CN109934417A - Boiler coke method for early warning based on convolutional neural networks - Google Patents
Boiler coke method for early warning based on convolutional neural networks Download PDFInfo
- Publication number
- CN109934417A CN109934417A CN201910232761.2A CN201910232761A CN109934417A CN 109934417 A CN109934417 A CN 109934417A CN 201910232761 A CN201910232761 A CN 201910232761A CN 109934417 A CN109934417 A CN 109934417A
- Authority
- CN
- China
- Prior art keywords
- layer
- convolutional
- sampling
- coking
- noncoking
- 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.)
- Granted
Links
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Abstract
The present invention discloses the boiler coke method for early warning based on convolutional neural networks, comprising the following steps: 1), acquisition pot slagging or noncoking data information；2), selecting step 1) data information in, the coking of same period or several measuring point temperature datas of noncoking；3) the convolutional neural networks model of coking or noncoking, is constructed comprising: input layer, convolutional layer, down-sampling layer, full articulamentum and output layer；4), from the boiler data source acquired in real time, several measuring point temperature datas of same period, input step 3 are randomly selected) coking or noncoking convolutional neural networks model, obtain the characteristics of image of the measuring point temperature data, judge its coking or noncoking.The present invention provides early warning scheme for boiler coke, can accurately predict boiler heating surface coking situation, to judge that boiler heating surface coking provides the guidance foundation of science.
Description
Technical field
The present invention relates to power plant boiler running optimizatin technical fields.
Background technique
Boiler coke refers to boiler in the process of running, heating surface surface due to fuel or combustion system reason caused by coke
Oil, dust attachment, and gradually develop, form the phenomenon that coke button covers.Boiler heating surface coking be concentrated mainly on vertical water wall,
The heating surfaces such as helical water-cooled wall, when heating surface coking is serious, the parameters such as unit load, attemperation water flow, operation of coal pulverizing mills
Also it will appear different degrees of change, reduce the service life of boiler heating surface, cause four pipes are revealed, unit is non-to stop, give power plant
Cause huge economic loss.Currently, being directed to boiler coke phenomenon, thermal power plant can not obtain having for boiler coke or noncoking
Data are imitated, only according to working experience, carries out disturbing coke using regular switching operation of coal pulverizing mills or spray coke-removing agent carries out actively
Decoking.
Summary of the invention
It is an object of that present invention to provide a kind of boiler coke method for early warning based on convolutional neural networks.
Based on above-mentioned purpose, the present invention mainly takes following technical scheme:
Boiler coke method for early warning based on convolutional neural networks, comprising the following steps:
1) data information of pot slagging or noncoking, is acquired；
2), selecting step 1) data information in, the coking of same period or several measuring point temperature datas of noncoking；
3), construct the convolutional neural networks model of coking or noncoking comprising: input layer, convolutional layer, down-sampling layer,
Full articulamentum and output layer,
The method and step of input layer are as follows:
A, several measuring point temperature datas by the coking of step 2) or noncoking supplement the data matrix for a × a, square respectively
Array element element is insufficient to be filled with 0；
B, each matrix element in step a in data matrix is normalized, obtains coking or noncoking analogy characteristics of image,
Normalize calculation formula are as follows:
Wherein: i, j are respectively the index of the row and column of data matrix in step a, xI, jFor i after normalization, the square of the position j
Array element element, vi,jFor normalize before i, the matrix element of the position j,To normalize the minimum matrix element in preceding j column,To normalize the maximal matrix element element in preceding j column；
4), from the boiler data source acquired in real time, several measuring point temperature datas of same period are randomly selected, are inputted
The convolutional neural networks model of coking or the noncoking of step 3), obtains the characteristics of image of the measuring point temperature data, judges the figure
As feature, meets the coking characteristics of image in convolutional neural networks model, be then coking；Meet in convolutional neural networks model
The characteristics of image of noncoking is then noncoking.
In step 3), convolutional layer: the analogy characteristics of image in step b is subjected to convolution algorithm, obtains the characteristic pattern of convolutional layer；
Down-sampling layer: down-sampling is carried out to the characteristic pattern of convolutional layer, obtains the characteristic pattern of down-sampling layer；
Full articulamentum: the characteristic pattern of down-sampling layer is flattened, one-dimensional vector is obtained；
Output layer: probability output is carried out using softmax function, includes the equal all neurons connecting with full articulamentum.
In step 3), convolutional layer is divided into convolutional layer C1 and convolutional layer C2, and down-sampling layer is divided into down-sampling layer S1 and down-sampling
Layer S2, specific steps are as follows:
Convolutional layer C1: the analogy characteristics of image in step b is subjected to convolution algorithm, obtains the characteristic pattern of convolutional layer C1；
Down-sampling layer S1: down-sampling is carried out to the characteristic pattern of convolutional layer C1, obtains the characteristic pattern of down-sampling layer S1；
Convolutional layer C2: the characteristic pattern of down-sampling layer S1 is subjected to convolution algorithm, obtains the characteristic pattern of convolutional layer C2；
Down-sampling layer S2: down-sampling is carried out to the characteristic pattern of convolutional layer C2, obtains the characteristic pattern of down-sampling layer S2.
In step 2), the measuring point temperature data of same period takes 123, in input layer step a, the size of data matrix
It is 12 × 12；The convolution kernel of convolutional layer C1 and convolutional layer C2 are mean μ=0, variances sigma2=1 standardization is just distributed very much；Convolution
Convolution kernel has 3 in layer C1, and convolution kernel is 3 × 3；Convolution kernel has 5 in convolutional layer C2, and convolution kernel is 3 × 3.
Convolutional layer C1, convolutional layer C2, full articulamentum are all made of relu activation primitive.
In step 2), the measuring point temperature data of same period takes 123, in input layer step a, the size of data matrix
It is 12 × 12.
Compared with prior art, the invention has the following advantages:
The present invention utilizes boiler measuring point real-time data base, and the thinking of image procossing is used to pre-process data to it,
Convolutional neural networks model is established, prediction classification is carried out to real time data using the image recognition algorithm of convolutional neural networks, is sentenced
Data when whether the disconnected real time data monitored is coking, provide early warning scheme for boiler coke, can accurately predict pot
Furnace heating surface coking situation, to judge that boiler heating surface coking provides the guidance foundation of science.
Detailed description of the invention
Fig. 1 is flow diagram of the invention；
Fig. 2 is the citing signal of down-sampling layer operation maximum pond method；
Fig. 3 is the coking gray feature figure of wall temperature data；
Fig. 4 is the noncoking gray feature figure of wall temperature data.
Specific embodiment
Embodiment
Boiler coke method for early warning based on convolutional neural networks, comprising the following steps:
1) data information of pot slagging or noncoking, is acquired；
2), selecting step 1) data information in, the coking of same period or several measuring point temperature datas of noncoking；
3), construct the convolutional neural networks model of coking or noncoking comprising: input layer, convolutional layer, down-sampling layer,
Full articulamentum and output layer,
The method and step of input layer are as follows:
A, 123 measuring point temperature datas of the coking of step 2) or noncoking are supplemented to the data square for 12 × 12 respectively
Battle array, matrix element is insufficient to be filled with 0；
B, by each matrix element normalization (carry out min-max standardization) in step a in data matrix, obtain coking or
Noncoking analogy characteristics of image,
Normalize calculation formula are as follows:
Wherein: i, j are respectively the index of the row and column of data matrix in step a, xI, jFor i after normalization, the square of the position j
Array element element, vI, jFor normalize before i, the matrix element of the position j,To normalize the minimum matrix element in preceding j column,To normalize the maximal matrix element element in preceding j column；
4), from the boiler data source acquired in real time, several measuring point temperature datas of same period are randomly selected, are inputted
The convolutional neural networks model of coking or the noncoking of step 3), obtains the characteristics of image of the measuring point temperature data, judges the figure
As feature, meets the coking characteristics of image in convolutional neural networks model, be then coking；Meet in convolutional neural networks model
The characteristics of image of noncoking is then noncoking.
Specifically, convolutional layer: by the analogy characteristics of image in step b and meeting mean μ=0, variances sigma in step 3)2=
The convolution kernel of 1 standardized normal distribution carries out convolution algorithm, obtains the characteristic pattern of convolutional layer；
Down-sampling layer: down-sampling is carried out to the characteristic pattern of convolutional layer, obtains the characteristic pattern of down-sampling layer；
Full articulamentum: Flatten (pressing) is carried out to the characteristic pattern of down-sampling layer, obtains one-dimensional vector；
Output layer: probability output is carried out using softmax function, includes all neurons connecting with full articulamentum.
Specifically, convolutional layer is divided into convolutional layer C1 and convolutional layer C2, the convolution of convolutional layer C1 and convolutional layer C2 in step 3)
Core is mean μ=0, variances sigma2=1 standardization is just distributed very much；Down-sampling layer is divided into down-sampling layer S1 and down-sampling layer S2, tool
Body step are as follows:
Convolutional layer C1: convolution kernel has 3, under convolution kernel is 3 × 3, and the analogy characteristics of image in step b is carried out convolution fortune
It calculates, obtains the characteristic pattern of convolutional layer C1；
Down-sampling layer S1: down-sampling is carried out to the characteristic pattern of convolutional layer C1, obtains the characteristic pattern of down-sampling layer S1；
Convolutional layer C2: convolution kernel has 5, under convolution kernel is 3 × 3, and the characteristic pattern of down-sampling layer S1 is carried out convolution algorithm,
Obtain the characteristic pattern of convolutional layer C2；
Down-sampling layer S2: down-sampling is carried out to the characteristic pattern of convolutional layer C2, obtains the characteristic pattern of down-sampling layer S2.
Convolutional layer C1, convolutional layer C2, full articulamentum are all made of relu activation primitive.
It is exemplified below:
Boiler coke method for early warning based on convolutional neural networks, process are as shown in Figure 1, comprising the following steps:
1) data information of pot slagging or noncoking, is acquired, specifically: find out the time of coke heavy and noncoking
Section: it is looked for from the coke dropping account of thermal power plant's worker's typing record and effectively falls burnt event and (manually disturb burnt, natural coke dropping etc. to generate
Fall big burnt event) time point, then at the time of being pushed forward 30 minutes until to be considered as coking within 3 hours before the moment serious
Period is considered as the period of noncoking to 3 hours after the moment at the time of time point was to pusher 30 minutes；
2), selecting step 1) data information in, the coking of same period or 123 measuring point temperature numbers of noncoking
According to, specifically: from the correspondence measuring point found out in the corresponding boiler wall temperature measuring point data library of thermal power plant in the period that step 1 determines
Wall temperature data, are classified as two parts: wall temperature data when coke heavy and the wall temperature data without coking；
3), construct the convolutional neural networks model of coking or noncoking comprising: input layer, convolutional layer, down-sampling layer,
Full articulamentum and output layer, convolutional layer are divided into convolutional layer C1 and convolutional layer C2, and down-sampling layer is divided into down-sampling layer S1 and down-sampling
Layer S2,
The method and step of input layer are as follows:
A, 123 measuring point temperature datas of the coking of step 2) or noncoking are supplemented to the data square for 12 × 12 respectively
Battle array, matrix element is insufficient to fill that (purpose of benefit 0 is easy for convolution algorithm when subsequent image feature extraction, boosting algorithm with 0
Versatility)；
B, each matrix element in step a in data matrix is normalized, obtains coking or noncoking analogy characteristics of image,
Normalize calculation formula are as follows:
Wherein: i, j are respectively the index of the row and column of data matrix in step a, xI, jFor i after normalization, the square of the position j
Array element element, vI, jFor normalize before i, the matrix element of the position j,To normalize the minimum matrix element in preceding j column,To normalize the maximal matrix element element in preceding j column；
Convolutional layer C1: convolution kernel has 3, under convolution kernel is 3 × 3, and analogy image meets min-max standardization in step b
Under conditions of, and meet mean μ=0, variances sigma2The analogy image in convolution kernel and step b that=1 standard is just being distributed very much is special
Sign carries out convolution algorithm, obtains the characteristic pattern of convolutional layer C1, and in sliding process, edge is not changed convolution kernel with 0 filling, the operation
Image size, the example are 12 × 12 × 3 data, three Color Channel (face of common RGB color figure by the operation export structure
Color is also one of feature of picture), this data generated can be understood as the characteristic pattern in 3 feature channel；
Down-sampling layer S1: down-sampling is carried out to the characteristic pattern of convolutional layer C1, obtains the characteristic pattern of down-sampling layer S1, down-sampling layer
Carrying out abstract processing to characteristics of image prevents over-fitting, increases the robustness of network, and down-sampling layer essence is a convolution operation,
But its effect is mutually simple, obscures to image, is blurred to unessential feature, and strengthens important feature (at this
Strengthen boiler high temperature region in example), process of mainly realizing in this example for maximum pond (maxpool) process, utilizes one 2 × 2
Matrix window, sliding step is 2 processions traversal, and the maximum value in 2 × 2 windows is extracted and forms new matrix such as
3 channels are successively carried out same steps, obtain one 6 × 6 × 3 matrix character figure by (citing is as shown in Figure 2)；
Convolutional layer C2: convolution kernel has 5, and under convolution kernel is 3 × 3, the characteristic pattern of down-sampling layer S1, which complies with standard, just to be divided very much
Under conditions of cloth, the characteristic pattern of down-sampling layer S1 is subjected to convolution algorithm, obtains the matrix character figure that convolutional layer C2 is 6 × 6 × 5；
Down-sampling layer S2: down-sampling is carried out to the characteristic pattern of convolutional layer C2, obtains the feature that down-sampling layer S2 is 3 × 3 × 5
Figure；
Full articulamentum is divided into full articulamentum 1 and full articulamentum 2 (can also only have full articulamentum 1), specifically:
Full articulamentum 1, full articulamentum neuron number are the size that image carries out image array after multilayer convolution operation, will
The each column of image extract, and are then spliced into 3 × 3 × 5=45 of one-dimensional vector, i.e., the input layer of full articulamentum 1 includes 45
Neuron；
Full articulamentum 2, chooses 15 neurons, complete 2 weights initialisation mode of articulamentum is similar with convolution kernel initialization, is
The standardized normal distribution matrix of one 45 (complete 1 neuronal quantity of articulamentum) × 15 (hidden layer neuron quantity)；
Output layer: carrying out probability output using softmax function, comprising the equal all neurons being connect with full articulamentum,
Above-mentioned convolutional layer and full articulamentum use relu activation primitive, and network optimization mode is stochastic gradient descent method (SGD), learning rate
It is set as 0.01, sample batch gives 50 training samples, all data repetitive exercises 200 times, by a large amount of sample number every time
It is trained according to convolutional neural networks model, constantly amendment error, and back transfer updates the weight of convolution kernel, finally obtains
The convolutional neural networks model of one relative ideal, and by model structure and first change the storage of the parameters such as weight；
4), from the boiler data source acquired in real time, 123 measuring point temperature datas of same period are randomly selected, it is defeated
The convolutional neural networks model for entering coking or the noncoking of step 3), obtains the characteristics of image of the measuring point temperature data, and judgement should
Characteristics of image meets the coking characteristics of image in convolutional neural networks model, then is coking；Meet in convolutional neural networks model
Noncoking characteristics of image, then be noncoking.As shown in Figures 3 and 4, in figure, ater be supplement and in truthful data
Existing 0 value, colour brightness is higher to show that boiler measuring point temperature is higher, and it is special that different data can form different image datas
Sign.
Claims (5)
1. the boiler coke method for early warning based on convolutional neural networks, which comprises the following steps:
1) data information of pot slagging or noncoking, is acquired；
2), selecting step 1) data information in, the coking of same period or several measuring point temperature datas of noncoking；
3) the convolutional neural networks model of coking or noncoking, is constructed comprising: input layer, convolutional layer, down-sampling layer, Quan Lian
Layer and output layer are connect,
The method and step of input layer are as follows:
A, several measuring point temperature datas by the coking of step 2) or noncoking supplement the data matrix for a × a, matrix element respectively
Element is insufficient to be filled with 0；
B, each matrix element in step a in data matrix is normalized, obtains coking or noncoking analogy characteristics of image,
Normalize calculation formula are as follows:Wherein: i, j are respectively the row and column of data matrix in step a
Index, xi,jFor i after normalization, the matrix element of the position j, vi,jFor normalize before i, the matrix element of the position j,For
Minimum matrix element before normalizing in j column,To normalize the maximal matrix element element in preceding j column；
4), from the boiler data source acquired in real time, several measuring point temperature datas of same period, input step are randomly selected
3) the convolutional neural networks model of coking or noncoking, obtains the characteristics of image of the measuring point temperature data, judges image spy
Sign, meets the coking characteristics of image in convolutional neural networks model, is then coking；Meet not tying in convolutional neural networks model
Burnt characteristics of image, then be noncoking.
2. the boiler coke method for early warning based on convolutional neural networks as described in claim 1, which is characterized in that step 3)
In, convolutional layer: the analogy characteristics of image in step b is subjected to convolution algorithm, obtains the characteristic pattern of convolutional layer；
Down-sampling layer: down-sampling is carried out to the characteristic pattern of convolutional layer, obtains the characteristic pattern of down-sampling layer；
Full articulamentum: the characteristic pattern of down-sampling layer is flattened, one-dimensional vector is obtained；
Output layer: probability output is carried out using softmax function.
3. the boiler coke method for early warning based on convolutional neural networks as claimed in claim 2, which is characterized in that step 3)
In, convolutional layer is divided into convolutional layer C1 and convolutional layer C2, and down-sampling layer is divided into down-sampling layer S1 and down-sampling layer S2, specific steps
Are as follows:
Convolutional layer C1: the analogy characteristics of image in step b is subjected to convolution algorithm, obtains the characteristic pattern of convolutional layer C1；
Down-sampling layer S1: down-sampling is carried out to the characteristic pattern of convolutional layer C1, obtains the characteristic pattern of down-sampling layer S1；
Convolutional layer C2: the characteristic pattern of down-sampling layer S1 is subjected to convolution algorithm, obtains the characteristic pattern of convolutional layer C2；
Down-sampling layer S2: down-sampling is carried out to the characteristic pattern of convolutional layer C2, obtains the characteristic pattern of down-sampling layer S2.
4. the boiler coke method for early warning based on convolutional neural networks as claimed in claim 3, which is characterized in that step 2)
In, the measuring point temperature data of same period takes 123, and in input layer step a, the size of data matrix is 12 × 12；Convolution
The convolution kernel of layer C1 and convolutional layer C2 is mean μ=0, variances sigma2=1 standardized normal distribution；There is convolution kernel in convolutional layer C1
3, convolution kernel is 3 × 3；Convolution kernel has 5 in convolutional layer C2, and convolution kernel is 3 × 3.
5. the boiler coke method for early warning based on convolutional neural networks as claimed in claim 4, which is characterized in that convolutional layer
C1, convolutional layer C2, full articulamentum are all made of relu activation primitive.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910232761.2A CN109934417B (en) | 2019-03-26 | 2019-03-26 | Boiler coking early warning method based on convolutional neural network |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910232761.2A CN109934417B (en) | 2019-03-26 | 2019-03-26 | Boiler coking early warning method based on convolutional neural network |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109934417A true CN109934417A (en) | 2019-06-25 |
| CN109934417B CN109934417B (en) | 2023-04-07 |
Family
ID=66988260
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910232761.2A Active CN109934417B (en) | 2019-03-26 | 2019-03-26 | Boiler coking early warning method based on convolutional neural network |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109934417B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113344909A (en) * | 2021-07-01 | 2021-09-03 | 中国石油大学(北京) | Method and device for identifying and displaying coking of flame-permeable high-temperature filter of thermal power boiler |
Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN1975382A (en) * | 2006-12-07 | 2007-06-06 | 华南理工大学 | Coal slagging scorification trend fast monitoring instrument |
| US20080116051A1 (en) * | 2006-09-29 | 2008-05-22 | Fisher-Rosemount Systems, Inc. | Main column bottoms coking detection in a fluid catalytic cracker for use in abnormal situation prevention |
| CN102607009A (en) * | 2012-02-20 | 2012-07-25 | 华北电力大学 | Fouling monitoring system for convection heating surface of boiler |
| CN102928056A (en) * | 2012-11-22 | 2013-02-13 | 中国人民解放军国防科学技术大学 | Method for measuring coking quantity of hydrocarbon fuel |
| CN103064289A (en) * | 2012-12-19 | 2013-04-24 | 华南理工大学 | Multiple-target operation optimizing and coordinating control method and device of garbage power generator |
| CN104361153A (en) * | 2014-10-27 | 2015-02-18 | 中国石油大学(北京) | Method for predicting coking amount of heavy oil catalytic cracking settler |
| CN105423273A (en) * | 2015-12-15 | 2016-03-23 | 天津鹰麟节能科技发展有限公司 | Spectroscopic boiler anti-coking system and control method |
| CN105654067A (en) * | 2016-02-02 | 2016-06-08 | 北京格灵深瞳信息技术有限公司 | Vehicle detection method and device |
| US20160177189A1 (en) * | 2014-12-18 | 2016-06-23 | Exxonmobil Research And Engineering Company | Reduction of coking in fccu feed zone |
| CN107092906A (en) * | 2017-05-01 | 2017-08-25 | 刘至键 | A kind of Chinese traditional medicinal materials recognition device based on deep learning |
| CN107256396A (en) * | 2017-06-12 | 2017-10-17 | 电子科技大学 | Ship target ISAR characteristics of image learning methods based on convolutional neural networks |
| CN107491781A (en) * | 2017-07-21 | 2017-12-19 | 国家电网公司 | A kind of crusing robot visible ray and infrared sensor data fusion method |
| CN107798336A (en) * | 2017-09-18 | 2018-03-13 | 广东电网有限责任公司东莞供电局 | A kind of infrared measurement of temperature image component recognition methods |
| US20180218515A1 (en) * | 2015-07-14 | 2018-08-02 | Unifai Holdings Limited | Computer vision process |
| CN108489912A (en) * | 2018-05-11 | 2018-09-04 | 东北大学 | A kind of coal constituent analysis method based on coal spectroscopic data |
| CN108921893A (en) * | 2018-04-24 | 2018-11-30 | 华南理工大学 | A kind of image cloud computing method and system based on online deep learning SLAM |
| CN109084613A (en) * | 2018-09-12 | 2018-12-25 | 东北电力大学 | Air cooling tubes condenser dust stratification status monitoring and cleaning control system and its regulation method based on convolutional neural networks and image recognition |
-
2019
- 2019-03-26 CN CN201910232761.2A patent/CN109934417B/en active Active
Patent Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20080116051A1 (en) * | 2006-09-29 | 2008-05-22 | Fisher-Rosemount Systems, Inc. | Main column bottoms coking detection in a fluid catalytic cracker for use in abnormal situation prevention |
| CN1975382A (en) * | 2006-12-07 | 2007-06-06 | 华南理工大学 | Coal slagging scorification trend fast monitoring instrument |
| CN102607009A (en) * | 2012-02-20 | 2012-07-25 | 华北电力大学 | Fouling monitoring system for convection heating surface of boiler |
| CN102928056A (en) * | 2012-11-22 | 2013-02-13 | 中国人民解放军国防科学技术大学 | Method for measuring coking quantity of hydrocarbon fuel |
| CN103064289A (en) * | 2012-12-19 | 2013-04-24 | 华南理工大学 | Multiple-target operation optimizing and coordinating control method and device of garbage power generator |
| CN104361153A (en) * | 2014-10-27 | 2015-02-18 | 中国石油大学(北京) | Method for predicting coking amount of heavy oil catalytic cracking settler |
| US20160177189A1 (en) * | 2014-12-18 | 2016-06-23 | Exxonmobil Research And Engineering Company | Reduction of coking in fccu feed zone |
| US20180218515A1 (en) * | 2015-07-14 | 2018-08-02 | Unifai Holdings Limited | Computer vision process |
| CN105423273A (en) * | 2015-12-15 | 2016-03-23 | 天津鹰麟节能科技发展有限公司 | Spectroscopic boiler anti-coking system and control method |
| CN105654067A (en) * | 2016-02-02 | 2016-06-08 | 北京格灵深瞳信息技术有限公司 | Vehicle detection method and device |
| CN107092906A (en) * | 2017-05-01 | 2017-08-25 | 刘至键 | A kind of Chinese traditional medicinal materials recognition device based on deep learning |
| CN107256396A (en) * | 2017-06-12 | 2017-10-17 | 电子科技大学 | Ship target ISAR characteristics of image learning methods based on convolutional neural networks |
| CN107491781A (en) * | 2017-07-21 | 2017-12-19 | 国家电网公司 | A kind of crusing robot visible ray and infrared sensor data fusion method |
| CN107798336A (en) * | 2017-09-18 | 2018-03-13 | 广东电网有限责任公司东莞供电局 | A kind of infrared measurement of temperature image component recognition methods |
| CN108921893A (en) * | 2018-04-24 | 2018-11-30 | 华南理工大学 | A kind of image cloud computing method and system based on online deep learning SLAM |
| CN108489912A (en) * | 2018-05-11 | 2018-09-04 | 东北大学 | A kind of coal constituent analysis method based on coal spectroscopic data |
| CN109084613A (en) * | 2018-09-12 | 2018-12-25 | 东北电力大学 | Air cooling tubes condenser dust stratification status monitoring and cleaning control system and its regulation method based on convolutional neural networks and image recognition |
Non-Patent Citations (4)
| Title |
|---|
| ZIWANG LIU: "Infrared Image Combined with CNN Based Fault Diagnosis for Rotating Machinery" * |
| 孙自强,顾幸生,俞金寿,党晓恒: "连续催化重整反应器结焦含量软测量" * |
| 董海鹰: "《智能控制理论及应用》", 30 September 2016 * |
| 陈军等: "煤粉炉结焦的预测技术", 《煤矿现代化》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113344909A (en) * | 2021-07-01 | 2021-09-03 | 中国石油大学(北京) | Method and device for identifying and displaying coking of flame-permeable high-temperature filter of thermal power boiler |
| CN113344909B (en) * | 2021-07-01 | 2023-12-08 | 中国石油大学(北京) | Method and device for identifying and displaying flame penetration height Wen Lvjing coking of thermal power boiler |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109934417B (en) | 2023-04-07 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110348376A (en) | A kind of pedestrian's real-time detection method neural network based | |
| CN108427920A (en) | A kind of land and sea border defense object detection method based on deep learning | |
| CN108985376A (en) | It is a kind of based on convolution-Recognition with Recurrent Neural Network rotary kiln sequence operating mode's switch method | |
| CN109447936A (en) | A kind of infrared and visible light image fusion method | |
| CN108664924A (en) | A kind of multi-tag object identification method based on convolutional neural networks | |
| CN111814875A (en) | Method for expanding ship samples in infrared image based on pattern generation countermeasure network | |
| CN110458060A (en) | A kind of vehicle image optimization method and system based on confrontation study | |
| CN110136210B (en) | Intelligent early warning system for leakage of monitoring device | |
| CN110298397A (en) | The multi-tag classification method of heating metal image based on compression convolutional neural networks | |
| CN109598200B (en) | Intelligent image identification system and method for molten iron tank number | |
| CN109934417A (en) | Boiler coke method for early warning based on convolutional neural networks | |
| CN109583498A (en) | A kind of fashion compatibility prediction technique based on low-rank regularization feature enhancing characterization | |
| CN114581560A (en) | Attention mechanism-based multi-scale neural network infrared image colorizing method | |
| CN113111758A (en) | SAR image ship target identification method based on pulse neural network | |
| CN107038419A (en) | A kind of personage's behavior method for recognizing semantics based on video sequence deep learning | |
| Wang et al. | A working condition diagnosis model of sucker rod pumping wells based on big data deep learning | |
| CN109977838B (en) | Flame combustion state detection method | |
| CN110287990A (en) | Microalgae image classification method, system, equipment and storage medium | |
| CN110674845B (en) | Dish identification method combining multi-receptive-field attention and characteristic recalibration | |
| CN108510483A (en) | A kind of calculating using VLAD codings and SVM generates color image tamper detection method | |
| CN107529647A (en) | A kind of cloud atlas cloud amount computational methods based on the unsupervised sparse learning network of multilayer | |
| Liu et al. | Improved design of risk assessment model for PPP project under the development of marine architecture | |
| Noh et al. | Data augmentation method for object detection in underwater environments | |
| Munshi et al. | Going, going, gone dark: Quantifying the scatter in the faintest dwarf galaxies | |
| CN113095181A (en) | Traffic sign identification method based on Defense-GAN |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |