JP7294242B2 - Method for predicting surface roughness, method for manufacturing steel strip, and method for generating trained machine learning model - Google Patents

Description

The present invention provides a method for predicting the surface roughness of a steel strip after a skin pass rolling process, a method for manufacturing a steel strip, and a method for generating a learned machine learning model. Regarding.

The manufacturing process for thin steel sheets consists of a cold rolling process in which surface oxides generated by hot rolling are removed in a pickling process, and then the steel strip is cold rolled to a predetermined target thickness. It has an annealing process for softening the hardened steel strip and a temper rolling process for adjusting the mechanical properties (eg, yield point elongation, yield stress, tensile strength, elongation, etc.) of the steel strip to be a product. A continuous cold rolling mill with 4 to 6 stands is used for the cold rolling process. However, in some cases, the thickness is reduced to a predetermined thickness by reverse rolling using a single-stand rolling mill.

The annealing step is a step of continuously heat-treating the steel strip in a continuous annealing line (CAL). The temper rolling step is a step of applying a light reduction with an elongation rate of about 1%, for example, by a temper rolling mill. The main purpose of the temper rolling process is to adjust the mechanical properties of the steel strip. Surface roughness transfer forms the surface roughness of the steel strip. The surface roughness of a steel strip affects not only the quality of the appearance of the steel strip, but also the quality of the characteristics of the steel strip, such as press formability and sharpness after painting. a useful operational indicator.

The temper rolling process is often arranged on the delivery side of the continuous annealing line. It is common to have equipment arrangements in which the quality rolling process is carried out. However, in the continuous annealing line, only the annealing process for heat-treating the steel strip is arranged, and after the annealing process, the steel strip is once wound into a coil and then temper-rolled in an independent line for batch-type temper rolling. .

Average roughness Ra (arithmetic mean roughness defined in JIS B 0601-1994) is generally used as a parameter representing the surface roughness of a steel strip. For example, steel strips for automobiles are often subjected to temper rolling with a target surface roughness Ra set in the range of about 0.5 μm to 2.0 μm. In addition, the surface roughness of the steel strip is required to have small variation between coils and small variation within the coil.

On the other hand, in Patent Document 1, the average roughness Ra and peak count PPI (parameters specified by the SAE911 standard and the number of uneven peaks in a measurement length of 1 inch) on the surface of the steel strip after temper rolling , values at a count level of ±0.63 μm), a regression equation is prepared in advance using the elongation rate, entry-side tension, exit-side tension, and coefficient of friction obtained from the experiment as input data, and the regression equation is used. It is disclosed that the predicted values of the average roughness Ra and the peak count PPI can be calculated. It also discloses changing the parameters selected from the elongation rate, entry-side tension, and exit-side tension so that Ra and PPI fall within the target ranges.

In Patent Document 2, in order to represent the surface roughness of the steel strip after temper rolling, a parameter called transfer efficiency is defined using the surface roughness of the rolling rolls and the surface roughness of the steel strip before temper rolling. Then, the transfer efficiency is expressed as a function of the steel strip thickness, 0.2% proof stress and elongation rate, and a method is disclosed in which the function is determined in advance by experiments.

JP-A-2008-6453 JP 2009-142879 A JP-A-60-201204

However, neither Patent Documents 1 nor 2 discloses predicting the surface roughness of a steel strip after temper rolling in consideration of the influence of the annealing process. The thermal history in the steel strip annealing process affects the transfer efficiency of the microscopic unevenness of the rolling rolls in the subsequent temper rolling process. Therefore, the conventional method of predicting the surface roughness of the steel strip without considering the influence of the annealing process has a problem that the surface roughness of the steel strip after the temper rolling process cannot be predicted with high accuracy. The present invention was completed in view of the above problems, and its object is to improve the surface roughness of a steel strip by considering the influence of the annealing process that affects the surface roughness of the steel strip after the temper rolling process. To provide a surface roughness prediction method capable of predicting the

Means for solving the above problems are as follows.
(1) A method for predicting the surface roughness of a steel strip after the temper rolling step manufactured in a manufacturing process including an annealing step and a temper rolling step, comprising at least one operating parameter of the annealing step; and at least one of the operating parameters of the skin pass rolling process are obtained as input data of the steel strip for predicting the surface roughness after the skin pass rolling process, and the surface roughness of the steel strip after the skin pass rolling process is output. A surface roughness prediction method, comprising: inputting the input data into a machine learning model as data to predict the surface roughness of the steel strip after the skin pass rolling process.
(2) The method of predicting surface roughness according to (1), wherein at least one attribute parameter relating to the chemical composition of the steel strip for predicting the surface roughness after the temper rolling step is further acquired as the input data. .
(3) The manufacturing process further includes a cold rolling process provided upstream of the annealing process, and further acquires at least one operational parameter of the cold rolling process as the input data, (1) Or the prediction method of the surface roughness as described in (2).
(4) predicting the surface roughness of the steel strip after the temper rolling process by changing the operating parameters of the temper rolling process, and performing the temper rolling process in which the predicted surface roughness is within the target range; The method for predicting surface roughness according to any one of (1) to (3), wherein operating parameters are specified.
(5) A method for manufacturing a steel strip, wherein the steel strip is manufactured in a manufacturing process including an annealing step and a skin pass rolling step in which the operating parameters specified by the surface roughness prediction method described in (4) are set. .
(6) A method for generating a learned machine learning model for outputting the surface roughness of a steel strip after the skin pass rolling process, which is manufactured in a manufacturing process including an annealing process and a skin pass rolling process, the method comprising: , at least one of the operating parameters of the skin pass rolling step, and the surface roughness of the steel strip after the skin pass rolling step as teacher data, after the skin pass rolling step A method for generating a trained machine learning model that generates a trained machine learning model that outputs surface roughness in .
(7) The method of generating a trained machine learning model according to (6), wherein the machine learning model is any one of neural network, decision tree learning, random forest, and support vector regression.

In the surface roughness prediction method according to the present invention, the surface roughness of the steel strip after the skin pass rolling process is predicted using a machine learning model that considers not only the effect of the skin pass rolling process but also the annealing process. Therefore, the surface roughness of the steel strip after the tempering process can be predicted with higher accuracy than the conventional method that does not consider the operational parameters of the annealing process. In addition, using the prediction method, the operating parameters of the skin pass rolling process are specified so that the surface roughness of the steel strip is within the target range, and the skin pass rolling process is performed with the specified operating parameters, so that the surface quality It is possible to manufacture a steel strip excellent in

BRIEF DESCRIPTION OF THE DRAWINGS It is a schematic diagram which shows the equipment example of the cold-rolling equipment 10 which can apply the manufacturing method of the steel strip which concerns on this embodiment. BRIEF DESCRIPTION OF THE DRAWINGS It is a schematic diagram which shows the equipment example of the continuous annealing equipment 40 to which the manufacturing method of the steel strip which concerns on this embodiment can be applied. 4 is a graph showing the thermal history of a steel strip in an annealing process; 1 is a schematic diagram showing an example of temper rolling equipment 60. FIG. 3 is a functional block diagram of a surface roughness prediction device 76; FIG. FIG. 7 is a flow diagram for explaining a process of specifying operation parameters for a temper rolling step by a surface roughness calculator 78;

In the method for predicting surface roughness according to the present embodiment, a steel sheet whose thickness has been reduced to a predetermined thickness by a cold rolling process is heat-treated in an annealing process, and then temper-rolled in a temper rolling process. Predict the surface roughness of thin steel sheets after temper rolling. At least after the cold rolling process, the thin steel sheet is coiled and then subjected to heat treatment and the like.

First, the manufacturing process of the steel strip 11 will be described. A slab cast in a steelmaking process is rough-rolled and finish-rolled in a hot-rolling process to reduce its thickness, and wound into a coil. The coiled steel strip 11 is cooled to near room temperature.

FIG. 1 is a schematic diagram showing an example of cold rolling equipment 10 to which the steel strip manufacturing method according to the present embodiment can be applied. The cold rolling equipment 10 shown in FIG. 1 is a pickling-tandem continuous pickling equipment in which a pickling equipment 20 composed of a pickling tank 22 and a rinsing tank 24 and a continuous cold rolling mill 32 are arranged on the same line. Cold rolling equipment 10 of the line.

A steel strip 11 that has been hot rolled and cooled to near room temperature is passed through a payoff reel 12, a welding machine 14, a No. A pickling process is performed by a pickling facility 20 having a pickling tank 22 and a rinsing tank 24 through a looper 16 and a tension leveler 18 to remove surface scales. After that, the steel strip 11 is no. 2 looper 26, trimmer 28, no. A cold rolling process is performed by a continuous cold rolling mill 32 through three loopers 30 . The continuous cold rolling mill 32 is, for example, a tandem rolling mill in which 5 to 6 rolling mills are continuously arranged. The tandem rolling mill produces coils with a thickness of about 0.5 to 2.4 mm, such as steel sheets for automobiles, and thin steel sheets with a thickness of about 0.1 to 0.5 mm, such as steel sheets for cans. Although it is classified as a tinplate mill, any mill can be used in the steel strip manufacturing method according to the present embodiment. The steel strip 11 cold-rolled in the cold-rolling process is coiled by a tension reel 34 and then sent to a continuous annealing facility 40 .

FIG. 2 is a schematic diagram showing an example of continuous annealing equipment 40 to which the steel strip manufacturing method according to the present embodiment can be applied. The continuous annealing equipment 40 is roughly divided into an entry side equipment 41 , a furnace body portion 50 and a delivery side equipment 58 . The entry side equipment 41 has a payoff reel 42 , a welder 44 , an electrolytic cleaner 46 , a tension leveler 48 and an entry side looper 49 . The furnace section 50 has a heating zone 53 , a soaking zone 54 and a cooling zone 55 . The annealing process in the present embodiment is a heat treatment process performed in the furnace section 50 . The delivery side equipment 58 has a delivery side looper 59 , a temper rolling equipment 60 , a side trimmer 70 and a tension reel 72 . The temper rolling process of the present embodiment is performed by a temper rolling facility 60 including incidental equipment that applies tension on the entry and exit sides of the temper rolling mill.

Continuous annealing equipment 40 further includes a process computer 74 and a surface roughness predictor 76 . The process computer 74 is connected to each device constituting the continuous annealing facility 40 and controls the annealing process and temper rolling process of the steel strip 11 . The process computer 74 is connected to a surface roughness prediction device 76, and transmits and receives data necessary for predicting the surface roughness of the steel strip 11 after the skin pass rolling process. The surface roughness prediction device 76 predicts the surface roughness of the steel strip 11 after the skin pass rolling process and specifies the operation parameters of the skin pass rolling process.

The annealing step is a step of raising the temperature of the steel strip 11 from around room temperature, maintaining it at a predetermined temperature, and then lowering the temperature to around room temperature. In the continuous annealing equipment 40 shown in FIG. 2, the annealing process is performed by a heating zone 53, a soaking zone 54 and a cooling zone 55. However, an overaging zone may be provided downstream of the cooling zone 55, and a reheating zone may be provided upstream of the overaging zone, and a final cooling zone may be provided downstream of the overaging zone. The overaging zone is arranged downstream of the cooling zone 55. After cooling the steel strip 11 to a predetermined temperature in the cooling zone 55, the steel strip 11 is reheated in a reheating zone using an induction heating device. This facility performs overaging treatment at a temperature of ~400°C. The annealing step includes a heat treatment step with a reheating zone, an overaging zone and a final cooling zone, if provided.

The heating zone 53 is equipment for raising the temperature of the steel strip 11, and heats the steel strip 11 to a preset temperature in the range of about 600 to 900° C. depending on the steel type. In the heating zone 53, direct fire or radiant combustion burners are used. Since these heating devices have a large heating capacity and a relatively quick response, it is easy to change the temperature rise history when changing the heat cycle. The soaking zone 54 is a facility for maintaining the steel strip 11 at a predetermined temperature, and has a heating capacity sufficient to compensate for the radiation heat of the furnace body. For this reason, the soaking zone 54 has a characteristic of having a long response time when changing the annealing temperature.

The cooling zone 55 is equipment for cooling the steel strip 11 to a predetermined temperature, and gas jet cooling, roll cooling, water cooling (water quench), or the like is used as cooling means. Gas jet cooling is a cooling means for blowing gas onto the surface of the steel strip 11 from a nozzle. Roll cooling is cooling means for cooling the steel strip 11 by bringing it into contact with a water-cooled roll. Water cooling is cooling means for cooling the steel strip 11 by immersing it in a water cooling tank installed downstream of the soaking zone 54 . Since the cooling rate of the steel strip 11 by these cooling devices is different, the cooling zone 55 is divided into a plurality of cooling zones such as a first cooling zone and a second cooling zone, and different cooling means are combined or the same type of cooling means is used. The conditions are changed to control the heat history during cooling of the steel strip 11 .

Thermometers for measuring the surface temperature of the steel strip 11 are installed at a plurality of positions in the heating zone 53 , the soaking zone 54 and the cooling zone 55 . In particular, in the cooling zone 55 where the temperature change of the steel strip 11 is large, thermometers are installed on the entrance side and the exit side of the cooling zone 55, and the surface temperature of the steel strip 11 at the position is measured to measure the cooling of the cooling zone 55. Actual speed values are calculated. As the thermometer, a radiation thermometer is used that continuously measures the surface temperature of the width center of the steel strip. However, the thermometer is not limited to the radiation thermometer, and a profile radiation thermometer that measures the temperature distribution in the sheet width direction may be used. Furnace thermometers are also installed to measure not only the surface temperature of the steel strip 11 but also the ambient temperature inside the furnace in each zone of the annealing process. The measured surface temperature of the steel strip 11 and ambient temperature are output to the process computer 74 .

FIG. 3 is a graph showing the thermal history of the steel strip in the annealing process. FIG. 3(a) is a graph showing the thermal history of the steel strip 11 subjected to the annealing process in the heating zone 53, soaking zone 54 and cooling zone 55 (first cooling zone and second cooling zone). FIG. 3(b) is a graph showing the thermal history of the steel strip 11 subjected to the annealing process in the heating zone 53, the soaking zone 54, the cooling zone 55, the reheating zone 90, the overaging zone 91 and the final cooling zone 92. be. In both FIGS. 3(a) and 3(b), the vertical axis is the steel strip temperature (° C.) and the horizontal axis is the time (min). In order to prevent variations in the quality of the steel strip 11 depending on the longitudinal position of the steel strip 11, the conveying speed of the steel strip 11 is kept constant during the annealing process. However, when steel strips 11 having different thicknesses, widths, steel grades, etc. are welded, the line speed may change before and after the welded portion. For this reason, the thermal history shown in FIG. 3 may not be constant in the longitudinal direction of the steel strip 11 .

FIG. 4 is a schematic diagram showing an example of the temper rolling facility 60. As shown in FIG. The temper rolling facility 60 is arranged downstream of the furnace body 50 . The temper rolling equipment 60 includes a temper rolling mill 63, an entry-side bridle roll 61, two tension gauges 62, an exit-side bridle roll 64, a load cell 66, a surface roughness meter 67, and a mill speed controller 68. and a bridle roll speed controller 69 . The entry-side bridle roll 61 and the exit-side bridle roll 64 are rolls that apply tension to the steel strip 11 . The tension gauges 62 are provided between the temper rolling mill 63 and the entry-side bridle roll 61 and between the temper rolling mill 63 and the exit-side bridle roll 64, respectively. The tension of the steel strip 11 is measured. A temper rolling process is performed by a temper rolling mill 63 , an entry-side bridle roll 61 and an exit-side bridle roll 64 .

As the temper rolling mill 63, a 4-high or 6-high rolling mill is usually used. Although FIG. 4 shows an example configured with a single-stand rolling mill, the temper rolling mill 63 may be configured with a two-stand rolling mill. In the case of using a two-stand rolling mill, the downstream rolling mill is targeted for surface roughness prediction by the surface roughness prediction method according to the present embodiment. This is because the final surface roughness of the product is imparted by the downstream rolling mill.

A load cell 66 measures the rolling load. A mill speed controller 68 also controls the peripheral speed of the work rolls of the temper rolling mill 63 . A bridle roll speed controller 69 controls the circumferential speeds of the entry-side and exit-side bridle rolls 64, respectively. The temper rolling process includes a dry skin skin rolling process that does not use a pressure conditioning liquid and a wet skin skin rolling process that performs skin skin rolling while supplying a water-based pressure conditioning liquid. can be applied to both dry and wet temper rolling processes.

In the temper rolling process, the steel strip 11 after the annealing process is rolled at a predetermined elongation rate. Here, the elongation rate is the elongation rate of the steel strip 11 before and after rolling, and is defined by the increase rate of the length of the steel strip 11 before and after the temper rolling process. In addition, under the condition that the elongation rate is small, the rolling reduction rate, which is the change rate of the sheet thickness, is approximately the same. Therefore, the reduction rate may be used instead of the elongation rate. The elongation rate is measured by the peripheral speed difference between the entry-side bridle roll 61 and the exit-side bridle roll 64 .

The elongation rate of the temper rolling process is controlled as follows, for example. A peripheral speed difference is given between the entry-side bridle roll 61 and the exit-side bridle roll 64 in accordance with the set target value of the elongation rate. The roll gap and roll peripheral speed of the temper rolling mill 63 are adjusted so that the tension measured by the tension gauges 62 before and after the temper rolling mill 63 falls within the target range. Specifically, while adjusting the roll peripheral speed so that the entry-side tension of the temper mill 63 becomes the target value, the roll gap is adjusted so that the exit-side tension of the temper mill 63 becomes the target value. However, not limited to the control method described above, for example, while adjusting the tension on the entry side and exit side of the temper rolling mill 63 by the peripheral speed of the rolls of the temper rolling mill 63 and the entry side bridle roll 61, It may be controlled to obtain the target elongation rate by adjusting the roll gap.

The upper limit and lower limit of the elongation rate of the steel strip 11 by the temper rolling mill 63 are determined in advance according to the steel type, size, etc. so as to satisfy the mechanical properties of the steel strip 11 . For example, a typical elongation rate range is 0.2% or more and 2.0% or less. If the elongation is 0.2% or less, the yield point elongation may remain even after temper rolling, and if it is 2.0% or more, work hardening progresses and ductility, which is a mechanical property of the product, decreases.

Next, the rolling rolls of the temper rolling mill 63 will be described. The rolling roll has an average surface roughness Ra of 0.5 to 10 μm. The roughness of the roll surface is finished by dulling, and as specific means, a shot blasting method, an electric discharge dulling method, a laser dulling method, an electron beam dulling method, or the like is used. Moreover, after dulling, the surface may be plated with chrome.

A value measured at a representative position on the surface of the work roll is used as the average surface roughness Ra of the rolling roll. Also, an average value of Ra measured at a plurality of positions on the roll surface may be used. When using the average value of a plurality of positions, for example, at least in the portion of the rolling roll that contacts the steel strip 11, 4 points in the circumferential direction at intervals of 90 °, 3 points in the center and both ends in the width direction, a total of 12 points It is preferable to use the average value of The average surface roughness Ra is measured with an evaluation length of 4 mm and a cutoff of 0.8 mm.

Next, the surface roughness of the steel strip 11 will be explained. In the present embodiment, the surface roughness of the steel strip 11 after temper rolling is the surface roughness of the steel strip 11 after temper rolling by the temper rolling mill 63 . The surface roughness of the steel strip 11 may be measured at any position in the production line after being skin-pass rolled by the skin-pass mill 63, for example, the skin-pass mill 63 and the delivery side bridle roll. 64 and may be measured at further downstream locations. Further, when the temper rolling equipment 60 is incorporated in a continuous annealing line, the surface roughness of the steel strip 11 may be measured before being wound by the tension reel 72, or may be measured separately after being wound once. It may be measured in process.

The surface roughness of steel strip 11 is preferably measured online by surface roughness meter 67 . As a result, the measured value of the surface roughness of the steel strip 11 after temper rolling can be immediately obtained, and the fluctuation of the surface roughness of the steel strip 11 in the longitudinal direction can also be grasped. As the online surface roughness meter 67, for example, the surface roughness meter disclosed in Patent Document 5 can be used. The measured surface roughness of the steel strip 11 is output to the process computer 74 .

In this embodiment, the average surface roughness Ra of the steel strip 11 is used as the surface roughness of the steel strip 11 . However, the surface roughness of the steel strip 11 is not limited to the average roughness Ra, and peak count PPI, maximum height Rmax, 10-point average height Rz, etc. may be used instead of the average roughness Ra.

The surface roughness of the steel strip 11 does not necessarily have to be measured by the online surface roughness meter 67. For example, a sample is taken from the steel strip 11 after temper rolling and measured by an offline surface roughness meter. surface roughness may be used. It is difficult to measure the surface roughness off-line continuously in the longitudinal direction of the steel strip 11 . For this reason, about 1 to 3 samples are taken from different positions in the longitudinal direction of the steel strip 11 to measure the surface roughness. Off-line measurements provide detailed surface roughness not available online. For example, the arithmetic mean height Sa, the root mean square height Sq, the maximum height Sz, etc., which are the surface roughness defined in ISO 25178, may be used by three-dimensional roughness measurement. A two-dimensional map (image) in which surface unevenness obtained as a result of three-dimensional roughness measurement is displayed by contour lines or color coding may be used as data indicating the surface roughness of the steel strip 11 . When surface roughness is measured off-line, the measured value is input to process computer 74 by operator 88 .

The process computer 74 is connected to each facility constituting the continuous annealing facility 40, controls the operation of each facility, sets the operating conditions of each facility, and receives measurement data obtained from measuring equipment provided in each facility. It has the function of collecting. Each piece of equipment may be provided with a separate control device, such as a mill speed controller 68 or a bridle roll speed controller 69, in which case the process computer 74 provides control of each piece of equipment via the controller. Set operating conditions and collect measurement data.

The process computer 74 issues commands to control the operation of each piece of equipment that makes up the continuous annealing equipment 40 . The process computer 74 also stores attribute parameters relating to the chemical composition of the steel strip 11 manufactured by the facility in the past, operational parameters that are the operating conditions of each facility of the continuous annealing facility 40, and Measurements taken by the measuring instrument are collected and stored. Furthermore, the process computer 74 also stores attribute parameters relating to the chemical composition of the steel strip 11 manufactured by the continuous annealing equipment 40 and operation parameters of each equipment of the continuous annealing equipment 40 . The surface roughness of the steel strip 11 after the temper rolling process is output to the process computer 74 and stored in the process computer 74 as soon as it is measured.

FIG. 5 is a functional block diagram of the surface roughness prediction device 76. As shown in FIG. The surface roughness prediction device 76 predicts the surface roughness of the steel strip 11 after the temper rolling process using the input data including the operational parameters of the annealing process and the operational parameters of the temper rolling process. Furthermore, the surface roughness predicting device 76 specifies the operating parameters of the skin pass rolling process in which the predicted surface roughness of the steel strip 11 is within the target range.

The surface roughness prediction device 76 is, for example, a general-purpose computer such as a workstation or personal computer. The surface roughness prediction device 76 has a surface roughness calculation section 78 , a storage section 80 and an input section 86 . The surface roughness calculation unit 78 is, for example, a CPU or the like, and uses programs and data stored in the storage unit 80 to execute predetermined calculations in order to realize the functions of the surface roughness prediction device 76. . Specifically, the surface roughness calculator 78 uses the learned machine learning model stored in the storage unit 80 to calculate the surface roughness of the steel strip 11 after the skin pass rolling process. Furthermore, the surface roughness calculator 78 specifies operation parameters for the skin pass rolling process that makes the predicted surface roughness after the skin pass rolling process within the target range, and outputs the operation parameters to the process computer 74 .

The storage unit 80 is, for example, a flash memory capable of updating and recording, a hard disk built in or connected via a data communication terminal, an information recording medium for a memory card, and a read/write device therefor. The storage unit 80 stores in advance a program for realizing the functions of the surface roughness prediction device 76, data used during execution of the program, and the like. Specifically, in the storage unit 80, a database 84 in which a plurality of data sets each including the operating parameters of the annealing process, the operating parameters of the temper rolling process, and the surface roughness of the steel strip 11 after the temper rolling process are stored as one set is stored. and a learned machine learning model 82 having input data of the operating parameters of the annealing process and the operating parameters of the skin pass rolling process and output data of the surface roughness of the steel strip 11 after the skin pass rolling process. there is In addition to the above, the database 84 may store attribute parameters relating to the chemical composition of the steel strip 11 and operation parameters of the cold rolling process. In this case, the input data of the learned machine learning model 82 are the operational parameters of the annealing process, the operational parameters of the temper rolling process, the attribute parameters related to the chemical composition of the steel strip 11, and the operational parameters of the cold rolling process.

The input unit 86 is, for example, an input device such as a keyboard and a mouse, or a reading device for an information recording medium for a memory card. The operational parameters of the annealing process, the operational parameters of the temper rolling process, and the surface roughness of the steel strip 11 after the temper rolling process, which are stored in the database 84, are output from the process computer 74 to the surface roughness calculator 78. , the operator 88 may input these values from the input unit 86 to the surface roughness calculator 78 .

Next, a method of generating a learned machine learning model used for predicting the surface roughness of the steel strip 11 after the temper rolling process will be described. The surface roughness calculator 78 acquires the operating parameters of the annealing process, the operating parameters of the skin pass rolling process, and the surface roughness of the steel strip 11 after the skin pass rolling process from the process computer 74, and uses these as one set of data. The set is stored in database 84 of storage unit 80 . The number of data sets stored in the database 84 is at least 100 or more, preferably 500 or more, and more preferably 1000 or more.

The surface roughness calculation unit 78 reads out a pre-stored machine learning model from the storage unit 80, performs machine learning on the machine learning model using the data set stored in the database 84 as teacher data, and converts the learned machine learning Generate a model. The machine learning model used in the surface roughness prediction method according to this embodiment is, for example, a neural network. However, the machine learning model is not limited to the neural network, and any one of decision tree learning, random forest, support vector regression, Gaussian process, and k nearest neighbor method may be used.

In the surface roughness prediction method according to the present embodiment, not only the operation parameters of the skin pass rolling process but also the operation parameters of the annealing process are used as teacher data for machine learning of the machine learning model, so that the annealing process The effect on the rolling process is reflected. As a result, the learned machine learning model that has undergone machine learning using the teacher data becomes a learned machine learning model that takes into consideration not only the factors that directly affect the temper rolling process but also the influence of the annealing process.

The longer the time spent passing through the heating zone 53 and the soaking zone 54 in the annealing process, the easier it is for the steel strip 11 to soften. Therefore, if the time to pass through the heating zone 53 and the soaking zone 54 is long, the microscopic unevenness of the rolling rolls is likely to be transferred to the surface of the steel strip 11 in the temper rolling process, and the microscopic unevenness of the rolling rolls is likely to be transferred. transfer efficiency is increased.

In addition, the longer the time passed through the heating zone 53 and the soaking zone 54, the more the elements contained in the steel strip 11 are concentrated on the surface, and the thickness and composition of the oxides formed on the surface of the steel strip 11 change. do. This change affects the coefficient of friction between the rolling rolls and the steel strip 11 in temper rolling, thereby changing the transfer efficiency of the microscopic unevenness of the rolling rolls. Also, regarding the soaking temperature of the soaking zone 54, the difference in the soaking temperature affects the deformation resistance and friction coefficient of the steel strip 11, so that the transfer efficiency of the microscopic unevenness of the rolling rolls changes. do.

Also, the cooling conditions such as the cooling rate in the cooling zone 55 in the annealing process affect the transformation behavior of the steel strip 11, and thereby the deformation resistance of the steel strip 11 after cooling changes. Furthermore, the temperature history in the cooling zone 55 affects the oxidation state of the surface of the steel strip 11, thereby changing the friction coefficient of temper rolling. Therefore, the difference in the cooling conditions of the cooling zone 55 also affects the transfer behavior of the microscopic unevenness of the rolling rolls.

Thus, the operating conditions of the heating zone 53, the soaking zone 54 and the cooling zone 55 in the annealing process affect the transfer efficiency of the microscopic unevenness of the rolling rolls. The prediction accuracy of the prediction method of the surface roughness of the steel strip 11 after the temper rolling process according to the present embodiment using the machine learning model is the prediction accuracy of the surface roughness of the steel strip 11 that does not consider the influence of the annealing process. higher than the predictive accuracy of the method.

As described above, the method for predicting the surface roughness of the steel strip 11 after the skin pass rolling process according to the present embodiment is based on the method of predicting the surface roughness of the steel strip 11 after the skin pass rolling process. Since it is a method for predicting the surface roughness of, the operating parameters for skin pass rolling specified using the prediction method are the operating parameters for skin pass rolling that take into account the deformation resistance of the steel strip 11 and the variation in the coefficient of friction. Become. Therefore, by performing the skin pass rolling process using the specified operating parameters for skin pass rolling, the variation in transfer behavior of the microscopic unevenness of the rolling rolls due to variations in deformation resistance and friction coefficient is suppressed. Therefore, it is possible to manufacture the steel strip 11 having excellent surface quality with small variations in surface roughness.

Next, data used as teacher data will be described. In the surface roughness prediction method according to the present embodiment, the operating parameters of the annealing process, the operating parameters of the skin pass rolling process, and the steel strip after the skin pass rolling process are used as teacher data for generating a learned machine learning model. A surface roughness of 11 is used.

<Operating parameters of the annealing process>
Using the example of the thermal history of the steel strip in the annealing process shown in FIG. 3, the following operating parameters for the annealing process can be used. For example, in the example shown in FIG. 3( a ), the time for the steel strip 11 to pass through the heating zone 53 and the amount of temperature rise may be used as the operating parameters of the heating zone 53 , or calculated from these values. may be used.

As the operation parameters of the soaking zone 54, the soaking temperature, which is the average temperature of the steel strip in the soaking zone 54, and the soaking time, which is the time for passing through the soaking zone 54, may be used. As the operating parameters of the cooling zone 55, the time for the steel strip 11 to pass through the first cooling zone and the amount of temperature drop may be used, or an average cooling rate calculated from these values may be used. Moreover, as the operating parameters of the cooling zone 55, the time for the steel strip 11 to pass through the second cooling zone and the amount of temperature drop may be used, or an average cooling rate calculated from these values may be used. However, when the temperature of the steel strip 11 is cooled to room temperature in the middle of the second cooling zone, the temperature difference measured by two or more radiation thermometers installed in the second cooling zone before reaching room temperature It is preferable to use a substantial cooling rate calculated from These values are output from heating zone 53, soaking zone 54 and cooling zone 55 to process computer 74 after measurement.

Also in the example shown in FIG. 3(b), the same operating parameters as above can be used for the heating zone 53, the soaking zone 54 and the cooling zone 55. As the operating parameters of the reheating zone, the passage time and temperature rise of the radiation thermometers placed on the inlet and outlet sides of the induction heating device may be used, or the average heating rate calculated from these values may be used. may be used. As the operating parameters of the overaging zone, the overaging temperature, which is the average temperature of the steel strip 11 in the overaging zone, and the time to pass through the overaging zone may be used. As the operating parameters of the final cooling zone, the time and temperature drop amount for the steel strip 11 to pass through the final cooling zone may be used, or the average cooling rate calculated from these values may be used. These values are also output to the process computer 74 after measurement.

At least one of the operational parameters of the annealing process as described above is used as teaching data. However, it is not limited to these operating parameters, and the control output value of the heating device in the heating zone 53 and the reheating zone 90 and the control output value of the cooling device in the cooling zone 55 and the final cooling zone 92 are also operating parameters. can be used as This is because these operational parameters are operational parameters used to control the temperature history of the steel strip 11 in the annealing process.

The operation parameters of the annealing process are not limited to the above parameters, and the line speed of the steel strip 11 at the entrance of the soaking zone 54, the average cooling speed in the cooling zone 55, and the injection pressure of the cooling device such as gas injection , dew point data in the annealing furnace, etc., and various measured values that are considered to affect the deformation resistance and surface oxide state of the steel strip 11 after the annealing process may be used.

<Operating parameters of temper rolling process>
As the operating parameters for the temper rolling process, the operating parameters used for predicting the surface roughness of the steel strip 11 in the prior art can be used. The operating parameters of the temper rolling process are, for example, the elongation rate, the entry-side tension and exit-side tension of the temper rolling mill, the work roll diameter of the rolling rolls, and the like. Moreover, the deformation resistance, friction coefficient, and rolling load of the steel strip 11 may be used as operating parameters for the temper rolling step. However, since it is difficult to measure the deformation resistance and friction coefficient of the steel strip 11, it is possible to use set values set according to the target steel strip 11 category as the deformation resistance and friction coefficient of the steel strip 11. good. Even if these set values are used, the operation parameters of the annealing process can be used together to correct the values to match the actual situation.

Furthermore, the average roughness Ra of the surface of the rolling rolls may be used as an operating parameter for the temper rolling step. Further, as the operating parameters of the skin pass rolling process, the cumulative rolling length and the cumulative rolling weight of the steel strip 11 subjected to the skin pass rolling may be used together with the average roughness Ra of the surface of the rolling rolls. By using the cumulative rolling length and the cumulative rolling weight as operating parameters, the effect of changes in surface roughness over time due to wear of the rolling rolls is reflected in the trained machine learning model.

<Attribute parameters of steel strip>
It is preferable to use at least one of the attribute parameters of the steel strip 11 in addition to the operational parameters of the annealing process and the temper rolling process as teacher data for the learned machine learning model. As an attribute parameter of the steel strip 11, the content of any element contained in the steel strip 11 can be used. The attribute parameter of the steel strip 11 is, for example, a value representing the content of the component composition of C, Si, Mn, P, S, Al, N, Ti, etc. of the steel strip 11 in mass %. Note that the distribution of these component contents in the longitudinal direction is generally constant, and one attribute parameter is used for one steel strip. In addition, one attribute parameter is also used for steel strip 11 produced from slabs cast from the same charge in the steelmaking process.

As attribute parameters of the steel strip 11, it is preferable to use the contents of C, Si, and Mn. This is because the C content has a large effect on the formation of the structure after the annealing process, affects the deformation resistance, and affects the transfer behavior of the rolling rolls to the surface of the steel strip in the temper rolling process. In addition, the content of Si and Mn affects the formation of the structure in the annealing process and affects the deformation resistance. This is because it affects the thickness and composition of the oxide film formed on the surface of the strip, and the transfer behavior of the rolling rolls to the surface of the steel strip in the skin pass rolling process through the coefficient of friction during skin pass rolling. Furthermore, since the thickness of the steel strip 11 also affects the transfer behavior from the rolling rolls to the steel strip 11 , the thickness may be used as an attribute parameter of the steel strip 11 . When strip 11 attribute parameters are used, these values are stored in process computer 74 and output to surface roughness predictor 76 .

<Operating parameters of the cold rolling process>
In addition to the operating parameters of the annealing process and the skin pass rolling process, at least one of the operating parameters of the cold rolling process is preferably used as teacher data for the learned machine learning model. The cold rolling process is a process on the upstream side of the annealing process, and is a process for reducing the thickness of the steel strip 11 to a predetermined plate thickness. As operating parameters for the cold rolling process, the total reduction in the cold rolling process, the rolling load and reduction in each pass, and the surface roughness of the work rolls used in the cold rolling can be used.

As the operating parameters for the cold rolling process, it is preferable to use the operating parameters for the final pass in the cold rolling process. This is because the final pass of the cold rolling process affects the surface roughness of the steel strip 11 on the entry side of the temper rolling mill 63 . Therefore, it is more preferable to use at least one operational parameter of the rolling reduction in the final pass of the cold rolling process, the rolling load, and the surface roughness of the work rolls in the final pass as the operational parameter of the cold rolling process. Further, as an operational parameter of the cold rolling process, the total reduction applied to the steel strip 11 in the cold rolling process may be used. This is because the total reduction ratio affects the behavior of recrystallization and grain growth of the internal structure of the steel strip 11 in the annealing process, and this behavior affects the deformation resistance in the temper rolling process. If cold rolling process operational parameters are used, these values are stored in process computer 74 and output to surface roughness predictor 76 .

Next, a method of predicting the surface roughness by the surface roughness predicting device 76 and a method of specifying the operating parameters of the skin pass rolling process will be described. The surface roughness calculator 78 of the surface roughness prediction device 76 predicts the surface roughness before the steel strip 11 is conveyed to the skin pass rolling facility 60 . The process computer 74 receives input data used for predicting the surface roughness of the steel strip 11 (operation parameter , setting values of operation parameters of the temper rolling process) and the target surface roughness range of the steel strip 11 to be manufactured are transmitted to the surface roughness calculator 78 .

When the input data is acquired from the process computer 74, the surface roughness calculation unit 78 reads out the learned machine learning model from the storage unit 80, and inputs the input data to the machine learning model to calculate the steel after the skin pass rolling process. A predicted value for the surface roughness of band 11 is determined. In this manner, the surface roughness prediction method according to the present embodiment predicts the surface roughness of the steel strip 11 after the temper rolling step.

Further, the surface roughness calculator 78 compares the calculated predicted value of surface roughness with the target surface roughness range, and if the predicted value exceeds the target surface roughness range, , the set values of the operation parameters of the skin pass rolling process among the input data are changed, and the predicted value of the surface roughness is obtained again. The surface roughness calculator 78 repeats this process until the predicted surface roughness value falls within the target surface roughness range.

The unit and range for changing the operating parameters of the skin pass rolling process may be stored in advance in the storage unit 80 or may be transmitted from the process computer 74 to the surface roughness calculation unit 78 . Further, if the operating parameters within the range of the target surface roughness cannot be identified within the range of the operating parameters, the surface roughness calculator 78 sends a signal indicating that the operating parameters cannot be identified to the process computer 74. may be output.

In this way, by specifying the operating parameters of the skin pass rolling process that are within the target surface roughness range and setting the operating parameters to the operating conditions of the skin pass rolling equipment 60, after the skin pass rolling process The surface roughness of the steel strip 11 can be controlled to be uniform and without variation, and the steel strip 11 having excellent surface quality can be manufactured.

After the surface roughness of steel strip 11 is measured, surface roughness calculator 78 acquires the measured value of the surface roughness of steel strip 11 from process computer 74 . The surface roughness calculator 78 stores in the database 84 a data set including the operational parameters of the annealing process, the operational parameters of the skin pass rolling process, and the measured values of the surface roughness. The surface roughness calculator 78 may update the learned machine learning model using teacher data including a new data set stored in the database 84 .

Next, the process of specifying the operation parameters of the temper rolling process by the surface roughness calculator 78 will be described. FIG. 6 is a flowchart for explaining the process of specifying the operation parameters of the skin pass rolling process by the surface roughness calculator 78. As shown in FIG. The flow shown in FIG. 6 is started, for example, on the condition that the steel strip 11 reaches a predetermined position before being transported to the temper rolling facility 60 after the annealing process is completed.

First, the surface roughness calculator 78 acquires the input data used to predict the surface roughness and the target surface roughness range from the process computer 74 (step S101). The surface roughness calculator 78 reads the learned machine learning model from the storage unit 80 (step S102), inputs input data to the learned machine learning model, and calculates the surface roughness after the skin pass rolling process. (step S103).

The surface roughness calculator 78 determines whether the calculated predicted value of surface roughness is within the target range (step S104). When the surface roughness calculation unit 78 determines that the calculated predicted value of the surface roughness is within the target range (step S104: Yes), the operation parameter of the skin pass rolling process included in the input data is used as the skin pass rolling process parameter. The operating conditions of the facility 60 are specified, the operating parameters are output to the process computer 74 (step S105), and the process of specifying the operating parameters for the skin pass rolling process ends. On the other hand, when the surface roughness calculator 78 determines that the calculated predicted value of the surface roughness exceeds the target range (step S104: No), the operating parameters of the skin pass rolling process among the input data are changed. (Step S106), the process from step S102 to step S104 is repeated again.

It should be noted that the target, change unit, and changeable range of the operating parameters of the skin pass rolling process in step S106 may be acquired from the process computer 74 at the time of processing in step S101, and the target within the range in which the operating parameters can be changed may be obtained. If the predicted value of the surface roughness within the range of the surface roughness to be calculated is not calculated, a signal indicating that the operation parameters of the skin pass rolling process cannot be specified is output to the process computer 74, and the process is terminated. do.

As described above, in the surface roughness prediction method according to the present embodiment, the surface roughness of the steel strip after the skin pass rolling process is predicted using a learned machine learning model that takes into consideration the influence of the annealing process. Therefore, the surface roughness of the steel strip can be predicted with higher accuracy than the method that does not consider the influence of the annealing process. Using this steel strip surface roughness prediction method, the operating parameters of the skin pass rolling process that allows the surface roughness of the steel strip to fall within the range of the target surface roughness are specified, and the skin pass rolling process is performed using the operating parameters. By carrying out, it is possible to manufacture the steel strip 11 having excellent surface quality within the target surface roughness range.

In the example shown in FIG. 2, the example in which the surface roughness prediction device 76 is connected to the process computer 74 has been described. However, the configuration is not limited to this, and the process computer 74 may have the function of the surface roughness prediction device 76 . In addition, the surface roughness prediction device 76 reads the learned machine learning model from the storage unit 80, inputs input data to the learned machine learning model, and predicts the surface roughness of the steel strip after the skin pass rolling process. explained with an example. However, not limited to this, the surface roughness calculation unit 78 reads out the machine learning model from the storage unit 80, generates a learned machine learning model using the data set stored in the database 84, and then Input data may be input to a learned machine learning model to obtain a predicted value of surface roughness after the skin pass rolling process.

An example in which a steel strip having a thickness of 0.6 to 1.2 mm and a width of 850 to 1570 mm was manufactured using the continuous annealing equipment 40 configured as shown in FIGS. The temper rolling mill used was a four-high rolling mill with a work roll diameter of 500 mm, and the target elongation rate was 0.8 to 1.2%. A work roll adjusted to have an average roughness of 2.4 μm by shot dulling was incorporated into a temper rolling mill, and a dry temper rolling process was performed without using a pressure conditioning liquid.

As a machine learning model, a neural network with three intermediate layers and five nodes each is used, and the operating parameters of the annealing process, the operating parameters of the temper rolling process, and the steel strip attribute parameters are used as input data. , generated a machine learning model that has been trained using the surface roughness after the skin pass rolling process as output data. A sigmoid function was used as the activation function.

In this example, the soaking temperature and soaking time in the soaking zone, the temperature difference on the surface of the steel strip at the inlet and outlet of the cooling zone, and the passage time through the cooling zone were used as the operating parameters of the annealing process. As operating parameters for the skin pass rolling process, strip thickness, strip width, elongation rate, rolling speed, entry-side tension, and total rolling elongation starting from the time when the work rolls were changed were used. Contents (% by mass) of C, Si, and Mn measured in the steelmaking process were used as attribute parameters of the steel strip. The average roughness Ra of the steel strip was used as the surface roughness of the steel strip. The average roughness Ra is a value measured by an online roughness meter installed in an inspection device downstream of the temper rolling mill, and is the value at the center of the strip width at a position 50 m from the tip of the steel strip. .

600 sets of these data were prepared, 450 sets were used for generating a trained machine learning model, and the remaining 150 sets were used to verify the prediction accuracy. When the prediction accuracy of the surface roughness of the steel strip predicted using the learned machine learning model was calculated, the average error of the predicted surface roughness was 0.1 μm, and the standard deviation of the error was 0.12 μm. rice field. The error of the surface roughness is the absolute value of the value calculated by "measured value of surface roughness-predicted value of surface roughness".

As a comparative example, the regression equation disclosed in Patent Document 1, which does not consider the influence of the annealing process, was obtained, and the surface roughness of the steel strip was predicted using this regression equation. Regarding the predicted surface roughness, the average error of the surface roughness and the standard deviation of the error were calculated in the same manner as in the example. From this result, it was confirmed that the surface roughness of the steel strip after the skin pass rolling process can be predicted with high accuracy by using the surface roughness prediction method according to the present embodiment. Furthermore, using a prediction method that can predict the surface roughness of the steel strip with such high accuracy, the operating parameters of the skin pass rolling process that make the predicted surface roughness within the target range are specified, and the specified operation By carrying out the skin pass rolling process with the parameters, it is possible to manufacture a steel strip with excellent surface quality within the target surface roughness range.

10 cold rolling equipment 11 steel strip 12 payoff reel 14 welding machine 16 No. 1 Looper 18 Tension Leveler 20 Pickling Equipment 22 Pickling Tank 24 Rinse Tank 26 No. 2 Looper 28 Trimmer 30 No. 3 looper 32 continuous cold rolling mill 34 tension reel 40 continuous annealing equipment 41 entry side equipment 42 payoff reel 44 welder 46 electrolytic cleaner 48 tension leveler 49 entry side looper 50 furnace body 53 heating zone 54 soaking zone 55 cooling zone 58 delivery side facility 59 delivery side looper 60 temper rolling facility 61 entry side bridle roll 62 tension gauge 63 temper rolling mill 64 delivery side bridle roll 66 load cell 67 surface roughness meter 68 mill speed controller 69 bridle roll speed controller 70 Side trimmer 72 Tension reel 74 Process computer 76 Surface roughness prediction device 78 Surface roughness calculation unit 80 Storage unit 82 Machine learning model 84 Database 86 Input unit 88 Operator 90 Reheating zone 91 Overaging zone 92 Final cooling zone

Claims (7)

A method for predicting the surface roughness of a steel strip manufactured in a manufacturing process including an annealing process and a temper rolling process after the temper rolling process,
At least one of the operating conditions of the heating zone, the soaking zone and the cooling zone in the annealing process, and at least one of the operating parameters of the temper rolling process of the steel strip for predicting the surface roughness after the temper rolling process Take as input data,
inputting the input data into a machine learning model whose output data is the surface roughness of the steel strip after the skin pass rolling process, and predicting the surface roughness of the steel strip after the skin pass rolling process; Forecast method. 2. The method of predicting surface roughness according to claim 1, further acquiring as said input data at least one attribute parameter relating to the chemical composition of the steel strip whose surface roughness after said skin pass rolling step is predicted. The manufacturing process further includes a cold rolling process provided upstream of the annealing process, and further acquires at least one operational parameter of the cold rolling process as the input data. 2. The method for predicting surface roughness according to 2. The operating parameters of the skin pass rolling process are changed to predict the surface roughness of the steel strip after the skin skin rolling process, and the operating parameters of the skin pass rolling process are set such that the predicted surface roughness is within the target range. A method for predicting surface roughness according to any one of claims 1 to 3, characterized. A method for manufacturing a steel strip, wherein the steel strip is manufactured in a manufacturing process including an annealing step and a skin pass rolling step in which the operation parameters specified by the surface roughness prediction method according to claim 4 are set. A method for generating a learned machine learning model for outputting the surface roughness of a steel strip after the temper rolling process of a steel strip manufactured in a manufacturing process including an annealing process and a temper rolling process, the heating zone in the annealing process , at least one of the operating conditions of the soaking zone and the cooling zone , at least one of the operating parameters of the temper rolling process, and the surface roughness of the steel strip after the temper rolling process, as teacher data, A learned machine learning model generation method for generating a learned machine learning model that outputs the surface roughness after the skin pass rolling step. 7. The method of generating a trained machine learning model according to claim 6, wherein said machine learning model is any one of neural network, decision tree learning, random forest and support vector regression.

Images