CN108734142A - A kind of core in-pile component surface roughness appraisal procedure based on convolutional neural networks - Google Patents
A kind of core in-pile component surface roughness appraisal procedure based on convolutional neural networks Download PDFInfo
- Publication number
- CN108734142A CN108734142A CN201810522834.7A CN201810522834A CN108734142A CN 108734142 A CN108734142 A CN 108734142A CN 201810522834 A CN201810522834 A CN 201810522834A CN 108734142 A CN108734142 A CN 108734142A
- Authority
- CN
- China
- Prior art keywords
- convolutional neural
- neural networks
- feature
- core
- value
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Computational Linguistics (AREA)
- Software Systems (AREA)
- General Physics & Mathematics (AREA)
- General Health & Medical Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Biophysics (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The core in-pile component surface roughness appraisal procedure based on convolutional neural networks that the invention discloses a kind of, includes the following steps：S1, acquisition video data；S2, image data is obtained；S3, image data is divided into training dataset and test data set；S4, training dataset input convolutional neural networks are trained, obtain feature recognition model and exports first identification feature；S5, first identification feature is classified according to roughness, obtains roughness grade number；S6, test data set input feature vector identification model is tested, exports secondary identification feature；S7, judge whether secondary identification feature meets roughness grade number；S8, by roughness grade number include in human-computer interaction interface.The present invention solves that artificial detection of the existing technology and assessment cause human input greatly and efficiency is low, and the problem of can not ensure real-time and the continuity requirement of material surface roughness measurement and assessment.
Description
Technical field
The present invention relates to nuclear industry technical fields, and in particular to a kind of core in-pile component surface based on convolutional neural networks
Roughness appraisal procedure.
Background technology
Before the year two thousand twenty, China will build about 32 million kilowatt nuclear power units again, including be completed Qinshan, Daya Gulf, ridge
The nuclear power generating sets such as Australia, total installation of generating capacity reach 40,000,000 kilowatts or more.Nuclear power station height radiates the regular of the in-service equipment of underwater environment
Maintenance and important leverage and a danger, arduous and time-consuming work that maintenance is in-service nuclear plant safety operation, need to solve
The problem of certainly high dose of radiation and underwater operation feasibility.
From the point of view of the current development in China, the safety core reliability verified and applied is improved, reduces operating personnel's as possible
Raying dosage improves working environment, solves some and seriously threaten China's nuclear plant safety hidden danger, and Devoting Major Efforts To Developing is needed to utilize water
Lower maintenance activity equipment, this is the strategic long-term objectives that can not be ignored.
During nuclear power plant's operation and maintenance, it is desirable that have under the specific environments such as radiation, underwater, remote control to machine
Tool equipment, structure such as are cut, are shaped, being repaired at the technique and equipment of processing, and spark erosion technique is nuclear power plant's operation, dimension
Shield, retired one of cutting, molding key technology in the process.
Electrical discharge machining is a kind of new process being processed using electric energy and thermal energy, is commonly called as electro-discharge machining.Electrical spark working
Difference lies in, electrical spark working working hour tool and workpiece not in contact with but by between tool and workpiece for work and general machining
The pulse feature spark discharge constantly generated, generation is local, instantaneous high temperature metal material, gradually get off by ablation when using electric discharge.
Due to there is visible spark to generate in discharge process, therefore claim electrical discharge machining.
Surface roughness is usually formed by by used processing method and other factors, such as knife in process
The plastic deformation of superficial layer metal and the high-frequency vibration in process system when tool is detached with friction, the chip between piece surface
Deng.Due to the difference of processing method and workpiece material, the depth, density, shape and texture that surface to be machined is left a trace have
Difference.No matter processed with which kind of processing method, always leaves fine scraggly trace in piece surface, appearance interlocks
The peak valley phenomenon of volt, the surface after roughing can with the naked eye see that the surface after finishing is remained to magnifying glass or microscope
It observes.
The prior art has the following disadvantages：
(1) in the prior art using manually core in-pile component surface is detected and is assessed, human input is big and efficiency
Low, simultaneously because there is radiation, big to human injury, safety is low；
(2) the part in-pile component of nuclear power station by high radiation, pressure, liquid vibration, high-temerature creep, friction due to being ground
The influence of the factors such as damage, hydrogen absorption, the roughness grade number for easily causing component surface increases, but the prior art cannot achieve reality
When, continuous detect and assess.
Invention content
For above-mentioned deficiency in the prior art, a kind of saving human input provided by the invention, efficient, real-time with
The high core in-pile component surface roughness appraisal procedure based on convolutional neural networks of continuity, solves of the existing technology
Artificial detection causes human input greatly with assessment and efficiency is low, and can not ensure the reality of material surface roughness measurement and assessment
The problem of when property and continuity require.
In order to reach foregoing invention purpose, the technical solution adopted by the present invention is：
A kind of core in-pile component surface roughness appraisal procedure based on convolutional neural networks, includes the following steps：
S1：Core in-pile component video surface data are acquired by image capture module；
S2：Sectional drawing frame by frame is carried out by video data Input Monitor Connector warning module, and to video data, obtains image data；
S3：Image data is divided into training dataset and test data set；
S4：Training dataset input convolutional neural networks are trained, feature recognition model is obtained and export first knowledge
Other feature；
S5：First identification feature is classified according to roughness, obtains roughness grade number；
S6：Test data set input feature vector identification model is tested, secondary identification feature is exported；
S7：Judge whether secondary identification feature meets roughness grade number, if then entering step S8, otherwise enters step
S2；
S8：Include the human-computer interaction interface in monitoring and warning module by roughness grade number, realizes that core in-pile component surface is thick
The detection and assessment of rugosity.
Further, in step S1, image capture module is underwater laser scanner.
Further, in step S2, monitoring and warning module is host computer.
Further, in step S3,60% sample of standard image data is randomly selected as training dataset, remaining
Sample be test data set.
Further, in step S4, convolutional neural networks include the 4 convolutional Neural sub-networks set gradually and maximum pond
Change layer.
Further, in step S4, the method that convolutional neural networks are trained includes the following steps：
S4-1：Data characteristics is extracted using the 1st and the 2nd convolutional Neural sub-network, and data characteristics is combined,
As low-level feature；
S4-2：Abstract high-rise expression is carried out to low-level feature using the 3rd and the 4th convolutional Neural sub-network, is obtained
High-level characteristic；
S4-3：Down-sampled processing is carried out to high-level characteristic using maximum pond layer, reduces the parameter of neural network, and export
Feature recognition model and first identification feature；
The calculation formula of identification feature is for the first time：
In formula, pl(i,j)For the first identification feature of l layers of i-th of convolution kernel；al(i,t)For l layers of i-th of convolution
T-th of activation value of core；W is the width of convolution kernel；J is constant.
Further, convolutional Neural sub-network includes the convolutional layer set gradually, active coating and batch normalization layer；
Convolutional layer carries out convolution algorithm to the regional area of input data, exports corresponding data feature values；
The calculation formula of convolution algorithm is：
In formula, yl(i,j)For the characteristic value of l layers of i-th of convolution kernel；For the jth of l layers of i-th of convolution kernel '
A weights；For l layers of i-th of regional area being convolved；W is the width of convolution kernel；
Active coating carries out nonlinear transformation to the data feature values that convolutional layer exports, and exports corresponding activation value；
The calculation formula of nonlinear transformation is：
al(i,j)=f (yl(i,j))=max { 0, yl(i,j)}
In formula, al(i,j)Y is exported for convolutional layerl(i,j)Activation value；yl(i,j)For the feature of l layers of i-th of convolution kernel
Value；
The characteristic value of input is normalized in batch normalization layer, reduces built-in variable transfer, improves convolution god
Through network of network training effectiveness, enhance the generalization ability of convolutional neural networks network；
The calculation formula of normalized is：
In formula, zl(i,j)Y is exported for convolutional layerl(i,j)Normalized value；Y is exported for convolutional layerl(i,j)Limits value；
γl(i)For scale value；βl(i)For bias；
In formula,Y is exported for convolutional layerl(i,j)Limits value；yl(i,j)For the characteristic value summation of convolution kernel；μBFor mean value；For standard deviation.
Further, in step S5, roughness is determined by the average arithmetic deviation value of profile.
Further, the calculation formula of the average arithmetic deviation value of profile is：
In formula, RaFor the average arithmetic deviation value of profile；L is contour surface length；Y is profile variation value；X is profile table
Face length axis.
This programme has the beneficial effect that：
Nuclear reactor surface topography image is obtained by underwater laser scanner, is automatically performed based on convolutional neural networks
The different grades of roughness features extraction in surface and identification process, improve efficiency, have saved human input, ensure that material table
Real-time and the continuity requirement of detection with the assessment of surface roughness, avoiding roughness grade number raising leads to nuclear reactor occur
Accident potential endangers the personal safety of staff, improves safety, adapts to the market demand of China's Nuclear Power Development, has wide
Wealthy application prospect, and generate significant economic results in society.
Description of the drawings
Fig. 1 is the core in-pile component surface roughness appraisal procedure flow chart based on convolutional neural networks；
Fig. 2 is the method flow diagram that convolutional neural networks are trained；
Fig. 3 is convolutional neural networks structure chart；
Fig. 4 is low roughness level diagram；
Fig. 5 is high roughness grade number schematic diagram；
Fig. 6 is special roughness grade number schematic diagram；
Fig. 7 is that embodiment one extracts characteristic profile；
Fig. 8 is that embodiment two extracts characteristic profile；
Fig. 9 is that embodiment three extracts characteristic profile.
Specific implementation mode
The specific implementation mode of the present invention is described below, in order to facilitate understanding by those skilled in the art this hair
It is bright, it should be apparent that the present invention is not limited to the ranges of specific implementation mode, for those skilled in the art,
As long as various change is in the spirit and scope of the present invention that the attached claims limit and determine, these variations are aobvious and easy
See, all are using the innovation and creation of present inventive concept in the row of protection.
In the embodiment of the present invention, a kind of core in-pile component surface roughness appraisal procedure based on convolutional neural networks, such as
Shown in Fig. 1, include the following steps：
S1：Core in-pile component video surface data are acquired by image capture module；
S2：Sectional drawing frame by frame is carried out by video data Input Monitor Connector warning module, and to video data, obtains image data；
S3：Image data is divided into training dataset and test data set；
S4：Training dataset input convolutional neural networks are trained, feature recognition model is obtained and export first knowledge
Other feature；
Convolutional neural networks are as shown in figure 3,4 convolutional Neural sub-networks including setting gradually and maximum pond layer；Volume
The method that product neural network is trained, as shown in Fig. 2, including the following steps：
S4-1：Data characteristics is extracted using the 1st and the 2nd convolutional Neural sub-network, and data characteristics is combined,
As low-level feature；
Convolutional Neural sub-network includes the convolutional layer set gradually, active coating and batch normalization layer；
Convolutional layer carries out convolution algorithm to the regional area of input data, exports corresponding data feature values；
The calculation formula of convolution algorithm is：
In formula, yl(i,j)For the characteristic value of l layers of i-th of convolution kernel；For the jth of l layers of i-th of convolution kernel '
A weights；For l layers of i-th of regional area being convolved；W is the width of convolution kernel；
Active coating carries out nonlinear transformation to the data feature values that convolutional layer exports, and exports corresponding activation value；
The calculation formula of nonlinear transformation is：
al(i,j)=f (yl(i,j))=max { 0, yl(i,j)}
In formula, al(i,j)Y is exported for convolutional layerl(i,j)Activation value；yl(i,j)For the feature of l layers of i-th of convolution kernel
Value；
The characteristic value of input is normalized in batch normalization layer, reduces built-in variable transfer, improves convolution god
Through network of network training effectiveness, enhance the generalization ability of convolutional neural networks network；
The calculation formula of normalized is：
In formula, zl(i,j)Y is exported for convolutional layerl(i,j)Normalized value；Y is exported for convolutional layerl(i,j)Limits value；
γl(i)For scale value；βl(i)For bias；
In formula,Y is exported for convolutional layerl(i,j)Limits value；yl(i,j)For the characteristic value summation of convolution kernel；μBFor mean value；For standard deviation；
S4-2：Abstract high-rise expression is carried out to low-level feature using the 3rd and the 4th convolutional Neural sub-network, to
It was found that the character representation of data distribution formula, obtains high-level characteristic；
S4-3：Down-sampled processing is carried out to high-level characteristic using maximum pond layer, reduces the parameter of neural network, and export
Feature recognition model and first identification feature；
The calculation formula of identification feature is for the first time：
In formula, pl(i,j)For the first identification feature of l layers of i-th of convolution kernel；al(i,t)For l layers of i-th of convolution
T-th of activation value of core；W is the width of convolution kernel；J is constant；
S5：First identification feature is classified according to roughness, obtains roughness grade number, roughness is averaged by profile
Arithmetic deviation value determines；
The calculation formula of the average arithmetic deviation value of profile is：
In formula, RaFor the average arithmetic deviation value of profile；L is contour surface length；Y is profile variation value；X is profile table
Face length axis；
S6：Test data set input feature vector identification model is tested, secondary identification feature is exported；
S7：Judge whether secondary identification feature meets roughness grade number, if then entering step S8, otherwise enters step
S2；
S8：Include the human-computer interaction interface in monitoring and warning module by roughness grade number, realizes that core in-pile component surface is thick
The detection and assessment of rugosity.
In the embodiment of the present invention, data characteristics is extracted using the 1st and the 2nd convolutional Neural sub-network, and to data spy
Sign is combined, and as low-level feature, corresponding window size is respectively 7x7 and 5x5, and step-length is 3 and 2, convolution kernel 32
It is a；Abstract high-rise expression is carried out to low-level feature using the 3rd and the 4th convolutional Neural sub-network, to find data point
The character representation of cloth, obtains high-level characteristic, select wicket, small step-length, more convolution kernels structure, corresponding window size
All it is 3x3, step-length 2, convolution kernel is 64.
What maximum pond layer carried out is down-sampled operation, and main purpose is to reduce the parameter of neural network, the high level of input
The window size of feature is 6x6, and the pondization for being 2 by size and step-length operates, and is downsampled to the first identification feature of 3x3.
Convolutional neural networks are trained based on image data, network is automatically performed surface different brackets in training process
Roughness features extraction and identification process.The maximum iteration of convolutional neural networks is 40 times, and learning rate is set as
0.0004.100 pictures of different roughness grade are taken, as shown in Fig. 4, Fig. 5 and Fig. 6, randomly select 60% sample of data set
, as training set, remaining sample is test set for this.
In order to more intuitively show the ability in feature extraction of convolutional neural networks, to the convolutional neural networks knot of data process
The feature of the extraction of the 1st convolutional Neural sub-network, the 3rd convolutional Neural sub-network and maximum pond layer is dropped after structure processing
It ties up and does visualization processing.Only signal characteristic Jing Guo a process of convolution is lengthy and jumbled, different grades of surface roughness characteristics without
Rule intersperses among in space, can not classify without apparent boundary and effectively, as shown in Figure 7；By convolutional neural networks knot
The processing of structure, the features of the 3rd convolutional Neural sub-network data are in the aggregation stage, roughness grade number Ra3.2, Ra25 with
The data characteristics line of demarcation of Ra100, but remaining roughness features are still partially in aliasing state, influence final classification effect
Fruit, as shown in Figure 8；The signal characteristic handled by maximum pond layer is clear, and roughness grade can be clearly distinguished open,
Roughness features of the same race cluster, different roughness character separation and be in divergent shape are conducive to classification, as shown in Figure 9.
A kind of saving human input provided by the invention, efficient, real-time and continuity are high based on convolutional Neural net
The core in-pile component surface roughness appraisal procedure of network, solving artificial detection of the existing technology and assessment causes manpower to be thrown
Enter greatly and efficiency is low, and can not ensure the problem of real-time of material surface roughness measurement and assessment and continuity require.
Claims (9)
1. a kind of core in-pile component surface roughness appraisal procedure based on convolutional neural networks, which is characterized in that including following
Step：
S1：Core in-pile component video surface data are acquired by image capture module；
S2：Sectional drawing frame by frame is carried out by video data Input Monitor Connector warning module, and to video data, obtains image data；
S3：Image data is divided into training dataset and test data set；
S4：Training dataset input convolutional neural networks are trained, feature recognition model is obtained and export first identification spy
Sign；
S5：First identification feature is classified according to roughness, obtains roughness grade number；
S6：Test data set input feature vector identification model is tested, secondary identification feature is exported；
S7：Judge whether secondary identification feature meets roughness grade number, if then entering step S8, otherwise enters step S2；
S8：Include the human-computer interaction interface in monitoring and warning module by roughness grade number, realizes core in-pile component surface roughness
Detection and assessment.
2. the core in-pile component surface roughness appraisal procedure according to claim 1 based on convolutional neural networks, special
Sign is, in the step S1, image capture module is underwater laser scanner.
3. the core in-pile component surface roughness appraisal procedure according to claim 1 based on convolutional neural networks, special
Sign is, in the step S2, monitoring and warning module is host computer.
4. the core in-pile component surface roughness appraisal procedure according to claim 1 based on convolutional neural networks, special
Sign is, in the step S3, randomly selects 60% sample of standard image data as training dataset, remaining sample
For test data set.
5. the core in-pile component surface roughness appraisal procedure according to claim 1 based on convolutional neural networks, special
Sign is, in the step S4, convolutional neural networks include the 4 convolutional Neural sub-networks set gradually and maximum pond layer.
6. the core in-pile component surface roughness appraisal procedure according to claim 5 based on convolutional neural networks, special
Sign is that in the step S4, the method that convolutional neural networks are trained includes the following steps：
S4-1：Data characteristics is extracted using the 1st and the 2nd convolutional Neural sub-network, and data characteristics is combined, as
Low-level feature；
S4-2：Abstract high-rise expression is carried out to low-level feature using the 3rd and the 4th convolutional Neural sub-network, obtains high level
Feature；
S4-3：Down-sampled processing is carried out to high-level characteristic using maximum pond layer, reduces the parameter of neural network, and export feature
Identification model and first identification feature；
The calculation formula of identification feature is for the first time：
In formula, pl(i,j)For the first identification feature of l layers of i-th of convolution kernel；al(i,t)For l layers of i-th of convolution kernel
T-th of activation value；W is the width of convolution kernel；J is constant.
7. the core in-pile component surface roughness appraisal procedure according to claim 6 based on convolutional neural networks, special
Sign is that the convolutional Neural sub-network includes the convolutional layer set gradually, active coating and batch normalization layer；
The convolutional layer carries out convolution algorithm to the regional area of input data, exports corresponding data feature values；
The calculation formula of convolution algorithm is：
In formula, yl(i,j)For the characteristic value of l layers of i-th of convolution kernel；For the jth of l layers of i-th of convolution kernel ' a power
Value；For l layers of j-th of regional area being convolved；W is the width of convolution kernel；
The active coating carries out nonlinear transformation to the data feature values that convolutional layer exports, and exports corresponding activation value；
The calculation formula of nonlinear transformation is：
al(i,j)=f (yl(i,j))=max { 0, yl(i,j)}
In formula, al(i,j)Y is exported for convolutional layerl(i,j)Activation value；yl(i,j)For the characteristic value of l layers of i-th of convolution kernel；
The characteristic value of input is normalized in the batch normalization layer, reduces built-in variable transfer, improves convolution god
Through network of network training effectiveness, enhance the generalization ability of convolutional neural networks network；
The calculation formula of normalized is：
In formula, zl(i,j)Y is exported for convolutional layerl(i,j)Normalized value；Y is exported for convolutional layerl(i,j)Limits value；γl(i)
For scale value；βl(i)For bias；
In formula,Y is exported for convolutional layerl(i,j)Limits value；yl(i,j)For the characteristic value summation of convolution kernel；μBFor mean value；For standard deviation.
8. the core in-pile component surface roughness appraisal procedure according to claim 1 based on convolutional neural networks, special
Sign is, in the step S5, roughness is determined by the average arithmetic deviation value of profile.
9. the core in-pile component surface roughness appraisal procedure according to claim 8 based on convolutional neural networks, special
Sign is that the calculation formula of the average arithmetic deviation value of the profile is：
In formula, RaFor the average arithmetic deviation value of profile；L is contour surface length；Y is profile variation value；X is that contour surface is long
Spend axis.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810522834.7A CN108734142A (en) | 2018-05-28 | 2018-05-28 | A kind of core in-pile component surface roughness appraisal procedure based on convolutional neural networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810522834.7A CN108734142A (en) | 2018-05-28 | 2018-05-28 | A kind of core in-pile component surface roughness appraisal procedure based on convolutional neural networks |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108734142A true CN108734142A (en) | 2018-11-02 |
Family
ID=63935616
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810522834.7A Pending CN108734142A (en) | 2018-05-28 | 2018-05-28 | A kind of core in-pile component surface roughness appraisal procedure based on convolutional neural networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108734142A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109003689A (en) * | 2018-05-28 | 2018-12-14 | 西南交通大学 | A kind of core in-pile component surface monitoring method based on convolutional neural networks |
| CN109461153A (en) * | 2018-11-15 | 2019-03-12 | 联想(北京)有限公司 | Data processing method and device |
| CN109840899A (en) * | 2018-12-20 | 2019-06-04 | 上海理工大学 | A kind of roughness grade number recognition methods based on depth convolutional neural networks |
| CN110110758A (en) * | 2019-04-15 | 2019-08-09 | 南京航空航天大学 | A kind of surface roughness classification method based on convolutional neural networks |
| CN111583502A (en) * | 2020-05-08 | 2020-08-25 | 辽宁科技大学 | Renminbi (RMB) crown word number multi-label identification method based on deep convolutional neural network |
| CN111709924A (en) * | 2020-06-11 | 2020-09-25 | 东北大学 | Intelligent 3D rock structural surface roughness extraction system and method |
| FR3116342A1 (en) * | 2020-11-19 | 2022-05-20 | Framatome | Method and electronic device for assisting in the monitoring of a nuclear reactor element, computer program and associated system |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1577056A1 (en) * | 2004-03-16 | 2005-09-21 | Waldrich Siegen Werkzeugmaschinenbau GmbH | Method and device for grinding a rotating roller |
| CN101634551A (en) * | 2009-08-18 | 2010-01-27 | 清华大学深圳研究生院 | Method and system for detecting surface roughness |
| CN102880771A (en) * | 2012-10-31 | 2013-01-16 | 贵州大学 | Method for predicting surface roughness of workpiece during high-speed cutting machining |
| CN103105820A (en) * | 2012-05-22 | 2013-05-15 | 华中科技大学 | Machining cutter abrasion state identification method of numerical control machine tool |
| CN103632735A (en) * | 2013-09-18 | 2014-03-12 | 魏玲 | Novel nuclear reactor pressure container material |
| CN103868930A (en) * | 2013-11-29 | 2014-06-18 | 中科华核电技术研究院有限公司 | Non-contact type detection equipment |
| CN104318249A (en) * | 2014-10-23 | 2015-01-28 | 刘奇元 | Grinding surface roughness grade identification method based on multi-classification support vector machine |
| CN107016440A (en) * | 2017-03-01 | 2017-08-04 | 西安交通大学 | The multiresolution deep neural network intelligent diagnosing method of machine driving failure |
| US20170270434A1 (en) * | 2016-03-17 | 2017-09-21 | Fanuc Corporation | Machine learning apparatus, laser machining system and machine learning method |
-
2018
- 2018-05-28 CN CN201810522834.7A patent/CN108734142A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP1577056A1 (en) * | 2004-03-16 | 2005-09-21 | Waldrich Siegen Werkzeugmaschinenbau GmbH | Method and device for grinding a rotating roller |
| CN101634551A (en) * | 2009-08-18 | 2010-01-27 | 清华大学深圳研究生院 | Method and system for detecting surface roughness |
| CN103105820A (en) * | 2012-05-22 | 2013-05-15 | 华中科技大学 | Machining cutter abrasion state identification method of numerical control machine tool |
| CN102880771A (en) * | 2012-10-31 | 2013-01-16 | 贵州大学 | Method for predicting surface roughness of workpiece during high-speed cutting machining |
| CN103632735A (en) * | 2013-09-18 | 2014-03-12 | 魏玲 | Novel nuclear reactor pressure container material |
| CN103868930A (en) * | 2013-11-29 | 2014-06-18 | 中科华核电技术研究院有限公司 | Non-contact type detection equipment |
| CN104318249A (en) * | 2014-10-23 | 2015-01-28 | 刘奇元 | Grinding surface roughness grade identification method based on multi-classification support vector machine |
| US20170270434A1 (en) * | 2016-03-17 | 2017-09-21 | Fanuc Corporation | Machine learning apparatus, laser machining system and machine learning method |
| CN107016440A (en) * | 2017-03-01 | 2017-08-04 | 西安交通大学 | The multiresolution deep neural network intelligent diagnosing method of machine driving failure |
Non-Patent Citations (4)
| Title |
|---|
| WEI ZHANG ET AL.: "A deep convolutional neural network with new training methods for bearing fault diagnosis under noisy environment and different working load", 《MECHANICAL SYSTEMS AND SIGNAL PROCESSING》 * |
| 卢男: "铸造表面粗糙度三维评价体系建立及检测方法研究", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 吴春亚: "基于机器视觉的磨削表面粗糙度检测", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
| 王爱囡: "表面粗糙度在线识别的理论与实践", 《合肥工业大学学报》 * |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109003689A (en) * | 2018-05-28 | 2018-12-14 | 西南交通大学 | A kind of core in-pile component surface monitoring method based on convolutional neural networks |
| CN109461153A (en) * | 2018-11-15 | 2019-03-12 | 联想(北京)有限公司 | Data processing method and device |
| CN109461153B (en) * | 2018-11-15 | 2022-04-22 | 联想(北京)有限公司 | Data processing method and device |
| CN109840899A (en) * | 2018-12-20 | 2019-06-04 | 上海理工大学 | A kind of roughness grade number recognition methods based on depth convolutional neural networks |
| CN110110758A (en) * | 2019-04-15 | 2019-08-09 | 南京航空航天大学 | A kind of surface roughness classification method based on convolutional neural networks |
| CN111583502A (en) * | 2020-05-08 | 2020-08-25 | 辽宁科技大学 | Renminbi (RMB) crown word number multi-label identification method based on deep convolutional neural network |
| CN111709924A (en) * | 2020-06-11 | 2020-09-25 | 东北大学 | Intelligent 3D rock structural surface roughness extraction system and method |
| FR3116342A1 (en) * | 2020-11-19 | 2022-05-20 | Framatome | Method and electronic device for assisting in the monitoring of a nuclear reactor element, computer program and associated system |
| WO2022106506A1 (en) * | 2020-11-19 | 2022-05-27 | Framatome | Method and electronic device for assisting with surveillance of an element of a nuclear reactor, associated computer program and associated system |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108734142A (en) | A kind of core in-pile component surface roughness appraisal procedure based on convolutional neural networks | |
| CN108805747A (en) | A kind of abnormal electricity consumption user detection method based on semi-supervised learning | |
| CN106897686A (en) | A kind of airborne LIDAR electric inspection process point cloud classifications method | |
| CN107730477A (en) | Substation equipment infrared intelligent monitoring system and monitoring image feature extracting method | |
| CN109238760B (en) | Online monitoring method of coal mill of intelligent power plant coal-fired generator set based on typical correlation analysis and slow characteristic analysis | |
| WO2016004749A1 (en) | Method for recognizing tool abrasion degree of large numerical control milling machine | |
| Li et al. | TireNet: A high recall rate method for practical application of tire defect type classification | |
| Zhao et al. | Detection and location of safety protective wear in power substation operation using wear-enhanced YOLOv3 algorithm | |
| CN108665464A (en) | A kind of foreign matter detecting method based on morphologic high tension electric tower and high-tension bus-bar | |
| CN109003689B (en) | A kind of core in-pile component surface monitoring method based on convolutional neural networks | |
| Rebhi et al. | Fabric defect detection using local homogeneity and morphological image processing | |
| Chang et al. | Application of fuzzy entropy to improve feature selection for defect recognition using support vector machine in high voltage cable joints | |
| CN106097138A (en) | A kind of electricity consumption anomaly data detection System and method for based on statistical model | |
| Mathias et al. | Detection of micro-cracks in electroluminescence images of photovoltaic modules | |
| CN110956077A (en) | Method for preventing misoperation and operation safety based on power distribution station room | |
| Lin et al. | Identification of icing thickness of transmission line based on strongly generalized convolutional neural network | |
| CN104915681B (en) | Take into account substation's clustering method of voltage dip monitoring information and its relation factor | |
| CN116542956B (en) | Automatic detection method and system for fabric components and readable storage medium | |
| CN106250458B (en) | A kind of Unit Cell point geometry information extracting method | |
| Zhang et al. | Research on image defect detection of silicon panel based on prewitt and canny operator | |
| Zhang et al. | Fault line selection method based on transfer learning depthwise separable convolutional neural network | |
| Fang et al. | Detection method of transmission line broken stock defects in aircraft inspection based on image processing technology | |
| CN112036410B (en) | Electric power equipment infrared image segmentation method based on deep learning and image gradient | |
| CN109033992A (en) | Heavy industrial district domain discovery algorithm based on active fiery point data | |
| Yan et al. | Infrared image segment and fault location for power equipment |
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: 20181102 |
|
| RJ01 | Rejection of invention patent application after publication |