Gas pipeline leakage acoustic detection device and method based on CPO-VMD and multi-feature extraction

Abstract

The invention discloses a gas pipeline leakage acoustic detection device and method based on CPO-VMD and multi-feature extraction, and relates to the technical field of gas pipeline leakage detection, and the device comprises a compressed air storage tank, a pressure control valve, a barometer, a pipeline, a seamless steel pipe, a broadband microphone, a data acquisition card and a data storage device. The method using the device comprises the following steps: collecting a sound wave signal of gas pipeline leakage; carrying out adaptive decomposition on the original signal in combination with a CPO algorithm and variational mode decomposition; performing reconstruction and noise reduction on the key modal component through the correlation coefficient; and multi-dimensional features of the reconstructed signals are extracted, the signal features are screened and preferentially selected, and gas pipeline leakage detection and classification are performed through a classification detector. Therefore, by adopting the method, the decomposition effect and the classification precision of the leakage sound wave signals can be effectively improved, and the strong adaptability of a method framework in different scenes is enhanced.

Description

Technical Field

[0001] The present invention relates to the technical field of gas pipeline leakage detection, and in particular to a gas pipeline leakage acoustic detection method based on CPO-VMD and multi-feature extraction. Background Art

[0002] Gas pipelines are one of the important transportation carriers for hazardous chemical raw materials such as natural gas and hydrogen. However, due to pipeline aging or external impact, valve seals may fail and cause leakage, or pipeline cavitation may cause leakage. In recent years, explosion accidents caused by gas pipeline leakage have occurred frequently, posing a serious threat to the safety of life and property. In order to further ensure the safe and stable operation of the pipeline system, an accurate and reliable gas pipeline leakage detection method is urgently needed.

[0003] According to different detection principles, gas pipeline leakage detection methods can be divided into distributed optical fiber method, negative pressure wave method, acoustic emission method and acoustic wave method. The optical fiber method can detect the location of the leak point, but the accuracy is not high. It can usually only determine the leak area and it is difficult to accurately locate the specific leak point. The negative pressure wave method is not sensitive enough to small-scale leaks and it is difficult to detect smaller leaks. Therefore, it is not effective for pipelines with smaller leaks. The acoustic wave method receives the acoustic wave signal generated by the gas pipeline leak through a microphone and identifies the leakage event based on the acoustic signal processing. Compared with other methods, the acoustic wave method has the advantages of high sensitivity, rapid response and non-contact detection, and is suitable for large-scale application in industrial scenarios.

[0004] Since the acoustic wave signals generated by leakage have highly nonlinear and non-stationary characteristics and are easily disturbed by environmental noise, it is very necessary to reduce the noise of the leakage acoustic wave signals. Commonly used noise reduction methods include Fourier transform and wavelet transform based on wave signal processing, but the noise reduction effect of Fourier transform is limited by the window function setting, while the noise reduction effect of wavelet transform is limited by the mother wave selection. Subsequently, noise reduction methods based on signal decomposition have been gradually developed, such as empirical mode decomposition (EMD) and ensemble empirical mode decomposition (EEMD). EMD and its evolution method can realize the adaptive decomposition of acoustic wave signals, but lack strict mathematical theoretical support, and the decomposition results have defects such as mode aliasing and end effect. Variational mode decomposition (VMD) can realize the decomposition of signals in a narrow frequency domain by constructing a set of Wiener filters, and the number of decomposition layers is controllable. The effectiveness of the VMD algorithm has been fully verified in many fields such as industrial equipment fault diagnosis, economic development forecasting, and climate and environmental forecasting. The number of decomposition layers K and the penalty factor α are the two core parameters of the VMD algorithm, which have a significant impact on the signal decomposition effect. However, manually setting parameters requires high experience from researchers, is time-consuming and complicated, and unreasonable parameter value settings will greatly weaken its decomposition effect and noise reduction performance. At the same time, the current VMD optimization decomposition algorithm generally has problems such as insufficient convergence and robustness, resulting in low repeatability of the decomposition algorithm results, and the decomposition effect still needs to be further improved.

[0005] Therefore, it is necessary to provide a gas pipeline leakage acoustic detection method with adaptive signal decomposition to make the detection results accurate and reliable, thereby ensuring the safe and stable operation of the gas pipeline system. Summary of the invention

[0006] The purpose of the present invention is to provide a gas pipeline leakage acoustic detection device and method based on CPO-VMD and multi-feature extraction, which can effectively improve the decomposition effect and classification accuracy of leakage acoustic wave signals and enhance the adaptability of the method framework in different scenarios.

[0007] To achieve the above object, the present invention provides a gas pipeline leakage acoustic detection device based on CPO-VMD and multi-feature extraction, comprising:

[0008] Compressed air storage tank, used to provide air flow, the generated air pressure provides the gas source conditions for the subsequent gas pipeline leakage experiment;

[0009] A pipeline, connected to the compressed air storage tank, for passage of air flow;

[0010] The pressure control valve and the pressure gauge are connected to the middle position of the pipeline. The pressure control valve is used to adjust the pressure of the compressed air, and the pressure gauge is used to monitor and display the air pressure in real time;

[0011] Three round-hole seamless steel pipes are connected to the tail of the pipeline and placed on a tripod. Leakage holes of different diameters are set on the three steel pipes to establish a gas leakage measurement data set with different leakage hole diameters.

[0012] A broadband microphone is used to receive the sound wave signal generated by pipeline leakage, facing the pipeline leakage hole of the perforated seamless steel pipe, and ensuring that the detection distance is fixed;

[0013] A data acquisition card is connected to the broadband microphone and the power supply, and is used to convert the analog sound wave signal received by the broadband microphone into a digital signal;

[0014] The data storage device is connected to the data acquisition card and is used to store the digital signals collected by the data acquisition card and process, analyze and visualize the data through corresponding software. The gas pipeline leakage acoustic detection method based on CPO-VMD and multi-feature extraction adopts the gas pipeline leakage acoustic detection device based on CPO-VMD and multi-feature extraction, including the following steps:

[0015] S1. Open the compressed air storage tank, detect the leakage sound wave signal emitted by the pipeline leakage in real time through a broadband microphone, and send the detected sound wave signal to the data storage device after being processed by a data acquisition card;

[0016] S2. Optimize the parameters of variational mode decomposition by CPO algorithm, introduce permutation entropy to construct fitness function as optimization target, and obtain K optimized intrinsic mode function components;

[0017] S3, obtaining the correlation coefficient between each intrinsic mode function component and the original signal, and reconstructing and denoising the key mode components based on the order of the correlation coefficients;

[0018] S4. Obtain the waveform, frequency domain and time domain characteristics of the leakage acoustic wave reconstruction signal, screen and select the signal characteristics, and perform gas pipeline leakage detection and classification through a classification detector.

[0019] Preferably, step S2 includes using the CPO algorithm for a position update strategy through four types of defense measures to optimize the two parameters of the decomposition layer number and the penalty factor; at the same time, introducing a cyclic population reduction strategy to dynamically adjust the model through the population size, and only activating threatened particles in each iteration process.

[0020] Preferably, the expression of the fitness function is as follows:

[0021] Fitness=m(h PE (IMF k )) / v(h PE (IMF k ));

[0022] In the formula, Fitness is the fitness function, m(h PE (IMF k )) represents the mean of the IMF permutation entropy, v(h PE (IMF k )) is the variance of the permutation entropy of each IMF.

[0023] Preferably, step S3 includes sorting the correlation coefficients between each intrinsic mode function component and the original signal from large to small: when the total number of intrinsic mode function components is greater than 2, the first two intrinsic mode function components ranked by the correlation coefficient are reconstructed as key mode components, and the remaining intrinsic mode function components are removed as noise; when the total number of intrinsic mode function components is equal to 2, the intrinsic mode function component with the largest correlation coefficient is used as the reconstructed signal.

[0024] Preferably, the classification detector comprises utilizing the multi-dimensional features of the extracted reconstructed signal to determine the gas pipeline leakage category through a support vector machine.

[0025] Therefore, the present invention adopts the above-mentioned gas pipeline leakage acoustic detection method based on CPO-VMD and multi-feature extraction, which has the following technical effects:

[0026] The present invention can effectively decompose the acoustic wave signal through the CPO-VMD algorithm, and has the significant advantages of strong robustness, high operating efficiency, strong convergence, strong decomposition effect, etc.; at the same time, through the multi-feature extraction scheme of time domain, frequency domain, and waveform characteristics, it realizes the re-screening and preferential selection of features, improves the classification accuracy and classification effect of leakage acoustic wave signals, and has wide applicability.

[0027] The technical solution of the present invention is further described in detail below through the accompanying drawings and embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0028] Figure 1 This is a schematic diagram of a gas pipeline leakage acoustic detection device based on CPO-VMD and multi-feature extraction;

[0029] Figure 2 This is a flow chart of the acoustic detection method for gas pipeline leakage based on CPO-VMD and multi-feature extraction;

[0030] Figure 3 This is a comparison chart of the running time of 100 iterations of different algorithms in an embodiment of a gas pipeline leakage acoustic detection device and method based on CPO-VMD and multi-feature extraction;

[0031] Figure 4 is the fitness function value of the PSO-VMD algorithm in the embodiment of the gas pipeline leakage acoustic detection device and method based on CPO-VMD and multi-feature extraction;

[0032] Figure 5 is the fitness function value of the SABO-VMD algorithm in the embodiment of the gas pipeline leakage acoustic detection device and method based on CPO-VMD and multi-feature extraction;

[0033] Figure 6 is the fitness function value of the CPO-VMD algorithm in the embodiment of the gas pipeline leakage acoustic detection device and method based on CPO-VMD and multi-feature extraction;

[0034] Figure 7 It is a simulation signal decomposition diagram of the PSO-VMD algorithm in an embodiment of a gas pipeline leakage acoustic detection device and method based on CPO-VMD and multi-feature extraction;

[0035] Figure 8 It is a simulation signal decomposition diagram of the SABO-VMD algorithm in an embodiment of a gas pipeline leakage acoustic detection device and method based on CPO-VMD and multi-feature extraction;

[0036] Fig. 9 It is a simulation signal decomposition diagram of the CPO-VMD algorithm in an embodiment of a gas pipeline leakage acoustic detection device and method based on CPO-VMD and multi-feature extraction;

[0037] Fig.10 It is a classification accuracy diagram of the combined feature ablation experiment under actual measurement conditions in an embodiment of a gas pipeline leakage acoustic detection device and method based on CPO-VMD and multi-feature extraction. DETAILED DESCRIPTION

[0038] The present invention can be explained in more detail by the following examples. The purpose of disclosing the present invention is to protect all changes and improvements within the scope of the present invention. The present invention is not limited to the following examples.

[0039] Embodiment 1

[0040] like Figure 1 As shown, the present invention provides a gas pipeline leakage acoustic detection device based on CPO-VMD and multi-feature extraction, including a compressed air storage tank, a pressure control valve, a pressure gauge, a pipeline, a seamless steel pipe, a broadband microphone, a data acquisition card and a data storage device.

[0041] Among them, the compressed air storage tank is used to provide air flow, and the generated air pressure provides the gas source conditions for the subsequent gas pipeline leakage experiment.

[0042] A pipe, connected to a compressed air storage tank, is used as a passage for air flow.

[0043] The pressure control valve and the barometer are connected to the middle of the pipeline. The pressure control valve is used to adjust the pressure of the compressed air to keep the air pressure in the pipeline at 0.3Mpa. The barometer is used to monitor and display the air pressure in real time to ensure the stability and accuracy of the air pressure during the experiment.

[0044] Three round-hole seamless steel pipes are connected to the tail of the pipeline and placed on a tripod. Leakage holes of different diameters are set on the three steel pipes. The diameters of the leakage holes are 0.4mm, 0.8mm, and 1.0mm respectively, which are used to establish a gas leakage measurement data set with different leakage hole diameters.

[0045] A wideband microphone is used to receive the sound wave signal generated by pipeline leakage. Under the background noise of the laboratory, a wideband microphone with a collection frequency of 20-20kHz is used to collect gas leakage sound wave signal data, while facing the pipeline leakage hole of the perforated seamless steel pipe and making the detection distance between them 3m.

[0046] The data acquisition card is connected to the broadband microphone and the power supply, and is used to convert the analog sound wave signal received by the broadband microphone into a digital signal, and then convert it into a digital form that can be processed by the subsequent data storage device. The sampling rate is set to 40 kHz, and a total of 171 leakage data are collected, each type contains 57 data, and each data contains 1460 sampling points.

[0047] The data storage device is connected to the data acquisition card and is used to store the digital signals collected by the data acquisition card and perform data processing, analysis and visualization through corresponding software.

[0048] In order to simulate the actual industrial gas pipeline leakage scenario, the gas pipeline leakage acoustic detection device can simulate three situations where the leakage aperture is 0.4, 0.8, and 1.0 mm respectively when the gas pressure is maintained at 0.3Mpa, and establish a measured data set of gas leakage with different leakage apertures to perform signal decomposition, reconstruction noise reduction, and multi-feature extraction to analyze the accuracy and effectiveness of leakage detection and classification. In this embodiment, the sampling rate is set to 40kHz, and a total of 171 leakage data are collected, each type contains 57 data, and each data contains 1460 sampling points.

[0049] Embodiment 2

[0050] like Figure 2 As shown, the present invention also provides a gas pipeline leakage acoustic detection method based on CPO-VMD and multi-feature extraction, using a gas pipeline leakage acoustic detection device based on CPO-VMD and multi-feature extraction, comprising the following steps:

[0051] S1. First, place the broadband microphone directly on the leaking hole of the seamless steel pipe. The broadband microphone is connected to the data acquisition card. The data acquisition card is connected to the power supply and data storage device. Then, open the compressed air storage tank. The broadband microphone detects the leakage sound wave signal emitted by the pipeline leak in real time. The detected sound wave signal is processed by the data acquisition card and sent to the data storage device.

[0052] S2. Use the CPO-VMD algorithm to decompose the original signal data, reconstruct and reduce noise, as follows:

[0053] The optimal decomposition layer number K and the parameter combination of the penalty factor α are used to perform variational mode decomposition (VMD) on the acoustic signal to obtain K optimized intrinsic mode functions u k (t)(IMF) component.

[0054] Among them, VMD algorithm processing is a method for nonlinear and non-stationary signal decomposition of the original signal, Hilbert transforms the analytical signal to obtain its unilateral spectrum, and at the same time, by minimizing the square of the second norm of the modal gradient, a variational problem is established, and the expression is:

[0055]

[0056] In the formula, U={u1,u2,...,u k},Ω={ω1,ω2,...,ω k},u k is the intrinsic mode function of the kth component, ω k is the estimated center frequency of the kth component; * represents the convolution operator, It represents the partial derivative with respect to time t, δ(t) is the Dirac function, K is the number of decomposition layers, and t is time.

[0057] In order to transform the constrained problem into an unconstrained problem, the penalty factor α and the Lagrange multiplier λ are introduced to obtain the augmented Lagrangian equation as follows:

[0058]

[0059] In order to further optimize the augmented Lagrangian equation, the initialization parameters and λ 1 is 0, and the alternating direction multiplier method ADMM is used to solve the saddle point of the problem and update the parameters and as follows:

[0060]

[0061] Where τ is the noise tolerance, and a balance is achieved between the accuracy of signal reconstruction and noise suppression by controlling the update rate of the Lagrange multiplier.

[0062] When satisfied When , the iteration stops. Where ε is the convergence tolerance.

[0063] To the final Perform inverse Fourier transform to obtain multiple narrowband intrinsic mode functions (IMFs).

[0064] Since the traditional VMD decomposition cannot adaptively adjust parameters, the crown porcupine optimization (CPO) algorithm is introduced to perform dual parameter optimization on (K, α) to adaptively obtain reasonable and effective VMD decomposition results.

[0065] Among them, the CPO algorithm is a new meta-heuristic algorithm that can model the four different protection mechanisms of the crested porcupine, so that it can jump out of the local optimal solution and improve the accuracy of the global optimal solution when performing parameter optimization tasks. In the VMD parameter optimization process, the CPO algorithm can be used for position update strategies through four types of defense measures, improve the accuracy of dual-parameter optimization, and thus improve the robustness of signal decomposition effects, and achieve dual-parameter optimization through multi-strategy synthesis.

[0066] In addition, the CPO algorithm also introduces a cyclic population reduction strategy, in which only the threatened particles are activated in each iteration, which promotes the convergence of optimization results and population diversity. It reduces unnecessary iterative calculations to reduce CPU time consumption during the optimization process, thereby improving the VMD dual-parameter optimization efficiency. The dynamic adjustment model of the population quantity is as follows:

[0067]

[0068] In the formula, N is the population size after update, N min is the minimum number of individuals in the newly generated population, N' is the number of candidate solutions, m is the current function value, % is the remainder operation, T max is the maximum number of function evaluations, and T is a variable that determines the number of loops.

[0069] Fitness is the core goal of algorithm optimization. Due to the unreasonable setting of fitness function, it is difficult to achieve reasonable and effective signal decomposition. The concept of permutation entropy is introduced to describe the degree of chaos of signal time series to facilitate nonlinear analysis of time series. The more regular the signal, the smaller the permutation entropy, and the more complex the signal, the larger the permutation entropy. The permutation entropy calculation of each IMF is as follows:

[0070]

[0071] In the formula, h PE (IMFk ) is the permutation entropy of the kth modal component, P i It refers to the probability of the ith permutation of the kth IMF signal sequence, where N is the length of the signal sequence.

[0072] In the VMD dual-parameter optimization process, different combinations of [K, α] values ​​will lead to changes in the permutation entropy of each modal component. By comparing the amplitude of the permutation entropy change, the separation of the effective signal and the noise signal is analyzed. In addition, the difference between the modal signals is caused by the variance and mean changes. By constructing the fitness function Fitness, the ratio of the mean to the variance of the permutation entropy of the gas leakage signal is quantified, and the minimization of the fitness function is used as the objective function for optimization. The formula is as follows:

[0073] Fitness=m(h PE (IMF k )) / v(h PE (IMF k ));

[0074] In the formula, m(h PE (IMF k )) represents the mean of the IMF permutation entropy, v(h PE (IMF k )) is the variance of the permutation entropy of each IMF.

[0075] Through the position update strategy of the CPO algorithm, the constructed fitness function is updated and adjusted. At the same time, the adaptive cyclic population reduction mechanism of the CPO algorithm is used to improve the iteration efficiency, and the best fitness and parameter combination (K, α) is obtained, thereby obtaining K optimized IMF components.

[0076] S3. Signal reconstruction and noise reduction are achieved through correlation coefficient. The correlation coefficient reflects the degree of correlation between two groups of signals, as follows:

[0077] The correlation coefficients between each IMF and the original signal are sorted by size. If the total number of IMFs is greater than 2, the IMFs with the top two correlation coefficients are reconstructed as key modal components, and the remaining IMFs are removed as noise signals. If the total number of IMFs is only 2, the IMF with the largest correlation coefficient is used as the reconstructed signal.

[0078] The correlation coefficient calculation formula is as follows:

[0079]

[0080] In the formula, is the correlation coefficient between each IMF and the original signal, is the mean of the intrinsic mode functions of each component, x(t) is the original signal, is the mean value of the original signal at each moment.

[0081] S4. Due to the influence of background noise, when a gas pipeline leaks, its acoustic signal amplitude will also be strongly affected. In order to fully reflect the characteristics of gas pipeline leakage, multi-feature extraction is performed to screen and select the signal features again, and the waveform (WF), frequency domain (FD) and time domain (TD) of the leakage acoustic wave signal are extracted respectively.

[0082] The waveform characteristics include a skewness factor, which is used to reflect the degree of skewness of the signal data distribution. The specific expression is as follows:

[0083]

[0084] In the formula, γ is the skewness factor, n is the number of samples, and x i is the value of a point in the signal data set at a certain time. is the mean of the signal data set, X std is the standard deviation of the signal data set.

[0085] The frequency domain characteristics include the center frequency fc, which reflects the concentration of the signal frequency; the spectral energy E, which reflects the distribution of the signal in the frequency domain; the spectral peak characteristic P jf , reflects the sharpness of energy distribution in the signal spectrum; the spectrum peak f peak , reflecting the maximum amplitude of the frequency component in the signal. The specific expression of the frequency domain feature is as follows:

[0086]

[0087] P jf =max(p x ) / E;

[0088] f peak =f max_index ;

[0089] In the formula, is the i-th element of the power spectral density, f i is the i-th element of the frequency vector f, f max_index is the frequency corresponding to the maximum value of the power spectral density.

[0090] The time domain features contain the mean Indicates the central tendency of signal strength; variance X var , which measures the degree of signal fluctuation; standard deviation X std , describing the discreteness of the signal; the average amplitude X ma , which measures the amplitude of the signal; energy X e , reflecting the overall strength of the signal. The specific expressions of each time domain feature are as follows:

[0091]

[0092] In the formula, f ik The i-th element in the k-th IMF component.

[0093] Then, the extracted multiple features are used to accurately detect pipeline leaks. Specifically, the multiple types of signal features that reflect the differences between leakage signals and conventional signals are combined into a one-dimensional feature matrix, which is input into a support vector machine (SVM) used as a classification detector to perform accurate and effective gas pipeline leak detection, and the leak classification results are displayed in a visual form.

[0094] Embodiment 3

[0095] In order to further verify the effectiveness of the CPO-VMD algorithm for acoustic detection of gas pipeline leakage, this embodiment establishes a simulation signal decomposition experiment and uses meta-heuristic algorithms such as SABO and GAO for comparative analysis.

[0096] The expression of the simulation signal is as follows:

[0097] x(t)=cos(2πf1t)+1.3sin(2πf2t)+1.5cos(2πf3t)+1.7sin(2πf4t)+μ;

[0098] Where μ is an additive white Gaussian noise with a signal-to-noise ratio of 20 dB, and the frequencies of each component are set to f1 = 10, f2 = 50, f3 = 150, f4 = 200, and the sampling frequency is f s The frequency is 1000Hz and the number of sampling points is 1000.

[0099] In order to fully perform global search, the population size is set to 20 and the number of iterations is set to 100. At the same time, to avoid accidental errors, the parameters (K, α) of VMD are adaptively optimized 10 times independently. The detailed parameter settings of CPO algorithm, SABO algorithm and CPO algorithm are shown in Table 1.

[0100] Table 1 Parameter settings of different metaheuristic algorithms

[0101]

[0102] The detailed parameter settings of the VMD algorithm are shown in Table 2.

[0103] Table 2 Parameter settings of different metaheuristic algorithms

[0104] Optimizing key parameters Parameter Value Penalty factor upper bound <![CDATA[α max =2250]]> Penalty factor lower bound <![CDATA[α min =100]]> Upper bound of decomposition level <![CDATA[K max =10]]> Lower bound of decomposition level <![CDATA[K min =2]]> Population size Agents_no=20 Maximum number of iterations Max_iteration = 100 Convergence tolerance <![CDATA[Tol=10 -7 ]]>

[0105] like Figure 2As shown in the figure, when different algorithms were subjected to 10 independent repeated experiments and 100 iterations of optimization each time, the CPO-VMD algorithm had relatively small data fluctuations, good stability, and more stable operating efficiency. In addition, the mean running time can reflect the overall computational efficiency level. Although the mean running time of the CPO-VMD algorithm is slightly higher than that of the PSO-VMD algorithm (14.54 seconds difference), the optimization strategy of the PSO algorithm is relatively simple, while the strategy of the CPO algorithm is more comprehensive, which significantly reduces redundant calculations. In general, the CPO-VMD algorithm shows relatively stable efficiency in terms of running time, which has significant advantages over the SABO-VMD algorithm, and the gap with the PSO-VMD algorithm is small, indicating that the CPO-VMD algorithm has outstanding performance in both optimization efficiency and computational stability.

[0106] This example also evaluates the fitness function, see Figure 4 , Figure 5 and Figure 6 It can be seen that the fitness functions of the PSO-VMD and SABO-VMD algorithms tend to different values ​​in different running times, indicating that they have not achieved stable convergence within 100 iterations and cannot guarantee the global optimal solution within a certain number of iterations; while the CPO-VMD algorithm successfully converged to the minimum fitness value of 22.6545, and all optimal solutions were obtained within the first 60 iterations, which shows that this method has significant advantages in convergence and has a high optimization efficiency.

[0107] See also Figure 7 , Figure 8 and Fig. 9 ,When the decomposition level K = 6, both the PSO-VMD algorithm and the SABO-VMD algorithm have over-decomposition of the same key center frequency; while the CPO-VMD algorithm effectively retains the four frequency domain peak values, avoiding over-decomposition, which has obvious advantages.

[0108] In addition, this embodiment also uses the CPO-VMD algorithm to analyze three types of leakage signals in the measured data, and obtains the minimum fitness value in the first 70 iterations. In addition, the reconstructed signal is compared with the original signal, and it is found that when the background noise is relatively strong, the signal-to-noise ratio of the reconstructed signal is improved more significantly, which fully verifies that the method of the present invention has a superior noise reduction effect when the noise intensity is large.

[0109] In order to analyze the classification accuracy, multi-dimensional features are extracted from the reconstructed signal of the measured data set to obtain a 171×3 feature matrix as the input of the SVM. The accuracy formula is as follows:

[0110]

[0111] Where Acc is the accuracy, TP is the true positive example, TN is the true negative example, FP is the false positive example, and FN is the false negative example.

[0112] Among the 69 data in the test set, 4 were misclassified, with an accuracy rate of 94.20%. This example also compares the classification effects of a single feature, a combination of two features, and a combination of three features. Fig.10 It can be seen that in the combined feature ablation experiment, the accuracy of the double combined features and the triple combined features are consistent. In general, the multi-feature extraction strategy is helpful to improve the accuracy of leakage acoustic wave detection.

[0113] Therefore, the present invention adopts the above-mentioned gas pipeline leakage acoustic detection device and method based on CPO-VMD and multi-feature extraction, which can detect the acoustic wave data of gas leakage under different leak hole diameters, is suitable for different scenarios, and can be applied to actual engineering.

[0114] Finally, it should be noted that the above embodiments are only used to illustrate the technical solution of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that they can still modify or replace the technical solution of the present invention with equivalents, and these modifications or equivalent replacements cannot cause the modified technical solution to deviate from the spirit and scope of the technical solution of the present invention.

Claims

1. Gas pipeline leakage acoustic detection device based on CPO-VMD and multi-feature extraction, characterized in that: include: Compressed air storage tank, used to provide air flow, the generated air pressure provides the gas source conditions for the subsequent gas pipeline leakage experiment; A pipeline, connected to the compressed air storage tank, for passage of air flow; The pressure control valve and the pressure gauge are connected to the middle position of the pipeline. The pressure control valve is used to adjust the pressure of the compressed air, and the pressure gauge is used to monitor and display the air pressure in real time; Three round-hole seamless steel pipes are connected to the tail of the pipeline and placed on a tripod. Leakage holes of different diameters are set on the three steel pipes to establish a gas leakage measurement data set with different leakage hole diameters. A broadband microphone is used to receive the sound wave signal generated by pipeline leakage, facing the pipeline leakage hole of the perforated seamless steel pipe, and ensuring that the detection distance is fixed; A data acquisition card is connected to the broadband microphone and the power supply, and is used to convert the analog sound wave signal received by the broadband microphone into a digital signal; The data storage device is connected to the data acquisition card and is used to store the digital signals collected by the data acquisition card and perform data processing, analysis and visualization through corresponding software.

2. A gas pipeline leakage acoustic detection method based on CPO-VMD and multi-feature extraction, using the gas pipeline leakage acoustic detection device based on CPO-VMD and multi-feature extraction according to claim 1, characterized in that: The following steps are involved: S1. Open the compressed air storage tank, detect the leakage sound wave signal emitted by the pipeline leakage in real time through a broadband microphone, and send the detected sound wave signal to the data storage device after being processed by a data acquisition card; S2. Optimize the parameters of variational mode decomposition by CPO algorithm, introduce permutation entropy to construct fitness function as optimization target, and obtain K optimized intrinsic mode function components; S3, obtaining the correlation coefficient between each intrinsic mode function component and the original signal, and reconstructing and denoising the key mode components based on the order of the correlation coefficients; S4. Obtain the waveform, frequency domain and time domain characteristics of the leakage acoustic wave reconstruction signal, screen and select the signal characteristics, and perform gas pipeline leakage detection and classification through a classification detector.

3. The gas pipeline leakage acoustic detection method based on CPO-VMD and multi-feature extraction according to claim 2 is characterized in that: Step S2 includes the CPO algorithm using four types of defense measures for position update strategy, optimizing the two parameters of decomposition layer number and penalty factor; at the same time, introducing a cyclic population reduction strategy to dynamically adjust the model through the population size, and only activating threatened particles in each iteration process.

4. The gas pipeline leakage acoustic detection method based on CPO-VMD and multi-feature extraction according to claim 2 is characterized in that: The expression of the fitness function is as follows: Fitness=m(h PE (IMF k )) / v(h PE (IMF k )); In the formula, Fitness is the fitness function, m(h PE (IMF k )) represents the mean of the IMF permutation entropy, v(h PE (IMF k )) is the variance of the permutation entropy of each IMF.

5. The gas pipeline leakage acoustic detection method based on CPO-VMD and multi-feature extraction according to claim 2 is characterized in that: Step S3, comprising sorting the correlation coefficients between each intrinsic mode function component and the original signal from large to small: when the total number of intrinsic mode function components is greater than 2, the first two intrinsic mode function components ranked by the correlation coefficient are reconstructed as key mode components, and the remaining intrinsic mode function components are removed as noise; When the total number of eigenmode function components is equal to 2, the eigenmode function component with the largest correlation coefficient is used as the reconstructed signal.

6. The gas pipeline leakage acoustic detection method based on CPO-VMD and multi-feature extraction according to claim 2 is characterized in that: The classification detector includes utilizing the multi-dimensional features of the extracted reconstructed signal to judge the type of gas pipeline leakage through a support vector machine.

Images