Methods for modeling and estimating the stability of coke after reaction, methods for estimating the stability of coke after reaction, and methods for producing coke.
Patent Information
- Application Number
- VN1202308404
- Authority / Receiving Office
- VN · VN
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2021-06-28
- Filing Date
- 2022-05-31
- Publication Date
- 2024-08-26
- Estimated Expiration
- 2042-05-31
AI Technical Summary
Current methods for estimating Coke Strength after Reaction (CSR) are time-consuming, leading to delayed identification of coke strength issues in blast furnaces, which can cause operational disruptions and inefficiencies in iron production.
A method using partial least squares regression to generate a model for estimating post-reaction coke strength, analyzing constituent elements via atomic spectroscopy techniques like X-ray fluorescence and ICP mass spectrometry, allowing for rapid and accurate prediction of coke strength post-reaction.
This approach significantly reduces the time required for CSR estimation, improving the accuracy and speed of coke strength evaluation, enabling timely adjustments to stabilize blast furnace operations and optimize coke production.
Smart Images
Figure VN1202308404_0
Abstract
Description
Method for generating a model for estimating post-reaction strength of coke, method for estimating post-reaction strength of coke, and method for manufacturing coke
[0001] The present invention relates to a method for generating a model for estimating post-reaction strength of coke, a method for estimating post-reaction strength of coke, and a method for producing coke.
[0002] In the production of pig iron in a blast furnace, coke is an essential material as a reducing agent for iron ore because it is porous and has the permeability to allow gases blown in from the bottom of the blast furnace to escape to the top.
[0003] In order to ensure gas permeability in a blast furnace, sufficient strength is required to prevent coke from spalling. Generally, values based on test results in room-temperature air, such as rotary drum strength tests, are used to measure coke strength. However, the inside of a blast furnace is a high-temperature environment exceeding 1000°C, containing gases that are reactive with coke, primarily carbon dioxide. Therefore, it is difficult to say that coke strength indicators based on ordinary test results represent the strength of coke in a blast furnace.
[0004] Here, coke strength after reaction (CSR) has been proposed as an index for accurately estimating the strength of coke in a blast furnace, and blast furnace operation is carried out based on CSR. "After reaction" refers to the time after coke is charged into a blast furnace and reacts with gases such as carbon dioxide in a high-temperature environment exceeding 1000°C. Furthermore, "strength after reaction" refers to the strength of coke after the reaction.
[0005] For example, if coke with a CSR lower than the target value is charged into a blast furnace, the coke will break down inside the furnace, producing coke dust. Coke dust not only blocks the flow of reducing gas inside the blast furnace, but also inhibits the smooth flow of the generated molten iron to the lower part of the blast furnace. This prevents efficient iron production, which in some cases could cause serious problems, such as shutting down operations. For this reason, it is important to accurately understand CSR.
[0006] The measurement procedure for CSR is specified in the ISO standard "ISO 18894 Coal and Coke Testing" and the ASTM standard "ASTM D 5341 Coal and Coke Testing" (hereinafter simply referred to as "ASTM standard"), which is a product specification and test method. For example, in the method specified in the ASTM standard, coke of a certain particle size is held in a carbon dioxide atmosphere at 1100°C for 2 hours, and then rotated 600 times in an I-type drum testing machine, and the lump residual ratio after that is measured as CSR. Usually, a lower limit is set for CSR, and operation is carried out so that the CSR does not fall below that lower limit.
[0007] One problem with CSR measurement is the length of time it takes to obtain CSR results. As mentioned above, the ASTM standard requires a two-hour reaction between coke and carbon dioxide, but the furnaces used to carry out this reaction are large, requiring even longer heating and cooling times. Furthermore, when considering the series of steps involved, such as preparing the coke in advance and measuring the results using a drum tester after heating in the furnace, the total time required can sometimes be as long as one day.
[0008] Furthermore, even if CSR measurements are started on coke immediately after it is discharged from the coke oven during normal operation, by the time the CSR measurement results are obtained, the coke has already been discharged from the coke oven and is being fed into the blast furnace. Therefore, even if an abnormality in the coke is discovered as a result of the CSR measurement, immediate action cannot be taken in blast furnace operations, resulting in the potential risk of delayed improvement measures and destabilization of blast furnace operations. For this reason, rapid CSR evaluation is required.
[0009] Various studies have been conducted to speed up the evaluation of CSR. Patent Document 1 discloses a method for estimating CSR using the intensity ratio of specific peaks by using Raman spectroscopy.
[0010] Furthermore, the use of the Coke Reaction Index (CRI), which has a high correlation with CSR, has also been considered. Patent Document 2 discloses a method in which the CRI of a coal blend is calculated by taking a weighted average of the CRI and total expansion coefficient of each single coal, and the CSR is estimated from the CRI of the coal blend and the drum strength of the coal blend.
[0011] Furthermore, Patent Document 3 discloses a method for estimating CSR using the abundance ratio of inorganic components in coal and various physical property values of raw coal.
[0012] JP 2019-163986 A JP 2005-232350 A JP 2001-172643 A
[0013] However, in any of the estimation methods disclosed in Patent Documents 1 to 3, the analytical values and physical property values to be applied cannot necessarily be obtained in a short time, and may take a long time depending on the parameters used for estimation.
[0014] Furthermore, the estimation methods disclosed in Patent Documents 2 and 3 use analytical values and physical property values obtained from coal before coking, but there is a problem in that operational errors that inevitably occur during the subsequent coking process and the resulting variations in operational data become sources of errors in estimating CSR.
[0015] Furthermore, the methods disclosed in Patent Documents 2 and 3 that use weighted averages or multiple regression analysis aim to determine the blending ratio of single coals on the premise that coke of a coal blend is produced from single coals, and the problem of CSR estimation error is unavoidable. Furthermore, it is impossible to estimate CSR for a coal blend coke whose blending ratio of single coals is unknown.
[0016] The present invention has been made in consideration of the above circumstances, and aims to provide a method for generating a coke post-reaction strength estimation model, a method for estimating coke post-reaction strength, and a method for manufacturing coke, which improve the accuracy of estimating coke post-reaction strength and enable quicker estimation of coke post-reaction strength.
[0017] The present invention, which solves the above-mentioned problems, has the following essential features: [1] A method for generating a model for estimating post-reaction strength of coke, comprising: a step of analyzing the constituent elements of coke; a step of acquiring the post-reaction strength of the coke; and a model generation step of generating a model for estimating post-reaction strength using the analyzed values of the constituent elements as explanatory variables and the post-reaction strength as a response variable. [2] The method for generating a model for estimating post-reaction strength of coke according to [1], in which the model generation step generates the post-reaction strength estimation model by partial least squares regression. [3] The method for generating a model for estimating post-reaction strength of coke according to [1] or [2], in which the analysis of the constituent elements is performed by atomic spectroscopy. [4] The method for generating a model for estimating post-reaction strength of coke according to [3], wherein the analysis of the constituent elements by atomic spectrometry is performed by any one of X-ray fluorescence spectroscopy, laser-induced breakdown spectroscopy, laser ablation ICP mass spectrometry, ICP mass spectrometry (wet dissolution - ICPMS), atomic absorption spectrometry (wet dissolution - atomic absorption spectrometry), or wet dissolution - ICP optical emission spectrometry. [5] The method for generating a model for estimating post-reaction strength of coke according to any one of [1] to [4], wherein the analytical values are numerical information based on any two or more elements of Fe, Ca, Mg, Al, Si, K, Ti, Mn, Cu, and Zn. [6] A method for estimating post-reaction strength of coke, which estimates the post-reaction strength of coke using the post-reaction strength estimation model generated by the method for estimating post-reaction strength of coke according to any one of [1] to [5]. [7] A method for producing coke, comprising estimating the post-reaction strength of the coke by the method for estimating the post-reaction strength of the coke described in [6], and changing the coke production conditions based on the estimated post-reaction strength. [8] The method for producing coke described in [7], wherein the change in the production conditions is a change in the coal blending ratio of the coke.
[0018] According to the present invention, it is possible to improve the accuracy of estimating the post-reaction strength of coke and realize rapid estimation of the post-reaction strength of coke. As a result, it is possible to stabilize blast furnace operation. Furthermore, it is possible to optimize the coal blend of coke based on the estimation result of the post-reaction strength of coke, thereby improving the quality of coke and stabilizing the operation of coke ovens.
[0019] 10 is a graph showing the relationship between actually measured post-reaction intensity and estimated post-reaction intensity.
[0020] A method for carrying out the present invention will be described below. The method for generating a model for estimating post-reaction strength of coke according to the present invention includes a step of analyzing the constituent elements of coke, a step of acquiring the post-reaction strength of the coke, and a model generation step of generating a model for estimating post-reaction strength using the analyzed values of the constituent elements as explanatory variables and the post-reaction strength as a response variable.
[0021] Here, some of the constituent elements of coke may have a certain correlation with each other, and the estimation accuracy may be degraded by a simple multiple regression analysis due to collinearity based on the correlation. Therefore, an analytical method that does not cause the collinearity problem in principle is desirable. A representative example of such an analytical method is partial least squares regression (PLS).
[0022] Partial least squares regression converts explanatory variables into mutually uncorrelated principal component axes, and then performs regression analysis with the target variable using only a small number of principal components. This is convenient when estimating post-reaction intensity from multiple analytical values where correlations are recognized between explanatory variables, and can achieve high estimation accuracy.
[0023] The explanatory variables are analytical values based on the physically observed quantities of the constituent elements of the coke. That is, partial least squares regression is performed using the post-reaction strength as the objective variable and the analytical values based on the physically observed quantities of the constituent elements of the coke as the explanatory variables. Specifically, for samples whose post-reaction strengths have been obtained, regression analysis is performed using the post-reaction strength as the objective variable and the corresponding analytical values as the explanatory variables, and a model equation, which is a regression formula that defines the relationship between the objective variable and the explanatory variables, is calculated. Thereafter, for samples whose post-reaction strengths are unknown, the analytical values obtained from the sample are substituted as explanatory variables into the calculated model equation to calculate the post-reaction strength, which is the objective variable.
[0024] The analytical values based on the physical observation quantities of the constituent elements designated as explanatory variables are preferably numerical information reflecting any two or more elements of Fe, Ca, Mg, Al, Si, K, Ti, Mn, Cu, and Zn. Since carbon, which is the matrix (base material) of the coke, may interfere with the analysis, Fe, which can be obtained by analyzing the coke ash after burning the coke, may be used. ２ O ３ , CaO, Na ２ O, MgO, Al ２ O ３ , SiO ２ , K. ２ O, TiO ２ It is more preferable to use numerical information on two or more of MnO, NiO, CuO and ZnO.
[0025] In addition, coke ash can be obtained, for example, by a combustion method described in the Japanese Industrial Standard "JIS M8815 Analysis Method for Coal Ash and Coke Ash" (hereinafter simply referred to as "JIS Standard"). Here, an example of the numerical information is the analytical value (%) of the constituent elements of the coke.
[0026] Here, the measured (obtained) post-reaction strength and the analyzed values (%) of the constituent elements of the coke ash after combustion for the 10 types of coke samples used in the partial least squares regression are shown in Table 1. Also, based on Table 1, Fe was used as an explanatory variable. ２ O ３ , CaO, Na ２ O, MgO, Al ２ O ３ , SiO ２ , K. ２ O, TiO ２ , MnO, NiO, CuO, and ZnO were set, and a model equation (Equation 1) based on partial least squares regression was generated. The regression equation based on Table 1 was finally expressed as a weighted sum (see Table 2) in the form shown in Equation 1.
[0027]
[0028] Y(CSR) = k ０ +k １ ×X1+k ２ ×X ２ +k ３ ×X ３+...k １２ ×X １２ (Formula 1)
[0029]
[0030] The relationship between the post-reaction strength estimated by substituting the analytical values of the constituent elements shown in Table 1 into the model formula and the actually measured post-reaction strength is shown in Figure 1. The horizontal axis in the figure shows the actually measured post-reaction strength (hereinafter referred to as "actually measured post-reaction strength"), and the vertical axis shows the estimated post-reaction strength (hereinafter referred to as "estimated post-reaction strength"). As shown in Figure 1, the correlation coefficient (R ２ ) is 0.9219, which shows that there is an extremely high correlation between the two. Therefore, the model formula based on partial least squares regression can predict the post-reaction strength with high accuracy.
[0031] The measured post-reaction strength is a value obtained by a method conforming to ASTM standards, and the analytical values of the constituent elements of the coke are analytical values of the coke ash after combustion conforming to JIS standards.
[0032] Furthermore, for analytical values based on physically observable quantities of the constituent elements of coke, it is preferable to perform elemental analysis using collected coke or post-combustion coke ash pulverized to 150 μm or less as an analytical sample. Because coke consists of a heterogeneous mixture of complex compounds and the elements vary depending on the analytical location, this method aims to eliminate bias in the sample and minimize variation in analytical values. The pulverized sample is then subjected to elemental analysis. Any analytical method capable of elemental analysis is sufficient, with atomic spectroscopy being preferred. Examples of atomic spectroscopy include compressing the pulverized sample into pellets and measuring them using X-ray fluorescence analysis, laser-induced breakdown spectroscopy, or laser ablation-ICP mass spectrometry (LA-ICPMS). Alternatively, a solution sample obtained by chemical dissolution can be measured using ICP atomic emission spectrometry (wet dissolution-ICP atomic emission spectrometry), ICP mass spectrometry (wet dissolution-ICPMS), or atomic absorption spectrometry (wet dissolution-atomic absorption spectrometry).
[0033] Furthermore, analytical values based on the physically observable quantities of the constituent elements of coke can be obtained within about one hour of collection if the pellets are analyzed directly (X-ray fluorescence analysis, laser-induced breakdown spectroscopy, or LA-ICPMS). Even if the pellets are analyzed as a solution sample (wet dissolution-ICP atomic emission spectroscopy, wet dissolution-ICPMS, or wet dissolution-atomic absorption spectroscopy), analytical results can be obtained within about two to three hours. Furthermore, even if the analysis is performed on coke ash after combustion, the combustion time is about two hours, so the post-reaction strength can be calculated within about three to five hours.
[0034] During actual operation, if the estimated post-reaction strength deviates from the operationally acceptable range of post-reaction strength, it becomes possible to quickly stop charging coke into the blast furnace, thereby stabilizing blast furnace operation. Furthermore, blending ratio management, such as changing the coal blending ratio of the coke, can be efficiently performed based on the estimated post-reaction strength of the coke, enabling optimization of the coal blending ratio. Furthermore, coke production conditions can be quickly changed by changing the coal blending ratio, thereby improving coke quality and stabilizing coke oven operation. Here, changing the coke production conditions based on the post-reaction strength of the coke corresponds to the "coke production method" in this invention. Furthermore, changing the coke production conditions includes changing the coal blending ratio based on the post-reaction strength of the coke.
[0035] As described above, in the method for generating a coke post-reaction strength estimation model according to the present invention, a coke post-reaction strength estimation model is generated in advance by a technique such as partial least squares regression, using analytical values based on physically observed quantities of the coke's constituent elements as explanatory variables and the coke's post-reaction strength as a response variable. Then, even during blast furnace operation, the coke's post-reaction strength can be quickly estimated by simply substituting the analytical values of the coke's constituent elements into the coke post-reaction strength estimation model, allowing for a prompt determination of whether the coke's post-reaction strength deviates from the operationally acceptable range. Furthermore, by using a coke post-reaction strength estimation model based on partial least squares regression, the coke post-reaction strength can be predicted with high accuracy. Here, the estimation of the coke post-reaction strength using the coke post-reaction strength estimation model corresponds to the "coke post-reaction strength estimation method" in the present invention.
[0036] The present inventors have considered that, in order to accurately estimate CSR (hereinafter also referred to simply as "post-reaction strength"), it is desirable not to use physical property values and analytical values before coking, which may fluctuate during the coking process, but to use parameters of the coke after it has become coke. As a result of careful investigation of the physical property values and analytical values of coke, they have concluded that the post-reaction strength of the coke can be estimated from analytical values of the constituent elements of the coke, which are routinely obtained in each coke production process.
[0037] Some correlation was observed between the strength after reaction and the constituent elements, so an attempt was made to estimate the strength after reaction using the least squares method based on the relationship between specific constituent elements and the strength after reaction. However, there are problems with accuracy when using only the analytical value of a single constituent element (simple regression analysis).
[0038] The analytical values of about a dozen constituent elements of coke are routinely obtained. Therefore, we thought that an estimation method that utilizes the analytical values of these multiple constituent elements, and even applies analytical values based on the physical observations of the constituent elements, such as spectral data obtained during coke analysis, would enable accurate determination of post-reaction strength. Therefore, we came up with the idea of applying multivariate analysis, which is used to summarize and predict multiple related variables. Specifically, we performed multivariate analysis, using post-reaction strength as the dependent variable and analytical values based on the physical observations of the constituent elements of coke as the explanatory variables.
[0039] Hereinafter, an example in which the post-reaction strength of coke is estimated using the post-reaction strength estimation model according to this embodiment will be described.
[0040] The analytical values (%) of the constituent elements in the coke ash after combustion for the three types of coke samples (A to C) are shown in Table 3. The analytical values of the constituent elements were obtained by incinerating the coke in accordance with JIS standards, crushing it into pellets with a size of 150 μm or less, and then subjecting the pellets to X-ray fluorescence analysis.
[0041]
[0042] The post-reaction strength was estimated by applying the analytical values of the constituent elements of these three types of coke (A to C) to a post-reaction strength estimation model, which was the result of partial least squares regression performed based on the post-reaction strength and analytical values of the constituent elements of the 10 types of coke shown in Table 1. The post-reaction strength was also measured according to the ASTM standard, and the estimated post-reaction strength and the measured post-reaction strength were compared. Note that the partial least squares regression was performed using "OriginPro2017" (registered trademark) and a cross-validation method was applied.
[0043] The results of the estimated post-reaction strength and the actually measured post-reaction strength are shown in Table 4. As a comparative example, the results of estimating the post-reaction strength by applying the analytical values of the constituent elements of three types of coke (A to C) to a multiple regression model that is the result of multiple regression performed based on the post-reaction strength and analytical values of the constituent elements of the 10 types of coke shown in Table 1 are also shown.
[0044]
[0045] The results of post-reaction intensity estimation using partial least squares regression (listed as "PLS" in Table 4) and multiple regression (listed as "multiple regression" in Table 4) both use Fe as an explanatory variable. ２ O ３ , CaO, Na ２ O, MgO, Al ２ O ３ , SiO ２ , K. ２ O, TiO ２ , MnO, NiO, CuO and ZnO were set, model equations for each were generated, and the strength after reaction was estimated.
[0046] The measured post-reaction strength and the results of post-reaction strength estimated by partial least squares regression were in excellent agreement, with σd (standard deviation of error), which indicates accuracy, being 0.98 for the three types of coke (A to C). On the other hand, the degree of deviation between the measured post-reaction strength and the results of post-reaction strength estimated by multiple regression was larger than in the case of the results of estimation by partial least squares regression, with σd being 5.3.
[0047] Next, estimation of the post-reaction strength by partial least squares regression was performed by reducing the number of constituent elements (number of oxide species) used as explanatory variables, and the results were compared with the actually measured post-reaction strength. That is, in Example 2, the generation of the post-reaction strength estimation model and the regression method were performed in the same manner as in Example 1, except that the number of constituent elements used as explanatory variables was reduced.
[0048] Specifically, a post-reaction strength estimation model was created based on the post-reaction strength and the analytical values of the constituent elements of the 10 types of coke shown in Table 1, and the post-reaction strength of the three types of coke (A to C) shown in Table 3 was estimated by applying the analytical values of the constituent elements to the post-reaction strength estimation model. ２ O ３ , CaO, MgO, Al ２ O ３ , SiO ２ and TiO ２ When the six constituent elements of Al are set (described as "Explanatory variables: 6 constituent elements" in Table 5), ２ O ３ , SiO ２ and TiO ２ The results are shown in Table 5 for the case where the three constituent elements were set (described as "explanatory variables: three constituent elements" in Table 5).
[0049]
[0050] As shown in Table 5, the measured post-reaction strengths and estimated post-reaction strengths of the three types of cokes (A to C) were in excellent agreement, and the σd (standard deviation of error), which indicates the accuracy, was 0.71 and 0.44 when the explanatory variables were six constituent elements and three constituent elements, respectively.
[0051] As described above, according to the present invention, by preparing a post-reaction strength estimation model in advance, it is possible to estimate the post-reaction strength of coke only from the analytical values of the constituent elements of the coke, without having to measure the post-reaction strength of the coke based on the ASTM standard each time. Furthermore, by using partial least squares regression as a method of multivariate analysis, it is possible to estimate the post-reaction strength with high accuracy.
[0052] In this example, the time required to determine the post-reaction strength of the coke by the method based on the ASTM standard was 9.5 hours, whereas the time required to estimate the post-reaction strength of the coke according to the present invention was approximately 3 hours. Therefore, the method for estimating the post-reaction strength of the coke according to the present invention also has a large contribution in terms of time.
Claims
1. A step of analyzing the constituent elements of coke, a step of obtaining the post-reaction strength of the coke, and a model generation step of generating a post-reaction strength estimation model using the analysis values of the constituent elements as explanatory variables and the post-reaction strength as the target variable. A method for generating a post-reaction strength estimation model of coke having these steps.
2. The method for generating a post-reaction strength estimation model of coke according to claim 1, wherein the model generation step generates the post-reaction strength estimation model by partial least squares regression.
3. The method for generating a post-reaction strength estimation model of coke according to claim 1 or 2, wherein the analysis of the constituent elements is performed by atomic spectrum analysis.
4. The analysis of the constituent elements by the atomic spectrum analysis method is performed by any one of fluorescence X-ray analysis, laser-induced breakdown spectroscopy, laser ablation ICP mass spectrometry, ICP mass spectrometry (wet digestion-ICPMS), atomic absorption spectrometry (wet digestion-atomic absorption spectrometry), or wet digestion-ICP emission spectrometry. The method for generating a post-reaction strength estimation model of coke according to claim 3.
5. The analysis value is numerical information based on any two or more elements of Fe, Ca, Mg, Al, Si, K, Ti, Mn, Cu, and Zn. The method for generating a post-reaction strength estimation model of coke according to any one of claims 1 to 4.
6. A method for estimating the post-reaction strength of coke by using the post-reaction strength estimation model generated by the method for generating a post-reaction strength estimation model of coke according to any one of claims 1 to 5.
7. A method for manufacturing coke, which estimates the post-reaction strength of coke by the method for estimating the post-reaction strength of coke according to claim 6 and changes the manufacturing conditions of coke based on the estimated post-reaction strength.
8. The method for manufacturing coke according to claim 7, wherein the change in the manufacturing conditions is a change in the coal blending ratio of coke.