CN105276988A - Method for controlling FeO content in sintered ore endmost section - Google Patents
Method for controlling FeO content in sintered ore endmost section Download PDFInfo
- Publication number
- CN105276988A CN105276988A CN201410307470.2A CN201410307470A CN105276988A CN 105276988 A CN105276988 A CN 105276988A CN 201410307470 A CN201410307470 A CN 201410307470A CN 105276988 A CN105276988 A CN 105276988A
- Authority
- CN
- China
- Prior art keywords
- feo content
- image
- tail section
- sintering deposit
- infrared
- 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
Landscapes
- Image Processing (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to a method for controlling the FeO content in a sintered ore endmost section. The control method comprises the steps that a double charge coupled device (CCD) image monitoring system collects images of the sintered ore endmost section, the double CCD image monitoring system comprises a colored CCD camera and an infrared CCD camera, the colored CCD camera collects the visible light images, and the infrared CCD camera collects the infrared images; the collected images are processed through a processing computer; and finally, the FeO content in the sintered ore is measured through a data measuring and analyzing system, and the data measuring and analyzing system comprises a fuzzy clustering system and a neural network system. By means of the control method, the problems that according to an existing method for measuring the FeO content of sintered ore, the labor intensity is high, and the quality deviation of the sintered ore is large due to the fact that the FeO content is judged by a fire looker through experience are solved.
Description
Technical field
The present invention relates to a kind of control method of sintering deposit plant-tail section FeO content.
Background technology
Current domestic sintering plant, the on-line checkingi of FeO in Sinter mainly relies on the experience of people, namely by watching sintering work at sintering machine tail to sintering machine section direct vision, then by rule of thumb FeO in Sinter is judged, although more domestic sintering plants have installed closed-circuit TV system at sintering machine tail point of observation, see that firer can observe plant-tail section by industrial television, but still need see that firer relies on experience to judge, namely by watching sintering work continuous observation tail industrial television picture, then by rule of thumb FeO in Sinter is judged, provide the scope of FeO content, and then carry out certain operation control.
Eachly see that firer is different due to experience accumulation, judge to there is certain deviation, also have following situation can produce sinter quality deviation in addition: 1. because iron ore resource variation is comparatively large, determine to bring numerous unfavorable factor to technique; 2. sinter quality testing result delayed (generally delayed 2 ~ 3 hours), cannot revise technological operation in time; 3. detect sample to have some limitations, can not complete reflection sintering result.The sampling sample that just extracting pole is a small amount of in a large amount of products is chemically examined, and can not reflect the actual index of whole bulk article completely.
Summary of the invention
The object of this invention is to provide, a kind of control method of sintering deposit plant-tail section FeO content, this control method can the grade separation of sintering deposit Fe0 content in quick online detection sintering machine production process, exception sintering situation in production process can be fed back in time, in order to solve existing sintering deposit Fe0 content measuring method owing to need see that firer relies on experience to judge, the problem that labour intensity is large and sinter quality deviation is large.
For achieving the above object, the solution of the present invention is: a kind of control method of sintering deposit plant-tail section FeO content, this control method adopts two ccd image monitoring system to gather the image of sintering deposit plant-tail section, and by process computer, image procossing is carried out to the image gathered, finally by measurement and analysis of data system, FeO in Sinter is measured, described pair of ccd image monitoring system comprises colourful CCD video camera and infrared CCD video camera, described measurement and analysis of data system comprises fuzzy clustering system and nerve network system, described control method specifically comprises the following steps:
(1) visible images of colourful CCD video camera Real-time Collection sintering deposit plant-tail section, the infrared image of infrared CCD video camera Real-time Collection sintering deposit plant-tail section, and described visible images and infrared image are sent to process computer;
(2) process computer carries out image procossing to the visible images received and infrared image, and from visible images, extract the geometric properties data of flourishing layer in sintering deposit plant-tail section, from infrared image, extract the physical features data of flourishing layer in sintering deposit plant-tail section;
(3) characteristic extracted in described step (2) is input in fuzzy clustering system carries out the classification of FeO content rating, and the characteristic extracted in grade separation result and step (2) is input to nerve network system, carry out the training of neutral net;
(4) the real-time visible images carrying out the classification of FeO content rating of colourful CCD video camera and infrared CCD camera acquisition and infrared image are input to the nerve network system of having trained, and emulate, obtain the FeO content rating of realtime graphic;
(5) characteristic extracted in the FeO content rating obtained in described step (4), real-time synchronization visible images and infrared image and step (2) is sent to terminal interface, regulate sintering process parameter, control sintering deposit plant-tail section FeO content.
According to the control method of sintering deposit plant-tail section FeO content of the present invention, in described step (2), described image procossing comprises image smoothing, top cap change, Threshold segmentation, closed operation and opening operation.
According to the control method of sintering deposit plant-tail section FeO content of the present invention, in described step (2), described geometric properties data comprise the position of flourishing layer, width and area, and described physical features data comprise average gray and the temperature of flourishing layer pore.
According to the control method of sintering deposit plant-tail section FeO content of the present invention, the method of carrying out neural metwork training in described step (3) is: using the input value of the characteristic of extraction as neutral net, using the output valve of the grade of correspondence as neutral net, select suitable neutral net, hidden layer and output layer number, learning rate and convergence algorithm neural network training.
The beneficial effect that the present invention reaches: the present invention can the grade separation of sintering deposit Fe0 content in on-line checkingi sintering machine production process rapidly, can feed back in time sintering situation abnormal in production process, overcome that deterministic process is in the past delayed, labour intensity greatly, only with deficiencies such as personnel's subjective experience judgements, there is obvious effect to the quality control of sintering deposit, can solid fuel consumption be reduced simultaneously.
Accompanying drawing explanation
Fig. 1 is system architecture schematic diagram of the present invention;
Fig. 2 is control method flow chart of the present invention;
Fig. 3 is the image after section visible images of the present invention takes filtering process;
Fig. 4 is the binary image that Fig. 3 is corresponding;
Fig. 5 is grade separation result figure of the present invention.
Detailed description of the invention
Below in conjunction with accompanying drawing, the present invention is further detailed explanation.
As shown in Figure 1, control method of the present invention adopts two ccd image monitoring system 1 to gather the image of sintering deposit plant-tail section, and by process computer 3, image procossing is carried out to the image gathered, analysis to measure is carried out finally by measurement and analysis of data system 2 pairs of FeO in Sinters, described pair of ccd image monitoring system 1 comprises colourful CCD video camera and infrared CCD video camera, and described measurement and analysis of data system 2 comprises fuzzy clustering system and nerve network system.
As shown in Figure 2, control method of the present invention specifically comprises the following steps:
(1) visible images of colourful CCD video camera Real-time Collection sintering deposit plant-tail section, the infrared image of infrared CCD video camera Real-time Collection sintering deposit plant-tail section, and described visible images and infrared image are sent to process computer;
(2) process computer carries out the image procossing such as image smoothing, top cap change, Threshold segmentation, closed operation and opening operation to the visible images received and infrared image, and from visible images, extract position, the geometric properties data such as width and area of flourishing layer in sintering deposit plant-tail section, from infrared image, extract the physical features data such as average gray and temperature of flourishing layer pore in sintering deposit plant-tail section;
(3) characteristic extracted in described step (2) is input in fuzzy clustering system carries out the classification of FeO content rating, and the characteristic extracted in grade separation result and step (2) is input to nerve network system, using the input value of the characteristic of extraction as neutral net, using the output valve of the grade of correspondence as neutral net, suitable neutral net, hidden layer and output layer number, learning rate and convergence algorithm is selected to carry out the training of neutral net;
(4) by the real-time visible images of the FeO content rating of colourful CCD video camera and infrared CCD camera acquisition and infrared image, and be input to the nerve network system of having trained and emulate, obtain the FeO content rating of realtime graphic;
(5) characteristic extracted in the FeO content rating obtained in described step (4), real-time synchronization visible images and infrared image and step (2) is sent to terminal interface, regulate sintering process parameter, to control sintering deposit plant-tail section FeO content.
Below for Baosteel stainless iron-smelter 1# sintering machine, specific embodiment of the invention process is described in detail:
(1) in production scene, two ccd image monitoring system distinguishes a large amount of visible images and the infrared light image of Real-time Collection sintering deposit plant-tail section by colourful CCD video camera and infrared CCD video camera, and is transferred to process computer.
(2) process computer carries out the image procossing such as image smoothing, top cap change, Threshold segmentation, closed operation and opening operation to the visible images received and infrared image, so as comparatively complete, clearly observe and measurement image information, as shown in Figure 3 and Figure 4.
(3) from the visible images after image procossing, the geometric properties such as the position of flourishing layer in section, width and area are extracted, the physical features such as average gray and temperature of the flourishing layer pore of section can be extracted from the infrared image after image procossing, and the characteristic of extraction is input in fuzzy clustering system carries out the classification of FeO content rating, detailed process is as follows:
The characteristic extracted is formed sample space X={x
1, x
2..., x
i..., x
n, and be input in fuzzy clustering system, adopt fuzzy C-mean algorithm method (FCM algorithm), characteristic is divided into c classification, c be greater than 1 integer, define sample point x
ibelong to the degree of jth (1≤j≤c) class.
The fuzzy clustering fuzzy matrix W=(w of sample space X
ij) (0≤w is described
ij≤ 1), element w
ijbe the i-th row jth column element of matrix W, represent the degree of membership that i-th sample point is under the jurisdiction of jth class.W has following character:
w
ij∈[0,1]；
In order to calculate the degree of membership of each sample point relative to cluster centre, objective definition function:
Wherein m (m>1) is Fuzzy Exponential, z
jrepresent the cluster centre of jth class,
sample point x
ito cluster centre z
jeuclidean distance.
FCM algorithm is by obtaining the fuzzy classification to sample set to the iteration optimization of object function.Iterative process is as follows:
1) W is initialized randomly
(0), initialize Z
(0), and calculate W
(0).Make iterations be k=1, select classification number c, as c=3, represent and sample is divided into three classifications, and select Fuzzy Exponential m (m>1).
2) W is calculated
(k), to a certain sample point i and cluster centre r, if d
ir(k) >0, then
if there is i and r, make d
ir(k)=0, then w
ir(k)=1, and to j ≠ r, w
ijk ()=0, k represents iterations.D
irthe Euclidean distance of sample point i to cluster centre r after (k) expression kth time iteration, w
ijthe value of fuzzy matrix after (k) expression kth time iteration.
3) Z is calculated
(k+1).
4) if || W
(k)-W
(k+1)|| < ε, illustrates that fuzzy matrix has changed very little, then stops iteration; Otherwise make k=k+1, go to step 2), wherein ε is positive number given in advance, by actual conditions value.Such as: if iteration terminates rear W for the first time
1=0.35, W after second time iteration
2=0.34, if setting ε=0.02, then iteration terminates; If setting ε=0.01, iteration continues.
(4) grade separation result and the infrared light image of extraction and the characteristic of visible images are input to nerve network system, using the input value of the characteristic of extraction as neutral net, using the output valve of the grade of correspondence as neutral net, select the parameter training neutral nets such as suitable neutral net, hidden layer and output layer number, learning rate and convergence algorithm, obtain comparatively correct each weight threshold.
The present embodiment selects BP neutral net, and the convergence algorithm for BP neutral net has gradient descent method, has the gradient descent method of momentum, has the gradient descent method of self adaptation lr, has momentum to add gradient descent method, tension gradient descent method, conjugate gradient method, the Scaled Conjugate Gradient Method scheduling algorithm of self adaptation lr.The detailed process that the present embodiment is trained neutral net is as follows:
1) parameter is determined: sample X=[x
1, x
2..., x
n]
tbe defined as neutral net input vector, wherein each element represents a characteristic, and namely this time image contains n feature.By the classification O=[o of image
1, o
2..., o
q]
tbe defined as and wish output vector, a reference quantity wherein in each element representative image generic, the actual output vector of neutral net is Y=[y
1, y
2..., y
q]
t, q is output layer unit number.Hidden layer output vector is B=[b
1, b
2..., b
p]
t, p is hidden layer unit number.Initialize the connection weights W of input layer to hidden layer
j=[w
j1, w
j2..., w
jt..., w
jn]
t, j=1,2 ... p.Beginningization hidden layer is to the connection weights V of output layer
k=[v
k1, v
k2..., v
kj..., v
kp]
t, k=1,2 ... q.
2) input pattern is propagated: calculate each neuronic activation value of hidden layer
θ
jfor the threshold value of hidden layer unit, activation primitive adopts S type function, namely
calculate the output valve of hidden layer j unit
calculate the activation value of an output layer kth unit
calculate the real output value y of an output layer kth unit
k=f (s
k) (k=1,2 ..., q), θ
kfor output layer unit threshold value.
3) the inverse of output error is propagated: the real output value of network and the output valve of hope different time when error is greater than limited numerical value in other words, will correct network, correction is here carried out from back to front, so be called error Back-Propagation.
The correction error of output layer is d
k=(o
k-y
k) y
k(1-y
k), the correction error of each unit of hidden layer is
be △ v for output layer to the correcting value of hidden layer connection weight
kj=α d
kb
j, b
j∈ B, the correcting value of output layer threshold value is △ θ
k=α d
k, wherein α is learning coefficient, α >0, and hidden layer is △ w to the correcting value of input layer connection weight
ji=β e
jx
i, x
i∈ X, the correcting value of hidden layer threshold value is △ θ
j=β e
j, wherein β is learning coefficient, 0< β <1.Usual learning coefficient is between 0.1-0.8.
(5) visible images of the FeO content rating obtained in real time and infrared light image are input in the neutral net trained, obtain FeO content rating result, as shown in Figure 5, wherein ordinate 1.1,1,0.9 represents A, B, C Three Estate respectively, middle waveform curve is fitted figure, is convenient to the variation tendency of observing grade.
(6) finally the result analyzed is fed back to terminal interface together with the information of image, just can realize the effect of real-time estimate and production control parameter, realize the control to sintering deposit plant-tail section FeO content.
The present invention installs a set of sintering deposit plant-tail section FeO content grade analysis measurement mechanism based on two CCD at rear of sintering machine, can the grade separation of sintering deposit Fe0 content in on-line checkingi sintering machine production process rapidly, can feed back in time sintering situation abnormal in production process, overcome that deterministic process is in the past delayed, labour intensity greatly, only with deficiencies such as personnel's subjective experience judgements, there is obvious effect to the quality control of sintering deposit, can solid fuel consumption be reduced simultaneously.
Claims (4)
1. the control method of a sintering deposit plant-tail section FeO content, it is characterized in that: this control method adopts two ccd image monitoring system to gather the image of sintering deposit plant-tail section, and by process computer, image procossing is carried out to the image gathered, finally by measurement and analysis of data system, FeO in Sinter is measured, described pair of ccd image monitoring system comprises colourful CCD video camera and infrared CCD video camera, described measurement and analysis of data system comprises fuzzy clustering system and nerve network system, and described control method specifically comprises the following steps:
(1) visible images of colourful CCD video camera Real-time Collection sintering deposit plant-tail section, the infrared image of infrared CCD video camera Real-time Collection sintering deposit plant-tail section, and described visible images and infrared image are sent to process computer;
(2) process computer carries out image procossing to the visible images received and infrared image, and from visible images, extract the geometric properties data of flourishing layer in sintering deposit plant-tail section, from infrared image, extract the physical features data of flourishing layer in sintering deposit plant-tail section;
(3) characteristic extracted in described step (2) is input in fuzzy clustering system carries out the classification of FeO content rating, and the characteristic extracted in grade separation result and step (2) is input to nerve network system, carry out the training of neutral net;
(4) the real-time visible images carrying out the classification of FeO content rating of colourful CCD video camera and infrared CCD camera acquisition and infrared image are input to the nerve network system of having trained, and emulate, obtain the FeO content rating of realtime graphic;
(5) characteristic extracted in the FeO content rating obtained in described step (4), real-time synchronization visible images and infrared image and step (2) is sent to terminal interface, regulate sintering process parameter, control sintering deposit plant-tail section FeO content.
2. the control method of sintering deposit plant-tail section FeO content according to claim 1, is characterized in that in described step (2), and described image procossing comprises image smoothing, top cap change, Threshold segmentation, closed operation and opening operation.
3. the control method of sintering deposit plant-tail section FeO content according to claim 1, it is characterized in that in described step (2), described geometric properties data comprise the position of flourishing layer, width and area, and described physical features data comprise average gray and the temperature of flourishing layer pore.
4. the control method of sintering deposit plant-tail section FeO content according to claim 1, it is characterized in that the method for carrying out neural metwork training in described step (3) is: using the input value of the characteristic of extraction as neutral net, using the output valve of the grade of correspondence as neutral net, select suitable neutral net, hidden layer and output layer number, learning rate and convergence algorithm neural network training.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410307470.2A CN105276988A (en) | 2014-06-30 | 2014-06-30 | Method for controlling FeO content in sintered ore endmost section |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201410307470.2A CN105276988A (en) | 2014-06-30 | 2014-06-30 | Method for controlling FeO content in sintered ore endmost section |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN105276988A true CN105276988A (en) | 2016-01-27 |
Family
ID=55146436
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201410307470.2A Pending CN105276988A (en) | 2014-06-30 | 2014-06-30 | Method for controlling FeO content in sintered ore endmost section |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105276988A (en) |
Cited By (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109839011A (en) * | 2019-02-15 | 2019-06-04 | 成昕 | A kind of continuous pallettype sintering machine tail machine vision Instructing manufacture process approach |
| CN109855548A (en) * | 2019-01-30 | 2019-06-07 | 中南大学 | Analyze the method and system of the red flame layer thickness of sintering cup test |
| CN111128313A (en) * | 2019-07-16 | 2020-05-08 | 中南大学 | Method and system for detecting FeO content of sinter |
| CN111292312A (en) * | 2020-02-26 | 2020-06-16 | 中南大学 | Sintering thermal state transverse heterogeneity on-line quantitative measurement method |
| CN113259631A (en) * | 2021-06-18 | 2021-08-13 | 上海交通大学 | On-site real-time video acquisition and analysis system based on sintering machine tail |
| CN113269138A (en) * | 2021-06-18 | 2021-08-17 | 上海交通大学 | FeO content detection method based on deep multi-source information fusion |
| CN113564348A (en) * | 2021-07-13 | 2021-10-29 | 北京科技大学 | Sintering production method based on machine vision and data driving |
| CN113739576A (en) * | 2020-05-28 | 2021-12-03 | 中冶长天国际工程有限责任公司 | Method and system for acquiring section image of tail of sintering machine |
| CN114880936A (en) * | 2022-05-06 | 2022-08-09 | 北京智冶互联科技有限公司 | Method for predicting FeO content, model training method, device, electronic equipment and medium |
-
2014
- 2014-06-30 CN CN201410307470.2A patent/CN105276988A/en active Pending
Cited By (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109855548A (en) * | 2019-01-30 | 2019-06-07 | 中南大学 | Analyze the method and system of the red flame layer thickness of sintering cup test |
| CN109839011A (en) * | 2019-02-15 | 2019-06-04 | 成昕 | A kind of continuous pallettype sintering machine tail machine vision Instructing manufacture process approach |
| CN111128313A (en) * | 2019-07-16 | 2020-05-08 | 中南大学 | Method and system for detecting FeO content of sinter |
| CN111128313B (en) * | 2019-07-16 | 2023-05-23 | 中南大学 | Method and system for detecting FeO content of sinter |
| CN111292312B (en) * | 2020-02-26 | 2022-08-12 | 中南大学 | Sintering thermal state transverse heterogeneity on-line quantitative measurement method |
| CN111292312A (en) * | 2020-02-26 | 2020-06-16 | 中南大学 | Sintering thermal state transverse heterogeneity on-line quantitative measurement method |
| CN113739576A (en) * | 2020-05-28 | 2021-12-03 | 中冶长天国际工程有限责任公司 | Method and system for acquiring section image of tail of sintering machine |
| CN113739576B (en) * | 2020-05-28 | 2023-06-27 | 中冶长天国际工程有限责任公司 | Method and system for acquiring tail section image of sintering machine |
| CN113269138A (en) * | 2021-06-18 | 2021-08-17 | 上海交通大学 | FeO content detection method based on deep multi-source information fusion |
| CN113259631A (en) * | 2021-06-18 | 2021-08-13 | 上海交通大学 | On-site real-time video acquisition and analysis system based on sintering machine tail |
| CN113564348A (en) * | 2021-07-13 | 2021-10-29 | 北京科技大学 | Sintering production method based on machine vision and data driving |
| CN114880936A (en) * | 2022-05-06 | 2022-08-09 | 北京智冶互联科技有限公司 | Method for predicting FeO content, model training method, device, electronic equipment and medium |
| CN114880936B (en) * | 2022-05-06 | 2023-01-06 | 北京智冶互联科技有限公司 | Method for predicting FeO content, model training method, device, electronic equipment and medium |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105276988A (en) | Method for controlling FeO content in sintered ore endmost section | |
| CN110598736B (en) | Power equipment infrared image fault positioning, identifying and predicting method | |
| CN108985376B (en) | Rotary kiln sequence working condition identification method based on convolution-cyclic neural network | |
| CN108921051B (en) | Pedestrian attribute identification network and technology based on cyclic neural network attention model | |
| CN111274860B (en) | Recognition method for online automatic tobacco grade sorting based on machine vision | |
| CN105678332B (en) | Converter steelmaking end point judgment method and system based on flame image CNN recognition modeling | |
| CN103902987B (en) | A kind of TV station symbol recognition method based on convolutional network | |
| CN106874929B (en) | Pearl classification method based on deep learning | |
| CN107644218B (en) | The working method that crowded region behavior analyzes and determines is realized based on image collecting function | |
| CN107860099A (en) | Frosting detection method, device, storage medium and the equipment of a kind of air-conditioning | |
| CN110222592B (en) | Construction method of time sequence behavior detection network model based on complementary time sequence behavior proposal generation | |
| CN109858516A (en) | A kind of fire and smog prediction technique, system and medium based on transfer learning | |
| CN106250896A (en) | The recognition methods of the positive and negative of online Nicotiana tabacum L. based on image collecting device | |
| CN112819802A (en) | Method for monitoring and predicting abnormal condition of blast furnace based on deep learning of tuyere information | |
| CN111299166A (en) | Zipper quality detection system and method based on cognitive vision | |
| CN108875118B (en) | Method and device for evaluating accuracy of prediction model of silicon content of blast furnace molten iron | |
| CN106354262A (en) | Optimized-neural-network gesture-recognition human-computer interaction method based on GL | |
| CN108181891A (en) | Industrial big data method for diagnosing faults based on intelligent core pivot element analysis | |
| CN108734277A (en) | A kind of targeted drug spray method based on convolutional neural networks | |
| CN110533100A (en) | A method of CME detection and tracking is carried out based on machine learning | |
| CN111222798B (en) | Complex industrial process key index soft measurement method | |
| CN110503647A (en) | Wheat plant real-time counting method based on deep learning image segmentation | |
| CN113077423B (en) | Laser selective melting pool image analysis system based on convolutional neural network | |
| CN112861443B (en) | Advanced learning fault diagnosis method integrated with priori knowledge | |
| CN110503092A (en) | The improvement SSD monitor video object detection method adapted to based on field |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20160127 |
|
| WD01 | Invention patent application deemed withdrawn after publication |