A defect detection model training method, a defect detection method, and related devices
By extracting the nonlinear vibration characteristics of the GIS disconnector base under variable frequency current excitation, constructing a variable frequency combined feature matrix and training a support vector machine model, the problem of low detection accuracy under a single power frequency current is solved, and higher defect detection accuracy and early defect discovery are achieved.
Patent Information
- Application Number
- CN202310574163.X
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2023-05-19
- Publication Date
- 2025-12-05
- Estimated Expiration
- 2043-05-19
AI Technical Summary
Existing technologies have low accuracy in detecting mechanical defects in GIS disconnect switches under a single power frequency current, making it difficult to effectively detect loosening of the disconnect switch base.
By acquiring the vibration signal of the loosening defect of the GIS disconnect switch base under variable frequency current excitation, nonlinear vibration feature quantities are extracted, a variable frequency combined feature matrix is constructed, and a support vector machine model is used for training. The penalty factor and Gaussian kernel function are optimized, and the parameters are optimized by using the gray wolf optimization algorithm to construct a defect detection model.
It improves the accuracy of mechanical defect detection in GIS disconnect switches, especially increasing the detection accuracy by 14.46% for minor defects, which helps to detect defects early and prevent them from worsening.
Smart Images
Figure CN116467605B_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the field of electrical equipment mechanical defect detection, and particularly relates to a defect detection model training method, a defect detection method and related equipment. BACKGROUND
[0002] The isolator is a core component unit of GIS (Gas Insulated Switchgear, gas insulated switchgear) equipment, and is applied in a large quantity and frequently operated. With the increase of the service life of the GIS equipment and the high-frequency switch operation, the wear of the components of the isolator gradually intensifies, and the vibration generated during the operation of the equipment can easily cause the loosening of the bolts of the isolator base, and long-term development can cause insulation failure. Therefore, the detection and research on the mechanical state of the GIS isolator are the key to ensure the safe and stable operation of the equipment. The GIS equipment is often caused by defects to cause damping to decrease and stiffness to soften, and to produce nonlinear dynamics behavior, and the appearance of this phenomenon is closely related to the frequency of the excitation current. At present, the detection and diagnosis of the mechanical defects of the GIS isolator are mostly based on a single power frequency current, and the defect detection accuracy under a single frequency current is low. SUMMARY
[0003] The purpose of the present application is to provide a defect detection model training method, a defect detection method and related equipment applied to the field of electrical equipment mechanical defect detection, which constructs a defect detection model by acquiring the vibration signal of the loosening defect of the GIS isolator base under the excitation of the variable frequency current. Since the nonlinear dynamics behavior of the GIS equipment under the defect state is closely related to the frequency of the excitation current, the nonlinear characteristics of the GIS isolator under the excitation of the variable frequency current are more significant, and compared with the detection of the GIS mechanical defect based on the single power frequency current in the prior art, the detection accuracy is improved.
[0004] To solve the above technical problems, the present application provides a defect detection model training method, comprising:
[0005] Acquiring the vibration signal of the loosening defect of the GIS isolator base under the excitation of the variable frequency current;
[0006] Extracting the nonlinear vibration characteristic quantity in the vibration signal;
[0007] Constructing a variable frequency combination feature matrix according to the nonlinear vibration characteristic quantity;
[0008] Inputting the variable frequency combination feature matrix into model training to obtain a defect detection model.
[0009] Optionally, the inputting of the variable frequency combination feature matrix into model training to obtain a defect detection model comprises:
[0010] input the variable frequency combination matrix into a support vector machine model for training;
[0011] call a grey wolf optimization algorithm to optimize a penalty factor and a Gaussian kernel function of the support vector machine model;
[0012] call a five-fold cross-validation method to verify the penalty factor and the Gaussian kernel function;
[0013] determine the support vector machine model after verification as the defect detection model.
[0014] Optionally, the variable frequency combination feature matrix is constructed according to the nonlinear vibration characteristic quantity, and the method comprises the following steps of:
[0015] an entropy weight method is called to construct the variable frequency combination feature matrix corresponding to each state label and the nonlinear vibration characteristic quantity under each frequency.
[0016] Optionally, the defect detection model is obtained by inputting the variable frequency combination feature matrix into a model for training, and the method comprises the following steps of:
[0017] the variable frequency combination feature matrix of a preset interval frequency band is input into the corresponding model to obtain the corresponding defect detection model.
[0018] Optionally, the nonlinear vibration characteristic quantity in the vibration signal is extracted, and the method comprises the following steps of:
[0019] waveform factors, pulse factors, margin factors, center of gravity frequencies, frequency standard deviations and amplitude square sums in the vibration signal are extracted.
[0020] To solve the above technical problems, the application further provides a defect detection method, which comprises the following steps of:
[0021] extracting a to-be-processed nonlinear vibration characteristic vector in a to-be-processed GIS disconnector base loosening defect vibration signal;
[0022] constructing a to-be-processed variable frequency combination feature matrix according to the nonlinear vibration characteristic vector;
[0023] inputting the to-be-processed variable frequency combination feature matrix into a defect detection model to obtain a defect state;
[0024] wherein the defect detection model is a model trained according to any one of the defect detection model training methods.
[0025] To solve the above technical problems, the application further provides a defect detection model training device, which comprises the following steps of:
[0026] a signal acquisition module is configured to acquire a vibration signal of a GIS disconnector base loosening defect under variable frequency current excitation;
[0027] The feature extraction module is configured to extract a nonlinear vibration characteristic quantity in the vibration signal.
[0028] The matrix construction module is configured to construct a variable-frequency combined feature matrix according to the nonlinear vibration characteristic quantity.
[0029] The defect detection module is configured to input the variable-frequency combined feature matrix into a model training to obtain a defect detection model.
[0030] To solve the above technical problems, the present application further provides a defect detection device, comprising:
[0031] The extraction module is configured to extract a to-be-processed nonlinear vibration characteristic vector in a to-be-processed GIS disconnector base loose defect vibration signal.
[0032] The construction module is configured to construct a to-be-processed variable-frequency combined feature matrix according to the nonlinear vibration characteristic vector.
[0033] The defect detection module is configured to input the to-be-processed variable-frequency combined feature matrix into a defect detection model to obtain a defect state.
[0034] The defect detection model is a model trained by the defect detection model training device.
[0035] To solve the above technical problems, the present application further provides a defect detection model training device, comprising:
[0036] The memory is configured to store a computer program.
[0037] The processor is configured to implement any one of the above defect detection model training methods and / or the defect detection method when executing the computer program.
[0038] To solve the above technical problems, the present application further provides a computer readable storage medium, wherein the computer readable storage medium stores computer executable instructions, and the computer executable instructions are executed by a processor to implement any one of the above defect detection model training methods and / or the defect detection method.
[0039] It can be seen that the present application obtains the vibration signal of the GIS disconnector base loosening defect under the variable frequency current excitation, extracts the nonlinear vibration characteristic quantity in the vibration signal, constructs a variable frequency combination feature matrix according to the nonlinear vibration characteristic quantity, inputs the variable frequency combination feature matrix into a model training to obtain a defect detection model. And by extracting the to-be-processed nonlinear vibration feature vector in the vibration signal of the to-be-processed GIS disconnector base loosening defect, constructing a to-be-processed variable frequency combination feature matrix according to the nonlinear vibration feature vector, and inputting the to-be-processed variable frequency combination feature matrix into the defect detection model to obtain the defect state. The method of the present application constructs a defect detection model by obtaining the vibration signal of the GIS disconnector base loosening defect under the variable frequency current excitation. Since the nonlinear dynamic behavior of the GIS device under the defect state is closely related to the frequency of the excitation current, the nonlinear characteristics of the disconnector defect under the variable frequency current excitation are more significant. Compared with the prior art of detecting the disconnector defect based on a single power frequency current, the detection accuracy is improved. BRIEF DESCRIPTION OF DRAWINGS
[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the drawings needed to be used in the embodiments or prior art description will be briefly introduced. Obviously, the drawings in the following description are only embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative labor on the basis of the provided drawings.
[0041] Figure 1 A flow chart of a defect detection model training method provided by an embodiment of the present application;
[0042] Figure 2 A scheme diagram provided by an embodiment of the present application;
[0043] Figure 3 A defect detection accuracy comparison diagram provided by an embodiment of the present application;
[0044] Figure 4 A structural block diagram of a defect detection model training device provided by an embodiment of the present application;
[0045] Figure 5 A structural block diagram of a defect detection device provided by an embodiment of the present application. DETAILED DESCRIPTION
[0046] With reference to the accompanying drawings, the technical solutions in the embodiments of the present application will be described clearly and completely. Obviously, the described embodiments are only a part but not all of the embodiments of the present application. Based on the embodiments in the present application, all the other embodiments obtained by a person of ordinary skill in the art without creative effort should fall into the scope of the present application.
[0047] The following will be described in detail Figure 1 , Figure 1 A flowchart of a defect detection model training method provided by an embodiment of the present application, which can include:
[0048] S101: Obtain a vibration signal of a disconnecting switch base under variable-frequency current excitation.
[0049] The present embodiment can obtain the vibration signal of the loosening defect of the GIS disconnecting switch base under variable-frequency current excitation. The present embodiment does not limit the acquisition method or the frequency range of the variable-frequency current, which can generally be 20-100 Hz. The present embodiment also does not limit the frequency gradient, which can generally be set as one frequency gradient per 10 Hz.
[0050] In the present embodiment, a state label can be set according to the defect state of the disconnecting switch base. The present embodiment does not limit the number of state labels, which can generally be four, including normal, mild, moderate and severe.
[0051] The present embodiment can obtain the vibration signal under each state label and each frequency. For example, when the current frequency is 10 Hz, the vibration signal under the normal state, the vibration signal under the mild defect state, the vibration signal under the moderate defect state and the vibration signal under the severe defect state are obtained.
[0052] S102: Extract a nonlinear vibration characteristic quantity in the vibration signal.
[0053] Further, the present embodiment can extract the nonlinear vibration characteristic quantity in the vibration signal under each state label and each frequency. The present embodiment does not limit the specific type of the nonlinear characteristic quantity, which can generally include six characteristic quantities, such as the waveform factor, the pulse factor, the margin factor, the center of gravity frequency, the frequency standard deviation and the amplitude square sum. The waveform factor is the ratio of the signal effective value to the rectified average value. The pulse factor is the ratio of the signal peak value to the rectified average value, which is used to detect whether there is an impact in the signal. The margin factor is the ratio of the signal peak value to the square root amplitude value, which is generally used to detect the wear condition of the equipment. The center of gravity frequency is generally used to describe the frequency of the signal component with larger components in the frequency spectrum, reflecting the distribution of the signal power spectrum. The frequency standard deviation is generally used to describe the dispersion degree of the power spectrum energy. The amplitude square sum is generally used to describe the amplitude variation characteristics of the frequency domain vibration signal.
[0054] The embodiment is not limited to a specific extraction method. The extraction method of the waveform factor can be shown in the following formula:
[0055]
[0056] In the formula, W is the waveform factor, N is the sample point number, x(n) is the time domain sequence of the signal, and n=1, 2, 3…, N.
[0057] In the embodiment, the extraction method of the general pulse factor can be shown in the following formula:
[0058]
[0059] In the formula, I is the pulse factor, N is the sample point number, max is the maximum function, x(n) is the time domain sequence of the signal, and n=1, 2, 3…, N.
[0060] In the embodiment, the extraction method of the general margin factor can be shown in the following formula:
[0061]
[0062] In the formula, L is the margin factor, N is the sample point number, max is the maximum function, x(n) is the time domain sequence of the signal, and n=1, 2, 3…, N.
[0063] In the embodiment, the extraction method of the general center of gravity frequency can be shown in the following formula:
[0064]
[0065] In the formula, FC is the center of gravity frequency, N is the sample point number, X(k) is the frequency domain signal after the signal is converted by FFT (Fast Fourier Transform), f(k) is the frequency value of the kth frequency domain signal point, and k=1, 2, 3…, N.
[0066] In the embodiment, the extraction method of the general frequency standard deviation can be shown in the following formula:
[0067]
[0068] In the formula, RVF is the frequency standard deviation, N is the sample point number, X(k) is the frequency domain signal after the signal is converted by FFT, FC is the center of gravity frequency, f(k) is the frequency value of the kth frequency domain signal point, and k=1, 2, 3…, N.
[0069] In the embodiment, the extraction method of the general amplitude square sum can be shown in the following formula:
[0070]
[0071] Wherein, E is the amplitude square sum, N is the sample point number, X(k) is the frequency domain signal after the signal is subjected to the FFT transformation, and k=1, 2, 3, …, N.
[0072] S103: Constructing a variable frequency combination feature matrix according to the nonlinear vibration characteristic quantity.
[0073] In the embodiment, the variable frequency combination feature matrix can be constructed according to the nonlinear vibration characteristic quantity. The embodiment is not limited to the specific way of constructing the variable frequency combination matrix. Generally, the entropy weight method can be used for feature fusion to construct the variable frequency combination feature matrix. In the embodiment, the specific construction method can be that the signal characteristic values obtained at different frequencies are assigned weights by using the entropy weight method, and the variable frequency combination feature matrix is constructed to highlight the influence of frequency on the defect signal characteristics. In the embodiment, the six feature vector sets {W, I, L, FC, RVF, E} extracted at a certain frequency can be defined as evaluation indexes {X1, X2, X3, X4, X5, X6}, wherein X i ={x1, x2, …, x n}, the standardized values of the index data are {Y1, Y2, Y3, Y4, Y5, Y6}, and the positive index calculation method can be as follows:
[0074]
[0075] In the formula, X ij is the i-th sample in the j-th evaluation index, Y ij is the i-th sample in the j-th standardized evaluation index, min(X i ) is the minimum value of the i-th sample in all j indexes, and max(X i ) is the maximum value of the i-th sample in all j indexes.
[0076] Further, the proportion of the i-th sample in the j-th standardized evaluation index is calculated, and the specific calculation method can be as follows:
[0077]
[0078] In the formula, p ij is the proportion of the i-th sample in the j-th standardized evaluation index, and Y ij is the i-th sample in the j-th standardized evaluation index.
[0079] Further, the information entropy of each index is calculated, and the specific calculation method can be as follows:
[0080]
[0081] In the formula, E j is the information entropy of the j-th index, and pij is the proportion of the i th sample in the j th evaluation index; wherein, when E j ≥ 0, if p ij = 0, E j = 0 is defined.
[0082] The weight of each index is calculated by information entropy, and the specific calculation method can be as shown in the following formula:
[0083]
[0084] In the formula, w j is the weight of the j th index, E j is the information entropy of the j th index, and in this embodiment, j = 1, 2, …, 6.
[0085] Finally, the weight of each feature sample is assigned, and the combination feature value in the variable frequency combination feature matrix is calculated, and the specific calculation method can be as shown in the following formula:
[0086]
[0087] In the formula, s i is the i th combination feature value in the variable frequency combination feature matrix, w j is the weight of the j th index, and X ij is the i th sample in the j th evaluation index.
[0088] In this embodiment, each state label and each frequency has its corresponding variable frequency combination matrix, and the number of variable frequency combination matrices is the product of the number of state labels and the number of frequencies.
[0089] S104: inputting the variable frequency combination feature matrix into a model to obtain a defect detection model.
[0090] In this embodiment, the variable frequency combination matrix can be input into a pre-set model as a training set to obtain a defect detection model, and the type of the model is not limited in this embodiment. Since the data in this embodiment has the characteristics of nonlinearity and small sample, a support vector machine (SVM) can be used as a training model, and the kernel function of the support vector machine is a Gaussian kernel function (RBF).
[0091] Further, the embodiment is not limited to a specific model training method. Since the penalty factor C and the RBF kernel function parameter sigma of the support vector machine have a great influence on the classification result, in the embodiment, the Grey Wolf Optimizer (GWO) which is simple and has strong global search ability can be used to globally optimize the parameters in the training process to enhance the learning and generalization ability of the SVM. The five-fold cross-validation method can be used to determine the optimal parameters, so as to obtain the optimal penalty factor C and RBF kernel function parameter sigma, and complete the training of the model.
[0092] In the embodiment, a plurality of frequency band intervals can be divided, and a corresponding defect detection model is trained according to each frequency band interval. The number of judgment intervals is not limited, and the interval range is not limited. The variable frequency combination matrix corresponding to each frequency band interval is input into the corresponding model to obtain the corresponding defect detection model.
[0093] The embodiment obtains the vibration signal of the GIS disconnector base loosening defect under the variable frequency current excitation to construct a defect detection model for defect detection of the disconnector base. Compared with the prior art of disconnector mechanical defect detection based on a single power frequency current, the detection accuracy is improved.
[0094] The following is a defect detection method provided by an embodiment of the application. The method can include:
[0095] Extracting a to-be-processed nonlinear vibration feature vector in a to-be-processed GIS disconnector base loosening defect vibration signal;
[0096] Constructing a to-be-processed variable frequency combination feature matrix according to the nonlinear vibration feature vector;
[0097] Inputting the to-be-processed variable frequency combination feature matrix into a defect detection model to obtain a defect state;
[0098] The defect detection model is a model trained according to any one of the defect detection model training methods.
[0099] The embodiment constructs a corresponding variable frequency combination feature matrix according to the to-be-processed GIS disconnector loosening defect vibration signal for defect detection, and inputs the variable frequency combination feature matrix into the model trained according to any one of the defect detection model training methods to realize defect detection. Compared with the prior art of disconnector defect detection based on a single power frequency current, the detection accuracy is improved.
[0100] The following is a specific embodiment provided by an embodiment of the application. The specific embodiment can include:
[0101] The waveform factor, pulse factor, margin factor, barycenter frequency, frequency standard deviation and amplitude square sum corresponding to the vibration signal of the GIS disconnector base loosening defect under the variable frequency current excitation mode of 20-100 Hz are extracted; wherein, each 10 Hz is set as a frequency gradient, and there are four state labels of normal, mild, moderate and severe;
[0102] The corresponding nonlinear characteristic quantities are fused under each frequency and state label by using the entropy weight method to construct a variable frequency combination feature matrix, and 4*9, i.e. 36 variable frequency combination feature matrices are obtained.
[0103] According to the applied excitation current frequency, the vibration signal is divided into four frequency bands, i.e. low frequency excitation band (20-40 Hz), medium frequency excitation band (50-70 Hz), high frequency excitation band (80-100 Hz) and full frequency excitation band (20-100 Hz), wherein the low frequency, medium frequency and high frequency each contain three frequencies, corresponding to 12 variable frequency combination feature matrices; the full frequency excitation band contains nine frequencies, corresponding to 36 variable frequency combination feature matrices.
[0104] The GIS mechanical defect state recognition SVM model is constructed for the variable frequency combination feature matrix under different excitation frequency bands. The low frequency excitation band corresponds to the low frequency band defect detection model, the medium frequency excitation band corresponds to the medium frequency band defect detection model, the high frequency excitation band corresponds to the high frequency band defect detection model, and the full frequency excitation band corresponds to the full frequency band defect detection model; wherein, the Gaussian kernel function is used as the kernel function, and the GWO algorithm is introduced in the training process to globally optimize the penalty factor C and the Gaussian kernel parameter σ, and the five-fold cross-validation method is used to determine the optimal parameters.
[0105] The nonlinear vibration characteristic vector in the vibration signal of the disconnector base loosening defect to be processed is extracted, and the variable frequency combination feature matrix to be processed is constructed according to the nonlinear vibration characteristic vector. The variable frequency combination feature matrix to be processed is input into the corresponding defect detection model, and the defect detection model will output the corresponding defect detection result. The overall scheme of the specific embodiment can be as shown in Figure 2 .
[0106] The variable frequency combination feature matrix and the feature matrix of the 50Hz signal are respectively input into the full frequency band defect detection model and the medium frequency band defect detection model to detect the defect recognition accuracy under the two excitation modes, and the defect recognition accuracy of the disconnector base loosening defect under different operating conditions under the two excitation modes is as shown in Figure 3 Compared with the single power frequency current excitation, the variable frequency current excitation can obtain higher defect detection accuracy, especially when the defect is mild, the accuracy of the variable frequency current excitation is increased by 14.46% compared with the single power frequency excitation, which is conducive to discovering the defect as early as possible when the defect has not developed to a serious state, and avoiding the defect from deteriorating.
[0107] The following describes the embodiments of the present application in detail Figure 4 , Figure 4 A structural block diagram of a defect detection model training device provided by an embodiment of the present application, which can include:
[0108] A signal acquisition module 100 is configured to acquire a vibration signal of a loosening defect of a GIS disconnector base under a variable-frequency current excitation.
[0109] A feature extraction module 200 is configured to extract a nonlinear vibration feature quantity in the vibration signal.
[0110] A matrix construction module 300 is configured to construct a variable-frequency combined feature matrix according to the nonlinear vibration feature quantity.
[0111] A defect detection module 400 is configured to input the variable-frequency combined feature matrix into a model training to obtain a defect detection model.
[0112] The defect detection model.
[0113] Based on the above embodiments, the defect detection model is constructed by acquiring a vibration signal of a GIS disconnector base loosening defect under a variable-frequency current excitation to detect defects of the disconnector base, which improves the detection accuracy compared with the prior art of detecting GIS disconnector mechanical defects based on a single power frequency current.
[0114] Based on the above embodiments, the defect detection module 400 can include:
[0115] A support vector machine unit is configured to input the variable-frequency combined matrix into a support vector machine model for training.
[0116] A grey wolf algorithm unit is configured to call a grey wolf optimization algorithm to optimize a penalty factor and a Gaussian kernel function of the support vector machine model.
[0117] A verification unit is configured to call a five-fold cross-validation method to verify the penalty factor and the Gaussian kernel function.
[0118] A determination unit is configured to determine the support vector machine model after verification as the defect detection model.
[0119] Based on the above embodiments, the matrix construction module 300 can include:
[0120] A weighted construction unit is configured to call an entropy weight method to construct the variable-frequency combined feature matrix according to each state label and the nonlinear vibration feature quantity under each frequency.
[0121] Based on the above embodiments, the defect detection module 400 can include:
[0122] The training model unit inputs the variable frequency combination feature matrix of the preset interval frequency band into the corresponding model to obtain the corresponding defect detection model.
[0123] Based on the above embodiments, the nonlinear vibration characteristic quantity in the vibration signal can include:
[0124] The extraction unit is configured to extract a waveform factor, a pulse factor, a margin factor, a barycentric frequency, a frequency standard deviation, and an amplitude square sum in the vibration signal.
[0125] The following will be described in combination with Figure 5 , Figure 5 The defect detection device provided by the embodiments of the present application has the structure diagram as shown in the figure, and the device can include:
[0126] The extraction module 500 is configured to extract a to-be-processed nonlinear vibration feature vector in a to-be-processed GIS disconnector base loose defect vibration signal.
[0127] The construction module 600 is configured to construct a to-be-processed variable frequency combination feature matrix according to the nonlinear vibration feature vector.
[0128]
[0129] The defect detection module 700 is configured to input the to-be-processed variable frequency combination feature matrix into a defect detection model to obtain a defect state.
[0130] The defect detection model is a model trained by the defect detection model training device.
[0131] Based on the above embodiments, the present application constructs a corresponding variable frequency combination feature matrix according to the to-be-processed GIS disconnector loose defect vibration signal for defect detection, and inputs the variable frequency combination feature matrix into a model trained by a defect detection model training device to realize defect detection.
[0132] Based on the above embodiments, the present application further provides a GIS mechanical defect detection device, which can include a memory and a processor, wherein the memory has a computer program stored therein, and the processor can realize the steps provided in the above embodiments when calling the computer program in the memory. Of course, the device can also include various necessary network interfaces, power supplies and other components.
[0133] The application further provides a computer readable storage medium, which has a computer program stored thereon, the computer program can implement the method provided by the embodiment of the application when executed by a terminal or a processor; the storage medium can include: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various storage medium capable of storing program codes.
[0134] The various embodiments are described in progressive manner in the specification, and each embodiment focuses on the difference from other embodiments, and the same or similar parts between the various embodiments can be referred to each other. For the device disclosed by the embodiments, since it corresponds to the method disclosed by the embodiments, the description is relatively simple, and the related parts can be referred to the method part.
[0135] Finally, it should be noted that, in this document, the relationship terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply that there is any such actual relationship or order between the entities or operations. Moreover, the terms "include", "contain" or any other variants thereof are intended to cover non-exclusive inclusion, so that the process, method, article or equipment including a series of elements not only includes those elements, but also includes other elements not explicitly listed or further includes other elements inherent in such process, method, article or equipment.
[0136] Without more limitations, the element defined by the statement "including a" does not exclude the presence of another identical element in the process, method, article or equipment including the element.
[0137] The above provides a detailed introduction to the defect detection model training method, defect detection method and related equipment provided by the application. The principle and implementation mode of the application are described by applying specific examples. The above embodiment description is only used to help understand the method and core idea of the application. For those skilled in the art, according to the idea of the application, the specific implementation mode and application range will be changed. In summary, the content of the specification should not be understood as a limitation of the application.
Claims
1. A defect detection model training method, characterized in that, The method comprises the following steps: obtaining a vibration signal of a loosening defect of a GIS disconnector base under a variable-frequency current excitation; extracting a nonlinear vibration characteristic quantity in the vibration signal; including: extracting a waveform factor, a pulse factor, a margin factor, a barycentric frequency, a frequency standard deviation and an amplitude square sum in the vibration signal; constructing a variable-frequency combined feature matrix according to the nonlinear vibration characteristic quantity; inputting the variable-frequency combined feature matrix into model training to obtain a defect detection model; including: inputting the variable-frequency combined feature matrix into a support vector machine model for training; calling a grey wolf optimization algorithm to optimize a penalty factor and a Gaussian kernel function of the support vector machine model; calling a five-fold cross-validation method to verify the penalty factor and the Gaussian kernel function; determining the verified support vector machine model as the defect detection model. 2.The defect detection model training method of claim 1, wherein, The variable-frequency combined feature matrix is constructed according to the nonlinear vibration characteristic quantity, including: calling an entropy weight method to construct the variable-frequency combined feature matrix corresponding to each state label and the nonlinear vibration characteristic quantity under each frequency. 3.The defect detection model training method of claim 1, wherein, The variable-frequency combined feature matrix is inputted into model training to obtain a defect detection model, including: inputting the variable-frequency combined feature matrix of a preset interval frequency band into corresponding model training to obtain corresponding defect detection models.
4. A defect detection method characterized by, The method comprises the following steps: extracting a to-be-processed nonlinear vibration characteristic vector in a to-be-processed GIS disconnector base loosening defect vibration signal; constructing a to-be-processed variable-frequency combined feature matrix according to the nonlinear vibration characteristic vector; inputting the to-be-processed variable-frequency combined feature matrix into a defect detection model to obtain a defect state; wherein the defect detection model is a model trained by the defect detection model training method according to any one of claims 1 to 3.
5. A defect detection model training apparatus characterized by comprising: The method comprises the following steps: a signal acquisition module for acquiring a vibration signal of a GIS disconnector base loosening defect under a variable-frequency current excitation; a feature extraction module for extracting a nonlinear vibration characteristic quantity in the vibration signal; including: extracting a waveform factor, a pulse factor, a margin factor, a barycentric frequency, a frequency standard deviation and an amplitude square sum in the vibration signal; a matrix construction module for constructing a variable-frequency combined feature matrix according to the nonlinear vibration characteristic quantity; a defect detection module for inputting the variable-frequency combined feature matrix into model training to obtain a defect detection model; including: inputting the variable-frequency combined feature matrix into a support vector machine model for training; calling a grey wolf optimization algorithm to optimize a penalty factor and a Gaussian kernel function of the support vector machine model; calling a five-fold cross-validation method to verify the penalty factor and the Gaussian kernel function; determining the verified support vector machine model as the defect detection model.
6. A defect detection apparatus characterized by comprising: The method comprises the following steps: an extraction module for extracting a to-be-processed nonlinear vibration characteristic vector in a to-be-processed GIS disconnector base loosening defect vibration signal; a construction module for constructing a to-be-processed variable-frequency combined feature matrix according to the nonlinear vibration characteristic vector; a defect detection module for inputting the to-be-processed variable-frequency combined feature matrix into a defect detection model to obtain a defect state; wherein the defect detection model is a model trained by the defect detection model training device of claim 5. 7.A defect detection model training device, comprising: The method comprises the following steps: a memory for storing a computer program; a processor for implementing the defect detection model training method according to any one of claims 1 to 3 and / or the defect detection method according to claim 4 when executing the computer program.
8. A computer-readable storage medium, characterized in that, The computer readable storage medium stores computer executable instructions, and the computer executable instructions are executed by the processor to implement the defect detection model training method according to any one of claims 1 to 3 and / or the defect detection method according to claim 4.