Surface crack prediction method, method for manufacturing steel strip, method for generating surface crack prediction model, and surface crack prediction device

Abstract

To provide a surface crack prediction method capable of predicting a surface crack of a steel strip after rolling.SOLUTION: In a surface crack prediction method for predicting a surface crack of a steel strip manufactured by performing a rolling process on a slab cast in a continuous casting process, input data including a component content of one or more of Cu, Ni, and Sn of the slab and one or more of rolling parameters of the rolling process is input to a surface crack prediction model, and surface crack data is output to predict the surface crack of the steel strip.SELECTED DRAWING: Figure 3

Description

[Technical Field]

[0001] The present invention relates to a surface crack prediction method for predicting surface cracks in a steel strip after rolling, a steel strip manufacturing method, a surface crack prediction model generation method, and a surface crack prediction device. [Background technology]

[0002] When scrap is used as the iron source for steelmaking, the Cu and Sn contained in the scrap become mixed into the steel. When the surface of the rolled material oxidizes during the furnace heating process in hot rolling, Cu concentrates at the interface between the scale (iron oxide) and the base steel, causing cracks on the surface during the hot rolling process after heating. Cu and Sn are called tramp elements and are difficult to remove through refining, so the amount of scrap used is limited due to the problem of surface cracks. In order to realize a recycling-oriented society, expanding the use of scrap is a major challenge for the steel industry.

[0003] Cu is less susceptible to oxidation than Fe, and its solid solubility in Fe is low (a few percent). For this reason, Cu precipitates as a metallic phase at the interface between the scale and the base steel during the scale formation process. Furthermore, the melting point of Cu is 1080°C, and because the material being rolled in typical hot rolling is heated to temperatures above this, the liquid phase of molten Cu penetrates the grain boundaries of the base steel, causing surface cracks during the width reduction and hot rolling processes, where large shear and tensile stresses are applied.

[0004] It is known that Sn reduces the solid solubility of Cu in steel, thereby promoting surface cracking caused by Cu. Conversely, Ni has the effect of increasing the solid solubility of Cu in steel. Non-Patent Document 1 discloses a technique for adding Ni to suppress surface cracking caused by Cu. However, using a large amount of Ni, which is a rare and expensive metal, increases the manufacturing cost.

[0005] In a hot rolling line, if a surface crack occurs in the width reduction process or hot rolling process on the upstream side, not only will the crack depth increase through the downstream processes, but scale will be pushed into the crack opening, causing surface defects due to poor descaling. For this reason, it is important to prevent surface cracks in all width reduction processes and hot rolling processes from upstream to downstream in a hot rolling line.

[0006] Patent Document 1 discloses a method for producing a hot-rolled steel sheet that does not generate surface defects by determining the boundary between the occurrence of surface defects during continuous casting and hot rolling using a relationship between the Cu content and the Sn content. Patent Document 2 discloses a method for preventing cracking in Cu- and Sn-containing steel by adding Si and setting the heating temperature to 1150°C or higher, thereby capturing the molten Cu and Sn precipitated at the interface between the scale and the base steel into the scale. Patent Document 3 proposes a hot rolling method for preventing surface cracking, in which the surface layer of the rolled material is cooled to a temperature below the crack initiation temperature just before it is bitten into the rolling rolls. [Prior art documents] [Patent documents]

[0007] [Patent Document 1] Patent No. 3180575 [Patent Document 2] Patent No. 3173914 [Patent Document 3] Patent No. 3025363 [Non-patent literature]

[0008] [Non-Patent Document 1] Steel Material Science of Tramp Elements: How to Deal Well with Scrap-Derived Impurity Elements, Iron and Steel Institute of Japan, March 1997 [Non-patent document 2] Theory and Practice of Plate Rolling (Revised Edition), Iron and Steel Institute of Japan, 2010, P165 Summary of the Invention [Problem to be solved by the invention]

[0009] However, in Patent Document 1, when steel containing 0.4 mass % Cu, which is considered to be in a range of components that does not cause surface cracks, is heated to 1100°C for production, the steel is processed in a state in which molten Cu penetrates grain boundaries, and in fact, surface cracks occur. Furthermore, there is no mention of width reduction or horizontal rolling conditions, which poses a problem in that it is not possible to set the reduction conditions in the reduction process so as to prevent surface cracks from occurring.

[0010] Patent Document 2 describes a technology for preventing the penetration of Cu into grain boundaries by adding Si, but it makes no mention of width reduction or horizontal rolling conditions, posing a problem in that it is not possible to set the reduction conditions in the reduction process so as to prevent surface cracks. Furthermore, the addition of Si makes the base steel interface uneven, which raises concerns about the occurrence of surface defects due to poor descaling.

[0011] Patent Document 3 is a technology that focuses on preventing surface cracks during the reduction process on a hot rolling line, but requires a cooling device. In the width reduction process using a sizing press, where slabs are transported intermittently, uneven cooling occurs in the surface temperature of the rolled material during processing, making it difficult to apply this technology to the width reduction process. Furthermore, lowering the temperature of the rolled material reduces the solid solubility of Cu in the steel, which could lead to Cu precipitation at the grain boundaries of the base steel, thereby potentially promoting surface cracks.

[0012] Furthermore, the techniques disclosed in Patent Documents 1 to 3 are techniques for preventing Cu-caused surface cracks during rolling, but do not disclose predicting such surface cracks. If Cu-caused surface cracks during rolling could be predicted, it would be possible to identify rolling conditions under which surface cracks are predicted not to occur, and by rolling under those rolling conditions, it may be possible to produce steel strips while suppressing the occurrence of surface cracks.

[0013] The present invention has been made in consideration of the problems of the prior art, and its object is to provide a surface crack prediction method and a surface crack prediction device that can predict surface cracks in a steel strip after rolling. Another object of the present invention is to provide a steel strip manufacturing method that uses a method for generating a surface crack prediction model used in the surface crack prediction method and the surface crack prediction method to identify rolling conditions in a rolling process that do not cause surface cracks in the steel strip, and to manufacture the steel strip under those rolling conditions. [Means for solving the problem]

[0014] The means for solving the above problems are as follows. [1] A surface crack prediction method for predicting surface cracks in a steel strip produced by performing a rolling process on a steel billet cast in a continuous casting process, the method comprising: inputting input data including the component contents of one or more of Cu, Ni, and Sn in the steel billet and one or more rolling parameters of the rolling process into a surface crack prediction model; outputting surface crack data to predict surface cracks in the steel strip. [2] The surface crack prediction method described in [1], wherein the surface crack prediction model is a trained machine learning model that has been trained using multiple data sets, each set consisting of a pair of actual values ​​of the input data and actual values ​​of the surface crack data, as training data. [3] The surface crack prediction method according to [1] or [2], wherein the steel strip is produced by subjecting the steel billet to a furnace heating process and a rolling process, and the input data further includes one or more heating parameters of the furnace heating process and / or one or more casting parameters of the continuous casting process. [4] A method for manufacturing a steel strip by performing a rolling process on a steel billet cast by continuous casting to produce a steel strip, the method comprising: identifying the rolling parameters that satisfy the target value for surface cracks predicted by the surface crack prediction method described in [1] or [2]; and manufacturing the steel strip under manufacturing conditions that include the identified rolling parameters. [5] A method for manufacturing a steel strip by performing a furnace heating process and a rolling process on a steel billet cast by continuous casting to manufacture the steel strip, the method comprising: identifying the heating parameters and / or rolling parameters that satisfy the target value for surface cracks predicted by the surface crack prediction method described in [3]; and manufacturing the steel strip under manufacturing conditions that include the identified heating parameters and / or rolling parameters. [6] A method for generating a surface crack prediction model that predicts surface cracks in a steel strip produced by performing a rolling process on a steel billet cast in a continuous casting process, the method comprising: training a machine learning model using multiple data sets, each set consisting of actual values ​​of input data including the content of one or more of Cu, Ni, and Sn components of the steel billet and one or more rolling parameters of the rolling process, and actual values ​​of surface crack data, as training data; and generating a surface crack prediction model that uses the input data as input and the surface crack data as output. [7] The method for generating a surface crack prediction model described in [6], wherein the steel strip is produced by performing a furnace heating process and a rolling process on the steel billet, and the input data further includes one or more heating parameters of the furnace heating process and / or one or more casting parameters of the continuous casting process. [8] A surface crack prediction device that predicts surface cracks in a steel strip produced by performing a rolling process on a steel billet cast in a continuous casting process, the surface crack prediction device having a surface crack prediction unit that inputs input data including the component content of one or more of Cu, Ni, and Sn in the steel billet and one or more rolling parameters of the rolling process into a surface crack prediction model, outputs surface crack data, and predicts surface cracks in the steel strip. [9] The surface crack prediction model is a trained machine learning model that has been trained using multiple data sets, each set consisting of the actual value of the input data and the actual value of the surface crack data, as training data.The surface crack prediction device described in [8].

[10] The surface crack prediction device described in [8] or [9], wherein the steel strip is produced by performing a furnace heating process and a rolling process on the steel billet, and the input data further includes one or more heating parameters of the furnace heating process and / or one or more casting parameters of the continuous casting process. [Effects of the Invention]

[0015] By using the surface crack prediction method according to the present invention, it becomes possible to predict surface cracks in a steel strip after rolling. In addition, by identifying rolling parameters that are predicted to cause the surface cracks to satisfy a target value and manufacturing the steel strip under manufacturing conditions that include the identified parameters, it becomes possible to manufacture the steel strip while suppressing surface cracks. [Brief explanation of the drawings]

[0016] [Figure 1] FIG. 1 is a schematic diagram showing the state of a steel billet (slab) being rolled by horizontal rolling. [Figure 2] FIG. 2 is a top view showing the state of a slab being reduced in width reduction. [Figure 3] FIG. 3 is a schematic diagram showing an example of a steel strip manufacturing facility including a surface crack prediction device capable of implementing the surface crack prediction method according to this embodiment. [Figure 4] FIG. 4 is a schematic diagram showing an example of the configuration of a surface crack prediction device. [Figure 5] FIG. 5 is a schematic diagram showing an example of a trained machine learning model, which is a surface crack prediction model using a neural network. [Figure 6] FIG. 6 is a flow diagram showing an example of the surface crack prediction method and steel strip manufacturing method according to this embodiment. DETAILED DESCRIPTION OF THE INVENTION

[0017] FIG. 1 is a schematic diagram showing the state of a slab 22 being rolled by horizontal rolling. FIG. 1(a) is a cross-sectional view showing the state of the slab 22 being rolled by horizontal rolling, and FIG. 1(b) is a top view showing the state of the slab 22 being rolled by horizontal rolling (the upper work roll 20 is not shown). FIG. 2 is a top view showing the state of the slab 22 being reduced in width. The mechanism of surface cracking caused by tramp elements will be explained using FIG. 1 and FIG. 2, taking Cu as an example.

[0018] Cu is less susceptible to oxidation than Fe, and its solid solubility in Fe is low (a few percent). Therefore, during the scale formation process in a heating furnace, Cu precipitates as a metallic phase at the interface between the scale and the base steel. Furthermore, the melting point of Cu is 1080°C, and in hot rolling, the steel billet 22 being rolled is heated to temperatures of 1080°C or higher. For this reason, during hot rolling, the liquid phase of molten Cu penetrates the grain boundaries of the base steel. In this state, large shear stresses and tensile stresses are applied by width reduction and horizontal rolling, causing surface cracks.

[0019] In horizontal rolling of a slab 22 shown in Figure 1, large shear stress 24 occurs at the position where contact with the work roll 20 begins, causing surface cracks 26 to form at that position. Furthermore, at the width edges, the material spreads in the width direction, causing tensile stress 25 in the rolling direction, causing surface cracks 26 to form at those positions as well. In width reduction of a slab 22 using width reduction dies 28 shown in Figure 2, tensile stress 25 occurs at the width edges, causing surface cracks 26 to form at those width edges.

[0020] On the other hand, in hot rolling, as disclosed in Non-Patent Document 2, the yield stress increases as the strain rate in the rolling process increases, and the yield stress decreases as the strain rate decreases. Therefore, even with the same reduction amount, if the strain rate in horizontal rolling and width rolling is slowed, the shear stress 24 and tensile stress 25 decrease, and the occurrence of surface cracks may be suppressed.

[0021] In this embodiment, "strain" is the logarithm of the sheet thickness ratio before and after rolling, and "strain rate" is the value obtained by dividing the strain by the time it takes for the sheet to pass through the contact length with the work roll 20. Strictly speaking, the strain rate becomes large locally at the position where contact with the work roll 20 begins. The maximum strain rate during rolling may be determined using numerical calculations such as finite element analysis, and this value may be used as the strain rate for the rolling process. Furthermore, "strain" during width reduction is the logarithm of the sheet width ratio before and after width reduction, and "strain rate" is the value obtained by dividing the strain by the time it takes for the sheet to contact the width reduction die 28. The strain rate during width reduction may also be determined using numerical calculations such as finite element analysis, as in rolling.

[0022] Regarding the heating temperature in the heating furnace, the melting point of Cu is approximately 1080°C, so if the slab temperature in the heating furnace is 1080°C or below, the concentrated Cu becomes solid and does not infiltrate the grain boundaries of the base steel, so surface cracks do not occur. Also, if the slab temperature in the heating furnace is 1200°C or above, the solubility of Cu in the base steel increases, reducing the amount of concentrated Cu, making surface cracks less likely to occur.

[0023] On the other hand, during the rolling process after the steel slab 22 is extracted from the heating furnace, the surface temperature of the steel slab 22 decreases, the solubility of Cu in the base steel decreases, and the supersaturated Cu concentrates, which may cause surface cracks. Furthermore, if the steel slab 22 remains in the heating furnace for a long time, the amount of Cu concentrates increases, making surface cracks more likely to occur.

[0024] Regarding the component contents of the steel slab 22, if the content of Cu, which acts as a tramp element, increases, the amount of Cu concentrated during the scale formation process increases, and infiltration into grain boundaries becomes more likely. Therefore, if the Cu content of the steel slab 22 increases, surface cracks become more likely to occur.

[0025] Furthermore, Ni increases the solubility of Cu in the base steel, while Sn decreases the solubility of Cu in the base steel. Therefore, the higher the Ni content and the lower the Sn content, the less likely surface cracks will occur.

[0026] Based on this mechanism of occurrence of surface cracks, the surface crack prediction method and surface crack prediction device according to this embodiment predict surface cracks in a steel strip caused by Cu, which is a tramp element. Hereinafter, an example will be described in which a surface crack prediction device 18 capable of implementing the surface crack prediction method according to this embodiment is applied to a steel strip manufacturing facility 100 that constitutes a hot rolling line.

[0027] The surface crack prediction device 18 according to this embodiment can be applied not only to the steel strip manufacturing equipment 100 but also to a thick plate manufacturing equipment for manufacturing thick steel plates and a hot rolling line directly connected to a thin slab continuous casting machine. In other words, the equipment to which the surface crack prediction device 18 according to this embodiment can be applied is not limited to the thickness of the steel strip 23 and the connection layout between the continuous casting machine and the steel strip manufacturing equipment 100.

[0028] 3 is a schematic diagram showing an example of a steel strip manufacturing facility 100 including a surface crack prediction device 18 capable of implementing the surface crack prediction method according to this embodiment. In this embodiment, a steel strip 23 is manufactured through a continuous casting process, a furnace heating process, and a rolling process. Therefore, the steel strip manufacturing facility 100 includes a continuous casting device 10 that performs the continuous casting process, a heating furnace 12 that performs the furnace heating process, a rolling mill 14 that performs the rolling process, a process computer 16, and a surface crack prediction device 18.

[0029] The continuous casting apparatus 10 is an apparatus that cools molten steel in a mold to form a billet, and cuts the billet to a predetermined length to produce a billet 22. The continuous casting apparatus 10 has a gas scarfing device that scarifies and removes impurities on the surface of the billet 22 using fuel gas and oxygen.

[0030] The heating furnace 12 is a device that heats the billet 22 to a target temperature. The billet 22 heated to the target temperature is then width-reduced by a width reduction device to a billet 22 with a predetermined width. The width reduction device includes a sizing press that forges the billet 22 in the width direction using a pair of left and right dies, and an edger that rolls the billet 22 in the width direction using a pair of left and right work rolls.

[0031] The steel slab 22 that has been reduced in width to a predetermined width dimension is rolled into a steel strip 23 of a predetermined thickness by a horizontal rolling mill. The horizontal rolling mill is composed of a roughing mill, an intermediate rolling mill, and a finishing rolling mill. The roughing mill, the intermediate rolling mill, and the finishing rolling mill are devices that roll the steel slab 22 in the thickness direction using a pair of upper and lower work rolls. The rolling mill 14 according to this embodiment includes a width reduction device that reduces the width of the steel slab 22 to a predetermined width dimension, and a horizontal rolling mill that rolls into a steel strip 23 of a predetermined thickness.

[0032] The rolling mill 14 is equipped with an induction heater for heating the ends of the billet 22, a protective cover for preventing temperature drops between rolling stands, and a tunnel furnace. Also, upstream of the finishing mill is a descaling device for injecting descaling water from a header.

[0033] The process computer 16 is, for example, a general-purpose computer such as a workstation or a personal computer. The process computer 16 is connected to each device of the steel strip manufacturing facility 100 by wire or wirelessly and controls the steel strip manufacturing process. The process computer 16 stores manufacturing conditions for operating the continuous casting apparatus 10, the heating furnace 12, and the rolling mill 14. The process computer 16 also collects and stores the elemental contents of the billet 22 manufactured by the continuous casting apparatus 10 and measurement values ​​measured during the operation of each device. The elemental contents of the billet 22 stored in the process computer 16 may be the elemental contents of molten steel measured during the steelmaking process.

[0034] In the steel strip 23 manufactured using such steel strip manufacturing equipment 100, surface cracks caused by Cu as described above may occur after rolling. These surface cracks occur when the liquid phase of molten Cu penetrates into the grain boundaries of the base steel, and in this state, large shear stresses and tensile stresses act on the surface in the rolling device 14.

[0035] Surface cracks on the steel strip 23 are determined over the entire width and length of both the front and back sides of the steel strip 23 after rolling, and surface crack data indicating the determination results is stored in the process computer 16. Surface cracks on the steel strip 23 may be detected, for example, by an image recognition method using a two-dimensional area camera, or may be detected visually by an operator.

[0036] The surface crack data may be binary data indicating the presence or absence of surface cracks in the steel strip 23 (x: surface cracks present, ◯: no surface cracks), or may be ternary data indicating the degree of surface cracks in the steel strip 23 (x: many surface cracks, △: few surface cracks, ◯: no surface cracks). The surface crack data may also be data indicating the number of surface cracks within a predetermined range. The presence or absence of surface cracks in the steel strip 23 is determined to be the presence of surface cracks when the number of surface cracks or the size of the surface cracks exceed a target value, and the target value may be changed depending on the quality required of the steel strip 23.

[0037] The surface crack prediction device 18 predicts surface cracks in a steel strip 23 to be manufactured using a steel strip manufacturing facility 100. The surface crack prediction device 18 acquires input data for the steel strip 23 to be manufactured from the process computer 16, the input data including the component contents of one or more of Cu, Ni, and Sn in the steel billet 22 and one or more rolling parameters of the rolling process. The surface crack prediction device 18 inputs the acquired input data into a surface crack prediction model and predicts surface cracks in the steel strip 23 by outputting surface crack data from the surface crack prediction model.

[0038] Furthermore, it is preferable that the surface crack prediction device 18 identifies rolling parameters that are predicted to cause the surface cracks of the manufactured steel strip 23 to satisfy the target value. The surface crack prediction device 18 outputs the identified rolling parameters to the process computer 16 so that the rolling parameters are set as the manufacturing conditions for the steel strip 23.

[0039] Next, we will explain the surface crack prediction device 18 that predicts surface cracks in the steel strip 23. Figure 4 is a schematic diagram showing an example configuration of the surface crack prediction device 18. The surface crack prediction device 18 is, for example, a general-purpose computer such as a workstation or a personal computer. The surface crack prediction device 18 has a control unit 30, an input unit 32, an output unit 34, and a storage unit 36. The control unit 30 is, for example, a CPU, and functions as an acquisition unit 38, a surface crack prediction unit 40, and a manufacturing condition identification unit 42 by executing a program stored in the storage unit 36.

[0040] The input unit 32 is, for example, a keyboard, a touch panel integrated with a display, etc. The output unit 34 is, for example, an LCD or CRT display, etc. The storage unit 36 ​​is, for example, an information recording medium such as an updatable flash memory, a built-in hard disk or a memory card connected via a data communication terminal, and a read / write device for the same.

[0041] The storage unit 36 ​​stores programs and data for realizing each function of the surface crack prediction device 18. The storage unit 36 ​​further stores a database 44 and a surface crack prediction model 46. The database 44 stores 200 or more, and more preferably 1000 or more, data sets, each set consisting of actual values ​​of input data for the surface crack prediction model and actual values ​​of surface crack data, for steel strips previously manufactured by the same steel strip manufacturing equipment 100.

[0042] The input data in the data set includes the content of one or more of Cu, Ni, and Sn in the steel billet 22 and one or more rolling parameters of the rolling process. The surface crack data in the data set is data indicating the presence or absence of surface cracks, the degree of surface cracks, or the number of surface cracks.

[0043] The content of one or more of Cu, Ni, and Sn in the steel billet 22 may be determined by measuring the concentration of the element in the molten steel used to cast the steel billet 22. The rolling parameters of the rolling process may include the reduction ratio, strain rate, rolling temperature, output value of the induction heater, presence or absence of a heat retention cover, number of headers used in the descaling device, presence or absence of a rolling lubricant spray, and the temperature inside the tunnel furnace between the rolling stands. All of these rolling parameters are set values ​​or calculated values. The data sets stored in the database 44 may be obtained from the process computer 16 or may be input to the storage unit 36 ​​by an operator via the input unit 32.

[0044] The surface crack prediction model 46 is a trained machine learning model that has been trained using a dataset stored in the database 44 as training data. FIG. 5 is a schematic diagram showing an example of a trained machine learning model that is the surface crack prediction model 46 using a neural network. The surface crack prediction model 46 can be generated, for example, using a machine learning model that uses a general neural network as shown in FIG. 5.

[0045] In Figure 5, L1, L2, and L3 represent the input layer, hidden layer, and output layer, respectively. Deep learning using a multi-layered neural network allows for the free selection of input data for operational parameters correlated with surface cracks without considering the problem of multicollinearity, thereby improving the accuracy of predicting surface cracks in steel strips 23. A neural network with two to three hidden layers and 18 to 512 nodes can be used. The number of nodes in the hidden layers can be determined so as to increase the accuracy of predicting surface cracks. However, the machine learning model is not limited to the neural network shown in Figure 5; other models such as random forests, support vector machines, XGBoost, and LightGBM can also be used.

[0046] 4, the processing executed by the acquisition unit 38 and the surface crack prediction unit 40 will be described. The acquisition unit 38 acquires, as input data, the content of one or more of Cu, Ni, and Sn in the steel billet 22 and rolling parameters from the process computer 16.

[0047] The acquisition unit 38 outputs the input data acquired from the process computer 16 to the surface crack prediction unit 40. When the surface crack prediction unit 40 acquires the input data from the acquisition unit 38, it reads the surface crack prediction model 46 from the storage unit 36 ​​and inputs the input data into the surface crack prediction model 46 to output surface crack data and predict surface cracks in the steel strip 23 after rolling. The surface crack prediction unit 40 may output the surface crack data to the output unit 34 and display it on the output unit 34. This allows the operator to visually check the output unit 34 to see whether or not there are surface cracks in the steel strip after rolling, or the number of surface cracks.

[0048] Next, the input data used to predict surface cracks will be described. As described above, the higher the Cu content of the steel slab 22, the more likely surface cracks will occur. Furthermore, Ni and Sn are elements that affect the solubility of Cu in the base steel. Therefore, the Cu, Ni, and Sn component contents affect the presence or absence of surface cracks. Therefore, by including the component contents of one or more of Cu, Ni, and Sn in the steel slab 22 in the input data of the surface crack prediction model 46, it becomes possible to predict surface cracks in the steel strip 23 after rolling with high accuracy.

[0049] Furthermore, when the reduction rate or strain rate in the rolling process changes, the stress applied to the billet 22 during rolling changes, which changes the likelihood of surface cracks occurring. Therefore, the reduction rate and strain rate affect the presence or absence of surface cracks. Furthermore, when the temperature of the billet 22 during rolling increases, the stress applied to the billet 22 during rolling changes, which changes the likelihood of surface cracks occurring. Therefore, the rolling temperature, the output value of the induction heating device, the presence or absence of a heat retention cover, the temperature inside the tunnel furnace, and the number of headers used in the descaling device, which all affect the temperature of the billet 22 during rolling, also affect the presence or absence of surface cracks.

[0050] Furthermore, the use of a rolling lubricant spray reduces the stress generated on the surface of the steel billet 22 during rolling, thereby suppressing surface cracks in the steel strip 23. Therefore, whether or not a rolling lubricant spray is used affects the presence or absence of surface cracks. For these reasons, one or more of the following rolling parameters are included in the input data for the surface crack prediction model 46. This makes it possible to predict with high accuracy the occurrence of surface cracks in the steel strip 23 after rolling.

[0051] <Rolling parameters> Reduction rate Strain Rate Rolling temperature Induction heating device output value Heat retention cover Tunnel kiln temperature Number of headers used by the descaling device Use of rolling lubricant spray

[0052] Note that, since parameters related to the stress generated on the surface of the steel billet 22 during rolling affect the presence or absence of surface cracks, the input data may include rolling parameters other than those mentioned above. This makes it possible to predict surface cracks on the steel strip 23 after rolling with even higher accuracy.

[0053] The input data for the surface crack prediction model 46 preferably further includes one or more heating parameters of the furnace heating process, such as the parameters described below.

[0054] <Heating parameters> Temperature of billet 22 when removed from heating furnace Time in heating furnace of billet 22 Temperature of billet 22 when charged into heating furnace Atmosphere composition inside the heating furnace

[0055] If the temperature at which the slab 22 is removed from the heating furnace during the furnace heating process is high, the degree of Cu penetration into the grain boundaries of the base steel increases, making surface cracks more likely to occur. Therefore, the temperature at which the slab 22 is removed from the heating furnace affects the presence or absence of surface cracks. Furthermore, if the slab 22 remains in the heating furnace for a long time, the amount of surface oxidation increases, increasing the amount of Cu concentration, making surface cracks more likely to occur. Therefore, the length of time the slab 22 remains in the heating furnace also affects the presence or absence of surface cracks.

[0056] Furthermore, if the temperature of the slab 22 when it is charged into the heating furnace is high, the surface oxidation of the slab 22 is likely to proceed, increasing the amount of Cu concentration and making surface cracks more likely to occur. For this reason, the temperature of the slab 22 when it is charged into the heating furnace affects the presence or absence of surface cracks. Furthermore, if the oxygen concentration, water vapor concentration, and CO2 concentration in the heating furnace are high, the surface oxidation of the slab 22 is likely to proceed, increasing the amount of Cu concentration and making surface cracks more likely to occur. For this reason, the atmosphere composition in the heating furnace affects the presence or absence of surface cracks.

[0057] As described above, the heating parameters affect the presence or absence of surface cracks after rolling, so by including one or more of the heating parameters in the input data of the surface crack prediction model 46, it becomes possible to predict the occurrence of surface cracks in the steel strip with even higher accuracy. Note that, since any parameter related to the heating temperature of the steel billet 22 affects the presence or absence of surface cracks, heating parameters other than those mentioned above may also be included. This makes it possible to predict the occurrence of surface cracks in the steel strip 23 after rolling with even higher accuracy.

[0058] The input data for the surface crack prediction model 46 preferably further includes one or more of the casting parameters of the continuous casting process, such as the parameters described below.

[0059] <Casting parameters> Whether or not the surface of the billet 22 is scarified by a gas scarfing device Cutting speed Temperature of steel piece 22 at the start of slicing

[0060] When the surface of the billet 22 is spalled by the gas scarfing device, the surface temperature of the billet 22 increases, making it easier for scale to form. The formation of this scale makes it easier for Cu to infiltrate into grain boundaries, which promotes surface cracking after rolling. For this reason, whether or not the surface of the billet 22 is spalled by the gas scarfing device affects whether or not surface cracks occur.

[0061] Furthermore, if the cutting speed is slow, the surface temperature of the steel billet 22 after cutting becomes high, which makes it easier for scale to form. The formation of this scale makes it easier for Cu to infiltrate into grain boundaries, which promotes surface cracking after rolling. For this reason, the cutting speed affects the presence or absence of surface cracks.

[0062] If the surface temperature of the slab 22 is high at the start of the laser cutting, the surface of the slab 22 is more likely to oxidize after the laser cutting process. The scale formed by this oxidation facilitates Cu infiltration into grain boundaries, which promotes surface cracks after rolling. Therefore, the surface temperature of the slab 22 at the start of the laser cutting affects the presence or absence of surface cracks after rolling. As such, the above-mentioned casting parameters affect the presence or absence of surface cracks after rolling, so by including one or more of the above-mentioned casting parameters in the input data of the surface crack prediction model 46, the occurrence of surface cracks in the steel strip can be predicted with even higher accuracy.

[0063] Next, the processing of the manufacturing condition specifying unit 42 will be described. The manufacturing condition specifying unit 42 specifies rolling parameters that will cause the surface cracks predicted by the surface crack prediction unit 40 to satisfy the target values. The manufacturing condition specifying unit 42 outputs the specified parameters to the process computer 16, thereby causing the specified parameters to be set as the manufacturing conditions of the steel strip manufacturing equipment 100. Below, the processing by the manufacturing condition specifying unit 42 will be described using an example in which the manufacturing condition specifying unit 42 specifies rolling parameters. Note that when the input data of the surface crack prediction model 46 includes heating parameters, the parameters specified by the manufacturing condition specifying unit 42 are not limited to rolling parameters, and heating parameters may be specified, or both heating parameters and rolling parameters may be specified.

[0064] 6 is a flow diagram showing an example of the surface crack prediction method and steel strip manufacturing method according to this embodiment. The flow shown in FIG. 6 is started, for example, by receiving an input from an operator via the input unit 32 to start the processing.

[0065] First, the acquisition unit 38 acquires input data for the steel strip 23 to be manufactured from the process computer 16 (step S101). The input data includes the content of one or more of Cu, Ni, and Sn in the steel billet 22 and one or more rolling parameters of the rolling process. The acquisition unit 38 outputs the acquired input data to the surface crack prediction unit 40.

[0066] When the surface crack prediction unit 40 acquires the input data from the acquisition unit 38, it reads out the surface crack prediction model 46 from the storage unit 36. The surface crack prediction unit 40 inputs the input data into the surface crack prediction model 46 and outputs surface crack data, thereby predicting surface cracks in the steel strip 23 to be manufactured (step S102). The surface crack prediction unit 40 outputs the surface crack prediction results to the manufacturing condition identification unit 42.

[0067] The manufacturing condition specifying unit 42 determines whether the surface cracks predicted for the steel strip 23 to be manufactured satisfy the target value (step S103). The target value for the surface cracks is predetermined and stored in the storage unit 36. If the predicted result of the surface cracks is equal to or less than the target value, the manufacturing condition specifying unit 42 determines that the surface cracks satisfy the target value (step S103: Yes). The manufacturing condition specifying unit 42 identifies the rolling parameters included in the input data as rolling parameters that can manufacture a steel strip 23 that satisfies the target value for the surface cracks (step S104).

[0068] The manufacturing condition specification unit 42 outputs the specified rolling parameters to the process computer 16 (step S105). As a result, the specified rolling parameters are reflected in the manufacturing conditions of the steel strip manufacturing equipment 100. Then, the steel strip 23 is manufactured under the manufacturing conditions in which the specified rolling parameters are reflected (step S106), and the flow shown in FIG. 6 ends. Note that if step S103: Yes is determined the first time through the processing of step S103, the rolling conditions in the rolling process are not changed, so the processing of steps S104 to S106 may be skipped. As a result, the steel strip 23 can be manufactured under manufacturing conditions including rolling parameters predicted to satisfy the target value for surface cracks, and therefore, a steel strip 23 that satisfies the target value for surface cracks can be manufactured.

[0069] On the other hand, if the acquired prediction result of surface cracks exceeds the target value, the manufacturing condition specifying unit 42 determines that the surface cracks do not satisfy the target value (step S103: No). The manufacturing condition specifying unit 42 changes the rolling parameters included in the input data in a direction that suppresses the occurrence of surface cracks (step S107). For example, if the rolling parameter is strain rate, the manufacturing condition specifying unit 42 changes the strain rate in a direction that slows down. Note that the unit by which the strain rate is changed and the range within which the strain rate is changed are predetermined.

[0070] The surface crack prediction unit 40 and the manufacturing condition specification unit 42 execute the processing of steps S102 and S103 again using the input data with the changed rolling parameters. This processing is executed repeatedly until it is determined in the processing of step S103 that the surface cracks satisfy the target values. This enables the manufacturing condition specification unit 42 to specify the rolling parameters that are predicted to cause the surface cracks to satisfy the target values.

[0071] If it is determined that the surface cracks do not satisfy the target values ​​in all ranges in which the rolling parameters can be changed, the manufacturing condition specifying unit 42 may display a message indicating that the rolling parameters cannot be specified or an error message on the output unit 34. After performing such processing, the flow shown in Fig. 6 ends.

[0072] Next, a method for generating the surface crack prediction model 46 will be described. The surface crack prediction model 46 is generated by machine learning a machine learning model using, as training data, a plurality of data sets, each set consisting of a set of actual values ​​of input data and actual values ​​of output data stored in the database 44. The number of data sets used as training data is preferably 200 or more, and more preferably 1000 or more.

[0073] Furthermore, it is preferable that the surface crack prediction model 46 be updated at predetermined intervals. A data set may be acquired from the process computer 16 each time a steel strip 23 is produced in the steel strip manufacturing facility 100 and stored in the database 44. Alternatively, an upper limit on the number of data sets to be stored in the database 44 may be set, and when the number of data sets reaches this upper limit, the oldest data set may be deleted each time a new data set is stored. It is then preferable that the surface crack prediction model 46 be updated at predetermined intervals by machine learning a machine learning model using the data sets stored in the database 44 as training data. This allows the surface crack prediction model 46 to be generated, reflecting the latest facility status, and by using this surface crack prediction model, it becomes possible to predict surface cracks after rolling with even higher accuracy.

[0074] As explained above, by using the surface crack prediction device 18 and surface crack prediction method according to this embodiment, it is possible to predict surface cracks in a steel strip 23 after rolling. Furthermore, rolling parameters are identified that will cause the surface cracks predicted by the surface crack prediction method to satisfy the target value, and the steel strip 23 is manufactured under operating conditions that include the specified parameters. As a result, the steel strip 23 is manufactured under manufacturing conditions that include the rolling parameters predicted to cause the surface cracks to satisfy the target value, making it possible to manufacture a steel strip 23 that satisfies the target value for surface cracks.

[0075] This embodiment is not limited to the above embodiment, and various modifications can be made. In the description of FIG. 6, an example was shown in which rolling parameters were specified so that predicted surface cracks would satisfy the target value, but this is not limiting, and the chemical composition and casting parameters of the billet 22 may also be specified. However, changing the chemical composition and casting parameters of the billet 22 may make it impossible to use a billet 22 that has already been produced. For this reason, it is preferable that the production condition specifying unit 42 specify rolling parameters and / or heating parameters so that predicted surface cracks would satisfy the target value.

[0076] In this embodiment, an example has been shown in which a trained machine learning model is used as the surface crack prediction model, but this is not limited to this. A multiple regression model may also be used as the surface crack prediction model. In this case, the input data of the machine learning model becomes the explanatory variable of the multiple regression model, and the surface crack data becomes the target variable of the multiple regression model. However, as described above, by using a machine learning model, it is possible to freely select operational parameters that are correlated with surface cracks as input data without considering the problem of multicollinearity, so it is preferable to use a trained machine learning model as the surface crack prediction model.

[0077] In the present embodiment, an example has been shown in which the steel strip manufacturing equipment 100 has a process computer 16 and a surface crack prediction device 18, but this is not limited to this. For example, the process computer 16 may have the function of the surface crack prediction device 18, and these may be configured as a single device. Furthermore, the surface crack prediction device 18 may be directly connected to the continuous casting device 10, the heating furnace 12, and the rolling device 14 by wire or wirelessly, and the acquisition unit 38 may acquire the operating conditions directly from these devices.

[0078] In the description of this embodiment, an example was shown in which the control unit 30 of the surface crack prediction device 18 shown in Figure 3 has an acquisition unit 38, a surface crack prediction unit 40, and a manufacturing condition identification unit 42, but this is not limited to this. For example, if input data for the steel strip 23 to be manufactured is already stored in the storage unit 36, the control unit 30 may not have the acquisition unit 38, and the surface crack prediction unit 40 may read the input data from the storage unit 36. Furthermore, if the surface crack prediction device 18 predicts surface cracks in the steel strip 23 to be manufactured, the control unit 30 may not have the manufacturing condition identification unit 42. [Example]

[0079] Example 1 Next, Example 1 will be described, in which surface cracks were confirmed in steel strips 1 to 4 produced by heating a steel billet cast by a continuous casting machine in three heating furnaces and rolling it in a rolling mill having one width reduction device, three roughing mills, and seven finishing mills. In Example 1, a surface crack prediction model of the input data and output data shown in Table 1 below was generated using 1,000 sets of data as training data, and the number of surface cracks in steel strips 1 to 4 to be produced from these was predicted using this surface crack prediction model. The predicted results for the number of surface cracks are shown in Table 1 below. The value shown in the output data column is the average of the predicted and measured values ​​for the number of surface cracks in 5 to 10 coils.

[0080] [Table 1]

[0081] As shown in Table 1, the number of surface cracks predicted using the surface crack prediction model was close to the number of surface cracks that actually occurred on the steel strips for steel strips 1 to 4. This result confirmed that the use of the surface crack prediction model makes it possible to predict surface cracks on steel strips after rolling.

[0082] In addition, a surface crack prediction model was generated using 1,000 sets of data as training data, in which the temperature of the slab at the time of removal from the heating furnace, which is a heating parameter, was added to the input data, and the number of surface cracks in steel strips 1 to 4 shown in Table 1 was predicted using this surface crack prediction model. The input data, output data, and prediction results for the number of surface cracks of the surface crack prediction model are shown in Table 2 below.

[0083] [Table 2]

[0084] As shown in Table 2, the predicted value of the number of surface cracks using the surface crack prediction model, which added the temperature of the slab at the time of removal from the heating furnace, which is a heating parameter, to the input data, was closer to the actual measured value of the number of surface cracks than the predicted value of the number of surface cracks shown in Table 1. From this result, it was confirmed that by adding the heating parameters of the furnace heating process to the input data, it is possible to predict the surface cracks of steel strips after rolling with even higher accuracy.

[0085] Furthermore, a surface crack prediction model with added input data was generated using 1,000 sets of data sets as training data, and the surface crack prediction model was used to predict the number of surface cracks in steel strips 1 to 4 shown in Table 1. The input data, output data, and prediction results for the number of surface cracks of the surface crack prediction model are shown in Table 3 below.

[0086] [Table 3]

[0087] As shown in Table 3, the predicted value of the number of surface cracks using the surface crack prediction model in which the Ni and Sn contents, casting parameters, and rolling parameters were further added to the input data was closer to the actually measured value of the number of surface cracks than the predicted value of the number of surface cracks shown in Table 2. From this result, it was confirmed that by adding the items shown in Table 3 to the input data, it is possible to predict the surface cracks of steel strips after rolling with even higher accuracy.

[0088] <Example 2> Next, Example 2 will be described, in which heating parameters and rolling parameters were identified to make the number of surface cracks equal to or less than a predetermined target value when the number of surface cracks predicted using the surface crack prediction model confirmed in Table 3 is greater than the predetermined target value. The predetermined target value for the number of surface cracks is 0.50 (cracks / 10 m). The results of Example 2 are shown in Table 4 below.

[0089] [Table 4]

[0090] For steel strip 5, the number of surface cracks predicted using the surface crack prediction model was 1.20 (cracks / 10 m), which was more than the target value of 0.50 (cracks / 10 m) and did not meet the target value for the number of surface cracks. Therefore, in order to bring the predicted number of surface cracks below the reference value of 0.50 (cracks / 10 m), the heating furnace extraction temperature was increased from 1150 (°C) to 1200 (°C) and the strain rate was reduced from 1.0 (1 / s) to 0.5 (1 / s). When the number of surface cracks in the steel strip to be produced was predicted using input data including the reset heating parameters and rolling parameters, the predicted number of surface cracks was 0.45 (cracks / 10 m), which was below the target value of 0.50 (cracks / 10 m).

[0091] Since the number of surface cracks predicted using the input data including the reset heating parameters and rolling parameters and the surface crack prediction model satisfied the target value, steel strip was manufactured under manufacturing conditions including the heating parameters and rolling parameters. As a result, the number of surface cracks (actual value) in the manufactured steel strip was 0.48 (pieces / 10 m), and the manufacturing of steel strip that satisfied the target value of 0.50 (pieces / 10 m) for the number of surface cracks was realized.

[0092] For steel strip 6, the number of surface cracks predicted using the surface crack prediction model was 1.00 (cracks / 10 m), which was more than the target value of 0.50 (cracks / 10 m) and did not meet the target value for the number of surface cracks. Therefore, in order to bring the predicted number of surface cracks to the target value of 0.50 (cracks / 10 m) or less, a rolling lubricant spray was used in the roughing mill and the output of the induction heating device in the rolling process was increased from 2000 kW to 3000 kW. When the number of surface cracks in the steel strip to be produced was predicted using input data including the reset rolling parameters, the predicted value was 0.40 (cracks / 10 m), which was less than the target value of 0.50 (cracks / 10 m).

[0093] Since the number of surface cracks predicted using the input data including the reset rolling parameters and the surface crack prediction model satisfied the target value, steel strip was manufactured under manufacturing conditions including the rolling parameters. As a result, the number of surface cracks (actual value) in the manufactured steel strip was 0.35 (pieces / 10 m), and the manufacturing of steel strip that satisfied the target value of 0.50 (pieces / 10 m) for the number of surface cracks was realized.

[0094] From the results of steel strips 5 and 6, the heating parameters and / or rolling parameters that make the number of surface cracks predicted by the surface crack prediction model below the target value are identified, and steel strips are manufactured under manufacturing conditions that include these parameters. This confirmed that it is possible to manufacture steel strips that satisfy the target value for surface cracks. [Explanation of symbols]

[0095] 10 Continuous casting equipment 12 Furnace 14 Rolling mill 16 Process Computer 18 Surface crack prediction device 20 Work Rolls 22 Steel billet 23 Steel strip 24 Shear Stress 25 Tensile stress 26 Surface cracks 28 Width reduction die 30 Control Unit 32 Input section 34 Output section 36 Storage area 38 Acquisition Department 40 Surface crack prediction section 42 Manufacturing condition specification department 44 databases 46 Surface crack prediction model

Claims

1. A surface crack prediction method for predicting surface cracks in a steel strip produced by performing a rolling process on a steel billet cast in a continuous casting process, comprising: A surface crack prediction method comprising inputting input data including the content of one or more of Cu, Ni and Sn components of the steel billet and one or more rolling parameters of the rolling process into a surface crack prediction model, and outputting surface crack data to predict surface cracks of the steel strip.

2. The surface crack prediction method described in claim 1, wherein the surface crack prediction model is a trained machine learning model that has been trained using multiple data sets, each set consisting of an actual value of the input data and an actual value of the surface crack data, as training data.

3. The steel strip is produced by subjecting the steel billet to a furnace heating process and a rolling process, The method for predicting surface cracks according to claim 1 or 2, wherein the input data further includes one or more heating parameters of a furnace heating process and / or one or more casting parameters of the continuous casting process.

4. A method for manufacturing a steel strip by performing a rolling process on a steel billet cast by continuous casting, Identifying the rolling parameters that satisfy the target value of the surface crack predicted by the surface crack prediction method according to claim 1 or 2, A method for manufacturing a steel strip, which manufactures the steel strip under manufacturing conditions including the specified rolling parameters.

5. A method for manufacturing a steel strip by carrying out a furnace heating process and a rolling process on a steel billet cast by continuous casting, Identifying the heating parameters and / or the rolling parameters that satisfy a target value for the surface crack predicted by the surface crack prediction method according to claim 3; A method for manufacturing a steel strip, comprising manufacturing the steel strip under manufacturing conditions including the specified heating parameters and / or the specified rolling parameters.

6. A method for generating a surface crack prediction model for predicting surface cracks in a steel strip produced by performing a rolling process on a steel billet cast in a continuous casting process, comprising: a machine learning model is trained using a plurality of data sets as training data, each set being a combination of actual values ​​of input data including the content of one or more elements of Cu, Ni, and Sn of the steel slab and one or more rolling parameters of the rolling step, and actual values ​​of surface crack data; A method for generating a surface crack prediction model, which generates a surface crack prediction model using the input data as input and the surface crack data as output.

7. The steel strip is produced by subjecting the steel billet to a furnace heating process and a rolling process, The method for generating a surface crack prediction model according to claim 6 , wherein the input data further includes one or more heating parameters of a furnace heating process and / or one or more casting parameters of the continuous casting process.

8. A surface crack prediction device for predicting surface cracks in a steel strip produced by performing a rolling process on a steel billet cast in a continuous casting process, A surface crack prediction device having a surface crack prediction unit that inputs input data including the content of one or more of Cu, Ni, and Sn components of the steel billet and one or more rolling parameters of the rolling process into a surface crack prediction model, outputs surface crack data, and predicts surface cracks of the steel strip.

9. The surface crack prediction model is a trained machine learning model that has been trained using multiple data sets, each set consisting of an actual value of the input data and an actual value of the surface crack data, as training data. The surface crack prediction device described in claim 8.

10. The steel strip is produced by subjecting the steel billet to a furnace heating process and a rolling process, 10. The surface crack prediction device according to claim 8 or claim 9, wherein the input data further includes one or more heating parameters of a furnace heating process and / or one or more casting parameters of the continuous casting process.

Images