Method for generating a steel sheet material prediction model, material prediction method, manufacturing method, and manufacturing apparatus

Abstract

The method for generating a steel sheet material prediction model according to the present application is a method for generating a steel sheet material prediction model in a steel sheet cooling device, the steel sheet cooling device including a water cooling device that cools a steel sheet by spraying cooling water on the heated steel sheet, and a temperature measuring device that measures the surface temperature of the steel sheet during cooling, wherein the method for generating a steel sheet material prediction model includes the step of generating a steel sheet material prediction model after the steel sheet passes through the cooling device by machine learning using a plurality of learning data sets that include, as input performance data, surface temperature information data sets including measured data of the surface temperature at a reference point set in advance on the steel sheet and measured data of the surface temperature at a reference point set based on the reference point, and, as output performance data, material information of the steel sheet after passing through the cooling device at a position corresponding to the reference point on the steel sheet corresponding to the input performance data.

Description

Technical Field

[0001] This invention relates to a method for generating a material prediction model for steel plates, a material prediction method, a manufacturing method, and manufacturing equipment. Background Technology

[0002] The required mechanical properties of steel plates, especially strength and toughness, have become particularly stringent in recent years. In steel plate manufacturing, in addition to direct quenching of the rolled high-temperature steel plate, quenching is also performed using heat treatment equipment, or cooling is stopped at a specified temperature during direct quenching or heat treatment, thereby ensuring the required characteristics of the steel plate. At this point, it is important to ensure that the cooling rate and cooling stop temperature are uniform across the entire width and length of the steel plate, thus guaranteeing uniform properties.

[0003] Here, quenching refers to a heat treatment method in which steel plates heated to temperatures above the Ac3 transformation point (the end temperature of the austenitic transformation after hot rolling), or steel plates that have been cooled after rolling and then reheated in a furnace to temperatures above the Ac3 transformation point, are rapidly cooled to temperatures below the martensitic transformation initiation temperature (Ms point) using a cooling device. Quenching is widely used, especially as a manufacturing method for high-strength steel plates. Furthermore, quenching without cooling / reheating the rolled steel plate is called direct quenching.

[0004] On the other hand, in order to generate internal structures such as ferrite, pearlite, and bainite during the cooling process, a heat treatment process is also used, in which the heated steel sheet is cooled to around 500°C. The technique of cooling the hot-rolled steel sheet and stopping the cooling at around 500°C is called accelerated cooling and is now a common heat treatment process. Furthermore, a technique has been developed where the hot-rolled steel sheet is temporarily cooled to room temperature, then reheated, and during the cooling process using a cooling device that cools the reheated steel sheet, the cooling is stopped at around 500°C, thereby controlling the complex internal structure of ferrite, pearlite, and bainite.

[0005] When using a cooling process to control the internal structure of steel sheets and produce steel sheet products, in-plane material variations can sometimes become a problem. Typically, during quality assurance related to the steel sheet material, mechanical test values ​​are obtained by taking test pieces from a portion of the manufactured steel sheet. However, in the presence of in-plane material variations as described above, the location of the test piece may not necessarily represent the material of the entire steel sheet. Therefore, there is room for improvement regarding the uniformity of the material across the entire width and length of the steel sheet.

[0006] In response, Patent Document 1 describes a material assurance system in the cooling equipment of a thick steel plate manufacturing production line. This system includes a temperature collection unit and performs material assurance for the thick steel plate based on the overall temperature mapping of the steel plate. The temperature collection unit uses temperature measurement units located at multiple positions, such as upstream of the water cooling device, the cooling start position, and the cooling stop position, to collect temperature data of the steel plate being cooled.

[0007] On the other hand, while Patent Document 2 focuses on hot-rolled steel strip, it describes a method where thermometers are installed on the steel strip manufacturing production line, including the cooling process. For each grid dividing the steel strip product along its length, width, and thickness directions, material property values ​​for each grid are estimated while referencing similarity to past operational data. Furthermore, Patent Document 2 describes using estimated values ​​as internal temperature information of the steel strip, derived using a heat transfer model based on measured surface temperatures of the steel strip.

[0008] Furthermore, Patent Document 3 describes a method for dividing the length of a hot-rolled steel strip into multiple regions during the cooling process, and calculating the temperature history of the steel strip for each region based on the cooling conditions for each region. Patent Document 3 also describes a method for predicting the microstructure state of each region using the calculated temperature history, and for predicting the material properties of the steel strip using data related to past manufacturing practices that approximate the predicted microstructure state. Moreover, Patent Document 3 also describes a method that assumes uniform temperature distribution along the thickness direction when calculating the temperature history, and a method for calculating the temperature distribution along the thickness direction by solving the heat conduction equation relative to the thickness direction.

[0009] Existing technical documents

[0010] Patent documents

[0011] Patent Document 1: Japanese Patent No. 5589260

[0012] Patent Document 2: Japanese Patent No. 5924362

[0013] Patent Document 3: Japanese Patent Application Publication No. 2012-171001 Summary of the Invention

[0014] The problem that the invention aims to solve

[0015] However, the system described in Patent Document 1 collects temperature data of the steel plate during cooling across its entire width and length. Even with the collection of temperature data from the entire surface of the steel plate, it is not always possible to accurately predict the material composition of the steel plate. In particular, the temperature data collected by the temperature collection unit is measurement data related to the surface of the steel plate, and therefore may not always represent the internal temperature history of the steel plate. There is room for improvement in the accuracy of predicting the material composition of the steel plate.

[0016] On the other hand, the method described in Patent Document 2 involves setting a grid that divides the thickness direction of the steel strip and using a heat transfer model to estimate the internal temperature of the steel strip. However, the heat transfer model in the cooling process of the steel strip is usually a method of solving a one-dimensional heat conduction equation in the thickness direction. If heat transfer analysis is to be performed in the length and width directions of the steel strip, the computer load becomes too large, making it difficult to use as a practical temperature estimation method. On the other hand, while it is practically feasible to ignore thermal movement in the length and width directions of the steel strip in the central part of the in-plane area, thermal movement in the length and width directions occurs near the front and rear ends or the ends in the width direction of the steel strip. Therefore, even if the surface temperature of the steel strip is the same, the temperature distribution in the thickness direction in the central part of the in-plane area of ​​the steel strip is different from the temperature distribution in the thickness direction near the front and rear ends or the ends in the width direction. Therefore, there is room for improvement in the one-dimensional heat transfer model in the thickness direction with surface temperature as the boundary condition, in terms of accurately estimating the internal temperature history of the steel strip.

[0017] Furthermore, in the method described in Patent Document 3, the thickness direction of the steel strip is divided, and the temperature history inside the steel strip is estimated by solving the heat conduction equation in the thickness direction. Based on the estimated temperature history, the microstructure state of each region is predicted. According to the method described in Patent Document 3, the correspondence between the prediction of not only the temperature history but also the microstructure inside the steel strip and the material inside the steel strip becomes clear. However, as mentioned above, there is room for improvement in the accuracy of material prediction when thermal movement occurs in the length and width directions of the steel strip.

[0018] This invention was made in view of the above-mentioned problems, and its object is to provide a method for generating a material prediction model for steel plates, capable of generating a high-precision material prediction model for steel plates after passing through a cooling device. Another object of this invention is to provide a method for predicting the material of steel plates, capable of predicting the material information of steel plates after passing through a cooling device with high precision. Furthermore, another object of this invention is to provide a method and equipment for manufacturing steel plates, capable of manufacturing steel plates with excellent material uniformity.

[0019] Technical solutions for solving the problem

[0020] The method for generating a material prediction model for steel plates involved in this invention is a method for generating a material prediction model for steel plates in a steel plate cooling device. The steel plate cooling device includes: a water cooling device for cooling the steel plate by spraying cooling water onto the heated steel plate; and a temperature measuring device for measuring the surface temperature of the steel plate during the cooling process. The method for generating the material prediction model for steel plates includes the following steps: generating a material prediction model for the steel plate after passing through the cooling device by using machine learning with multiple learning data. The multiple learning data includes a surface temperature information dataset including measured surface temperature data at a pre-set reference point on the steel plate and measured surface temperature data at a reference point set based on the reference point as input performance data, and the material information of the steel plate after passing through the cooling device at the position corresponding to the reference point on the steel plate corresponding to the input performance data is used as output performance data.

[0021] The reference point can be set at least one position relative to the reference point in the length direction of the steel plate, and at least one position in the width direction.

[0022] The material prediction model may include attribute information parameters selected from the attribute information of the steel plate as the input performance data.

[0023] The material prediction model may include at least one operational performance data selected from the operational performance data of the water cooling device as the input performance data.

[0024] As the machine learning method, machine learning methods selected from neural networks, decision tree learning, random forests, and support vector regression can be used.

[0025] The material prediction method for steel plates involved in this invention is a method for predicting the material of steel plates in a steel plate cooling device. The steel plate cooling device includes: a water cooling device for cooling the steel plate by spraying cooling water onto the heated steel plate; and a temperature measuring device for measuring the surface temperature of the steel plate during the cooling process. The material prediction method for steel plates includes the following steps: using a material prediction model generated by machine learning to predict the material information of the steel plate after passing through the cooling device. The machine learning takes a dataset of surface temperature information, including the surface temperature at a pre-set prediction reference point on the steel plate and the surface temperature at a prediction reference point set based on the prediction reference point, as input data, and takes the material information of the steel plate after passing through the cooling device at the position corresponding to the prediction reference point on the steel plate as output data.

[0026] The steel plate manufacturing method of the present invention includes the following steps: using the steel plate material prediction method of the present invention, determining whether the material of the steel plate after passing through the cooling equipment is qualified.

[0027] The steel plate manufacturing equipment involved in this invention includes: a steel plate cooling device, comprising a water cooling device and a temperature measuring device, wherein the water cooling device cools the steel plate by spraying cooling water onto the heated steel plate, and the temperature measuring device measures the surface temperature of the steel plate during the cooling process; and a material prediction unit, which outputs the material information of the steel plate after passing through the cooling device. The material prediction unit outputs the material information of the steel plate using a machine learning model. The machine learning model takes a dataset of surface temperature information, including the surface temperature at a pre-set prediction reference point on the steel plate and the surface temperature at a prediction reference point set based on the prediction reference point, as input data, and takes the material information of the steel plate after passing through the cooling device at the position corresponding to the prediction reference point on the steel plate as output data.

[0028] Invention Effects

[0029] According to the method for generating a material prediction model for steel plates of the present invention, a material prediction model for steel plates that can predict the material information of steel plates after passing through a cooling device with high accuracy can be generated. Furthermore, the material prediction method for steel plates of the present invention can predict the material information of steel plates after passing through a cooling device with high accuracy. Additionally, the steel plate manufacturing method and manufacturing equipment of the present invention can manufacture steel plates with excellent material uniformity. Attached Figure Description

[0030] Figure 1 The figure shows an example of a cooling device for a steel plate, as an embodiment of the present invention, configured in a steel plate manufacturing equipment that performs online and offline heat treatment processes.

[0031] Figure 2 This is a diagram showing the structure of a cooling device for a steel plate, as an embodiment of the present invention, applicable to an offline heat treatment process.

[0032] Figure 3 It means Figure 2 The diagram shows the structure of the water-cooling device.

[0033] Figure 4 It means Figure 2 The diagram shown is a block diagram of the structure of the control computer.

[0034] Figure 5 This diagram illustrates the method of establishing a correspondence between the positional information along the length of a steel plate and its surface temperature.

[0035] Figure 6 This is a diagram showing an example of measuring the surface temperature of a steel plate.

[0036] Figure 7 This is a diagram representing an example of a reference point.

[0037] Figure 8 This is a diagram representing an example of a reference point.

[0038] Figure 9 This is a block diagram showing the structure of the material prediction model generation unit as one embodiment of the present invention.

[0039] Figure 10 This is a diagram illustrating the operation of the material prediction section of a steel plate, which is an embodiment of the present invention.

[0040] Figure 11 This diagram illustrates an example of surface temperature information obtained through a surface temperature measuring device configured at the inlet and outlet of a cooling device.

[0041] Figure 12 This is a diagram showing an example of yield stress and tensile strength.

[0042] Figure 13 This is a diagram showing the inlet temperature, outlet temperature, tensile strength, and yield stress of the cooling device.

[0043] Figure 14 This is a graph illustrating an example of the relationship between actual and predicted values ​​of tensile strength and yield stress.

[0044] Figure 15 This is a diagram illustrating the structure of the hot rolling production line in the embodiment. Detailed Implementation

[0045] Hereinafter, with reference to the accompanying drawings, a method for generating a material prediction model for a steel plate, a material prediction method, a manufacturing method, and a cooling device, which are embodiments of the present invention, will be described.

[0046] [Steel plate manufacturing process]

[0047] First, refer to Figure 1 (a) and (b) describe the manufacturing process of the steel plate as one embodiment of the present invention.

[0048] Figure 1 (a) is a diagram illustrating an example of a cooling device for a steel plate, as an embodiment of the present invention, configured on a hot rolling production line performing an online heat treatment process. Figure 1As shown in (a), in this example, firstly, the slab, which is a casting, is heated to a specified heating temperature using heating equipment in a hot rolling production line, and then reversibly rolled using one or two rolling mills. Next, the steel plate rolled to the specified dimensions by the rolling mill is conveyed from the rolling mill to a cooling equipment while maintaining a high temperature. In the cooling equipment, the steel plate is cooled to a pre-set cooling stop temperature by accelerated cooling, and then cooled to near room temperature in a cooling bed (a site for air-cooling the steel plate to near room temperature). Then, the steel plate is transferred to a finishing equipment, where shape correction, defect inspection, and product collection are performed as needed. Alternatively, there are cases where steel plates cooled using cooling equipment in an online heat treatment process are reheated to a tempering process at temperatures of 140–680°C before being conveyed to a cooling bed. In the tempering process, the steel plate is not cooled using cooling equipment after reheating, which differs from the offline heat treatment process described below.

[0049] In this embodiment, a test piece collection device is provided in the finishing equipment. This device collects test piece samples for material inspection from a steel plate cooled to near room temperature. The test piece collection device is a device that collects test piece samples from a position within the surface of the steel plate pre-set according to product standards, specifications, etc. As the test piece collection device, a laser cutting machine or a shearing machine is used to further process the collected test piece samples into a shape corresponding to the test items of the material test (for example, a JIS 4 test piece if it is a tensile test).

[0050] Figure 1 (b) is a diagram illustrating an example of a heat treatment apparatus equipped with a cooling device for a steel plate, as described in one embodiment of the present invention, used in an offline heat treatment process. The offline heat treatment process utilizes steel plates rolled to a specified size in a hot rolling production line. The hot rolling production line has the same equipment structure as described above. However, in an online heat treatment process, it is not necessarily required to perform an online heat treatment process in a hot rolling production line. Figure 1 (b) As shown, in this example, the steel sheet rolled by the hot rolling production line is cooled to near room temperature on a cooling bed and then conveyed to a pretreatment device. The pretreatment device is used before the offline heat treatment process to perform shape correction, cutting to specified dimensions, and sometimes descaling based on shot peening. However, the pretreatment in the pretreatment device is not a mandatory step. Afterward, the steel sheet is transferred to a heat treatment device to perform the offline heat treatment process. The heat treatment device has heating equipment, which heats the steel sheet to a specified temperature and then cools it using cooling equipment. The heat-treated steel sheet is air-cooled to near room temperature on a cooling bed and then conveyed to a finishing device. The finishing device is the same as that used in the online heat treatment process.

[0051] [Cooling equipment]

[0052] Next, refer to Figure 2 The structure of a cooling device for a steel plate, which is one embodiment of the present invention, will be described.

[0053] Figure 2 This is a diagram illustrating the structure of a cooling device for a steel plate, as an embodiment of the present invention, suitable for offline heat treatment processes. (See diagram for example.) Figure 2 As shown, the offline heat treatment equipment 1 includes, as its main components, a heating furnace 2 for heating a steel plate S at a temperature below 100°C to a specified temperature, a cooling device 3 for cooling the steel plate S heated in the heating furnace 2, and a control computer 10 for controlling the operation of the offline heat treatment equipment 1 including the cooling device 3.

[0054] The cooling device 3 includes a water-cooling device 4 that sprays cooling water W onto the steel plate S, and a temperature measuring device 5 that measures the surface temperature of the steel plate S during the cooling process. In this embodiment, the temperature measuring device 5 includes a temperature measuring device 51 located upstream of the water-cooling device 4, a temperature measuring device 52 located midway through the water-cooling device 4, and a temperature measuring device 53 located downstream of the water-cooling device 4. However, it is sufficient to install a temperature measuring device at at least one of the upstream, midway, and downstream locations of the water-cooling device 4. Furthermore, Figure 2 The water cooling device 4 shown is a device that includes water cooling nozzles 41a and 41b and restraint devices (restraint rollers 42a and 42b), but in this embodiment, the restraint devices are not necessary.

[0055] A steel plate S is loaded into a heating furnace 2. This steel plate S is hot-rolled to a specified thickness (e.g., 30 mm) and width (e.g., 2000 mm) in a hot rolling production line located in a different location from the offline heat treatment equipment 1, and then cooled to approximately room temperature. The steel plate S is then heated to a specified temperature (e.g., 910°C) in the heating furnace 2. The steel plate S, extracted from the heating furnace 2, is conveyed by multiple table rollers 6 located on the outlet side of the heating furnace 2 and sent to a cooling device 3. Figure 2 In order to explain the structure of the cooling device 3 in detail, the cooling device 3 is depicted as larger than the heating furnace 2. In reality, the length of the heating furnace 2 is about 60 to 80 m, the length of the cooling device 3 is about 20 to 25 m, and the length of the steel plate S is longer than the length of the cooling device 3. Therefore, during the stage where the front end of the steel plate S passes through the cooling device 3, the stable part and the tail end of the steel plate S are located inside the heating furnace 2.

[0056] Generally, in the offline heat treatment equipment 1, the steel plate S is extracted from the heating furnace 2 and conveyed at a roughly constant speed until the cooling by the cooling device 3 is completed. Therefore, the temperature difference at the beginning of cooling at the front and rear ends of the steel plate S is small. That is, when the heating temperature of the steel plate S is set to T0, the distance from the heating furnace 2 to the water cooling device 4 of the cooling device 3 is set to L0, and the conveying speed of the steel plate S is set to V0, the front end of the steel plate S is extracted at temperature T0 and cooled over a cooling time L0 / V0. In the offline heat treatment equipment 1, because the distance L0 from the heating furnace 2 to the water cooling device 4 is short, even if the front end of the steel plate S is extracted from the heating furnace 2 and reaches the inlet of the water cooling device 4, the rear end of the steel plate S is also maintained at temperature T0 within the heating furnace 2. Therefore, the rear end of the steel plate S is also extracted at temperature T0 in the same way as the front end and cooled over a cooling time L0 / V0, thus maintaining a constant cooling start temperature throughout the entire length of the steel plate S. In this way, the offline heat treatment equipment 1 is advantageous for manufacturing steel plates with small in-plane temperature deviations for thin steel plates that are prone to temperature drop due to cooling, and thus has the advantage of easily ensuring the uniformity of the steel plate material.

[0057] The cooling equipment used in the online heat treatment process differs from that used in the offline heat treatment process. The heating furnace 2 is not positioned close to the cooling equipment. Therefore, during the stage where the front end of the steel plate S passes through the cooling equipment 3, the stable part and the tail end of the steel plate S are in a state of reflux. Consequently, regarding the reflux time until the start of cooling, the tail end of the steel plate S is longer than the front end. When the length of the steel plate S is set to L and the conveying speed of the steel plate S is set to V, a reflux time difference of L / V occurs between the tail end and the front end. Even if the temperature of the rolled steel plate is uniform, because the tail end refluxes additionally due to the reflux time difference, a difference in the cooling start temperature occurs between the front end and the tail end, easily resulting in a temperature distribution along the length direction within the surface of the steel plate S. The online heat treatment equipment creates conditions where the material within the surface of the steel plate S easily becomes non-uniform.

[0058] [Water cooling device]

[0059] Next, refer to Figure 3 The structure of water cooling device 4 will be described.

[0060] Figure 3 It means Figure 2 A diagram showing the structure of the water-cooling device 4. (See diagram for reference.) Figure 3As shown, the water-cooling device 4 includes multiple water-cooling nozzles 41a and 41b arranged in pairs along the conveying direction of the steel plate S in the vertical direction. Water-cooling nozzles 41a spray cooling water W downwards towards the upper surface of the steel plate S. On the other hand, water-cooling nozzles 41b spray cooling water W upwards towards the lower surface of the steel plate S. For example, water-cooling nozzles 41a and 41b constitute an upper and lower pair of water-cooling nozzles; a cooling area defined by this pair is called a cooling zone, and a collection of one or more cooling zones is called a region. Figure 2 In the example shown, the cooling area (the area cooled by water cooling device 4) consists of 6 cooling zones. Figure 3 In the example shown, the cooling area consists of four cooling zones. However, even if the number of cooling zones is not these, the effect of the invention will not be compromised. Alternatively, the cooling area may consist of multiple cooling zones, separated by air-cooled zones without water-cooling nozzles.

[0061] As water-cooled nozzles 41a and 41b, it is preferable to have cooling water flow adjustment valves, which can adjust the amount of cooling water W sprayed toward the steel plate S. This allows adjustment of the flow rate of cooling water W sprayed toward each cooling zone. Furthermore, it is preferable to be able to adjust the amount of cooling water W sprayed toward the steel plate S from the upper and lower paired water-cooled nozzles 41a and 41b to different values. The amount of cooling water W sprayed from each of these water-cooled nozzles is controlled by the water-cooled flow control device 43 for each water-cooled nozzle based on the water flow set value set by the control computer 10.

[0062] The operating parameters of the water cooling device 4 include the volume of cooling water W sprayed from at least one pair of water cooling nozzles 41a, 41b (cooling water volume) and the speed of the steel plate S conveyed by the table roller 6 (conveyor speed). A higher volume of cooling water increases the cooling rate and temperature drop of the steel plate S. Conversely, a lower conveyor speed of the steel plate S increases the temperature drop. Furthermore, by combining these operating parameters, the cooling stop temperature and cooling rate are controlled as cooling conditions to obtain the desired material.

[0063] In addition to these, the operating parameters of the water cooling device 4 also include the balance of cooling water volume in each cooling zone (e.g., increasing the cooling water volume in the upstream cooling zone and decreasing the cooling water volume in the downstream cooling zone). This is because the cooling rate can be controlled according to the temperature range of the steel plate S. The balance of cooling water volume in each cooling zone can be represented by the ratio of the amount of cooling water sprayed in each cooling zone. Furthermore, the number of cooling zones spraying cooling water W can also be varied. Depending on the number of cooling zones used, different cooling stop temperatures can be controlled while maintaining the same cooling rate. The cooling zones used are determined by labels or values ​​used to judge the use / non-use of each cooling zone, and these labels or values ​​can also be used as operating parameters of the water cooling device 4.

[0064] The greater the water volume, the higher the thermal conductivity of the cooling water W ejected from the water-cooling nozzles 41a and 41b. Therefore, the cooling rate can be adjusted by changing the cooling water volume, thereby controlling the material properties of the steel plate S. The water-cooling nozzles 41a and 41b can be slit-type nozzles or flat spray nozzles capable of uniformly spraying a large flow rate of cooling water W across the width of the steel plate S. Alternatively, multi-hole spray nozzles or mist nozzles can also be used.

[0065] Furthermore, it is not necessary to use nozzles capable of adjusting the cooling water volume for each water-cooling nozzle as water-cooling nozzles 41a and 41b. This is because, when the water-cooling device 4 has multiple cooling zones, the cooling conditions can be changed by altering the number of cooling zones from which the cooling water W is sprayed.

[0066] The water cooling device 4 sometimes includes water cooling nozzles 41a and 41b, along with a constraint device having at least a pair of constraint rollers that constrain the steel plate S during cooling. By constraining the strain that may occur in the steel plate S during water cooling, it is advantageous to maintain the uniformity of cooling and ensure the uniformity of the material composition of the steel plate S. Figure 3 The structure of the restraint device is described.

[0067] The restraining device is located in the cooling zone, adjacent to the water-cooling zone. The restraining rollers 42a and 42b constituting the restraining device are arranged such that their axial direction is approximately perpendicular to the conveying direction of the steel plate S, so that the steel plate S is restrained by a pair of upper and lower rollers. During the cooling process by the water-cooling device 4, the steel plate S experiences strain due to thermal contraction and phase transformation. The restraining rollers 42a and 42b are designed to restrain the deformation of the steel plate S so that it does not buckle due to such strain. This suppresses the deformation of the steel plate S during cooling, preventing uneven cooling by the cooling water W and resulting in inhomogeneous material within the surface of the steel plate S. In addition to restraining the steel plate S, the restraining device also functions as a dewatering roller. This prevents water splashing on the upper surface of the steel plate S downstream of the restraining device from interfering with temperature measurement when the temperature measuring device 5 is installed in the cooling equipment 3.

[0068] In addition, Figure 3 In the water-cooling device 4 shown, a dewatering purging nozzle 7 is provided on the outlet side of the constraint device at the downstream end. The dewatering purging nozzle 7 sprays dewatering purging air 7a at an angle toward the constraint roller 42a, so that the cooling water W leaking from the gap formed between the constraint roller 42a and the steel plate S does not flow further downstream. It has the effect of suppressing the expansion of temperature deviation of the steel plate S by using dewatering purging air 7a, thereby suppressing the deterioration of the material uniformity of the steel plate S.

[0069] [Control computer]

[0070] Return to Figure 2 In addition to obtaining information such as the heating temperature, thickness, width, and weight of the steel plate S from the host computer 11, the control computer 10 also obtains information related to the target range of cooling stop temperature (target cooling stop temperature) and the target range of cooling rate (target cooling rate) required to obtain the desired material. Then, the control computer 10 calculates the operating conditions for achieving these conditions and determines the operating parameters of each device in the water cooling system 4.

[0071] Figure 4 It means Figure 2 The diagram shows the structure of the control computer 10. Figure 4 As shown, the control computer 10 obtains the attribute information of the steel plate S, which is the object of heat treatment, from the host computer 11. The attribute information of the steel plate S includes not only dimensional information such as thickness, width, length, and weight, but also information related to its composition (C content, Si content, Mn content, Cr content, Mo content, etc.) and target values ​​for the mechanical properties of the heat-treated steel plate S (yield stress, tensile strength, elongation, toughness, hardness, etc.).

[0072] In addition to the property information of the steel plate S, the control computer 10 also obtains information related to the target cooling stop temperature and the target cooling rate from the host computer 11. Then, the control computer 10 performs heat transfer calculations based on an internal model in the water cooling condition calculation unit 10a, and determines the operating conditions of the water cooling device 4, including the flow rate of the cooling water W from the water cooling nozzles 41a and 41b in the cooling area, the cooling area of ​​the sprayed cooling water W, and the conveying speed of the steel plate S in the cooling equipment 3, in a manner that satisfies the target cooling stop temperature and the target cooling rate set as cooling conditions.

[0073] The operating conditions of the water cooling device 4, set by the water cooling condition calculation unit 10a, are sent to the water cooling flow control device 43. The water cooling flow control device 43 generates commands for the operating pressure of the cooling water pump, the number of working units, the number of manifolds upstream of the piping system installed on the water cooling nozzles 41a and 41b, the opening degree of the flow regulating valve, and the rotational speed of the motor driving the table rollers 6, and sets the operating conditions of the water cooling device 4. Furthermore, when the cooling equipment 3 is equipped with constraint rollers 42a and 42b and water removal purging nozzles 7 as constraint devices, the control computer 10 also sets their operating conditions.

[0074] [Temperature measuring device]

[0075] Return to Figure 2 The cooling device 3 of this embodiment includes a temperature measuring device 5 for measuring the surface temperature of the steel plate S during the cooling process. The temperature measuring device 51, located upstream of the water cooling device 4, is positioned approximately 1-2 m away from the inlet of the water cooling device 4. This is for measuring the cooling start temperature during the heat treatment process of the steel plate S. Conversely, the temperature measuring device 53, located downstream of the water cooling device 4, is positioned approximately 5-10 m away from the outlet of the water cooling device 4. This is for measuring the cooling stop temperature during the heat treatment process of the steel plate S. Furthermore, when the temperature measuring device 51 is not located upstream of the water cooling device 4, the inlet of the cooling device 3 coincides with the inlet of the water cooling device 4. Conversely, when the temperature measuring device 53 is not located downstream of the water cooling device 4, the outlet of the cooling device 3 coincides with the outlet of the water cooling device 4.

[0076] The temperature measuring device 5 used in this embodiment has the function of measuring the surface temperature of the steel plate S during the cooling process. The cooling process refers to the temperature change of the steel plate S that occurs between the inlet and outlet of the cooling device 3. Therefore, the temperature measuring device 5 can be installed at any position between the inlet and outlet of the cooling device 3. Figure 2In the example shown, in addition to the temperature measuring device 51 located upstream of the water cooling device 4 and the temperature measuring device 53 located downstream of the water cooling device 4, a temperature measuring device 52 is also provided behind the three upstream zones of the water cooling device 4. However, the temperature measuring device 52 between the cooling zones can be located at any position from the inlet to the outlet of the water cooling device 4, or more than two temperature measuring devices can be provided inside the water cooling device 4.

[0077] The temperature measuring device 5 can be a device for measuring the surface temperature of the upper surface of the steel plate S or a device for measuring the surface temperature of the lower surface. The temperature measuring device 5 can be either a contact type or a non-contact type. In the case of a contact type, a thermocouple is preferred; in the case of a non-contact type, a radiation thermometer is preferred. Besides a conventional radiation thermometer that pre-determines the emissivity of the steel plate S and converts luminance data into temperature data, a bicolor radiation thermometer can also be used, which measures the radiance of two different wavelengths and converts the ratio of these wavelengths into the temperature of the object. Furthermore, a more preferred embodiment is a temperature measuring device capable of measuring the temperature distribution along the width direction of the steel plate S. Specifically, a scanning thermometer can be used, which arranges multiple radiation thermometers in a direction orthogonal to the transport direction (length direction) of the steel plate and scans the temperature measurement points along the width direction. Alternatively, a camera-type thermometer can be used, which acquires an image of the steel plate S and converts the luminance data of the image into temperature.

[0078] Furthermore, the temperature measuring device 5 is preferably located in a position where it is unlikely to be affected by the cooling water W from the water-cooled nozzles 41a and 41b. Figure 2 In the example shown, the temperature measuring device 52 is positioned at a location divided by the constraint rollers 42a and 42b, and is not directly supplied with cooling water W based on the water-cooling nozzles 41a and 41b. This reduces interference with the temperature measurement performed by the temperature measuring device 52. However, when using a temperature measuring device such as a fountain pyrometer that can measure the surface temperature of the steel plate S even in the presence of cooling water W, the temperature measuring device 52 can also be positioned at a location where cooling water W is directly supplied.

[0079] In this embodiment, the surface temperature of the steel plate S measured by the temperature measuring device 5 is correlated with the positional information within the surface of the steel plate S. Regarding the positional information in the width direction of the steel plate S, since the width direction position for obtaining the temperature data from the radiation thermometers is predetermined, the correspondence between the temperature data and the position in the width direction of the steel plate S is clear. When the temperature measuring device 5 uses a scanning thermometer to scan in the width direction, since the scanning position in the width direction is determined by the temperature measuring device 5, the position in the width direction of the steel plate S is correlated with the measured temperature data.

[0080] On the other hand, regarding the positional information of the steel plate S along its length, by determining the conveying distance from the front end of the steel plate S, a correspondence can be established with the temperature data obtained by the temperature measuring device 5. For example, as... Figure 5 As shown, when the temperature measuring device 5 is configured, the arrival of the front end of the steel plate S at the temperature measuring device 5 can be determined by the phased increase of the measured temperature data. Then, using the signal that the front end of the steel plate S has arrived at the temperature measuring device 5 as a trigger, the subsequent conveying distance is calculated based on the rotational speed and diameter of the table roller 6 that conveys the steel plate S. This conveying distance becomes the position information in the length direction from the front end of the steel plate S. The measured value of the surface temperature of the steel plate S measured by the temperature measuring device 5 and the position information in the length direction are sent, for example, to the surface temperature information generation device 54, and a correspondence is established between the measured value of the surface temperature of the steel plate S and the position information in the surface of the steel plate. In addition, the method of establishing a correspondence between the surface temperature of the steel plate S and the position information in the surface of the steel plate S is not limited to this method. When measuring the entire surface of the steel plate S using a camera-type two-dimensional thermometer, since the temperature and position information of the entire surface of the steel plate S are obtained, the surface temperature and position information in the surface of the steel plate S are obtained simultaneously.

[0081] Figure 6 This example illustrates the measurement of the surface temperature of steel plate S. Regarding the surface temperature information within the surface of steel plate S, a two-dimensional coordinate system is established on the surface of steel plate S, thereby establishing a correspondence between the measured temperature and the position information within the surface of steel plate S. Furthermore, the surface temperature information corresponding to the position information is sent to control computer 10 or host computer 11 and stored in at least one of the storage devices.

[0082] [Benchmark point and reference point]

[0083] In the method for generating the material prediction model of the steel plate in this embodiment, a temperature measuring device 5 installed in the cooling device 3 is used to obtain measured data of the surface temperature at a pre-set reference point on the steel plate S, and measured data of the surface temperature at a reference point set based on the reference point. Hereinafter, the measured data of the surface temperature at the reference point and the reference point are referred to as the surface temperature information dataset.

[0084] In this embodiment, the reference point refers to a point arbitrarily set within the surface of the cooled steel plate S, and its position within the surface of the steel plate S is determined. That is, the reference point is a point whose position is determined based on the distance from the front or rear end of the surface of the steel plate S, and the distance from one end of the steel plate S in the width direction or the other end of the steel plate S.

[0085] Figure 7 An example representing a reference point. For example... Figure 7 As shown, with the direction from the front end to the rear end of the steel plate S relative to its travel direction defined as the x-axis and the direction to the right of the travel direction of the steel plate S defined as the y-axis, the reference point PA is a point whose position is determined by the coordinates (x1, y1). Multiple reference points can also be set relative to a single steel plate S. Figure 7 In the example shown, the planar shape of the steel plate S is not necessarily rectangular. Therefore, a coordinate system is established with the origin at a predetermined distance Lt (e.g., Lt can be set to approximately 0.05 to 0.3 m) from the center of the frontmost part of the steel plate S in the width direction towards the rearmost part, excluding the area where the width is reduced at the frontmost end of the steel plate S. The area where the width is reduced at the frontmost end of the steel plate S will not be a steel plate product, and therefore can be excluded from the target area for material prediction of the steel plate S in this embodiment.

[0086] On the other hand, a reference point is a point set based on a datum point, referring to a point whose positional relationship with a datum point within the steel plate surface is determined. Multiple reference points can be set relative to a single datum point. Figure 8 Examples of reference points are provided. Reference points PB1 and PB2 are set at the tail end and front end sides of the travel direction, respectively, at a distance dx from the reference point PA determined by coordinates (x1, y1). Additionally, reference points PB3 and PB4 are set at the right and left ends of the travel direction, respectively, at a distance dy from the reference point PA in the width direction. At least one reference point needs to be set relative to a reference point, and it can be set in either the travel direction or the width direction of the steel plate S. Furthermore, reference points can also be set at an angle relative to the travel direction of the steel plate S, and do not necessarily need to be set in either the travel direction or the width direction of the steel plate S. However, it is preferable that reference points be set relative to the reference point in both the travel direction and the width direction of the steel plate S, and more preferably, two or more reference points are set in each direction.

[0087] The distance between the reference point and the reference point can be set arbitrarily within the range of 0.1 to 200 mm. Preferably, it is 1 to 50 mm, more preferably 5 to 20 mm. The distance between the reference point and the reference point can also be varied according to the thickness of the steel plate S. For example, relative to the thickness H of the steel plate S, the distance between the reference point and the reference point can be set from a range of 0.5H to 3.0H. As described later, in this embodiment, since the behavior of thermal movement inside the steel plate is indirectly determined by the difference in surface temperatures measured at the reference point and the reference point, it is difficult to detect the difference in surface temperatures when the distance between the reference point and the reference point is too short. On the other hand, it is difficult to determine the material distribution based on the position within the steel plate surface when the distance between the reference point and the reference point is too long. Furthermore, since the thermal movement between the reference point and the reference point depends on time, the distance between the reference point and the reference point can also be set according to the target cooling rate. For example, the distance between the reference point and the reference point can be shortened when the target cooling rate is high, and lengthened when the target cooling rate is low.

[0088] Furthermore, in this embodiment, as described above, the temperature measuring device 5 is used to establish a correspondence between the surface temperature of the steel plate S and its positional information within the surface. Therefore, based on the positional information of the reference point and the reference point, the surface temperature data of the reference point and the reference point can be determined, thereby enabling the construction of a surface temperature information dataset.

[0089] [Steel plate material information]

[0090] As an embodiment of the present invention, the method for generating a material prediction model for a steel plate includes the above-mentioned surface temperature information dataset in the input data, and takes the material information of the steel plate S after passing through the cooling device 3 at the position corresponding to the reference point on the steel plate S corresponding to the input data as the output data.

[0091] The actual data on the material properties of steel plate S are obtained from steel plate S after passing through cooling device 3, and can be obtained from steel plate S cooled to near room temperature. Specifically, in Figure 1 In the examples shown in (a) and (b), material information can be obtained from the steel plate S after it has passed through cooling device 3 and is being transported to the cooling bed or finishing equipment. From the steel plate S cooled to near room temperature, test samples are collected using the test sample collection device of the finishing equipment to obtain material information. In the quality assurance of the steel plate S that becomes the product, test items for checking the material of the steel plate S and the location for collecting test samples are pre-set. Furthermore, the location for collecting test samples is mostly pre-set and determined according to the size, standard, and specifications of the steel plate S. Therefore, the material information of the steel plate S is obtained in a corresponding manner to the location information of the steel plate S. Additionally, in Figure 1 In the online heat treatment process shown in (a), when the steel plate cooled by the cooling equipment undergoes a tempering process for reheating and is then conveyed to the cooling bed, the actual data of the material information of the steel plate S is obtained from the steel plate S after tempering. This is because, in the tempering process, although the steel plate S is not cooled using the cooling equipment, the material information of the steel plate S changes due to the tempering process.

[0092] In this embodiment, the material information of the steel plate S refers to information related to the mechanical properties of the steel plate S manufactured through a heat treatment process. This information related to mechanical properties refers to information obtained from tests typically performed to determine the mechanical properties of the steel plate S, such as tensile tests, compression tests, bending tests, Charpy impact tests, CTOD tests, DWTT tests, and fatigue tests. In the case of tensile tests, information based on standards such as JIS or ISO can include tensile strength, yield strength, and elongation (permanent elongation, elongation at break, total elongation, etc.). Additionally, upper yield point, lower yield point, 0.2% yield strength, and reduction of area are also information obtained from tensile tests and thus constitute the material information of the steel plate S. In the case of Charpy impact tests, V-notch test pieces are collected from test specimens, and the absorbed energy and brittle fracture surface area obtained at each test temperature during pendulum failure can be used as the material information of the steel plate S. In CTOD tests, the crack tip opening amount (limiting CTOD value) at each test temperature during unstable failure can be used as the material information of the steel plate S. In addition, in general fatigue failure, the number of fracture repetitions and fatigue limit values ​​obtained for each set stress amplitude can also be used as the material information of the steel plate S.

[0093] When test samples are obtained from multiple locations on a steel plate S, material information is obtained for each test sample collection location corresponding to the collection location within the surface of the steel plate S. In this case, material information of the steel plate S with each test sample collection location as a reference point can be obtained. When multiple test samples are collected from the same location on the steel plate S, multiple types of material information are obtained using each test sample. In this case, the multiple obtained material information can be set as a dataset and designated as the material information of the steel plate S.

[0094] [Material Prediction Model Generation Department]

[0095] The steel plate material prediction apparatus according to one embodiment of the present invention includes a material prediction model generation unit. The material prediction model generation unit generates a material prediction model for the steel plate S after passing through the cooling device 3 using machine learning with multiple learning data sets. These multiple learning data sets include a surface temperature information dataset, comprising measured surface temperature data at a pre-set reference point on the steel plate S in the cooling device 3 and measured surface temperature data at a reference point set based on that reference point, as input performance data. The output performance data is the material information of the steel plate S after passing through the cooling device 3 at the position corresponding to the reference point on the steel plate S corresponding to these input performance data sets. The material information of the steel plate S after passing through the cooling device 3, as output performance data, is not necessarily limited to the material information after cooling to room temperature, as long as it has passed through the cooling device 3. However, when performing a tempering process where the steel plate S is reheated after passing through the cooling device 3, the material information of the steel plate S after the tempering process is completed is used as output performance data.

[0096] Figure 9 This illustrates the structure of the material prediction model generation unit, which is one embodiment of the present invention. For example... Figure 9 As shown, the material prediction model generation unit 20, as an embodiment of the present invention, includes a database unit 20a and a machine learning unit 20b. The database unit 20a acquires measured data of the surface temperature at a reference point on the steel plate S and measured data of the surface temperature at a reference point corresponding to the reference point as the actual value of the surface temperature information dataset, and acquires the material information of the steel plate S at the position corresponding to the reference point after passing through the cooling device 3.

[0097] The surface temperature information dataset stored in the database unit 20a consists of surface temperature information data corresponding to the locations where material information is obtained, which serve as reference points on the steel plate S. Therefore, when there are multiple locations on a single steel plate S where material information is obtained, the surface temperature information datasets and material information corresponding to each reference point are stored in the database unit 20a in a corresponding manner. That is, the number of datasets stored in the database unit 20a is the same as the number of reference points from a single steel plate S.

[0098] The actual data on the material information of steel plate S is obtained by determining the positional information within the surface of the steel plate based on the location of the test piece sample collected as described above. However, the test piece sample used to obtain the material information of steel plate S has a fixed size depending on the test method, and therefore does not strictly correspond to the position of the reference point constituting the surface temperature information dataset. However, in general quality assurance considerations, it is common knowledge to evaluate based on the premise that the material is uniform within the test piece sample. Therefore, in this embodiment, it is not necessary for the position of the actual data on the material information of the steel plate to strictly correspond to the reference point constituting the surface temperature information dataset. Therefore, the "position corresponding to the reference point on the steel plate" mentioned above only needs to mean that the coordinates of the reference point determined within the surface of the steel plate S are included in the range of the test piece sample collected by the test piece collection device. Preferably, the central part of the position of the test piece used to obtain the material information is within 200 mm of the reference point constituting the surface temperature information dataset.

[0099] In the database section 20a, parameters (attribute information parameters) related to the property information of the steel plate S can also be accumulated as input data. As described above, the property information of the steel plate S includes not only dimensional information such as thickness, width, length, and weight, but also information related to its composition (C, Si, Mn, Cr, and Mo content) and target values ​​for the mechanical properties of the heat-treated steel plate S (yield stress, tensile strength, elongation, toughness, hardness, etc.). The composition information of the steel plate S includes not only the contents of C, Si, Mn, Cr, and Mo, but also the contents of Nb, Ni, V, W, Sn, and Cu. However, since a set of property information is established for each steel plate S, even if multiple reference points are set for a single steel plate S and multiple surface temperature data sets are obtained, the property information parameters of the same steel plate S are still associated with these surface temperature data sets. The property information parameters of steel plate S are used to generate the material prediction model because it is advantageous in generating a high-accuracy material prediction model even when the property information of steel plate S changes significantly. However, when generating a material prediction model for steel plate S of the same steel grade (with a common range of compositional management), it is not necessary to include the property information parameters of steel plate S in the input of the material prediction model. This is because, with a common range of compositional management, even if the actual values ​​of the composition of steel plate S change, the material information of steel plate S will not deviate significantly.

[0100] Database unit 20a can also store performance data (operational performance data) of the operating performance parameters of water cooling device 4. As described above, the operating performance parameters of water cooling device 4 can be set or measured values ​​of operating parameters that affect the cooling state of steel plate S, such as the volume of cooling water W sprayed from water cooling nozzles 41a and 41b (cooling water volume), the speed of steel plate S based on table roller 6 (conveyor speed), information related to the balance of each cooling zone with respect to the cooling water volume, and the number of cooling zones sprayed with cooling water W. This is because the cooling state of steel plate S affects the material properties of steel plate S after cooling. Alternatively, the ratio of the upper and lower water volumes of cooling water W sprayed from water cooling nozzles 41a and 41b can also be used as an operating performance parameter of water cooling device 4. This is because the upper and lower water volume ratio of cooling water W affects the cooling state of steel plate S due to warping caused by the steel plate S.

[0101] When using operating conditions related to cooling water volume as the operational performance parameters of the water cooling device 4, the cooling water volume of each water cooling nozzle can be used as the operational performance data of the water cooling device 4 by assigning identification numbers to each water cooling zone and the upper and lower water cooling nozzles 41a and 41b. However, the sum of the cooling water volumes in the water cooling zones or the sum of the cooling water volumes in multiple water cooling zones arbitrarily selected from the water cooling zones can also be used as the operational performance data of the water cooling device 4. In particular, in the multiple cooling zones on the front side (upstream side) of the cooling zone, the temperature change of the steel plate S is large, which has a significant impact on the material of the steel plate S. Therefore, the sum of the cooling water volumes in 2 to 3 cooling zones on the front side of the cooling zone can also be used.

[0102] Furthermore, if flow meters are installed on a per-unit basis, such as in the water-cooled zone, water-cooled nozzles, or water-cooled manifolds located upstream of the piping system of the water-cooled nozzles, the actual operating data obtained from the flow meters can also be used as the operating performance data of the water-cooling device 4. However, the set value of the cooling water volume set in the water-cooling condition calculation unit 10a can also be used. This is because it is believed that if the set value and the actual value of the water-cooled nozzles are compared in advance, the actual cooling water volume is less likely to deviate significantly from the set value.

[0103] Furthermore, the cooling rate of the steel plate S and the conveying speed of the steel plate S within the cooling equipment 3 can also be used as operational performance data for the water cooling device 4. This is because, depending on the cooling rate of the steel plate S, the temperature gradient generated along the length of the steel plate S changes, and the shape of the steel plate S changes due to the gradient of thermal strain along the length, thereby affecting the temperature history of the steel plate S and the material of the steel plate S after passing through the cooling equipment.

[0104] In addition, the cooling stop temperature of the steel plate S can also be included as operational performance data for the water cooling device 4. This is because when the cooling stop temperature is low, the cooling zone that enters the nuclear boiling stage becomes a condition where temperature deviation is likely to occur, thus leading to a deterioration in the uniformity of the material within the surface of the steel plate S.

[0105] Based on the above, when using the operational performance data of the water cooling device 4 as input performance data for the steel plate material prediction model, it is preferable to include at least one of the following: cooling water volume, the ratio of the upper and lower water volumes of the cooling water W, the cooling rate of the steel plate S, and the conveying speed of the steel plate S within the cooling equipment 3; more preferably, it should include multiple operational performance data. This is because it is beneficial to predict the in-plane material distribution of the steel plate S caused by multiple factors.

[0106] When using the operational performance data of the water cooling device 4 as described above, the average values ​​of the surfaces of the steel plate S can be calculated, and the calculated values ​​are used as representative values ​​in the operational performance data of the water cooling device 4. In this case, when multiple surface temperature data sets are obtained for a single steel plate S, the same representative value is used as the operational performance data of the water cooling device 4 to establish a correspondence with each surface temperature data set. Furthermore, during the cooling process of the steel plate S, if the cooling water volume or the number of cooling zones used changes according to the position along the length of the steel plate, the cooling water volume and the number of cooling zones used when passing through the cooling device 3 at the reference point can be stored in the database unit 20a as operational performance data of the water cooling device 4.

[0107] The input data for the material prediction model M is not limited to the above. It can also include actual or set values ​​of temperature and residence time in various zones such as the heating zone or soaking zone of the heating furnace 2 in the offline heat treatment equipment 1. This is because the surface roughness or oxide state of the steel plate S affects the wettability of the cooling water W, and the temperature distribution within the surface of the steel plate S changes during the cooling process, thus indirectly affecting the material properties of the steel plate S. Furthermore, when a constraint device is configured in the cooling zone, the operational data of the constraint device, such as the constraint force of the constraint rollers 42a and 42b on the steel plate S, and the set or measured values ​​of the pressing position, can also be included in the input data for the material prediction model M. This is because the shape of the steel plate S changes during the cooling process due to these constraints, affecting the uniformity of the material within the surface of the steel plate S after passing through the cooling equipment 3. When a water-removing purging nozzle 7 is provided on the outlet side of the constraint device at the downstream end of the water-cooling device 4, the operational performance data of the water-removing purging nozzle 7, such as the purging pressure and gas injection volume, can also be included in the input performance data of the material prediction model M. This is because if the operating conditions of the water-removing purging nozzle 7 are inappropriate, the temperature deviation of the steel plate S will increase, and the material uniformity of the steel plate S will deteriorate.

[0108] The material prediction model generation unit 20 can be located inside the control computer 10, or it can be installed in the host computer 11 that provides manufacturing instructions to the control computer 10. Alternatively, it can be a separate piece of hardware capable of communicating with both the control computer 10 and the host computer 11. It can also be installed in the material determination unit, which will be described later, to determine whether a material is qualified or not.

[0109] As described above, the data includes the actual data of the surface temperature information dataset corresponding to the reference point pre-set on the steel plate S corresponding to the sample collection position of the test piece, the material information of the steel plate S at the position corresponding to the reference point after passing through the cooling device, and the operation data of the steel plate S obtained as needed, as well as the operation data of the water cooling device 4, which constitute a set of datasets for each pre-set reference point and are stored in the storage device of the database unit 20a. In addition, the set of datasets for each reference point may also include one or more operation data selected from the operation data of the heating furnace 2, the operation data of the constraint device, and the operation data of the water purging nozzle 7. The number of sample collection positions set in the surface of the steel plate S is usually set to about 1 to 10. Therefore, in this embodiment, about 1 to 10 datasets are stored in the storage device of the database unit 20a for one steel plate.

[0110] In the database section 20a, at least 50 datasets are stored in each partition for the same specifications, steel type, and size. Preferably, at least 100 datasets are stored, more preferably at least 500. Furthermore, when including any one of the different steel plates S in terms of specifications, steel type, and size, it is preferable to store at least 2000 datasets. If the specifications or steel type of the steel plate S are different, the composition has a greater impact on the material properties of the heat-treated steel plate S. Therefore, it is preferable that the datasets stored in the database section 20a include attribute information parameters of the steel plate S.

[0111] For the data accumulated in Database 20a, there are situations where filtering is required, and data with outliers can be removed. This is because accumulating highly reliable data improves the prediction accuracy of the material. The number of datasets accumulated in Database 20a can also be capped at a certain limit, and the datasets accumulated in Database 20a can be updated appropriately within this limit.

[0112] Machine learning unit 20b uses the dataset stored in database unit 20a to generate a material prediction model M for the steel plate S after passing through cooling device 3 by using machine learning with multiple learning data. This multiple learning data uses a dataset of surface temperature information corresponding to a pre-set reference point on the steel plate S as input data, and uses the material information of the steel plate after passing through cooling device 3 at the position corresponding to the reference point on the steel plate S as output data. Alternatively, machine learning can be performed by including one or more operational data selected from the attribute information parameters of steel plate S, operational data of water cooling device 4, operational data of constraint device, operational data of heating furnace 2, and operational data of dewatering purging nozzle 15 stored in database unit 18a as needed, within the aforementioned input data.

[0113] The machine learning model used to generate the material prediction model M can be any machine learning model, as long as it can obtain prediction accuracy with practically sufficient material information. For example, commonly used neural networks (including deep learning, convolutional neural networks, etc.), decision tree learning, random forests, support vector regression, etc., can be used. Alternatively, an ensemble model composed of multiple models can also be used.

[0114] Furthermore, the material prediction model M can also utilize a machine learning model that not only outputs the material information of the steel plate S as numerical values, but also determines whether it falls within the predetermined allowable range of material information, and binarizes the result into qualified / unqualified data as output performance data. In this case, classification models such as k-nearest neighbors or logistic regression can be used.

[0115] Furthermore, the material prediction model M can be updated to a new model through relearning, for example, every month or every year. This is because the more data stored in the database 20a increases, the more accurate the material prediction becomes. By updating the material prediction model M based on the latest data, a material prediction model M that reflects changes in operating conditions over time can be generated.

[0116] Here, we will explain the reason for using the measured surface temperature data at a pre-set reference point on the steel plate S and the measured surface temperature data at a reference point set based on that reference point as input to the material prediction model M.

[0117] In this embodiment, the steel plate used for material prediction is typically 3–100 mm thick, 1000–4000 mm wide, and 4000–20000 mm long. For such steel plates, conventional material prediction models estimate the internal temperature distribution based on measurements of the steel plate's surface temperature during cooling, using heat transfer calculations, and predict the steel plate's material based on a pre-determined correlation between the steel plate's internal thermal history and the material composition after cooling. In this case, the heat transfer calculation used to estimate the internal temperature distribution is based on a one-dimensional heat conduction equation along the thickness direction. That is, information related to the steel plate's surface temperature is only used to estimate the internal temperature from the measurement location in the direction perpendicular to the steel plate surface. On the other hand, when the steel plate thickness reaches 5 mm or more, not only the heat transfer behavior in the thickness direction but also the effect of in-plane thermal movement on the internal temperature of the steel plate cannot always be ignored. In particular, in areas where heat conduction from the end face of the steel plate may occur, such as the front end, rear end, and width end, in-plane thermal movement within the steel plate needs to be considered.

[0118] In this embodiment, not only is measured surface temperature data from a pre-set reference point on the steel plate used, but also measured surface temperature data from a reference point set based on that reference point. Therefore, the behavior of in-plane thermal movement within the steel plate is reflected in the difference between the surface temperatures at the reference point and the reference point. That is, if the surface temperature at the reference point is higher than the surface temperature at the reference point, it is inferred that the thermal movement within the steel plate occurs from directly below the reference point towards directly below the reference point. Compared to conventional methods that only use surface temperature information at the reference point, by combining the surface temperature information at the reference point, a material prediction model reflecting information related to in-plane thermal movement within the steel plate can be obtained. In this embodiment, information related to heat transfer behavior in the thickness direction of the steel plate is obtained from the surface temperature at the reference point, and information related to in-plane heat transfer behavior within the steel plate is obtained from the relationship between the surface temperatures at the reference point and the reference point. Therefore, this surface temperature information becomes information for determining the internal heat transfer behavior of the steel plate. It is known that the internal heat transfer behavior during the heat treatment process of the steel plate has a significant impact on the material properties of the steel plate after heat treatment. Machine learning methods, which use surface temperature information datasets as input and steel plate material information as output, can quantitatively reflect their correlation in a material prediction model, thereby generating a high-precision material prediction model for steel plates.

[0119] Furthermore, by setting multiple reference points relative to a pre-defined benchmark, behavior related to thermal movement within the steel plate can be reflected with high precision. Additionally, by setting at least one reference point along the length of the steel plate and at least one reference point along its width, information related to the direction of thermal movement generated internally directly below the benchmark point can be reflected. For example, in... Figure 8 Among the reference points shown, if there is a difference in surface temperature at reference point PB1, reference point PA, and reference point PB2 located along the length of the steel plate, thermal movement along the length of the steel plate occurs. Conversely, if there is a difference in surface temperature at reference point PB3, reference point PA, and reference point PB4 located along the width of the steel plate, thermal movement along the width of the steel plate occurs. Therefore, information related to the direction of thermal movement within the steel plate can be reflected based on the surface temperature information at the reference points. For this reason, in this embodiment, the surface temperature information dataset can be constructed not only using measured surface temperature data at the reference points and measured surface temperature data at reference points set based on those reference points, but also using the difference between the measured surface temperature data at the reference points and the surface temperature at reference points set based on those reference points. This is because the difference in surface temperature between the reference points and the reference points is also information related to thermal movement within the steel plate.

[0120] By obtaining reference points and surface temperature information at multiple locations during the cooling process of the steel plate, the temporal variations in the internal thermal movement behavior of the steel plate can be reflected in the material prediction model, thereby improving the prediction accuracy of the material prediction model. The surface temperature information at the reference points and these reference points is preferably obtained at 2 to 10 locations during the period from the start to the end of cooling of the steel plate S. This is because obtaining surface temperature information at more than two locations allows for the acquisition of information correlated with the temporal variations in the internal thermal movement behavior of the steel plate S, and even obtaining surface temperature information at more than ten locations does not significantly improve the prediction accuracy of the material prediction model.

[0121] [Material Prediction Department]

[0122] As one embodiment of the present invention, the material prediction part of the steel plate can be provided in... Figure 2 The control computer 10 or host computer 11 shown is used for control. Alternatively, it can be set to control... Figure 1 The finishing equipment shown includes a computer for controlling the finishing process, which oversees the entire finishing process. This computer can be integrated as part of these computers or configured as separate hardware. Alternatively, it can be integrated into a tablet terminal capable of communicating with these computers. The material prediction unit is equipped with the ability to communicate with... Figure 2 The unit that communicates with the control computer 10 or the host computer 11 shown is capable of obtaining the surface temperature information of the steel plate S obtained in the cooling device 3. Hereinafter, refer to... Figure 10 The operation of the material prediction unit of the steel plate, which is one embodiment of the present invention, will be explained.

[0123] Figure 10 This is a diagram illustrating the operation of the material prediction section of a steel plate, which is an embodiment of the present invention. Figure 10 The operation shown begins at the following moments: the material prediction unit obtains information about the steel plate, such as the product number or manufacturing number, from the control computer 10 or the host computer 11 to identify the steel plate that has passed through the cooling device 3; and the material prediction unit obtains information about the prediction reference point within the surface of the steel plate, which serves as the location of the material to be predicted. Furthermore, the "prediction reference point" used in the material prediction unit is different from the reference point used when accumulating performance data in the database unit 20a of the material prediction model generation unit 20; it can specify any position of the steel plate that is the object of material prediction.

[0124] Next, based on the information of the steel plate being identified, the material prediction unit obtains the surface temperature information collected during passage through the cooling device 3, which is stored in the control computer 10 or the host computer 11. The surface temperature information of the steel plate is actual data corresponding to the surface temperature of the steel plate's in-plane position information. On the other hand, if a prediction reference point is specified, the material prediction unit sets prediction reference points based on that reference point. The prediction reference points used in the material prediction unit are set relative to the prediction reference point with the same positional relationship as the reference points used to accumulate actual data in the database section 20a of the material prediction model generation unit 20. That is, the material prediction unit applies reference points that are the same number as the reference points corresponding to the reference point accumulated in the database section 20a of the material prediction model generation unit 20, and whose distance and direction from the reference point are also in the same positional relationship.

[0125] Based on the position information of the prediction reference points and prediction benchmarks set in this way, the material prediction unit obtains measured data of the surface temperature of the steel plate corresponding to these positions from the surface temperature information of the steel plate. This constitutes a surface temperature information dataset that serves as input to the material prediction model M. Furthermore, since the surface temperature information of the steel plate is discrete information corresponding to the coordinates within the surface of the steel plate, in this embodiment, the distance between the prediction reference points and prediction benchmarks is at least twice as large as the distance between the partitions (collection intervals of surface temperature information) related to the positions within the surface of the steel plate from which the surface temperature information is obtained. On the other hand, when the property information parameters of the steel plate and the operating parameters of the water cooling device 4 are used as input to the material prediction model M, each set of measured data is sent from the storage device of the host computer 11 or the control computer 10 connected to the host computer 11 to the material prediction unit.

[0126] Through the above processing, the material prediction unit inputs the surface temperature information dataset as input data into the material prediction model M, thereby outputting the material information of the steel plate after passing through the cooling equipment at the position corresponding to the prediction reference point set on the steel plate.

[0127] On the other hand, in the material prediction department, while changing the prediction reference points used for material prediction within the surface of the steel plate, material information corresponding to each prediction reference point is output, thereby obtaining the material prediction results for the entire width and length of the steel plate. Furthermore, based on these full-width and full-length material prediction results, the quality of the steel plate's material can be determined. Specifically, an additional processing step can be added: determining whether the entire surface of the steel plate meets the specified material standards; discarding portions that do not meet the material standards by cutting them off, or allocating them to products with specifications different from the initial plan. This processing step refers to an additional step that differs from the initial production plan. Therefore, it is possible to prevent the shipment of steel plates with uneven material quality, and to provide steel plates with uniform material quality throughout the surface.

[0128] <Example 1>

[0129] In this embodiment, Figure 1 In the offline heat treatment equipment 1 shown, a steel plate at room temperature with its surface oxide scale removed by shot peening in a pretreatment equipment is used. The steel plate is then heated to 930°C in a nitrogen atmosphere using a heating furnace 2, and cooled using a cooling device 3 to a target temperature of 430°C, which serves as the cooling stop temperature, to produce quenched and tempered steel. The cooling device 3 is located downstream of the heating furnace 2 and contains seven pairs of water-cooling nozzles 41a and 41b and eight pairs of constraint rollers 42a and 42b, constituting a water-cooling unit 4. Flat spray nozzles are used as the water-cooling nozzles 41a and 41b.

[0130] In this embodiment, a scanning surface thermometer capable of measuring the temperature in the width direction of the steel plate is installed at the inlet of the cooling device 3, located 2.0m away from the heating furnace 2, and another scanning surface thermometer is installed at a location 3.0m away from the outlet of the water cooling device 4. Thus, the cooling start temperature is measured across the entire width and length of the steel plate at the inlet of the cooling device 3, and the cooling stop temperature is measured across the entire width and length of the steel plate at the outlet of the cooling device 3. These temperatures are then stored as surface temperature information of the steel plate in the storage device of the host computer 11 via the control computer 10. Furthermore, in this embodiment, the measured cooling start temperature and cooling stop temperature of the steel plate are 910℃±10℃ and 450℃±50℃ respectively at the center of the steel plate's surface.

[0131] In this embodiment, the steel plate used for material prediction is a thick steel plate with a thickness of 12 mm and a tensile strength of 780 MPa (standard: yield stress above 685 MPa, tensile strength 780-930 MPa). When generating the material prediction model, 100 thick steel plates classified into the same partition according to size and standard are used to obtain training data. For each thick steel plate, 1 to 5 reference points are set at positions within the steel plate surface, such as the front end, tail end, width end, and center. Based on the set reference points, 2 reference points are set in both the length and width directions of the steel plate, with a distance of 50 mm between the reference points and the reference points. Thus, with four reference points relative to one reference point, a set of surface temperature information datasets consisting of surface temperature information from five points obtained by a surface temperature measuring device is generated. That is, after a cooling process performed by the cooling device 3, a set of surface temperature information datasets consisting of 10 points of surface temperature information is generated for one reference point.

[0132] On the other hand, as attribute information parameters selected from the steel plate's property information, the steel plate's composition, width, and length are selected. Regarding the steel plate's composition, the C, Si, Mn, Cr, and Mo contents are set as weight percent and included in the steel plate's attribute information parameters. This information is stored in the host computer 11. In addition to these alloy components, the steel plate also contains P, Ti, S, Al, N, etc., but these do not significantly affect the prediction accuracy of the material prediction model for the steel plate shown below within the aforementioned partitions. Therefore, these alloy components are not included in the attribute information parameters of this embodiment. Furthermore, in this embodiment, as operational performance data of the water cooling device 4, the total spray water volume of the water cooling device 4, the number of spray zones, and the conveying speed of the steel plate within the cooling equipment 3 are included in the operational performance data. Moreover, as material information of the steel plate used in this embodiment after passing through the cooling equipment 3, the tensile strength and yield stress are obtained by using tensile tests on test specimens collected from positions corresponding to reference points during the finishing process.

[0133] In this embodiment, heat treatment is performed on the aforementioned 100 steel plates, and the actual data of material information at each reference point, the actual data of surface temperature information dataset for the reference points, the aforementioned attribute information parameters, and the actual data of operation of the aforementioned water cooling device 4 are stored in the database unit 20a. Then, by using machine learning with these learning data, a material prediction model M for the steel plates after passing through the cooling device 3 is generated. As the machine learning algorithm, a neural network is used, with three intermediate layers and five nodes in each layer. The activation function used is a sigmoid function.

[0134] The generated material prediction model M is sent to a tablet terminal capable of communicating with the computer controlling the finishing process, serving as the material prediction model for the material prediction unit internally configured therein. Furthermore, the material prediction unit in this embodiment has the following functions: while making various changes to the prediction reference points set within the surface of the steel plate, it acquires material information at the positions corresponding to the prediction reference points, thereby outputting a material prediction result for the entire width and length of the steel plate. Additionally, the predicted material information and the surface temperature information acquired during the cooling process can be displayed and output as images on the aforementioned tablet terminal.

[0135] Figure 11 (a) shows an example of surface temperature information obtained during the cooling process by a surface temperature measuring device disposed at the inlet of the cooling equipment. Figure 11 (b) shows an example of surface temperature information obtained by a surface temperature measuring device disposed at the outlet of a cooling device. Additionally, Figure 12 (a) indicates an example where the yield stress is shown as material information in this embodiment. Figure 12 (b) shows an example demonstrating tensile strength. Thus, the material prediction unit can predict the material information of the steel plate across its entire width and length. The aforementioned flat panel terminal has the function of displaying surface temperature and material information using contour lines based on such a color image; for example, it can display... Figure 13 Images (a) to (d) are shown. By using such material prediction results, it is easy to determine whether there are areas in the surface of the steel plate where the material does not meet the specified specifications.

[0136] exist Figure 13 In the example shown in (a), it was observed that the surface cooling initiation temperature of the steel plate was slightly lower at the width end of the thick steel plate. This is believed to be because, although it is assumed that the thick steel plate is heated approximately uniformly within the furnace, heat dissipation also occurs from the sides of the thick steel plate during its journey from the furnace to the surface temperature measuring device located at the inlet of the cooling equipment. Furthermore, in Figure 13 As shown in (b), a decrease in the cooling stop temperature of the thick steel plate surface was observed at the ends in the width direction, with slightly higher cooling stop temperatures in the 1 / 4w and 3 / 4w regions of the thick steel plate's width direction. Furthermore, in the length direction of the thick steel plate, the cooling stop temperature at the front end was slightly higher than that at the rear end. It is believed that the temperature decrease at the ends of the steel plate is due to overcooling occurring upstream of the water cooling system, as well as the influence of the flow or fluctuation of cooling water, a characteristic of the water cooling system. Additionally, the temperature variation in the length direction of the steel plate is also presumed to be caused by such characteristics of the water cooling system.

[0137] about Figure 13The predicted tensile strength and yield stress values ​​shown in (c) and (d) indicate that the tensile strength decreases and the yield stress increases at the ends of the steel plate in the width direction. Furthermore, the yield stress decreases in the 1 / 4w and 3 / 4w regions of the steel plate's width direction. Moreover, in the length direction of the steel plate, the yield stress at the front end is lower than that at the rear end. This trend is consistent with... Figure 13 (b) shows that the surface temperature information obtained at the outlet of the cooling device indicates a tendency for higher yield stress at lower cooling stop temperatures.

[0138] Next, regarding Figure 13 For the steel plates shown in (a) to (d), test specimens were collected from 30 randomly selected locations, including the vicinity of the front and rear ends in the in-plane region and the vicinity of both ends in the width direction, and tensile tests were performed. The results were compared with... Figure 13 The results obtained by comparing the material information shown in (c) and (d) are shown below. Figure 14 (a), (b). For example... Figure 14 As shown in (a), the standard deviation of the predicted and measured values ​​for tensile strength is 4.6 MPa. Additionally, as... Figure 14 As shown in (b), the standard deviation of the error between the predicted and measured values ​​for yield stress is 9.1 MPa. This confirms that, according to the present invention, practically sufficient accuracy in predicting material properties can be obtained.

[0139] As described above, for example, at the stage of steel plate production, quality assurance was previously based solely on material information from the parts that underwent mechanical testing. In contrast, according to this embodiment, material quality can be guaranteed across the entire width and length. Furthermore, if the material prediction model generated through this embodiment is used, for example, before the heat treatment process of the steel plate, by adjusting the operating parameters of the water cooling device according to the deviation in the composition of the steel plate, it is also possible to manufacture steel plates with minimal material deviation.

[0140] <Example 2>

[0141] The following describes the results of applying the steel plate material prediction method according to this embodiment to the output of other material information. In this embodiment, the method used... Figure 1The offline heat treatment equipment 1 shown manufactures steel plates with excellent wear resistance. The steel plates heat-treated using the offline heat treatment equipment 1 use steel raw materials with the following composition (by mass%): C: 0.12–0.50%, Si: 0.01–1.0%, Mn: 0.01–2.5%, P: less than 0.040%, S: less than 0.040%, Cr: 0.01–3.0%, Ti: 0.001–1.5%, B: 0.0001–0.010%, Al: less than 0.10%, N: less than 0.050%, with the balance being Fe and unavoidable impurities. The steel raw materials are first hot-rolled using a hot rolling production line at a heating temperature of 1150–1250°C and a cumulative reduction rate of 90–97% in the temperature range above the Ar3 phase transformation point to produce steel plates with a thickness of 12–13 mm, which are then cooled to room temperature. In this embodiment, using… Figure 2 The heating furnace 2 of the offline heat treatment equipment 1 shown reheats the hot-rolled steel plate and quenches the steel plate using the cooling equipment 3. As the heat treatment conditions for the steel plate, the cooling start temperature is set to 930℃±10℃, and the operating conditions of the water cooling device 4 are set to make the target cooling stop temperature 250℃±50℃.

[0142] In this embodiment, under the aforementioned conditions, and with changes made to the total water spray volume of the water-cooling device 4, the number of spray zones, and the conveying speed of the steel plates within the cooling equipment 3, 100 steel plates were heat-treated. The following material information was selected for the steel plates in this embodiment.

[0143] • Brinell hardness HBW of the steel plate surface obtained through hardness testing

[0144] • Absorbed energy (J) at -40°C obtained through Charpy impact test

[0145] • Repeated yield strength (MPa) obtained through repeated stress-strain tests

[0146] • Fatigue crack propagation velocity (m / cycle) within the stress expansion coefficient range ΔK1 = 15 MPa√m obtained through fatigue crack propagation tests.

[0147] The material information of the steel plate is obtained from the steel plate after passing through the cooling equipment 3. Test pieces for the above tests are collected from the steel plate transported to the finishing equipment in the offline heat treatment process at positions corresponding to the reference points. The test piece sample collected corresponding to one reference point is 200×200mm, and test pieces for the above tests are collected from the collected test piece samples. The reference points set on the steel plate are the front end, the rear end, the width end (working side, driving side), and the center of the surface of the steel plate. Reference points 1 to 5 are set individually according to the steel plate to be manufactured. Two reference points are set in the length direction and two in the width direction of the steel plate corresponding to the set reference points, and the distance between the reference points and the reference points is 150mm. Therefore, four reference points are established relative to one reference point.

[0148] In this embodiment, the temperature measuring device for measuring the surface temperature of the steel plate is positioned at two locations: one 1 m away from the inlet of the water cooling device 4 and the other 5 m downstream of the water cooling device 4. This allows for the measurement of the cooling start temperature and cooling stop temperature of the steel plate. Specifically, in this embodiment, as a surface temperature information dataset for the steel plate, five temperature data points, including the temperature of the reference point, are obtained at two locations corresponding to a reference point, thus generating a surface temperature information dataset with ten surface temperature data points for each reference point.

[0149] As described above, the surface temperature information dataset obtained for each reference point is used as learning data and stored in the database unit 20a along with the information for identifying the steel plate and determining the reference point. Furthermore, the database unit 20a stores actual data on the width of the steel plate and actual data related to the composition of the steel plate, such as C content, Si content, Mn content, and Cr content, as attribute information parameters of the steel plate obtained for each plate. Although the steel raw material used in this embodiment also contains other components, these components have little impact on the prediction accuracy of the material prediction model due to the combination with the operating parameters of the water cooling device 4, etc. Therefore, the material prediction model is generated based on the main component composition of the steel raw material.

[0150] Furthermore, in the database section 20a, as operational performance data of the water cooling device 4, performance data such as the total spray volume of the water cooling device 4, the number of spray zones, and the conveying speed of the steel plate within the cooling equipment 3 are stored. On the other hand, regarding the material information obtained for each reference point of the steel plate, the aforementioned test results using test pieces collected from test piece samples are stored in the database section 20a. In addition, in the database section 20a, based on the information identifying the steel plate and the information determining the reference points, a correspondence is established between the surface temperature information dataset, the property information parameters of the steel plate, the operational performance data of the water cooling device 4, and the performance data of the material information of the steel plate.

[0151] Then, in the machine learning unit 20b of the material prediction model generation unit 20, a material prediction model M for the steel plate after passing through the cooling device 3 is generated by machine learning using multiple training data. This multiple training data uses surface temperature information datasets, steel plate attribute information parameters, and operational performance data of the water cooling device 4 as input performance data, and outputs the material information of the steel plate corresponding to these input performance data. The machine learning algorithm uses a neural network with 5 intermediate layers and 8 nodes each. The activation function used is a sigmoid function.

[0152] The material prediction model M of the steel plate generated by the material prediction model generation unit 20 is sent to a tablet terminal that can communicate with the control computer of the finishing equipment, serving as the material prediction model of the material prediction unit internally configured therein. The control computer of the finishing equipment has a host computer 11 common to the control computer 10 of the heat treatment equipment, and is structured to acquire performance data obtained by the control computer 10 of the heat treatment equipment and performance data obtained by the surface temperature measuring device. Thus, in the material prediction unit, it is possible to acquire measured data of the surface temperature at the prediction reference point and prediction reference point of the steel plate obtained in the offline heat treatment equipment 1, performance data of the property information parameters of the steel plate, and operational performance data of the water cooling device 4. By inputting these data into the material prediction model M, it is possible to output the material information of the steel plate after passing through the cooling device at the position corresponding to the prediction reference point of the steel plate.

[0153] In this embodiment, six steel plates identical to those described above were prepared for testing and subjected to heat treatment again using the offline heat treatment equipment 1. After the steel plates passed through the cooling device 3 of the offline heat treatment equipment 1, the material prediction unit obtained the surface temperature information of the entire steel plate measured by the surface temperature measuring device. In the material prediction unit, while changing the prediction reference point of the steel plate within the steel plate surface, the material information of the steel plate at the position corresponding to each prediction reference point was output. Thus, the predicted value of the material information of the entire steel plate surface was obtained from the material prediction unit.

[0154] Test samples were collected from three randomly selected locations—the front end, the rear end, and the in-plane—of steel plates manufactured for testing. Material information of the steel plates was obtained through the same tests described above. Specifically, 30 data points of material information were collected from five locations each from six steel plates. Furthermore, the predicted values ​​of the material information at the prediction reference points corresponding to the locations of the collected test samples were output to the material prediction unit and compared with the actual data of the material information. Then, the error between the predicted and actual values ​​related to the material information of the steel plates was calculated, and its standard deviation σ was determined, yielding the following results.

[0155] • The standard deviation of the prediction error σ is 5.2 HBW relative to the average of the actual Brinell hardness data of 366 HBW.

[0156] The standard deviation of the prediction error σ is 2.3 J relative to the average of the actual absorbed energy data of 22 J.

[0157] • The standard deviation σ of the prediction error is 43.6 MPa, relative to the average of the actual data on repeated yield strength of 1057 MPa.

[0158] • The average value of the actual data relative to the fatigue crack propagation rate is 4.12 × 10⁻⁶. -9 m / cycle, the standard deviation of the prediction error σ is 0.75×10 -9 m / cycle

[0159] As can be seen from the above, the material prediction model M of the steel plate generated in this embodiment can predict the material information of the steel plate determined by hardness test, Charpy impact test, repeated stress-strain test and fatigue crack propagation test with sufficient accuracy in practical applications.

[0160] <Example 3>

[0161] The following describes the results of applying the material prediction method for steel plates according to this embodiment to wear-resistant steel with excellent bending workability. In this embodiment, the method utilizes an online heat treatment process... Figure 15 The cooling equipment 3 of the hot rolling production line shown is used for quenching treatment.

[0162] Figure 15 The hot rolling production line shown includes a heating furnace 2, a rolling mill 30, and a cooling device 3. In the heating furnace 2, the steel slab is heated to a specified temperature. The rolling mill 30 is a reversible mill, used for multi-pass rolling to produce steel plates of a specified thickness and width. The steel plates rolled to the specified dimensions by the rolling mill 30 are heated to a high temperature and then subjected to a heat treatment process using the cooling device 3. Temperature measuring devices 51 and 53 are installed upstream and downstream of the cooling device 3. The cooling device 3, installed in the in-line heat treatment equipment, also uses… Figure 2 The cooling device shown is the same as the one shown in Cooling Device 3.

[0163] In this embodiment, after performing an online heat treatment process using the cooling equipment 3 of the hot rolling production line, the steel plate is tempered. During tempering, the steel plate is not cooled using the cooling equipment, and material information is obtained from the tempered steel plate. In the steel plate tempered through the above manufacturing process, the steel raw material composition, by mass%, contains C: 0.06–0.25%, Si: 0.01–0.8%, Mn: 0.5–2%, P: less than 0.010%, S: less than 0.003%, Al: 0.005–0.1%, N: 0.0005–0.008%, Mo: 0.01–1%, with the balance consisting of Fe and unavoidable impurities. The steel raw material is hot-rolled using a hot rolling production line at a heating temperature of 1100°C and a reduction rate of 40–50% in the non-recrystallization zone to produce a steel plate with a thickness of 10 mm. Then, it is further processed by... Figure 2 The cooling equipment shown in Figure 3 has the same structure as the cooling equipment shown in the figure. It is heat treated under the following conditions: cooling start temperature of 750-780°C, target cooling stop temperature of 200-250°C, and an average cooling rate of 65-70°C / second at 500-700°C. Then, the steel plate is reheated at 580-600°C to perform tempering treatment.

[0164] In this embodiment, during the online heat treatment process described above, 100 steel plates were quenched by altering the total water spray volume, the number of spray zones, and the conveying speed of the steel plates within the cooling equipment 3. Then, after tempering the steel plates under the same tempering conditions, they were air-cooled in a cooling bed, and test samples were collected in the finishing equipment. The material information of the steel plates was obtained through the following tests.

[0165] • Tensile strength (MPa) and yield stress (MPa) obtained through tensile testing

[0166] • The ultimate bending radius obtained through bending tests (the ratio of the bending radius to the thickness of the steel plate).

[0167] The material information of the steel plate is obtained from the steel plate after quenching in the cooling equipment 3 of the hot rolling production line and then tempering by reheating. Test pieces for the above tests are collected from the steel plate after tempering and transported to the finishing equipment at positions corresponding to reference points. Each test piece collected corresponding to a reference point is 150×150mm, and test pieces for the above tests are collected from these sample pieces. Reference points are set on the steel plate at five points: the front end, the rear end, the width-direction end (working side, driving side), and the center of the surface. Reference points 1 to 5 are individually set according to the steel plate to be manufactured. Two reference points are set in the width direction of the steel plate corresponding to the set reference points, with a distance of 80mm between the reference points and the reference points. Therefore, two reference points are established corresponding to each reference point.

[0168] The temperature measuring device for determining the surface temperature of the steel plate is positioned at two locations: 2m away from the inlet of the water-cooling device 4 located downstream of the hot rolling production line and 5m downstream of the water-cooling device 4. This allows for the measurement of the cooling start temperature and cooling stop temperature of the steel plate. In this embodiment, as a surface temperature information dataset for the steel plate, three temperature data points, including the temperature of the reference point, are obtained at two locations corresponding to a reference point, thus resulting in a surface temperature information dataset with six surface temperature data points for each reference point.

[0169] As described above, the surface temperature information dataset obtained for each reference point is used as learning data and stored in the database section 20a along with the information for identifying the steel plate and determining the reference point. Furthermore, the database section 20a stores performance data on the width and length of the steel plate, as well as performance data related to the composition of the steel plate, including C, Si, Mn, and Mo content, as attribute information parameters for each steel plate. Although the steel raw material used in this embodiment also contains other components, these components have little impact on the prediction accuracy of the material prediction model when combined with the operating parameters of the water cooling device 4, thus limiting the material prediction model to the main component composition of the steel raw material. Moreover, the database section 20a stores performance data on the total spray water volume, the number of spray zones, and the conveying speed of the steel plate within the cooling equipment 3, as operational performance data of the water cooling device 4. On the other hand, regarding the material information obtained for each reference point of the steel plate, the aforementioned test results using test pieces collected from test piece samples are stored in the database section 20a. In addition, in the database section 20a, based on the information of identifying the steel plate and the information of determining the reference point, a corresponding data set of surface temperature information, attribute information parameters of the steel plate, operational performance data of the water cooling device 4, and performance data of the material information of the steel plate are established.

[0170] Then, in the machine learning unit 20b of the material prediction model generation unit 20, a material prediction model M for the steel plate after tempering treatment by the cooling equipment 3 of the hot rolling production line is generated by machine learning using multiple learning data. This multiple learning data uses surface temperature information datasets, steel plate attribute information parameters, and operational performance data of the water cooling device 4 as input performance data, and outputs the material information of the steel plate corresponding to these input performance data as output performance data. The machine learning algorithm uses a neural network with four intermediate layers and six nodes. The activation function is a sigmoid function. Furthermore, the tempering treatment operation conditions performed on the steel plate after passing through the cooling equipment 3 are preset according to the steel plate standard and steel grade, resulting in small deviations in the tempering temperature during reheating. Therefore, compared to the operation parameters during the cooling process of the steel plate in the cooling equipment 3, the impact on the material information of the steel plate is relatively small, and a high-precision material information model of the steel plate can be generated even without using the tempering process operation parameters as input to the material prediction model.

[0171] The material prediction model M of the steel plate generated by the material prediction model generation unit 20 is sent to a tablet terminal that can communicate with the control computer of the finishing equipment, serving as the material prediction model of the material prediction unit internally configured therein. The control computer of the finishing equipment has a host computer 11 that is common to the control computer 10 of the hot rolling production line, and is structured to acquire performance data obtained by the control computer 10 of the hot rolling production line and performance data obtained by the surface temperature measuring device. Thus, in the material prediction unit, it is possible to acquire measured data of the surface temperature at the prediction reference point and prediction reference point of the steel plate obtained in the cooling equipment of the hot rolling production line, performance data of the attribute information parameters of the steel plate S, and operational performance data of the water cooling device 4. By inputting these data into the material prediction model M, it is possible to output the material information of the steel plate after passing through the cooling equipment at the position corresponding to the prediction reference point of the steel plate.

[0172] In this embodiment, 10 steel plates identical to those described above are prepared for testing and subjected to quenching and tempering treatment using the cooling equipment of the hot rolling production line. After the steel plates pass through the cooling equipment 3 of the hot rolling production line, the material prediction unit obtains the surface temperature information of the entire steel plate measured by the surface temperature measuring device. In the material prediction unit, while changing the prediction reference point within the steel plate surface, the material information of the steel plate at the position corresponding to each prediction reference point is output. Thus, the predicted value of the material information for the entire steel plate surface is obtained from the material prediction unit. Test samples are collected from three randomly selected positions within the steel plate surface from the steel plates manufactured for testing, and the material information of the steel plates is obtained through the same tests as described above. That is, a total of 30 material information data points are obtained from each of the 10 steel plates at three different positions. Furthermore, the predicted value of the material information at the prediction reference point corresponding to the position where the test sample was collected is output to the material prediction unit and compared with the actual material information data. Then, the error between the predicted and actual values ​​related to the material information of the steel plate is statistically analyzed, and its standard deviation σ is calculated, yielding the following results.

[0173] • The standard deviation σ of the prediction error is 12.3 MPa, relative to the average tensile strength data of 1005 MPa.

[0174] • The standard deviation of the prediction error σ is 9.2 MPa, relative to the average yield stress of 970 MPa.

[0175] • The average of the actual data relative to the ultimate bending radius (the ratio relative to the plate thickness) is 2.5, and the standard deviation of the prediction error is 0.34.

[0176] As can be seen from the above, the material prediction model M of the steel plate generated in this embodiment can predict the material information of the steel plate determined by tensile and bending tests with sufficient accuracy in practical applications.

[0177] The embodiments of the invention made by the inventors of this invention have been described above, but the invention is not limited to the description and drawings that constitute a part of the disclosure of the invention based on these embodiments. That is, all other embodiments, examples, and applications made by those skilled in the art based on these embodiments are included within the scope of this invention.

[0178] Industrial applicability

[0179] According to the present invention, a method for generating a material prediction model for a steel plate can be provided, which can generate a material prediction model for a steel plate with high accuracy in predicting the material information of the steel plate after passing through a cooling device. Furthermore, according to the present invention, a method for predicting the material of a steel plate can be provided, which can predict the material information of the steel plate after passing through a cooling device with high accuracy. Further, according to the present invention, a method and apparatus for manufacturing a steel plate capable of producing a steel plate with excellent material uniformity can be provided.

[0180] Label Explanation

[0181] 1. Offline heat treatment equipment

[0182] 2 Heating Furnace

[0183] 3. Cooling equipment

[0184] 4. Water cooling system

[0185] 5, 51, 52, 53 Temperature measuring devices

[0186] 6 rollers

[0187] 7. Water removal and purging nozzles

[0188] 7a Water removal purge air

[0189] 10. Control computer

[0190] 10a Water-cooled Condition Calculation Unit

[0191] 11 Host Computer

[0192] 20 Material Prediction Model Generation Department

[0193] 20a Database Department

[0194] 20b Machine Learning Department

[0195] 30 rolling mill

[0196] 41a, 41b water-cooled nozzles

[0197] 42a, 42b constraint rollers

[0198] 43 Water cooling flow control device

[0199] 54 Surface Temperature Information Generation Device

[0200] M Material Prediction Model

[0201] PA reference point

[0202] PB1, PB2, PB3, PB4 reference points

[0203] S steel plate

[0204] W cooling water.

Claims

1. A method for generating a material prediction model for a steel plate, which is a method for generating a material prediction model for a steel plate in a steel plate cooling device, wherein the steel plate cooling device comprises: a water cooling device for cooling the steel plate by spraying cooling water onto the heated steel plate; and a temperature measuring device for measuring the surface temperature of the steel plate during the cooling process, wherein... The method for generating the material prediction model of the steel plate includes the following steps: using machine learning with multiple learning data, a material prediction model of the steel plate after passing through the cooling device is generated. The multiple learning data includes a surface temperature information dataset, including measured surface temperature data at a pre-set reference point on the steel plate and measured surface temperature data at a reference point set based on the reference point, as input performance data. The material information of the steel plate after passing through the cooling device at the position corresponding to the reference point on the steel plate corresponding to the input performance data is used as output performance data.

2. The method for generating a material prediction model for steel plates according to claim 1, wherein, At least one reference point is set relative to the datum point in the length direction of the steel plate, and at least one reference point is set in the width direction.

3. The method for generating a material prediction model for steel plates according to claim 1, wherein, The material prediction model includes attribute information parameters selected from the attribute information of the steel plate as the input performance data.

4. The method for generating a material prediction model for steel plates according to claim 2, wherein, The material prediction model includes attribute information parameters selected from the attribute information of the steel plate as the input performance data.

5. The method for generating a material prediction model for steel plates according to any one of claims 1 to 4, wherein, The material prediction model includes at least one operational performance data selected from the operational performance data of the water cooling device as the input performance data.

6. The method for generating a material prediction model for steel plates according to any one of claims 1 to 4, wherein, As the machine learning method, a machine learning approach selected from neural networks, decision tree learning, random forests, and support vector regression is used.

7. The method for generating a material prediction model for steel plates according to claim 5, wherein, As the machine learning method, a machine learning approach selected from neural networks, decision tree learning, random forests, and support vector regression is used.

8. A method for predicting the material properties of a steel plate, which is a method for predicting the material properties of a steel plate in a steel plate cooling device, wherein the steel plate cooling device comprises: a water cooling device for cooling the steel plate by spraying cooling water onto the heated steel plate; and a temperature measuring device for measuring the surface temperature of the steel plate during the cooling process, wherein... The method for predicting the material of the steel plate includes the following steps: using a material prediction model generated by machine learning to predict the material information of the steel plate after passing through the cooling device. The machine learning takes a dataset of surface temperature information, including the surface temperature at a pre-set prediction reference point on the steel plate and the surface temperature at a prediction reference point set based on the prediction reference point, as input data, and takes the material information of the steel plate after passing through the cooling device at the position corresponding to the prediction reference point on the steel plate as output data.

9. A method for manufacturing a steel plate, comprising the following steps: using the material prediction method for steel plates according to claim 8 to determine whether the material of the steel plate after passing through the cooling equipment is qualified or not.

10. A steel plate manufacturing equipment, comprising: A steel plate cooling device includes a water cooling unit and a temperature measuring device. The water cooling unit cools the heated steel plate by spraying cooling water onto it, and the temperature measuring device measures the surface temperature of the steel plate during the cooling process. The material prediction unit outputs the material information of the steel plate after it has passed through the cooling equipment. The material prediction unit uses a machine learning model to output the material information of the steel plate. The machine learning model takes a dataset of surface temperature information, including the surface temperature at a pre-set prediction reference point on the steel plate and the surface temperature at a prediction reference point set based on the prediction reference point, as input data, and takes the material information of the steel plate after passing through the cooling device at the position corresponding to the prediction reference point on the steel plate as output data.

Images