CN108844735A - Epicyclic gearbox fault detection method based on convolution coder and Min formula distance - Google Patents
Epicyclic gearbox fault detection method based on convolution coder and Min formula distance Download PDFInfo
- Publication number
- CN108844735A CN108844735A CN201810653611.4A CN201810653611A CN108844735A CN 108844735 A CN108844735 A CN 108844735A CN 201810653611 A CN201810653611 A CN 201810653611A CN 108844735 A CN108844735 A CN 108844735A
- Authority
- CN
- China
- Prior art keywords
- fault
- layer
- feature vector
- epicyclic gearbox
- feature
- 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
-
- 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
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Acoustics & Sound (AREA)
- Testing Of Devices, Machine Parts, Or Other Structures Thereof (AREA)
Abstract
The present invention relates to a kind of epicyclic gearbox fault detection method based on convolution coder and Min formula distance, this method comprises the following steps：(1) epicyclic gearbox vibration signal is obtained；(2) epicyclic gearbox vibration signal is input to Feature Selection Model trained in advance, the Feature Selection Model is one-dimensional convolution coder；(3) Feature Selection Model extracts fault feature vector, Min formula distance of fault feature vector Yu known fault category feature vector is sought, according to Min formula apart from acquiring size fault type.Compared with prior art, the present invention can achieve very high diagnosis accuracy, and this method can adapt to the needs of real-time, have good practical value.
Description
Technical field
The present invention relates to power system device maintenance areas, are based on convolution coder and Min formula distance more particularly, to one kind
Epicyclic gearbox fault detection method.
Background technique
The status monitoring of power system device is a part indispensable in electric system.Epicyclic gearbox is as wind-force
The important transmission device of generator, it is made of sun gear, planetary gear, planet wheel bearing, gear ring and planet carrier, can be compact
Space in obtain high torque ratio.Due to its complicated vibration transmission path, multiple tooth engagement effect, signal it is non-stationary
And the reasons such as working background noise is big, cause its fault diagnosis to have the characteristics that itself and difficult point.The fault diagnosis of gear-box
It is typically based on two aspects：Fault diagnosis based on index and the fault diagnosis based on data-driven.
Fault diagnosis based on index needs to carry out physical modeling to research object, obtains the time domain and frequency domain of object data
Etc. indexs, then fault diagnosis is achieved the purpose that by each index of comparative analysis.But it largely makes an uproar since gear-box signal generally comprises
Sound, signal is there are modulation phenomenon, the disadvantages of to diagnose fault that there are accuracy using index low, stability is poor.
With network and hardware technological development, the acquisition and storage of data are more convenient, and the failure based on data-driven is examined
It is broken into as a new developing direction.Different with based on calibration method is referred to, data-driven method is not necessarily to carry out physical modeling, but
Processing and analysis appropriate are carried out by data of the intelligent algorithm to acquisition to extract data characteristics, to find between data
Rule.
Intelligence learning algorithm includes supervised learning and unsupervised learning.Supervised learning needs data sensitivity height
The disadvantages of a large amount of training data.Unsupervised learning can automatically extract data characteristics in the case where no label, and algorithm adapts to
Property is stronger, has become the new research hotspot of area of pattern recognition.
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.Traditional intelligent algorithm category
In supervised learning algorithm, a large amount of manpower is needed to remove production data label, not only wasting manpower and material resources, the model of generation adapts to model
Enclose that there is also limitations.
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 convolutional encoding
The epicyclic gearbox fault detection method of device and Min formula distance.
The purpose of the present invention can be achieved through the following technical solutions：
A kind of epicyclic gearbox fault detection method based on convolution coder and Min formula distance, this method include following step
Suddenly：
(1) epicyclic gearbox vibration signal is obtained；
(2) epicyclic gearbox vibration signal is input to Feature Selection Model trained in advance, the feature extraction mould
Type is one-dimensional convolution coder；
(3) Feature Selection Model extracts fault feature vector, seek fault feature vector and known fault category feature to
Min formula distance of amount, according to Min formula apart from acquiring size fault type.
The one-dimensional convolution coder includes sequentially connected input layer, the first convolutional layer, the first pond layer, volume Two
Lamination, the second pond layer, feature vector layer and output layer, the input layer are used for input planet gear case vibration signal, institute
The first convolutional layer, the first pond layer, the second convolutional layer and the second pond layer stated successively carry out adopting under convolution-down-sampling-convolution-
Sample operation, the Feature Mapping figure head and the tail connection of the second pond layer is formed fault feature vector by the feature vector layer, described
Output layer fault feature vector is connected entirely and exports class identical with input layer epicyclic gearbox vibration signal dimension
Vibration signal, the class vibration signal infinite tendency epicyclic gearbox vibration signal.
First convolutional layer and the second convolutional layer be specially：
If l layers are convolutional layer in one-dimensional convolutional neural networks, then the calculation formula for corresponding to convolutional layer is：
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 be specially：
If l+1 layers are pond layer in one-dimensional convolutional neural networks, then the calculation formula for corresponding to pond layer is：
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 is specially：
yl+1=f (ul+1)=f (Wl+1xl+1+bl+1),
yl+1Indicate the class vibration signal of output layer output, xl+1Indicate l+1 layers of Feature Mapping, Wl+1Indicate output layer
Weight,Indicate the biasing of output layer, f () is activation primitive.
Step (3) Feature Selection Model extracts fault feature vector：Epicyclic gearbox vibration signal is input to one
Convolution coder is tieed up, the fault feature vector of one-dimensional convolution coder feature vector layer output is extracted.
Step (3) Min formula distance obtains in the following way：
If fault feature vector is a (x11, x12... ..., x1n), the corresponding known fault category feature of a certain fault type
Vector is b (x21, x22... ..., x2n), then Min formula distance of fault feature vector and the known fault category feature vector is d12：
Wherein, p is Minkowski index.
Step (3) is specially apart from acquiring size fault type according to Min formula：Respectively obtain fault feature vector with it is multiple
Min formula distance of known fault category feature vector, choose Min formula corresponding to the minimum value known fault category feature vector therefore
Hinder fault type of the type as epicyclic gearbox.
Compared with prior art, the invention has the advantages that：
(1) present invention realizes the extraction of fault feature vector using one-dimensional convolution coder, and then combines formula distance in Min real
The diagnosis of existing fault type, convolution autocoder are a kind of efficiently unsupervised recognition methods that developed recently gets up, it is combined
The advantages of convolutional neural networks and autocoder respectively, the local sensing and weight of convolution algorithm share characteristic and improve meter
Calculate efficiency；Autocoding makes algorithm training process unsupervisedization, does not need a large amount of time and manpower removes production data label,
Unsupervised learning also reduces over-fitting to a certain extent.
(2) present invention can achieve very high diagnosis accuracy, and this method can adapt to the needs of real-time, have
Good practical value.
Detailed description of the invention
Fig. 1 is that the present invention is based on the flow chart elements of convolution coder and the epicyclic gearbox fault detection method of Min formula distance
Figure；
Fig. 2 is the structural schematic diagram of the one-dimensional convolution coder of the present invention；
Fig. 3 is the one-dimensional convolution coder training flow chart of the present invention；
Fig. 4 is epicyclic gearbox planetary gear vibrational waveform figure；
Fig. 5 is vibratory impulse waveform；
Fig. 6 is the feature vector chart under the effect of the first pond layer；
Fig. 7 is variation of the mean square error with the number of iterations；
Fig. 8 is epicyclic gearbox mean failure rate feature vector chart；
Fig. 9 is that the training of different Minkowski index p is accurate；
Figure 10 is the fault feature vector figure of test data.
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 convolution coder and Min formula distance, this method
Include the following steps：
(1) epicyclic gearbox vibration signal is obtained；
(2) epicyclic gearbox vibration signal is input to Feature Selection Model trained in advance, Feature Selection Model one
Tie up convolution coder；
(3) Feature Selection Model extracts fault feature vector, seek fault feature vector and known fault category feature to
Min formula distance of amount, according to Min formula apart from acquiring size fault type.
As shown in Fig. 2, one-dimensional convolution coder includes 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, input layer be used for input planet gear case vibration signal, first
Convolutional layer, the first pond layer, the second convolutional layer and the second pond layer successively carry out convolution-down-sampling-convolution-down-sampling operation,
The Feature Mapping figure head and the tail connection of second pond layer is formed fault feature vector by feature vector layer, output layer by fault signature to
Amount is connected entirely and is exported class vibration signal identical with input layer epicyclic gearbox vibration signal dimension, class vibration signal without
Limit approach epicyclic gearbox vibration signal.
First convolutional layer and the second convolutional layer are specially：
If l layers are convolutional layer in one-dimensional convolutional neural networks, then the calculation formula for corresponding to convolutional layer is：
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.
First pond layer and the second pond layer are specially：
If l+1 layers are pond layer in one-dimensional convolutional neural networks, then the calculation formula for corresponding to pond layer is：
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, and activation primitive uses ReLU activation primitive, ReLU activation primitive
It is as follows：
F (x)=max (0, x).
Output layer is specially：
yl+1=f (ul+1)=f (Wl+1xl+1+bl+1),
yl+1Indicate the class vibration signal of output layer output, xl+1Indicate l+1 layers of Feature Mapping, Wl+1Indicate output layer
Weight,Indicate the biasing of output layer, f () is activation primitive.
This above-mentioned one-dimensional convolution coder is named as CAE-c1 (k1)-s1-c2 (k2)-s2.Wherein, c1 (k1) and c2 (k2)
It respectively indicates first convolutional layer to handle to obtain by the c1 convolution kernels having a size of k1 × 1, second convolutional layer is by c2
Convolution kernel having a size of k2 × 1 handles to obtain, and s1 and s2 indicate the pond factor.The one-dimensional each convolution operation of convolution coder is mobile
Step number is 1, and pondization is using average pondization operation.Using BP back-propagation algorithm training network, training flow chart is as shown in Figure 3.
Step (3) Feature Selection Model extracts fault feature vector：Epicyclic gearbox vibration signal is input to one
Convolution coder is tieed up, the fault feature vector of one-dimensional convolution coder feature vector layer output is extracted.
Step (3) Min formula distance obtains in the following way：
If fault feature vector is a (x11, x12... ..., x1n), the corresponding known fault category feature of a certain fault type
Vector is b (x21, x22... ..., x2n), then Min formula distance of fault feature vector and the known fault category feature vector is d12：
Wherein, p is Minkowski index.
Step (3) is specially apart from acquiring size fault type according to Min formula：Respectively obtain fault feature vector with it is multiple
Min formula distance of known fault category feature vector, choose Min formula corresponding to the minimum value known fault category feature vector therefore
Hinder fault type of the type as epicyclic gearbox.
The data source of this method validity is verified in gear case of blower analog platform.Test planetary gear is mounted on speed change
In case gear-box, accelerometer is mounted on gear box casing to measure vibration signal.It can be changed by speed control
Motor speed, speed setting range are 0~60Hz.The sample frequency of signal is 12kHz.Failure planet gear distress
Including abrasion, spot corrosion, broken teeth failure.Planetary gear health status includes：Normally, abrasion, spot corrosion and broken teeth situation, work as drive
Dynamic motor speed is 40Hz, and the planetary gear time domain waveform of acquisition is as shown in Figure 4.
Motor speed is set as 1800r/min, when sample rate is 12000, different health status are taken respectively
Data.Different health status primary planet wheel data acquire 100 groups of data, altogether include 400 groups of data samples.
1. convolution autocoder parameter is chosen：
Rule of thumb, choose c1 value is that 4, c2 value is 8, s1 value and s2 value is 10 to first choice.
When each sample contains at least one impact signal, this section of sample can more effective representing fault feature.Work as sample
This can make each sample contain at least one impact signal comprising 2400 data points, therefore the input ruler of convolution autocoder
Degree takes 2400 × 1.
As shown in figure 5, being the impact signal enlarged drawing of a certain broken teeth signal, an impact signal known to observation is opened from impact
Begin to impact to terminate comprising about 20 data points.During convolution kernel local shape factor, it is believed that when convolution kernel completely covers
Lid terminates since impact to impact, can effectively extract shock characteristic, therefore selecting k1 value is 21.By first layer convolution sum
After pond, obtains a group profile of the signal under the effect of the first pond layer and reach, figure is as shown in Figure 6, it is seen that one
A impact includes about 10 data points, and k1 value takes 9.Therefore network structure is CAE-4 (21) -10-8 (9) -10.
2. the feature extraction of convolution autocoder：
The arrangement of sample data such as table 1.50% sample is randomly selected to train network, remaining 50% is the most to be detected
Sample carry out the accuracy of test network.
1 sample arrangement of table
Convolution autocoder is trained using training data.Fig. 7 is that the variation of mean square error in iterative process is bent
Line, it can be seen that when the number of iterations reaches 1000 times, the variation of mean square error tends towards stability, therefore can consider network at this time
It tends towards stability, choosing the number of iterations is 1000 as iteration stopping condition.
The set of the feature vector of training data is obtained by trained network.Since there are noises for experimental data, extract
Feature vector similarly contain noise factor, noise is generally the random value of very little, in order to reduce the influence of noise, with feature
Based on maximum value in vector set, retain maximum 90% feature vector value, so that the influence of noise is reduced, it is every kind prominent
The feature of health status is distinguished.
There is the feature vector set of crack failure and broken teeth failure to be averaged respectively on normal condition, wear-out failure, tooth root
Value, obtains averaged feature vector, as shown in Figure 8.It can be seen that the average characteristics difference between them is obvious.Illustrate using average
Vector can efficiently differentiate different fault types, and the averaged feature vector of extraction can be used as the index for judging healthy type
Vector.
3. the selection of Minkowski index：
According to p=1~10, Min formula distance of the training sample feature vector with label and indicator vector is sought, is obtained
Training precision variation under different p values is as shown in Figure 9.When p value is 4, training precision highest, therefore take p=4.
4. Gernral Check-up：
The fault feature vector figure of the data is obtained using one group of test data as input, as shown in Figure 10.From amplitude and
It is observed on waveform shape, the waveform and the crannied known fault category feature vector of tooth root are the most similar, which may
Belonging to tooth root has crack failure.
Calculate Min formula distance between this feature vector sum indicator vector, obtain its have with normal, abrasion, tooth root crack and
Min formula distance of the indicator vector of broken teeth failure is respectively 2.61,6.90,2.09,2.61.Judge that it has crack failure for tooth root,
Verified, which is that tooth root has crack fault data really.
All 200 groups of test datas are obtained into the classification situation of table 3 by feature extraction and Min formula Distance Judgment.Final
Accuracy of classifying is 96%.Wherein in 50 groups of data of broken teeth failure, there are 2 groups to be judged as normal, 6 groups are judged as that tooth root has crack event
Barrier, reason may is that broken teeth failure and tooth root have crack failure to belong to local failure.
In operation, computer processor is intel pentium g2030, inside saves as the DDR3 memory of 2GB.1000 institutes of iteration
It takes time about 7 minutes.If 1 second vibration data is inputted trained network, obtains feature vector and formula distance in Min calculates
Total time less than 0.07 second.In view of network only needs training primary, network query function speed can adapt to the need of real-time diagnosis
It wants.
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.Tool
There is 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 (8)
1. a kind of epicyclic gearbox fault detection method based on convolution coder and Min formula distance, which is characterized in that this method
Include the following steps：
(1) epicyclic gearbox vibration signal is obtained；
(2) epicyclic gearbox vibration signal is input to Feature Selection Model trained in advance, the Feature Selection Model is
One-dimensional convolution coder；
(3) Feature Selection Model extracts fault feature vector, seeks fault feature vector and known fault category feature vector
Min formula distance, according to Min formula apart from acquiring size fault type.
2. a kind of epicyclic gearbox fault detection side based on convolution coder and Min formula distance according to claim 1
Method, which is characterized in that the one-dimensional convolution coder include sequentially connected input layer, the first convolutional layer, the first pond layer,
Second convolutional layer, 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 operation of product-down-sampling, the feature vector layer Feature Mapping figure head and the tail of the second pond layer are connected formed fault signature to
Fault feature vector is connected entirely and is exported and input layer epicyclic gearbox vibration signal dimension phase by amount, the output layer
Same class vibration signal, the class vibration signal infinite tendency epicyclic gearbox vibration signal.
3. a kind of epicyclic gearbox fault detection side based on convolution coder and Min formula distance according to claim 2
Method, which is characterized in that first convolutional layer and the second convolutional layer be specially：
If l layers are convolutional layer in one-dimensional convolutional neural networks, then the calculation formula for corresponding to convolutional layer is：
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.
4. a kind of epicyclic gearbox fault detection side based on convolution coder and Min formula distance according to claim 2
Method, which is characterized in that the first pond layer and the second pond layer be specially：
If l+1 layers are pond layer in one-dimensional convolutional neural networks, then the calculation formula for corresponding to pond layer is：
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.
5. a kind of epicyclic gearbox fault detection side based on convolution coder and Min formula distance according to claim 2
Method, which is characterized in that output layer is specially：
yl+1=f (ul+1)=f (Wl+1xl+1+bl+1),
yl+1Indicate the class vibration signal of output layer output, 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.
6. a kind of epicyclic gearbox fault detection side based on convolution coder and Min formula distance according to claim 2
Method, which is characterized in that step (3) Feature Selection Model extracts fault feature vector and is specially：Epicyclic gearbox vibration signal is defeated
Enter the fault feature vector that one-dimensional convolution coder feature vector layer output is extracted to one-dimensional convolution coder.
7. a kind of epicyclic gearbox fault detection side based on convolution coder and Min formula distance according to claim 1
Method, which is characterized in that formula distance in step (3) Min obtains in the following way：
If fault feature vector is a (x11, x12... ..., x1n), the corresponding known fault category feature vector of a certain fault type
For b (x21, x22... ..., x2n), then Min formula distance of fault feature vector and the known fault category feature vector is d12：
Wherein, p is Minkowski index.
8. a kind of epicyclic gearbox fault detection side based on convolution coder and Min formula distance according to claim 1
Method, which is characterized in that step (3) is specially apart from acquiring size fault type according to Min formula：Fault feature vector is obtained respectively
With Min formula distance of multiple known fault category feature vectors, Min formula is chosen apart from minimum value known fault category feature vector institute
Fault type of the corresponding fault type as epicyclic gearbox.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810653611.4A CN108844735A (en) | 2018-06-22 | 2018-06-22 | Epicyclic gearbox fault detection method based on convolution coder and Min formula distance |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810653611.4A CN108844735A (en) | 2018-06-22 | 2018-06-22 | Epicyclic gearbox fault detection method based on convolution coder and Min formula distance |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108844735A true CN108844735A (en) | 2018-11-20 |
Family
ID=64203116
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810653611.4A Pending CN108844735A (en) | 2018-06-22 | 2018-06-22 | Epicyclic gearbox fault detection method based on convolution coder and Min formula distance |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108844735A (en) |
Cited By (4)
| 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 |
| CN112380782A (en) * | 2020-12-07 | 2021-02-19 | 重庆忽米网络科技有限公司 | Rotating equipment fault prediction method based on mixed indexes and neural network |
| CN113075546A (en) * | 2021-03-24 | 2021-07-06 | 河南中烟工业有限责任公司 | Motor vibration signal feature extraction method and system |
| CN113570473A (en) * | 2021-06-25 | 2021-10-29 | 深圳供电局有限公司 | Equipment fault monitoring method and device, computer equipment and storage medium |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1906453A (en) * | 2004-01-21 | 2007-01-31 | 三菱电机株式会社 | Device diagnosis device, freezing cycle device, fluid circuit diagnosis method, device monitoring system, and freezing cycle monitoring system |
| CN104155108A (en) * | 2014-07-21 | 2014-11-19 | 天津大学 | Rolling bearing failure diagnosis method base on vibration temporal frequency analysis |
| CN107506695A (en) * | 2017-07-28 | 2017-12-22 | 武汉理工大学 | Video monitoring equipment failure automatic detection method |
| CN107844067A (en) * | 2017-12-07 | 2018-03-27 | 国家电网公司 | A kind of gate of hydropower station on-line condition monitoring control method and monitoring system |
-
2018
- 2018-06-22 CN CN201810653611.4A patent/CN108844735A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1906453A (en) * | 2004-01-21 | 2007-01-31 | 三菱电机株式会社 | Device diagnosis device, freezing cycle device, fluid circuit diagnosis method, device monitoring system, and freezing cycle monitoring system |
| CN104155108A (en) * | 2014-07-21 | 2014-11-19 | 天津大学 | Rolling bearing failure diagnosis method base on vibration temporal frequency analysis |
| CN107506695A (en) * | 2017-07-28 | 2017-12-22 | 武汉理工大学 | Video monitoring equipment failure automatic detection method |
| CN107844067A (en) * | 2017-12-07 | 2018-03-27 | 国家电网公司 | A kind of gate of hydropower station on-line condition monitoring control method and monitoring system |
Non-Patent Citations (2)
| Title |
|---|
| 吴怀城 等: ""一种换流变双有载分接开关故障检测方法"", 《电力科学与工程》 * |
| 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 (5)
| 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 |
| CN112380782A (en) * | 2020-12-07 | 2021-02-19 | 重庆忽米网络科技有限公司 | Rotating equipment fault prediction method based on mixed indexes and neural network |
| CN113075546A (en) * | 2021-03-24 | 2021-07-06 | 河南中烟工业有限责任公司 | Motor vibration signal feature extraction method and system |
| CN113570473A (en) * | 2021-06-25 | 2021-10-29 | 深圳供电局有限公司 | Equipment fault monitoring method and device, computer equipment and storage medium |
| CN113570473B (en) * | 2021-06-25 | 2024-02-09 | 深圳供电局有限公司 | Equipment fault monitoring method, device, computer equipment and storage medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108844735A (en) | Epicyclic gearbox fault detection method based on convolution coder and Min formula distance | |
| Li et al. | Understanding and improving deep learning-based rolling bearing fault diagnosis with attention mechanism | |
| Zhao et al. | Deep convolutional neural network based planet bearing fault classification | |
| Xiang et al. | Fault diagnosis of rolling bearing under fluctuating speed and variable load based on TCO spectrum and stacking auto-encoder | |
| Abdul et al. | A hybrid temporal feature for gear fault diagnosis using the long short term memory | |
| CN110334764B (en) | Rotary machine intelligent fault diagnosis method based on integrated depth self-encoder | |
| CN109800861A (en) | A kind of equipment fault recognition methods, device, equipment and computer system | |
| Syed et al. | Feature extraction using Discrete Wavelet Transform for fault classification of planetary gearbox–A comparative study | |
| CN106197999B (en) | A kind of planetary gear method for diagnosing faults | |
| CN111504635B (en) | Planetary gear fault diagnosis method based on differential evolution probability neural network | |
| CN108256556A (en) | Wind-driven generator group wheel box method for diagnosing faults based on depth belief network | |
| CN109029974A (en) | A kind of epicyclic gearbox fault detection method based on one-dimensional convolutional neural networks | |
| Gecgel et al. | Gearbox fault diagnostics using deep learning with simulated data | |
| CN114778112A (en) | Audio identification and fault diagnosis method for mechanical fault of wind turbine generator system | |
| CN104330255A (en) | Gear fault diagnosing method based on multi-sensor information fusion | |
| CN112229633A (en) | Fan bearing fault diagnosis method based on multivariate feature fusion | |
| Du et al. | Fault diagnosis of rotating machinery based on 1D–2D joint convolution neural network | |
| CN112364706A (en) | Small sample bearing fault diagnosis method based on class imbalance | |
| Chen et al. | A visualized classification method via t-distributed stochastic neighbor embedding and various diagnostic parameters for planetary gearbox fault identification from raw mechanical data | |
| Rao et al. | A new strategy for rotating machinery fault diagnosis under varying speed conditions based on deep neural networks and order tracking | |
| CN114034481A (en) | Fault diagnosis system and method for rolling mill gearbox | |
| Han et al. | An intelligent fault diagnosis method of variable condition gearbox based on improved DBN combined with WPEE and MPE | |
| CN114091525A (en) | Rolling bearing degradation trend prediction method | |
| CN113537152A (en) | Flow field state fault detection method based on deep neural network | |
| CN113916535B (en) | Bearing diagnosis method, system, equipment and medium based on time frequency and CNN |
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: 20181120 |
|
| RJ01 | Rejection of invention patent application after publication |