Pipeline leakage positioning method based on CEEMDAN-LMS adaptive noise elimination

Abstract

The invention provides a pipeline leakage positioning method based on CEEMDAN-LMS adaptive noise elimination, and the method comprises the steps: obtaining two leakage signals, including a first leakage signal and a second leakage signal, of the upstream and downstream of a pipeline leakage point; sequentially decomposing the first leakage signal into k IMF components by adopting a CEEMDAN algorithm; screening the k IMF components to obtain n signal components and m noise components; reconstructing the m noise components to obtain a first reconstructed signal, and sequentially decomposing the first reconstructed signal into k IMF'components by adopting a CEEMDAN algorithm; performing LMS adaptive filtering on each signal component according to the IMF'component corresponding to the signal component in the same order to obtain n denoised signal components; reconstructing the n denoised signal components to obtain a first noise reduction signal; performing the same processing as the first leakage signal on the second leakage signal to obtain a second noise reduction signal; and according to the two noise reduction signals, calculating the position of the leakage point by adopting a two-point positioning model. According to the method, the positioning precision of the pipeline leakage point can be improved.

Description

Technical Field

[0001] The invention belongs to the technical field of pipeline leakage locating, and in particular relates to a pipeline leakage locating method based on CEEMDAN-LMS adaptive noise elimination. Background Art

[0002] As a core component of urban infrastructure, the safe and stable operation of water supply networks is directly linked to economic development, livelihood security, and public safety. As pipeline networks expand and their service life increases, leakage problems caused by aging, ground subsidence, and vandalism are becoming increasingly prominent. Currently, pipeline leak detection technologies primarily include infrared thermal imaging, ground-penetrating radar, negative pressure wave detection, transient wave detection, and acoustic detection. Acoustic detection has become the mainstream solution due to its non-invasiveness, high sensitivity, and flexible deployment. This technology utilizes accelerometers deployed upstream and downstream of the pipeline leak to capture the acoustic vibration signals caused by the leak. The time delay between the two signals is estimated based on a cross-correlation function, allowing the leak location to be deduced. However, in practice, the complex environment in which pipelines operate (such as traffic vibration, electromagnetic interference, and fluid turbulence) often causes leakage signals to be drowned out by strong background noise, significantly increasing delay estimation errors and directly impacting positioning accuracy. Therefore, effectively extracting leakage signatures from noisy signals is a key challenge in improving positioning reliability.

[0003] To address the problem of noise suppression, noise reduction methods based on signal decomposition are widely used. Patent CN113864665A proposes a fluid pipeline leak location method based on adaptive ICA and an improved RLS filter. Its core is to use the complete adaptive noise ensemble empirical mode decomposition (CEEMDAN) algorithm to decompose the leakage signal into multiple components. Then, by calculating the Euclidean distance between each component and its corresponding leakage vibration signal, irrelevant IMF components are filtered out. The filtered components are then subjected to RLS filtering and further processing to locate the leak point. When traditional empirical mode decomposition (EMD) decomposes nonlinear and non-stationary signals, different frequency components will be mixed in the same modal component (IMF), resulting in blurred signal characteristics. Although the CEEMDAN algorithm alleviates this problem by gradually adding adaptive noise, in high-noise environments, residual mode aliasing can still cause high-frequency noise to infiltrate the effective components, affecting the accuracy of Euclidean distance filtering. In addition, during RLS filtering, the IMF components may retain noise, causing the convergence of the RLS adaptive filter to deviate, reducing the accuracy of leak location.

[0004] In the noise reduction process based on CEEMDAN, the screening and reconstruction strategy of IMF components is a key link in determining the denoising effect. In the existing technology, IMF screening mostly relies on a single indicator. For example, by calculating the correlation coefficient between the IMF and the original signal or the kurtosis value of each component, the components with high correlation or large kurtosis with the leakage signal are selected for reconstruction. However, when the noise intensity is high, the noise will significantly change the extreme value distribution, zero-crossing density and envelope characteristics of the signal, resulting in deviations in the calculation results of the above statistical indicators, and then causing the problem of IMF misselection or omission. In addition, the existing reconstruction method usually directly reconstructs the IMF after screening, ignoring the residual noise components in each component. For example, the traditional scheme directly decomposes CEEMDAN and directly reconstructs the IMF determined to be dominated by the leakage signal. However, all the IMFs obtained by the actual decomposition contain noise pollution to varying degrees. Direct reconstruction will result in residual noise in the reconstructed signal, affecting the accuracy of subsequent delay estimation. These problems restrict the positioning accuracy of acoustic detection technology in strong noise scenarios. Summary of the Invention

[0005] In view of the shortcomings of the existing technology, the present invention provides a pipeline leakage location method based on CEEMDAN-LMS adaptive noise elimination, the method comprising:

[0006] S1: Acquire two leakage signals upstream and downstream of the pipeline leakage point, including a first leakage signal and a second leakage signal;

[0007] S2: Use the CEEMDAN algorithm to decompose the first leakage signal into k IMF components in sequence;

[0008] S3: Filter the k IMF components of the first leakage signal to obtain n signal components and m noise components, where n+m=k;

[0009] S4: Reconstruct the m noise components to obtain the reconstructed signal, and use the CEEMDAN algorithm to decompose the reconstructed signal into k IMF′ components in sequence;

[0010] S5: Each signal component of the first leakage signal is subjected to LMS adaptive filtering according to its corresponding IMF′ component of the same order to obtain n denoised signal components;

[0011] S6: reconstructing n denoised signal components to obtain a first denoised signal;

[0012] S7: Perform the same processing as steps S2 to S6 on the second leakage signal to obtain a second noise reduction signal;

[0013] S8: Calculate the position of the leakage point using a two-point positioning model based on the first noise reduction signal and the second noise reduction signal to complete pipeline leakage positioning.

[0014] Preferably, the two leakage signals are collected by two acceleration sensors installed upstream and downstream of the pipeline leakage point, and are expressed as:

[0015] x1(t)=s(t)+n1(t)

[0016] x2(t)=αs(t-τ)+n2(t)

[0017] Wherein, x1(t) is the first leakage signal collected by the first accelerometer at time t, x2(t) is the second leakage signal collected by the second accelerometer at time t, n1(t) is the noise signal in the first leakage signal collected by the first accelerometer at time t, n2(t) is the noise signal in the second leakage signal collected by the second accelerometer at time t, s(t) is the leakage source signal emitted by the leakage point at time t, s(t-τ) is the leakage source signal emitted by the leakage point at time t-τ, α is the attenuation factor, and τ is the time delay.

[0018] Preferably, the process of sequentially decomposing the first leakage signal into k IMF components using the CEEMDAN algorithm includes:

[0019] S21: Add multiple Gaussian white noises to the first leakage signal to obtain a noisy signal sequence;

[0020] S22: performing EMD decomposition on the noisy signal sequence to obtain the current IMF component and the current order residual component;

[0021] S23: Add multiple Gaussian white noises to the residual component of the current order and perform EMD decomposition again to obtain the next IMF component and the residual component of the next order;

[0022] S24: Repeat S23 until the residual component cannot be decomposed any further, and obtain the final residual component and k IMF components of the first leakage signal.

[0023] Preferably, the process of screening the k IMF components of the first leakage signal includes:

[0024] respectively calculating a first screening factor and a second screening factor for each IMF component of the first leakage signal;

[0025] Setting a first screening factor threshold and a second screening factor threshold;

[0026] Each IMF component of the first leakage signal is screened respectively. If a first screening factor of the IMF component is greater than a first screening factor threshold and a second screening factor is greater than a second screening factor threshold, the IMF component is determined to be a signal component; otherwise, the IMF component is determined to be a noise component.

[0027] Preferably, the process of calculating the first screening factor and the second screening factor of each IMF component of the first leakage signal includes:

[0028] Calculating a cross-correlation function between an IMF component of the first leakage signal and the second leakage signal;

[0029] According to the cross-correlation function between the IMF component of the first leakage signal and the second leakage signal, the initial first screening factor and the initial second screening factor of the IMF component are calculated as follows:

[0030] CR1=|R(τ) max |+|R(τ+1)|+|R(τ-1)|

[0031]

[0032] Among them, CR1 is the initial first screening factor, CR2 is the initial second screening factor, R(τ) max is the maximum value of the cross-correlation function between the IMF component of the first leakage signal and the second leakage signal, τ is the time delay, and R(τ+1) is R(τ) max The adjacent value of the cross-correlation function delayed by 1 time point, R(τ-1) is R(τ) max The adjacent values ​​of the cross-correlation function at one time point ahead;

[0033] The initial first screening factor of the IMF component is normalized to obtain the final first screening factor of the IMF component, and the initial second screening factor of the IMF component is forward-normalized and normalized to obtain the final second screening factor of the IMF component.

[0034] Preferably, each signal component of the first leakage signal is subjected to LMS adaptive filtering according to its corresponding IMF′ component of the same order, and is expressed as:

[0035] y(n)=w T (n)x(n)

[0036] e(n)=d(n)-y(n)

[0037] w(n+1)=w(n)-μe(n)x(n)

[0038] Where x(n) is the IMF′ component of the same order as d(n) at time n, y(n) is the denoised signal component output at time n, w(n) is the coefficient vector at time n, w(n+1) is the updated coefficient vector at time n+1, e(n) is the error at time n, d(n) is the signal component of the first leakage signal at time n, and μ is the step size factor.

[0039] Preferably, the process of calculating the location of the leak point using the two-point positioning model includes:

[0040] Calculating a cross-correlation function between the first noise reduction signal and the second noise reduction signal;

[0041] calculating a time delay based on a cross-correlation function between the first noise reduction signal and the second noise reduction signal;

[0042] According to the time delay, the two-point positioning model is used to calculate the location of the leak point.

[0043] Preferably, the formula for calculating the position of the leakage point using the two-point positioning model is:

[0044]

[0045] Wherein, L is the distance between the first acceleration sensor and the second acceleration sensor, L1 is the distance from the leakage point to the first acceleration sensor, c is the signal propagation speed, and τ is the time delay.

[0046] The beneficial effects of the present invention are:

[0047] 1. This method uses the maximum value of the cross-correlation function and its adjacent values ​​to calculate two filtering factors, reflecting the leakage signal content and suppressing the randomness of noise, respectively. This solves the problem of traditional methods relying on a single indicator being susceptible to noise interference and leading to misjudgment. This method can effectively distinguish signal components from noise components, thereby reducing the deviation of subsequent delay estimation.

[0048] 2. To address the existing flaw of directly reconstructing signal components, which often results in residual noise in the reconstructed signal, the present invention reconstructs the noise components determined by the initial CEEMDAN decomposition and performs a secondary CEEMDAN decomposition to extract a reference signal with a narrower frequency band that closely matches the noise characteristics of the original signal. This method, combined with LMS adaptive filtering, achieves deep denoising of the signal components. This method effectively improves the denoising effect of leakage signals, significantly enhancing the accuracy of locating pipeline leaks. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following introduction is made to the drawings of the relevant technical solutions of the embodiments of the present invention. It should be understood that the drawings introduced below are only for the convenience of clearly describing some embodiments of the technical solutions of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without any creative work.

[0050] Figure 1 1 is a flow chart of a method for pipeline leak location based on CEEMDAN-LMS adaptive noise elimination according to an embodiment of the present invention;

[0051] Figure 2 Schematic diagram of the principle of the LMS adaptive noise elimination method in an embodiment of the present invention;

[0052] Figure 3 2 is a flow chart of a method for combining CEEMDAN with LMS adaptive filtering in an embodiment of the present invention;

[0053] Figure 4 1 is a schematic diagram of the cross-correlation function results of signals after CEEMDAN processing and CEEMDAN-LMS processing in an embodiment of the present invention;

[0054] Figure 5 1 is a schematic diagram of the correct estimation times of various methods under different signal-to-noise ratios in an embodiment of the present invention. DETAILED DESCRIPTION

[0055] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments in the present invention, all other embodiments obtained by those skilled in the art without making creative work are within the scope of protection of the present invention. For the step numbers in the following embodiments, they are only set for the convenience of explanation and description, and no limitation is made to the order between the steps. The execution order of each step in the embodiment can be adaptively adjusted according to the understanding of those skilled in the art.

[0056] The present invention provides a pipeline leakage location method based on CEEMDAN-LMS adaptive noise elimination, the method steps are as follows: Figure 1 shown.

[0057] S1: Acquire two leakage signals upstream and downstream of the pipeline leakage point, including a first leakage signal and a second leakage signal.

[0058] When a water supply pipeline leaks, the acoustic vibration signals generated by the leak will propagate from the leak point along the pipeline in both upstream and downstream directions. These two acoustic vibration signals carrying the leak source signal can be collected by two acceleration sensors installed upstream and downstream of the leak point. The two acquired leakage signals are expressed as:

[0059] x1(t)=s(t)+n1(t)

[0060] x2(t)=αs(t-τ)+n2(t)

[0061] Wherein, x1(t) is the first leakage signal collected by the first accelerometer at time t, x2(t) is the second leakage signal collected by the second accelerometer at time t, n1(t) is the noise signal in the first leakage signal collected by the first accelerometer at time t, n2(t) is the noise signal in the second leakage signal collected by the second accelerometer at time t, s(t) is the leakage source signal emitted by the leakage point at time t, s(t-τ) is the leakage source signal emitted by the leakage point at time t-τ, α is the attenuation factor, and τ is the time delay.

[0062] In the embodiment of the present invention, s(t) is uncorrelated with n1(t) and n2(t), and the cross-correlation function of x1(t) and x2(t) can be expressed as:

[0063]

[0064] in, is the cross-correlation function of x1(t) and x2(t), R ss [τ-(τ1-τ2)] is the autocorrelation function of the leakage source signal s(t), α is the attenuation factor, is the cross-correlation function of n1(t) and n1(t), τ is the time delay, τ1 is the time delay of the first leakage signal, and τ2 is the time delay of the second leakage signal.

[0065] In the embodiment of the present invention, when τ=τ1-τ2, R ss [τ-(τ1-τ2)] reaches its maximum value, so when the two leakage signals x1(t) and x2(t) contain a small amount of noise, the time delay can be directly determined based on the cross-correlation function of the leakage signals. However, in actual situations, due to the influence of factors such as the pipeline's own characteristics, the surrounding environment, and signal transmission interference, the obtained leakage signal often contains a lot of high-frequency noise, and the noise disturbance term Larger, resulting in The maximum value of will deviate, which will cause the accuracy of time delay estimation to seriously decrease. Therefore, it is necessary to denoise the leakage signal and improve the signal-to-noise ratio.

[0066] S2: Use the CEEMDAN algorithm to decompose the first leakage signal into k IMF components in sequence.

[0067] Before using the cross-correlation function to estimate delay, embodiments of the present invention require denoising the leakage signal. This method uses the CEEMDAN algorithm to sequentially decompose the first leakage signal. By introducing adaptive noise (dynamically adjusting the noise amplitude based on signal characteristics) and staged residual decomposition, the computational complexity is reduced while ensuring decomposition completeness. The process of sequentially decomposing the first leakage signal into k IMF components using the CEEMDAN algorithm includes:

[0068] S21: Add multiple Gaussian white noises to the first leakage signal to obtain a noisy signal sequence.

[0069] Add multiple Gaussian white noises ω to the first leakage signal i (t), and obtain the noise signal sequence x1(t)+ε0ω i (t), i=1,···,I, where ε0 is the noise coefficient, ω i (t) is the Gaussian white noise added for the i-th time.

[0070] S22: Perform EMD decomposition processing on the noisy signal sequence to obtain the current IMF component and the residual component of the current order.

[0071] For the noise signal sequence x1(t)+ε0ω i (t) Perform EMD decomposition, extract the first modal component sequence and take the mean, and obtain the first IMF component of the first leakage signal: The residual component of order 1 is calculated as r1(t)=x1(t)-IMF1(t).

[0072] S23: Add multiple Gaussian white noises to the residual component of the current order and perform EMD decomposition again to obtain the next IMF component and the residual component of the next order. The specific steps include:

[0073] Step 1: Add multiple Gaussian white noises to the first-order residual component for EMD decomposition, take the first modal component sequence and take the mean, and obtain the second IMF component of the first leakage signal, which is expressed as:

[0074]

[0075] Where: IMF2(t) is the second IMF component of the first leakage signal, E1(·) is the first modal component extracted after the first leakage signal is decomposed by EMD, r1(t) is the first-order residual component, ω i (t) is the Gaussian white noise added for the i-th time, and ε1 is the noise coefficient;

[0076] Step 2: Calculate the k-order (k ≥ 2) residual component:

[0077] r k (t) = r k-1 (t)-IMF k (t)

[0078] Among them, r k (t) is the residual component of order k, r k-1 (t) is the residual component of order k-1, IMF k (t) is the kth IMF component of the first leakage signal;

[0079] Step 3: Add multiple Gaussian white noises to the k-order residual component for EMD decomposition, and take the average value to obtain the k+1th IMF component IMF of the leakage signal x1(t) k+1 (t), expressed as:

[0080]

[0081] Among them, the IMF k+1 (t) is the k+1th IMF component of the first leakage signal, E k (·) is the kth modal component extracted from the first leakage signal after EMD decomposition, r k (t) is the residual component of order k, ω i (t) is the Gaussian white noise added for the i-th time, ε k is the noise coefficient of order k.

[0082] S24: Repeat S23 until the residual component cannot be decomposed any further, and obtain the final residual component and k IMF components of the first leakage signal.

[0083] In this embodiment of the present invention, the first leakage signal is sequentially decomposed into k IMFs using the CEEMDAN algorithm. These IMFs can be divided into IMFs dominated by the leakage signal and IMFs dominated by the noise signal. In the prior art, IMF screening often relies on a single metric. For example, by calculating the correlation coefficient between the IMF and the original signal or the kurtosis value of each component, components with high correlation or large kurtosis with the leakage signal are selected for reconstruction. However, when the noise intensity is high, the noise can significantly change the extreme value distribution, zero-crossing density, and envelope characteristics of the signal, resulting in deviations in the calculation results of a single metric, which in turn leads to the problem of misselection or omission of IMFs.

[0084] S3: Screening the k IMF components of the first leakage signal to obtain n signal components and m noise components, where n+m=k.

[0085] In this embodiment of the present invention, if the primary component of an IMF is a leakage signal, the maximum absolute value (MAVCC) of the cross-correlation function between the IMF and the second leakage signal is large. However, if the primary component of an IMF is a noise signal, the MAVCC between the IMF and the second sensor leakage signal is small. Based on this characteristic, a filtering factor is constructed to measure the peak value of the cross-correlation function as a screening indicator. However, signal simulations have shown that due to the randomness of noise, the MAVCC of some noise-dominated IMFs is also large. Further analysis of the cross-correlation function in this case reveals that the cross-correlation function has a large value only at the MAVCC point, with small values ​​at adjacent points. However, when the MAVCC of the leakage-dominated IMF is large, the cross-correlation function has a large value not only at the MAVCC point but also at adjacent points. Therefore, the present invention comprehensively considers the MAVCC and its adjacent values ​​to construct two filtering factors, CR1 and CR2. CR1 reflects the correlation between the IMF and the second leakage signal, while CR2 eliminates MAVCCs caused by random noise. The specific process of screening the k IMF components of the first leakage signal includes:

[0086] S31: Calculate the first filtering factor and the second filtering factor of each IMF component of the first leakage signal respectively. The specific steps of calculating the first filtering factor and the second filtering factor of each IMF component of the first leakage signal include:

[0087] S311: Calculate the cross-correlation function between the IMF component of the first leakage signal and the second leakage signal:

[0088]

[0089] Among them, R 1i (τ) is the cross-correlation function between the IMF component of the first leakage signal and the second leakage signal, IMF(t) is the IMF component of the first leakage signal, t is the time variable, τ is the time delay, and T is the duration of the entire signal;

[0090] S312: Calculate an initial first screening factor and an initial second screening factor of the IMF component according to the cross-correlation function of the IMF component of the first leakage signal and the second leakage signal. The calculation formula is:

[0091] CR1=|R(τ) max |+|R(τ+1)|+|R(τ-1)|

[0092]

[0093] Among them, CR1 is the initial first screening factor, CR2 is the initial second screening factor, R(τ) maxis the maximum value of the cross-correlation function between the IMF component after the decomposition of the leakage signal and the second leakage signal, τ is the time delay, and R(τ+1) is R(τ) max The adjacent value of the cross-correlation function delayed by 1 time point, R(τ-1) is R(τ) max The adjacent values ​​of the cross-correlation function at one time point ahead;

[0094] S313: performing normalization processing on the initial first screening factor of the IMF component to obtain the final first screening factor of the IMF component, and performing forward and normalization processing on the initial second screening factor of the IMF component to obtain the final second screening factor of the IMF component. The normalization processing formula is:

[0095]

[0096] Among them, NCR1 is the final first screening factor, NCR2 is the final second screening factor, CR1 is the initial first screening factor, and CR′2 is the second screening factor after CR1 is positively converted.

[0097] S32: Setting a first screening factor threshold and a second screening factor threshold; preferably, in the embodiment of the present invention, the first screening factor threshold is 0.8, and the second screening factor threshold is 0.8.

[0098] S33: Filter each IMF component of the first leakage signal respectively. If the first filtering factor of the IMF component is greater than the first filtering factor threshold and the second filtering factor is greater than the second filtering factor threshold, determine that the IMF component is a signal component; otherwise, determine that the IMF component is a noise component.

[0099] In this embodiment of the present invention, k IMF components of the first leakage signal are screened. Each IMF component is determined to be either a signal component or a noise component. After the screening is complete, n signal components and m noise components are obtained. Furthermore, the sum of the number of signal components and the number of noise components equals the number of IMF components, i.e., n + m = k.

[0100] S4: Reconstruct the m noise components to obtain a reconstructed signal, and use the CEEMDAN algorithm to decompose the reconstructed signal into k IMF′ components in sequence.

[0101] After the first leakage signal of the embodiment of the present invention is decomposed by the CEEMDAN algorithm, the first leakage signal is only decomposed into a few components due to its limited bandwidth, while the background noise bandwidth is wider than the leakage signal and contributes to all components, so the filtered signal components need to be further denoised. The present invention uses an LMS adaptive filter to further remove the noise of the signal components. According to the principle of the LMS adaptive filter, the filtered signal components serve as the desired signal of the filter. However, if the noise components are directly reconstructed into a wider frequency band signal as the reference signal of the filter, the denoising effect is often poor because the LMS filter is more suitable for narrowband signals.

[0102] In an embodiment of the present invention, m noise components of a first leakage signal are reconstructed to obtain a reconstructed signal, and the CEEMDAN algorithm is used to sequentially decompose the reconstructed signal into k IMF′ components, making the reconstructed signal suitable for an LMS filter. The order of the IMF′ components corresponds to the order of the IMF components, and due to the effect of white noise added by CEEMDAN, the components after the two decompositions decrease at the same frequency, so that the signal components can correspond to the IMF′ components of the same order in the same frequency band.

[0103] S5: Each signal component of the first leakage signal is subjected to LMS adaptive filtering according to its corresponding IMF′ component of the same order to obtain n denoised signal components.

[0104] In the embodiment of the present invention, the LMS adaptive noise cancellation can be regarded as a decorrelation system. The principle is as follows: Figure 2 As shown. The expected signal d is a mixture of the leakage signal s and the noise signal v1, the reference signal is the noise v2, v1 and v2 are correlated, and the correlation between the noise and the leakage signal is small, then the output of the system is equal to the error signal e, e = s + v1 - v', where v' is the estimate of the noise v2 through adaptive filtering. In the present invention, the signal component of the first leakage signal is used as the expected signal in the LMS adaptive noise filtering, and the IMF' component corresponding to the same order as the signal component of the first leakage signal is used as the reference signal, and LMS adaptive filtering is performed to obtain the denoised signal component. The method flow is as follows: Figure 3 As shown, each signal component of the first leakage signal is subjected to LMS adaptive filtering according to its corresponding IMF' component of the same order, which is expressed as:

[0105] y(n)=w T (n)x(n)

[0106] e(n)=d(n)-y(n)

[0107] w(n+1)=w(n)-μe(n)x(n)

[0108] Where x(n) is the IMF′ component of the same order as d(n) at time n, y(n) is the denoised signal component output at time n, w(n) is the coefficient vector at time n, w(n+1) is the updated coefficient vector at time n+1, e(n) is the error at time n, d(n) is the signal component of the first leakage signal at time n, and μ is the step size factor.

[0109] S6: Reconstruct the n denoised signal components to obtain a first denoised signal.

[0110] S7: Perform the same processing as steps S2 to S6 on the second leakage signal to obtain a second noise reduction signal.

[0111] The embodiment of the present invention performs the same processing as steps S2 to S6 on the second leakage signal. The specific processing process includes: using the CEEMDAN algorithm to decompose the second leakage signal into k IMF components in sequence; screening the k IMF components of the second leakage signal to obtain n signal components and m noise components, where n+m=k; reconstructing the m noise components to obtain a reconstructed signal, and using the CEEMDAN algorithm to decompose the reconstructed signal in sequence into k IMF′ components; each signal component of the second leakage signal is individually subjected to LMS adaptive filtering according to the corresponding IMF′ components of the same order to obtain n denoised signal components; and reconstructing the n denoised signal components to obtain a second noise reduction signal.

[0112] S8: Calculate the position of the leakage point using a two-point positioning model based on the first noise reduction signal and the second noise reduction signal to complete pipeline leakage positioning.

[0113] The embodiment of the present invention uses a two-point positioning model to calculate the location of the leak point based on the first and second noise reduction signals. The specific process includes: calculating the cross-correlation function of the first and second noise reduction signals; calculating the time delay between the two noise reduction signals based on the peak position of the cross-correlation function of the first and second noise reduction signals; and calculating the location of the leak point based on the time delay using the two-point positioning model. The calculation formula of the two-point positioning model is:

[0114]

[0115] Wherein, L is the distance between the first acceleration sensor and the second acceleration sensor, L1 is the distance from the leakage point to the first acceleration sensor, c is the signal propagation speed, and τ is the time delay.

[0116] In this embodiment of the present invention, the distance between the first and second acceleration sensors is L = L1 + L2, where L1 is the distance from the leak point to the first acceleration sensor, and L2 is the distance from the leak point to the second acceleration sensor. This embodiment of the present invention calculates the L1 value and then locates a location between the two acceleration sensors that is L1 away from the first acceleration sensor. This location is the leak point, thereby locating the pipeline leak.

[0117] The present invention conducted simulation experiments. First, a Gaussian random signal with a sampling rate of 8000 Hz and 8000 sampling points was generated. A bandpass filter with a passband frequency of 400-600 Hz and a cutoff frequency of 300-700 Hz was designed. Finally, the generated Gaussian random signal was filtered to obtain the leakage source signal s(t). The added noise was random noise with an SNR of -18 dB. The two leakage signals were time-delayed by 300 sampling points. The SNRs of the initial signal, the signal processed by CEEMDAN, and the signal processed by CEEMDAN-LMS were compared. Table 1 shows the comparison results. Directly reconstructing the signal components obtained after CEEMDAN decomposition yielded signal-to-noise ratios of -8.46 dB and -8.91 dB for the two signals. However, after processing using the present invention, the signal-to-noise ratios of the two signals were reduced to -6.54 dB and -6.94 dB, improvements of 1.9 dB and 1.97 dB, respectively.

[0118] Table 1 SNR before and after leakage signal processing

[0119]

[0120] The embodiment of the present invention calculates the cross-correlation function of the noise reduction signal after CEEMDAN processing and CEEMDAN-LMS processing respectively, and the function results are as follows: Figure 4 As shown in the figure, when the delay estimation errors of the two methods are both 0, after CEEMDAN denoising, the peak value of the cross-correlation function is 0.099, which is 0.061 higher than that before denoising. After denoising by the present invention, the peak value is 0.146, which is 0.108 higher, and the peak value is more prominent.

[0121] The present invention compares different pipeline leak location methods, including direct CC, EMD-CC, EEMD-CC, CEEMDAN-CC, and a combination of robust local mean decomposition and singular value decomposition with CC (RLMD-SVD-CC). Table 2 shows the performance indicators, which are tested 20 times for each method with a signal-to-noise ratio of -16 dB. N is the number of estimated delay values ​​within the interval [-315, -285] in the 20 trials, SNR1, SNR2, and MAVCC are all averaged over the 20 trials, SNR1 is the SNR of the first noise reduction signal, and SNR2 is the SNR of the second noise reduction signal. It can be seen that the present invention correctly estimates the delay value the most times, while also achieving the greatest improvement in signal-to-noise ratio and the largest cross-correlation function peak, validating the superiority of the present invention.

[0122] Table 2 Performance indicators after processing by different methods

[0123]

[0124] In order to further analyze the impact of SNR on the performance of different methods, the present invention compares the delay estimation results of the SNR variation range of [-18, -13] dB. 20 independent experiments are conducted under each SNR. The results are shown in the figure below. Figure 5 As shown in Figure 2, the number of correct delay estimates for each method decreases as the signal-to-noise ratio decreases, but the rate of decrease varies. CC decreases the fastest, while the method proposed in this paper decreases the slowest. Therefore, the method proposed in this paper has the best noise immunity and is superior to the other three methods.

[0125] The embodiment of the present invention introduces the maximum value of the cross-correlation function and its adjacent values ​​to calculate two screening factors that respectively reflect the leakage signal content and suppress the randomness of the noise, thereby solving the problem that the traditional method is susceptible to noise interference and leads to misjudgment when relying on a single indicator. The method can effectively distinguish between signal components and noise components, thereby reducing the deviation of subsequent delay estimation. The embodiment of the present invention addresses the defect of the prior art that directly reconstructs the signal components, resulting in the reconstructed signal still having residual noise. The present invention reconstructs the noise components determined by the initial CEEMDAN decomposition and performs a secondary CEEMDAN decomposition to extract a reference signal with a narrower frequency band that highly matches the noise characteristics of the original signal, and then combines LMS adaptive filtering to achieve deep denoising of the signal components. This method can effectively improve the denoising effect of the leakage signal, thereby significantly improving the positioning accuracy of the pipeline leakage point.

[0126] Although the embodiments of the present invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and variations may be made to the embodiments of the present invention without departing from the principles and spirit of the present invention, and that the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A pipeline leak location method based on CEEMDAN-LMS adaptive noise elimination, characterized in that: The following steps are involved: S1: Acquire two leakage signals upstream and downstream of the pipeline leakage point, including a first leakage signal and a second leakage signal; S2: Use the CEEMDAN algorithm to decompose the first leakage signal into k IMF components in sequence; S3: Filter the k IMF components of the first leakage signal to obtain n signal components and m noise components, where n+m=k; S4: Reconstruct the m noise components to obtain the reconstructed signal, and use the CEEMDAN algorithm to decompose the reconstructed signal into k IMF′ components in sequence; S5: Each signal component of the first leakage signal is subjected to LMS adaptive filtering according to its corresponding IMF′ component of the same order to obtain n denoised signal components; S6: reconstructing n denoised signal components to obtain a first denoised signal; S7: Perform the same processing as steps S2 to S6 on the second leakage signal to obtain a second noise reduction signal; S8: Calculate the position of the leakage point using a two-point positioning model based on the first noise reduction signal and the second noise reduction signal to complete pipeline leakage positioning.

2. The pipeline leakage location method based on CEEMDAN-LMS adaptive noise elimination according to claim 1 is characterized in that: The two leakage signals are collected by two acceleration sensors installed upstream and downstream of the pipeline leakage point, and are expressed as: x1(t)=s(t)+n1(t) x2(t)=αs(t-τ)+n2(t) Wherein, x1(t) is the first leakage signal collected by the first acceleration sensor at time t, x2(t) is the second leakage signal collected by the second acceleration sensor at time t, n1(t) is the noise signal in the first leakage signal collected by the first acceleration sensor at time t, n2(t) is the noise signal in the second leakage signal collected by the second acceleration sensor at time t, s(t) is the leakage source signal emitted by the leakage point at time t, and s(t-τ) is the leakage signal emitted by the leakage point at time t. τ The leakage source signal emitted by the leakage point at time , α is the attenuation factor, and τ is the time delay.

3. The pipeline leakage location method based on CEEMDAN-LMS adaptive noise elimination according to claim 1 is characterized in that: The process of using the CEEMDAN algorithm to decompose the first leakage signal into k IMF components in sequence includes: S21: Add multiple Gaussian white noises to the first leakage signal to obtain a noisy signal sequence; S22: performing EMD decomposition on the noisy signal sequence to obtain the current IMF component and the current order residual component; S23: Add multiple Gaussian white noises to the residual component of the current order and perform EMD decomposition again to obtain the next IMF component and the residual component of the next order; S24: Repeat S23 until the residual component cannot be decomposed any further, and obtain the final residual component and k IMF components of the first leakage signal.

4. The pipeline leakage location method based on CEEMDAN-LMS adaptive noise elimination according to claim 1 is characterized in that: The process of screening the k IMF components of the first leakage signal includes: respectively calculating a first screening factor and a second screening factor for each IMF component of the first leakage signal; Setting a first screening factor threshold and a second screening factor threshold; Each IMF component of the first leakage signal is screened respectively. If a first screening factor of the IMF component is greater than a first screening factor threshold and a second screening factor is greater than a second screening factor threshold, the IMF component is determined to be a signal component; otherwise, the IMF component is determined to be a noise component.

5. The pipeline leakage location method based on CEEMDAN-LMS adaptive noise elimination according to claim 4 is characterized in that: The process of calculating the first screening factor and the second screening factor of each IMF component of the first leakage signal includes: Calculating a cross-correlation function between an IMF component of the first leakage signal and the second leakage signal; According to the cross-correlation function between the IMF component of the first leakage signal and the second leakage signal, the initial first screening factor and the initial second screening factor of the IMF component are calculated as follows: CR1=|R(τ) max |+|R(τ+1)|+|R(τ-1)| Among them, CR1 is the initial first screening factor, CR2 is the initial second screening factor, R(τ) max is the maximum value of the cross-correlation function between the IMF component of the first leakage signal and the second leakage signal, τ is the time delay, and R(τ+1) is R(τ) max The adjacent value of the cross-correlation function delayed by 1 time point, R(τ-1) is R(τ) max The adjacent values ​​of the cross-correlation function at one time point ahead; The initial first screening factor of the IMF component is normalized to obtain the final first screening factor of the IMF component, and the initial second screening factor of the IMF component is forward-normalized and normalized to obtain the final second screening factor of the IMF component.

6. The pipeline leakage location method based on CEEMDAN-LMS adaptive noise elimination according to claim 1 is characterized in that: Each signal component of the first leakage signal is subjected to LMS adaptive filtering according to its corresponding IMF' component of the same order as that of the first leakage signal, which is expressed as follows: y(n)=w T (n)x(n) e(n)=d(n)-y(n) w(n+1)=w(n)-μe(n)x(n) Where x(n) is the IMF′ component of the same order as d(n) at time n, y(n) is the denoised signal component output at time n, w(n) is the coefficient vector at time n, w(n+1) is the updated coefficient vector at time n+1, e(n) is the error at time n, d(n) is the signal component of the first leakage signal at time n, and μ is the step size factor.

7. The pipeline leakage location method based on CEEMDAN-LMS adaptive noise elimination according to claim 1 is characterized in that: The process of calculating the location of the leak using the two-point positioning model includes: Calculating a cross-correlation function between the first noise reduction signal and the second noise reduction signal; calculating a time delay based on a cross-correlation function between the first noise reduction signal and the second noise reduction signal; According to the time delay, the two-point positioning model is used to calculate the location of the leak point.

8. The pipeline leakage location method based on CEEMDAN-LMS adaptive noise elimination according to claim 7 is characterized in that: The formula for calculating the location of the leakage point using the two-point positioning model is: Wherein, L is the distance between the first acceleration sensor and the second acceleration sensor, L1 is the distance from the leakage point to the first acceleration sensor, c is the signal propagation speed, and τ is the time delay.

Images