CN106443447B - A kind of aerogenerator fault signature extracting method based on iSDAE - Google Patents
A kind of aerogenerator fault signature extracting method based on iSDAE Download PDFInfo
- Publication number
- CN106443447B CN106443447B CN201610871333.0A CN201610871333A CN106443447B CN 106443447 B CN106443447 B CN 106443447B CN 201610871333 A CN201610871333 A CN 201610871333A CN 106443447 B CN106443447 B CN 106443447B
- Authority
- CN
- China
- Prior art keywords
- isdae
- training
- aerogenerator
- data
- fault
- 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
- 238000000034 method Methods 0.000 title claims abstract description 27
- 230000009467 reduction Effects 0.000 claims abstract description 35
- 238000012549 training Methods 0.000 claims abstract description 28
- 241000255581 Drosophila <fruit fly, genus> Species 0.000 claims description 21
- 238000004804 winding Methods 0.000 claims description 7
- 239000000796 flavoring agent Substances 0.000 claims description 6
- 235000019634 flavors Nutrition 0.000 claims description 6
- 230000006870 function Effects 0.000 claims description 6
- 238000004088 simulation Methods 0.000 claims description 6
- 230000008859 change Effects 0.000 claims description 3
- 235000013305 food Nutrition 0.000 claims description 3
- 238000012804 iterative process Methods 0.000 claims description 3
- 239000000654 additive Substances 0.000 claims description 2
- 230000000996 additive effect Effects 0.000 claims description 2
- 238000003745 diagnosis Methods 0.000 abstract description 14
- 238000005516 engineering process Methods 0.000 abstract description 3
- 210000002569 neuron Anatomy 0.000 description 6
- 238000005457 optimization Methods 0.000 description 5
- 238000000605 extraction Methods 0.000 description 3
- 238000012544 monitoring process Methods 0.000 description 3
- 238000010606 normalization Methods 0.000 description 3
- 230000001360 synchronised effect Effects 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 230000006872 improvement Effects 0.000 description 2
- 230000008569 process Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 208000037656 Respiratory Sounds Diseases 0.000 description 1
- 241000255588 Tephritidae Species 0.000 description 1
- 230000004913 activation Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000004891 communication Methods 0.000 description 1
- 238000005260 corrosion Methods 0.000 description 1
- 230000007797 corrosion Effects 0.000 description 1
- 238000002405 diagnostic procedure Methods 0.000 description 1
- 230000005611 electricity Effects 0.000 description 1
- 230000005284 excitation Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
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
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Testing And Monitoring For Control Systems (AREA)
- Complex Calculations (AREA)
Abstract
The invention discloses one kind to be based on iSDAE (improved Stacked Denoising Auto-Encoders, improve storehouse noise reduction autocoder) aerogenerator fault signature extracting method, mainly solve the problems, such as that existing fault diagnosis technology is caused rate of correct diagnosis not high by the artificial limitation for extracting feature.Of the invention the specific implementation steps are as follows: (1) accident analysis；(2) data acquire；(3) data prediction；(4) training improves storehouse noise reduction autocoder；(5) feature exports.This method can automatically learning data feature, the distributed nature for obtaining initial data indicates, and has certain noise resisting ability, has good robustness, effectively improves aerogenerator fault diagnosis accuracy.
Description
Technical field
The present invention relates to one kind, and based on iSDAE, (improved Stacked Denoising Autoencoders is improved
Storehouse noise reduction autocoder) aerogenerator fault signature extracting method, belong to generator state monitoring and fault diagnosis
Technical field.
Background technique
Aerogenerator is the important component of main aircraft power source, it be responsible for instrument on aircraft, instrument, radar,
Various control systems etc. provide power supply on illumination, radio communication and machine.Any one of aerogenerator link breaks down, no
It only will affect its normal operation, while being possible to will lead to aircraft and being unable to normal flight, even will cause great aviation when serious
Accident.Thus carry out the research of aerogenerator fault diagnosis technology, the potential faults being likely to occur to aerogenerator in a deep going way
It makes in time, accurately and rapidly judge, it is ensured that the safe operation of aircraft has extremely important realistic meaning and huge
Economic benefit.
What is be now widely used aboard is rotating-rectifier type three-stage brushless synchronous generator, it is mainly by secondary excitation
Three machine, AC exciter and main generator parts form, wherein pilot exciter is rotary magnetic pole type permanent magnet generator, exchange
Exciter is revolving-armature type synchronous generator, and main generator is rotary pole formula synchronous generator.The structure of aerogenerator
Sufficiently complex, failure mode is various, by carrying out fault mode, influence and HAZAN to aerogenerator, determines aviation
The main fault mode of generator has faults in rotating rectifiers, machines under rotor winding faults, stator winding faults, shaft failure and axis
Hold failure etc., wherein each fault mode can be divided into different fault types again, for example faults in rotating rectifiers can be divided into again
Single tube failure, two-tube failure etc., bearing fault can be divided into spot corrosion, crackle etc. again.
In fault diagnosis field, there is method based on model, based at signal for the main diagnostic method of these failures
The method of reason and method based on artificial intelligence, currently used common practice is to combine signal processing with artificial intelligence,
Particularly, fault-signal is typically first acquired, signal processing then is carried out to collected fault-signal, it is artificial to extract event
Hinder feature, is finally used to the fault signature of extraction to train classifier, to carry out failure modes.But current feature extraction
Method is generally depended on and is manually extracted, and time and effort consuming is affected by noise jamming, and does not have universality, for
By the artificial problem for extracting feature and being limited and causing rate of correct diagnosis not high in existing fault diagnosis technology, the invention proposes
A kind of aerogenerator fault signature extracting method based on iSDAE, this method can automatically carry out data characteristics study, obtain
Distributed nature to initial data indicates, and has certain noise resisting ability, has good robustness, effectively improves
Aerogenerator fault diagnosis accuracy.
Summary of the invention
The aerogenerator fault signature extracting method based on iSDAE that the invention proposes a kind of, this method is applied to
Generator state monitoring and fault diagnosis field can automatically carry out data characteristics study, obtain the distribution of initial data
Character representation, and have certain noise resisting ability, there is good robustness, effectively improve aerogenerator fault diagnosis
Accuracy.
The present invention to achieve the above object, adopts the following technical scheme that
A kind of aviation alternator rotating rectifier on-line fault diagnosis method based on iSDAE, includes the following steps:
(1) accident analysis.Fault mode, influence and HAZAN are carried out on aerogenerator, determine aerogenerator
Chife failure models and required acquisition diagnostic signal.Through analyzing, aerogenerator mainly has faults in rotating rectifiers, turns
The fault modes such as sub- winding failure, stator winding faults, shaft and bearing fault, the diagnostic signal that need to be acquired are that main generator is defeated
Voltage signal, AC exciter exciting current signal, fuselage shaking signal and shaft torsion signal out.
(2) data acquire.Fault simulation experiment is carried out on generator failure simulation experiment platform, by institute in step (1)
The diagnostic signal of four kinds of need acquisition is stated, respectively through voltage sensor, current sensor, vibrating sensor and torque sensor,
Computer, which is connected to, with data collecting card again carries out data acquisition.
(3) data prediction.Since the dimension of four kinds of diagnostic signals collected in step (2) is different, in order to make signal
With unified statistical distribution, four kinds of signals are normalized, then by four in the case of each fault type
Kind signal forms column vector, generates sample.
(4) training improves storehouse noise reduction autocoder.Sample obtained in above-mentioned steps (3) is passed through into unsupervised side
Formula trains storehouse noise reduction autocoder, and the distributed nature for learning initial data indicates.
(4.1) series of noise reduction autocoder needed for setting simultaneously carries out noise reduction autocoder weight and biasing initially
Change.Noise reduction autocoder is a kind of neural networks with single hidden layer, since traditional noise reduction autocoder uses random initializtion
The method of network weight and biasing is impacted to entire autocoder performance, and the present invention uses drosophila optimization algorithm
(Fruit Fly Optimization Algorithm, FFOA) first encodes network weight and biasing, and search obtains one
A more excellent solution, is then trained this network parameter as the initial parameter of autocoder again, finally trains optimal
Network parameter.
(4.1-a) initialization.Noise reduction autocoder weight and biasing are encoded, determine the rule of initial drosophila population
Mould, maximum number of iterations, and the initial position of drosophila population is initialized.
(4.1-b) smell random search.The primary iteration number g=0 of drosophila algorithm is enabled, drosophila in iterative process is set
Body is looked for food the random flight direction rand () and arbitrary width in stage in smell.
(4.1-c) determines flavor concentration decision content, and calculates the odorousness value of drosophila individual, at this point, by reality output
Error E between value and exact value is as taste decision function.
(4.1-d) vision positioning.Seek the minimum individual of odorousness (i.e. error E) as optimum individual, and records this
When individual position and flavor concentration, at the same time, entire drosophila group is flown to using sharp vision to optimal location.
(4.1-e) iteration optimizing.Judge whether to reach termination condition, i.e. whether the number of iterations reaches maximum number of iterations.
Terminate algorithm if meeting, if being unsatisfactory for continuing to repeat step (4.1-b) to step (4.1-e), recycles the process.Until repeatedly
When generation number reaches maximum number of iterations, terminate algorithm.
(4.2) noise reduction autocoder training.Training sample is inputted, and additive Gaussian noise is manually added in the sample,
Make autocoder that there is certain noise resisting ability, calculates autocoder output.Since autocoder is using unsupervised
Training method reconstructed error is found out with output according to input it is expected that obtaining and inputting identical output, constantly adjust weight and
Biasing, so that reconstructed error is minimum.
(4.3) complete level-one noise reduction autocoder training after, save weight and the biasing of coded portion, at this time noise reduction from
The hidden layer output of dynamic encoder is the level one data feature learnt, and certainly using this data characteristics as next stage noise reduction
The training sample of dynamic encoder, repeats step (4.1) to step (4.3), until completing the noise reduction autocoder of setting number
Training, as improvement storehouse noise reduction autocoder.
(5) feature exports.Above-mentioned steps complete the distributed nature study of initial data, remain the spy of initial data
The feature learnt can be input to classifier and carry out failure modes by reference breath.
The present invention has the beneficial effect that:
The aerogenerator fault signature extracting method based on iSDAE that the invention proposes a kind of, this method is applied to
Generator state monitoring and fault diagnosis field can automatically carry out data characteristics study, obtain the distributed nature of data
Indicate, and have certain noise resisting ability that there is good robustness, it is correct to effectively improve aerogenerator fault diagnosis
Rate.
Detailed description of the invention
Fig. 1 feature extraction flow chart
Fig. 2 iSDAE network structure
Specific embodiment
Technical solution of the present invention is described in detail with reference to the accompanying drawing.As shown in Figure 1, a kind of based on iSDAE's
Aerogenerator fault signature extracting method specific embodiment is as follows:
(1) accident analysis.Fault mode, influence and HAZAN are carried out on aerogenerator, determine aerogenerator
Fault mode and required acquisition diagnostic signal.Through analyzing, aerogenerator mainly have faults in rotating rectifiers, rotor around
The fault modes such as group failure, stator winding faults, shaft and bearing fault, the diagnostic signal that need to be acquired are main generator output electricity
Press signal, AC exciter exciting current signal, fuselage shaking signal and shaft torsion signal.
(2) data acquire.Fault simulation experiment is carried out on generator failure simulation experiment platform, by institute in step (1)
The diagnostic signal of four kinds of need acquisition is stated, respectively through voltage sensor, current sensor, vibrating sensor and torque sensor,
Computer, which is connected to, with data collecting card again carries out data acquisition.
(3) data prediction.Since the dimension of four kinds of diagnostic signals collected in step (2) is different, in order to make signal
With unified statistical distribution, four kinds of signals are normalized, specific normalization formula is as follows:
Wherein: XnewFor the fault-signal after normalization, X is the fault-signal before normalization, XmeanFor sample average, Xstd
For the standard deviation of sample.
Then continuous 200 points of four kinds of signals in the case of each fault type are formed into 1 column vector yn, each
Column vector forms a sample set Y={ y as a sample, by all samples1, y2, y3...yn...yN, total N number of sample.
(4) training improves storehouse noise reduction autocoder.Sample obtained in above-mentioned steps (3) is passed through into unsupervised side
Formula trains storehouse noise reduction autocoder, and the distributed nature for learning initial data indicates.As shown in Fig. 2, specific training step
It is as follows:
(4.1) series of noise reduction autocoder needed for setting simultaneously carries out noise reduction autocoder weight and biasing initially
Change.The present invention sets the series of noise reduction autocoder as 4 grades, and the neuron number of first order noise reduction autocoder is 800-
600-800, the neuron number of the second level are 600-400-600, and the neuron number of the third level is 400-200-400, the 4th
The neuron number of grade is 200-100-200, and the feature finally learnt is the hidden layer of afterbody noise reduction autocoder
Output vector, used neuron activation functions are sigmoid function, and formula is as follows:
Wherein, e is natural constant.
Noise reduction autocoder is a kind of neural networks with single hidden layer, since traditional noise reduction autocoder is using random first
The method of beginningization network weight and biasing is impacted to entire autocoder performance, and the present invention uses drosophila optimization algorithm
First network weight and biasing are encoded, search obtains a more excellent solution, then using this network parameter as autocoder
Initial parameter be trained again, finally train optimal network parameter.Specific drosophila optimization algorithm optimization weight and threshold
Steps are as follows for value:
(4.1-a) initialization.Noise reduction autocoder weight and biasing are encoded, determine the rule of initial drosophila population
Mould, maximum number of iterations, and the initial position of drosophila population is initialized.
(4.1-b) smell random search.The primary iteration number g=0 of drosophila algorithm is enabled, drosophila in iterative process is set
Body is looked for food the random flight direction rand () and arbitrary width in stage in smell.
(4.1-c) determines flavor concentration decision content, and calculates the odorousness value of drosophila individual, at this point, by reality output
Error E between value and exact value is as taste decision function.
(4.1-d) vision positioning.Seek the minimum individual of odorousness (i.e. error E) as optimum individual, and records this
When individual position and flavor concentration, at the same time, entire drosophila group is flown to using sharp vision to optimal location.
(4.1-e) iteration optimizing.Judge whether to reach termination condition, i.e. whether the number of iterations reaches maximum number of iterations.
Terminate algorithm if meeting, if being unsatisfactory for continuing to repeat step (4.1-b) to step (4.1-e), recycles the process.Until repeatedly
When generation number reaches maximum number of iterations, terminate algorithm.
(4.2) noise reduction autocoder training.Training sample is inputted, and Gaussian noise is manually added in the sample, is made certainly
Dynamic encoder has certain noise resisting ability, calculates autocoder output.Since autocoder uses unsupervised instruction
The mode of white silk finds out reconstructed error with output according to input it is expected that obtaining and inputting identical output.Constantly adjustment weight and partially
It sets, so that reconstructed error is minimum.Specific step is as follows:
(4.2-a) and allowable error ε and learning rate α are set, then carries out DAE network training.Input N number of trained sample
This, calculates the output of noise reduction autocoder.
(4.2-b) is since noise reduction autocoder uses unsupervised training method, it is expected that obtaining identical defeated with input
Out, reconstructed error is found out with output according to input, reconstructed error formula is
Wherein, Y indicates training sample, hW, b(Y) output valve of the training sample through network query function is indicated.
(4.2-c) adjusts weight and biasing according to reconstructed error, specific formula is as follows:
Wherein, WijIndicate the network weight of j-th of neuron of the i-th layer network, biIndicate i-th layer of biasing,Indicate J (W, b) to WijLocal derviation is sought,Indicate J (W, b) to biLocal derviation is sought, l indicates iteration time
Number.
Whether (4.2-d) decision errors, which meet allowable error ε, requires or whether reaches the number of iterations, wants if be not able to satisfy
It asks and then repeats step (4.2-b) to step (4.2-d), it is expected to require or reach the number of iterations knot until whole network exports to meet
Shu Xunhuan.
(4.3) complete level-one noise reduction autocoder training after, save weight and the biasing of coded portion, at this time noise reduction from
The hidden layer output of dynamic encoder is the level one data feature learnt, and certainly using this data characteristics as next stage noise reduction
The training sample of dynamic encoder, repeats step (4.1) to step (4.3), until completing the noise reduction autocoder of setting number
Training, as improvement storehouse noise reduction autocoder.
(5) feature exports.Above-mentioned steps complete the distributed nature study of initial data, remain the spy of initial data
The feature learnt can be input to classifier and carry out failure modes by reference breath.
Claims (3)
1. a kind of aerogenerator fault signature extracting method based on iSDAE, which is characterized in that comprise the steps of:
Step 1: accident analysis carries out fault mode, influence and HAZAN to aerogenerator, determines aerogenerator
Chife failure models and required acquisition diagnostic signal, through analyzing, aerogenerator mainly have faults in rotating rectifiers, turn
Sub- winding failure, stator winding faults, shaft and bearing fault, the diagnostic signal that need to be acquired are main generator output voltage letter
Number, AC exciter exciting current signal, fuselage shaking signal and shaft torsion signal；
Step 2: data acquisition carries out fault simulation experiment, by four kinds in step 1 on generator failure simulation experiment platform
The diagnostic signal that need to be acquired, respectively through voltage sensor, current sensor, vibrating sensor and torque sensor, then
Computer, which is connected to, with data collecting card carries out data acquisition；
Step 3: data prediction, since the dimension of four kinds of diagnostic signals collected in step 2 is different, in order to have signal
There is unified statistical distribution, four kinds of diagnostic signals are normalized, it then will be in the case of each fault type
Four kinds of diagnostic signals form column vector, generate sample；
Step 4: training iSDAE, sample obtained in above-mentioned steps three is automatic by unsupervised mode training storehouse noise reduction
Encoder, the distributed nature for learning initial data indicate；
Step 5: feature output, above-mentioned steps four complete the distributed nature study of initial data, remain initial data
The feature learnt can be input to classifier and carry out failure modes by characteristic information.
2. a kind of aerogenerator fault signature extracting method based on iSDAE according to claim 1, feature exist
In the specific implementation steps are as follows by iSDAE described in step 4:
Step 2.1: the series of iSDAE needed for setting simultaneously carries out iSDAE weight and biasing initialization；
Step 2.2:iSDAE training, input sample are simultaneously manually added additive Gaussian noise in the sample, have iSDAE certain
Noise resisting ability, calculate iSDAE output, due to iSDAE use unsupervised training method, it is expected that obtain with input it is identical
Output finds out reconstructed error with output according to input, weight and biasing is constantly adjusted, so that reconstructed error is minimum；
Step 2.3: after completing level-one iSDAE training, saving weight and the biasing of coded portion, the hidden layer of iSDAE is defeated at this time
Out it is the level one data feature learnt, and using this data characteristics as the training sample of next stage iSDAE, repeats step
2.1 complete iSDAE training until completing the iSDAE training of setting series to step 2.3.
3. a kind of aerogenerator fault signature extracting method based on iSDAE according to claim 1, feature exist
In iSDAE training step described in step 4 is as follows:
Step 3.1: initialization encodes the weight and biasing of iSDAE, determines that scale, the maximum of initial drosophila population change
Generation number, and the initial position of drosophila population is initialized；
Step 3.2: smell random search, enabling the primary iteration number of drosophila algorithm is zero, sets drosophila individual in iterative process
Smell look for food the stage random flight direction and arbitrary width；
Step 3.3: determine flavor concentration decision content, and calculate the odorousness value of drosophila individual, at this point, by real output value with
Error between exact value is as taste decision function；
Step 3.4: vision positioning seeks the minimum individual of odorousness as optimum individual, and records position individual at this time
And flavor concentration, at the same time, entire drosophila group is flown to using sharp vision to optimal location；
Step 3.5: iteration optimizing judges whether to reach termination condition, i.e. whether the number of iterations reaches maximum number of iterations, if full
It is sufficient then terminate algorithm, continue to repeat step 3.2 if being unsatisfactory for step 3.5, until the number of iterations reaches maximum number of iterations
When, terminate algorithm.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610871333.0A CN106443447B (en) | 2016-09-26 | 2016-09-26 | A kind of aerogenerator fault signature extracting method based on iSDAE |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610871333.0A CN106443447B (en) | 2016-09-26 | 2016-09-26 | A kind of aerogenerator fault signature extracting method based on iSDAE |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106443447A CN106443447A (en) | 2017-02-22 |
| CN106443447B true CN106443447B (en) | 2019-05-21 |
Family
ID=58171685
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610871333.0A Active CN106443447B (en) | 2016-09-26 | 2016-09-26 | A kind of aerogenerator fault signature extracting method based on iSDAE |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106443447B (en) |
Families Citing this family (15)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107316046B (en) * | 2017-03-09 | 2020-08-25 | 河北工业大学 | Fault diagnosis method based on incremental compensation dynamic self-adaptive enhancement |
| CN107316087B (en) * | 2017-07-03 | 2021-08-17 | 中国航空工业集团公司西安飞机设计研究所 | Method for judging fault use of aviation product |
| CN107247231A (en) * | 2017-07-28 | 2017-10-13 | 南京航空航天大学 | A kind of aerogenerator fault signature extracting method based on OBLGWO DBN models |
| CN108241298A (en) * | 2018-01-09 | 2018-07-03 | 南京航空航天大学 | A kind of aerogenerator method for diagnosing faults based on FWA-RNN models |
| CN109190304B (en) * | 2018-10-16 | 2020-12-01 | 南京航空航天大学 | Method for extracting fault features and identifying faults of gas circuit components in all-enveloping line of aircraft engine |
| CN109598336A (en) * | 2018-12-05 | 2019-04-09 | 国网江西省电力有限公司信息通信分公司 | A kind of Data Reduction method encoding neural network certainly based on stack noise reduction |
| CN109613428A (en) * | 2018-12-12 | 2019-04-12 | 广州汇数信息科技有限公司 | It is a kind of can be as system and its application in motor device fault detection method |
| CN109858345B (en) * | 2018-12-25 | 2021-06-11 | 华中科技大学 | Intelligent fault diagnosis method suitable for pipe expansion equipment |
| CN109599872B (en) * | 2018-12-29 | 2022-11-08 | 重庆大学 | Power system probability load flow calculation method based on stack noise reduction automatic encoder |
| CN110018417B (en) * | 2019-05-24 | 2020-05-15 | 湖南大学 | Motor fault diagnosis method, system and medium based on radial stray magnetic flux detection |
| CN110261773B (en) * | 2019-07-01 | 2021-04-13 | 南京航空航天大学 | Aviation generator fault symptom extraction method and system |
| US20220155771A1 (en) * | 2020-10-22 | 2022-05-19 | Massachusetts Institute Of Technology | Modular, general purpose, automated, anomalous data synthesizers for rotary plants |
| CN112434298B (en) * | 2021-01-26 | 2021-07-06 | 浙江大学 | Network threat detection system based on self-encoder integration |
| CN113049035B (en) * | 2021-03-12 | 2022-05-27 | 辽宁工程技术大学 | Transformer state monitoring system based on Internet of things |
| CN113625164A (en) * | 2021-08-02 | 2021-11-09 | 南京航空航天大学 | Aviation generator fault feature extraction method, system, medium and computing device |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104748962A (en) * | 2015-04-03 | 2015-07-01 | 西安交通大学 | Planetary gear box intelligent diagnosis method based on stacking automatic encoding machine |
| CN105184368A (en) * | 2015-09-07 | 2015-12-23 | 中国科学院深圳先进技术研究院 | Distributed extreme learning machine optimization integrated framework system and method |
| CN105957092A (en) * | 2016-05-31 | 2016-09-21 | 福州大学 | Mammary gland molybdenum target image feature self-learning extraction method for computer-aided diagnosis |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US9367428B2 (en) * | 2013-10-14 | 2016-06-14 | Nec Corporation | Transparent performance inference of whole software layers and context-sensitive performance debugging |
-
2016
- 2016-09-26 CN CN201610871333.0A patent/CN106443447B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN104748962A (en) * | 2015-04-03 | 2015-07-01 | 西安交通大学 | Planetary gear box intelligent diagnosis method based on stacking automatic encoding machine |
| CN105184368A (en) * | 2015-09-07 | 2015-12-23 | 中国科学院深圳先进技术研究院 | Distributed extreme learning machine optimization integrated framework system and method |
| CN105957092A (en) * | 2016-05-31 | 2016-09-21 | 福州大学 | Mammary gland molybdenum target image feature self-learning extraction method for computer-aided diagnosis |
Non-Patent Citations (1)
| Title |
|---|
| 基于堆栈降噪自动编码模型的动态纹理分类方法;汪彩霞等;《现代电子技术》;20150315;第38卷(第6期);20-24 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106443447A (en) | 2017-02-22 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106443447B (en) | A kind of aerogenerator fault signature extracting method based on iSDAE | |
| CN109580215B (en) | Wind power transmission system fault diagnosis method based on deep generation countermeasure network | |
| Zhang et al. | The multiclass fault diagnosis of wind turbine bearing based on multisource signal fusion and deep learning generative model | |
| CN107941537B (en) | A kind of mechanical equipment health state evaluation method | |
| CN105354587B (en) | A kind of method for diagnosing faults of wind-driven generator group wheel box | |
| CN107909118A (en) | A kind of power distribution network operating mode recording sorting technique based on deep neural network | |
| CN106991666B (en) | A kind of disease geo-radar image recognition methods suitable for more size pictorial informations | |
| CN113641486B (en) | Intelligent turnout fault diagnosis method based on edge computing network architecture | |
| CN107392304A (en) | A kind of Wind turbines disorder data recognition method and device | |
| CN107247231A (en) | A kind of aerogenerator fault signature extracting method based on OBLGWO DBN models | |
| CN106979126A (en) | Wind power generating set high wind speed section effective wind speed method of estimation based on SVR | |
| CN111680875A (en) | Unmanned aerial vehicle state risk fuzzy comprehensive evaluation method based on probability baseline model | |
| CN105242205A (en) | Aviation three-level AC power generator rotary rectifier online fault diagnosis method | |
| CN108154223A (en) | Power distribution network operating mode recording sorting technique based on network topology and long timing information | |
| CN110866448A (en) | Flutter signal analysis method based on convolutional neural network and short-time Fourier transform | |
| CN107991876A (en) | Aero-engine condition monitoring data creation method based on production confrontation network | |
| CN102609764A (en) | CPN neural network-based fault diagnosis method for stream-turbine generator set | |
| CN109100648A (en) | Ocean current generator impeller based on CNN-ARMA-Softmax winds failure fusion diagnosis method | |
| CN110362045A (en) | A kind of marine double-fed fan motor unit fault distinguishing method considering maritime meteorology factor | |
| CN114021433A (en) | Construction method and application of dominant instability mode recognition model of power system | |
| CN111553112A (en) | Power system fault identification method and device based on deep belief network | |
| Fan et al. | Defect identification detection research for insulator of transmission lines based on deep learning | |
| CN115587290A (en) | Aero-engine fault diagnosis method based on variational self-coding generation countermeasure network | |
| CN111795819A (en) | Gear box fault diagnosis method integrating vibration and current signal collaborative learning | |
| CN110261773B (en) | Aviation generator fault symptom extraction method and system |
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 |