Pipeline leakage vibration signal noise reduction method based on parameter self-information

Abstract

The invention belongs to the technical field of structural health monitoring and vibration signal processing, and discloses a pipeline leakage vibration signal noise reduction method based on parameter self-information, and the method comprises the steps: forming an ASSA through an improved SAA algorithm, and optimizing a modal decomposition number and a penalty factor in VMD-SI; then, the effective IMF component and the invalid IMF component are distinguished by calculating the normalized self-information distance between the intrinsic mode function and the original signal; and reconstructing the filtered signal by using the effective IMF component. According to the method, the global optimization capacity is improved through the ASSA algorithm, the situation of local optimal stagnation caused by a traditional algorithm is avoided, and vibration signal irregular interference caused by the fact that the pipeline is in a complex environment is eliminated; self-information is introduced to avoid common modal aliasing, excessive decomposition and other problems in a traditional method; the method can better adapt to the characteristics of non-stationary signals, and improves the signal-to-noise ratio of the signals, thereby extracting the leakage characteristics more accurately.

Description

Technical Field

[0001] The present invention belongs to the technical field of structural health monitoring and vibration signal processing, and particularly relates to a method for reducing the noise of pipeline leakage vibration signals based on parametric self-information. Background Art

[0002] Pipelines carry key tasks such as water supply and energy transportation, and are crucial for the orderly operation of cities. Pipelines are buried underground for a long time and are always in a liquid-filled state, with a relatively high internal pressure. When a pipeline bursts or leaks, the liquid inside the pipeline will shoot out of the pipeline in the form of a high-pressure water column, which will not only cause serious waste of resources, but also cause the soil structure around the pipeline to loosen, resulting in accidents such as foundation settlement and building subsidence. During the pipeline leakage detection process, it will be affected by various noises, leading to problems such as false alarms and missed detections. Therefore, it is of great significance to perform noise reduction preprocessing on the collected signals to provide signals with a large number of leakage characteristics.

[0003] When a pipeline cracks, the internal pressure becomes locally uneven, and the fluid jets out from the crack and rubs against the pipe wall, exciting the pipeline vibration. The characteristics of the vibration signal can reflect the damage information of the structure, and vibration analysis methods have been widely used in the field of pipeline leakage damage assessment. However, due to the long-term exposure of pipelines to complex environments and being affected by various factors, their vibration signals show irregularity, increasing the difficulty of leakage detection. When traditional optimization algorithms deal with complex optimization problems, they often fall into local optimal solutions due to the complexity of the problem space, resulting in the inability to find the global optimal solution, and they are also unable to separate the noise and signals within the same frequency band. For example, empirical mode decomposition (EMD) can effectively suppress the noise of vibration signals, but EMD has problems such as insufficient envelope, over-decomposition, and mode mixing during use, and the EMD method cannot distinguish between effective IMF components and ineffective IMF components, posing a risk of noise aliasing. Summary of the Invention

[0004] To solve the above technical problems, the present invention provides a method for reducing the noise of pipeline leakage vibration signals based on parametric self-information. By improving the SAA algorithm to form the ASSA (Adaptive Sparrow Search Algorithm), the number of mode decompositions and the penalty factor in VMD (Variational Mode Decomposition) are optimized; then, by calculating the normalized self-information distance between the intrinsic mode function and the original signal, the effective IMF components and ineffective IMF components are distinguished; and the filtered signal is reconstructed using the effective IMF components.

[0005] The method for reducing the noise of pipeline leakage vibration signals based on parametric self-information according to the present invention includes the following steps:

[0006] Step 1: Collect the original vibration signal S(t) under different damage conditions of the pipeline j ;

[0007] Step 2: Add adaptive sine-cosine and Cauchy-Gaussian mutation to the Sparrow Search Algorithm (SSA) to form the ASSA algorithm, and use the original vibration signals S(t) corresponding to different damage conditions j as the search space to optimize the number of modal decompositions and the penalty factor in VMD-SI;

[0008] Step 3: Input the obtained number of modal decompositions and penalty factor into the VMD-SI algorithm to obtain the optimal decomposition parameters;

[0009] Step 4: Calculate the correlation coefficient between the intrinsic mode function and the original vibration signal based on the optimal decomposition parameters, and calculate the self-information of the intrinsic mode function and the original vibration signal accordingly;

[0010] Step 5: Normalize the obtained self-information, obtain the effective IMF components based on the normalized self-information distance between the intrinsic mode function and the original signal, reconstruct the effective IMF components, and achieve vibration signal denoising.

[0011] Furthermore, Step 2 is specifically as follows:

[0012] Step 201: Introduce an adaptive search factor γ into the SSA 1 ,

[0013]

[0014] The new discoverer position The iterative formula is as follows:

[0015]

[0016] where Cons is used to adjust the overall amplitude of the adaptive search factor; ξ is the adjustment coefficient; γ 2 and γ 3 are random numbers in [0, 2π]; t is the current iteration number, M is the maximum iteration number; H best represents the global optimal position in the current population, that is, the position where the sparrow with the highest fitness value is located; MZ is the safety value;

[0017] Step 202: Adopt an adaptive Cauchy-Gaussian mutation strategy, where the probability density function f 1 (x) of the standard Cauchy distribution and the probability density function f 2 (x) of the standard Gaussian distribution are as follows:

[0018]

[0019] The update of the new follower position is as follows:

[0020]

[0021] where x is the independent variable; Cauchy(0, 1) is the standard Cauchy distribution function; Gauss(0, 1) is the standard Gaussian distribution function; γ 4 and γ 5 are adaptive dynamic adjustment parameters.

[0022] Step 203: Take the vibration signals corresponding to different damage levels as the search space, input them into the ASSA algorithm, set the upper and lower limit ranges of the modal decomposition number k to [3, 10], set the upper and lower limit ranges of the penalty factor α to [100, 2500], set the number of iterations, and obtain the optimal values of k and α.

[0023] Furthermore, in Step 3, input the optimized modal decomposition number and penalty factor into the VMD-SI algorithm to obtain the optimal decomposition parameters, and decompose the original signal S(t) j into k IMF components.

[0024] Furthermore, Step 4 is specifically as follows:

[0025] Step 401: Calculate the correlation R between the IMF components and the original signal,

[0026]

[0027] Step 402: Calculate the self-information I(IMF k , S(t) j ),

[0028] I(IMF k , S(t) j ) = -log 2 R(IMF k , S(t) j )

[0029] where j = 1, 2..m, m represents the maximum damage level; k represents the number of reconstructed components.

[0030] Furthermore, Step 5 is specifically as follows:

[0031] Step 501: Normalize the self-information of the damage parameters:

[0032]

[0033] Step 502: Calculate the maximum slope Slop of the distance between adjacent IMF original signals S(t). The maximum slope is the boundary between the effective components and the ineffective components.

[0034] Slop = MAX|D(i + 1) - D(i)|, i = 1, 2,...k - 1;

[0035] Step 503: Retain the effective components, reconstruct the signal, and obtain the denoised signal.

[0036]

[0037] where D(i) represents the value of the self-information normalization of the i-th IMF component, and I(IMF i , S(t) j ) represents the self-information of the i-th IMF component, and min([I(IMF, S(t) j )] all ) represents the minimum value of the self-information of all components, and max([I(IMF, S(t) j )] all ) represents the maximum value of the self-information of all components. represents the reconstructed signal, and n is the number of effective components used for reconstruction.

[0038] The beneficial effects of the present invention are as follows: The method of the present invention adds adaptive sine-cosine and Cauchy-Gaussian mutation to the traditional SSA algorithm to form the ASSA algorithm, which can improve the global optimization ability in the early stage of algorithm iteration, enhance the speed of obtaining the optimal solution in the later stage of algorithm iteration, and avoid the local optimal stagnation situation brought by the traditional algorithm. It can effectively optimize the penalty factor and mode decomposition number parameters in VMD-SI, and eliminate the irregularity interference of the vibration signal caused by the pipeline in a complex environment; introduce self-information, and form the variational mode decomposition-self-information VMD-SI decomposition algorithm through the VMD and SI algorithms, avoiding common problems such as mode mixing and over-decomposition in traditional methods. At the same time, this method can better adapt to the characteristics of non-stationary signals, improve the signal-to-noise ratio of the signal, and thus extract the leakage characteristics more accurately. At the same time, it can better show the correlation between the decomposed IMF and the original signal, and reconstructing the IMF can achieve the denoising effect. Description of the Drawings

[0039] Figure 1 is the flowchart of this method;

[0040] Figure 2 is the algorithm logic structure diagram;

[0041] Figure 3 is the experimental equipment diagram;

[0042] Figure 4 is the vibration signal diagram corresponding to different working conditions;

[0043] Figure 5 Iteration curve diagram of the signal corresponding to different working conditions;

[0044] Figure 6It is the IMF decomposition diagram;

[0045] Figure 7 The correlation coefficient diagram;

[0046] Figure 8 The normalized processing result diagram;

[0047] Figure 9 The signal comparison diagram;

[0048] Figure 10 The evaluation index result diagram. Specific implementation manners

[0049] In order to make the content of the present invention be more clearly understood, the present invention will be further described in detail below according to specific embodiments in conjunction with the accompanying drawings.

[0050] As Figure 1 and Figure 2 shown, a pipeline leakage vibration signal noise reduction method based on parametric self-information according to the present invention includes the following steps:

[0051] Step 1, collect the original vibration signal S(t) under different damage conditions of the pipeline j ;

[0052] Step 2, add adaptive sine-cosine and Cauchy-Gaussian mutations to the sparrow search algorithm SSA to form the ASSA algorithm, and use the original vibration signals S(t) corresponding to different damage conditions j as the search space to optimize the number of modal decompositions and the penalty factor in VMD-SI;

[0053] Step 3, input the obtained number of modal decompositions and the penalty factor into the VMD-SI algorithm to obtain the optimal decomposition parameters;

[0054] Step 4, calculate the correlation coefficient between the intrinsic mode function and the original vibration signal based on the optimal decomposition parameters, and calculate the self-information between the intrinsic mode function and the original vibration signal accordingly;

[0055] Step 5, perform normalization processing on the obtained self-information, obtain the effective IMF components according to the normalized self-information distance between the intrinsic mode function and the original signal, and reconstruct the effective IMF components to realize vibration signal noise reduction.

[0056] In this embodiment, taking the vibration signals under the leakage conditions of four damage levels of the pipeline simulated by four valve openings as an example for experiments, the present invention will be further described in detail. Among them, the leakage level of signal 1 is the largest, and the leakage level of signal 4 is the smallest; signal 1 corresponds to level 1, signal 2 corresponds to level 2, signal 3 corresponds to level 3, and signal 4 corresponds to level 4.

[0057] The experimental pipeline is a water supply pipeline, and different opening degrees of the fire hydrant valve are set to simulate different cracking states. The experimental system consists of vibration sensors, data transmission cables, signal acquisition cards, and a central control system. The vibration signal propagates from the leakage point to both sides of the pipeline. The vibration sensors are deployed near the leakage opening to collect the pipeline vibration signals under the impact of water flow. The experimental equipment is as Figure 3 shown. The vibration signal is amplified, band-pass filtered, and analog-to-digital converted in the signal processing system, and then transmitted to the central control system via the RS232 bus. The collected vibration signal is as Figure 4 shown.

[0058] Step 2 specifically includes:

[0059] Step 201, by introducing an adaptive search factor γ 1 , which has a large weight in the early stage of algorithm iteration, a slow decreasing speed, and is conducive to improving the global optimization ability; in the later stage of algorithm iteration, the weight is small, enhancing the advantage of the algorithm in local development to increase the speed of obtaining the optimal solution:

[0060]

[0061] The iterative formula for the new discoverer's position is as follows:

[0062]

[0063] where Cons is used to adjust the overall amplitude of the adaptive search factor, and usually Cons = 1; ξ is the adjustment coefficient, ξ = 1.6; γ 2 and γ 3 are random numbers in [0, 2π], the former determines the moving distance of the sparrow, and the latter is used to control the influence of the optimal individual on the next position of the sparrow; t is the current iteration number, M is the maximum iteration number; H best represents the global optimal position in the current population, that is, the position where the sparrow with the highest fitness value is located; MZ is the safety value, and its value is usually between 0.5 and 1. When R 2 is less than MZ, it means that the current environment is relatively safe, and the sparrows can conduct extensive global searches to find better solutions; when R 2 is greater than MZ, it means that there may be predators in the environment, and the sparrows need to quickly transfer to a safer area. At this time, local search or random walk will be triggered;

[0064] Step 202, in the later stage of the SSA algorithm iteration, the followers will forage around the optimal discoverer, and there may be competition for food, resulting in the situation of local optimal stagnation; to solve this problem, an adaptive Cauchy-Gaussian mutation strategy is adopted. The Cauchy distribution and the standard Gaussian are both continuous probability distributions, with a small value at the origin and a slow rate of approaching zero. The probability density function f of the standard Cauchy distribution 1(x) and the probability density function f of the Gaussian distribution 2 are as follows:

[0065]

[0066] Use adaptive Cauchy-Gaussian mutation to perturb the individuals in the sparrow position update, expand the search scale of the sparrow algorithm, and enhance the ability of the algorithm to jump out of the local optimum.

[0067] The updated position of the new follower is as follows:

[0068]

[0069] where x is the independent variable; Cauchy(0, 1) is the standard Cauchy distribution function; Gauss(0, 1) is the standard Gaussian distribution function; γ 4 and γ 5 are adaptive dynamic adjustment parameters. During the algorithm iteration process, γ 4 gradually decreases, and γ 5 gradually increases, enabling the algorithm to jump out of the current stagnation and enhancing the local development and global exploration capabilities of the algorithm;

[0070] Step 203: Take the vibration signals corresponding to the four damage levels as the search space, input them into the ASSA algorithm, set the upper and lower limit ranges of the modal decomposition number k to [3, 10], the upper and lower limit ranges of the penalty factor α to [100, 2500], and set the number of iterations to 100 times to obtain the optimal values of k and α; the algorithm iteration curve is as Figure 5 shown. The SSA iteration curves of Signal 2 and Signal 4 are a straight line, falling into the local optimum; the fitness functions of Signal 1 and Signal 3 converge after 71 iterations and 46 iterations respectively, with a slow convergence speed. This is because SSA is a probability-based random search algorithm. When the search algorithm finds a relatively good solution, a large number of followers will flock to this position, causing the entire sparrow population to stagnate and resulting in falling into the local optimum rather than the global optimum. In addition, in the later stage of the SSA algorithm iteration, the followers will forage around the optimal discoverer, resulting in local optimum stagnation and slowing down the algorithm convergence speed. The ASSA iteration curves of working conditions 1, 2, 3, and 4 do not fall into the local optimum and converge after 12, 19, 37, and 21 iterations respectively. This is because the local development and global development capabilities of the ASSA algorithm are improved, and the probability of the algorithm jumping out of the local optimum is increased, accelerating the algorithm convergence speed.

[0071] Among them, step 3 specifically includes:

[0072] Step 301: Input the obtained modal decomposition number and penalty factor into the VMD-SI algorithm to obtain the optimal decomposition parameters; input the original signal S(t)j Decompose it into k IMF components. Taking Signal 1 and Signal 4 as examples, the IMF components and spectrograms of Signal 1 are as shown in Figure 6 (a) of Figure 6 and (b) of Figure 6 ; the IMF components and spectrograms of Signal 4 are as shown in Figure 6 (c) of Figure 6 and (d) of

[0073]

[0074] As shown in Figure 7 , the IMF5 of Signal 1, the IMF2 of Signal 2, the IMF2 of Signal 3, and the IMF1 of Signal 4 have the highest similarity, all greater than 0.7. In the figure, y represents the changing trend of the first-order function fitting the IMF component, and R 2 indicates whether the fitting degree is high; Figure 7 As shown in

[0075] Step 302, calculate the self-information I(IMF k , S(t) j ) of the IMF component and the original signal,

[0076] I(IMF k , S(t) j ) = -log 2 R(IMF k , S(t) j )

[0077] where j = 1, 2..m represents the damage level. k represents the number of reconstructed components.

[0078] Among them, Step 4 specifically includes:

[0079] Step 401, perform normalization processing on the self-information of damage parameters,

[0080]

[0081] The result is as shown in Figure 8 . Figure 8As shown, the slopes of the normalized values of Signal 1 and Signal 3 are the largest at IMF4 to IMF6, and the slopes of the normalized values of Signal 2 and Signal 4 are the largest at IMF5 to IMF6. This is because the smaller the correlation coefficient, the greater the self-information of the IMF. The smaller the self-information, the stronger the correlation between the IMF and the original signal, that is, the IMF can better capture the important features and information in the original signal.

[0082] Step 402: Calculate the maximum slope Slop of the distance between adjacent IMF original signals S(t). The maximum slope is the boundary between the effective component and the ineffective component.

[0083] Slop = MAX|D(i + 1) - D(i)|, i = 1, 2,... k - 1

[0084] Step 403: Retain the effective components and reconstruct the signal to obtain the denoised signal.

[0085]

[0086] Taking the maximum slope as the demarcation point, select the effective IMF components for reconstruction. The filtered signals and spectrograms of Signal 1 and Signal 4 are as Figure 9 shown. Figure 9 As shown, after ASSA-VMD-SI processing, the amplitude of the vibration signal shows a local decrease compared with the original signal. The amplitudes of the reconstructed Signal 1 and Signal 4 are 0.05V and 0.032V respectively. This is because there are a large number of noise signals in the original signal that are irrelevant to the damage information. The reconstructed signal eliminates the spike interference and makes the signal smoother. It can be found from the spectrograms of Signal 1 and Signal 4 that the trend presented by the spectrogram of the reconstructed signal is the same as that of the original signal as a whole; as shown in the enlarged part, in the low-frequency stage, the spectrogram of the reconstructed signal almost reproduces the spectrogram of the original signal, and in the high-frequency stage, the amplitude of the spectrogram of the reconstructed signal is smaller than that of the original signal spectrogram. Most of the signals containing damage information in the original signal are distributed in the low-frequency band, and most of the noise signals are distributed in the high-frequency band.

[0087] Among them, D(i) represents the value of the self-information normalization of the i-th IMF component, and I(IMF i , S(t) j ) represents the self-information of the i-th IMF component. min([I(IMF, S(t) j )] all ) represents the minimum value of the self-information of all components, and max([I(IMF, S(t) j )] all ) represents the maximum value of the self-information of all components. represents the reconstructed signal, and n is the number of components used for reconstruction.

[0088] For the original signals under four working conditions, this method is used to denoise and reconstruct them. Calculate the signal-to-noise ratio (SNR), mean square error (MSE), mean absolute error (MAE), and root mean square error (RMSE) of the four reconstructed signals, and compare the output results with other optimization methods.

[0089] Table 1 Comparison of the running times of the algorithms

[0090]

[0091]

[0092] The comparison of the running times of the algorithms is shown in Table 1. The optimized algorithm proposed in the present invention has the shortest average running time. Because the adaptive sine and cosine and Cauchy mutation optimize the ability to find the optimal solution. In addition, the probability of the algorithm jumping out of the local optimum is also enhanced, thus accelerating the convergence speed of the algorithm.

[0093] The four evaluation indexes of Signals 1-4 are as Figure 10 shown. Figure 10 shown. Under the same signal, compared with other denoising methods (VMD, EMD, Fourier), the method adopted in the present invention outputs the largest SNR, and reduces the MSE, MAE, and RMSE, verifying that the ASSA-VMD-SI method has good performance in denoising pipeline leakage vibration signals.

[0094] The above is only the preferred solution of the present invention, and is not used as a further limitation of the present invention. All equivalent changes made by using the content of the specification and drawings of the present invention are within the protection scope of the present invention.

Claims

1. A pipeline leakage vibration signal denoising method based on parameter self-information, characterized in that: The following steps are involved: Step 1: Collect the original vibration signal S(t) of the pipeline under different damage conditions j ; Step 2: Add adaptive sine-cosine and Cauchy-Gaussian variation to the sparrow search algorithm SSA to form the ASSA algorithm, and convert the original vibration signal S(t) corresponding to different damage conditions into j As the search space, optimize the modal decomposition number and penalty factor in VMD-SI; Step 3, input the obtained modal decomposition number and penalty factor into the VMD-SI algorithm to obtain the optimal decomposition parameters; Step 4: Calculate the correlation coefficient between the intrinsic mode function and the original vibration signal based on the optimal decomposition parameter, and calculate the self-information of the intrinsic mode function and the original vibration signal; Step 5: Normalize the obtained self-information, obtain the effective IMF component according to the normalized self-information distance between the intrinsic mode function and the original signal, reconstruct the effective IMF component, and realize vibration signal denoising.

2. According to the method for reducing noise of pipeline leakage vibration signal based on parameter self-information of claim 1, it is characterized in that: Step 2 is as follows: Step 201, introducing an adaptive search factor γ1 into SSA, New discoverer location The iteration formula is as follows: Among them, Cons is used to adjust the overall amplitude of the adaptive search factor; ξ is the adjustment coefficient; γ2 and γ3 are random numbers in [0, 2π]; t is the current iteration number, M is the maximum iteration number; H best It represents the global optimal position in the current population, that is, the position of the sparrow with the highest fitness value; MZ is the safety value; Step 202, adopting an adaptive Cauchy-Gaussian mutation strategy, wherein the probability density function f1(x) using the standard Cauchy distribution and the probability density function f2(x) using the standard Gaussian distribution are as follows: New follower location update as follows: Among them, x is the independent variable; Cauchy (0, 1) is the standard Cauchy distribution function; Gauss (0, 1) is the standard Gaussian distribution function; γ4 and γ5 are adaptive dynamic adjustment parameters; Step 203, using vibration signals corresponding to different damage levels as the search space, inputting the ASSA algorithm, setting the upper and lower limits of the modal decomposition number k to [3, 10], the upper and lower limits of the penalty factor α to [100, 2500], setting the number of iterations, and obtaining the optimal values ​​of k and α.

3. The pipeline leakage vibration signal denoising method based on parameter self-information according to claim 1 is characterized in that: In step 3, the optimized modal decomposition number and penalty factor are input into the VMD-SI algorithm to obtain the optimal decomposition parameters and the original signal S(t) j Decomposed into k IMF components.

4. The pipeline leakage vibration signal denoising method based on parameter self-information according to claim 1 is characterized in that: Step 4 is as follows: Step 401, calculate the correlation R between the IMF component and the original signal, Step 402, calculate the self-information I (IMF k ,S(t) j ), I(IMF k ,S(t) j )=-log2R(IMF k ,S(t) j ) Wherein, j=1,2..m, m represents the maximum damage level; k represents the number of reconstructed components.

5. The pipeline leakage vibration signal denoising method based on parameter self-information according to claim 4 is characterized in that: Step 5 is as follows: Step 501, normalize the damage parameter self-information: Step 502, calculate the maximum slope Slop of the distance between adjacent IMF original signals S(t), where the maximum slope is the boundary between the effective component and the invalid component: Slop=MAX|D(i+1)-D(i)|,i=1,2,...k-1; Step 503, retain the effective component, reconstruct the signal, and obtain the denoised signal: Where D(i) represents the self-information normalized value of the i-th IMF component, I(IMF i ,S(t) j ) represents the self-information of the ith IMF component, min([I(IMF,S(t) j )] all ) represents the minimum value of the self-information of all components, max([I(IMF,S(t) j )] all ) represents the maximum value of the self-information of all components, represents the reconstructed signal, n is the number of effective components used for reconstruction, IMF q represents the qth IMF component, 1≤q≤n.

Images