High-voltage circuit breaker operating mechanism voiceprint state detection method and device based on hesitant fuzzy number and medium

Abstract

The invention relates to a hesitant fuzzy number-based high-voltage circuit breaker operating mechanism voiceprint state detection method and device, and a medium. Extracting features of an original voiceprint data sequence of the high-voltage circuit breaker operating mechanism, wherein the features comprise a Mel-frequency cepstrum coefficient, short-time energy and a zero-crossing rate; forming a symptom feature vector by the symptom parameters, inputting the symptom feature vector into a fuzzy depth residual shrinkage network, constructing a membership function, and outputting a membership value of each state; constructing hesitant fuzzy numbers according to the membership values to form a collective hesitant fuzzy evaluation matrix; determining the evidence weight of each input evaluation model through an optimal-worst method, deriving a state risk weight through a TOPSIS method in combination with a language Z number, and performing weighted fusion on the collective hesitation fuzzy evaluation matrix to obtain a health index; and dividing the health indexes into normal, attention and dangerous health levels by adopting a K-means clustering algorithm. Compared with the prior art, the method has the advantages of high accuracy, high robustness, high flexibility and the like.

Description

Technical Field

[0001] This invention relates to the field of high-voltage circuit breaker testing, and in particular to a method, equipment, and medium for detecting the acoustic signature status of a high-voltage circuit breaker operating mechanism based on hesitant fuzzy numbers. Background Technology

[0002] As a critical piece of equipment in the power system, the operating mechanism of high-voltage circuit breakers directly affects the safe and stable operation of the power grid. Traditional condition monitoring methods mainly rely on physical signals such as vibration, temperature, and torque. However, under complex operating conditions, such as environments with strong electromagnetic interference and mixed mechanical vibrations, these signals are easily affected by external factors, leading to distorted monitoring data. Furthermore, physical signals can only reflect local characteristics of the equipment and are insufficient to comprehensively characterize complex faults such as wear and loosening of mechanical components, making it impossible to accurately assess the overall health status of the equipment.

[0003] In recent years, voiceprint monitoring technology has gradually become an effective supplementary method. This technology collects the operating sounds of equipment through audio sensors and uses signal processing and machine learning algorithms to extract features for status assessment. Compared with traditional methods, voiceprint information can more sensitively capture anomalies in mechanical components, especially in characterizing dynamic operating states. With technological advancements, various models have been introduced into the field of voiceprint monitoring. For example, feedforward neural networks are used for basic voiceprint feature extraction and classification; convolutional neural networks improve classification accuracy by automatically learning local features from two-dimensional spectrograms; recurrent neural networks and their variants focus on processing temporal dependencies in voiceprint signals, making them suitable for dynamic operating condition analysis; and hybrid models combining fuzzy logic and neural networks attempt to address the problem of uncertainty representation in the monitoring process.

[0004] Despite these limitations, existing voiceprint monitoring technologies still have significant shortcomings. When environmental noise interferes, background noise in industrial settings easily contaminates voiceprint features, leading to misjudgments. Voiceprint monitoring technologies also suffer from limitations in recognizing overlapping multiple states; a single model struggles to distinguish the boundary features of normal, attentive, and hazardous states under complex operating conditions. Furthermore, uncertainty handling is inadequate; traditional membership functions only support single-state assignment and cannot simultaneously quantify the probability distribution of equipment across multiple states, thus limiting the accuracy of risk assessment.

[0005] Chinese patent application CN111638449A discloses a method, device, and readable storage medium for fault diagnosis of distribution automation switches. Based on the characteristic quantities of the distribution automation switch operation status evaluation index system, it establishes a fuzzy evaluation model factor set and hierarchical structure. Then, based on the distribution automation switch fault information and status evaluation standards, it establishes the membership function of the evaluation factors, performs comprehensive weighted fuzzy calculations, and finally establishes a distribution automation switch status evaluation model based on hierarchical fuzzy comprehensive evaluation. However, the weight allocation in this application is only based on the analytic hierarchy process (AHP), without considering the dynamic reliability differences of different features or models, thus reducing the robustness of the evaluation. Therefore, how to optimize the characteristics of acoustic signals and manage multi-model uncertainties in the detection of high-voltage circuit breaker operating mechanisms to improve the accuracy and robustness of reliability status evaluation is a technical problem that needs to be solved. Summary of the Invention

[0006] The purpose of this invention is to overcome the shortcomings of the existing technology by providing a method, device and medium for detecting the acoustic signature status of a high-voltage circuit breaker operating mechanism based on hesitant fuzzy numbers. By efficiently extracting acoustic signature features and combining them with an advanced fuzzy deep residual shrinking network and flexible clustering method, the status of the high-voltage circuit breaker operating mechanism can be accurately assessed, ensuring the reliability and accuracy of monitoring, thereby optimizing maintenance strategies and extending the service life of the equipment.

[0007] The objective of this invention can be achieved through the following technical solutions:

[0008] According to one aspect of the present invention, a method for detecting the voiceprint status of a high-voltage circuit breaker operating mechanism based on hesitant fuzzy numbers is provided. The specific steps include: S1, acquiring the original voiceprint data sequence of the high-voltage circuit breaker operating mechanism through an array microphone of an edge device, and extracting symptom parameters, including Mel frequency cepstral coefficients, short-time energy, and zero-crossing rate; S2, forming a symptom feature vector from the symptom parameters, inputting it into a fuzzy deep residual shrinking network, constructing membership functions including normal state, attentive state, and dangerous state, and outputting the membership values ​​of each state; S3, constructing hesitant fuzzy numbers based on the membership values ​​of each state, integrating the hesitant fuzzy numbers of evaluation models with different symptom parameter inputs, and forming a collective hesitant fuzzy evaluation matrix; S4, determining the evidence weights of each input evaluation model using the optimal-worst method, deriving the state risk weights using the TOPSIS method combined with the language Z-number, and weighting and fusing the collective hesitant fuzzy evaluation matrix based on the evidence weights and state risk weights to obtain a health index; and using a K-means clustering algorithm to divide the health index into normal, attentive, and dangerous health levels, outputting the health status.

[0009] Furthermore, the calculation of the Mel frequency cepstral coefficients in S1 includes: processing the audio signal in frames, passing it through a Mel filter bank after Fourier transform, taking the logarithmic energy and performing discrete cosine transform for dimensionality reduction; the calculation of the short-time energy includes: taking the mean of the sum of squares of the audio signal sample values ​​within the time window; the calculation of the zero-crossing rate includes: counting the number of symbol changes of adjacent audio samples within the time window.

[0010] Furthermore, the symptom feature vector in S2 includes a single feature vector corresponding to each symptom parameter individually, a double feature vector combining any two symptom parameters, and a full feature vector combining three symptom parameters; the single feature vector includes a Mel frequency cepstral coefficient feature vector, a short-time energy feature vector, and a zero-crossing rate feature vector; the double feature vector includes a feature vector combining Mel frequency cepstral coefficients with short-time energy, a feature vector combining short-time energy with zero-crossing rate, and a feature vector combining Mel frequency cepstral coefficients with zero-crossing rate; the full feature vector is composed of a combination of Mel frequency cepstral coefficients, short-time energy, and zero-crossing rate.

[0011] Furthermore, in the fuzzy deep residual shrinkage network, the fuzzy model parameters of each symptom feature vector are fine-tuned through a fully connected layer, and the evaluation values ​​of the three states corresponding to each symptom feature vector are expressed as follows:

[0012]

[0013] Among them, z N z is the evaluation value for the normal state. A z is the evaluation value for the state of attention. D F is the assessment value for a hazardous condition. I (p) represents the input feature value of the p-th neuron in the l-th layer of the neural network, W plq,N W plq,A and W plq,D The weight vectors for normal, attentive, and dangerous states, respectively, b lq,N b lq,A and b lq,D This is a bias term.

[0014] Furthermore, the membership function is obtained based on the evaluation values ​​of the three states corresponding to each symptom feature vector. The expression for the membership function y(z) is:

[0015]

[0016] Where a1, a2, and a3 are the baseline offsets of the membership functions controlling normal, alert, and dangerous states, respectively; b1, b2, and b3 are the rates of change of the membership functions controlling normal, alert, and dangerous states, respectively; and z N z is the evaluation value for the normal state.A z is the evaluation value for the state of attention. D This is the assessment value for a hazardous condition.

[0017] Furthermore, the hesitation fuzzy number in S3 is obtained from the membership degree corresponding to each symptom feature vector. The hesitation fuzzy number is a triple, including the membership degree of the normal state, the attention state, and the danger state. The collective hesitation fuzzy evaluation matrix includes the membership degree groups of the three states corresponding to each symptom feature vector, and the expression is:

[0018]

[0019] in, For the collective hesitation fuzzy assessment matrix, M is the number of types of symptom feature vectors, which is 7, (h N,i (z N ),h A,i (z A ),h D,i (z D ) represents the membership group corresponding to the feature vector of the i-th symptom class, h N,i (z N ) represents the membership degree of the normal state corresponding to the i-th type of symptom feature vector, h A,i (z A ) represents the membership degree of the attentional state corresponding to the feature vector of the i-th symptom class, h D,i (z D ) represents the membership degree of the dangerous state corresponding to the feature vector of the i-th type of symptom.

[0020] Further, the optimal-worst method in S4 specifically involves obtaining the relative performance priority of each membership group to the optimal and worst membership groups through the selected optimal and worst membership groups, and solving for the evidence weights in the optimization problem based on the relative performance priority; the optimization problem is to minimize the maximum deviation, expressed as:

[0021]

[0022] Among them, w B The evidence weight for the optimal membership group, w W For the evidence weight of the worst membership group, AB i As the relative performance priority between the best membership group and the i-th membership group, A i W represents the relative performance priority between the worst membership group and the i-th membership group. i Let be the evidence weight for the i-th membership group.

[0023] Furthermore, the K-means clustering algorithm in S4 involves randomly selecting initial cluster centers based on a preset number of clusters, calculating the Euclidean distance from the current health index to each initial cluster center, assigning the current health index to the nearest cluster, updating the cluster center to the cluster mean, and repeating the iterative operation of the health index to each cluster center until the iteration result is stable, and outputting the operating mechanism state corresponding to the current nearest cluster.

[0024] According to a second aspect of the present invention, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program, and the processor executes the program to implement the method described thereon.

[0025] According to a third aspect of the present invention, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the method described thereon.

[0026] Compared with the prior art, the present invention has the following beneficial effects:

[0027] (1) Improve the accuracy and robustness of state assessment: By extracting three acoustic features—Mel frequency cepstral coefficients, short-time energy, and zero-crossing rate—and combining them with a fuzzy deep residual shrinkage network to construct a state membership function, the operating status of the high-voltage circuit breaker operating mechanism can be comprehensively captured. Compared with the traditional single signal monitoring method, multi-feature fusion significantly reduces the interference of complex environment on monitoring results and improves the accuracy of state classification. By using single-feature, dual-feature, and full-feature symptom feature vectors, the importance of each feature is clearly analyzed while the features are interactively verified, which improves the anti-interference ability and model interpretability. At the same time, by constructing hesitant fuzzy numbers to integrate multi-model evaluation results, the robustness of the system to uncertainty and noise is further enhanced, ensuring that the evaluation results are more reliable.

[0028] (2) Optimizing the flexibility and scientific nature of health level classification: The best-worst BWM method is used to determine the evidence weights of each assessment model, and the state risk weights are derived by combining the TOPSIS method and the language Z number. This achieves dual optimization of model reliability and state risk. The health index HI is dynamically classified by the K-means clustering algorithm, which can adaptively and accurately classify the equipment status into three levels: normal, attention and danger. This assessment method based on data-driven and expert knowledge avoids the problem of strong subjectivity in traditional threshold setting, while adapting to state changes under different working conditions, providing a scientific basis for maintenance decisions.

[0029] (3) Achieve efficient real-time monitoring and intelligent maintenance: Feature extraction, fuzzy reasoning, multi-model fusion and clustering classification are integrated into edge devices to support real-time acquisition and status assessment of the acoustic signals of high-voltage circuit breaker operating mechanisms. Through cloud collaboration, the system can quickly output health levels and trigger early warnings, significantly shortening fault response time. In addition, the collaborative design of multiple features and multiple models gives the system strong generalization ability and can be widely applied to different types of high-voltage circuit breakers, providing efficient and reliable technical support for the intelligent operation and maintenance of power systems. Attached Figure Description

[0030] Figure 1 The flowchart shows a method for detecting the acoustic signature of a high-voltage circuit breaker operating mechanism based on hesitant fuzzy numbers.

[0031] Figure 2 Diagram of microphone array;

[0032] Figure 3 The zero-crossing rate evaluation curve for the membership function;

[0033] Figure 4 This is a structural diagram of a fuzzy depth residual shrinkage network;

[0034] Figure 5 This is a schematic diagram of the grade classification results. Detailed Implementation

[0035] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0036] Example 1

[0037] This embodiment proposes a method for detecting the acoustic signature status of a high-voltage circuit breaker operating mechanism based on hesitant fuzzy numbers. On an edge device, three key symptom parameters—Mel frequency cepstral coefficients, short-time energy, and zero-crossing rate—are extracted from the raw acoustic signature data sequence collected by an array microphone. A fuzzy deep residual shrinking network is used to establish a state membership function based on these symptom parameters to evaluate the status of the high-voltage circuit breaker operating mechanism. Hesitant fuzzy numbers are constructed to represent the probabilities of different operating states, and hesitant fuzzy numbers from multiple evaluation models are integrated into a collective hesitant fuzzy evaluation matrix. A flexible clustering method is proposed to objectively identify health levels. By integrating model evidence weights and state risk weights, different operating states are comprehensively evaluated and classified, thereby completing the acoustic signature status assessment of the high-voltage circuit breaker operating mechanism.

[0038] like Figure 1 The image shows a method for detecting the acoustic signature status of a high-voltage circuit breaker operating mechanism based on hesitant fuzzy numbers. The specific steps include:

[0039] S1. Collect the original acoustic data sequence of the high-voltage circuit breaker operating mechanism through the array microphone of the edge device, and extract the symptom parameters, including Mel frequency cepstral coefficients, short-time energy and zero-crossing rate;

[0040] S2. Combine the symptom parameters into a symptom feature vector, input it into a fuzzy deep residual shrinking network, construct a membership function including normal state, attention state and dangerous state, and output the membership value of each state.

[0041] S3. Construct hesitant fuzzy numbers based on the membership values ​​of each state, integrate the hesitant fuzzy numbers of the evaluation model with different symptom parameter inputs, and form a collective hesitant fuzzy evaluation matrix.

[0042] S4. Determine the evidence weights of each input evaluation model using the best-worst method, derive the state risk weights using the TOPSIS method combined with the language Z-number, and weight and fuse the collective hesitation fuzzy evaluation matrix based on the evidence weights and state risk weights to obtain the health index; use the K-means clustering algorithm to divide the health index into normal, attentive and dangerous health levels, and output the health status.

[0043] like Figure 2 The diagram shows the microphone layout on an edge device. Mel frequency cepstral coefficients, short-time energy, and zero-crossing rate are extracted from the voiceprint information collected by the microphones. The calculation of the Mel frequency cepstral coefficients involves frame-by-frame processing of the audio signal, Fourier transform, passing it through a Mel filter bank, taking the logarithmic energy, and performing discrete cosine transform for dimensionality reduction. The expression is:

[0044]

[0045] Where MFCC is the Mel frequency cepstral coefficient, F(x) k f is the Fourier transform result of the k-th frame audio signal. k Let N be the response of the Mel filter bank, and N be the frame length.

[0046] The calculation of short-time energy E(t) involves taking the average of the sum of squares of the audio signal sample values ​​within the time window, expressed as:

[0047]

[0048] Where x(n) represents the audio signal samples within the time window, and N is the window length.

[0049] The calculation of the zero-crossing rate ZCR(t) includes the number of symbol changes between adjacent audio samples within a statistical time window, expressed as:

[0050]

[0051] Where x(n) is the audio signal sample within the time window, I is the indicator function used to determine whether adjacent samples have opposite signs, and N is the window length.

[0052] The symptom feature vectors in S2 include single feature vectors corresponding to each symptom parameter individually, dual feature vectors combining any two symptom parameters, and full feature vectors combining three symptom parameters. Single feature vectors include Mel-frequency cepstral coefficient feature vectors, short-time energy feature vectors, and zero-crossing rate feature vectors. Dual feature vectors include feature vectors combining Mel-frequency cepstral coefficients and short-time energy, short-time energy and zero-crossing rate features, and Mel-frequency cepstral coefficients and zero-crossing rate features. The full feature vector is composed of Mel-frequency cepstral coefficients, short-time energy, and zero-crossing rate features. Single feature vector input verifies the discriminative power of a single feature and identifies the most sensitive feature. Dual feature vector input reveals synergistic or redundant relationships between features, utilizing complementary relationships. The full feature vector input integrates all information and serves as a benchmark for comparison, verifying whether feature fusion improves performance. Using different input combinations improves the system's anti-interference capability and model interpretability, and the comparison results of different combinations provide clear optimization directions for subsequent feature engineering.

[0053] In the fuzzy deep residual shrinking network, the symptom feature vectors are fine-tuned using fuzzy model parameters through fully connected layers. The evaluation values ​​for the three states corresponding to each symptom feature vector are expressed as follows:

[0054]

[0055] Among them, z N z is the evaluation value for the normal state. A z is the evaluation value for the state of attention. D F is the assessment value for a hazardous condition. l (p) represents the input feature value of the p-th neuron in the l-th layer of the neural network, W plq,N W plq,A and W plq,D The weight vectors for normal, attentive, and dangerous states, respectively, b lq,N b lq,A and b lq,D This is a bias term.

[0056] like Figure 3 The figure shows the zero-crossing rate evaluation curve of the membership function. The membership function is obtained based on the evaluation values ​​of the three states corresponding to each symptom feature vector. The expression of the membership function y(z) is:

[0057]

[0058] Where a1, a2, and a3 are the baseline offsets of the membership functions controlling normal, alert, and dangerous states, respectively; b1, b2, and b3 are the rates of change of the membership functions controlling normal, alert, and dangerous states, respectively; and z N z is the evaluation value for the normal state. A z is the evaluation value for the state of attention. D This is the assessment value for a hazardous condition. For example... Figure 4 The diagram shown is a structural diagram of a fuzzy depth residual shrinkage network.

[0059] The hesitant fuzzy number in S3 is obtained from the membership degrees corresponding to each symptom feature vector. The hesitant fuzzy number is a triple, including the membership degrees of the normal state, the attentive state, and the dangerous state, and its expression is:

[0060] h = y(z) N )∨y(z A )∨y(z D )= <h N (z N ),h A (z A ),h D (z D )>,

[0061] Among them, y(z) N ), y(z A ) and y(z D ) are the membership degrees of the normal state, attention state, and danger state, respectively, calculated through membership functions. <h N (z N ),h A (z A ),h D (z D )> is a triple representing the membership degrees of three states. Specifically: h N (z N ) represents the membership degree in the normal state, h A (z A ) represents the membership degree of the attention state, h D (z D ) represents the membership degree of a dangerous state.

[0062] The collective hesitation fuzzy evaluation matrix includes membership groups of the three states corresponding to each symptom feature vector, with M types of symptom feature vectors. Each model will provide the membership degrees of the three states, expressed as:

[0063]

[0064] in, For the collective hesitation fuzzy assessment matrix, M is the number of types of symptom feature vectors, which is 7, (h N,i (z N ),h A,i (z A ),h D,i (z D ) represents the membership group corresponding to the feature vector of the i-th symptom class, h N,i (z N ) represents the membership degree of the normal state corresponding to the i-th type of symptom feature vector, h A,i (z A ) represents the membership degree of the attentional state corresponding to the feature vector of the i-th symptom class, h D,i (z D ) represents the membership degree of the dangerous state corresponding to the feature vector of the i-th type of symptom.

[0065] The optimal-worst method in S4 specifically involves obtaining the relative performance priority of each membership group to the optimal and worst membership groups, respectively, based on the selected optimal and worst membership groups. The evidence weights are then calculated using these relative performance priorities in the optimization problem. The optimization problem is to minimize the maximum deviation, expressed as:

[0066]

[0067] AB = {AB1, AB2, ..., AB} i ,…,AB M},

[0068] AW = {A1W, A2W, ..., A i W,…,A M W},

[0069] Among them, w B The evidence weight for the optimal membership group, w W For the evidence weight of the worst membership group, AB i As the relative performance priority between the best membership group and the i-th membership group, a i W represents the relative performance priority between the worst membership group and the i-th membership group. i Let be the evidence weight for the i-th membership group.

[0070] The state risk weights are derived by combining the TOPSIS method with the language Z-number. Based on the evidence weights and state risk weights, the collective hesitation fuzzy evaluation matrix is ​​weighted and fused to obtain the health index. The specific formula is as follows:

[0071]

[0072] Where: δj is the risk weight of the j-th operation state, s ij These are elements in the state risk estimation distance matrix, where Q is the number of operational states (normal, caution, danger).

[0073] After combining the model evidence weights, the health assessment information for each model is aggregated as follows:

[0074]

[0075] Among them, w i h is the weight of the i-th evaluation model, reflecting the importance of that model in the overall evaluation. N,i h A,i , and h D,i These represent the membership degrees of the i-th evaluation model to the normal state, the attention state, and the dangerous state, respectively. It is the weighted average membership of all evaluation models to the normal state. It is the weighted average membership of all evaluation models to the attentional state. It is the weighted average membership degree of all assessment models to the hazardous state.

[0076] The final health index is obtained by integrating the evidence weights and state risk weights of the model. The specific formula is as follows:

[0077]

[0078] in, For health information matrix.

[0079] The K-means clustering algorithm in S4 randomly selects initial cluster centers based on a preset number of clusters, calculates the Euclidean distance from the current health index to each initial cluster center, assigns the current health index to the nearest cluster, and updates the cluster center to the cluster mean. This process is repeated iteratively to perform the corresponding operation of the health index to each cluster center until the iteration result is stable, and then outputs the operating mechanism state corresponding to the current nearest cluster.

[0080] The K-means algorithm, as an unsupervised iterative clustering technique, classifies the health status of high-voltage circuit breaker operating mechanisms into a predefined number of health levels (normal, caution, and danger) of k = Q = 3 using a three-dimensional health index vector generated by a voiceprint time window. Each sample corresponds to a health assessment result of a voiceprint signal time window, specifically represented as a health index vector HI = (hN, hA, hD), where hN, hA, and hD are used to weight and fuse the membership degree H.

[0081] Besides the number of clusters k, two parameters need to be predefined: initial cluster centers and a distance metric. k initial cluster center vectors HI′ are selected from a set of health indices. i(i = 1, 2, ..., k). The difference between each sample and the k initial cluster center vectors is measured using Euclidean distance as follows:

[0082]

[0083] Where P is the sample size, and HI is the health assessment information of the sample. The health index vector.

[0084] Next, each sample is assigned to the nearest cluster center vector, generating a new health level cluster. The cluster mean is calculated and updated with the new cluster center vector for the next iteration. This process is repeated until the iteration is stable. This classification model is then deployed to edge devices, and the classification results are transmitted in real-time to a cloud server for real-time status monitoring of the high-voltage circuit breaker operating mechanism.

[0085] In this way, device status can be automatically classified into different health levels based on voiceprint characteristics. This method can not only quickly identify abnormal situations but also detect potential problems in advance, thereby enabling timely maintenance and preventing larger failures.

[0086] Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working process of the described module can be referred to the corresponding process in the foregoing method embodiments, and will not be repeated here.

[0087] Example 2

[0088] This embodiment experimentally verifies the method of Embodiment 1, and obtains the following results: Figure 5 The diagram shows the classification results. The horizontal axis represents the number of tests, reflecting the multiple assessments of the acoustic signature status of the high-voltage circuit breaker operating mechanism. The vertical axis represents the tests under different operating conditions. The decision boundary is the boundary line between the three states: hazardous, caution, and normal. In the 10 sets of tests, each test point falls within the hazardous, caution, or normal zone based on its assessment results, clearly indicating the classification of the equipment status under different test scenarios.

[0089] This invention constructs a multi-dimensional uncertainty modeling and dynamic decision-making fusion mechanism, deeply embedding hesitant fuzzy theory into the voiceprint status assessment system. After extracting the dynamic correlation between voiceprint features and equipment health status through a fuzzy deep residual network, and addressing the multi-valued and temporally uncertain characteristics of high-voltage circuit breaker mechanical states, a three-dimensional hesitant fuzzy number is constructed to synchronously represent the membership probabilities of equipment in normal, attentive, and dangerous states, overcoming the limitations of traditional methods that rely on a single membership degree. Based on this, a dynamic weight fusion mechanism is designed, integrating the credibility weights of the assessment model with the priority weights of state risks to achieve objective clustering of health levels. By combining the semantic expression capabilities of fuzzy logic, the feature mining advantages of deep learning, and the fusion mechanism of multi-criteria decision-making, a state assessment system with autonomous adaptability is formed. This solves the problem of misjudgment caused by environmental noise interference in traditional voiceprint monitoring and overcomes the limitations of single models in recognizing overlapping features of multiple states under complex operating conditions, achieving accurate quantification of equipment health status and synergistic optimization of risk warning.

[0090] The electronic device of this invention includes a central processing unit (CPU), which can perform various appropriate actions and processes according to computer program instructions stored in read-only memory (ROM) or loaded from a storage unit into random access memory (RAM). The RAM may also store various programs and data required for device operation. The CPU, ROM, and RAM are interconnected via a bus. Input / output (I / O) interfaces are also connected to the bus.

[0091] Multiple components in the device are connected to an I / O interface, including: input units such as a keyboard, mouse, etc.; output units such as various types of displays, speakers, etc.; storage units such as disks, optical disks, etc.; and communication units such as network interface cards, modems, wireless transceivers, etc. The communication unit allows the device to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks. The processing unit performs the various methods and processes described above, such as the method of the present invention. For example, in some embodiments, the method of the present invention may be implemented as a computer software program tangibly contained in a machine-readable medium, such as a storage unit. In some embodiments, part or all of the computer program may be loaded and / or installed on the device via ROM and / or the communication unit. When the computer program is loaded into RAM and executed by the CPU, one or more steps of the method of the present invention described above may be performed. Alternatively, in other embodiments, the CPU may be configured to execute the method of the present invention by any other suitable means (e.g., by means of firmware).

[0092] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Complex Programmable Logic Devices (CPLDs), and so on.

[0093] The program code used to implement the methods of the present invention can be written in any combination of one or more programming languages. This program code can be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the program code causes the functions / operations specified in the flowcharts and / or block diagrams to be implemented. The program code can be executed entirely on the machine, partially on the machine, as a standalone software package partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0094] In the context of this invention, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. Machine-readable media can include, but are not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0095] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in the present invention, and these modifications or substitutions should all be covered within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for detecting the voiceprint status of a high-voltage circuit breaker operating mechanism based on hesitant fuzzy numbers, characterized in that, The specific steps include: S1. Acquiring the original voiceprint data sequence of the high-voltage circuit breaker operating mechanism through the array microphone of the edge device, and extracting symptom parameters, including Mel frequency cepstral coefficients, short-time energy, and zero-crossing rate; S2. Combining the symptom parameters into a symptom feature vector, inputting it into a fuzzy deep residual shrinking network, constructing membership functions including normal state, attention state, and dangerous state, and outputting the membership value of each state; S3. Constructing hesitant fuzzy numbers based on the membership values ​​of each state, integrating the hesitant fuzzy numbers of the evaluation models with different symptom parameter inputs, and forming a collective hesitant fuzzy evaluation matrix; S4. Determining the evidence weights of each input evaluation model through the optimal-worst method, deriving the state risk weights through the TOPSIS method combined with the language Z-number, and performing weighted fusion of the collective hesitant fuzzy evaluation matrix based on the evidence weights and state risk weights to obtain the health index; using the K-means clustering algorithm, dividing the health index into normal, attention, and dangerous health levels, and outputting the health status.

2. The method for detecting the voiceprint status of a high-voltage circuit breaker operating mechanism based on hesitant fuzzy numbers according to claim 1, characterized in that, The calculation of the Mel frequency cepstral coefficients in S1 includes: processing the audio signal in frames, passing it through a Mel filter bank after Fourier transform, taking the logarithmic energy and performing discrete cosine transform for dimensionality reduction; the calculation of the short-time energy includes: taking the mean of the sum of squares of the audio signal sample values ​​within the time window; the calculation of the zero-crossing rate includes: counting the number of symbol changes of adjacent audio samples within the time window.

3. The method for detecting the acoustic signature status of a high-voltage circuit breaker operating mechanism based on hesitant fuzzy numbers according to claim 1, characterized in that, The symptom feature vector in S2 includes a single feature vector corresponding to each symptom parameter individually, a double feature vector combining any two symptom parameters, and a full feature vector combining three symptom parameters. The single feature vector includes a Mel frequency cepstral coefficient feature vector, a short-time energy feature vector, and a zero-crossing rate feature vector. The double feature vector includes a feature vector combining Mel frequency cepstral coefficients and short-time energy, a feature vector combining short-time energy and zero-crossing rate, and a feature vector combining Mel frequency cepstral coefficients and zero-crossing rate. The full feature vector is composed of a combination of Mel frequency cepstral coefficients, short-time energy, and zero-crossing rate.

4. The method for detecting the acoustic signature status of a high-voltage circuit breaker operating mechanism based on hesitant fuzzy numbers according to claim 1, characterized in that, In the fuzzy deep residual shrinkage network, the fuzzy model parameters of each symptom feature vector are fine-tuned through a fully connected layer, and the evaluation values ​​of the three states corresponding to each symptom feature vector are expressed as follows: Among them, z N z is the evaluation value for the normal state. A z is the evaluation value for the state of attention. D F is the assessment value for a hazardous condition. l (p) represents the input feature value of the p-th neuron in the l-th layer of the neural network, W plq,N W plq,A and W plq,D The weight vectors for normal, attentive, and dangerous states, respectively, b lq,N b lq,A and b lq,D This is a bias term.

5. The method for detecting the acoustic signature status of a high-voltage circuit breaker operating mechanism based on hesitant fuzzy numbers according to claim 4, characterized in that, The membership function is obtained based on the evaluation values ​​of the three states corresponding to each symptom feature vector. The expression for the membership function y(z) is: Where a1, a2, and a3 are the baseline offsets of the membership functions controlling normal, alert, and dangerous states, respectively; b1, b2, and b3 are the rates of change of the membership functions controlling normal, alert, and dangerous states, respectively; and z N z is the evaluation value for the normal state. A z is the evaluation value for the state of attention. D This is the assessment value for a hazardous condition.

6. The method for detecting the voiceprint status of a high-voltage circuit breaker operating mechanism based on hesitant fuzzy numbers according to claim 1, characterized in that, The hesitation fuzziness number in S3 is obtained from the membership degree corresponding to each symptom feature vector. The hesitation fuzziness number is a triple, including the membership degree of the normal state, the attentive state, and the dangerous state. The collective hesitation fuzziness evaluation matrix includes the membership degree groups of the three states corresponding to each symptom feature vector, and the expression is: in, For the collective hesitation fuzzy assessment matrix, M is the number of types of symptom feature vectors, which is 7, (h N,i (z N ),h A,i (z A ),h D,i (z D ) represents the membership group corresponding to the feature vector of the i-th symptom class, h N,i (z N ) represents the membership degree of the normal state corresponding to the i-th type of symptom feature vector, h A,i (z A ) represents the membership degree of the attentional state corresponding to the feature vector of the i-th symptom class, h D,i (z D ) represents the membership degree of the dangerous state corresponding to the feature vector of the i-th type of symptom.

7. The method for detecting the acoustic signature status of a high-voltage circuit breaker operating mechanism based on hesitant fuzzy numbers according to claim 1, characterized in that, The optimal-worst method in S4 specifically involves obtaining the relative performance priority of each membership group to the optimal and worst membership groups through the selected optimal and worst membership groups, and then solving for the evidence weights based on these relative performance priorities in an optimization problem. The optimization problem is to minimize the maximum deviation, expressed as: Among them, w B The evidence weight for the optimal membership group, w W For the evidence weight of the worst membership group, AB i As the relative performance priority between the best membership group and the i-th membership group, a i W represents the relative performance priority between the worst membership group and the i-th membership group. i Let be the evidence weight for the i-th membership group.

8. The method for detecting the acoustic signature status of a high-voltage circuit breaker operating mechanism based on hesitant fuzzy numbers according to claim 1, characterized in that, The K-means clustering algorithm in S4 is as follows: randomly select initial cluster centers according to the preset number of clusters, calculate the Euclidean distance from the current health index to each initial cluster center, assign the current health index to the nearest cluster, and update the cluster center to the intra-cluster mean. Repeat the iteration to perform the corresponding operation of the health index to each cluster center until the iteration result is stable, and output the operating mechanism status corresponding to the current nearest cluster.

9. An electronic device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the program, it implements the method as described in any one of claims 1 to 8.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1 to 8.

Images