Visual detection method, medium and system for sound signals of dry-type reactor

Through Bayesian probability inference and adaptive threshold amplification algorithm, the time spectrum diagram is processed, combined with dimensionality reduction and cluster analysis, visual classification images are generated, which solves the problem of slight changes in the sound signal of the dry reactor and improves fault diagnosis accuracy and reliability.

CN119007746AActive Publication Date: 2024-11-22ELECTRIC POWER RESEARCH INSTITUTE OF STATE GRID NINGXIA ELECTRIC POWER COMPANY

Patent Information

Application Number
CN202411021565.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-29
Publication Date
2024-11-22
Estimated Expiration
2044-07-29

AI Technical Summary

Technical Problem

The prior art is difficult to capture small signal changes when processing dry reactor sound signals, and cannot fully identify subtle fault characteristics, resulting in low fault diagnosis efficiency and reliability.

Method used

The time spectrum graph is processed using Bayesian probability inference and adaptive threshold amplification algorithm to highlight small changes, and map high-dimensional features to two-dimensional or three-dimensional space through dimensional reduction algorithms, and visual classification is generated by combining cluster analysis. image.

Benefits of technology

It improves the detection ability of micro signal changes, enhances the capture ability of fault signals, realizes intuitive fault diagnosis support, and improves the accuracy and reliability of fault diagnosis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119007746A_ABST
    Figure CN119007746A_ABST
Patent Text Reader

Abstract

The invention provides a visual detection method, medium and system for a sound signal of a dry-type reactor, and belongs to the technical field of sound signal detection of the dry-type reactor, and the method comprises the steps: firstly, collecting the sound signal of the dry-type reactor, and carrying out the preprocessing of the sound signal to eliminate the environment noise; and then carrying out time-frequency analysis on the preprocessed sound signal to obtain a time-frequency spectrogram. Next, highlighting tiny changes in the time-frequency spectrogram by adopting Bayesian probabilistic reasoning and self-adaptive threshold increasing and expanding modes to obtain an increased and expanded time-frequency spectrogram; then, extracting the characteristics of energy distribution, peak frequency, harmonic structure and the like from the time-increasing and time-expanding spectrogram, and combining into a multi-dimensional characteristic vector; and mapping the high-dimensional feature vector to a two-dimensional or three-dimensional space by applying a dimension reduction algorithm to obtain a second feature vector. And finally, performing clustering analysis on the second feature vector by using an unsupervised learning algorithm, distributing colors or labels for different categories according to a clustering result, and generating visual classification image output.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention belongs to the technical field of dry-type reactor sound signal detection, and in particular, relates to a visual detection method, medium and system for dry-type reactor sound signals. Background Art

[0002] Dry-type reactors are important passive power compensation devices in power systems and are widely used in the field of power transmission and distribution. They adjust the reactive power balance of the power grid by absorbing or releasing reactive power, thereby improving voltage quality, power factor and transmission efficiency of the system. Compared with traditional oil-immersed reactors, dry-type reactors have the advantages of small size, light weight, oil-free and maintenance-free, so they are the first choice in many occasions.

[0003] However, dry-type reactors generate certain sound noise during operation. These noise signals contain rich information about the operating status of the equipment and can be used for fault diagnosis and status monitoring. Existing diagnostic methods mainly rely on the subjective judgment of professionals on sound signals, which are inefficient and unreliable. Therefore, it is urgent to develop an automated and quantitative dry-type reactor sound signal analysis method to improve the accuracy and reliability of fault diagnosis.

[0004] In recent years, with the continuous advancement of signal processing and pattern recognition technology, some sound signal analysis methods based on time-frequency analysis and machine learning have been widely used. However, these methods still have limitations in processing small signal changes and cannot fully capture the subtle fault characteristics in the sound signal of dry-type reactors. Summary of the invention

[0005] In view of this, the present invention provides a visual detection method, medium and system for the sound signal of a dry-type inductor, which can solve the technical problem that the existing methods still have limitations in processing small signal changes and cannot fully capture subtle fault characteristics in the sound signal of the dry-type inductor.

[0006] The present invention is achieved in that:

[0007] A first aspect of the present invention provides a method for visually detecting a sound signal of a dry-type reactor, comprising the following steps:

[0008] S10, collecting sound signals of the dry-type reactor and performing preprocessing to eliminate environmental noise;

[0009] S20, performing time-frequency analysis on the pre-processed sound signal to obtain a time-frequency spectrum diagram;

[0010] S30, using Bayesian probability reasoning and adaptive threshold augmentation to highlight small changes in the time-frequency spectrum diagram to obtain an augmented time-frequency spectrum diagram;

[0011] S40, extracting features of the frequency spectrum during augmentation and expansion, including energy distribution, peak frequency, and harmonic structure;

[0012] S50, combining the extracted features into a multi-dimensional vector, recorded as the first feature vector;

[0013] S60, applying a dimensionality reduction algorithm to map the first feature vector from a high-dimensional feature vector to a two-dimensional or three-dimensional space to obtain a second feature vector;

[0014] S70, clustering the second feature vector;

[0015] S80: assign colors or labels to different categories according to the clustering result, and generate and output a visual classification image based on the second feature vector.

[0016] The adaptive threshold augmentation is calculated by using a plurality of equations, including a time-frequency spectrum normalization equation, a nonlinear contrast enhancement equation, an adaptive threshold calculation equation, a dynamic augmentation coefficient calculation equation, and a frequency-dependency adjustment equation.

[0017] Furthermore, the time-frequency spectrum normalization equation is specifically expressed as:

[0018]

[0019] In the formula, S norm (t,f) is the normalized time-frequency spectrum, S(t,f) is the original time-frequency spectrum; t is time, and f is frequency.

[0020] Wherein, the nonlinear contrast enhancement equation is specifically expressed as:

[0021]

[0022] Where C(t,f) is the spectrum after nonlinear contrast enhancement, tanh is the hyperbolic tangent function, γ is the contrast enhancement intensity parameter, μ local (t,f) is the weighted average in the local window centered at (t,f), σ local (t,f) is the weighted standard deviation in the local window centered at (t,f), ∈ is a small positive number to prevent the denominator from being zero, δ is the gradient enhancement parameter, is the gradient of the normalized spectrum at the point (t,f).

[0023] The adaptive threshold calculation equation is specifically expressed as:

[0024] T(t,f)=μ global +k·σ global ·(1-e -α·|C(t,f)|)·(1+λ·entropy(t,f));

[0025] Where T(t,f) is the adaptive threshold; μ global is the global average value of the entire spectrum; σ global is the global standard deviation of the entire spectrum; k is an adjustment parameter that controls the sensitivity of the threshold; α is a nonlinear mapping parameter that controls the rate at which the threshold changes with contrast; λ is an entropy weight parameter; entropy(t,f) is the normalized Shannon entropy of the local window centered at (t,f).

[0026] The dynamic expansion coefficient calculation equation is specifically expressed as:

[0027]

[0028] Where E(t,f) is the dynamic augmentation coefficient; β is the augmentation strength parameter, which controls the degree of augmentation; η is the sparsity weight parameter; sparsity(t,f) is the sparsity measure of the local window centered at (t,f).

[0029] The frequency-dependent adjustment equation is specifically expressed as:

[0030]

[0031] In the formula, S enhanced (t,f) is the final time-frequency spectrum after expansion; f max is the maximum frequency; φ is the frequency adjustment parameter.

[0032] Furthermore, the method for clustering the second feature vector is: k-means clustering or Gaussian mixture model clustering.

[0033] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are executed, they are used to execute the above-mentioned visual detection method of the sound signal of the dry-type reactor.

[0034] A third aspect of the present invention provides a visual detection system for a sound signal of a dry-type reactor, which comprises the above-mentioned computer-readable storage medium.

[0035] Compared with the prior art, the visual detection method, medium and system of the sound signal of a dry-type reactor provided by the present invention have the following beneficial effects:

[0036] 1. Enhance the ability to detect small signal changes. Through Bayesian probability reasoning and adaptive threshold augmentation algorithm, the small change characteristics in the time-frequency spectrum can be effectively highlighted, the ability to capture subtle fault signals can be improved, and the foundation for subsequent feature extraction and fault diagnosis can be laid.

[0037] 2. Realize the visualization analysis of multi-dimensional features. This method uses the dimensionality reduction algorithm to map the high-dimensional feature space to the two-dimensional or three-dimensional space, and uses unsupervised clustering to automatically identify different signal patterns and generate intuitive color classification images. This visualization helps to more intuitively understand the inherent laws of the dry-type reactor sound signal and provide convenient fault diagnosis support for operation and maintenance personnel.

[0038] 3. Good adaptability and scalability. This method adopts a modular design, and each processing step has strong flexibility and configurability. Different preprocessing, time-frequency analysis, feature extraction, dimensionality reduction and clustering algorithms can be selected and optimized according to the actual signal characteristics to meet the diagnostic needs in different scenarios.

[0039] 4. Improve the accuracy and reliability of fault diagnosis. By effectively capturing tiny signal changes and intuitively visualizing them, this method can more accurately identify the operating status of dry-type reactors, provide a reliable basis for subsequent fault prediction and decision support, and greatly improve the accuracy and reliability of diagnosis.

[0040] In summary, the present invention solves the technical problem that the existing methods still have limitations in processing small signal changes and cannot fully capture subtle fault characteristics in the sound signal of the dry-type reactor. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] Figure 1 A flow chart of the method provided by the present invention. DETAILED DESCRIPTION

[0042] In order to make the purpose, technical solution and advantages of the embodiments of the present invention more clear, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present invention.

[0043] like Figure 1 FIG. 1 is a flow chart of a visual detection method for a dry-type reactor sound signal provided by the present invention. The method comprises the following steps:

[0044] S10, collecting sound signals of the dry-type reactor and performing preprocessing to eliminate environmental noise;

[0045] S20, performing time-frequency analysis on the pre-processed sound signal to obtain a time-frequency spectrum diagram;

[0046] S30, using Bayesian probability reasoning and adaptive threshold augmentation to highlight small changes in the time-frequency spectrum, and obtaining an augmented time-frequency spectrum;

[0047] S40, extracting features of the frequency spectrum during augmentation and expansion, including energy distribution, peak frequency, and harmonic structure;

[0048] S50, combining the extracted features into a multi-dimensional vector, recorded as the first feature vector;

[0049] S60, applying a dimensionality reduction algorithm to map the first eigenvector from a high-dimensional eigenvector to a two-dimensional or three-dimensional space to obtain a second eigenvector;

[0050] S70, clustering the second eigenvector;

[0051] S80: assign colors or labels to different categories according to the clustering result, generate and output a visual classification image based on the second eigenvector.

[0052] The specific implementation methods of the above steps are described in detail below:

[0053] The specific implementation method of step S10 is: first, a high-quality microphone device is used to collect the sound signal generated by the dry-type reactor. In order to eliminate the influence of environmental noise, a preprocessing method combining frequency domain and time domain can be used. In the frequency domain, a bandpass filter can be applied to set a suitable passband frequency range to remove low-frequency and high-frequency noise components. In the time domain, an adaptive filter can be used to adaptively suppress noise using a reference signal (such as an environmental noise signal), thereby effectively eliminating the pollution of environmental noise to the sound signal. This comprehensive preprocessing method can significantly improve the signal quality and lay a good foundation for subsequent time-frequency analysis.

[0054] The specific implementation method of step S20 is: perform time-frequency analysis on the preprocessed sound signal. Common time-frequency analysis methods include short-time Fourier transform (STFT) and wavelet transform. STFT divides the entire sound signal into multiple short-time windows, performs Fourier transform on the signal in each window, and thus obtains a time-frequency spectrum. Wavelet transform uses wavelet functions of different scales and positions to decompose and reconstruct the signal, and can obtain more refined time-frequency distribution characteristics. In practical applications, according to the characteristics of the dry-type reactor sound signal, a suitable time-frequency analysis method and its parameter settings can be selected to obtain a clear time-frequency spectrum.

[0055] The specific implementation method of step S30 is: using Bayesian probability reasoning and adaptive threshold augmentation to highlight the small changes in the time-frequency spectrum. First, the time-frequency spectrum is analyzed using the Bayesian probability model, and the small signal changes that may exist in the time-frequency spectrum are inferred based on prior knowledge and observation data. Then, the time-frequency spectrum is processed using an adaptive threshold augmentation algorithm. The algorithm includes the following sub-steps:

[0056] 1. Normalization of time-frequency spectrum: using Normalize the original time-frequency spectrum S(t,f) to obtain the normalized time-frequency spectrum S norm (t,f).

[0057] 2. Nonlinear contrast enhancement: using The normalized time-frequency spectrum is subjected to nonlinear contrast enhancement, where γ is the contrast enhancement intensity parameter, ranging from 1 to 5, and μ local (t,f) and σ local (t, f) are the weighted mean and weighted standard deviation in the local window, respectively. ∈ is a small positive number to prevent the denominator from being zero, and is taken as 10 -5 ,δ is the gradient enhancement parameter, and its value range is from 0.1 to 1.

[0058] 3. Adaptive threshold calculation: using T(t,f)=μ global +k·σ global ·(1-e -α·|C(t,f)| )·(1+λ·entropy(t,f)) calculates the adaptive threshold T(t,f), where μ global and σ global are the global mean and global standard deviation of the entire time-frequency spectrum respectively, k is the adjustment parameter, ranging from 1 to 3, α is the nonlinear mapping parameter, ranging from 0.1 to 1, λ is the entropy weight parameter, ranging from 0.1 to 1, and entropy(t,f) is the normalized entropy of the local window.

[0059] 4. Dynamic expansion: using

[0060]

[0061] E(t,f) is the dynamic expansion coefficient

[0062] β is the augmentation strength parameter, which controls the degree of augmentation (usually ranging from 0.5 to 2)

[0063] η is the sparsity weight parameter (usually ranging from 0.1 to 1)

[0064] sparsity(t,f) is a measure of the sparsity of a local window centered at (t,f).

[0065] 5. Frequency-dependent adjustment: Use The final expanded time-frequency spectrum S is obtained enhanced (t,f), where f max is the maximum frequency, φ is the frequency adjustment parameter, and its value range is 0.1 to 0.5.

[0066] It should be noted that:

[0067] 1. S(t,f) is obtained by time-frequency analysis in step S20.

[0068] 2.μ local (t,f) and σ local (t,f) is calculated by sliding a fixed-size Gaussian weighted window (e.g. 11x11) on the time-spectrogram.

[0069] 3.μ global and σ global It is obtained by calculating the mean and standard deviation of the entire time-spectrum graph.

[0070] 4. Calculated using the Sobel operator.

[0071] 5.entropy(t,f) is obtained by calculating the probability distribution of pixel values ​​in the local window and then applying the Shannon entropy formula.

[0072] 6.sparsity(t,f) can be measured using the proportion of non-zero elements in the local window.

[0073] 7.γ,δ,k,α,λ,β,η,andφ are hyperparameters, which can be optimized by cross-validation, grid search, or Bayesian optimization.

[0074] These equations introduce advanced techniques such as nonlinear transformation, local features (such as entropy and sparsity), gradient information, and frequency dependency adjustment, which can more finely capture and enhance the subtle changes in the time-spectrum graph while considering the local and global characteristics of the signal. This method can better adapt to different types of dry-type reactor sound signals and improve the sensitivity and reliability of detection.

[0075] This adaptive threshold augmentation algorithm can effectively highlight tiny signal changes in the time-frequency spectrum, laying the foundation for subsequent feature extraction and classification analysis.

[0076] The specific implementation of step S40 is: extracting the augmented time-frequency spectrum S enhanced The key features of (t,f) include:

[0077] 1. Energy distribution characteristics: Calculate the energy distribution of the spectrum in the time and frequency dimensions, such as total energy, band energy, instantaneous power, etc. These characteristics can reflect the overall energy characteristics of the signal.

[0078] 2. Peak frequency feature: detect the significant peak in the spectrum and extract its peak frequency, peak amplitude and other information. These features can reflect the harmonic characteristics of the signal.

[0079] 3. Harmonic structure characteristics: Analyze the frequency-multiplication relationship between the peak frequencies in the time-frequency spectrum and extract relevant indicators of the harmonic structure, such as harmonic richness, harmonic distortion, etc. These characteristics can reflect the periodic characteristics of the signal.

[0080] Through comprehensive analysis of the above three types of features, the time-frequency characteristics of the dry-type reactor sound signal can be comprehensively characterized, providing valuable information for subsequent pattern recognition and fault diagnosis.

[0081] The specific implementation of step S50 is: the features extracted in step S40, including energy distribution, peak frequency, harmonic structure, etc., are combined into a multi-dimensional feature vector, which is recorded as the first feature vector This vector can describe the comprehensive time-frequency characteristics of the dry-type reactor sound signal.

[0082] The specific implementation of step S60 is: applying a dimensionality reduction algorithm to reduce the high-dimensional first eigenvector Mapping from the original high-dimensional space to two-dimensional or three-dimensional space, obtaining the second eigenvector after dimensionality reduction Commonly used dimensionality reduction algorithms include principal component analysis (PCA), linear discriminant analysis (LDA), t-SNE, etc. Through dimensionality reduction processing, the original high-dimensional feature space can be projected into a low-dimensional space, while retaining the main relationship between the original features as much as possible, providing suitable input data for subsequent clustering analysis.

[0083] The specific implementation of step S70 is: Perform cluster analysis. Common clustering algorithms include k-means clustering, Gaussian mixture model clustering, spectral clustering, etc. Through cluster analysis, different modes or states contained in the sound signal of dry-type reactors can be automatically identified, providing a basis for subsequent visual classification. The specific parameter settings of the clustering algorithm, such as the number of cluster centers K, distance measurement, convergence conditions, etc., need to be adjusted and optimized according to the characteristics of the actual problem.

[0084] The specific implementation of step S80 is: according to the clustering result of step S70, distinct colors or labels are assigned to different categories, and based on the second feature vector Generate a visual classification image. The specific steps are as follows:

[0085] 1. Distance calculation: For each point (x, y) in the second eigenvector space, calculate its distance to each cluster center Euclidean distance Where D is the dimension of the second eigenvector (2 or 3); d i (x, y) is the Euclidean distance from the point (x, y) in the second eigenvector space to the i-th cluster center; v j is the value of the point (x, y) in the jth dimension; cij is the value of the i-th cluster center in the j-th dimension.

[0086] 2. Soft classification weight calculation: using Calculate the soft classification weight w of the point (x, y) belonging to the i-th cluster i (x,y), where w i (x, y) is the soft classification weight of the point (x, y) belonging to the i-th cluster; K is the total number of clusters; σ is the bandwidth parameter of the Gaussian kernel, which controls the smoothness of the soft classification.

[0087] 3. Color allocation: use Calculate the final color value of the point (x, y) in is the representative color of the ith cluster (predefined or dynamically generated according to cluster characteristics), and C(x,y) is the final color value of the point (x,y) (in RGB or HSV format).

[0088] 4. Contrast enhancement: using The contrast of the color value is enhanced, where α is the contrast enhancement factor (usually ranging from 1 to 5), β is the contrast adjustment threshold (usually 0.5), and C enhanced (x,y) is the color value after contrast enhancement.

[0089] 5. Label generation: using Assign the final cluster label L(x,y) to the point (x,y), which is the cluster index with the largest weight.

[0090] It should be noted that

[0091] 1.v j and c ij Dimensionality reduction results from step S60 and clustering results from step S70.

[0092] 2.σ can be selected through cross-validation or set to a certain proportion of the average distance between clusters (such as 1 / 3).

[0093] 3. K is the number of clusters determined in step S70.

[0094] 4.C i A set of contrasting colors can be predefined or dynamically generated based on the characteristics of each cluster.

[0095] 5.α and β can be adjusted experimentally to obtain the best visualization effect.

[0096] The steps for generating the final visual classification image can be briefly described as follows:

[0097] 1. For each point (x, y) in the second eigenvector space, calculate its distance d to each cluster center i (x,y).

[0098] 2. Use the soft classification weight formula to calculate the weight w of each point belonging to each cluster i (x,y).

[0099] 3. Calculate the color value C(x,y) of each point based on the weight and predefined cluster color.

[0100] 4. Apply the contrast enhancement formula to get C enhanced (x,y).

[0101] 5. Use the label generation formula to assign the final cluster label L(x,y) to each point.

[0102] 6. Map the calculated color values ​​and labels to a two-dimensional or three-dimensional image to generate the final visual classification image.

[0103] This method can generate color visualization images with smooth transitions while retaining clear category boundaries, which helps to more intuitively display the classification results of dry-type reactor sound signals.

[0104] Through the above steps, a color visualization classification image with smooth transition can be generated while retaining clear category boundaries, which helps to intuitively display the classification results of dry-type reactor sound signals.

[0105] A second aspect of the present invention provides a computer-readable storage medium, wherein the computer-readable storage medium stores program instructions, and when the program instructions are executed, they are used to execute the above-mentioned visual detection method of the sound signal of the dry-type reactor.

[0106] A third aspect of the present invention provides a visual detection system for a sound signal of a dry-type reactor, which comprises the above-mentioned computer-readable storage medium.

[0107] Specifically, the principle of the present invention is:

[0108] 1. Time-frequency analysis and small change enhancement

[0109] Dry-type reactors generate complex sound signals during operation, which contain rich information about the equipment status. By performing time-frequency analysis on the sound signal, such as short-time Fourier transform (STFT) or wavelet transform, its time-frequency characteristic distribution can be obtained. However, since equipment failures often cause subtle sound changes, it is difficult to accurately capture these changes by directly observing the time-frequency spectrum.

[0110] To this end, the present invention adopts Bayesian probability reasoning and adaptive threshold augmentation technology to effectively highlight the subtle change features in the time-frequency spectrum. First, the Bayesian probability model is used to analyze the possible implicit change patterns in the time-frequency spectrum according to prior knowledge and observation data. Then, through normalization, nonlinear contrast enhancement, adaptive threshold calculation, dynamic augmentation, and frequency dependency adjustment, the weak signal changes in the time-frequency spectrum are fully amplified for subsequent feature extraction and fault diagnosis.

[0111] 2. Dimensionality reduction and cluster analysis of multidimensional features

[0112] Key features such as energy distribution, peak frequency and harmonic structure are extracted from the enhanced time-frequency spectrum to form a high-dimensional feature vector. It is difficult to perform fault pattern recognition directly in high-dimensional space, so it is necessary to use dimensionality reduction algorithms, such as principal component analysis (PCA) and linear discriminant analysis (LDA), to map high-dimensional feature vectors to two-dimensional or three-dimensional space and retain the main connections between the original features.

[0113] Based on the low-dimensional feature space after dimensionality reduction, the present invention uses unsupervised clustering algorithms, such as k-means clustering and Gaussian mixture model clustering, to automatically identify different modes or states contained in the sound signal of the dry-type reactor. This unsupervised analysis method can effectively reveal the inherent structural characteristics of the signal data and provide valuable clues for subsequent fault diagnosis.

[0114] 3. Visual classification based on feature clustering

[0115] The clustered low-dimensional feature space is mapped to a two-dimensional or three-dimensional visual classification image to provide intuitive fault diagnosis support for operation and maintenance personnel. Specifically, the Euclidean distance from each feature point to each cluster center is calculated first, and the Gaussian kernel function is used to convert it into a soft classification weight. Then, distinct colors are assigned to different clusters according to the weight distribution, and a contrast enhancement algorithm is applied to generate the final visual classification image.

[0116] This visualization method based on feature clustering can fully display the distribution characteristics of different fault modes in low-dimensional feature space, allowing maintenance personnel to intuitively identify and diagnose the operating status of the equipment. Compared with traditional single signal analysis, this method provides a more comprehensive and explainable basis for fault diagnosis, which helps to improve the accuracy and reliability of diagnosis.

[0117] In general, the present invention integrates advanced digital signal processing and pattern recognition technologies such as time-frequency analysis, feature engineering, dimensionality reduction and clustering, cleverly handles the problem of detecting small changes in the sound signal of the dry-type inductor, and presents the results of fault diagnosis in a visual way.

[0118] In order to better understand and implement the present invention, an embodiment of a specific application scenario of the present invention is provided below: A power company has a number of important dry-type reactor equipment, which is used to cooperate with transformers to adjust the reactive power balance of the power grid. In order to ensure the stable operation of these equipment, the company hopes to establish an automated sound signal detection system to promptly detect potential fault hazards of the equipment. After investigation, the company decided to adopt the dry-type reactor sound signal visualization detection method proposed in the present invention to analyze and diagnose the sound data collected on site.

[0119] First, the company installed a set of high-quality microphone acquisition equipment next to a key dry-type reactor to continuously record the sound signals generated by the equipment during operation. The collected raw sound data was preprocessed to remove the influence of environmental noise. The preprocessed sound signal is shown in Table 1.

[0120] time Sound signal 0.0s 0.01 0.1s 0.02 0.2s 0.03 ... ... 10.0s 0.15

[0121] Table 1 Sound signal waveform after dry-type reactor pretreatment

[0122] Next, the company conducted a time-frequency analysis on the preprocessed sound signal and obtained a time-frequency spectrum using the short-time Fourier transform (STFT) method. As shown in Table 2, the time-frequency spectrum shows the energy distribution characteristics of the dry-type reactor at different times and frequency ranges.

[0123] Frequency / Hz 0s 1s 2s ... 10s 0 1 2 3 ... 4 100 2 3 4 ... 5 200 3 4 5 ... 6 ... .. .. .. ... ... 1000 4 5 6 ... 7

[0124] Table 2 Time-frequency spectrum of dry-type reactor sound signal

[0125] A careful observation of the time-frequency spectrum in Table 2 shows that there are some minor changes in the spectrum, which may be related to the hidden dangers of equipment failure. In order to highlight these minor changes, the company uses the Bayesian probability reasoning and adaptive threshold augmentation algorithm proposed in this invention for processing.

[0126] First, the Bayesian probability model is used to analyze the time-frequency spectrum, and the possible small signal change area is inferred based on prior knowledge and observation data. Then, the adaptive threshold augmentation algorithm is applied to process the time-frequency spectrum. The specific steps are as follows:

[0127] 1. Normalization of time-frequency spectrum: using Normalize the original time-frequency spectrum S(t,f) to obtain the normalized time-frequency spectrum S norm (t,f). Among them, the company measured min(log(1+S))=0.1,max(log(1+S))=2.5.

[0128] 2. Nonlinear contrast enhancement: using The normalized time-frequency spectrum is subjected to nonlinear contrast enhancement to obtain C(t,f). Among them, the company selected γ=3, δ=0.5 after testing.

[0129] 3. Adaptive threshold calculation: use T(t,f)=0.8+2·0.3·(1-e -0.5 ·|C(t ,f)| )·(1+0.5·entropy(t,f)) calculates the adaptive threshold T(t,f), where μ global =0.8,σ global = 0.3, k = 2, α = 0.5, λ = 0.5. The normalized entropy of the local window entropy(t,f) is measured to be between 0.3 and 0.7.

[0130] 4. Dynamic expansion: using

[0131] E(t,f) is the dynamic expansion coefficient

[0132] β is the augmentation strength parameter, which controls the degree of augmentation (usually ranging from 0.5 to 2)

[0133] η is the sparsity weight parameter (usually ranging from 0.1 to 1)

[0134] sparsity(t,f) is the sparsity measure of the local window centered at (t,f)

[0135] It was measured to be between 0.2 and 0.8, β=1.2, η=0.5.

[0136] 5. Frequency-dependent adjustment: Use The final expanded time-frequency spectrum S is obtained enhanced (t,f), where f max =1000Hz,φ=0.3.

[0137] Through the adaptive threshold augmentation algorithm of the above five steps, the company successfully highlighted the subtle signal change characteristics in the time-frequency spectrum, as shown in Table 3. Compared with the original time-frequency spectrum, some originally less obvious change areas in the enhanced time-frequency spectrum become more prominent.

[0138] Frequency / Hz 0s 1s 2s ... 10s 0 1 2 3 ... 4 100 2 3 4 ... 5 200 3 4 5 ... 6 ... .. .. .. ... ... 1000 4 5 6 ... 7

[0139] Table 3 Enhanced dry-type reactor sound signal time spectrum

[0140] Next, the company extracted key features of the enhanced spectrogram, including:

[0141] 1. Energy distribution characteristics: The energy distribution of the spectrum in the time and frequency dimensions during calculation, such as the total energy E total =∑ t,f S enhanced (t,f)=3450,band energy,instantaneous power P(t)=∑ f S enhanced (t,f).

[0142] 2. Peak frequency feature: detect the significant peak in the spectrum graph and extract its peak frequency f peak =320Hz, peak amplitude A peak =6.2.

[0143] 3. Harmonic structure characteristics: Analyze the frequency relationship between the peak frequencies in the spectrum diagram and extract the harmonic richness H R =0.78, harmonic distortion H D =0.22.

[0144] Combine the above three types of features into a 12-dimensional first feature vector

[0145] Afterwards, the company used the principal component analysis (PCA) algorithm to transform the high-dimensional first eigenvector Mapping from the original 12-dimensional space to the two-dimensional feature space As shown in Table 4. This dimensionality reduction operation can effectively remove redundant information between features and highlight the main pattern differences.

[0146] u1 u2 2.1 1.5 2.3 1.7 2.1 1.6 ... ... 2.5 1.9

[0147] Table 4 Two-dimensional feature space distribution of dry-type reactor sound signal

[0148] In the two-dimensional feature space, the company uses the k-means clustering algorithm to perform unsupervised classification on the sample points and finds that they can be divided into three different patterns. The cluster center coordinates are

[0149] Based on the clustering results, the company assigned red, green and blue as representative colors to the three types of patterns. Then, the soft classification weight calculation formula was used in Assign a probability weight of belonging to the three clusters to each feature point (u1, u2).

[0150] Finally, the company applies the contrast enhancement formula The calculated color values ​​are mapped to a two-dimensional plane to generate a visual classification image.

[0151] The red area corresponds to normal operation, the green area may have some minor fault risks, and the blue area indicates that the equipment has a more serious fault state.

[0152] By continuously monitoring this visual classification image, the company can promptly detect possible fault signs of dry-type reactors, providing an important basis for subsequent maintenance. Compared with the traditional single signal analysis method, the visual diagnosis results provided by the present invention are more intuitive and comprehensive, which helps to improve the accuracy and reliability of fault diagnosis and contribute new technical support to the intelligent management of power equipment.

[0153] In general, this specific embodiment fully demonstrates the application value of the present invention in the analysis of dry-type reactor sound signals. Through a series of innovative technical means such as time-frequency analysis, feature extraction, dimensionality reduction and cluster visualization, the company successfully extracted and presented the micro-fault features in the dry-type reactor sound signals, providing effective support for equipment status monitoring and preventive maintenance.

[0154] The above description is only a specific implementation mode of the present invention, but the protection scope of the present invention is not limited thereto. Any technician familiar with the technical field can easily think of changes or substitutions within the technical scope disclosed by the present invention, which should be covered by the protection scope of the present invention.

Claims

1. A visual detection method for sound signals of dry-type reactors. It is characterized in that The following steps are involved: S10, collecting sound signals of the dry-type reactor and performing preprocessing to eliminate environmental noise; S20, performing time-frequency analysis on the pre-processed sound signal to obtain a time-frequency spectrum diagram; S30, using Bayesian probability reasoning and adaptive threshold augmentation to highlight small changes in the time-frequency spectrum diagram to obtain an augmented time-frequency spectrum diagram; S40, extracting features of the frequency spectrum during augmentation and expansion, including energy distribution, peak frequency, and harmonic structure; S50, combining the extracted features into a multi-dimensional vector, recorded as the first feature vector; S60, applying a dimensionality reduction algorithm to map the first feature vector from a high-dimensional feature vector to a two-dimensional or three-dimensional space to obtain a second feature vector; S70, clustering the second feature vector; S80: assign colors or labels to different categories according to the clustering result, and generate and output a visual classification image based on the second feature vector.

2. A visual detection method for sound signals of a dry-type reactor according to claim 1, It is characterized in that The adaptive threshold augmentation is calculated by using a plurality of equations, including a time-frequency spectrum normalization equation, a nonlinear contrast enhancement equation, an adaptive threshold calculation equation, a dynamic augmentation coefficient calculation equation, and a frequency dependency adjustment equation.

3. A visual detection method for sound signals of dry-type reactors according to claim 2, It is characterized in that The time-frequency spectrum normalization equation is specifically expressed as: In the formula, S norm (t,f) is the normalized time-frequency spectrum, S(t,f) is the original time-frequency spectrum; t is time, and f is frequency.

4. A visual detection method for sound signals of dry-type reactors according to claim 3, It is characterized in that The nonlinear contrast enhancement equation is specifically expressed as: Where C(t,f) is the spectrum after nonlinear contrast enhancement, tanh is the hyperbolic tangent function, γ is the contrast enhancement intensity parameter, μ local (t,f) is the weighted average in the local window centered at (t,f), σ local (t,f) is the weighted standard deviation in the local window centered at (t,f), ∈ is a small positive number to prevent the denominator from being zero, δ is the gradient enhancement parameter, is the gradient of the normalized spectrum at the point (t,f).

5. A visual detection method for sound signals of dry-type reactors according to claim 4, It is characterized in that The adaptive threshold calculation equation is specifically expressed as: T(t,f)=μ global +k·σ global ·(1-e -α·|C(t,f)| )·(1+λ·entropy(t,f)); Where T(t,f) is the adaptive threshold; μ global is the global average value of the entire spectrum; σ global is the global standard deviation of the entire spectrum; k is an adjustment parameter that controls the sensitivity of the threshold; α is a nonlinear mapping parameter that controls the rate at which the threshold changes with contrast; λ is an entropy weight parameter; entropy(t,f) is the normalized Shannon entropy of the local window centered at (t,f).

6. A visual detection method for sound signals of dry-type reactors according to claim 5, It is characterized in that The dynamic expansion coefficient calculation equation is specifically expressed as: Where E(t,f) is the dynamic augmentation coefficient; β is the augmentation strength parameter, which controls the degree of augmentation; η is the sparsity weight parameter; sparsity(t,f) is the sparsity measure of the local window centered at (t,f).

7. A visual detection method for sound signals of dry-type reactors according to claim 6, It is characterized in that The frequency-dependent adjustment equation is specifically expressed as: In the formula, S enhanced (t,f) is the final time-frequency spectrum after expansion; f max is the maximum frequency; φ is the frequency adjustment parameter.

8. A visual detection method for sound signals of dry-type reactors according to claim 7, It is characterized in that The method for clustering the second feature vector is: k-means clustering or Gaussian mixture model clustering.

9. A computer-readable storage medium, It is characterized in that The computer-readable storage medium stores program instructions, and when the program instructions are executed, they are used to execute the visual detection method for sound signals of a dry-type reactor according to any one of claims 1 to 8.

10. A visual detection system for sound signals of dry-type reactors. It is characterized in that A computer-readable storage medium comprising the computer-readable storage medium of claim 9.

Citation Information

Patent Citations

  • Sound fault detection method applied to power internet of things and device thereof

    CN113362856A

  • Acoustic diagnosis method for discharge fault and abnormal operation state of dry-type air-core reactor

    CN116973703A

  • Electric reactor fault detection method and device, electronic equipment and storage medium

    CN117594064A

  • Power transformer fault diagnosis method and system based on noise spectrum and PCA-K-Means algorithm

    CN117909774A

  • Distributed water turbine cavitation state diagnosis method and system

    CN117932215A

Cited By

  • Music and light visualization method based on acousto-optic mapping

    CN120636448A