JP2021030300A - Warp prediction method in hot rolling process, warp control method, hot-rolled steel sheet manufacturing method, warp prediction model generation method and hot rolling equipment - Google Patents

Abstract

To predict a warp in the finish rolling more accurately in hot rolling equipment, and thereby enabling suppressing the warp more effectively.SOLUTION: Hot rolling equipment includes a heating step 1 of heating a slab, a coarse rolling step 3 of coarsely rolling the slab after the heating in the heating step 1, and a finish rolling step 4 of performing finish rolling of a rolled material 8 after finishing the coarse rolling. A warp prediction method predicts a warp at the tip end of the rolled material 8 in the finish rolling step 4 in the hot rolling equipment. The method contains at least one parameter of operation parameters of at least the coarse rolling step 3 of the heating step 1 and the coarse rolling step 3 as input data. The method has a warp prediction model 21A being the model where the warp information about the warp at the tip end of the rolled material 8 in the finish rolling step 4 is output data, and learned by machine learning. The warp at the tip end of the rolled material 8 in the finish rolling step 4 is predicted by using the warp prediction model 21A.SELECTED DRAWING: Figure 1

Description

The present invention is a technique relating to a hot rolling facility constituting a hot rolling line for performing a hot rolling process (hot rolling process), and in particular, for predicting and suppressing warpage of the tip of a rolled material in finish rolling. Regarding technology.

In the hot rolling line for producing hot-rolled steel sheets, for example, a continuously cast slab having a thickness of about 200 to 300 mm is heated to about 1100 to 1300 ° C. in a heating furnace. After the heat treatment, the slab is roughly rolled into a sheet bar having a thickness of about 20 to 80 mm by a plurality of or a single rough rolling machine. Then, the rolled material composed of the sheet bar is finished rolled to a thickness of about 1 to 20 mm by a finishing rolling machine having a plurality of stands.

In rough rolling and finish rolling, it is known that when the steel sheet (rolled material) to be rolled or the rolling mill has vertical asymmetry, the steel sheet is warped in the rolling direction. Warpage in hot rolling includes the temperature difference between the upper and lower surfaces of the rolled material, the peripheral speed difference between the upper and lower work rolls, the diameter difference between the upper and lower work rolls, the upper and lower friction coefficient difference between the work roll and the rolled material, and the difference in the rolled material. It is known that it occurs due to the approach angle and the like.
However, in hot rolling, there are few operating conditions (operating parameters) that can be freely changed in order to control warpage. If the upper and lower work roll diameter difference is designed according to a specific material in order to suppress the warp, warpage may occur when, for example, a steel plate made of another material having a different plate thickness is rolled.

Further, the angle of entry of the rolled material into the rolling mill can be indirectly changed by changing the height of the entry side table. However, similarly, if the approach angle of the rolled material is set according to a specific material in order to control the warp, there is a possibility that the warp of other materials cannot be suppressed. The coefficient of friction can also be changed by finishing the upper and lower work rolls so that the polishing roughness is different, but it is difficult to apply for the same reason.

Controlling the warpage in rolling through the vertical temperature difference of the rolled material is an effective method, but it is necessary to newly install a cooling device for adjusting the vertical temperature difference of the rolled material. Here, since the temperature deviation within the plate thickness of the rolled material is alleviated with the passage of time, it is preferable to arrange the above-mentioned cooling device in the vicinity of the rolling mill entry side in the finish rolling. However, there are many peripheral equipment such as crop shears, thermometers, and side guides near the entrance side of the rolling mill, and it is difficult to install a cooling device at an effective position.

The difference in peripheral speed of the work roll can be changed by driving the drive motors of the upper and lower rolls independently. However, the finishing rolling mill often has a structure in which two work rolls up and down are rotated by one drive motor, and a large equipment modification is required to change the peripheral speed difference.
In addition, since there are many factors that cannot be accurately measured, such as the approach angle of the rolled material near the entrance side of the rolling mill and the vertical temperature difference, it has been conventionally recognized that it is difficult to stably suppress the warpage during rolling.

The downward warp (downward warp) is restrained and suppressed by the transport table. Therefore, with respect to the warp in the vertical direction, the downward warp rarely occurs, and the actual warp direction is often the upward warp. Further, since the warp is suppressed by the mass of the steel sheet itself on the exit side of the rolling mill, the warp often occurs on the rolling tip side and not on the tail end side.

In addition, if warpage occurs during rolling, there is a concern that the next stand may be poorly bitten or the peripheral equipment may be damaged. Therefore, the warp generated in rough rolling may be corrected by a flatness straightening machine such as a roller leveler before being conveyed to the finishing rolling machine. Further, as described above, since the warp of the rolling tail end side tends to be relatively small, in the reverse type rough rolling mill, after rolling in the upstream direction of the line, the work roll gap is opened in the downstream direction of the line. It is possible to plate and transfer to the next rolling mill with the rolling tip flat.

On the other hand, in the finishing rolling mill, there is no space for arranging the straightening apparatus between the rolling mills, and if warpage occurs, it directly leads to poor biting into the next stand and damage to peripheral equipment. In particular, the strength of steel sheets is high in recent high-strength high-tensile materials. In addition, the thick line pipe material has a large plate thickness and high rigidity. For this reason, when the steel sheet comes into contact with the equipment around the rolling mill due to warpage, it often leads to breakage.

Conventionally, as a method for improving warpage in rolling, techniques such as those described in Patent Documents 1 and 2 have been disclosed.
In Patent Document 1, a relational expression is prepared in advance for obtaining the tip warp amount of the rolled material by using the vertical set temperature difference and the slab staying time in the heating zone of the heating furnace and the tropics as variables. Then, the temperature difference between the heating zone of the heating furnace and the temperature difference between the upper and lower parts of the tropics is set so that the tip warp amount becomes the target value.

Further, Patent Document 2 covers a hot rolling line in which a hot rolling method including a rough rolling step, a cooling step of water-cooling the rolled material after rough rolling to a predetermined temperature, and a subsequent finish rolling step is carried out. And. Then, in Patent Document 2, in the cooling step, water cooling is performed so as to control the difference between the upper surface temperature and the lower surface temperature of the rolled material at the time of stopping cooling within a range of 100 ° C. according to the steel type of the rolled material, and the finish rolling step Then, the upper and lower surfaces of the rolled material are asymmetrically cooled based on the warp state in the latter pass to control the tip warp generated in the second pass.

Japanese Patent No. 3251455 Japanese Unexamined Patent Publication No. 6-71326

In the method described in Patent Document 1, the vertical temperature difference in the heating furnace is changed according to the target material whose warpage should be controlled. However, the vertical temperature in the heating furnace affects the warpage in rolling through the vertical difference in the deformation resistance of the rolled material and the vertical difference in the thickness and composition of the oxide scale generated in the heating furnace. Therefore, it is necessary to change the vertical temperature difference in the heating furnace according to the chemical composition of the target material whose warpage should be controlled. Generally, the inside of the heating furnace has a length of about 30 to 50 m, and a plurality of slabs are heated at the same time. Therefore, when the upper and lower temperatures of the heating furnace are set for a specific target material, the chemical components in the heating furnace are set. On the contrary, the warp may increase for other slabs with different values.

In addition, it is necessary to operate the furnace according to the preset residence time of each band in the heating furnace. However, for example, if the rolling line is stopped due to an operation trouble with the preceding material and the residence time exceeds the planned residence time, The effect of warpage control cannot be expected at all. In this way, the effectiveness of changing the operating conditions of the heating furnace in order to control the warp is doubtful.

Further, in Patent Document 1, warpage at the time of finish rolling is predicted from the operation parameters of the heating furnace. However, in the case of predicting the warpage in the finish rolling of the hot rolling process, since many steps are intervened between the heating furnace and the finish rolling process, the operation of the heating furnace is performed as in the method described in Patent Document 1. Prediction of warpage from parameters alone may result in poor prediction accuracy.
Further, since the method described in Patent Document 2 focuses on the vertical temperature difference of the rolled material, which is one of the causes of warpage, it seems to be useful for suppressing warpage, but the temperature of the upper surface and the lower surface. The tolerance of the difference is as wide as 100 ° C, and its effectiveness is doubtful. Further, as described in the examples, a water cooling time of about 80 seconds is required after rough rolling, and the rolling cannot be performed for a time corresponding to the water cooling treatment, which hinders productivity.

The present invention has been made by paying attention to the above points, and provides a technique capable of more accurately predicting the warp of a rolled material in finish rolling and effectively suppressing the warp in a hot rolling facility. The purpose is.

The present inventors have diligently studied a method for preventing tip warpage by analyzing the state of occurrence of tip warpage in finish rolling of a hot rolling line while comparing it with operation data. As a result, it was found that it is possible to control the tip warpage in the finish rolling by manipulating the operating condition group of the slab heating step, the rough rolling step and the finish rolling step. Furthermore, it has been found that the tip warpage amount can be controlled more effectively by manipulating at least one of the parameters of the number of rough rolling passes and the number of descaling as the operating conditions for rough rolling.

Then, in order to solve the problem, one aspect of the present invention includes a heating step of heating the slab, a rough rolling step of rough rolling the slab after heating in the heating step, and a finish rolling of the rolled material after the rough rolling. A method for predicting warpage of the tip of a rolled material in the finish rolling process in a hot rolling process including a finishing rolling process, in which one or two or more parameters selected from the operation parameters of the rough rolling process are used as input data. Predicting the warp of the tip of the rolled material in the finish rolling process by using the warp prediction model learned by machine learning, which includes the warp information about the warp of the tip of the rolled material in the finish rolling process as output data. The gist is to do.
Further, in the aspect of the present invention, the warp of the rolled material in the finish rolling process is predicted by the method of predicting the warp of the above aspect, and the operating conditions of the rough rolling process are reset so that the predicted warp becomes small. Is the gist.
Further, an aspect of the present invention is a method for manufacturing a hot-rolled steel sheet using the method for controlling warpage of the above aspect.

Further, an aspect of the present invention includes a heating step of heating the slab, a rough rolling step of rough rolling the slab after heating in the heating step, and a finish rolling step of finishing rolling the rolled material after the rough rolling. It is a method of generating a warp prediction model used for predicting the warp of the tip of a rolled material in the finish rolling process in the inter-rolling process, and at least in the operation record data in the heating process and the rough rolling process. One or two or more operation record data selected from the operation record data of the above are used as input record data, and the record data of the amount of warpage of the tip of the rolled material in the above finish rolling process in the hot rolling process using the input record data. The gist is to acquire a plurality of training data with the output actual data as the output result data, and to generate a warp prediction model by machine learning using the acquired plurality of training data.

Further, in the embodiment of the present invention, a heating furnace for heating the slab, a rough rolling machine for rough rolling the slab, a rough rolling facility having a descaling facility for descaling the slab, and a rolled material after rough rolling are finished. It is a hot-rolling facility having a finishing rolling facility for rolling, and includes a warp prediction unit for predicting warpage information of the tip of a rolled material in the finish rolling facility using a warpage prediction model. It is a model learned by machine learning, includes one or two or more parameters selected from the operation parameters of the rough rolling equipment as input data, and outputs warp information regarding the warp of the tip of the rolled material in the finish rolling equipment. The gist is that it is a learning model that uses data.

According to the aspect of the present invention, in a hot-rolling facility for producing a hot-rolled steel sheet by heating a slab in a heating step (heating furnace) and rolling the heated slab by a group of rough rolling mills and a group of finishing rolling mills. , It is possible to more effectively predict the tip warp of the rolled material in finish rolling and suppress the occurrence of warpage at the tip of the rolled material.

It is a figure which shows the process (hot rolling process) of the hot rolling process which concerns on embodiment based on this invention. It is a figure which shows the heat spreading equipment which concerns on embodiment based on this invention. It is a figure which shows the structure of the warp control. It is a figure which shows the warp detection example. It is a figure which shows the example of the amount of warpage. It is a schematic diagram which shows the neural network which concerns on embodiment based on this invention. It is a figure which shows the example of the relationship between the tip warp amount and the number of descaling. It is a figure which shows the example of the relationship between the tip warp amount and the number of rough passes.

Hereinafter, embodiments according to the present invention will be described with reference to the drawings.
In the present embodiment, a hot-rolling facility for manufacturing a hot-rolled steel sheet will be described as an example.
(Constitution)
In the hot rolling equipment (configuration of the hot rolling line) of the present embodiment, as shown in FIG. 1, the heating step 1, the width reduction step 2, the rough rolling step 3, and the finish rolling step 4 are carried out in this order. A hot-rolled steel sheet is manufactured in. Reference numeral 5 indicates a steelmaking process for manufacturing a slab (steel piece). As shown in FIG. 2, the hot rolling equipment of the present embodiment includes a rough rolling equipment 15 including a heating furnace 10, a width reduction device 11, a rough rolling machine 13 and a descaling device 14, a finishing rolling machine 16, a runout table 6, and a runout table 6. It has a down coiler 7. As the hot rolling equipment, other known equipment such as a flatness straightening machine provided downstream of the rough rolling equipment 15 may be separately provided. The inside of the heating furnace 10 is divided in the order of pre-tropical, heating zone, and even-tropical, and heating is performed to a target heating temperature while transporting the slab inside the heating furnace 10.

<Heating step 1>
In the heating step 1, the slab is heat-treated using the heating furnace 10. In the heating step 1, the slab cast in the steelmaking step 5 is charged into the heating furnace 10 and heated under preset operating conditions (operating parameters), and the slab after heating to the target heating temperature is heated in the heating furnace 10. Extract from.
<Width reduction process 2>
In the width reduction step 2, the width reduction device 11 performs the width reduction treatment of the slab. The width reduction device 11 is a device that reduces the slab from both sides in the plate width direction under preset operating conditions (operating parameters).

<Rough rolling process 3>
In the rough rolling step 3, the rough rolling process of the slab is performed by using the rough rolling machine 13 and the descaling device 14. The rough rolling mill 13 is an apparatus for rough rolling a slab into a sheet bar. In the rough rolling step 3, the number of rough rolling passes can be adjusted by performing reverse rolling as necessary. The descaling device 14 is a device that injects high-pressure water toward the slab in order to remove the scale generated by heating and rough rolling.
<Finishing rolling process 4>
In the finish rolling step 4, the sheet bar (rolled material 8) after rough rolling is finish-rolled to the finish plate thickness by the finish rolling mill 16 of a plurality of stands.

<Downstream process>
The rolled material 8 (steel plate) after finish rolling is cooled to a predetermined temperature while being conveyed on the runout table 6, and then wound into a coil by a down coiler 7. The coil is then cooled in the coil yard to room temperature.
<War control unit 20>
The heat spreading equipment of the present embodiment further includes a warp control unit 20. As shown in FIG. 3, the warp control unit 20 includes a warp prediction unit 21 and a warp suppression unit 22.
<War Prediction Unit 21>
The warp prediction unit 21 is based on at least one or two or more parameters selected from the operation parameters in the rough rolling process 3 among the operation parameters in the heating process 1 and the operation parameters in the rough rolling process 3. The process of predicting the warp of the tip of the rolled material 8 in No. 4 is performed.

The predicted warp of the tip of the rolled material 8 is, for example, the warp on the exit side of the first stand in the finishing rolling mill 16. The warp on the exit side of the stand after the second stand may be targeted. Further, the predicted warp position may be set at two or more points. When the predicted warp positions are set to two positions, for example, the warp of the finishing rolling mill 16 at two points, the warp on the exit side of the first stand and the warp on the exit side of the second stand, is predicted. When predicting warpage at two or more locations, one warp prediction model 21A may be used to predict two or more warpages, but individual warp prediction models 21A with different input data may be used to individually predict warpage. Is also good. However, the warp information may be output as an image showing the curvature of the warp of the tip of the rolled material 8 or the warp.

Further, the warp information predicted (output) by the warp prediction unit 21 is, for example, a warp amount.
The warp prediction unit 21 includes a warp prediction model 21A for predicting warpage, and obtains warp information by the warp prediction model 21A.
The warp prediction model 21A is a model learned by known machine learning such as a neural network. The warp prediction model 21A includes at least one or two or more parameters selected from the operation parameters in at least the rough rolling step 3 of the heating step 1 and the rough rolling step 3 as input data, and the rolled material in the finish rolling step 4 8 This is a learning model using warp information (for example, the amount of warp) related to the warp of the tip as output data. In the present embodiment, the amount of warpage will be described as an example of warpage information. The warp prediction model 21A is expressed, for example, in the form of a functional expression.

As the input data of the warp prediction model 21A, the parameters in both the 1 or 2 or more parameters selected from the operation parameters of the heating process 1 and the 1 or 2 or more parameters selected from the operation parameters of the rough rolling process 3 are used. It is preferable to have. Prediction accuracy is further improved when both the operation parameter of the heating process 1 and the operation parameter of the rough rolling process 3 are used as input data.
As the operation parameters of the rough rolling step 3 constituting the input data, at least one parameter of the number of rough rolling passes and the number of descaling in rough rolling can be exemplified. These parameters have a high correlation with warpage (see Examples) and have a relatively high degree of freedom of change.
Further, as the operation parameters of the heating step 1 constituting the input data, for example, the slab temperature at the time of charging the heating furnace 10, the pre-tropical heating zone of the heating furnace 10, the heating zone, and the temperature inside the heating furnace can be exemplified.

In addition to the above parameters, the input data of the warp prediction model 21A includes, for example, other parameters among the operation parameters in the heating step 1, the width reduction step 2, the rough rolling step 3, and the finish rolling step 4, especially the warp. May include other parameters that are presumed to be correlated with. Further, the operation parameters of the steelmaking process 5 may be included as the input data.

Then, when each operation parameter constituting the input data of the warp prediction model 21A is input, the warp prediction unit 21 inputs the parameter as input data to the warp prediction model 21A (for example, a relational expression) set in advance. , Warp information (for example, warp amount) is calculated.

<War suppressor 22>
When the warp suppressing unit 22 determines, for example, that the warp amount calculated (predicted) by the warp prediction unit 21 exceeds a preset allowable threshold, the predicted warp amount becomes smaller, and the operation parameter of hot rolling Of these, the value of the operation parameter that correlates with the amount of warpage is reset. Therefore, the processing of the warp control unit 20 is a feedforward control.
There is a correlation with the amount of warpage, and the operating parameters (operating conditions to be reset) that are reset to suppress the amount of warpage may be the same parameters as the input data of the warp prediction model 21A, or may be different operating conditions. Is also good.

As the operation parameter that has a correlation with the amount of warpage and is reset to suppress the amount of warpage, a parameter having a correlation with a known warp (for example, a parameter as described in the background art) may be adopted. However, there is a correlation with the amount of warpage, and the operation parameter in the rough rolling step 3 is preferable as the operation parameter to be reset in order to suppress the amount of warpage.
In particular, it is preferable that the operating parameters are at least one of the number of rough rolling passes and the number of descaling in the rough rolling step 3. This is because these parameters have a relatively high degree of freedom of change.

As a method of resetting the parameters, for example, the correlation between the parameter to be reset and the amount of warpage is obtained in advance, and the current value is changed in the direction in which the warp is estimated to be small based on the obtained correlation. By doing so, the reset value of the parameter to be reset is determined.
When the parameter to be reset is the input data of the warp prediction model 21A, the warp prediction unit 21 is operated every time the parameter is reset until the warp amount becomes equal to or less than the preset allowable threshold value. Then, it is good to repeat the prediction of the amount of warpage again.
When performing such processing, the warp suppressing unit 22 is configured to include, for example, an operating condition resetting unit 22A, a warp re-evaluation unit 22B, and an end determination unit 22C (FIG. 3).

For example, the operation condition resetting unit 22A warps the current reset parameter value based on the relationship between the reset parameter acquired in advance and the warp amount for the operation parameter to be reset (also referred to as the reset parameter). Change to the direction in which the amount is estimated to be smaller.
When the operating condition resetting unit 22A resets the parameters, the warp re-evaluation unit 22B inputs the reset parameters to the warp prediction unit 21 and acquires the warp amount as a predicted value from the warp prediction unit 21. .. Specifically, the input parameters of the warp prediction model 21A are reset, and an operation command is supplied to the warp prediction unit 21. In addition to the reset operation parameters, the previous operation parameters are used as the input data of the warp prediction unit 21.
The end determination unit 22C determines whether or not the acquired warp amount is larger than the allowable threshold value. When the acquired warp amount is larger than the permissible threshold value, the end determination unit 22C supplies an operation command to the operation condition reset unit 22A. When the acquired warp amount is equal to or less than the allowable threshold value, the end determination unit 22C ends the process of the warp suppression unit 22.

<About the warp prediction model 21A>
The warp prediction model 21A in the present embodiment will be described.
The warp prediction model 21A is a learning model learned by machine learning as described above.
As described above, the generated warp prediction model 21A includes, as input data, one or two or more parameters selected from at least one or more parameters selected from the operation parameters in the rough rolling step 3 of the heating step 1 and the rough rolling step 3, and finish rolling. This is a model in which the warp information (for example, the amount of warp) regarding the warp of the tip of the rolled material 8 in the step 4 is used as output data. The warp prediction model 21A is expressed in the form of a relational expression for obtaining the amount of tip warp generated during finish rolling.

The warp prediction model 21A is generated by the following processing.
First, the rolling operation at the target hot rolling facility is repeated, and the operation record data of the hot rolling facility at that time is used as the input record data (explanatory variable), and the finishing rolling process 4 when the input record data is used. The actual data of the warp of the tip of the rolled material 8 at the preset position in is used as the output actual data, and the input actual data and the output actual data are acquired. By repeating this, a plurality of learning data are acquired. Each learning data consists of a set of one or more input actual data and output actual data.

As the operation record data to be used, for example, the operation parameters in the heating step 1 and the operation parameters in the rough rolling process 3 may be adopted. As the operation parameters in the heating step 1, for example, the slab temperature at the time of charging the heating furnace 10, the temperature inside the heating furnace 10 in the pre-tropics, the heating zone and the soaking tropics, and the slabs in each of the pre-tropics, the heating zones and the soaking tropics. Adopt the staying time. For example, the number of rolling passes and the number of descaling are adopted as the operation parameters in the rough rolling step 3.

Here, the above operating parameters are taken as an example, but the explanatory variables for predicting the tip warpage amount are not limited to these. For example, other operating parameters described above include:
・ Steel making process 5: Distinguishing continuous casting machine, slab casting speed, slab temperature at the time of casting completion ・ Heating process 1: Slab temperature at the time of charging, slab temperature at the time of extraction ・ Width rolling process 2: Width rolling amount, slab transport Speed, slab transfer pitch ・ Rough rolling process 3: Rolling rate of each pass, rolling speed, work roll diameter, rolling temperature ・ Finish rolling process 4: Rolling rate of each pass, rolling load, rolling speed, work roll diameter, between stands Amount of cooling water

To acquire the actual data of the amount of tip warpage in finish rolling, for example, as shown in FIG. 4, the tip of the steel plate between the finish rolling stands 30 is photographed by the area camera 18, and the captured image data is quantified by image processing. Obtained at. Reference numeral 30A represents a work roll, and 30B represents a backup roll. Further, the warp height of the steel sheet may be directly measured using a distance meter such as a microwave. Here, as shown in FIG. 5, the tip warp amount H is defined as the difference in height between the tip 8A of the steel sheet (rolled material 8) and the tip 8A of the steel sheet at a position separated by a certain horizontal distance L. To do. The horizontal distance L from the steel plate tip 8A for defining the tip warp amount H may be determined from the camera field of view ARA in which the steel sheet 8 does not interfere with the peripheral equipment between the rolling mills. Generally, the distance between the finishing rolling mills 16 is about 4 m, so the horizontal distance L may be about 2 to 4 m. When the horizontal distance L is 2 m or less, the amount of warpage H becomes small and evaluation becomes difficult. As an index of the amount of warp H, the warp shape may be approximated by, for example, a quadratic curve and quantified as the warp curvature instead of the difference in the height direction.

Further, the warp measuring device (camera 18) does not need to be permanently installed, and once the relational expression (warp prediction model 21A) between the warp amount H of the tip 8A and the operation performance data is obtained, after that. May be removed.
After acquiring a plurality of learning data, machine learning is performed using the acquired plurality of learning data to generate a warp prediction model 21A.

For example, as shown in FIG. 6, a relational expression expressing the warp prediction model 21A for obtaining the tip warp amount H of finish rolling can be generated by using a neural network which is one of the machine learning methods. Neural networks are widely used in the fields of mathematical phenomenon prediction and image recognition with strong non-linearity, and are used here as function approximators. The learning method of the number of intermediate layers and nodes, the activation function f, and the weighting coefficient w of each neuron may be determined so that the prediction accuracy of the tip warp amount H is high. FIG. 6 shows an example in which one intermediate layer, five nodes, and an activation function using a sigmoid function are used to determine the amount of tip warpage on the exit side of a certain finishing rolling stand.

The relational expression (warp prediction model 21A) for obtaining the tip warpage amount of finish rolling does not need to be created for a material that does not have a concern of warpage, and may be prepared only for the target material whose warpage should be controlled. In addition, if the plate thickness, plate width, chemical composition, etc. of the target material whose warpage should be controlled are widely distributed and the occurrence of warpage is different, the plate thickness, plate width, chemical composition of the target material whose warpage should be controlled are different. A plurality of warp prediction models 21A may be prepared and used properly according to the above.
The machine learning technique used to generate the warp prediction model 21A is not limited to the neural network, and other known machine learning techniques may be adopted. Examples of other machine learning techniques include techniques such as decision tree learning, random forest, and support vector regression. In addition, as a machine learning technique, a technique such as a Gaussian process or a k-nearest neighbor method may be used.

(Operation and others)
In the present embodiment, the relational expression which is the warp prediction model 21A for obtaining the tip warp amount of the finish rolling is obtained in advance by machine learning using the data acquired in the actual operation.
Then, the warp prediction unit 21 inputs the operating conditions (operation parameters) to be used into the warp prediction model 21A before the hot rolling operation using the hot rolling equipment, and predicts the amount of warpage in finish rolling. When the warp amount predicted by the warp prediction unit 21 exceeds the permissible threshold value, the warp suppression unit 22 sets and changes a part of the operation parameters in the direction of reducing the warp. As a method of changing the setting of a part of the operation parameters, an appropriate operation parameter may be searched by trial and error, or a known optimization calculation method may be applied.

After that, the heat extension operation will be carried out. Since the operating conditions are set in advance so that the amount of warpage is small, it is possible to prevent defects due to plate warpage in hot rolling. The warp suppressing unit 22 may execute the heat spreading operation after the start of the heat spreading operation as long as it is before the step in which the parameter to be reset is used.
Here, even after the warp prediction model 21A is set, the learning data for the warp prediction model 21A is acquired during the actual operation, and the warp prediction model 21A is updated as appropriate. You can do it.

When the inventor investigated the relationship between the measured tip warpage of the finished rolling and the operation data, the slab temperature when the heating furnace 10 was charged, the pre-tropical, heating zone, and average tropical furnace temperatures of the heating furnace 10 were investigated. A particularly strong correlation was observed between the slab staying time in each of the pre-tropical, heating zone and average tropics, the number of rolling passes and the number of descaling in the 13 groups of rough rolling mills. The reason for showing a strong correlation with the operating conditions of the heating furnace 10 is that when the rolled material 8 is uniformly heated in the heating furnace 10, the temperature deviation in the plate thickness cross section becomes small and the occurrence of warpage is suppressed. The effect of soaking the rolled material 8 in the heating furnace 10 is suggested.
In addition, the number of passes and the number of descaling in rough rolling are affected by changes in the coefficient of friction due to peeling of the oxide scale and surface irregularities, and the effect of vertical temperature deviation due to watering of the cooling water on the upper surface of the steel sheet during descaling. It is suggested.

Next, an example of a method for determining the number of descaling times in rough rolling for suppressing warpage of a target material whose tip warp should be controlled will be described.
In the state after the extraction of the heating furnace 10 and before the start of rough rolling, the slab temperature at the time of charging the heating furnace 10, the pre-tropical, heating zone and uniform tropical furnace temperature of the heating furnace 10, each pre-tropical, heating zone and uniform tropical Each operating parameter of the slab staying time in each is fixed. Further, since the planned number of passes of the rough rolling mill 13 is determined in advance according to the plate thickness, plate width and deformation resistance after finish rolling, among the explanatory variables of the relational expression for obtaining the tip warpage amount, in rough rolling. Only the number of descaling is undecided. Therefore, the tip warp amount is calculated by assuming the number of descaling in rough rolling as many times as the number of descaling in rough rolling can be taken, and the absolute difference between the tip warp amount and the control target value is absolute. The number of descaling times when the value is the smallest is adopted. However, when the amount of warpage cannot be sufficiently reduced only by adjusting the number of descaling times, it is possible to review the number of rough rolling passes, such as adjusting the reduction rate per pass. The operating condition resetting unit 22A also includes such a function.

The descaling removes the surface oxidation scale by injecting high-pressure water from the header arranged on the entry side of the rough rolling mill 13, and the number of times is equal to the number of rough rolling passes at the maximum. The minimum number of times can be determined in advance based on the past operation results of the target material whose tip warp should be controlled, and in consideration of the presence or absence of biting defects of the surface oxidation scale due to insufficient descaling.
Here, the inventor found that descaling in rough rolling not only affects the temperature difference between the upper and lower surfaces, but also affects the peeling of the oxide scale and the change in the coefficient of friction due to surface irregularities, which is simply the case. We obtained a new finding that it has a stronger correlation with warpage than when it is cooled.

(Other)
(1) In the hot rolling operation, the following parameters can be exemplified as operation parameters that can affect the warp. Therefore, it is possible to control the warp by changing such operation parameters.
-Steelmaking process 5: Casting speed, cooling water volume, lead time (time from extraction from continuous casting machine to charging into heating furnace)
-Heating process 1: Slab charging temperature, furnace time, furnace temperature, gas flow rate / width reduction process 2: width reduction amount, transfer speed / rough rolling process 3: number of rough rolling passes (also referred to as the number of rough passes), Reduction schedule, number of descaling, rolling speed, pick-up (amount of deviation between the center of plate thickness and the center of roll gap when the rolled material 8 bites into the rolling mill)
・ Finish rolling process 4: Rolling schedule, number of passes, rolling speed

(2) As the operation parameters to be reset by the warp suppressing unit 22, the number of rough rolling passes and the number of descaling in rough rolling are exemplified.
Regarding the number of descaling in rough rolling, in the case of a device configuration that can turn on / off the descaling individually on the top and bottom, the operation parameter is to change the difference between the number of upper descaling and the number of lower descaling, the number ratio, etc. May be reset.
Further, the operation parameter reset by the warp suppressing unit 22 is not limited to the above-mentioned operation parameter, and may be a parameter having a correlation with other warpages. For example, as such operation parameters, the different speed rates of the upper and lower work rolls, the pickup, the furnace temperature of the heating furnace 10, and the like can be exemplified.

(effect)
In this embodiment, the following effects are obtained.
(1) In the present embodiment, the heating step 1 for heating the slab, the rough rolling step 3 for rough rolling the slab after heating in the heating step 1, and the finish rolling step 4 for finish rolling the rolled material 8 after rough rolling. This is a warp prediction method for predicting the warp of the tip of the rolled material 8 in the finish rolling step 4 in the hot rolling process including the above, and is a model learned by machine learning. It has a warp prediction model 21A that includes one or more parameters selected from the operation parameters and uses warp information about the warp of the tip of the rolled material 8 in the finish rolling step 4 as output data, and uses the warp prediction model 21A. , Predict the warp of the tip of the rolled material 8 in the finish rolling step 4.
According to this configuration, since the warp is predicted by the operation parameters in the rough rolling that is closer to the finish rolling than the heating furnace 10, the accuracy of the warp prediction is improved.

(2) In the present embodiment, as the input data of the warp prediction model 21A, one or two or more parameters selected from the operation parameters of the heating step 1 and one or two or more parameters selected from the operation parameters of the rough rolling process 3 are used. And have.
According to this configuration, the warp of the tip portion of the rolled material 8 in the finish rolling is predicted by the combination of the operating conditions in the heating furnace 10 and the operating conditions in the rough rolling, so that the warp can be predicted more accurately.
(3) The present embodiment includes at least one parameter of the number of rough rolling passes and the number of descaling in rough rolling as the operation parameters of the rough rolling step 3 constituting the input data.
According to this configuration, the warp of the tip portion of the rolled material 8 in the finish rolling is predicted by the combination of the operating conditions in the heating furnace 10 and the operating conditions in the rough rolling, so that the warp can be predicted more accurately.

(4) The present embodiment has one or two or more parameters selected from the operating parameters of the heating step 1 as input data, and as the operating parameters of the heating step 1 constituting the input data, when the heating furnace 10 is charged. Includes one or more parameters selected from the slab temperature, the pre-tropical, heating zone, and even-tropical furnace temperature of the heating furnace 10.
According to this configuration, the warp of the tip portion of the rolled material 8 in the finish rolling is predicted by the combination of the operating conditions in the heating furnace 10 and the operating conditions in the rough rolling, so that the warp can be predicted more accurately.

(5) In the present embodiment, the warp prediction unit 21 predicts the warp amount of the rolled material 8 in the finish rolling step 4, and 1 is selected from the operation parameters of the rough rolling step 3 so that the predicted warp amount becomes small. Or reset two or more parameters.
According to this configuration, it is possible to suppress warpage in finish rolling in a feedforward manner.

(6) In the present embodiment, the parameters to be reset are at least one of the number of rough rolling passes and the number of descaling in the rough rolling step 3.
According to this configuration, it is possible to suppress warpage in finish rolling in a feedforward manner.
Further, according to this configuration, the number of rough rolling passes and the number of descaling in the rough rolling step 3 have a relatively wide range that can be set among the operating parameters of hot rolling, so that the finishing rolling other than warpage can be performed. It is possible to suppress the warp more effectively while suppressing the influence of.
(7) A hot-rolled steel sheet is manufactured by using the warp control of the present embodiment.
According to this configuration, the warp can be controlled more effectively, so that the yield of hot-rolled steel sheet production is improved.

(8) In the present embodiment, the heating step 1 for heating the slab, the rough rolling step 3 for rough rolling the slab after heating in the heating step 1, and the finish rolling step 4 for finish rolling the rolled material 8 after rough rolling. This is a method for generating a warp prediction model 21A used for predicting the warp of the tip of the rolled material 8 in the finish rolling step 4 in the hot rolling process having the above, and is the operation record data at least in the heating step 1. And one or more operation record data selected from the operation record data in the rough rolling process 3 is used as the input record data, and the tip of the rolled material 8 in the finish rolling process 4 in the hot rolling process using the input record data. A plurality of training data are acquired using the actual data of the amount of warpage of the part as output actual data, and the warp prediction model 21A is generated by machine learning using the acquired plurality of learning data.
According to this configuration, it is possible to reliably generate a prediction model for suppressing warpage.

(9) In the present embodiment, machine learning selected from neural networks, decision tree learning, random forest, and support vector regression is used as machine learning to generate the warp prediction model 21A.
According to this configuration, machine learning for generating the warp prediction model 21A can be reliably performed.

(10) In the present embodiment, a heating furnace 10 for heating the slab, a rough rolling machine 13 for rough rolling the slab, a rough rolling facility 15 having a descaling facility for descaling the slab, and rolling after rough rolling. It is a hot rolling facility having a finish rolling facility for finish rolling the material 8, and includes a warp prediction unit 21 for predicting warpage information of the tip of the rolled material 8 in the finish rolling step 4 using the warp prediction model 21A. The warp prediction model 21A is a model learned by machine learning, includes one or two or more parameters selected from the operation parameters of the rough rolling step 3 as input data, and includes the tip of the rolled material 8 in the finishing rolling step 4. This is a learning model that uses the warp information related to the warp of the part as output data.

According to this configuration, the warp of the tip portion of the rolled material 8 in the finish rolling is predicted by the combination of the operating conditions in the heating furnace 10 and the operating conditions in the rough rolling, so that the warp can be predicted more accurately. The operation condition resetting unit 22A for resetting the operation parameters of the rough rolling equipment 15 is provided in the direction in which the warp predicted by the warp prediction unit 21 is estimated to be small.
Further, according to the present embodiment, it is possible to suppress warpage in finish rolling in a feedforward manner.

(11) The present embodiment has a warp re-evaluation unit 22B that causes the warp prediction unit 21 to re-estimate the warp using the operation parameters reset by the operation condition resetting unit 22A.
According to this configuration, it is possible to more reliably suppress warpage in finish rolling in a feedforward manner.

Examples of the present invention are shown below.
Here, the present invention is applied to a hot rolling apparatus having stands R1, R2 of two rough rolling mills including one reversible rolling mill and first to seventh stands F1 to F7 of seven finishing rolling mills. The applied embodiment will be described. Table 1 shows the equipment specifications of the rolling mill.

(Example 1)
An example of the present invention and a conventional example were compared with each other using a material made of carbon steel having a finished plate thickness of 18 mm and a finished plate width of 1500 mm as a warp control target material for finish rolling.
First, the relational expression (warp prediction model 21A) for obtaining the tip warp amount of the first stand (F1), which is the first stage of finish rolling, is used as the operation parameter of the heating step 1, and the slab temperature at the time of charging the heating furnace 10 and heating. Explanatory variables are the temperature inside the furnace 10 in the pre-tropical, heating zone and uniform tropical zone, the staying time of the slab in each of the pre-tropical, heating zone and uniform tropical zone, and the number of rolling passes and the number of descaling as the operation parameters of the rough rolling process 3. Created as.

The amount of tip warpage was measured using an area camera 18 for photographing the state of the steel plates between the rolling mills. Regarding the definition of the tip warp amount H, the horizontal distance L from the tip of the steel plate in FIG. 5 was set to 3 m. The relational expression for obtaining the tip warp amount was created using a neural network. Here, as a condition of the neural network, the intermediate layer is set to one layer and the number of nodes is set to five. The sigmoid function was used as the activation function. The operation data of the coils of the above dimensions and steel grade was prepared for 500 coils, 450 coils were used for model learning, and the model prediction accuracy was verified with the remaining 50 coils. The amount of tip warpage for 500 coils was 45.5 mm on average, and the standard deviation was 125.3 mm. The model prediction accuracy was an error average of 10.0 mm and a standard deviation of 25.8 mm.

Next, for the target material whose warpage should be controlled, the tip warpage amount was determined using the relational expression for obtaining the tip warpage amount after the extraction of the heating furnace 10 and before the start of rough rolling. Then, the number of descaling times was determined and rolling was performed so that the amount of tip warpage was closest to the control target value. The number of rough rolling passes was 7. An example thereof is shown in FIG. The warp control target value was set to 0 mm. The allowable warp threshold was set to a warp control target value of ± 5 mm. When a total of 400 coils were rolled by applying the technique of the present invention, the amount of tip warpage of the first stand (F1) was 10.4 mm on average and 26.3 mm with standard deviation. In addition, there was no defective biting into the next stand or damage to the equipment around the rolling mill.

Further, as a comparative example, for materials of the same size and steel grade, the amount of tip warpage is obtained in advance using the vertical set temperature difference and the slab residence time in each of the pre-tropical, heating zone and average tropics of the heating furnace 10 as explanatory variables. A relational expression was created, and the vertical temperature difference between the pre-tropical heating furnace 10 and the heating furnace 10 and the average tropical zone was set so that the tip warpage amount became the target value, and rolling was performed on a total of 250 coils. However, the amount of tip warpage of the first stand (F1) was 20.5 mm on average and 70.2 mm with a standard deviation. In addition, poor engagement with the next stand occurred in the 6 coils. Further, another rolled material 8 rolled at the same rolling timing also caused a large warp, resulting in poor engagement with the next stand.
As described above, the amount of tip warpage has been significantly reduced by applying the hot rolling method using the warp prediction according to the present invention. The amount of tip warpage of finish rolling can be controlled stably at a low level, and troubles such as poor biting of the next stand and equipment damage can be prevented.

(Example 2)
An example of the present invention and a conventional example were compared in a rolling cycle made of carbon steel having a finished plate thickness of 22 mm and a finished plate width of 1600 mm.
First, the tip warp amount H of the first stand (F1) for 600 coils of the target material was measured using the area camera 18. Regarding the definition of the tip warp amount H, the horizontal distance L from the tip of the steel plate in FIG. 5 was set to 3 m. The relational expression for obtaining the tip warp amount H uses a neural network, and the operating parameters of the steelmaking process 5 are the slab casting speed, and the operating parameters of the heating process 1 are the slab temperature at the time of charging the heating furnace 10 and the prediction of the heating furnace 10. In-core temperature of each of tropical, heating zone and uniform tropical, staying time of slab in each of pre-tropical, heating zone and uniform tropical, slab width reduction amount as operation parameter of width reduction process 2, rolling as operation parameter of rough rolling process 3. The number of passes and the number of descaling were created as explanatory variables. Here, as the condition of the neural network, the intermediate layer is set to 3 layers, and the number of nodes is set to 5 each. The sigmoid function was used as the activation function. The operation data of the coils of the above dimensions and steel grade was prepared for 500 coils, 450 coils were used for model learning, and the model prediction accuracy was verified with the remaining 50 coils. The amount of tip warpage for 500 coils was 65.3 mm on average, and the standard deviation was 151.0 mm. The model prediction accuracy was an error average of 12.0 mm and a standard deviation of 18.8 mm.

Next, for the target material whose warpage should be controlled, the tip warpage amount was determined using the relational expression for obtaining the tip warpage amount after the end of width reduction and before the start of rough rolling. Then, the number of descaling times was determined and rolling was performed so that the tip warpage amount was closest to the control target value. The control target value was set to 0 mm, and the warp tolerance threshold was set to the warp control target value of ± 5 mm. When a total of 400 coils were rolled by applying this technique, the amount of tip warpage of the first stand (F1) was 15.0 mm on average and 28.7 mm with standard deviation. There was no failure to bite into the next stand or damage to the equipment around the rolling mill.

Further, as a comparative example, the tip warp amount is obtained in advance for materials of the same size and steel grade, and the vertical set temperature difference and the slab residence time in each of the pre-tropical, heating zone and average tropics of the heating furnace 10 are obtained as explanatory variables. A relational expression was created, and the vertical temperature difference between the pre-tropical heating furnace 10 and the heating furnace 10 and the soaking tropics was set so that the tip warpage amount became the target value, and rolling was performed on a total of 300 coils. However, poor engagement with the next stand occurred in the 5 coils. The tip warp amount of the first stand (F1) was 44.5 mm on average and the standard deviation was 92.5 mm.
As shown in the results of Examples 1 and 2 above, the amount of tip warpage was significantly reduced by applying the hot rolling method based on the present invention. It was found that the amount of tip warpage of finish rolling can be controlled stably at a low level, and troubles such as poor biting of the next stand and equipment damage can be prevented.

(Example 3)
Further, for a material made of carbon steel having a finishing plate thickness of 18 mm and a finishing plate width of 1500 mm, the relationship between the number of rough passes in rough rolling and the amount of tip warpage is obtained when finishing rolling is performed in the hot rolling equipment shown in Table 1. Therefore, the number of rough passes was set to 6 to 13 times, and the amount of tip warpage was determined.
The result is shown in FIG.
As can be seen from FIG. 8, it was confirmed that there is a strong correlation between the number of coarse passes and the amount of tip warpage.

1 Heating process 2 Width reduction process 3 Rough rolling process 4 Finish rolling process 5 Steelmaking process 8 Rolled material (steel plate)
10 Heating furnace 11 Width reduction device 13 Rough rolling machine 14 Descaling device 15 Rough rolling equipment 16 Finishing rolling machine 20 Warp control unit 21 Warp prediction unit 21A Warp prediction model 22 Warp suppression unit 22A Operating condition resetting unit 22B Warp re-evaluation unit 22C end judgment unit

Claims (11)

The finish rolling step in the hot rolling step including a heating step of heating the slab, a rough rolling step of rough rolling the slab after heating in the heating step, and a finish rolling step of finishing rolling the rolled material after rough rolling. It is a method of predicting the warp of the tip of the rolled material in
Warpage learned by machine learning that includes one or more parameters selected from the operation parameters of the rough rolling process as input data and uses warp information related to the warp of the tip of the rolled material in the finish rolling process as output data. A method for predicting warpage in a hot rolling process, which comprises predicting the warp of the tip of a rolled material in the finishing rolling process using a prediction model. The input data of the warp prediction model is characterized by having one or two or more parameters selected from the operation parameters of the heating process and one or two or more parameters selected from the operation parameters of the rough rolling process. The method for predicting warpage in the hot rolling process according to claim 1. The hot rolling according to claim 1 or 2, wherein the operation parameters of the rough rolling process constituting the input data include at least one parameter of the number of rough rolling passes and the number of descaling in rough rolling. Warp prediction method in the process. The input data has one or more parameters selected from the operation parameters of the heating process, and the operation parameters of the heating process constituting the input data include the slab temperature at the time of charging the heating furnace and the prediction of the heating furnace. Warpage in the heat spreading process according to any one of claims 1 to 3, wherein one or more parameters selected from tropical, heating zone, and even tropical furnace temperatures are included. Prediction method. The amount of warpage of the rolled material in the finish rolling process is predicted by the method for predicting warpage according to any one of claims 1 to 4.
A method for controlling warpage in a hot rolling process, which comprises resetting one or more parameters selected from the operation parameters of the rough rolling process so that the predicted amount of warpage becomes small. The warp control method in the hot rolling process according to claim 5, wherein the parameter to be reset is at least one of the number of rough rolling passes and the number of descaling in the rough rolling process. A method for manufacturing a hot-rolled steel sheet using the warp control method according to claim 5 or 6. The finish rolling in the hot rolling process including a heating step of heating the slab, a rough rolling step of rough rolling the slab after heating in the heating step, and a finish rolling step of finishing rolling the rolled material after the rough rolling. A method for generating a warp prediction model used to predict the warp of the tip of a rolled material in a process.
At least one or two or more operation record data selected from the operation record data in the heating process and the operation record data in the rough rolling process are used as input record data, and the above in the hot rolling process using the input record data. Acquire multiple learning data using the actual data of the amount of warpage of the tip of the rolled material in the finish rolling process as the output actual data.
A method for generating a warp prediction model, which comprises generating a warp prediction model by machine learning using a plurality of acquired learning data. The method for generating a warp prediction model according to claim 8, wherein machine learning selected from a neural network, decision tree learning, random forest, and support vector regression is used as the machine learning for generating the warp prediction model. It has a heating furnace for heating slabs, a rough rolling machine for rough rolling slabs, a rough rolling facility having a descaling facility for descaling the slab, and a finish rolling facility for finish rolling the rolled material after rough rolling. It is a hot rolling facility
It is equipped with a warp prediction unit that predicts the warp information of the tip of the rolled material in the finish rolling equipment using the warp prediction model.
The warp prediction model is a model learned by machine learning, includes one or more parameters selected from the operation parameters of the rough rolling equipment as input data, and includes the tip of the rolled material in the finishing rolling equipment. A hot-rolling facility characterized by being a learning model that uses warp information related to warp as output data. The input data of the warp prediction model is characterized by having one or two or more parameters selected from the operating parameters of the heating furnace and one or two or more parameters selected from the operating parameters of the rough rolling equipment. The hot rolling equipment according to claim 10.

Images