CN109029974A - A kind of epicyclic gearbox fault detection method based on one-dimensional convolutional neural networks - Google Patents
A kind of epicyclic gearbox fault detection method based on one-dimensional convolutional neural networks Download PDFInfo
- Publication number
- CN109029974A CN109029974A CN201810654805.6A CN201810654805A CN109029974A CN 109029974 A CN109029974 A CN 109029974A CN 201810654805 A CN201810654805 A CN 201810654805A CN 109029974 A CN109029974 A CN 109029974A
- Authority
- CN
- China
- Prior art keywords
- layer
- neural networks
- layers
- convolutional neural
- epicyclic gearbox
- 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
- 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
- G01M13/021—Gearings
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F03—MACHINES OR ENGINES FOR LIQUIDS; WIND, SPRING, OR WEIGHT MOTORS; PRODUCING MECHANICAL POWER OR A REACTIVE PROPULSIVE THRUST, NOT OTHERWISE PROVIDED FOR
- F03D—WIND MOTORS
- F03D17/00—Monitoring or testing of wind motors, e.g. diagnostics
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01M—TESTING STATIC OR DYNAMIC BALANCE OF MACHINES OR STRUCTURES; TESTING OF STRUCTURES OR APPARATUS, NOT OTHERWISE PROVIDED FOR
- G01M15/00—Testing of engines
- G01M15/02—Details or accessories of testing apparatus
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Sustainable Development (AREA)
- Sustainable Energy (AREA)
- Chemical & Material Sciences (AREA)
- Combustion & Propulsion (AREA)
- Mechanical Engineering (AREA)
- General Engineering & Computer Science (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The present invention relates to a kind of epicyclic gearbox fault detection methods based on one-dimensional convolutional neural networks, and this method comprises the following steps: (1) obtaining epicyclic gearbox vibration signal；(2) epicyclic gearbox vibration signal is input to fault identification model trained in advance, the fault identification model is one-dimensional convolutional neural networks；(3) fault identification model identifies the failure of epicyclic gearbox and exports fault type.Compared with prior art, the present invention is able to achieve accurate, the quick diagnosis of epicyclic gearbox failure.
Description
Technical field
The present invention relates to power system device maintenance area, more particularly, to a kind of based on one-dimensional convolutional neural networks
Epicyclic gearbox fault detection method.
Background technique
Wind energy is current most promising one of new energy, the epicyclic gearbox biography important as wind-driven generator
Dynamic device, it is made of planetary gear, sun gear, gear ring and planet carrier, and high torque ratio can be obtained in compact space.
Since it vibrates transmission path complexity, the originals such as non-stationary and working background noise is big of multiple tooth engagement effect, signal
Cause causes its fault diagnosis to have the characteristics that itself and difficult point, conventional method it is carried out time domain or frequency-domain analysis it is difficult to extract
Effective fault message.With Internet technology, the development of technology of Internet of things, the acquisition and storage of data are more convenient, are based on
The fault diagnosis of data-driven becomes a new developing direction.Different with conventional method, data-driven method is without carrying out object
Reason modeling directly carries out processing and analysis appropriate to the data of acquisition to extract information characteristics, to find fault observer.
In fault diagnosis technology, the characteristic features of data how are effectively extracted, the precision of diagnosis is played to pass
Important role.Important algorithm of the traditional neural network as feature extraction is examined in fault diagnosis field and electric system
It is widely studied and is applied in survey field.But traditional neural network algorithm has the shortcomings that be difficult to overcome, such as algorithm sheet
Body computational efficiency is low, and diagnostic accuracy is difficult to reach requirement, needs the initial data such as to pre-process.
Early stage due to data and the computing capability of lacking training, will train high-performance refreshing in the case where not generating over-fitting
It is highly difficult, the development of recent GPU through network, so that convolutional neural networks (Convolutional Neural
Networks, CNNs) it studies and emerges in large numbers.CNNs is a kind of efficient identification method that developed recently gets up, pattern-recognition,
The fields such as medicine, which achieve, to be widely applied.In area of pattern recognition, CNNs is mainly used to identification displacement, scaling and other shapes
Formula distorts the X-Y scheme of invariance, due to that can directly input original this method avoid the pretreatment complicated early period to image
Beginning image, thus obtained more being widely applied.
Summary of the invention
It is an object of the present invention to overcome the above-mentioned drawbacks of the prior art and provide one kind to be based on one-dimensional convolution
The epicyclic gearbox fault detection method of neural network.
The purpose of the present invention can be achieved through the following technical solutions:
A kind of epicyclic gearbox fault detection method based on one-dimensional convolutional neural networks, this method comprises the following steps:
(1) epicyclic gearbox vibration signal is obtained；
(2) epicyclic gearbox vibration signal is input to fault identification model trained in advance, the fault identification mould
Type is one-dimensional convolutional neural networks；
(3) fault identification model identifies the failure of epicyclic gearbox and exports fault type.
The one-dimensional convolutional neural networks include sequentially connected input layer, the first convolutional layer, the first pond layer,
Two convolutional layers, the second pond layer, feature vector layer and output layer, the input layer is for input planet gear case vibration letter
Number, the first convolutional layer, the first pond layer, the second convolutional layer and the second pond layer successively carries out convolution-down-sampling-volume
The Feature Mapping figure head and the tail connection of second pond layer is formed feature vector by the operation of product-down-sampling, the feature vector layer,
Feature vector is connected entirely and exports fault type class vector by the output layer.
First convolutional layer and the second convolutional layer is one-dimensional convolutional layer.
First convolutional layer and the second convolutional layer specifically:
If l layers are convolutional layer in one-dimensional convolutional neural networks, then the calculation formula of convolutional layer is corresponded to are as follows:
Indicate l layers of j-th of Feature Mapping,Indicate l-1 layers of ith feature mapping, M indicates that l-1 layers of feature are reflected
The number penetrated,Indicate l layers of trainable convolution kernel,Indicate l layers of biasing, * is convolution operation, and f () is activation primitive.
The first pond layer and the second pond layer specifically:
If l+1 layers are pond layer in one-dimensional convolutional neural networks, then the calculation formula of pond layer is corresponded to are as follows:
Indicate l+1 layers of j-th of Feature Mapping,Indicate l layers of j-th of Feature Mapping,Indicate l+1 layers inclined
It sets, down () is down-sampling function, and f () is activation primitive.
The output layer specifically:
yl+1=f (ul+1)=f (Wl+1xl+1+bl+1),
yl+1Indicate fault type class vector, xl+1Indicate l+1 layers of Feature Mapping, Wl+1Indicate the weight of output layer,Indicate the biasing of output layer, f () is activation primitive.
Compared with prior art, the present invention has the advantage that
Fault identification model of the present invention uses one-dimensional convolutional neural networks, and one-dimensional convolutional neural networks are only to local progress
Perception, and between parameter be it is shared, this local sensing and parameter sharing operation can reduce number of parameters, Neng Gou great
It is big to improve computational efficiency, to meet the on-line monitoring of modern electric equipment fault diagnosis and the requirement of real-time diagnosis.
Detailed description of the invention
Fig. 1 is the flow diagram of the epicyclic gearbox fault detection method based on one-dimensional convolutional neural networks；
Fig. 2 is the structural block diagram of the one-dimensional convolutional neural networks of the present invention；
Fig. 3 is the schematic diagram of the one-dimensional convolutional neural networks of the present invention；
Fig. 4 is the one-dimensional convolutional neural networks training flow chart of the present invention；
Fig. 5 is epicyclic gearbox planetary gear vibrational waveform figure；
Fig. 6 is variation of the mean square error with iteration.
Specific embodiment
The present invention is described in detail with specific embodiment below in conjunction with the accompanying drawings.Note that the following embodiments and the accompanying drawings is said
Bright is substantial illustration, and the present invention is not intended to be applicable in it object or its purposes is defined, and the present invention does not limit
In the following embodiments and the accompanying drawings.
Embodiment
As shown in Figure 1, a kind of epicyclic gearbox fault detection method based on one-dimensional convolutional neural networks, this method packet
Include following steps:
(1) epicyclic gearbox vibration signal is obtained；
(2) epicyclic gearbox vibration signal is input to fault identification model trained in advance, the fault identification mould
Type is one-dimensional convolutional neural networks；
(3) fault identification model identifies the failure of epicyclic gearbox and exports fault type.
As shown in Fig. 2, one-dimensional convolutional neural networks include sequentially connected input layer, the first convolutional layer, the first pond
Layer, the second convolutional layer, the second pond layer, feature vector layer and output layer, the input layer shake for input planet gear case
Dynamic signal, first convolutional layer, the first pond layer, the second convolutional layer and the second pond layer successively carry out adopting under convolution-
Sample-convolution-down-sampling operation, the Feature Mapping figure head and the tail connection of the second pond layer is formed feature vector by feature vector layer, defeated
Feature vector is connected entirely and exports fault type class vector by layer out.This network structure is named as CN8 (k1)-s2-
CN16(k3)-s4.Wherein, k1 and k3 respectively represents the size of convolution kernel, and s2 and s4 represent the diminution ruler of pond layer down-sampling
Degree.
First convolutional layer and the second convolutional layer are one-dimensional convolutional layer.
First convolutional layer and the second convolutional layer specifically:
If l layers are convolutional layer in one-dimensional convolutional neural networks, then the calculation formula of convolutional layer is corresponded to are as follows:
Indicate l layers of j-th of Feature Mapping,Indicate l-1 layers of ith feature mapping, M indicates that l-1 layers of feature are reflected
The number penetrated,Indicate l layers of trainable convolution kernel,Indicate l layers of biasing, * is convolution operation, and f () is activation letter
Number.
First pond layer and the second pond layer specifically:
If l+1 layers are pond layer in one-dimensional convolutional neural networks, then the calculation formula of pond layer is corresponded to are as follows:
Indicate l+1 layers of j-th of Feature Mapping,Indicate l layers of j-th of Feature Mapping,Indicate l+1 layers inclined
It sets, down () is down-sampling function, and f () is activation primitive.
Output layer specifically:
yl+1=f (ul+1)=f (Wl+1xl+1+bl+1),
yl+1Indicate fault type class vector, xl+1Indicate l+1 layers of Feature Mapping, Wl+1Indicate the weight of output layer,Indicate the biasing of output layer, f () is activation primitive.
As a result, one-dimensional convolutional neural networks schematic diagram as shown in figure 3, the figure there was only one layer of convolutional layer and one layer of pond
Illustrate the working principle of one-dimensional convolutional neural networks for change layer.Input data first passes around convolution, then carries out down-sampling behaviour
Make, the Feature Mapping figure head and the tail connection after down-sampling is operated forms feature vector, and output layer is connected feature vector entirely
And export fault type class vector.
When establishing one-dimensional convolutional neural networks, training convolutional neural networks first initialize the parameter of neural network,
Network is trained using BP algorithm, iteration reaches sets requirement until network error or the number of iterations.Finally we examine net
Whether the training precision of network is sufficiently high, if meeting required precision, saves network model, if not satisfied, then resetting convolution
The core size core pond factor, training process are as shown in Figure 4.
For the validity for verifying this method, the present embodiment has carried out case verification.
Data source is tested planetary gear and is mounted in gearbox gear roller box in gear case of blower analog platform, acceleration
Meter is mounted on gear box casing to measure vibration signal.Motor speed, rotary frequency can be changed by speed control
Rate setting range is 0~60Hz.The sample frequency of signal is 12kHz.Failure planetary gear failure include abrasion, spot corrosion,
Broken teeth failure.Planetary gear health status includes: normal, abrasion, spot corrosion and broken teeth situation, when driving motor revolving speed is
The planetary gear time domain waveform of 40Hz, acquisition are as shown in Figure 5.
It acquires in data procedures, setting motor speed is respectively 30Hz, 40Hz and 50Hz.Under every kind of revolving speed not
600 samples are acquired with health status, altogether include 7200 samples, and each sample includes 1200 data points.
Parameter involved in the present invention includes the contracting of Feature Mapping the figure number, the size of convolution kernel, pond layer of convolutional layer
Small scale, learning rate and the number of iterations.
In order to study influence of the various parameters to classification accuracy rate, a monovolume lamination list pond layer is constructed first
Network structure CN8 (k1)-s2, input layer and output layer are identical with aforementioned network.Next has studied k1 and s2 to the shadow of accuracy
It rings.After determining k1 and s2, CN8 (k1)-s2-CN16 (k3)-s4, research k3 and s4 influences accuracy.Determine network structure
Afterwards, the influence of last research learning rate η and the number of iterations to accuracy.
1. long-pending core size and pond layer reduce the selection of the factor:
Simple structure network structure CN8 (k1)-s2 first, initial learning rate are 0.05, and convolutional neural networks training changes
Generation number is 100 times.The sample of extraction 10% is for training from planetary gear sample at random, and remaining 90% sample is for testing
Neural network accuracy.The value for taking s2 is 10, i.e., the pond factor of first pond layer is 10 times.Study influence of the k1 to accuracy.Table
1 lists corresponding accuracy under each k1 value.In view of training precision and measuring accuracy, it is 51 that we, which choose k1 value, construction
Network structure CN8 (51)-s2, other conditions are constant, study influence of the s2 to accuracy.
Table 2 lists corresponding accuracy under each s2 value, obtains when s2 chooses 25, reaches best effects.
- 10 accuracy of table 1CN8 (k1)
Table 2CN8 (51)-s2 accuracy
Obtaining (51) -25 structure of CN8 has best effects, deepens network, tectonic network CN8 (51) -25- on this basis
CN16(k3)-s4.With same method, enabling s4 first is 5, obtains k3, later re-optimization s4.Obtained result is respectively such as table 3
With shown in table 4.Available highest accuracy rate when being 9 that obtain k3 value be 2, s4 value, i.e. network structure is CN8 (51) -25-
CN16(2)-9。
Table 3CN8 (51) -25-CN16 (k3) -5 accuracy
Table 4CN8 (51) -25-CN16 (2)-s4 accuracy
2. the selection of the selection of learning rate η and training the number of iterations:
The convergence rate of learning rate influence network.Learning rate is excessively high, will lead to network paralysis, weight and convolution kernel and is learning
It is zeroed out during practising, learning rate is too low, causes network can not fast convergence.Under suitable learning rate, the increasing of the number of iterations
Add the error that can reduce network, network is made to have higher accuracy.At neural network iteration 100 times, table 5 lists convolution
Measuring accuracy under neural network difference learning rate.With the increase of network the number of iterations, the error of network is gradually reduced, Fig. 6
Indicate the mean square error of convolutional neural networks with the variation of the number of iterations.
As can be seen from Table 5, when learning rate is 0.25, measuring accuracy reaches highest, is 96.33%.It can by Fig. 6
To find out, with increasing for the number of iterations, mean square error is gradually reduced, and when iteration is 500 times, the error change of network becomes
Must be very slow, it is inefficient to be further added by the number of iterations, therefore selects the number of iterations for 500 times.
Measuring accuracy under the different learning rate η of table 5
In an experiment, computer processor is intel pentium g2030, inside saves as the DDR3 memory of 2GB.Iteration 100 times
Required time about 3.5 minutes.If 1 second vibration data is inputted trained network, it is total that propagated forward obtains class vector
Time was less than 0.045 second.In view of network only needs training primary, network query function speed can adapt to the needs of real-time diagnosis.
Determine that network structure is CN8 (51) -25-CN16 (2) -9, i.e. C1 layers and C3 layers contains 8 and 16 features respectively
Mapping graph, k1 and k2 size are respectively 51 × 1 and 2 × 1, and the pond factor of pond layer is respectively 25 and 9.Value is 0.25.With
Machine extracts 10% sample from planetary gear sample, i.e. every kind of health status randomly selects 180 samples, and totally 720 samples are used
In training, remaining 90% sample, i.e. 1620 samples of every kind of health status, totally 6480 samples are used for test network precision.One
Tieing up convolutional neural networks training the number of iterations is 500 times.
Higher for the error rate of broken teeth failure modes, in 1620 groups of test samples, 60 groups are assigned to normal condition, and 24 groups
Spot corrosion situation is assigned to, reason may be that one side broken teeth signal and normal signal all have a large amount of ambient noise, cause to classify
Mistake, another aspect broken teeth and spot corrosion belong to local fault, and fault signature is similar, so as to cause classification error.
In order to exclude the randomness of experiment, 5 groups of experiments have been carried out respectively, obtaining average training precision is 99.86%, is put down
Equal measuring accuracy is 98.17%.
It can be seen that this method according to above-mentioned embodiment to be not necessarily to carry out in pretreated situation vibration data, vibration
Sample of signal trains network, and can achieve very high diagnosis accuracy, and this method can adapt to the needs of real-time.
With good practical value.
Above embodiment is only to enumerate, and does not indicate limiting the scope of the invention.These embodiments can also be with other
Various modes are implemented, and can make in the range of not departing from technical thought of the invention it is various omit, displacement, change.
Claims (6)
1. a kind of epicyclic gearbox fault detection method based on one-dimensional convolutional neural networks, which is characterized in that this method includes
Following steps:
(1) epicyclic gearbox vibration signal is obtained；
(2) epicyclic gearbox vibration signal is input to fault identification model trained in advance, the fault identification model is
One-dimensional convolutional neural networks；
(3) fault identification model identifies the failure of epicyclic gearbox and exports fault type.
2. a kind of epicyclic gearbox fault detection method based on one-dimensional convolutional neural networks according to claim 1,
It is characterized in that, the one-dimensional convolutional neural networks include sequentially connected input layer, the first convolutional layer, the first pond layer,
Two convolutional layers, the second pond layer, feature vector layer and output layer, the input layer is for input planet gear case vibration letter
Number, the first convolutional layer, the first pond layer, the second convolutional layer and the second pond layer successively carries out convolution-down-sampling-volume
The Feature Mapping figure head and the tail connection of second pond layer is formed feature vector by the operation of product-down-sampling, the feature vector layer, institute
Feature vector is connected entirely and exports fault type class vector by the output layer stated.
3. a kind of epicyclic gearbox fault detection method based on one-dimensional convolutional neural networks according to claim 2,
It is characterized in that, first convolutional layer and the second convolutional layer are one-dimensional convolutional layer.
4. a kind of epicyclic gearbox fault detection method based on one-dimensional convolutional neural networks according to claim 2,
It is characterized in that, first convolutional layer and the second convolutional layer specifically:
If l layers are convolutional layer in one-dimensional convolutional neural networks, then the calculation formula of convolutional layer is corresponded to are as follows:
Indicate l layers of j-th of Feature Mapping,Indicate l-1 layers of ith feature mapping, M indicates l-1 layers of Feature Mapping
Number,Indicate l layers of trainable convolution kernel,Indicate l layers of biasing, * is convolution operation, and f () is activation primitive.
5. a kind of epicyclic gearbox fault detection method based on one-dimensional convolutional neural networks according to claim 2,
It is characterized in that, the first pond layer and the second pond layer specifically:
If l+1 layers are pond layer in one-dimensional convolutional neural networks, then the calculation formula of pond layer is corresponded to are as follows:
Indicate l+1 layers of j-th of Feature Mapping,Indicate l layers of j-th of Feature Mapping,Indicate l+1 layers of biasing,
Down () is down-sampling function, and f () is activation primitive.
6. a kind of epicyclic gearbox fault detection method based on one-dimensional convolutional neural networks according to claim 2,
It is characterized in that, the output layer specifically:
yl+1=f (ul+1)=f (Wl+1xl+1+bl+1),
yl+1Indicate fault type class vector, xl+1Indicate l+1 layers of Feature Mapping, Wl+1Indicate the weight of output layer,Table
Show the biasing of output layer, f () is activation primitive.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810654805.6A CN109029974A (en) | 2018-06-22 | 2018-06-22 | A kind of epicyclic gearbox fault detection method based on one-dimensional convolutional neural networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810654805.6A CN109029974A (en) | 2018-06-22 | 2018-06-22 | A kind of epicyclic gearbox fault detection method based on one-dimensional convolutional neural networks |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109029974A true CN109029974A (en) | 2018-12-18 |
Family
ID=64610856
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810654805.6A Pending CN109029974A (en) | 2018-06-22 | 2018-06-22 | A kind of epicyclic gearbox fault detection method based on one-dimensional convolutional neural networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109029974A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109829408A (en) * | 2019-01-23 | 2019-05-31 | 中国科学技术大学 | Intelligent lightening recognition device based on convolutional neural networks |
| CN109932174A (en) * | 2018-12-28 | 2019-06-25 | 南京信息工程大学 | A kind of Fault Diagnosis of Gear Case method based on multitask deep learning |
| CN110657984A (en) * | 2019-09-27 | 2020-01-07 | 重庆大学 | Planetary gearbox fault diagnosis method based on reinforced capsule network |
| CN111060316A (en) * | 2019-12-10 | 2020-04-24 | 华南理工大学 | Rolling bearing state monitoring method and system based on convolutional neural network model |
| CN111413075A (en) * | 2020-04-02 | 2020-07-14 | 重庆交通大学 | Fan base bolt loosening diagnosis method of multi-scale one-dimensional convolution neural network |
| CN112113755A (en) * | 2020-09-02 | 2020-12-22 | 西安交通大学 | Mechanical fault intelligent diagnosis method based on deep convolution-kurtosis neural network |
| CN112766200A (en) * | 2021-01-26 | 2021-05-07 | 福州大学 | Planetary gear box small sample fault diagnosis method based on one-dimensional convolutional neural network |
| CN115578365A (en) * | 2022-10-26 | 2023-01-06 | 西南交通大学 | Tooth pitch detection method and equipment for adjacent racks of toothed rail railway |
| CN116292367A (en) * | 2023-03-22 | 2023-06-23 | 山东科技大学 | Power plant fan system abnormal condition detection method based on one-dimensional convolution |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103544705A (en) * | 2013-10-25 | 2014-01-29 | 华南理工大学 | Image quality testing method based on deep convolutional neural network |
| CN108038466A (en) * | 2017-12-26 | 2018-05-15 | 河海大学 | Multichannel human eye closure recognition methods based on convolutional neural networks |
-
2018
- 2018-06-22 CN CN201810654805.6A patent/CN109029974A/en active Pending
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103544705A (en) * | 2013-10-25 | 2014-01-29 | 华南理工大学 | Image quality testing method based on deep convolutional neural network |
| CN103544705B (en) * | 2013-10-25 | 2016-03-02 | 华南理工大学 | A kind of image quality test method based on degree of depth convolutional neural networks |
| CN108038466A (en) * | 2017-12-26 | 2018-05-15 | 河海大学 | Multichannel human eye closure recognition methods based on convolutional neural networks |
Non-Patent Citations (1)
| Title |
|---|
| 李东东 等: ""基于一维卷积神经网络和soft-max分类器的风电机组行星齿轮箱故障检测"", 《电机与控制应用》 * |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109932174A (en) * | 2018-12-28 | 2019-06-25 | 南京信息工程大学 | A kind of Fault Diagnosis of Gear Case method based on multitask deep learning |
| CN109829408A (en) * | 2019-01-23 | 2019-05-31 | 中国科学技术大学 | Intelligent lightening recognition device based on convolutional neural networks |
| CN110657984A (en) * | 2019-09-27 | 2020-01-07 | 重庆大学 | Planetary gearbox fault diagnosis method based on reinforced capsule network |
| CN111060316A (en) * | 2019-12-10 | 2020-04-24 | 华南理工大学 | Rolling bearing state monitoring method and system based on convolutional neural network model |
| CN111413075A (en) * | 2020-04-02 | 2020-07-14 | 重庆交通大学 | Fan base bolt loosening diagnosis method of multi-scale one-dimensional convolution neural network |
| CN112113755A (en) * | 2020-09-02 | 2020-12-22 | 西安交通大学 | Mechanical fault intelligent diagnosis method based on deep convolution-kurtosis neural network |
| CN112766200A (en) * | 2021-01-26 | 2021-05-07 | 福州大学 | Planetary gear box small sample fault diagnosis method based on one-dimensional convolutional neural network |
| CN115578365A (en) * | 2022-10-26 | 2023-01-06 | 西南交通大学 | Tooth pitch detection method and equipment for adjacent racks of toothed rail railway |
| CN115578365B (en) * | 2022-10-26 | 2023-06-20 | 西南交通大学 | Method and equipment for detecting tooth pitch of adjacent racks of rack rail |
| CN116292367A (en) * | 2023-03-22 | 2023-06-23 | 山东科技大学 | Power plant fan system abnormal condition detection method based on one-dimensional convolution |
| CN116292367B (en) * | 2023-03-22 | 2023-11-10 | 山东科技大学 | Power plant fan system abnormal condition detection method based on one-dimensional convolution |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109029974A (en) | A kind of epicyclic gearbox fault detection method based on one-dimensional convolutional neural networks | |
| Wu et al. | Intelligent fault diagnosis of rotating machinery based on one-dimensional convolutional neural network | |
| Shao et al. | A deep learning approach for fault diagnosis of induction motors in manufacturing | |
| Wu et al. | Rub-impact fault diagnosis of rotating machinery based on 1-D convolutional neural networks | |
| CN108830127A (en) | A kind of rotating machinery fault feature intelligent diagnostic method based on depth convolutional neural networks structure | |
| CN103398843B (en) | Based on the epicyclic gearbox sun gear Fault Classification of many classification Method Using Relevance Vector Machines | |
| CN112418013A (en) | Complex working condition bearing fault diagnosis method based on meta-learning under small sample | |
| Zhong et al. | Multi-fault rapid diagnosis for wind turbine gearbox using sparse Bayesian extreme learning machine | |
| CN108827605A (en) | A kind of mechanical breakdown characteristic automatic extraction method based on improvement sparseness filtering | |
| Huang et al. | Residual gated dynamic sparse network for gearbox fault diagnosis using multisensor data | |
| CN107392304A (en) | A kind of Wind turbines disorder data recognition method and device | |
| Di et al. | Ensemble deep transfer learning driven by multisensor signals for the fault diagnosis of bevel-gear cross-operation conditions | |
| Li et al. | Joint attention feature transfer network for gearbox fault diagnosis with imbalanced data | |
| CN105590144A (en) | Wind speed prediction method and apparatus based on NARX neural network | |
| Huang et al. | Deep residual networks-based intelligent fault diagnosis method of planetary gearboxes in cloud environments | |
| CN112146882B (en) | Bearing fault diagnosis method based on transfer learning vibration signal image recognition | |
| CN112113755A (en) | Mechanical fault intelligent diagnosis method based on deep convolution-kurtosis neural network | |
| Zheng et al. | An end-to-end CNN-BiLSTM attention model for gearbox fault diagnosis | |
| Grezmak et al. | Multi-stream convolutional neural network-based fault diagnosis for variable frequency drives in sustainable manufacturing systems | |
| CN112160877A (en) | Fan bearing fault diagnosis method based on SFA and CNN | |
| CN108844735A (en) | Epicyclic gearbox fault detection method based on convolution coder and Min formula distance | |
| Li et al. | Early gear pitting fault diagnosis based on bi-directional LSTM | |
| Xing et al. | Multi-sensor signals with parallel attention convolutional neural network for bearing fault diagnosis | |
| Fan et al. | A novel metric-based model with the ability of zero-shot learning for intelligent fault diagnosis | |
| CN107527093B (en) | Wind turbine generator running state diagnosis method and device |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181218 |