Soft sensing method for CO content in blast furnace top based on collaborative analysis of multi-view data

Abstract

The present invention provides a soft measurement method for the CO content of a blast furnace top based on collaborative analysis of multi-perspective data, and relates to the technical field of blast furnace ironmaking process. First, image data and related physical variables of the blast furnace tuyere and furnace top are collected, and image matrices and physical variable matrices of the tuyere and furnace top are obtained through grayscale processing and normalization processing; then, collaborative correlation analysis is performed on the above data to obtain a correlation information index matrix for prediction; then, the index matrix is ​​further processed to further process information for calculating and removing associated data; a prediction model for the CO content of the blast furnace top is established using the information after removing associated data, and parameters for predicting the CO content of the blast furnace top are obtained; finally, after grayscale processing and removing associated data information on the data acquired in real time, the CO content of the blast furnace top is predicted according to the parameters of the blast furnace top CO content prediction model. The present invention can improve the accuracy of CO measurement of the blast furnace top and reduce its measurement difficulty and time lag.

Description

Technical Field

[0001] The present invention relates to the technical field of blast furnace ironmaking processes, and in particular to a soft measurement method for CO content in a blast furnace top based on collaborative analysis of multi-perspective data. Background Art

[0002] CO concentration is a key indicator of gas distribution and reduction reaction inside the blast furnace, and directly reflects the efficiency of the chemical reaction in the blast furnace. A high CO content generally indicates high reaction efficiency and optimal gas flow, while a low CO content indicates poor reaction kinetics or uneven gas distribution, resulting in reduced output. Therefore, measuring the CO content helps to actively adjust operating parameters, which is crucial to maintaining safe, continuous and stable operation of the blast furnace, energy conservation, emission reduction, and quality improvement. It also helps to promote high-quality development of the steel industry and accelerate the realization of the industry's "carbon peak". However, as the most energy-consuming process in modern processing industry and one of the largest and most complex production processes, the blast furnace ironmaking process is characterized by a continuous production environment of high temperature, high pressure, high dust, multi-phase, and multi-field coupling, which leads to great difficulties in real-time measurement of CO content and seriously increases the measurement time and cost requirements. Summary of the Invention

[0003] The technical problem to be solved by the present invention is to address the deficiencies of the above-mentioned existing technologies and provide a soft measurement method for the CO content in the blast furnace top based on collaborative analysis of multi-perspective data. The method can solve the problems of long measurement time, high cost and failure to fully utilize on-site production data and improve the accuracy of CO content measurement in the furnace top.

[0004] In order to solve the above technical problems, the technical solution adopted by the present invention is:

[0005] A soft sensing method for CO content in a blast furnace top based on collaborative analysis of multi-view data includes the following steps:

[0006] Step 1: Collect images of the tuyere raceway, furnace top, and related physical variable data as well as CO content data of the furnace top during blast furnace production;

[0007] Step 1.1: Collect image data of the flame in the tuyere raceway of the blast furnace and its related physical variables;

[0008] Step 1.2: Collect image data of the blast furnace top flame and its related physical variables;

[0009] Step 1.3: Collect blast furnace top CO content data;

[0010] Step 2: Perform collaborative analysis on the data obtained in step 1. The objective function model of data collaborative analysis is:

[0011]

[0012] Among them, X i 、X j are images at adjacent moments under the same viewing angle, X a 、X b is the image of the adjacent moment from another perspective, and the size of the image matrix is ​​m×n; is the projection matrix, which is used to analyze the correlation information of multiple perspectives of samples, l is the number of samples, w ij 、w ab is the weight of the physical variable under different perspectives, β1 and β2 are coefficient parameters; y i is the kinetic energy of the wind, y a is the CO content at the furnace top;

[0013] w ij 、w ab The settings are as follows:

[0014]

[0015] Among them, r i 、r j is the relevant physical variable of the tuyere raceway, r a 、r b is the relevant physical variable of the furnace top, σ is the core parameter;

[0016] Solving the data collaborative analysis objective function model to obtain U, V and a constant deviation b;

[0017] Step 3: Based on the projection matrices U and V obtained in step 2, calculate the index Q of the correlation information under multiple perspectives;

[0018] First we get P = UV T , P is an m×n matrix; convert the matrix P into a column vector, that is: Q=Vec(P),

[0019] Step 4: The correlation information index Q obtained in step 2 is applied to the image of the furnace top to obtain correlation information and calculate the key information for prediction after removing the correlation data;

[0020] Step 5: Using the bilinear regression method, the prediction model for the CO content of each sample furnace top is established as follows:

[0021]

[0022] Among them, u pj 、v pj They are the left projection matrix u p , right projection matrix v p The j-th column vector of ; Set as u p=[u p1 ,u p2 ,...,u pj ,...,u pk ]、v p =[v p1 ,v p2 ,...,v pj ,...,v pk ];X s is the key information for CO content calculation, b p is a constant deviation, α is a system parameter, y s is the label vector corresponding to the CO content at the top of the furnace in the CO content prediction model, q = Vec(u p (v p ) T ).

[0023] Solve the furnace top CO content prediction model to obtain u p 、v p and b p ;

[0024] Step 6: Use the furnace top camera to collect the furnace top image data during the blast furnace industrial production process. Based on the solutions of steps 2 to 5, the furnace top CO content is measured online during the ironmaking blast furnace production process. The calculation model for the furnace top CO content is:

[0025]

[0026] Furthermore, the specific steps for solving the objective function model in step 2 are as follows:

[0027] Step 2.1: Set the initial projection matrix V;

[0028] Step 2.2: Fix V and find U and b. The specific steps are as follows:

[0029] Step 2.2.1: Combine the multi-view image operations in equation (1) to obtain:

[0030]

[0031] in, u1, u2…u k is the column vector of U; E=[S,A] mk×2l ,A=[A1,A2,…,A a ,,A l ] mk×l , S=[s1,s2,…,s i ,,s l ] mk×l , Y is the label vector corresponding to the viewing angle,

[0032] Step 2.2.2: Calculate the operation from the perspective of physical variables and obtain:

[0033]

[0034] Where F = β1SL d S T +β2AL f A T , L d 、L f are the Laplace matrices corresponding to the physical variables of the furnace top and the physical variables of the tuyere, L d 、L f The calculation method is: L = DW, L represents L d or L f , D is a diagonal matrix, W is the corresponding adjacency matrix;

[0035] Step 2.2.3: Derivative (5) and set it to 0, and we get:

[0036]

[0037] Where G = I-1 / (2l)e T e, all elements in e are 1; I is the identity matrix;

[0038] Step 2.3: Fix U and find V and b. The specific steps are as follows:

[0039] Step 2.3.1: Convert equation (1) to:

[0040]

[0041] in, v1, v2…v k is the column vector of V; H=[J,K] nk×2l ,J=[j1,j2,…,j i ,…,j l ] nk×l , K=[k1,k2,…,k a ,…,k l ] nk×l ,

[0042] Step 2.3.2: Calculate the operation from the perspective of physical variables and obtain:

[0043]

[0044] Where M = β1JL d J T +β2KL f K T ;

[0045] Step 2.3.3: Derivative (9) and set it to 0, and we get:

[0046]

[0047] Step 2.4: Repeat steps 2.2 to 2.3 for c iterations to obtain the projection vectors U and V and the constant deviation b.

[0048] Furthermore, in step 3, after obtaining Q, the elements of the vector Q are sorted from large to small, and their corresponding associations are arranged from strong to weak, so as to obtain the corresponding association rules.

[0049] Furthermore, the specific steps of solving the furnace top CO content prediction model in step 5 are as follows:

[0050] Step 5.1: Fix v p , find u p and b p , the specific steps are as follows:

[0051] Define as follows:

[0052] q′=Vec(v p (u p ) T ) T (15)

[0053] Step 5.1.1: Solve the prediction model for the CO content at the furnace top and rewrite Equation (14) as follows:

[0054]

[0055] Formula (16) is further transformed into:

[0056]

[0057] Among them, e p The elements of are all 1, I is the identity matrix; T=[t1,t2,…,t i ,…,t l ] mk×l ,

[0058] Step 5.1.2: Derivative (17) and set it to 0, and we get:

[0059]

[0060] Step 5.2: Fix u p , find v p and b p , the specific steps are as follows:

[0061] Step 5.2.1: Rewrite equation (14) as:

[0062]

[0063] in, N=[n1,n2,…,n i ,…n l ] mk×l ,

[0064] Step 5.2.2: Derivative (20) and set it to 0, and we get:

[0065]

[0066] Step 5.3: Repeat steps 5.1 to 5.2 for R iterations to obtain the left and right projection vectors u. p 、v p and constant deviation b p .

[0067] Furthermore, the specific steps of step 6 include:

[0068] Step 6.1: Use the furnace top camera to collect the image data of the furnace top during the blast furnace industrial production process online, and perform grayscale and normalization processing on the image data;

[0069] Step 6.2: Process the image in step 6.1 using the associated information index obtained in step 2 to obtain the image X with key information. s ;

[0070] Step 6.3: u obtained in step 5 p 、v p and b p As well as the system parameters α and the key information of the furnace top image X s As the input of the CO content calculation model of the furnace top, the CO content in the blast furnace top is measured and calculated.

[0071] The beneficial effect produced by adopting the above technical scheme is that: the soft measurement method of blast furnace top CO content based on multi-perspective data collaborative analysis provided by the present invention realizes the measurement and calculation of blast furnace top CO content using tuyere and furnace top images and physical variable data collected at the industrial site, replaces sensor measurement, solves the difficulties such as sensor installation difficulty, high cost, poor application conditions and easy failure; At the same time, compared with sampling measurement, the delay of the method of the present invention is low and the response speed is fast, which is more in line with the technical requirements of the blast furnace ironmaking industry. Based on the principle of the process of blast furnace ironmaking process and the characteristics of strong coupling, the collected images are uniformly gray-processed, the light and dark features of the original image are retained, and physical variables are introduced. The blast furnace production data are correlated and collaboratively analyzed from multiple perspectives, and then the key information for measuring and calculating the blast furnace CO content is extracted to improve the accuracy of the measurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0072] Figure 1 A schematic diagram of a blast furnace ironmaking process provided by an embodiment of the present invention;

[0073] Figure 2 Flowchart of a soft-sensing method for CO content in a blast furnace top based on collaborative analysis of multi-view data provided by an embodiment of the present invention;

[0074] Figure 3 A comparison chart of measured values ​​and true values ​​provided by an embodiment of the present invention;

[0075] Figure 4 A diagram showing the relative errors between the measured values ​​and the true values ​​provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0076] The following embodiments of the present invention are described in further detail with reference to the accompanying drawings and examples. The following examples are used to illustrate the present invention but are not intended to limit the scope of the present invention.

[0077] like Figure 1 The figure shows a schematic diagram of the production process of blast furnace ironmaking. In the entire process, first, iron ore, coke and other furnace materials are loaded from the upper part of the blast furnace in a certain proportion. High-temperature hot air, coal powder, etc. are blown into the bottom of the blast furnace, where oxygen and coke react to produce carbon monoxide gas. Subsequently, carbon monoxide gradually flows upward in the blast furnace and undergoes a reduction reaction with the iron ore in the upper layer to reduce the iron oxide to iron. At the same time, impurities in the iron ore react with limestone to form slag, which falls to the bottom of the blast furnace together with molten iron and residual coke. When the temperature at the bottom of the blast furnace reaches above 1500°C, the molten iron and slag are separated, and the molten iron is discharged through the iron outlet, while the slag is discharged through the slag outlet. Blast furnace ironmaking not only produces the main product of pig iron, but also produces by-products such as blast furnace gas (CO is the main combustible component) and slag. The blast furnace gas is recycled through the treatment system. As Figure 2 As shown, the method of this embodiment is as follows.

[0078] Step 1: Collect images of the tuyere raceway, furnace top, and related physical variable data as well as CO content data of the furnace top during blast furnace production;

[0079] Step 1.1: Collect image data of the flame in the tuyere raceway of the blast furnace and its related physical variables;

[0080] Step 1.2: Collect image data of the blast furnace top flame and its related physical variables;

[0081] Step 1.3: Collect blast furnace top CO content data;

[0082] In this embodiment, the number of various samples collected under the stable operation condition of the blast furnace is 1500.

[0083] Step 2: Perform collaborative analysis on the data obtained in step 1. The objective function model is:

[0084]

[0085] Among them, X i 、X j are images at adjacent moments under the same viewing angle, X a 、X b is the image of the adjacent moment from another perspective, and the size of the image matrix is ​​m×n; is the projection vector, which is used to analyze the correlation information of multiple perspectives of the sample, l is the number of samples, y i is the kinetic energy of the wind, y a is the CO content at the furnace top; w ij 、w ab is the weight of the physical variable under different perspectives, and β1 and β2 are coefficient parameters.

[0086] Solve the objective function model to obtain U, V and a constant deviation b.

[0087] In this embodiment, m, n, and k are 258, 255, and 15, respectively. The number of samples is l=1500. U and V are matrices of 258×15 and 255×15, respectively. β1=1000, β2=1, and b=-0.0243.

[0088] w ij 、w ab The settings are as follows:

[0089]

[0090] In this embodiment, σ=0.5.

[0091] The specific steps to solve the objective function model are as follows:

[0092] Step 2.1: Set the initial projection matrix V.

[0093] Step 2.2: Fix V and find U and b. The specific steps are as follows:

[0094] Step 2.2.1: Combine the multi-view image operations in equation (1) to obtain:

[0095]

[0096] in, u1, u2…u k is the column vector of U; E=[S,A] 3870×3000 ,A=[A1,A2,…,A a ,…,A l ] 3870×1500 , S=[s1,s2,…,s i ,…,s l ] 3870×1500 , Y is the label vector corresponding to the viewing angle,

[0097] Step 2.2.2: Calculate the operation from the perspective of physical variables and obtain:

[0098]

[0099] Where F = β1SL d S T +β2AL f A T , L d 、L f are the Laplace matrices corresponding to the physical variables of the furnace top and the physical variables of the tuyere, L d 、L f The calculation method is: L = DW, L represents L d or L f , D is a diagonal matrix, W is the corresponding adjacency matrix.

[0100] Step 2.2.3: Derivative (5) and set it to 0, and we get:

[0101]

[0102] Where G = I-1 / (2l)e T e, all elements in e are 1, and I is the identity matrix.

[0103] Step 2.3: Fix U and find V and b. The specific steps are as follows:

[0104] Step 2.3.1: Convert equation (1) to:

[0105]

[0106] in, v1, v2…v k is the column vector of V; H=[J,K] 3825×1500 ,J=[j1,j2,…,j i ,…,j l ] 3825×1500 , K=[k1,k2,…,k a ,…,k l ] 3825×1500

[0107] Step 2.3.2: Calculate the operation from the perspective of physical variables and obtain:

[0108]

[0109] Where M = β1JL d J T +β2KL f K T .

[0110] Step 2.3.3: Derivative (9) and set it to 0, and we get:

[0111]

[0112]

[0113] Step 2.4: Repeat steps 2.2 to 2.3 above for c iterations to obtain the projection vectors U and V and the constant deviation b.

[0114] Step 3: Calculate the index of the associated information under multiple perspectives. The specific steps are as follows:

[0115] Step 3.1: Based on the projection vectors U and V obtained in step 2, we get:

[0116] P=UV T (12)

[0117] Step 3.2: The m×n matrix P obtained by formula (12) is converted into a column vector, that is:

[0118] Q=Vec(P) (13)

[0119] in,

[0120] Step 3.3: Sort the elements of vector Q from large to small, and arrange the corresponding association rules from strong to weak, so as to obtain the index of the corresponding association analysis.

[0121] Step 4: The correlation information index Q obtained in step 2 is applied to the image of the furnace top to obtain the correlation information and the key information for prediction after removing the correlation data.

[0122] Step 5: Using the bilinear regression method, the prediction model for the CO content of each sample furnace top is established as follows:

[0123]

[0124] Among them, u pj 、v pj They are the left projection matrix u p , right projection matrix v p The j-th column vector of ; Set as u p =[u p1 ,u p2 ,...,u pj ,...,u pk ]、v p =[v p1 ,v p2 ,...,v pj ,...,v pk ];X s is the key information for CO content calculation, α is a system parameter, and in this embodiment, α=10000. s is the label vector corresponding to the CO content at the top of the furnace in the CO content prediction model, q = Vec(u p (v p ) T ).

[0125] Solve the furnace top CO content prediction model to obtain u p 、v p and b p .

[0126] The specific steps for solving the furnace top CO content prediction model are as follows:

[0127] Step 5.1: Fix v p , find u p and b p , the specific steps are as follows:

[0128] Define as follows:

[0129] q′=Vec(v p (u p ) T ) T (15)

[0130] Step 5.1.1: Solve the prediction model for the CO content at the furnace top and rewrite Equation (14) as follows:

[0131]

[0132] Formula (16) is further transformed into:

[0133]

[0134] Among them, e p The elements of are all 1, I is the identity matrix; T=[t1,t2,…,t i ,…,t l ] mk×l ,

[0135] Step 5.1.2: Derivative (17) and set it to 0, we can obtain:

[0136]

[0137] Step 5.2: Fix u p , find v p and b p , the specific steps are as follows:

[0138] Step 5.2.1: Rewrite equation (14) as:

[0139]

[0140] in, N=[n1,n2,…,n i ,n l ] mk×l ,

[0141] Step 5.2.2: Derivative (20) and set it to 0, we can obtain:

[0142]

[0143] Step 5.3: Repeat steps 5.1 to 5.2 for 100 iterations to obtain the left and right projection vectors u. p 、v p and constant deviation b p .

[0144] Step 6: Calculate the CO content in the furnace top of the blast furnace online during the ironmaking process. The specific steps are as follows:

[0145] Step 6.1: Use the furnace top camera to collect the image data of the furnace top during the blast furnace industrial production process online, and perform grayscale and normalization processing on the image data;

[0146] Step 6.2: Process the image in step 6.1 using the associated information index obtained in step 2 to obtain the image X with key information. s ;

[0147] Step 6.3: u obtained in step 5 p 、v p and b p As well as the system parameters α and the key information of the furnace top image X s As the input of the calculation model, the prediction calculation of the CO content in the blast furnace top is realized. The calculation formula is:

[0148]

[0149] Figure 3 、 Figure 4 They are the comparison diagram of measured value and true value and the relative error diagram of measured value. Figure 3 、 Figure 4 It can be seen that the present invention has achieved good results in measuring the CO content in the blast furnace top.

[0150] In this example, the RMSE between the sample measured value and the true value during the overall measurement process was 0.181%. This indicates that the present invention can accurately calculate and measure the CO content at the top of a blast furnace during ironmaking, thereby assisting operators in understanding the operating status of the blast furnace and ensuring its safe and stable operation.

[0151] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some or all of the technical features therein. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope defined by the claims of the present invention.

Claims

1. A soft-sensing method for CO content in a blast furnace top based on collaborative analysis of multi-view data, characterized by: The steps include: Step 1: Collect images of the tuyere raceway, furnace top, and related physical variable data as well as CO content data of the furnace top during blast furnace production; Step 1.1: Collect image data of the flame in the tuyere raceway of the blast furnace and its related physical variables; Step 1.2: Collect image data of the blast furnace top flame and its related physical variables; Step 1.3: Collect blast furnace top CO content data; Step 2: Perform collaborative analysis on the data obtained in step 1. The objective function model of data collaborative analysis is: Among them, X i 、X j are images at adjacent moments under the same viewing angle, X a 、X b is the image of the adjacent moment from another perspective, and the size of the image matrix is ​​m×n; is the projection matrix, which is used to analyze the correlation information of multiple perspectives of samples, l is the number of samples, w ij 、w ab is the weight of the physical variable under different perspectives, β1 and β2 are coefficient parameters; y i is the kinetic energy of the wind, y a is the CO content at the furnace top; w ij 、w ab The settings are as follows: Among them, r i 、r j is the relevant physical variable of the tuyere raceway, r a 、r b is the relevant physical variable of the furnace top, σ is the core parameter; Solving the data collaborative analysis objective function model to obtain U, V and a constant deviation b; Step 3: Based on the projection matrices U and V obtained in step 2, calculate the index Q of the correlation information under multiple perspectives; First we get P = UV T , P is an m×n matrix; convert the matrix P into a column vector, that is: Q=Vec(P), Step 4: The correlation information index Q obtained in step 2 is applied to the image of the furnace top to obtain correlation information and calculate the key information for prediction after removing the correlation data; Step 5: Using the bilinear regression method, the prediction model for the CO content of each sample furnace top is established as follows: Among them, u pj 、v pj They are the left projection matrix u p , right projection matrix v p The j-th column vector of ; Set as u p =[u p1 ,u p2 ,...,u pj ,...,u pk ]、v p =[v p1 ,v p2 ,...,v pj ,...,v pk ];X s is the key information for CO content calculation, b p is a constant deviation, α is a system parameter, y s is the label vector corresponding to the CO content at the top of the furnace in the CO content prediction model, q = Vec(u p (v p ) T ); Solve the furnace top CO content prediction model to obtain u p 、v p and b p ; Step 6: Use the furnace top camera to collect the furnace top image data during the blast furnace industrial production process. Based on the solutions of steps 2 to 5, the furnace top CO content is measured online during the ironmaking blast furnace production process. The calculation model for the furnace top CO content is:

2. The soft sensing method for blast furnace top CO content based on multi-view data collaborative analysis according to claim 1 is characterized in that: The specific steps for solving the objective function model in step 2 are as follows: Step 2.1: Set the initial projection matrix V; Step 2.2: Fix V, calculate U and b, and we get: in, u1, u2…u k is the column vector of U; E=[S,A] mk×2l ,A=[A1,A2,…,A a ,…,A l ] mk×l , S=[s1,s2,…,s i ,…,s l ] mk×l , F=β1SL d S T +β2AL f A T , L d 、L f are the Laplace matrices corresponding to the physical variables of the furnace top and the physical variables of the tuyere, L d 、L f The calculation method is: L = DW, L represents L d or L f , D is a diagonal matrix, W is the corresponding adjacency matrix; Y is the label vector of the corresponding view, G=I-1 / (2l)e T e, all elements in e are 1; I is the identity matrix; Step 2.3: Fix U, calculate V and b, and we get: in, v1, v2…v k is the column vector of V; H=[J,K] nk×2l ,J=[j1,j2,…,j i ,…,j l ] nk×l , K=[k1,k2,…,k a ,…,k l ] nk×l , M=β1JL d J T +β2KL f K T ; Step 2.4: Repeat steps 2.2 to 2.3 for c iterations to obtain the projection vectors U and V and the constant deviation b.

3. The soft sensing method for blast furnace top CO content based on multi-view data collaborative analysis according to claim 2 is characterized in that: The specific steps of step 2.2 are as follows: Step 2.2.1: Combine the multi-view image operations in equation (1) to obtain: Step 2.2.2: Calculate the operation from the perspective of physical variables and obtain: Step 2.2.3: Derivative (5) and set it to 0 to obtain (6) and (7).

4. The soft sensing method for blast furnace top CO content based on multi-view data collaborative analysis according to claim 2 is characterized in that: The specific steps of step 2.3 are as follows: Step 2.3.1: Convert equation (1) to: Step 2.3.2: Calculate the operation from the perspective of physical variables and obtain: Step 2.3.3: Derivative (9) and set it to 0 to obtain (10) and (11).

5. The soft sensing method for blast furnace top CO content based on multi-view data collaborative analysis according to claim 2 is characterized in that: In step 3, after obtaining Q, the elements of vector Q are sorted from large to small, and their corresponding associations are arranged from strong to weak, so as to obtain the corresponding association rules.

6. The soft sensing method for blast furnace top CO content based on multi-view data collaborative analysis according to claim 5 is characterized in that: The specific steps of solving the furnace top CO content prediction model in step 5 are as follows: Step 5.1: Fix v p , find u p and b p , the specific steps are as follows: Define as follows: q′=Thing(in p (at p ) T ) T (15) Step 5.1.1: Solve the prediction model for the CO content at the furnace top and rewrite Equation (14) as follows: Formula (16) is further transformed into: Among them, e p The elements of are all 1, I is the identity matrix; T=[t1,t2,…,t i ,…,t l ] mk×l , Step 5.1.2: Derivative (17) and set it to 0, and we get: Step 5.2: Fix u p , find v p and b p , the specific steps are as follows: Step 5.2.1: Rewrite equation (14) as: in, N=[n1,n2,…,n i ,…n l ] mk×l , Step 5.2.2: Derivative (20) and set it to 0, and we get: Step 5.3: Repeat steps 5.1 to 5.2 for R iterations to obtain the left and right projection vectors u. p 、v p and constant deviation b p .

7. The soft sensing method for blast furnace top CO content based on multi-view data collaborative analysis according to claim 6 is characterized in that: The specific steps of step 6 include: Step 6.1: Use the furnace top camera to collect the image data of the furnace top during the blast furnace industrial production process online, and perform grayscale and normalization processing on the image data; Step 6.2: Process the image in step 6.1 using the associated information index obtained in step 2 to obtain the image X with key information. s ; Step 6.3: u obtained in step 5 p 、v p and b p As well as the system parameters α and the key information of the furnace top image X s As the input of the CO content calculation model of the furnace top, the CO content in the blast furnace top is measured and calculated.

Images