Method for locating near-field acoustic emission sources based on orthogonal matching pursuit under sparse array

Abstract

The present application relates to the field of structural damage online monitoring, in particular to a near-field acoustic emission source positioning method based on orthogonal matching pursuit under a sparse array, comprising the following steps: step 1, acoustic emission signal acquisition and preprocessing; step 2, dimension reduction process of the near-field acoustic emission source; step 3, near-field acoustic emission source positioning based on orthogonal matching pursuit; and step 4, near-field acoustic emission source positioning. The present application provides a crack tip positioning method without judging the sensor wave arrival time, which can effectively improve the positioning accuracy.

Description

Technical Field

[0001] This invention relates to the field of online monitoring of structural damage, and more specifically to a near-field acoustic emission source localization method based on orthogonal matched pursuit in a sparse array. Background Technology

[0002] Complex equipment structures are often subjected to complex alternating loads during operation, making them prone to fatigue cracks and other damage. Therefore, to prevent dangerous accidents during maintenance downtime, conducting research on high-precision remote online monitoring and location of fatigue cracks is a key method to ensure structural safety, develop efficient and reasonable maintenance cycles, and improve economic efficiency.

[0003] Acoustic emission (AE) is an effective method for remote online structural monitoring, exhibiting good detection capabilities for small-scale cracks. Current methods often directly perceive and identify crack states by analyzing the characteristic parameters of AE signals under different crack states. However, the characteristic parameters of AE signals vary significantly between different materials, and even within the same material, there is considerable dispersion in these parameters. Therefore, this method is less effective for crack detection and identification.

[0004] Acoustic emission (AE) is a continuous monitoring method. If the acoustic emission source generated by crack tip propagation can be located, the degree of fatigue crack damage can still be determined with long-term monitoring data. For crack tip location, the most commonly used method is based on time-of-arrival (TOA) differences, including triangulation and time-of-arrival mapping. However, in real-world environments, sensor signals are easily affected by noise, and acoustic emission signals also exhibit dispersion, making it difficult to determine accurate TOA values. Machine learning methods demonstrate excellent location performance when dealing with complex structures and anisotropic materials, but their model training requires large amounts of data, resulting in significant training costs. Therefore, for research on crack tip location using acoustic emission technology, there is an urgent need for an acoustic emission location method that can avoid the interference of inaccurate TOA values ​​while improving computational efficiency. Summary of the Invention

[0005] To address the problems existing in the prior art, the purpose of this invention is to provide a near-field acoustic emission source localization method based on orthogonal matching pursuit under a sparse sensor array that can effectively expand the aperture of the sensor array, improve the monitoring range of near-field acoustic emission signals, and directly realize two-dimensional joint estimation of the source angle and distance.

[0006] To achieve the above objectives, the technical solution adopted by the present invention is: a near-field acoustic emission source localization method based on orthogonal matched pursuit in sparse arrays, comprising the following steps:

[0007] Step 1: Based on the geometric characteristics of the monitored object and guided by the optimal number of virtual array elements obtained by differential sparse sensor array, complete the installation and arrangement of sparse sensor arrays for key monitoring parts of the structure, collect the sudden acoustic emission signals under the sparse sensor array, and decompose the sudden acoustic emission signals to obtain the acoustic emission narrowband signals.

[0008] Step 2: Based on the aforementioned acoustic emission narrowband signal, construct a near-field source model oriented towards the sparse sensor array. Select sensors at specific locations in the sparse sensor array and construct a special fourth-order cumulant matrix containing only the angle term information in the near-field acoustic emission source. Based on the near-field source model, establish a near-field acoustic emission source dimensionality reduction model based on the fourth-order cumulant matrix. Obtain virtual extended array elements through sparse array difference.

[0009] Step 3: Based on the steering vector of the virtual extended array element, establish the overcomplete steering vector matrix of the virtual array, convert the signal model of the virtual array into a sparse representation, sparsely reconstruct and solve the angle term information of the near-field acoustic emission source under the sparse array, and establish an estimation model of the angle term information of the near-field acoustic emission source.

[0010] Step 4: Construct the corresponding covariance matrix based on the acoustic emission signal received by the sensor, establish a MUSIC algorithm model for the near-field acoustic emission source, input the angle term information of the near-field acoustic emission source into the MUSIC algorithm model to obtain a near-field MUSIC model that only includes distance term information, perform spectral peak search on the near-field MUSIC model to obtain the positioning information of the near-field acoustic emission source.

[0011] The above-described near-field acoustic emission source localization method based on orthogonal matched pursuit under sparse arrays uses a sparse sensor array that is a compressed symmetric nested array structure. The sparse sensor array structure includes three levels of subarrays: a first subarray, a second subarray, and a third subarray. The array element at the center of the first subarray is the reference point of the sparse sensor array structure. The array element of the first subarray is located at a multiple of 2M²-1, and the array element spacing is d = λ / 4, where λ is the signal wavelength. The number of array elements in the second and third subarrays is defined as N², and the array element spacing is (2M²-1)d.

[0012] The above-mentioned near-field acoustic emission source localization method based on orthogonal matched pursuit under sparse arrays involves setting the cutoff frequency according to the dispersion curve of the guided wave under specific materials and structures to obtain an acoustic emission signal containing only A0 and S0 modes. The Shannon wavelet transform method is used to obtain the acoustic emission narrowband signal at the dominant frequency f.

[0013] The near-field acoustic emission source localization method based on orthogonal matched pursuit under the above sparse array, wherein the near-field source model is:

[0014]

[0015] Where, x i (t) represents the original time-domain signal received by the i-th array element, K is the number of signal sources, ω is the angular frequency, d is the element spacing, i is the element number, and s k (t) represents the k-th narrowband AE source signal, and p represents the coordinate value. i Let n be the coordinates of the i-th element. i (t) represents the noise signal received by the i-th array element, λ is the signal wavelength, and θ k and r k Then it represents the angle and distance of the k-th source from the phase origin.

[0016] The above-described near-field acoustic emission source localization method based on orthogonal matched pursuit using a sparse array defines the fourth-order cumulant received by the sparse sensor array as follows: in, Let m,n,f,q∈[-Q,Q] be the fourth-order cumulant of the k-th acoustic emission source.

[0017] Let n = -m, q = -f, and construct a special fourth-order cumulant:

[0018] Construct the special fourth-order cumulant matrix C4:

[0019] The expressions for each symbol in the fourth-order cumulant matrix C4 are as follows: B(θ)=[b(θ1),…,b(θ K )],

[0020] The above-described near-field acoustic emission source localization method based on orthogonal matched pursuit under sparse arrays further includes the following steps in constructing the special fourth-order cumulant matrix C4: defining c4(k) = c4(m,n), where k = mn are the virtual element coordinates of the difference array, and vectorizing C4 as follows: Where y is the received signal vector of the virtual array after differential sparse array; The source signal vector of the virtual array; This is the guiding vector matrix of the virtual array.

[0021] The above-described near-field acoustic emission source localization method based on orthogonal matched pursuit under sparse arrays, wherein the sparse array is differentially obtained to obtain virtual extended array elements S V Represented as: S V ={Pv-Y , Pv -Y+1 , ..., 0, ..., Pv Y-1 , Pv Y}, where Pv i Let be the coordinates of the i-th virtual array element;

[0022] definition D>>K is an overcomplete dictionary from the perspective of the information source, where Given the possible incident angles, and D as the number of incident angles, construct the overcomplete steering vector matrix A of the virtual array. Θ :

[0023]

[0024] The aforementioned near-field acoustic emission source localization method based on orthogonal matched pursuit under sparse arrays transforms the solution of the signal model into an l1-norm convex optimization problem: min||p Θ ||1 in, For the single-shot signal received by the virtual array element after redundancy removal, p Θ It is a sparse vector with K non-zero elements.

[0025] The above-described near-field acoustic emission source localization method based on orthogonal matching pursuit under sparse arrays uses the orthogonal matching pursuit algorithm to sparsely reconstruct and solve the angle term information of the near-field acoustic emission sources under sparse arrays. The optimal number corresponding to the selected atom, A Θ,j A represents Θ Update the index set Γ to the j-th column of (θ). t =Γ t-1 ∪{γ t and residual The angle information of the near-field acoustic emission source is estimated.

[0026] The above-described near-field acoustic emission source localization method based on orthogonal matched pursuit under sparse arrays, wherein the covariance matrix R XX for: The covariance matrix R is obtained through eigenvalue decomposition. XX The signal is decomposed into a signal subspace and a noise subspace. Based on the MUSIC algorithm, the source is estimated by using the orthogonality between the noise subspace and the signal vector corresponding to the MUSIC algorithm, thus establishing the spatial spectrum function of the near-field acoustic emission source.

[0027] The aforementioned near-field acoustic emission source localization method based on orthogonal matched pursuit under sparse arrays inputs the estimated angle information of the near-field acoustic emission source into the spatial spectrum function to obtain a special spatial spectrum function containing only the distance term:

[0028] Where θ k Let k be the estimated acoustic emission source angle, k = 1, 2, ..., K.

[0029] The above-described near-field acoustic emission source localization method based on orthogonal matched pursuit under sparse arrays performs spectral peak search on the special spatial spectral function containing only the distance term, realizes the estimation and pairing of distance information under each angle term, and completes the localization of the near-field acoustic emission source.

[0030] The beneficial effects of this invention's near-field acoustic emission source localization method based on orthogonal matched pursuit in a sparse array are: this invention provides a crack tip localization method that eliminates the need to determine the sensor's wave arrival time, effectively improving localization accuracy. By employing the Shannon wavelet transform method to decompose the signal, a narrowband acoustic emission signal with only A0 and S0 modes at a certain dominant frequency f is obtained.

[0031] Meanwhile, in the implementation of this method, the sensor array adopts a sparse structure, which effectively expands the array aperture and enables joint estimation of the angle and distance information of the acoustic emission source, thereby improving the positioning and monitoring accuracy. Subsequently, a dimensionality reduction method for near-field acoustic emission sources based on fourth-order cumulants is proposed. By constructing a special fourth-order cumulant, the angle and distance information are separated, realizing the dimensionality reduction process of the acoustic emission source. This achieves the separation of the angle and distance terms in the near-field source estimation, avoiding the extremely high computational complexity caused by two-dimensional variable search.

[0032] In order to make full use of all the virtual array elements obtained by the differential of the sparse sensor array, a near-field acoustic emission source angle information estimation model based on orthogonal matching pursuit was established. By vectorizing the special fourth-order cumulative matrix, redundant elements in the virtual sensor were removed, which improved the performance and computational efficiency of acoustic emission source angle estimation.

[0033] Finally, given the angle information of the source to be estimated, the near-field MUSIC algorithm is used to effectively estimate the distance term, thus completing the effective estimation of the near-field acoustic emission source. A comparison with the 2D-MUSIC and NLA-MUSIC algorithms shows that the proposed method exhibits the best performance across multiple variables, including signal-to-noise ratio, snapshot number, grid density, and computation time. Attached Figure Description

[0034] Figure 1 This is a flowchart of the method of the present invention;

[0035] Figure 2 This is the dispersion curve of the 7075-T6 aluminum plate used in the embodiments of the present invention;

[0036] Figure 3 This is the waveform and spectrum of the acoustic emission narrowband signal in this invention;

[0037] Figure 4 This is the acoustic emission source propagation model under sparse sensor array in this invention;

[0038] Figure 5 This is a sparse representation of the acoustic emission signal of the present invention in the search space;

[0039] Figure 6 This invention compares physical and virtual arrays with different sparse array structures.

[0040] Figure 7 This is a performance comparison of different sparse arrays of the present invention;

[0041] Figure 8 This is a performance comparison of different estimation methods of the present invention;

[0042] Figure 9 This is a comparison of the computational efficiency of different acoustic emission positioning methods of the present invention;

[0043] Figure 10 This is the result of a localization experiment using a single acoustic emission source according to the present invention. Detailed Implementation

[0044] To enable those skilled in the art to better understand the technical solution of the present invention, the technical solution of the present invention will be described below in conjunction with specific embodiments and accompanying drawings.

[0045] Example 1

[0046] Figure 1 This is a flowchart of a near-field acoustic emission source localization method based on orthogonal matched pursuit under sparse arrays. The overall method is divided into four parts, which will be described in detail below:

[0047] Part 1, Step 1: Acquisition and preprocessing of acoustic emission signals.

[0048] In this embodiment, the acoustic emission signal acquisition device mainly includes a piezoelectric ceramic sensor, a signal amplifier, a multi-channel signal acquisition instrument, and acquisition software. Acoustic emission signals are acquired by setting a signal threshold triggering mechanism. To achieve accurate positioning of the acoustic emission source in a sparse array, appropriate signal preprocessing is necessary based on the propagation characteristics of the acoustic emission signal. As a broadband complex guided wave signal generated due to internal deformation or damage of materials, the original acoustic emission signal exhibits multimode and dispersion phenomena. Its propagation process can be described by the Rayleigh-Lamb wave equation:

[0049]

[0050]

[0051] in, ω is the angular frequency, k = ω / c p Let be the wave number, cp be the phase velocity, and cL and cT be the transverse and longitudinal wave velocities, respectively.

[0052] Assuming the crack originates in the 7075-T6 aluminum plate, the corresponding dispersion curve is shown in Figure 2. Figure 2 In the middle (a), the phase velocity is represented in the dispersion curve of the 7075-T6 aluminum plate; Figure 2 In Figure (b), the group velocity in the dispersion curve of the 7075-T6 aluminum plate is shown. It can be seen that when the cutoff frequency-thickness product is 1860 kHz × mm, only the zero-order symmetric wave S0 and the zero-order antisymmetric wave A0 exist in the plate. For burst-type AE signals generated by metal fracture and other reasons, their signal energy is usually concentrated between 40 and 500 kHz. Therefore, to obtain a narrowband AE signal at a certain frequency, Shannon wavelet transform is used. The Shannon continuous complex wavelet mother function is defined as:

[0053] Where fb is the frequency band of the wavelet, and fc is the center frequency of the wavelet.

[0054] The Fourier transform of the wavelet mother function yields the Fourier transform of the Shannon continuous complex wavelet function:

[0055] Where, ω c =2πf c ω b =2πf b , where a is the scaling factor and b is the delay factor, which is usually 0.

[0056] Therefore, by setting reasonable center frequency fc, bandwidth fb, and scaling factor a, the zero-order mode in the AE signal can be extracted, while also reducing high-frequency noise interference and improving the accuracy of AE source localization. The acoustic emission signal waveforms and spectra before and after wavelet transform are shown below. Figure 3 As shown, Figure 3 The first packet of the original signal in (a1) Figure 3 In the middle (a2), the spectrum of the original narrowband acoustic emission signal is represented. Figure 3 (b1) represents the first packet after wavelet transform of the narrowband acoustic emission signal. Figure 3 (b2) represents the signal spectrum after wavelet transform of the acoustic emission narrowband signal.

[0057] Part 2, Step 2: Dimensionality reduction process of near-field acoustic emission sources.

[0058] AE source propagation model under sparse array as follows Figure 4 As shown in the figure, the sensor positions of this symmetrical sparse array can be represented as S = {p} -Q ,p -Q+1 ,…,0,…,p Q-1 ,p Q}, where p i Let p be the coordinate value of the i-th element. Therefore, the distance of the i-th element from the phase origin is p. i d, the aperture D of the array can be expressed as 2p Q d. Therefore, the signal acquired by each element in the array after Shannon wavelet transform is: i = -Q, -Q+1, ..., Q-1, Q. Where, x i (t) represents the original time-domain signal received by the i-th array element, K is the number of signal sources, ω is the angular frequency, d is the element spacing, i is the element number, and s k (t) represents the k-th narrowband AE source signal, p i Let n be the coordinates of the i-th element. i (t) represents the noise signal received by the i-th array element, λ is the signal wavelength, and θ k and r k Then it represents the angle and distance of the k-th source from the phase origin.

[0059] To achieve dimensionality reduction processing of near-field acoustic emission sources, it is necessary to solve for the fourth-order cumulant of the signal acquired by the sparse array:

[0060]

[0061] in, Let m,n,f,q∈[-Q,Q] be the fourth-order cumulant of the k-th acoustic emission source.

[0062] To separate angle and distance information while adhering to the definition of a sparse array differential process, let n = -m and q = -f. Then, the expression for the fourth-order cumulant can be transformed into:

[0063] As can be seen from the above formula, the exponent term It only contains the angle term information of the source to be estimated, and p m -p p It also has a sparse array difference form. Therefore, based on this formula, a fourth-order cumulant matrix C4 can be constructed:

[0064] in, B(θ)=[b(θ1),…,b(θK )],

[0065] Define c4(k) = c4(m,n), where k = mn are the coordinates of the virtual elements of the difference array. It can be seen that this contains many redundant elements, and the difference array of the sparse array is not a continuous array of elements. To remove the redundant virtual elements, matrix C4 needs to be vectorized, resulting in: Where y is the received signal vector of the virtual array after differential sparse array; The source signal vector of the virtual array; This is the guiding vector matrix of the virtual array.

[0066] Part 3, Step 3: Near-field acoustic emission source localization based on orthogonal matching pursuit.

[0067] For the AE signal sensed by a one-dimensional sparse array, its energy distribution in the search domain space can be considered sparsely distributed, such as... Figure 5 As shown, ideally, the position vector is non-zero only at the grid points where the AE event occurs, while the position vector is zero at most other grid points. Therefore, by combining the fourth-order cumulant matrix C4 obtained above after dimensionality reduction of the near-field acoustic emission source, the traditional two-dimensional spectral peak search for near-field sources is transformed into a more efficient problem of finding the optimal sparse solution under an overcomplete atom library.

[0068] To construct the overcomplete dictionary matrix of the sparse array, the virtual array elements obtained from the sparse array difference can be represented as S. V ={Pv -Y , Pv -Y+1 , ..., 0, ..., Pv Y-1 , Pv Y}, where Pv i Let be the coordinates of the i-th virtual element. Simultaneously define... D>>K is an overcomplete dictionary from the perspective of the information source, where Let A be the possible angle of incidence, and D be the number of angles of incidence. Therefore, the overcomplete steering vector matrix A of the virtual array can be constructed. Θ :

[0069]

[0070] Therefore, the virtual array model can be transformed into a sparse representation: in, The single-shot signal received by the virtual array element after redundancy removal; p Θ It is a sparse vector with K non-zero elements.

[0071] Solving the above equation is usually expressed as an l1 norm convex optimization problem: min||p Θ ||1

[0072] To solve this problem, the orthogonal matching pursuit algorithm is used to achieve sparse reconstruction of the AE signal under a sparse array. The specific steps of the algorithm are as follows:

[0073] (1) Based on the special fourth-order cumulant matrix C4 under the constructed sparse array, vectorization, redundancy removal and other processes are then performed to obtain the virtual array S after sparse array difference. V ;

[0074] (2) Based on the constructed overcomplete dictionary Θ, construct the overcomplete guided vector matrix A of the virtual array. Θ ;

[0075] (3) Define the initial information for the OMP algorithm: residual Index set Reconstruct column set

[0076] (4) Set the initial iteration number t = 1, and calculate... The optimal number corresponding to the selected atom, A Θ,j A represents Θ The j-th column of (θ);

[0077] (5) Update the index set Γ t =Γ t-1 ∪{γ t}, and rebuild A at the same time t =[A t-1 A Θ,j ];

[0078] (6) Update residuals At the same time, t = t + 1;

[0079] (7) If t≤K, then the loop will continue; while when t>K, then Γ t The angle in the middle is the source estimation angle.

[0080] Part 4, Step 4: Near-field acoustic emission source localization.

[0081] The above process yields an estimate of the angle term for the acoustic emission source, while the distance term is estimated using the near-field MUSIC algorithm. First, the covariance matrix R of the acoustic emission signal under sparse array conditions is established. XX :

[0082]

[0083] R is obtained through eigenvalue decompositionXX The signal is decomposed into a signal subspace and a noise subspace, and then the spatial spectrum function of the near-field acoustic emission source is established based on the MUSIC algorithm:

[0084]

[0085] The obtained near-field acoustic emission source angle information is substituted into the spatial spectrum function of the near-field acoustic emission source to obtain a special spatial spectrum function containing only the range term, thereby enabling the estimation of the range term:

[0086]

[0087] Where θ k Let k be the estimated acoustic emission source angle, k = 1, 2, ..., K.

[0088] In the algorithm verification, three sparse arrays were selected to verify the performance of different sparse array structures, such as... Figure 6 As shown, Figure 6 In the middle (a), a comparison is made between a physical array and a virtual array of coprime symmetric arrays (CSA). Figure 6 (b) shows a comparison between a physical array and a virtual array of Symmetric Nested Arrays (SNA). Figure 6 (c) represents a comparison between physical and virtual arrays of Compressed Symmetric Nested Arrays (CSNA), including Coprime Symmetric Arrays (CSA), Symmetric Nested Arrays (SNA), and Compressed Symmetric Nested Arrays (CSNA). It can be seen that different array structures have different physical element apertures and virtual element numbers. Subsequently, the performance of the three arrays is verified by considering signal-to-noise ratio, number of signal sources, and number of snapshots, as follows: Figure 7 As shown, Figure 7 In the middle (a), the signal-to-noise ratio of different sparse arrays is compared. Figure 7 In the middle (b), the number of information sources for different sparse arrays is compared. Figure 7 (c) represents the comparison of snapshot counts for different sparse arrays. The results show that when the signal-to-noise ratio (SNR) is low, the DOA and range estimation performance of all three arrays are poor, while their performance improves significantly with increasing SNR. Among the three arrays, the CSNA array performs best due to having the most virtual elements. Since angle estimation directly affects range determination, the CSNA array also shows the best range estimation performance. Similar results were observed in the subsequent comparison of the number of information sources and snapshot counts for the three sparse arrays.

[0089] The proposed NLA-OMP algorithm is compared with near-field 2D-MUSIC and the NLA-MUSIC algorithm based on fourth-order cumulants in terms of signal-to-noise ratio, number of snapshots, grid search density, and computation time. Figure 8 He Ru Figure 9 As shown, Figure 8 In the middle (a), the signal-to-noise ratio is compared between different estimation methods. Figure 8 (b) shows a comparison of the number of snapshots using different estimation methods. Figure 8 (c) represents a comparison of grid density for different estimation methods. In the comparison, a 7-element CSNA array was selected, with two near-field sources (10°, 5λ) and (32°, 7λ). The simulated source type was set to narrowband non-Gaussian signal, and Gaussian white noise was applied. It can be seen that at low signal-to-noise ratios (SNR), 2D-MUSIC exhibits the best performance in angle estimation, while the performance of the other two methods improves with increasing SNR. A similar performance is observed in the comparison of snapshot count; when the number of snapshots is large, the NLA-OMP algorithm slightly outperforms the 2D-MUSIC algorithm. However, in the comparison of computation time, although the 2D-MUSIC algorithm has a performance advantage over the other two methods under the same conditions, this comes at the cost of lower computational efficiency. The NLA-OMP algorithm, on the other hand, has the best computational time performance. Therefore, overall, only when the AE detection environment is significantly affected by noise can the 2D-MUSIC algorithm be chosen, sacrificing computational efficiency for computational accuracy. In most other application scenarios, the NLA-OMP algorithm can achieve excellent computational efficiency while meeting estimation accuracy requirements, making it the best choice.

[0090] To verify the localization capability of the proposed method under real acoustic emission signals, experimental verification was conducted, and the results are as follows: Figure 10 As shown in the figure. The test specimen was a 600×600×10mm 7075-T6 aluminum alloy plate. The acoustic emission signal was simulated by the breaking of a 0.5mm pencil, during which the pencil elongated by approximately 2.5mm. The positioning results show that, among the eight randomly selected broken pencil locations, the first seven achieved effective positioning. However, while the acoustic emission source angle information was obtainable at the eighth location, its estimation error was the worst among the eight locations, and the distance estimation could not be obtained. Overall, the method proposed in this invention not only exhibits superior near-field acoustic emission positioning performance but also significantly improves computational efficiency.

[0091] Example 2

[0092] A near-field acoustic emission source localization method based on orthogonal matched pursuit in sparse arrays includes the following steps:

[0093] Step A: Arrange sensors in a sparse array to collect and preprocess burst-type acoustic emission signals under the sparse sensor array arrangement.

[0094] Step A1: Based on the geometric characteristics of the monitored object, and guided by the optimal number of virtual array elements obtained by sparse array differential, complete the installation and arrangement of sparse sensor arrays for key monitoring parts of the structure.

[0095] The sparse sensor array is configured as a compressed symmetric nested array structure, which is divided into three levels of subarrays: a first subarray, a second subarray, and a third subarray. The center element of the first subarray serves as the reference point for the entire array, and the element size of the first subarray is defined as 2M²-1, with an element spacing of d = λ / 4, where λ is the signal wavelength. The number of elements in the second and third subarrays is defined as N², and the element spacing is (2M²-1)d.

[0096] Step A2: Acquire sudden acoustic emission signals using devices such as multi-channel signal acquisition instruments, signal amplifiers, and filters.

[0097] Step A3: Use the Shannon wavelet transform method to decompose the signal to obtain a narrowband acoustic emission signal with only A0 and S0 modes at a certain frequency f as the main frequency. Set the cutoff frequency according to the dispersion curve of the guided wave under specific materials and structures to obtain an acoustic emission signal containing only A0 and S0 modes. Then, use the Shannon wavelet transform method to obtain the narrowband acoustic emission signal at the main frequency f.

[0098] Step B: Establish a dimensionality reduction model of the near-field acoustic emission source based on fourth-order cumulants.

[0099] Step B1: Based on the collected acoustic emission signals, construct a near-field source model for sparse arrays, and obtain the signal propagation model under the near-field acoustic emission source:

[0100]

[0101] Where x i (t) represents the original time-domain signal received by the i-th array element, K is the number of signal sources, ω is the angular frequency, d is the element spacing, i is the element number, and s k (t) represents the k-th narrowband AE source signal, and p is the coordinate value. i Let n be the coordinates of the i-th element. i (t) represents the noise signal received by the i-th array element, λ is the signal wavelength, and θ k and r k Then it represents the angle and distance of the k-th source from the phase origin.

[0102] Step B2: Select sensors at special locations in the sparse array to construct a special fourth-order cumulant matrix C4 containing only the angle information from the near-field acoustic emission source. This also achieves the purpose of the sparse array differential process to obtain virtual extended array elements.

[0103] The fourth-order cumulant received by the sparse array sensor is defined as:

[0104] in, Let m,n,f,q∈[-Q,Q] be the fourth-order cumulant of the k-th acoustic emission source.

[0105] To separate angle information from distance information and achieve dimensionality reduction of the acoustic emission source, taking n = -m and q = -f, a special fourth-order cumulant can be constructed:

[0106] Therefore, the special fourth-order cumulative matrix C4 for sparse sensor arrays is defined as follows:

[0107]

[0108] The expressions for each symbol in the fourth-order cumulant matrix C4 are as follows: B(θ)=[b(θ1),…,b(θ K )],

[0109] In the special fourth-order cumulant matrix C4 construction method, c4(k) = c4(m,n) is defined, where k = mn are the coordinates of the virtual elements of the difference array. To remove redundant elements in the virtual sensor, C4 is vectorized as follows:

[0110] Where y is the received signal vector of the virtual array after differential processing of the one-dimensional sparse array. The source signal vector of the virtual array. This is the guiding vector matrix of the virtual array.

[0111] Step C: Establish a near-field acoustic emission source angle information estimation model based on orthogonal matching pursuit.

[0112] Step C1: Based on the virtual extended array element steering vector obtained in step B2 Establish the overcomplete steering vector matrix A of the virtual array Θ The virtual array signal model is transformed into a sparse representation, making it an l1-norm convex optimization problem min||p Θ ||1

[0113] The virtual array element obtained by sparse array difference can be represented as S V ={Pv -Y , Pv -Y+1 , ..., 0, ..., Pv Y-1 , Pv Y}, where Pv i Let be the coordinates of the i-th virtual element. Simultaneously define... D>>K is an overcomplete dictionary from the perspective of the information source, where Let A be the possible angle of incidence, and D be the number of angles of incidence. Therefore, the overcomplete steering vector matrix A of the virtual array can be constructed. Θ :

[0114]

[0115] The solution to the signal propagation model can then be transformed into an l1 norm convex optimization problem: min||p Θ ||1 in, For the single-shot signal received by the virtual array element after redundancy removal, p Θ It is a sparse vector with K non-zero elements.

[0116] Step C2: Use the orthogonal matching pursuit algorithm to realize the sparse reconstruction and solution of the angle information of the near-field acoustic emission source under sparse array.

[0117] Calculate using the orthogonal matching pursuit algorithm The optimal number corresponding to the selected atom, A Θ,j A represents Θ The j-th column of (θ). Then update the index set Γ. t =Γ t-1 ∪{γ t and residual The angle information of the near-field acoustic emission source is estimated, and the estimated value of the angle information of the near-field acoustic emission source is obtained.

[0118] Step D: Establish a near-field acoustic emission source distance estimation model based on the MUSIC algorithm.

[0119] Step D1: Construct the corresponding covariance matrix R based on the acoustic emission signal received by the sensor. XX A MUSIC algorithm model for near-field acoustic emission sources was established.

[0120] Covariance matrix R XX : R is obtained through eigenvalue decomposition XX The signal is decomposed into a signal subspace and a noise subspace, and then the spatial spectrum function of the near-field acoustic emission source is established based on the MUSIC algorithm:

[0121] Step D2: Input the near-field acoustic emission source angle information obtained in step C3 into the near-field MUSIC model established in step D1 to obtain a near-field MUSIC model that only includes distance information.

[0122] The estimated angle information of the near-field acoustic emission source is input into the spatial spectrum function of the near-field acoustic emission source established based on the MUSIC algorithm to obtain a special spatial spectrum function that contains only the distance term. This is achieved by performing eigenvalue decomposition on the covariance function to obtain the signal subspace and the noise subspace, which are orthogonal to each other. The MUSIC algorithm mainly uses the orthogonality between the noise subspace and its corresponding signal vector for source estimation. Where θ k Let k be the estimated acoustic emission source angle, k = 1, 2, ..., K.

[0123] Step D3: Perform spectral peak search on the near-field MUSIC model obtained in step D2 that only contains the distance term. That is, perform spectral peak search on the spatial spectral function that only contains the distance term to estimate and match the distance information under each angle term information, and complete the localization of the near-field acoustic emission source.

[0124] The above embodiments are merely illustrative of the inventive concept and features of the present invention, intended to enable those skilled in the art to understand the content of the present invention and implement it accordingly, and should not be construed as limiting the scope of protection of the present invention. All equivalent changes or modifications made based on the essence of the present invention should be covered within the scope of protection of the present invention.

Claims

1. A near-field acoustic emission source localization method based on orthogonal matched pursuit in a sparse array, characterized in that, Includes the following steps: Step 1: Based on the geometric characteristics of the monitored object and guided by the optimal number of virtual array elements obtained by differential sparse sensor array, complete the installation and arrangement of sparse sensor arrays for key monitoring parts of the structure, collect the sudden acoustic emission signals under the sparse sensor array, and decompose the sudden acoustic emission signals to obtain the acoustic emission narrowband signals. Step 2: Based on the aforementioned narrowband acoustic emission signal, construct a near-field source model oriented towards the sparse sensor array. Select sensors at specific locations within the sparse sensor array to construct a special fourth-order cumulant matrix containing only the angle term information from the near-field acoustic emission source. Establish a near-field acoustic emission source dimensionality reduction model based on the fourth-order cumulant matrix according to the near-field source model. Obtain virtual extended array elements through sparse array differencing. The near-field source model is as follows: , i = - Q , - Q +1, …, Q -1, Q . , ,in, x i ( t () represents the original time-domain signal received by the i-th array element. K For the number of information sources, Angular frequency, d For the spacing between array elements, i For array element sequence number, s k ( t ) is the first k A narrowband AE source signal, where p represents the coordinate value. p i For the first i The coordinate values ​​of each array element. n i ( t ) is the first i The noise signal received by each array element λ For the signal wavelength, θ k and r k Then it is the first k The angle and distance of each signal source from the phase origin, and the fourth-order cumulant received by the sparse sensor array are defined as follows: ,in, For the first k The fourth-order cumulant of an acoustic emission source m , n,f, q [- Q , Q ]; Pick n =- m , q =- f This constructs a special fourth-order cumulant: ; Construct the special fourth-order cumulant matrix C4: ， The expressions for each symbol in the fourth-order cumulant matrix C4 are as follows: , , ; Step 3: Based on the steering vector of the virtual extended array element, establish the overcomplete steering vector matrix of the virtual array, convert the signal model of the virtual array into a sparse representation, sparsely reconstruct and solve the angle term information of the near-field acoustic emission source under the sparse array, and establish an estimation model of the angle term information of the near-field acoustic emission source. Step 4: Construct the corresponding covariance matrix based on the acoustic emission signal received by the sensor, establish a MUSIC algorithm model for the near-field acoustic emission source, input the angle term information of the near-field acoustic emission source into the MUSIC algorithm model to obtain a near-field MUSIC model that only includes distance term information, perform spectral peak search on the near-field MUSIC model to obtain the positioning information of the near-field acoustic emission source.

2. The near-field acoustic emission source localization method based on orthogonal matched pursuit in a sparse array according to claim 1, characterized in that: The sparse sensor array is a compressed symmetric nested array structure, comprising a first subarray, a second subarray, and a third subarray in three levels. The array element at the center of the first subarray serves as the reference point for the sparse sensor array structure, and the array elements of the first subarray are positioned at multiples of 2. M 2-1, the element spacing is d=λ / 4, λ Given the signal wavelength, the number of elements in the second and third subarrays is defined as follows: N 2, the element spacing is (2 M 2-1) d .

3. The near-field acoustic emission source localization method based on orthogonal matched pursuit in sparse arrays according to claim 1, characterized in that: The signal decomposition process sets the cutoff frequency based on the dispersion curve of the guided wave under specific materials and structures, obtaining an acoustic emission signal containing only the A0 and S0 modes. The dominant frequency is then obtained using the Shannon wavelet transform method. f The acoustic emission narrowband signal.

4. The near-field acoustic emission source localization method based on orthogonal matched pursuit in sparse arrays according to claim 1, characterized in that: The construction of the special fourth-order cumulative matrix C4 also includes: defining... ,in k = m - n Given the virtual element coordinates of the difference array, C4 is vectorized as follows: , where y is the received signal vector of the virtual array after differential sparse array; The source signal vector of the virtual array; This is the guiding vector matrix of the virtual array.

5. The near-field acoustic emission source localization method based on orthogonal matched pursuit in a sparse array according to claim 4, characterized in that: The sparse array differentially obtains the virtual extended array element S. V Represented as: S V ={ Pv -Y , Pv -Y+1 , …, 0,…, Pv Y-1, Pv Y },in Pv i Let be the coordinates of the i-th virtual array element; definition , D >> K For the source perspective, an overcomplete dictionary, in which For the possible angles of incidence, D To determine the number of incident angles, construct the overcomplete steering vector matrix of the virtual array. : ， 。 6. The near-field acoustic emission source localization method based on orthogonal matched pursuit in a sparse array according to claim 5, characterized in that: The solution of the signal model is transformed into l 1-norm convex optimization problem: ,in, This refers to the single-shot signal received by the virtual array element after redundancy removal. It is a type of... K A sparse vector with n non-zero elements.

7. The near-field acoustic emission source localization method based on orthogonal matched pursuit in a sparse array according to claim 6, characterized in that: The sparse reconstruction and solution of the angle term information of the near-field acoustic emission source under sparse array is calculated using the orthogonal matching pursuit algorithm: The optimal number corresponding to the selected atom is then identified. express The j Columns, update index set and residual We obtain the estimated angle information of the near-field acoustic emission source.

8. The near-field acoustic emission source localization method based on orthogonal matched pursuit in sparse arrays according to claim 7, characterized in that: The covariance matrix R XX for: The covariance matrix is ​​obtained through eigenvalue decomposition. R XX The signal is decomposed into a signal subspace and a noise subspace. Based on the MUSIC algorithm, the source is estimated by using the orthogonality between the noise subspace and the signal vector corresponding to the MUSIC algorithm, thus establishing the spatial spectrum function of the near-field acoustic emission source. .

9. The near-field acoustic emission source localization method based on orthogonal matched pursuit in sparse arrays according to claim 8, characterized in that: Substituting the estimated angle information of the near-field acoustic emission source into the spatial spectrum function yields a special spatial spectrum function containing only the distance term: ,in Given the already estimated acoustic emission source angle, k =1, 2, …, K .

10. The near-field acoustic emission source localization method based on orthogonal matched pursuit in a sparse array according to claim 9, characterized in that: By performing a spectral peak search on the special spatial spectral function that contains only a distance term, the distance information under each angle term is estimated and matched, thus completing the localization of the near-field acoustic emission source.

Images