Transformer sound source localization imaging method, system and device based on sparse Bayesian learning model and storage medium

Through the sound source positioning method based on the sparse Bayesian learning model, the problem of signal-to-noise ratio change in the transformer sound source positioning is solved, and high-resolution and accurate acoustic imaging is achieved, which is suitable for complex electromagnetic environments.

CN119044891AInactive Publication Date: 2024-11-29GUANGDONG POWER GRID CO LTD DONGGUAN POWER SUPPLY BUREAU

Patent Information

Application Number
CN202411143558.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-20
Publication Date
2024-11-29
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

In the positioning of transformer sound source, the existing acoustic imaging technology has a long distance from the microphone array and the complex electromagnetic environment, which leads to a change in the data signal-to-noise ratio, affecting the accuracy of acoustic imaging, making it difficult to achieve high-resolution sound source positioning.

Method used

Using a sparse Bayesian learning model method, using variable Bayesian inference by establishing spatial coordinate systems, constructing propagation matrix, short-time Fourier transform and multi-layer Bayesian model, combining noise model and likelihood function Methods A close posterior distribution of signal, noise and variance variables was obtained, and a high-resolution sound pressure map was generated to locate the transformer sound source.

Benefits of technology

It improves the resolution and accuracy of transformer acoustic imaging, reduces noise interference during data acquisition, and is suitable for high-resolution sound source positioning in complex environments.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119044891A_ABST
    Figure CN119044891A_ABST
Patent Text Reader

Abstract

The invention discloses a transformer sound source localization imaging method, system and device based on a sparse Bayesian learning model and a storage medium, and relates to the field of acoustic imaging, in particular to the field of transformer sound source localization imaging based on the sparse Bayesian learning model. According to the invention, the problem of low accuracy caused by large noise interference in the existing transformer sound source localization imaging method is solved. The method comprises the following steps: S1, predefining parameters; s2, constructing an over-complete representation model; s3, constructing a sparse Bayesian learning model; s4, calculating a joint distribution function; s5, updating the signal variable; updating a signal variance variable; updating a noise variance variable; s6, calculating a relative error between the current update and the last update; s7, outputting the expectation of the signal variable in the final update as the estimated value of the signal; calculating the signal energy of each grid point according to the estimated value of the signal, superposing a target plane sound pressure image and a transformer image, and displaying the sound source position of the transformer. The method is suitable for the high-resolution imaging field of transformer sound source localization.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the field of acoustic imaging, and in particular to the field of transformer sound source localization imaging based on a sparse Bayesian learning model. Background Art

[0002] The spatial distribution of noise sources in the operating state of the transformer can reflect its health status. Accurately locating the transformer sound source is of great significance for transformer fault diagnosis and health status monitoring. Acoustic imaging can intuitively display the transformer sound source distribution by generating sound pressure diagrams and superimposing them with video images. It is an effective means of locating the transformer sound source.

[0003] Existing acoustic imaging technology generally uses beamforming algorithm or CLEAN algorithm. The distance between the microphone array and the transformer is far, and the electromagnetic environment of the substation is complex. The data acquisition process of the microphone array is easily interfered, causing the data signal-to-noise ratio to change, which seriously affects the accuracy of acoustic imaging. Therefore, high-resolution acoustic imaging technology plays a key role in the accurate positioning of the transformer sound source. Summary of the invention

[0004] The present invention aims to improve the resolution of acoustic imaging and reduce noise interference when collecting data.

[0005] The technical solution described in the present invention is a transformer sound source localization imaging method based on a sparse Bayesian learning model, and the method comprises the following steps:

[0006] S1. Establish a spatial coordinate system, determine the spatial coordinates of the microphone array reference point and the spatial coordinates of all microphones, determine the target plane, grid the target plane, determine the spatial coordinates of all grid points in the gridded target plane, bring the spatial coordinates of all microphones, the spatial coordinates of the grid points and the target frequency into the Green function to construct a propagation matrix;

[0007] S2. Collect the time domain signal of the microphone array according to the propagation matrix described in S1, convert the time domain signal into a time-frequency domain signal using a short-time Fourier transform, select the time-frequency domain data of the time-frequency domain signal and construct an over-complete representation model;

[0008] S3. Constructing a sparse Bayesian learning model according to the overcomplete representation model of S2, wherein the sparse Bayesian learning model includes a multi-layer Bayesian model of the signal, a noise model and a likelihood function;

[0009] S4. Determine the joint distribution of all variables according to the multi-layer Bayesian model, noise model and likelihood function described in S3;

[0010] S5. According to the joint distribution described in S4, the approximate posterior distribution of the signal variable is obtained using the variable Bayesian inference method, wherein the approximate posterior distribution is a multidimensional complex Gaussian distribution, and the signal variable is updated by calculating the expectation and covariance matrix of the multidimensional complex Gaussian distribution;

[0011] Using a variable Bayesian inference method to obtain an approximate posterior distribution of the signal variance variable, the approximate posterior distribution is a gamma distribution, and the signal variance variable is updated by calculating the expectation of the gamma distribution;

[0012] Using a variable Bayesian inference method to obtain an approximate posterior distribution of the noise variance variable, the approximate posterior distribution is a gamma distribution, and the noise variance variable is updated by calculating the expectation of the gamma distribution;

[0013] S6. Obtain the relative error between the current update and the last update of the signal variable, signal variance variable and noise variance variable described in S5, and determine whether to stop updating according to the update judgment condition, wherein the update judgment condition is:

[0014] When the relative error is greater than 10 -4 If the number of updates is less than 2000, repeat S5 to update.

[0015] When the relative error is less than or equal to 10 -4 Or if the number of updates is greater than or equal to 2000, stop updating and execute S7.

[0016] S7. Output the expectations of the current signal variable, signal variance variable and noise variance variable as the estimated value of the signal, calculate the signal energy at each grid point position according to the estimated value, generate a target plane sound pressure map, read the transformer image, superimpose the target plane sound pressure map and the transformer image, and obtain the transformer sound source position.

[0017] Furthermore, the Green's function in S1 is:

[0018]

[0019] Among them, r k represents the distance from the kth grid point to the microphone array reference point, r mk represents the distance of the kth grid point to the mth microphone, τ mk represents the relative delay of the kth grid point to the mth microphone,

[0020]

[0021] e is the speed of sound propagation and f is the frequency of the sound wave.

[0022] Furthermore, the overcomplete representation model in S2 is:

[0023] Yi,t =A i X i,t +W i,t

[0024] Among them, Y i,t represents the measurement data at the tth moment on the i-th frequency band, A i represents the propagation matrix corresponding to the ith frequency band, X i,t represents the signal matrix at the tth time on the i-th frequency band, W i,t represents the measurement noise matrix at the tth moment in the ith frequency band.

[0025] Furthermore, in the multi-layer Bayesian model of the signal described in S3, the first layer uses a zero-mean complex Gaussian model to describe the signal:

[0026]

[0027] Among them, x .,l is the lth column of the signal matrix, L is the number of windows in each frame, γ -1 is the variance vector of the signal, Γ is a diagonal matrix, and the elements on the diagonal are γ;

[0028] The second layer of the multi-layer Bayesian model of the signal uses a gamma distribution to describe the variable γ:

[0029]

[0030] Among them, κ is the kth element in Υ, a and b are hyperparameters and can be set to 10 -3 ;

[0031] The noise model is:

[0032]

[0033] Among them, w .,l is the lth column of the noise matrix, ρ -1 is the variance variable, I M is the identity matrix;

[0034] We further describe ρ using the gamma distribution:

[0035] p(ρ)=G(ρ|c,d)

[0036] The likelihood function is:

[0037]

[0038] Furthermore, the method for determining the joint distribution of all variables described in S4 is:

[0039]

[0040] Among them, Θ is the set of all hidden variables, ‖·‖ F is the matrix norm.

[0041] A transformer sound source localization imaging system based on sparse Bayesian learning includes the following modules:

[0042] Module 1 is used to establish a spatial coordinate system, determine the spatial coordinates of the microphone array reference point and the spatial coordinates of all microphones, determine the target plane, grid the target plane, determine the spatial coordinates of all grid points in the gridded target plane, and bring the spatial coordinates of all microphones, the spatial coordinates of the grid points and the target frequency into the Green function to construct a propagation matrix;

[0043] Module 2 is used to collect the time domain signal of the microphone array according to the propagation matrix, convert the time domain signal into a time-frequency domain signal using short-time Fourier transform, select the time-frequency domain data of the time-frequency domain signal and construct an over-complete representation model;

[0044] Module 3, used to construct a sparse Bayesian learning model based on the overcomplete representation model, wherein the sparse Bayesian learning model includes a multi-layer Bayesian model of the signal, a noise model and a likelihood function;

[0045] Module 4, for determining the joint distribution of all variables based on the multi-layer Bayesian model, the noise model and the likelihood function of the signal;

[0046] Module 5 is used to obtain an approximate posterior distribution of the signal variable using a variable Bayesian inference method according to the joint distribution, wherein the approximate posterior distribution is a multidimensional complex Gaussian distribution, and update the signal variable by calculating the expectation and covariance matrix of the multidimensional complex Gaussian distribution;

[0047] It is also used to obtain an approximate posterior distribution of the signal variance variable using a variable Bayesian inference method, wherein the approximate posterior distribution is a gamma distribution, and the signal variance variable is updated by calculating the expectation of the gamma distribution;

[0048] It is also used to obtain an approximate posterior distribution of the noise variance variable using a variable Bayesian inference method, wherein the approximate posterior distribution is a gamma distribution, and the noise variance variable is updated by calculating the expectation of the gamma distribution;

[0049] Module 6 is used to obtain the relative error between the current update and the last update of the signal variable, signal variance variable and noise variance variable. Module 6 includes two submodules 61 and 62;

[0050] Submodule 61 is used to -4 If the number of updates is less than 2000, repeat S5 to update;

[0051] Submodule 62 is used to determine the relative error when the relative error is less than or equal to 10 -4 Or if the number of updates is greater than or equal to 2000, stop updating and execute submodule 7;

[0052] Submodule 7 is used to output the expectations of the current signal variable, signal variance variable and noise variance variable as the estimated value of the signal, calculate the signal energy at each grid point position according to the estimated value, generate a target plane sound pressure map, read the transformer image, superimpose the target plane sound pressure map and the transformer image, and obtain the transformer sound source position.

[0053] A transformer sound source localization imaging device based on a sparse Bayesian learning model comprises one or more processors and a memory, wherein the memory is used to store a computer executable program, the processor reads part or all of the computer executable program from the memory and executes it, and the transformer sound source localization imaging method can be implemented when the processor executes part or all of the computer executable program.

[0054] A computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the transformer sound source localization imaging method can be implemented.

[0055] The present invention can achieve the following beneficial effects:

[0056] (1) The broadcast matrix A of the present invention f The propagation characteristics of the signal between different microphones are described, providing a basis for subsequent data processing and model building.

[0057] (2) The overcomplete representation model described in the present invention contains more basic elements, which helps to describe and reconstruct signals more accurately and is applicable to complex noise environments.

[0058] (3) The sparse Bayesian learning model described in the present invention determines the joint distribution of all variables by describing the Bayesian model of signals and noise, and obtains the approximate posterior distribution of each variable by minimizing the distance (Kullback-Leibler), thereby achieving high-resolution acoustic imaging.

[0059] (4) The joint distribution function calculation module described in the present invention provides a basis for updating the signal and variance variables.

[0060] The imaging method of the present invention improves the resolution and accuracy of transformer acoustic imaging, and is applicable to the field of high-resolution imaging for transformer sound source positioning. BRIEF DESCRIPTION OF THE DRAWINGS

[0061] Figure 1 The present invention is a flowchart of the sound source localization imaging method. DETAILED DESCRIPTION

[0062] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0063] See also Figure 1 This embodiment is described. In this embodiment, a 64-element multi-arm spiral microphone array is used to perform acoustic imaging on an oil-immersed 10kV transformer.

[0064] S1: Predefined parameters. Establish a spatial coordinate system, determine the spatial coordinates of the microphone array reference point and the spatial coordinates of all microphones, determine the target plane, set the center point of the microphone array as the coordinate origin, establish a coordinate system based on the coordinate origin, determine the spatial coordinates of all microphones, select the target plane, the target plane covers the main components of the transformer, grid the target plane, determine all grid points, and bring the microphone spatial coordinates, target grid point coordinates and target frequency into the following Green's function to construct the propagation matrix A f , the propagation matrix A f The element in the mth row and kth column of is:

[0065]

[0066] Among them, r k represents the distance from the kth grid point to the microphone array reference point, r mk represents the distance of the kth grid point to the mth microphone, τ mk represents the relative delay of the kth grid point to the mth microphone,

[0067]

[0068] e is the speed of sound propagation, which is set to 343 m / s in this embodiment, and f is the frequency.

[0069] S2: Process the microphone array data and build an over-complete representation model. The microphone array time domain signal is collected, and the time domain signal is converted into a time-frequency domain signal using a short-time Fourier transform. In this embodiment, the sampling frequency is 192kHz, the window function is a 4096-point Hanning window, and the overlap rate between windows is 90%. The time domain data of each window is subjected to a 4096-point fast Fourier transform to obtain the time-frequency domain data. The time-frequency domain data of every 20 windows is set as one frame, and an over-complete representation model is built for the time-frequency domain data within one frame:

[0070] Yf,t =A f X f,t +W f,t (0.2)

[0071] Among them, Y i,t represents the measurement data at the tth moment on the i-th frequency band, A i represents the propagation matrix corresponding to the ith frequency band, X i,t represents the signal matrix at the tth time on the i-th frequency band, W i,t represents the measurement noise matrix at the tth moment in the ith frequency band.

[0072] S3: Construct a sparse Bayesian learning model. A sparse Bayesian learning model is constructed based on the overcomplete representation model, in which the block sparse characteristics of the signal matrix are expressed through two layers of conjugate priors, that is, the variables in the signal matrix are described using a zero-mean complex Gaussian distribution, and the variables in the same row share variance parameters, and the variance parameters are regarded as new variables, and the variance variables are described using a gamma distribution. In addition, the variables in the measurement noise matrix are described using a zero-mean complex Gaussian distribution. On this basis, the likelihood function can be obtained according to the overcomplete representation model.

[0073] The method for constructing a multi-layer Bayesian model of the signal is:

[0074] The first layer uses a zero-mean complex Gaussian model to describe the signal, that is:

[0075]

[0076] Among them, x ·,l is the lth column of the signal matrix, L is the number of windows in each frame, γ -1 is the variance vector of the signal, Γ is a diagonal matrix, and the elements on the diagonal are γ.

[0077] The second layer uses the gamma distribution to describe the variable γ, that is:

[0078]

[0079] Among them, γ k is the kth element in γ, a and b are hyperparameters, which are set to 10 in this implementation. -3 . Then, construct the noise model.

[0080] In this implementation, assuming that the noise is Gaussian white noise, then:

[0081]

[0082] Among them, w ·,l is the lth column of the noise matrix, ρ -1 is the variance variable, I Mis the identity matrix.

[0083] Furthermore, using the gamma distribution to describe ρ, we have:

[0084] p(ρ)=G(ρ|c,d)(0.6)

[0085] Among them, c and d are hyperparameters, which are set to 10 in this implementation. -3 .

[0086] Finally, the likelihood function is given:

[0087]

[0088] Among them, y ·,l Y f,t The first column of .

[0089] S4: Calculate the joint distribution function. In this embodiment, the joint distribution function is calculated using the multi-layer Bayesian model of the signal, the noise model and the likelihood function. For the convenience of expression, the log-likelihood function is given in this embodiment, that is:

[0090] Step 4: Determine the joint distribution of all variables based on the Bayesian model of signal and noise and the likelihood function. The present invention uses the variable Bayesian inference method to obtain the approximate posterior distribution of each variable by minimizing the Kullback-Leibler distance between the variable distribution and the actual distribution of the variable according to the joint distribution.

[0091]

[0092] Among them, Θ is the set of all hidden variables, ‖·‖ F is the matrix norm.

[0093] S5: Update signal variables. According to the joint distribution and variable Bayesian inference method, the approximate posterior distribution of the signal variable is obtained, which is a multidimensional complex Gaussian distribution. The signal variable is updated by calculating the expectation and covariance matrix of the multidimensional complex Gaussian distribution. In this embodiment, there are:

[0094]

[0095] Among them, <·> represents expectation, μ ·,l is the l-th column estimate of the signal.

[0096] Update the signal variance variable. According to the joint distribution and variable Bayesian inference method, the approximate posterior distribution of the signal variance variable is obtained, which is a gamma distribution. The signal variance variable is updated by calculating the expectation of the gamma distribution. That is, in this embodiment, the update of Σ involves the variable γ, and its update rule is:

[0097]

[0098] Update the noise variance variable. According to the joint distribution and variable Bayesian inference method, the approximate posterior distribution of the noise variance variable is obtained, which is a gamma distribution. The expected updated noise variance variable of the gamma distribution is calculated. In this embodiment, there are:

[0099]

[0100] S6: Calculate the relative error between this update and the last update. If the relative error is greater than a predefined threshold and the number of updates is less than a predefined maximum number of updates, repeat steps 5 to 7. If the relative error is less than or equal to the predefined threshold or the number of updates is greater than or equal to the predefined maximum number of updates, stop updating. In this embodiment, the error is defined as:

[0101]

[0102] in, <X f,t > old is the estimated value of the signal at the last update. The predefined threshold of the error in this implementation is 10 -4 , the predefined maximum number of updates is 2000.

[0103] S7: Output the expectation of the signal variable in the last update as the estimated value of the signal. Calculate the signal energy of each grid point according to the estimated value of the signal, generate the target plane sound pressure map, read the transformer image, superimpose the target plane sound pressure map and the transformer image, and display the transformer sound source position. The positioning result in this embodiment shows that the main sound source of the oil-immersed 10kV transformer is the insulator.

Claims

1. A transformer sound source localization imaging method based on a sparse Bayesian learning model, characterized in that: The method comprises the following steps: S1. Establish a spatial coordinate system, determine the spatial coordinates of the microphone array reference point and the spatial coordinates of all microphones, determine the target plane, grid the target plane, determine the spatial coordinates of all grid points in the gridded target plane, bring the spatial coordinates of all microphones, the spatial coordinates of the grid points and the target frequency into the Green function to construct a propagation matrix; S2. Collect the time domain signal of the microphone array according to the propagation matrix described in S1, convert the time domain signal into a time-frequency domain signal using a short-time Fourier transform, select the time-frequency domain data of the time-frequency domain signal and construct an over-complete representation model; S3. Constructing a sparse Bayesian learning model according to the overcomplete representation model of S2, wherein the sparse Bayesian learning model includes a multi-layer Bayesian model of the signal, a noise model and a likelihood function; S4. Determine the joint distribution of all variables according to the multilayer Bayesian model, noise model and likelihood function of the signal described in S3; S5. According to the joint distribution described in S4, the approximate posterior distribution of the signal variable is obtained using the variable Bayesian inference method, wherein the approximate posterior distribution is a multidimensional complex Gaussian distribution, and the signal variable is updated by calculating the expectation and covariance matrix of the multidimensional complex Gaussian distribution; Using a variable Bayesian inference method to obtain an approximate posterior distribution of the signal variance variable, the approximate posterior distribution is a gamma distribution, and the signal variance variable is updated by calculating the expectation of the gamma distribution; Using a variable Bayesian inference method to obtain an approximate posterior distribution of the noise variance variable, the approximate posterior distribution is a gamma distribution, and the noise variance variable is updated by calculating the expectation of the gamma distribution; S6. Obtain the relative error between the current update and the last update of the signal variable, signal variance variable and noise variance variable described in S5, and determine whether to stop updating according to the update judgment condition, wherein the update judgment condition is: When the relative error is greater than 10 -4 If the number of updates is less than 2000, repeat S5 to update. When the relative error is less than or equal to 10 -4 Or if the number of updates is greater than or equal to 2000, stop updating and execute S7; S7. Output the expectations of the current signal variable, signal variance variable and noise variance variable as the estimated value of the signal, calculate the signal energy at each grid point position according to the estimated value, generate a target plane sound pressure map, read the transformer image, superimpose the target plane sound pressure map and the transformer image, and obtain the transformer sound source position.

2. The transformer sound source localization imaging method according to claim 1, characterized in that: The Green's function described in S1 is: Among them, r k represents the distance from the kth grid point to the microphone array reference point, r mk represents the distance of the kth grid point to the mth microphone, τ mk represents the relative delay of the kth grid point to the mth microphone, e is the speed of sound propagation and f is the frequency of the sound wave.

3. The transformer sound source localization imaging method according to claim 1, characterized in that: The overcomplete representation model described in S2 is: Y i,t =A i X i,t +W i,t Among them, Y i,t represents the measurement data at the tth moment on the i-th frequency band, A i represents the propagation matrix corresponding to the ith frequency band, X i,t represents the signal matrix at the tth time on the i-th frequency band, W i,t represents the measurement noise matrix at the tth moment in the ith frequency band.

4. The transformer sound source localization imaging method according to claim 1, characterized in that: In the multi-layer Bayesian model of the signal described in S3, the first layer uses a zero-mean complex Gaussian model to describe the signal: Among them, x .,l is the lth column of the signal matrix, L is the number of windows in each frame, γ -1 is the variance vector of the signal, Γ is a diagonal matrix, and the elements on the diagonal are γ; The second layer of the multi-layer Bayesian model of the signal uses a gamma distribution to describe the variable γ: Among them, κ is the κth element in Υ, a and b are hyperparameters and can be set to 10 -3 ; The noise model is: Among them, w ·,l is the lth column of the noise matrix, ρ -1 is the variance variable, I M is the identity matrix; We further describe ρ using the gamma distribution: p(ρ)=G(ρ|c,d) The likelihood function is:

5. The transformer sound source localization imaging method according to claim 1, characterized in that: The method described in S4 to determine the joint distribution of all variables is: Among them, Θ is the set of all hidden variables, ‖·‖ F is the matrix norm.

6. A transformer sound source localization imaging system based on sparse Bayesian learning, characterized in that: Includes the following modules: Module 1 is used to establish a spatial coordinate system, determine the spatial coordinates of the microphone array reference point and the spatial coordinates of all microphones, determine the target plane, grid the target plane, determine the spatial coordinates of all grid points in the gridded target plane, and bring the spatial coordinates of all microphones, the spatial coordinates of the grid points and the target frequency into the Green function to construct a propagation matrix; Module 2 is used to collect the time domain signal of the microphone array according to the propagation matrix, convert the time domain signal into a time-frequency domain signal using short-time Fourier transform, select the time-frequency domain data of the time-frequency domain signal and construct an over-complete representation model; Module 3, used to construct a sparse Bayesian learning model based on the overcomplete representation model, wherein the sparse Bayesian learning model includes a multi-layer Bayesian model of the signal, a noise model and a likelihood function; Module 4, for determining the joint distribution of all variables based on the multi-layer Bayesian model, the noise model and the likelihood function of the signal; Module 5 is used to obtain an approximate posterior distribution of the signal variable using a variable Bayesian inference method according to the joint distribution, wherein the approximate posterior distribution is a multidimensional complex Gaussian distribution, and update the signal variable by calculating the expectation and covariance matrix of the multidimensional complex Gaussian distribution; It is also used to obtain an approximate posterior distribution of the signal variance variable using a variable Bayesian inference method, wherein the approximate posterior distribution is a gamma distribution, and the signal variance variable is updated by calculating the expectation of the gamma distribution; It is also used to obtain an approximate posterior distribution of the noise variance variable using a variable Bayesian inference method, wherein the approximate posterior distribution is a gamma distribution, and the noise variance variable is updated by calculating the expectation of the gamma distribution; Module 6 is used to obtain the relative error between the current update and the last update of the signal variable, signal variance variable and noise variance variable. Module 6 includes two submodules 61 and 62; Submodule 61 is used to -4 If the number of updates is less than 2000, repeat S5 to update; Submodule 62 is used to determine the relative error when the relative error is less than or equal to 10 -4 Or if the number of updates is greater than or equal to 2000, stop updating and execute submodule 7; Submodule 7 is used to output the expectations of the current signal variable, signal variance variable and noise variance variable as the estimated value of the signal, calculate the signal energy at each grid point position according to the estimated value, generate a target plane sound pressure map, read the transformer image, superimpose the target plane sound pressure map and the transformer image, and obtain the transformer sound source position.

7. A transformer sound source localization imaging device based on a sparse Bayesian learning model, characterized in that: The invention comprises one or more processors and a memory, wherein the memory is used to store a computer executable program, and the processor reads part or all of the computer executable program from the memory and executes it. When the processor executes part or all of the computer executable program, the sound source localization imaging method according to any one of claims 1 to 5 can be implemented.

8. A computer-readable storage medium, characterized in that: A computer program is stored in a computer-readable storage medium, and when the computer program is executed by a processor, the sound source localization imaging method according to any one of claims 1 to 5 can be implemented.

Citation Information

Patent Citations

  • Sparse hole diameter ISAR imaging method based on SBL-ADMM

    CN110244303A

  • High-resolution sound source positioning method based on Green function correction

    CN110850371A

  • Sound source imaging method, system and device and storage medium

    CN115291169A

Cited By

  • Broadband acoustic imaging method and device applied to transformer direct current magnetic bias detection

    CN119667343A

  • Bayesian asynchronous measurement noise source positioning method

    CN119959866A

  • Rotary machinery composite fault diagnosis method and system

    CN120316630A

  • Sagittal plane auditory localization method and device based on Bayesian reasoning and medium

    CN120468776A

  • A method, device, and medium for sagittal auditory localization based on Bayesian inference

    CN120468776B