A damage locating method based on acoustic emission array

Abstract

The application relates to the technical field of structural damage positioning, and discloses a damage positioning method based on an acoustic emission array, which comprises the following steps: firstly, the acoustic emission signal is subjected to adaptive inverse filtering and delay compensation through a virtual time reversal mirror technology; secondly, a narrowband single-frequency signal is extracted based on Morlet wavelet transformation to inhibit frequency dispersion and multi-modal interference; and finally, a long-short-time kurtosis ratio enhancement operator is adopted to realize high-precision adaptive pickup of the arrival time in a strong noise environment. Through multi-stage cooperative processing, the application can effectively overcome multi-modal effects, frequency dispersion and noise interference, realize high-precision, high-robustness and high-real-time acoustic source positioning in a complex industrial scene with sparse sensor arrangement, non-uniform medium and variable noise, and significantly improve the detection capability for hidden defects and the system adaptability.

Description

Technical Field

[0001] This invention relates to the field of structural damage localization technology, and in particular to a damage localization method based on an acoustic emission array. Background Technology

[0002] Hydrogen is considered the most promising clean fuel because it does not produce pollutants such as carbon dioxide, sulfur dioxide, or soot during its use. Currently, the mainstream method of transporting hydrogen is using long-tube trailers to transport high-pressure gaseous hydrogen. In addition, there are pure hydrogen pipelines and liquid hydrogen tankers, among which pure hydrogen pipelines are expected to become the main route for large-scale, long-distance hydrogen transportation.

[0003] However, hydrogen molecules are much smaller than natural gas molecules, making them extremely prone to leakage. Hydrogen leaks can also lead to catastrophic consequences such as combustion and explosion. Therefore, flaw detection and leak location in hydrogen transportation are crucial for the development and security of hydrogen energy. Current transportation technologies face multiple bottlenecks: high-pressure hydrogen storage containers endure pressures of 10MPa-70MPa for extended periods, making it difficult to detect hidden defects such as weld fatigue cracks and valve seal aging in a timely manner; liquid hydrogen transportation requires maintaining a low temperature of -253℃, and there is a lack of real-time monitoring methods for issues such as damaged tank insulation and frozen pipelines; traditional detection technologies, such as infrared imaging, are susceptible to environmental temperature interference, and mass spectrometers have response times as long as 30 seconds, failing to meet the real-time requirements of dynamic transportation scenarios. Therefore, a more efficient and accurate detection technology is urgently needed to promptly detect and locate potential leaks or structural damage, minimizing energy loss and safety accidents, and dynamically ensuring the safety of hydrogen transportation in real time.

[0004] Existing hydrogen safety detection technologies are generally divided into direct detection methods and indirect detection methods. Direct detection methods locate leaks by placing moving detectors inside the pipe or static sensors at leak-prone points. These can be mainly divided into electrical hydrogen sensors and optical hydrogen sensors. Although there are many types of hydrogen sensors, each with its own advantages and disadvantages, their fixed locations and focus on only one parameter—hydrogen concentration—make it difficult to achieve real-time location in various complex environments. Furthermore, the uncertainty in calculating the leak location based on hydrogen inversion is relatively high. Indirect detection technologies, on the other hand, provide more accurate location and more timely monitoring. When gas leakage or escape occurs in hydrogen storage and transportation equipment, the leak area not only generates local acoustic signals but also causes significant changes in environmental parameters such as temperature and pressure in the surrounding environment. By monitoring the dynamic evolution of these parameters, the specific location and rate of the leak can be indirectly inferred.

[0005] Acoustic emission (AE) technology is an indirect detection technique that locates and detects leaks by capturing sound waves, using sensors for signal detection and delay estimation, and analyzing the leak signal. It is primarily based on acoustic emission (AE), a phenomenon where a material emits transient elastic waves due to the rapid release of energy locally. Common methods for AE-based leak location include cross-correlation location, time-difference location, pulse arrival identification kurtosis method, area location, time-reversal focusing, and multimodal analysis. Compared to traditional damage detection and location methods, AE technology offers high sensitivity in hydrogen leak detection. It effectively addresses the lower accuracy of traditional pipeline leak detection methods in complex environments or long-distance pipelines, and it allows for continuous real-time monitoring without disrupting normal pipeline operation.

[0006] Therefore, there are many research applications in hydrogen flaw detection. One paper combines acoustic emission testing with deep learning technology to construct a high-precision dataset of hydrogen storage tank failures. Using a carefully designed dataset, a multimodal 12-layer classification model combining waveform and spectral data achieved superior performance compared to single-modal models, overcoming the high computational cost and real-time monitoring limitations of traditional Type III hydrogen storage tank diagnosis due to its heavy reliance on simulation. Another paper focuses on acoustic signal source flaw detection of carbon fiber materials widely used in hydrogen storage tanks, analyzing the impact of reduced signal-to-noise ratio on positioning accuracy. Acoustic signals with different signal-to-noise ratios were measured three times on a carbon fiber pre-cast plate, and positioning was achieved through a sensor grid. Yet another paper develops a multipath model that reconstructs late-arriving signals from edge reflections based on the first arrival of the acoustic emission signal, integrating reflection-based and modal-based acoustic emission techniques. Experiments on aluminum plates verified an average positioning error of 2.8 cm with no blind spots; however, this model is only applicable to thin isotropic plates and does not consider the actual situation of multiple interface plates.

[0007] In summary, while acoustic emission technology has made significant progress in hydrogen damage detection and localization, it still faces the following challenges and limitations: sensor layout issues and large demand, insufficient accuracy due to signal processing and data analysis, environmental interference and noise management, and poor environmental adaptability.

[0008] Traditional Time Difference of Arrival (TDOA) acoustic emission localization, a common method, determines the location of a sound source based on the time difference between the arrival times of sound signals at different sensors. Depending on the environment, it is categorized into one-dimensional (linear) localization, two-dimensional (surface) localization, and three-dimensional localization, using geometric calculations to pinpoint the sound source's location. One study employed TDOA, combining complete set empirical mode decomposition with adaptive noise, Hilbert weighted cross-correlation, and neural networks to investigate a precise method for locating pipeline leaks based on acoustic emission. While this improved accuracy, the location error for leaks in the middle of the pipeline remained relatively large, exceeding 10% for some sources, with a maximum error reaching 12.78%. Traditional TDOA is susceptible to dispersion effects and signal attenuation, and the arrival time information can become chaotic when multiple acoustic emission events occur simultaneously—this is its most critical problem.

[0009] To address this, the Time-Reversal Focusing Method (TRFM) was developed. The propagation process of acoustic emission signals is either radiation or scattering. However, according to the reciprocity theorem of sound waves, after a sound source emits a signal, which is received by a sensor, the signal is reversed in time and emitted again. The signal received again is thus the inverse time history signal, which is the core principle of the TRFM. Another paper proposes a novel acoustic emission localization method and detection system, using the time-reversal focusing method and a fiber Bragg grating sensor network to complete multi-source acoustic emission localization. While this method achieves simultaneous localization of multi-source acoustic emissions and solves the problem of information confusion in multi-source events with the traditional TDOA method, it suffers from limited monitoring conditions, limited material availability, and the imposition of fixed-fixed boundary conditions, which severely restricts environmental conditions and makes it difficult to adapt to complex real-world environments. Another paper uses a time-reversal (TR) focusing imaging method for initial source localization to locate acoustic emission sources within anisotropic carbon fiber reinforced polymer plates. However, its simulation and experimental data do not completely match the theoretical curves, and systematic deviations still exist.

[0010] To reduce the number of sensors, some studies have proposed using an algorithm combining time-of-flight analysis and modal localization to assess source location. Compared with other traditional methods, this array can reduce the number of sensors required to detect large structures, proving the feasibility of array-based localization. El yamine Dris used a probabilistic method to determine the arrival time of acoustic emission (AE) waves in each sensor, considering uncertainty, through continuous wavelet transform. Experimental results show that the accuracy of the probabilistic algorithm is higher than that of the geometric algorithm. Building on this, another paper developed a novel acoustic emission source localization method based on Bayesian inversion. The probabilistic mapping intuitively displays the confidence level of the location prediction, which helps in determining the inspection area. Another paper further applies the Bayesian update algorithm based on artificial neural networks to the impact localization of composite laminates. However, the Bayesian algorithm and artificial neural networks require large datasets. To achieve ideal results, a sufficiently large training dataset must be created, and the model is likely to overfit, reducing its generalization ability and causing the estimated impact location to exhibit considerable randomness and significant deviation from the actual impact location.

[0011] The most common initial detection method for acoustic emission signals is a fixed threshold, where the point where the signal amplitude first exceeds a selected value is marked as the starting point of the signal. This leads to a series of problems, such as difficulty in threshold selection, noise interference, and signal-to-noise ratio dependence. The continuous wavelet transform (CWT) method used by the authors reduced the error by about 1.3 mm, but the computation time was extremely long, resulting in a lack of real-time performance. Professor Li mainly studied how to use near-entropy theory to achieve arrival time picking of microseismic data under low signal-to-noise ratio conditions. Based on the concept of near-entropy, he proposed a new picking method that can effectively distinguish between signal and noise and improve the signal-to-noise ratio. Another paper uses a decision tree-based ensemble learning method to estimate the arrival time of the first wave of acoustic emission. Another paper uses Dynamically Optimized Time Window (DOTW) to determine a small, dynamically accurate pickup window based on signal characteristics to avoid impulse noise interference, reducing the false pickup rate from 57.14% to 4.76%. However, this method is not ideal for handling other types of noise.

[0012] In summary, existing technologies employing time-reversal focusing combined with dispersion compensation are susceptible to noise and external interference during transmission, leading to altered signal characteristics. Traditional methods often involve denoising the received acoustic signal, resulting in poor positioning stability. Furthermore, existing technologies struggle to adapt to complex and unpredictable situations in real-world applications. Their limitations hinder the acquisition of effective and accurate positioning results, preventing their practical application. Additionally, some positioning methods and models are only applicable to specific situations, lacking universality. For example, deep learning-based array positioning schemes are typically only suitable for acoustic sensor arrays at specific locations; if the sensor positions change in practice, the model needs to be retrained. Moreover, some specific applications require high real-time monitoring; acoustic emission positioning demands rapid response, which most existing technologies may struggle to meet. Summary of the Invention

[0013] Therefore, the technical problem to be solved by the present invention is to overcome the defects of the existing acoustic emission positioning method, which is susceptible to noise interference, has poor environmental adaptability, chaotic positioning of multi-source events, strong dependence on sensor layout, insufficient universality and difficulty in meeting real-time requirements.

[0014] To address the aforementioned technical problems, this invention provides a damage localization method based on an acoustic emission array, comprising: Real-time acquisition of raw time-domain signals passively and synchronously captured by each acoustic emission sensor in a sparse acoustic emission sensor array set in a preset monitoring area; Acoustic emission mode analysis was performed on each original time-domain signal to obtain the basic symmetric mode and antisymmetric mode, as well as the group velocity corresponding to each mode; Using the Molay wavelet as the mother wavelet, a continuous wavelet transform is performed on the original time-domain signal to obtain the wavelet transform coefficients, and the continuous wavelet transform modulus is calculated to construct the wavelet coefficient scale spectrum. Based on the ridge line in the wavelet coefficient scale spectrum, the wave arrival time is characterized, and the single-frequency coefficients are extracted in the characteristic frequency scale. The basic symmetric single-frequency waveform and the antisymmetric single-frequency waveform are separated and constructed, and the mode type with higher energy proportion is obtained as the target single-frequency waveform. Perform Fourier transform on the target single-frequency waveform to obtain the frequency domain signal, and then perform virtual time reversal processing and time domain reconstruction to obtain the reconstructed signals of each acoustic emission sensor. The reconstructed signals of all acoustic emission sensors are coherently superimposed to obtain the virtual time-reversal mirror focusing signal of each acoustic emission sensor; Based on a preset step size, the short-term window and the long-term window are used to slide on the virtual time reversal mirror focusing signal to obtain the short-term kurtosis and long-term kurtosis in the window after each slide. Based on the ratio of short-term kurtosis to long-term kurtosis, a feature enhancement curve is constructed, and the time of the local maximum point where the forward slope is extremely large and the backward slope change rate is the largest is obtained, which is used as the precise arrival time of the virtual time reversal mirror focusing signal. Based on the precise arrival time of the virtual time-reversal mirror focusing signal, the virtual time-reversal mirror focusing signals of various acoustic emission sensors are fused using the delay summation imaging method to generate a damage localization image, and the position corresponding to the amplitude peak is obtained as the location of the damage source.

[0015] Compared with the prior art, the above-described technical solution of the present invention has the following advantages: The damage localization method based on acoustic emission arrays described in this invention integrates virtual time-reversal mirror focusing technology, Morlet wavelet-based single-frequency signal extraction technology, and arrival time precision picking technology based on long-short time kurtosis ratio enhancement operator to form a multi-stage collaborative processing system. This effectively overcomes the shortcomings of existing acoustic emission array localization technologies, such as poor localization stability, weak noise resistance, poor environmental adaptability, and insufficient real-time performance. The virtual time-reversal mirror technology enables adaptive inverse filtering and delay compensation, enhancing the reconstruction effect of the damage signal in the medium. Combined with modal analysis and single-frequency extraction, it effectively suppresses waveform dispersion and multimodal interference, thereby significantly improving the spatial resolution accuracy and localization stability of the damage source. Furthermore, the virtual time-reversal mirror technology is implemented solely through virtual computation based on the reciprocity principle, making it applicable to arbitrary array geometry and non-uniform media. It does not require prior knowledge of the Green's function or propagation path information, avoiding the strong dependence on sensor layout and the problem of retraining models inherent in traditional methods. This significantly improves its applicability in complex monitoring environments such as hydrogen energy storage and transportation equipment.

[0016] Meanwhile, the long-short time kurtosis ratio method combines local sensitivity and global stability, enabling high-precision and adaptive acquisition of acoustic emission signal arrival time in noisy environments. It effectively suppresses the interference of echoes on arrival time extraction and enhances the robustness of the positioning system. Attached Figure Description

[0017] To make the content of this invention easier to understand, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings, wherein: Figure 1 This is a flowchart of the damage localization method based on acoustic emission array of the present invention; Figure 2 This is a schematic diagram of the virtual time reversal mirror technology. Figure 3 This is a flowchart for calculating the ratio of short-term kurtosis to long-term kurtosis; Figure 4 This is a comparison chart of the success rate curves for different location methods; Figure 5 This is a comparison chart of the average positioning error under different noise levels when using a virtual time-reversal mirror and an enhanced virtual time-reversal mirror for positioning. Figure 6 This is a schematic diagram of the acoustic emission transducer distribution in the positioning experiment; Figure 7 This is a flowchart of the damage localization process; Figure 8 This is an example diagram of acoustic emission signal and arrival time extraction; Figure 9 It is a damage localization image; Figure 10 This is a comparison chart of the average positioning errors of different methods indoors and outdoors. Detailed Implementation

[0018] The present invention will be further described below with reference to the accompanying drawings and specific embodiments, so that those skilled in the art can better understand and implement the present invention. However, the embodiments described are not intended to limit the present invention.

[0019] This invention aims to address several key issues in existing acoustic emission (AE) array localization technologies, such as poor localization accuracy due to multimodal and dispersion effects, inaccurate arrival time acquisition in complex noise environments, and limitations of traditional methods in terms of real-time performance and sensor layout requirements. To address these issues, this invention proposes a damage localization method based on an acoustic emission array. Utilizing enhanced virtual time-reversal mirror technology, it first achieves adaptive inverse filtering and delay compensation through time-reversal focusing theory, improving the reconstruction effect and localization accuracy of the damage signal in the medium. Then, it combines modal analysis and single-frequency extraction of the acoustic emission signal to suppress waveform dispersion and multi-type noise interference, enhancing the robustness of localization. Finally, it introduces the long-short time kurtosis ratio (S / L-Kurt) method to achieve high-precision, adaptive acquisition of arrival time in strong noise environments, suppressing echo interference with arrival time extraction and enhancing localization stability. Through this multi-stage collaborative processing, this invention can achieve high-precision, high-real-time sound source localization in complex industrial scenarios (such as damage monitoring of hydrogen energy storage and transportation equipment), significantly improving the detection capability for hidden defects and system adaptability, providing reliable technical support for structural health monitoring and safety early warning.

[0020] Reference Figure 1 The flowchart of the damage localization method based on acoustic emission array of the present invention is shown below, and the specific steps are shown in S101 to S109.

[0021] S101: Real-time acquisition of the raw time-domain signals passively and synchronously captured by each acoustic emission sensor in the sparse acoustic emission sensor array set in the preset monitoring area, including: S101-1: Arrange a sparse acoustic emission sensor array consisting of multiple acoustic emission sensors in a preset monitoring area; S101-2: Acoustic emission signals generated by damage sources within a preset monitoring area are passively captured in real time using a sparse acoustic emission sensor array. Simultaneously, the raw time-domain electrical signals acquired by each acoustic emission sensor are recorded, yielding the raw time-domain signals corresponding to each acoustic emission sensor, expressed as: ; in, Indicates the first Each acoustic emission sensor in The original time-domain signal at time [time]. express The initial sound source signal at time [time]. This represents the convolution operator. Indicates the location from the initial sound source. To the Location of each acoustic emission sensor The impulse response function, Indicates transmission delay time; , This represents the total number of sensors in a sparse acoustic emission sensor array.

[0022] Reference Figure 2 The diagram shown illustrates the principle of virtual time-reversal mirror technology, where the acoustic emission transducer is the acoustic emission sensor of this application. Specifically, virtual time-reversal mirror technology is a technique for localizing sound sources using adaptive inverse filtering and focusing; it re-emits the signal at the receiver along the opposite direction of propagation, essentially a process of sampling a closed time-reversal cavity using a finite number of sensors in a sparse array. Assuming the location in a two-dimensional region... The initial sound source signal generated when damage occurs at the location is Then located The The raw time-domain signals collected by each sensor are When the received signal After time reversal, it becomes If this inverted signal If retransmitted from the receiver, the sound will be transmitted from the initial source by the first... The signal caused by the retransmission of a sensor is written as: Considering the reciprocity of the propagation medium, the impulse response functions of opposite propagation paths are the same. Therefore, the time-reversed signal is similar to the original sound source signal, differing only by a product of real functions, i.e. In this way, the time delay in the received signal is adaptively removed. Therefore, when the signal waveform is virtually reversed in the time domain, all independent signals from the sparse array will reach their maximum values ​​at the same time; then, all time-reversed mirror signals are emitted simultaneously, causing the original sound sources to coherently superimpose and form an enhanced reconstructed sound source at the initial real sound source location after spatiotemporal transformation.

[0023] S102: Perform acoustic emission mode analysis on each raw time-domain signal to obtain the basic symmetric and antisymmetric modes, as well as the group velocity corresponding to each mode, including: S102-1: Based on the material thickness, basic longitudinal wave, shear wave velocity, and Lamb wave angular frequency of the monitored object in the preset monitoring area, construct the Rayleigh-Lamb equation; S102-2: Based on the Rayleigh-Lamb equation, obtain the basic symmetric and antisymmetric modes of acoustic emission signal propagation in the monitored object, and the group velocities of the basic symmetric and antisymmetric modes at different angular frequencies; Here, group velocity is the derivative of angular frequency with respect to wave number.

[0024] Specifically, in shell or layered structures, acoustic emission waveforms typically propagate in the form of Lamb waves, exhibiting complex behaviors such as dispersion and multimodal patterns. Therefore, the accuracy of damage localization is severely affected by the aforementioned dispersion attenuation and multimodal characteristics, leading to spurious focus peaks or artifacts in the images. Generally, group velocity and phase velocity in dispersive media exhibit significant frequency dependence, which can be calculated using the following Rayleigh-Lamb equation, expressed as: ; in, It is half the thickness of the shell or sheet metal. Wave number; coefficient and Defined as: , ; It is angular frequency. and These are the basic longitudinal wave velocity and shear wave velocity of the monitored material, respectively.

[0025] Among the existing modes, two modes are easier to detect and separate: the fundamental symmetric mode (S0) and the antisymmetric mode (A0). By solving the Rayleigh-Lamb equations, the phase velocities of each mode can be obtained. Group speed , respectively represented as , And apply it to subsequent positioning.

[0026] S103: Using the Moray wavelet as the mother wavelet, perform continuous wavelet transform on the original time-domain signal to obtain the wavelet transform coefficients, and calculate the wavelet continuous transform modulus to construct the wavelet coefficient scaling spectrum, including: S103-1: Using the Moray wavelet as the mother wavelet, perform continuous wavelet transform on the original time-domain signal, and decompose the original time-domain signal by wavelet scaling and translation to obtain the wavelet transform coefficients; S103-2: Take the modulus values ​​of the wavelet transform coefficients and calculate the continuous wavelet transform modulus; S103-3: Construct a two-dimensional wavelet coefficient scale spectrum with scale as the vertical axis, time as the horizontal axis, and continuous wavelet transform modulus as the energy value.

[0027] Specifically, to suppress dispersion and multimode uncertainty, narrowband burst waveforms of the acoustic emission signal should be extracted before the time reversal operation. This typically requires using time-frequency representation to analyze and extract narrowband or single-frequency components. Continuous wavelet transform, as the most commonly used linear time-frequency method, can display the scale spectrum of specific frequencies in short-time waveforms. The continuous wavelet transform coefficients are defined as follows: ; in, The scaling parameter determines the wavelet. Analysis frequency or support width; The translation parameter represents the position of the wavelet in the time domain. Therefore, the continuous wavelet transform can decompose the signal through wavelet scaling and translation, while providing an ideal trade-off between time and frequency resolution.

[0028] In particular, the Morlet wavelet can separate amplitude and phase, making it suitable for describing the transient characteristics of acoustic emission signals. Therefore, this embodiment uses it as the mother wavelet. The Morlet wavelet expression is as follows: ; The wavelet is essentially a Gaussian window pulse burst signal with a center frequency. and bandwidth .

[0029] S104: Based on the ridge lines in the wavelet coefficient scale spectrum, the wave arrival time is characterized, and single-frequency coefficients are extracted in the characteristic frequency scale. Basic symmetric and antisymmetric single-frequency waveforms are constructed separately, and the mode type with the higher energy proportion is obtained as the target single-frequency waveform, including: S104-1: Retrieve the maximum value of the continuous wavelet transform modulus in the wavelet coefficient scaling spectrum at each time point, connect the maximum value points corresponding to each time point in sequence to obtain the ridge line of the wavelet coefficient scaling spectrum, and use the starting point of the ridge line on the time axis to characterize the wave arrival time. S104-2: Select the frequency range in which only the basic symmetric mode and antisymmetric mode exist as the characteristic frequency range, and map it to obtain the corresponding characteristic frequency scale; S104-3: Extract wavelet transform coefficients as single-frequency coefficients from the characteristic frequency scale. Based on the group velocity difference between the basic symmetric mode and the antisymmetric mode, separate the coefficient components corresponding to the basic symmetric mode and the antisymmetric mode from the single-frequency coefficients. S104-4: After performing continuous wavelet inverse transform and time-domain reconstruction on the two coefficient components respectively, the basic symmetrical single-frequency waveform and the antisymmetric single-frequency waveform are obtained; S104-5: Obtain the single-frequency waveform with a higher energy proportion from the basic symmetrical single-frequency waveform and the antisymmetric single-frequency waveform, and use it as the target single-frequency waveform.

[0030] Specifically, for dispersion analysis, consider two frequencies that are very close in phase. and The sound source, the distance of propagation The resulting signal can be represented as , and Frequency and The corresponding wave number, Represents the imaginary unit. Assume... , , , The above signal can be rewritten as: ;when , Substituting the above equation into the definition of continuous wavelet transform coefficients, the continuous wavelet transform modulus can be calculated as follows: ; Therefore, the frequency components of a signal at any given time can be represented by the instantaneous energy concentration at various scales. The first arrival time of a waveform can be represented by the wavelet coefficient scaling spectrum. and Ridge characterization at a location; ridge amplitude in the plane The position on corresponds to the frequency Time of arrival of the wave Therefore, single-frequency coefficients can be extracted at a specific frequency scale using Morlet wavelet transform to obtain the waveform of the basic mode, which can then be used for subsequent sound source localization.

[0031] S105: Perform Fourier transform on the target single-frequency waveform to obtain the frequency domain signal, then perform virtual time reversal processing and time domain reconstruction to obtain the reconstructed signals of each acoustic emission sensor, including: S105-1: Perform Fourier transform on the target single-frequency waveform to convert the time-domain target single-frequency waveform into a frequency-domain signal; S105-2: Based on the principle of acoustic reciprocity, frequency domain virtual time reversal processing is performed on the frequency domain signal to adaptively remove the time delay in the propagation process of the acoustic emission signal from the damage source to the sensor, thus obtaining the frequency domain inverted signal; S105-3: Perform an inverse Fourier transform on the frequency domain inverted signal to convert it back to the time domain signal, complete the time domain reconstruction, and obtain the reconstructed signal.

[0032] S106: Coherently superimpose the reconstructed signals of all acoustic emission sensors to obtain the virtual time-reversal mirror focusing signal of each acoustic emission sensor.

[0033] Specifically, unlike active ultrasound imaging, the signals from acoustic emission arrays are passively acquired; although the sound sources occur simultaneously, the exact time of occurrence of the damage source in the received signal is unknown. Furthermore, in passive scenarios, there is no receiver or transmitter at the sound source, and the Green's function is typically unknown. Therefore, the directly received acoustic emission signals cannot be directly used to form a localization image, thus introducing a virtual time-reversal mirror in the frequency domain; the virtual time-reversal mirror focusing signal can be described as: ; in, and These represent the initial sound source signals respectively. and the Raw time-domain signal from each sensor The Fourier transform of the time reversal process can be conveniently represented as a virtual signal processing technique that depends solely on the received signal, without requiring prior knowledge of the Green's function of the transmitting structure or the spectral characteristics of the initial sound source.

[0034] S107: Based on a preset step size, slide short-time windows and long-time windows across the virtual time-reversal mirror focusing signal to obtain the short-term and long-term kurtosis within each window after each slide, including: S107-1: Using the virtual time-reversal mirror focusing signal as the input signal, and setting a short-time window and a long-time window on the input signal based on a preset step size; S107-2: Control the short-time window to slide along the time axis of the input signal with a preset step size, calculate the sample mean of the signal within the short-time window corresponding to each time index, and then obtain the short-term kurtosis based on the sample mean, expressed as: ; S107-3: Control the long-term window to slide along the time axis of the input signal with a preset step size, calculate the sample mean of the signal within the long-term window corresponding to each time index, and then obtain the long-term kurtosis based on the sample mean, expressed as: ; in, Indicates the first Short-term kurtosis in a short-term window after the second slide. Indicates the length of the short-time window. Indicates the first time window The signal amplitude at each sampling point This represents the sample mean of a short-time window; Indicates the first Long-term kurtosis in a long-term window after the second sliding. Indicates the length of the long-term window. Indicates the first time window The signal amplitude at each sampling point This represents the sample mean over a long-term window. , For short-term windows, For long-term windows; For the first After the second slide Sample variance within the window The square of, , express The sample mean of the window; express The first in the window The signal amplitude at each sampling point.

[0035] In this embodiment of the invention, the virtual time-reversal mirror technique and the single-frequency signal extraction based on acoustic emission mode analysis can effectively compensate for the multimodal and dispersion effects of acoustic emission signals during propagation. However, in the subsequent delay-summing beamforming or signal focusing process, the interference of random noise and the superposition of unexpected echoes still lead to a decrease in the spatial resolution of damage localization. The fundamental reason is the insufficient accuracy of picking up the effective arrival time in the original signal. To solve this problem, this invention proposes a method for accurate signal arrival time picking based on the long-short-time kurtosis ratio (S / L-Kurt) enhancement operator. Kurtosis is a measure of the tail weight of the input data distribution and is very effective in signal recognition. This method constructs a characteristic curve by calculating the kurtosis ratio of the signal within the short-time window and the long-time window, and locates the signal arrival time on this curve. It is particularly suitable for processing non-Gaussian signals such as acoustic emission signals in low signal-to-noise ratio environments.

[0036] Specifically, assuming Set a short time window for the input signal. and long window The input continuous time-domain signal is divided into several sliding windows, with the window sliding step size typically set to 1 to ensure time resolution. Then, the nth... Short-term kurtosis in the short-time window after the second slide With the Long-term kurtosis in a long-term window after the second slide The short-time window is used to capture the local transient features of the signal, while the long-time window is used to characterize the background statistical properties of the signal over a longer time scale.

[0037] S108: Based on the ratio of short-term kurtosis to long-term kurtosis, a feature enhancement curve is constructed, and the time of the local maximum point where the forward slope is extremely large and the rate of change of the backward slope is extremely large is obtained. This time is used as the precise arrival time of the virtual time-reversal mirror focusing signal, including: S108-1: Calculate the ratio of short-term kurtosis to long-term kurtosis under each time index to obtain the short-term kurtosis ratio corresponding to each time index, and construct a feature enhancement curve with time as the horizontal axis and the short-term kurtosis ratio as the vertical axis; S108-2: Retrieve all local maxima on the feature enhancement curve, and calculate the average forward slope within the preset window before each local maximum and the average backward slope within the preset window after each local maximum. S108-3: Select the local maximum point where the forward slope is extremely large and the change in slope between the front and back is most significant, and determine the time corresponding to this point as the precise arrival time of the virtual time-reversal mirror focusing signal.

[0038] Specifically, before the signal arrives, the background noise typically approximates a Gaussian distribution, with a low and relatively stable kurtosis value. When the acoustic emission signal arrives, its non-Gaussian, pulse-like characteristics cause a sharp increase in kurtosis within a short time window, while the kurtosis changes relatively slowly in a long time window due to the inclusion of a large amount of foreground and background. Therefore, by calculating the ratio of the two (S / L-Kurt), a feature enhancement curve can be constructed. Thus, the final arrival time accurate picking based on the long-short time kurtosis ratio (S / L-Kurt) enhancement operator can be expressed as: ; in, It is a small constant to avoid division by zero errors.

[0039] The ratio of short-term kurtosis to long-term kurtosis is essentially a normalization process that can suppress the influence of overall fluctuations in background noise levels on the detection threshold, while highlighting the local kurtosis abrupt changes caused by signal arrival in the form of a ratio spike.

[0040] Specifically, all local maxima are found on the feature enhancement curve, and the average forward slope within a certain window before each maximum point and the average backward slope within a certain window after each maximum point are calculated. The actual signal arrival time usually corresponds to a local maximum point where the forward slope is extremely large and the change in slope between the two points is most significant. The forward slope can be expressed as: ; in, , This represents the number of windows traced backward.

[0041] Reference Figure 3 The diagram shows the flowchart for calculating the ratio of short-term kurtosis to long-term kurtosis. This embodiment significantly suppresses Gaussian noise interference by using the ratio of short-term and long-term dual-window kurtosis, providing a reliable time reference for subsequent damage localization, beamforming, and other processing.

[0042] S109: Based on the precise arrival time of the virtual time-reversal mirror focusing signal, the virtual time-reversal mirror focusing signals from various acoustic emission sensors are fused using a delay-summing imaging method to generate a damage localization image. The location corresponding to the amplitude peak value is obtained as the damage source location, including: S109-1: Divide the preset monitoring area into multiple nodes according to the preset grid, and mark the physical coordinates of each node and each acoustic emission sensor; S109-2: Combine the precise arrival time with the group velocity of the corresponding mode to calculate the propagation time of the acoustic emission waveform from each node in the monitoring area to each acoustic emission sensor; S109-3: Based on the propagation time, the virtual time-reversal mirror focusing signals of each acoustic emission sensor are weighted and summed using the delay summation imaging method to obtain the image intensity values ​​of each node in the monitoring area; S109-4: Using the physical coordinates of each node as the position reference and the image intensity value as the amplitude, generate a two-dimensional or three-dimensional damage location image. The physical coordinates of the node corresponding to the amplitude peak of the damage location image are the location of the damage source.

[0043] Specifically, the virtual time-reversal mirror focusing technology is implemented solely through virtual computation based on the reciprocity principle. This process is applicable to automatic delay compensation in any array geometry and non-uniform media, requiring no prior information about the medium or propagation path. After obtaining the virtual time-reversal acoustic signals, these signals can be directly used to reconstruct the acoustic field through spatiotemporal transformation. This technology employs the time-delay summation method as the spatiotemporal transformation process. The time-delay summation method is widely used in ultrasonic imaging and Lamb wave beamforming applications to visualize the reconstructed acoustic field; it calculates pixel values ​​in the image by weighted summing of signals from different sensors. When combined with virtual time-reversal mirror focusing signals from multiple sensors, the time-delay summation method can be used to form a comprehensive damage detection image for localization. Mathematically, at any location... The image intensity at the time delay summation point can be calculated using the acoustic emission signal to obtain the image intensity value of each node within the monitoring area, expressed as: ; in, The coordinates within the monitoring area are: The image intensity value at the node, , This represents the total number of sensors in a sparse acoustic emission sensor array; Indicates the first The focusing signal of the virtual time-reversal mirror of the acoustic emission sensor is represented as: , , , ; Indicates the first The raw time-domain signal collected by the acoustic emission sensor The Fourier transform of the Fourier transform is expressed as: ; express The complex conjugate of . Represented as: ; Indicates the initial sound source signal Fourier transform; express The complex conjugate; Indicates the location of the sound source To the Sensor locations impulse response function Fourier transform, Indicates the location of the sound source To the Sensor locations impulse response function Fourier transform, express The complex conjugate; This indicates the acoustic emission waveform from any position to the [missing information]. The propagation time of an acoustic emission sensor, , Indicates the first The location of each acoustic emission sensor The group velocity represents the mode type corresponding to the target single-frequency waveform.

[0044] This invention provides embodiments for achieving high-precision and robust damage source localization in complex media and noisy environments. It effectively integrates the advantages of inverse filtering in time-reversal focusing, the anti-dispersion capability of modal analysis and single-frequency extraction, and the local sensitivity and global stability of the long-short time kurtosis ratio, thereby significantly improving the environmental adaptability and signal interpretation accuracy of the localization system. In multimodal time-reversal focusing, the Virtual Time Reversal Mirror (VTRM) technology is used to achieve adaptive delay compensation and coherent signal superposition in the frequency domain, enhancing the accuracy of sound source reconstruction. In modal analysis and single-frequency extraction, based on the Rayleigh-Lanmu equation and Moray wavelet transform, the basic modal waveforms in the acoustic emission signal are extracted and separated, suppressing dispersion and multimodal interference. After calculating the long-short time kurtosis ratio, the signal kurtosis characteristic curve is calculated through a sliding window, and combined with local extremum slope analysis, achieving high stability and adaptive picking of arrival time in strong noise environments. Finally, the outputs are fused using the Delayed Summation (DAS) imaging method to generate a high-contrast damage localization image and obtain the damage localization result. In actual industrial scenarios with sparse sensor arrangement, non-uniform medium, and variable noise, this invention can significantly improve the accuracy and robustness of localization.

[0045] To fully demonstrate the significant effects of this embodiment in practical applications, this embodiment verifies its great potential in acoustic emission localization of simulated damage to hydrogen energy storage and transportation equipment by employing the proposed multimodal time-reversal focusing and pulse arrival identification kurtosis method. Specifically, this embodiment uses steel plate samples and pencil lead fracture (PLB) as simulated sound sources to collect multiple sets of acoustic emission signals under different sensor array layouts and noise environments. A comparative experiment is conducted using the method proposed in this invention and existing representative methods. (Refer to...) Figure 4 The image shows a comparison of the success rate curves for different localization methods; refer to... Figure 5 The figure shows a comparison of the average positioning errors of virtual time-reversal mirror and enhanced virtual time-reversal mirror under different noise levels. Based on the positioning curves and statistical analysis of experimental data, the positioning error of this method is less than 1.51 cm in 95% of the test scenarios. The positioning sensitivity is on average 1.8 cm lower than the time-difference positioning method based on triangulation, and the standard deviation or variance is lower than that of the original virtual time-reversal mirror, significantly outperforming other diagnostic methods. The results show that this method exhibits excellent positioning performance in all test groups, effectively overcoming multimodal dispersion and random noise interference, and significantly improving the spatial resolution accuracy and image contrast of the damage source. Furthermore, through dynamic testing under different noise levels, the method of this invention demonstrates higher stability and noise resistance than the original virtual time-reversal mirror.

[0046] To verify the method of the present invention, a series of repeatable tests were performed on a steel plate sample with dimensions of 100cm × 100cm × 0.3cm. The mass density, Young's modulus, and Poisson's ratio of the steel plate were set to 7930 kg / m³. The values ​​were 2.11 × 10¹¹ Pa and 0.22. In addition to the samples, the experimental system included a data acquisition system (Softland DS2-8B), a sparse acoustic emission array consisting of four piezoelectric sensors (Softland RS-54A), and four 40 dB preamplifiers. PZTs were fixed to the surface of the plate, and coupling agent was filled between the sensors and the monitored surface. The sensors were 8 mm in diameter and 15.5 mm thick, with a receiving frequency range of 1 MHz. The sampling rate of the acquisition system was set to 2.5 MHz. A standard Hsu-Nielsen source (2H type pencil lead fracture source (PLB)) was used to simulate stable and repeatable damage sources in the plate. The entire experiment was conducted at a constant room temperature.

[0047] The entire monitoring area is divided into 400 pre-set nodes using a 5cm x 5cm grid, such as... Figure 6 The diagram shown is a schematic of the acoustic emission transducer distribution for the positioning experiment, with the lower left corner set as the origin of the plate; the acoustic emission transducer in this embodiment is an acoustic emission sensor.

[0048] Table 1. Location numbering of sensor array layout configuration in different test groups

[0049] Reference Figure 7 The diagram shows the damage localization flowchart, which can be divided into three parts: signal acquisition and modal processing, optimization of signal processing accuracy, and visualization of the damage location obtained through signal reconstruction. In this embodiment, the enhanced virtual time-reversal mirror method provided by this invention is used for damage localization; the enhanced virtual time-reversal mirror in this embodiment combines Morloet wavelet transform and long-short time kurtosis ratio to achieve higher accuracy damage localization.

[0050] Specifically, enhancing virtual time-reversal mirror technology includes: synchronously capturing signals from different locations using an acoustic emission array; and using Moray wavelet transform at the characteristic frequencies of the signal. Extracting single-frequency components This data is then applied to subsequent positioning. Since only S0 and A0 modes exist in the frequency range below 500kHz in the figure below, single-frequency coefficients are extracted at this frequency scale using Morlet wavelet transform to further obtain the waveforms of the basic modes. According to the sampling theorem, the frequency range analyzed in the waveform wavelet plot is set to 1MHz, and the total wavelet scale is 1000 to obtain a resolution of 1kHz in the frequency domain. Furthermore, the bandwidth of the Morlet wavelet is... Choose a sufficiently large one. To ensure that each center frequency is The frequency window does not overlap with adjacent windows. Therefore, to meet the resolution requirements of the aforementioned time-frequency analysis, the dimensionless bandwidth and center frequency parameters of the Morlet wavelet transform are respectively... and Based on the experimental material properties, only A0 and S0 modes exist at sound frequencies below 500kHz. This embodiment selects a 150kHz signal component for subsequent analysis; at 150kHz, the waveforms of the S0 and A0 modes were extracted using continuous wavelet transform and are shown in the figure. At this frequency, the Lamb wave of the S0 mode propagates at a group velocity of 5050.3m / s, much faster than the 2745.4m / s of the A0 mode. Therefore, the first arrival time of the S0 mode is earlier than that of the A0 mode. (Refer to...) Figure 8 The image shown is an example of acoustic emission signal and time of arrival extraction. However, considering that most of the energy propagates in the Lamb wave in the A0 mode rather than the S0 mode, a 150kHz A0 waveform was used in the subsequent reconstruction of the final VTRM image. Subsequently, the extracted single-frequency signal was reversed using a virtual time-reversal mirror focusing technique; and based on the long-short time kurtosis ratio enhancement operator, accurate time of arrival was achieved; based on the virtual re-emission reversed signal, a delay summation method was used to reconstruct the intensity values ​​of the localization image, revealing the source location using the peak amplitude in the image; this was combined with VTRM signals from multiple sensors to form a comprehensive damage detection image for localization. (Refer to...) Figure 9 The image shown is a damage localization image. In this way, the enhanced virtual time-reversal mirror method can suppress multimodal effects, eliminate random noise interference and unwanted echo superposition. Therefore, it can improve image contrast between the focal and non-focal areas and provide more accurate localization results.

[0051] Under the same environment, this example was compared with data obtained by representative time-difference localization methods, modal acoustic emission localization methods, and virtual time-reversal mirror techniques; (Refer to...) Figure 10The figure shows a comparison of the average positioning errors of different methods indoors and outdoors. The data shows that the enhanced virtual time-reversal mirror method has higher accuracy than the time-difference positioning method and modal acoustic emission, effectively eliminating artifacts and achieving higher image contrast and more stable performance than the original virtual time-reversal mirror technology. In low signal-to-noise ratio environments, the average positioning error of this invention is smaller than that of the original virtual time-reversal mirror technology, demonstrating the superiority of this invention.

[0052] This invention integrates virtual time-reversal mirror focusing technology, Morlet wavelet-based single-frequency signal extraction technology, and arrival time precision picking technology based on long-short-time kurtosis ratio (LSR) enhancement operator to form a multi-stage collaborative processing system. This effectively overcomes the shortcomings of existing acoustic emission array positioning technologies, such as poor positioning stability, weak noise resistance, poor environmental adaptability, and insufficient real-time performance. The virtual time-reversal mirror technology enables adaptive inverse filtering and delay compensation, enhancing the reconstruction effect of damaged signals in the medium. Combined with modal analysis and single-frequency extraction, it effectively suppresses waveform dispersion and multimodal interference, thereby significantly improving the spatial resolution accuracy and positioning stability of the damage source. The LSR method combines local sensitivity with global stability, enabling high-precision, adaptive picking of acoustic emission signal arrival time in noisy environments. It effectively suppresses echo interference in arrival time extraction, enhancing the robustness of the positioning system. The virtual time-reversal mirror technology is implemented through virtual computation based solely on the reciprocity principle. It is applicable to arbitrary array geometries and non-uniform media, requiring no prior knowledge of the Green's function or propagation path information. This avoids the strong dependence on sensor layout and the problems of model retraining inherent in traditional methods, significantly improving its applicability in complex monitoring environments such as hydrogen storage and transportation equipment. Furthermore, this method avoids excessively long computation times through multi-stage collaborative processing and delayed summation imaging, enabling rapid response and meeting the real-time requirements of structural health monitoring and safety early warning. In summary, this application can achieve high-precision, high-robustness, strong noise resistance, and high real-time acoustic emission array positioning in complex industrial scenarios, providing reliable technical support for early detection and safety warning of hidden defects.

[0053] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0054] This application is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of this application. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart... Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0055] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0056] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0057] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A damage localization method based on acoustic emission array, characterized in that, include: Real-time acquisition of raw time-domain signals passively and synchronously captured by each acoustic emission sensor in a sparse acoustic emission sensor array set in a preset monitoring area; Acoustic emission mode analysis was performed on each original time-domain signal to obtain the basic symmetric mode and antisymmetric mode, as well as the group velocity corresponding to each mode; Using the Molay wavelet as the mother wavelet, a continuous wavelet transform is performed on the original time-domain signal to obtain the wavelet transform coefficients, and the continuous wavelet transform modulus is calculated to construct the wavelet coefficient scale spectrum. Based on the ridge line in the wavelet coefficient scale spectrum, the wave arrival time is characterized, and the single-frequency coefficients are extracted in the characteristic frequency scale. The basic symmetric single-frequency waveform and the antisymmetric single-frequency waveform are separated and constructed, and the mode type with higher energy proportion is obtained as the target single-frequency waveform. Perform Fourier transform on the target single-frequency waveform to obtain the frequency domain signal, and then perform virtual time reversal processing and time domain reconstruction to obtain the reconstructed signals of each acoustic emission sensor. The reconstructed signals of all acoustic emission sensors are coherently superimposed to obtain the virtual time-reversal mirror focusing signal of each acoustic emission sensor; Based on a preset step size, the short-term window and the long-term window are used to slide on the virtual time reversal mirror focusing signal to obtain the short-term kurtosis and long-term kurtosis in the window after each slide. Based on the ratio of short-term kurtosis to long-term kurtosis, a feature enhancement curve is constructed, and the time of the local maximum point where the forward slope is extremely large and the backward slope change rate is the largest is obtained, which is used as the precise arrival time of the virtual time reversal mirror focusing signal. Based on the precise arrival time of the virtual time-reversal mirror focusing signal, the virtual time-reversal mirror focusing signals of various acoustic emission sensors are fused using the delay summation imaging method to generate a damage localization image, and the position corresponding to the amplitude peak is obtained as the location of the damage source.

2. The damage localization method based on acoustic emission array according to claim 1, characterized in that, Real-time acquisition of raw time-domain signals passively and synchronously captured by each acoustic emission sensor in a sparse acoustic emission sensor array set in a preset monitoring area, including: A sparse acoustic emission sensor array consisting of multiple acoustic emission sensors is arranged in a preset monitoring area; Acoustic emission signals generated by damage sources within a preset monitoring area are passively captured in real time using a sparse acoustic emission sensor array. Simultaneously, the raw time-domain electrical signals acquired by each acoustic emission sensor are recorded, resulting in the raw time-domain signals corresponding to each acoustic emission sensor, expressed as follows: ； in, Indicates the first Each acoustic emission sensor in The original time-domain signal at time [time]. express The initial sound source signal at time [time]. This represents the convolution operator. Indicates the location from the initial sound source. To the Location of each acoustic emission sensor The impulse response function, Indicates transmission delay time; , This represents the total number of sensors in a sparse acoustic emission sensor array.

3. The damage localization method based on acoustic emission array according to claim 1, characterized in that, Acoustic emission mode analysis was performed on each raw time-domain signal to obtain the fundamental symmetric and antisymmetric modes, as well as the group velocity corresponding to each mode, including: Based on the material thickness, basic longitudinal wave, shear wave velocity, and Lamb wave angular frequency of the monitored object in the preset monitoring area, the Rayleigh-Lamb equation is constructed. Based on the Rayleigh-Lamb equation, the basic symmetric and antisymmetric modes of acoustic emission signal propagation in the monitored object are obtained, as well as the group velocities of the basic symmetric and antisymmetric modes at different angular frequencies; Here, group velocity is the derivative of angular frequency with respect to wave number.

4. The damage localization method based on acoustic emission array according to claim 1, characterized in that, Using the Moray wavelet as the mother wavelet, a continuous wavelet transform is performed on the original time-domain signal to obtain the wavelet transform coefficients. The continuous wavelet transform modulus is then calculated to construct the wavelet coefficient scaling spectrum, including: Using the Molay wavelet as the mother wavelet, a continuous wavelet transform is performed on the original time-domain signal. The original time-domain signal is decomposed by wavelet scaling and translation to obtain the wavelet transform coefficients. The modulus of the continuous wavelet transform is calculated by taking the modulus values ​​of the wavelet transform coefficients. A two-dimensional wavelet coefficient scale spectrum is constructed with scale as the vertical axis, time as the horizontal axis, and continuous wavelet transform modulus as the energy value.

5. The damage localization method based on acoustic emission array according to claim 1, characterized in that, Based on the ridge lines in the wavelet coefficient scaling spectrum to characterize wave arrival time, and extracting single-frequency coefficients at the characteristic frequency scale, a basic symmetric single-frequency waveform and an antisymmetric single-frequency waveform are constructed. The mode type with the higher energy proportion is then selected as the target single-frequency waveform, including: The maximum value of the continuous wavelet transform modulus in the wavelet coefficient scaling spectrum is retrieved point by point in time. The maximum value points corresponding to each time point are connected in sequence to obtain the ridge line of the wavelet coefficient scaling spectrum. The wave arrival time is represented by the starting point of the ridge line on the time axis. The frequency range containing only basic symmetric and antisymmetric modes is selected as the characteristic frequency range, and the corresponding characteristic frequency scale is obtained by mapping. Wavelet transform coefficients are extracted as single-frequency coefficients at the characteristic frequency scale. Based on the group velocity difference between the basic symmetric mode and the antisymmetric mode, the coefficient components corresponding to the basic symmetric mode and the antisymmetric mode are separated from the single-frequency coefficients. After performing continuous wavelet inverse transform and time-domain reconstruction on the two coefficient components respectively, we obtain the basic symmetrical single-frequency waveform and the antisymmetric single-frequency waveform. The single-frequency waveform with the higher energy proportion between the basic symmetrical single-frequency waveform and the antisymmetric single-frequency waveform is used as the target single-frequency waveform.

6. The damage localization method based on acoustic emission array according to claim 1, characterized in that... The target single-frequency waveform is subjected to Fourier transform to obtain the frequency domain signal. Then, virtual time reversal processing and time domain reconstruction are performed to obtain the reconstructed signals of each acoustic emission sensor, including: Perform a Fourier transform on the target single-frequency waveform to convert the time-domain target single-frequency waveform into a frequency-domain signal; Based on the principle of acoustic reciprocity, frequency domain virtual time reversal processing is performed on the frequency domain signal to adaptively remove the time delay in the propagation process of the acoustic emission signal from the damage source to the sensor, thus obtaining the frequency domain inverted signal. Perform an inverse Fourier transform on the frequency-domain inverted signal to convert it back to a time-domain signal, thus completing the time-domain reconstruction and obtaining the reconstructed signal.

7. The damage localization method based on acoustic emission array according to claim 1, characterized in that, Based on a preset step size, short-term and long-term windows are used to slide across the virtual time-reversal mirror focusing signal, respectively, to obtain the short-term and long-term kurtosis within each window after each slide, including: Using the virtual time-reversal mirror focusing signal as the input signal, and setting short-time and long-time windows on the input signal based on a preset step size; The short-time window slides along the time axis of the input signal with a preset step size. The sample mean is calculated for the signal within the short-time window corresponding to each time index. Then, the short-term kurtosis is obtained based on the sample mean, expressed as: ; The long-term window slides along the time axis of the input signal with a preset step size. The sample mean is calculated for the signal within the long-term window corresponding to each time index. The long-term kurtosis is then obtained based on the sample mean, expressed as: ; in, Indicates the first Short-term kurtosis in a short-term window after the second slide. Indicates the length of the short-time window. Indicates the first time window The signal amplitude at each sampling point This represents the sample mean of a short-term window; Indicates the first Long-term kurtosis in a long-term window after the second sliding. Indicates the length of the long-term window. Indicates the first time window The signal amplitude at each sampling point This represents the sample mean over a long-term window. , For short-term windows, For long-term windows; For the first After the second slide Sample variance within the window The square of, , express The sample mean of the window; express The first in the window The signal amplitude at each sampling point.

8. The damage localization method based on acoustic emission array according to claim 1, characterized in that, Based on the ratio of short-term kurtosis to long-term kurtosis, a feature enhancement curve is constructed, and the time of the local maximum point where the forward slope is extremely large and the rate of change of the backward slope is extremely large is obtained. This time is used as the precise arrival time of the virtual time-reversal mirror focusing signal, including: Calculate the ratio of short-term kurtosis to long-term kurtosis for each time index to obtain the short-term kurtosis ratio for each time index, and construct a feature enhancement curve with time as the horizontal axis and the short-term kurtosis ratio as the vertical axis. Retrieve all local maxima on the feature enhancement curve, and calculate the average forward slope within the preset window before each local maximum and the average backward slope within the preset window after each local maximum. The local maxima point with the largest forward slope and the most significant change in slope before and after is selected, and the time corresponding to this point is determined as the precise arrival time of the virtual time-reversal mirror focusing signal.

9. The damage localization method based on acoustic emission array according to claim 1, characterized in that, Based on the precise arrival time of the virtual time-reversal mirror focusing signal, a delay-summing imaging method is used to fuse the virtual time-reversal mirror focusing signals from various acoustic emission sensors to generate a damage localization image, including: The preset monitoring area is divided into multiple nodes according to a preset grid, and the physical coordinates of each node and each acoustic emission sensor are calibrated. By combining the precise arrival time with the group velocity of the corresponding mode, the propagation time of the acoustic emission waveform from each node in the monitoring area to each acoustic emission sensor is calculated; Based on the propagation time, the virtual time-reversal mirror focusing signals of each acoustic emission sensor are weighted and summed using the delayed summation imaging method to obtain the image intensity values ​​of each node in the monitoring area. Using the physical coordinates of each node as the location reference and the image intensity value as the amplitude, a two-dimensional or three-dimensional damage localization image is generated. The physical coordinates of the node corresponding to the amplitude peak of the damage localization image are the location of the damage source.

10. The damage localization method based on acoustic emission array according to claim 9, characterized in that, The image intensity values ​​of each node within the monitoring area are represented as follows: ； in, The coordinates within the monitoring area are: The image intensity value at the node, , This represents the total number of sensors in a sparse acoustic emission sensor array; Indicates the first The focusing signal of the virtual time-reversal mirror of the acoustic emission sensor is represented as: , , , ; Indicates the first The raw time-domain signal collected by the acoustic emission sensor The Fourier transform of the Fourier transform is expressed as: ; express The complex conjugate of . Represented as: ; Indicates the initial sound source signal Fourier transform; express The complex conjugate; Indicates the location of the sound source To the Sensor locations impulse response function Fourier transform, Indicates the location of the sound source To the Sensor locations impulse response function Fourier transform, express The complex conjugate; This indicates the acoustic emission waveform from any position to the [missing information]. The propagation time of an acoustic emission sensor, , Indicates the first The location of each acoustic emission sensor The group velocity represents the mode type corresponding to the target single-frequency waveform.

Images