CN110146812A - A kind of Method of Motor Fault Diagnosis based on the study of characteristic node increment type width - Google Patents
A kind of Method of Motor Fault Diagnosis based on the study of characteristic node increment type width Download PDFInfo
- Publication number
- CN110146812A CN110146812A CN201910401213.8A CN201910401213A CN110146812A CN 110146812 A CN110146812 A CN 110146812A CN 201910401213 A CN201910401213 A CN 201910401213A CN 110146812 A CN110146812 A CN 110146812A
- Authority
- CN
- China
- Prior art keywords
- matrix
- proc
- node
- training
- characteristic
- 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
Classifications
-
- 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
- G01R31/34—Testing dynamo-electric machines
-
- 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
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Control Of Electric Motors In General (AREA)
Abstract
The present invention discloses a kind of Method of Motor Fault Diagnosis based on the study of characteristic node increment type width, and this method diagnoses induction conductivity failure using characteristic node increment type width learning method.Characteristic node increment type width learns the simple for structure of (IBL), can effectively re -training network.The intelligent diagnosing method of a threephase motor is constituted present invention incorporates feature extraction (particle group optimizing-variation mode decomposition and Time-domain Statistics feature), the study of characteristic node increment type width and Non-negative Matrix Factorization.The experimental results showed that this method is better than other algorithms when diagnosing threephase motor failure.In addition, passing through, IBL error that Non-negative Matrix Factorization (NMF) simplifies is small, system is more stable.
Description
Technical field
The present invention relates to Diagnosing Faults of Electrical fields, are based on characteristic node increment type width more particularly, to one kind
The three-phase induction motor method for diagnosing faults of habit.
Background technique
Three phase induction motor (TPIM) provides main drive for our daily lifes.Since TPIM is at low cost, volume
Small, sturdy and durable, maintenance cost is low, and more and more scholars study TPIM.Although TPIM is reliable, they also can
By some adverse effects, these adverse effects will lead to failure, cause major accident.Before major accident generation,
It is necessary to monitor its operation conditions by people.Document show induction conductivity there are winding imbalance, stator or rotor unbalance,
Rotor bar fracture, the failures such as eccentric and bearing defect.
With the development of machine learning, the fault diagnosis research that machine learning is applied to conventional motors is more and more.It is deep
Spend belief network (deep belief network, DBN), extreme learning machine (extreme learn machine, ELM) and volume
Product neural network (conventional neural network, CNN), is currently widely used for direct current generator and alternating current generator
Fault diagnosis in.Although deep learning network is very powerful, due to being related to a large amount of hyper parameter and complicated structure, these
Deep learning network often takes considerable time in the training process.Further, since structure is complicated for deep learning, theoretically divide
It is extremely difficult to analyse deep structure.Most of work require adjustment parameter or increase more layers to improve precision, so in this way
With regard to needing stronger and stronger computing resource.In order to improve the training performance of machine learning, there is researcher to propose a kind of width
Learning method.Unlike above, width learning structure only has two layers, and one layer is input layer, it contain mappings characteristics and
Enhance node.The other is output layer.Although it is a simple structure, it can by increase feature node come
Improve performance.Therefore, it can be applied to the diagnosis of induction conductivity, improve the training speed and accuracy rate of diagnosis.
Fast Fourier Transform (FFT) (FFT) is not suitable for non-stationary signal;The defect of Short Time Fourier Transform (STFT) can when
Between and frequency generate internal association;Wavelet transformation (WT) will lead to energy leakage loss.In the recent period, Dragomiretskiy et al. is mentioned
A kind of variation mode decomposition (VMD) method is gone out, this method assumes that each extraction mode has limited bandwidth and matching
Centre frequency nearby compress.Degree of rarefication before each subpattern is selected as the midbandwidth in spectrum domain.However, VMD
Modulation capability is largely dependent upon intrinsic parameter setting in practical applications.Punish the son in the different configurations and VMD of α
The quantity difference of component K will lead to a variety of different decomposabilitys.Therefore, it is necessary to Optimal Parameters α and K.
Traditional data processing and system model training were carried out in the experimental stage, after the completion of model training, diagnosis
System model cannot be modified.The reconstruct of Diagnosing Faults of Electrical model will spend a large amount of training time, especially depth
Model is practised, this will limit its application significantly.
Summary of the invention
For overcome the deficiencies in the prior art, the present invention proposes a kind of motor based on the study of characteristic node increment type width
Method for diagnosing faults, this method diagnose three-phase motor failure using characteristic node increment type width learning method (IBL).IBL
It is simple for structure, can effectively train and re -training network;It can effectively save the training time, improve fault diagnosis system
Accuracy rate and stability.
To achieve the goals above, technical scheme is as follows:
One kind learning Method of Motor Fault Diagnosis based on characteristic node increment type width, comprising the following steps:
S1: data acquisition and data processing;Acquire stator winding A current signal, stator winding B current signal or stator around
Any two groups of current signals and an acoustic signals in group C current signal, are denoted as x1,x2,x3, two-way electric current collected is believed
Number and acoustic signals be filtered;Filtered data are divided into two groups, wherein one group of current signal data carries out Time-domain Statistics
Acoustic data signal is carried out Time-domain Statistics feature and particle group optimizing-variation mode decomposition by feature respectively;After finally handling
Data be divided into three groups of independent data sets: including training dataset, validation data set and test data set;It is denoted as respectively
xk-Proc-Train、xk-Proc-ValiAnd xk-Proc-Test;
S2: model training, i.e., it will treated xk-Proc-TrainWidth study is carried out, training obtains system model, width
Practise the process of training are as follows:
Use treated training dataset xk-Proc-TrainTrain width learning network, width learning network it is defeated
It is the accuracy rate of diagnosis out;When the accuracy rate of output be more than or equal to setting target accuracy rate when to get arrive training pattern;When defeated
When accuracy rate out is less than the target accuracy rate of setting, system enters incremental learning;
S3: characteristic node incremental learning will treated x that is, by increasing the number of characteristic nodek-Proc-ValiIt carries out
Incremental learning, characteristic node incremental learning process are as follows:
Use treated validation data set xk-Proc-ValiCome training characteristics node increment type width learning network, feature
What node increment type width learning network exported is the accuracy rate of fault diagnosis;When the accuracy rate of output is quasi- not in the target of setting
When true rate ± M%, system continues incremental learning;As target accuracy rate ± M% of the accuracy rate of output in setting to get to spy
Levy node incremental training model;M takes 2.5 in the present invention.
S4: the model obtained by step S3 training is simplified using NMF method, more stable model is obtained, passes through
Test data set xk-Proc-TestOutput matrix is obtained, the fault diagnosis accuracy rate of motor can be obtained with faulty tag comparison.
Preferably, the data acquisition modes in the step S1 are as follows: current signal, microphone acquisition are acquired using oscillograph
Voice signal.
Preferably, in the step S1, it is respectively divided into three independences again after carrying out limit filtration to the data of acquisition
Data set.
Preferably, to the data handling procedure of acoustic signals in the step S1 are as follows: use particle group optimizing-variation mode
It decomposes (PSO-VDM) and Time-domain Statistics method (TDSF) extracts signal characteristic;
After particle group optimizing-variation mode decomposition, due to each intrinsic mode function dimension after variation mode decomposition
It is constant, it needs to reduce dimension, then counts its feature using Sample Entropy (SE), i.e., calculate each natural mode letter using Sample Entropy
Several characteristic features;Its characteristic results saves as x respectivelyk-SE-Train、xk-SE-ValiAnd xk-SE-Test;In order to guarantee all spies
Sign all contributes, xk-SE-Train、xk-SE-ValiAnd xk-SE-TestIn each feature carry out [0,1] normalized.
Preferably, step S1 further include: 10 Time-domain Statistics features are increased separately to two groups of current signals of acquisition, and are carried out
[0,1] normalized, then merge with the acoustic characteristic handled by Sample Entropy to obtain treated training dataset, test
Card data set and test data set, treated training dataset, validation data set and test data set are named as respectively
xk-Proc-Train、xk-Proc-ValiAnd xk-Proc-Test。
Preferably, 10 features are increased to current signal and voice signal are as follows: average value, standard deviation, root mean square, peak value,
The degree of bias, kurtosis, crest factor, gap factor, form factor and impulse ratio.
Preferably, the process of width learning training specifically:
Use treated training dataset xk-Proc-TrainIt trains width learning network, enables X=xk-Proc-Train, i.e.,
Input feature vector set X, if it has N number of sample, each sample is M dimension;
For n Feature Mapping, map feature Zi is expressed as formula (1):
Zi=φ (XWei+βei), i=1 ..., n (1)
Wherein WeiIndicate the random weight matrix of i-th of input feature vector, βeiIndicate the random offset of i-th of input feature vector
Matrix, φ are mapping function, Zn≡[Z1..., Zn] indicate the mapping sets of all characteristic nodes;
For enhancing node, HmIndicate the Enhanced feature of m group enhancing node:
Hm≡ξ(ZnWhm+βhm) (2)
WhmIndicate the random weight matrix of m group enhancing node, βhmIndicate the random offset matrix of m group enhancing node,
ξ is enhancing mapping function, Hm≡[H1..., Hm] indicate all mapping sets for enhancing nodes;All enhancing connection weights indicate
For Wm≡[Wh1..., Whm];
Therefore, output matrix Y is expressed as equation:
Y=[Zn|Hm]Wm (3)
Y is to belong toOutput matrix;
W can be calculated using formula (3)m=[Zn|Hm]+Y。
Preferably, during the model training of step S2, when the accuracy rate of output is less than the target accuracy rate of setting,
Setting value can be greater than or equal to by increasing the accuracy rate that characteristic node number exports model, then obtain training pattern;
Increase characteristic node in learning process, if the composite joint of initial input feature vector and enhancing node is Am=
[Zn|Hm], the additional eigenmatrix increased after characteristic node are as follows:
WhereinTo increase the connection weight after characteristic node;Deviation after being characterized node can obtain
The pseudo inverse matrix of new matrix is expressed as follows out:
Wherein transition matrix
Intermediary matrix
Wherein
New weight are as follows:
By the validation data set x of processingk-Proc-ValiAs input set X, spy can be obtained based on input X and new weight
Levy node increment type width learning model.
It preferably, further include being carried out to the model obtained by step S4 before testing test manifold conjunction in step S4
NMF structure simplifies, if the weight matrix before simplifying isSince the data set of input is normalization, so
Weight matrix is nonnegative number matrix, it is assumed that has nonnegative matrixWith an other nonnegative matrixThen
It arrives:
Wm≈IWr (7)
Wherein WmIt is original matrix, right matrix WrIt is coefficient matrix, left matrix I is basis matrix;
New weight matrix Wr≈I+Wm, the model of step S4 acquisition can be simplified using new weight matrix.
Compared with existing, the invention has the benefit that the present invention proposes one kind based on characteristic node increment type width
Learning method, this method can improve rate of accurateness by increasing the number of characteristic node come re -training model.Due to increasing
The width learning training speed of characteristic node formula is fast, can greatly improve its application field with on-line training.In addition, in order to improve
Precision, the present invention realize the extraction of useful fault signature using signal processing method.
Detailed description of the invention
Fig. 1 is the structural diagram of the present invention.
Fig. 2 characteristic node increment type width learning network schematic diagram.
Specific embodiment
The attached figures are only used for illustrative purposes and cannot be understood as limitating the patent;In order to better illustrate this embodiment, attached
Scheme certain components to have omission, zoom in or out, does not represent the size of actual product;
To those skilled in the art, it is to be understood that certain known features and its explanation, which may be omitted, in attached drawing
's.The following further describes the technical solution of the present invention with reference to the accompanying drawings and examples.The technology provided according to the present invention
Scheme is as shown in Figure 1, this method includes four steps: (a) data acquisition and data processing, (b) width learns, (c) feature section
Point increment type width study, (d) NMF structure simplifies.
A kind of Method of Motor Fault Diagnosis based on the study of characteristic node increment type width, includes the following steps:
Step 1: the acquisition of threephase motor data and data processing
Digital signal is acquired using digital device.They are stator winding A current signal, stator winding B electric current letter respectively
Number and acoustic signals, be denoted as x respectively1,x2,x3.Interference is reduced using digital clipping filter.For each signal, with sound
For signal, signal x3The data of acquisition are divided into three independent data sets, including training dataset, validation data set and survey
Try data set.
Original signal is decomposed using PSO-VMD are as follows: training data, verify data and test data.Training data
It should be labeled as xk-PSO-VMD-Train, verify data is labeled as xk-PSO-VMD-Vali, test data is labeled as xk-PSO-VMD-Test.Consider
Into extraction feature, there are irrelevant informations and redundancy, using Sample Entropy (SE) statistic algorithm to xk-PSO-VMD-Train、
xk-PSO-VMD-ValiAnd xk-PSO-VMD-TestCarry out feature extraction.As a result x is saved as respectivelyk-SE-Train、xk-SE-ValiAnd xk-SE-Test.It removes
Outside signal characteristic, it joined 10 Time-domain Statistics features (TDSF) in signal characteristic.In order to ensure all characteristics have
Uniform contribution, each characteristic are [0,1] by normalization.Last xk-Proc-Train、xk-Proc-ValiAnd xk-Proc-TestIt will be respectively
Labeled as treated training dataset, validation data set and test data set.
Step 2: width learning training
In width study module, first by using width learning model training dataset xk-Proc-Train.Secondly training
BL model output afterwards is the training accuracy rate of training pattern.If training accuracy rate is greater than the target percentage (TP) of setting,
Model will be completed.Otherwise, width study will carry out the study of increment type width by increasing characteristic node.
Width study is based on traditional random vector function neural network.Use treated training dataset
xk-Proc-TrainIt trains width learning network, enables X=xk-Proc-Train, i.e. input feature vector set X, if it has N number of sample, various kinds
It originally is M dimension;
For n Feature Mapping, map feature ZiIt is expressed as formula (1):
Zi=φ (XWei+βei), i=1 ..., n (1)
Wherein WeiIndicate the random weight matrix of i-th of input feature vector, βeiIndicate the random offset of i-th of input feature vector
Matrix, φ are mapping function, Zn≡[Z1..., Zn] indicate the mapping sets of all characteristic nodes;
For enhancing node, HmIndicate the Enhanced feature of m group enhancing node:
Hm≡ξ(ZnWhm+βhm) (2)
WhmIndicate the random weight matrix of m group enhancing node, βhmIndicate the random offset matrix of m group enhancing node,
ξ is enhancing mapping function, Hm≡[H1..., Hm] indicate all mapping sets for enhancing nodes;All enhancing connection weights indicate
For Wm≡[Wh1..., Whm];
Therefore, output matrix Y is expressed as equation:
Y=[Zn|Hm]Wm (3)
Y is to belong toOutput matrix;
W can be easily calculated using formula (3)m=[Zn|Hm]+Y。
Step 3: increasing the study of characteristic node increment type width
Under certain conditions, additional increase node is needed in order to improve the accuracy rate of system, increased in learning process
" characteristic node ".Assuming that originally when input feature value and characteristic node composite joint be Am=[Zn|Hm], it is additional to increase spy
Feature after levying node are as follows:
WhereinTo increase the connection weight after characteristic node;It, can to increase the deviation after characteristic node
Show that the pseudo inverse matrix of new matrix is expressed as follows:
Wherein transition matrix
Intermediary matrix
Wherein
New weight are as follows:
By the validation data set x of processingk-Proc-ValiAs input set X, spy can be obtained based on input X and new weight
Levy node increment type width learning model.Characteristic node increasable algorithm does not need to calculate entire Am+1, and need to only calculate additional increasing
Incremental learning can be realized by adding the pseudoinverse of characteristic node, so as to improve the speed of network re -training.
Step 4: NMF structure simplifies the error rate for reducing Diagnostic system of motor fault
After increasing characteristic node in incremental learning, due to the input data that initialization is insufficient or excessive, Ke Nengcun
In the node or data of redundancy.In general, this structure can be simplified by a series of low-rank approximation methods.Present invention choosing
It selects Non-negative Matrix Factorization (NMF) and the offer structure simplification of node increment type width learning model is provided.
If the weight matrix before simplifying isSince the data set of input is normalization, so weight
Matrix is nonnegative number matrix, it is assumed that has nonnegative matrixWith an other nonnegative matrixIt can then obtain
It arrives:
Wm≈IWr (7)
Wherein WmTwo minor matrixs can be resolved into.M is the dimension of Enhanced feature value, and n is sample number, and r is contraction.WmIt is
One original matrix.The matrix W on the rightrIt is coefficient matrix.The matrix I on the left side is known as fundamental matrix.The column vector of original matrix is
The weighted sum of all column vectors in left matrix, weight coefficient are the element that right matrix corresponds to column vector.R should be generally selected to be less than m, from
And realize the dimensionality reduction of original matrix, with coefficient matrix WrThe dimensionality reduction matrix of data characteristics is obtained instead of original matrix:
Wr≈I+Wm
Thus it can simplify the simplified model of step S4 acquisition using new weight matrix.
The invention proposes a kind of new Method of Motor Fault Diagnosis.This method by feature extraction, width study, feature section
The study of point increment type width and NMF-IBL are combined and are diagnosed to electrical fault, are improved the measuring accuracy of system, are reduced
Average test error shortens training time and retraining time.Firstly, this method is from winding A&B electric current and acoustic signal
Raw sample data is extracted, die filter, PSO-VMD, SampEn, TDSF and normalized these sample numbers are then passed through
According to.Secondly, the present invention is by these, treated that data are input in width learning network model, and is trained to its network.
If measuring accuracy is undesirable, retraining model is learnt using characteristic node increment type width.Finally, using NMF method come
Simplify network structure.The experimental results showed that the motor based on characteristic node increment type width study and Non-negative Matrix Factorization (NMF)
Method for diagnosing faults is effective in terms of improving diagnostic accuracy and training speed.Innovative point of the invention is as follows:
1. proposing the method for a kind of new diagnosis three phase induction motor stator and rotor fault.
2. the Feature Extraction Technology combined using PSO-VMD, SampEn, TDSF, improves the accuracy of system diagnostics.
3. proposing that a kind of re -training method based on the study of characteristic node increment width improves measuring accuracy and training
Speed.
4. simplifying IBL structure using Non-negative Matrix Factorization (NMF), the average test error of system is reduced.
To sum up: present invention incorporates feature extractions (particle group optimizing-variation mode decomposition and Time-domain Statistics feature), feature
The study of node increment type width constitutes the intelligent diagnosing method of a threephase motor with Non-negative Matrix Factorization.Experimental result table
Bright this method is better than other algorithms when diagnosing threephase motor failure.In addition, simplified by Non-negative Matrix Factorization (NMF)
IBL error is small, system is more stable.
Claims (9)
1. one kind learns Method of Motor Fault Diagnosis based on characteristic node increment type width, which comprises the following steps:
S1: data acquisition and data processing;Acquire stator winding A current signal, stator winding B current signal or stator winding C
Any two groups of current signals and an acoustic signals, are denoted as x in current signal1,x2,x3, by two-way current signal collected and
Acoustic signals are filtered;Filtered data are divided into two groups, wherein one group of current signal data carries out Time-domain Statistics feature,
Acoustic data signal is carried out to Time-domain Statistics feature and particle group optimizing-variation mode decomposition respectively;Finally will treated number
According to being divided into three groups of independent data sets: including training dataset, validation data set and test data set;It is denoted as respectively
xk-Proc-Train、xk-Proc-ValiAnd xk-Proc-Test;
S2: model training, i.e., it will treated xk-Proc-TrainWidth study is carried out, training obtains system model, width study instruction
Experienced process are as follows:
Use treated training dataset xk-Proc-TrainTrain width learning network, the output of width learning network is
The accuracy rate of diagnosis;When the accuracy rate of output be more than or equal to setting target accuracy rate when to get arrive training pattern;When output
When accuracy rate is less than the target accuracy rate of setting, system enters incremental learning;
S3: characteristic node incremental learning will treated x that is, by increasing the number of characteristic nodek-Proc-ValiCarry out increment
Study, characteristic node incremental learning process are as follows:
Use treated validation data set xk-Proc-ValiCome training characteristics node increment type width learning network, characteristic node
What increment type width learning network exported is the accuracy rate of fault diagnosis;When the accuracy rate of output is not in the target accuracy rate of setting
When ± M%, system continues incremental learning;As target accuracy rate ± M% of the accuracy rate of output in setting to get arriving feature section
Point incremental training model;
S4: the model obtained by step S3 training is simplified using NMF method, more stable model is obtained, passes through test
Data set xk-Proc-TestOutput matrix is obtained, the fault diagnosis accuracy rate of motor can be obtained with faulty tag comparison.
2. the method according to claim 1, wherein the data acquisition modes in the step S1 are as follows: use and show
Wave device acquires current signal, microphone collected sound signal.
3. the method according to claim 1, wherein carrying out clipping filter to the data of acquisition in the step S1
Three independent data sets are respectively divided into after wave again.
4. method according to claim 1-3, which is characterized in that the data of acoustic signals in the step S1
Treatment process are as follows: it is special that signal is extracted using particle group optimizing-variation mode decomposition (PSO-VDM) and Time-domain Statistics method (TDSF)
Sign;
After particle group optimizing-variation mode decomposition, since each intrinsic mode function dimension is constant after variation mode decomposition,
It needs to reduce dimension, then counts its feature using Sample Entropy (SE), i.e., calculate each intrinsic mode function using Sample Entropy
Characteristic features;Its characteristic results saves as x respectivelyk-SE-Train、xk-SE-ValiAnd xk-SE-Test;In order to guarantee all features all
It contributes, xk-SE-Train、xk-SE-ValiAnd xk-SE-TestIn each feature carry out [0,1] normalized.
5. according to the method described in claim 4, it is characterized in that, step S1 further include: to two groups of current signal difference of acquisition
Increase by 10 Time-domain Statistics features, and carry out [0,1] normalized, then is closed with the acoustic characteristic handled by Sample Entropy
And obtain that treated training dataset, validation data set and test data set, treated training dataset, verify data
Collection and test data set are named as x respectivelyk-Proc-Train、xk-Proc-ValiAnd xk-Proc-Test。
6. according to the method described in claim 5, it is characterized in that, increasing by 10 features to current signal and voice signal are as follows:
Average value, standard deviation, root mean square, peak value, the degree of bias, kurtosis, crest factor, gap factor, form factor and impulse ratio.
7. the method according to claim 1, wherein the process of width learning training specifically:
Use treated training dataset xk-Proc-TrainIt trains width learning network, enables X=xk-Proc-Train, that is, input
Characteristic set X, if it has N number of sample, each sample is M dimension;
For n Feature Mapping, map feature ZiIt is expressed as formula (1):
Zi=φ (XWei+βei), i=1 ..., n (1)
Wherein WeiIndicate the random weight matrix of i-th of input feature vector, βeiIndicate the random offset matrix of i-th of input feature vector,
φ is mapping function, Zn≡[Z1..., Zn] indicate the mapping sets of all characteristic nodes;
For enhancing node, HmIndicate the Enhanced feature of m group enhancing node:
Hm≡ξ(ZnWhm+βhm) (2)
WhmIndicate the random weight matrix of m group enhancing node, βhmIndicate that the random offset matrix of m group enhancing node, ξ are to increase
Strong mapping function, Hm≡[H1..., Hm] indicate all mapping sets for enhancing nodes;All enhancing connection weights are expressed as Wm≡
[Wh1..., Whm];
Therefore, output matrix Y is expressed as equation:
Y=[Zn|Hm]Wm (3)
Y is to belong toOutput matrix;
W can be calculated using formula (3)m=[Zn|Hm]+Y。
8. the method according to the description of claim 7 is characterized in that during the model training of step S2, when the standard of output
When true rate is less than the target accuracy rate of setting, the accuracy rate that model can be made to export by increase characteristic node number, which is greater than or equal to, to be set
Definite value then obtains training pattern;
Increase characteristic node in learning process, if the composite joint of initial input feature vector and enhancing node is Am=[Zn|
Hm], the additional eigenmatrix increased after characteristic node are as follows:
WhereinTo increase the connection weight after characteristic node;Deviation after being characterized node can obtain new square
The pseudo inverse matrix of battle array is expressed as follows:
Wherein transition matrix
Intermediary matrix
Wherein
New weight are as follows:
By the validation data set x of processingk-Proc-ValiAs input set X, feature section can be obtained based on input X and new weight
Point increment type width learning model.
9. according to the method described in claim 8, it is characterized in that, also being wrapped before testing test manifold conjunction in step S4
It includes and the model progress NMF structure obtained by step S4 is simplified, if the weight matrix before simplifying isDue to
The data set of input is normalization, so weight matrix is nonnegative number matrix, it is assumed that there is nonnegative matrixIn addition
One nonnegative matrixThen obtain:
Wm≈IWr (7)
Wherein WmIt is original matrix, right matrix WrIt is coefficient matrix, left matrix I is basis matrix;
New weight matrix Wr≈I+Wm, the model of step S4 acquisition can be simplified using new weight matrix.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910401213.8A CN110146812B (en) | 2019-05-15 | 2019-05-15 | Motor fault diagnosis method based on feature node incremental width learning |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910401213.8A CN110146812B (en) | 2019-05-15 | 2019-05-15 | Motor fault diagnosis method based on feature node incremental width learning |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110146812A true CN110146812A (en) | 2019-08-20 |
CN110146812B CN110146812B (en) | 2021-07-13 |
Family
ID=67595255
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910401213.8A Active CN110146812B (en) | 2019-05-15 | 2019-05-15 | Motor fault diagnosis method based on feature node incremental width learning |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110146812B (en) |
Cited By (10)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110673578A (en) * | 2019-09-29 | 2020-01-10 | 华北电力大学(保定) | Fault degradation degree determination method and device, computer equipment and storage medium |
CN110749793A (en) * | 2019-10-31 | 2020-02-04 | 杭州中恒云能源互联网技术有限公司 | Dry-type transformer health management method and system based on width learning and storage medium |
CN111899905A (en) * | 2020-08-05 | 2020-11-06 | 哈尔滨工程大学 | Fault diagnosis method and system based on nuclear power device |
CN112215281A (en) * | 2020-10-12 | 2021-01-12 | 浙江大学 | Fan blade icing fault detection method |
CN112308159A (en) * | 2020-11-05 | 2021-02-02 | 湖南科技大学 | Image identification and classification method based on prediction increment width learning |
CN112508058A (en) * | 2020-11-17 | 2021-03-16 | 安徽继远软件有限公司 | Transformer fault diagnosis method and device based on audio characteristic analysis |
CN112861328A (en) * | 2021-01-22 | 2021-05-28 | 东北电力大学 | Generator damping evaluation device and method based on random response signals |
CN113419519A (en) * | 2021-07-14 | 2021-09-21 | 北京航空航天大学 | Electromechanical product system or equipment real-time fault diagnosis method based on width learning |
CN113688786A (en) * | 2021-09-10 | 2021-11-23 | 广东电网有限责任公司广州供电局 | PSO (particle swarm optimization) width learning-based voltage sag multiple disturbance source identification method |
CN114123896A (en) * | 2021-11-30 | 2022-03-01 | 江南大学 | Permanent magnet synchronous motor control method and system based on incremental width learning system |
Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20020038510A1 (en) * | 2000-10-04 | 2002-04-04 | Orbotech, Ltd | Method for detecting line width defects in electrical circuit inspection |
CN107644231A (en) * | 2017-09-19 | 2018-01-30 | 广东工业大学 | A kind of generator amature method for diagnosing faults and device |
CN107944274A (en) * | 2017-12-18 | 2018-04-20 | 华中科技大学 | A kind of Android platform malicious application off-line checking method based on width study |
CN108734301A (en) * | 2017-06-29 | 2018-11-02 | 澳门大学 | A kind of machine learning method and machine learning device |
CN108960339A (en) * | 2018-07-20 | 2018-12-07 | 吉林大学珠海学院 | A kind of electric car induction conductivity method for diagnosing faults based on width study |
CN109522802A (en) * | 2018-10-17 | 2019-03-26 | 浙江大学 | The pump of application experience mode decomposition and particle swarm optimization algorithm is made an uproar removing method |
EP3480714A1 (en) * | 2017-11-03 | 2019-05-08 | Tata Consultancy Services Limited | Signal analysis systems and methods for features extraction and interpretation thereof |
-
2019
- 2019-05-15 CN CN201910401213.8A patent/CN110146812B/en active Active
Patent Citations (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20020038510A1 (en) * | 2000-10-04 | 2002-04-04 | Orbotech, Ltd | Method for detecting line width defects in electrical circuit inspection |
CN108734301A (en) * | 2017-06-29 | 2018-11-02 | 澳门大学 | A kind of machine learning method and machine learning device |
CN107644231A (en) * | 2017-09-19 | 2018-01-30 | 广东工业大学 | A kind of generator amature method for diagnosing faults and device |
EP3480714A1 (en) * | 2017-11-03 | 2019-05-08 | Tata Consultancy Services Limited | Signal analysis systems and methods for features extraction and interpretation thereof |
CN107944274A (en) * | 2017-12-18 | 2018-04-20 | 华中科技大学 | A kind of Android platform malicious application off-line checking method based on width study |
CN108960339A (en) * | 2018-07-20 | 2018-12-07 | 吉林大学珠海学院 | A kind of electric car induction conductivity method for diagnosing faults based on width study |
CN109522802A (en) * | 2018-10-17 | 2019-03-26 | 浙江大学 | The pump of application experience mode decomposition and particle swarm optimization algorithm is made an uproar removing method |
Non-Patent Citations (2)
Title |
---|
SAIBIAO JIANG ET.AL: "An Efficient Fault Diagnostic Method for Three-Phase Induction Motors Based on Incremental Broad Learning and Non-Negative Matrix Factorization", 《IEEE ACCESS》 * |
贾晨 等: "基于宽度学习方法的多模态信息融合", 《智能系统学报》 * |
Cited By (16)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110673578A (en) * | 2019-09-29 | 2020-01-10 | 华北电力大学(保定) | Fault degradation degree determination method and device, computer equipment and storage medium |
CN110749793A (en) * | 2019-10-31 | 2020-02-04 | 杭州中恒云能源互联网技术有限公司 | Dry-type transformer health management method and system based on width learning and storage medium |
CN111899905B (en) * | 2020-08-05 | 2022-11-01 | 哈尔滨工程大学 | Fault diagnosis method and system based on nuclear power device |
CN111899905A (en) * | 2020-08-05 | 2020-11-06 | 哈尔滨工程大学 | Fault diagnosis method and system based on nuclear power device |
CN112215281A (en) * | 2020-10-12 | 2021-01-12 | 浙江大学 | Fan blade icing fault detection method |
CN112308159A (en) * | 2020-11-05 | 2021-02-02 | 湖南科技大学 | Image identification and classification method based on prediction increment width learning |
CN112308159B (en) * | 2020-11-05 | 2023-04-07 | 湖南科技大学 | Image identification and classification method based on prediction increment width learning |
CN112508058A (en) * | 2020-11-17 | 2021-03-16 | 安徽继远软件有限公司 | Transformer fault diagnosis method and device based on audio characteristic analysis |
CN112508058B (en) * | 2020-11-17 | 2023-11-14 | 安徽继远软件有限公司 | Transformer fault diagnosis method and device based on audio feature analysis |
CN112861328A (en) * | 2021-01-22 | 2021-05-28 | 东北电力大学 | Generator damping evaluation device and method based on random response signals |
CN113419519A (en) * | 2021-07-14 | 2021-09-21 | 北京航空航天大学 | Electromechanical product system or equipment real-time fault diagnosis method based on width learning |
CN113419519B (en) * | 2021-07-14 | 2022-05-13 | 北京航空航天大学 | Electromechanical product system or equipment real-time fault diagnosis method based on width learning |
CN113688786A (en) * | 2021-09-10 | 2021-11-23 | 广东电网有限责任公司广州供电局 | PSO (particle swarm optimization) width learning-based voltage sag multiple disturbance source identification method |
CN113688786B (en) * | 2021-09-10 | 2022-07-12 | 广东电网有限责任公司广州供电局 | PSO (particle swarm optimization) width learning-based voltage sag multiple disturbance source identification method |
CN114123896B (en) * | 2021-11-30 | 2022-05-24 | 江南大学 | Permanent magnet synchronous motor control method and system based on incremental width learning system |
CN114123896A (en) * | 2021-11-30 | 2022-03-01 | 江南大学 | Permanent magnet synchronous motor control method and system based on incremental width learning system |
Also Published As
Publication number | Publication date |
---|---|
CN110146812B (en) | 2021-07-13 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110146812A (en) | A kind of Method of Motor Fault Diagnosis based on the study of characteristic node increment type width | |
CN109977920B (en) | Water turbine set fault diagnosis method based on time-frequency spectrogram and convolutional neural network | |
Jiang et al. | An efficient fault diagnostic method for three-phase induction motors based on incremental broad learning and non-negative matrix factorization | |
CN108960339A (en) | A kind of electric car induction conductivity method for diagnosing faults based on width study | |
Lei et al. | EEMD method and WNN for fault diagnosis of locomotive roller bearings | |
Wu et al. | Investigation of engine fault diagnosis using discrete wavelet transform and neural network | |
CN110702411B (en) | Residual error network rolling bearing fault diagnosis method based on time-frequency analysis | |
Huang et al. | Application of wavelet neural networks on vibration fault diagnosis for wind turbine gearbox | |
CN109740523A (en) | A kind of method for diagnosing fault of power transformer based on acoustic feature and neural network | |
CN108229544A (en) | A kind of NPC three-level photovoltaic inverters open fault diagnostic method | |
CN110705456A (en) | Micro motor abnormity detection method based on transfer learning | |
Jiang et al. | A fault diagnostic method for induction motors based on feature incremental broad learning and singular value decomposition | |
CN112329914B (en) | Fault diagnosis method and device for buried transformer substation and electronic equipment | |
CN112083328A (en) | Fault diagnosis method, system and device for high-voltage circuit breaker | |
CN107886078A (en) | A kind of Threshold Denoising method based on layered self-adapting threshold function table | |
Chen et al. | Research on fault diagnosis method of rotating machinery based on deep learning | |
CN112327208A (en) | Fault diagnosis method and device for turn-to-turn short circuit of phase modulator rotor winding | |
Chen et al. | Understanding IFRA for detecting synchronous machine winding short circuit faults based on image classification and smooth grad-CAM++ | |
Wang et al. | Feature extraction techniques of non-stationary signals for fault diagnosis in machinery systems | |
Zhou et al. | Three-phase asynchronous motor fault diagnosis using attention mechanism and hybrid CNN-MLP by multi-sensor information | |
Sun et al. | A Novel Fault Diagnosis of GIS Partial Discharge Based on Improved Whale Optimization Algorithm | |
CN113591792A (en) | Transformer fault identification method based on self-organizing competitive neural network algorithm | |
Zhao et al. | An intelligent fault diagnosis method for a diesel engine valve based on improved wavelet packet-Mel frequency and convolutional neural network | |
Lan et al. | A Novel Motor Fault Diagnosis Method Based on Generative Adversarial Learning with Distribution Fusion of Discrete Working Conditions. | |
Zhu et al. | Aiming to Complex Power Quality Disturbances: A Novel Decomposition and Detection Framework |
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 |