Creation method of strength estimation model for sintered ore, strength estimation method for sintered ore and producing method of sintered ore

Abstract

To provide a creation method of a strength estimation model of sintered ore capable of estimating the strength of the sintered ore more accurately than ever.SOLUTION: A method for creating a strength estimation model of sintered ore includes a strength acquisition process for obtaining the strength of the sintered ore, a domain fraction calculation process for dividing the texture images of the sintered ore into multiple domains and calculating the area fraction of the domains for each first texture, and a model creation process for creating a strength estimation model for the sintered ore using the area fraction as an explanatory variable and the strength as a target variable, wherein the domain is defined as a region consisting of the continuous first texture and one or both of the textures and voids other than the first texture dispersed within the first texture.SELECTED DRAWING: None

Description

[Technical field]

[0001] The present invention relates to a method for creating a sintered ore strength estimation model, a method for estimating sintered ore strength, and a method for producing sintered ore. [Background technology]

[0002] Sintered ore is a blast furnace charge in blast furnace operation. The raw material for sintered ore is a sintered ore blended raw material obtained by blending multiple brands of fine ores with appropriate amounts of auxiliary raw material powder, miscellaneous raw materials, and solid fuel. The sintered ore blended raw material is mixed and granulated by adding water as a binder, and the granulated raw material obtained is charged into a Dwight Lloyd type (DL) sintering machine or the like and fired to produce sintered ore. The produced sintered ore is discharged, crushed by a crusher, and sized by a sieve (crushing and sieving process). In order to maintain good air permeability in the blast furnace, the product sintered ore needs to be somewhat coarse, so the undersieve is returned to the raw material as return ore, and the oversieve is sent to the blast furnace as product sintered ore. Therefore, in order to increase the ratio of product sintered ore to the production volume of sintered ore (yield of product sintered ore), it is necessary to increase the ratio of sintered ore that is somewhat coarse, and for this purpose, the produced sintered ore needs to have high strength.

[0003] In general, the strength of sintered ore can be evaluated using a drop strength index determined by a drop strength test. The drop strength index is a value corresponding to the sieve particle size after a certain drop impact is applied to the sintered ore. However, even if the drop strength index is measured, the cause of improving or deteriorating the strength is often unknown. In other words, in order to efficiently improve the yield, it is not sufficient to directly measure the strength of sintered ore using a method such as a drop strength measurement. Therefore, it is required to be able to explain the reason for the change in strength using an index other than strength.

[0004] In light of the above discussion, an index capable of estimating the strength of sintered ore with higher accuracy and a method for estimating strength with higher accuracy using the index are being studied.

[0005] Patent Document 1 discloses a strength estimation method for determining the strength of sintered ore from the phase fraction of the mineral phase. Specifically, the phase fraction of the mineral phase is determined by applying Rietveld analysis to a diffraction pattern obtained by X-ray diffraction. Among the phase fractions, SFCA (Silico-ferrite of calcium and aluminum, Ca2(Fe,Ca)6(Fe,Al,Si)6O 20 A method for estimating strength from the phase fraction of

[0006] On the other hand, Patent Document 2 proposes a learning model that estimates the structural composition of a composite structure formed by a mineral phase from a cross-sectional image of sintered ore. Here, examples of the structural composition of the composite structure disclosed include a first composite structure in which either hematite or magnetite is surrounded by calcium ferrite, a second composite structure in which a composite structure of calcium ferrite and silicate slag is formed, a third composite structure in which a composite structure of either hematite or magnetite and calcium ferrite is formed, and a fourth composite structure in which a composite structure of either hematite or magnetite and silicate slag is formed. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] JP 2018-179690 A [Patent Document 2] JP 2021-166002 A Summary of the Invention [Problem to be solved by the invention]

[0008] However, it has been found that the method for evaluating the strength of sintered ore disclosed in Patent Document 1 has low accuracy.

[0009] Moreover, the method described in Patent Document 2 could not be used to accurately estimate the strength of sintered ore.

[0010] The present invention has been made in view of the above circumstances, and has an object to provide a method for creating a sintered ore strength estimation model that can estimate the strength of sintered ore with higher accuracy than conventional methods. [Means for solving the problem]

[0011] As a result of intensive research, the present inventors have found that the above object can be achieved by adopting the following configuration.

[0012] 1. A strength obtaining process for obtaining the strength of sintered ore; a domain fraction calculation step of dividing the structural image of the sintered ore into a plurality of domains and calculating an area fraction of the domains for each first structure of the domain; A model creation process for creating a strength estimation model for sintered ore by using the area fraction as an explanatory variable and the strength as a response variable, A method for creating a strength estimation model for sintered ore, wherein the domain is defined as a region consisting of a continuous first structure and one or both of structures other than the first structure and voids dispersed in the first structure.

[0013] 2. A method for creating a strength estimation model for sintered ore as described in 1 above, wherein the domains consist of a domain having primary hematite as the first structure, a domain having secondary hematite as the first structure, a domain having magnetite as the first structure, and a domain having calcium ferrite as the first structure.

[0014] 3. A method for estimating the strength of sintered ore, comprising estimating the strength of sintered ore using a strength estimation model for sintered ore created by the method for creating a strength estimation model for sintered ore described in 1 or 2 above.

[0015] 4. A method for producing sintered ore, comprising changing production conditions for sintered ore based on the strength of the sintered ore estimated by the method for estimating the strength of sintered ore described in 3 above.

[0016] 5. The method for producing sintered ore according to claim 4, wherein the change in the production conditions is a change in the ratio of components constituting the sintered ore blending raw material. Effect of the Invention

[0017] According to the present invention, it is possible to provide a method for creating a sintered ore strength estimation model that can estimate the strength of sintered ore with higher accuracy than conventional methods. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0018] The present inventors have conducted a detailed study on the reason why the strength of sintered ore cannot be accurately determined from the phase fraction of the mineral phase, and have obtained the following findings.

[0019] Generally, sintered ore is mainly composed of four mineral phases, hematite (Fe2O3), magnetite (Fe3O4), calcium ferrite (CaO·Fe2O3), and silicate slag (CaO·SiO2), as well as voids. It has been known that the amount of calcium ferrite phase among them affects the drop strength of sintered ore. Furthermore, calcium ferrite phases are mainly classified into SFCA phase and SFCA-I phase due to differences in composition and crystal structure, and it has been reported that the phase fraction of the SFCA phase shows a stronger correlation with the strength of sintered ore than other calcium ferrite phases (Patent Document 1). However, in sintered ore, multiple mineral phases are not distributed as single phases, but many mixed phases in which the mineral phases are mixed in a complex manner are formed. Therefore, it is expected that the strength of the mixed phase region changes depending on the type and amount of the mineral phase that constitutes the mixed phase, and the strength of the sintered ore also changes depending on the proportion of the mixed phase in the sintered ore. For these reasons, when a method for predicting the strength of sintered ore from the phase fraction of each single phase is used, the prediction accuracy of the strength of sintered ore is low.

[0020] Therefore, it is important to define a domain, which is a region including a continuous first structure, as a concept corresponding to a mixed phase region, and to determine the fraction of the domain for each first structure.

[0021] An embodiment of the present disclosure will be described below.

[0022] A method for creating a strength estimation model for sintered ore according to an embodiment of the present disclosure includes a strength acquisition step, a domain fraction calculation step, and a model creation step.

[0023] (Sintered ore) As the sintered ore according to the present embodiment, any sintered ore can be used. For example, the sintered ore can be produced by any method. An example of a method for producing sintered ore will be described below.

[0024] In one embodiment of the present disclosure, sintered ore can be produced by adding water as a binder to a sintered ore blended raw material, mixing and granulating the mixture, and firing the resulting granulated raw material.

[0025] Sintered ore is produced using a sintered ore blended raw material. Examples of components constituting the sintered ore blended raw material include fine ore, auxiliary raw material powder, miscellaneous raw materials, and solid fuel. As the fine ore, multiple brands of fine ore may be used. Examples of the auxiliary raw material powder include limestone, silica stone, and serpentine. Examples of the miscellaneous raw materials include dust, scale, and return ore. Examples of the solid fuel include fine coke. The ratio of the above components can be appropriately changed.

[0026] First, the sintered ore blend raw material is granulated, and then sintered in a sintering machine to obtain a sintered cake. The sintering machine includes, for example, a Dwight Lloyd type sintering machine and a sintering pot. The sintering pot includes, for example, a Greenawalt type sintering machine.

[0027] Next, a sintered ore sample is collected from the sintered cake. The collected sintered ore sample is then pulverized. The pulverization method is not particularly limited as long as it does not affect the mineral phase. Specifically, a method using a pulverizing device such as a vibration mill, a ball mill, a rotary mill, a stamp mill, or a crusher is common. In addition, pulverization may be performed by two or more methods. For example, the sintered cake may be coarsely pulverized with a crusher and then further pulverized with a rotary mill.

[0028] In addition, the sintered ore can be cooled before or after the sintered cake is crushed. For example, when using a sintered cake discharged from an actual sintering device, a method can generally be used in which the obtained sintered cake is coarsely crushed by a crusher and then introduced into a cooler for cooling. On the other hand, when the sintered cake is obtained using a laboratory-scale sintering pot, a method can also be used in which the sintered cake is crushed after cooling.

[0029] The method for creating a strength estimation model for sintered ore according to the present embodiment may include a step of collecting sintered ore. The collecting step may be a step of collecting sintered ore after crushing. The method for collecting sintered ore is not particularly limited, but for example, a method of collecting sintered ore using a dedicated box-shaped sampler may be used. A method of collecting sintered ore flowing on a belt conveyor may also be used. For example, a method of manually collecting sintered ore from the belt conveyor using a ladle or the like may be used, or a method of continuously collecting a portion of sintered ore flowing on the belt conveyor using a robot arm may be used. Furthermore, a method of collecting sintered ore from a sample container in which a sample whose grain size has been selected for strength measurement is placed may also be used. For example, a method of continuously collecting selected samples using an arm may be used, or a method of an operator periodically and continuously collecting samples from the sample container by hand may be used.

[0030] When it is necessary to evaluate the variation in texture when collecting sintered ore, it is preferable to collect multiple samples from the same lot in order to improve the evaluation accuracy of the variation. In addition, in order to further improve the estimation accuracy of the strength of the sintered ore, it is preferable to collect multiple sintered ore samples from the upper and lower layers of the sintered cake of the same lot. Specifically, it is preferable to collect at least one sintered ore sample from each of the upper and lower layers of the sintered cake of the same lot. Here, the upper layer of the sintered cake refers to a layer above the half-height position of the sintered cake, and the lower layer of the sintered cake refers to a layer below the half-height position of the sintered cake.

[0031] [Strength acquisition process] In the strength acquisition step, the strength of the sintered ore is acquired. Although the method for acquiring the strength is not limited, it is preferable to measure the strength of the sintered ore.

[0032] As the strength, it is preferable to use the Shatter Index (SI) defined in JIS M 8711. Alternatively, the Tumble Index (TI) defined in JIS M 8712 may be used. Furthermore, an index correlated with any of the above indexes may be used.

[0033] [Domain fraction calculation process] In the domain fraction calculation step, the structural image of the sintered ore, the strength of which has been acquired in the strength acquisition step, is divided into a plurality of domains, and the area fraction of the domains is calculated for each first structure of the domain. Here, the calculated area fraction of the domain is called the domain fraction. The domain is a region consisting of a continuous first structure and one or both of structures other than the first structure and voids dispersed in the first structure. The strength acquisition step and the domain fraction calculation step may be performed first, simultaneously, or in parallel.

[0034] (Image of sintered ore structure) The structural image of the sintered ore is a structural image of any surface of the sintered ore.

[0035] Here, an example of a method for acquiring a structural image will be described. First, a sintered ore sample is pulverized by the above-mentioned method, and a desired surface of the pulverized sintered ore is exposed as an observation surface by, for example, polishing, a cutter, or a laser, and the observation surface is polished until it becomes a mirror surface. At this time, since the sintered ore is amorphous and easily crumbles, it is preferable to embed the sintered ore in a resin in advance. Next, a structural image is acquired for the observation surface of the polished sintered ore.

[0036] The structure image can be obtained, for example, using a microscope. Examples of the microscope include an optical microscope, a scanning electron microscope, and a stereo microscope. From the viewpoint that the color of the image can be used as a criterion when dividing the domains, it is preferable to obtain the image using an optical microscope or a stereo microscope. In contrast, when a scanning electron microscope is used for the image, the type of image is not limited, and an image such as a secondary electron image or a reflected electron image can be obtained. From the viewpoint of improving the accuracy of dividing the domains, multiple types of images may be obtained. From the same viewpoint, when a scanning electron microscope is used, a map of crystal orientation, etc. may be obtained using EBSD (Electron Backscatter Diffraction), and an element map may be obtained using EDX (Energy Dispersive X-ray Spectroscopy). For convenience, the map of crystal orientation, etc. and the element map are also referred to as a structure image hereinafter.

[0037] In addition to the above-mentioned techniques, the structure image can also be obtained using a non-destructive method such as performing X-ray CT (Computed Tomography) on a desired surface of the sintered ore.

[0038] The upper and lower limits of the observation magnification of the structure image are not particularly limited. However, by increasing the observation magnification, it becomes easier to identify the structure. Therefore, the observation magnification of the structure image is preferably 50 times or more, and more preferably 100 times or more. On the other hand, if the observation magnification is too high, the measurement time for obtaining the structure image becomes long, and it is time-consuming to ensure the representativeness of the sintered ore structure. Therefore, from the viewpoint of measurement efficiency, the observation magnification of the structure image is preferably 200 times or less.

[0039] In addition, the observation magnification affects how finely the domains are divided in the domain division described later. As described later, the domains contain continuous structures. However, by increasing the observation magnification, a region where one continuous structure was originally observed may be observed as a region containing multiple continuous structures. In this case, the region of the continuous structure is subdivided, and the domains are divided finely. Therefore, by appropriately setting the observation magnification, the fineness of the domain division can be controlled. As described above, since the intensity is determined for each region of the mixed phase, it is preferable to divide the domains finely so that the mixed phase and the domain correspond to each other, and this can further improve the accuracy of the intensity estimation. From the above viewpoints, the observation magnification of the tissue image is preferably 200 times or less.

[0040] The upper and lower limits of the size of the observation area of ​​the structure image are not particularly limited, but in order to ensure the representativeness of the sintered ore structure, it is preferable that the observation area of ​​the structure image is the entire observation surface. 2 It is preferable that the thickness is more than 100 mm. 2 The upper limit of the size of the observation area of ​​the tissue image is, for example, 225 mm 2 It can be said that:

[0041] The tissue image can be acquired by photographing while continuously changing the field of view. More specifically, the tissue image is preferably an image acquired by photographing images while continuously changing the field of view of the microscope and connecting the photographed images. In this way, by acquiring one tissue image while changing the field of view, it becomes easy to acquire a tissue image that has both a high observation magnification and a wide observation area.

[0042] (domain) In the present invention, a domain is defined as a region consisting of a continuous first structure and one or both of structures other than the first structure and voids dispersed in the first structure. By dividing the structure image of sintered ore into domains defined in this way, the state of the mixed phase of the sintered ore can be evaluated, and an index that is highly correlated with strength can be obtained. For example, in the conventional technology described in Patent Document 2, since the structure image was not divided into the domains, an index suitable for estimating the strength of sintered ore could not be obtained. Note that the same position in the structure image is not divided to belong to two or more domains, so that the domains do not overlap.

[0043] The structure of the sintered ore may be any structure. Examples of the structure of the sintered ore include primary hematite, secondary hematite, magnetite, calcium ferrite, and silicate slag, but generally, the structure of the sintered ore is mainly composed of the above five types of structures. Therefore, it is preferable that the structure of the sintered ore is composed of the above five types of structures and other structures. In addition, one or more of the above five types of structures may be further classified. In addition, the domain can have any sintered ore structure as the first structure, but it is preferable that the first structure is primary hematite, secondary hematite, magnetite, or calcium ferrite.

[0044] A domain is a region that includes a continuous first structure. As described above, various types of structures are distributed in a complex mixture in sintered ore, and among the structures, a structure that is continuous in a certain portion of a structure image is defined as a first structure. A structure that is continuous in a portion where a mixed phase exists best reflects the melting state (whether or not melting occurred in the sintering process) of the region corresponding to the domain, and therefore best represents the properties of the mixed phase generated by sintering. Furthermore, among the structures included in a domain, the first structure has the highest area fraction in the domain. From this perspective, the first structure is considered to be a structure that represents the properties of the mixed phase.

[0045] Depending on the difference in the first structure, the domains are divided into domains that have undergone melting during the sintering process (also called molten domains) and domains that have not undergone melting (also called unmelted domains).

[0046] It is important that a domain consists of a first structure, and one or both of structures other than the first structure and voids. If a domain is defined as a region that does not include any structure other than the first structure and does not include voids, the phase fraction of each single phase is evaluated, and therefore the state of a mixed phase cannot be evaluated. In addition, the amount of voids, which is thought to affect strength, cannot be reflected in the value of the domain fraction.

[0047] Here, in the domain, the tissues other than the first tissue and the voids are dispersed in the first tissue. In other words, if there is one or both of a tissue other than the continuous tissue and a void contained in the continuous tissue, the tissue and the void are also combined with the continuous tissue to form one domain. In addition, when a certain tissue or void exists in a region sandwiched between two or more continuous tissues, the tissue or void may be dispersed in any of the continuous tissues. The tissue in which the tissue or void is dispersed can be determined using an index such as the distance between the tissue or void and the continuous tissue.

[0048] In the present invention, the domains preferably include a domain having a primary hematite as a first structure, a domain having a secondary hematite as a first structure, a domain having a magnetite as a first structure, and a domain having a calcium ferrite as a first structure. Here, the above four types of domains are defined on the assumption that the structure of sintered ore is composed of primary hematite, secondary hematite, magnetite, calcium ferrite, silicate slag, and other structures.

[0049] (1) Domains with primary hematite as the primary texture During the sintering process, a melt is generated at a position where the parent mineral hematite, calcium oxide, and SiO2 coexist. However, there are regions where the parent mineral hematite remains unreacted as primary hematite. This domain corresponds to the above region and is classified as an unmelted domain. This domain may contain calcium ferrite and silicate slag.

[0050] (2) Domains with secondary hematite as the primary texture During the sintering process, a melt is generated at a position where the parent ore hematite, calcium oxide, and SiO2 coexist, and there is a region where hematite (secondary hematite) crystallizes in a skeletal form under high oxygen partial pressure. This domain corresponds to the above region and is classified as a molten domain. This domain may contain calcium ferrite and silicate slag.

[0051] (3) Domains with Magnetite as their primary organization During the sintering process, the parent ore hematite is reduced to generate magnetite, and there is a region where the magnetite remains without reacting with the surrounding molten liquid. This domain corresponds to the region and is classified as an unmelted domain. This domain may contain calcium ferrite and silicate slag.

[0052] (4) Domains with calcium ferrite as the primary structure During the sintering process, a melt is generated at a position where the parent ore hematite, calcium oxide, and SiO2 coexist, and there is a region where the melt crystallizes as calcium ferrite and silicate slag. This domain corresponds to the region and is classified as a molten domain. This domain may contain silicate slag.

[0053] The above-mentioned domains (1) to (4) may contain a structure other than the first structure in which the sintered ore blend raw material does not undergo a sintering reaction and remains partially, for example, SiO2.

[0054] In addition to the above-mentioned domains (1) to (4), it is also possible to provide a domain having a first structure other than the primary hematite, secondary hematite, magnetite, and calcium ferrite contained in the sintered ore.

[0055] Below, we describe how to partition a tissue image into domains.

[0056] Dividing a tissue image into domains refers to identifying which tissue exists at every point in the tissue image, and based on the result, determining which domain every point in the tissue image belongs to, or whether the point does not belong to any domain. Here, if the tissue image is electronic data, for example, pixels can be used as the points. Specific examples of cases in which the points do not belong to any domain include the following two cases. (A) There is no organization at the location and the location is not a gap within the definition of a domain as described above. (B) When a texture is present at the location, other textures and voids are not dispersed within the texture, and the texture is not dispersed within other textures. When the tissue at the point is continuous over a wide range, other tissues and / or voids are often dispersed in the tissue, and therefore it does not usually fall under (B). On the other hand, when the tissue at the point occupies a small area, it is highly likely that other tissues and voids are not dispersed in the tissue at the point. In other words, when the tissue at the point occupies a small area and exists away from other tissues (isolated), it is highly likely that it falls under (B).

[0057] There is no particular limitation on the method for dividing a tissue image into domains, but for example, when a tissue image is obtained using an optical microscope and the tissue image is divided into the domains (1) to (4), the method can be performed as follows.

[0058] First, the sintered ore structures, hematite, magnetite, and calcium ferrite, are identified based on the brightness and color of the optical microscope images. The hematite region is white, and the magnetite region is reddish.

[0059] Next, the region where magnetite is continuous is combined with one or both of structures other than magnetite and voids dispersed in the magnetite to form (3) a domain having magnetite as the first structure.

[0060] Next, hematite that is observed to retain the shape of the original ore and has a rounded granular texture is identified as primary hematite.Then, a region in which primary hematite is continuous is combined with one or both of the texture other than primary hematite and voids dispersed in the primary hematite to define (1) a domain in which primary hematite is the first texture.

[0061] Next, among the hematite, hematite having an angular shape other than the primary hematite is identified as secondary hematite, and a region in which the secondary hematite is continuous is combined with one or both of structures other than the secondary hematite and voids dispersed in the secondary hematite to define (2) a domain having secondary hematite as a first structure.

[0062] Then, the region not belonging to any of the above (1) to (3) and in which calcium ferrite is continuous is combined with one or both of other structures and voids dispersed in the calcium ferrite to form (4) a domain having calcium ferrite as the first structure.

[0063] In this way, it is possible to divide the domains in accordance with the definition of each domain. Also, for every domain, the first texture of the domain is the texture that occupies the highest area fraction in the domain.

[0064] Here, it is preferable to use machine learning for dividing the domains from the viewpoint of performing the work quickly and easily and further eliminating arbitrariness. At this time, an image obtained by dividing a structural image of sintered ore into a plurality of domains is used as a teacher image. The machine learning includes well-known techniques such as deep learning. Specifically, it is preferable to use U-net, which is one of FCNs (Fully Convolutional Networks) that apply CNN (Convolutional Neural Network). U-net is a network suitable for image segmentation, and has a property of extracting a feature map by convolution of an image, and a property of performing deconvolution processing after retaining the feature map.

[0065] (Domain Fraction) In the domain fraction calculation step, after dividing the tissue image into domains by the above-mentioned method, the domain fraction is calculated for each first tissue of the domain. For example, the ratio of the total area of ​​a domain having a first tissue to the total area of ​​all domains is calculated as the area fraction, and the area fraction can be set as the domain fraction of the domain having the first tissue as the tissue.

[0066] [Model creation process] In the model creation step, a strength estimation model for the sintered ore is created using the domain fraction calculated in the domain fraction calculation step as an explanatory variable and the strength of the sintered ore acquired in the strength acquisition step as a response variable. Hereinafter, the strength estimation model may be simply referred to as a model.

[0067] The method of creating the model is not limited, and for example, the model can be created by machine learning. In one embodiment of the present invention, as the machine learning, a multiple regression analysis is performed to create the model. Specifically, for the sintered ore sample whose strength has been acquired, the strength is used as the objective variable, and the domain fraction of the corresponding sintered ore sample is used as the explanatory variable to perform multiple regression analysis, whereby the relationship between the objective variable and the explanatory variable can be calculated as a regression equation. The regression equation is used as the model equation of the strength estimation model.

[0068] The method for estimating the strength of sintered ore according to the present invention is a method for estimating the strength of sintered ore using a strength estimation model for sintered ore created by the method for creating a strength estimation model for sintered ore. A sintered ore sample with unknown strength is targeted, and a domain fraction obtained from the sintered ore sample is substituted into a model formula as an explanatory variable to calculate the strength, which is a target variable. According to the strength estimation method, even during the operation of a blast furnace, the strength of sintered ore can be accurately estimated by simply calculating the domain fraction for sintered ore with unknown strength and substituting the domain fraction into a strength estimation model for sintered ore created in advance, and the yield of product sintered ore can be predicted.

[0069] The method for producing sintered ore according to the present invention is a method for changing the production conditions of sintered ore based on the strength of sintered ore estimated by the sintered ore strength estimation method. By reviewing the production conditions of sintered ore based on the estimated strength, it is possible to improve the yield of product sintered ore. Here, the production conditions that can be changed include, but are not limited to, the amount and type of gaseous fuel, whether or not oxygen is blown into the sintering furnace, and negative pressure. For example, if it is found that it is better to reduce the domains in which magnetite is the first structure according to the estimated strength of the sintered ore, it is better to blow oxygen into the sintering furnace. Here, it is preferable that the change in the production conditions is a change in the ratio of the components that constitute the sintered ore blended raw material. As described above, examples of the components that constitute the sintered ore blended raw material include fine ore, auxiliary raw material powder, miscellaneous raw materials, and solid fuel. For example, if it is found that it is better to increase the domains in which calcium ferrite is the first structure according to the estimated strength of the sintered ore, it is better to increase the basicity (CaO / SiO2) of the raw material. EXAMPLES

[0070] Hereinafter, a method for creating a strength estimation model for sintered ore and a method for estimating strength of sintered ore according to an embodiment of the present disclosure will be described based on examples.

[0071] Example 1 A model for estimating the strength of sintered ore was created as follows.

[0072] First, three types of sintered cakes were obtained by changing the amount of coke in the sintered ore blend raw material and sintering at different heat values. A laboratory-scale sintering pot was used for sintering. Next, sintered ore samples (No. 1 to 3) were taken from each sintered cake.

[0073] The sintered ore samples were cut in half, embedded in resin, and the surfaces to be evaluated (surfaces to be observed) were mirror-polished. The polishing was performed with #120 to #400 SiC paper, followed by polishing with 9 μm, 3 μm, 1 μm, and 0.25 μm diamond, and colloidal silica polishing.

[0074] The microstructure images of the sintered ore were obtained for the observation surface of the polished sintered ore. The microstructure images of the sintered ore were prepared by taking images of the entire observation surface of the sintered ore at 50x magnification using an optical microscope equipped with an imaging function while continuously changing the field of view, and then linking the obtained optical microscope images.

[0075] The structure images were divided into the domains (1) to (4) by the above-mentioned method, and the domain fraction of each sintered ore sample was calculated. In addition, the drop strength index defined in JIS M 8711 was measured for each sintered ore sample.

[0076] The measured strengths and calculated domain fractions of the sinter samples are shown in Table 1.

[0077] [Table 1]

[0078] A model formula was derived by performing multiple regression analysis using the domain fraction of the sintered ore sample shown in Table 1 as the explanatory variable and the strength as the objective variable, to create a sintered ore strength estimation model. The domain fraction of the sintered ore sample shown in Table 1 was then substituted into the model formula to estimate the strength of the sintered ore (hereinafter, the estimated strength of the sintered ore is referred to as "estimated strength"). The coefficient of determination obtained using the measured strength of the sintered ore (hereinafter, "actual strength") and the estimated strength was 0.94. In this way, a sintered ore strength estimation model could be created by performing multiple regression analysis.

[0079] Example 2 Next, the strength of sintered ore was estimated using the created strength estimation model.

[0080] First, three types of sintered ore samples (A, B, and C) used for estimating strength were prepared as follows. First, the coke content of the sintered ore blend raw material was set to 4.0%, 4.5%, and 5.0%, respectively, and sintered in a laboratory-scale sintering pot to obtain sintered cakes with a diameter of 300 mm and a height of 600 mm. Next, the sintered cakes were cooled and then dropped to be crushed, and the sintered ore samples with a diameter of about 20 mm were collected.

[0081] The sintered ore samples were cut in half, embedded in resin, and the surfaces to be evaluated (surfaces to be observed) were mirror-polished. The polishing was performed with #120 to #400 SiC paper, followed by polishing with 9 μm, 3 μm, 1 μm, and 0.25 μm diamond, and colloidal silica polishing.

[0082] The microstructure images of the sintered ore were obtained for the observation surface of the polished sintered ore. The microstructure images of the sintered ore were prepared by taking images of the entire observation surface of the sintered ore at 50x magnification using an optical microscope equipped with an imaging function while continuously changing the field of view, and then linking the obtained optical microscope images.

[0083] Then, the structure image was divided into the above domains (1) to (4) by the trained network (U-net), and the area fraction of each domain was calculated as the domain fraction. As the teacher image of the trained network, an optical microscope image of sintered ore divided into the above domains (1) to (4) by the above method was used.

[0084] Next, the domain fraction of each sintered ore sample was substituted into the strength estimation model for sintered ore created in Example 1 to obtain the estimated strength of the sintered ore sample. In addition, the drop strength index defined in JIS M 8711 was measured to obtain the actual strength. The estimated strength and the actual strength were compared for each sintered ore sample.

[0085] On the other hand, as a comparative example, the strength was estimated by the following method. First, an XRD-Rietveld analysis was performed on a sintered ore sample other than sintered ore samples A, B, and C, with reference to the method described in JP 2013-122403 A, to determine the phase fractions of hematite, magnetite, calcium ferrite, and di-calcium silicate. In addition, the drop strength index defined in JIS M 8711 was measured, and a regression equation between the phase fraction and the drop strength index was created. Thereafter, the phase fractions of sintered ore samples A, B, and C were determined, and the estimated strength was obtained by substituting the phase fractions into the regression equation.

[0086] The results of the estimated strength and the measured strength are shown in Table 2. Compared to the strength estimated by the method of the comparative example, the strength estimated by the strength estimation model according to the present invention was in extremely good agreement with the measured strength. The coefficient of determination was 0.91, indicating that the regression equation fits well.

[0087] [Table 2]

[0088] As described above, according to the present invention, by preparing a strength estimation model for sintered ore in advance and calculating the domain fraction of the sintered ore, it is possible to accurately estimate the strength of the sintered ore. In other words, the domain fraction is an excellent index for explaining the strength of the sintered ore. Furthermore, by reviewing the manufacturing conditions of the sintered ore based on the estimated strength, it is possible to improve the yield of the product sintered ore.

Claims

1. a strength obtaining step of obtaining the strength of the sintered ore; a domain fraction calculation step of dividing the structural image of the sintered ore into a plurality of domains and calculating an area fraction of the domains for each first structure of the domain; A model creation process for creating a strength estimation model for sintered ore by using the area fraction as an explanatory variable and the strength as a response variable, A method for creating a strength estimation model for sintered ore, wherein the domain is defined as a region consisting of a continuous first structure and one or both of structures other than the first structure and voids dispersed in the first structure.

2. 2. The method for creating a strength estimation model of sintered ore according to claim 1, wherein the domains are a domain having primary hematite as the first structure, a domain having secondary hematite as the first structure, a domain having magnetite as the first structure, and a domain having calcium ferrite as the first structure.

3. A method for estimating the strength of sintered ore, comprising: estimating the strength of sintered ore using a strength estimation model for sintered ore created by the method for creating a strength estimation model for sintered ore according to claim 1 or 2.

4. A method for producing sintered ore, comprising changing production conditions for sintered ore based on the strength of the sintered ore estimated by the method for estimating the strength of sintered ore according to claim 3.

5. The method for producing sintered ore according to claim 4, wherein the change in the production conditions is a change in the ratio of components constituting the sintered ore blending raw material.

Images