CN109214356A - A kind of fan transmission system intelligent fault diagnosis method based on DCNN model - Google Patents
A kind of fan transmission system intelligent fault diagnosis method based on DCNN model Download PDFInfo
- Publication number
- CN109214356A CN109214356A CN201811145766.3A CN201811145766A CN109214356A CN 109214356 A CN109214356 A CN 109214356A CN 201811145766 A CN201811145766 A CN 201811145766A CN 109214356 A CN109214356 A CN 109214356A
- Authority
- CN
- China
- Prior art keywords
- model
- sample
- dcnn
- layer
- dcnn model
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2218/00—Aspects of pattern recognition specially adapted for signal processing
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M13/00—Testing of machine parts
- G01M13/02—Gearings; Transmission mechanisms
- G01M13/028—Acoustic or vibration analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Abstract
The present invention discloses a kind of based on depth convolutional neural networks (Deep Convolutional Neural Network, DCNN) the fan transmission system intelligent fault diagnosis method of model, the vibration signal that wind turbine generator drive system normally and under malfunction is run is acquired first, handled with vibration data of the Winger-Ville harmonic analysis to acquisition, extract under different operating statuses when-frequency characteristic pattern is as sample set;DCNN model is established, selects a certain number of characteristic patterns as training sample, DCNN model is trained;The remaining sample that finally characteristic pattern sample is concentrated tests trained DCNN model as test sample, optimizes to the structural parameters and training parameter of model, selects best recognition performance, structural parameters of optimal stability and training parameter.Such method can realize the intelligent diagnostics identification to fan transmission system failure, improve the accuracy and efficiency to fault identification.
Description
Technical field
The invention belongs to intelligent diagnostics fields, are related to a kind of based on DCNN (Deep Convolutional Neural
Network, depth convolutional neural networks) model fan transmission system intelligent fault diagnosis method, more particularly to wind turbine
The vibrating failure diagnosis of group transmission chain system.
Background technique
As emerging industry, China starts late in terms of fault diagnosis of wind turbines research, and swift and violent increased wind-powered electricity generation
Installed capacity is compared, the relevant significant backwardness of fault diagnosis technology, in the research of wind-powered electricity generation failure modes and recognition methods, electromechanical coupling
All various aspects such as dependent diagnostic research, intelligent diagnostics and the trend prediction technology research of collaboration system are all weaker, apart from wind-powered electricity generation row
Industry also differs farther out the urgent need of fault diagnosis technology.
External fault diagnosis of wind turbines is mainly for transmission system and electrical system expansion research.Fault diagnosis at present
There are mainly two types of methods, and one is the method for diagnosing faults based on signal processing, and another kind is the fault diagnosis side based on model
Method, the Analysis on Fault Diagnosis of wherein most are all based on vibration signal, also have and are analyzed by oil liquid, noise, electricity etc.
Diagnosis.With the rise of artificial intelligence, the country such as America and Europe also starts for artificial intelligence technology to be introduced into wind-power electricity generation fault diagnosis
In, and achieve good results.China starts late to fault diagnosis of wind turbines research, and present research emphasis mainly exists
On transmission chain, the analysis method of use is also based on vibration analysis, while domestic also having some scholars to carry out based on time-frequency figure
Study on Intelligent Fault Diagnosis, but at present application in feature vector be all arteface on the basis of time-frequency figure, directly
Research using time-frequency figure as basis of characterization is almost without and the shallow-layers network hidden layer quantity such as BP neural network, SVM used
Limited, feature learning ability to express is limited, and training easily falls into local extremum.
Summary of the invention
The purpose of the present invention is to provide a kind of fan transmission system intelligent fault diagnosis method based on DCNN model,
It can realize the intelligent diagnostics identification to fan transmission system failure, improve the accuracy and efficiency to fault identification.
In order to achieve the above objectives, solution of the invention is:
A kind of fan transmission system intelligent fault diagnosis method based on DCNN model, includes the following steps:
Step 1, the vibration signal that acquisition wind turbine generator drive system normally and under malfunction is run, with Winger-
Ville harmonic analysis handles the vibration signal of acquisition, extract under different operating statuses when-frequency characteristic pattern is as sample
This collection, and it is divided into training sample and test sample;
Step 2, DCNN model is established, DCNN model is trained using the training sample of step 1;
Step 3, it is tested using the test sample of step 1 DCNN model trained to step 2, to the structure of model
Parameter and training parameter optimize, and select best recognition performance, structural parameters of optimal stability and training parameter.
In above-mentioned steps 1, Winger-Ville harmonic analysis detailed process is: calculating vibration signal is instantaneous right first
Claim correlation function, Fourier transformation then is carried out to it, obtains the time-frequency figure for being able to reflect fault signature.
In above-mentioned steps 2, the process being trained to DCNN model is: firstly, the training parameter of setting network, initialization
The weight of network and biasing, training sample characteristic pattern successively after convolutional layer, sample level, the processing of full articulamentum, are transmitted to defeated
Layer out, the input that each layer of output is next layer;Then by the layer-by-layer backpropagation of error between reality output and desired output,
And error is assigned to each layer, the weight and biasing of network are adjusted, until meeting the condition of convergence.
After adopting the above scheme, the present invention proposes on the basis of conventional vibration signal characteristic abstraction to blower power train
The vibration signal of system carries out the processing of Winger-Ville harmonic analysis, is reflected using the time-frequency figure that WV harmonic analysis generates
The time-frequency characteristics of vibration signal, the operating status of accurate characterization fan transmission system;WV is composed using convolutional neural networks simultaneously
The time-frequency figure that parser generates adequately is learnt and is expressed, and is avoided artificial participation and is caused being not enough to for feature representation
And the deficiency of identification accuracy and efficiency;Finally failure is identified with trained convolutional neural networks model, thus real
Now to the intelligent diagnostics of wind turbine generator drive system failure.The invention avoids the artificial deficiency for participating in causing feature representation and
The problem for identifying accuracy and efficiency deficiency has higher accuracy rate and more preferably stability compared with existing recognition methods.
Detailed description of the invention
Fig. 1 is flow chart of the invention;
Fig. 2 is DCNN model training flow chart;
Fig. 3 is DCNN model structure schematic diagram;
Fig. 4 is DCNN model structure detail drawing.
Specific embodiment
Below with reference to attached drawing, technical solution of the present invention and beneficial effect are described in detail.
The fan transmission system intelligent fault diagnosis method based on DCNN model that the present invention provides a kind of, key step are situated between
It continues as follows:
1, vibration signal pre-processes, and generates WV spectrum analysis and is distributed time-frequency figure
When wind turbine transmission chain breaks down in practical application, impact vibration signal is generally produced, is had apparent
Nonlinear and nonstationary feature, traditional signal processing means are difficult to sensitively react mutation and non-stationary characteristic.Therefore one is needed
Effective method is planted to handle vibration signal.
The present invention is calculated first by handling with vibration signal of the Wigner-Ville harmonic analysis to acquisition
The transient symmetric correlation function of vibration signal, then carries out Fourier transformation to it, time-frequency image is obtained, then by image with ash
Degree form is shown, and time-frequency figure is compressed to suitable size, construction feature figure sample set, and the purpose of compression is to reduce
The size of DCNN input feature vector Tu Gewei, improves the training speed of network, while guaranteeing that the useful information in time-frequency figure is not flooded
Not yet.The time-frequency bandwidth product of the WV spectrum analysis Distribution Algorithm has reached the lower bound that Heisenberg uncertainty principle provides, and makes
It will not lose the amplitude and phase information of signal with very high time frequency resolution, have very to nonlinear and non local boundary value problem
Good time-frequency ability to express.
2, DCNN model is established, selects a certain number of characteristic patterns as training sample, DCNN is trained
Network training parameter setting is as follows: batch=32, epoch=80, and learning rate λ=0.0002 uses Adam algorithm
Optimizer, cross entropy is as error function.As shown in Fig. 2, the training process of DCNN model mainly includes that the positive of data passes
Two parts of backpropagation with error are broadcast, firstly, the training parameter of setting network, initializes weight and the biasing of network, it will
The time-frequency figure that characteristic pattern sample is concentrated is divided into training sample and test sample, and training sample characteristic pattern is inputted network, is successively passed through
After crossing convolutional layer, sample level, the processing of full articulamentum, it is transmitted to output layer, the input that each layer of output is next layer;Then,
By the layer-by-layer backpropagation of error between reality output and desired output, and error is assigned to each layer, to the weight of network and partially
It sets and is adjusted, until meeting the condition of convergence, to realize the Training of network.
Adequately learnt and expressed with the time-frequency figure that convolutional neural networks can generate WV harmonic analysis,
DCNN is a kind of deep learning network model containing multiple hidden layers, can be transmitted by layer-by-layer feature, low-level feature is transformed to
High-level characteristic, to realize the study and expression of feature, while compared with the shallow-layers network such as BP neural network, SVM, DCNN is to complexity
The study ability to express of feature is stronger, and arithmetic speed faster, avoids the problems such as training falls into local extremum;DCNN can successfully by
The similar fault sample that time-frequency figure differs greatly is identified as one kind, embodies its stronger generalization ability and assembility;DCNN
The similar foreign peoples's fault sample of time-frequency figure can be successfully distinguished, its stronger feature extraction and recognition capability are embodied.
3, trained DCNN model is tested
10 kinds of organization plans of planned network, respectively test it using characteristic pattern sample set, for test network
The stability of performance eliminates the influence of enchancement factor, and experiment is all repeated 10 times every time, with the minimum value, mean value, mark of test result
The time that quasi- poor and iteration once consumes is as evaluation index.DCNN structure is as shown in Figure 3 and Figure 4, input feature vector figure it is big
Small is 1024 × 1024, and then convolution sample level is alternately present three times, and first convolutional layer uses 64@2 × 2 of convolution kernel, second
A convolutional layer uses 128@2 × 2 of convolution kernel, and third convolutional layer uses 192@2 × 2 of convolution kernel, and pond layer is all made of mean value pond
Changing size is 2 × 2.Network architecture parameter and training parameter are optimized by test sample, network can be effectively improved
The stability of correct recognition rata and recognition performance to failure.
The above examples only illustrate the technical idea of the present invention, and this does not limit the scope of protection of the present invention, all
According to the technical idea provided by the invention, any changes made on the basis of the technical scheme each falls within the scope of the present invention
Within.
Claims (3)
1. a kind of fan transmission system intelligent fault diagnosis method based on DCNN model, it is characterised in that include the following steps:
Step 1, the vibration signal that acquisition wind turbine generator drive system normally and under malfunction is run, with Winger-Ville
Harmonic analysis handles the vibration signal of acquisition, extract under different operating statuses when-frequency characteristic pattern as sample set,
And it is divided into training sample and test sample;
Step 2, DCNN model is established, DCNN model is trained using the training sample of step 1;
Step 3, it is tested using the test sample of step 1 DCNN model trained to step 2, to the structural parameters of model
It is optimized with training parameter, selects best recognition performance, structural parameters of optimal stability and training parameter.
2. a kind of fan transmission system intelligent fault diagnosis method based on DCNN model as described in claim 1, feature
Be: in the step 1, Winger-Ville harmonic analysis detailed process is: calculating the transient symmetric of vibration signal first
Then correlation function carries out Fourier transformation to it, obtain the time-frequency figure for being able to reflect fault signature.
3. a kind of fan transmission system intelligent fault diagnosis method based on DCNN model as described in claim 1, feature
Be: in the step 2, the process being trained to DCNN model is: firstly, the training parameter of setting network, initializes net
The weight of network and biasing, training sample characteristic pattern successively after convolutional layer, sample level, the processing of full articulamentum, are transmitted to output
Layer, the input that each layer of output is next layer;Then by the layer-by-layer backpropagation of error between reality output and desired output, and
Error is assigned to each layer, the weight and biasing of network are adjusted, until meeting the condition of convergence.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811145766.3A CN109214356A (en) | 2018-09-29 | 2018-09-29 | A kind of fan transmission system intelligent fault diagnosis method based on DCNN model |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201811145766.3A CN109214356A (en) | 2018-09-29 | 2018-09-29 | A kind of fan transmission system intelligent fault diagnosis method based on DCNN model |
Publications (1)
Publication Number | Publication Date |
---|---|
CN109214356A true CN109214356A (en) | 2019-01-15 |
Family
ID=64982163
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201811145766.3A Pending CN109214356A (en) | 2018-09-29 | 2018-09-29 | A kind of fan transmission system intelligent fault diagnosis method based on DCNN model |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN109214356A (en) |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110595780A (en) * | 2019-09-20 | 2019-12-20 | 西安科技大学 | Bearing fault identification method based on vibration gray level image and convolution neural network |
CN110702411A (en) * | 2019-09-23 | 2020-01-17 | 武汉理工大学 | Residual error network rolling bearing fault diagnosis method based on time-frequency analysis |
CN111426950A (en) * | 2020-03-19 | 2020-07-17 | 燕山大学 | Wind driven generator fault diagnosis method of multi-scale space-time convolution depth belief network |
CN111709190A (en) * | 2020-06-24 | 2020-09-25 | 国电联合动力技术有限公司 | Wind turbine generator operation data image identification method and device |
CN113554085A (en) * | 2021-07-20 | 2021-10-26 | 云南电力试验研究院(集团)有限公司 | System and method for sensing vibration safety situation of induced draft fan |
CN114233581A (en) * | 2021-12-13 | 2022-03-25 | 山东神戎电子股份有限公司 | Intelligent patrol alarm system for fan engine room |
CN114757239A (en) * | 2022-06-15 | 2022-07-15 | 浙江大学 | Fan fault migratable diagnosis method based on data enhancement and capsule neural network |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101872165A (en) * | 2010-06-13 | 2010-10-27 | 西安交通大学 | Method for fault diagnosis of wind turbines on basis of genetic neural network |
CN106124212A (en) * | 2016-06-16 | 2016-11-16 | 燕山大学 | Based on sparse coding device and the Fault Diagnosis of Roller Bearings of support vector machine |
CN107560849A (en) * | 2017-08-04 | 2018-01-09 | 华北电力大学 | A kind of Wind turbines Method for Bearing Fault Diagnosis of multichannel depth convolutional neural networks |
CN107657250A (en) * | 2017-10-30 | 2018-02-02 | 四川理工学院 | Bearing fault detection and localization method and detection location model realize system and method |
CN107979554A (en) * | 2017-11-17 | 2018-05-01 | 西安电子科技大学 | Radio signal Modulation Identification method based on multiple dimensioned convolutional neural networks |
-
2018
- 2018-09-29 CN CN201811145766.3A patent/CN109214356A/en active Pending
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101872165A (en) * | 2010-06-13 | 2010-10-27 | 西安交通大学 | Method for fault diagnosis of wind turbines on basis of genetic neural network |
CN106124212A (en) * | 2016-06-16 | 2016-11-16 | 燕山大学 | Based on sparse coding device and the Fault Diagnosis of Roller Bearings of support vector machine |
CN107560849A (en) * | 2017-08-04 | 2018-01-09 | 华北电力大学 | A kind of Wind turbines Method for Bearing Fault Diagnosis of multichannel depth convolutional neural networks |
CN107657250A (en) * | 2017-10-30 | 2018-02-02 | 四川理工学院 | Bearing fault detection and localization method and detection location model realize system and method |
CN107979554A (en) * | 2017-11-17 | 2018-05-01 | 西安电子科技大学 | Radio signal Modulation Identification method based on multiple dimensioned convolutional neural networks |
Non-Patent Citations (1)
Title |
---|
李兵等著: "时频分析技术", 《机械故障信号的数学形态学分析与智能分类》 * |
Cited By (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110595780A (en) * | 2019-09-20 | 2019-12-20 | 西安科技大学 | Bearing fault identification method based on vibration gray level image and convolution neural network |
CN110595780B (en) * | 2019-09-20 | 2021-12-14 | 西安科技大学 | Bearing fault identification method based on vibration gray level image and convolution neural network |
CN110702411A (en) * | 2019-09-23 | 2020-01-17 | 武汉理工大学 | Residual error network rolling bearing fault diagnosis method based on time-frequency analysis |
CN110702411B (en) * | 2019-09-23 | 2020-11-10 | 武汉理工大学 | Residual error network rolling bearing fault diagnosis method based on time-frequency analysis |
CN111426950A (en) * | 2020-03-19 | 2020-07-17 | 燕山大学 | Wind driven generator fault diagnosis method of multi-scale space-time convolution depth belief network |
CN111426950B (en) * | 2020-03-19 | 2020-11-27 | 燕山大学 | Wind driven generator fault diagnosis method of multi-scale space-time convolution depth belief network |
CN111709190A (en) * | 2020-06-24 | 2020-09-25 | 国电联合动力技术有限公司 | Wind turbine generator operation data image identification method and device |
CN113554085A (en) * | 2021-07-20 | 2021-10-26 | 云南电力试验研究院(集团)有限公司 | System and method for sensing vibration safety situation of induced draft fan |
CN113554085B (en) * | 2021-07-20 | 2023-04-18 | 云南电力试验研究院(集团)有限公司 | System and method for sensing vibration safety situation of induced draft fan |
CN114233581A (en) * | 2021-12-13 | 2022-03-25 | 山东神戎电子股份有限公司 | Intelligent patrol alarm system for fan engine room |
CN114757239A (en) * | 2022-06-15 | 2022-07-15 | 浙江大学 | Fan fault migratable diagnosis method based on data enhancement and capsule neural network |
CN114757239B (en) * | 2022-06-15 | 2022-08-30 | 浙江大学 | Fan fault migratable diagnosis method based on data enhancement and capsule neural network |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN109214356A (en) | A kind of fan transmission system intelligent fault diagnosis method based on DCNN model | |
Li et al. | A novel deep autoencoder and hyperparametric adaptive learning for imbalance intelligent fault diagnosis of rotating machinery | |
CN109977920A (en) | Fault Diagnosis of Hydro-generator Set method based on time-frequency spectrum and convolutional neural networks | |
CN101634605B (en) | Intelligent gearbox fault diagnosis method based on mixed inference and neural network | |
Chen et al. | A novel fault diagnosis model for gearbox based on wavelet support vector machine with immune genetic algorithm | |
CN106197999B (en) | A kind of planetary gear method for diagnosing faults | |
CN105372087B (en) | Method for diagnosing faults based on multiple sensor signals analysis | |
CN104792522A (en) | Intelligent gear defect analysis method based on fractional wavelet transform and BP neutral network | |
CN109000930B (en) | Turbine engine performance degradation evaluation method based on stacking denoising autoencoder | |
CN101246200B (en) | Analog PCB intelligent test system based on neural network | |
CN108197648A (en) | A kind of Fault Diagnosis Method of Hydro-generating Unit and system based on LSTM deep learning models | |
CN105628425A (en) | Rotation machinery early stage fault diagnosis method based on heredity annealing optimization multi-core support vector machine | |
CN109978079A (en) | A kind of data cleaning method of improved storehouse noise reduction self-encoding encoder | |
CN107884475A (en) | A kind of city gas pipeline method for diagnosing faults based on deep learning neutral net | |
CN103245907B (en) | A kind of analog-circuit fault diagnosis method | |
CN113935460A (en) | Intelligent diagnosis method for mechanical fault under class imbalance data set | |
CN105758644A (en) | Rolling bearing fault diagnosis method based on variation mode decomposition and permutation entropy | |
CN103886405B (en) | Boiler combustion condition identification method based on information entropy characteristics and probability nerve network | |
Zhang et al. | Short term wind energy prediction model based on data decomposition and optimized LSSVM | |
CN109472288A (en) | A kind of pump-storage generator vibration hybrid feature extraction and classification method | |
CN109657789A (en) | Gear case of blower failure trend prediction method based on wavelet neural network | |
CN101221213A (en) | Analogue circuit fault diagnosis neural network method based on particle swarm algorithm | |
CN103995237A (en) | Satellite power supply system online fault diagnosis method | |
CN102289682A (en) | Transformer fault diagnosis method based on integrated learning Bagging algorithm | |
CN104748962A (en) | Planetary gear box intelligent diagnosis method based on stacking automatic encoding machine |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
PB01 | Publication | ||
PB01 | Publication | ||
SE01 | Entry into force of request for substantive examination | ||
SE01 | Entry into force of request for substantive examination | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190115 |
|
RJ01 | Rejection of invention patent application after publication |