KR102542332B1 - System for manufacturing hot rolled steel and operating method of the same - Google Patents

Abstract

A hot-rolled steel sheet production system according to an embodiment of the present invention includes calculation modules for determining process parameters for controlling a pickling process of a hot-rolled steel sheet, storage for storing learning models for learning the calculation modules, and a communication unit connected to a network. and a processor that controls the learning models to learn the calculation modules and transmits at least one of the calculation modules to an external server that operates the pickling apparatus that performs the pickling process through the communication unit. .

Description

Hot-rolled steel sheet production system and operation method thereof

The present invention relates to a hot-rolled steel sheet production system and an operation method thereof.

Steel plates include general steel and high carbon steel, and bearing shells for automobiles produced using steel plates may receive continuous and repeated loads on their surfaces. Therefore, strict surface quality is required for steel sheets used in the production of bearing shells and the like.

It is necessary to strictly manage the thickness of the internal defect layer present in the steel sheet, and a pickling process or the like for removing the internal defect layer may be applied to the steel sheet. The pickling process is a process of removing surface defects by washing the hot-rolled steel sheet with an acidic solution such as hydrochloric acid, and the internal defect layer can be removed by sufficiently washing the hot-rolled steel sheet in the pickling process.

Registered Patent Publication No. 10-2077281

One of the tasks to be achieved by the technical idea of the present invention is to predict the thickness of an internal defect layer occurring in a hot-rolled steel sheet and operate a pickling treatment device by learning an arithmetic module that controls the pickling process according to the thickness of the internal defect layer. It is an object of the present invention to provide a hot-rolled steel sheet production system and its operation method capable of improving the productivity of the hot-rolled steel sheet production process by providing it to the side.

A hot-rolled steel sheet production system according to an embodiment of the present invention includes a prediction calculation module for predicting the thickness of an internal defect layer occurring in a hot-rolled steel sheet, and process parameters for controlling a pickling process for removing the internal defect layer in the pickling process. a storage for storing a process calculation module that determines to be used in the process calculation module, and a learning model for training the process calculation module; a communication unit connected to a network to which an external server operating the pickling apparatus performing the pickling process is connected; and controlling the learning models stored in the storage to train the process calculation module so that a difference between the thickness of the internal defect layer predicted by the prediction calculation module and the thickness of the internal defect layer to be removed by the pickling process is reduced. a processor transmitting the process calculation module to the external server through a communication unit; The process parameters include the concentration of the pickling solution contained in the pickling tank of the pickling treatment device, the temperature of the pickling solution, whether or not an accelerator is used, whether or not a brush contacting the surface of the hot-rolled steel sheet is used, and the pickling It may include at least two of the feed rates of the hot-rolled steel sheet in the processing device.

In an operating method of a hot-rolled steel sheet production system according to an embodiment of the present invention, a process calculation module for determining process parameters for controlling a pickling process to be used in the pickling process is learned, and the pickling process is performed inside the hot-rolled steel sheet. a step of removing the defect layer; transmitting the process calculation module to an external system including a pickling device that performs the pickling process; Receiving, from the external system, the thickness of the remaining internal defect layer included in the pickling steel sheet after the pickling process and the process parameters measured by the pickling apparatus during the pickling process; and applying the received process parameters as weights or variables to the stored function so as to reduce a difference between the thickness of the internal defect layer predicted by the prediction operation module and the thickness of the remaining internal defect layer received from the external system. learning the module; The process parameters include the concentration of the pickling solution contained in the pickling tank of the pickling treatment device, the temperature of the pickling solution, whether or not an accelerator is used, whether or not a brush contacting the surface of the hot-rolled steel sheet is used, and the pickling It may include at least two of the feed rates of the hot-rolled steel sheet in the processing device.

According to one embodiment of the present invention, a hot-rolled steel sheet production system capable of improving the productivity of a pickling process by controlling a pickling treatment device to apply optimal process parameters to each region of a hot-rolled steel sheet and an operation method thereof are provided. can

In addition, optimal process parameters can be determined more accurately by measuring the thickness of the internal defect layer of the pickling steel sheet that has completed the pickling process and learning calculation modules for determining process parameters using the measured thickness of the internal defect layer. .

The various beneficial advantages and effects of the present invention are not limited to the above, and will be more easily understood in the process of describing specific embodiments of the present invention.

1 is a view for explaining a manufacturing process of a hot-rolled steel sheet according to an embodiment of the present invention.
2 is a diagram simply showing a hot-rolled steel sheet according to an embodiment of the present invention.
3 and 4 are block diagrams showing a hot-rolled steel sheet production system according to an embodiment of the present invention.
5 is a diagram for explaining an initial learning method of a hot-rolled steel sheet production system according to an embodiment of the present invention.
6 is a diagram for explaining a learning model included in a hot-rolled steel sheet production system according to an embodiment of the present invention.
7 is a diagram for explaining an operating method of a hot-rolled steel sheet production system according to an embodiment of the present invention.
8 is a diagram for explaining an operating method of a hot-rolled steel sheet production system according to an embodiment of the present invention.
9 is a diagram for explaining an operating method of a hot-rolled steel sheet production system according to an embodiment of the present invention.
10 is a diagram for explaining an operating method of a hot-rolled steel sheet production system according to an embodiment of the present invention.
11 is a flowchart illustrating an operating method of a hot-rolled steel sheet production system according to an embodiment of the present invention.
12 and 13 are views simply illustrating an automated pickling system interlocked with the hot-rolled steel sheet production system according to an embodiment of the present invention.

Hereinafter, preferred embodiments of the present invention will be described in detail with reference to the accompanying drawings. The same reference numerals are used for the same components in the drawings, and redundant descriptions of the same components will be omitted.

1 is a view for explaining a manufacturing process of a hot-rolled steel sheet according to an embodiment of the present invention.

1 may be a diagram simply illustrating a hot-rolled steel sheet production apparatus for producing a hot-rolled steel sheet, cooling and winding the hot-rolled steel sheet to produce a hot-rolled coil. Referring to FIG. 1 , a hot rolling process may be performed in which a slab heated in a heating furnace is rolled to a predetermined thickness using a roughing mill and a finishing mill (or a finishing mill). The hot-rolled steel sheet (strip) produced through the hot rolling process can be moved to the run-out table (ROT) cooling zone, which is a cooling section. The hot-rolled steel sheet may be cooled by cooling water sprayed from the ROT cooling table, and the cooling temperature and cooling time may be determined according to the quality required for the hot-rolled steel sheet.

After cooling, the hot-rolled steel sheet may be wound into a coil in a winding machine for storage and/or transport convenience. Hot-rolled coils (HC, Hot Coil), in which hot-rolled steel sheet is wound into a coil form, are deposited in the yard, air-cooled, and shipped as products. Then, a pickling process to remove the internal defect layer that exists in the hot-rolled steel sheet will be performed. can In one embodiment, the hot-rolled steel sheet manufacturing process described with reference to FIG. 1 and the pickling process for removing the internal defect layer of the hot-rolled steel sheet may be performed by different entities.

2 is a diagram simply showing a hot-rolled steel sheet according to an embodiment of the present invention.

Referring to FIG. 2 , a hot-rolled steel sheet may be divided into a plurality of regions along the longitudinal direction (X-axis direction). For example, a hot-rolled steel sheet may include a first region, a second region, a third region, and the like that are sequentially disposed along the longitudinal direction. In one embodiment shown in FIG. 2, the first region, the second region, and the third region may be sequentially wound. In other words, the first region may be wound first, then the second region, and the third region may be wound last. The first to third regions are defined according to the winding order, and the length and/or area of each of the first to third regions is not limited as shown in FIG. 2 .

In one embodiment, the hot-rolled steel sheet may include surface defects. For example, the surface defect may include at least some of scale, an internal oxide layer, and a decarburized layer. Scale occurs during the rolling process and may be present on the surface of the material. The internal defect layer may be included under the surface of the material, that is, inside the material, and may be defined as a concept including an internal oxidation layer and a decarburized layer. In the internal defect layer, components such as chromium (Cr), manganese (Mn), silicon (Si), zinc (Zn), magnesium (Mg), and aluminum (Al), which have higher oxygen affinity than iron (Fe), are formed in the base material. Oxidation can occur during the process. The decarburized layer may be generated in the process of combining steel carbon with atmospheric and scale oxygen and then discharging them into the atmosphere in the form of a gas. The thickness of the internal defect layer may vary depending on the composition of the hot-rolled steel sheet, the temperature at which the hot-rolled steel sheet is wound into the hot-rolled coil (HC), the cooling time after winding, the width, thickness, and length of the hot-rolled steel sheet. Surface defects, such as an internal defect layer, may be a factor in reducing durability of products produced using hot-rolled steel sheets.

For example, the temperature of the wound hot-rolled coil (HC) may be about 500 to 600 ° C., and the wound hot-rolled coil (HC) may be cooled by air-cooling while being exposed to air. In the hot-rolled coil HC in a wound state, the first and third regions exposed to the outside are cooled relatively quickly, whereas the second region not exposed to the outside may be cooled relatively slowly. Accordingly, the thickness of the internal defect layer occurring in the second region of the hot-rolled steel sheet may be greater than the thickness of the internal defect layer occurring in each of the first region and the second region of the hot-rolled steel sheet. Accordingly, the thickness of the internal defect layer may vary according to regions of the hot-rolled steel sheet.

Also, according to embodiments, the thickness of the internal defect layer may vary according to regions in the width direction of the hot-rolled steel sheet. For example, regions adjacent to corners of the hot-rolled steel sheet may be cooled relatively quickly in the width direction, and a difference in thickness of the internal defect layer according to the cooling rate may also appear in the width direction.

The pickling process for removing the internal defect layer may be performed by transferring the hot-rolled steel sheet while the hot-rolled steel sheet is immersed in the pickling solution. For example, the pickling process may be performed over a sufficient period of time regardless of the area of the hot-rolled steel sheet so that the internal defect layer included in the hot-rolled steel sheet can be sufficiently removed. However, the method as described above leads to an increase in the time of the pickling process and/or the amount of pickling solution introduced into the pickling process, which may result in a decrease in productivity.

In one embodiment, the thickness of the internal defect layer generated in the hot-rolled steel sheet may be determined according to the components of the hot-rolled steel sheet, the temperature of the hot-rolled steel sheet during winding, the temperature change of the hot-rolled steel sheet after winding, and the like. Therefore, when the subject producing the hot-rolled steel sheet and the subject performing the pickling process are separated, the subject producing the hot-rolled steel sheet predicts the thickness of the internal defect layer in advance and provides it to the subject performing the pickling process. process can be carried out efficiently.

According to an embodiment of the present invention, the thickness of the internal defect layer included in the hot-rolled steel sheet can be predicted, and process parameters capable of controlling the pickling process with optimum efficiency can be calculated according to the predicted thickness of the internal defect layer. Process parameters may be provided to a subject performing the pickling process. Alternatively, the main body producing the hot-rolled steel sheet predicts the thickness of the internal defect layer and provides it to the main body performing the pickling process, and the main body performing the pickling process uses the thickness of the internal defect layer to control the pickling process. It is also possible to calculate the necessary process parameters.

According to embodiments, a calculation module for controlling the pickling process based on process parameters may be provided by learning and tuning the thickness of the internal defect layer with a machine learning model or the like on the side of the hot-rolled steel sheet producer. . Therefore, productivity can be improved by shortening the time of the pickling process and reducing the amount of pickling solution used. In addition, it is possible to reduce the thickness deviation of the internal defect layer appearing for each region in the pickled steel sheet after the pickling process is completed.

3 and 4 are block diagrams showing a hot-rolled steel sheet production system according to an embodiment of the present invention.

First, referring to FIG. 3 , the hot-rolled steel sheet production system 1 according to an embodiment of the present invention may include a first system 10 and a network 20 connected to the first system 10. . The first system 10 may be operated by an entity that produces a hot-rolled steel sheet and winds the hot-rolled steel sheet to produce a hot-rolled coil.

In one embodiment shown in FIG. 3, the network 20 may mean a wired Internet network, a wireless Internet network, or a wireless local area network (WLAN) such as Wi-Fi (wireless fidelity). can The wired Internet network or the wireless Internet network is an Internet TCP/IP protocol and various services existing in the upper layer, that is, HTTP (HyperText Transfer Protocol), Telnet, FTP (File Transfer Protocol), DNS (Domain Name System) ), Simple Mail Transfer Protocol (SMTP), Simple Network Management Protocol (SNMP), Network File Service (NFS), Network Information Service (NIS), and the like.

In one embodiment, the first system 10 may include arithmetic modules 11 that determine process parameters for controlling the pickling process. In addition, the first system 10 may include a learning model 12 for training at least some of the calculation modules 11 .

For example, the first system 10 may be connected to the second system through the network 20 . The second system may include a pickling treatment device that performs the pickling process of the hot-rolled steel sheet, and may be operated by a second subject that performs the pickling process, different from the first subject. For example, the pickling treatment device of the second system may perform a pickling process by unwinding a hot-rolled coil wound with a hot-rolled steel sheet to form a strip again, transporting the hot-rolled steel sheet in the form of a strip, and immersing the hot-rolled steel sheet in a pickling solution inside a pickling tank. The pickled steel sheet after the pickling process may be wound into a pickling coil shape again.

The first system 10 may transmit at least one of the calculation modules 11 to the second system so that the second system can effectively control the pickling device. For example, the calculation module transmitted to the second system may determine process parameters for controlling the pickling apparatus. The process parameters may include the concentration and temperature of the pickling solution contained in the pickling tank of the pickling treatment device, whether or not an accelerator is used, whether a brush is used in contact with the surface of the hot-rolled steel sheet, and the transfer speed of the hot-rolled steel sheet in the pickling treatment device. can For example, the calculation module transmitted to the second system may adjust the transfer speed of the hot-rolled steel sheet during the pickling process in consideration of the thickness deviation of the internal defect layer present in the hot-rolled steel sheet. For example, while a region in a hot-rolled steel sheet where the thickness of the internal defect layer is expected to be large is immersed in the pickling solution, the feed rate decreases, and an area in the hot-rolled steel sheet where the thickness of the internal defect layer is expected to be small is immersed in the pickling solution. During this time, the feed rate may increase.

The first system 10 may learn and transmit the calculation modules 11 to the second system so that the pickling process performed in the second system is optimized. Also, the first system 10 may receive data necessary for learning the calculation modules 11 as feedback from the second system. The first system 10 may optimize the calculation modules 11 by training the calculation modules 11 using data fed back from the second system. The data fed back by the first system 10 are equipment parameters acquired by the second system from the pickling treatment device while the pickling process is in progress, and residual internal defects measured at at least one location of the pickled steel sheet after the pickling process is completed. It may include at least one of the thickness of the layer.

In operation, the calculation modules 11 of the first system 10 may predict the thickness of the internal defect layer included in the hot-rolled steel sheet. For example, the calculation modules 11 may predict the thickness of the internal defect layer according to the length direction of the hot-rolled steel sheet. The first system 10 may transmit the thickness of the internal defect layer predicted by the calculation modules 11 to the second system through the network 20 .

As described above, the second system may receive and store at least some of the calculation modules 11 of the first system 10 . In an embodiment, the calculation module of the second system may use the thickness of the internal defect layer predicted by the first system 10 to determine process parameters for controlling the pickling apparatus. The pickling apparatus may be controlled by the process parameters determined by the calculation module of the second system.

For example, process parameters input to the pickling apparatus may be determined as values capable of optimizing a pickling process for each of regions defined along the length direction of the hot-rolled steel sheet. For example, the feed rate of the hot-rolled steel sheet during the pickling process for the first region or the third region corresponding to the end of the hot-rolled steel sheet is determined during the pickling process for the second region located between the first region and the third region. It may be different from the feed rate of the steel plate. For example, the conveyance speed of the hot-rolled steel sheet during the pickling process for the first region or the third region may be higher than the conveyance speed of the hot-rolled steel sheet during the pickling process for the second region located between the first region and the third region. . In one embodiment, scale present on the surface of the hot-rolled steel sheet may be removed together during the pickling process.

The calculation modules 11 of the first system 10 may include a first module that predicts the phase fraction before winding the hot-rolled steel sheet. In addition, the calculation modules 11 may include a second module for predicting the temperature in each of the regions along the length direction of the hot-rolled steel sheet. Also, the calculation modules 11 may include a third module that predicts the thickness of the internal defect layer included in each of the regions along the length direction of the hot-rolled steel sheet.

In an embodiment, the second module may predict the temperature using the phase fraction calculated by the first module. The temperature predicted by the second module may be a temperature predicted in each of the regions according to an elapsed time after the hot-rolled steel sheet is wound in the form of a hot-rolled coil (HC). Also, the third module may predict the thickness of the internal defect layer included in each of the regions using the temperature calculated by the second module. The thickness of the internal defect layer predicted by the third module is transmitted to the second system, and the calculation module of the second system uses the thickness of the internal defect layer calculated by the third module to determine process parameters for controlling the pickling treatment device. can decide

In addition, in one embodiment, the calculation modules 11 of the first system 10 may include a fourth module that predicts the thickness of the internal defect layer removed by the pickling device based on at least one of the process parameters. can Also, the calculation modules 11 may include a fifth module for determining process parameters for controlling the pickling apparatus. For example, the fifth module determines process parameters by using the thickness of the internal defect layer calculated by the fourth module to be removed by the pickling device and the thickness of the internal defect layer calculated by the third module to be present in the hot-rolled steel sheet. can decide The first system 10 transmits the fourth and fifth modules to the second system, and the second system stores the fourth and fifth modules and can use them to control the pickling apparatus.

The learning model 12 may train at least one of the calculation modules 11 . For example, the learning model 12 may train the calculation modules 11 by comparing a predetermined reference value with an output value output from at least one of the calculation modules 11 . The learning model 12 may train the calculation modules 11 until the reference value and the output value match or the difference between the reference value and the output value is within a predetermined range. For example, the learning model 12 may calculate a difference between a reference value and an output value using a predetermined loss function, and may train the calculation modules 11 according to whether the difference is within a predetermined range. In an embodiment, the learning model 12 may use a mean squared error, a cross entropy error, or the like as a loss function for calculating a difference between a reference value and an output value.

In one embodiment, the second system may obtain data required for the learning model 12 to train at least one of the calculation modules 11, and may feed the data back to the first system 10. For example, the second system may obtain device parameters corresponding to process parameters input to the pickling device at at least one measurement point in time while the pickling device is performing the pickling process. The device parameters may have different values from the process parameters due to mechanical errors and/or operational delays of the pickling device.

Taking the hot-rolled steel sheet feed rate as an example among the process parameters input to the pickling device, the feed rate of the hot-rolled steel sheet input to the pickling device and the actual feed rate for transferring the hot-rolled steel sheet in the pickling device may not match. can In addition, among the process parameters input to the pickling apparatus, the concentration of the pickling solution and/or the temperature of the pickling solution may change during pickling of the hot-rolled steel sheet. Accordingly, the device parameters may be obtained by measuring the actual transport speed of the hot-rolled steel sheet, the concentration of the pickling solution, the temperature of the pickling solution, and whether or not an accelerator is used at at least one measurement point during the pickling process.

In addition, the second system may obtain the thickness of the remaining internal defect layer included in the pickled steel sheet after the pickling process is completed. For example, the thickness of the remaining internal defect layer may be obtained by designating at least one measurement location on the pickled steel sheet where the pickling process is completed and measuring the thickness of the internal defect layer at the measurement location. According to embodiments, the measurement position may be designated as a plurality of different positions along the length direction of the pickling steel sheet.

The second system transmits the location information of the pickling steel sheet indicating the measurement location, the thickness of the residual internal defect layer obtained at the measurement location, and the device parameters obtained during the pickling process for the measurement location through the network 20 to the first system 10. ) can be transmitted.

The learning model 12 of the first system 10 may learn at least some of the calculation modules 11 using the data received from the second system. As described above, the learning model 12 may train at least some of the calculation modules 11 using a loss function. For example, a fourth module for predicting the thickness of the internal defect layer among the calculation modules 11 may be trained by applying the thickness of the remaining internal defect layer according to the position of the pickling steel sheet transmitted by the second system as a reference value. . When learning of the calculation modules 11 is completed, the first system 10 may transmit at least one of the learned calculation modules 11, for example, a fourth module and a fifth module to the second system. there is.

The second system may update the previously stored calculation module by using the new calculation module received from the first system 10 . For example, the second system may overwrite the existing calculation module with a new calculation module received from the first system 10 .

4 may be a simple block diagram of a first system 30 included in a hot-rolled steel sheet production system according to an embodiment of the present invention. The first system 30 may include a first server 31 and a hot-rolled steel sheet production device 35 and the like, and the first server 31 may control and manage the hot-rolled steel sheet production device 35 . The hot-rolled steel sheet production device 35 may be a device that produces a hot-rolled steel sheet by rolling a slab heated in a heating furnace, and cools and winds the hot-rolled steel sheet.

The first server 31 may include a communication unit 32 , a storage 33 , a processor 34 , and the like. The communication unit 32 may connect the first server 31 and the network to enable communication. The storage 33 may store data required for operation of the first server 31 and management of the hot-rolled steel sheet production device 35 . The processor 34 may control the communication unit 32 , the storage 33 , and the hot-rolled steel sheet production device 35 .

For example, the storage 33 may store calculation modules that execute predetermined calculations and learning models for optimizing the calculation modules. At least one of the calculation modules may be interlocked with the hot-rolled steel sheet production device 35 and perform calculation using data obtained from the hot-rolled steel sheet production device 45 . also. At least one of the arithmetic modules may execute an arithmetic operation for controlling a device included in another system connected to the first system 30 through a network. In one embodiment, at least one of the calculation modules is included in a second system connected to the first system 30 through a network and performs a pickling process of the hot-rolled steel sheet produced by the hot-rolled steel sheet production device 35. It is possible to execute calculations for optimally controlling the processing unit.

5 is a diagram for explaining an initial learning method of a hot-rolled steel sheet production system according to an embodiment of the present invention.

Referring to FIG. 5 , the first system SYS1 may include calculation modules 110 and a learning model 120 . The calculation modules 110 may include a prediction module 111 and a parameter calculation module 112 . In one embodiment, the calculation modules 110 may include first to fifth modules M1 to M5, and the prediction module 111 may include first to third modules M1 to M3. , and the parameter calculation module 112 may include a fourth module M4 and a fifth module M5. The first to fifth modules M1 to M5 may be implemented with hardware such as a circuit or software such as a source code.

For example, when the first to fifth modules M1 to M5 are implemented as software such as source code, the first to fifth modules M1 to M5 calculate output data using input data. calculations can be executed. Operations of each of the first to fifth modules M1 to M5 may be determined and optimized by the learning model 120 .

The learning model 120 may include first to fourth learning models LM1 to LM4 and a feedback learning model LMT. The first learning model LM1 may be a model for initially learning the first module M1, and the second learning model LM2 may be a model for initially learning the second module M2. Similarly, the third learning model LM3 may be a model for initially learning the third module M3, and the fourth learning model LM4 may be a model for initially learning the fourth module M4. can The feedback learning model LMT may be a model for learning at least one of the first to fifth modules M1 to M5 after the pickling process is completed. In one embodiment, the feedback learning model (LMT) may simultaneously train the first to fifth modules M1 to M5.

The learning model 120 may learn each of the first to fifth modules M1 to M5 using a deviation correction method or reinforcement learning. In the case of using the deviation correction method, the learning model 120 may train the first to fifth modules M1 to M5 by correcting an error between a measurement value input from the outside and a predicted value output from the module. When the learning model 120 uses reinforcement learning, the learning model 120 may be a deep neural network (DNN), but is not necessarily limited to this example.

An embodiment in which the learning model 120 learns each module using reinforcement learning will be described later with reference to FIG. 6 . In addition, a learning method of the feedback learning model (LMT) will be described later with reference to FIG. 10 . Hereinafter, the initial learning method of the first system SYS1 will be described in detail with reference to FIG. 5 .

In one embodiment, the first module M1 may predict the phase fraction before winding the hot-rolled steel sheet. The first learning model LM1 may receive the first process condition CD1 and the first phase fraction value EPF. The first phase fraction value EPF may include a phase transformation fraction or phase transformation amount actually measured under the first process condition CD1. The first process condition CD1 may include a cooling rate of the hot-rolled steel sheet, a temperature of the hot-rolled steel sheet, and components of the hot-rolled steel sheet.

The first learning model LM1 may input the first process condition CD1 to the first module M1. The first module M1 may calculate a second phase fraction value PPF by predicting the phase fraction of the hot-rolled steel sheet using the first process condition CD1. The second phase fraction value PPF output by the first module M1 may be transmitted to the first learning model LM1.

The first learning model LM1 may learn the first module M1 by comparing the first phase fraction value EPF and the second phase fraction value PPF. For example, if the first phase fraction value (EPF) and the second phase fraction value (PPF) do not match, the first learning model (LM1) receives the first process condition (CD1) from the first module (M1) and The first function predicting the fraction may be adjusted. The first learning model LM1 is configured such that the first phase fraction value EPF and the second phase fraction value PPF are identical or the difference between the first phase fraction value EPF and the second phase fraction value PPF is The first function of the first module M1 may be adjusted to be less than or equal to a predetermined value. In this case, the first learning model LM1 may calculate the difference between the first phase fraction value EPF and the second phase fraction value PPF by using a predetermined loss function. In an embodiment, the first learning model LM1 may learn the first module M1 by adjusting a weight or the like used by the first module M1 for calculation.

The second learning model LM2 may receive the second process condition CD2, the environmental condition FV indicating the surrounding environment, and the first temperature value ET measured from the hot-rolled steel sheet. The second process condition CD2 may include a time point at which the temperature of the hot-rolled steel sheet is measured after winding, and components of the hot-rolled steel sheet. The environmental condition FV may include a factor value representing the surrounding environment that affects the cooling rate of the hot-rolled steel sheet after coiling. For example, if the surrounding environment is air, the factor value of the surrounding environment may be '1', if the surrounding environment is wind, the factor value of the surrounding environment may be '2', and if the surrounding environment is water, the factor value of the surrounding environment may be may be '3'. The first temperature value ET may include a value actually measured for the temperature of the hot-rolled coil.

The second module M2 may receive the second process condition CD2 and the environmental condition FV from the second learning model LM2. For example, the second module M2 may predict the temperature of the hot-rolled steel sheet using the second process condition CD2 and the environmental condition FV, and outputs a second temperature value PT by predicting the temperature of the hot-rolled steel sheet. can do. The second temperature value PT may include a value predicted according to an elapsed time after winding the temperature of each region defined in the longitudinal direction of the hot-rolled steel sheet. The second module M2 may output the second temperature value PT as the second learning model LM2.

The second learning model LM2 may learn the second module M2 by comparing the first temperature value ET and the second temperature value PT. Similar to the first learning model LM1, the second learning model LM2 may compare the first temperature value ET and the second temperature value PT using a loss function. For example, when the first temperature value ET and the second temperature value PT do not match or the difference between the first temperature value ET and the second temperature value PT is greater than a predetermined value, the second learning The model LM2 may adjust the second function for calculating the second temperature value PT in the second module M2. The second learning model LM2 is configured such that the first temperature value ET and the second temperature value PT are identical or the difference between the first temperature value ET and the second temperature value PT is less than or equal to a predetermined value. It is possible to learn the second module (M2) so that.

The third learning model (LM3) is a component of the hot-rolled steel sheet (ING), an oxygen partial pressure (OPP) around the hot-rolled steel sheet, a second temperature value (PT) that is an output value of the second module (M2), and an internal measurement from the hot-rolled steel sheet. The first thickness ED of the defect layer may be input. The first thickness ED of the internal defect layer may refer to a value actually measured for the thickness of the internal defect layer of the hot-rolled steel sheet.

The third module (M3) receives the component (ING) of the hot-rolled steel sheet, the oxygen partial pressure (OPP) around the hot-rolled steel sheet, and the second temperature value (PT) of the hot-rolled steel sheet as input data from the third learning model (LM3). can The third module M3 may calculate second thickness information PD of the internal defect layer. The second thickness information PD is obtained by the third module M3 by using at least one of the component ING of the hot-rolled steel sheet, the oxygen partial pressure around the hot-rolled steel sheet OPP, and the second temperature value PT. It may be the thickness of the internal defect layer predicted to exist. The second thickness information PD of the internal defect layer output from the third module M3 may be transmitted to the third learning model LM3.

The third learning model LM3 may learn the third module M3 by comparing the first thickness ED of the internal defect layer with the second thickness information PD. For example, if the first thickness (ED) and the second thickness information (PD) of the internal defect layer do not match, or if the difference is greater than a predetermined value, the third learning model (LM3) is internal in the third module (M3). A third function for calculating the second thickness information PD of the defect layer may be adjusted. The third learning model LM3 uses the third module M3 so that the first thickness ED, which is the measured value of the internal defect layer, and the second thickness information PD, which is the predicted value, match or the difference is less than or equal to a predetermined value. can be learned

The fourth learning model LM4 receives the transport speed of the hot-rolled steel sheet (PV), the characteristics of the pickling solution contained in the pickling tank (AC), and the first thickness of the internal defect layer removed by the pickling treatment device (EPA). can The characteristics (AC) of the pickling solution may include the concentration of the pickling solution in the pickling tank, the temperature of the pickling solution, whether or not an accelerator is used, and the like. The transport speed (PV) of the hot-rolled steel sheet may mean a moving speed of the hot-rolled steel sheet in the pickling tank. The first thickness (EPA) of the internal defect layer may mean the thickness of the internal defect layer actually removed in the pickling process performed by the feed rate (PV) of the hot-rolled coil and the characteristic (AC) of the pickling solution.

The fourth module M4 may receive the transfer speed (PV) of the hot-rolled steel sheet and the characteristic (AC) of the pickling solution from the fourth learning model LM4. The fourth module (M4) calculates the second thickness (PPA1) of the internal defect layer expected to be removed by the pickling treatment device using at least one of the transport speed (PV) of the hot-rolled steel sheet and the characteristic (AC) of the pickling solution. can In one embodiment, the fourth module (M4) receives the pickling time determined according to the transfer speed (PV) and the characteristic (AC) of the pickling solution as input values, and removes the internal defect layer using the input values. 2 Calculation to calculate the thickness (PPA1) can be performed. The fourth module M4 may output the second thickness PPA1 of the internal defect layer as the fourth learning model LM4.

The fourth learning model LM4 may learn the fourth module M4 by comparing the second thickness PPA1 of the internal defect layer with the first thickness EPA of the internal defect layer. For example, if the first thickness EPA of the internal defect layer and the second thickness PPA1 of the internal defect layer do not match or the difference is greater than a predetermined value, the fourth learning model LM4 performs the fourth module M4 In , weights and variables used to calculate the second thickness PPA1 of the internal defect layer may be adjusted.

The fifth module M5 may determine process parameters input to the pickling apparatus. For example, the fifth module M5 may find optimal process parameters by repeatedly calling the fourth module M4 using an optimization technique (eg, golden section method). The fifth module M5 may select an optimization technique for finding optimal process parameters or modify the optimization technique.

For example, the fifth module (M5) determines the transfer speed (PV) for transferring the hot-rolled steel sheet in the pickling process as the first speed and the concentration included in the characteristic (AC) of the pickling solution as the first concentration, and then determines the fourth The module M4 may be called to receive the second thickness PPA1 of the internal defect layer expected to be removed by the pickling treatment device. When the second thickness (PPA1) of the internal defect layer does not match the target of the pickling process, the fifth module (M5) corrects at least one of the feed rate (PV) and the characteristic (AC) of the pickling solution and then removes it again. 4 The second thickness PPA1 of the internal defect layer may be received by calling the module M4. In this way, by repeatedly calling the fourth module (M4) while modifying the feed rate (PV) and/or the characteristic (AC) of the pickling solution, the process parameters optimized for the target to be achieved by the pickling process are set in the fifth module (M5). ) can be determined.

6 is a diagram for explaining a learning model included in a hot-rolled steel sheet production system according to an embodiment of the present invention.

Referring to FIG. 6 , when the learning model learns each module using reinforcement learning, the learning model may be implemented as a deep neural network (DNN) or the like. The learning model may include an input layer (IL), a hidden layer (HL), and an output layer (OL). For example, a plurality of nodes included in the input layer IL, the hidden layer HL, and the output layer OL may be fully connected to each other. The input layer IL may include a plurality of input nodes x1-xi, and the number of input nodes x1-xi may correspond to the number of input data. The output layer OL may include a plurality of output nodes y1-yj, and the number of output nodes y1-yj may correspond to the number of output data.

The hidden layer HL may include first to third hidden layers HL1 to HL3, and the number of hidden layers HL1 to HL3 may be variously modified. For example, the learning model 120 may be learned by adjusting weights of each of the hidden nodes included in the hidden layer HL. Also, unlike the embodiment shown in FIG. 6 , at least one of the hidden nodes of the hidden layers HL1 to HL3 may recurrently send an output value to a hidden node of the same layer.

7 is a diagram for explaining an operating method of a hot-rolled steel sheet production system according to an embodiment of the present invention.

Referring to FIG. 7 , the hot-rolled steel sheet production system 200 according to an embodiment of the present invention may include arithmetic modules 210 to 230. For example, the calculation modules 210 to 230 may be stored and executed in a first system among systems included in the hot-rolled steel sheet production system 200 . The first system may be a system connected to a second system that directly controls the pickling apparatus through a network.

In one embodiment, the calculation modules 210 to 230 may predict the thickness of the internal defect layer included in the hot-rolled steel sheet. In one embodiment, in order to calculate the thickness of the internal defect layer, a cooling rate of the hot-rolled steel sheet along the length direction of the hot-rolled steel sheet may be required. If the phase transformation is not completed before winding the hot-rolled steel sheet, heat generation due to the phase transformation may occur after winding. Therefore, the temperature change of the hot-rolled steel sheet after coiling may be calculated in consideration of the phase fraction before coiling of the hot-rolled steel sheet.

In an embodiment, the first module 210 may calculate the phase fraction (PPF) according to the cooling time using the first process condition CD1. The first process condition CD1 may include a cooling rate of the hot-rolled steel sheet, a temperature of the hot-rolled steel sheet, and components of the hot-rolled steel sheet.

In a hot-rolled steel sheet, a phase transformation from austenite to pearlite may occur during a cooling process, and transformation exotherm may occur due to the phase transformation. The phase transformation of hot-rolled high-carbon steel may not be completed during cooling due to slow pearlite transformation, and additional phase transformation may occur in the hot-rolled steel sheet after coiling. As a result, the hot-rolled steel sheet may be exposed to a high-temperature oxidizing atmosphere for a long time after winding, and the thickness of the internal defect layer may increase.

The second module 220 may predict the temperature (PT) according to the elapsed time after winding of the hot-rolled steel sheet using the second process condition (CD2) and the environmental condition (FV). For example, the second module 220 may predict the temperature PT in each of the regions along the length direction of the hot-rolled steel sheet. The second process condition CD2 may include a time point at which the temperature of the hot-rolled steel sheet is measured after winding, and components of the hot-rolled steel sheet.

In order to predict the temperature of the rolled hot-rolled steel sheet, the transformation heat generation described above should be considered, so the second module 220 may receive the phase fraction (PPF) from the first module 210 . Based on the phase fraction (PPF), the phase fraction at the time of winding can be known, and the amount of phase transformation additionally occurring in the hot-rolled steel sheet after winding can be known. Accordingly, the second module 220 may predict the temperature PT of the hot-rolled steel sheet according to the elapsed time after winding the hot-rolled steel sheet in consideration of the amount of phase transformation additionally occurring in the rolled hot-rolled steel sheet. In one embodiment shown in FIG. 8 , time point t1 may indicate a temperature at which transformation exotherm due to phase transformation is reflected.

The third module 230 may predict the thickness (PD) of the internal defect layer included in the hot-rolled steel sheet by using the component (ING) of the hot-rolled steel sheet, the oxygen partial pressure (OPP) around the hot-rolled steel sheet, and the like. In one embodiment, the third module 230 receives the temperature (PT) according to the elapsed time after winding in each of the regions of the hot-rolled steel sheet from the second module 220, and internally in each of the regions of the hot-rolled steel sheet. The thickness (PD) of the defect layer can be predicted.

However, according to embodiments, without the first module 210 predicting the phase fraction (PPF) and the second module 220 predicting the temperature PT of the hot-rolled steel sheet based on the phase fraction (PPF), 3 module 230 may predict the thickness PD of the internal defect layer. Without the first module 210 and the second module 220, the temperature of the hot-rolled steel sheet is actually measured and provided to the third module 230 so that the third module 230 can predict the thickness PD of the internal defect layer. can do. Hereinafter, it will be described with reference to FIG. 8 .

8 is a diagram for explaining an operating method of a hot-rolled steel sheet production system according to an embodiment of the present invention.

In the embodiment shown in FIG. 8 , the third module 230 without the first module 210 and the second module 220 may estimate the thickness PD of the internal defect layer present in the hot-rolled steel sheet. As described with reference to FIG. 7 , in order for the third module 230 to predict the thickness PD of the internal defect layer, data on the temperature PT of the hot-rolled steel sheet may be required. In the embodiment shown in FIG. 8 , the temperature change on the surface of the hot-rolled coil HC may be directly measured using the infrared camera 240 installed above the hot-rolled coil HC. The control unit 250 interworking with the infrared camera 240 may generate data including temperature change of the hot-rolled coil HC according to time and transmit the generated data to the third module 230 . According to embodiments, the temperature change may be first measured using the infrared camera 240 before the hot-rolled steel sheet is wound into the hot-rolled coil HC.

In FIG. 8, it has been described that the temperature of the hot-rolled coil HC is measured using one infrared camera 240, but the number and location of the infrared cameras 240 may be variously modified according to embodiments. For example, a plurality of infrared cameras 240 may be disposed to measure temperatures at different locations of the hot-rolled coil HC. In this case, according to embodiments, some of the infrared cameras 240 may measure the temperature of the hot-rolled coil HC from a side of the hot-rolled coil HC, not from the top.

The infrared camera 240 detects thermal infrared rays emitted from the hot-rolled coil HC, and the controller 250 may detect a temperature change of the hot-rolled coil HC using the thermal infrared rays detected by the infrared camera 240. there is. The first module 210 predicts the phase fraction by directly measuring the temperature change of the hot-rolled coil HC using the infrared camera 240 and transmitting it to the third module 230, and the temperature change according to the phase fraction The third module 230 may predict the thickness PD of the internal defect layer without the second module 220 predicting . Therefore, it is possible to reduce the amount of computation as a whole in the system. In addition, since the actual temperature measured from the hot-rolled coil (HC) and/or the hot-rolled steel sheet can be used in the learning process of the third module 230, the accuracy of the third module 230 can be improved.

However, with the thermal infrared rays detected by the infrared camera 240, only the surface temperature of the hot-rolled coil (HC) or the hot-rolled steel sheet can be measured. Accordingly, the control unit 250 may further include a calculation module that calculates the internal temperature as well as the surface temperature of the hot-rolled coil HC or the hot-rolled steel sheet by using the thermal infrared rays detected by the infrared camera 240 .

9 is a diagram for explaining an operating method of a hot-rolled steel sheet production system according to an embodiment of the present invention.

Referring to FIG. 9 , the hot-rolled steel sheet production system 300 may include calculation modules 310 and 320 . For example, the calculation modules 310 and 320 are stored in a first server of a first system for producing a hot-rolled steel sheet, among systems included in the hot-rolled steel sheet production system 300, and learned by the first server. can be modules. In addition, the calculation modules 310 and 320 may be stored in a second system including a pickling apparatus that is connected to the first system through a network and performs a pickling process of the hot-rolled steel sheet. For example, the calculation modules 310 and 320 may be transmitted to the second system in a state in which learning has been completed by the learning model of the first system and stored in the second system.

The hot-rolled steel sheet production system 300 may determine process parameters for controlling the pickling apparatus. The hot-rolled steel sheet production system 300 may receive the thickness PD of the internal defect layer. The thickness PD of the internal defect layer received by the hot-rolled steel sheet production system 300 is not the thickness of the internal defect layer actually measured in the hot-rolled steel sheet, similar to the embodiment described with reference to FIG. 7, but the hot-rolled steel sheet It may be the thickness of the internal defect layer predicted to be included in , and may include a change in the thickness of the internal defect layer according to regions of the hot-rolled steel sheet. The hot-rolled steel sheet production system 300 may calculate optimal process parameters PPV2 based on the thickness PD of the internal defect layer.

In one embodiment, the calculation modules 310 and 320 may include a first module 310 and a second module 320 and the like. The second module 320 may receive the thickness PD of the internal defect layer. The thickness PD of the internal defect layer is the thickness of the internal defect layer expected to exist in the hot-rolled steel sheet, and may not exactly match the thickness of the internal defect layer actually present in the hot-rolled steel sheet. Also, the second module 320 may receive the target thickness GD of the remaining internal defect layer from the outside. The residual internal defect layer may be an internal defect layer present in the hot-rolled steel sheet after the pickling process is completed. For example, when the internal defect layer is to be completely removed by a pickling process, the target thickness (GD) of the remaining internal defect layer may be zero. According to embodiments, the target thickness GD of the remaining internal defect layer may be greater than zero. The second module 320 may periodically call the first module 310 to calculate the optimal process parameters PPV2.

The second module 320 may output initial values of process parameters to the first module 310 . Initial values of the process parameters may include characteristics of the pickling solution (AC) measured from the pickling tank, and a conveying speed (PPV1) of the hot-rolled steel sheet. The initial values of the process parameters may be determined as arbitrary values or values measured in real time from the pickling apparatus. When the initial values of the process parameters are arbitrarily determined, the initial values of the process parameters may be determined by the first system operating the hot-rolled steel sheet production apparatus or the second system operating the pickling treatment apparatus. The first module 310 may calculate the thickness PPA2 of the internal defect layer expected to be removed by the pickling apparatus using at least one of the characteristics AC of the pickling solution and a certain transfer speed PPV1. . Alternatively, the first module 310 is expected to remain after the pickling process by using the difference between the thickness of the internal defect layer predicted to exist in the hot-rolled steel sheet (PD) and the thickness of the internal defect layer expected to be removed (PPA2). It is also possible to calculate the predicted thickness of the internal defect layer.

The second module 320 may receive the thickness PPA2 of the internal defect layer from the first module 310 . The second module 320 determines whether the characteristic (AC) of the pickling solution and the arbitrarily set transfer speed (PPV1) meet the optimal process parameters (PPV2) based on the thickness (PPA2) of the internal defect layer. The second module 320 may call the fourth module M4 until the characteristic AC of the pickling solution and the transfer speed PPV1 meet the optimal pickling process parameter PPV2.

[Table 1] is a table for explaining how the second module 320 calculates the optimal process parameters PPV2. For convenience of description, it is assumed that the characteristics (AC) of the pickling solutions in [Table 1] are the same. The second module 320 determines a target removal thickness indicating how much the internal defect layer should be removed based on the difference between the thickness PD of the internal defect layer expected to exist in the hot-rolled steel sheet and the target thickness GD of the remaining internal defect layer. (GPA) can be calculated.

The first region, the second region, and the third region defined in [Table 1] may be regions continuously arranged in the longitudinal direction of the hot-rolled steel sheet, and the first region and the third region are adjacent to the end of the hot-rolled steel sheet. can be picked up Referring to [Table 1], the target removal thickness (GPA) in the first region is 1 μm, which is the difference between the thickness (PD) of the internal defect layer expected to exist in the hot-rolled steel sheet and the target thickness (GD) of the remaining internal defect layer. (=3 μm-2 μm).

Referring to [Table 1], in the first module 310, when the transport speed PPV1 is 15 mpm, the thickness PPA2 of the internal defect layer expected to be removed by the pickling apparatus may be 2 μm. Since the thickness of the internal defect layer expected to be removed by the pickling treatment device (PPA2) is greater than the target removal thickness (GPA), the second module 320 increases the transport speed (PPV1) to 20 mpm, which is faster than 15 mpm, so that the first output to module 310. When the transport speed (PPV1) is 20 mpm, if the thickness (PPA2) of the internal defect layer expected to be removed by the pickling device is calculated by the first module 310 as a value other than 1 μm, the second module 320 The transfer speed (PPV1) may be changed again to a value different from 20 mpm and output to the first module 310. On the other hand, when the transport speed (PPV1) is 20 mpm and the thickness of the internal defect layer (PPA2) calculated by the first module 310 is 1 μm, the thickness of the internal defect layer expected to be removed by the pickling device (PPA2) Since A coincides with the target removal thickness (GPA), the second module 320 may determine a transport speed of 20 mpm among optimal process parameters for the first region of the hot-rolled steel sheet.

items Area 1 area 2 tertiary area PD 3μm 10 μm 3μm GD 2μm 2μm 2μm GPA 1 μm 8μm 1 μm PPV1 15 mpm 10mpm 15 mpm PPA2 2μm 4μm 2μm PPV2 20mpm 5 mpm 20mpm

mpm = meters per minute

In the embodiment shown in FIG. 9, it has been described that the calculation modules 310 and 320 of the second system calculate optimal process parameters, but it is not necessarily limited to this form. According to embodiments, calculation modules of the first system connected to the second system through a network may calculate optimal process parameters. The second system may receive the optimal process parameters calculated by the first system through the network and control the pickling apparatus using the received optimal process parameters.

10 is a diagram for explaining an operating method of a hot-rolled steel sheet production system according to an embodiment of the present invention.

Referring to FIG. 10 , the first system 400 may include calculation modules 411 to 415 (410) and a learning model 420. The first system 400 may be connected to a second system including a pickling apparatus that performs a pickling process of removing an internal defect layer of a hot-rolled steel sheet through a network or the like.

In one embodiment, the first system 400 is configured to measure the thickness (ERD) of the residual internal defect layer included in the hot-rolled steel sheet and the process parameters at at least one measurement point while the pickling device performs the pickling process. Device parameters (EPV) may be received. In one embodiment, there may be a plurality of measurement points, and they may be set according to a predetermined cycle. The equipment parameters EPV are process parameters actually measured in the pickling apparatus, and may have different values from process parameters input to the pickling apparatus due to machine errors, operation delays, and the like. For example, the conveying speed of the hot-rolled steel sheet included in the process parameters may be different from the conveying speed of the hot-rolled steel sheet included in the apparatus parameters EPV.

While the pickling process is in progress, the pickling treatment device may unwind and transport the hot-rolled steel sheet wound in a coil form and immerse it in a pickling solution. The pickled steel sheet that has passed through the pickling solution can be re-wound. For example, before the pickling steel sheet is wound again, the pickling treatment device may measure the thickness (ERD) of the remaining internal defect layer in the pickling steel sheet. In addition, the pickling treatment device may transmit, to the fifth module M5, device parameters EPV applied to control the pickling process for a specific position of the pickled steel sheet, in which the thickness of the residual internal defect layer ERD is measured. .

According to an embodiment, at least one measurement location may be designated on the pickled steel sheet, and a sample may be collected by cutting a portion of the pickled steel sheet at the measurement location. The thickness of the remaining internal defect layer can be measured by observing the cross section of the sample taken from the pickled steel sheet under a microscope. The measurement location may correspond to the location information (PI) of the pickled steel sheet, and the thickness of the remaining internal defect layer may correspond to the thickness (ERD) of the remaining internal defect layer at a location corresponding to the location information (PI). there is.

The fifth module 415 may output the device parameters EPV to the fourth module 414 . The fourth module 414 may calculate the thickness PPA2 of the internal defect layer expected to be removed by the pickling device based on the device parameters EPV. The fifth module 415 may receive the thickness PPA2 of the internal defect layer from the fourth module 414 . The fifth module M5 may calculate the expected thickness PRD of the remaining internal defect layer based on the thickness PPA2 of the internal defect layer expected to be removed by the pickling treatment device. The predicted thickness PRD of the residual internal defect layer may be a thickness of the residual internal defect layer expected to remain in the pickled steel sheet after the pickling process is performed according to the device parameters EPV.

The learning model LMT may receive the expected thickness PRD of the internal defect layer from the fifth module 415 . The learning model (LMT) may compare the thickness (ERD) of the residual internal defect layer directly measured from the pickled steel sheet with the expected thickness (PRD) of the residual internal defect layer calculated by the fourth module 414 . In the learning model (LMT), the thickness of the residual internal defect layer actually measured (ERD) and the expected thickness of the residual internal defect layer calculated by the fourth module 414 (PRD) match or the difference is less than a predetermined value. As much as possible, at least some of the calculation modules 411 to 415 may be trained. For example, the learning model LMT may simultaneously train all of the arithmetic modules 411 to 415 or selectively train only a specific model.

11 is a flowchart illustrating an operating method of a hot-rolled steel sheet production system according to an embodiment of the present invention.

Referring to FIG. 11 , the hot-rolled steel sheet production system may transmit at least some of the calculation modules to an external system (S110). For example, a subject that transmits at least some of the calculation modules to an external system may be a first system that produces a hot-rolled steel sheet and winds the hot-rolled steel sheet into a hot-rolled coil. Also, the external system that receives at least some of the calculation modules may be a second system that receives the hot-rolled coil and performs the pickling process.

The external system may perform the pickling process using the arithmetic modules received in step S110. The hot-rolled steel sheet production system may receive, from an external system, the thickness of the remaining internal defect layer included in the pickled steel sheet after the pickling process, and the process parameters measured by the pickling treatment device during the pickling process (S120). For example, the process parameters received in step S120 may have values actually measured in the pickling treatment device, and thus may be different from values input to the pickling treatment device by an external system to control the pickling treatment device. Process parameters received in step S120 may be defined as device parameters.

The hot-rolled steel sheet production system may use the process parameters received in step S120 to calculate the thickness of the internal defect layer expected to be removed by the pickling treatment device (S130). In addition, the hot-rolled steel sheet production system may compare the thickness of the internal defect layer calculated in step S130 with the thickness of the remaining internal defect layer received in step S120, and learn calculation modules based on the comparison result (S140). The hot-rolled steel sheet production system may include a learning model required for learning of the calculation modules, and the learning model may adjust weights, coefficients, etc. used in calculation in at least one of the calculation modules.

When learning is completed, the hot-rolled steel sheet production system may transmit at least one of the calculation modules to an external system (S150). The external system may perform the pickling process in a manner optimized for the thickness deviation of the internal defect layer present in the hot rolled steel sheet by controlling the pickling treatment device using the calculation module received from the hot rolled steel sheet production system.

12 and 13 are views simply illustrating an automated pickling system interlocked with the hot-rolled steel sheet production system according to an embodiment of the present invention.

As described above, the hot-rolled steel sheet production system according to an embodiment of the present invention may predict the thickness of the internal defect layer included in the produced hot-rolled steel sheet and transmit the predicted thickness to an external system performing the pickling process. In other words, the external system that receives the thickness of the internal defect layer predicted by the hot-rolled steel sheet production system may be an automated pickling system including a pickling treatment device that performs a pickling process on the hot-rolled steel sheet.

In addition, as described with reference to FIG. 11, the hot-rolled steel sheet production system according to an embodiment of the present invention receives the thickness of the remaining internal oxide layer from the pickled steel sheet on which the pickling process has been completed, and compares it to the previously predicted thickness of the internal defect layer. Calculation modules can be learned by comparison. In order to measure the thickness of the residual internal defect layer in the pickled steel sheet, a sample obtained by cutting a part of the pickled steel sheet at at least one measurement position specified in the pickled steel sheet may be used.

However, the method of taking samples from the pickled steel sheet and predicting the thickness of the remaining internal defect layer therefrom may result in physical damage to the pickled steel sheet. In addition, in order to minimize physical damage to the pickled steel sheet, the number of measurement locations must be reduced, and in this case, the thickness of the residual internal defect layer over the pickled steel sheet may not be accurately measured. In the pickling treatment devices 500 and 600 according to the embodiments shown in FIGS. 12 and 13, the surface state is detected using the surface defect detection devices 430 and 530 instead of directly taking samples. By doing so, the thickness of the remaining internal defect layer can be measured.

Referring first to FIG. 12 , the pickling apparatus 500 may include a coiler CL, an uncoiler UCL, a pickling tank TK, a first meter MI1, a second meter MI2, and the like. there is. The uncoiler UCL may enter the strip-shaped hot-rolled steel sheet 510 from which the hot-rolled coil HC is unwound into the pickling tank TK. The hot-rolled steel sheet 510 may be pickled while passing through the pickling solution contained in the pickling tank TK. The pickling tank TK stores a pickling solution necessary for the pickling process, and according to embodiments, a brush or the like contacting the surface of the hot-rolled steel sheet 510 may be further included.

While the pickling process is in progress, the pickling treatment apparatus 500 may be controlled according to the optimal process parameters PPV2 according to the above-described embodiments. For example, the conveying speed of the hot-rolled steel sheet 510 may vary in each of regions along the length direction of the hot-rolled steel sheet 510 according to the process parameters PPV2 . The first meter MI1 connected to the pickling tank TK may measure the device parameters EPV at at least one measurement point while the pickling treatment device 500 is performing the pickling process. The device parameters EPV correspond to the process parameters PPV2, but at least some of the device parameters EPV and process parameters PPV2 may have different values due to machine errors, operation delays, and the like. . The device parameters EPV obtained by the first meter MI1 may be transmitted to the fifth module 415 as described above with reference to FIG. 10 .

The second measuring instrument MI2 may designate at least one measurement position on the pickling steel plate 520 and cut a portion of the pickling steel plate 520 at the measurement position to collect a sample. The thickness of the remaining internal defect layer can be measured by observing the cross section of the sample taken from the pickled steel sheet 520 under a microscope. The measurement location may correspond to the location information (PI) of the pickled steel sheet 520, and the thickness of the remaining internal defect layer may correspond to the thickness (ERD) of the remaining internal defect layer at a location corresponding to the location information (PI). may apply.

In one embodiment shown in FIG. 12, the surface state of the hot-rolled steel sheet 510 is detected by taking a surface image of the hot-rolled steel sheet 510, or the surface state of the hot-rolled steel sheet 510 is detected by measuring the glossiness of the surface of the hot-rolled steel sheet 510. A surface defect detection device 530 may be included in the pickling device 500 . The surface defect detection device 530 may detect a surface state and glossiness of the hot-rolled steel sheet 510 in at least one of a width direction and a length direction. From the surface state detected by the surface defect detection device 530, the amount of the internal defect layer removed from the hot-rolled steel sheet 510 by the pickling process of the pickling device 500 can be determined. The amount of samples to be collected may be reduced, or the second measuring instrument MI2 may not be selectively used.

An operation of the acid pickling apparatus 600 according to the embodiment shown in FIG. 13 may be similar to that described with reference to FIG. 12 . However, in the embodiment shown in FIG. 13 , the surface defect detection device 630 may be disposed at the rear end of the pickling tank TK. Accordingly, the surface defect detection device 630 may detect the surface state and/or glossiness of the pickled steel sheet 620 passing through the pickling tank TK.

In the embodiment shown in FIG. 12, since the surface defect detection device 630 is disposed at the rear end of the pickling tank TK, the state of the pickling steel sheet 620 that has passed through the pickling tank TK and the pickling process has been completed is accurately detected. can be detected. Accordingly, the second measuring device MI2 may be omitted or the number of samples collected by the second measuring device MI2 may be reduced.

In addition, unlike the second measuring instrument MI2 that measures the thickness (ERD) of the remaining internal defect layer with samples taken from specific locations of the pickled steel sheet 620, the surface defect detection device 630 uses the pickled steel sheet 620 ) The entire surface image or gloss can be obtained. Therefore, the thickness of the remaining internal defect layer can be sufficiently measured in the longitudinal direction and/or the width direction of the pickled steel sheet 620 .

In this way, by measuring the thickness of the remaining internal defect layer over the pickling steel sheet 620, the learning amount for optimizing the fourth module 414 and the fifth module 415 can be increased. As the amount of learning of the fourth module 414 and the fifth module 415 increases, the fifth module 415 is executed in a system linked to the pickling device 600 and outputs optimal process parameters (PPV2) accuracy can also be improved.

The present invention is not limited by the above-described embodiments and accompanying drawings, but is intended to be limited by the appended claims. Therefore, various forms of substitution, modification, and change will be possible by those skilled in the art within the scope of the technical spirit of the present invention described in the claims, which also falls within the scope of the present invention. something to do.

Claims (18)

A prediction calculation module for predicting the thickness of an internal defect layer occurring in a hot-rolled steel sheet, a process calculation module for determining process parameters for controlling a pickling process for removing the internal defect layer to be used in the pickling process, and the process calculation a storage for storing a learning model for learning a module;
a communication unit connected to a network to which an external server operating the pickling apparatus performing the pickling process is connected; and
The learning models stored in the storage control the process calculation module to learn so that the difference between the thickness of the internal defect layer predicted by the prediction calculation module and the thickness of the internal defect layer to be removed by the pickling process is reduced, and the communication unit Through, a processor for transmitting the process calculation module to the external server; including,
The process parameters include the concentration of the pickling solution contained in the pickling tank of the pickling treatment device, the temperature of the pickling solution, whether or not an accelerator is used, whether or not a brush in contact with the surface of the hot-rolled steel sheet is used, and the hot rolling in the pickling treatment device. A hot-rolled steel sheet production system comprising at least two of the feed rates of the steel sheet.
According to claim 1,
The processor controls one of the learning models to learn the predictive operation module using loss functions that compare a predicted value of the predictive operation module with an actual value measured from the hot-rolled steel sheet,
The loss functions calculate a difference between the predicted value and the measured value by applying the predicted value and the measured value to at least one of a mean square error and a cross entropy error.
According to claim 1,
The prediction calculation module includes a first module for predicting a phase fraction before winding the hot-rolled steel sheet.
According to claim 3,
The processor compares a first phase fraction value measured in the hot-rolled steel sheet with a second phase fraction value predicted by the first module from at least one of a temperature of the hot-rolled steel sheet and a component of the hot-rolled steel sheet, and The hot-rolled steel sheet production system for learning the first module to reduce the difference between the phase fraction value and the second phase fraction value.
According to claim 1,
The prediction calculation module includes a second module for predicting the temperature of each of a plurality of regions along the length direction of the hot-rolled steel sheet.
According to claim 5,
The processor uses at least one of a first temperature value measured from the hot-rolled steel sheet, a phase fraction predicted before winding the hot-rolled steel sheet, an elapsed time after winding the hot-rolled steel sheet, and a component of the hot-rolled steel sheet The hot-rolled steel sheet production system of comparing the second temperature value predicted by the second module and learning the second module to reduce the difference between the first temperature value and the second temperature value.
According to claim 1,
The prediction operation module includes a third module that predicts a thickness of an internal defect layer included in each of a plurality of regions along the length direction of the hot-rolled steel sheet.
According to claim 7,
The processor determines the first thickness of the internal defect layer measured from the hot-rolled steel sheet, the temperature predicted according to the elapsed time after winding in each of the regions, the components of the hot-rolled steel sheet, and the oxygen partial pressure around the hot-rolled steel sheet. The third module compares the second thickness of the internal defect layer predicted by the third module using at least one in each of the regions and learns the third module to reduce the difference between the first thickness and the second thickness. steel sheet production system.
According to claim 7,
The third module predicts the thickness of the internal defect layer included in the hot-rolled steel sheet using a temperature change output by an infrared camera for photographing thermal infrared rays emitted from the surface of the hot-rolled steel sheet Hot-rolled steel sheet production system.
According to claim 1,
The hot-rolled steel sheet production system receiving, through the communication unit, at least one of a transport speed of the hot-rolled steel sheet in the pickling apparatus and a characteristic of a pickling solution contained in a pickling tank of the pickling apparatus.
According to claim 10,
The process calculation module includes a fourth module for predicting the thickness of the internal defect layer removed by the pickling apparatus by applying at least one of the process parameters as a weight or a variable to a stored function.
According to claim 11,
The processor learns the fourth module so that the difference between the first thickness of the residual internal defect layer measured from the pickled steel sheet on which the pickling process is completed and the second thickness of the internal defect layer calculated by the fourth module is reduced. hot-rolled steel sheet production system.
According to claim 11,
The process calculation module includes a fifth module for determining the process parameters.
According to claim 13,
The processor determines the thickness of the internal defect layer calculated by the external server using the process parameters measured from the pickling device during the pickling process, and the external surface from the pickled steel sheet subjected to the pickling process to the hot-rolled steel sheet. Receives the thickness of the remaining internal defect layer measured by the server from the external server;
The hot-rolled steel sheet production system for learning at least one of the calculation modules to reduce a difference between the thickness of the internal defect layer and the thickness of the remaining internal defect layer.
According to claim 14,
The processor is configured so that the difference between the thickness of the remaining internal defect layer and the thickness of the internal defect layer determined using the surface state of the pickling steel sheet detected by taking a surface image of the pickling steel sheet by the external server is reduced. A hot-rolled steel sheet production system that trains at least one of the arithmetic modules.
According to claim 13,
The processor transmits the fourth module and the fifth module to the external server through the communication unit.
training a process calculation module for determining process parameters for controlling a pickling process to be used in the pickling process, wherein the pickling process is a process of removing an internal defect layer of a hot-rolled steel sheet;
transmitting the process calculation module to an external system including a pickling device that performs the pickling process;
Receiving, from the external system, the thickness of the remaining internal defect layer included in the pickling steel sheet after the pickling process and the process parameters measured by the pickling apparatus during the pickling process; and
The process calculation module applies the received process parameters as weights or variables to the stored function to reduce the difference between the thickness of the internal defect layer predicted by the prediction calculation module and the thickness of the remaining internal defect layer received from the external system. Step of learning; including,
The process parameters include the concentration of the pickling solution contained in the pickling tank of the pickling treatment device, the temperature of the pickling solution, whether or not an accelerator is used, whether or not a brush in contact with the surface of the hot-rolled steel sheet is used, and the hot rolling in the pickling treatment device. A method of operating a hot-rolled steel sheet production system including at least two of the feed rates of the steel sheet.
According to claim 17,
A method of operating a hot-rolled steel sheet production system for transmitting the process calculation module to the external system when the learning of the process calculation module is completed.

Images