CN111879397A - Fault diagnosis method for energy storage mechanism of high-voltage circuit breaker - Google Patents
Fault diagnosis method for energy storage mechanism of high-voltage circuit breaker Download PDFInfo
- Publication number
- CN111879397A CN111879397A CN202010900568.4A CN202010900568A CN111879397A CN 111879397 A CN111879397 A CN 111879397A CN 202010900568 A CN202010900568 A CN 202010900568A CN 111879397 A CN111879397 A CN 111879397A
- Authority
- CN
- China
- Prior art keywords
- energy storage
- circuit breaker
- sound
- signal
- vibration
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 69
- 238000004146 energy storage Methods 0.000 title claims abstract description 55
- 230000007246 mechanism Effects 0.000 title claims abstract description 44
- 238000003745 diagnosis Methods 0.000 title claims abstract description 35
- 230000008569 process Effects 0.000 claims abstract description 27
- 239000011159 matrix material Substances 0.000 claims abstract description 23
- 238000012549 training Methods 0.000 claims abstract description 18
- 230000008859 change Effects 0.000 claims abstract description 16
- 230000006870 function Effects 0.000 claims abstract description 12
- 230000005236 sound signal Effects 0.000 claims description 29
- 238000011176 pooling Methods 0.000 claims description 10
- 230000000694 effects Effects 0.000 claims description 9
- 238000000354 decomposition reaction Methods 0.000 claims description 8
- 230000000877 morphologic effect Effects 0.000 claims description 8
- 210000002569 neuron Anatomy 0.000 claims description 5
- 230000011218 segmentation Effects 0.000 claims description 5
- 230000004044 response Effects 0.000 claims description 4
- 238000010200 validation analysis Methods 0.000 claims description 4
- 229910003460 diamond Inorganic materials 0.000 claims description 3
- 239000010432 diamond Substances 0.000 claims description 3
- 239000003607 modifier Substances 0.000 claims description 3
- 238000013507 mapping Methods 0.000 claims description 2
- 230000008901 benefit Effects 0.000 abstract description 4
- 238000010606 normalization Methods 0.000 abstract 1
- 238000013527 convolutional neural network Methods 0.000 description 32
- 238000010586 diagram Methods 0.000 description 8
- 238000004422 calculation algorithm Methods 0.000 description 6
- 239000013598 vector Substances 0.000 description 6
- 230000007797 corrosion Effects 0.000 description 5
- 238000005260 corrosion Methods 0.000 description 5
- 238000011160 research Methods 0.000 description 5
- 238000004364 calculation method Methods 0.000 description 4
- 238000000605 extraction Methods 0.000 description 4
- 238000005070 sampling Methods 0.000 description 4
- 230000004913 activation Effects 0.000 description 3
- 238000001914 filtration Methods 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000000513 principal component analysis Methods 0.000 description 3
- 238000012360 testing method Methods 0.000 description 3
- 230000000295 complement effect Effects 0.000 description 2
- 238000010276 construction Methods 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000004088 simulation Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000012795 verification Methods 0.000 description 2
- NAWXUBYGYWOOIX-SFHVURJKSA-N (2s)-2-[[4-[2-(2,4-diaminoquinazolin-6-yl)ethyl]benzoyl]amino]-4-methylidenepentanedioic acid Chemical compound C1=CC2=NC(N)=NC(N)=C2C=C1CCC1=CC=C(C(=O)N[C@@H](CC(=C)C(O)=O)C(O)=O)C=C1 NAWXUBYGYWOOIX-SFHVURJKSA-N 0.000 description 1
- 238000009825 accumulation Methods 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 238000013016 damping Methods 0.000 description 1
- 230000007423 decrease Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000002405 diagnostic procedure Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 238000009499 grossing Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000004615 ingredient Substances 0.000 description 1
- 238000009434 installation Methods 0.000 description 1
- 238000009413 insulation Methods 0.000 description 1
- 230000023886 lateral inhibition Effects 0.000 description 1
- 230000003278 mimic effect Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000009024 positive feedback mechanism Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000004321 preservation Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000035939 shock Effects 0.000 description 1
- 230000003595 spectral effect Effects 0.000 description 1
- 238000001228 spectrum Methods 0.000 description 1
- 230000001629 suppression Effects 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
- G01H11/00—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves by detecting changes in electric or magnetic properties
- G01H11/06—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves by detecting changes in electric or magnetic properties by electric means
- G01H11/08—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves by detecting changes in electric or magnetic properties by electric means using piezoelectric devices
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01H—MEASUREMENT OF MECHANICAL VIBRATIONS OR ULTRASONIC, SONIC OR INFRASONIC WAVES
- G01H17/00—Measuring mechanical vibrations or ultrasonic, sonic or infrasonic waves, not provided for in the preceding groups
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/048—Activation functions
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The invention relates to a fault diagnosis method for an energy storage mechanism of a high-voltage circuit breaker; firstly, removing background noise from an acquired acoustic signal by adopting morphology, providing a time scale alignment method based on kurtosis and envelope similarity to ensure the synchronism of acoustic vibration signals, then constructing a two-dimensional image characteristic matrix for the acoustic vibration signals after data expansion by using a Pearson correlation coefficient, finally training the characteristic matrix by using CNN, improving a CNN model structure by adopting local mean normalization and kernel function decorrelation, and reducing the influence of large data change on the diagnosis accuracy of a circuit breaker energy storage mechanism in an energy storage process. The invention has the advantages of high total diagnosis accuracy up to 98.1%, good generalization performance and obvious advantages compared with the traditional method.
Description
Technical Field
The invention relates to a fault diagnosis method for an energy storage mechanism of a high-voltage circuit breaker, in particular to a fault diagnosis method for the energy storage mechanism of the circuit breaker, which constructs a CNN characteristic matrix by using a sound vibration signal.
Background
As an important control and protection device in a power system, a high-voltage circuit breaker is important for fault diagnosis of the circuit breaker because whether the circuit breaker can reliably operate directly affects safety and stability of the power system. The research of the current circuit breaker is focused on the switching-on and switching-off process: the mechanical fault is identified by utilizing coil current, insulation pull rod displacement and vibration signals, the research focuses on the problem of the breaker in the operation process, the problem of the energy storage process is not deeply researched, quantitative judgment basis is lacked, and how to accurately find the fault type and the development change rule of the energy storage mechanism is worth deeply discussing.
The vibration signal is commonly used for diagnosing the fault of the circuit breaker, and is characterized by rich state information and high signal-to-noise ratio, but in practical application, the vibration signal has a saturation phenomenon when the amplitude is large, and high-frequency impact failure caused by a charge accumulation effect is easy to generate. The sound signal can effectively avoid the saturation failure phenomenon due to the wide measuring frequency band, the sound pick-up is convenient to install, and the influence of the installation mode on the signal is small. The sound vibration belongs to homologous signals and is generated by the vibration of the components of the circuit breaker, and the homologous complementary characteristics of the sound vibration and the components of the circuit breaker can be utilized to exert respective advantages and jointly diagnose the fault of the energy storage mechanism of the circuit breaker.
For a sound and vibration signal combined method, Zhao billows and the like, in a high-voltage circuit breaker vibration and sound combined fault diagnosis method, improved Ensemble Empirical Mode Decomposition (EEMD) is carried out on sound and vibration signals, and then two-dimensional spectral entropy is obtained from the decomposed IMF and is used as a feature vector for fault diagnosis. And in the universal circuit breaker fault vibration sound diagnosis method based on multi-feature fusion and improved QPSO-RVM, decomposing the sound vibration signals by utilizing a complementary general empirical mode decomposition algorithm, and solving the sample entropy, the energy coefficient and the power spectrum entropy of the IMF components as feature vectors to perform fault diagnosis. HE Mengyuan et al, in Research of circulation breaker in ingredient family method based on double clustering, utilizes fuzzy peak value to optimize C-means clustering, and combines SVM to diagnose the vibro-acoustic signal. The method achieves certain effects, but has the following defects:
(1) the difference of the sound vibration signals is not considered, and the sound vibration signals are combined mechanically, so that the diagnosis accuracy is not high enough.
(2) The method for extracting the features depends on manual selection and expert knowledge, and is too subjective and easy to cause omission of fault information. The extracted features also need to be selected and then a proper classifier is selected for fault classification, a proper feature extraction method needs to be selected continuously in a targeted mode, and the generalization capability is poor.
With the continuous development of the deep learning theory, the CNN is used as a typical deep learning algorithm, has strong feature self-learning capability, good adaptivity and strong parallel processing capability, and has obvious advantages in fault diagnosis of mechanical equipment. And in the high-voltage circuit breaker fault diagnosis based on the convolutional neural network algorithm, the vibration signals are converted into frequency domain signals by using fast Fourier transform, and the fault diagnosis is carried out by adopting a one-dimensional convolutional neural network (1DCNN) after the preprocessing. And the method comprises the steps of performing a fault classification on a rotating machine by using a Softmax regression model, and the like, wherein in the intelligent fault diagnosis method for the rotating machine based on the one-dimensional convolutional neural network, a vibration signal is directly used as an input, 1DCNN is adopted to extract features, and a Softmax regression model is adopted to realize the fault classification. In the wind turbine generator set planetary gearbox fault detection based on the one-dimensional convolutional neural network and the Soft-Max classifier, the Lidongdong and the like adopt EEMD to decompose vibration signals, then use Hilbert transform to extract fault characteristics, and use CNN to realize fault identification. Most of the researches are fault diagnosis performed on a single signal based on 1DCNN, the diagnosis process is complicated, the accuracy is not high enough, and the self-learning capability of the CNN is not maximized.
Disclosure of Invention
The invention aims to provide a fault diagnosis method for a circuit breaker energy storage mechanism by utilizing a sound vibration signal to construct a CNN characteristic matrix.
The invention adopts the following technical scheme:
a fault diagnosis method for an energy storage mechanism of a high-voltage circuit breaker comprises the following steps:
(1) collecting vibration signals and sound signals of the circuit breaker in the energy storage process;
(2) carrying out morphological denoising on the collected sound signal;
(3) carrying out sound vibration time scale alignment on the sound signal subjected to the morphological denoising and the vibration signal;
(4) performing overlapped data expansion on the data processed in the step (3);
(5) constructing a CNN characteristic matrix of the acoustic vibration signal from the data obtained in the step (4);
(6) optimizing the CNN model;
(7) and inputting the optimized CNN model to identify the fault of the energy storage mechanism.
Wherein the step (1) is specifically as follows: simultaneously collecting vibration signals and sound signals in the energy storage process of the circuit breaker by using a piezoelectric transducer with the frequency range of 1-10kHz and a sound pick-up with the frequency range of 20-20 kHz; the vibration sensor is adsorbed on a connecting fixing piece below a spring vibration body of the high-voltage circuit breaker energy storage mechanism, and the sound sensor is arranged at a sound source of 30-50 cm.
Wherein the step (2) specifically comprises: performing morphological denoising on the sound signal by adopting an opening-closing and closing-opening operation mixed filter based on morphology, and selecting an ellipse b1And diamond b2Structural elements, width of 1/40-1/10, b of maximum value of selected signal1Is set to 12, b2Set to 5.
Wherein the step (3) is specifically as follows:
(a) performing Hilbert transform on the sound signal and the vibration signal to obtain an envelope, dividing the sound vibration signal into N equal parts, calculating the kurtosis of each part, comparing the kurtosis values of each section of signal, and searching a signal section with obvious kurtosis value change difference so as to determine the time of the change;
(b) after the corresponding time periods of the sound signal and the vibration signal are determined, the similarity is judged by utilizing a Min's distance formula, namely q roots of the sum of the sound and vibration signal envelope absolute difference value q powers, after the time with the highest similarity is found, the time of subtracting the vibration starting time from the sound starting time is delta T, and the sound signal can be aligned with the vibration signal by advancing the delta T.
Wherein the step (4) is specifically as follows: for a signal x with the length of N, setting the sample length as L and the overlapping rate as lambda, and adopting the following method:
(A) the maximum number of partitionable samples under the current signal length is calculated:
wherein
The operator is a round-down operator;
(B) solving each segmentation sample; the position of the ith sample in the original signal is represented as follows:
xi＝X[(i-1)×L×(1-λ)+(0:1)×L]，i∈[1,n]
wherein x isiIs the segmented sample data.
The sample length L is set to 1024 and λ is 0.5.
Wherein the step (5) is specifically as follows: the correlation of the sound signal and the vibration signal change is described by using the Pearson product moment correlation coefficient, and the value of the correlation coefficient is used as a matrix element.
Wherein the step (6) is specifically as follows: performing wavelet decomposition on the kernel function to obtain multi-resolution wavelet coefficients, and selecting the wavelet decomposition coefficients in mutually orthogonal directions to process convolution kernel error modifiers to remove the correlation of the kernel function; at the same time, adding an LRN after each convolution-pooling layer reduces the effect of more parameters needed by the model.
Wherein, in the step (6), the LRN layer normalized response is expressed as:
wherein n represents the number of adjacent mapping kernels passing through the same spatial position; n is the total number of cores of the layer;
represents the ReLU nonlinear neuron output for the ith nucleus at (x, y); k, α, β are validation set superparameters with values of 2, 0.0002 and 0.5, respectively.
Wherein, Early-Stopping mechanism is introduced into the model full-connection layer in the training process of the step (6), and the coefficients are 0.5 and 0.01 respectively.
And (6) acquiring multiple groups of sound signals and vibration signals for training the model under the conditions that the high-voltage circuit breaker energy storage mechanism is normal, the spring is corroded, the mechanism is jammed and the spring falls off.
The invention has the beneficial effects that:
1) according to the invention, the vibration signal and the sound signal are organically combined, so that the fault identification accuracy of the energy storage mechanism of the high-voltage circuit breaker is improved.
2) The method comprises the steps of effectively editing the sound vibration signal by using morphological denoising and a time scale alignment method based on kurtosis-envelope similarity, acquiring a large amount of data required by training by adopting data expansion, and constructing a sound vibration signal characteristic matrix by combining Pearson product moment correlation coefficients, so that a new idea is developed for fault identification of a circuit breaker energy storage mechanism.
3) A CNN fault diagnosis method is introduced, and the model is improved by combining the characteristics of the energy storage process, so that the generalization performance of the model is improved.
Drawings
FIG. 1 is a flow chart of the method of the present invention.
FIG. 2 is a comparison graph of results before and after noise reduction.
Fig. 3 is a schematic diagram of the hysteresis vibration of an acoustic signal.
Fig. 4 is a schematic diagram of data expansion.
FIG. 5 is a two-dimensional map of a feature matrix.
Fig. 6 is a schematic diagram of CNN structure.
Fig. 7 is a diagram of the spring corrosion operating state in the energy storage process of the high-voltage circuit breaker.
Fig. 8 is a mechanism jamming operation state diagram in the energy storage process of the high-voltage circuit breaker.
Fig. 9 is a diagram of the spring falling operation state in the energy storage process of the high-voltage circuit breaker.
FIG. 10 is a schematic diagram of model training.
Fig. 11 is a PCA visualization.
FIG. 12 is a graph illustrating the accuracy of the model.
FIG. 13 is a graph showing loss values of the model
FIG. 14 is a graph comparing the diagnostic effects of different models.
Detailed Description
The technical solutions are described below clearly and completely with reference to the embodiments of the present invention and the accompanying drawings, and it is obvious that the described embodiments are only a part of the embodiments of the present invention, not all of the embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
1 sound vibration signal construction CNN characteristic matrix
1.1 Circuit breaker mechanical Fault diagnosis procedure
The change rule of the sound vibration signal is closely related to the mechanical state in different states of the energy storage process, the implicit relation between the sound vibration signal and the mechanical state can be accurately reflected through the sound vibration characteristic matrix, and therefore fault diagnosis of the circuit breaker is achieved.
Because the noise of the acoustic signal background is greatly interfered, firstly, morphology is adopted to remove background noise, then time scale alignment is carried out on the acquired acoustic vibration signal based on kurtosis and envelope similarity, overlapped data is adopted to expand capacity and increase matrix dimension after alignment, finally, a CNN characteristic matrix of the acoustic vibration signal is constructed by using a Pearson product moment correlation coefficient, a CNN optimization model is input to carry out fault identification, and the diagnosis process is shown in figure 1.
1.2 Acoustic Signal morphological De-noising
Morphology can be based on the geometrical characteristics of the sound signals in the energy storage process, and sound signal matching, signal extraction, detail preservation and noise suppression are realized by performing morphology transformation on structural elements. The method only needs algebraic operation without calculating frequency domain information, has considerable speed, has efficient nonlinear filtering function, and is suitable for filtering noise signals in the energy storage process of the circuit breaker.
Opening and closing operations are used for smoothing signals in different modes, and the opening operations can be used for filtering peak noise and removing signal edge burrs; the closed operation is used to smooth and suppress the valley signal noise. Aiming at the characteristics of sound signals in the energy storage process of the circuit breaker, the research adopts an open-close and close-open operation mixed filter based on morphology, and selects an ellipse b1And diamond b2Structural elements, width selected from 1/40 to 1/10, b of maximum value of signal1Is set to 12, b2The setting is 5, and the denoising effect of the partial sound signal under the condition that the voltage of the energy storage mechanism is lower is shown in fig. 2.
The signal-to-noise ratio before and after denoising is improved from 13.56dB to 24.38dB through verification of the upper graph and multiple groups of actual signals, and the denoising result is obvious. The background noise of the denoised acoustic signal is obviously reduced, and the detail information of the original signal is reserved.
1.3 time scale alignment of sound and vibration
Due to the difference in the propagation speed of the sound vibration signal and the difference in the placement position of the sensor, when the circuit breaker starts to store energy, the collected sound wave lags behind the vibration signal Δ T, as shown in fig. 3.
In order to accurately judge the fault of the energy storage mechanism of the circuit breaker, the sound vibration signals of the circuit breaker at the same moment must be analyzed and compared. The kurtosis is taken as a dimensionless parameter, is particularly sensitive to signal impact, and can be used for detecting the kurtosis of the envelope curve of the sound vibration signal. The method includes the steps of carrying out Hilbert transform on a sound vibration signal to obtain an envelope, dividing the sound vibration signal into N equal parts, calculating the kurtosis of each part, comparing the kurtosis values of each part of the signal, and searching a signal section with obvious kurtosis value change difference so as to determine the time of the change. The kurtosis calculation formula is as follows:
in formula (1), E refers to the mean value; x is a sound signal; μ is the mean of the signal x; σ is the standard deviation of the signal x.
After the time period corresponding to the sound-vibration signal is determined, a Min's distance formula is utilized, as shown in formula (2), namely, the similarity is judged according to the q-th root of the sum of the sound and vibration signal envelope absolute difference value q-th power, after the time with the highest similarity is found, the time when the sound starts minus the vibration start time is delta T, and the sound signal can be aligned with the vibration signal by advancing the delta T.
In formula (2), c and d are the sound and vibration signal data points respectively; n is the data point dimension and q is the distance adjustment parameter.
1.4 overlapped data expansion
By adopting the overlapped sample segmentation, the relevance of adjacent samples can be completely reserved, the feature loss caused by sample truncation is avoided, a large amount of data is provided for constructing a sound vibration combined feature matrix, the dimensionality of the feature matrix is increased, and a foundation is laid for accurate fault diagnosis of CNN.
For a signal x with the length of N, setting the sample length as L and the overlap ratio as lambda, and adopting the capacity expansion and division method as follows:
(1) the maximum number of partitionable samples under the current signal length is calculated:
wherein
The rounding-down operator.
(2) Each segmented sample is evaluated. The position of the ith sample in the original signal is represented as follows:
xi＝X[(i-1)×L×(1-λ)+(0:1)×L]，i∈[1,n](3)
in the formula (3), xiFig. 4 is a schematic diagram of capacity expansion for the segmented sample data.
The short sample segmentation length can improve the convergence speed of the model and save the training time, but the loss of nonlinear characteristic information is easy to cause; too long a sample segmentation length may affect the convergence speed of the model and affect the real-time performance of the diagnosis. By experimental analysis, the sample length was set to 1024 and λ was 0.5.
1.5 construction of the Acoustic-vibration Joint matrix
The rows and columns of the CNN characteristic matrix respectively represent values corresponding to sampling points, and the sample number and the parameter setting of the acoustic vibration signals are the same. When the breaker breaks down, the values of the sound and vibration signals change, the corresponding fault types are reflected by the complex and various sound and vibration change relations, the correlation of the sound and vibration signal changes is described by adopting Pearson product moment correlation coefficients, and the correlation coefficient values are used as matrix elements.
The Pearson product-moment correlation coefficient is used for measuring the change relation between two variables, is independent of specific values of the variables, and is a non-parameter statistic. If the mean values of the vibro-acoustic signals tend to be greater than or less than their respective mean values at the same time, the correlation coefficient is positive; otherwise, the correlation coefficient is negative. The calculation is as follows:
in the formula (4), { xiI ═ 1,2, …, n } and { y ═ yiI ═ 1,2, …, n } represent the values of the sound and vibration signal sample points, respectively; x and y are the mean values of the samples in this group.
The Pearson coefficients are normalized to fit the feature matrix, since they range from [ -1,1 ]. In order to facilitate observation of values of matrix elements, 40 sample data continuously collected in the energy storage spring falling state are selected to calculate a characteristic matrix, and a two-dimensional graph is drawn as shown in fig. 5.
The correlation of the change of the sound vibration signal under the state that the energy storage spring of the circuit breaker falls off can be observed from the graph shown in fig. 5, but the fault type cannot be directly identified from the two-dimensional graph, and at the moment, fault diagnosis can be carried out by means of the strong feature extraction and identification capability of the CNN.
2CNN diagnostic methods and optimizations
2.1 CNN diagnostic principle
CNN, as a multi-layer neural network structure, is composed of an input layer, a convolutional layer, a pooling layer, a full-link layer, and an output layer, as shown in fig. 6.
The convolution layer performs convolution on input data through convolution kernel, and constructs a feature vector by using a nonlinear activation function, wherein the calculation process is as follows:
in the formula (5), the first and second groups,
represents the l-th layer input; n is a radical ofjRepresenting an input feature vector; l represents the l-th network; k represents a convolution kernel; b represents the bias of the convolution kernel. And a modified linear unit (ReLU) is selected as a nonlinear activation function, so that the network sparsity is improved, and the overfitting problem is suppressed. The calculation of ReLU is as follows:
in the formula (6), the first and second groups,
is that
An activation value of;
representing the output value of the convolution operation.
The pooling layer contains the average pooling and the maximum pooling, and its transformation function is as follows:
in formula (7), W is the kernel width;
is the value of the t neuron in the ith feature of the l layer;
the value for the l +1 th neuron.
The CNN output layer is used for fully connecting the output of the last pooling layer, and then a Softmax classifier is used for solving the multi-classification problem, wherein the model is as follows:
O＝f(bo+fvwo) (8)
in the formula (8), boIs a deviation vector; f. ofvIs a feature vector; w is aoIs a weight matrix.
2.2CNN model optimization
The convolutional neural network has strong generalization capability due to a special structure, but for the circuit breaker, mechanical parts start, move and stop according to a certain sequence in the energy storage process to generate a series of composite signals superposed by shock waves, and the improvement of CNN generalization performance and accuracy is limited due to the fact that the measured signals change greatly.
Correlation exists among CNN convolution kernels, and the smaller the correlation is, the more comprehensive the features of convolution extraction are, and the better the generalization and accuracy are. In order to improve the generalization and the accuracy, the wavelet decomposition is carried out on the kernel function to obtain a multi-resolution wavelet coefficient, and the wavelet decomposition coefficients in the mutually orthogonal directions are selected to process the convolution kernel error modifier to remove the correlation of the kernel function. At the same time, adding an LRN after each convolution-pooling layer reduces the effect of more parameters needed by the model. The LRN layer may mimic the "lateral inhibition" mechanism of an organism, with a positive feedback mechanism of response, and the normalized response may be expressed as:
in the formula (9), n represents the same spatial position passing through the vicinityMapping the number of kernels; n is the total number of cores of the layer;
represents the ReLU nonlinear neuron output for the ith nucleus at (x, y); k, α, β are validation set superparameters with values of 2, 0.0002 and 0.5, respectively.
In the training process, in order to avoid the over-fitting phenomenon, an Early-Stopping mechanism is introduced into a model full-connection layer, the coefficients of the Early-Stopping mechanism are 0.5 and 0.01 respectively, and the model structure disclosed by the invention is as follows:
TABLE 1 CNN Structure Table
3 Experimental results and analytical verification
3.1 Fault simulation and parameter setting
A piezoelectric (CK 8605) sensor with the frequency range of 1-10kHz and a (WM-025N) sound pickup with the frequency range of 20-20kHz are selected, and vibration and sound signals of the energy storage process of the circuit breaker are collected simultaneously. Wherein the vibration sensor is adsorbed on the connecting fixing piece below the spring vibration body, and the sound sensor is arranged at the position of 30-50cm of the sound source. The method is characterized in that the ZN65-12 type circuit breaker is subjected to energy storage normal, spring corrosion, mechanism jamming and spring falling tests respectively, and multiple groups of sound and vibration signals are collected in different states.
The circuit breaker energy storage in-process is changed into the old spring of corrosion and is tested and simulated the spring corrosion, adopts the plank to block pivot mechanism and increases damping simulation mechanism bite, unscrews the fastening screw of circuit breaker spring lower extreme and simulates the spring trouble that drops. The field test patterns are shown in fig. 7 to 9.
The number of CNN convolution layers is 2, parameters are set to be 32@2 x 2 and 64@2 x 2, the maximum pooling with the size of 2 x 2 and the step length of 2 is adopted by the pooling layer, nodes of the two fully-connected layers are set to be 256 and 64, and Softmax is adopted as a classifier for classification. An RMSprop optimizer is adopted, the initial learning rate is set to be 0.03, and the attenuation rate is 0.99; the number of iterations is 50, and Dropout is set for the fully connected layer by a factor of 0.5. The training steps of the model are shown in fig. 10.
3.2 results of the experiment
350 groups of data are collected for each type of normal energy storage, spring corrosion, mechanism jamming and spring falling, 220 groups of data are used as training, 130 groups of data are used as testing, and each group of data comprises 50000 sampling points. For the ZN65-12 model breaker, the time from the output torque of the switch energy storage motor to the energy storage holding of the energy storage holding engine is 10s, so the sampling frequency is set to be 40kHz, 400000 points can be obtained in one energy storage period of the breaker, and 8 96 multiplied by 96 characteristic matrixes are formed. The model is trained for 50 times, the minimum Mean Square Error (MSE) is used as a loss function, and the formula is as follows:
in the formula (10), the first and second groups,
for the prediction of the ith sample,
the value is classified for the ith sample. To verify the learned effect, the first two principal components are extracted using Principal Component Analysis (PCA) to visualize the learned features of the penultimate layer (fully connected layer), as shown in fig. 11.
As can be seen from fig. 11, the learning features of the models in the respective states are clustered in the corresponding regions (circles in the figure indicate cluster centers), which indicates that the models are well distinguishable.
The results of classification using Soft-Max as a classifier are shown in fig. 12 (the upper half of the graph indicates accuracy, and the lower half indicates loss value).
As can be seen from fig. 12 and 13, as the number of times of training increases, the model recognition accuracy gradually increases, and becomes stable after 35 times of training, and the accuracy is not improved. Meanwhile, the loss value gradually decreases, and the training effect of the convolutional neural network is continuously optimized at the moment. Although the training times are set to be 50 times, the accuracy is not improved after the iteration is performed for 40 times, and the loss value is correspondingly reduced to the minimum, because the Early-Stopping mechanism is introduced into the model, the training of the model is stopped when the accuracy and the loss value of the model are not obviously changed, and the overfitting phenomenon is effectively avoided.
After training, the diagnosis result of the model is shown in table 2, wherein the lowest classification accuracy of mechanism jamming is 96.9%, the accuracy of spring falling is 99.2%, the total classification accuracy of the model is 98.1% by calculating an average value, and the CNN fault diagnosis method based on the sound-vibration combined image can accurately reflect the operation state of the energy storage mechanism.
TABLE 2CNN diagnostic results
3.3 model generalization Performance validation
Due to differences in sources and structures of data in an actual circuit breaker monitoring process, fault data of the same type and different representations need to be classified. Therefore, the model ZN63-12 circuit breaker is used instead, the sampling rate is changed from 40kHz to 30kHz, the model ZD-530 piezoelectric sensor and the model WM-025N waterproof sound pickup are used instead, and the mounting position of the sensor is changed simultaneously.
To verify the generalization ability of the optimized model, a control was performed using the original model, and the diagnosis results are shown in fig. 14.
As can be seen from fig. 14, under the condition that the data source, the acquisition parameter setting, and the sensor position are changed, the overall diagnosis accuracy of the optimized CNN model still reaches 97.3%, which is much higher than that of the unoptimized model, and thus, the optimized model has stronger adaptive capacity to the fresh sample and better generalization capability.
Comparing the generalization ability of the model with that of other models, decomposing the original signal by adopting an EEMD algorithm aiming at the same training sample, and diagnosing by adopting LSTM, BP and SVM after extracting the characteristics, wherein the identification result is as follows:
TABLE 3 comparison of results of different diagnostic models
Note: accuracy 1 is for raw data and accuracy 2 is for fluctuating data.
As can be seen from table 3, the diagnosis accuracy of the conventional intelligent diagnosis method is significantly reduced under the condition that the data structure and the source are changed, wherein LSTM is the most serious, the classification accuracy is reduced by 13%, while the CNN model still maintains very high accuracy due to its excellent generalization performance, which is reduced by only 0.6%, the accuracy of the data from different sources in the algorithm is still as high as 97.5%, which is improved by 17.5% compared with other methods, and the generalization capability and the fault identification accuracy of the algorithm are both greatly improved.
The preferred embodiments of the invention disclosed above are intended to be illustrative only. The preferred embodiments are not intended to be exhaustive or to limit the invention to the precise embodiments disclosed. Obviously, many modifications and variations are possible in light of the above teaching. The embodiments were chosen and described in order to best explain the principles of the invention and the practical application, to thereby enable others skilled in the art to best utilize the invention. The invention is limited only by the claims and their full scope and equivalents.
Claims (10)
1. A fault diagnosis method for an energy storage mechanism of a high-voltage circuit breaker is characterized by comprising the following steps:
(1) collecting vibration signals and sound signals of the circuit breaker in the energy storage process;
(2) carrying out morphological denoising on the collected sound signal;
(3) carrying out sound vibration time scale alignment on the sound signal subjected to the morphological denoising and the vibration signal;
(4) performing overlapped data expansion on the data processed in the step (3);
(5) constructing a CNN characteristic matrix of the acoustic vibration signal from the data obtained in the step (4);
(6) optimizing the CNN model;
(7) and inputting the optimized CNN model to identify the fault of the energy storage mechanism.
2. The method for diagnosing the fault of the energy storage mechanism of the high-voltage circuit breaker according to claim 1, wherein the step (1) is specifically as follows: simultaneously collecting vibration signals and sound signals in the energy storage process of the circuit breaker by using a piezoelectric transducer with the frequency range of 1-10kHz and a sound pick-up with the frequency range of 20-20 kHz; the vibration sensor is adsorbed on a connecting fixing piece below a spring vibration body of the high-voltage circuit breaker energy storage mechanism, and the sound sensor is arranged at a sound source of 30-50 cm.
3. The method for diagnosing the fault of the energy storage mechanism of the high-voltage circuit breaker according to claim 1, wherein the step (2) is specifically as follows: performing morphological denoising on the sound signal by adopting an opening-closing and closing-opening operation mixed filter based on morphology, and selecting an ellipse b1And diamond b2A structural element.
4. The method for diagnosing the fault of the energy storage mechanism of the high-voltage circuit breaker according to claim 1, wherein the step (3) is specifically as follows:
(a) performing Hilbert transform on the sound signal and the vibration signal to obtain an envelope, dividing the sound vibration signal into N equal parts, calculating the kurtosis of each part, comparing the kurtosis values of each section of signal, and searching a signal section with obvious kurtosis value change difference so as to determine the time of the change;
(b) after the corresponding time periods of the sound signal and the vibration signal are determined, the Min's distance formula is utilized, the time with the highest similarity is found, the vibration starting time is subtracted from the sound starting time to obtain the delta T, and the sound signal is advanced by the delta T to be aligned with the vibration signal.
5. The method for diagnosing the fault of the energy storage mechanism of the high-voltage circuit breaker according to the claim 1, wherein the step (4) is specifically as follows: for a signal x with the length of N, setting the sample length as L and the overlapping rate as lambda, and adopting the following method:
(A) the maximum number of partitionable samples under the current signal length is calculated:
wherein
The operator is a round-down operator;
(B) solving each segmentation sample; the position of the ith sample in the original signal is represented as follows:
xi＝X[(i-1)×L×(1-λ)+(0:1)×L]，i∈[1,n]
wherein x isiIs the segmented sample data.
6. The method for diagnosing the fault of the energy storage mechanism of the high-voltage circuit breaker according to the claim 1, wherein the step (5) is specifically as follows: the correlation of the sound signal and the vibration signal change is described by using the Pearson product moment correlation coefficient, and the value of the correlation coefficient is used as a matrix element.
7. The method for diagnosing the fault of the energy storage mechanism of the high-voltage circuit breaker according to the claim 1, wherein the step (6) is specifically as follows: performing wavelet decomposition on the kernel function to obtain multi-resolution wavelet coefficients, and selecting the wavelet decomposition coefficients in mutually orthogonal directions to process convolution kernel error modifiers to remove the correlation of the kernel function; at the same time, adding an LRN after each convolution-pooling layer reduces the effect of more parameters needed by the model.
8. The method of claim 7, wherein in the step (6), the normalized LRN layer response is expressed as:
wherein n represents the number of adjacent mapping kernels passing through the same spatial position; n is the total number of cores of the layer;
represents the ReLU nonlinear neuron output for the ith nucleus at (x, y); k, α, β are validation set hyper-parameters.
9. The method for diagnosing the fault of the energy storage mechanism of the high-voltage circuit breaker according to the claim 7, wherein an Early-Stopping mechanism is introduced into the model full connection layer in the training process of the step (6), and the coefficients are 0.5 and 0.01 respectively.
10. The method for diagnosing the fault of the energy storage mechanism of the high-voltage circuit breaker according to claim 7, wherein in the step (6), a plurality of groups of sound signals and vibration signals are collected for training the model under the conditions that the energy storage mechanism of the high-voltage circuit breaker is normal, the spring is rusted, the mechanism is jammed and the spring is fallen.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010900568.4A CN111879397B (en) | 2020-09-01 | 2020-09-01 | Fault diagnosis method for energy storage mechanism of high-voltage circuit breaker |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010900568.4A CN111879397B (en) | 2020-09-01 | 2020-09-01 | Fault diagnosis method for energy storage mechanism of high-voltage circuit breaker |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111879397A true CN111879397A (en) | 2020-11-03 |
| CN111879397B CN111879397B (en) | 2022-05-13 |
Family
ID=73199080
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010900568.4A Active CN111879397B (en) | 2020-09-01 | 2020-09-01 | Fault diagnosis method for energy storage mechanism of high-voltage circuit breaker |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111879397B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112241605A (en) * | 2019-07-17 | 2021-01-19 | 华北电力大学（保定） | Method for identifying state of circuit breaker energy storage process by constructing CNN characteristic matrix through acoustic vibration signals |
| CN112731127A (en) * | 2020-12-08 | 2021-04-30 | 贵州电网有限责任公司 | High-voltage circuit breaker multi-dimensional fusion diagnosis device and method |
| CN113239075A (en) * | 2021-05-13 | 2021-08-10 | 中国公路工程咨询集团有限公司 | Construction data self-checking method and system |
| CN113252323A (en) * | 2021-04-27 | 2021-08-13 | 国网安徽省电力有限公司亳州供电公司 | Circuit breaker mechanical fault identification method and system based on human ear hearing characteristics |
| CN113642484A (en) * | 2021-08-18 | 2021-11-12 | 湖南师范大学 | Magnetotelluric signal noise suppression method and system based on BP neural network |
| CN114323691A (en) * | 2021-12-28 | 2022-04-12 | 中国科学院工程热物理研究所 | Gas circuit fault diagnosis device and method for compressed air energy storage system |
| WO2022264595A1 (en) * | 2021-06-16 | 2022-12-22 | 株式会社日立製作所 | Acoustic circuit breaker diagnosis device and method |
Citations (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH02272326A (en) * | 1989-04-14 | 1990-11-07 | Hitachi Ltd | Diagnostic system for state of apparatus/equipment |
| US20050171774A1 (en) * | 2004-01-30 | 2005-08-04 | Applebaum Ted H. | Features and techniques for speaker authentication |
| CN101360869A (en) * | 2005-11-21 | 2009-02-04 | 弗吉尼亚科技知识产权公司 | Active/passive distributed absorber for vibration and sound radiation control |
| WO2010089911A1 (en) * | 2009-02-03 | 2010-08-12 | 株式会社アクション・リサーチ | Oscillation generation device and method |
| WO2011046474A2 (en) * | 2009-09-24 | 2011-04-21 | Общество С Ограниченной Ответственностью "Цeнтp Речевых Технологий" | Method for identifying a speaker based on random speech phonograms using formant equalization |
| CN104713714A (en) * | 2015-02-09 | 2015-06-17 | 云南电网有限责任公司电力科学研究院 | High-voltage circuit breaker operating characteristic analysis method based on grid multi-density clustering |
| WO2016053748A1 (en) * | 2014-09-29 | 2016-04-07 | Sikorsky Aircraft Corporation | Vibration signatures for prognostics and health monitoring of machinery |
| CN205620511U (en) * | 2016-01-19 | 2016-10-05 | 华北电力大学（保定） | Novel on -line monitoring and control of high voltage circuit breaker trouble device |
| CN106017879A (en) * | 2016-05-18 | 2016-10-12 | 河北工业大学 | Universal circuit breaker mechanical fault diagnosis method based on feature fusion of vibration and sound signals |
| CN106646205A (en) * | 2015-10-30 | 2017-05-10 | 国网山西省电力公司电力科学研究院 | Random big-disturbance signal removing algorithm for analyzing circuit breaker fault through sound and vibration combination |
| CN106960581A (en) * | 2017-04-25 | 2017-07-18 | 中国计量大学 | Speed measurer for motor vehicle based on voice signal |
| CA2965340A1 (en) * | 2016-05-11 | 2017-11-11 | Mahmoud Ismail | An enhanced system and method for conducting pca analysis on data signals |
| CN107766816A (en) * | 2017-10-18 | 2018-03-06 | 河海大学 | A kind of Mechanical Failure of HV Circuit Breaker recognition methods based on LVQ neutral nets |
| CN107796602A (en) * | 2016-08-31 | 2018-03-13 | 华北电力大学（保定） | A kind of circuit breaker failure diagnostic method of sound and vibration signal fused processing |
| CN108509701A (en) * | 2018-03-21 | 2018-09-07 | 南京航空航天大学 | A kind of direct intelligent diagnosing method of rotating machinery fault based on vibration signal |
| KR20180130294A (en) * | 2017-05-29 | 2018-12-07 | 부경대학교 산학협력단 | Method for diagnosing machine fault based on sound |
| CN109142534A (en) * | 2018-10-18 | 2019-01-04 | 吉林化工学院 | Based on rattle signal to the method for pachyrhizus quality evaluation |
| CN110263846A (en) * | 2019-06-18 | 2019-09-20 | 华北电力大学 | The method for diagnosing faults for being excavated and being learnt based on fault data depth |
| CN110534118A (en) * | 2019-07-29 | 2019-12-03 | 安徽继远软件有限公司 | Transformer/reactor method for diagnosing faults based on Application on Voiceprint Recognition and neural network |
| CN110796121A (en) * | 2019-11-21 | 2020-02-14 | 杭州电力设备制造有限公司 | Method for diagnosing mechanical fault of high-voltage circuit breaker by S transformation and optimized random forest |
| CN110926782A (en) * | 2019-12-06 | 2020-03-27 | 国网河南省电力公司三门峡供电公司 | Circuit breaker fault type judgment method and device, electronic equipment and storage medium |
| CN111259947A (en) * | 2020-01-13 | 2020-06-09 | 国网浙江省电力有限公司信息通信分公司 | Power system fault early warning method and system based on multi-mode learning |
| CN111325112A (en) * | 2020-01-31 | 2020-06-23 | 贵州大学 | Cutter wear state monitoring method based on depth gate control circulation unit neural network |
| CN111458141A (en) * | 2020-04-10 | 2020-07-28 | 中国工程物理研究院机械制造工艺研究所 | Vibration source number estimation method based on empirical mode decomposition and singular value decomposition |
| CN111487046A (en) * | 2020-02-27 | 2020-08-04 | 广西电网有限责任公司电力科学研究院 | Fault diagnosis method for circuit breaker voiceprint and vibration entropy feature fusion |
-
2020
- 2020-09-01 CN CN202010900568.4A patent/CN111879397B/en active Active
Patent Citations (25)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH02272326A (en) * | 1989-04-14 | 1990-11-07 | Hitachi Ltd | Diagnostic system for state of apparatus/equipment |
| US20050171774A1 (en) * | 2004-01-30 | 2005-08-04 | Applebaum Ted H. | Features and techniques for speaker authentication |
| CN101360869A (en) * | 2005-11-21 | 2009-02-04 | 弗吉尼亚科技知识产权公司 | Active/passive distributed absorber for vibration and sound radiation control |
| WO2010089911A1 (en) * | 2009-02-03 | 2010-08-12 | 株式会社アクション・リサーチ | Oscillation generation device and method |
| WO2011046474A2 (en) * | 2009-09-24 | 2011-04-21 | Общество С Ограниченной Ответственностью "Цeнтp Речевых Технологий" | Method for identifying a speaker based on random speech phonograms using formant equalization |
| WO2016053748A1 (en) * | 2014-09-29 | 2016-04-07 | Sikorsky Aircraft Corporation | Vibration signatures for prognostics and health monitoring of machinery |
| CN104713714A (en) * | 2015-02-09 | 2015-06-17 | 云南电网有限责任公司电力科学研究院 | High-voltage circuit breaker operating characteristic analysis method based on grid multi-density clustering |
| CN106646205A (en) * | 2015-10-30 | 2017-05-10 | 国网山西省电力公司电力科学研究院 | Random big-disturbance signal removing algorithm for analyzing circuit breaker fault through sound and vibration combination |
| CN205620511U (en) * | 2016-01-19 | 2016-10-05 | 华北电力大学（保定） | Novel on -line monitoring and control of high voltage circuit breaker trouble device |
| CA2965340A1 (en) * | 2016-05-11 | 2017-11-11 | Mahmoud Ismail | An enhanced system and method for conducting pca analysis on data signals |
| CN106017879A (en) * | 2016-05-18 | 2016-10-12 | 河北工业大学 | Universal circuit breaker mechanical fault diagnosis method based on feature fusion of vibration and sound signals |
| CN107796602A (en) * | 2016-08-31 | 2018-03-13 | 华北电力大学（保定） | A kind of circuit breaker failure diagnostic method of sound and vibration signal fused processing |
| CN106960581A (en) * | 2017-04-25 | 2017-07-18 | 中国计量大学 | Speed measurer for motor vehicle based on voice signal |
| KR20180130294A (en) * | 2017-05-29 | 2018-12-07 | 부경대학교 산학협력단 | Method for diagnosing machine fault based on sound |
| CN107766816A (en) * | 2017-10-18 | 2018-03-06 | 河海大学 | A kind of Mechanical Failure of HV Circuit Breaker recognition methods based on LVQ neutral nets |
| CN108509701A (en) * | 2018-03-21 | 2018-09-07 | 南京航空航天大学 | A kind of direct intelligent diagnosing method of rotating machinery fault based on vibration signal |
| CN109142534A (en) * | 2018-10-18 | 2019-01-04 | 吉林化工学院 | Based on rattle signal to the method for pachyrhizus quality evaluation |
| CN110263846A (en) * | 2019-06-18 | 2019-09-20 | 华北电力大学 | The method for diagnosing faults for being excavated and being learnt based on fault data depth |
| CN110534118A (en) * | 2019-07-29 | 2019-12-03 | 安徽继远软件有限公司 | Transformer/reactor method for diagnosing faults based on Application on Voiceprint Recognition and neural network |
| CN110796121A (en) * | 2019-11-21 | 2020-02-14 | 杭州电力设备制造有限公司 | Method for diagnosing mechanical fault of high-voltage circuit breaker by S transformation and optimized random forest |
| CN110926782A (en) * | 2019-12-06 | 2020-03-27 | 国网河南省电力公司三门峡供电公司 | Circuit breaker fault type judgment method and device, electronic equipment and storage medium |
| CN111259947A (en) * | 2020-01-13 | 2020-06-09 | 国网浙江省电力有限公司信息通信分公司 | Power system fault early warning method and system based on multi-mode learning |
| CN111325112A (en) * | 2020-01-31 | 2020-06-23 | 贵州大学 | Cutter wear state monitoring method based on depth gate control circulation unit neural network |
| CN111487046A (en) * | 2020-02-27 | 2020-08-04 | 广西电网有限责任公司电力科学研究院 | Fault diagnosis method for circuit breaker voiceprint and vibration entropy feature fusion |
| CN111458141A (en) * | 2020-04-10 | 2020-07-28 | 中国工程物理研究院机械制造工艺研究所 | Vibration source number estimation method based on empirical mode decomposition and singular value decomposition |
Non-Patent Citations (8)
| Title |
|---|
| ANTONI J 等: "Fast computation of the kurtogram for the detection of transient faults", 《MECHANICAL SYSTEMS AND SIGNAL PROCESSING,》 * |
| RAN WANG 等: "A Non-Contact Fault Diagnosis Method for Rolling Bearings Based on Acoustic Imaging and Convolutional Neural Networks", 《IEEE ACCESS》 * |
| 张重远 等: "基于Mel时频谱-卷积神经网络的变压器铁芯声纹模式识别方法", 《高电压技术》 * |
| 胡永涛: "基于多特征融合及深度信念网络的轴承故障诊断", 《中国优秀硕士/博士学位论文全文数据库》 * |
| 赵书涛 等: "Fault Diagnosis of Oil Pumping Machine Retarder Based on Sound Texture-Vibration Entropy Characteristics and Gray Wolf Optimization-Support Vector Machine", 《SHOCK AND VIBRATION》 * |
| 赵书涛 等: "声振信号联合1D-CNN的大型电机故障诊断方法", 《哈尔滨工业大学学报》 * |
| 赵书涛 等: "断路器控制线圈电流特征诊断故障的技术分析及展望", 《华北电力大学学报》 * |
| 陶鹏 等: "基于Gumbel copula联合概率分布的电力系统综合净负荷预测", 《华北电力大学学报》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112241605A (en) * | 2019-07-17 | 2021-01-19 | 华北电力大学（保定） | Method for identifying state of circuit breaker energy storage process by constructing CNN characteristic matrix through acoustic vibration signals |
| CN112241605B (en) * | 2019-07-17 | 2023-12-19 | 华北电力大学（保定） | Breaker energy storage process state identification method for constructing CNN feature matrix by using acoustic vibration signals |
| CN112731127A (en) * | 2020-12-08 | 2021-04-30 | 贵州电网有限责任公司 | High-voltage circuit breaker multi-dimensional fusion diagnosis device and method |
| CN113252323A (en) * | 2021-04-27 | 2021-08-13 | 国网安徽省电力有限公司亳州供电公司 | Circuit breaker mechanical fault identification method and system based on human ear hearing characteristics |
| CN113239075A (en) * | 2021-05-13 | 2021-08-10 | 中国公路工程咨询集团有限公司 | Construction data self-checking method and system |
| WO2022264595A1 (en) * | 2021-06-16 | 2022-12-22 | 株式会社日立製作所 | Acoustic circuit breaker diagnosis device and method |
| CN113642484A (en) * | 2021-08-18 | 2021-11-12 | 湖南师范大学 | Magnetotelluric signal noise suppression method and system based on BP neural network |
| CN113642484B (en) * | 2021-08-18 | 2022-06-28 | 湖南师范大学 | Magnetotelluric signal noise suppression method and system based on BP neural network |
| CN114323691A (en) * | 2021-12-28 | 2022-04-12 | 中国科学院工程热物理研究所 | Gas circuit fault diagnosis device and method for compressed air energy storage system |
| CN114323691B (en) * | 2021-12-28 | 2023-06-23 | 中国科学院工程热物理研究所 | Air path fault diagnosis device and method for compressed air energy storage system |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111879397B (en) | 2022-05-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111879397B (en) | Fault diagnosis method for energy storage mechanism of high-voltage circuit breaker | |
| CN109827777B (en) | Rolling bearing fault prediction method based on partial least square method extreme learning machine | |
| Hu et al. | A novel fault diagnosis technique for wind turbine gearbox | |
| CN110929847A (en) | Converter transformer fault diagnosis method based on deep convolutional neural network | |
| CN111504635B (en) | Planetary gear fault diagnosis method based on differential evolution probability neural network | |
| CN104569747A (en) | System and method for checking insulativity of power-off cable | |
| CN114169374B (en) | Cable-stayed bridge stay cable damage identification method and electronic equipment | |
| CN113252322B (en) | Method and device for diagnosing mechanical fault of circuit breaker under noise-containing background | |
| CN109918417B (en) | Time sequence data self-adaptive segmentation, dimension reduction and characterization method based on wavelet transformation and application | |
| CN112766140A (en) | Transformer fault identification method based on kernel function extreme learning machine | |
| CN111950526A (en) | Fault diagnosis method for energy storage mechanism of circuit breaker based on deep learning | |
| Chen et al. | Fault identification method for planetary gear based on DT-CWT threshold denoising and LE | |
| Li et al. | Gear pitting fault diagnosis using raw acoustic emission signal based on deep learning | |
| Zhao et al. | Fault diagnosis method for energy storage mechanism of high voltage circuit breaker based on CNN characteristic matrix constructed by sound-vibration signal | |
| CN110348468A (en) | A kind of bearing inferior health recognition methods of the strong reconstruct edge noise reduction autocoder of Method Using Relevance Vector Machine optimization | |
| CN113203914A (en) | Underground cable early fault detection and identification method based on DAE-CNN | |
| Shi et al. | Intelligent fault diagnosis of bearings based on feature model and Alexnet neural network | |
| CN116776075A (en) | Fan blade health monitoring system and monitoring method thereof | |
| CN112636328B (en) | Medium-voltage distribution network equipment load state identification method | |
| CN116699400A (en) | Generator rotor short-circuit fault monitoring system, method and readable storage medium | |
| Li et al. | A robust fault diagnosis method for rolling bearings based on deep convolutional neural network | |
| CN115563480A (en) | Gear fault identification method for screening octave geometric modal decomposition based on kurtosis ratio coefficient | |
| Wu et al. | A fault diagnosis system for a mechanical reducer gear-set using Wigner-Ville distribution and an artificial neural network | |
| CN115410599A (en) | Method and system for identifying running state of transformer based on voiceprint perception | |
| CN115204272A (en) | Industrial system fault diagnosis method and equipment based on multi-sampling rate data |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |