CN107013449B - The method and system of voice signal identification compressor fault based on deep learning - Google Patents
The method and system of voice signal identification compressor fault based on deep learning Download PDFInfo
- Publication number
- CN107013449B CN107013449B CN201710253920.8A CN201710253920A CN107013449B CN 107013449 B CN107013449 B CN 107013449B CN 201710253920 A CN201710253920 A CN 201710253920A CN 107013449 B CN107013449 B CN 107013449B
- Authority
- CN
- China
- Prior art keywords
- data
- compressor
- layer
- framing
- sampled points
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 45
- 238000013135 deep learning Methods 0.000 title claims abstract description 17
- 238000009432 framing Methods 0.000 claims abstract description 37
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 28
- 238000012549 training Methods 0.000 claims abstract description 21
- 230000005236 sound signal Effects 0.000 claims abstract description 11
- 238000003062 neural network model Methods 0.000 claims abstract description 5
- 230000011218 segmentation Effects 0.000 claims description 8
- 238000013528 artificial neural network Methods 0.000 claims description 5
- 230000004913 activation Effects 0.000 claims description 4
- 238000012795 verification Methods 0.000 claims description 3
- 230000004888 barrier function Effects 0.000 claims description 2
- 238000005070 sampling Methods 0.000 claims description 2
- 238000004458 analytical method Methods 0.000 description 6
- 238000000605 extraction Methods 0.000 description 5
- 230000008569 process Effects 0.000 description 5
- 238000013461 design Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000003672 processing method Methods 0.000 description 3
- 230000015556 catabolic process Effects 0.000 description 2
- 238000001514 detection method Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 235000013399 edible fruits Nutrition 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000007781 pre-processing Methods 0.000 description 1
- 238000012360 testing method Methods 0.000 description 1
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F04—POSITIVE - DISPLACEMENT MACHINES FOR LIQUIDS; PUMPS FOR LIQUIDS OR ELASTIC FLUIDS
- F04B—POSITIVE-DISPLACEMENT MACHINES FOR LIQUIDS; PUMPS
- F04B51/00—Testing machines, pumps, or pumping installations
Landscapes
- Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- General Engineering & Computer Science (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
The invention discloses the method and system that the voice signal based on deep learning identifies compressor fault；Sound signal collecting：The operational sound of compressor is acquired using microphone；Set sample rate M；Data overlap framing：The M sampled point that each second acquires is subjected to overlapping framing；Increase data volume, be that the input of convolutional neural networks increases sample, the length per frame data is set as N sampled points, and framing step-length is P sampled points, is overlapped N P sampled points；To which the data of each second can be divided into the new data set of overlapping；Neural network model is built：Convolutional neural networks are built；Model training：The voice data acquired in advance under four kinds of working conditions of compressor is trained convolutional neural networks after dividing with data overlap framing method；Fault identification：The compressor real-time working sound of acquisition is input in trained convolutional neural networks model after dividing according to data framing method, exports current working condition.
Description
Technical field
The present invention relates to the method and system that the voice signal based on deep learning identifies compressor fault.
Background technology
The use of analysis of vibration signal diagnosing compressor fault is current main stream approach in practical engineering application, it is logical
The vibration signal of acquisition compressor in the process of running is crossed, breakdown judge is carried out after carrying out feature extraction.Common vibration signal
Processing method includes mainly time-domain analysis, frequency-domain analysis, time frequency analysis, multiresolution analysis, empirical modal analysis etc..Such methods
The characteristics of be to extract big measure feature using classical signal processing method, then use machine learning to carry out tagsort, it is this kind of
Method needs to be grasped the signal processing method of specific area and is combined with Practical Project, more demanding to technical staff.
At the same time, complicated Feature Engineering is also one of the difficult point of such method, and how extracting can effectively distinguish
The feature of different faults type is a cumbersome and difficult task, and good feature is very big on recognition effect influence, in fact,
Many Fault Classifications will lean on good feature to be identified, if can be got not from collected initial data automatically
Same feature, will greatly reduce Feature Engineering workload.
Invention content
The purpose of the present invention is exactly to solve the above-mentioned problems, to provide the voice signal identification compressor based on deep learning
The method and system of failure, automatically from the feature extracted in original sound signal under different faults, are simplified using deep learning
Identification process so that layman passes through contactless sound detection, identification under the premise of not needing professional knowledge
It has the failure of compressor.
To achieve the goals above, the present invention adopts the following technical scheme that：
The method of voice signal identification compressor fault based on deep learning, including：
Step (1)：Sound signal collecting：The operational sound of compressor is acquired using microphone；Set sample rate
M；
Step (2)：Data overlap framing：The M sampled point that each second acquires is subjected to overlapping framing；Increase data volume,
Increase sample for the input of convolutional neural networks, the length per frame data is set as N sampled points, and framing step-length is P sampled points, weight
Folded N-P sampled points；It can be divided into the data of each secondFrame has the new data set of overlapping；Indicate downward
Rounding；
Step (3)：Neural network model is built：Convolutional neural networks are built；
Step (4)：Model training：The voice data acquired in advance under four kinds of working conditions of compressor, process and step
(2) after same data overlap framing method segmentation, convolutional neural networks are trained；
Step (5)：Fault identification：By the compressor real-time working sound of acquisition, according to the same data framing of step (2)
It after method segmentation, is input in trained convolutional neural networks model, the current working condition of output compressor.
The sample rate of the step (1) is 50KHz.
The step of step (2) is：The 50K sampled point that each second is acquired is again regular, increases data volume, for volume
The input of product neural network increases sample, and the length per frame data is set as 10K sampled points, framing step-length 2K sampled points, overlapping
8K sampled points, the 1st frame data are 0K to 10K sampled points；2nd frame data are 2K to 12K sampled points；3rd frame data
For 4K to 14K sampled points；4th frame data are 6K- 16K sampled points；5th frame data are 8K to 18K sampled points；
And so on, the 20th frame data are 38K- 48K sampled points；21st frame data are 40K- 50K sampled points；Each second
Data, which can be divided into 21 frames, the new data set of overlapping.
The step of step (3) is：
First layer is input layer, uses 10K sampled point of original sound signal；
The second layer is convolutional layer, and convolution kernel size is 32, and convolution nuclear volume is 16, and convolution step-length is 8, is activated using ReLU
Function, the subsequent maximum pond layer of addition, pond size 4, step-length 2, the second layer network terminate, and network output is connected to third layer；
In third layer convolutional layer, convolution kernel size is 3, convolution nuclear volume 32, and convolution step-length is 1, and letter is activated using ReLU
Number, the subsequent maximum pond layer of addition, pond size 2, step-length 1, third layer network terminate, and network output is connected to the 4th layer；
In 4th layer of convolutional layer, convolution kernel size is 3, convolution nuclear volume 32, and convolution step-length is 1, and letter is activated using ReLU
Maximum pond layer is then added in number, and pond size 2, step-length 1, four-layer network network terminates, and network output carries out flattening operation；
Layer 5 is full articulamentum, data dimension 64；
Layer 6 is output layer, and output dimension is consistent with setting fault type number, uses softmax activation primitives.
The step of step (4) is：Four kinds of working conditions include：Normal operation, the event of intake valve failure, air outlet valve
Barrier and bearing fault；
Convolutional neural networks are trained using the voice data acquired in advance under four kinds of working conditions of compressor：
Each state acquires 5 minutes initial data under 50K sample rates, and data are divided according to 50K sampled points, are obtained
A length of 1 second data at 300 groups；
Then per second data according to the data framing method of step (2), being divided into 21 frames again has the initial data of overlapping,
It can obtain within 5 minutes 6300 groups of initial data for having overlapping；
The lower 6300 groups of data of each state, then share 25200 groups of data under 4 kinds of states；
By data set according to 2:1 ratio random division is that training set collects with verification, when network training, by every group of data
10000 sampled points are inputted with corresponding Status Type, are carried out more wheel traversals using error backpropagation algorithm, are reached setting work
When making Status Type recognition accuracy, terminate training.
The step of step (5) is：
1 second original sound real time data is acquired, is then divided into real time data according to the data framing method of step (2)
Have 21 frame data of overlapping, per frame data 10000 sampled points, by 21 frame data input trained convolutional neural networks model into
Row fault identification exports 21 judging results, then, each type of number in 21 output types is counted, by maximum number
Type exported as final recognition result.
The system of voice signal identification compressor fault based on deep learning, including：
Sound signal collecting module：The operational sound of compressor is acquired using microphone；Set sample rate M；
Data overlap framing module：The M sampled point that each second acquires is subjected to overlapping framing；Increase data volume, for volume
The input of product neural network increases sample, and the length per frame data is set as N sampled points, and framing step-length is P sampled points, is overlapped N-
P sampled points；It can be divided into the data of each secondFrame has the new data set of overlapping；Expression takes downwards
It is whole；
Neural network model builds module：Convolutional neural networks are built；
Model training module：The voice data acquired in advance under four kinds of working conditions of compressor, by same with step (2)
After the data overlap framing method segmentation of sample, convolutional neural networks are trained；
Fault identification module：By the compressor real-time working sound of acquisition, according to the same data framing method of step (2)
It after segmentation, is input in trained convolutional neural networks model, the current working condition of output compressor.
Beneficial effects of the present invention：
Voice signal when 1 equipment is run is the important information for reflecting operating status, and veteran engineer can pass through
Trouble location and type are judged in the sense of hearing, in essence, voice signal be caused by vibrating, but Relative Vibration signal and
Speech, voice signal are easier to acquire.Under normal circumstances, sensor needs are directly contacted with compressor when acquiring vibration signal, and
It is contacted completely without with equipment when collected sound signal, moreover, using this non-contact sensor of voice signal, to equipment sheet
Body damages smaller.
2 voice signals of the invention by acquiring compressor operating, using deep learning, automatically from original sound signal
Extract the feature under different faults, simplify identification process so that layman under the premise of not needing professional knowledge,
By contactless sound detection, the failure of compressor is identified.
3 use compressor fault recognition methods of the present invention, can reduce technical staff and be identified in compressor fault
The professional knowledge that aspect needs, the method for voice signal when directly acquiring compressor operating using microphone that the present invention uses
It is more convenient, smaller is influenced on running equipment.After model training is good, technical staff only need to be by microphone close to operation
In compressor, acquire its voice signal, the computer of the other end can observe directly its corresponding operating status.
The expansion of 4 present invention is very strong, can be simply by increasing new state when needing the state detected to increase
The method of training set is expanded, and final output layer also need to only increase corresponding states output, theoretically, can be increased arbitrary more
The state of type, as long as training set is sufficiently large.
Description of the drawings
Fig. 1 is the data preprocessing method that this programme uses, which illustrates the framing side of data by taking 1s data volumes as an example
Method；
Fig. 2 is the basic structure of convolutional neural networks；
Fig. 3 is network training and identification process.
Specific implementation mode
The invention will be further described with embodiment below in conjunction with the accompanying drawings.
The present invention mainly uses two methods, and the identification of existing compressor fault is improved and is simplified.First, this hair
The bright direct voice signal that compressor is acquired using microphone, this simplifies the acquisition methods of signal, secondly, by the present invention in that
With the convolutional neural networks in deep learning, automatically extracting for fault signature is realized, avoids complicated Feature Engineering.This
Sample, even layman can also use trained identification model to carry out breakdown judge.
Biggest advantage of the present invention is, the Automatic Feature Extraction function by making full use of convolutional neural networks outstanding and non-
Linear Mapping function so that data prediction is simple, while characteristic of human nature being avoided to extract.Core of the invention is depth convolution god
Design through network.
As shown in figure 3, technical solution of the present invention includes the content of four aspects：Data acquire, and data prediction, network is set
Meter and model training, fault identification.
Data time related cost is low, implements easy microphone and is acquired, sample rate 50KHz.
Data prediction need to only carry out having the data framing of overlapping to work, and be exactly the 50K for acquiring each second specifically
A sampled point is again regular, is that the input of convolutional neural networks increases sample, embodiment is as schemed the purpose is to increase data volume
Shown in 1.In the present invention, the length per frame data is set as 10K sampled points, and framing step-length 2K sampled points (are overlapped 8K samplings
Point), in this way, the data of each second, which can be divided into 21 frames, the new data set of overlapping.
The present invention directly carries out simple framing operation using the original sound data of acquisition to be identified, and do not wrap
Containing feature extraction work because the convolutional neural networks that the present invention uses can carry out feature extraction automatically, this be the present invention with
The maximum of other any relative identifying methods premised on feature extraction is different.
The design of convolutional neural networks, convolutional neural networks basic structure such as Fig. 2 institutes that the present invention uses are carried out in next step
Show, this structure design is simple, and effect is fine.It is specific as follows per layer network parameter：
First layer is input layer, uses 10K sampled point of original sound signal；
The second layer is convolutional layer, and convolution kernel size is 32, and convolution nuclear volume is 16, and convolution step-length is 8, is activated using ReLU
Function, the subsequent maximum pond layer of addition, pond size 4, step-length 2, the second layer network terminate, and network output is connected to third layer；
In third layer convolutional layer, convolution kernel size is 3, convolution nuclear volume 32, and convolution step-length is 1, and letter is activated using ReLU
Number, the subsequent maximum pond layer of addition, pond size 2, step-length 1, third layer network terminate, and network output is connected to the 4th layer；
4th layer of all parameters of convolution layer network are consistent with third layer, the difference is that network output will carry out flattening operation；
Layer 5 is full articulamentum, data dimension 64；
Layer 6 is output layer, and output dimension is consistent with fault type number, is 4 in the present invention, and letter is activated using softmax
Number.
It is the specific design parameter for the depth convolutional neural networks that the present invention uses above, there it can be seen that Home Network
Network is clear in structure simple.
After designing network, need to be trained network, present invention is generally directed to four kinds of states of compressor into
Row training, including：It runs well, intake valve failure, air outlet valve failure, bearing fault.
Training data uses the voice data under prior collected four kinds of states.Specifically, each state is adopted in 50K
5 minutes initial data are acquired under sample rate, and data are divided according to 50K sampled points, obtain a length of 1 second data at 300 groups, then
Per second data according to aforementioned data framing method, being divided into 21 frames again has the initial data of overlapping, can obtain within such 5 minutes
6300 groups of initial data for having overlapping.
The lower 6300 groups of data of each state, then share 25200 groups of data under 4 kinds of states.By data set according to 2:1 ratio with
Machine is divided into training set and collects with verification, when network training, 10000 sampled points of every group of data and corresponding Status Type are defeated
Enter, more wheel traversals are carried out using classical error backpropagation algorithm.Finally, the present invention realizes training on data set used
98% or more accuracy rate on collection, test set.
In the fault identification stage, the present invention uses following methods：
First, 1 second original sound data is acquired, then data are divided into 21 frame numbers of overlapping according to preceding method
According to per 10000 sampled point of frame data, by the trained model progress fault identification of 21 frame data input, output 21 judges knot
Then fruit counts each type of number in 21 output types, the maximum number of type is defeated as final recognition result
Go out.
Above-mentioned, although the foregoing specific embodiments of the present invention is described with reference to the accompanying drawings, not protects model to the present invention
The limitation enclosed, those skilled in the art should understand that, based on the technical solutions of the present invention, those skilled in the art are not
Need to make the creative labor the various modifications or changes that can be made still within protection scope of the present invention.
Claims (7)
1. the method for the voice signal identification compressor fault based on deep learning, characterized in that including：
Step (1)：Sound signal collecting：The operational sound of compressor is acquired using microphone；Set sample rate M；
Step (2)：Data overlap framing：The M sampled point that each second acquires is subjected to overlapping framing；Increase data volume, for volume
The input of product neural network increases sample, and the length per frame data is set as N sampled points, and framing step-length is P sampled points, is overlapped N-
P sampled points；It can be divided into the data of each secondFrame has the new data set of overlapping；Expression takes downwards
It is whole；
Step (3)：Neural network model is built：Convolutional neural networks are built；
Step (4)：Model training：The voice data acquired in advance under four kinds of working conditions of compressor, by same with step (2)
After the data overlap framing method segmentation of sample, convolutional neural networks are trained；Four kinds of working conditions include：Normal fortune
Turn, intake valve failure, air outlet valve failure and bearing fault；
Step (5)：Fault identification：By the compressor real-time working sound of acquisition, according to the same data framing method of step (2)
It after segmentation, is input in trained convolutional neural networks model, the current working condition of output compressor.
2. the method for the voice signal identification compressor fault based on deep learning as described in claim 1, characterized in that
The sample rate of the step (1) is 50KHz.
3. the method for the voice signal identification compressor fault based on deep learning as described in claim 1, characterized in that
The step (2) the specific steps are：The 50K sampled point that each second is acquired is again regular, increases data volume, for volume
The input of product neural network increases sample, and the length per frame data is set as 10K sampled points, framing step-length 2K sampled points, overlapping
8K sampled points, the 1st frame data are 0K to 10K sampled points；2nd frame data are 2K to 12K sampled points；3rd frame data
For 4K to 14K sampled points；4th frame data are 6K- 16K sampled points；5th frame data are 8K to 18K sampled points；
And so on, the 20th frame data are 38K- 48K sampled points；21st frame data are 40K- 50K sampled points；Each second
Data, which can be divided into 21 frames, the new data set of overlapping.
4. the method for the voice signal identification compressor fault based on deep learning as described in claim 1, characterized in that
The step (3) the specific steps are：
First layer is input layer, uses 10K sampled point of original sound signal；
The second layer is convolutional layer, and convolution kernel size is 32, and convolution nuclear volume is 16, and convolution step-length is 8, and letter is activated using ReLU
Number, the subsequent maximum pond layer of addition, pond size 4, step-length 2, the second layer network terminate, and network output is connected to third layer；
In third layer convolutional layer, convolution kernel size is 3, and convolution nuclear volume 32, convolution step-length is 1, using ReLU activation primitives, with
The maximum pond layer of addition, pond size 2, step-length 1, third layer network terminate afterwards, and network output is connected to the 4th layer；
In 4th layer of convolutional layer, convolution kernel size is 3, and convolution nuclear volume 32, convolution step-length is 1, using ReLU activation primitives, with
Maximum pond layer is added afterwards, pond size 2, step-length 1, four-layer network network terminates, and network output carries out flattening operation；
Layer 5 is full articulamentum, data dimension 64；
Layer 6 is output layer, and output dimension is consistent with setting fault type number, uses softmax activation primitives.
5. the method for the voice signal identification compressor fault based on deep learning as described in claim 1, characterized in that
The step (4) the specific steps are：Using the voice data acquired in advance under four kinds of working conditions of compressor to volume
Product neural network is trained：
Each state acquires 5 minutes initial data under 50K sample rates, and data are divided according to 50K sampled points, obtain 300 groups
1 second data of Shi Changwei；
Then per second data according to the data framing method of step (2), being divided into 21 frames again has an initial data of overlapping, 5 points
Clock can obtain 6300 groups of initial data for having overlapping；
The lower 6300 groups of data of each state, then share 25200 groups of data under 4 kinds of states；
By data set according to 2:1 ratio random division is that training set collects with verification, when network training, by the 10000 of every group of data
A sampled point is inputted with corresponding Status Type, is carried out more wheel traversals using error backpropagation algorithm, is reached setting work shape
When state type identification accuracy rate, terminate training.
6. the method for the voice signal identification compressor fault based on deep learning as described in claim 1, characterized in that
The step (5) the specific steps are：
1 second original sound real time data is acquired, real time data has then been divided into weight according to the data framing method of step (2)
21 frame data are inputted trained convolutional neural networks model and carry out event by 21 folded frame data per 10000 sampled point of frame data
Barrier identification, exports 21 judging results, then, each type of number in 21 output types is counted, by the maximum number of class
Type is exported as final recognition result.
7. the system of the voice signal identification compressor fault based on deep learning, characterized in that including：
Sound signal collecting module：The operational sound of compressor is acquired using microphone；Set sample rate M；
Data overlap framing module：The M sampled point that each second acquires is subjected to overlapping framing；Increase data volume, is convolution god
Input through network increases sample, and the length per frame data is set as N sampled points, and framing step-length is P sampled points, and overlapping N-P is adopted
Sampling point；It can be divided into the data of each secondFrame has the new data set of overlapping；Indicate downward rounding；
Neural network model builds module：Convolutional neural networks are built；
Model training module：The voice data acquired in advance under four kinds of working conditions of compressor, by with data overlap framing
After the same data overlap framing method segmentation of module, convolutional neural networks are trained；Four kinds of working conditions include：
Normal operation, intake valve failure, air outlet valve failure and bearing fault；
Fault identification module：By the compressor real-time working sound of acquisition, according to the same data point of data overlap framing module
It after frame method segmentation, is input in trained convolutional neural networks model, the current working condition of output compressor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710253920.8A CN107013449B (en) | 2017-04-18 | 2017-04-18 | The method and system of voice signal identification compressor fault based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710253920.8A CN107013449B (en) | 2017-04-18 | 2017-04-18 | The method and system of voice signal identification compressor fault based on deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107013449A CN107013449A (en) | 2017-08-04 |
| CN107013449B true CN107013449B (en) | 2018-08-24 |
Family
ID=59448416
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710253920.8A Active CN107013449B (en) | 2017-04-18 | 2017-04-18 | The method and system of voice signal identification compressor fault based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107013449B (en) |
Families Citing this family (20)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| TWI742312B (en) * | 2017-10-02 | 2021-10-11 | 宏達國際電子股份有限公司 | Machine learning system, machine learning method and non-transitory computer readable medium for operating the same |
| CN109754359B (en) | 2017-11-01 | 2021-12-07 | 腾讯科技（深圳）有限公司 | Pooling processing method and system applied to convolutional neural network |
| CN109959889A (en) * | 2017-12-25 | 2019-07-02 | 通用电气公司 | The monitoring method and system of magnetic resonance imaging system cold head working condition |
| CN108872960A (en) * | 2018-05-09 | 2018-11-23 | 南京航空航天大学 | Sliding framing method for HRRP target identification |
| KR102067601B1 (en) | 2018-06-14 | 2020-01-17 | 엘지전자 주식회사 | Linear compressor and method for controlling linear compressor |
| CN109065030B (en) * | 2018-08-01 | 2020-06-30 | 上海大学 | Convolutional neural network-based environmental sound identification method and system |
| CN109086888B (en) * | 2018-09-05 | 2020-10-27 | 北京机械设备研究所 | Automobile engine fault determination method and device based on voice recognition |
| CN109348086B (en) * | 2018-11-05 | 2020-09-15 | 重庆大学 | Intelligent wireless camera image synchronous identification and compression method |
| CN110360091B (en) * | 2019-06-05 | 2020-08-28 | 合肥通用机械研究院有限公司 | Refrigeration compressor measurement and control system and method based on neural network |
| CN112326210A (en) * | 2019-07-17 | 2021-02-05 | 华北电力大学（保定） | Large motor fault diagnosis method combining sound vibration signals with 1D-CNN |
| CN110852154B (en) * | 2019-09-29 | 2022-10-14 | 广东石油化工学院 | Rolling bearing fault diagnosis method and device based on deep learning and sound waveform images and readable storage medium |
| CN110992985A (en) * | 2019-12-02 | 2020-04-10 | 中国科学院声学研究所东海研究站 | Identification model determining method, identification method and identification system for identifying abnormal sounds of treadmill |
| CN111027260A (en) * | 2019-12-24 | 2020-04-17 | 大连圣力来监测技术有限公司 | Reciprocating compressor fault diagnosis method based on neural network |
| CN111581425A (en) * | 2020-04-28 | 2020-08-25 | 上海鼎经自动化科技股份有限公司 | Equipment sound classification method based on deep learning |
| CN112857767B (en) * | 2021-01-18 | 2022-03-11 | 中国长江三峡集团有限公司 | Hydro-turbo generator set rotor fault acoustic discrimination method based on convolutional neural network |
| CN114396376A (en) * | 2021-12-01 | 2022-04-26 | 石化盈科信息技术有限责任公司 | Fault diagnosis auxiliary method, device, equipment and system for reciprocating compressor |
| CN114371649A (en) * | 2022-01-10 | 2022-04-19 | 江苏大学 | Spray flow regulation and control system and method based on convolutional neural network |
| CN115288994B (en) * | 2022-08-03 | 2024-01-19 | 西安安森智能仪器股份有限公司 | Improved DCGAN-based compressor abnormal state detection method |
| CN115640915B (en) | 2022-12-19 | 2023-03-10 | 成都秦川物联网科技股份有限公司 | Intelligent gas pipe network compressor safety management method and Internet of things system |
| CN117367570B (en) * | 2023-11-02 | 2024-04-12 | 中国人民解放军海军工程大学 | Intelligent fault diagnosis method for air compressor based on single-point acoustic signals |
Family Cites Families (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US6301572B1 (en) * | 1998-12-02 | 2001-10-09 | Lockheed Martin Corporation | Neural network based analysis system for vibration analysis and condition monitoring |
| CN102072144A (en) * | 2011-03-11 | 2011-05-25 | 大连大学 | Vibration and noise online monitoring and fault diagnosis system of scroll compressor |
| CN103603794B (en) * | 2013-11-19 | 2016-03-02 | 中国石油天然气集团公司 | A kind of gas storage note adopts compressor bank adaptive failure diagnostic method and equipment |
| CN104156628B (en) * | 2014-08-29 | 2017-05-31 | 东南大学 | A kind of naval vessel radiation signal recognition methods based on Multiple Kernel Learning discriminant analysis |
| US20160245279A1 (en) * | 2015-02-23 | 2016-08-25 | Biplab Pal | Real time machine learning based predictive and preventive maintenance of vacuum pump |
| CN106503798A (en) * | 2015-09-08 | 2017-03-15 | 中国科学院沈阳自动化研究所 | Based on rough set and the method for diagnosing faults of the pump of BP neural network |
-
2017
- 2017-04-18 CN CN201710253920.8A patent/CN107013449B/en active Active
Also Published As
| Publication number | Publication date |
|---|---|
| CN107013449A (en) | 2017-08-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107013449B (en) | The method and system of voice signal identification compressor fault based on deep learning | |
| CN108426713A (en) | Rolling bearing Weak fault diagnostic method based on wavelet transformation and deep learning | |
| CN110567720B (en) | Method for diagnosing depth confrontation of fault of fan bearing under unbalanced small sample scene | |
| CN110940539B (en) | Machine equipment fault diagnosis method based on artificial experience and voice recognition | |
| CN108181107B (en) | The Wind turbines bearing mechanical method for diagnosing faults of meter and more class objects | |
| CN104655423B (en) | A kind of Fault Diagnosis of Roller Bearings based on time-frequency domain multi-dimensional vibration Fusion Features | |
| CN109740523A (en) | A kind of method for diagnosing fault of power transformer based on acoustic feature and neural network | |
| CN111238814B (en) | Rolling bearing fault diagnosis method based on short-time Hilbert transform | |
| CN110135492B (en) | Equipment fault diagnosis and abnormality detection method and system based on multiple Gaussian models | |
| CN110031226A (en) | A kind of diagnostic method and device of bearing fault | |
| CN105841961A (en) | Bearing fault diagnosis method based on Morlet wavelet transformation and convolutional neural network | |
| CN105424366A (en) | Bearing fault diagnosis method based on EEMD adaptive denoising | |
| CN103234767A (en) | Nonlinear fault detection method based on semi-supervised manifold learning | |
| CN106017876A (en) | Wheel set bearing fault diagnosis method based on equally-weighted local feature sparse filter network | |
| CN103674234B (en) | State early warning method and system for abnormal vibration of wind generating set | |
| CN106228200A (en) | A kind of action identification method not relying on action message collecting device | |
| CN103900824A (en) | Method for diagnosing faults of diesel engine based on instant rotary speed clustering analysis | |
| CN106096562A (en) | Separate based on vibration signal blind sources and the gearbox of wind turbine method for diagnosing faults of sparse component analysis | |
| CN105044590A (en) | Single measurement node simulation circuit fault diagnosis method | |
| CN111370027A (en) | Off-line embedded abnormal sound detection system and method | |
| CN112364706A (en) | Small sample bearing fault diagnosis method based on class imbalance | |
| CN110261116A (en) | A kind of Bearing Fault Detection Method and device | |
| CN113327632B (en) | Unsupervised abnormal sound detection method and device based on dictionary learning | |
| CN103267652B (en) | Intelligent online diagnosis method for early failures of equipment | |
| CN113532848A (en) | Fault diagnosis system for planetary gear box |
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 | ||
| CB03 | Change of inventor or designer information |
Inventor after: Shao Peng Inventor after: Zhang Zhen Inventor after: Shi Yunfei Inventor after: Liang Bo Inventor before: Zhang Zhen Inventor before: Shi Yunfei Inventor before: Liang Bo |
|
| CB03 | Change of inventor or designer information | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant | ||
| CB03 | Change of inventor or designer information |
Inventor after: Zhang Zhen Inventor after: Shao Peng Inventor after: Shi Yunfei Inventor after: Liang Bo Inventor before: Shao Peng Inventor before: Zhang Zhen Inventor before: Shi Yunfei Inventor before: Liang Bo |
|
| CB03 | Change of inventor or designer information | ||
| CP01 | Change in the name or title of a patent holder |
Address after: 250103 room 1-101, office building, 2269 development road, high tech Zone, Ji'nan, Shandong Patentee after: Shandong Wanteng Digital Technology Co.,Ltd. Address before: 250103 room 1-101, office building, 2269 development road, high tech Zone, Ji'nan, Shandong Patentee before: SHANDONG WANTENG ELECTRONIC TECHNOLOGY CO.,LTD. |
|
| CP01 | Change in the name or title of a patent holder |