Apparatus for estimating inclusion adhesion rate, method for estimating inclusion adhesion rate, method for learning an estimation model, method for determining operating conditions, and method for manufacturing steel products.

Abstract

To provide a defect generation factor estimation apparatus and a defect generation factor estimation method which can estimate defect generation factors based on a relationship between a molten steel solidification process and inclusion adhesion, a defect generation factor estimation model learning method, an operating condition determination method and a steel product production method.SOLUTION: A defect generation factor estimation apparatus comprises: defect generation depth calculation means for calculating the depth position in a slab stage of surface defects in a product; defect generation rate calculation means for calculating a defect generation rate per slab depth; inclusion adhesion rate calculation means for calculating an inclusion adhesion rate per distance from a solidification start position upon slab casting based on the defect generation rate per slab depth; and defect generation factor estimation means for, by use of a pre-learned estimation model with the production conditions of a product as input data and with each inclusion adhesion rate as output data, selecting some of the production conditions and estimating an inclusion adhesion rate upon gradually changing the selected production conditions to estimate defect generation factors.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a defect occurrence factor estimation device, a defect occurrence factor estimation method, a learning method for a defect occurrence factor estimation model, an operation condition determination method, and a method for manufacturing a steel product.

Background Art

[0002] Patent Document 1 discloses a support system that supports an operator in performing an operation for estimating the cause of quality abnormality. In the method proposed in Patent Document 1, attention is paid to the correlation between operation variables and quality in past manufacturing performance data, operation variables having a strong correlation with quality are extracted, and a time-series chart of the operation variables is displayed to the operator.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] In the method proposed in Patent Document 1, in order to identify the cause of quality abnormality, cause estimation is performed by paying attention to the relationship based on the correlation between quality and manufacturing conditions from product quality data and manufacturing condition data. However, simply looking at the correlation between operation factors and quality cannot estimate at which stage during the solidification process of molten steel inclusions that become the cause of defects adhered and became surface defects in the final product. In order to carry out an operation to eliminate surface defects in the final product, it is extremely important to grasp the relationship between the solidification process of molten steel and inclusion adhesion.

[0005] The present invention has been made in view of the above, and aims to provide a defect occurrence cause estimation device, a defect occurrence cause estimation method, a defect occurrence cause estimation model learning method, an operating condition determination method, and a steel product manufacturing method that can estimate defect occurrence cause based on the relationship between the solidification process of molten steel and the adhesion of inclusions. [Means for solving the problem]

[0006] To solve the above-mentioned problems and achieve the objective, the defect occurrence cause estimation apparatus according to the present invention comprises: a defect occurrence depth calculation means for calculating the depth position of surface defects in a product at the slab stage; a defect occurrence rate calculation means for calculating the defect occurrence rate for each slab depth based on the depth position; an inclusion adhesion rate calculation means for calculating the inclusion adhesion rate for each distance from the solidification start position during slab casting based on the defect occurrence rate for each slab depth; and a defect occurrence cause estimation means for estimating the defect occurrence cause by using a pre-learned estimation model that takes the manufacturing conditions of the product as input data and the inclusion adhesion rate as output data, selecting some manufacturing conditions, and estimating the inclusion adhesion rate when the selected manufacturing conditions are changed step by step.

[0007] In the defect occurrence cause estimation device according to the present invention, the estimation model is pre-trained using manufacturing conditions, including the distance from the solidification start position and the operating conditions in continuous casting, as input data, and the inclusion adhesion rate for each distance from the solidification start position as output data.

[0008] The defect occurrence cause estimation device according to the present invention calculates the depth position of the defect occurrence depth by taking into account at least the amount of grinding by the slab grinder in slab finishing, the amount of scale-off in hot rolling, and the amount of plate thickness reduction due to pickling.

[0009] The defect occurrence cause estimation device according to the present invention displays the relationship between one or more manufacturing conditions and the inclusion adhesion rate as one-dimensional or multi-dimensional data.

[0010] In the defect occurrence cause estimation device according to the present invention, the inclusion adhesion rate is the probability that inclusions, which are bubbles or minute solids in the molten steel, adhere during the casting process.

[0011] The defect occurrence cause estimation apparatus according to the present invention, in the above invention, the manufacturing conditions include one or more of the following: tundish injection gas flow rate, casting speed, casting width, casting thickness, slab length, and slag basicity.

[0012] To solve the above-mentioned problems and achieve the objective, the defect occurrence cause estimation method according to the present invention includes the steps of: a computer-based defect occurrence depth calculation means calculating the depth position of a surface defect in the product at the slab stage; a computer-based defect occurrence rate calculation means calculating the defect occurrence rate for each slab depth based on the depth position; a computer-based inclusion adhesion rate calculation means calculating the inclusion adhesion rate for each distance from the solidification start position during slab casting based on the defect occurrence rate for each slab depth; and a computer-based defect occurrence cause estimation means estimating the defect occurrence cause by selecting some manufacturing conditions and estimating the inclusion adhesion rate when the selected manufacturing conditions are changed stepwise, using a pre-learned estimation model that takes the manufacturing conditions of the product as input data and the inclusion adhesion rate as output data.

[0013] To solve the above-mentioned problems and achieve the objective, the learning method for a defect occurrence cause estimation model according to the present invention includes the steps of: a computer-based defect occurrence depth calculation means calculating the depth position of a surface defect in the product at the slab stage; a computer-based defect occurrence rate calculation means calculating the defect occurrence rate for each slab depth based on the depth position; a computer-based inclusion adhesion rate calculation means calculating the inclusion adhesion rate for each distance from the solidification start position during slab casting based on the defect occurrence rate for each slab depth; and a computer-based model learning means learning an estimation model that estimates the inclusion adhesion rate, using the manufacturing conditions of the product as input data and the inclusion adhesion rate as output data.

[0014] To solve the above-mentioned problems and achieve the objective, the method for determining operating conditions according to the present invention determines the operating conditions based on the defect occurrence factors estimated by the defect occurrence factor estimation method described above.

[0015] To solve the above-mentioned problems and achieve the objective, the method for manufacturing steel products according to the present invention manufactures steel products based on the operating conditions determined by the above-described method for determining operating conditions.

[0016] To solve the above-mentioned problems and achieve the objective, the steel product manufacturing method according to the present invention estimates the inclusion adhesion rate for each distance from the solidification start position during slab casting based on the operating conditions after casting using the above-described defect occurrence factor estimation method, and then manufactures the steel product by changing the operating conditions of the subsequent process based on the estimated inclusion adhesion rate. [Effects of the Invention]

[0017] In the defect occurrence factor estimation device and the defect occurrence factor estimation method according to the present invention, paying attention to the reproducibility of the solidification process of molten steel and the adhesion of inclusions, from a large number of past operation performance data, an estimation model is used to estimate at which stage during the solidification process of molten steel inclusions that become the types of defects adhere and appear as surface defects of the product. As a result, since the relationship between the solidification process of molten steel and inclusion adhesion can be grasped, operations for manufacturing products without surface defects can be considered. Further, according to the learning method of the defect occurrence factor estimation model according to the present invention, an estimation model for estimating defect occurrence factors can be constructed based on the relationship between the solidification process of molten steel and inclusion adhesion. Further, according to the operation condition determination method according to the present invention, operations for manufacturing products without surface defects can be carried out. Further, according to the manufacturing method of steel products according to the present invention, steel products without surface defects can be manufactured.

Brief Description of the Drawings

[0018] [Figure 1] FIG. 1 is a diagram showing a schematic configuration of a defect occurrence factor estimation device according to an embodiment of the present invention. [Figure 2] FIG. 2 is a flowchart showing the procedure of a defect occurrence factor estimation method executed by the defect occurrence factor estimation device according to an embodiment of the present invention. [Figure 3] FIG. 3 is a diagram showing an example of a manufacturing process of a stainless steel product (steel product). [Figure 4] FIG. 4 is an example of the present invention, and is a diagram showing an example of manufacturing performance data read from a database in the data reading step of the defect occurrence factor estimation method. [Figure 5] FIG. 5 is an example of the present invention, and is a stacked graph showing the presence or absence of defects for each slab depth calculated in the defect occurrence rate calculation step of the defect occurrence factor estimation method. [Figure 6] FIG. 6 is an example of the present invention, and is a line graph showing the defect occurrence rate for each slab depth calculated in the defect occurrence rate calculation step of the defect occurrence factor estimation method. [Figure 7]FIG. 7 is an example of the present invention and shows an image of solidification of molten steel and flow of molten steel in a mold. [Figure 8] FIG. 8 is an example of the present invention and is a line graph showing the inclusion adhesion rate for each meniscus distance calculated in the inclusion adhesion rate calculation step of the defect occurrence factor estimation method. [Figure 9] FIG. 9 is an example of the present invention and is a histogram showing the model accuracy of the estimation model constructed in the model learning step of the defect occurrence factor estimation method. [Figure 10] FIG. 10 is an example of the present invention and is a contour diagram showing the relationship between the distance from the meniscus, the tundish blown gas flow rate, and the inclusion adhesion rate estimated in the defect occurrence factor estimation step of the defect occurrence factor estimation method. [Figure 11] FIG. 11 is an example of the present invention and is a graph showing the relationship between the distance from the meniscus and the inclusion adhesion rate estimated in the defect occurrence factor estimation step of the defect occurrence factor estimation method. [Figure 12] FIG. 12 is an example of the present invention and is a graph showing the relationship between the slag basicity and the inclusion adhesion rate estimated in the defect occurrence factor estimation step of the defect occurrence factor estimation method. **Embodiments for Carrying Out the Invention**

[0019] The defect occurrence factor estimation device, defect occurrence factor estimation method, learning method of the defect occurrence factor estimation model, operation condition determination method, and steel product manufacturing method according to the embodiments of the present invention will be described with reference to the drawings. Hereinafter, an example in the case where the present invention is applied to a steel process for manufacturing thick plate steel will be described. However, the present invention is not limited to the following embodiments, and the constituent elements in the following embodiments include those that can be replaced by those skilled in the art and are easy, or those that are substantially the same.

[0020] (Defect Occurrence Factor Estimation Device) A defect cause estimation device according to an embodiment will be described with reference to Figure 1. The information processing device 10 for realizing the defect cause estimation device includes an arithmetic processing unit 101, a storage unit (ROM: Read Only Memory) 103, a temporary storage unit (RAM: Random Access Memory) 104, a bus wiring 105, and a database (DB) 120.

[0021] The arithmetic processing unit 101 is implemented using electronic circuits such as a CPU (Central Processing Unit) and executes the defect occurrence cause estimation program 102 in the memory unit 103 to perform various arithmetic processes necessary for defect occurrence cause estimation. The arithmetic processing unit 101 includes a data reading unit 106, a defect occurrence depth calculation unit 107, a defect occurrence rate calculation unit 108, an inclusion adhesion rate calculation unit 109, a model learning unit 110, and a defect occurrence cause estimation unit 111 as functional blocks that function through the execution of the defect occurrence cause estimation program 102.

[0022] The data reading unit 106 executes the data reading step described below. The defect occurrence depth calculation unit 107 executes the defect occurrence depth calculation step described below. The defect occurrence rate calculation unit 108 executes the defect occurrence rate calculation step described below. The inclusion adhesion rate calculation unit 109 executes the inclusion adhesion rate calculation step described below. The model learning unit 110 executes the model learning step described below. The defect occurrence cause estimation unit 111 executes the defect occurrence cause estimation step described below. Details of each step are described below (see Figure 2 below).

[0023] The memory unit 103 stores the defect cause estimation program 102. The database 120 stores past manufacturing performance data. The information processing device 10 may also be connected to a display device 20 that displays the estimated defect cause results and an input device 30 that receives input from the operator, for example, as shown in Figure 1.

[0024] (Method for estimating the causes of defects) The defect occurrence cause estimation method performed by the defect occurrence cause estimation device according to the embodiment will be described with reference to Figure 2. The defect occurrence cause estimation method includes a data reading step, a defect occurrence depth calculation step, a defect occurrence rate calculation step, an inclusion adhesion rate calculation step, a model learning step, and a defect occurrence cause estimation step. In the defect occurrence cause estimation method according to the embodiment, the steps from the data reading step to the model learning step and the defect occurrence cause estimation step may be performed at different times. That is, an estimation model may be constructed in advance by the data reading step to the model learning step, and the defect occurrence cause estimation step using the estimation model may be performed at a different time.

[0025] The following describes each step of the defect occurrence cause estimation method according to the embodiment. The defect occurrence cause estimation method starts, for example, when the manufacturing of the target material is completed and the defect occurrence distribution is measured by the defect occurrence cause estimation device.

[0026] <Data reading step> In the data reading step, the data reading unit 106 reads the continuous casting manufacturing performance data stored in the database 120 (step S1). This manufacturing performance data is data compiled for each product (final product) that includes manufacturing conditions related to the change in slab (semi-finished product) thickness in a large number of coils (steel strips) manufactured in the past, manufacturing conditions related to the occurrence of surface defects, and whether or not defects occurred on the coil surface. The manufacturing performance data is also composed of matrix data, for example, where the rows represent products manufactured in the past and the columns represent manufacturing conditions and inspection results (presence or absence of defects), etc. (see Figure 4 below).

[0027] <Steps for calculating defect depth> In the defect occurrence depth calculation step, the defect occurrence depth calculation unit 107 calculates the depth position (defect occurrence depth position) of the surface defect in the product at the slab stage based on the manufacturing conditions read in the data reading step (step S2). The "defect occurrence depth position" refers to the depth position within the slab where the inclusions that appear as surface defects in the final product are attached.

[0028] In the defect depth calculation step, the depth at which a surface defect occurred in the product at the slab stage is determined by calculating the amount of thickness reduction in the product at each stage based on manufacturing conditions related to the change in slab thickness. Examples of "manufacturing conditions related to the change in slab thickness" include the amount of reduction in hot rolling, heating temperature, acid concentration and pickling rate in pickling, and the number of grinder passes for the slab and coil, which are manufacturing conditions that cause scale removal, grinding, and dissolution of the product surface.

[0029] In the defect depth calculation step, the depth position is calculated by taking into account, for example, the amount of grinding by the slab grinder during slab finishing, the amount of scale-off during hot rolling, and the amount of plate thickness reduction due to pickling (amount of dissolution during pickling). In addition, in the defect depth calculation step, the defect depth at the slab stage is calculated for all products read in the data reading step.

[0030] <Steps for calculating defect rate> In the defect occurrence rate calculation step, the defect occurrence rate calculation unit 108 calculates the defect occurrence rate for each predetermined slab depth based on the depth position at the slab stage (step S3). In the defect occurrence rate calculation step, for each slab for which the defect occurrence depth was calculated in the defect occurrence depth calculation step, the number of products with defects and the total number of products are totaled at a predetermined slab depth (a predetermined depth pitch, for example, 0.5 mm). Then, the defect occurrence rate (defect occurrence ratio) for each predetermined slab depth is calculated by dividing the number of products with defects by the total number of products. Note that in the defect occurrence rate calculation step, the defect occurrence rate for each predetermined slab depth is calculated for all products read in the data reading step. Here, in the above explanation, the predetermined depth pitch in the depth direction was assumed to be constant, but it does not necessarily have to be a constant pitch as long as the distribution of the defect occurrence rate in the slab depth direction is known.

[0031] <Steps for calculating the inclusion adhesion rate> In the inclusion adhesion rate calculation step, the inclusion adhesion rate calculation unit 109 converts the slab depth determined in the defect occurrence rate calculation step into the distance from the solidification start position during slab casting that results in a shell thickness equivalent to the slab depth (hereinafter referred to as "distance from the meniscus") (step S4). Based on this, the inclusion adhesion rate for each distance from the meniscus is calculated based on the defect occurrence rate for each slab depth.

[0032] Here, the inclusion adhesion rate calculated in the inclusion adhesion rate calculation step is the probability that inclusions, which are bubbles or minute solids in the molten steel, adhere during the casting process. In the inclusion adhesion rate calculation step, the inclusion adhesion rate for each distance from the meniscus is calculated for all products read in the data reading step.

[0033] <Model Learning Steps> In the model learning step, the model learning unit 110 takes the product manufacturing conditions as input data and the inclusion adhesion rate as output data to learn a defect occurrence cause estimation model (hereinafter referred to as the "estimation model") that estimates the inclusion adhesion rate (step S5).

[0034] In the model learning step, the estimation model is trained based on the operational data read in the data reading step and the inclusion adhesion rate for each distance from the meniscus calculated in the inclusion adhesion rate calculation step. More specifically, in the model learning step, the estimation model is trained using manufacturing conditions, including the distance from the meniscus and the operating conditions in continuous casting, as input data, and the inclusion adhesion rate for each distance from the meniscus as output data.

[0035] The training of the estimation model in the model training step can be carried out using regression methods such as random forests, linear regression, support vector machines, and neural networks. However, the estimation model may be trained using methods other than the above regression methods, as long as they estimate the target variable from the explanatory variables.

[0036] <Steps for estimating the cause of defects> In the defect occurrence cause estimation step, the defect occurrence cause estimation unit 111 inputs the values ​​of the explanatory variables for the operating conditions to be estimated and the distance from the meniscus to the estimation model constructed in the model learning step. This allows the unit to calculate the inclusion adhesion rate at a distance from the meniscus under those operating conditions (step S6).

[0037] In the defect cause estimation step, a pre-trained estimation model is used, with the product's manufacturing conditions as input data and the inclusion adhesion rate as output data. A selection of manufacturing conditions is then used, and the inclusion adhesion rate is estimated when the selected manufacturing conditions are changed stepwise. This estimates the cause of the defect. The manufacturing conditions input to the estimation model include, for example, one or more of the following: tundish injection gas flow rate, casting speed, casting width, casting thickness, slab length, and slag basicity.

[0038] Furthermore, in the defect occurrence cause estimation step, the relationship between, for example, one or more manufacturing conditions and the inclusion adhesion rate is displayed as one-dimensional or multi-dimensional data (see Figures 10 to 12 below).

[0039] (Examples) Examples of the present invention will be described with reference to Figures 3 to 12. Below, an example of applying the defect occurrence cause estimation method according to the present invention to the manufacture of stainless steel products will be described.

[0040] Figure 3 shows an example of the manufacturing process for stainless steel products. In the manufacturing of stainless steel products, the process involves refining, casting, slab finishing, hot rolling, pickling and annealing 1 (first pickling and annealing), cold rolling, and pickling and annealing 2 (second pickling and annealing), followed by a slitting and inspection process where defects are determined. The product is then shipped in coil form. The following describes each step of the defect occurrence cause estimation method according to the present invention (see Figure 2).

[0041] <Data reading step> First, in the data reading step, manufacturing performance data as shown in Figure 4 is read. The manufacturing performance data in this figure is matrix data where the rows represent coil products manufactured in the past, and the columns represent the manufacturing conditions and inspection results for each process. In addition, in the manufacturing performance data in this figure, the first row contains the coil number, and the first column contains the name of the item. The number of samples (rows) of this manufacturing performance data is 2300 coils.

[0042] <Steps for calculating defect depth> In the defect depth calculation step, the amount of thickness reduction in the slab due to peeling, grinding, and melting in each process is calculated from the coil data in the second row of the manufacturing performance data in Figure 4 using the following equations (1) to (5). Then, the amount of thickness reduced in slab equivalent from the completion of casting to the product is calculated using the following equation (6). This allows for the calculation of the defect depth at the slab stage of the coil.

[0043] (A) Slab grinding amount in slab finishing = ●● [mm] / 1 pass × number of grinder passes ···(1) (B) Surface scale-off amount during heating in hot rolling = ●● [mm] ···(2) (C) Amount of solution dissolved in pickling during pickling annealing 1 = ●● [g / m 2 ]×Specific gravity 7.5×●●[g / m3 ] / Plate width [m]×(Standard pickling speed [mpm] / Pickling speed [mpm]) ···(3) (D) Amount of material removed by the coil grinder in pickling and annealing 1 = ●● [μm] / 1 pass × number of grinder passes ... (4) (E) Amount of solution dissolved in pickling during pickling annealing 2 = ●● [g / m 2 ]×Specific gravity 7.5×●●[g / m 3 ] / Plate width [m] ···(5) (F) Slab thickness reduction from casting completion to product = (A) + (B) + (C) × slab thickness / pickled and annealed plate thickness + (D) × slab thickness / pickled and annealed plate thickness + (E) × slab thickness / product thickness ... (6)

[0044] Furthermore, in the defect depth calculation step, the above formulas (1) to (6) are used to calculate the defect depth at the slab stage for all 2300 coils from the third row onward in the manufacturing performance data in Figure 4.

[0045] <Steps for calculating defect rate> In the defect rate calculation step, the defect rate is calculated at 0.5 mm intervals in slab depth based on the defect depth data at the slab stage of each coil. Figure 5 shows the aggregated results of defect depth as a stacked bar graph, where the number of defect-free slabs (= number of coils; however, if multiple coils are generated from a single slab, it is converted to the number of slabs) and the number of defective slabs are stacked. In this figure, the horizontal axis represents the slab depth (0.5 mm interval), and the vertical axis represents the number of slabs. Figure 6 shows the defect rate calculated based on the aggregated results in Figure 5, with the horizontal axis representing the slab depth (0.5 mm interval) and the vertical axis representing the defect rate.

[0046] <Steps for calculating the inclusion adhesion rate> In the inclusion adhesion rate calculation step, the slab depth shown in Figure 6 is converted to the distance in the withdrawal direction from the meniscus (solidification start point) of the mold of the continuous casting machine using the following formula (7).

[0047] Slab depth (shell thickness) [m] = solidification constant [ms]-1 / 2 ] × (Distance from meniscus [m] / Casting speed [ms] -1 ]^(1 / 2) ···(7)

[0048] The meaning of equation (7) above will be explained with reference to Figure 7. This figure shows an image of the solidification and flow of molten steel in a mold. As shown in the figure, inclusions that cause defects adhere to the shell within the mold, and the shell thickness at the time of inclusion adhesion is precisely when the slab is removed through each process, and is exposed as a defect on the surface at the product stage. Therefore, the distance from the meniscus and the shell thickness in equation (7) above are synonymous with the slab depth. In addition, the "defect occurrence rate at a given slab depth" can be considered as the "inclusion adhesion rate at a given distance from the meniscus".

[0049] Figure 8 shows the result of converting the slab depth on the horizontal axis of Figure 6 to distance from the meniscus, with the horizontal axis representing distance from the meniscus and the vertical axis representing the inclusion adhesion rate. Note that, due to the conversion of slab depth to distance from the meniscus, the defect occurrence rate on the vertical axis of Figure 6 is shown as the inclusion adhesion rate in Figure 8. As shown in Figure 8, it can be seen that inclusions are likely to adhere near the meniscus and at positions 30-40 mm, 70-80 mm, and 90-120 mm from the meniscus, respectively. In other words, the figure suggests that the molten steel flow carries inclusions toward these positions.

[0050] In the model training step, the explanatory variables (input data) were meniscus distance, tundish injection gas flow rate, casting speed, casting width, casting thickness, slab length, and slag basicity, and the objective variable was the inclusion adhesion rate. An estimation model was trained using the machine learning method Random Forest. Figure 9 is a histogram showing the model accuracy of the trained estimation model, with the horizontal axis representing the inclusion adhesion rate estimated by the estimation model and the vertical axis representing the number of validation samples (frequency). In the same figure, the histograms shown with dot hatching represent coils estimated to be defect-free (no inclusion adhesion), and the histograms shown with diagonal hatching represent coils estimated to be defective (inclusion adhesion).

[0051] In Figure 9, the peak of the histogram representing coils estimated to be defect-free and the peak of the histogram representing coils estimated to be defective (shown with diagonal hatching) are clearly separated to the left and right. Therefore, it can be said that the trained estimation model is able to accurately estimate whether or not a coil has defects.

[0052] <Steps for estimating the cause of defects> In the defect occurrence cause estimation step, the values ​​of the explanatory variables—casting speed, casting width, casting thickness, slab length, and slag basicity—were fixed to the average values ​​from actual operations and inputted into the pre-trained estimation model described above. In addition, two explanatory variables, meniscus distance and tundish injection gas flow rate, were input while being varied. In this way, the defect occurrence cause estimation step estimated the inclusion adhesion rate by fixing some input data and gradually changing other input data while inputting them into the estimation model.

[0053] Figure 10 is a contour plot showing the relationship between the distance from the meniscus, the tundish gas flow rate, and the inclusion adhesion rate, as estimated by the method described above. In this figure, the vertical axis represents the distance from the meniscus, the horizontal axis represents the tundish gas flow rate, and the density of the dots represents the inclusion adhesion rate. In this figure, the distance from the meniscus increases as you move upwards. The tundish gas flow rate increases as you move to the right. The inclusion adhesion rate increases as the dots become denser.

[0054] By referring to Figure 10, it can be seen that the inclusion adhesion rate is high near the meniscus, and that the inclusion adhesion rate can be reduced by increasing the tundish injection gas flow rate. Therefore, by referring to the results in the figure, operators can understand the relationship between the solidification process of molten steel and inclusion adhesion, and obtain suggestions for considering the optimal molten steel flow to suppress the amount of inclusion adhesion, i.e., surface defects in the product.

[0055] In Figure 10, the relationship between multiple explanatory variables (distance from the meniscus, tundish injection gas flow rate) and the inclusion adhesion rate was estimated. However, it is also possible to estimate the relationship between a single explanatory variable and the inclusion adhesion rate.

[0056] For example, Figure 11 is a graph showing the relationship between the distance from the meniscus and the inclusion rate, which are among the explanatory variables. In this figure, the horizontal axis represents the distance from the meniscus, and the vertical axis represents the inclusion rate. The solid, dashed, and dotted lines show the inclusion rates when the tundish injection gas flow rate (TD injection gas flow rate) is 4 L / min, 24 L / min, and 58 L / min, respectively. In this way, the defect rate for each distance from the meniscus (= slab depth) can be estimated based on the operating conditions after casting.

[0057] Figure 12 is a graph showing the relationship between slag basicity, one of the explanatory variables, and the inclusion adhesion rate. In this figure, the horizontal axis represents slag basicity, and the vertical axis represents the inclusion adhesion rate. In this way, the defect adhesion rate for each slag basicity can be estimated based on the operating conditions after casting.

[0058] Furthermore, the previous explanation assumed that, based on the estimation results of the defect occurrence cause estimation step, the relationship between the solidification process of molten steel and inclusion adhesion would be understood, operating conditions to eliminate surface defects in the product would be determined, and this would be used in subsequent operations. Alternatively, for example, in the defect occurrence cause estimation step, the inclusion adhesion rate during slab casting may be estimated based on the operating conditions after casting, and the operating conditions of subsequent processes may be changed based on the estimation results. In other words, during product manufacturing, the inclusion adhesion rate may be estimated at an intermediate stage, and the operating conditions of subsequent processes may be determined (changed) according to the estimation results.

[0059] In this case, as shown in Figure 11 above, for example, the defect adhesion rate for each distance from the meniscus (= slab depth) is estimated based on the operating conditions after casting. Next, from the estimated defect adhesion rate, the slab depth amount at which the defect adhesion rate exceeds a threshold set from a quality assurance perspective (for example, the defect adhesion rate is 20% within 2 mm from the slab surface) is read. Then, the amount of slab grinding in the subsequent slab finishing process is determined so that the slab surface is ultimately removed by the amount of slab depth read.

[0060] In other words, the amount to be ground with a slab grinder is determined by subtracting the predicted amount of surface scale-off during hot rolling and the predicted amount of pickling dissolution during pickling and annealing from the amount to be removed from the slab. If the determined amount to be ground with a slab grinder exceeds the predetermined maximum grinder

[0061] Furthermore, as shown in Figure 12 above, for example, the defect adhesion rate for each slag basicity in the refining process is estimated. Here, the slag basicity is constrained from operational considerations other than inclusion generation, such as the cost of slag production. For example, if the slag basicity is constrained to 1.25-2.00%, the value of 1.6% slag basicity, which results in the smallest inclusion adhesion rate, is read from Figure 12. Then, the read slag basicity (1.6%) is determined as the set value for the slag basicity in the refining process.

[0062] In this way, by manufacturing products under predetermined operating conditions, it is possible to estimate at which depth of the slab the probability of defects occurring is highest, and by changing the operating conditions appropriately based on that estimation, it is possible to manufacture products without surface defects.

[0063] The defect generation cause estimation apparatus and defect generation cause estimation method according to the embodiments described above focus on the reproducibility of the solidification process of molten steel and the adhesion of inclusions. Then, an estimation model is used that estimates, from a large amount of past operational data, at what stage in the solidification process of molten steel inclusions that cause defects adhere and appear as surface defects in the product. This makes it possible to understand the relationship between the solidification process of molten steel and the adhesion of inclusions, and to consider operations for producing products without surface defects.

[0064] Furthermore, according to the learning method for the defect occurrence cause estimation model of the embodiment, an estimation model can be constructed that estimates the cause of defects based on the relationship between the solidification process of molten steel and the adhesion of inclusions.

[0065] Furthermore, according to the operating condition determination method of the embodiment, operating conditions can be determined based on the defect occurrence factors estimated by the defect occurrence factor estimation method described above. This makes it possible to carry out operations to manufacture products without surface defects.

[0066] Furthermore, according to the manufacturing method for steel products according to the embodiment, steel products can be manufactured based on the operating conditions determined by the above-described method for determining operating conditions. This makes it possible to manufacture steel products without surface defects.

[0067] Furthermore, according to the steel product manufacturing method of the embodiment, the above-described defect occurrence factor estimation method allows for the estimation of the inclusion adhesion rate for each distance from the solidification start position during slab casting, based on the operating conditions after casting. Based on the estimated inclusion adhesion rate, the operating conditions of subsequent processes can be changed to manufacture the steel product. This makes it possible to manufacture steel products without surface defects.

[0068] Although embodiments applying the invention made by the present inventors have been described above, the present invention is not limited by the descriptions and drawings that constitute part of the disclosure of the present invention in this embodiment. That is, other embodiments, examples, and operational techniques made by those skilled in the art based on this embodiment are all included in the scope of the present invention. [Explanation of symbols]

[0069] 10 Information Processing Devices 101 Arithmetic Processing Unit 102 Defect Cause Estimation Program 103 Memory unit (ROM) 104 Temporary storage (RAM) 105 Bus wiring 106 Data reading unit 107 Defect Occurrence Depth Calculation Unit 108 Defect Occurrence Rate Calculation Unit 109 Inclusion adhesion rate calculation section 110 Model Learning Department 111 Defect Occurrence Factor Estimation Unit 120 Databases (DB) 20 Display device 30 Input devices

Claims

1. A defect depth calculation means for calculating the depth position at the slab stage for the surface of steel products with and without surface defects, A defect rate calculation means calculates the defect rate for each slab depth by dividing the number of products with surface defects at each depth position by the total number of products, which consists of the sum of the number of products with surface defects and the number of products without surface defects. An inclusion adhesion rate calculation means calculates the inclusion adhesion rate for each distance from the solidification start position during slab casting that results in a shell thickness equivalent to the slab depth, based on the defect occurrence rate for each slab depth. A learning means for learning an estimation model that estimates the inclusion adhesion rate, using manufacturing conditions including the distance from the solidification start position and the operating conditions in continuous casting as input data, and the inclusion adhesion rate for each distance from the solidification start position as output data, An estimation means for estimating the inclusion adhesion rate when any part of the manufacturing conditions is selected using the estimation model and the selected manufacturing conditions are changed in stages, Equipped with, The aforementioned manufacturing conditions include one or more of the following: tundish blowing gas flow rate, casting speed, casting width, casting thickness, slab length, and slag basicity. Inclusion adhesion rate estimation device.

2. The inclusion adhesion rate estimation device according to claim 1, wherein the defect occurrence depth calculation means calculates the depth position by taking into account at least the amount of grinding by the slab grinder in slab finishing, the amount of scale-off in hot rolling, and the amount of plate thickness reduction due to pickling.

3. The inclusion adhesion rate estimation apparatus according to claim 1, wherein the estimation means displays the relationship between one or more manufacturing conditions and the inclusion adhesion rate in a graph.

4. The inclusion adhesion rate estimation device according to claim 1, wherein the aforementioned inclusion adhesion rate is the probability that inclusions, which are bubbles or minute solids in the molten steel, adhere during the casting process.

5. The computer's defect depth calculation means includes the steps of calculating the depth position at the slab stage for the surface of steel products with surface defects and steel products without surface defects, The defect rate calculation means provided by the computer calculates the defect rate for each slab depth by dividing the number of products with surface defects at each depth position by the total number of products, which consists of the sum of the number of products with surface defects and the number of products without surface defects. The computer includes a means for calculating the inclusion adhesion rate, which calculates the inclusion adhesion rate for each distance from the solidification start position during slab casting that results in a shell thickness equivalent to the slab depth, based on the defect occurrence rate for each slab depth. The estimation means provided by the computer takes manufacturing conditions, including the distance from the solidification start position and the operating conditions in continuous casting, as input data, and the inclusion adhesion rate for each distance from the solidification start position as output data, and learns an estimation model for estimating the inclusion adhesion rate. The steps include: using the estimation model described above, selecting some arbitrary manufacturing conditions and estimating the inclusion adhesion rate when the selected manufacturing conditions are changed stepwise; Includes, The aforementioned manufacturing conditions include one or more of the following: tundish blowing gas flow rate, casting speed, casting width, casting thickness, slab length, and slag basicity. Method for estimating inclusion rate.

6. The computer's defect depth calculation means includes the steps of calculating the depth position at the slab stage for the surface of steel products with surface defects and steel products without surface defects, The defect rate calculation means provided by the computer calculates the defect rate for each slab depth by dividing the number of products with surface defects at each depth position by the total number of products, which consists of the sum of the number of products with surface defects and the number of products without surface defects. The computer includes a means for calculating the inclusion adhesion rate, which calculates the inclusion adhesion rate for each distance from the solidification start position during slab casting that results in a shell thickness equivalent to the slab depth, based on the defect occurrence rate for each slab depth. The computer's model learning means takes manufacturing conditions, including the distance from the solidification start position and operating conditions in continuous casting, as input data, and the inclusion adhesion rate for each distance from the solidification start position as output data, to learn an estimation model for estimating the inclusion adhesion rate. Includes, The aforementioned manufacturing conditions include one or more of the following: tundish blowing gas flow rate, casting speed, casting width, casting thickness, slab length, and slag basicity. Training methods for estimation models.

7. A method for determining operating conditions, which determines operating conditions based on the inclusion adhesion rate estimated by the inclusion adhesion rate estimation method described in claim 5.

8. A method for manufacturing steel products, comprising manufacturing steel products based on operating conditions determined by the method for determining operating conditions described in claim 7.

9. The method for estimating the inclusion adhesion rate described in claim 5 estimates the inclusion adhesion rate for each distance from the solidification start position during slab casting based on the tundish injection gas flow rate, and the amount of grinding by the slab grinder in the slab finishing process is changed based on the estimated inclusion adhesion rate and the predicted value of the surface scale-off amount in hot rolling or the predicted value of the pickling dissolution amount in pickling annealing, in order to manufacture a steel product. A method for manufacturing steel products.

Images