CN109766874A - A kind of fan trouble classifying identification method based on deep learning algorithm - Google Patents
A kind of fan trouble classifying identification method based on deep learning algorithm Download PDFInfo
- Publication number
- CN109766874A CN109766874A CN201910108062.7A CN201910108062A CN109766874A CN 109766874 A CN109766874 A CN 109766874A CN 201910108062 A CN201910108062 A CN 201910108062A CN 109766874 A CN109766874 A CN 109766874A
- Authority
- CN
- China
- Prior art keywords
- signal
- fault
- vibration signal
- group
- deep learning
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02E—REDUCTION OF GREENHOUSE GAS [GHG] EMISSIONS, RELATED TO ENERGY GENERATION, TRANSMISSION OR DISTRIBUTION
- Y02E40/00—Technologies for an efficient electrical power generation, transmission or distribution
- Y02E40/70—Smart grids as climate change mitigation technology in the energy generation sector
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Landscapes
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
- Control Of Positive-Displacement Air Blowers (AREA)
Abstract
The present invention provides a kind of fan trouble classifying identification method based on deep learning, is used for fan bearing fault diagnosis, acquires the vibration signal of Wind turbines bearing in real time first, and interception is the segment of specified duration T after being filtered；Then time frequency analysis is carried out to the signal segment of interception, and time frequency analysis result is normalized；Clustering is carried out to the time frequency analysis result after normalization followed by deep learning algorithm network, judges whether fan operation is normal based on the analysis results；Maintenance measure suggestion or shutdown inspection are finally exported according to fault type.This method carries out feature identification for the time frequency analysis result of vibration signal, more comprehensively efficiently utilizes the Time-Frequency Information of vibration signal, improves the recognition accuracy of neural network, to reduce the shutdown loss due to caused by failure.
Description
Technical field
The present invention relates to Fault Diagnosis of Fan method more particularly to a kind of fan trouble classification based on deep learning algorithm
Recognition methods.
Background technique
The progress of science and technology greatly have stimulated demand of the mankind to the energy, wind energy as not public hazards renewable energy by
To countries in the world extensive concern.Wind power generating set is typically mounted at offshore islands, the grass that wind resource is abundant but has inconvenient traffic
Former pasture, mountain area and highlands.The natural environment in these areas is more severe, is not suitable for the mankind and hangs up one's hat, and wind direction, wind
Speed is changeable, therefore blower is arranged must emphasis consideration maintenance and repair problem.Axis and bearing are the transmission parts of blower core, but
It is the multiple component of failure.These operation troubles mainly include that rotor misalignment failure, bearing imbalance fault, permanent shaft are curved
Bent failure, provisional shaft bending failure etc..Fan bearing fault diagnosis is carried out generally by bear vibration data at present
What analysis was realized.Conventional method is by carrying out time domain, frequency domain or time frequency analysis, selected part time domain, frequency domain to vibration signal
Characteristic value is as the index for judging whether to break down.
Artificial intelligence approach is important class in Fault Diagnosis of Fan, such as BP neural network, support vector machines and minimum two
Multiply vector machine etc..Patent 201610503720.9 discloses a kind of Fault Diagnosis of Fan method based on extreme learning machine.The party
Method, as sampling feature vectors, generates training set and test set using the time domain charactreristic parameter of fan bearing vibration signal.It is instructing
Practice and carry out the study of ELM disaggregated model on collection, the sample of test set is substituted into classifier and verifies its classification.Patent
201610374216.3 disclosing a kind of Fault Diagnosis of Fan method based on support vector machines.This method wavelet packet is to wind-powered electricity generation
Set gear box vibration signal carries out feature extraction, forms total training set and test set.It will be total using binary tree sort method
Training set and total test set split into 3 sub- training sets and 3 sub- test sets, and concentrate respectively to each sub- training set and son verifying
Classification be identified.Model parameter finally is determined using improvement PSO algorithm, establishes the fault diagnosis model of LS-SVM algorithm,
Test set is substituted into classifier, the correspondence classification of sample in each test set is obtained.Patent 201710142440.4 discloses one
Kind gear case of blower fault diagnosis model method for building up and device, this method first carry out vibration signal at smooth and noise reduction
Reason, then to treated, vibration signal is decomposed, and extracts the feature vector of vibration signal.Then by the feature vector of extraction
It is divided into training dataset and test data set, and is optimized using parameter of the drosophila algorithm to radial basis neural network,
Finally gear case of blower failure is diagnosed using radial base neural net.Above-mentioned artificial intelligence diagnosis' method generally merely with
The temporal signatures value or frequency domain character value of vibration signal, and training set and test are that data volume is limited, thus there are networks
The disadvantages such as training convergence is full, fault identification low efficiency and accuracy rate are low.
Summary of the invention
For the fault category for rapidly and accurately diagnosing Wind turbines bearing, a kind of wind based on deep learning algorithm is proposed
Machine failure modes recognition methods, has first carried out data volume expansion to original vibration signal, then using artificial neural network to vibration
The time frequency analysis result of dynamic signal is identified, accurately to determine bearing fault classification and alarm as early as possible and take corresponding dimension
Shield measure finally reduces downtime, improves fan operation efficiency.
The object of the present invention is achieved like this:
A kind of fan trouble classifying identification method based on deep learning algorithm, comprising the following steps:
Step a, the vibration signal of Wind turbines bearing is acquired in real time；
Step b, it is filtered to collecting vibration signal data, and intercepts as the segment of specified duration T；
Step c, time frequency analysis is carried out to the signal segment of interception；
Step d, time frequency analysis result is normalized；
Step e, cluster point is carried out to the time frequency analysis result after normalization using trained deep learning algorithm network
Analysis；
Step f, judge whether fan operation is normal according to cluster analysis result, if normal, then continue to acquire vibration data,
A is entered step, then export fault type if abnormal, and is alarmed；
Step g, output maintenance measure suggestion continues step a, otherwise shutdown inspection if not needing shutdown inspection.
The above-mentioned fan trouble classifying identification method based on deep learning algorithm, deep learning algorithm network in step e
Training step includes:
Step e1, vibration signal and the segmentation of Wind turbines bearing are acquired, vibration signal therein includes fan operation event
Signal when barrier and fan operation are normal, operation troubles type includes rotor misalignment failure, bearing imbalance fault, permanent
Shaft bending failure, provisional shaft bending failure；
Step e2, fault-signal is subjected to manual sort and marks fault type information；
Step e3, the fault-signal after label is expanded, increases the data volume of training network；
Step e4, the fault-signal after arranging primary fault signal and expanding, establishes fault-signal database；
Step e5, sample set is extracted from fault-signal database；
Step e6, the sample set training deep learning algorithm network of extraction is utilized；
Step e7, trained network model parameter is saved.
The step of step e3 expands the fault-signal after label include:
Step e3-1, the good vibration signal of K group echo is extracted, vibration signal therein includes fan operation failure and blower
Signal when normal operation, operation troubles type include rotor misalignment failure, bearing imbalance fault, permanent shaft bending
Failure, provisional shaft bending failure；
Step e3-2, N group white noise is added in K group vibration signal respectively, is extended for K*N group vibration signal, wherein N group is white
The power spectrum degree series of noise are specified arithmetic sequence；
Step e3-3, M group white Gaussian noise is added in K group vibration signal respectively, is extended for K*M group vibration signal, wherein M
The mean value and standard deviation of group white Gaussian noise are designated value；
Step e3-4, to the K*N+K*M group vibration signal after expansion plus fault type label and noise token；
Step e3-5, the K*N+K*M group vibration signal for expanding and marking is arranged, Signals Data Base is established and saves.
The above-mentioned fan trouble classifying identification method based on deep learning algorithm, the time frequency analysis side in the step c
Method is Short Time Fourier Transform or wavelet transformation or Wigner-Ville distribution method.
The above-mentioned fan trouble classifying identification method based on deep learning algorithm, the deep learning in the step e are calculated
Method network is convolutional neural networks or deepness belief network.
A kind of deep learning algorithm network training method towards fan trouble classifying identification method, comprising:
Step e1, vibration signal and the segmentation of Wind turbines bearing are acquired, vibration signal therein includes fan operation event
Signal when barrier and fan operation are normal, operation troubles type includes rotor misalignment failure, bearing imbalance fault, permanent
Shaft bending failure, provisional shaft bending failure；
Step e2, fault-signal is subjected to manual sort and marks fault type information；
Step e3, the fault-signal after label is expanded, increases the data volume of training network；
Step e4, the fault-signal after arranging primary fault signal and expanding, establishes fault-signal database；
Step e5, sample set is extracted from fault-signal database；
Step e6, the sample set training deep learning algorithm network of extraction is utilized；
Step e7, trained network model parameter is saved.
A kind of fault-signal progress extending method towards fan trouble classifying identification method, comprising:
Step e3-1, the good vibration signal of K group echo is extracted, vibration signal therein includes fan operation failure and blower
Signal when normal operation, operation troubles type include rotor misalignment failure, bearing imbalance fault, permanent shaft bending
Failure, provisional shaft bending failure；
Step e3-2, N group white noise is added in K group vibration signal respectively, is extended for K*N group vibration signal, wherein N group is white
The power spectrum degree series of noise are specified arithmetic sequence；
Step e3-3, M group white Gaussian noise is added in K group vibration signal respectively, is extended for K*M group vibration signal, wherein M
The mean value and standard deviation of group white Gaussian noise are designated value；
Step e3-4, to the K*N+K*M group vibration signal after expansion plus fault type label and noise token；
Step e3-5, the K*N+K*M group vibration signal for expanding and marking is arranged, Signals Data Base is established and saves.
Compared with prior art, with the beneficial effects of the present invention are:
The first, the present invention carries out Classification and Identification with time frequency analysis result of the deep learning algorithm to vibration signal, contains
The time domain and frequency domain information of vibration signal, more comprehensively, resolution is high for information；
The second, the present invention expands initial vibration signal, and white noise and Gauss white noise are added in original signal
Sound does not influence the frequency and amplitude of the vibration signal generated by failure, but increases the data volume of network training, improves network
Trained performance and speed is conducive to the accuracy and speed that improve fault vibration type identification.
Detailed description of the invention
Fig. 1 is a kind of flow chart of the fan trouble classifying identification method based on deep learning algorithm of the present invention.
Fig. 2 is the flow chart of deep neural network training method.
Fig. 3 is the flow chart of fault-signal data volume extending method.
Specific embodiment
Below in conjunction with the drawings and specific embodiments, the present invention will be described in further detail.
Specific embodiment one
The present embodiment is the fan trouble classifying identification method embodiment based on deep learning algorithm.
The fan trouble classifying identification method based on deep learning algorithm of the present embodiment, as shown in Figure 1, including following step
It is rapid:
Step a, the vibration signal of Wind turbines bearing is acquired in real time；
Step b, it is filtered to collecting vibration signal data, and intercepts as the segment of specified duration T；
Step c, time frequency analysis is carried out to the signal segment of interception；
Step d, time frequency analysis result is normalized；
Step e, cluster point is carried out to the time frequency analysis result after normalization using trained deep learning algorithm network
Analysis；
Step f, judge whether fan operation is normal according to cluster analysis result, if normal, then continue to acquire vibration data,
A is entered step, then export fault type if abnormal, and is alarmed；
Step g, output maintenance measure suggestion continues step a, otherwise shutdown inspection if not needing shutdown inspection.
Specific embodiment two
The present embodiment is the fan trouble classifying identification method embodiment based on deep learning algorithm.
The fan trouble classifying identification method based on deep learning algorithm of the present embodiment, on the basis of specific embodiment one
On, the training step of deep learning algorithm network in step e is further limited, as shown in Figure 2, comprising:
Step e1, vibration signal and the segmentation of Wind turbines bearing are acquired, vibration signal therein includes fan operation event
Signal when barrier and fan operation are normal, operation troubles type includes rotor misalignment failure, bearing imbalance fault, permanent
Shaft bending failure, provisional shaft bending failure；
Step e2, fault-signal is subjected to manual sort and marks fault type information；
Step e3, the fault-signal after label is expanded, increases the data volume of training network；
Step e4, the fault-signal after arranging primary fault signal and expanding, establishes fault-signal database；
Step e5, sample set is extracted from fault-signal database；
Step e6, the sample set training deep learning algorithm network of extraction is utilized；
Step e7, trained network model parameter is saved.
The step of step e3 expands the fault-signal after label include:
Specific embodiment three
The present embodiment is the fan trouble classifying identification method embodiment based on deep learning algorithm.
The fan trouble classifying identification method based on deep learning algorithm of the present embodiment, on the basis of specific embodiment two
On, step e3 is further limited, as shown in Figure 3, comprising:
Step e3-1, the good vibration signal of K group echo is extracted, vibration signal therein includes fan operation failure and blower
Signal when normal operation, operation troubles type include rotor misalignment failure, bearing imbalance fault, permanent shaft bending
Failure, provisional shaft bending failure；
Step e3-2, N group white noise is added in K group vibration signal respectively, is extended for K*N group vibration signal, wherein N group is white
The power spectrum degree series of noise are specified arithmetic sequence；
Step e3-3, M group white Gaussian noise is added in K group vibration signal respectively, is extended for K*M group vibration signal, wherein M
The mean value and standard deviation of group white Gaussian noise are designated value；
Step e3-4, to the K*N+K*M group vibration signal after expansion plus fault type label and noise token；
Step e3-5, the K*N+K*M group vibration signal for expanding and marking is arranged, Signals Data Base is established and saves.
Specific embodiment four
The present embodiment is the fan trouble classifying identification method embodiment based on deep learning algorithm.
The fan trouble classifying identification method based on deep learning algorithm of the present embodiment, on the basis of specific embodiment one
On, further limiting the Time-Frequency Analysis Method in step c is Short Time Fourier Transform or wavelet transformation or Wigner-Ville distribution
Method.
Specific embodiment five
The present embodiment is the fan trouble classifying identification method embodiment based on deep learning algorithm.
The fan trouble classifying identification method based on deep learning algorithm of the present embodiment, on the basis of specific embodiment one
On, further limiting the deep learning algorithm network in step e is convolutional neural networks or deepness belief network.
Specific embodiment six
The present embodiment is the deep learning algorithm network training method embodiment towards fan trouble classifying identification method.
The deep learning algorithm network training method towards fan trouble classifying identification method of the present embodiment, comprising:
Step e1, vibration signal and the segmentation of Wind turbines bearing are acquired, vibration signal therein includes fan operation event
Signal when barrier and fan operation are normal, operation troubles type includes rotor misalignment failure, bearing imbalance fault, permanent
Shaft bending failure, provisional shaft bending failure；
Step e2, fault-signal is subjected to manual sort and marks fault type information；
Step e3, the fault-signal after label is expanded, increases the data volume of training network；
Step e4, the fault-signal after arranging primary fault signal and expanding, establishes fault-signal database；
Step e5, sample set is extracted from fault-signal database；
Step e6, the sample set training deep learning algorithm network of extraction is utilized；
Step e7, trained network model parameter is saved.
Specific embodiment seven
The present embodiment is that the fault-signal towards fan trouble classifying identification method carries out extending method embodiment.
The fault-signal towards fan trouble classifying identification method of the present embodiment carries out extending method, comprising:
Step e3-1, the good vibration signal of K group echo is extracted, vibration signal therein includes fan operation failure and blower
Signal when normal operation, operation troubles type include rotor misalignment failure, bearing imbalance fault, permanent shaft bending
Failure, provisional shaft bending failure；
Step e3-2, N group white noise is added in K group vibration signal respectively, is extended for K*N group vibration signal, wherein N group is white
The power spectrum degree series of noise are specified arithmetic sequence；
Step e3-3, M group white Gaussian noise is added in K group vibration signal respectively, is extended for K*M group vibration signal, wherein M
The mean value and standard deviation of group white Gaussian noise are designated value；
Step e3-4, to the K*N+K*M group vibration signal after expansion plus fault type label and noise token；
Step e3-5, the K*N+K*M group vibration signal for expanding and marking is arranged, Signals Data Base is established and saves.
Specific embodiment eight
The present embodiment is the fan trouble classifying identification method embodiment based on deep learning algorithm.
The fan trouble classifying identification method based on deep learning algorithm of the present embodiment, specific steps process such as Fig. 1 institute
Show, bearing vibration signal is acquired and intercepted in real time first.The energy of fan bearing vibration signal concentrates on 1000Hz-
Between 6000Hz, therefore sample frequency is set to 20480Hz, when interception a length of 0.2s.Then to the vibration signal piece of each acquisition
Duan Jinhang time frequency analysis, and time frequency analysis result i.e. 4096 × 4096 matrix is normalized.Finally utilize depth
Learning algorithm network carries out clustering to the matrix after normalization, and judges whether fan operation is normal based on the analysis results.
If normal operation, continue to acquire vibration data, if discovery operation troubles, exports fault type and send a warning message.Root
Maintenance measure suggestion is exported according to fault type, according to suggesting judging whether to need shutdown inspection, then shutdown inspection if necessary,
Otherwise continue to acquire vibration data.
In the present embodiment, the training step of deep learning algorithm network includes:
The first step, acquires the vibration signal of Wind turbines bearing and segmentation, sample frequency 20480Hz, and when segmentation is a length of
0.2s.The signal when vibration signal of acquisition includes fan operation failure and normal fan operation, and operation troubles type packet
Include rotor misalignment failure, bearing imbalance fault, permanent shaft bending failure and provisional shaft bending failure；
Fault-signal is carried out manual sort and marks fault type information by second step, equally by normal operation when vibration
Dynamic signal is labeled as normal；
Third step expands the fault-signal after the second step mark, increases the data volume of training network；
4th step, the fault-signal after arranging primary fault signal and expanding, establishes fault-signal database；
5th step extracts sample set from fault-signal database, generally takes total data, if only diagnostics division is needed to classify
The failure of type can only extract the data for marking corresponding fault type；
6th step utilizes the sample set training deep learning algorithm network of extraction；
7th step saves trained network model parameter.
In the present embodiment, the step of expanding the fault-signal after label include:
The first step, extracts the good vibration signal of K group echo, and vibration signal therein includes fan operation failure and blower fortune
Signal when row is normal, operation troubles type include rotor misalignment failure, bearing imbalance fault, the event of permanent shaft bending
Hindering, provisional shaft bending failure, various types of signal proportional assignment, normal operation vibration signal quantity account for 50% in K group signal,
Rotor misalignment fault vibration number of signals accounts for 12.5%, and bearing imbalance fault vibration signal quantity accounts for 12.5%, permanently
Shaft bending fault vibration number of signals accounts for 12.5%, and provisional shaft complete failure vibration signal quantity accounts for 12.5%, K value
Greater than 200；
K group vibration signal is added N group white noise respectively, is extended for K*N group vibration signal, wherein N group white noise by second step
The power spectrum degree series of sound are specified arithmetic sequence, and the power spectral density value of maximum white noise is no more than power spectrum signal
Density maxima, i.e. white noise cannot flood signal completely, and N value is greater than 10；
K group vibration signal is added M group white Gaussian noise respectively, is extended for K*M group vibration signal, wherein M group by third step
The mean value and standard deviation of white Gaussian noise are designated value, and M value is greater than 10；
4th step, to the K*N+K*M group vibration signal after expansion plus fault type label and noise token；
5th step arranges the K*N+K*M group vibration signal for expanding and marking, establishes Signals Data Base and save.
In the present embodiment, Time-Frequency Analysis Method is Short Time Fourier Transform or wavelet transformation or Wigner-Ville distribution
Method.
In the present embodiment, deep learning algorithm network is convolutional neural networks or deepness belief network.
The foregoing is only a preferred embodiment of the present invention, all within the spirits and principles of the present invention, made
Any modifications, equivalent replacements, and improvements etc., should all be included in the protection scope of the present invention.
Claims (7)
1. a kind of fan trouble classifying identification method based on deep learning algorithm, which comprises the following steps:
Step a, the vibration signal of Wind turbines bearing is acquired in real time；
Step b, it is filtered to collecting vibration signal data, and intercepts as the segment of specified duration T；
Step c, time frequency analysis is carried out to the signal segment of interception；
Step d, time frequency analysis result is normalized；
Step e, clustering is carried out to the time frequency analysis result after normalization using trained deep learning algorithm network；
Step f, judge whether fan operation is normal according to cluster analysis result, if normal, then continue to acquire vibration data, enter
Step a then exports fault type if abnormal, and alarms；
Step g, output maintenance measure suggestion continues step a, otherwise shutdown inspection if not needing shutdown inspection.
2. the fan trouble classifying identification method according to claim 1 based on deep learning algorithm, which is characterized in that step
The training step of deep learning algorithm network includes: in rapid e
Step e1, acquire Wind turbines bearing vibration signal and segmentation, vibration signal therein include fan operation failure and
Signal when fan operation is normal, operation troubles type include rotor misalignment failure, bearing imbalance fault, permanent shaft
Flexural failure, provisional shaft bending failure；
Step e2, fault-signal is subjected to manual sort and marks fault type information；
Step e3, the fault-signal after label is expanded, increases the data volume of training network；
Step e4, the fault-signal after arranging primary fault signal and expanding, establishes fault-signal database；
Step e5, sample set is extracted from fault-signal database；
Step e6, the sample set training deep learning algorithm network of extraction is utilized；
Step e7, trained network model parameter is saved.
3. the fan trouble classifying identification method according to claim 2 based on deep learning algorithm, which is characterized in that step
The step of rapid e3 expands the fault-signal after label include:
Step e3-1, the good vibration signal of K group echo is extracted, vibration signal therein includes fan operation failure and fan operation
Signal when normal, operation troubles type include rotor misalignment failure, bearing imbalance fault, the event of permanent shaft bending
Barrier, provisional shaft bending failure；
Step e3-2, N group white noise is added in K group vibration signal respectively, is extended for K*N group vibration signal, wherein N group white noise
Power spectrum degree series be specified arithmetic sequence；
Step e3-3, M group white Gaussian noise is added in K group vibration signal respectively, is extended for K*M group vibration signal, wherein M group is high
The mean value and standard deviation of this white noise are designated value；
Step e3-4, to the K*N+K*M group vibration signal after expansion plus fault type label and noise token；
Step e3-5, the K*N+K*M group vibration signal for expanding and marking is arranged, Signals Data Base is established and saves.
4. the fan trouble classifying identification method according to claim 1 based on deep learning algorithm, which is characterized in that institute
The Time-Frequency Analysis Method in step c stated is Short Time Fourier Transform or wavelet transformation or Wigner-Ville distribution method.
5. the fan trouble classifying identification method according to claim 1 based on deep learning algorithm, which is characterized in that institute
Deep learning algorithm network in the step e stated is convolutional neural networks or deepness belief network.
6. a kind of deep learning algorithm network training method towards fan trouble classifying identification method characterized by comprising
Step e1, acquire Wind turbines bearing vibration signal and segmentation, vibration signal therein include fan operation failure and
Signal when fan operation is normal, operation troubles type include rotor misalignment failure, bearing imbalance fault, permanent shaft
Flexural failure, provisional shaft bending failure；
Step e2, fault-signal is subjected to manual sort and marks fault type information；
Step e3, the fault-signal after label is expanded, increases the data volume of training network；
Step e4, the fault-signal after arranging primary fault signal and expanding, establishes fault-signal database；
Step e5, sample set is extracted from fault-signal database；
Step e6, the sample set training deep learning algorithm network of extraction is utilized；
Step e7, trained network model parameter is saved.
7. a kind of fault-signal towards fan trouble classifying identification method carries out extending method characterized by comprising
Step e3-1, the good vibration signal of K group echo is extracted, vibration signal therein includes fan operation failure and fan operation
Signal when normal, operation troubles type include rotor misalignment failure, bearing imbalance fault, the event of permanent shaft bending
Barrier, provisional shaft bending failure；
Step e3-2, N group white noise is added in K group vibration signal respectively, is extended for K*N group vibration signal, wherein N group white noise
Power spectrum degree series be specified arithmetic sequence；
Step e3-3, M group white Gaussian noise is added in K group vibration signal respectively, is extended for K*M group vibration signal, wherein M group is high
The mean value and standard deviation of this white noise are designated value；
Step e3-4, to the K*N+K*M group vibration signal after expansion plus fault type label and noise token；
Step e3-5, the K*N+K*M group vibration signal for expanding and marking is arranged, Signals Data Base is established and saves.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910108062.7A CN109766874A (en) | 2019-02-02 | 2019-02-02 | A kind of fan trouble classifying identification method based on deep learning algorithm |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910108062.7A CN109766874A (en) | 2019-02-02 | 2019-02-02 | A kind of fan trouble classifying identification method based on deep learning algorithm |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109766874A true CN109766874A (en) | 2019-05-17 |
Family
ID=66456006
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910108062.7A Pending CN109766874A (en) | 2019-02-02 | 2019-02-02 | A kind of fan trouble classifying identification method based on deep learning algorithm |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109766874A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110276303A (en) * | 2019-06-25 | 2019-09-24 | 湖南科技大学 | Rotor misalignment quantitative identification method based on VMD and DBN |
| CN110399854A (en) * | 2019-07-31 | 2019-11-01 | 中南大学 | Rolling bearing fault classification method based on hybrid feature extraction |
| CN110953179A (en) * | 2020-02-26 | 2020-04-03 | 浙江上风高科专风实业有限公司 | Fan state monitoring method and system based on inverse wavelet transform |
| CN111120388A (en) * | 2020-03-31 | 2020-05-08 | 浙江上风高科专风实业有限公司 | Fan state combined monitoring method and system |
| CN111141517A (en) * | 2020-01-02 | 2020-05-12 | 上海电气风电集团股份有限公司 | Fan fault diagnosis method and system |
| CN112052796A (en) * | 2020-09-07 | 2020-12-08 | 电子科技大学 | Permanent magnet synchronous motor fault diagnosis method based on deep learning |
| CN113052716A (en) * | 2019-12-27 | 2021-06-29 | 新疆金风科技股份有限公司 | Abnormity early warning method and device for main bearing of wind generating set |
| CN113139251A (en) * | 2021-04-23 | 2021-07-20 | 桂林电子科技大学 | Variable working condition rolling bearing fault diagnosis method for optimizing theme correlation analysis |
| CN115615540A (en) * | 2022-12-20 | 2023-01-17 | 潍坊百特磁电科技有限公司 | Carrier roller fault identification method, equipment and medium of permanent magnet self-discharging iron remover |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102226740A (en) * | 2011-04-18 | 2011-10-26 | 中国计量学院 | Bearing fault detection method based on manner of controlling stochastic resonance by external periodic signal |
| CN106198000A (en) * | 2016-07-11 | 2016-12-07 | 太原理工大学 | A kind of rocker arm of coal mining machine gear failure diagnosing method |
| CN108256556A (en) * | 2017-12-22 | 2018-07-06 | 上海电机学院 | Wind-driven generator group wheel box method for diagnosing faults based on depth belief network |
| CN109299705A (en) * | 2018-10-24 | 2019-02-01 | 电子科技大学 | Rotary machinery fault diagnosis method based on one-dimensional depth residual error convolutional neural networks |
-
2019
- 2019-02-02 CN CN201910108062.7A patent/CN109766874A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102226740A (en) * | 2011-04-18 | 2011-10-26 | 中国计量学院 | Bearing fault detection method based on manner of controlling stochastic resonance by external periodic signal |
| CN106198000A (en) * | 2016-07-11 | 2016-12-07 | 太原理工大学 | A kind of rocker arm of coal mining machine gear failure diagnosing method |
| CN108256556A (en) * | 2017-12-22 | 2018-07-06 | 上海电机学院 | Wind-driven generator group wheel box method for diagnosing faults based on depth belief network |
| CN109299705A (en) * | 2018-10-24 | 2019-02-01 | 电子科技大学 | Rotary machinery fault diagnosis method based on one-dimensional depth residual error convolutional neural networks |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110276303A (en) * | 2019-06-25 | 2019-09-24 | 湖南科技大学 | Rotor misalignment quantitative identification method based on VMD and DBN |
| CN110399854A (en) * | 2019-07-31 | 2019-11-01 | 中南大学 | Rolling bearing fault classification method based on hybrid feature extraction |
| CN113052716A (en) * | 2019-12-27 | 2021-06-29 | 新疆金风科技股份有限公司 | Abnormity early warning method and device for main bearing of wind generating set |
| CN113052716B (en) * | 2019-12-27 | 2023-03-03 | 新疆金风科技股份有限公司 | Abnormity early warning method and device for main bearing of wind generating set |
| CN111141517A (en) * | 2020-01-02 | 2020-05-12 | 上海电气风电集团股份有限公司 | Fan fault diagnosis method and system |
| CN110953179A (en) * | 2020-02-26 | 2020-04-03 | 浙江上风高科专风实业有限公司 | Fan state monitoring method and system based on inverse wavelet transform |
| CN111120388A (en) * | 2020-03-31 | 2020-05-08 | 浙江上风高科专风实业有限公司 | Fan state combined monitoring method and system |
| CN112052796A (en) * | 2020-09-07 | 2020-12-08 | 电子科技大学 | Permanent magnet synchronous motor fault diagnosis method based on deep learning |
| CN113139251A (en) * | 2021-04-23 | 2021-07-20 | 桂林电子科技大学 | Variable working condition rolling bearing fault diagnosis method for optimizing theme correlation analysis |
| CN113139251B (en) * | 2021-04-23 | 2023-03-14 | 桂林电子科技大学 | Variable working condition rolling bearing fault diagnosis method for optimizing theme correlation analysis |
| CN115615540A (en) * | 2022-12-20 | 2023-01-17 | 潍坊百特磁电科技有限公司 | Carrier roller fault identification method, equipment and medium of permanent magnet self-discharging iron remover |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109766874A (en) | A kind of fan trouble classifying identification method based on deep learning algorithm | |
| CN109376801B (en) | Wind driven generator blade icing diagnosis method based on integrated deep neural network | |
| CN104459388B (en) | Permanent magnetic direct-drive wind power generation system integrated fault diagnosis method | |
| CN109255333B (en) | Fault hybrid diagnosis method for rolling bearing of large wind turbine generator | |
| CN108827605A (en) | A kind of mechanical breakdown characteristic automatic extraction method based on improvement sparseness filtering | |
| CN110046562A (en) | A kind of wind power system health monitor method and device | |
| CN106934421A (en) | Converting station electric power transformer fault detecting system and detection method based on 2DPCA and SVM | |
| CN103176128A (en) | Method and system for forcasting state of wind generating set and diagnosing intelligent fault | |
| CN109297713A (en) | It is a kind of based on steadily with the steam turbine hostdown diagnostic method of Non-stationary vibration signal feature selecting | |
| CN110131109A (en) | A kind of pneumatic equipment bladess unbalance detection based on convolutional neural networks | |
| CN112857767B (en) | Hydro-turbo generator set rotor fault acoustic discrimination method based on convolutional neural network | |
| CN110188867A (en) | Steam turbine hostdown diagnostic method based on integrated empirical mode decomposition and convolutional neural networks | |
| CN111412114B (en) | Wind turbine generator impeller imbalance detection method based on stator current envelope spectrum | |
| CN109100648A (en) | Ocean current generator impeller based on CNN-ARMA-Softmax winds failure fusion diagnosis method | |
| CN113466634B (en) | Ground fault waveform identification method based on fault indicator | |
| CN105065212A (en) | Checking method and system of wind generation sets of wind power plant | |
| CN110595778A (en) | Wind turbine generator bearing fault diagnosis method based on MMF and IGRA | |
| CN107091986A (en) | ANFIS wind-powered electricity generation double fed asynchronous machine air-gap eccentric fault diagnostic classification methods | |
| CN106570790A (en) | Wind farm output power data restoration method considering segmental characteristics of wind speed data | |
| CN115994325A (en) | Fan icing power generation data enhancement method based on TimeGAN deep learning method | |
| Chen et al. | Acoustical damage detection of wind turbine yaw system using Bayesian network | |
| CN113298134A (en) | BPNN-based remote non-contact health monitoring system and method for fan blade | |
| CN115481657A (en) | Wind generating set communication slip ring fault diagnosis method based on electric signals | |
| Yu et al. | Early fault diagnosis model design of reciprocating compressor valve based on multiclass support vector machine and decision tree | |
| CN111539381B (en) | Construction method of wind turbine bearing fault classification diagnosis model |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190517 |
|
| WD01 | Invention patent application deemed withdrawn after publication |