Corrosion defect quantitative evaluation method fused with spatial-temporal distribution of multiple acoustic emission sources

Abstract

The invention discloses a corrosion defect quantitative evaluation method fusing spatial and temporal distribution of multiple acoustic emission sources, and belongs to the field of nondestructive testing. The method comprises the following steps: carrying out acoustic emission event judgment on collected multichannel acoustic emission original signals, and screening effective acoustic emission event signal segments; preprocessing all the effective acoustic emission event signals, including band-pass filtering, window function processing and adjustable Q factor wavelet transform, and performing time reversal focusing analysis to obtain global amplitude response distribution of each event signal and obtain an acoustic emission source position corresponding to each event signal; all global amplitude responses are accumulated, the boundary contour and the geometric parameter set of the corrosion area are extracted through a spatial clustering algorithm and convex hull fitting after binarization processing, and the boundary contour and the parameter set of the corrosion area are obtained. According to the method, information of a plurality of acoustic emission events in a section of signal is effectively fused, the accuracy of corrosion defect positioning is improved, and extraction and quantitative characterization of a corrosion defect contour are realized.

Description

Technical Field

[0001] This invention relates to a defect detection method based on acoustic emission signals, and more particularly to a quantitative evaluation method for corrosion defects that integrates the spatiotemporal distribution of multiple acoustic emission sources. This method is applicable to the location and contour parameter extraction of corrosion defects in metal plate structures, and belongs to the field of nondestructive testing technology. Background Technology

[0002] Metal plate structures are widely used in aerospace, shipbuilding, bridge and other engineering fields. During long-term use, they are prone to corrosion, cracks and other damage, which affect structural safety. Acoustic emission technology has the advantages of high real-time performance and high sensitivity, and is suitable for early damage detection of metal plate structures. However, existing acoustic emission methods are mostly focused on corrosion determination and type identification [1], and are insufficient in the extraction and characterization of defect geometric information, which makes it difficult to meet the quantitative detection needs in actual engineering.

[0003] During corrosion, acoustic emission sources are often distributed in the defect area in a multi-point, discrete manner, exhibiting characteristics of random timing and unknown distribution. The signals excited by these discrete sources usually have weak amplitude and short duration, and are subject to reflection from structural boundaries and interference from environmental noise. Existing studies have attempted to use the statistical characteristics and amplitude distribution of acoustic emission signals for analysis to assess the extent of corrosion [2]. However, such methods are easily affected by reflection from structural boundaries and noise interference, and are difficult to accurately reflect the actual size characteristics of defects, requiring a signal processing method with good robustness.

[0004] As a sound source reconstruction technique based on signal back propagation, the time reversal method can realize the spatiotemporal focusing of energy at the sound source location. In recent years, this method has shown excellent stability and accuracy in the localization of acoustic emission sources. ZexingYu et al. [3] proposed a step-by-step time reversal method to realize the accurate localization of acoustic emission sources in stiffened plates, effectively improving the localization accuracy and robustness of acoustic emission sources under conditions of wave scattering and uneven energy distribution. R. Zeman et al. [4] proposed a time reversal localization method based on image source model, which improved the localization accuracy under complex boundary conditions by modeling multipath reflection and suppressed the interference of noise and reflection errors on the focusing quality. The above studies show that even under complex boundary conditions and noise interference, the time reversal method still shows good localization stability and engineering adaptability, providing important technical support for the localization of corrosion defects in metal plate structures using acoustic emission technology.

[0005] Although the time reversal method has shown good focusing ability in acoustic emission source localization[5], most methods are limited to the reconstruction of a single or multiple concurrent sources. Given the discrete spatial distribution and random temporal excitation characteristics of acoustic emission sources during corrosion, it is difficult to obtain complete defect geometric features by relying solely on the localization of a single source. In order to achieve accurate localization and contour extraction of corrosion defects, a signal processing and feature extraction method that integrates the spatiotemporal distribution characteristics of multiple acoustic emission sources is needed to address the acoustic emission source features during corrosion.

[0006] This invention proposes a quantitative evaluation method for corrosion defects based on the spatiotemporal distribution of multiple acoustic emission sources. This method integrates the time-reversal focusing results of signals generated by multiple acoustic emission sources with spatial statistical analysis to extract the high-density aggregation characteristics of acoustic emission sources in the corrosion region, thereby achieving a quantitative characterization of the location and contour of corrosion defects. Summary of the Invention

[0007] This invention addresses the shortcomings of existing acoustic emission detection methods in extracting geometric features of corrosion defects by proposing a quantitative evaluation method for corrosion defects that integrates the spatiotemporal distribution of multiple acoustic emission sources. This method performs event identification and extraction on a long-term acquired raw signal, sequentially performing time-reversal focusing on multiple acoustic emission event signal segments to reconstruct the location of the acoustic emission source corresponding to each event signal segment. Considering that corrosion acoustic emission sources are concentrated in the defect area, all source location information is statistically analyzed, and their spatial distribution characteristics are extracted through cluster analysis, thereby achieving the localization of corrosion defects and estimation of geometric parameters in metal plate structures.

[0008] The basic principle of the method of this invention is as follows:

[0009] 1) Acoustic emission signal extraction and enhancement

[0010] Given the weak amplitude of the acoustic emission signal from corrosion and its susceptibility to noise, a dual criterion based on energy threshold and frequency domain peak determination is employed to automatically filter out effective acoustic emission event signal segments from the background noise, and simultaneously extract the corresponding signals from all sensor channels. Subsequently, each extracted signal segment undergoes preprocessing, including bandpass filtering, window function processing, and adjustable Q-factor wavelet transform, to improve the signal-to-noise ratio and enhance time-frequency characteristics, providing high-quality data for subsequent time-reversal analysis.

[0011] First, the raw signals collected over a long period of time... Event detection is performed (based on channel m). A sliding window method is used for traversal analysis. Let the window length be... The overlap rate between adjacent windows is sequentially process the i-th signal segment Perform energy and frequency domain determination. When the root mean square value of this segment... satisfy:

[0012]

[0013] in The standard deviation of background noise. If the coefficient is empirical, then this signal segment is considered a candidate signal segment. .

[0014] right Perform spectrum analysis, if in a specific frequency band Memory in normalized main peak amplitude satisfy:

[0015]

[0016] in If the amplitude threshold is reached, the segment is determined to be a valid acoustic emission event signal segment, and its time index is recorded. .

[0017] Based on the time index of the valid signal segment determined in channel m Extract the synchronization signal segments from the remaining channels within the same time range to form a multi-channel signal group for this event. .

[0018] To suppress low-frequency background disturbances and high-frequency noise, the multi-channel signal group Feature band identification and filtering are performed. The multi-channel signal is sequentially divided into C(M,2) channel pairs (M being the number of channels). Each channel pair is then divided into two sub-segments, A and B, with an overlap rate of [missing information]. Calculate their cross-power spectra respectively:

[0019]

[0020] Where f is the frequency; (i=1,2) represents the spectrum of the corresponding signal segment. Indicates complex conjugation.

[0021] In the frequency domain, the tiny delay between two signals This will cause its phase difference spectrum to exhibit a linear trend, that is:

[0022]

[0023] To quantitatively identify characteristic frequency bands, the first derivative of the cross-spectral phase difference spectrum within a certain frequency interval must satisfy:

[0024]

[0025] in For the set change threshold, then the frequency band It was determined to be a characteristic frequency band. Based on this, a bandpass filter was designed for... Frequency constraint processing is performed to obtain This is done to preserve the main propagation components and suppress noise interference.

[0026] To reduce spectral leakage and boundary effects, the filtered signal... Apply window function To further improve the signal-to-noise ratio and enhance the dominant propagation mode, a multi-scale decomposition of the corrosion acoustic emission signal is performed using an adjustable Q-factor wavelet transform to separate different frequency band components and enhance the dominant frequency characteristics, taking into account the multi-scale characteristics of the signal. The iterative decomposition process is expressed as follows:

[0027]

[0028] in, Let J be the high-frequency wavelet coefficients of the j-th layer. For the low-frequency scaling coefficients of the Jth layer, and These represent the high-pass and low-pass decomposition operators, respectively.

[0029] For the selected multiscale coefficients and Reconstructing the time-domain signal using inverse adjustable Q-factor wavelet transform yields the denoised and enhanced signal:

[0030]

[0031] The reconstructed signal group As input for subsequent time reversal focusing analysis.

[0032] 2) Time-reversal focusing analysis of single-event acoustic emission signals

[0033] Based on the denoised and enhanced single-event signal obtained, it is necessary to further determine the source location corresponding to the event signal. Therefore, this invention employs a time-reversal focusing method to locate the source of the single-event signal. According to the Mindlin board theory, if an acoustic emission source exists at position O... Then the time-domain response of the signal after it propagates to the sensor location M is:

[0034]

[0035] in This represents the distance between point M and point O; This represents the nth discrete angular frequency component; This refers to the wavenumber of the flexural wave at the corresponding frequency. It is a zero-order Hankel function of the second kind.

[0036] right After performing frequency domain complex conjugation, the position is obtained. The frequency domain expression of the time-reversed signal is:

[0037]

[0038] in This represents the signal spectrum.

[0039] Will Reverse launch to any position on the board Then, its frequency domain response is expressed as:

[0040]

[0041] After frequency domain compensation and inverse Fourier transform, the time reversal result at point P is obtained:

[0042]

[0043] After calculating the time reversal result at point P on the plate, the above time reversal process is applied to all spatial locations of the discrete mesh on the plate surface. The above steps are executed sequentially to obtain a time-reversed wave field. Among them, in the moment of focus At the actual sound source location, the wave field will exhibit local energy peaks.

[0044] Since the focusing results of the true source location are consistent across different signal channels, while noise interference is random, multi-channel signal superposition processing can enhance the energy peak of the true source point and suppress background noise, thereby effectively improving the stability and accuracy of the positioning results. At the focusing moment... The channel C signal is located at any position on the board. The amplitude of the time-reversal wave field is The sum of the focusing amplitudes of the multi-channel signals is expressed as:

[0045]

[0046] Where M represents the number of sensor channels. Through the above superposition operation, the global amplitude response of the event signal at the time of time-domain focusing is obtained. The peak position is the source localization result of the multi-channel signal of the event.

[0047] 3) Distribution statistics and clustering of multiple acoustic emission sources

[0048] After locating the source of a single-event signal, the overall spatiotemporal distribution characteristics of multiple acoustic emission event signals are considered. Since the corrosion process generates continuous and dense acoustic emission activity within the defect region, the source locations of all event signals exhibit a high-density distribution within the defect region. Therefore, by statistically analyzing and clustering the source location results of multiple event signals, the focusing response of the defect region can be effectively enhanced, enabling the localization of corrosion defects and estimation of their geometric parameters.

[0049] Assume the original signal The number of valid acoustic emission events is N, and the global amplitude response of the signal of the k-th event is... Then the cumulative focused response of all event signals is:

[0050]

[0051] By spatially accumulating the localization results of all event signal sources, the focusing response of the defect area is significantly enhanced, and interference signals caused by random noise or isolated events are suppressed, thus forming a spatial probability distribution map of the acoustic emission sources. To automatically identify high-density clusters of acoustic emission sources from this distribution map, it is necessary to convert it into a discrete point set and perform spatial clustering analysis. A threshold is set. right Binarization is performed to obtain the set of high-amplitude points:

[0052]

[0053] in This is the normalized amplitude.

[0054] Spatial clustering analysis of set P was performed using the density-based spatial clustering algorithm (DBSCAN):

[0055]

[0056] in, This is the cluster radius threshold. Let K be the kth cluster, and K be the number of clusters identified.

[0057] Select the cluster with the largest area Using convex hull operators Calculate its minimum convex hull:

[0058]

[0059] in, This represents the set of boundary points of the convex hull. The geometric envelope corresponding to the convex hull B is the edge contour of the eroded region.

[0060] Based on the convex hull B, the geometric center of the region Calculated using the following formula:

[0061]

[0062] Where A is the area of ​​the polygon, and satisfies The closing condition.

[0063] Least-squares elliptic fitting is performed on the convex hull B to obtain the major axis a and minor axis b (a≥b) of the defect region. Parameter set As a quantitative evaluation index for corrosion defects.

[0064] This invention is implemented based on an acoustic emission detection experimental system, such as... Figure 1 As shown. The system mainly includes: a host computer 1, an acoustic emission signal analyzer (DS5, Soft Island Technology) 2, preamplifiers 3-6, acoustic emission sensors S1-S4 (RS2A) 7-10, and a metal plate sample 11 to be tested. During the corrosion experiment, a corrosion zone 12 is set on the surface of the metal plate sample 11, and a corrosion reaction occurs by dripping a corrosion solution through a fixed quartz tube.

[0065] Acoustic emission sensors 7-10 are fixed to the surface of the metal plate sample 11 using a coupling agent to collect acoustic emission signals generated during corrosion in real time. The output terminals of sensors 7-10 are connected to the input terminals of preamplifiers 3-6, and the amplified signals are input to acoustic emission signal analyzer 2. Analyzer 2 synchronously acquires and converts multi-channel signals to analog-to-digital signals, and transmits them to host computer 1 via a data cable. Host computer 1 is responsible for data storage and signal processing, sequentially performing event extraction and enhancement, time-reversal focusing of single-event signals, and statistical and cluster analysis of sound source distribution on the acquired data, thereby realizing the location of corrosion defects and estimation of geometric parameters in the metal plate structure.

[0066] To ensure signal quality, sensors 7-10 are evenly distributed at key locations on the metal plate sample 11 during installation and are tightly bonded to the sample surface using a coupling agent. Preamplifiers 3-6 are used to suppress energy attenuation and noise interference during long-distance transmission. The sampling rate and gain of the DS5 analyzer 2 are adjustable to ensure effective acquisition of the target acoustic emission signal.

[0067] The implementation process of the corrosion defect detection method based on the spatiotemporal distribution of acoustic emission signals proposed in this invention is as follows: Figure 2 As shown, it is achieved through the following steps:

[0068] Step 1: During the corrosion experiment, an array of acoustic emission sensors arranged at the four corners of the metal plate is used to acquire four channels of raw acoustic emission signal data.

[0069] Step 2: Use the sliding window method to process the original signal. To determine acoustic emission events, candidate signal segments are first identified and selected in the time domain according to equation (1). Then, according to equation (2), the normalized main peak amplitude is analyzed in the frequency domain to determine whether it meets the set threshold, thereby determining the effective acoustic emission event signal segment. The effective signal segment time index is determined based on the reference channel m. Extract the synchronization signal segments from the remaining channels within the same time range to form a multi-channel signal group for this event. .

[0070] Step 3: For each valid acoustic emission event signal group Execute in sequence:

[0071] ① Combine the M channels in pairs sequentially to form C(M,2) channel pairs. Calculate the cross-power spectrum of each channel pair according to equations (3)-(5). Phase difference spectrum Extract its frequency features;

[0072] ② Perform statistical analysis on all results, count the number of occurrences of each frequency, and select the frequency range with high and concentrated frequency occurrences as the characteristic frequency band. Design a bandpass filter and filter the signal;

[0073] ③ For the filtered signal Apply window function The windowed signal is obtained, and its Q-factor wavelet transform is performed according to equations (6) and (7) to obtain the reconstructed signal. ;

[0074] ④ Based on the Mindlin plate theory, the reconstructed signal is time-reversed according to equations (8)-(11), and the time-reversed wavefield of each channel signal is obtained by calculating equation (12). At the moment of focus The amplitude values ​​of the signal wavefields of each channel are normalized and superimposed to obtain the global amplitude response of the four-channel signal for a single event. This enables the localization of sound sources for single-event signals.

[0075] Step 4: Repeat Step 3 until all valid acoustic emission event signals have been processed and the sound source localization results of all event signals are obtained.

[0076] Step 5: According to equations (13) and (14), the global amplitude responses of all event signals are superimposed point by point to form the cumulative focused response. Then, amplitude threshold segmentation and binarization are performed to obtain the set P of high amplitude points of the source distribution.

[0077] Step 6: Based on equations (15) and (16), use the DBSCAN density clustering algorithm to perform spatial clustering on the high-amplitude point set P, and select the cluster with the largest area. The convex hull B is calculated and used as the boundary profile of the corrosion region. The geometric center of the defect region is calculated according to equation (17). The least-squares ellipse fitting was performed on the convex hull B to obtain the major axis a and minor axis b of the defect region, thus realizing the quantitative characterization of corrosion defects.

[0078] The present invention has the following advantages: (1) By accumulating the focusing results of multiple acoustic emission event signals, the feature signals of the defect area are significantly enhanced, thereby improving the stability and accuracy of detection. (2) By utilizing the spatial clustering of acoustic emission sources within the defect area, the boundary of the corrosion defect area is extracted by combining cluster analysis and convex hull method, and the geometric parameters are obtained by ellipse fitting, thereby realizing the localization of corrosion defects and the characterization of contour parameters. Attached Figure Description

[0079] Figure 1 This is a schematic diagram of the acoustic emission detection system.

[0080] Figure 2 This is a flowchart of a corrosion defect detection method.

[0081] Figure 3 This is the original received signal for the four channels.

[0082] Figure 4 This is the acoustic emission determination process.

[0083] Figure 5 These are the statistical results for the characteristic frequency bands.

[0084] Figure 6 This is the result of the inversion of the S1 channel signal.

[0085] Figure 7 This represents the global amplitude response distribution.

[0086] Figure 8 This represents the cumulative focused response distribution.

[0087] Figure 9 The results show the boundary identification of the corrosion defect area. Detailed Implementation

[0088] The following example uses acoustic emission detection of corrosion defects in metal plates, combined with... Figures 3 to 9The implementation process of the proposed corrosion defect detection method based on the spatiotemporal distribution of acoustic emission signals is described in detail:

[0089] The sample used was an aluminum plate with dimensions of 500mm × 500mm × 3mm. During the experiment, the system sampling rate was set to 6MHz, and the single acquisition time was 60s. The corrosion experiment was carried out by adding a 5% NaOH solution (by mass) to a quartz tube fixed on the plate surface. The volume of each drop was 5 mL, and acoustic emission signal acquisition began immediately after the drop. The relative position of the sensor to the corrosion area is shown in the figure. Figure 1 As shown, the acquired multi-channel signals serve as input data for subsequent analysis.

[0090] The specific implementation steps are given below.

[0091] Step 1: Conduct three corrosion experiments, using four acoustic emission sensors to collect multi-channel acoustic emission signals. The raw acoustic emission signals collected from each channel in one experiment are shown below. Figure 3 As shown.

[0092] Step 2: Process the acquired acoustic emission raw signal from channel 1. Perform acoustic emission event determination and set the window length. 10ms, overlap rate between adjacent windows The value is 20%. Calculate the root mean square value of each signal segment. Standard deviation of background noise and normalized amplitude spectrum When the judgment condition is met, the signal segment is determined to be a valid acoustic emission event signal segment. The judgment process is as follows: Figure 4 As shown, using channel 1 as a reference, the effective acoustic emission event signal segments of the other three channels within the corresponding time period are extracted.

[0093] Step 3: For each valid event signal group Let the overlap rate of the two segments before and after each channel signal be denoted as . With a value of 0.4, the four-channel signals were combined pairwise to obtain six sets of cross-spectral phase difference results, among which the results for channels 1 and 2 are as follows: Figure 5 As shown in (a). Statistical analysis of the cross-spectral phase difference spectra yields the following results. Figure 5 As shown in (b), the characteristic frequency band of the signal is 5–130 kHz. Based on this, a bandpass filter is designed, and a rectangular window is applied to the filtered signal. Multi-scale decomposition and reconstruction are performed using an adjustable Q-factor wavelet transform (Q=2, J=8) to obtain the signal. .

[0094] Step 4: [Regarding...] Time reversal is performed, and the result of reversing the channel S1 signal is as follows: Figure 6 As shown. Figure 6a) The time reversal results at different reversal distances r constitute a time reversal wave field. . Figure 6 b) shows the curve of the peak value of the inversion result as a function of propagation distance. The first maximum value of its envelope corresponds to the focusing distance. The inversion result corresponding to this distance is as follows: Figure 6 As shown in c), the time corresponding to its peak value is the focusing moment. Focusing on the moment The spatial distribution of the lower amplitude is as follows Figure 6 As shown in d), the amplitude distribution of the four-channel signals at the focusing time... Normalized superposition yields the global amplitude response distribution. ,like Figure 7 As shown. Among them, The extreme value is located at (273, 238), which is within the actual corrosion area inside the red dashed line.

[0095] Step 5: Overlay all To form a cumulative focused response distribution ,like Figure 8 As shown. Set the normalized amplitude threshold. If the value is 0.9, threshold segmentation and binarization are performed to obtain the high-amplitude point set P.

[0096] Step 6: Perform cluster analysis on set P using the DBSCAN density clustering algorithm, and set a cluster radius threshold. The area is 10mm, so select the largest cluster. Calculate using convex hull operator The minimum convex hull B has a geometric envelope that is the boundary contour of the eroded region, such as... Figure 9 As shown in the diagram, the area within the red dashed line represents the actual corrosion region, and the green line represents the boundary contour of the corrosion region. Finally, least-squares ellipse fitting is performed on the convex hull B to obtain the set of shape and size parameters of the corrosion region. The results of the three experiments are shown in Table 1.

[0097] Table 1. Statistical Table of Recognition Results

[0098]

[0099] As shown in Table 1, the distance error between the results calculated by the corrosion defect detection method based on the spatiotemporal distribution of acoustic emission signals and the actual corrosion center location is within 4 mm, and the axial error is within 8 mm. This result proves the effectiveness and accuracy of the method.

[0100] The above are typical applications of the present invention, but the applications of the present invention are not limited thereto.

[0101] References

[0102] [1]Riccioli F, Alkhateeb S, Mol A, et al. Feasibility assessment ofnon-contact acoustic emission monitoring of corrosion-fatigue damage insubmerged steel structures[J]. Ocean Engineering, 2024, 312: 119296.

[0103] [2]Niu Y, Wang E, Li Z. A study on moment tensor inversion ofacoustic emission response on damaging localization of gas-bearing coal underload[J]. Scientific Reports, 2022, 12(1): 16360.

[0104] [3]Yu Z, Sun J, Xu C, et al. Locating of acoustic emission source forstiffened plates based on stepwise time-reversal processing with time-domainspectral finite element simulation[J]. Structural Health Monitoring, 2023, 22(2): 927-947.

[0105] [4]Zeman R, Kober J, Scalerandi M, et al. Hybrid experimental / computational approach to Time Reversal source localization in thin platesusing image source method[J]. Applied Acoustics, 2024, 218: 109873.

[0106] [5]Sun X, Fan S, Liu C. Multitype damage imaging in concrete modelingbased on time reversal technique[J]. Buildings, 2022, 12(4): 430.

Claims

1. A quantitative evaluation method for corrosion defects that integrates the spatiotemporal distribution of multiple acoustic emission sources, characterized in that, This is achieved through the following steps: Step 1: During the corrosion experiment, an array of acoustic emission sensors arranged at the four corners of the metal plate is used to acquire four channels of raw acoustic emission signal data. Step 2: Use the sliding window method to process the original signal. Perform acoustic emission event determination, and identify and filter candidate signal segments in the time domain. In the frequency domain, the normalized main peak amplitude is analyzed to determine whether it meets a set threshold, thereby identifying the effective acoustic emission event signal segment; based on the time index of the effective signal segment determined in the reference channel m... Extract the synchronization signal segments from the remaining channels within the same time range to form a multi-channel signal group for this event. ; Step 3: For each valid acoustic emission event signal group Execute in sequence: Step 31: Combine the M channels in pairs sequentially to form C(M,2) channel pairs; calculate the cross-power spectrum of the signal for each channel pair. Phase difference spectrum Extract its frequency features; Step 32: Perform statistical analysis on all results, count the number of occurrences of each frequency, and select the frequency range with high and concentrated frequency occurrences as the characteristic frequency band. Design a bandpass filter and filter the signal; Step 33, filter the signal Apply window function The windowed signal is obtained, and a wavelet transform with an adjustable Q factor is performed on it to obtain the reconstructed signal. ; Step 34: Based on the Mindlin plate theory, perform time reversal processing on the reconstructed signal and calculate the time-reversed wavefield of each channel signal. At the moment of focus The amplitude values ​​of the signal wavefields of each channel are normalized and superimposed to obtain the global amplitude response of the four-channel signal for a single event. To achieve single-event signal source localization; Step 4: Repeat Step 3 until all valid acoustic emission event signals have been processed and the sound source localization results of all event signals are obtained; Step 5: Superimpose the global amplitude responses of all event signals point by point to form a cumulative focused response. Then, amplitude threshold segmentation and binarization are performed to obtain the set P of high amplitude points of the source distribution; Step 6: Use the DBSCAN density clustering algorithm to perform spatial clustering on the set P of high amplitude points, and select the cluster with the largest area. And calculate its convex hull B, as the boundary contour of the eroded region; Calculate the geometric center of the defect region The least-squares ellipse fitting was performed on the convex hull B to obtain the major axis a and minor axis b of the defect region, thus realizing the quantitative characterization of corrosion defects.

2. The quantitative evaluation method for corrosion defects based on the spatiotemporal distribution of multiple acoustic emission sources according to claim 1, characterized in that, In step two, the raw signals acquired over a long period of time are processed. Event detection is performed. A sliding window method is used for traversal analysis; the window length is set to... The overlap rate between adjacent windows is sequentially process the i-th signal segment Perform energy and frequency domain determination. When the root mean square value of this segment... satisfy: ,in The standard deviation of background noise. If the coefficient is empirical, then this signal segment is considered a candidate signal segment. The calculation process for analyzing whether the normalized main peak amplitude in the frequency domain meets the set threshold is as follows: right Perform spectrum analysis, if in a specific frequency band Memory in normalized main peak amplitude satisfy: ,in If the amplitude threshold is reached, the segment is determined to be a valid acoustic emission event signal segment, and its time index is recorded. ; Based on the time index of the valid signal segment determined in channel m Extract the synchronization signal segments from the remaining channels within the same time range to form a multi-channel signal group for this event. .

3. The quantitative evaluation method for corrosion defects based on the spatiotemporal distribution of multiple acoustic emission sources according to claim 1, characterized in that, In step 31, the cross-power spectrum of the signal for each channel is calculated. Phase difference spectrum The process of extracting its frequency features is as follows: The multi-channel signal is sequentially divided into C(M,2) channel pairs, where M is the number of channels. Each channel pair is divided into two sub-segments, A and B, with an overlap ratio of [missing information]. Calculate their cross-power spectra respectively: where f is the frequency; (i=1,2) represents the spectrum of the corresponding signal segment. Indicates complex conjugation; In the frequency domain, the tiny delay between two signals This will cause its phase difference spectrum to exhibit a linear trend, that is: To quantitatively identify characteristic frequency bands, the first derivative of the cross-spectral phase difference spectrum within a certain frequency interval must satisfy: ,in For the set change threshold, then the frequency band It was determined to be a characteristic frequency band; Design a bandpass filter, for Frequency constraint processing is performed to obtain This is to preserve the propagation components and suppress noise interference.

4. The quantitative evaluation method for corrosion defects based on the spatiotemporal distribution of multiple acoustic emission sources according to claim 1, characterized in that, In step 33, an adjustable Q-factor wavelet transform is performed to obtain the reconstructed signal. The process is as follows: A multi-scale decomposition of the signal is performed using adjustable Q-factor wavelet transform to separate components of different frequency bands and enhance the main frequency characteristics; Its iterative decomposition process is expressed as: ,in, Let J be the high-frequency wavelet coefficients of the j-th layer. For the low-frequency scaling coefficients of the Jth layer, and These represent the high-pass and low-pass decomposition operators, respectively. For the selected multiscale coefficients and Reconstructing the time-domain signal using inverse adjustable Q-factor wavelet transform yields the denoised and enhanced signal: 。 5. The quantitative evaluation method for corrosion defects based on the spatiotemporal distribution of multiple acoustic emission sources according to claim 1, characterized in that, In step 34, according to the Mindlin plate theory, if there is an acoustic emission source at position O... Then the time-domain response of the signal after it propagates to the sensor location M is: ,in This represents the distance between point M and point O; This represents the nth discrete angular frequency component; This refers to the wavenumber of the flexural wave at the corresponding frequency. It is a zeroth-order Hankel function of the second kind; right After performing frequency domain complex conjugation, the position is obtained. The frequency domain expression of the time-reversed signal is: ,in The signal spectrum; Will Reverse launch to any position on the board Then, its frequency domain response is expressed as: After frequency domain compensation and inverse Fourier transform, the time reversal result at point P is obtained: After calculating the time reversal result at point P on the plate, the time reversal result is then applied to all spatial locations of the discrete grid on the plate surface. The above steps are executed sequentially to obtain a time-reversed wave field. ; among them, in the moment of focus At the actual sound source location, the wave field will exhibit local energy peaks; At the moment of focus The channel C signal is located at any position on the board. The amplitude of the time-reversal wave field is The sum of the focusing amplitudes of the multi-channel signals is expressed as: Where M is the number of sensor channels; through the above superposition operation, the global amplitude response of the event signal obtained at the time of time-domain focusing is obtained. The peak position is the source localization result of the multi-channel signal of the event.

6. The quantitative evaluation method for corrosion defects based on the spatiotemporal distribution of multiple acoustic emission sources according to claim 1, characterized in that, In step five, let the original signal be... The number of valid acoustic emission events is N, and the global amplitude response of the signal of the k-th event is... Then the cumulative focused response of all event signals is: By spatially accumulating the location results of all event signal sources, the focusing response of the defect area is significantly enhanced, and interference signals caused by random noise or isolated events are suppressed, thus forming a spatial probability distribution map of the acoustic emission sources. Convert to a discrete point set and perform spatial clustering analysis; set a threshold. right Binarization is performed to obtain the set of high-amplitude points: ,in This is the normalized amplitude.

7. The quantitative evaluation method for corrosion defects based on the spatiotemporal distribution of multiple acoustic emission sources according to claim 1, characterized in that, In step six, a density-based spatial clustering algorithm is used to perform spatial clustering analysis on set P: ,in, This is the cluster radius threshold. Let K be the k-th cluster, and K be the number of clusters identified. Select the cluster with the largest area Using convex hull operators Calculate its minimum convex hull: ,in, This represents the set of boundary points of the convex hull; the geometric envelope corresponding to the convex hull B is the edge contour of the eroded region. Based on the convex hull B, the geometric center of the region Calculated using the following formula: Where A is the area of ​​the polygon, and satisfies The closing condition; Least-squares elliptic fitting is performed on the convex hull B to obtain the major axis a and minor axis b of the defect region, where a ≥ b; parameter set As a quantitative evaluation index for corrosion defects.

8. The quantitative evaluation method for corrosion defects based on the spatiotemporal distribution of multiple acoustic emission sources according to claim 1, characterized in that, The characteristic frequency band of the effective acoustic emission event signal mentioned in step three refers to the frequency band of the signal group that is selected by performing cross-spectral phase difference analysis on the multi-channel signal of the effective acoustic emission event segment, counting the number of occurrences of each frequency, and selecting the frequency band range with higher occurrence frequency and concentrated distribution as the characteristic frequency band of the signal group.

9. The quantitative evaluation method for corrosion defects based on the spatiotemporal distribution of multiple acoustic emission sources according to claim 1, characterized in that, The cumulative focusing response mentioned in step five refers to the global amplitude response distribution of all effective acoustic emission events in the original signal, which is obtained by superimposing the spatial coordinates point by point.

Images