Power equipment fault sound source positioning system and method
By combining microphone array components with VMD and C-Beamforming algorithms to select the optimal sub-bands, and combining MVDR beamforming with LSTM neural networks, the problem of accurately locating the fault sound source of power equipment in complex noise environments was solved, improving the robustness and accuracy of the positioning system.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-12-01
- Publication Date
- 2026-03-10
AI Technical Summary
In complex acoustic environments, the sound signals of power equipment faults are submerged by strong background noise, causing a sharp drop in the signal-to-noise ratio of traditional acoustic detection methods, resulting in insufficient positioning accuracy and a high misjudgment rate, which affects practical engineering applications.
By employing a microphone array component combined with variational mode decomposition (VMD) and C-beamforming algorithms to select the optimal sub-band, and combining MVDR beamforming with LSTM neural networks, the system achieves accurate localization and distance estimation of fault sound sources, shields interference noise, and improves the robustness of the positioning system.
It significantly improves the ability to identify and locate fault sound sources in complex noise environments, optimizes the clarity of sound source maps, and enhances the engineering application value of the system.
Smart Images
Figure CN121633995A_ABST
Abstract
Description
Technical Field
[0001] This application belongs to the field of sound source localization technology, and in particular relates to a sound source localization system and method for power equipment faults. Background Technology
[0002] In power equipment condition monitoring and fault diagnosis, acoustic detection methods are increasingly becoming an important means of identifying defects such as partial discharge due to their advantages such as non-contact operation and resistance to electromagnetic interference. However, this method faces significant challenges in practical field applications: the strong background noise prevalent in complex acoustic environments (such as equipment operation, environmental wind and rain, traffic horns, and other equipment vibrations) often drowns out or severely interferes with the weak sound source signals generated by power equipment faults, such as... Figure 2 The interference noise waveform diagram is shown. Figure 3 The waveform of the fault sound source is shown. This sharp drop in signal-to-noise ratio not only makes it extremely difficult for traditional acoustic detection methods to extract and identify fault signals, but also directly causes insufficient sound source localization accuracy and an increased rate of misjudgment of fault features, seriously restricting the reliability and widespread application of this technology in practical engineering. Summary of the Invention
[0003] To address the problems in the background art, this application provides a power equipment fault sound source localization system and method, which realizes accurate location and distance estimation of fault sound sources in complex noise environments, and improves the robustness and engineering practicality of the localization system.
[0004] The technical solution to the technical problem solved in this application is as follows:
[0005] According to one aspect of this application, a power equipment fault sound source localization system is provided, comprising a microphone array assembly, an industrial camera, and a signal processing module; the microphone array assembly includes a microphone bracket on which multiple MEMS microphones are uniformly arranged to form a microphone array for acquiring a mixed signal of fault sound source and interference noise; the industrial camera is located at the center of the microphone bracket; the signal processing module is configured to perform the following operations: decompose the microphone acquired signal using variational mode decomposition (VMD), select the optimal sub-band using envelope spectrum kurtosis as an index, process the optimal sub-band signal using the C-Beamforming algorithm to obtain a sound source map, thereby realizing fault sound source enhancement and localization.
[0006] Preferably, the signal processing module acquires multiple IMF components during variational mode decomposition (VMD) and selects the IMF component with the largest envelope spectrum kurtosis as the optimal subband.
[0007] Preferably, the system further includes processing the optimal sub-band signal using the Beamforming algorithm to obtain a sound source map, and then using an image overlay method to combine the sound source maps obtained by the C-Beamforming algorithm and the Beamforming algorithm, so that C-Beamforming can shield interference noise on the basis of Beamforming, thereby making the result clearer.
[0008] Preferably, the microphone array is a circular array or a diagonal double rectangular array.
[0009] Preferably, the signal processing module further includes a distance estimation unit, which adopts a distance estimation model based on MVDR beamforming and LSTM neural network. The MVDR beamforming combines the signals collected by the microphone array, suppresses signals from non-selected directions, and enhances signals from selected directions, thereby enabling focused sound pickup in a specified direction and improving the signal-to-noise ratio of the received signal. The LSTM neural network estimates the distance to the fault sound source based on the signal processed by MVDR.
[0010] According to another aspect of this application, a method for locating the sound source of a power equipment fault is provided, which includes the following steps:
[0011] S1. A microphone array is used to collect mixed audio signals under power equipment fault scenarios. The mixed audio signals include fault sound source signals and background noise interference signals.
[0012] S2. Perform VMD decomposition on the mixed audio signal to obtain multiple IMF components, calculate the envelope kurtosis of each IMF component, and select the optimal subband.
[0013] S3. The C-Beamforming algorithm is used to perform beamforming processing on the optimal sub-band signal to generate a sound source map and realize the directional location of the fault sound source.
[0014] S4. Establish a distance estimation model based on MVDR beamforming and LSTM neural network. Process the optimal sub-band signal through MVDR beamforming algorithm, extract signal features and input them into LSTM neural network, and output the distance estimation result of the fault sound source.
[0015] Preferably, the microphone array in step S1 is a circular array or a diagonal double rectangular array.
[0016] Preferably, in step S2, the optimal subband is the IMF component with the largest envelope spectrum kurtosis value.
[0017] Preferably, step S3 further includes processing the optimal sub-band signal using the Beamforming algorithm to obtain a sound source map, and then using an image overlay method to combine the sound source maps obtained by the C-Beamforming algorithm and the Beamforming algorithm, so that C-Beamforming can shield interference noise on the basis of Beamforming and optimize the clarity of the sound source map.
[0018] Compared to existing technologies, the power equipment fault sound source localization system and method described in this application adopts a technical solution combining VMD and C-Beamforming. Through adaptive extraction of optimal sub-bands and interference shielding mechanisms, it significantly improves the identification capability of fault sound sources in complex noise environments, and the sound source image clarity is superior to the traditional Beamforming algorithm. At the same time, the sound source images obtained by the Beamforming algorithm and the C-Beamforming algorithm are superimposed to optimize the sound source image clarity. By fusing MVDR beamforming and LSTM neural networks, it achieves synergistic optimization of azimuth localization and distance estimation, solving the problem of insufficient ranging accuracy under low signal-to-noise ratio. By optimizing the main lobe suppression and localization robustness through the preferred diagonal double rectangular array configuration, it provides the optimal solution for the arrangement of microphone arrays in practical engineering and enhances the engineering application value of the system. Attached Figure Description
[0019] Figure 1 This is a schematic diagram of a power equipment fault sound source localization system according to this application;
[0020] Figure 2 This is a waveform diagram of interference noise;
[0021] Figure 3 The waveform diagram of the fault sound source;
[0022] Figure 4 For acquiring signal waveform diagrams;
[0023] Figure 5 This is a schematic diagram of VMD decomposition subbands;
[0024] Figure 6 A three-dimensional sound source diagram of Beamforming;
[0025] Figure 7 A three-dimensional sound source diagram of C-Beamforming;
[0026] Figure 8 This is a sound source diagram for Beamforming;
[0027] Figure 9 C-Beamforming sound source diagram;
[0028] Figure 10-11Simulation diagrams of various microphone arrays.
[0029] In the diagram: 1. Microphone array assembly; 11. Microphone stand; 12. MEMS microphone; 13. Industrial camera; 2. Computer; 3. Fault noise playback speaker; 4. Interference noise playback speaker. Detailed Implementation
[0030] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those of ordinary skill in the art without creative effort are within the scope of protection of this application. Unless otherwise defined, the technical or scientific terms used herein should have the ordinary meaning understood by those skilled in the art to which this disclosure pertains. Any aspects not detailed in this application are well-known technologies to those skilled in the art.
[0031] Example 1:
[0032] This embodiment provides a power equipment fault sound source localization system. Taking a laboratory simulation as an example, the power equipment fault sound source localization system includes a microphone array assembly 1, an industrial camera 13, and a signal processing module. The microphone array assembly 1 includes a microphone bracket 11, on which multiple MEMS microphones 12 are uniformly arranged to form a microphone array. A microphone array is an advanced technology for capturing sound signals. It consists of multiple miniature microphones and can receive sound signals from different directions. The microphone array provides a more accurate and clearer sound signal than a single microphone because it can suppress ambient noise and echoes and can focus on receiving sounds from different directions. In this example, it is used to acquire a mixed signal of fault sound source and interference noise. The fault sound source and interference noise signals can be pre-recorded and played back during signal acquisition by setting fault noise playback speaker 3 and interference noise playback speaker 4. The industrial camera 13 is set at the center of the microphone bracket 11 to acquire visual data. The microphone array is preferably a circular array or a diagonal double rectangular array. Taking a circular array as an example, the microphone bracket 11 is set as a circle with a diameter of 26cm. Multiple MEMS microphones 12 are evenly arranged on the edge of the microphone bracket 11. The distance between the microphone array assembly 1 and the fault noise playback speaker 3 and interference noise playback speaker 4 is 1m.
[0033] The fault noise signal, interference noise signal, and mixed signal of fault sound source and interference noise can be collected by the microphone array component 1, respectively. The waveforms of the three signals are shown in the figure below. Figure 2 , 3As shown in Figure 4, it can be seen that due to the influence of interference noise, the interference sound with a larger sound intensity is highlighted, while the fault sound source is submerged.
[0034] To process the signals acquired by the microphone array, a computer 2 is set up, and a signal processing module is installed in the computer 2. Each MEMS microphone 12 and industrial camera 13 is connected to the computer 2 via a line. The signal processing module is configured to perform the following operations: decompose the microphone acquired signals using variational mode decomposition (VMD), such as... Figure 5 As shown, the optimal sub-band is selected using envelope kurtosis as an indicator. The optimal sub-band signal is then processed using the C-Beamforming algorithm to obtain a source map, thereby achieving fault source enhancement and localization. Figure 7 and Figure 9 As shown.
[0035] In this embodiment, the signal processing module acquires multiple IMF components during variational mode decomposition (VMD), such as... Figure 5 As shown in Table 1, after VMD decomposition of the signal, three IMFs are obtained. Among the three IMF components, the IMF2 component with the largest envelope spectrum kurtosis is selected as the optimal subband.
[0036] Table 1. Kurtosis of envelope spectrum for each component
[0037] In a preferred embodiment, the system further includes processing the optimal sub-band signal using a Beamforming algorithm to obtain a sound source map, such as... Figure 6 and Figure 8 As shown, the source map obtained by the C-Beamforming algorithm highlights the sound source location more clearly than that obtained by the Beamforming algorithm. To make the sound source more intuitive to observe, the source maps obtained by the C-Beamforming and Beamforming algorithms are overlaid using an image overlay method. This allows C-Beamforming to filter out interference noise on top of Beamforming, thus making the result clearer.
[0038] In a preferred embodiment, the signal processing module further includes a distance estimation unit. The distance estimation unit adopts a distance estimation model based on MVDR beamforming and LSTM neural network. The MVDR beamforming combines the signals collected by the microphone array, suppresses signals from non-selected directions, and enhances signals from selected directions, thereby enabling focused sound pickup in a specified direction and improving the signal-to-noise ratio of the received signal. The LSTM neural network estimates the distance to the fault sound source based on the signal processed by MVDR.
[0039] Example 2:
[0040] This application provides a method for locating the sound source of power equipment faults, based on the system of Embodiment 1, including the following steps:
[0041] S1. A microphone array is used to collect mixed audio signals under power equipment fault scenarios. The mixed audio signals include fault sound source signals and background noise interference signals.
[0042] S2. Perform VMD decomposition on the mixed audio signal to obtain multiple IMF components, calculate the envelope kurtosis of each IMF component, and select the IMF component with the largest envelope kurtosis value as the optimal subband.
[0043] S3. The C-Beamforming algorithm is used to perform beamforming processing on the optimal sub-band signal to generate a sound source map and realize the directional location of the fault sound source.
[0044] In a preferred embodiment, this step further includes processing the optimal sub-band signal using the Beamforming algorithm to obtain a sound source map, and then using an image overlay method to combine the sound source maps obtained by the C-Beamforming algorithm and the Beamforming algorithm, so that C-Beamforming can shield interference noise on the basis of Beamforming and optimize the clarity of the sound source map.
[0045] S4. Establish a distance estimation model based on MVDR beamforming and LSTM neural network. Process the optimal sub-band signal through MVDR beamforming algorithm, extract signal features and input them into LSTM neural network, and output the distance estimation result of the fault sound source.
[0046] The Minimum Variance Distortionless Response (MVDR) beamforming algorithm is based on the minimum mean square error criterion. It minimizes the total output energy of the beamformer while keeping the target direction gain constant, thus minimizing the output interference and noise power and suppressing interference and noise signals to recover the target speech.
[0047] The specific process of MVDR beamforming is as follows:
[0048] Assuming the desired target position P' in space is (r0, θ0), and the interference signal ij (j=1,2, Given that the interference position is (rj, θj) and the array element noise is n(t), the formula for the signal of the nth array element at the receiving end is as follows:
[0049]
[0050] In the above formula, a(θ) is the receiving steering vector from the (r,θ) direction, and the formula for a(θ) is as follows:
[0051]
[0052] In the above formula, under the constraint w"a(0)=1, the result with the minimum noise is calculated, and the optimized objective function is formulated as follows:
[0053]
[0054] The MVDR weight optimization problem is represented in the formula as follows:
[0055]
[0056] The essence of MVDR beamforming is to solve for the weight coefficients of each array element. The Lagrange multiplier method is used to solve this problem, and the formula is as follows:
[0057]
[0058] After differentiating the formula, when the result is 0, the formula is as follows:
[0059]
[0060] The optimal values of the array weights are determined according to the MVDR criterion, as shown in the following formula:
[0061]
[0062] In a preferred embodiment, the microphone array in step S1 is a circular array or a diagonal double rectangular array. The array configuration refers to the arrangement of the microphones; different configurations can adapt to different application scenarios. Figure 10 (ac) shows the schematic diagram and two-dimensional / three-dimensional positioning diagram of the diagonal double rectangular array. Figure 10 (df) shows the linear array diagram and the two-dimensional and three-dimensional positioning diagram. Figure 10 (gi) shows the schematic diagram of the circular array and the two-dimensional and three-dimensional positioning diagrams. Figure 11 (jl) shows the schematic diagram of the cross array and the two-dimensional and three-dimensional positioning diagram. Figure 11(mo) shows the schematic diagram of the rectangular array and the two-dimensional and three-dimensional positioning diagrams. It can be seen that the diagonal double rectangular array best approximates the true sound source location and has the best suppression effect on the side lobes. The uniform straight array has narrower main lobes, but a larger main lobe area, resulting in lower accuracy. Although it can accurately locate the main sound source, interference sources appear in other locations, leading to unsatisfactory suppression of the side lobes and errors. The cross-shaped array is relatively uniform, and the linear array has a more prominent main lobe, but inaccurate positioning still exists. Due to the excessive number of main lobes, the rectangular array has the most significant suppression effect on the side lobes, but the local main lobe area is too large to achieve precise positioning. Therefore, the proposed double rectangular diagonal array has the best robustness and is most suitable as a microphone array for fault distance models.
Claims
1. A power equipment fault sound source positioning system, characterized in that, The microphone array assembly, the industrial camera and the signal processing module are included; the microphone array assembly includes a microphone support, a plurality of MEMS microphones are uniformly arranged on the microphone support, forming a microphone array for collecting mixed signals of fault sound sources and interference noises; the industrial camera is arranged at the center of the microphone support; the signal processing module is configured to perform the following operations: using variational mode decomposition (VMD) to decompose the microphone collected signals, using envelope spectrum kurtosis as an index to select the optimal sub-band, processing the optimal sub-band signals through a C-Beamforming algorithm to obtain a sound source map, realizing fault sound source enhancement and positioning.
2. A power equipment fault sound source positioning system according to claim 1, characterized in that, The signal processing module obtains a plurality of IMF components when performing variational mode decomposition (VMD), and selects the IMF component with the maximum envelope spectrum kurtosis as the optimal sub-band.
3. The power equipment fault sound source positioning system of claim 1, wherein, It also includes processing the optimal sub-band signals through a Beamforming algorithm to obtain a sound source map, and using an image superposition method with the sound source maps obtained by the C-Beamforming algorithm and the Beamforming algorithm, so that the C-Beamforming algorithm shields interference noises on the basis of the Beamforming algorithm, and the result is more clear.
4. The power equipment fault sound source positioning system of claim 1, wherein, The microphone array is a circular array or a diagonal double rectangular array.
5. The power equipment fault sound source positioning system of claim 1, wherein, The signal processing module further includes a distance estimation unit, which uses a distance estimation model based on MVDR beamforming and LSTM neural network, the MVDR beamforming combines the signals collected by the microphone array, suppresses non-selected direction signals, enhances selected direction signals, and thus can realize focusing on a specified direction, improve the signal-to-noise ratio of the received signal, and the LSTM neural network realizes fault sound source distance estimation based on the signals processed by the MVDR.
6. A method of locating a sound source of a fault in an electrical power device, characterized in that The following steps are included: S1, using a microphone array to collect mixed audio signals in a power equipment fault scene, the mixed audio signals containing fault sound source signals and background noise interference signals; S2, performing VMD decomposition on the mixed audio signals to obtain a plurality of IMF components, calculating the envelope spectrum kurtosis of each IMF component, and selecting the optimal sub-band; S3, using a C-Beamforming algorithm to perform beamforming processing on the optimal sub-band signals to generate a sound source map and realize the azimuth positioning of the fault sound source; S4, establishing a distance estimation model based on MVDR beamforming and LSTM neural network, processing the optimal sub-band signals through the MVDR beamforming algorithm, extracting signal features and inputting them into the LSTM neural network, and outputting the distance estimation result of the fault sound source.
7. A method of locating a sound source of a fault in an electrical device according to claim 6, characterized in that, The microphone array in step S1 is a circular array or a diagonal double rectangular array.
8. A method of locating a sound source of a fault in an electrical device according to claim 7, characterized in that, The optimal sub-band in step S2 is the IMF component with the maximum envelope spectrum kurtosis.
9. A method of locating a sound source of a fault in an electrical device according to claim 8, characterized in that, In step S3, the optimal sub-band signal is processed by the Beamforming algorithm to obtain a sound source map, and the sound source map obtained by the C-Beamforming algorithm and the Beamforming algorithm is superimposed by an image superposition method, so that the C-Beamforming algorithm shields the interference noise on the basis of the Beamforming algorithm and optimizes the clarity of the sound source map.