A system and method for imaging and locating transformer fault sound sources

Abstract

This application provides a system and method for imaging and locating the sound source of a transformer fault. The system includes: a microphone array module composed of N array elements arranged at equal phase intervals along M spiral arms; an array optimization module that performs topological optimization of the spatial coordinates of the microphone array elements using a particle swarm optimization algorithm to generate optimal array layout parameters; a signal acquisition and transmission module that uses the microphone array module to acquire and transmit sound signals; and a signal processing module that receives the transmitted sound signals, performs fast Fourier transform (FFT) time-frequency analysis and multiple signal classification, and improves the spatial spectrum estimation of the multiple signal classification by coherent superposition of subarray power spectra to obtain an acoustic energy cloud map. By optimizing the microphone array structure and improving the signal processing algorithm, this application improves the accuracy and imaging effect of sound source localization. Particularly in low signal-to-noise ratio environments, it can effectively reduce the impact of sidelobes and improve the accuracy and reliability of sound source localization.

Description

Technical Field

[0001] The present application relates to the field of acoustic imaging technology, and in particular to a system and method for imaging and locating a transformer fault sound source. Background Art

[0002] In power systems, transformers are critical transmission and transformation equipment, and their operating status directly impacts the reliability and safety of the power grid. According to statistics, power outages caused by transformer failures account for approximately 30% of power system failures. Traditional fault detection methods (such as oil chromatography and partial discharge detection) suffer from shortcomings such as response lag and high invasiveness. However, fault location technology based on acoustic signals, due to its non-contact and real-time capabilities, has become a hot topic in the industry.

[0003] Existing technologies face several bottlenecks in array structure and algorithms. Traditional acoustic positioning systems often use regular arrays (such as uniform linear, rectangular, or circular arrays). These arrays present several challenges. Specifically, their uniform spacing leads to grating lobe effects when locating wideband sound sources, which can easily lead to false source localization. Conventional circular arrays also exhibit high sidelobe levels in the 6kHz band, significantly interfering with mainlobe resolution.

[0004] In addition, the existing MUSIC (Multiple Signal Classification) algorithm has insufficient noise resistance in low signal-to-noise ratio (SNR<10dB) environments, resulting in reduced positioning accuracy. Summary of the Invention

[0005] In view of one of the defects in the prior art, the purpose of this application is to provide a transformer fault sound source imaging and positioning system and method.

[0006] In a first aspect of the present application, a transformer fault sound source imaging and positioning system is provided, comprising:

[0007] The microphone array module consists of N array elements arranged at equal phase intervals along M spiral arms;

[0008] an array optimization module, which uses a particle swarm optimization algorithm to perform adaptive topology optimization on the microphone array, with the coordinates of each array element being constrained to not exceed half the aperture of the microphone array, and with the weighted sum of the main lobe width, maximum side lobe height, and uniformity of array element spacing of the microphone array as the particle fitness, to obtain a microphone array with an optimal array layout;

[0009] A signal acquisition and transmission module, which uses the microphone array with the optimal array layout to acquire sound signals and transmit the sound signals;

[0010] The signal processing module receives the transmitted sound signal, performs fast Fourier transform time-frequency analysis and multiple signal classification, and obtains the sound signal by dividing N array elements into S groups. The coherent superposition of the power spectra of the sub-arrays improves the spatial spectrum estimation of the multiple signal classification and obtains an acoustic energy cloud map for fault sound source location.

[0011] Optionally, the microphone array module uses 64 array elements to achieve balanced main lobe width and side lobe level;

[0012] The 64 array elements are arranged at equal phase intervals along the 8 spiral arms, and the operating frequency band is 2000 Hz;

[0013] The angle between adjacent spiral arms is 45 degrees;

[0014] The polar coordinate equation of the i-th array element is:

[0015] r i ( θ )= aθ i + b ；

[0016] in a is the spiral growth rate, b is the initial radius offset, θ i is the polar angle of the i-th array element, 1≤i≤64;

[0017] The phase difference between adjacent elements of a single spiral arm satisfies:

[0018] Δφ=2πk / N ;

[0019] Where N=64 is the total number of array elements and k=8 is the number of spiral arms.

[0020] Optionally, the array optimization module includes:

[0021] Position determination submodule: determines the position of the 8 array elements on a single spiral arm, i The position of each element is:

[0022] ;

[0023] ;

[0024] Coordinate constraint submodule: Determine the i The array element is in the k The coordinate constraints on the spiral arms are:

[0025] ;

[0026] ;

[0027] D is the array aperture;

[0028] Fitness definition submodule: Establish the fitness function definition as follows:

[0029] F=α·main lobe width+β·maximum side lobe height+γ·uniformity of array element spacing;

[0030] Where α, β, and γ are weighting coefficients, and α + β + γ = 1. Array element spacing uniformity is an indicator that measures the uniformity of microphone distribution in the array. It directly affects the array's beamforming performance and refers to the degree of discreteness of the spacing between adjacent array elements.

[0031] Iterative optimization submodule: Iterative optimization is performed through particle swarm method to generate the optimal array layout parameters.

[0032] Optionally, the iterative optimization submodule generates optimal array layout parameters through iterative optimization using a particle swarm optimization method, including:

[0033] Setting parameters of the particle swarm method, including population size, particle dimension, and respective threshold ranges, wherein each particle represents a microphone array distribution;

[0034] Randomly generate multiple particles and initialize the position X of each particle i , speed V i , get the particle's previous optimal position P ibest ;

[0035] For each particle, calculate its performance indicators, including the main lobe width, the maximum side lobe height, and the uniformity of the array element spacing, to obtain the fitness of each particle;

[0036] According to the fitness and the number of iterations, it is determined whether the termination condition is met. If so, the search is terminated. Otherwise, the previous best position P of the particle is updated. ibest , calculate the global optimal position G best , update the speed and position of all particles; return to calculate the fitness of particles;

[0037] The global optimal position G of the last iteration best In order to finally obtain the optimal array layout, obtain G best The corresponding array element coordinates and spiral parameters (a, b) are used as the optimal array layout parameters.

[0038] Optionally, the signal processing module includes:

[0039] Fourier transform submodule: converting the sound signal from the time domain to the frequency domain using Fourier transform to obtain a frequency domain sound signal;

[0040] Extraction submodule: extracting a target frequency band containing a sound signal from the frequency domain sound signal;

[0041] Music submodule: performs spectrum estimation on the target frequency band to obtain the noise subspace and fault location;

[0042] The superposition submodule divides the N microphone main array into S non-overlapping groups, from which all S-1 groups are selected to form S sub-arrays;

[0043] The main array and all the sub-arrays are imaged, and all the images are summed to form a final acoustic cloud map.

[0044] Optionally, the superposition submodule divides the 64 microphones into four non-overlapping groups, each group containing 16 microphones; three groups are used to form a subarray;

[0045] When performing acoustic imaging calculations, in addition to imaging the original 64-element main array, it is also necessary to image the 48-element sub-array;

[0046] The power values ​​of the main array and sub-array are normalized and summed to form the final acoustic cloud map.

[0047] Optionally, it also includes a photoacoustic fusion imaging module, which collects optical images, and spatially aligns the acoustic energy cloud map with the optical image based on the SIFT feature point matching algorithm, and outputs a multimodal fusion visualization result.

[0048] A second aspect of the present application provides a method for imaging and locating a transformer fault sound source, comprising:

[0049] A microphone array is formed by arranging N array elements at equal phase intervals along M spiral arms;

[0050] With the constraint that the coordinates of each array element do not exceed half of the aperture of the microphone array, and the weighted sum of the main lobe width, the maximum height of the side lobe, and the uniformity of the array element spacing of the microphone array as the particle fitness, the particle swarm optimization algorithm is used to perform adaptive topology optimization on the microphone array to obtain a microphone array with an optimal array layout;

[0051] Collecting sound signals using the microphone array with the optimal array layout, and transmitting the sound signals;

[0052] Receive the transmitted sound signal, perform fast Fourier transform time-frequency analysis and multiple signal classification, and obtain the sound signal by dividing N array elements into S groups. The coherent superposition of the power spectra of the sub-arrays improves the spatial spectrum estimation of the multiple signal classification and obtains an acoustic energy cloud map for fault sound source location.

[0053] According to a third aspect of the present application, a terminal is provided, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor can be used to execute the system or the method when executing the program.

[0054] In a fourth aspect of the present application, a computer-readable storage medium is provided, on which a computer program is stored. When the program is executed by a processor, it can be used to execute the method described, or run the system described.

[0055] The transformer fault sound source imaging and positioning system of the present application uses a particle swarm optimization algorithm to optimize the microphone array, effectively solving the sidelobe interference problem of the traditional array during the sound source positioning process, thereby achieving a better acoustic imaging effect.

[0056] Other technical effects brought about by the additional features will be further explained in the corresponding embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0057] Other features, objects and advantages of the present application will become more apparent upon reading the detailed description of non-limiting embodiments with reference to the following drawings:

[0058] Figure 1 1 is a schematic structural diagram of a transformer fault sound source imaging and locating system according to an exemplary embodiment;

[0059] Figure 2 Schematic diagram of a two-dimensional Archimedean spiral MEMS microphone array according to an exemplary embodiment;

[0060] Figure 3 Graph showing the impact of topology on microphone array performance according to an exemplary embodiment, where (a) main lobe width, (b) maximum side lobe height (dB);

[0061] Figure 4 Graph showing the effect of the number of microphones on acoustic imaging quality according to an exemplary embodiment, where (a) main lobe width, (b) maximum side lobe height (dB);

[0062] Figure 5 is a flow chart of a PSO algorithm according to an exemplary embodiment;

[0063] Figure 6 Figures 2 and 3 show cloud images before and after the improvement of the MUSIC algorithm according to an exemplary embodiment, wherein (a) a two-dimensional cloud image generated by the MUSIC algorithm, (b) a three-dimensional cloud image generated by the MUSIC algorithm, (c) a two-dimensional cloud image generated by the improved algorithm, and (d) a three-dimensional cloud image generated by the improved algorithm.

[0064] Figure 7The figure is a flow chart of a method for imaging and locating a transformer fault sound source according to an exemplary embodiment. DETAILED DESCRIPTION

[0065] The present application is described in detail below with reference to specific embodiments. The following examples will help those skilled in the art to further understand the present application, but are not intended to limit the present application in any form. It should be noted that, without departing from the concept of the present application, a number of variations and improvements may be made by those skilled in the art, and these all fall within the scope of protection of the present application. Parts not described in detail in the following examples may be implemented using existing technologies.

[0066] In existing technology, uniform spacing causes the array to produce grating lobe effects when locating wideband sound sources, which can easily lead to false sound source localization. Conventional 64-element circular arrays have high sidelobe levels in the 6kHz band, seriously interfering with mainlobe resolution. Uniform spacing refers to the spatial distribution characteristics of the array elements in a Cartesian coordinate system (rectangular coordinate system). To address these issues, embodiments of the present application provide a transformer fault sound source imaging and localization system to address these issues.

[0067] Reference Figure 1 As shown, a transformer fault sound source imaging and positioning system 100 includes:

[0068] The microphone array module 110 is composed of N array elements arranged at equal phase intervals along M spiral arms;

[0069] An array optimization module 120 performs adaptive topology optimization on the microphone array using a particle swarm optimization algorithm, with the constraint that the coordinates of each array element do not exceed half the aperture of the microphone array and the weighted sum of the main lobe width, the maximum side lobe height, and the uniformity of the array element spacing as the particle fitness, to obtain a microphone array with an optimal array layout.

[0070] The signal collection and transmission module 130 uses a microphone array with an optimal array layout to collect sound signals and transmit the sound signals;

[0071] The signal processing module 140 receives the transmitted sound signal, performs fast Fourier transform time-frequency analysis and multiple signal classification, and obtains the sound signal by dividing the N array elements into S groups. The coherent superposition of the power spectra of the sub-arrays improves the spatial spectrum estimation of the multiple signal classification and obtains an acoustic energy cloud map for fault sound source location.

[0072] Here, "equal phase spacing" specifically refers to the regular angular arrangement of array elements in a polar coordinate system, intended to ensure phase consistency of the array's beam directivity within a specific frequency band. As can be seen above, "uniform spacing" refers to the spatial distribution characteristics of array elements in a Cartesian (rectangular) coordinate system. Therefore, the spiral array in the aforementioned embodiment of this application is non-uniform, primarily manifested in the diversity of spatial sampling intervals. This contrasts sharply with regular arrays (such as uniform linear and circular arrays).

[0073] Specifically, the microphone array is arranged in a non-uniform pattern along the spiral arms, thereby disrupting the periodic structure and avoiding repetitive spatial sampling. Furthermore, through the iterative update process of the particle swarm optimization algorithm, the array element layout is optimized to meet the requirements of a narrow mainlobe, low sidelobes, and high uniformity, effectively reducing the sidelobe level. Furthermore, the coherent superposition of the subarray power spectra suppresses localized noise and enhances the robustness of the system. The combination of these three technologies effectively addresses the problem of false sound source localization and the interference with mainlobe resolution caused by high sidelobe levels in the 6kHz band, which is a common problem in existing technologies.

[0074] The main lobe and side lobes in the acoustic cloud map can be used to evaluate the performance of a microphone array. The largest peak in the acoustic cloud map corresponds to the location of the sound source and is usually called the main lobe, while the remaining peaks are collectively referred to as side lobes. The narrower the main lobe, the higher the resolution, indicating that the array is more accurate in localizing the sound source. The fewer side lobes and the lower the peaks, the better the array's ability to suppress ambient noise.

[0075] In order to obtain the best performance of the microphone array, in some specific embodiments of the present application, the microphone array uses 64 array elements, which are arranged at equal phase intervals along 8 spiral arms, with the angle between adjacent spiral arms being 45 degrees, and the operating frequency band being 2000 Hz. Figure 2 As shown, such 64 array elements form a two-dimensional Archimedean spiral MEMS microphone array.

[0076] Specifically, analysis of the acoustic signal characteristics generated by transformer faults revealed that the signal strength at the mid-frequency band of 2000Hz increases significantly when a fault occurs. Therefore, setting the operating frequency band to 2000Hz can effectively identify transformer faults.

[0077] Specifically, the microphone array consists of eight spiral arms spaced at 45-degree intervals, achieving uniform omnidirectional coverage of the horizontal sound field and eliminating directional blind spots. The non-uniform sampling characteristics of the spiral structure, combined with a beamforming algorithm (multiple signal classification), suppress sidelobe levels to below -15 decibels, significantly improving the localization signal-to-noise ratio. The 64 array elements are arranged symmetrically, with eight elements per arm, meeting the Nyquist criterion (the theoretical anti-aliasing frequency of 6800Hz is greater than the operating frequency of 2000Hz), achieving optimized spatial undersampling and effectively preventing spatial aliasing of the sound field. Furthermore, a phase constraint formula of Δφ=2πk / N=π / 4 ensures phase coherence among the spiral elements as the spiral grows, enhancing beam focusing capabilities on rotationally symmetric transformer sound sources. Ultimately, a lateral resolution of less than 0.5 meters and a dynamic range exceeding 30 dB are achieved in the 2000Hz band.

[0078] Specifically, the relationship between the number of microphones and the main lobe width and side lobe level is as follows: Figure 3 Figure (a) and (b) Figure 4 As shown in Figures (a) and (b), as the number of microphones increases, the main lobe width gradually increases and the side lobe level gradually decreases.

[0079] When the number of microphones varies between 36 and 64, the mainlobe width remains nearly constant, while the sidelobe levels gradually decrease. When the number of microphones exceeds 64, the mainlobe width increases significantly, but the sidelobe levels do not decrease significantly. Taking both mainlobe width and sidelobe levels into consideration, a total of 64 microphones was selected to ensure the optimal overall performance of the microphone array.

[0080] Furthermore, in some specific implementations of the present application, the microphone array is generated as follows:

[0081] The first spiral arm is initially generated, and the remaining seven spiral arms are created by rotating them 45°. The polar coordinate equation of the i-th element is

[0082] r i (θ)=aθ i +b ;

[0083] Where a is the spiral growth rate, b is the initial radius offset, θ i is the polar angle of the i-th array element, 1≤i≤64;

[0084] The phase difference between adjacent array elements satisfies

[0085] Δφ=2πk / N ;

[0086] Where N=64 is the total number of array elements and k=8 is the number of spiral arms.

[0087] For example, the microphone is IM69D120. According to the temperature reliability test of JESD22A-103E and JESD22A-108D standards, IM69D120 can arrive It can operate and store data in an environment with excellent temperature stability. arrive temperature requirements.

[0088] In this application, the two-dimensional Archimedean spiral MEMS microphone array used has the ability to suppress sidelobe interference. The non-periodic layout of the array elements along the Archimedean spiral effectively avoids the problem of repeated spatial sampling of the uniform array, thereby reducing the sidelobe energy. In addition, the array can eliminate the grating lobe effect and prevent the occurrence of false positioning. This is because the geometric characteristics of the spiral equation break the phase periodicity caused by the uniform spacing, thereby avoiding the generation of grating lobes. The array also exhibits wide-band adaptability and can take into account both high and low frequency performance. The elements in the central area of ​​the spiral array are relatively dense, which is suitable for adapting to the short wavelength of high-frequency signals; while the peripheral area is relatively sparse to adapt to the long wavelength of low-frequency signals. For different frequency bands, the array can dynamically optimize its response to achieve frequency band focusing.

[0089] The sound propagation model can be divided into two types: near-field and far-field models. The critical value is generally calculated using the following formula:

[0090] ;

[0091] Where L is the distance from the array to the sound source, D is the array aperture, λ is the wavelength of the acoustic signal, f is the frequency of the acoustic signal, and v is the speed of sound in air. When a transformer is defective, the characteristic frequency of the sound is primarily distributed in the range of 100 to 20,300 Hz. The critical distance is between 1 and 25 meters. To ensure that the microphone array can capture sufficiently clear sound signals, in some specific embodiments of this application, the operating distance between the microphone array and the dry-type transformer is set to 1.5 meters. Based on this analysis, the sound wave propagation model is considered to be a near-field model.

[0092] This application explicitly proposes that the acoustic field model be a near-field model, which accurately simulates the physical properties of sound wave propagation and optimizes array design, signal processing algorithms, and system performance. In the near-field model, sound waves behave as spherical waves, and their wavefront curvature and sound pressure amplitude attenuation patterns differ significantly from those in the far-field plane wave model. Therefore, determining that the acoustic field model is a near-field model is a key prerequisite for achieving high-precision sound source localization.

[0093] The periodic arrangement of traditional uniform arrays (such as uniform linear and circular arrays) results in insufficient spatial sampling frequency, generating grating lobes when localizing broadband sound sources, leading to false sound source alarms. To further improve localization accuracy, in some specific embodiments of this application, the array optimization module uses a particle swarm optimization (PSO) algorithm to adaptively optimize the array topology. In this embodiment, the array optimization module includes the following submodules.

[0094] Position determination submodule: The particle swarm optimization algorithm (PSO) is used to determine the layout of the eight array elements on a single spiral arm. From the inside to the outside of the spiral arm, the array elements are numbered from 1 to 8.

[0095] The position of the i-th element of a single spiral arm is set as follows:

[0096] ;

[0097] ;

[0098] Specifically, a is the spiral growth rate, and b is the initial radius offset. The spiral growth rate is a core parameter that describes the geometric expansion rate of a spiral. In acoustic array design, it specifically refers to the radius expansion per unit angular increment of the spiral arm. The spiral growth rate represents the linear increment in radius r for each radian (approximately 57.3°) increase in the polar angle θ of the spiral. The initial radius offset is a key parameter in spiral array design that controls the starting position of each spiral arm and directly affects the spatial symmetry and acoustic performance of the array.

[0099] Eight spiral arms are generated by fixing the angle interval θ = 45°, and the starting radius of each arm is dynamically adjusted by a and b. The term binds the element index i to the spiral arm angle interval θ, ensuring that the phase difference between the arms satisfies .

[0100] It is worth noting that the parameters of traditional spiral arrays (such as a and b) are usually fixed based on experience or the half-wavelength rule, while this application uses the PSO algorithm to use a and b as optimization variables.

[0101] Coordinate constraint submodule: Considering the portability and practicality of the array, the array aperture is set to D = 0.5m. Therefore, the coordinate constraint of the i-th array element on the k-th spiral arm of the particle swarm algorithm is as follows:

[0102] ;

[0103] ;

[0104] Fitness function submodule:

[0105] The fitness of the PSO algorithm is determined by the main lobe width and the maximum side lobe height. The fitness function of the designed PSO algorithm is defined as

[0106] F=α·main lobe width+β·maximum side lobe height+γ·uniformity of array element spacing;

[0107] Where α, β, and γ are weighting coefficients, and α+β+γ=1.

[0108] Iterator module:

[0109] The optimal array layout parameters are generated through PSO iterative optimization.

[0110] In this application, the cloud diagram of the optimized array at different source frequencies is calculated through the embodiment, and the main lobe width and the maximum side lobe height are as follows: Figure 6 The optimized array demonstrates significant performance advantages over other arrays at all frequencies. In particular, within the source frequency range below 6 kHz, the maximum sidelobe height drops below -5 dB, surpassing other arrays.

[0111] The embodiment proposed in this application has been optimized in the microphone array structure to achieve higher sound source localization accuracy; at the same time, it can also improve the signal-to-noise ratio during the monitoring process. In addition, this embodiment enhances the fault tolerance of the hardware. In a traditional array, damage to a single microphone may cause the positioning error to increase by up to 25%. In contrast, even if any microphone in the optimized array fails, the impact is limited to 1 to 2 sub-arrays, and the other sub-arrays can still provide valid data.

[0112] In order to obtain the optimal microphone array layout parameters, in some specific embodiments of the present application, the optimal array layout parameters can be generated through iterative optimization, and the following steps can be adopted, such as Figure 5 As shown:

[0113] (1) Define the particle position and velocity equations as follows:

[0114] ;

[0115] Where c1 and c2 are two positive constants (usually c1=c2=1.5); r1 and r2 are two random functions in the range of (0, 1), which introduce randomness to avoid local optimality; w is the inertia weight, usually between 0.4 and 0.9, and is set to 0.8 in this application; p id Best previous position, g dRepresents the global best position. The best previous position represents the best position that the i-th particle has ever reached during the iteration process (i.e., the position where the particle has historically achieved the best objective function value); the global best position represents the individual historical best position (p) of all particles during the iteration process. id ). That is, if a particle’s p id The corresponding objective function value is the best in the current group, then the p id Will become the g of the entire group d .

[0116] (2) Select parameters suitable for the PSO algorithm.

[0117] The parameters that need to be set include the population size and particle dimension that affect the convergence speed, as well as the reasonable range to judge whether they have crossed the boundary. The number of particles is set to 30, the dimension of each particle is 2×64=128, and the size of each array element is ( x , y ) coordinates, and 64 is the number of array elements.

[0118] (3) Initialize the particle position X i , speed V i and the particle's previous best position P ibest .

[0119] According to the coordinate constraints mentioned above, 30 initial array layouts, i.e., 30 particles, are randomly generated.

[0120] The initial velocity of each particle is set to 0 or a random small value Each particle is initialized with its random position and random velocity to start its optimization process. The initial position defaults to the initial previous best position P ibest。

[0121] (4) Evaluate the fitness of particles

[0122] For each particle, its performance indicators are calculated: main lobe width, maximum side lobe height and array element spacing uniformity.

[0123] Specifically, the main lobe width is calculated, including:

[0124] First, generate the beam pattern:

[0125] ;

[0126] in, N : Number of array elements; w n :No. n The weighting coefficient of each array element; dn :No. n The position coordinates of each array element;k =2 π / λ :wavenumber( λ is wavelength); u ( θ ): direction vector

[0127] Then, locate the main lobe peak:

[0128] Find the maximum value of the directional pattern B max =max B ( θ ), the corresponding angle is θ peak .

[0129] Next, determine the half-power point:

[0130] Search on both sides of the main lobe to find B ( θ )≥ B max / Angle range (power drops by half) θ 1 and θ 2

[0131] Finally, the main lobe width is obtained:

[0132] ;

[0133] Specifically, the maximum sidelobe height is calculated as the ratio of the maximum sidelobe amplitude after excluding the main lobe to the main lobe.

[0134] Specifically, the calculation of array element spacing uniformity includes:

[0135] First, calculate the distance between adjacent elements:

[0136] Calculate the spacing between adjacent elements along the spiral arm d i

[0137] ;

[0138] Next, statistical uniformity indicators:

[0139] Use standard deviation to calculate the degree of dispersion of adjacent intervals

[0140] ;

[0141] Standard Deviation: ;

[0142] (5) Update P ibest and G best

[0143] Calculate the fitness of each particle according to the fitness function and update P ibest and G best Then return a fitness value to assign to the current position.

[0144] Specifically, if the calculated new fitness is less than the fitness of the particle at the previous best position, the previous best position is set as the current position and its fitness is updated. Otherwise, the previous best position and fitness are retained. Repeat the above operation for all particles.

[0145] Compare the fitness values ​​of all particles at the current previous best position and the global best position G best The fitness of the smallest one is updated as the current global optimal position G best .

[0146] (6) Update speed and position

[0147] The velocity of a moving particle is a real-valued vector that controls the direction and speed of the particle. best The velocity of the impact will determine the next position of the particle. Update the velocity and position of each particle using the particle position and velocity equations defined above.

[0148] (7) Determine whether the termination conditions are met:

[0149] The search terminates if the following conditions are met: the number of iterations reaches the maximum allowed number or the fitness value meets certain conditions. Otherwise, return to step 4.

[0150] For example, the termination condition is set as: reaching the maximum number of iterations of 150, or G iterations of 20 consecutive times. best Change Δ F <0.01.

[0151] (8) Output the optimal result

[0152] Get the optimal array layout G best The corresponding array element coordinates, spiral parameters (a, b), and optimized performance index values, including main lobe width, side lobe height and uniformity.

[0153] In order to realize the collection, storage and transmission of sound signals, in the specific implementation of the present application, the signal collection and transmission module is composed of a microphone array, an FPGA core board and a data storage and transmission module.

[0154] For example, the FPGA core board uses a Cyclone IV series chip, and a 245 logic chip controls the MEMS microphone array to collect sound signals. The distance between the microphone and the dry-type transformer is set to 1.5 meters, the sampling rate is 48kHz, and the signal acquisition time is 10 seconds.

[0155] Sound pressure levels were recorded using an Aiwa AWA-5636-0 sound level meter. Sound pressure level is a logarithmic measurement of the instantaneous pressure of a sound wave propagating through air relative to a reference sound pressure (typically the threshold of human hearing), expressed in decibels (dB). The meter records the sound pressure intensity at a distance of 1.5 meters from the transformer fault source for subsequent signal processing and fault diagnosis. The sound level meter measures the sound pressure level range under normal and fault conditions at a distance of 1.5 meters to establish a fault diagnosis threshold. The collected sound signal data is stored in RAM and then transferred to a computer via USB for processing.

[0156] In order to accurately locate the fault sound source, in some specific implementations of the present application, the signal processing module includes a Fourier transform submodule, an extraction submodule, a Music submodule and a superposition submodule.

[0157] In the Fourier transform submodule, the sound signal is converted from the time domain to the frequency domain using Fourier transform to obtain the frequency domain sound signal.

[0158] Specifically, the mathematical expression of the sound signal collected by the microphone in the time domain is:

[0159] ;

[0160] Where s(t) represents the signal received by the array, s0(t) represents the source signal, and n(t) represents the sum of all ambient noise and microphone background noise. α represents the steering vector of the array, as shown in the following formula:

[0161] ;

[0162] f is the frequency of the sound signal, v is the speed of sound in air, d M is the distance from the Mth microphone to the sound source, which is calculated from the coordinates of the optimized microphone array. j is the imaginary unit, and M is the total number of microphones in the array. T is the transpose operator.

[0163] After being transformed into the frequency domain through Fourier transform, the sound signal is represented as , w Refers to angular frequency, which characterizes the rate of phase change of the frequency domain component.

[0164] The sound signals collected on site contain various information, such as the target sound source, environmental noise, microphone background noise, etc. This reduces the positioning accuracy of existing algorithms.

[0165] The sound signals of different defects have specific effective characteristic frequency bands. Therefore, in the extraction submodule, the target frequency band required for each defect is intercepted, and the remaining frequency bands are not involved in the imaging calculation, thereby achieving the purpose of suppressing interference.

[0166] Specifically, the target frequency band refers to the frequency range of acoustic signals strongly associated with transformer defects (such as partial discharge and loose windings). Due to differences in the physical mechanisms of different defects, the frequency distribution of acoustic waves generated by the vibration or discharge process may vary. The target frequency band can be determined by combining experimental data with signal analysis. When a transformer fault occurs within this frequency band, the signal amplitude in this frequency band increases. Specifically, a bandpass filter or frequency domain mask can be used to extract the target frequency band.

[0167] In the MUSIC submodule:

[0168] The frequency domain sound signal contains all frequency bands, and the target frequency band containing the sound signal must be extracted from it. The extracted frequency domain signal is marked as s g (ω).

[0169] The covariance matrix calculation formula is as follows:

[0170] ;

[0171] Perform eigenvalue decomposition on the covariance matrix to obtain the noise subspace. The covariance matrix can be decomposed into two parts related to the signal and noise, as shown below:

[0172] ;

[0173] in U s is the subspace composed of eigenvectors corresponding to large eigenvalues, i.e., the signal subspace. U N It is a subspace composed of eigenvectors corresponding to small eigenvalues, that is, the noise subspace.

[0174] The spectrum estimation equation of the MUSIC algorithm is:

[0175] ;

[0176] Where: a is the direction vector of the array, The MUSIC algorithm locates the fault by traversing the candidate locations in the spatial spectrum and finding the peak points.

[0177] In the superposition submodule, considering that microphone damage may reduce positioning accuracy, a sub-array superposition strategy is adopted to further optimize the imaging algorithm.

[0178] The 64 microphones are divided into four non-overlapping groups, each containing 16 microphones. Three groups are used to form a subarray.

[0179] When performing acoustic imaging calculations, in addition to imaging the original 64-element main array, it is also necessary to image the 48-element sub-array.

[0180] The power values ​​of the main array and sub-array are normalized and summed to form the final acoustic cloud map.

[0181] The above-mentioned embodiments of the present application can effectively improve the accuracy and imaging effect of transformer fault sound source positioning, especially in a low signal-to-noise ratio environment, and can effectively reduce the sidelobe effect and improve the accuracy and reliability of sound source positioning.

[0182] To achieve more comprehensive and intuitive fault diagnosis results, some specific implementations of this application utilize a camera mounted at the center of the array to capture optical images. Using the SIFT feature point matching algorithm, the acoustic energy cloud map and the optical image are spatially aligned, maintaining an error tolerance of less than 0.5 pixels. Dynamic adjustments are made based on the acoustic energy amplitude, achieving acoustic imaging visualization that integrates multimodal information.

[0183] Specifically, in the multimodal acoustic imaging visualization process, the fusion of the acoustic energy cloud map and the optical image uses a pixel-based weighted fusion method, and the formula is as follows:

[0184] ;

[0185] I acoustic( x , y ) represents the acoustic energy cloud at coordinate ( x , y ) is the normalized energy value at ; I optical( x , y ) represents the optical image at coordinates ( x , y ) pixel value at ; w ( x , y ) is the dynamic weight coefficient, which is determined by the acoustic energy amplitude E ( x , y )Decide.

[0186] ;

[0187] inE min, E Max is the minimum and maximum energy value in the current acoustic cloud image. E The threshold is set to 3 times the mean value of the background noise.

[0188] Weight w ( x , y ) Follow the following rules to adjust:

[0189] when E ( x , y )> E When the threshold is reached, the weight is increased accordingly, and the weight increases rapidly as the energy increases.

[0190] when E ( x , y )≤ E When the threshold is reached, the weight is reduced, and the weight decays rapidly as the energy decreases.

[0191] The above-mentioned embodiments of the present application realize the fusion of multimodal information and provide more comprehensive and intuitive fault diagnosis results.

[0192] Based on the same technical concept, in other embodiments of the present application, a transformer fault sound source imaging and positioning method is provided, such as Figure 7 As shown, including:

[0193] S100, using N array elements arranged at equal phase intervals along M spiral arms to form a microphone array;

[0194] S200, using a particle swarm optimization algorithm to perform adaptive topology optimization on the microphone array, with the coordinates of each array element being constrained to not exceed half of the aperture of the microphone array, and with the weighted sum of the main lobe width, the maximum side lobe height, and the uniformity of array element spacing of the microphone array as the particle fitness, to obtain a microphone array with an optimal array layout;

[0195] S300, using a microphone array to collect sound signals and transmit the sound signals;

[0196] S400, receiving the transmitted sound signal, performing fast Fourier transform time-frequency analysis and multiple signal classification, and dividing N array elements into S groups to obtain The coherent superposition of the power spectra of the sub-arrays improves the spatial spectrum estimation of the multiple signal classification and obtains an acoustic energy cloud map for fault sound source location.

[0197] The above-mentioned embodiments of this application address the pain points of existing technologies and construct a highly accurate and robust transformer fault location method. By optimizing the microphone array structure and improving the signal processing algorithm, the accuracy and imaging effect of sound source localization are improved. This effectively reduces the impact of side lobes, particularly in low signal-to-noise ratio environments, and improves the accuracy and reliability of sound source localization. Furthermore, through optical-acoustic fusion imaging technology, multimodal information fusion is achieved, providing more comprehensive and intuitive fault diagnosis results.

[0198] The specific implementation technologies of the modules / units in the above examples of the present application can refer to the corresponding modules / sub-modules of the transformer fault sound source imaging and positioning system in the above embodiments, which will not be repeated here.

[0199] Based on the same technical concept, in other embodiments of the present application, a terminal includes a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the program, the terminal can be used to run the system or execute the method.

[0200] Based on the same technical concept, in other embodiments of the present application, a computer-readable storage medium stores a computer program, which can be used to execute a method or run a system when executed by a processor.

[0201] Optionally, the memory is used to store programs. The memory may include volatile memory (volatile memory), such as random-access memory (RAM), such as static random-access memory (SRAM) and double data rate synchronous dynamic random access memory (DDR SDRAM). The memory may also include non-volatile memory (non-volatile memory), such as flash memory. The memory is used to store computer programs (such as applications and functional modules that implement the above-mentioned methods), computer instructions, etc. These computer programs and computer instructions may be partitioned and stored in one or more memories. Furthermore, these computer programs, computer instructions, data, etc. can be accessed by the processor.

[0202] The aforementioned computer programs, computer instructions, etc. may be partitioned and stored in one or more memories, and the aforementioned computer programs, computer instructions, data, etc. may be called by a processor.

[0203] The processor is configured to execute the computer program stored in the memory to implement the various steps of the method involved in the above embodiment. For details, please refer to the relevant description in the above method embodiment.

[0204] The processor and memory can be independent structures or integrated structures. When the processor and memory are independent structures, the memory and processor can be coupled via a bus.

[0205] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of an entirely hardware embodiment, an entirely software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0206] The present application is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0207] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0208] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operational steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing the instructions executed on the computer or other programmable device for implementing the process. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0209] The above describes some specific embodiments of the present application. It should be understood that the present application is not limited to the specific embodiments described above, and those skilled in the art may make various variations or modifications within the scope of the claims, which do not affect the substantive content of the present application. The above preferred features may be used in any combination as long as they do not conflict with each other.

Claims

1. A transformer fault sound source imaging and positioning system, characterized in that: include: The microphone array module consists of N array elements arranged at equal phase intervals along M spiral arms; an array optimization module, which uses a particle swarm optimization algorithm to perform adaptive topology optimization on the microphone array, with the coordinates of each array element being constrained to not exceed half the aperture of the microphone array, and with the weighted sum of the main lobe width, maximum side lobe height, and uniformity of array element spacing of the microphone array as the particle fitness, to obtain a microphone array with an optimal array layout; A signal acquisition and transmission module, which uses the microphone array with the optimal array layout to acquire sound signals and transmit the sound signals; The signal processing module receives the transmitted sound signal, performs fast Fourier transform time-frequency analysis and multiple signal classification, and obtains the sound signal by dividing N array elements into S groups. The power spectra of the sub-arrays are coherently superimposed to improve the spatial spectrum estimation of the multiple signal classification, and an acoustic energy cloud map is obtained for fault sound source location; The microphone array module uses 64 array elements to achieve balanced main lobe width and side lobe level; The 64 array elements are arranged at equal phase intervals along the 8 spiral arms, and the operating frequency band is 2000 Hz; The angle between adjacent spiral arms is 45 degrees; The polar coordinate equation of the i-th array element is: r i (θ)=aθ i +b; Where a is the spiral growth rate, b is the initial radius offset, θ i is the polar angle of the i-th array element, 1≤i≤64; The phase difference between adjacent elements of a single spiral arm satisfies: Δφ=2πk / N ; Where N=64 is the total number of array elements, k=8 is the number of spiral arms; The array optimization module includes: Position determination submodule: determines the position of the 8 array elements on a single spiral arm, i The position of each element is: ； ； Coordinate constraint submodule: Determine the i The array element in k The coordinate constraints on the spiral arms are: ； ； D is the array aperture; Fitness definition submodule: Establish the fitness function definition as follows: F=α·main lobe width+β·maximum side lobe height+γ·uniformity of array element spacing; Where α, β, and γ are weighting coefficients, and α + β + γ = 1. Array element spacing uniformity is an indicator that measures the uniformity of microphone distribution in the array. It directly affects the array's beamforming performance and refers to the degree of discreteness of the spacing between adjacent array elements. Iterative optimization submodule: Iterative optimization through particle swarm method to generate optimal array layout parameters; The calculation of array element spacing uniformity includes: First, calculate the distance between adjacent array elements: Calculate the spacing between adjacent elements along the spiral arm d i : ； Next, statistical uniformity indicators: Use standard deviation to calculate the degree of dispersion of adjacent intervals; ； Standard Deviation: .

2. A transformer fault sound source imaging and positioning system according to claim 1, characterized in that: The iterative optimization submodule generates optimal array layout parameters through iterative optimization using a particle swarm optimization method, including: Setting parameters of the particle swarm method, including population size, particle dimension, and respective threshold ranges, wherein each particle represents a microphone array distribution; Randomly generate multiple particles and initialize the position X of each particle i , speed V i , get the particle's previous optimal position P ibest ; For each particle, calculate its performance indicators, including the main lobe width, the maximum side lobe height, and the uniformity of the array element spacing, to obtain the fitness of each particle; According to the fitness and the number of iterations, it is determined whether the termination condition is met. If so, the search is terminated. Otherwise, the previous best position P of the particle is updated. ibest , calculate the global optimal position G best , update the speed and position of all particles; return to calculate the fitness of particles; The global optimal position G of the last iteration best In order to finally obtain the optimal array layout, obtain G best The corresponding array element coordinates and spiral parameters (a, b) are used as the optimal array layout parameters.

3. A transformer fault sound source imaging and positioning system according to claim 1, characterized in that: The signal processing module includes: Fourier transform submodule: converting the sound signal from the time domain to the frequency domain using Fourier transform to obtain a frequency domain sound signal; Extraction submodule: extracting a target frequency band containing a sound signal from the frequency domain sound signal; Music submodule: performs spectrum estimation on the target frequency band to obtain the noise subspace and fault location; The superposition submodule divides the N microphone main array into S non-overlapping groups, from which all S-1 groups are selected to form S sub-arrays; The main array and all the sub-arrays are imaged, and all the images are summed to form a final acoustic cloud map.

4. A transformer fault sound source imaging and positioning system according to claim 3, characterized in that: The superposition submodule divides the 64 microphones into four non-overlapping groups, each group containing 16 microphones; three groups are used to form a subarray; When performing acoustic imaging calculations, in addition to imaging the original 64-element main array, it is also necessary to image the 48-element sub-array; The main array and sub-array power values ​​are normalized and summed to form the final acoustic cloud map.

5. The transformer fault sound source imaging and positioning system according to claim 1, characterized in that: It also includes a photoacoustic fusion imaging module, which collects optical images, and spatially aligns the acoustic energy cloud map with the optical image based on the SIFT feature point matching algorithm, and outputs a multimodal fusion visualization result.

6. A transformer fault sound source imaging and positioning method based on the system according to any one of claims 1 to 5, characterized in that: include: A microphone array is formed by arranging N array elements at equal phase intervals along M spiral arms; With the constraint that the coordinates of each array element do not exceed half of the aperture of the microphone array, and the weighted sum of the main lobe width, the maximum height of the side lobe, and the uniformity of the array element spacing of the microphone array as the particle fitness, the particle swarm optimization algorithm is used to perform adaptive topology optimization on the microphone array to obtain a microphone array with an optimal array layout; Collecting sound signals using the microphone array with the optimal array layout, and transmitting the sound signals; Receive the transmitted sound signal, perform fast Fourier transform time-frequency analysis and multiple signal classification, and divide N array elements into S groups to obtain The coherent superposition of the power spectra of the sub-arrays improves the spatial spectrum estimation of the multiple signal classification and obtains an acoustic energy cloud map for fault sound source location.

7. A terminal comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When executing the program, the processor can be used to run the system according to any one of claims 1 to 5, or to execute the method according to claim 6.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, it can be used to perform the method according to claim 6, or run the system according to any one of claims 1 to 5.

Images