.5.10: Perform annealing operation, T _(t+1) = λT _t (t is the number of iterations);

S2.5.11: If the termination condition is met, output the optimal solution, otherwise return to step S2.5.2;

S2.5.12: Find the optimal W(t), and find the optimal estimate of the source signal s(t).

Add the memory to record the optimal solution, compare the optimal solution obtained each time with the optimal solution recorded before, and eliminate the repeated optimal solution, effectively avoiding the roundabout search method, thereby reducing the search time and solving the problem. The time-consuming problem of the annealing particle swarm method is solved.

S2.6: Analyze the positioning signals S ₁ and S ₂ through wavelet singular points to obtain the singular points of the signals; determine the time difference Δt between the leakage source signal reaching the upstream and downstream acoustic emission sensors according to the difference between the sampling points of the two singular points;

Since the propagation speed of the acoustic emission signal in the material is affected by many factors such as material type, anisotropy, structural shape and size, and internal medium, the propagation speed becomes a variable. And because the acoustic emission signal also has dispersion phenomenon and is affected by the frequency of the wave, it is difficult to determine the sound velocity of the leakage sound wave in actual working conditions.

The sound velocity of pipeline leakage is not only affected by the pipeline material, but also by different media and different working conditions, and the wave velocity detected by different methods is different. At present, there is no unified method at home and abroad to calculate the wave velocity of pipeline leakage sound waves , the wave velocity values detected or calculated by researchers are also inconsistent. For example, Sun Liying of Tianjin University et al. in the article "Research on the Propagation and Attenuation Characteristics of Acoustic Emission Waves in Liquid-Filled Pipelines" (Piezoelectricity and Acousto-Optics, August 2008, Volume 30, Issue 4, Pages 401-403) It is believed that when the acoustic emission wave is in a steel pipeline, the wave velocity is 3300m/s when the medium is air, and 1500m/s under the action of water load; and Shen Gongtian in "Acoustic Emission Detection Technology and Application" (Science Press, 2015 edition, No. 242-243) in the book, it is believed that the acoustic emission signal generated by steel pipeline leakage has a wave velocity of 880-960m/s when the medium is air; and Didem Ozevin in "Novel leak localization in pressurized pipeline networks using Acoustic emission and geometric connectivity" (International Journal of Pressure Vessels & Piping, 2012, Vol. 92, No. 2, Pages 63-69) uses modal calculations to calculate the acoustic emission in PVC pipes where the medium is air and the wave velocity is 1479m/s. How to obtain a relatively objective and accurate sound velocity of the leakage sound wave is another important key to be solved in this patent.

The cross-correlation technique is applicable not only to the time difference or time delay measurement between discontinuous waves, but also to the time difference or time delay measurement between continuous waves. This technology has been successfully applied to the location of leakage sources in pipeline acoustic emission detection. The correlator detects pipeline leaks based on the principle of cross-correlation technology, but care should be taken to use the correct sound velocity.

The placement of the two sensors of the correlator is the same as that of the acoustic emission pipeline leak detection sensor. Correlator is used to locate and detect pipeline leakage, and cross-correlation function analysis is usually realized by dual-channel fast Fourier transform (FFT) analysis. From the inverse Fourier transform of cross-correlation spectrum G _AB (v) in frequency domain v, the time The cross-correlation function R _AB (τ) in the domain τ:

In the formula, G _AB (v) is the Fourier transform of A(t)B(t+τ), where A(t) represents a wave, B(t+τ) represents another wave with a delay time of τ, and G The result of _AB (v) is leaking location data. This method can obtain relatively accurate location for the single-point leakage of the pipeline under test, and the location error for multi-point leakage is also relatively large. However, when the probability of the positioning result is greater than 4%, it can be considered that the result is relatively objective and accurate, which can be verified from multiple experimental results (see experimental data table 1). Therefore, we decided to use the correlator to measure and determine the wave velocity.

The working condition of experiment 1 is: a PE pipe with a length of 2030cm, the internal medium is compressed air, the pressure is 0.5MPa, the pipeline is buried, the leakage hole is 1600cm away from the upstream sensor of the pipeline, and the leakage hole diameter is 2mm.

The working condition of experiment 2 is: a steel pipe with a pipe length of 3600cm, the internal medium is compressed air, the pressure is 0.57MPa, the pipe is empty, the leak hole is 1600cm away from the upstream sensor of the pipe, and the leak hole diameter is 1mm.

The working condition of experiment 3 is: a steel pipe with a pipe length of 4275cm, the internal medium is compressed air, the pressure is 0.28MPa, the pipe is empty, the leak hole is 1706cm away from the upstream sensor of the pipe, and the leak hole diameter is 1mm.

Table 1 Correlator pipeline leak location experiment data

Calculating and determining the wave velocity of the leakage signal specifically includes the following steps:

S3: Determine the propagation velocity v of the leakage source signal in the pipeline;

S3.1: Use the correlator to collect the acoustic signal of the pipeline leakage source;

S3.2: According to the collected acoustic signal of the pipeline leakage source, through the analysis of the location result detection and the probability analysis of the location result by the correlator, determine the propagation velocity v of the leakage source acoustic signal in the pipeline medium; in order to ensure the objectivity of the data, the correlation The instrument analyzes the positioning result detection and the probability analysis of the positioning result. Under the same working condition, three batches of collected data are taken, and the average speed obtained by analyzing 10 sets of data in each batch Finally, the propagation velocity v of the acoustic signal of the leakage source in the pipeline medium is calculated.

S4: The time difference Δt of the leakage source signal calculated by step S2.6 reaching the upstream and downstream acoustic emission sensors and the velocity v of the leakage source signal propagated in the pipeline medium obtained by step S3.2, according to the cross-correlation positioning formula (1 ) to calculate the location of the leak source.

In the formula: l is the location of the leakage source of the detected pipeline, that is, the distance from the leakage point to the upstream acoustic emission sensor (m); L is the distance between two acoustic emission sensors (m).

The above data analysis and processing processes are realized by Matlab programming.

The beneficial effects of the present invention are: the present invention provides a precise positioning method for two-point leak detection of pressure pipelines, which uses the acoustic emission leak detection system and the correlator leak detection system to collect pipeline data for the same object at the same time and under the same environment. Leakage signal: the upstream and downstream pipeline leakage source signals are collected by the acoustic emission instrument, the influence of noise on the leakage source signal is reduced by using wavelet denoising technology, and the leakage signal detected by the detection signal is blindly sourced by using the particle swarm optimization algorithm based on the idea of simulated annealing. Separation processing, embedding memory at the same time, constructing and applying memory simulated annealing particle swarm blind separation method, not only can eliminate the dispersion characteristics of the signal caused by multi-point leakage in the pipeline, but also separate more accurate leakage source signals, greatly reducing the separation time, and This determines the time when the leak source signal arrives at the upstream and downstream sensors; then uses the leak signal collected by the correlator leak detection system to analyze the precise propagation speed of the acoustic signal in the pipeline; finally calculates the position of the leak source based on the cross-correlation positioning algorithm. The invention can accurately locate the leakage of the pressure pipeline, and provides a method for locating multi-point leakage of the pipeline with low cost, convenient use, and the ability to identify tiny leakage.

Description of drawings

The present invention will be further described below in conjunction with drawings and embodiments.

Fig. 1 is a schematic flow sheet of a preferred embodiment of the present invention;

Fig. 2 is a schematic diagram of a pressure pipeline two-point leakage device;

Fig. 3 is the spectrum diagram of sensor No. 1;

Fig. 4 is sensor No. 2 spectrum diagram;

Fig. 5 is the waveform diagram of No. 1 sensor;

Fig. 6 is sensor No. 2 waveform diagram;

Fig. 7 is an acoustic emission detection effective voltage RMS positioning diagram;

Fig. 8 is the ASL positioning diagram of the average signal level of the acoustic emission detection;

Fig. 9 is the observation signal after sensor No. 1 (upstream sensor) noise reduction;

Fig. 10 is the observation signal after sensor No. 2 (downstream sensor) noise reduction;

Figure 11 is the two new observation signals of sensors No. 1 and No. 2;

Fig. 12 is sensor No. 1, No. 2 separation signals;

Fig. 13 is a curve display diagram of a correlator detecting a pipeline leak point;

Fig. 14 is a probability diagram of the location result of the pipeline leakage point by the correlator.

detailed description

The present invention will be described in detail in conjunction with accompanying drawing now. This figure is a simplified schematic diagram only illustrating the basic structure of the present invention in a schematic manner, so it only shows the components relevant to the present invention.

As shown in Figure 1, a kind of pressure pipeline two-point leak detection accurate positioning method of the present invention specifically comprises the following steps,

S1: Build a detection system;

Install two acoustic emission sensors on the upstream and downstream of the pipeline to be tested, and connect the acoustic emission sensors to the acoustic emission instrument to build an acoustic emission leak detection system; at the same time, install the two correlator sensors on the same upstream and downstream of the pipeline to be tested position, and connect the two sensors with the correlator to build a correlator leak detection system;

S2: Determine the time difference Δt between the leakage source signal reaching the upstream and downstream acoustic emission sensors;

S2.1: Acquire the original signal of the leakage source of the pipeline through the acoustic emission leakage detection system;

S2.2: Screen the original signals of the upstream and downstream leakage sources of the pipeline collected by the acoustic emission leak detection system, and extract the mixed positioning signal data with relatively high effective value voltage RMS value and average signal level ASL value and relatively concentrated peak value as rough positioning signal data;

S2.3: Use wavelet denoising technology to denoise the coarse positioning signal to obtain the observation signal;

Noise reduction processing specifically includes,

S2.3.1: Wavelet decomposition of signals, select appropriate wavelet bases for different signals, and determine the levels to be decomposed, and then perform decomposition calculations;

S2.3.2: Threshold quantization of wavelet decomposition high-frequency coefficients, select an appropriate threshold to quantify the high-frequency coefficients under each decomposition scale, the threshold selects soft thresholds to quantify it;

S2.3.3: Wavelet reconstruction, perform one-dimensional wavelet reconstruction according to the high-frequency coefficients of each layer of wavelet decomposition and the low-frequency coefficients of the bottom layer.

S2.4: Mix the observed signal obtained after noise reduction with the matrix randomly generated in Matlab to form a new observed signal;

S2.5: Import the new observation signal into Matlab for whitening processing, and then use the blind source separation method of memory simulated annealing particle swarm to perform blind source separation to obtain the separated upstream and downstream leakage source positioning signals S ₁ and S ₂ ;

The specific steps of the memory simulated annealing particle swarm blind source separation include,

S2.5.1: Initial parameter setting: Set the number of particles in the particle swarm to n, and initialize each particle, the weight is w, the cognitive factor and social learning factor are c ₁ and c ₂ respectively, and a certain number of The unmixing matrix W(t) of is used as the initial particle, at the same time, the initial velocity of each particle is randomly generated, and the individual extremum and the global extremum are initialized, given the starting temperature T, the termination temperature T ₀ and the simulated annealing speed λ ;

S2.5.2: Separate the signal according to the position of the particle, perform centering and whitening operations on y(t), and use the maximum likelihood estimation function as the objective function to calculate the fitness value of each particle;

S2.5.3: Take the fitness value of each particle as the individual optimal extremum p _i of the particle, and select the optimal value among the individual extremums as the global extremum p _g ;

S2.5.4: Judging whether the termination condition is satisfied, if it is satisfied, the calculation is terminated, otherwise continue;

S2.5.5: Compare the fitness value of each particle with the individual extremum p _i and the global extremum p _g , and take the optimal value to update the individual extremum p _i and the global extremum p _g of each particle;

S2.5.6: Update the velocity and position of the particle, and limit the range of the given maximum velocity and maximum position respectively, and calculate the fitness value of each updated particle;

S2.5.7: Calculate the variable ΔE of the fitness value caused by the two particle positions before and after, if ΔE<0, accept the new position; if exp(-ΔE/T)<δ, δ∈(0,1) random number, also accept the new position, otherwise reject and return to step S2.5.2;

S2.5.8: Embed and set the initial position and adaptive value of the memory variable, which is the optimal position and adaptive value of the first cycle;

S2.5.9: Compare the new position and the adaptation value with the stored position and adaptation value in the memory, if the new position and the adaptation value are the same as the position and the adaptation value in the memory, return to step S2.5.2, otherwise record it into the memory;

S2.5.10: Perform annealing operation, T _(t+1) = λT _t (t is the number of iterations);

S2.5.11: If the termination condition is met, output the optimal solution, otherwise return to step S2.5.2;

S2.5.12: Find the optimal W(t), and find the optimal estimate of the source signal s(t).

S2.6: Analyze and locate the signal through the wavelet singular point to obtain the singular point of the signal; determine the time difference Δt of the leakage source signal reaching the upstream and downstream acoustic emission sensors according to the sampling point difference between the two singular points;

S3: Determine the precise propagation speed v of the leakage source signal in the pipeline;

S3.1: Use the correlator to collect the acoustic signal of the pipeline leakage source;

S3.2: According to the collected acoustic emission signal data of the pipeline leakage source, analyze the location results through the correlator and analyze the probability analysis of the location results to determine the propagation velocity v of the leakage source acoustic signal in the pipeline medium;

S4: The time difference Δt of the leakage source signal calculated by step S2.6 reaching the upstream and downstream acoustic emission sensors and the velocity v of the leakage source signal propagated in the pipeline medium obtained by step S3.2, according to the cross-correlation positioning formula (1 ) to calculate the location of the leak source.

In the formula: l is the location of the leakage source of the detected pipeline, that is, the distance (m) from the leakage point to the upstream acoustic emission sensor; L is the distance (m) between two acoustic emission sensors.

Carry out the simulated leakage experiment according to the above steps, as shown in Figure 2, first, set up the detection system, in this embodiment, a section of steel pipeline with a pipe diameter of DN150 is used, its nominal diameter is 150mm, the length of the experimental pipeline is 44m, and the pressure is 0.1 MPa, the medium inside the pipeline is tap water; the downstream acoustic emission sensors are respectively placed at 1m and 43m, as shown in the figure, acoustic emission sensor No. 1 and acoustic emission sensor No. 2, and the two leakage holes are respectively set at 19m and 33m from the zero point. The hole diameter is 1mm, and the simulated leakage experiment is carried out.

Collected through the leak test, Figures 3-6 are the signal spectrum diagrams obtained by the acoustic emission leak detector in the two-point leak test of the liquid-filled pipeline at a pressure of 0.1 MPa and a leak aperture of 1 mm (Figures 3 and 4 ) and waveform diagrams (Figure 5, Figure 6), as well as the effective value voltage RMS positioning diagram (Figure 7), and the average signal level ASL positioning diagram (Figure 8).

Among them, the abscissa of the signal spectrogram represents frequency (Hz), and the ordinate represents power (dB); the abscissa of the waveform diagram represents time (s), and the ordinate represents voltage (mV); The coordinates represent the distance (mm) from the leak point to the upstream acoustic emission sensor, the ordinate represents the effective value voltage RMS (V); the abscissa of the average signal level ASL location map represents the distance (mm) from the leak point to the upstream acoustic emission sensor, The ordinate represents the average signal level (dB).

The effective value voltage RMS location map and the average signal level ASL location map provide a reference basis for extracting pipeline acoustic emission leakage signal parameters, so that part of the signal can be extracted as rough location signal data.

From Figure 7 and Figure 8, it can be seen that there are mixed positioning signals with high amplitude and relatively concentrated peaks near the two places of 19m and 33m, but there are many noise signals at the same time; these mixed positioning signals are extracted through the pipeline acoustic emission leakage signal parameters Part of the signal data in the positioning signal data is used as rough positioning signal data (see Table 2).

Table 2 Data table of coarse positioning signal for experimental location of pipeline two-point leakage source

The above content is the content of S2.1-S2.2.

S2.3 Use wavelet denoising technology to denoise the coarse positioning signal to obtain the observation signal.

In order to reduce the influence of surrounding noise, the original signal is denoised by wavelet denoising, and this step is realized by Matlab software.

The wavelet signal denoising process is divided into the following steps:

S2.3.1: Wavelet decomposition of the signal. First of all, it is necessary to select the appropriate wavelet base for different signals, and determine the level to be decomposed, and then carry out the decomposition calculation.

S2.3.2: Threshold quantization of high-frequency coefficients of wavelet decomposition. It is necessary to select an appropriate threshold to quantify the high-frequency coefficients at each decomposition scale, and in the present invention, select a soft threshold to quantize it.

S2.3.3: Wavelet reconstruction. One-dimensional wavelet reconstruction is carried out according to the high-frequency coefficients of each layer and the low-frequency coefficients of the bottom layer decomposed by wavelet.

Import the two coarse positioning signals obtained from the upstream and downstream of the leaking pipeline from the acoustic emission leak detection system into the wavelet analysis module in the MATLAB toolbox, and perform noise analysis on the obtained pipeline leakage acoustic emission signal through the above steps S2.3.1-S2.3.3. After denoising processing, the obtained signals after noise reduction are observation signals I ₁ and I ₂ , where I ₁ represents the signal obtained by the upstream sensor, and I ₂ represents the signal obtained by the downstream sensor. As shown in Fig. 9 and Fig. 10, they are respectively the observation signal diagrams of the rough positioning signals received by the upstream and downstream acoustic emission sensors of the pipeline after denoising. The abscissas in FIGS. 9 and 10 both represent sampling points, and the ordinates represent amplitude values (V).

S2.4 Mix the observed signals I ₁ , I ₂ obtained after denoising with matrices randomly generated in Matlab to form new observed signals L ₁ , L ₂ ; since the number of observed signals is greater than or equal to the number of source signals, use This step changes the original underdetermined blind source separation problem into a positive definite blind source separation problem.

In this step, the number of observed signals is greater than or equal to the number of source signals, so that the original underdetermined blind source separation problem is changed into a positive definite blind source separation problem. As shown in Fig. 11, two new observation signals L ₁ and L ₂ are obtained, which are observation signal diagrams formed by upstream and downstream signals and a randomly generated mixing matrix, respectively.

S2.5 Import the new observation signals L ₁ , L ₂ into Matlab for whitening processing, and then use the blind source separation method of memory simulated annealing particle swarm to perform blind source separation to obtain the separated upstream and downstream leakage source positioning signals S ₁ , S ₂ ; the blind source separation can be simplified and the blind source separation algorithm can be improved through the whitening process, making the blind source separation process more convenient.

S2.6 Analyze the positioning signals S ₁ and S ₂ according to the singular point of the upstream and downstream positioning signals, and obtain the time difference Δt between the arrival of the acoustic emission signal of the leakage source to the upstream and downstream sensors. It can be known from the singular point of the positioning signal of the upstream and downstream (Figure 12), the singular point of the positioning signal of the upstream No. 1 sensor is near the 270th sampling point, and the singular point of the positioning signal of the downstream No. 2 sensor is near the 890th sampling point , it can be obtained that the time difference Δt between the signal transmission to the upstream and downstream sensors is 0.0062s. Wherein, the abscissas in FIG. 11 and FIG. 12 both represent sampling points, and the ordinates both represent amplitude values (V).

S3: Determine the precise propagation speed v of the acoustic emission signal in the pipeline;

S3.1: Place the two sensors of the correlator on the upstream and downstream of the pipeline to be detected. The positions of the two sensors of the correlator are the same as those of the acoustic emission instrument. The two sensors of the correlator (correlator acoustic sensor No. Acoustic sensor No. 2) is placed as shown in Figure 2, and a correlator is used to collect pipeline leakage acoustic signals;

S3.2: According to the collected sound signal of pipeline leakage, the detection and probability analysis of the positioning results are analyzed by the correlator; for the same working condition, the average speed obtained from the analysis of three batches of collected data That is, the propagation velocity v of the acoustic emission of the leakage source in the pipeline medium.

Use the correlator to set parameters, and analyze the data for the wave speed whose positioning probability result is greater than 4%, and extract the corresponding wave speed. By analyzing the curve display graph of the correlator to detect the pipeline leak point and the probability display of the location result of the pipeline leak point by the correlator (as shown in Figure 13 and Figure 14), wherein, the abscissa of Figure 13 and Figure 14 represents the correlation between the leak point and the upstream The distance (cm) of the acoustic wave sensor of the correlator, the ordinate represents the amplitude value (V), and the location value with a higher probability of occurrence in the three batches of location data in the correlator is extracted, as well as the corresponding wave velocity v, as shown in Table 3.

Table 3 Correlator positioning data table

From Table 3, we can calculate that the wave velocity of the leakage sound wave under this working condition is 1150m/s.

S4: The time difference Δt of the leakage source signal calculated by step S2.6 reaching the upstream and downstream acoustic emission sensors and the velocity v of the leakage source signal propagated in the pipeline medium obtained by step S3.2, according to the cross-correlation positioning formula (1 ) to calculate the location of the leak source.

Put the obtained wave velocity v and time difference Δt into formula (1), and one of the leak location points can be obtained as 19.44m. And use this method to calculate other leak location points and calculate the relative error before and after data processing, as shown in Table 4 and Table 5.

Table 4 Leakage source location results and relative errors at 19m

serial number Coarse positioning leak location (m) Measurement relative error (%) Leakage point after treatment (m) Relative error after processing (%) 1 15.96 16.0 17.25 9.2 2 21.02 10.6 19.78 4.1 3 19.84 4.4 19.55 2.9 4 19.42 2.2 19.32 1.7 5 19.47 2.5 19.26 1.4 6 20.33 7.0 19.67 3.5 7 20.71 9.0 19.44 2.3 8 20.98 10.4 19.9 4.7 9 19.16 0.8 19.21 1.1 10 21.18 11.5 20.13 5.9 11 17.41 8.4 17.65 7.1 12 16.55 12.9 17.37 8.6 13 21.01 10.6 19.78 4.1 14 19.39 2.1 19.21 1.1 average value 19.5 7.7 19.1 4.1

Table 5 Leakage source location results and relative errors at 33m

After analyzing and processing all the data, it can be seen from Table 4 and Table 5 that the position of the leak point after analysis is basically consistent with the design 19m and 33m, and the positioning accuracy is better than the leakage detection results of the acoustic emission instrument and the correlator. it is good. The relative error measured at the 19m leak point is 7.7%, and the relative error after processing is 4.1%; the relative error measured at the 33m leak point is 76.4%, and the relative error after processing is 4.0%. It can be seen that after data processing using this method, the error in leak location can be greatly reduced, and the method is low in cost and easy to use.

A method for detecting and accurately locating two-point leakage sources of pressure pipelines provided by the present invention has been introduced in detail above. The principles and implementation methods of the present invention have been described by using specific experimental examples. It should be noted that the above descriptions are only preferred embodiments of the present invention, and are not intended to limit the present invention. All modifications, equivalent replacements and improvements made within the spirit and principles of the present invention shall be included within the protection scope of the present invention.

Claims (3)

1. A kind of 1. 2 leak detection accurate positioning methods of pressure pipeline, it is characterised in that：Comprise the following steps,

   S1：Build detecting system；

   Two acoustic emission sensors are arranged on to the upstream and downstream for being detected pipeline, and acoustic emission sensor is connected with Acoustic radiating instrument Connect and be built into sound emission leak detection system；Meanwhile pairwise correlation instrument sensor is arranged on and is detected ducts upstream and downstream Same position, and two sensorses is connected with correlator and be built into correlator leak detection system；

   S2：It is determined that leakage source signal reaches the time difference Δ t of the acoustic emission sensor of upstream and downstream two；

   S2.1：The source of leaks primary signal of pipeline is gathered by sound emission leak detection system；

   S2.2：The pipeline upstream and downstream source of leaks primary signal collected to sound emission leak detection system carries out filtering screening, carries Take RMS voltage RMS value and average signal level ASL values are of a relatively high and the mixed positioning signal data of peak value Relatively centralized As coarse positioning signal data；

   S2.3：Noise reduction process is carried out to coarse positioning signal using Wavelet Denoising Technology, obtains observation signal；

   S2.4：The matrix generated at random in the observation signal and Matlab that are obtained after noise reduction is mixed to form new observation signal；

   S2.5：New observation signal is imported in Matlab and carries out whitening processing, recycles the blind of memory simulated annealing population Source separation method carries out the leak position signal S of the upstream and downstream after blind source separating is separated₁、S₂；

   S2.6：By wavelet singular point analysis positioning signal, singular points are obtained；It is poor according to the sampled point of two singular points Value determines that leakage source signal reaches the time difference Δ t of two acoustic emission sensors of upstream and downstream；

   S3：It is determined that the spread speed v of leakage source signal in the duct；

   S3.1：Pipeline source of leaks acoustical signal is gathered using correlator；

   S3.2：According to the pipe leakage source acoustical signal data collected, detected by correlator analyzing and positioning result and tied with positioning Fruit probability analysis, determine spread speed v of the source of leaks acoustical signal in pipeline medium；

   S4：The leakage source signal being calculated by step S2.6 travel to upstream and downstream acoustic emission sensor time difference Δ t and The spread speed v of the leakage source signal obtained by step S3.2 in the duct, leakage is calculated according to cross-correlation ranging formula (1) Source position；

   <mrow> <mi>l</mi> <mo>=</mo> <mfrac> <mrow> <mi>L</mi> <mo>-</mo> <mi>&amp;Delta;</mi> <mi>t</mi> <mo>&amp;CenterDot;</mo> <mi>v</mi> </mrow> <mn>2</mn> </mfrac> <mo>-</mo> <mo>-</mo> <mo>-</mo> <mrow> <mo>(</mo> <mn>1</mn> <mo>)</mo> </mrow> </mrow>

   In formula：L is to be detected pipe leakage source position, i.e. distance (m) of the leakage point to upstream acoustic emission sensor；L is two sound The distance between emission sensor (m).
2. 2. 2 leak detection accurate positioning methods of pressure pipeline as claimed in claim 1, it is characterised in that：The step Noise reduction process in S2.3 specifically includes,

   S2.3.1：The wavelet decomposition of signal, to the different suitable wavelet basis of signal behavior, and determine the layer to be decomposed It is secondary, decomposition computation is then carried out again；

   S2.3.2：The threshold value quantizing of wavelet decomposition high frequency coefficient, a suitable threshold value is selected under each decomposition scale High frequency coefficient is quantified, and the threshold value selects soft-threshold to carry out quantification treatment to it；

   S2.3.3：Wavelet reconstruction, carried out according to the low frequency coefficient of the high frequency coefficient of each layer of wavelet decomposition and the bottom one-dimensional small Reconstructed wave.
3. 3. 2 leak detection accurate positioning methods of pressure pipeline as claimed in claim 1, it is characterised in that：The step The specific steps of memory simulated annealing population blind source separating described in S2.5 include,

   S2.5.1：Initial parameter sets：Population population is set as n, and each particle is initialized, weight w, is recognized Know that the factor and social learning's factor are respectively c₁、c₂, a number of solution hybrid matrix W (t) will be randomly generated and be used as initial grain Son, can simultaneously randomly generate the initial velocity of each particle, and initialize individual extreme value and global extremum, give starting temperature Spend T, final temperature T₀With simulated annealing speed λ；

   S2.5.2：Signal is separated according to the position of particle, centralization and whitening operation are carried out to y (t), according to Maximum-likelihood estimation Function calculates the adaptive value of each particle with this as object function；

   S2.5.3：Using the adaptive value of each particle as the optimal extreme value p of particle individual_i, and choose optimal value in individual extreme value and make For global extremum p_g；

   S2.5.4：Judge whether to meet end condition, terminate and calculate if meeting, otherwise continue；

   S2.5.5：By each particle adaptive value and individual extreme value p_iWith global extremum p_gIt is compared, takes optimal value to update each grain The individual extreme value p of son_iWith global extremum p_g；

   S2.5.6：The speed of more new particle and position, and in the range of limiting given maximal rate and maximum position respectively, and Calculate the adaptive value of each more new particle；

   S2.5.7：The variable Δ E of the adaptive value caused by former and later two particle positions is calculated, if Δ E ＜ 0, receive new position； If the random number between exp (- Δ E/T) ＜ δ, δ ∈ (0,1), also receives new position, otherwise refuse and return to step S2.5.2；

   S2.5.8：It is embedded in and memory variable initial position and adaptive value is set, the optimal location as circulated for the first time is with fitting It should be worth；

   S2.5.9：Compare storage location and adaptive value in new position and adaptive value and memory, if new position and adaptive value and Position then return to step S2.5.2 identical with adaptive value in memory, on the contrary it is recorded into memory；

   S2.5.10：Carry out annealing operation, T_(t+1)=λ T_t(t is iterations)；

   S2.5.11：If meeting end condition, optimal solution is exported, otherwise return to step S2.5.2；

   S2.5.12：Ask to obtain that W (t) is optimal, solve source signal s (t) optimal estimation.