Multi-snapshot partial discharge sound source localization method based on non-convex joint sparse constraint

Abstract

The invention relates to a multi-snapshot partial discharge sound source localization method based on non-convex joint sparse constraint, and belongs to the technical field of partial discharge sound source localization. According to the method, a microphone array collects sound signals radiated to a surrounding space through discharge of equipment, a positioning problem is constructed as an optimization model of Laplacian norm combined sparse constraint, and a single snapshot model is solved cooperatively by using differential convex programming and Nesterov accelerated soft threshold iteration. And according to the joint sparsity, popularizing the single snapshot to a positioning algorithm capable of directly processing multi-sampling data, and finally reconstructing sound field distribution of a detection space, and visualizing the sound field distribution into an acoustic imaging graph, so that the position and the number of discharge sources are intuitively determined. By introducing the Laplace non-convex sparse constraint, the spatial resolution and the imaging quality of the algorithm are remarkably improved, and meanwhile, the positioning robustness of the algorithm in noise and interference environments is effectively enhanced through combined solution of multiple snapshot data.

Description

Technical Field

[0001] This invention belongs to the field of partial discharge sound source localization technology, and relates to a multi-shot partial discharge sound source localization method based on non-convex joint sparse constraints. Background Technology

[0002] Acoustic detection technology, characterized by its non-invasiveness, high efficiency, and resistance to electromagnetic interference, is an emerging method for condition monitoring and fault detection of power equipment. The sound source localization method based on planar microphone arrays can achieve sound field visualization and has been gradually promoted in the field of power inspection in recent years for the detection and localization of partial discharge in equipment, showing broad engineering application prospects.

[0003] Beamforming algorithms (such as time-delay summation) are a typical sound source localization technique. Due to their simple principle and good robustness, handheld audio-visual equipment and other inspection devices have been developed. However, this method is limited by the Rayleigh criterion, resulting in low spatial resolution, which makes it difficult to meet the high-precision requirements of practical engineering scenarios. Deconvolution algorithms (such as DAMAS) effectively improve resolution by deconvolving the beamforming results, but they rely on a complete cross-spectral matrix, resulting in high computational complexity, which is not conducive to field applications with high real-time requirements. Machine learning and deep learning-based methods improve adaptability to complex environments through data-driven approaches, but they face problems such as scarce labeled data and limited model generalization ability in partial discharge detection, which restricts their reliable application in industrial scenarios. In recent years, the rise of compressed sensing theory has promoted the rapid development of compressed beamforming algorithms. These methods are based on the prior knowledge of the sparsity of sound sources in spatial distribution, and can still achieve super-resolution localization under conditions with few array elements. They are particularly suitable for typical sparse sound source scenarios such as partial discharge (where only a few points in the detection area usually experience partial discharge simultaneously).

[0004] In terms of sparsity modeling and constraint function selection, the commonly used L1 norm constraint is a convex relaxation, which has limited sparse excitation capability and may lead to localization errors and increased sidelobes. Research has shown that non-convex Laplace norm constraint functions have strong sparsity induction capabilities and demonstrate good potential in suppressing sidelobes and improving multi-source resolution. However, current research on this topic mainly focuses on the theoretical level, with few application verifications and adaptations in partial discharge detection in power industry scenarios. Furthermore, related research concentrates on single-measurement vector models, failing to effectively process multi-sampled data and exhibiting insufficient robustness in low signal-to-noise ratio or interference environments. Therefore, it is necessary to develop a sound source localization method with strong sparse excitation capability, suitable for multi-shot data acquisition modes, to solve or mitigate the above problems and improve the accuracy and practicality of partial discharge acoustic localization technology. Summary of the Invention

[0005] In view of this, the purpose of this invention is to provide a multi-shot partial discharge sound source localization method based on non-convex joint sparse constraints, which alleviates the problems of poor localization accuracy and weak anti-interference ability of traditional sound source localization algorithms in partial discharge detection. This method can improve resolution by suppressing sidelobes and enhance environmental adaptability by utilizing multi-shot sampling data, thereby achieving accurate and robust partial discharge sound source detection and localization, and thus improving the accuracy and practicality of partial discharge acoustic localization technology.

[0006] To achieve the above objectives, the present invention provides the following technical solution: Solution 1: A multi-shot partial discharge sound source localization method based on non-convex joint sparse constraints, specifically including the following steps: S1: Establish a partial discharge acoustic signal model for the target detection area based on a planar microphone array; S2: Model the partial discharge source localization problem as a least squares model with joint Laplace norm and sparse constraints; S3: The model constructed in step S2 is solved using a localization algorithm. Specifically, first, the objective function of a single measurement vector is considered and solved using a soft thresholding iterative algorithm accelerated by differential convex programming and Nesterov, resulting in a single-shot version of the localization algorithm. Then, the objective function of multiple measurement vectors is decoupled into multiple independent subproblems, and based on joint sparsity, the soft thresholding operator is generalized to a row vector thresholding operator on the basis of the single-shot algorithm, resulting in a localization algorithm that can directly process multi-shot sampled data. S4: Based on the solution results of the localization algorithm, the sound pressure distribution of the sound field is generated into an acoustic imaging map, realizing the visual localization of the partial discharge source.

[0007] Furthermore, in step S1, establishing the partial discharge signal model of the target detection area specifically includes the following steps: S11: Establish a point with the center of the microphone array as the origin. The Cartesian coordinate system, the array plane is located in flat, The positive axis points perpendicularly to the target detection area; the microphone array includes Each array element has a coordinate set represented as... The target detection area is divided into The spacing is A uniform grid, the grid coordinate set is Accordingly, The set of coordinates of a real sound source is ; A real sound source can fall on any grid node coordinate, i.e., satisfy... Furthermore, the greater the sound intensity at that grid point, the more likely a sound source exists at that coordinate. The higher the probability; S12: A continuous time interval obtained by the microphone array Measurement signal of quick snapshot The quantity is known; The sound intensity matrix of potential sound sources at each grid coordinate. For the quantity to be determined, among which Represents complex numbers; the transfer matrix between array elements and the grid is... The error matrix is Using Gaussian white noise for simulation, a linear partial discharge signal model can be constructed as follows:

[0008] Among them, matrix The Middle line, number The elements of a column are defined as follows ,in The imaginary unit, For wave number, The center frequency of the partial discharge signal. The speed at which sound travels through the air. Represents an exponential function; Indicates the first The microphone and the first Euclidean distance between grids This represents the L2 norm.

[0009] Furthermore, in step S2, the partial discharge source localization problem is modeled as a least squares model with joint Laplace norm and sparsity constraints. Specifically, this includes: if the position of the sound source does not change during the sampling time, then the sound intensity matrix... It exhibits consistent row sparsity, i.e. The number of rows in a non-zero row vector is equal to the number of sound sources. Based on this prior information, the following objective function can be constructed:

[0010] in, The Frobenius norm is represented by the following method: , Represents matrix vectorization; Let (Laplace, L1) be the mixture norm, defined as follows: , For matrix The Middle Line number The elements corresponding to the column, For control parameters; This is a regularization parameter used to balance the degree of data fit and the sparsity of the solution.

[0011] Furthermore, in step S3, when a partial discharge source exists at a certain grid location, the sound intensity of that grid node is significantly higher than that of the surrounding grids; therefore, by solving the above objective function to obtain the sound intensity of the grid source, and then searching for the local maxima of the grid sound intensity amplitude, the number and coordinates of the partial discharge sources can be obtained.

[0012] Furthermore, in step S3, the specific steps for obtaining the single-shot version of the localization algorithm are as follows: S301: Sound intensity matrix any row vector in There exists a single measurement vector objective function as follows:

[0013] in, For the regularization parameters of the single snapshot model; This type of non-convex optimization problem can be solved by using the principle of convex difference programming to decompose the original objective function into the difference of two convex functions. By processing these two sub-functions separately, an approximate optimal solution to the overall objective function can be obtained. Therefore, the objective function can be rewritten as follows: ,in:

[0014] S302: Obviously, the function It is a classic LASSO problem, in the first... In the next iteration, by ignoring or introducing a constant term through the formulation, it is equivalent to:

[0015] in, The iteration step size, Defined as a matrix The largest eigenvalue, This represents the conjugate transpose operation; For functions Within the majorization-minimization framework, according to The first-order concavity condition can be linearized as:

[0016] in, Indicates the inner product; gradient matrix The elements in the middle are calculated using the following formula:

[0017] in, Represents the modulus of a complex number; The sign function for complex numbers is defined as follows: ; S303: Will and Substituting the approximate expression into the overall objective function; similarly, performing the complete square, the solution can be expressed as:

[0018] in, To facilitate the representation of the defined matrix ; This represents the soft thresholding operator. The Middle The value of each element is defined as ; S304: The Nesterov strategy is adopted to accelerate the above iterative solution process, based on current and historical iteration information. Construct the following momentum term :

[0019] in, This is a momentum parameter; By replacing the object of the soft thresholding operator with the result of momentum acceleration, it is possible to obtain an approximate optimal solution at the optimal first-order speed. Repeat the soft thresholding and momentum acceleration steps described above until the iteration stopping condition is met, thus obtaining the single-snapshot version of the localization algorithm.

[0020] Furthermore, in step S3, the specific steps for obtaining a positioning algorithm that can directly process multi-shot sampling data are as follows: S311: For multi-shot signals The objective function can be generalized to the following multi-measurement vector form:

[0021] in, The (L2, L1) mixed norm is represented by the following method: , Representation matrix The Each row vector; correspondingly, It should be noted that, in order to maintain consistency of dimensions, Should be in accordance with The formula is calculated element by element; Due to the sound intensity matrix It exhibits joint sparsity, therefore, for The first in row vectors Based on the properties of the Frobenius norm and the (L2, L1) mixture norm, the multi-measurement vector objective function can be decoupled into multiple independent sub-functions; then the... Sub-objective functions corresponding to each row vector for:

[0022] S312: The The solution at the next iteration is represented as ,in, Scaling factor; substitute Zhongde:

[0023] beg right The first-order partial derivatives are:

[0024] Setting it to zero, we get exist There is a minimum value at; when (Right now When the sound intensity is high, If it has a non-negative physical meaning, then the objective function is... There is a minimum value at; when (Right now When ), There is a minimum value at that point; substitute it into... The expression, the first The solution at the next iteration is:

[0025] This expression is a generalization of the soft thresholding operator from its application to elements to its application to row vectors. It is named the row vector thresholding iterative operator and its notation is defined as follows: The symbol of an element is defined as follows: When row vector When the vector contains only one element, the row vector thresholding iteration operator and the soft thresholding operator are equivalent and can be uniformly written as:

[0026] Similarly, repeat the row vector thresholding iteration operation. With the momentum acceleration step, the original multi-snapshot objective function can be directly solved, yielding a multi-snapshot version of the localization algorithm. The momentum acceleration term corresponding to the multi-shot model is defined as follows: ; S313: For the multi-snapshot localization algorithm, the regularization parameters are dynamically updated using the following strategy:

[0027] Among them, symbols Indicates to After sorting in descending order, the first Larger values ​​can serve as a benchmark for thresholds.

[0028] Scheme 2: A multi-shot partial discharge sound source localization system based on non-convex joint sparse constraints, comprising a microphone array, a partial discharge sound source localization module, and an acoustic imager; the partial discharge sound source localization module acquires the target signal of the sound source to be detected through the microphone array, then uses a multi-shot partial discharge sound source localization method based on non-convex joint sparse constraints to solve for the sound intensity distribution of the sound field, and finally sends the sound intensity distribution of the sound field to the acoustic imager to generate an acoustic image map to visualize the number and location of the sound sources.

[0029] The beneficial effects of this invention are as follows: This invention collects acoustic signals radiated into the surrounding space by the discharge of the device using a microphone array, constructs the localization problem as an optimization model with Laplace norm joint sparsity constraints, and uses differential convex programming and Nesterov-accelerated soft threshold iteration to solve the single-shot model collaboratively. Then, based on joint sparsity, the single-shot model is generalized to a localization algorithm that can directly process multi-sample data, ultimately reconstructing the sound field distribution of the detection space and visualizing it as an acoustic image, thereby intuitively determining the location and number of discharge sources. This invention uses Laplace non-convex joint sparsity constraints to induce strong sparsity, thereby improving the spatial resolution and imaging quality of the algorithm. Simultaneously, the joint solution of multi-shot data effectively enhances its localization robustness in noisy and interference environments.

[0030] This invention is applicable to the partial discharge detection and location of unshielded equipment such as insulators and bushings in open spaces such as substations and converter stations, providing an efficient, intuitive, and non-invasive detection technology for the safe operation and maintenance of power equipment.

[0031] Other advantages, objectives, and features of the invention will be set forth in part in the description which follows, and in part will be apparent to those skilled in the art from the following examination, or may be learned from practice of the invention. The objectives and other advantages of the invention can be realized and obtained through the following description. Attached Figure Description

[0032] To make the objectives, technical solutions, and advantages of the present invention clearer, the preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings, wherein: Figure 1 This is a flowchart of the multi-fastshot partial discharge sound source localization method based on non-convex joint sparse constraints of the present invention; Figure 2 This is a planar microphone array used to collect partial acoustic signals in the experiment. Figure 3The results of two actual experiments conducted to verify the effectiveness of the method of the present invention are as follows: (a) is the experimental result of locating the sound source of the dual loudspeakers in a semi-anechoic laboratory; (b) is the experimental result of locating the sound source of the partial discharge of the defective insulator in a high-voltage test hall. Detailed Implementation

[0033] The following specific examples illustrate the implementation of the present invention. Those skilled in the art can easily understand other advantages and effects of the present invention from the content disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and various details in this specification can be modified or changed based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic representations of the basic concept of the present invention. Unless otherwise specified, the following embodiments and features can be combined with each other.

[0034] The accompanying drawings are for illustrative purposes only and are schematic diagrams, not actual pictures. They should not be construed as limiting the invention. To better illustrate the embodiments of the invention, some parts in the drawings may be omitted, enlarged, or reduced, and do not represent the actual product dimensions. It is understandable to those skilled in the art that some well-known structures and their descriptions may be omitted in the drawings.

[0035] In the accompanying drawings of the embodiments of the present invention, the same or similar reference numerals correspond to the same or similar components. In the description of the present invention, it should be understood that if terms such as "upper," "lower," "left," "right," "front," and "rear" indicate the orientation or positional relationship based on the orientation or positional relationship shown in the drawings, they are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, the terms used to describe positional relationships in the drawings are only for illustrative purposes and should not be construed as limiting the present invention. For those skilled in the art, the specific meaning of the above terms can be understood according to the specific circumstances.

[0036] Figure 1 This document presents a flowchart illustrating the overall process of the multi-shot partial discharge sound source localization method based on non-convex joint sparse constraints of this invention. The flowchart includes key steps such as transforming the physical problem of partial discharge localization into a mathematical optimization model, designing a solution method for the single-sample data model, extending the algorithm to process multi-sample data, and generating the localization results. Please refer to [link / reference]. Figure 1 The present invention will be further described below with reference to the accompanying drawings and embodiments: Step S1: Establish a partial discharge acoustic signal model of the target detection area based on the planar microphone array, specifically including the following steps: S11: Establish a point with the center of the microphone array as the origin. The Cartesian coordinate system, the array plane is located in flat, The positive axis points perpendicularly to the target detection area; the microphone array includes Each array element has a coordinate set represented as... The target detection area is divided into The spacing is A uniform grid, the grid coordinate set is Accordingly, The set of coordinates of a real sound source is ; S12: The real sound source may be located at any grid node coordinate (i.e., If the sound intensity at a grid point is greater, then the presence of a sound source at that coordinate (i.e., The higher the probability of (), the better; S13: A continuous time interval obtained by the microphone array Measurement signal of quick snapshot It is a known quantity ( (representing complex numbers); The sound intensity matrix of potential sound sources at each grid coordinate. The variable to be determined is: The transfer matrix between the array element and the mesh is: The error matrix is Using Gaussian white noise for simulation, a linear partial discharge signal model can be constructed as follows:

[0037] Among them, matrix The Middle line, number The elements of a column are defined as follows , The imaginary unit, For wave number, The center frequency of the partial discharge signal. The speed at which sound travels through the air. Represents an exponential function; Indicates the first The microphone and the first Euclidean distance between grids This represents the L2 norm.

[0038] Step S2: Model the partial discharge source localization problem as a least squares model with joint Laplace norm and sparse constraints, specifically including the following steps: S21: If the position of the sound source does not change during the sampling time, then the sound intensity matrix... It exhibits consistent row sparsity, i.e. The number of rows in a non-zero row vector is equal to the number of sound sources. Based on this prior information, the following objective function can be constructed:

[0039] in, The Frobenius norm is represented by the following method: , Represents matrix vectorization; Let (Laplace, L1) be the mixture norm, defined as follows: , For matrix The Middle Line number The elements corresponding to the column, For control parameters; This is a regularization parameter used to balance the degree of data fit and the sparsity of the solution; S22: When a partial discharge source exists at a certain grid location, the sound intensity of that grid node is significantly higher than that of the surrounding grids. Therefore, by solving the above objective function to obtain the sound intensity of the grid source, and then searching for the local maxima of the grid sound intensity amplitude, the number and coordinates of the partial discharge sources can be obtained.

[0040] Step S3: First, consider the objective function of a single measurement vector. Solve it using differential convex programming and a Nesterov-accelerated soft thresholding iterative algorithm to obtain the single-snapshot version of the localization algorithm. Specifically, this includes the following steps: S31: Sound intensity matrix any row vector in There exists a single measurement vector objective function as follows:

[0041] in, For the regularization parameters of the single snapshot model; This type of non-convex optimization problem can be solved by using the principle of convex difference programming to decompose the original objective function into the difference of two convex functions. By processing these two sub-functions separately, an approximate optimal solution to the overall objective function can be obtained. Therefore, the objective function can be rewritten as follows: ,in:

[0042] S32: Obviously, the function It is a classic LASSO problem, in the first... In the next iteration, by ignoring or introducing a constant term through the formulation, it is equivalent to:

[0043] in, The iteration step size, Defined as a matrix The largest eigenvalue, Indicates conjugate transpose; For functions Within the majorization-minimization framework, according to The first-order concavity condition can be linearized as:

[0044] in, Indicates the inner product; gradient matrix The elements in the middle are calculated using the following formula:

[0045] in, Represents the modulus of a complex number; The sign function for complex numbers is defined as follows: ; S33: Will and Substituting the approximate expression into the overall objective function; similarly, performing the complete square, the solution can be expressed as:

[0046] in, To facilitate the representation of the defined matrix ; This represents the soft thresholding operator. The Middle The value of each element is defined as ; S34: Employ the Nesterov strategy to accelerate the above iterative solution process, based on current and historical iteration information. Construct the following momentum term :

[0047] in, This is a momentum parameter; By replacing the object of the soft thresholding operator with the result of momentum acceleration, it is possible to obtain an approximate optimal solution at the optimal first-order speed. Repeat the above soft thresholding and momentum acceleration steps until the iteration stopping condition is met, thus obtaining the single-shot version of the positioning algorithm described in this invention.

[0048] Step S4: Decouple the multi-measurement vector objective function into multiple independent sub-problems, and based on joint sparsity, extend the soft thresholding operator to a row vector thresholding operator on the basis of the single-shot algorithm to obtain a localization algorithm that can directly process multi-shot sampling data. Specifically, this includes the following steps: S41: For multi-shot signals The objective function can be generalized to the following multi-measurement vector form:

[0049] in, The (L2, L1) mixed norm is represented by the following method: , Representation matrix The Each row vector; correspondingly, To maintain dimensional consistency, according to The formula is calculated element by element; Due to the sound intensity matrix It exhibits joint sparsity, therefore, for The first in row vectors Based on the properties of the Frobenius norm and the (L2, L1) mixture norm, the multi-measurement vector objective function can be decoupled into multiple independent sub-functions; then the... Sub-objective functions corresponding to each row vector for:

[0050] S42: No. The solution at the next iteration is represented as ,in, Scaling factor; substitute Zhongde:

[0051] beg right The first-order partial derivatives are:

[0052] Setting it to zero, we get exist There is a minimum value at; when (Right now When the sound intensity is high, If it has a non-negative physical meaning, then the objective function is... There is a minimum value at; when (Right now When ), There is a minimum value at that point; substitute it into... The expression, the first The solution at the next iteration is:

[0053] Clearly, this expression is a generalization of the soft thresholding operator from its application to elements to its application to row vectors. We will name it the row vector thresholding iterative operator and define its notation as follows: The symbol of an element is defined as follows: When row vector When the vector contains only one element, the row vector thresholding iteration operator and the soft thresholding operator are equivalent and can be uniformly written as:

[0054] Similarly, repeat the row vector thresholding iteration operation. With the momentum acceleration step, the original multi-snapshot objective function can be directly solved, yielding a multi-snapshot version of the localization algorithm. The momentum acceleration term corresponding to the multi-shot model is defined as follows: ; S43: For the multi-snapshot localization algorithm, the regularization parameters are dynamically updated using the following strategy:

[0055] Among them, symbols Indicates to After sorting in descending order, the first Larger values ​​can serve as a benchmark for thresholds.

[0056] Step S5: Based on the solution results of the localization algorithm, generate an acoustic imaging map of the sound pressure distribution of the sound field to realize the visual localization of the partial discharge source.

[0057] Figure 2 The experimental setup showcased a planar microphone array with 128 channels, arranged in a multi-arm spiral shape.

[0058] Figure 3 The localization results of the method of the present invention in two actual experimental scenarios are shown; (a) is an acoustic imaging image of locating a dual-speaker sound source in a semi-anechoic laboratory. The results show that both sound sources can be identified and the location estimation is accurate; (b) is an acoustic imaging image of locating a partial discharge sound source of a defective insulator in a high-voltage test hall. There is a gap caused by human damage on the edge of the upper umbrella layer of the insulator. The results show that the insulator has a discharge phenomenon and the defect location can be clearly locked.

[0059] The beneficial effects of the present invention are demonstrated through experiments: Experimental setup: First, audio was played using two speakers in a semi-anechoic laboratory to conduct a dual-source localization experiment to verify the effectiveness of the method of this invention. The experimental results are as follows: Figure 3 As shown in (a); to further illustrate its effectiveness in detecting and locating partial discharge sources in engineering scenarios, a partial discharge location experiment of defective insulators was conducted in this high-voltage test hall. The experimental results are as follows. Figure 3 As shown in (b); from Figure 3 As can be seen, the method of the present invention can detect different types of sound source signals in real-world scenarios and accurately estimate the number and location of sound sources. Furthermore, no other interference sources were identified within the 30 dB imaging range, verifying the effectiveness of the present invention.

[0060] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for locating multi-shot partial discharge sound sources based on non-convex joint sparse constraints, characterized in that, The method specifically includes the following steps: S1: Establish a partial discharge acoustic signal model for the target detection area based on a planar microphone array; S2: Model the partial discharge source localization problem as a least squares model with joint Laplace norm and sparse constraints; S3: The model constructed in step S2 is solved using a localization algorithm. Specifically, first, the objective function of a single measurement vector is considered and solved using a soft thresholding iterative algorithm accelerated by differential convex programming and Nesterov, resulting in a single-shot version of the localization algorithm. Then, the objective function of multiple measurement vectors is decoupled into multiple independent subproblems, and based on joint sparsity, the soft thresholding operator is generalized to a row vector thresholding operator on the basis of the single-shot algorithm, resulting in a localization algorithm that can directly process multi-shot sampled data. S4: Based on the solution results of the localization algorithm, the sound pressure distribution of the sound field is generated into an acoustic imaging map, realizing the visual localization of the partial discharge source.

2. The multi-shot partial discharge sound source localization method according to claim 1, characterized in that, Step S1, establishing the partial discharge signal model of the target detection area specifically includes the following steps: S11: Establish a point with the center of the microphone array as the origin. The Cartesian coordinate system, the array plane is located in flat, The positive axis points perpendicularly to the target detection area; the microphone array includes Each array element has a coordinate set represented as... The target detection area is divided into The spacing is A uniform grid, the grid coordinate set is Accordingly, The set of coordinates of a real sound source is The real sound source may fall on any grid node coordinate, i.e., satisfy... Furthermore, the greater the sound intensity at that grid point, the more likely a sound source exists at that coordinate. The higher the probability; S12: A continuous time interval obtained by the microphone array Measurement signal of quick snapshot The quantity is known; The sound intensity matrix of potential sound sources at each grid coordinate. For the quantity to be determined, among which Represents complex numbers; the transfer matrix between array elements and the grid is... The error matrix is The partial discharge signal is simulated using Gaussian white noise; therefore, the following linear partial discharge signal model is constructed: Among them, matrix The Middle line, number The elements of a column are defined as follows ,in The imaginary unit, For wave number, The center frequency of the partial discharge signal. The speed at which sound travels through the air. Represents an exponential function; Indicates the first The microphone and the first Euclidean distance between grids This represents the L2 norm.

3. The multi-shot partial discharge sound source localization method according to claim 2, characterized in that, In step S2, the partial discharge source localization problem is modeled as a least squares model with joint Laplace norm and sparsity constraints. Specifically, if the position of the sound source does not change during the sampling time, then the sound intensity matrix... It exhibits consistent row sparsity, i.e. The number of rows in a non-zero row vector is equal to the number of sound sources. Based on this prior information, the following objective function is constructed: in, The Frobenius norm is represented by the following method: , Represents matrix vectorization; Let (Laplace, L1) be the mixture norm, defined as follows: , For matrix The Middle Line number The elements corresponding to the column, For control parameters; This is a regularization parameter used to balance the degree of data fit and the sparsity of the solution.

4. The multi-shot partial discharge sound source localization method according to claim 3, characterized in that, In step S3, the sound intensity of the grid sound source is obtained by solving the objective function, and then the local maxima of the grid sound intensity amplitude are searched to obtain the number and coordinates of the partial discharge sound source.

5. The multi-shot partial discharge sound source localization method according to claim 3, characterized in that, In step S3, the specific steps to obtain the single-shot version of the localization algorithm are as follows: S301: Sound intensity matrix any row vector in There exists a single measurement vector objective function as follows: in, For the regularization parameters of the single snapshot model; The original objective function is decomposed into the difference of two convex functions using the principle of differential convex programming. By processing these two sub-functions separately, an approximate optimal solution to the overall objective function is obtained. Therefore, the objective function is rewritten as follows: ,in: S302: Obviously, the function It is a classic LASSO problem, in the first... In the next iteration, by ignoring or introducing a constant term through the formulation, it is equivalent to: in, The iteration step size, Defined as a matrix The most prominent feature; This represents the conjugate transpose operation; For functions Within the majorization-minimization framework, according to The first-order concavity condition is linearized as follows: in, Indicates the inner product; gradient matrix The elements in the middle are calculated using the following formula: in, Represents the modulus of a complex number; The sign function for complex numbers is defined as follows: ; S303: Will and Substituting the approximate expression into the overall objective function; similarly, performing the complete square, the solution is expressed as: in, To facilitate the representation of the defined matrix ; This represents the soft thresholding operator. The Middle The value of each element is defined as ; S304: The Nesterov strategy is adopted to accelerate the above iterative solution process, based on current and historical iteration information. Construct the following momentum term : in, This is a momentum parameter; By replacing the object of the soft thresholding operator with the result of momentum acceleration, an approximate optimal solution can be obtained at the optimal first-order speed. Repeat the soft thresholding and momentum acceleration steps until the iteration stopping condition is met, thus obtaining the single-snapshot version of the localization algorithm.

6. The multi-shot partial discharge sound source localization method according to claim 5, characterized in that, In step S3, the specific steps for obtaining a localization algorithm that can directly process multi-shot sampling data are as follows: S311: For multi-shot signals The objective function is generalized to the following multi-measurement vector form: in, The (L2, L1) mixed norm is represented by the following method: , Representation matrix The Each row vector; correspondingly, To maintain dimensional consistency, according to The formula is calculated element by element; for The first in row vectors Based on the properties of the Frobenius norm and the (L2, L1) mixture norm, the multi-measurement vector objective function is decoupled into multiple independent sub-functions; then the... Sub-objective functions corresponding to each row vector for: S312: The The solution at the next iteration is represented as ,in, Scaling factor; substitute Zhongde: beg right The first-order partial derivatives are: Setting it to zero, we get exist There is a minimum value at; when Right now At that time, due to the sound intensity If it has a non-negative physical meaning, then the objective function is... There is a minimum value at; when Right now At that time, There is a minimum value at that point; substitute it into... The expression, the first The solution at the next iteration is: This expression is a generalization of the soft thresholding operator from its application to elements to its application to row vectors. It is named the row vector thresholding iterative operator and its notation is defined as follows: The symbol of an element is defined as follows: When row vector When the vector contains only one element, the row vector thresholding iteration operator and the soft thresholding operator are equivalent and can be uniformly written as: Similarly, repeat the row vector thresholding iteration operation. Following the momentum acceleration step, the original multi-snapshot objective function is directly solved to obtain the multi-snapshot version of the localization algorithm, where... The momentum acceleration term corresponding to the multi-shot model is defined as follows: ; S313: For the multi-snapshot localization algorithm, the regularization parameters are dynamically updated using the following strategy: Among them, symbols Indicates to After sorting in descending order, the first Larger values ​​can serve as a benchmark for thresholds.

7. A system for implementing the multi-shot partial discharge sound source localization method according to any one of claims 1 to 6, characterized in that, The system includes a microphone array, a partial discharge sound source localization module, and an acoustic imager. The partial discharge sound source localization module acquires the target signal of the sound source to be detected through the microphone array, and then uses a multi-shot partial discharge sound source localization method based on non-convex joint sparse constraints to solve for the sound intensity distribution of the sound field. Finally, the sound intensity distribution of the sound field is sent to the acoustic imager to generate an acoustic image map to visualize the number and location of the sound sources.

Images