CN103761372B - A kind of multi-electrical level inverter Fault Diagnosis Strategy based on pivot analysis with more classification Method Using Relevance Vector Machines - Google Patents
A kind of multi-electrical level inverter Fault Diagnosis Strategy based on pivot analysis with more classification Method Using Relevance Vector Machines Download PDFInfo
- Publication number
- CN103761372B CN103761372B CN201410005022.7A CN201410005022A CN103761372B CN 103761372 B CN103761372 B CN 103761372B CN 201410005022 A CN201410005022 A CN 201410005022A CN 103761372 B CN103761372 B CN 103761372B
- Authority
- CN
- China
- Prior art keywords
- sample
- fault
- model
- probability
- test
- 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.)
- Active
Links
Abstract
The invention discloses a kind of multi-electrical level inverter Fault Diagnosis Strategy for Method Using Relevance Vector Machine (the PCA mRVM) that classified more based on pivot analysis.Original sample is carried out dimensionality reduction by the Diagnostic Strategy by pivot analysis first, multiple principal components with fault signature is extracted, to form training sample；Fault diagnosis is carried out to training sample secondly by more classification Method Using Relevance Vector Machines, and exports the probability of each fault category, diagnostic result is used as using the fault category of maximum probability.The Fault Diagnosis Strategy of the present invention, which is used to have in larger samples space and more multi-class fault diagnosis, obvious advantage, it is high with model degree of rarefication, the advantages such as computation complexity is low, more importantly mRVM can export membership probability of all categories, output has probability statistics meaning, is easy to analyze uncertain problem.
Description
Technical field：
The present invention relates to the fault diagnosis of the multi-electrical level inverter in field of power electronics, more particularly to one kind to be based on PCA-
MRVM method for diagnosing faults.
Background technology：
With the development of Power Electronic Technique and the reduction of power electronic devices production cost, high-voltage high-power converter quilt
It is widely used in various electrical equipments, such as high-power AC motor transmission, active power filtering and new-energy grid-connected etc..And
In order to meet the power system development demand increasingly improved, multi-electrical level inverter is as a kind of new converter, because of its main electricity
The advantages that line structure is simply and control circuit is very flexible is most living as the high-power conversion aspect of field of power electronics mesohigh
One branch of jump, has obtained rapid popularization in the industrial production.Wherein Cascade H bridge type inverter is because it does not need a large amount of pincers
Position diode or electric capacity, without balanced capacitance voltage, it is easy to modularization and extension in structure, operating efficiency is high, therefore obtains
Relatively broad application.Although Cascade H bridge type inverter is that same number level and required device are exported in multi-electrical level inverter
Minimum one kind, but the switching device quantity required for it is still multiplied with the increase of system level number.Largely make
The possibility (switch open or short circuit) of failure generation is added with power semiconductor, and with the rise of voltage, therefore
The probability that barrier occurs also becomes big.Although the generation of Cascade H bridge type multi-electrical level inverter is Power Electronic Technique in high pressure, high-power
The application of occasion provides many facilities, but once breaks down, and gently then causes enterprise to stop production, heavy then can cause disaster sexual behavior
Therefore bring huge loss to society.Research shows that the failure of switching device accounts for whole in the frequency conversion speed-adjusting system of inverter power supply
The 82.5% of individual drive system failure, it is the link for being easiest to break down in drive system.
Multi-electrical level inverter open fault diagnostic method mainly has at present：Rothenhagen proposes the event of IGBT open circuits
Hindering diagnostic method, this method needs substantial amounts of test point and sensor, so circuit can be made to become more complicated, and cost is higher,
And easily caused wrong diagnosis by external disturbance；Soup clear spring et al. propose according to detection inverter outlet side PWM voltages and
Output current polarity diagnoses the Fault Identification scheme of open-circuit fault of power tubes, and this method divides the mode of operation of inverter
Area, it cannot be only used for single power tube and break down, the situation that also diagnosable multiple power tubes break down, but due to output electricity
Flow by load effect, cause this method generalization ability poor, and increasing with inverter level number, mode of operation is divided
Area can also become extremely complex；Other Yang Zhong woodss et al. propose can by detecting the frequency content of inverter direct-flow side electric current
To realize the fault diagnosis of inverter switching device pipe, when each switching tube is working properly, there is no harmonic components in DC side electric current, and
When there is switching tube to break down, harmonic components then occur in DC side.Although this method can be carried out with switch tube failure
Detection, but the position of breakdown switch pipe can not be specifically navigated to, and when circuit is more complicated, easily led by external disturbance
Cause wrong diagnosis.Analyzed more than, currently for the troubleshooting issue of multi-electrical level inverter, be primarily present following difficulty：
(1) the more difficult extraction of fault signature sample；(2) sample data dimension is larger, and processing is more difficult and computationally intensive；(3) current diagnosis
Method is only capable of exporting diagnostic accuracy, it is impossible to exports the concrete outcome of each classification, inconvenience analysis uncertain problem etc..
The content of the invention：
One kind that the technical problems to be solved by the invention are the technical problem for above-mentioned prior art and proposed is fitted
For cascaded H-bridges multi-electrical level inverter Fault Diagnosis Strategy, its purpose is exactly to overcome extensive energy present in prior art
The shortcomings of power is poor, cost is high, diagnosis speed is slow, frequency domain is transformed into by using FFT by sample data from time domain, consequently facilitating therefore
Hinder the extraction of feature samples, carry out dimensionality reduction secondly by PCA, reduce amount of calculation etc., finally using mRVM, can export of all categories
Membership probability, output have probability statistics meaning, are easy to analyze uncertain problem.
To achieve these goals, present invention employs following technical scheme：
1st, a multi-electrical level inverter simulation model is constructed
One single-phase level inverter circuit system model of cascaded H-bridges five is built using Matlab/Simulink, then added
Failsafe link, the single switching transistor open fault test of diverse location is carried out successively, extract the fault characteristic signals of a cycle,
Establish fault signature collection；
2nd, data prediction
The more obvious frequency domain of fault signature is converted a signal into using FFT and extracts fault characteristic signals；
3rd, failure pivot is extracted
Because fault signature sample dimension is larger, the computation complexity directly handled is higher, and processing speed is slower, therefore
Dimensionality reduction is carried out to sample using PCA herein, extracted on the influential principal component of classification, so as to significantly improve processing speed；
4th, the model training stage
MRVM parameters are initialized, the present invention uses kernel function of the RBF kernel functions as mRVM, and uses cross-validation method
Determine that core is joined.Then the fault signature sample after simplification is substituted into mRVM models to be trained；Training then uses test specimens after terminating
This is tested the Diagnostic Strategy, if the nicety of grading that training obtains reaches the required precision of setting, training is completed；If instruction
Practice result and be not up to given requirements, then jump to characteristic extraction step and change mRVM core ginseng.The required precision of training can be with
According to condition of work sets itself.5th, the real-time diagnosis stage
At this stage, directly sample data to be sorted is imported in the model trained, you can obtain each sample
Probability output result, and using the sample of wherein maximum probability as classification diagnosis result.
The technical effects of the invention are that：PCA-mRVM strategies of the present invention, first, examined in traditional RVM classification
Dimensionality reduction link is added on the basis of disconnected algorithm, to meet to have a case that larger samples, and PCA of the present invention is calculated
Method, the feature samples extracted are more accurate and more representative；Secondly, the Diagnostic Strategy can directly export it is of all categories into
Member's probability, output have probability statistics meaning, are easy to analyze uncertain problem.Because the Diagnostic Strategy is only with output end electricity
Signal is pressed as its input signal, therefore avoids a large amount of uses of sensor, and reduces the complexity and structure of system
Cost.
Brief description of the drawings
Fig. 1 is multi-electrical level inverter real-time fault diagnosis policy construction figure of the present invention, wherein：
1：Threephase asynchronous machine
2：Cascaded H-bridges multi-electrical level inverter
3：Voltage sampling circuit
Fig. 2 is the model training process flow diagram flow chart of Fault Diagnosis Strategy of the present invention；
Fig. 3 is the output voltage waveform of inverter single switching transistor failure in the present invention；
Fig. 4 is that output voltage waveforms are carried out with the spectrogram after FFT；
Fig. 5 is layering Bayesian model structure chart；
Fig. 6 is Fast Type-II ML program flow diagram.
Embodiment：
In order that the technical means, the inventive features, the objects and the advantages of the present invention are easy to understand, tie below
Conjunction is specifically illustrating, further the explaination present invention.
A kind of multi-electrical level inverter real-time fault diagnosis policy construction figure is as shown in Figure 1.Diagnostic Strategy is mainly by four parts
Composition：Data prediction and feature extraction, classification diagnosis model, output diagnostic result, switching mode calculating.It is of the invention main
Three parts before research, switching mode are calculated and put aside here.The strategy works principle is as follows：Output electricity first to inverter
Pressure signal is pre-processed and to its feature extraction, so as to reduce the dimension of sample；Secondly, the data after processing are brought into and instructed
The mRVM models (when the model is applied to strange sample only need to training is once when initial) perfected carry out diagnostic classification；So
After export membership probability of all categories, and using the sample of wherein maximum probability as classification diagnosis result.
The waveform that Surin Khomifoi et al. analyze cascade connection multi-level its output voltage is and switched on-off close
Related, by taking single-phase cascaded H-bridges as an example, in the event of a short trouble, some switched voltage output will be zero always.And
In the event of an open fault, output voltage is (average in a cycle).When switch breaks down (open circuit or short circuit)
Desired output voltage waveforms can not be produced.Analysis is understood, when different H-bridge units break down, its fault signature is different, output
Phase voltage signal is closely related with fault type and position, and exports phase voltage and do not change with load.Therefore the present invention adopts
Fault diagnosis signal is used as by the use of output voltage.
PCA-mRVM multi-electrical level inverters Diagnostic Strategy proposed by the present invention is mainly introduced below, mainly includes model structure
Build, data prediction, extraction failure pivot, model training stage and test phase etc..Fig. 2 is is proposed Fault Diagnosis Strategy
Training process flow chart, specific step is as follows.
Step1：Model construction
One single-phase level inverter circuit system model of cascaded H-bridges five is built using Matlab/Simulink, utilized
SHPWM (Subharmonic PWM) is controlled, frequency of modulated wave 50Hz, carrier frequency 4kHz, and sample frequency is
40kHz.Then failsafe link is added, carries out the single switching transistor open fault test of diverse location successively, extraction inverter is every
The output voltage signal of kind failure a cycle, as shown in figure 3, establishing fault signature collection；
Step2：Data prediction
Sample the discrete output voltage signal of obtained a cycleAfter FFT, data set is obtained
Xa×b, wherein a is observation number, and b is harmonic wave number (corresponding 0 arrives b-1 subharmonic).FFT formula：Here Wb=e-j2π/b, k=0,1 ..., b/2-1.By pretreated failure
Sample data Xa×bCorresponded with its tag along sort vector l, as input sample data [X, l], Fig. 4 gives one group of sample
Pretreated spectrogram, above-mentioned Xa×bI.e. X here9×400；Step3：Extract failure pivot
(1) pretreated data sample X ∈ R are obtained by step 2a×b, wherein a is corresponding data sample points, and b is corresponding
Sample dimension.
(2) covariance matrix is calculated, solves covariance matrix RXEigenvalue λ and corresponding feature value vector P：
Covariance matrix：RX=E { [X-E (X)] [X-E (X)]T}；
By solving equation | λ I-RX|=0 He | λiI-RX|pi=0, i=1,2 ..., b try to achieve eigenvalue λ and characteristic value to
Measure P.Wherein, λiFor RXIth feature value, meet λ1≥λ2≥…≥λb, piIt is corresponding to eigenvalue λiCharacteristic vector, P
=[p1,p2,…,pb]T。
(3) relative principal component number m is selected according to contribution rate of accumulative total：
Wherein, CL is a reference value, and its value can generally choose CL=85% with User Defined.
(4) score matrix is calculated：
T=X × P, wherein, T=[t1,t2,…,tb]；
X principal component models are after pivot analysis:
Wherein, m is the number of taken pivot, tiIt is pivot score vector, piIt is load vector, is X covariance matrixes again
Characteristic vector.XpThe Main change of data is represented for preceding m pivot, E is error matrix.
Step4：The model training stage
(1) to pivot sample XpClass label l ∈ { 1,2 ... C } are set gradually, C is that failure is total, data sample now
This is Xp∈RN×M, and training sample (X is set respectivelytrain,ltrain) and test sample (Xtest,ltest)。
(2) kernel function is selected：The present invention chooses Radial basis kernel function (RBF), and wherein σ is the high wide parameter of kernel function.
(3) m-RVM model constructions
Introduce auxiliary regression target Y ∈ RC×NWith weight parameter W ∈ RN×C, obtain standard noise regression model：
Multinomial probability Copula is introduced, regressive object is converted into class label：
Resulting multinomial probability likelihood function is as follows, wherein u~N (0,1):
It is openness in order to ensure model, zero-mean is introduced for weight vectors, variance isStandard normal prior distribution.
By Study first αncThe matrix of composition is designated as A ∈ RN×C,αncObedience hyper parameter is τ, and υ Gamma is distributed, and Fig. 5 gives layering
Bayesian model structure.
Bayesian model according to Fig. 5 derives posterior probability, wherein AcFor the diagonal matrix as derived from A c row：
Obtaining method by maximum a-posteriori estimation can obtain：
The update method such as following formula based on maximum a posteriori (MAP) of weight when therefore giving classification：
According to above formula, the E-step forms of auxiliary variable can be derived, for i classes such as following formula：
ForSuch as following formula：
The Posterior probability distribution of weight vectors Study first such as following formula：
Gamma distribution average be：
(4) m-RVM model learnings
The present invention adds or removed sample by one empty model of construction and into it to obtain degree of rarefication, addition or removal
Sample is the contribution margin to model according to them, here using quick type-II maximum likelihoods (Fast Type-II ML) parameter
Update method, Fig. 6 give Fast Type-II ML program flow diagrams.
Marginal likelihood function be P (Y | K,A)=log ∫ P (Y | K, W) P (W | A) dW, asked for by logarithmic, wherein
It can be decomposed intoWhereinWhen representing i-th of sample of removalValue, can be with
Derive：
Marginal likelihood function can be further decomposed such as following formula now：
WhereinFor the sparse factor,For quantizing factor.Pass through solutionCan
Obtain stationary point：
Contribution factor θ is defined to the information contribution amount of model for the ease of i-th of sample of descriptioni：
If often step iterative process has active set Ο, it is made up of D dimension active samples.If some sample θ values are just, but
It is but to be not included in Ο, this sample is just added in Ο by we；, whereas if some sample is in Ο, but
It is that θ values are negative, then removes the sample.Therefore in the learning process of model, MAP update method formula amendments are as follows, wherein
K*∈RD×N,A*∈RD×D, D ＜ ＜ N：
By formula (10) and giveNow provide αiInitialization formula it is as follows：
(5) m-RVM end conditions
When m-RVM models meet following three conditions in the training stage, then training terminates：
(a)Include and contribute sample；
(b)Reject all non-contribution samples；
(c)|logA(K)-logA(K-1)| ＜ ε, i.e. Study first matrix A no longer update in an iterative process.
Step5：Test phase
First by test sample (Ttest,ltest) after processing by step 3 and step 4, bring the mRVM models trained into
In, the probability output result of each sample is obtained, and using the sample of wherein maximum probability as classification diagnosis result, can specifically join
See formula (1) (2).
The present invention applies example：
In order to verify the validity of carried Diagnostic Strategy and diagnosis performance, the present invention has built a single-phase five level cascade
H bridge inverters, and obtain 8 switching tube normal works and respectively open circuit when output voltage signal (only examined as experimental data
Consider an only switching tube simultaneously to break down), experimental data set is in ma(m is obtained under conditions of=0.86aIt is modulating wave and load
The amplitude modulation ratio of ripple), it is controlled using SHPWM (Subharmonic PWM), frequency of modulated wave 50Hz, carrier frequency
For 4kHz, sample frequency is that 40kHz number of samples N is 800, with the performance of fault diagnosis of the test present invention.Obtained after sampling
Output voltage signal passes through FFT into frequency domain form.
(1) model training
FFT is carried out to data first, trained and test sample, because the first half cycle after FFT is with after
The frequency spectrum of half period is identical, therefore only chooses the data of first half cycle, i.e. training sample Ttrain∈ (270*400), ltrain∈
(9*270), test sample Ttest∈ (180*400), ltest∈(9*180).Feature extraction is carried out using PCA, CL takes 85%, obtained
Sample after to dimensionality reduction is Here
MRVM models use RBF kernel functions, and by repeatedly training, it is σ=0.23 to take wherein optimal core ginseng.Table 1 divides for fault type
And tag along sort is set.
The fault type of table 1 is divided and tag along sort is set
| Sequence number | Fault category | Tag along sort |
| 1 | Normally | [1,0,0,0,0,0,0,0,0]T |
| 2 | H1S1 opens a way | [0,1,0,0,0,0,0,0,0]T |
| 3 | H1S2 opens a way | [0,0,1,0,0,0,0,0,0]T |
| 4 | H1S3 opens a way | [0,0,0,1,0,0,0,0,0]T |
| 5 | H1S4 opens a way | [0,0,0,0,1,0,0,0,0]T |
| 6 | H2S1 opens a way | [0,0,0,0,0,1,0,0,0]T |
| 7 | H2S2 opens a way | [0,0,0,0,0,0,1,0,0]T |
| 8 | H2S3 opens a way | [0,0,0,0,0,0,0,1,0]T |
| 9 | H2S4 opens a way | [0,0,0,0,0,0,0,0,1]T |
(2) fault diagnosis platform test
Above-mentioned test data is brought into the model trained, obtains model probability output result, takes one of which Q (such as
Shown in lower).When row vector in Q represents corresponding test data as inputting with column vector, the state of inversion system
Opened for normal, H1S1 open circuits, H1S2 open circuits, H1S3 open circuits, H1S4 open circuits, H2S1 open circuits, H2S2 open circuits, H2S3 open circuits and H2S4
The target output value on road, it specifically can refer to the division in table 1.With test sample class label ltestAfter comparing, it can not only obtain
It is 100% to the Model Diagnosis accuracy rate, and by comparing the nonzero value exported in Q, it can be deduced that the situation belongs to other
The possibility of classification, laid a good foundation for further analysis uncertain problem.
General principle, principal character and the advantages of the present invention of the present invention has been shown and described above.The technology of the industry
Personnel are it should be appreciated that the present invention is not limited to the above embodiments, and the simply explanation described in above-described embodiment and specification is originally
The principle of invention, various changes and modifications of the present invention are possible without departing from the spirit and scope of the present invention, these changes
Change and improvement all fall within the protetion scope of the claimed invention.The claimed scope of the invention by appended claims and its
Equivalent defines.
Claims (1)
1. a kind of convergence strategy based on pivot analysis with the multi-electrical level inverter fault diagnosis of more classification Method Using Relevance Vector Machines, including
Following steps：
(1) data set of multi-electrical level inverter switch element characteristic signal is established：
(1a) builds a multi-level inverter circuit system model using Matlab/Simulink, in order to illustrate to need, below with
Single-phase cascaded H-bridges five-electrical level inverter is to illustrate object；
(1b) is controlled using SHPWM, frequency of modulated wave 50Hz, carrier frequency 4kHz, sample frequency 40kHz；
(1c) and then failsafe link is set, successively for the single switching transistor open fault of H bridge multi-electrical level inverter diverse locations
Test, the output voltage signal of the every kind of failure a cycle of inverter is extracted, establish the data set of fault-signal；
(2) fault signature extraction is carried out to the data set of fault-signal：
The characteristic signal of time domain is transformed into frequency domain using the method for Fast Fourier Transform (FFT) and carries out harmonic amplitude analysis by (2a),
Each mode signal extraction result after FFT；
(2b) utilizes the frequency domain harmonic amplitude Characteristics that FFT is extracted；
(3) classification diagnosis is carried out using mRVM
(3a) is by pivot sample XpIt is divided into two parts, respectively training sample XtrainWith test sample Xtest, set gradually classification mark
L ∈ { 1,2 ... C } are signed, C is failure sum；
(3b) selectes kernel function：Radial basis kernel function is chosen, wherein σ is the high wide parameter of kernel function,Adjust the parameter of kernel function；
The structure of (3c) mRVM models
Introduce auxiliary regression target Y ∈ RC×NWith weight parameter W ∈ RN×C, obtain standard noise regression model：
Multinomial probability Copula is introduced, regressive object is converted into class label：
Resulting multinomial probability likelihood function is as follows, wherein u~N (0,1):
It is openness in order to ensure model, zero-mean is introduced for weight vectors, variance isStandard normal prior distribution, by priori
Parameter alphancThe matrix of composition is designated as A ∈ RN×C,αncObedience hyper parameter is τ, and υ Gamma is distributed；
Posterior probability, wherein A are derived according to Bayesian modelcFor as A c row derived from diagonal matrix,
Obtaining method by maximum a-posteriori estimation can obtain：
The update method such as following formula based on maximum a posteriori (MAP) of weight when therefore giving classification：
According to above formula, the E-step forms of auxiliary variable can be derived, for i classes such as following formula：
ForSuch as following formula：
The Posterior probability distribution of weight vectors Study first such as following formula：
Gamma distribution average be：
(3d) m-RVM model measurements
Degree of rarefication is obtained by constructing an empty model and sample being added or removed into it, it is foundation to add or remove sample
Their contribution margins to model, using quick type-II maximum likelihoods (Fast Type-II ML) parameter updating method；
Marginal likelihood function is P (Y | K, A)=log ∫ P (Y | K, W) P (W | A) dW, is asked for by logarithmic, wherein
It can be decomposed intoWhereinWhen representing i-th of sample of removalValue, can derive
Go out：
Marginal likelihood function is further decomposed such as following formula：
WhereinFor the sparse factor,For quantizing factor, pass through solutionIt must can stay
Point：
Contribution factor θ is defined to the information contribution amount of model for the ease of i-th of sample of descriptioni：
If often step iterative process has active set Ο, it is made up of M dimension active samples, if some sample θ values are just, but
It is not included in Ο, this sample is just added in Ο by we；, whereas if some sample is in Ο, but θ values
It is negative, then removes the sample；Therefore in the learning process of model, MAP update method formula amendments are as follows, wherein K*∈RM ×N,A*∈RM×M, M ＜ ＜ N：
By above-mentioned stationary point and giveNow provide αiInitialization formula it is as follows：
The test of (3e) sample
First by test sample (Ttest,ltest) after processing by step (3c) and step (3d), bring the mRVM moulds trained into
In type, the probability output result of each sample is obtained, and using the sample of wherein maximum probability as classification diagnosis result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410005022.7A CN103761372B (en) | 2014-01-06 | 2014-01-06 | A kind of multi-electrical level inverter Fault Diagnosis Strategy based on pivot analysis with more classification Method Using Relevance Vector Machines |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410005022.7A CN103761372B (en) | 2014-01-06 | 2014-01-06 | A kind of multi-electrical level inverter Fault Diagnosis Strategy based on pivot analysis with more classification Method Using Relevance Vector Machines |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN103761372A CN103761372A (en) | 2014-04-30 |
| CN103761372B true CN103761372B (en) | 2018-02-02 |
Family
ID=50528608
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410005022.7A Active CN103761372B (en) | 2014-01-06 | 2014-01-06 | A kind of multi-electrical level inverter Fault Diagnosis Strategy based on pivot analysis with more classification Method Using Relevance Vector Machines |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN103761372B (en) |
Families Citing this family (24)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104698397B (en) * | 2015-03-16 | 2018-02-23 | 浙江万里学院 | A kind of multi-electrical level inverter method for diagnosing faults |
| CN104849689B (en) * | 2015-05-20 | 2017-11-28 | 国网河南省电力公司电力科学研究院 | A kind of cascade connection multi-level current transformer vertical compression faulty sensor method for early warning |
| CN105044624B (en) * | 2015-08-11 | 2018-02-23 | 上海海事大学 | Seven electrical level inverters and method for diagnosing faults of tape jam diagnostic function |
| CN105589037A (en) * | 2016-03-16 | 2016-05-18 | 合肥工业大学 | Ensemble learning-based electric power electronic switch device network fault diagnosis method |
| CN105811794B (en) * | 2016-05-06 | 2018-03-30 | 上海海事大学 | The fault tolerant control method of the reference voltage signal reconstruct of multi-electrical level inverter |
| JP6293388B1 (en) * | 2016-06-21 | 2018-03-14 | 三菱電機株式会社 | Load abnormality detection device |
| CN106291162A (en) * | 2016-07-20 | 2017-01-04 | 江南大学 | A kind of method for diagnosing faults of photovoltaic diode clamp formula three-level inverter |
| CN106682303B (en) * | 2016-12-26 | 2019-09-13 | 江南大学 | A kind of three-level inverter method for diagnosing faults based on empirical mode decomposition and decision tree RVM |
| CN106651197A (en) * | 2016-12-29 | 2017-05-10 | 中国民航大学 | Civil aviation industry effective competition evaluation method based on SCP (Structure-Conduct-Performance) normal form and kernel principal component analysis |
| CN107704863B (en) * | 2017-05-22 | 2021-06-15 | 上海海事大学 | PCA principal component rearrangement fault feature representation method |
| CN107256546B (en) * | 2017-05-23 | 2020-09-15 | 上海海事大学 | Ocean current machine blade attachment fault diagnosis method based on PCA convolution pooling SOFTMAX |
| CN107271925B (en) * | 2017-06-26 | 2019-11-05 | 湘潭大学 | Five level converter Fault Locating Method of modularization based on depth convolutional network |
| CN107505133B (en) * | 2017-08-10 | 2019-05-28 | 滁州学院 | The probability intelligent diagnosing method of rolling bearing fault based on adaptive M RVM |
| CN108020781B (en) * | 2017-12-19 | 2020-05-29 | 上海电机学院 | Breaker fault diagnosis method |
| CN108628185A (en) * | 2018-06-26 | 2018-10-09 | 上海海事大学 | Five-electrical level inverter fault diagnosis and fault-tolerant control method and semi-physical emulation platform |
| CN111400850B (en) * | 2018-12-28 | 2024-04-09 | 阿里巴巴集团控股有限公司 | Equipment fault analysis method, device, equipment and storage medium |
| CN109795713B (en) * | 2019-02-14 | 2020-08-18 | 四川大学 | Fault diagnosis method for aileron actuator based on Simulink model |
| CN109884449B (en) * | 2019-02-26 | 2021-07-16 | 广东省智能机器人研究院 | Real-time detection method for open-circuit fault of three-phase inverter bridge arm of motor driving system |
| CN110018411B (en) * | 2019-05-14 | 2021-06-15 | 集美大学 | Inverter circuit fault diagnosis method based on symmetric component analysis |
| CN111091094B (en) * | 2019-12-18 | 2023-02-24 | 上海海事大学 | Fault diagnosis method of photovoltaic cascade grid-connected inverter based on PCA-SVM secondary classification |
| CN113075469A (en) * | 2020-01-06 | 2021-07-06 | 株洲中车时代电气股份有限公司 | Inversion overcurrent fault diagnosis method, device and system |
| CN112803484B (en) * | 2021-02-26 | 2023-08-11 | 上海海事大学 | Multi-level grid-connected inverter fault-tolerant control method based on mean current compensation model |
| CN113176468B (en) * | 2021-06-10 | 2022-09-13 | 阳光电源股份有限公司 | Inverter detection method and application device thereof |
| CN115166517B (en) * | 2022-09-07 | 2022-11-25 | 沈阳科来沃电气技术有限公司 | Inverter fault diagnosis method for motor transmission system |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2013079188A1 (en) * | 2011-11-28 | 2013-06-06 | Ipsogen | Methods for the diagnosis, the determination of the grade of a solid tumor and the prognosis of a subject suffering from cancer |
-
2014
- 2014-01-06 CN CN201410005022.7A patent/CN103761372B/en active Active
Non-Patent Citations (5)
| Title |
|---|
| 利用主元方法进行传感器故障检测的行为分析;黄孝彬 等;《传感技术学报》;20031231(第4期);第419-423页 * |
| 基于PCA-SVM模型的多电平逆变系统故障诊断;刘远 等;《电力系统保护与控制》;20130201;第41卷(第3期);第66-72页摘要，第1-4节，图1-6，表1-4 * |
| 基于多分类相关向量机的变压器故障诊断新方法;尹金良 等;《电力系统保护与控制》;20130301;第41卷(第5期);第77-82页第1-3节，图1-图2 * |
| 基于相对主元分析的动态数据窗口故障检测方法;王天真 等;《电工技术学报》;20130131;第28卷(第1期);第142-148页第2节 * |
| 相对主元分析方法及其在故障检测中的应用;王天真 等;《系统仿真学报》;20070731;第19卷(第13期);第2889-2894页第2.2节 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN103761372A (en) | 2014-04-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN103761372B (en) | A kind of multi-electrical level inverter Fault Diagnosis Strategy based on pivot analysis with more classification Method Using Relevance Vector Machines | |
| NL2020015B1 (en) | Fault diagnosis method of three-level inverter based on empirical mode decomposition and decision tree RVM | |
| CN106443297B (en) | The decision tree SVM method for diagnosing faults of photovoltaic diode Clamp three-level inverter | |
| Khomfoi et al. | Fault diagnostic system for a multilevel inverter using a neural network | |
| US20220198244A1 (en) | Method for diagnosing open-circuit fault of switching transistor of single-phase half-bridge five-level inverter | |
| CN107632258A (en) | A kind of fan converter method for diagnosing faults based on wavelet transformation and DBN | |
| CN109782091A (en) | Multi-electrical level inverter method for diagnosing faults based on depth small echo extreme learning machine | |
| CN110456217B (en) | MMC fault positioning method based on WPD-FOA-LSSVM dual model | |
| CN106093678B (en) | A method of quick and precisely diagnosing flexible HVDC transmission system converter fault | |
| CN106291162A (en) | A kind of method for diagnosing faults of photovoltaic diode clamp formula three-level inverter | |
| CN113125992B (en) | NPC three-level inverter fault diagnosis method and system based on DBN | |
| CN113159119A (en) | Fault diagnosis method for medium-voltage modular multilevel converter | |
| CN103267941A (en) | Method for testing fault modes of integrated switching current circuit | |
| CN101782625A (en) | Power electronic system fault diagnostic method based on Gradation-boosting algorithm | |
| Kou et al. | Data‐driven design of fault diagnosis for three‐phase PWM rectifier using random forests technique with transient synthetic features | |
| CN113341345A (en) | MMC switch tube open-circuit fault diagnosis method based on feature extraction and random forest | |
| CN106019173A (en) | Real-time fault detection method applied to voltage source converter | |
| Xia et al. | A learning-based method for speed sensor fault diagnosis of induction motor drive systems | |
| Malik et al. | Transfer learning-based novel fault classification technique for grid-connected PV inverter | |
| Cui et al. | T-type inverter fault diagnosis based on GASF and improved AlexNet | |
| CN109782158A (en) | A kind of Analog circuit diagnosis method based on multiclass classification | |
| Deng et al. | A High-Accuracy-Light-AI Data-Driven Diagnosis Method for Open-Circuit Faults in Single-Phase PWM Rectifiers | |
| CN116566061A (en) | Grid-connected inverter system stability on-line monitoring method and system | |
| Zheng et al. | Fault diagnosis of photovoltaic inverters using hidden Markov model | |
| Zhang | Fault diagnosis for PWM inverter based on LMD |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |