Bridge cable local damage detection method based on magnetic flux leakage array signal fusion

Abstract

The invention provides a bridge inhaul cable local damage detection method based on magnetic flux leakage array signal fusion, and relates to the technical field of bridge inhaul cable nondestructive testing, and the main steps are as follows: collecting magnetic flux leakage signals; preprocessing a single-channel magnetic flux leakage signal; reducing the dimension of the multi-channel signal by using a principal component analysis method; and carrying out envelope extraction on the dimension-reduced signal and removing a low-frequency noise signal. According to the method, a principal component analysis method is introduced in a magnetic flux leakage signal processing and analysis process, data dimension reduction processing is performed on bridge cable magnetic flux leakage array signals by using a principal component analysis technology, damage is detected and positioned, and most important data characteristics are selected and reserved from disordered signal data; by reserving signal components with higher contribution rate to damage diagnosis, the accuracy of damage signal identification can be improved, and the diagnosis efficiency is improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of non-destructive testing of bridge cables, and specifically relates to a method for detecting local damage of bridge cables based on magnetic flux leakage array signal fusion. Background Art

[0002] Bridges are essential components of modern transportation and infrastructure, and their structural health and safety are crucial to public and economic stability. The cable system, typically composed of multiple parallel steel wires, is a crucial component in supporting and maintaining the bridge. As bridges age, monitoring and maintenance of the cable system becomes increasingly important to ensure structural integrity and safety.

[0003] Magnetic flux leakage (MFL) testing is a non-invasive method commonly used to monitor the health of bridge cables. By measuring the magnetic field distribution on the wire surface, potential problems such as wire breakage, corrosion, or wear can be detected. However, analyzing and interpreting MFL signals is a complex task, especially when the wires are numerous and arranged in parallel. Traditional signal processing methods may not provide sufficient accuracy and reliability.

[0004] Existing technologies for processing magnetic flux leakage array signals from bridge cables present several technical challenges. Parallel steel wires can interfere with each other. During sensor detection, the inevitable vibration of the wire ropes causes the sensors to pick up vibration noise. This noise is widespread in engineering practice and difficult to distinguish from low-frequency noise, easily leading to false alarms or missed detections. Furthermore, processing large amounts of magnetic flux leakage data requires highly complex algorithms and computing resources, making traditional detection methods inefficient. Furthermore, existing technologies are still not accurate or sensitive enough to detect small-scale damage or predict the health of steel wires. Summary of the Invention

[0005] (1) Technical problems solved The purpose of the present invention is to overcome the shortcomings of the existing technology and provide a method for detecting local damage of bridge cables based on leakage magnetic array signal fusion. Unlike the original damage signal diagnosis method, the method studied in the present invention introduces the principal component analysis method in the leakage magnetic array signal processing and analysis process. By using the principal component analysis technology to perform data dimensionality reduction processing on the leakage magnetic array signal of the bridge cable and detect and locate the damage, the accuracy of damage signal identification can be improved and the safe use performance of the bridge cable can be enhanced.

[0006] (2) Technical solution To achieve the above-mentioned object, the present invention provides a method for detecting local damage of bridge cables based on magnetic flux leakage array signal fusion, comprising the following steps: S1, collecting magnetic flux leakage signals; After the bridge cable to be tested is saturated magnetized, the circumferential arrangement of the cable outside the cable is used to m A Hall sensor obtains the radial magnetic leakage information on its surface and obtains the original magnetic leakage signal x ( t , m ),in, t is the total number of sampling points; S2, single-channel magnetic flux leakage signal preprocessing; S2 includes the following sub-steps: 2-1. Using moving average filter to analyze magnetic flux leakage signal x ( t , m ) is processed to obtain the detrended signal y ( t , m ); 2-2. Filtered signal y ( t , m ) is formed after cubic spline interpolation y ( T , M ), thereby improving the resolution of the signal, where M and T Respectively represent m and t Interpolated data; 2-3, yes y ( T , M ) is subjected to median filtering to obtain Y ( T , M ); S3, using principal component analysis to reduce the dimensionality of the multi-channel signal; S3 includes the following sub-steps: 3-1. Create a data matrix to centralize the data of each signal channel; The preprocessed data matrix Y ( T , M ), where each row represents a sample, i.e., a signal collected by a channel, and each column represents a feature, i.e., a sensor channel, creating a Y ( T , M ) Matrices with the same number of rows A ( T , M ), where each element is equal to Y ( T , M ) corresponds to the mean of the feature, by Y ( T ,M ) by subtracting the mean of the corresponding dimension from each feature, which can adjust the mean of the data to zero and realize the centralization of the data; 3-2. Calculate the covariance matrix; 3-3. Calculate the eigenvalues and eigenvectors of the covariance matrix; Calculate the covariance matrix C The eigenvector matrix of V and the eigenvalue matrix , then according to the eigenvalue Sort the eigenvectors by size and store them in the matrix V In the eigenvalues, the eigenvalues are sorted in descending order and stored in the column vector d , calculate the sum of eigenvalues S Used for subsequent calculation of contribution rate; 3-4. Screening and retention of eigenvalues; Select the appropriate principal component according to the importance of the eigenvalue, and set a cumulative variance contribution rate threshold. G , loop from 1 to the number of eigenvalues M Iterate the length of , and in each iteration, calculate the current number of iterations i The sum of the previous eigenvalues and the sum of the total eigenvalues S The ratio of r ; 3-5. Data dimensionality reduction; According to the number of eigenvalues after screening f , from the eigenvector matrix V Before taking f The eigenvectors form a new transformation matrix V’ ,pass Project the original signal matrix onto the filtered f The dimensionality reduction matrix is obtained on the subspace composed of principal components. Z ( T , f ), the signal waveform after feature extraction can more clearly distinguish the damage signal from the noise signal; S4, performing envelope extraction on the dimension-reduced signal and removing low-frequency noise signals; S4 includes the following sub-steps: 4-1. Perform envelope detection on the dimensionality-reduced signal and extract the envelope of the signal; 4-2. Setting the threshold L The signal envelope peak is compared with the threshold. The part exceeding the threshold is regarded as the damage signal, and the part below the threshold is regarded as the noise signal. The part below the threshold is set to zero to eliminate the noise signal, and the part exceeding the threshold is retained to enhance the characteristics of the damage signal. 4-3. Through threshold judgment, all signals exceeding the threshold havel From this, we can judge that there are l The damage can be detected locally, thus realizing the qualitative diagnosis of local damage of bridge cables.

[0007] Preferably, the signal after detrending in 2-1 y ( t , m ) is calculated as:

[0008] in, b ( t , m )for x ( t , m ) is a trend signal after moving average.

[0009] Preferably, in 2-3 Y ( T , M ) is calculated as:

[0010] in, k is the window width of the median filter, median () is the median function.

[0011] Preferably, the data matrix calculated in 3-2 is Y ( T , M )'s covariance matrix C , can be expressed using the following formula;

[0012] in, Y is the data matrix Y ( T , M ) is the transpose of the data matrix ( T , M ), covariance matrix C It can reflect the relationship between variables in the data set.

[0013] Preferably, in 3-4, when r > G , indicating that the current eigenvalue has reached the set contribution rate requirement, the cycle ends, and the f eigenvalues, where:

[0014] (3) Beneficial effects Beneficial effects of the present invention: The present invention provides a method for detecting local damage of bridge cables based on magnetic flux leakage array signal fusion. By introducing a principal component analysis method in the process of magnetic flux leakage signal processing and analysis, the principal component analysis technology is used to perform data dimensionality reduction processing on the magnetic flux leakage array signals of the bridge cables and to detect and locate damage. The most important data features are selected and retained in the chaotic signal data. By retaining the signal components with higher contribution rates to damage diagnosis, the accuracy of damage signal identification can be improved, thereby improving the diagnostic efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0015] Figure 1 The figure is a flow chart of a method for detecting local damage of bridge cables based on magnetic flux leakage array signal fusion; Figure 2 This is the damage distribution map of the bridge cable sample; Figure 3 This is a schematic diagram of the arrangement and protection of bridge cable wires; Figure 4 The original magnetic leakage signal diagram is collected for 16 Hall sensors; Figure 5 It is the waveform diagram of the original magnetic flux leakage signal after preprocessing; Figure 6 This is the signal waveform diagram of 16-channel signals after dimension reduction by principal component analysis technology; Figure 7 The signal waveform is obtained by performing envelope extraction on the signal waveform after dimensionality reduction; Figure 8 Signal waveform diagram for threshold judgment and continued noise signal removal; Figure 9 This is a signal waveform diagram for continuing to perform envelope extraction on the signal after threshold judgment. DETAILED DESCRIPTION

[0016] The following will provide a clear and complete description of the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of them. All other embodiments obtained by ordinary technicians in this field based on the embodiments of the present invention without making any creative efforts shall fall within the scope of protection of the present invention.

[0017] See also Figure 1-9 The present invention provides a method for detecting local damage of bridge cables based on magnetic flux leakage array signal fusion, comprising the following steps: S1, collecting magnetic flux leakage signals; After the bridge cable to be tested is saturated magnetized, the circumferential arrangement of the cable outside the cable is used to mA Hall sensor obtains the radial magnetic leakage information on its surface and obtains the original magnetic leakage signal x ( t , m ),in, t is the total number of sampling points; like Figure 2 As shown, in this embodiment, a broken wire cable sample was customized. A parallel steel wire bridge cable with a length of 4 meters and a diameter of 95 mm (including PE sheath) was selected. Three two-wire broken wire samples and three three-wire broken wire samples were made on the cable. The six damages were arranged at equal distances, and the damage width was 5 mm. Non-magnetic aluminum materials, hard plastics and other fillers were used at the fracture to prevent cross-section contact and extrusion narrowing. The schematic diagram of the arrangement and protection of the bridge cable steel wires is shown in the figure. Figure 3 shown.

[0018] In this embodiment, m =16, t =3000, the sixteen signals collected by the Hall sensor array are as follows Figure 4 shown.

[0019] S2, single-channel magnetic flux leakage signal preprocessing; S2 includes the following sub-steps: 2-1. Using moving average filter to analyze magnetic flux leakage signal x ( t , m ) is processed to obtain the detrended signal y ( t , m ), the detrended signal y ( t , m ) is calculated as:

[0020] in, b ( t , m )for x ( t , m ) Trend signal after moving average; 2-2. Filtered signal y ( t , m ) is formed after cubic spline interpolation y ( T , M ), thereby improving the resolution of the signal, where M and T Respectively represent m and t Interpolated data; 2-3, yesy ( T , M ) is subjected to median filtering to obtain Y ( T , M ), Y ( T , M ) is calculated as:

[0021] in, k is the window width of the median filter, median () is the median function; In this embodiment, after the 16-channel sensor signals are subjected to wavelet filtering and trend processing, the signal image after the vibration noise is further removed is as follows: Figure 5 shown.

[0022] S3, using principal component analysis to reduce the dimensionality of the multi-channel signal; S3 includes the following sub-steps: 3-1. Create a data matrix to centralize the data of each signal channel; The preprocessed data matrix Y ( T , M ), where each row represents a sample, i.e., a signal collected by a channel, and each column represents a feature, i.e., a sensor channel. Y ( T , M ) Matrices with the same number of rows A ( T , M ), where each element is equal to Y ( T , M ) corresponds to the mean of the feature. Y ( T , M ) by subtracting the mean of the corresponding dimension from each feature, which can adjust the mean of the data to zero and realize the centralization of the data; 3-2. Calculate the covariance matrix; Calculate the data matrix Y ( T , M )'s covariance matrix C , can be expressed using the following formula;

[0023] in, Y is the data matrix Y ( T , M ), is the transpose of the data matrix ( T , M ), covariance matrix C It can reflect the relationship between variables in the data set; 3-3. Calculate the eigenvalues and eigenvectors of the covariance matrix; Calculate the covariance matrix C The eigenvector matrix of V and the eigenvalue matrix Next, according to the eigenvalue Sort the eigenvectors by size and store them in the matrix V In the eigenvalues, the eigenvalues are sorted in descending order and stored in the column vector d , calculate the sum of eigenvalues S Used for subsequent calculation of contribution rate.

[0024] 3-4. Screening and retention of eigenvalues; Select the appropriate principal component based on the importance of the eigenvalue. By setting a cumulative variance contribution rate threshold G , loop from 1 to the number of eigenvalues M Iterate the length of , and in each iteration, calculate the current number of iterations i The sum of the previous eigenvalues and the sum of the total eigenvalues S The ratio of r ;when r > G , indicating that the current eigenvalue has reached the set contribution rate requirement, the cycle ends, and the f eigenvalues, where:

[0025] 3-5. Data dimensionality reduction; According to the number of eigenvalues after screening f , from the eigenvector matrix V Before taking f The eigenvectors form a new transformation matrix V’ ,pass Project the original signal matrix onto the filtered f The dimensionality reduction matrix is obtained on the subspace composed of principal components. Z ( T , f ), the signal waveform after feature extraction can more clearly distinguish the damage signal from the noise signal; In this embodiment, the data after dimension reduction by principal component analysis algorithm is as follows Figure 6 As shown in the figure, it can be clearly found that the signal peak at the damaged area is much larger than the signal size in the non-damaged area.

[0026] S4, performing envelope extraction on the dimension-reduced signal and removing low-frequency noise signals; S4 includes the following sub-steps: 4-1. Perform envelope detection on the dimensionality-reduced signal and extract the envelope of the signal; In this embodiment, the signal waveform after the dimension reduction signal is extracted by envelope is as follows: Figure 7 shown.

[0027] 4-2. Setting the threshold L The signal envelope peak is compared with the threshold. The part exceeding the threshold is regarded as the damage signal, and the part below the threshold is regarded as the noise signal. The part below the threshold is set to zero to eliminate the noise signal, and the part exceeding the threshold is retained to enhance the characteristics of the damage signal. In this embodiment, the signal waveform after threshold determination and clearing after envelope extraction is as follows: Figure 8 shown.

[0028] 4-3. Through threshold judgment, all signals exceeding the threshold have l From this, we can judge that there are l Local damage can be detected to achieve qualitative diagnosis of local damage of bridge cables; In this example, Figure 9 As shown in the figure, through threshold judgment, there are 6 signals exceeding the threshold, which indicates that there are 6 damages on the bridge cable, thus achieving qualitative diagnosis of local damage to the bridge cable.

[0029] While embodiments of the present invention have been shown and described, it will be appreciated by those skilled in the art that various changes, modifications, substitutions, and variations may be made to these embodiments without departing from the principles and spirit of the invention, and that the scope of the invention is defined by the appended claims and their equivalents.

Claims

1. A method for detecting local damage of bridge cables based on magnetic flux leakage array signal fusion, characterized in that: The following steps are involved: S1, collecting magnetic flux leakage signals; After the bridge cable to be tested is saturated magnetized, the circumferential arrangement of the cable outside the cable is used to m A Hall sensor obtains the radial magnetic leakage information on its surface and obtains the original magnetic leakage signal x ( t , m ),in, t is the total number of sampling points; S2, single-channel magnetic flux leakage signal preprocessing; S2 includes the following sub-steps: 2-1. Using moving average filter to analyze magnetic flux leakage signal x ( t , m ) is processed to obtain the detrended signal y ( t , m ); 2-2. Filtered signal y ( t , m ) is formed after cubic spline interpolation y ( T , M ), thereby improving the resolution of the signal, where M and T Respectively represent m and t Interpolated data; 2-3, yes y ( T , M ) is subjected to median filtering to obtain Y ( T , M ); S3, using principal component analysis to reduce the dimensionality of the multi-channel signal; S3 includes the following sub-steps: 3-1. Create a data matrix to centralize the data of each signal channel; The preprocessed data matrix Y ( T , M ), where each row represents a sample, i.e., a signal collected by a channel, and each column represents a feature, i.e., a sensor channel, creating a Y ( T , M ) Matrices with the same number of rows A ( T , M ), where each element is equal to Y ( T , M ) corresponds to the mean of the feature, by Y ( T , M ) by subtracting the mean of the corresponding dimension from each feature, which can adjust the mean of the data to zero and realize the centralization of the data; 3-2. Calculate the covariance matrix; 3-3. Calculate the eigenvalues and eigenvectors of the covariance matrix; Calculate the covariance matrix C The eigenvector matrix of V and the eigenvalue matrix , then according to the eigenvalue Sort the eigenvectors by size and store them in the matrix V In the eigenvalues, the eigenvalues are sorted in descending order and stored in the column vector d , calculate the sum of eigenvalues S Used for subsequent contribution rate calculation; 3-4. Screening and retention of eigenvalues; Select the appropriate principal component according to the importance of the eigenvalue, and set a cumulative variance contribution rate threshold. G , loop from 1 to the number of eigenvalues M Iterate the length of , and in each iteration, calculate the current number of iterations i The sum of the previous eigenvalues and the sum of the total eigenvalues S The ratio of r ; 3-5. Data dimensionality reduction; According to the number of eigenvalues after screening f , from the eigenvector matrix V Before taking f The eigenvectors form a new transformation matrix V’ ,pass , projecting the original signal matrix onto the filtered f The dimensionality reduction matrix is obtained on the subspace composed of principal components. Z ( T , f ), the signal waveform after feature extraction can more clearly distinguish the damage signal from the noise signal; S4, performing envelope extraction on the dimension-reduced signal and removing low-frequency noise signals; S4 includes the following sub-steps: 4-1. Perform envelope detection on the dimensionality-reduced signal and extract the envelope of the signal; 4-2. Setting the threshold L The signal envelope peak is compared with the threshold. The part exceeding the threshold is regarded as the damage signal, and the part below the threshold is regarded as the noise signal. The part below the threshold is set to zero to eliminate the noise signal, and the part exceeding the threshold is retained to enhance the characteristics of the damage signal. 4-3. Through threshold judgment, all signals exceeding the threshold have l From this, we can judge that there are l The damage can be detected locally, thus realizing the qualitative diagnosis of local damage of bridge cables.

2. The method for detecting local damage of bridge cables based on magnetic flux leakage array signal fusion according to claim 1 is characterized in that: The signal after detrending in 2-1 y ( t , m ) is calculated as:

3. Among them, b ( t , m )for x ( t , m ) is a trend signal after moving average.

4. The method for detecting local damage of bridge cables based on magnetic flux leakage array signal fusion according to claim 2 is characterized in that: 2-3 Y ( T , M ) is calculated as:

5. Among them, k is the window width of the median filter, median () is the median function.

6. The method for detecting local damage of bridge cables based on magnetic flux leakage array signal fusion according to claim 3 is characterized in that: The data matrix calculated in 3-2 Y ( T , M )'s covariance matrix C , can be expressed using the following formula; 7. Among them, Y is the data matrix Y ( T , M ), is the transpose of the data matrix ( T , M ), covariance matrix C It can reflect the relationship between variables in the data set.

8. The method for detecting local damage of bridge cables based on magnetic flux leakage array signal fusion according to claim 4 is characterized in that: In 3-4 above, when r > G , indicating that the current eigenvalue has reached the set contribution rate requirement, the cycle ends, and the f eigenvalues, where: 。

Images