CN110796121A - Method for diagnosing mechanical fault of high-voltage circuit breaker by S transformation and optimized random forest - Google Patents
Method for diagnosing mechanical fault of high-voltage circuit breaker by S transformation and optimized random forest Download PDFInfo
- Publication number
- CN110796121A CN110796121A CN201911146141.3A CN201911146141A CN110796121A CN 110796121 A CN110796121 A CN 110796121A CN 201911146141 A CN201911146141 A CN 201911146141A CN 110796121 A CN110796121 A CN 110796121A
- Authority
- CN
- China
- Prior art keywords
- circuit breaker
- voltage circuit
- transformation
- random forest
- time
- 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.)
- Pending
Links
- 238000007637 random forest analysis Methods 0.000 title claims abstract description 65
- 238000000034 method Methods 0.000 title claims abstract description 46
- 230000009466 transformation Effects 0.000 title claims abstract description 44
- 239000011159 matrix material Substances 0.000 claims abstract description 38
- 238000000354 decomposition reaction Methods 0.000 claims abstract description 30
- 238000003745 diagnosis Methods 0.000 claims abstract description 25
- 239000013598 vector Substances 0.000 claims abstract description 21
- 238000005457 optimization Methods 0.000 claims abstract description 12
- 238000005070 sampling Methods 0.000 claims description 25
- 238000012549 training Methods 0.000 claims description 19
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical group [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 claims description 8
- 238000003066 decision tree Methods 0.000 claims description 8
- 238000005461 lubrication Methods 0.000 claims description 8
- 238000012545 processing Methods 0.000 claims description 7
- 230000009471 action Effects 0.000 claims description 3
- 230000008569 process Effects 0.000 abstract description 6
- 230000009286 beneficial effect Effects 0.000 abstract description 3
- 239000000284 extract Substances 0.000 abstract description 2
- 238000010586 diagram Methods 0.000 description 17
- 230000000694 effects Effects 0.000 description 7
- 238000004458 analytical method Methods 0.000 description 4
- 238000006243 chemical reaction Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000010276 construction Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000007812 deficiency Effects 0.000 description 2
- 230000036039 immunity Effects 0.000 description 2
- 238000003909 pattern recognition Methods 0.000 description 2
- 238000012706 support-vector machine Methods 0.000 description 2
- 238000013519 translation 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
- 238000012952 Resampling Methods 0.000 description 1
- 230000001133 acceleration Effects 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000005315 distribution function Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 238000013138 pruning Methods 0.000 description 1
- 238000000926 separation method Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/08—Feature extraction
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01R—MEASURING ELECTRIC VARIABLES; MEASURING MAGNETIC VARIABLES
- G01R31/00—Arrangements for testing electric properties; Arrangements for locating electric faults; Arrangements for electrical testing characterised by what is being tested not provided for elsewhere
- G01R31/327—Testing of circuit interrupters, switches or circuit-breakers
- G01R31/3271—Testing of circuit interrupters, switches or circuit-breakers of high voltage or medium voltage devices
- G01R31/3275—Fault detection or status indication
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/004—Artificial life, i.e. computing arrangements simulating life
- G06N3/006—Artificial life, i.e. computing arrangements simulating life based on simulated virtual individual or collective life forms, e.g. social simulations or particle swarm optimisation [PSO]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
- G06F2218/12—Classification; Matching
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Biomedical Technology (AREA)
- General Health & Medical Sciences (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Molecular Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Biology (AREA)
- Signal Processing (AREA)
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The method for diagnosing the mechanical fault of the high-voltage circuit breaker by S transformation and optimization of random forests belongs to the technical field of judgment of the mechanical fault of the high-voltage circuit breaker. In the invention, the vibration signal is firstly transformed by S to obtain a time-frequency matrix of the signal, and then the time-frequency domain characteristic vector of the signal is extracted by utilizing the local singular value decomposition. The S transformation has strong characteristic expression capability, and can comprehensively and visually display the time-frequency domain characteristics of the vibration signals of the circuit breaker; the S transformation has strong anti-noise capability, and noise interference can be obviously reduced in the S transformation process; the local singular value decomposition method decomposes the time-frequency matrix after S transformation, extracts the maximum singular value of each part after decomposition as a characteristic vector, can highlight the characteristics of signals and is more beneficial to fault diagnosis. The random forest model finds the number of the optimal trees through the comprehensive influence of the number of the trees on the generalization error and the diagnosis accuracy, and further improves the accuracy of HVCBs state identification.
Description
Technical Field
The invention belongs to the technical field of judgment of mechanical faults of high-voltage circuit breakers, and particularly relates to a method for diagnosing the mechanical faults of the high-voltage circuit breakers by S transformation and optimization of random forests.
Background
High Voltage Circuit Breakers (HVCBs) are applied to power systems in many ways and have a complex structure, and are susceptible to external factors. Most of the high-voltage circuit breaker faults belong to mechanical faults, and the analysis of vibration signals generated by HVCBs actions can find out various circuit breaker faults caused by poor mechanical properties. The circuit breaker vibration signal has the characteristics of high complexity, easy noise interference, non-stability and non-linearity, and a time-frequency analysis method is mostly adopted for signal processing. Common methods include wavelet transformation, empirical mode decomposition, ensemble empirical mode decomposition, local mean decomposition, and the like. The wavelet transform has better local feature representation capability, but the difficulty of selecting the wavelet base is higher in practical application. The separation capability of empirical mode decomposition and local mean decomposition on components with similar frequencies is poor, and a mode aliasing phenomenon exists; ensemble empirical mode decomposition adds white noise to suppress modal aliasing, but also increases the amount of computation and decomposes out multiple components that exceed the true signal composition. The S transformation characteristic expression capability and the anti-noise capability are strong, the time-frequency characteristic of the vibration signal can be visually displayed, and the S transformation characteristic expression method is suitable for analyzing and processing the vibration signal with the complex time-frequency characteristic.
Due to the existing fault types and the low fault degrees of the mechanical faults of the high-voltage circuit breakers, training samples containing all fault types are difficult to obtain. Currently, common methods for HVCBs state recognition include Support Vector Machines (SVMs), Neural Networks (NNs), and the like. The SVM is not easy to fall into local optimal solution and over-learning, is suitable for the problem of small sample classification, but has low training speed and difficult parameter selection. The NN self-learning capability and the nonlinear pattern recognition capability are strong, but the training speed is slow and the NN self-learning capability and the nonlinear pattern recognition capability are easy to fall into local optimal solutions.
The method has the advantages that parameters needing to be optimized in Random Forest (RF) are few, noise immunity is good, accuracy in processing small sample data is high, overfitting is not prone to occurring, and the method is suitable for fault diagnosis of the high-voltage circuit breaker. In the prior patent CN201810457787, a circuit breaker mechanical fault diagnosis method and system based on random forest adopts a traditional random forest classifier, and does not perform optimized selection on parameters of a random forest model, so that the classification effect is poor. In addition, only the time domain feature vector of the circuit breaker vibration signal data is extracted in the patent, so that comprehensive feature information cannot be obtained, the effect of feature extraction is poor, and the accuracy of final fault diagnosis is low. Meanwhile, only 7 time domain features are extracted, and the quantity is too small, so that the classifier identification is not facilitated.
Therefore, there is a need in the art for a new solution to solve this problem.
Disclosure of Invention
The technical problem to be solved by the invention is as follows: the method for diagnosing the mechanical fault of the high-voltage circuit breaker by using the S transformation and the optimized random forest is used for solving the technical problems that a circuit breaker mechanical fault diagnosis method using a traditional random forest classifier is poor in classification effect and low in fault diagnosis accuracy.
The method for diagnosing the mechanical fault of the high-voltage circuit breaker by the S transformation and the optimized random forest comprises the following steps which are sequentially carried out,
the method comprises the following steps of firstly, carrying out signal acquisition on the vibration of the high-voltage circuit breaker in four mechanical states of a normal state, an iron core jamming state, a screw loosening state and a crank arm insufficient lubrication state, wherein the zero point of a coordinate of a sampling signal is the initial moment of the action of the circuit breaker, the signal sampling rate is the same, and the recording starting point of the vibration signal is the same as the time length of the signal acquisition;
step two, respectively carrying out S transformation on the vibration signals in the four mechanical states to obtain a time-frequency characteristic matrix of the signals;
step three, local singular value decomposition
Selecting sampling points in the middle section of the total sampling time, performing Singular Value Decomposition (SVD) processing on a time-frequency characteristic matrix between a sampling starting point and the selected sampling points, dividing the time-frequency characteristic matrix into a plurality of sub-matrices along a time domain and a frequency domain at equal intervals, calculating and obtaining each maximum singular value respectively, constructing a characteristic vector, removing a numerical value smaller than a set threshold value in the characteristic vector, and obtaining a high-voltage circuit breaker vibration signal characteristic value without noise;
step four, a random forest classifier model is built in Matlab, 20 groups of vibration signals are taken as training samples of the classifier from the four mechanical state vibration signals obtained in the step one, S transformation and local singular value decomposition are carried out on the training samples, the characteristic values of the vibration signals of the high-voltage circuit breakers without noise are obtained respectively, the obtained characteristic value vectors of the vibration signals of the high-voltage circuit breakers without noise are input into the random forest classifier model to train and optimize classifier parameters, and the trained random forest fault state classifier is obtained;
and fifthly, inputting the acquired vibration signals of the high-voltage circuit breaker into a random forest fault state classifier trained in the fourth step through S transformation and local singular value decomposition, and inputting the characteristic value vector of the vibration signals of the high-voltage circuit breaker without noise to obtain the state of the high-voltage circuit breaker reflected by the signals.
And setting the sampling rate of the sampling signals in the step one to be 25.6 kS/s.
The parameters of the random forest classifier model optimization in the fourth step comprise the number of trees, the maximum depth of the trees and the maximum characteristic number of a single decision tree.
Through the design scheme, the invention can bring the following beneficial effects:
in the invention, the vibration signal is firstly transformed by S to obtain a time-frequency matrix of the signal, and then the time-frequency domain characteristic vector of the signal is extracted by utilizing the local singular value decomposition. The S transformation has strong characteristic expression capability, and can comprehensively and visually display the time-frequency domain characteristics of the vibration signals of the circuit breaker; the S transformation has strong anti-noise capability, and noise interference can be obviously reduced in the S transformation process; the local singular value decomposition method decomposes the time-frequency matrix after S transformation, extracts the maximum singular value of each part after decomposition as a characteristic vector, can highlight the characteristics of signals and is more beneficial to fault diagnosis.
The random forest comprises a plurality of parameters, wherein the influence of the number of trees on the diagnosis accuracy is the largest, the more the number of trees is, the more the classification is accurate, but the occupied memory and the training time are correspondingly increased, so that the difficulty in selecting the number of the most optimal trees is higher. The generalization error can measure the classification accuracy, the calculation speed, the generalization capability and the like of the random forest model, and the generalization error is inversely proportional to the classification effect of the random forest. Therefore, the optimal tree number is found through the comprehensive influence of the tree number on the generalization error and the diagnosis accuracy. The method can ensure the fault diagnosis accuracy of the random forest model and simultaneously reduce the memory occupation and training time to the maximum extent. Optimizing the number of trees by taking the generalized error and the diagnosis accuracy as comprehensive indexes to form an Optimized Random Forest (ORF), and further improving the accuracy of HVCBs state identification.
Drawings
The invention is further described with reference to the following figures and detailed description:
fig. 1 is a diagram of a system-acquired normal-state measured vibration signal in the method for diagnosing mechanical faults of a high-voltage circuit breaker by using the S transformation and the optimized random forest.
Fig. 2 is a diagram of a vibration signal actually measured in the state of jamming of an iron core acquired by a system in the method for diagnosing mechanical faults of the high-voltage circuit breaker by S transformation and optimization of random forests.
Fig. 3 is a diagram of measured vibration signals of a screw loosening state acquired by a system in the method for diagnosing mechanical faults of the high-voltage circuit breaker by S transformation and optimization of random forests.
FIG. 4 is a diagram of measured vibration signals of a crank arm under lubrication condition acquired by a system in the method for diagnosing mechanical faults of the high-voltage circuit breaker by S transformation and optimized random forests.
Fig. 5 is a time-frequency matrix diagram after S transformation of a normal state vibration signal in the method for S transformation and optimized random forest diagnosis of mechanical faults of a high-voltage circuit breaker according to the present invention.
Fig. 6 is a time-frequency matrix diagram after the vibration signal of the iron core jam state is subjected to S transformation in the method for diagnosing the mechanical fault of the high-voltage circuit breaker by using the S transformation and the optimized random forest.
Fig. 7 is a time-frequency matrix diagram after the vibration signal of the screw loosening state is subjected to S transformation in the method for diagnosing the mechanical fault of the high-voltage circuit breaker by using the S transformation and the optimized random forest.
Fig. 8 is a time-frequency matrix diagram after the vibration signal of the crank arm lubrication deficiency state is subjected to S conversion in the method for diagnosing the mechanical fault of the high-voltage circuit breaker by using the S conversion and the optimized random forest.
Fig. 9 is a characteristic vector diagram after local singular value decomposition is performed on a time-frequency matrix diagram of a vibration signal in a normal state in the method for diagnosing the mechanical fault of the high-voltage circuit breaker by using the S transformation and the optimized random forest.
Fig. 10 is a characteristic vector diagram after local singular value decomposition is performed on a time-frequency matrix diagram of an iron core jamming state vibration signal in the method for diagnosing the mechanical fault of the high-voltage circuit breaker by using the S transformation and the optimized random forest.
Fig. 11 is a characteristic vector diagram after local singular value decomposition is performed on a time-frequency matrix diagram of a vibration signal of a screw loosening state in the method for diagnosing the mechanical fault of the high-voltage circuit breaker by using the S transformation and the optimized random forest.
FIG. 12 is a characteristic vector diagram after local singular value decomposition of a time-frequency matrix diagram of a crank arm lubrication deficiency vibration signal in the method for S-transform and optimized random forest diagnosis of mechanical faults of a high-voltage circuit breaker according to the invention.
Fig. 13 is a tree number optimizing diagram in the method for diagnosing mechanical faults of a high-voltage circuit breaker by using the S transformation and the optimized random forest.
Fig. 14 is a fault diagnosis flowchart of the method for diagnosing mechanical faults of a high-voltage circuit breaker by using the S transformation and the optimized random forest according to the invention.
Detailed Description
The method for diagnosing the mechanical fault of the high-voltage circuit breaker by S transformation and optimization of the random forest comprises the following steps:
1. high voltage circuit breaker vibration signal acquisition
The analysis object is a high-voltage circuit breaker, the piezoelectric acceleration sensor and the NI9234 data acquisition card are used for acquiring and recording vibration signal data, and the vibration signal data form a signal acquisition system together. The zero point of the coordinate of the sampling signal is the time when the circuit breaker is about to act (the trigger card sends a signal acquisition instruction to the signal acquisition card), the sampling rate of the signal is set to be 25.6kS/s, and the recording starting point of the vibration signal in 4 mechanical states is the same as the time length of the signal acquisition. The signal data acquired by the system is divided into four states as shown in fig. 1-4: a normal state; jamming the iron core; loosening the screw; poor lubrication.
2. Vibration signal S transformation
The S transformation is a time-frequency analysis method, a Gaussian window with the window width inversely proportional to the frequency is added to a signal, the complexity of window function selection and the limitation of window width fixation are effectively improved, and the time-frequency characteristic of the signal can be directly reflected. The original signals of fig. 1 to 4 need to be subjected to the following S conversion process.
For a one-dimensional continuous-time signal h (t), its continuous S-transform S (τ, f) is defined as
In formulas (1) and (2): w (τ -t, f) is a Gaussian window function; tau is the translation factor parameter controlling the position of the Gaussian window, t is time, f is frequency, and i is an imaginary unit.
Given the Fourier transform of S (τ, f) to R (β, f), there are
In the formula (3), β is a translation frequency, H (f) is a Fourier transform of h (t), and H (f) is a Fourier transform of h (t)
In the formula (4), f is not equal to 0
Let f → nLT, τ → jT (L is data length, T is sampling interval time, n is number of sampling points, j is number of sampling cycles), obtain discrete form of S transform
In equation (5), m represents the mth sampling period.
The time-frequency modulus value matrix obtained by performing S conversion on the vibration signals in the four HVCBs states is shown in fig. 5 to 8. Compared with a normal signal, the energy fluctuation of the iron core jam fault occurs later, and the energy distribution on a frequency domain is basically consistent; the energy fluctuation of the loosening fault of the base screw is small along with the time change, and the energy is more concentrated on a low-frequency part; the vibration time of the crank arm lubrication fault is longer, and the energy of the low-frequency part is higher.
3. Local singular value decomposition
The amplitude of the vibration signal in the second half of the total sampling time is small and has no obvious difference, so that in order to more accurately obtain effective information, only the matrix before a certain sampling point is subjected to SVD processing. For the whole time-frequency matrix, the time domain and the frequency domain are respectively divided into a plurality of sub-matrixes at equal intervals, the maximum singular value is respectively calculated, and the eigenvector is constructed.
As can be seen from the time-frequency matrix after S transformation in fig. 5, when the time is greater than 2000ms, the amplitudes of the vibration signals in the four states are small and have no obvious difference. Therefore, in order to obtain effective information more accurately, the SVD processing is only performed on the matrix before 2000 ms. For the whole time-frequency matrix, 20 sub-matrices are equally divided along the time domain, 10 sub-matrices are equally divided along the frequency domain, the maximum singular values are respectively calculated, and the constructed eigenvectors are shown in fig. 9-12. The singular value decomposition method of the matrix is as follows:
assuming that the rank of matrix X is r and the dimension of the matrix is mxn, the singular value decomposition of matrix X is in the form:
in formula (6), the matrices U and VHRespectively, m × m, n × n orthogonal matrices, sigma is r × r diagonal matrix, diagonal data of matrix sigma is singular value of matrix X and is not equal to 0,
if the zero singular value of X is removed in the formula (1), a simplified singular value matrix X can be obtained:
in formula (7): u. ofiIs the ith vector of U, viIs a number Vi vectors, δiIs the ith singular value of the matrix X.
Assuming that the collected signals are y (i) (1, 2, …, N), reconstructing the signals by using a matrix, wherein the matrix A after reconstruction is as follows:
in the formula (8), M is the number of rows of the matrix a, and N is the total number of signals.
Singular value decomposition is carried out on the matrix 2 by using the formula (1), and a group of non-zero singular values delta is obtained after decompositioni. Depending on the transmissibility of the information, if y (i) contains noise, the matrix A also contains noise, all the information being at the singular value δiIn (1). At the set of non-zero singular values deltaiIn the method, the fact that the value of P is larger represents a useful signal, the fact that P is smaller represents noise, and if the value of P is smaller, the fact that the noise signal is removed is equivalent to the fact that the noise signal is removed, and therefore the characteristic value of the vibration signal of the high-voltage circuit breaker without noise is obtained.
4. Construction and parameter optimization of random forest model
And respectively taking 20 groups of vibration signals of the high-voltage circuit breaker in a normal state, an iron core jamming state, an insufficient lubrication state and a screw loosening state as training samples of the classifier, carrying out S transformation and local singular value decomposition on the training samples, inputting the obtained feature vectors into a random forest classifier model for training, and obtaining the random forest fault state classifier which is finally trained. In order to improve the classification accuracy, a random forest model needs to be constructed first, and then parameters in the random forest model need to be optimized.
1) Random forest construction
The random forest is a novel strong classifier based on decision trees and integrated learning, integrates two ideas of self-service sampling and random subspaces, and is good in noise immunity, few in parameters needing optimization and small in influence of overfitting. The construction process of the random forest can be divided into the following three steps:
step 1: and for an original feature set N with m samples, samples in the N are extracted in a returning mode by utilizing a self-help resampling technology to form a new sample set serving as a training set, the samples are extracted for k times, and the number of the samples in each training set is the same as that of the samples in the N and is m.
Step 2: and selecting a training set to establish a decision tree model, and allowing the decision tree to grow without pruning. And stopping growing when the number of samples in the training set is 1 and the splitting cannot be continued or all the samples point to the same label when the entropy of the training set is 0.
Step 3: and performing the steps once on each training set to generate k sub decision trees which jointly form a random forest.
2) Random forest classification
After test set data is input into a random forest, k results are generated together, the final result is determined by voting, and the voting formula is
In the formula (9), x is data to be measured; y is a target classification; r (x) is the result of the classification; r isi(x) An ith decision tree model; arg is an average value; i is an indicative function.
3) Random forest parameter optimization
The random forest relates to adjustable parameters including the number of trees, the maximum depth of trees, the maximum feature number of a single decision tree, and the like. The parameters have different degrees of influence on the random forest performance, wherein the influence of the number of trees on the diagnosis accuracy is large, the more the number of trees is, the more the classification is accurate, but the occupied memory and the training time are correspondingly increased, so that the difficulty in selecting the number of the optimal trees is large.
The generalization error can measure the classification accuracy, the calculation speed, the generalization capability and the like of the random forest model, and the generalization error is inversely proportional to the classification effect of the random forest. Therefore, the optimal tree number can be found through the comprehensive influence of the tree number on the generalization error and the diagnosis accuracy. The generalized error is calculated as follows:
PE*=Px,y(K(x,y)<0)
in the formula (10), PE*To generalize errors; px,yProbability distribution function for classifying the data to be measured and the target; k (x, y) is a margin function; thetakRandom vectors distributed independently from the k-th number; r (x, theta)k) A model of the kth tree; j is a certain classification of the data x to be measured.
In order to examine the optimizing effect of the generalized error and diagnosis accuracy comprehensive index on the number of trees, random forest RF was trained, and the result is shown in fig. 13. In the process of changing the number of trees from 0 to 160, the generalization error of ORF is firstly reduced and then increased, and the fault diagnosis accuracy is firstly rapidly increased and then slowly increased. The change conditions of the generalization error and the diagnosis accuracy are comprehensively considered, when the number of the trees is 80, the generalization error reaches the lowest value of 0.203, the diagnosis accuracy is up to 96.27%, and the improvement of the diagnosis accuracy is not obvious when the number of the trees is more than 80, so that the comprehensive classification effect of the RF is optimal when the number of the trees is 80. The entire fault diagnosis process is shown in fig. 14.
Claims (3)
- The method for diagnosing the mechanical fault of the high-voltage circuit breaker by S transformation and optimized random forest is characterized by comprising the following steps: comprises the following steps which are sequentially carried out,the method comprises the following steps of firstly, carrying out signal acquisition on the vibration of the high-voltage circuit breaker in four mechanical states of a normal state, an iron core jamming state, a screw loosening state and a crank arm insufficient lubrication state, wherein the zero point of a coordinate of a sampling signal is the initial moment of the action of the circuit breaker, the signal sampling rate is the same, and the recording starting point of the vibration signal is the same as the time length of the signal acquisition;step two, respectively carrying out S transformation on the vibration signals in the four mechanical states to obtain a time-frequency characteristic matrix of the signals;step three, local singular value decompositionSelecting sampling points in the middle section of the total sampling time, performing Singular Value Decomposition (SVD) processing on a time-frequency characteristic matrix between a sampling starting point and the selected sampling points, dividing the time-frequency characteristic matrix into a plurality of sub-matrices along a time domain and a frequency domain at equal intervals, calculating and obtaining each maximum singular value respectively, constructing a characteristic vector, removing a numerical value smaller than a set threshold value in the characteristic vector, and obtaining a high-voltage circuit breaker vibration signal characteristic value without noise;step four, a random forest classifier model is built in Matlab, 20 groups of vibration signals are taken as training samples of the classifier from the four mechanical state vibration signals obtained in the step one, S transformation and local singular value decomposition are carried out on the training samples, the characteristic values of the vibration signals of the high-voltage circuit breakers without noise are obtained respectively, the obtained characteristic value vectors of the vibration signals of the high-voltage circuit breakers without noise are input into the random forest classifier model to train and optimize classifier parameters, and the trained random forest fault state classifier is obtained;and fifthly, inputting the acquired vibration signals of the high-voltage circuit breaker into a random forest fault state classifier trained in the fourth step through S transformation and local singular value decomposition, and inputting the characteristic value vector of the vibration signals of the high-voltage circuit breaker without noise to obtain the state of the high-voltage circuit breaker reflected by the signals.
- 2. The S-transform-based high-voltage circuit breaker mechanical fault optimization random forest diagnosis method as claimed in claim 1, wherein the method comprises the following steps: and setting the sampling rate of the sampling signals in the step one to be 25.6 kS/s.
- 3. The S-transform-based high-voltage circuit breaker mechanical fault optimization random forest diagnosis method as claimed in claim 1, wherein the method comprises the following steps: the parameters of the random forest classifier model optimization in the fourth step comprise the number of trees, the maximum depth of the trees and the maximum characteristic number of a single decision tree.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911146141.3A CN110796121A (en) | 2019-11-21 | 2019-11-21 | Method for diagnosing mechanical fault of high-voltage circuit breaker by S transformation and optimized random forest |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911146141.3A CN110796121A (en) | 2019-11-21 | 2019-11-21 | Method for diagnosing mechanical fault of high-voltage circuit breaker by S transformation and optimized random forest |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110796121A true CN110796121A (en) | 2020-02-14 |
Family
ID=69445740
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911146141.3A Pending CN110796121A (en) | 2019-11-21 | 2019-11-21 | Method for diagnosing mechanical fault of high-voltage circuit breaker by S transformation and optimized random forest |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110796121A (en) |
Cited By (8)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111458145A (en) * | 2020-03-30 | 2020-07-28 | 南京机电职业技术学院 | Cable car rolling bearing fault diagnosis method based on road map characteristics |
CN111638047A (en) * | 2020-05-25 | 2020-09-08 | 国网河北省电力有限公司电力科学研究院 | GIS equipment mechanical fault diagnosis system and method |
CN111879397A (en) * | 2020-09-01 | 2020-11-03 | 国网河北省电力有限公司检修分公司 | Fault diagnosis method for energy storage mechanism of high-voltage circuit breaker |
CN112083328A (en) * | 2020-09-08 | 2020-12-15 | 广东电网有限责任公司广州供电局 | Fault diagnosis method, system and device for high-voltage circuit breaker |
CN112083327A (en) * | 2020-09-08 | 2020-12-15 | 广东电网有限责任公司广州供电局 | Mechanical fault diagnosis method and system for high-voltage vacuum circuit breaker |
CN112327149A (en) * | 2020-10-19 | 2021-02-05 | 西安工程大学 | SSD-SRAE-based high-voltage circuit breaker fault diagnosis method |
CN113567131A (en) * | 2021-07-29 | 2021-10-29 | 沈阳建筑大学 | Bearing fault diagnosis method based on S transformation and miniature convolution neural network model |
CN115342900A (en) * | 2022-08-15 | 2022-11-15 | 东北石油大学 | Laser self-mixing interference micro-vibration measurement method and system based on random forest |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108062571A (en) * | 2017-12-27 | 2018-05-22 | 福州大学 | Diagnosing failure of photovoltaic array method based on differential evolution random forest grader |
CN108680348A (en) * | 2018-05-14 | 2018-10-19 | 国网山东省电力公司莱芜供电公司 | A kind of breaker mechanical fault diagnosis method and system based on random forest |
CN109597401A (en) * | 2018-12-06 | 2019-04-09 | 华中科技大学 | A kind of equipment fault diagnosis method based on data-driven |
CN109886464A (en) * | 2019-01-20 | 2019-06-14 | 东北电力大学 | The low information loss short-term wind speed forecasting method of feature set is generated based on optimization singular value decomposition |
-
2019
- 2019-11-21 CN CN201911146141.3A patent/CN110796121A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108062571A (en) * | 2017-12-27 | 2018-05-22 | 福州大学 | Diagnosing failure of photovoltaic array method based on differential evolution random forest grader |
CN108680348A (en) * | 2018-05-14 | 2018-10-19 | 国网山东省电力公司莱芜供电公司 | A kind of breaker mechanical fault diagnosis method and system based on random forest |
CN109597401A (en) * | 2018-12-06 | 2019-04-09 | 华中科技大学 | A kind of equipment fault diagnosis method based on data-driven |
CN109886464A (en) * | 2019-01-20 | 2019-06-14 | 东北电力大学 | The low information loss short-term wind speed forecasting method of feature set is generated based on optimization singular value decomposition |
Non-Patent Citations (2)
Title |
---|
方匡南: "《随机森林组合预测理论及其在金融中的应用》", 31 May 2012, 厦门大学出版社 * |
黄南天等: "基于S变换和极限学习机的高压断路器机械故障诊断", 《高压电器》 * |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111458145A (en) * | 2020-03-30 | 2020-07-28 | 南京机电职业技术学院 | Cable car rolling bearing fault diagnosis method based on road map characteristics |
CN111638047A (en) * | 2020-05-25 | 2020-09-08 | 国网河北省电力有限公司电力科学研究院 | GIS equipment mechanical fault diagnosis system and method |
CN111879397A (en) * | 2020-09-01 | 2020-11-03 | 国网河北省电力有限公司检修分公司 | Fault diagnosis method for energy storage mechanism of high-voltage circuit breaker |
CN112083328A (en) * | 2020-09-08 | 2020-12-15 | 广东电网有限责任公司广州供电局 | Fault diagnosis method, system and device for high-voltage circuit breaker |
CN112083327A (en) * | 2020-09-08 | 2020-12-15 | 广东电网有限责任公司广州供电局 | Mechanical fault diagnosis method and system for high-voltage vacuum circuit breaker |
CN112327149A (en) * | 2020-10-19 | 2021-02-05 | 西安工程大学 | SSD-SRAE-based high-voltage circuit breaker fault diagnosis method |
CN113567131A (en) * | 2021-07-29 | 2021-10-29 | 沈阳建筑大学 | Bearing fault diagnosis method based on S transformation and miniature convolution neural network model |
CN113567131B (en) * | 2021-07-29 | 2024-05-03 | 沈阳建筑大学 | Bearing fault diagnosis method based on S transformation and miniature convolutional neural network model |
CN115342900A (en) * | 2022-08-15 | 2022-11-15 | 东北石油大学 | Laser self-mixing interference micro-vibration measurement method and system based on random forest |
CN115342900B (en) * | 2022-08-15 | 2024-04-30 | 东北石油大学 | Random forest-based laser self-mixing interference micro-vibration measurement method and system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110796121A (en) | Method for diagnosing mechanical fault of high-voltage circuit breaker by S transformation and optimized random forest | |
CN109612708B (en) | Power transformer on-line detection system and method based on improved convolutional neural network | |
CN112084974B (en) | Multi-label rolling bearing fault diagnosis method based on meta-learning | |
CN110657985B (en) | Gearbox fault diagnosis method and system based on singular value spectrum manifold analysis | |
CN110261109B (en) | Rolling bearing fault diagnosis method based on bidirectional memory cyclic neural network | |
CN106597260A (en) | Simulation circuit fault diagnosis method based on continuous wavelet analysis and ELM network | |
CN111738309B (en) | Multi-scale analysis and integrated learning gas sensor fault mode identification method | |
CN111412977A (en) | Preprocessing method for vibration sensing data of mechanical equipment | |
CN110796120A (en) | Time domain feature-based circuit breaker mechanical fault XGboost diagnosis method | |
CN106444666B (en) | Dynamic process monitoring method based on the dynamic distributed pca model of weighting type | |
CN113408458A (en) | Circuit breaker fault diagnosis method based on vibration signal characteristic entropy fusion | |
CN106771589B (en) | A kind of implementation method of the real-time Waterfall plot of spectrum analysis | |
CN109948726B (en) | Power quality disturbance classification method based on deep forest | |
CN110515096A (en) | Satellite navigation interference signal identification device and its method based on convolutional neural networks | |
CN111680725A (en) | Gas sensor array multi-fault isolation algorithm based on reconstruction contribution | |
CN113537102B (en) | Feature extraction method of microseismic signals | |
CN112462355A (en) | Sea target intelligent detection method based on time-frequency three-feature extraction | |
CN116559650A (en) | Automatic change-over switch fault identification method based on multidimensional entropy distance | |
CN117349614A (en) | Frequency stability prediction method based on self-attention mechanism and space-time diagram convolution network | |
CN111025100A (en) | Transformer ultrahigh frequency partial discharge signal mode identification method and device | |
CN112560888B (en) | Method and system for evaluating short circuit resistance of transformer winding based on octyl geometric transformation | |
CN113158781B (en) | Lightning trip type identification method | |
CN110057918A (en) | Damage of composite materials quantitative identification method and system under strong noise background | |
CN116610990A (en) | Method and device for identifying hidden danger of breaker based on characteristic space differentiation | |
CN111414970A (en) | Wind power gear box abnormal data classification method |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200214 |