JP7803308B2 - H-beam shape prediction method, H-beam manufacturing method, rolling equipment, and shape prediction model generation method - Google Patents

Description

The present invention relates to a shape prediction method for H-beams that predicts the amount of flange warpage of H-beams, a manufacturing method for H-beams, rolling equipment, and a method for generating a shape prediction model.

H-shaped steel beams, which are used as beams and columns in buildings, are generally manufactured by rolling. Figure 1 is a schematic cross-sectional view showing the cross-sectional shape of an H-shaped steel beam 10. The flanges 12 are the relatively thick sections of the H-shaped steel beam 10, while the web 14 is the relatively thin section where the pair of flanges 12 are joined. There are numerous combinations of dimensions for the H-shaped steel beam 10, such as the thickness tf and width B of the flanges 12 and the thickness tw and height H of the web 14, ranging from several dozen for constant inside dimension H-shaped steel (JIS H) to several hundred for constant outside dimension H-shaped steel. Here, the outside dimension is the web height H of the H-shaped steel beam 10, and the inside dimension is the dimension obtained by subtracting the flange thickness from the outside dimension.

Furthermore, the strength classification of the H-shaped steel 10 is broadly divided into three classes based on tensile strength: 400 N ^{/ mm2} class, 490 N/mm2 class, and higher, and the chemical composition of the H-shaped steel 10 is adjusted appropriately for each steel type according to this strength classification. When the H-shaped steel 10 is actually used as a member, the dimensions and strength of the H-shaped steel 10 are selected according to the performance requirements of the building.

Such H-shaped steel 10 is primarily manufactured by hot rolling. H-shaped steel 10 manufactured by rolling is called rolled H-shaped steel. To manufacture rolled H-shaped steel, material is preheated in a heating furnace and then subjected to rough rolling, intermediate rolling, and finish rolling using multiple rolling mills to form the H-shaped steel 10 into the target dimensions.

For example, in rough rolling, the raw steel billet is rolled in multiple passes using a rough rolling mill equipped with a pair of upper and lower rolls each equipped with multiple grooves, roughly shaping it to approach the target dimensions. Next, in intermediate rolling, multiple passes are performed using an intermediate universal rolling mill equipped with a pair of upper and lower horizontal rolls and a pair of left and right hard rolls, and an edger rolling mill equipped with a pair of upper and lower horizontal rolls, bringing it even closer to the target dimensions. Finally, in finish rolling, one pass is performed using a finish universal rolling mill equipped with a pair of upper and lower horizontal rolls and a pair of left and right hard rolls, shaping it into an H-beam product of the target dimensions.

The Japanese Industrial Standard JIS G3192 (Shape, dimensions, mass, and tolerances of hot-rolled steel sections) and corresponding foreign and international standards specify the allowable values for the cross-sectional dimensions of steel sections. For example, for H-section steel 10, the representative values and allowable values for each dimension, such as web height H, flange width B, web thickness tw, and flange thickness tf, are specified, and it is required that all of these dimensions be controlled to fall within the allowable range. Therefore, rolling conditions such as the roll spacing for each rolling mill and each pass are adjusted and modified so that each dimension falls within the allowable range.

Furthermore, because the flange 12 of the H-beam 10 is thicker than the web 14, the flange 12 is more difficult to cool than the web 14 during the cooling process during rolling, resulting in a higher temperature. Residual stress caused by this temperature difference between the flange 12 and the web 14 can cause the web 14 to buckle and deform into a wavy shape after cooling. Therefore, for example, a cooling device may be installed before the finishing mill to cool the outer surface of the flange 12 of the H-beam 10 during the hot rolling process. This eliminates the temperature difference between the flange 12 and the web 14, thereby suppressing buckling deformation of the web 14. Furthermore, when manufacturing higher-strength H-beams, an accelerated cooling device may be installed on the outlet side of the finishing mill to cool the flange 12 in order to control the structure of the steel material.

However, if such cooling of the flange 12 creates a large temperature difference between the outer and inner surfaces of the flange 12, thermal stress caused by this temperature difference or stress caused by phase transformation may cause flange warping, in which the flange 12 bends outward or inward. Figure 2 is a schematic cross-sectional view of an H-beam with flange warping. Figure 2(a) shows an outwardly bowed state in which the flange has bent outward, while Figure 2(b) shows an inwardly bowed state in which the flange has bent inward. If the amount of flange warping as shown in Figure 2 becomes large, additional processes such as pressing may be required for the manufactured H-beam 10, resulting in problems such as increased manufacturing costs and reduced productivity.

As a technology for preventing flange warpage in H-shaped steel 10, Patent Document 1 discloses a flange shape control method for structural steel in which the flange temperature during finish rolling that results in good flange warpage is determined in advance through experiments depending on the dimensions and type of H-shaped steel 10. According to Patent Document 1, the amount of flange warpage can be reduced by adjusting the flange cooling conditions so that the flange temperature is within ±25°C of the predetermined flange temperature.

Patent Document 2 discloses a method for manufacturing H-section steel that uses a temperature calculation model to predict the thermal history under certain cooling conditions, predicts the shape from the predicted thermal history using a shape calculation model, and adjusts the cooling conditions so that the predicted shape is satisfactory. According to Patent Document 2, by manufacturing H-section steel based on cooling conditions discovered by repeatedly performing thermal history calculations and shape calculations, it is possible to prevent shape defects caused by cooling.

Japanese Patent Application Publication No. 7-323320 Japanese Patent Application Publication No. 8-323402

However, with the technology disclosed in Patent Document 1, it is necessary to conduct experiments to determine the flange temperature during finish rolling that will result in good flange warpage for the many different H-section steel dimensions, which requires a great deal of effort. Furthermore, flange warpage is also affected by stresses that arise from phase transformation during cooling. Even if the thermal history and temperature distribution are the same, different steel types will experience different stresses due to phase transformation. Therefore, with the technology disclosed in Patent Document 1, it is necessary to determine the flange temperature during finish rolling that will result in good flange warpage for each steel type, which requires a great deal of effort.

The technology disclosed in Patent Document 2 presents an example of a shape prediction model, which uses an equation expressed as a linear sum using the dimensions of the H-section steel and the flange temperature during finish rolling. However, it is difficult to accurately predict the shape of an H-section steel, which has a two-dimensional cross-sectional shape, using such an equation. Furthermore, flange warpage can change not only due to temperature but also due to the influence of rolling conditions such as the roll spacing of each rolling mill used to adjust dimensions, i.e., the influence of plastic deformation caused by rolling. Therefore, it is difficult to accurately predict flange warpage using a shape prediction model that does not take into account the influence of rolling. Accurately predicting the shape of an H-section steel, taking into account the influence of not only temperature but also rolling, requires two- or three-dimensional structural analysis, which imposes an extremely high computational load for online prediction and is beyond the capabilities of current computers.

The present invention has been made to solve the above-mentioned problems, and its purpose is to provide a shape prediction method for H-section steel that can accurately predict the amount of flange warpage of H-section steel, a manufacturing method for H-section steel that uses this shape prediction method for H-section steel, rolling equipment, and a method for generating a shape prediction model.

The means for solving the above problems are as follows.
[1] A shape prediction method for H-shaped steel that predicts the amount of flange warpage of H-shaped steel manufactured in a rolling facility having a plurality of rolling mills that hot-roll heated steel billets into H-shaped steel, a cooling device that cools the H-shaped steel, and a shape meter that measures the flange warpage of the H-shaped steel after cooling, wherein input data including rolling operation parameters of the rolling mills, cooling operation parameters of the cooling device, and attribute parameters of the H-shaped steel are input into a shape prediction model, and the flange warpage amount is output to predict the flange warpage amount.
[2] A method for manufacturing H-shaped steel, comprising: specifying cooling operation parameters that cause the amount of flange warpage predicted using the shape prediction method for H-shaped steel described in [1] to fall within a predetermined allowable range; and manufacturing the H-shaped steel under manufacturing conditions that include the specified cooling operation parameters.
[3] A rolling facility having a plurality of rolling mills that hot-roll heated steel billets into H-shaped steel, a cooling device that cools the H-shaped steel, a shape meter that measures the amount of flange warpage of the H-shaped steel after cooling, and a shape prediction device for H-shaped steel that predicts the amount of flange warpage, wherein the shape prediction device has a shape prediction unit that inputs input data including rolling operation parameters of the rolling mills, cooling operation parameters of the cooling device, and attribute parameters of the H-shaped steel into a shape prediction model and outputs the amount of flange warpage.
[4] A method for generating a shape prediction model that predicts the amount of flange warpage of H-shaped steel produced by rolling equipment having a plurality of rolling mills that hot-roll heated steel billets into H-shaped steel, a cooling device that cools the H-shaped steel, and a shape meter that measures the flange warpage of the H-shaped steel after cooling, the method comprising: training a machine learning model using as training data a plurality of data sets each consisting of actual values of rolling operation parameters of the rolling mills for H-shaped steel produced in the past, actual values of cooling operation parameters of the cooling device, actual values of attribute parameters of the H-shaped steel, and actual values of the flange warpage amount; and generating a shape prediction model that uses the rolling operation parameters, the cooling operation parameters, and the attribute parameters as inputs and outputs the flange warpage amount.
[5] The method for generating a shape prediction model described in [4], wherein the machine learning model is one of a neural network, decision tree learning, random forest, and support vector regression.

According to the present invention, the amount of flange warpage of H-shaped steel is predicted using a shape prediction model for steel that takes into account the effects of not only cooling but also rolling, allowing for highly accurate prediction of the amount of flange warpage of H-shaped steel. Furthermore, by using the predicted results of flange warpage to control the cooling during H-shaped steel manufacturing, it becomes possible to consistently manufacture H-shaped steel with flange warpage within the allowable range.

FIG. 1 is a cross-sectional schematic diagram showing the cross-sectional shape of an H-beam. FIG. 2 is a cross-sectional schematic diagram of an H-section steel in which flange warping has occurred. FIG. 3 is a schematic diagram showing the general configuration of a rolling facility including an H-beam shape prediction device capable of implementing the H-beam shape prediction method according to this embodiment. FIG. 4 is a schematic diagram showing an outline of the roll arrangement of the finishing rolling mill. FIG. 5 is a functional block diagram of the shape prediction device for steel sections. FIG. 6 is a flowchart showing the flow of the cooling operation parameter specification process performed by the cooling operation parameter specification unit.

One embodiment of the present invention will now be described with reference to the drawings. Note that the embodiment shown below is an example of an apparatus and method that embodies the technical concept of the present invention, and the technical concept of the present invention is not limited to the quality, shape, structure, arrangement, etc. of the components of the embodiment described below.

One embodiment of the present invention will be described with reference to Figures 3 to 6. Figure 3 is a schematic diagram showing the general configuration of a rolling facility 100 including an H-beam shape prediction device that can implement the H-beam shape prediction method according to this embodiment.

As shown in FIG. 3, the rolling equipment 100 includes a heating furnace 20 that heats the steel slabs to a predetermined temperature, a roughing mill 22, an intermediate mill 24, and a finishing mill 28 that roll the steel slabs heated in the heating furnace 20, a rolling control device 30 that controls the operation of these multiple rolling mills, cooling devices 32 and 34 that cool the H-beam 10 during rolling, thermometers 36, 38, and 40 that measure the temperature of the H-beam 10 during rolling at the inlet and outlet sides of the cooling devices 32 and 34, a cooling control device 42 that controls the operation of the cooling devices 32 and 34, a shape meter 44 that measures the amount of flange warp of the H-beam 10 after cooling, a process computer 46, and an H-beam shape prediction device 48.

Slabs of steel are loaded into the heating furnace 20 and heated to a temperature above the austenite temperature range (e.g., 1100-1300°C). The slabs extracted from the heating furnace 20 are transported to the rolling mill by multiple table rolls installed on the outlet side of the heating furnace 20, and are hot rolled into H-beam steel 10 by the roughing mill 22, intermediate mill 24, and finishing mill 28. Furthermore, before and after the rolling mill, the H-beam steel 10 is cooled in cooling devices 32, 34 while being transported by the table rolls.

Rough rolling involves roughly rolling the steel sheet 5 to 25 passes using a rough rolling mill 22, which has a pair of upper and lower grooved rolls each equipped with multiple grooves. Rough rolling is also called breakdown rolling.

In intermediate rolling, one or more intermediate universal rolling mills 25 and one or more edger rolling mills 26 are often used in combination as the intermediate rolling mill 24. In intermediate rolling, the intermediate universal rolling mill 25 and the edger rolling mill 26 are used to perform approximately 5 to 25 passes of reverse rolling until the product reaches approximately the cross-sectional dimensions.

The intermediate universal rolling mill 25 has a total of four rolls: a pair of upper and lower horizontal rolls driven to rotate about a horizontal axis, and a pair of left and right hard rolls that rotate freely about a vertical axis. The intermediate universal rolling mill 25 has the function of simultaneously reducing the web thickness tw and flange thickness tf of the H-section steel 10, rolling the flange 12 so that, for example, the flange 12 is inclined outward by a maximum of approximately 10°. Horizontal rolls with a width corresponding to the height H of the web 14 of the H-section steel 10 are selected and installed in the intermediate universal rolling mill 25. The intermediate universal rolling mill 25 is designed to allow for arbitrary adjustment of the spacing between the upper and lower horizontal rolls, the relative axial positions of the horizontal rolls, and the spacing between the side surfaces of the horizontal rolls and the hard roll. Specifically, the spacing between the upper and lower horizontal rolls can be adjusted between 0 and 200 mm, the relative axial positions of the horizontal rolls can be displaced between -7 and +7 mm, and the spacing between the side surfaces of the horizontal rolls and the hard roll can be adjusted between 0 and 300 mm.

The edger rolling mill 26 has a pair of upper and lower horizontal rolls that are driven to rotate around a horizontal axis. The diameter of the horizontal rolls of the edger rolling mill 26 is, for example, approximately 800 to 1500 mm. Each horizontal roll is provided with a groove, and the tip of the flange 12 is pressed down from above and below to reduce the flange width B.

In finish rolling, the intermediately rolled material is rolled by the finish rolling mill 28, usually in one pass, to finish the cross-sectional dimensions, such as web thickness tw, flange thickness tf, web height H, and flange width B, to the target cross-sectional dimensions of the H-section steel product. A finish universal rolling mill with a configuration similar to that of the intermediate universal rolling mill 25 is used as the finish rolling mill 28. Like the intermediate universal rolling mill, the finish rolling mill 28 is capable of freely adjusting the spacing between the upper and lower horizontal rolls, the relative axial positions of the horizontal rolls, and the spacing between the side of the horizontal roll and the hard roll. Furthermore, in the finish rolling mill 28, the roll angle between the side of the horizontal roll and the vertical roll is set so that the flange after finish rolling is inclined by a maximum of 1°.

Figure 4 is a schematic diagram showing the roll arrangement of the finishing rolling mill 28. As shown in Figure 4, the hard rolls are provided with a roll angle (θ) for tilting the flange 12. By adjusting the position, spacing, and roll angle of each roll, the roll spacing (1) of the web 14 and the roll spacings (2) to (5) of the four flange legs are adjusted. Note that the roll spacing (1) of the web 14 and the roll spacings (2) to (5) of the four flange legs are also the same in the intermediate universal rolling mill 25.

Changing rolling conditions such as the distance between the horizontal rolls and the hard rolls so that each dimension of the H-beam 10 falls within the tolerance range changes the effect of plastic deformation on the cross section of the section steel, which in turn changes the final inclination of the flange 12, i.e., the magnitude of flange warpage. As such, flange warpage is affected not only by temperature but also by rolling conditions, so the influence of rolling conditions must be taken into account in order to accurately predict the amount of flange warpage. In particular, the finishing rolling mill 28 has a particularly strong correlation with flange warpage, as it performs the final rolling to incline the flange 12.

Referring again to Figure 3, the cooling devices 32, 34 cool the H-beam 10 under specified cooling conditions. The cooling devices 32, 34 are separated by a specified length (e.g., 10 m), and each separation is called a bank. The cooling devices 32, 34 shown in Figure 3 each have four banks, but this is not limited to this configuration, and the number of banks may be increased or decreased from four. The cooling operation parameters that define the cooling conditions for the cooling devices 32, 34 include the water flow density of the cooling water and the cooling time. The cooling time can be calculated by multiplying the length per bank by the number of banks and dividing the resulting cooling length by the conveying speed.

Increasing the cooling water flow density can increase the cooling rate and temperature drop of the H-shaped steel 10. Also, extending the cooling time can increase the temperature drop of the H-shaped steel 10. By adjusting the water flow density and cooling time, which are cooling operation parameters of the cooling devices 32 and 34, the surface temperature of the flange 12 after cooling can be controlled.

Furthermore, the water flow density and cooling time may be changed independently for each cooling device installed before and after finish rolling, such as cooling devices 32 and 34. This is because flange warpage is correlated with the temperature of the flange 12 after finish rolling, and the mechanical properties of the steel material are correlated with the cooling stop temperature of the flange 12.

The cooling devices 32, 34 can be general flat spray nozzles, square spray nozzles, or multi-hole jet nozzles. If it is difficult to cool the entire flange with a single flat spray or square spray nozzle, multiple nozzles can be lined up across the width of the flange 12 for cooling.

The thermometers 36, 38, and 40 are installed on the inlet and outlet sides of the cooling devices 32 and 34 and measure the surface temperature of the flange 12 of the H-shaped steel 10 before and after cooling by the cooling devices 32 and 34. The thermometers 36, 38, and 40 are scanning thermometers that scan temperature measurement points across the width of the flange 12, or devices that measure the temperature of the flange 12 of the H-shaped steel 10 by placing multiple spot thermometers across the width of the flange 12. If the latter method is used, it is preferable to measure the temperature by placing spot thermometers at at least five different positions across the width of the flange 12 in order to grasp the temperature distribution across the width of the flange 12.

In particular, flange warpage correlates with the flange temperature after finish rolling. According to the inventors' research, flange warpage correlates particularly with the temperature at the 1/2F position of the flange 12 (the center position in the flange width direction), so it is preferable to measure the temperature at least at the 1/2F position of the flange.

The temperature information measured by the thermometers 36, 38, and 40 is output to the process computer 46. The temperature information output from the thermometers 36, 38, and 40 may be measured values at representative points on the flange 12 (for example, five points equally spaced from the 1/2F position to the upper and lower ends of the flange in the width direction). Alternatively, the data obtained from the thermometers may be approximated by a polynomial of degree two or higher, and parameters that can identify the function of the approximated polynomial may be output as temperature information. This temperature information is used to confirm that the temperature of the H-beam 10 is within a specified range as a result of controlling the cooling devices 32 and 34.

The shape meter 44 measures the amount of flange warpage of the H-beam 10 after it has been cooled by the cooling device 34. A one-dimensional laser displacement meter capable of measuring the distance between one point can be used as the shape meter 44, for example. The amount of warpage of the flange 12 is measured by scanning the flange 12 in the width direction with the laser displacement meter. When scanning the flange 12 in the width direction with the laser displacement meter in this way, it is preferable to set the measurement pitch to 20 to 50 mm to improve measurement accuracy.

Furthermore, multiple one-dimensional laser displacement meters may be installed in the flange width direction, and flange warpage may be measured using these multiple laser displacement meters. In this case, the shape of the flange 12 at multiple different positions in the width direction can be measured without scanning the laser displacement meters as described above. In this case as well, it is preferable to install the laser displacement meters at intervals of 20 to 50 mm in the flange width direction to improve measurement accuracy.

The shape data measured by the shape meter 44 is output to the process computer 46. The process computer 46 records the distance between the fillet portion and the flange widthwise end for the shape data acquired from the shape meter 44 as the actual value of the flange warpage. The process computer 46 may also record parameters that can identify a polynomial function that approximates the warpage shape of the flange 12 as the actual value of the flange warpage.

The shape meter 44 does not necessarily have to be positioned on an extension of the conveying direction of the H-beam 10 as shown in Figure 3. For example, it may be installed so that it can be measured on a cooling bed or the like that can measure the shape of the H-beam 10 after it has passed through the cooling device 34. For example, if the flange warpage is measured when the average temperature within the flange cross section falls to 500°C or below, the flange warpage shape due to thermal strain will stabilize, allowing the flange warpage to be measured with high accuracy.

The process computer 46 can be a general-purpose computer such as a workstation or personal computer. The process computer 46 is connected to the rolling control device 30, cooling control device 42, thermometers 36-40, and shapemeter 44 via wired or wireless connections, and controls the manufacturing process of the structural steel. The process computer 46 also acquires attribute parameters of the H-beam 10 from a host computer. The attribute parameters include the target dimensional values of the H-beam 10 (web height H, flange width B, web thickness tw, flange thickness tf), steel type classification (40k steel, 50k steel, and three types above), information on the chemical composition (contents of C, Si, Mn, Cr, Mo, V, etc.), and information on the target mechanical properties of the H-beam 10 (yield stress, tensile strength, elongation, toughness, hardness, etc.).

In addition to attribute parameters, the process computer 46 stores information regarding the heating temperature of the H-beam 10 and the cooling stop temperature at the outlet of the cooling devices 32, 34 required to obtain the desired material properties. The process computer 46 performs heat transfer calculations based on an internal model and identifies the cooling operation parameters of the cooling device, namely the water flow density and cooling time, required to achieve the cooling stop temperature.

The cooling operation parameters determined in this manner are sent to the cooling control device 42. Based on the cooling operation parameters, the cooling control device 42 controls the operating pressure and number of operating cooling water pumps in the cooling devices 32 and 34, the opening degree of the cooling nozzle flow valves, and the rotational speed of the motor that drives the table roll.

The process computer 46 also sets rolling operation parameters, such as the number of passes and roll spacing, for each rolling mill according to the attribute information of the H-beam 10. Typical rolling operation parameter settings are set as table values associated with the attribute information of the H-beam 10 based on past rolling performance. However, if the rolls are rearranged, rolling the preceding and succeeding rolled materials using the same parameters may not produce the same rolling results. For example, even if the same setting values are used before and after the roll rearrangement, the reference state (e.g., the zero point setting state) that serves as the basis for the setting values actually changes, so the same cross-sectional dimensions may not be obtained if the rolling conditions are used as is. Furthermore, even if the rolls are not rearranged, changes in the roll spacing for H-beams 10 with different cross-sectional dimensions can change the mill stretch behavior of the rolling mill and the friction characteristics between the housing and the roll chock, resulting in nonlinear characteristics in the relationship between the roll spacing and the rolling load. In such cases, even if the roll opening setting is the same, the cross-sectional dimensions of the H-beam may vary. Therefore, in such cases, the operator can reset and change the table values each time.

The process computer 46 sets the gap between the horizontal rolls and the relative axial positions of the horizontal rolls for each pass in the roughing mill 22 as rolling operation parameters for the rolling process. The process computer 46 also sets the web roll spacing and flange four-leg roll spacing in the intermediate universal mill 25, and the gap between the horizontal rolls, horizontal inclination of the horizontal rolls, and the relative axial positions of the horizontal rolls in the edger mill 26 as rolling operation parameters for the rolling process. The process computer 46 also sets the web roll spacing and flange four-leg roll spacing in the finishing universal mill as rolling operation parameters for the rolling process.

The process computer 46 outputs the set rolling operation parameters to the rolling control device 30. The rolling control device 30 controls the roll spacing, roll positions, etc. of the roughing mill 22, intermediate mill 24, and finishing mill 28 based on the rolling operation parameters.

The process computer 46 also collects and stores temperature information and shape data of the H-beam 10 from the thermometers 36-40 and shape meter 44. The actual values of these rolling operation parameters, cooling operation parameters, temperature information, flange warpage data, and attribute parameters are associated with the identification number of the H-beam 10 being manufactured and stored in the database of the process computer 46.

The H-beam shape prediction device 48 acquires from the process computer 46 the rolling operation parameters, cooling operation parameters, and attribute parameters of the H-beam 10 manufactured by the rolling equipment 100, inputs the input data including these into a shape prediction model, and predicts the amount of flange warp of the H-beam 10 by outputting the amount of flange warp of the H-beam 10. The H-beam shape prediction device 48 also identifies cooling operation parameters that will bring the predicted amount of flange warp of the H-beam 10 within the allowable range, and outputs the identified cooling operation parameters to the process computer 46 to set them as cooling operation parameters for the manufacture of the H-beam 10.

Next, we will explain the H-beam shape prediction device 48 that predicts the amount of flange warp of the H-beam 10. Figure 5 is a functional block diagram of the H-beam shape prediction device 48. The H-beam shape prediction device 48 can be a general-purpose computer such as a workstation or a personal computer. The H-beam shape prediction device 48 has a control unit 50, an input unit 52, an output unit 54, and a storage unit 56. The control unit 50 is, for example, a CPU, and functions as a data acquisition unit 58, a shape prediction unit 60, a cooling operation parameter identification unit 62, and a shape prediction model generation unit 64 by executing programs stored in the storage unit 56.

The input unit 52 is, for example, a keyboard, a touch panel integrated with a display, or the like. The output unit 54 is, for example, an LCD or CRT display. The storage unit 56 is, for example, an updatable flash memory, a built-in hard disk or a memory card connected via a data communication terminal, or an information recording medium and its read/write device. The storage unit 56 stores programs and data for realizing each function of the H-beam shape prediction device 48. The storage unit 56 also stores a database 66 and a shape prediction model 68. The database 66 stores 500 or more, and preferably 2000 or more, data sets each consisting of rolling operation parameters, cooling operation parameters, attribute parameters, and flange warpage amounts for H-beams 10 previously manufactured using the same rolling equipment 100 (with the same roll sets for the roughing mill 22, intermediate mill 24, and finishing mill 28).

The shape prediction model 68 is a trained machine learning model that has been trained using the dataset stored in the database 66 as training data. The shape prediction model 68 of this embodiment is a trained machine learning model that receives input data including rolling operation parameters, cooling operation parameters, and attribute parameters, and outputs the amount of flange warpage.

Next, we will explain the processing performed by the data acquisition unit 58 and shape prediction unit 60. The data acquisition unit 58 acquires rolling operation parameters, cooling operation parameters, and attribute parameters as input data from the process computer 46.

The data acquisition unit 58 acquires the set values for the web roll spacing and the four flange roll spacing of the finishing mill 28 from the process computer 46 as rolling operation parameters. The finishing mill 28 is the final rolling mill that finishes the product to the target dimensions, and the web roll spacing and the four flange roll spacing of this rolling mill directly affect the cross-sectional shape of the product. Furthermore, since this rolling mill is the last of the rolling mills to tilt the flange 12, it has a particularly strong correlation with flange warpage. Therefore, it is preferable to use the web roll spacing and the four flange roll spacing of the finishing mill 28 as rolling operation parameters, which improves the prediction accuracy of the flange warpage of the H-beam 10.

The data acquisition unit 58 may also acquire from the process computer 46 the gap between the horizontal rolls and the relative axial position of the horizontal rolls in each pass of the roughing mill 22 as rolling operation parameters. The gap between the horizontal rolls in each pass of the roughing mill 22 directly affects the cross-sectional dimensions of the rolled material at the end of rough rolling. Furthermore, changing the relative axial positions of the horizontal rolls causes horizontal shifts in the vertical direction of the cross section of the rolled material, changing the cross-sectional dimensions of the rolled material. Thus, since the gap between the horizontal rolls and the relative axial positions of the horizontal rolls in each pass of the roughing mill 22 affect flange warpage, the accuracy of predicting the amount of flange warpage of the H-beam 10 is further improved by including the gap between the horizontal rolls and the relative axial positions of the horizontal rolls in each pass of the roughing mill 22 as rolling operation parameters.

The data acquisition unit 58 may also acquire from the process computer 46 rolling operation parameters such as the set values for the web roll spacing and the flange four-leg roll spacing for each pass of the intermediate universal rolling mill 25, the opening between the horizontal rolls for each pass of the edger rolling mill 26, the horizontal inclination of the horizontal rolls, and the relative axial position of the horizontal rolls.

In the intermediate universal rolling mill 25, the set values for the web roll spacing and the roll spacing of the four flange legs in each pass affect the variation in flange thickness at four locations: top, bottom, left, and right. The gap between the upper and lower horizontal rolls in each pass of the edger rolling mill 26 affects the flange width. Furthermore, the horizontal tilt of the horizontal rolls of the edger rolling mill 26 (the difference in the height displacement of the roll axis centers on the working and driving sides of the rolling mill), also known as the leveling amount, creates a difference in the roll gap between the left and right at the positions where the left and right flanges are pressed down, thereby affecting the variation in left and right flange width. Furthermore, the relative axial position of the horizontal rolls can cause the left and right flange width dimensions to become uneven. As such, the setting values for each roll of the intermediate universal rolling mill 25 and the edger rolling mill 26 also affect flange warpage, so by including the average value of the roll gap dimension between the four flange legs in each pass of the intermediate universal rolling mill 25, the gap between the horizontal rolls in each pass of the edger rolling mill 26, the horizontal inclination of the horizontal rolls, and the relative axial position of the horizontal rolls in the rolling operation parameters, the prediction accuracy of the amount of flange warpage of the H-beam 10 is improved.

The data acquisition unit 58 acquires the water flow rate and cooling time of the cooling devices 32, 34 from the process computer 46 as cooling operation parameters. In the cooling devices 32, 34, the higher the water flow rate and the longer the cooling time, the greater the temperature difference between the inner and outer surfaces of the flange, resulting in greater flange warpage. Therefore, by using the water flow rate and cooling time of the cooling devices 32, 34 as cooling operation parameters, the accuracy of predicting the amount of flange warpage of the H-shaped steel 10 is improved.

Furthermore, the data acquisition unit 58 acquires the target dimension values and steel type classification of the H-shaped steel 10 from the process computer 46 as attribute parameters. By including the target dimension values and steel type classification of the H-shaped steel 10 in the input data for the shape prediction model, the shape prediction model takes into account the target dimension values and steel type classification. Therefore, the same shape prediction model can be used to predict flange warpage for H-shaped steel 10 with different target dimension values and steel type classifications. The data acquisition unit 58 may also acquire, as attribute parameters, information on chemical composition (e.g., C, Si, Mn, Cr, Mo, V content) and target values of mechanical properties of the H-shaped steel 10 (e.g., yield stress, tensile strength, elongation, toughness, hardness). By including information on chemical composition and target values of mechanical properties, the shape prediction model takes into account the target values of chemical composition and mechanical properties. Therefore, the same shape prediction model can be used to predict flange warpage for H-shaped steel 10 with different target values of chemical composition and mechanical properties. The data acquisition unit 58 outputs the acquired input data to the shape prediction unit 60.

When the shape prediction unit 60 acquires input data from the data acquisition unit 58, it reads the shape prediction model 68 from the storage unit 56, inputs the input data into the shape prediction model 68, and causes it to output the flange warp amount. In this way, by outputting the flange warp amount from the shape prediction model 68, the shape prediction unit 60 predicts the flange warp amount of the H-beam 10 to be manufactured. The shape prediction unit 60 may also output the output flange warp amount to the output unit 54 and display the flange warp amount on the output unit 54. This allows the operator to visually check the output unit 54 to confirm the predicted value of the flange warp amount.

Next, the processing of the cooling operation parameter identification unit 62 and the shape prediction model generation unit 64 will be described. The cooling operation parameter identification unit 62 identifies cooling operation parameters that will bring the flange warpage amount predicted by the shape prediction unit 60 within the allowable range, and sets the cooling operation parameters as cooling conditions by outputting them to the process computer 46.

Figure 6 is a flow diagram showing the flow of the cooling operation parameter identification process performed by the cooling operation parameter identification unit 62. The flow shown in Figure 6 is started, for example, by receiving input from an operator to start the process.

First, the data acquisition unit 58 acquires rolling operation parameters for the H-section steel to be manufactured from the process computer 46 (step S101). The data acquisition unit 58 also acquires attribute parameters for the H-section steel 10 to be manufactured from the process computer 46 (step S102). The data acquisition unit 58 outputs the acquired rolling operation parameters and attribute parameters to the shape prediction unit 60.

The cooling operation parameter specification unit 62 sets arbitrary cooling operation parameters (step S103). The cooling operation parameter specification unit 62 outputs the set cooling operation parameters to the shape prediction unit 60. The shape prediction unit 60 reads the shape prediction model 68 from the storage unit 56 and inputs the acquired rolling operation parameters, cooling operation parameters, and attribute parameters into the shape prediction model 68 to output the flange warp amount of the H-shaped steel 10 and predict the flange warp amount of the H-shaped steel 10 to be manufactured (step S104). The shape prediction unit 60 outputs the output predicted value of the flange warp amount of the H-shaped steel 10 to the cooling operation parameter specification unit 62.

The allowable range of flange warpage may be predetermined and stored in the storage unit 56, or may be input by the operator via the input unit 52. After acquiring the predicted flange warpage value and the allowable range of flange warpage, the cooling operation parameter identification unit 62 determines whether the predicted flange warpage value is within the allowable range of flange warpage (step S105). If the cooling operation parameter identification unit 62 determines that the predicted flange warpage value is outside the allowable range (step S105: No), it changes the cooling operation parameters in accordance with predetermined conditions (step S106), returns the process to step S104, and repeatedly executes steps S104 to S106 until the predicted flange warpage value falls within the allowable range.

On the other hand, if it is determined in step S105 that the predicted value of the flange warpage is within the allowable range (step S105: Yes), the cooling operation parameter identification unit 62 identifies the cooling operation parameters used to predict the flange warpage as cooling operation parameters that can bring the flange warpage within the allowable range (step S107), and the cooling operation parameter determination process flow shown in Figure 6 ends. The cooling operation parameter identification unit 62 outputs the identified cooling operation parameters to the process computer 46, and sets the identified cooling operation parameters as the cooling operation parameters of the cooling devices 32, 34.

In this way, by specifying the cooling operation parameters, it is possible to identify cooling operation parameters that can control the amount of flange warpage within the allowable range. Then, by manufacturing the H-shaped steel 10 under manufacturing conditions that include these cooling operation parameters, it becomes possible to consistently manufacture H-shaped steel 10 with an amount of flange warpage within the allowable range, thereby suppressing increases in manufacturing costs and decreases in productivity for the H-shaped steel 10.

Next, we will explain how to generate a shape prediction model used to predict the amount of flange warpage. The data acquisition unit 58 acquires from the process computer 46 the actual values of the rolling operation parameters, the actual values of the cooling operation parameters, the actual values of the attribute parameters, and the actual values of the amount of flange warpage for previously manufactured H-section steel 10, and stores these as a set of data in the database 66 of the storage unit 56. It is preferable that the number of data sets stored in the database 66 be at least 500 or more, and more preferably 2000 or more.

The shape prediction model generation unit 64 reads a pre-stored machine learning model from the storage unit 56, and trains the machine learning model using the dataset stored in the database 66 as training data to generate a trained machine learning model. This trained machine learning model becomes the shape prediction model. Note that the machine learning model used in the H-beam shape prediction method and H-beam shape prediction device 48 according to this embodiment may be any of the commonly used neural networks, decision tree learning, random forests, and support vector regression. Furthermore, instead of the flange warp amount, the output of the shape prediction model may be a machine learning model that outputs binary data, either pass or fail, indicating whether the flange warp amount is within the allowable range. In this case, a classification model such as k-nearest neighbor or logistic regression may be used.

The shape prediction model may be updated to a new one, for example, every month or year, by conducting machine learning again. The data acquisition unit 58 acquires performance data each time an H-shaped steel 10 is manufactured and stores it in the database 66. As performance data for newly manufactured H-shaped steel 10 is stored in the database 66, the amount of performance data stored increases. The greater the amount of performance data, the more accurate the shape prediction becomes. Therefore, by periodically conducting machine learning to update the shape prediction model, it becomes possible to predict the amount of flange warpage of the H-shaped steel 10 with even greater accuracy.

As explained above, the shape prediction method for H-shaped steel and the shape prediction model used in the H-shaped steel shape prediction device 48 according to this embodiment include rolling operation parameters in their input data. As mentioned above, the rolling operation parameters of the rolling process affect the amount of flange warpage. Therefore, by including the rolling operation parameters in the input data of the shape prediction model that predicts the amount of flange warpage, the shape prediction model becomes a shape prediction model that takes the rolling process into account, and the prediction accuracy of the amount of flange warpage increases. Furthermore, if the amount of flange warpage can be predicted with high accuracy in this way, it is possible to prevent the production of H-shaped steel 10 with an amount of flange warpage that exceeds the allowable range, thereby preventing increases in the manufacturing costs of H-shaped steel 10 and decreases in productivity.

Furthermore, by including attribute parameters of the H-shaped steel 10 in the input data for the shape prediction model, it becomes possible to predict the amount of flange warpage of H-shaped steel 10 with different target cross-sectional dimensions and steel type classifications using the same shape prediction model. Furthermore, by identifying cooling operation parameters that bring the amount of flange warpage predicted by the shape prediction model within the allowable range and manufacturing the H-shaped steel 10 under manufacturing conditions that include these cooling operation parameters, it becomes possible to consistently manufacture H-shaped steel 10 with an amount of flange warpage within the allowable range.

While the rolling equipment 100 shown in Figure 3 includes an example having a process computer 46 and an H-beam shape prediction device 48, this is not limiting. For example, the process computer 46 may have the functions of the H-beam shape prediction device 48, and these may be configured as a single device. Furthermore, in the H-beam shape prediction device 48 shown in Figure 5, the control unit 50 includes a cooling operation parameter identification unit 62 and a shape prediction model generation unit 64, but this is not limiting. If the flange warp amount is predicted using the H-beam shape prediction device 48, the control unit 50 does not need to include the cooling operation parameter identification unit 62. Furthermore, if the shape prediction model 68 is generated externally and stored in the storage unit 56 via the data acquisition unit 58, the H-beam shape prediction device 48 does not need to include the shape prediction model generation unit 64.

[Example 1]
To verify the effects of the present invention, constant outer dimension rolled H-beams having a material of SM400A, a web height of 900 mm, a flange width of 350 mm, a web thickness of 16 mm, and a flange thickness of 28 mm were manufactured using the H-beam manufacturing equipment 100 shown in Figure 3. The raw material for the section steel was rolled to the desired cross section through 17 passes in the rough rolling mill, 21 passes in the intermediate rolling mill, and one pass in the finishing rolling mill.

The cooling device 32 before the finishing rolling mill 28 uses water as a coolant and is composed of seven rows of diagonal spray nozzles arranged in the flange width direction, with six of these rows of nozzles being used to cool the outer flange surface of the H-beam 10. The nozzles are arranged at 50 mm intervals in the flange width direction, with the center of the lowest nozzle positioned 50 mm above the bottom edge of the flange. The nozzle spray pattern is a 50 mm x 50 mm square, and the nozzles are arranged at 100 mm intervals in the conveying direction. The total length of the cooling device 32 is 40 m, and it is divided into four banks, each 10 m long. The finishing rolling mill 28 is installed 30 m downstream from the outlet of this cooling device 32.

The cooling device 34 after the finishing mill 28 is installed 13 m downstream from the exit of the finishing mill. It uses water as a coolant and is composed of seven rows of multi-hole jet nozzles arranged in the flange width direction. The nozzles in the first to sixth rows from the bottom are used to cool the outer surface of the flange of the H-beam 10. The total length of the cooling device is 40 m, and it is divided into a total of four banks, each 10 m long.

In addition, thermometer 36 was installed 5 m upstream from the inlet of cooling device 32, thermometer 38 was installed 5 m upstream from the inlet of cooling device 34 (8 m downstream from finishing rolling mill 28), thermometer 40 was installed 10 m downstream from the outlet of cooling device 34, and shapemeter 44 was installed 35 m downstream from the outlet of cooling device 34, and each measured the temperature and flange warpage of H-beam 10. The temperatures measured by thermometers 36, 38, and 40 were used to confirm that the temperature of the H-beam cooled by cooling devices 32 and 34 was within the specified range.

<Example 1>
In Example 1 of the present invention, a shape prediction model was used that was machine-learned into a neural network model using approximately 5,000 recent historical H-section steel rolling operation parameters, cooling operation parameters, attribute parameters, and actual values of flange warpage as learning data.

The rolling operation parameters used were the roll spacing between the horizontal rolls and the relative axial position of the horizontal rolls in each pass of rough rolling, the web roll spacing and the roll spacing between the four flange legs in each pass of the intermediate universal rolling mill in intermediate rolling, the gap between the horizontal rolls in each pass of the edger rolling mill in intermediate rolling, the relative axial position of the horizontal rolls and their relative vertical position in relation to the vertical rolls, and the web roll spacing and the roll spacing between the four flange legs in the finish universal rolling mill in finish rolling. The cooling operation parameters used were the water flow rate density and cooling time of the cooling water in each cooling device 32, 34. The attribute parameter information used target values for the dimensions of the H-beam 10, the steel type classification, and target values for the mechanical properties of the H-beam 10 (yield stress, tensile strength, elongation, toughness).

In Example 1, calculations were performed repeatedly using the shape prediction model according to the flowchart shown in Figure 6, and cooling operation parameters were identified that would result in flange warpage falling within the allowable range of less than ±1.5 mm. Note that flange warpage in which the flange collapses toward the web (inward warpage) is indicated by a "- dimension," and flange collapses outward (outward warpage) is indicated by a "+ dimension."

The cooling operation parameter to be reset was the cooling time, and the finally specified cooling operation parameters were a cooling time of 6.7 seconds when the water flow rate of the cooling water in the cooling device 32 was 1600 L/ ^m² ·min, and a cooling time of 8.0 seconds when the water flow rate of the cooling water in the cooling device 34 was 1000 L/m²· ^min .

When a constant outer diameter rolled H-section steel with material SM400, web height 900 mm, flange width 350 mm, web thickness 16 mm, and flange thickness 28 mm was manufactured under manufacturing conditions including the cooling operation parameters identified as described above, the shape prediction model predicted a flange warpage of +0.4 mm, while the actual flange warpage at the longitudinal center measured with the shape meter 44 was +0.4 mm on average on both sides. These results confirm that flange warpage can be predicted with high accuracy by using a shape prediction model including rolling operation parameters, and that by manufacturing H-section steel 10 using cooling operation parameters identified using the prediction results, rolled H-section steel with a flange warpage of less than ±1.5 mm, within the allowable range, can be manufactured.

<Comparative Example 1-1>
In Comparative Example 1-1, constant outer dimension rolled H-section steel was manufactured using SM400A material, a web height of 900 mm, a flange width of 350 mm, a web thickness of 16 mm, and a flange thickness of 28 mm, in the same manner as in Example 1. The manufacturing equipment was the same as that used in Invention Example 1.

In Comparative Example 1-1, a shape prediction model was used that was machine-learned into a neural network model using approximately 5,000 recent historical cooling operation parameters, attribute parameters, and actual flange warpage values for H-beam 10 as learning data. In other words, in Comparative Example 1-1, a shape prediction model was used in which the input data did not include rolling operation parameters.

In Comparative Example 1-1, calculations were also performed repeatedly using the shape prediction model of the steel section according to the flowchart shown in Figure 6, and cooling operation parameters that satisfied the requirement for a flange warpage of less than ±1.5 mm were identified. The cooling operation parameters that were finally identified were a cooling time of 5.3 seconds when the cooling water flow rate of the cooling device 32 was 1600 L/ ^m² min. Furthermore, a cooling time of 10.0 seconds when the cooling water flow rate of the cooling device 34 was 1000 L/ ^m² min.

When a constant outer diameter rolled H-section steel with material SM400, web height 900 mm, flange width 350 mm, web thickness 16 mm, and flange thickness 28 mm was manufactured under manufacturing conditions including the cooling operation parameters identified as described above, the flange warpage predicted by the shape prediction model was +0.5 mm, while the flange warpage at the longitudinal center actually measured with the shape meter 44 was +1.6 mm on average on both sides. These results confirm that flange warpage cannot be predicted with high accuracy when using a shape prediction model that does not include rolling operation parameters, and that even when H-section steel 10 is manufactured using cooling operation parameters identified using these prediction results, it is not possible to manufacture rolled H-section steel with a flange warpage of less than ±1.5 mm, which is within the allowable range.

<Comparative Example 1-2>
In Comparative Example 1-2, constant outer dimension rolled H-beams were manufactured using the material SM400A, a web height of 900 mm, a flange width of 350 mm, a web thickness of 16 mm, and a flange thickness of 28 mm, as in Example 1. The manufacturing equipment was the same as in Compliant Example 1. In Comparative Example 1-2, flange warpage was predicted using an equation expressed as a linear sum using the dimensions of the H-beam and the flange temperature during finish rolling, as disclosed in Patent Document 2, and cooling operation parameters that satisfied the requirement for a flange warpage of less than ±1.5 mm were identified by performing repeated calculations.

The cooling operation parameters finally specified were a cooling time of 3.3 seconds when the water flow rate of the cooling water in the cooling device 32 was 1600 L/ ^m² ·min, and a cooling time of 20.0 seconds when the water flow rate of the cooling water in the cooling device 34 was 1000 L/ ^m² ·min.

When a constant outer diameter rolled H-section steel with material SM400, web height 900 mm, flange width 350 mm, web thickness 16 mm, and flange thickness 28 mm was manufactured under manufacturing conditions including the cooling operation parameters identified as described above, the flange warpage predicted by the equation expressed as a linear sum was +0.6 mm, while the actual flange warpage at the longitudinal center measured with a shape meter 44 was +1.8 mm on average on both sides. These results confirm that the equation expressed as a linear sum disclosed in Patent Document 2 cannot predict flange warpage with high accuracy, and that even when H-section steel 10 is manufactured using cooling operation parameters identified using the prediction results, it is not possible to manufacture rolled H-section steel with a flange warpage of less than ±1.5 mm, which is within the allowable range.

As described above, in Example 1, by using a shape prediction model in which rolling operation parameters, cooling operation parameters, and attribute parameters are input data and flange warpage is output data, the flange warpage of the H-section steel 10 after passing through the cooling device can be predicted with higher accuracy than in Comparative Examples 1-1 and 1-2. Furthermore, it was confirmed that by manufacturing the H-section steel 10 using cooling operation parameters specified using the prediction results, rolled H-section steel with a flange warpage within the allowable range can be manufactured.

[Example 2]
In Example 2, constant outer dimension rolled H-beams having a material of SM490A, a web height of 1000 mm, a flange width of 400 mm, a web thickness of 19 mm, and a flange thickness of 40 mm were manufactured using the same rolling equipment as in Example 1. The difference from Example 1 was the dimensions and material, i.e., the steel type, of the H-beams 10 to be manufactured. In addition, in Example 2, the flange thickness was as large as 40 mm, making offline correction of flange warpage difficult, so flange warpage was strictly controlled, and flange warpage of less than ±1.0 mm was deemed acceptable.

<Example 2>
In Example 2, calculations were repeatedly performed using the same shape prediction model as in Example 1 according to the flowchart shown in Fig. 6, and cooling operation parameters were identified that would result in a flange warpage of less than ±1.0 mm. The cooling operation parameters finally identified were a cooling time of ^6.1 seconds when the cooling water flow rate of the cooling device 32 was 1600 L/ ^m² min, and a cooling time of 10.0 seconds when the cooling water flow rate of the cooling device 34 was 1000 L/m² min.

When a constant outer diameter rolled H-section steel with material SM490, web height 1000 mm, flange width 400 mm, web thickness 19 mm, and flange thickness 40 mm was manufactured under manufacturing conditions including the cooling operation parameters identified as described above, the shape prediction model predicted a flange warpage of -0.6 mm, while the actual flange warpage at the longitudinal center measured with the shape meter 44 was -0.6 mm on average on both sides. These results confirm that flange warpage can be predicted with high accuracy by using a shape prediction model including rolling operation parameters, and that by manufacturing H-section steel 10 using cooling operation parameters identified using these prediction results, rolled H-section steel with a flange warpage of less than ±1.0 mm, within the allowable range, can be manufactured.

<Comparative Example 2-1>
In Comparative Example 2-1, as in Compliant Example 2, a constant outer dimension rolled H-section steel was manufactured with material SM490A, a web height of 1000 mm, a flange width of 400 mm, a web thickness of 19 mm, and a flange thickness of 40 mm.

In Comparative Example 2-1, as in Comparative Example 1-1, a shape prediction model was used in which the input data did not include rolling operation parameters. Following the flowchart shown in FIG. 6 , repeated calculations were performed using the shape prediction model for the section steel, and cooling operation parameters that satisfied the requirement for flange warpage of less than ±1.0 mm were identified. The cooling operation parameters finally identified were a cooling time of 5.0 seconds when the cooling water flow rate density of the cooling device 32 was 1600 L/ ^m² ·min. Furthermore, a cooling time of 10.7 seconds when the cooling water flow rate density of the cooling device 34 was 1000 L/ ^m² ·min.

When a constant outer diameter rolled H-section steel with material SM490, web height 1000 mm, flange width 400 mm, web thickness 19 mm, and flange thickness 40 mm was manufactured under manufacturing conditions including the cooling operation parameters identified as described above, the flange warpage predicted by the shape prediction model was +-0.65 mm, while the flange warpage at the longitudinal center actually measured with the shape meter 44 was -1.2 mm on average on both sides. These results confirmed that flange warpage cannot be predicted with high accuracy when using a shape prediction model that does not include rolling operation parameters, and that even when H-section steel 10 is manufactured using cooling operation parameters identified using these prediction results, it is not possible to manufacture rolled H-section steel with a flange warpage of less than ±1.0 mm, which is within the allowable range.

<Comparative Example 2-2>
In Comparative Example 2-2, constant outer dimension rolled H-beams were manufactured using the material SM490A, a web height of 1000 mm, a flange width of 400 mm, a web thickness of 19 mm, and a flange thickness of 40 mm, as in Compliant Example 2. The manufacturing equipment was the same as in Example 1. In Comparative Example 2-2, flange warpage was predicted using an equation expressed as a linear sum using the dimensions of the H-beam 10 and the flange temperature during finish rolling, as disclosed in Patent Document 2, and cooling operation parameters that satisfied the requirement for a flange warpage of less than ±1.0 mm were identified by performing repeated calculations.

The cooling operation parameters finally specified were a cooling time of 4.0 seconds when the water flow rate of the cooling water in the cooling device 32 was 1600 L/ ^m² ·min, and a cooling time of 11.5 seconds when the water flow rate of the cooling water in the cooling device 34 was 1000 L/ ^m² ·min.

When a constant outer diameter rolled H-section steel made of SM490A steel, with a web height of 1000 mm, flange width of 400 mm, web thickness of 19 mm, and flange thickness of 40 mm, was manufactured under manufacturing conditions including the cooling operation parameters identified as described above, the flange warpage predicted by the equation expressed as a linear sum was -0.8 mm, while the flange warpage at the longitudinal center actually measured with a shape meter 44 was -1.6 mm on average on both sides. These results confirm that flange warpage cannot be predicted with high accuracy when using the equation expressed as a linear sum disclosed in Patent Document 2, and that even when H-section steel 10 is manufactured using cooling operation parameters identified using the prediction results, it is not possible to manufacture rolled H-section steel with a flange warpage of less than ±1.0 mm, which is within the allowable range.

As described above, in Example 2, even if the dimensions and steel type of the H-shaped steel 10 change, the flange warpage of the H-shaped steel 10 after passing through the cooling device can be predicted with higher accuracy than in Comparative Examples 2-1 and 2-2 by using a shape prediction model in which rolling operation parameters, cooling operation parameters, and attribute parameters are input data and the flange warpage amount is output data. Furthermore, it was confirmed that by manufacturing the H-shaped steel 10 using cooling operation parameters specified using the prediction results, rolled H-shaped steel with a flange warpage amount within the allowable range can be manufactured.

DESCRIPTION OF SYMBOLS 10 H-section steel 12 Flange 14 Web 20 Heating furnace 22 Roughing rolling mill 24 Intermediate rolling mill 25 Intermediate universal rolling mill 26 Edger rolling mill 28 Finishing rolling mill 30 Rolling control device 32 Cooling device 34 Cooling device 36 Thermometer 38 Thermometer 40 Thermometer 42 Cooling control device 44 Shape meter 46 Process computer 48 H-section steel shape prediction device 50 Control unit 52 Input unit 54 Output unit 56 Storage unit 58 Data acquisition unit 60 Shape prediction unit 62 Cooling operation parameter identification unit 64 Shape prediction model generation unit 100 Rolling equipment

Claims (5)

A method for predicting the amount of flange warpage of an H-shaped steel manufactured by a rolling facility having a plurality of rolling mills for hot-rolling heated steel billets into H-shaped steel, a cooling device for cooling the H-shaped steel, and a shape meter for measuring the flange warpage of the H-shaped steel after cooling, comprising:
the plurality of rolling mills including a roughing mill, an intermediate universal mill, an edger mill, and a finishing mill;
a setting value of the roll spacing of the four flange legs; an opening between the horizontal rolls in at least one pass of the roughing rolling mill, a horizontal inclination of the horizontal rolls, and a relative axial position of the horizontal rolls; a setting value of the roll spacing of the four flange legs; an opening between the horizontal rolls in at least one pass of the edger rolling mill, a horizontal inclination of the horizontal rolls, and a relative axial position of the horizontal rolls; cooling operation parameters of the cooling device; and attribute parameters of the H-beam, input data including : operation parameters of the rolling mill; cooling operation parameters of the cooling device; and attribute parameters of the H-beam; Identifying the cooling operation parameters that cause the flange warpage amount predicted using the shape prediction method for H-section steel according to claim 1 to fall within a predetermined allowable value range,
A method for manufacturing H-beam steel, which manufactures H-beam steel under manufacturing conditions including specified cooling operation parameters. A rolling facility comprising: a plurality of rolling mills for hot rolling heated steel billets into H-shaped steel; a cooling device for cooling the H-shaped steel; a shape meter for measuring the amount of flange warpage of the H-shaped steel after cooling; and an H-shaped steel shape prediction device for predicting the amount of flange warpage,
the plurality of rolling mills including a roughing mill, an intermediate universal mill, an edger mill, and a finishing mill;
The shape prediction device has a shape prediction unit that inputs input data including rolling operation parameters of the rolling mill including at least one of the gap between the horizontal rolls in at least one pass of each pass of the roughing rolling mill, the axial relative position of the horizontal rolls, the roll spacing of the web in at least one pass of each pass of the intermediate universal rolling mill, the set value of the roll spacing of the four flange legs, the gap between the horizontal rolls in at least one pass of each pass of the edger rolling mill, the horizontal inclination of the horizontal rolls, and the axial relative position of the horizontal rolls, cooling operation parameters of the cooling device, and attribute parameters of the H-shaped steel into a shape prediction model and outputs the flange warpage amount. A method for generating a shape prediction model for predicting the amount of flange warpage of H-shaped steel produced by rolling equipment having a plurality of rolling mills for hot-rolling heated steel billets into H-shaped steel, a cooling device for cooling the H-shaped steel, and a shape meter for measuring the flange warpage of the H-shaped steel after cooling, comprising:
the plurality of rolling mills including a roughing mill, an intermediate universal mill, an edger mill, and a finishing mill;
a machine learning model is trained using as training data a plurality of data sets each including actual values of rolling operation parameters of the rolling mill, including at least one of the gap between the horizontal rolls in at least one pass of each pass of the roughing mill for H-section steel previously manufactured, the axial relative position of the horizontal rolls, the roll spacing of the web in at least one pass of each pass of the intermediate universal rolling mill, the set value of the roll spacing of the four flange legs, the gap between the horizontal rolls in at least one pass of each pass of the edger rolling mill, the horizontal inclination of the horizontal rolls, and the axial relative position of the horizontal rolls, actual values of cooling operation parameters of the cooling device, actual values of attribute parameters of the H-section steel, and actual values of the flange warpage amount;
a method for generating a shape prediction model, the method receiving the rolling operation parameters, the cooling operation parameters, and the attribute parameters as inputs and generating a shape prediction model having the flange warpage amount as output; The method for generating a shape prediction model described in claim 4, wherein the machine learning model is one of a neural network, decision tree learning, random forest, and support vector regression.