JP2022192003A - Surface layer hardness prediction model and predictive control method for surface layer hardness of steel plate using the same, control command device, steel plate manufacturing line, and steel plate manufacturing method - Google Patents

Abstract

To accurately predict material properties of a manufactured steel plate, especially, surface layer hardness.SOLUTION: A surface layer hardness prediction model fitted for a steel plate manufacturing line to manufacture a steel plate is subjected to machine learning using a predetermined machine learning algorithm that is inputted with carbon equivalent of the steel plate, rolling process data in a rolling process for the steel plate, and cooling process data in a cooling process for the steel plate, and outputs material testing data representing surface layer hardness of the steel plate, so as to predict the surface layer hardness of the steel plate.SELECTED DRAWING: Figure 1

Description

TECHNICAL FIELD The present invention relates to a technique for predicting and controlling the surface layer hardness of a steel sheet, and more particularly to a surface layer hardness prediction model, a method for predicting and controlling the surface layer hardness of a steel sheet using the same, a control command device, a steel sheet manufacturing line, and a steel sheet manufacturing method.

Generally, in the steel sheet manufacturing process, after the chemical components of crude steel are adjusted, the crude steel is reheated in a heating facility, hot rolled, and then cooled under controlled conditions. The target material properties of the steel sheet are determined according to the chemical composition, heating conditions, rolling conditions, and/or controlled cooling conditions. However, these conditions are likely to vary due to various disturbance factors, and as a result, the material properties of the steel sheet vary, leading to a decrease in yield. Therefore, it is extremely important for quality management and quality control to predict the material properties of steel sheets manufactured under conditions that vary due to disturbance factors.

Therefore, conventionally, several techniques for predicting the material properties of steel sheets have been proposed. For example, in Patent Document 1 below, when a steel ingot is hot-rolled after reheating and air-cooled or accelerated cooling is performed to produce a thick plate, the composition of the steel ingot, the original thickness of the steel ingot, the reheating A technique is disclosed for predicting and calculating the material quality by constructing a hierarchical neural network in which the conditions, rolling conditions, cooling conditions, and heat treatment conditions are input signals and the material quality is the output signal.

In addition, in Patent Document 2 below, for each process, a predicted material quality value based on process conditions is calculated, and if the calculated predicted material quality value does not satisfy the guaranteed material quality value, recalculation is performed to correct the process conditions, and the target A technique for obtaining a steel material having a material quality is disclosed.

By the way, the surface layer hardness of a steel sheet affects its ductility, and is an important material property to ensure sour resistance in a wet hydrogen sulfide environment (sometimes referred to as a "sour environment"). is one. A "surface layer" of a steel plate generally refers to a layer or portion from the surface to a depth of about 2.0 mm. However, the surface layer hardness of a steel sheet is easily affected by external disturbance factors. It is an easy characteristic. In addition, the composition of the oxide scale is susceptible to the distortion of the surface layer, the condition of the outside air, the condition of the injection of cooling water, trace elements, and the like. In particular, in the fields of civil engineering and construction industrial machinery, steel sheets for sour line pipes are required, so it is extremely important to control the surface layer hardness of steel sheets to a desired value.

For example, Patent Document 3 below discloses a technique for controlling the surface layer hardness of a steel sheet by controlling the cooling rate of 0.5 mm below the surface layer of the steel sheet in controlled cooling to 80° C./s or less.

Further, Patent Document 4 below discloses a technique for controlling the surface layer hardness of a steel sheet by reducing the water density in the first half of the cooling process to cool the steel sheet, thereby suppressing the cooling rate of the surface of the steel sheet.

JP-A-8-240587 Japanese Patent No. 2509481 Japanese Patent No. 6521197 JP-A-57-152430

As described above, the surface layer hardness of a steel sheet is one of the important material properties of the manufactured steel sheet. The surface layer hardness of such a steel sheet is typically measured by pressing an indenter into the surface of the steel sheet after the cooling process. For example, in a technique called Brinell hardness measurement, the size of an indentation produced when an indenter is pressed into the surface of a steel sheet is measured. In this case, the hardness of the portion up to a depth of several millimeters from the surface of the steel sheet affects the measurement. For example, FIG. 12 shows changes over time (temperature history) in temperature at the surface and at a depth of 1 mm from the surface when a steel plate with a thickness of 25 mm is cooled. As can be seen from the figure, when comparing the steel plate surface and the position at a depth of 1 mm from the surface, the temperature histories are different, and accordingly the hardness is also different. Therefore, there is a problem that it is not possible to accurately predict the surface layer hardness of a steel sheet simply by predicting the state of the structure at a specific position in the thickness direction of the steel sheet (for example, the center of the thickness or the surface) as in the conventional method. be.

Further, in general, when trying to produce a high-strength and high-toughness steel sheet, it is effective to rapidly cool the steel sheet in a high temperature state after hot rolling at a high cooling rate. However, if the cooling rate is too high, a problem may arise that the surface layer of the steel sheet hardens more than expected.

Furthermore, it is known that when cooling a steel plate, the cooling capacity changes depending on the thickness of scale generated on the surface of the steel plate. For example, as shown in FIG. 13, when the thickness of the scale increases, the cooling rate increases, which causes the surface of the steel sheet to harden more than expected. In particular, in a situation where red scale occurs, the cooling rate is high as shown in FIG. 14, and the surface layer of the steel sheet hardens more than expected. Red scale refers to scale having a red surface appearance due to generation of powdery Fe2O3 on the surface layer. Such red scales are generated by pulverizing the scales on the surface layer of the steel sheet into powder during the rolling process. In addition, there is a problem that the thicker the scale, the more easily the scale is pulverized, and the more easily the red scale is generated.

In view of the above, the techniques disclosed in Patent Documents 1 and 2 described above are for predicting the material properties related to the entire plate thickness of the steel sheet, and the material properties of local portions such as surface layer hardness can be accurately estimated. Not suitable for prediction. In particular, these techniques do not consider predicting the state of scale formed on the surface of the steel sheet, that is, both the thickness of the scale and the presence or absence of red scale. If the results were used to control the material, there was a problem that the surface hardness of the steel sheet would become higher than expected.

In addition, the techniques for controlling the surface layer hardness of the steel sheet disclosed in Patent Documents 3 and 4 described above do not take into account the influence that the cooling ability of the steel sheet surface layer changes depending on the state of the scale on the surface of the steel sheet. There is a problem that the surface layer hardness of the steel sheet becomes higher than expected.

Accordingly, an object of the present invention is to propose a technique capable of accurately predicting the material properties of a steel sheet to be manufactured, particularly the surface layer hardness thereof.

More specifically, one of the objects of the present invention is to consider the influence of the thickness and state of scale formed immediately before the start of cooling on the surface layer hardness of the steel sheet in the steel sheet manufacturing process. A surface hardness prediction model that enables the production of steel sheets with little variation in surface hardness by accurately predicting the It is to provide a steel plate manufacturing method.

The present invention for solving the above-mentioned problems includes the matters specifying the invention or the technical features described below.

The present invention according to one aspect is a surface layer hardness prediction model adapted to a steel plate manufacturing line for manufacturing steel plates. The surface layer hardness prediction model inputs the carbon equivalent of the steel sheet, the rolling process data in the rolling process for the steel sheet, and the cooling process data in the cooling process for the steel sheet, and generates material test data indicating the surface layer hardness of the steel sheet. The output is constructed by performing machine learning using a predetermined machine learning algorithm to predict the surface layer hardness of the steel sheet.

Here, the rolling process data may include the surface temperature of the steel sheet when the final descaling is applied, and the elapsed time from the application of the final descaling to the end of rolling. In addition, the rolling process data may further include at least one of specifications of rolling equipment that performs the rolling process, a conveying speed of the steel sheet, dimensions of the steel sheet, and descaling information regarding descaling of the steel sheet.

Also, the cooling process data may include a cooling start temperature, a cooling stop temperature, and a calculated cooling rate of the steel plate surface layer of the steel plate. Further, the cooling process data may include at least one of the specifications of cooling equipment for performing the cooling process, the conveying speed of the steel plate, and the dimensions of the steel plate.

Also, the material test data may be values measured by any one of Vickers hardness measurement method, Brinell hardness measurement method, and Rockwell hardness measurement method.

Also, the predetermined machine learning algorithm may be at least one of neural network, decision tree learning, random forest, and support vector regression.

Also, the steel plate may be at least one of a thick steel plate, a steel plate for line pipes, and a steel plate for civil engineering and construction.

The present invention according to another aspect is a control command device for controlling a steel plate manufacturing line for manufacturing steel plates. The control command device inputs the carbon equivalent of the steel sheet, the rolling process data in the rolling process for the steel sheet, and the cooling process data in the cooling process for the steel sheet, and outputs material test data indicating the surface layer hardness of the steel sheet. As a surface hardness prediction model subjected to machine learning using a predetermined machine learning algorithm so as to predict the surface hardness of the steel sheet, and the prediction value output from the surface hardness prediction model, the cooling a surface layer hardness control section for changing and resetting the process data.

The surface layer hardness prediction model can output the predicted value based on the carbon equivalent, the rolling process data, and the cooling process data during operation of the steel sheet manufacturing line. Further, the surface layer hardness control unit can change and reset the cooling process data when the predicted value output from the surface layer hardness prediction model does not satisfy a predetermined target value.

Further, the surface layer hardness control unit can change and reset operation parameters related to the injection amount of cooling water and/or the conveying speed of the steel sheet in the cooling process data.

According to another aspect of the present invention, there is provided a control command device, and under the control of the control command device, rolling is performed until a predetermined thickness is achieved while descaling scales formed on the surface of a heated steel plate. and a cooling facility for cooling the rolled steel sheet at a predetermined cooling rate under the control of the control command device. The control command device includes a surface layer hardness prediction model. Then, the surface layer hardness prediction model inputs the carbon equivalent of the steel sheet, the rolling process data in the rolling process for the steel sheet, and the cooling process data in the cooling process for the steel sheet, and a material test that indicates the surface layer hardness of the steel sheet. Using the data as an output, machine learning is performed using a predetermined machine learning algorithm so as to predict the surface layer hardness of the steel sheet, and a system is constructed.

Moreover, the present invention according to another aspect may be a method of predicting and controlling the surface layer hardness of the steel sheet using the surface layer hardness prediction model. In the method, the surface layer hardness prediction model subjected to the machine learning is based on the carbon equivalent, the rolling process data, and the cooling process data during operation of the steel sheet production line. Includes outputting hardness predictions.

Further, the method further comprises changing and resetting the cooling process data based on the output predicted value, and performing the cooling process based on the reset cooling process data. can contain.

Further, the resetting may include changing operation parameters related to the injection amount of cooling water and/or the conveying speed of the steel plate in the cooling process data.

Moreover, the present invention according to another aspect can be a steel sheet manufacturing method. The steel sheet manufacturing method includes a rolling process of rolling a heated steel sheet to a predetermined thickness, and a cooling process of cooling the rolled steel sheet at a predetermined cooling rate. Further, in the steel sheet manufacturing method, the cooling data is changed and reset based on the predicted value output from the surface layer hardness prediction model, and based on the reset cooling data, Including changing operating parameters.

In this specification and the like, "means" does not simply mean physical means, but also includes the case where the functions of the means are realized by software. Also, the function of one means may be realized by two or more physical means, or the functions of two or more means may be realized by one physical means.

According to the present invention, it is possible to accurately predict and control the surface layer hardness of a steel sheet, thereby making it possible to manufacture a steel sheet with no variation in quality.

Other technical features, objects, effects or advantages of the present invention will be made clear by the following embodiments described with reference to the accompanying drawings. The effects described in this specification are only examples and are not limiting, and other effects may also occur.

BRIEF DESCRIPTION OF THE DRAWINGS It is a figure which shows an example of a schematic structure of the steel plate production line to which the surface layer hardness prediction model which concerns on one Embodiment of this invention is applied. BRIEF DESCRIPTION OF THE DRAWINGS It is a figure which shows an example of a schematic structure of the cooling equipment in the steel plate manufacturing line to which the surface layer hardness prediction model which concerns on one Embodiment of this invention is applied. BRIEF DESCRIPTION OF THE DRAWINGS It is a figure which shows an example of a schematic structure of the cooling equipment in the steel plate manufacturing line to which the surface layer hardness prediction model which concerns on one Embodiment of this invention is applied. BRIEF DESCRIPTION OF THE DRAWINGS It is a figure which shows an example of a schematic structure of the control command apparatus in the steel plate production line to which the prediction technique of the steel plate surface layer hardness which concerns on one Embodiment of this invention is applied. FIG. 4 is a diagram showing the presence or absence of red scale on a steel sheet in a rolling process. FIG. 4 is a diagram showing the average value of the cooling rate and the hardness at a predetermined depth from the surface of the steel sheet in the cooling process. It is a figure for demonstrating the division|segmentation point of the board|plate thickness (depth) direction of a steel plate. FIG. 4 is a diagram for explaining the calculated cooling temperature of the steel sheet surface layer used in the steel sheet surface layer hardness prediction technique according to the embodiment of the present invention. FIG. 4 is a diagram showing the relationship between the Vickers hardness and the cooling rate at a predetermined depth from the surface of the steel sheet. 4 is a flowchart for explaining a prediction process of surface layer hardness of a steel sheet by a learned surface layer hardness prediction model according to one embodiment of the present invention. 4 is a chart for explaining examples and comparative examples using a steel sheet manufacturing line to which a surface layer hardness prediction model according to an embodiment of the present invention is applied; FIG. 4 is a diagram showing temperature histories at a surface of a steel sheet and a position at a predetermined depth from the surface when a steel sheet in a high temperature state is subjected to a cooling process. FIG. 4 is a diagram showing the temperature history of a steel sheet with a predetermined scale thickness when a steel sheet in a high temperature state undergoes a cooling process. FIG. 4 is a diagram showing temperature histories of steel sheets with and without occurrence of red scale when a steel sheet in a high temperature state undergoes a cooling process.

BEST MODE FOR CARRYING OUT THE INVENTION Hereinafter, embodiments of the present invention will be described with reference to the drawings. However, the embodiments described below are merely examples, and are not intended to exclude various modifications and application of techniques not explicitly described below. The present invention can be implemented in various modifications (for example, by combining each embodiment) without departing from the scope of the invention. In addition, in the description of the drawings below, the same or similar parts are denoted by the same or similar reference numerals. The drawings are schematic and do not necessarily correspond to actual dimensions, proportions, and the like. Even between the drawings, there are cases where portions with different dimensional relationships and ratios are included.

FIG. 1 is a diagram showing an example of a schematic configuration of a steel sheet manufacturing line to which a surface layer hardness prediction model according to one embodiment of the present invention is applied. In the present disclosure, an example in which a model for predicting the surface layer hardness of a steel sheet is applied to a steel sheet manufacturing line will be described, but the model can also be applied to hot rolling and shaped steel manufacturing lines. Moreover, the steel plate may be at least one of a thick steel plate, a steel plate for line pipes, and a steel plate for civil engineering and construction.

As shown in the figure, the steel sheet production line 1 includes a control command device 10, a heating facility 20, a rolling facility 30, a descaling facility 40, a radiation thermometer 50, and a cooling facility 60.

The control command device 10 is a device that issues a command for overall control of the operation of the steel plate manufacturing line 1 . The control command device 10 controls each facility and equipment in the steel plate manufacturing line 1 according to, for example, a predetermined control sequence program. The control command device 10 is composed of, for example, a process computer 110 that controls each processing facility and a host computer 120 that supervises them (see FIG. 4). Further, as will be described later, the control command device 10 has a surface layer hardness prediction model 121 subjected to machine learning for predicting the surface layer hardness of the steel plate to be manufactured. The control command device 10 uses the surface layer hardness prediction model 121 to predict the surface layer hardness of the steel sheet, and based on the prediction result, resets the operation parameters of the cooling process, thereby producing a steel sheet with less variation in surface layer hardness. enable manufacturing.

The heating equipment 20 is equipment including a furnace for heat-treating a plate-like slab (steel plate). During operation, the heating equipment 20 is heated by, for example, a heating device (not shown) to maintain a high temperature, heats the steel plate brought in from the carry-in port, and carries it out from the carry-out port.

The rolling facility 30 is a facility for rolling a steel sheet heated to a predetermined temperature. The rolling equipment 30 is composed of, for example, a plurality of rolling mills 31, that is, a roughing mill 31a and a roughing mill 31b. It consists of a work roll and a backup roll that receives a rolling load and suppresses deflection of the work roll, but is not limited to this. The steel sheet is rolled in each of the rough rolling mill 31a and the rough rolling mill 31b, for example, by moving back and forth in the conveying direction, and after rolling, is conveyed to the next step (cooling step).

The descaling equipment 40 is equipment for removing (descaling) scale formed on the surface of the heated steel plate according to the descaling information. The descaling equipment 40 includes, for example, injection nozzles 41 provided on the entry and exit sides of the rolling mill 31 so as to face the upper and lower surfaces of the steel sheet, respectively. The descaling information includes, for example, information on descaling injection results (water volume, injection pressure, injection angle, nozzle shape, etc.). The descaling equipment 40 injects a high-pressure stream of water from the injection nozzle 41 onto the surface of the steel sheet being rolled by the rolling equipment 30, thereby removing the generated scale. Although the descaling facility 40 is described in this disclosure as being configured separately from the rolling facility 30 , the descaling facility 40 may be configured as part of the rolling facility 30 .

The radiation thermometer 50 is an instrument for non-contact measurement of the surface temperature of the steel sheet being processed using infrared rays. The temperature measured by the radiation thermometer 50 is notified to the control command device 10 . The control command device 10 controls the steel plate manufacturing line 1 based on the notified temperature.

The cooling equipment 60 is equipment for cooling the rolled steel sheet in a high temperature state at a predetermined cooling rate. The cooling facility 60 is typically divided into several cooling zones Z, and cools the steel sheet by jetting cooling water onto the surface of the steel sheet, as will be described later.

2 and 3 are diagrams showing an example of a schematic configuration of a cooling facility in a steel sheet manufacturing line to which a surface layer hardness prediction model according to an embodiment of the present invention is applied. More specifically, FIG. , shows an example of a schematic configuration of a cooling facility using a draining purge, and FIG. 3 shows an example of a schematic configuration of a cooling facility using a draining roll.

A cooling facility 60 shown in FIG. 2 is a facility having a plurality of cooling zones Z partitioned by draining purges 61 . Each cooling zone Z is provided with an injection nozzle 63 connected to a flow control valve 62 . In order to cool the steel plate, the injection nozzle 63 injects cooling water, the injection flow rate of which is adjusted by the flow control valve 62, toward the steel plate. The drain purge 61 purges water remaining on the surface of the steel sheet by injecting high-pressure air onto the surface of the steel sheet. Radiation thermometers 50 are provided on the entrance side and the exit side of each cooling zone Z. As shown in FIG. The temperature measured by the radiation thermometer 50 is notified to the control command device 10 . The control command device 10 controls the steel plate manufacturing line 1 based on the notified temperature. The control command device 10 adjusts the water flow rate of the flow rate adjustment valve 62 and the air flow rate of the drain purge 61 based on the notified temperature.

Also, the cooling facility 60 shown in FIG. 3 has a plurality of cooling zones Z partitioned by draining rolls 64 . Similarly, each cooling zone Z is provided with an injection nozzle 63 connected to a flow control valve 62, and radiation thermometers 50 are provided on the inlet and outlet sides of each cooling zone Z. The steel plate is cooled by cooling water jetted from the jet nozzle 63 under the control of the control command device 10 .

FIG. 4 is a diagram showing an example of a schematic configuration of a control command device in a steel sheet manufacturing line to which a surface layer hardness prediction model according to one embodiment of the present invention is applied. As shown in the figure, the control command device 10 includes, for example, a process computer 110, a host computer 120, and a database .

Process computer 110 is a computer for controlling the processing equipment. In this example, the heating equipment process computer 110a for controlling the heating equipment 20, the rolling equipment process computer 110b for controlling the rolling equipment 30 and the descaling equipment 40, and the cooling equipment for controlling the cooling equipment 60 A facility process computer 110c is provided. For example, the cooling facility process computer 110c controls the cooling rate based on cooling process data (for example, operating parameters) set based on the predicted surface layer hardness. These process computers 110a to 110c are connected to a host computer 120 via a communication network such as a local LAN.

The host computer 120 is a computer that controls each of the process computers 110a to 110c and controls the steel plate manufacturing line 1 in an integrated manner. In this example, the host computer 120 can include a surface layer hardness prediction model 121 and a surface layer hardness controller 122 . The surface hardness control unit 122 instructs the process computer 110 so that the surface hardness predicted by the surface hardness prediction model 121 falls within the target value.

The database 130 accumulates operation parameters of each process of the steel sheet manufacturing line 1 and various performance data acquired during operation. The performance data includes, for example, rolling process data and cooling process data. In addition, the database 130 stores material test data measured in advance by the actual composition of the steel sheet and a predetermined measurement method. The material test data is, for example, the surface layer hardness of the steel plate. In this example, the database 130 is connected to the host computer 120, but is not limited to this and may be directly connected to the process computers 110a to 110c.

Next, the surface layer hardness prediction model 121 will be described. The surface layer hardness prediction model 121 of the present disclosure is constructed so as to accurately predict the surface layer hardness (objective variable) of a steel sheet by using various input parameters (explanatory variables) related to the occurrence and growth of scale. More specifically, the surface layer hardness prediction model 121 uses the steel plate carbon equivalent, rolling process data, and cooling process data as inputs (training data), and the material test actual values of the steel plate as outputs (labels). Machine learning is performed to predict the surface layer hardness of the steel sheet according to the learning algorithm. Each input and output will be described below.

ただし、Ｃ、Ｓｉ、Ｍｎ、Ｎｉ、Ｃｒ、Ｍｏは、及びＶは、それぞれ、炭素、ケイ素、マンガン、ニッケル、クロム、モリブデン、及びバナジウム成分の含有量を示す。なお、炭素当量は、上記の式に限られず、ロイド式で算出された値であっても良い。 (Carbon equivalent of steel plate)
The surface layer hardness prediction model 121 is given a carbon equivalent, which is information about the material of the steel sheet. The carbon equivalent is a value that contributes to predicting the structural state of the steel sheet as well as predicting the susceptibility to scale components, particularly red scale. The carbon equivalent (Ceq) of a steel sheet is a value calculated by the following formula (1).

However, C, Si, Mn, Ni, Cr, Mo, and V indicate contents of carbon, silicon, manganese, nickel, chromium, molybdenum, and vanadium components, respectively. Note that the carbon equivalent is not limited to the above formula, and may be a value calculated by Lloyd's formula.

(Rolling process data)
The surface layer hardness prediction model 121 is further provided with rolling process data regarding the rolling process of the steel sheet. The rolling process data includes the surface temperature of the steel sheet when the final descaling is applied and the elapsed time from the application of the final descaling to the end of rolling in the rolling process. The surface temperature of the steel sheet when the final descaling is applied in the rolling process means the surface temperature immediately before the rolling bite of the rolling pass to which the final descaling is applied in the rolling process. In addition, the elapsed time from the application of the final descaling in the rolling process to the end of rolling is the time from the time when the steel plate is caught in the work rolls of the rolling stand in the rolling pass to which the final descaling is applied in the rolling process. It is the elapsed time until the steel plate is bitten in the last pass. As a result, it is possible to grasp the scale generation state starting from the end of scale removal by the final descaling, and further the generation state of red scale due to the scale being pulverized during rolling biting.

The rolling process data may be based on actual measurement data output from various measuring instruments, for example. For example, the host computer 120 can acquire measured data output from various measuring devices (not shown) via the rolling facility process computer 110b. Alternatively, the rolling process data may be based on a rolling data set acquired as general rolling operation data. Rolling data sets include, for example, specifications of rolling equipment (descaling arrangement, number of rolling stands, etc.), rolling speed, dimensions of steel plate (thickness, width, length, shape, etc.), position information of steel plate (longitudinal direction, width direction position, etc.), descaling injection results (water volume, injection pressure, injection angle, nozzle shape, etc.), pass schedule, etc., but are not limited to these.

The growth of scale is generally arranged by diffusion control and is represented by the following formula (2).
ξ2=a×exp(−Q/RT)×t (2)
is the scale thickness, a is a constant, Q is the activation energy, R is a constant, and t is the elapsed time. As is clear from the formula (2), the higher the surface temperature of the steel sheet and the longer the elapsed time, the more the scale grows and becomes thicker.

FIG. 5 is a diagram showing the presence or absence of red scale on the steel sheet in the rolling process. The generation of red scale on a steel sheet is related to the surface temperature of the steel sheet when the final descaling is applied and the elapsed time from the application of the final descaling to the end of rolling in the rolling process. In the figure, "○" is a plot of the relationship between the surface temperature and the elapsed time when there was no red scale, and "×" is the surface temperature when the red scale was generated and the elapsed time. It plots the relationship with time. As shown in the figure, when the surface temperature at the final descaling is high and the elapsed time from the final descaling pass to the end of the rolling process is long, red scale is likely to occur on the steel sheet.

In addition, FIG. 6 is a diagram showing the cooling rate at a position 1 mm deep from the surface of the steel plate in the cooling process and the average value of the hardness, and shows the case where red scale is generated and the case where it is not generated. ing. In the cooling process, red scale occurred when the surface temperature of the steel sheet at the time of final descaling was 820° C. or higher and the elapsed time from the final descaling pass to the end of rolling was 60 s or longer. As shown in the figure, when red scale occurs, the cooling rate is fast and the surface layer hardness is high.

Moreover, as shown in the formula (2), the higher the surface temperature of the steel sheet and the longer the elapsed time, the thicker the scale grows. That is, the higher the surface temperature of the steel sheet when the final descaling is applied and the longer the elapsed time from the application of the final descaling to the end of rolling, the more the scale grows and becomes thicker. Also, the thicker the scale, the faster the cooling rate and the higher the surface layer hardness. From this, by adding the surface temperature of the steel sheet when the final descaling in the rolling process is applied and the elapsed time from the application of the final descaling to the end of rolling to the inputs to the surface layer hardness prediction model 121, Not only the thickness of the scale, but also the presence or absence of red scale caused by pulverization of the scale can be grasped, and the surface layer hardness of the steel sheet can be predicted with higher accuracy.

As shown in FIG. 1, the steel sheet rolling facility 30 is provided with descaling facilities 40 for removing scales from the surface of the steel sheet at the entrance and exit sides of the rolling mill 31 . Under the control of the control command device 10, the rolling equipment 30 rolls the steel sheet so that the finishing temperature at the end of rolling is within the target range in order to satisfy the requirements for the material characteristics of the steel sheet. At this time, when the thickness of the steel sheet is reduced by the rolling process, the temperature drop on the surface of the steel sheet due to the descaling water increases, so there is concern that the finishing temperature may fall below the target range. In such a situation, the descaling facility 40 suppresses injection of descaling water according to predetermined descaling information so that the finishing temperature does not fall below the target lower limit value, according to instructions from the operator, correction calculations by the process computer 110, and the like. do. Therefore, in the rolling process, the surface temperature of the steel sheet when the final descaling is applied and the elapsed time from the application of the final descaling to the end of rolling are determined by the thickness of the scale growing on the surface of the steel sheet and the thickness of the red scale. It is an effective index to determine whether or not the phenomenon occurs.

(Cooling process data)
Further, the surface layer hardness prediction model 121 is provided with cooling process data regarding the cooling process of the steel sheet. Cooling process data includes, for example, a cooling start temperature and a cooling stop temperature. The cooling start temperature is the surface temperature of the steel sheet measured by the radiation thermometer 50 installed on the inlet side of the cooling facility 60, and the cooling stop temperature is measured by the radiation thermometer 50 installed on the outlet side of the cooling facility 60. It is the measured surface temperature of the steel plate.

Further, the cooling process data is provided with the calculated cooling rate of the surface layer of the steel sheet in the cooling process. The calculated cooling rate of the steel sheet surface layer is the temperature calculated by a predetermined arithmetic expression at each point (division point) divided in the thickness (depth) direction of the steel sheet as shown in FIG. 7, for example. The calculated cooling rate of the surface layer of the steel sheet is calculated, for example, by calculating the difference based on the temperature at each division point. More specifically, as shown in FIG. 8, the calculated cooling rate of the steel sheet surface layer in the cooling process is the amount of temperature drop from the cooling start temperature point to the specific temperature point in the obtained temperature history of the steel sheet surface. Defined as the divided value. Here, the specific temperature point is the point when the surface temperature first reaches the cooling stop temperature, and the elapsed time is the time required for temperature change from the cooling start temperature point to the specific temperature point. Further, the calculated cooling rate of the steel sheet surface layer is preferably the cooling rate at the thickness position where the surface layer hardness is measured, for example, from the temperature history at a depth of about 1 mm or 2 mm from the surface of the steel sheet It can be a calculated cooling rate.

The cooling process data may be based on actual measurement data output from various measuring instruments. For example, the host computer 120 can acquire actual measurement data output from various measuring devices via the cooling facility process computer 110c. Alternatively, the cooling data may be calculated from a cooling data set acquired as general cooling operation data. Cooling data sets include, for example, cooling equipment specifications (nozzle shape and arrangement, etc.), strip threading speed, steel strip dimensions (thickness, strip width, strip length, shape, etc.), and strip position information (longitudinal direction, width direction, etc.). position, etc.), cooling water injection record (water volume, injection pressure, injection angle, cooling water injection pattern, etc.), steel plate temperature measured at an arbitrary position, etc., but are not limited to these.

The cooling process data can be an important input parameter for determining the microstructure state of the steel sheet surface layer. FIG. 9 is a diagram showing the relationship between the Vickers hardness and the cooling rate at a position 1 mm deep from the surface of a steel plate of a certain steel type. As is clear from the figure, the slower the cooling rate, the lower the Vickers hardness. Further, when the cooling start temperature is low and ferrite is partially formed before cooling starts, the hardness decreases, and when the cooling stop temperature is low, the hardness increases slightly. This is because the tempering temperature of the steel sheet is low after cooling is stopped. In this way, the cooling start temperature, the cooling stop temperature, and the calculated cooling rate of the steel plate surface layer determine the structure state of the steel plate surface layer, in addition to the carbon equivalent of the steel plate, which is information about the material, and the information about the occurrence of scale in the rolling process. It is an important input parameter for accurately predicting the hardness of the steel plate surface layer.

Next, the output of the surface layer hardness prediction model 121 in machine learning will be described. For the output (label) of the surface hardness prediction model 121 when machine learning is applied to the surface hardness prediction model 121, material test data measured by, for example, Vickers hardness, Brinell hardness, and Rockwell hardness are used as the surface hardness of the steel plate. be able to.

Vickers hardness is measured by pressing an indenter from the surface of a steel plate to a predetermined hardness measurement position with a predetermined measurement load. The predetermined hardness measurement position depends on the required specifications, and can be, for example, a position at a depth of 1 mm or 2 mm from the steel plate surface. Moreover, the measurement load is, for example, 10 kgf. For example, when measuring the Vickers hardness, first, the longitudinal end of the steel sheet is sampled as a test piece using a shear or gas cutter after the cooling process, and the sampled test piece is a cross section in the thickness direction. Polished with a polishing machine. Subsequently, the position where the hardness is to be measured is identified while confirming the polished surface with a microscope, and then the hardness is measured by pressing the indenter.

Brinell hardness and Rockwell hardness are values measured by grinding the surface of the steel sheet by several millimeters. For example, in the measurement of Brinell hardness and Rockwell hardness, first, a test piece is sampled and the surface of the sampled test piece is ground in the same manner as described above. The ground surface is then smoothed further and its hardness is measured by indenting it with an indenter. Conditions for hardness measurement may follow the conditions specified in the JIS standard, but are not limited thereto. For example, the applied load and holding time may be determined according to the index requested by the consumer.

The surface hardness prediction model 121 is subjected to machine learning for predicting the surface hardness of the steel sheet according to a predetermined machine learning algorithm based on the above-described input (training data) and output (label). In the present disclosure, the machine-learned surface layer hardness prediction model 121 may be referred to as a learned surface layer hardness prediction model 121 . Machine learning algorithms can be applied, for example, neural networks, decision tree learning, random forests, support vector regression, and the like. Also, the surface layer hardness prediction model 121 may be configured as an ensemble model in which a plurality of models are combined.

The performance data accumulated in the database 130 is used as training data for constructing the learned surface hardness prediction model 121 . The performance data may be, for example, data output from various measuring instruments during normal steel sheet manufacturing line operations. The host computer 120 provides the training data and labels read from the database 130 to the surface layer hardness prediction model 121 before machine learning, and performs machine learning. The learned surface layer hardness prediction model 121 constructed in this way can predict the surface layer hardness of the steel sheet based on the actual data during operation in the steel sheet manufacturing line 1 and output the prediction result.

FIG. 10 is a flowchart for explaining processing for controlling the surface layer hardness of a steel sheet in a steel sheet manufacturing line to which the surface layer hardness prediction model according to one embodiment of the present invention is applied. Such processing consists of prediction processing by the surface layer hardness prediction model 121 and control processing by the surface layer hardness control unit 122 based on the predicted result.

As shown in the figure, the control command device 10 first acquires performance data in a process prior to the cooling process, such as the rolling process (S1001). Furthermore, the control command device 10 acquires operation parameters in the cooling process (S1002). Subsequently, the control command device 10 provides the learned surface hardness prediction model 121 with the carbon equivalent and the acquired performance data and operation parameters, whereby the learned surface hardness prediction model 121 predicts the surface hardness of the steel sheet, The result (predicted value) is output (S1003). Prediction by the learned surface layer hardness prediction model 121 is typically performed for each cooling zone Z.

Next, the control command device 10 determines whether or not the predicted value satisfies the target value (S1004). When determining that the predicted value does not satisfy the target value (No in S1004), the control command device 10 changes and resets the cooling process data (S1005). For example, the control command device 10 changes the operating parameters related to the injection amount of cooling water and/or the conveying speed of the steel plate. Thereby, the cooling facility 60 performs cooling processing for each cooling zone according to the reset operation parameters. For example, an increase in the injection amount of cooling water increases the cooling rate of the surface temperature of the steel sheet, or a decrease in the injection amount of cooling water slows down the cooling rate. is controlled to approach Also, alternatively or additionally, the conveying speed of the steel sheet is adjusted, and similarly the surface layer hardness of the steel sheet is controlled to approach the target value. The learned surface layer hardness prediction model 121 predicts the surface layer hardness of the steel sheet based on the reset operating parameters, and outputs the result (predicted value) (S1003).

On the other hand, when the control command device 10 determines that the predicted value by the learned surface hardness prediction model 121 satisfies the target value (Yes in S1004), the prediction processing by the learned surface hardness prediction model 121 is terminated. Thereby, the cooling facility 60 will continue the cooling process according to the current operational parameters.

As described above, the control command device 10 predicts the surface layer hardness of the steel sheet by the learned surface layer hardness prediction model 121 based on the acquired performance data, and if the predicted hardness is off the target, the hardness is within the target. Since the operation parameters in the cooling process are reset so that the surface layer hardness of the steel sheet can be controlled to an appropriate value.

Further, by manufacturing steel sheets in the steel sheet manufacturing line 1 using such a learned surface layer hardness prediction model 121, it is possible to efficiently manufacture steel sheets with little variation in surface layer hardness.

In the steel plate manufacturing line 1 shown in FIG. 1, a slab (steel billet) is reheated in a heating equipment 20, and then the reheated steel plate is hot-rolled while being descaled in a rolling equipment 30. was cooled in the cooling equipment 60 as shown in 1, thereby manufacturing a predetermined number of steel plates. The cooling equipment 60 was provided with 10 cooling zones Z(1) to (10) partitioned by draining rolls 64. As shown in FIG.

The steel sheet manufactured in the steel sheet manufacturing line 1 had a thickness of 15 mm after the rolling process, and was cooled from 800°C to 400°C in the cooling process. The cooling rate was 1.0 m ³ /m ² ·min in cooling zones Z (1) to (5) and 2.0 m ³ /m ² ·min in cooling zones Z (6) to (10). .

In addition, the carbon equivalent of the steel sheet, rolling process data, and cooling process data are input (training data) to the surface hardness prediction model 121, and machine learning is performed using the reference surface hardness as an output (label). A finished surface layer hardness prediction model 121 was constructed.

The rolling process data was data relating to the equipment specifications of the rolling equipment 30, the pass schedule, the temperature of the steel sheet, the conveying speed, and the dimensions of the steel sheet. Further, the cooling process data was data relating to equipment specifications, steel plate temperature, conveying speed, and steel plate dimensions in the cooling process.

In order to evaluate the learned surface layer hardness prediction model 121, 100 steel sheets were manufactured in the steel plate production line 1 while giving various condition parameters as inputs (test data) to the learned surface layer hardness prediction model 121, and from there, A sample piece was taken and its surface layer hardness was measured. As a sample piece, the longitudinal end portion of the steel sheet after the cooling process was sheared by a shearing machine. The surface layer hardness of the sample piece is determined by measuring the Vickers hardness at a depth of 1 mm from the surface of the sample piece with a load of 10 kgf. did.

FIG. 11 shows a table for explaining examples and comparative examples using a steel sheet production line to which the surface layer hardness prediction model according to one embodiment of the invention is applied. In the table shown in the same figure, for each of Comparative Examples 1 to 5 and Examples 1 to 6, the presence or absence of application of the surface hardness prediction model 121, the presence or absence of use of various input parameters, the presence or absence of occurrence of red scale, and the predicted value Variation in (surface layer hardness) and yield are shown. The surface layer hardness is the Vickers hardness measured at a position 1 mm below the surface with a measurement load of 10 kgf. In the table, T _LAST is the steel sheet surface temperature when the final descaling is applied, and t _LAST is the elapsed time from the application of the final descaling to the end of rolling.

Comparative Example 1 shows a case where a steel plate is manufactured in a conventional steel plate manufacturing line to which the surface layer hardness prediction model 121 is not applied. As shown in the table, in Comparative Example 1, the surface layer hardness of the steel sheet varied greatly, and the surface layer hardness of the steel sheet with red scale was particularly high, resulting in a decrease in yield.

Comparative Examples 2 to 5 show cases where steel sheets are manufactured in the steel sheet manufacturing line 1 to which the surface layer hardness prediction model 121 is applied. In Comparative Examples 2 to 5, the input parameters of the steel plate surface temperature (T _LAST ) when the final descaling was applied and the elapsed time (t _LAST ) from the application of the final descaling to the end of rolling were not used. not In Comparative Example 2, among the input parameters, the carbon equivalent and rolling process parameters were used, while the cooling process parameters were not used. Also, in Comparative Example 3, while the carbon equivalent and rolling process parameters were used, the rolling process parameters were not used. Also, in Comparative Example 4, the rolling process parameter and the cooling process parameter were used, while the carbon equivalent was not used. Furthermore, in Comparative Example 5, the same input parameters as in Comparative Example 2 were used.

As shown in the table, in Comparative Examples 2 to 4, the surface layer hardness of the steel sheets varied greatly, and the yield decreased. Also, in Comparative Example 5, although the input parameters were the same as in Comparative Example 2, red scale did not occur. Therefore, the variation in the surface layer hardness of Comparative Example 5 was smaller than that of Comparative Example 2, but the yield was still low.

Examples 1 and 2 show cases where steel sheets are manufactured in the steel sheet manufacturing line 1 to which the surface layer hardness prediction model 121 is applied. However, in Examples 1 and 2, among the rolling process data, the steel sheet surface temperature when the final descaling is applied and the elapsed time from the application of the final descaling to the end of rolling are not included in the input. As shown in the table, in Example 1, red scale was observed, but the variation in surface layer hardness was small and the yield was good. This means that the surface layer hardness is predicted by the surface layer hardness prediction model 121 in consideration of the scale thickness and the possibility of occurrence of red scale. On the other hand, in Example 2, since the conditions were such that red scale was not generated, the variation was smaller than in Example 1, and the yield was improved.

Examples 3 and 4 show cases where steel sheets are manufactured in the steel sheet manufacturing line 1 to which the surface layer hardness prediction model 121 is applied. However, in Example 3, among the input parameters, the steel plate surface temperature (T _LAST ) when final descaling is applied is not used, and in Example 4, among the input parameters, final descaling is applied. The elapsed time (t _LAST ) from the end of rolling was not used. As shown in the table, in Examples 3 and 4, the prediction accuracy of the scale thickness and the red scale was improved, and the prediction accuracy of the surface layer hardness was further improved. Improved.

Examples 5 and 6 show cases where steel sheets are manufactured while predicting and controlling the surface layer hardness using all input parameters in the steel sheet manufacturing line 1 to which the surface layer hardness prediction model 121 is applied. As shown in the table, in Examples 5 and 6, similarly, the scale thickness immediately before the start of cooling that determines the surface layer hardness and the presence or absence of red scale generation can be accurately grasped, and the hardness variation is further reduced. Yield was 100%. As a result, it was found that the surface layer hardness prediction model 121 was able to predict the surface layer hardness with higher accuracy.

Each of the above embodiments is an example for explaining the present invention, and is not intended to limit the present invention only to these embodiments. The present invention can be embodied in various forms without departing from the gist thereof.

For example, in the methods disclosed herein, steps, actions or functions may be performed in parallel or in a different order without conflicting results. The steps, acts and functions described are provided as examples only and some of the steps, acts and functions may be omitted or combined together without departing from the scope of the invention. one, and other steps, operations, or functions may be added.

In addition, although various embodiments are disclosed in this specification, specific features (technical matters) in one embodiment may be added to other embodiments or Certain features in the embodiments can be substituted and such forms are included in the gist of the invention.

DESCRIPTION OF SYMBOLS 1... Steel plate manufacturing line 10... Control command device 110... Process computer 110a... Process computer for heating equipment 110b... Process computer for rolling equipment 110c... Process computer for cooling equipment 120... Host computer 121... Surface layer hardness prediction model 122... Surface layer hardness control Part 20 Heating equipment 30 Rolling equipment 40 Descaling equipment 41 Injection nozzle 50 Radiation thermometer 60 Cooling equipment 61 Draining purge 62 Flow control valve 63 Injection nozzle 64 Draining roll

Claims (16)

A surface layer hardness prediction model adapted to a steel plate manufacturing line for manufacturing a steel plate,
The carbon equivalent of the steel sheet, the rolling process data in the rolling process for the steel sheet, and the cooling process data in the cooling process for the steel sheet are input, and the material test data indicating the surface layer hardness of the steel sheet is output. Machine learning was performed using a predetermined machine learning algorithm to predict surface hardness,
Surface layer hardness prediction model. The rolling process data includes the surface temperature of the steel sheet when the final descaling is applied, and the elapsed time from the application of the final descaling to the end of rolling,
The surface layer hardness prediction model according to claim 1. The rolling process data further includes at least one of specifications of rolling equipment that performs the rolling process, conveying speed of the steel sheet, dimensions of the steel sheet, and descaling information on descaling for the steel sheet.
The surface layer hardness prediction model according to claim 2. The cooling process data includes the cooling start temperature of the steel sheet, the cooling stop temperature, and the calculated cooling rate of the steel sheet surface layer,
The surface layer hardness prediction model according to any one of claims 1 to 3. The cooling process data further includes at least one of the specifications of the cooling equipment that performs the cooling process, the conveying speed of the steel plate, and the dimensions of the steel plate.
The surface layer hardness prediction model according to claim 4. The material test data is a value measured by any one of Vickers hardness measurement method, Brinell hardness measurement method, and Rockwell hardness measurement method,
The surface layer hardness prediction model according to claim 1. the predetermined machine learning algorithm is at least one of neural network, decision tree learning, random forest, and support vector regression;
The surface layer hardness prediction model according to claim 1. A control command device for controlling a steel plate manufacturing line for manufacturing a steel plate,
The carbon equivalent of the steel sheet, the rolling process data in the rolling process for the steel sheet, and the cooling process data in the cooling process for the steel sheet are input, and the material test data indicating the surface layer hardness of the steel sheet is output. a surface hardness prediction model machine-learned using a predetermined machine-learning algorithm to predict hardness;
a surface hardness control unit that changes and resets the cooling process data based on the predicted value output from the surface hardness prediction model;
Control commander. The surface layer hardness prediction model outputs the predicted value based on the carbon equivalent, the rolling process data, and the cooling process data during operation of the steel sheet production line,
The surface layer hardness control unit changes and resets the cooling process data when the predicted value output from the surface layer hardness prediction model does not satisfy a predetermined target value.
9. The control command device according to claim 8. The surface layer hardness control unit changes and resets operation parameters related to the injection amount of cooling water and/or the conveying speed of the steel plate in the cooling process data,
10. The control command device according to claim 9. a control command device;
A rolling facility that rolls the heated steel sheet to a predetermined thickness while descaling scales formed on the surface of the heated steel sheet under the control of the control command device;
A steel plate manufacturing line comprising a cooling facility that performs cooling treatment at a predetermined cooling rate on the rolled steel plate under the control of the control command device,
The control command device includes a surface layer hardness prediction model,
The surface layer hardness prediction model is
The carbon equivalent of the steel sheet, the rolling process data in the rolling process for the steel sheet, and the cooling process data in the cooling process for the steel sheet are input, and the material test data indicating the surface layer hardness of the steel sheet is output. machine-learned using a predetermined machine-learning algorithm to predict hardness,
Steel plate production line. A method for predicting and controlling the surface layer hardness of the steel sheet using the surface layer hardness prediction model according to any one of claims 1 to 7,
The machine-learned surface layer hardness prediction model predicts the surface layer hardness of the steel sheet based on the carbon equivalent, the rolling process data, and the cooling process data during operation of the steel sheet production line. including outputting
Method. changing and resetting the cooling process data based on the output predicted value;
further comprising performing the cooling process based on the reset cooling process data;
13. The method of claim 12. The resetting includes changing operation parameters related to the injection amount of cooling water and / or the conveying speed of the steel plate in the cooling process data,
14. The method of claim 13. A steel plate manufacturing method comprising a rolling step of rolling a heated steel plate to a predetermined thickness and a cooling step of cooling the rolled steel plate at a predetermined cooling rate,
The cooling data is changed and reset based on the predicted value output from the surface layer hardness prediction model according to claim 1, and the operating parameters in the cooling process are changed based on the reset cooling data. including to
Steel plate manufacturing method. The steel plate manufacturing method according to claim 15, wherein the steel plate is at least one of a thick steel plate, a steel plate for line pipes, and a steel plate for civil engineering and construction.

Images