Program, information processing device, and information processing method

Abstract

To provide a program, an information processing device, and an information processing method capable of suppressing variation in work for each worker.SOLUTION: A computer (information processing device) executes: a first calculation related to an oxygen heating operation for increasing a molten steel temperature in an initial stage of refining; a second calculation related to a slag-making agent adding operation; a third calculation related to a component change of molten steel by vacuum degassing treatment; a fourth calculation related to addition of ferroalloy; a fifth calculation related to addition of scrap; a sixth calculation related to electric heating for compensating for a decrease in the molten steel temperature due to the addition of ferroalloy and scrap; a seventh calculation related to the molten steel temperature decreased due to the vacuum degassing treatment; and an eighth calculation related to electric heating for compensating for the decrease in the molten steel temperature due to the vacuum degassing treatment.SELECTED DRAWING: Figure 16

Description

Technical Field

[0001] The present invention relates to a program, an information processing apparatus, and an information processing method for scheduling a molten steel refining process.

Background Art

[0002] In the molten steel refining process, the molten steel melted in a converter, an electric furnace, etc. is tapped into a ladle, and oxygen in the molten steel is reduced while adding a slag-forming agent, deslagging deoxidation, vacuum degassing treatment, etc. In parallel with these, the weight and composition of the molten steel are adjusted while adding scrap and alloy iron when the amount of molten steel is insufficient. Further, regarding the molten steel temperature, not only a good temperature is maintained from when it is tapped into the ladle until the end of the refining process, but in addition, in order to adjust the temperature required for the casting operation performed after the refining process, the temperature of the molten steel is controlled by operations called oxygen heating and electric heating. Therefore, the refining process is a complex operation that simultaneously adjusts three items: the weight, chemical composition, and temperature control of the molten steel.

[0003] Therefore, for example, in Patent Document 1, using the estimated slag components, the molten steel components after the determined alloy input amount is input are predicted, and component adjustment is performed based on the alloy input amount when the predicted molten steel components reach the target component values. A component adjustment method is disclosed.

[0004] Also, in Patent Document 2, a method for managing the temperature of molten steel is disclosed in which calculations are performed using a predetermined calculation formula to predict the temperature after vacuum degassing from the molten steel temperature and weight before vacuum degassing, the refractory temperature, and the treatment time.

Prior Art Documents

Patent Documents

[0005]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0006] However, these documents do not describe a method for managing all three items of the weight, chemical composition, and temperature of molten steel. In actual operations, variations may occur in the work due to differences in the proficiency of workers and the like, which may pose an obstacle to ensuring stable quality.

[0007] On the other hand, in order to suppress variations in work, it is necessary to support the work throughout the entire refining process. However, Patent Document 1 is limited to the content of the component adjustment method, and Patent Document 2 is limited to the content of the temperature control of molten steel, and there is a possibility that the above-mentioned problems cannot be completely solved.

[0008] The present disclosure has been made in view of such circumstances, and its object is to provide a program, an information processing apparatus, and an information processing method capable of suppressing variations in work for each worker in the refining process, and specifically, characterized by supporting all three items of the weight, chemical composition, and temperature management of molten steel.

Means for Solving the Problems

[0009] The program according to the present disclosure causes a computer to execute calculations related to each operation in the refining process, obtains the required time necessary for oxygen heating operation, slag-forming agent addition operation, scrap addition operation, ferroalloy addition operation, vacuum degassing treatment operation, and electric heating operation, obtains the added weight of scrap, the type, added weight, and number of additions of the ferroalloy to be added, and causes the computer to perform a process of scheduling the entire refining process.

Effects of the Invention

[0010] According to the present disclosure, it is possible to support the work of workers in all steps of the refining process and suppress variations in work for each worker.

Brief Description of the Drawings

[0011]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

Figure 13

Figure 14

Figure 15

Figure 16

Figure 17

Figure 18

Figure 19

Figure 20

Figure 21

Figure 22

Figure 23

Figure 24

Mode for Carrying Out the Invention

[0012] Hereinafter, a program, an information processing apparatus, and an information processing method according to an embodiment of the present invention will be described. First, with reference to FIG. 1, an outline of the actual refining process operation will be described.

[0013] Generally, in the steel manufacturing process, the molten steel melted in a converter or an electric furnace is tapped into a ladle and then transferred to a ladle refining furnace for the refining process. The first operation is to collect a sample of the molten steel, and by this operation, component analysis is performed to confirm the initial components. The temperature of the molten steel immediately after tapping is lower compared to the electric furnace process, and in some cases, an operation called oxygen heating-up is carried out to raise the temperature of the molten steel in a short time. Oxygen heating-up is an operation of supplying oxygen while adding Al or Si to the molten steel, and by utilizing the heat generated when the added Al or Si oxidizes, the temperature of the molten steel is raised in a relatively short time. The required time for this operation is set to 10 minutes in this embodiment, and calculating the temperature rise due to this operation is the first calculation.

[0014] After that, a slag layer is provided on the molten steel by adding and dissolving a slag-forming agent. In this embodiment, the operation time for this is 2 minutes and is constant each time, and since there is no change in the weight and temperature of the molten steel, there is nothing particularly to be calculated. However, in the scheduling for organizing the entire refining process, 2 minutes is used as the operation time, so for convenience, it is called the second calculation. The slag layer provided on the molten steel becomes an electric resistance source during subsequent electric heating-up, and by applying electricity here, Joule heat is generated and used for raising the temperature of the molten steel. Further, by blowing Ar gas from the bottom of the ladle, the molten steel is stirred, and at the contact surface with the slag layer, mainly oxides contained in the molten steel are captured, contributing to the reduction of impurities in the molten steel. In addition, due to the chemical reaction with the molten steel, effects such as reducing sulfur (S) contained in the molten steel are brought about.

[0015] When the addition of the slag-forming agent is completed and the slag layer is formed on the molten steel, the full-scale refining operation begins. First, the weight and chemical composition of the molten steel are confirmed. Generally, the initial weight of the molten steel is measured by a load cell or the like installed on a crane for lifting the ladle or a trolley on which the ladle is loaded when tapping from the electric furnace. Also, as described above, the initial components are confirmed immediately after tapping.

[0016] The initial weight of the molten steel is compared with the target weight. When the initial weight is greater, the addition of scrap is not carried out because the required weight is ensured. However, when the initial weight is smaller, a comparison with the target weight is made taking into account the weight increase due to the addition of ferroalloy in the subsequent process. Furthermore, in the calculation for determining the ferroalloy to be added, the component changes due to the vacuum degassing treatment are taken into account, improving the calculation accuracy of the ferroalloy to be added. Although the addition weights of the ferroalloy and scrap thus calculated, a limit is set such that the maximum weight that can be added at one time is 4000 kg in order to prevent a temperature drop of the molten steel due to excessive addition and the remaining unmelted added materials. In accordance with this limit, in this embodiment, when the material to be added exceeds 4000 kg, the addition is carried out in multiple times. Calculations related to the addition weights and addition times of such scrap and ferroalloy are performed by the calculation related to the component changes of the molten steel by vacuum degassing treatment / the third calculation, the calculation related to the weight and addition times of the ferroalloy to be added / the fourth calculation, and the calculation related to the scrap addition / the fifth calculation, which will be described later.

[0017] When the ferroalloy and scrap, whose weights and numbers have been thus calculated, are added, the temperature of the molten steel drops each time. However, in order to avoid the solidification of the molten steel even when the temperature of the molten steel drops, specifically, it is necessary to preheat the molten steel by energization to raise the temperature so that it does not fall below the allowable minimum temperature of 1540°C. For this purpose, it is necessary to calculate the target temperature during energization heating, and also to calculate the required time to heat up to the calculated target temperature. These calculations are the sixth calculation related to the energization heating of the molten steel.

[0018] While maintaining the temperature of the molten steel by energization heating, the addition of ferroalloy and scrap is carried out to adjust the weight and chemical composition of the molten steel. After these operations are completed, next, vacuum degassing treatment is performed to reduce the gas components such as oxygen and hydrogen contained in the molten steel. This treatment is carried out by exposing the ladle to a reduced-pressure atmosphere. During this period, energization heating cannot be carried out, so the temperature of the molten steel drops. Calculating this temperature drop of the molten steel is the seventh calculation.

[0019] After the vacuum degassing process, the temperature of the molten steel decreases. Therefore, in order to raise the temperature of the molten steel to the final refining temperature, energization heating is performed again, and calculating the time required for this is the eighth calculation.

[0020] In this way, each operation in the refining process is carried out. By calculating the time required for each operation, and further calculating the weight of alloy iron and scrap to be added, the number of addition times, etc., it is possible to assist the operator. The purpose of this embodiment is to calculate the details of these operation contents to assist the operator. There are also values that are always constant, such as the second calculation, that is, the time for adding the slag-forming agent. When summarizing the changes that each operation gives to the weight, composition, and temperature of the molten steel, it becomes as shown in FIG. 2.

[0021] The weight of the molten steel starts from the initial weight, and finally a weight exceeding the required weight is needed, but it increases due to the addition of alloy iron and scrap. The chemical composition of the molten steel starts from the initial composition, and finally it is necessary to adjust all components within the target component range, which is done by adding alloy iron. However, since some components increase or decrease during the vacuum degassing process, addition taking this into account is required. The temperature of the molten steel decreases when it is tapped from the electric furnace into the ladle. In addition, it decreases due to the addition of scrap and alloy iron, and furthermore, although not shown in FIG. 2, it also decreases just by leaving it without any operation. Therefore, the temperature is raised by oxygen heating and energization heating.

[0022] As described above, adding scrap and alloy iron is a complex operation because it affects all three items of weight, composition, and temperature. When further organized according to the order of operations for implementing this embodiment, it becomes as shown in FIG. 3. In other words, the purpose of this embodiment is to calculate the blank spaces in this table.

[0023] Next, the configuration and functions of the information processing apparatus required for performing such various calculations will be described. FIG. 4 is a functional block diagram showing a configuration example of the information processing apparatus 100 according to this embodiment. The information processing apparatus 100 includes a control unit 10, a storage unit 20, a power supply unit 30, a display unit 40, a reading unit 50, and a mass storage unit 60. Each component is connected via a bus.

[0024] The control unit 10 includes an arithmetic processing unit such as a CPU, MPU (Micro-Processing Unit), or GPU (Graphics Processing Unit), and reads and executes the control program P stored in the storage unit 20 to perform various information processes, control processes, etc. related to the information processing device 100. Note that the control program P can be deployed to be executed on a single computer, or at one site, or distributed over a plurality of sites and executed on a plurality of computers interconnected by a communication network. In this embodiment, the control unit 10 is described as a single processor, but it may also be a multi-processor.

[0025] The storage unit 20 includes memory elements such as a RAM (Random Access Memory) and a ROM (Read Only Memory), and stores the control program P or data etc. necessary for the control unit 10 to execute processing. Further, the storage unit 20 temporarily stores data etc. necessary for the control unit 10 to execute arithmetic processing.

[0026] Furthermore, various information necessary for calculations is stored in the storage unit 20. The storage unit 20 stores various information necessary for the first to eighth calculations, and the basic information that requires input for each refining charge is shown in FIG. 5.

[0027] There are three types of ladle sizes, which are selected from "large, medium, and small" according to the weight of the molten steel. The steel grade is a symbol representing the steel grade to be refined. There are two types of molten steel weights. The initial weight is the actual amount of molten metal at the start of refining, while the target weight is the amount of molten metal required at the end of refining. In this embodiment, since the initial weight is greater than the target weight, there is no problem. However, if the initial weight is less than the target weight, the weight increase due to the addition of ferroalloy described later is calculated. If it is still less than the target weight after that, scrap is added to ensure a molten steel weight equal to or greater than the target weight.

[0028] There are four types of molten steel temperatures. The initial temperature is the temperature at the start of refining. In the case of this embodiment, it is the measured value of 1552 °C. The target temperature of 1600 °C at the start of vacuum degassing treatment is given to raise the temperature before the start of treatment because the temperature of the molten steel decreases during vacuum degassing treatment. The target temperature of 1650 °C at the end of refining is the temperature required at the end of the refining process and is the molten steel temperature required mainly for purposes such as ensuring quality when performing the next process, which is the casting operation. The allowable minimum temperature is set to avoid solidification of the molten steel, and it must not fall below this temperature even after the addition of alloy iron and scrap. Therefore, if it is expected to fall below the allowable minimum temperature after these additions, it is necessary to perform energized heating in advance.

[0029] For oxygen heating, there are two types: adding Si and oxidizing it in the case of Si-killed steel, and adding Al and oxidizing it in the case of Al-killed steel. In both cases, the amount of oxygen used is 100 Nm 3 and is a constant value. The CaO weight, the weight other than CaO, and the total weight of these, which are the amounts of slag-forming agent used, are values determined in advance according to the initial ladle weight. Since there are multiple energizing facilities corresponding to the ladle size for energized heating, the values of the AC voltage and AC current of the energizing facility to be used are given. Vacuum degassing treatment is a treatment that exposes the molten steel to a reduced-pressure environment to reduce the gas components in the molten steel. The high-vacuum duration means the time of exposure to an atmosphere below a predetermined pressure, and the total time means the total time from the start to the end of the vacuum degassing treatment. Also, during vacuum degassing treatment, Ar gas is blown in from the bottom of the ladle to help reduce the gas components in the molten steel, and this flow rate is also given as one of the treatment conditions. Also, there are operations where the working time is determined to be constant each time. Oxygen heating is 10 minutes, and the addition of slag-forming agent, alloy iron, and scrap is 2 minutes, and a fixed time is given.

[0030] These values shown in FIG. 5 are stored in the storage unit 20 and are used for the first to eighth calculations described later.

[0031] Also, the range of the target chemical components is stored in the storage unit 20, and an example is shown in FIG. 6. In the present embodiment, there are components for which both the lower limit value and the upper limit value are defined, such as C, Si, Mn, Cr, Ni, Mo, and V, and components for which only the upper limit value is defined, such as S. These target ranges of chemical components are given for each steel type, respectively.

[0032] Furthermore, necessary information regarding alloy iron added so that the chemical components of the molten steel are within the target range and scrap added when the weight of the molten steel is insufficient is stored in the storage unit 20, and an example is shown in FIG. 7.

[0033] FIG. 7 is a chart showing an example of a component table of additives added to molten steel in the refining process. The additives include alloy iron used for adjusting chemical components in addition to scrap, and the alloy iron is further divided into 1 to 3 groups. Group 1 is for adding Si, Ni, and Mo, Group 2 is for adding Mn and Cr, and Group 3 is for adding C and V. When adding these, they are carried out in the order of Groups 1, 2, and 3. Also, for alloy iron, for example, when increasing the Si content of the molten steel, there are two types, low-carbon Fe-Si and CaSi, depending on whether it is Si-deoxidized steel or Al-deoxidized steel. Similarly, when increasing the Mn and Cr contents, there is a "selection priority order" that determines which alloy iron to use preferentially. This is determined according to the amount of impurity elements contained other than Mn and Cr. When increasing the Mn content, the addition weights are calculated in the order of priority: 1st / Si-Mn, 2nd / high-carbon Fe-Mn, 3rd / low-carbon Fe-Mn. At this time, for example, when increasing the Mn in the molten steel, considering that the addition of priority: 1st / Si-Mn is accompanied by an increase in Si contained in this by 10.9 to 13.1%, and the addition of priority: 2nd / high-carbon Fe-Mn is accompanied by an increase in C contained in this by 6.9 to 7.0%, the respective addition weights are calculated. Similarly, when increasing the Cr content, the addition weights are calculated in the order of priority: 1st / high-carbon Fe-Cr, 2nd / low-carbon Fe-Cr, and considering that the addition of priority: 1st / high-carbon Fe-Cr is accompanied by an increase in C contained in this by 7.46 to 7.85% and an increase in Si contained in this by 2.38 to 3.92%, the addition weight is calculated.

[0034] In the component table, values from 0 to 100% are set for each additive as the "molten steel weight reflection rate". These values indicate the degree of influence on the weight increase of the molten steel when actually added. For example, CaSi and the shock absorber set at 0% mean that the weight increase of the molten steel is 0 kg regardless of the added weight. Similarly, for low-carbon Fe-Si set at 50%, 50% of the added weight becomes the weight increase of the molten steel, and for other additives with 100%, 100% of the added weight leads to an increase in the molten steel weight. The reason why the molten steel weight reflection rate is not 100% lies in the oxygen contained in the molten steel. The alloy iron forms oxides and is absorbed by the slag, which does not lead to an increase in the molten steel weight, so values less than 100% are given based on empirical rules. Scrap is set at 90% because it is stored in a state containing some slag, and it is set assuming that the weight ratio is constant at 10%.

[0035] Note that these data shown in Fig. 7 are constantly fixed regardless of the refining charge.

[0036] Returning to Fig. 4, the display unit 40 is a liquid crystal display or an organic EL (electroluminescence) display, etc., and displays various information according to the instructions of the control unit 10. Also, the display unit 40 may be configured to have a touch panel function and be able to receive setting inputs from the operator.

[0037] The reading unit 50 reads a portable storage medium 200a including a CD (Compact Disc)-ROM or a DVD (Digital Versatile Disc)-ROM. The control unit 10 may read the control program P from the portable storage medium 200a via the reading unit 50 and store it in the large-capacity storage unit 60. Also, the control unit 10 may download the control program P from another computer via a network or the like and store it in the large-capacity storage unit 60. Further, the control unit 10 may read the control program P from the semiconductor memory 200b.

[0038] The large-capacity storage unit 60 includes a recording medium such as an HDD (Hard Disk Drive) or an SSD (Solid State Drive). The large-capacity storage unit 60 includes a learning model DB (database) 61, a training data DB 62, and a learning model management DB 63.

[0039] The learning model DB 61 stores a component variation prediction model 611 (first learning model), a current-carrying heat-up prediction model 612 (second learning model), and a temperature drop prediction model 613 (third learning model). Hereinafter, for convenience, the component variation prediction model 611, the current-carrying heat-up prediction model 612, and the temperature drop prediction model 613 are also collectively referred to as prediction models 611 to 613.

[0040] The component variation prediction model 611 is a model for predicting the component change of molten steel that occurs before and after vacuum degassing, and is used in the fourth calculation for calculating the type and weight of alloy iron added to the molten steel. The current-carrying heat-up prediction model 612 is a model for predicting the time required to heat up to the target temperature when the molten steel is heated by passing an electric current, and is used in the sixth calculation and the eighth calculation. The temperature drop prediction model 613 is a model for predicting the temperature drop of molten steel during vacuum degassing, and is used in the seventh calculation.

[0041] The training data DB 62 stores training data for constructing (creating) the prediction models 611 to 613. Specifically, it stores component training data 621, which is training data for the component variation prediction model 611, heat-up training data 622, which is training data for the current-carrying heat-up prediction model 612, and temperature training data 623, which is training data for the temperature drop prediction model 613.

[0042] The learning model management DB63 stores information about the learned prediction models 611 to 613. Specifically, it stores the files of the learned prediction models 611 to 613, the date and time information when the prediction models 611 to 613 were generated, etc., and an example is shown in FIG. 8. The learning model management DB63 includes a model ID column, a learning model column, a type column, and a generation date and time column. In the model ID column, IDs for identifying each of the learned prediction models 611 to 613 are stored. The learning model column stores the files of the learned prediction models 611 to 613. Also, the type column stores the type of the learning model. The type is classified, for example, by the output data output by each of the prediction models 611 to 613. The generation date and time column stores the date and time information when the prediction models 611 to 613 were generated.

[0043] The prediction models 611 to 613 are learned learning models generated by machine learning, and these will be sequentially described.

[0044] The component variation prediction model 611, which is the first learning model, is generated based on the vacuum degassing treatment conditions performed in the past, the actual component variation amounts obtained as the measured values at that time (hereinafter referred to as component variation amounts) before and after the vacuum degassing treatment, etc., and outputs the component variation amounts before and after the vacuum degassing predicted under the conditions of the charge. The component variation prediction model 611 is composed of a neural network in which a plurality of neurons are connected. The component variation prediction model 611 has an input layer, an intermediate layer, and an output layer, and when input data at the time of predicting component variations such as the molten steel components at the initial stage of the refining process, the vacuum degassing conditions, the molten steel weight at the initial stage of the refining process, and the amount of slag-making agent used are input, it is learned to output the amounts of each component of the molten steel that has varied during the vacuum degassing (hereinafter simply referred to as the component variation amount after degassing). Here, the component variation amount after degassing is the difference between the amount of molten steel components after the vacuum degassing and the amount of molten steel components before the vacuum degassing.

[0045] The type of neural network that constitutes the component variation prediction model 611 may be arbitrary. For example, the component variation prediction model 611 uses a model of Elastic Net regression. Elastic Net regression is one of the regularized linear regressions and was created as a compromise between the well-known Ridge regression and Lasso regression to solve the problem that there is a limit to the number of explanatory variables that can be incorporated into the Lasso regression model.

[0046] The training data DB62 in FIG. 4 stores component training data 621 for creating the component variation prediction model 611, and an example thereof is shown in FIG. 9. The component training data 621 includes a training ID column, an input data column, and an output data column. The training ID column stores an ID for identifying each training data, and the molten steel components at the initial stage of the refining process in past refining charges, the molten steel components after vacuum degassing treatment, the molten steel weight at the initial stage of the refining process, the usage amount of slag-forming agents (CaO, others), the high vacuum duration, the total time, and the Ar flow rate as the vacuum degassing treatment conditions are stored, and the actual component variation amount measured after the vacuum degassing treatment performed under these conditions is stored.

[0047] An example of the calculation result using this component variation prediction model 611 is shown in FIG. 10. In FIG. 10, C, Si, and S have negative values, and Mn and Cr have positive values, indicating the values (%) by which C, Si, and S contained in the molten steel decrease and the values (%) by which Mn and Cr increase due to the vacuum degassing treatment. Such component variation amounts are used when calculating the added weight of alloy iron in the fourth calculation, and the details will be described later.

[0048] Returning to FIG. 4, the energization heating prediction model 612, which is the second learning model, is generated based on the measured data of the temperature rise of the molten steel during energization heating performed in the past (hereinafter referred to as the measured temperature rise value), etc., and outputs the time required for energization heating predicted under the conditions of the charge and the rising temperature of the molten steel. The energization heating prediction model 612 is composed of a neural network in which a plurality of neurons are connected. The energization heating prediction model 612 has an input layer, an intermediate layer, and an output layer, and when the ladle size, the weight of the molten steel at the initial stage of the refining process, the amount of slag-forming agent used, the elapsed time since the start of refining, the temperature of the molten steel after oxygen heating, the energization conditions, the input energy by energization, and the measured temperature of the molten steel before and after energization are input, it calculates, as output, the time required for energization heating and the rising temperature of the molten steel.

[0049] The type of neural network constituting the energization heating prediction model 612 may be arbitrary. In this embodiment, the energization heating prediction model 612 uses a random forest model, but random forest is a method of supervised machine learning. Random forest is an ensemble learning method consisting of a large number of decision trees, and is a collection of many decision trees created randomly.

[0050] The energization heating operation is performed not only before the vacuum degassing treatment but also after the vacuum degassing treatment, and the energization heating prediction model 612 is used for both. In this embodiment, for the purpose of improving the calculation speed, the energization time is extended at 5-minute intervals such as 5 minutes, 10 minutes, 15 minutes..., and the temperature of the molten steel for each energization time is calculated. However, before the vacuum degassing treatment, the energization heating time and the molten steel temperature when the target temperature is exceeded for the first time are adopted as the calculation results, while after the vacuum degassing treatment, the energization heating time and the molten steel temperature when the temperature closest to the target temperature is reached are adopted as the calculation results.

[0051] In the training data DB62 of FIG. 4, heating-up training data 622 for creating the energization heating prediction model 612 is stored, and an example thereof is shown in FIG. 11. The heating-up training data 622 includes a training ID column, an input data column, and an output data column. In the training ID column, an ID for identifying each training data is stored. In the input data column, input data such as ladle size, molten steel weight at the initial stage of the refining process, amount of slag-forming agent used (CaO, others), elapsed time since the start of refining, molten steel temperature after oxygen heating, energization conditions (AC voltage / AC current), input energy by energization, and molten steel temperature before / after energization in the past refining charge is stored. Here, the molten steel temperature after oxygen heating is the measured molten steel temperature value when the oxygen heating operation is carried out in the refining charge.

[0052] Returning to FIG. 4, the temperature decrease prediction model 613, which is the third learning model, is generated based on the measured value data of the molten steel temperature decrease value during the vacuum degassing process performed in the past, etc., and outputs the predicted molten steel temperature decrease value in the vacuum degassing process predicted under the conditions of the charge.

[0053] The temperature decrease prediction model 613 is composed of a neural network in which a plurality of neurons are combined. The temperature decrease prediction model 613 has an input layer, an intermediate layer, and an output layer, and is learned to output the vacuum degassing end temperature when the ladle size, molten steel weight and temperature at the initial stage of refining, amount of slag-forming agent used (CaO, others), and vacuum degassing start temperature are input.

[0054] The type of neural network constituting the temperature decrease prediction model 613 may be arbitrary. In this embodiment, the temperature decrease prediction model 613 uses a random forest model, similar to the energization heating prediction model 612. However, the total time of the vacuum degassing process is 20 minutes, and this value is constant even in the refining of different charges. Therefore, unlike the case of energization heating, the vacuum degassing time is a constant value of 20 minutes, and the temperature decrease of the molten steel is predicted for this.

[0055] In the training data DB62 of FIG. 4, temperature training data 623 for creating the temperature drop prediction model 613 is stored, and an example thereof is shown in FIG. 12. The temperature training data 623 stores an ID for identifying each training data and temperature drop data in past refining charges. Specifically, the input data (problem data) includes the ladle size, the weight and temperature of the molten steel at the start of refining, the amount of slag-forming agent used (CaO, others), the target temperature at the start of vacuum degassing, the target temperature at the end of refining, and the output data (answer data) includes the temperature at the end of the vacuum degassing process.

[0056] As described above, the ladle size represents information on the size of the ladle used for refining. In this embodiment, the case where the ladle size is stored in the input data series of the heat-up training data 622 and the temperature training data 623 will be taken as an example for explanation, but it is not limited thereto. For example, as information related to the ladle, detailed information such as the physical properties of the refractory material linked with the management number of the ladle, the number of uses or the total use time after relining the refractory, and the total weight of the molten steel processed, etc., are considered factors that affect the temperature rise of the molten steel. Therefore, these may be included in the input data of the heat-up training data 622 and the temperature training data 623.

[0057] FIG. 13 is a flowchart showing the procedure of the generation process of the component variation prediction model 611. The control unit 10 acquires a plurality of input data (problem data) and answer data (actual component variation amount) corresponding to the problem data from the component training data 621 (see FIG. 9) of the training data DB62 (step S6111). The control unit 10 generates a component variation prediction model 611 that outputs the component variation amount after degassing using the acquired problem data and the answer data corresponding to the problem data (step S6112).

[0058] The control unit 10 sequentially inputs the problem data into the component variation prediction model 611 and sequentially outputs the component variation amount after degassing. The control unit 10 compares the output component variation amount after degassing with the answer data (measured component variation amount) included in the component training data 621, and optimizes various parameters of the intermediate layer so that the two are approximated to generate the component variation prediction model 611.

[0059] Specifically, for the model of elastic net regression used by the component variation prediction model 611, for example, the regression coefficients are obtained using the following equation (1). In equation (1), the first term is the objective function of ordinary linear regression, and the second term is a term for penalizing the increase in the regression coefficients, which is called the regularization term.

[0060]

Number

[0061] β: Regression coefficient α, λ: Regularization parameters x: Explanatory variable y: Objective variable

[0062] First, the regularization parameters α (0 < α < 1) and λ in equation (1) are explored. For the exploration of the regularization parameters α and λ, K - Fold Cross Validation is used. Also, the model is evaluated using MSE (Mean Squared Error), and the regularization parameters α and λ are determined. When the regularization parameters α and λ are determined, the component variation prediction model 611 is generated by obtaining the regression coefficients using equation (1) and the component training data 621.

[0063] The control unit 10 stores the generated component variation prediction model 611 in the learning model management DB63 of the mass storage unit 60 (step S6113) and ends a series of processes. Specifically, the control unit 10 assigns a model ID to the generated component variation prediction model 611, and stores the file of the component variation prediction model 611 and the generation date and time as one record in the learning model management DB63 in association with the assigned model ID.

[0064] FIG. 14 is a flowchart showing the procedure of the energization heat rise prediction model 612 generation process. The control unit 10 acquires a plurality of input data (problem data) and answer data (measured temperature rise value) corresponding to the problem data from the heat rise training data 622 (see FIG. 11) of the training data DB62 (step S6121). The control unit 10 generates an energization heat rise prediction model 612 that outputs the rising temperature of the molten steel during energization heating using the acquired problem data and answer data (step S6122).

[0065] The control unit 10 sequentially inputs the problem data into the energization heat rise prediction model 612 and sequentially outputs the rising temperature of the molten steel. The control unit 10 compares the output rising temperature of the molten steel with the answer data (measured temperature rise value) included in the heat rise training data 622, and optimizes various parameters of the intermediate layer so that the two are approximated to generate the energization heat rise prediction model 612.

[0066] Specifically, the control unit 10 generates a plurality (M) of sample sets using the heat rise training data 622 by bootstrap sampling, and randomly extracts a plurality of feature quantities (explanatory variables) used for machine learning for the sample sets. Further, the control unit 10 generates M decision trees using the M sets of sample sets and explanatory variables, and learns each decision tree using the heat rise training data 622. Thereby, a random forest model (energization heat rise prediction model 612) is generated. When actually performing prediction using the random forest model, the results of each decision tree are averaged (ensembled) to obtain the final result.

[0067] The control unit 10 stores the generated energization heat rise prediction model 612 in the learning model management DB63 of the mass storage unit 60 (step S6123) and ends a series of processes. Specifically, the control unit 10 assigns a model ID to the generated energization heat rise prediction model 612, and stores the file of the energization heat rise prediction model 612 and the generation date and time as one record in the learning model management DB63 in association with the assigned model ID.

[0068] FIG. 15 is a flowchart showing the procedure of the generation process of the temperature drop prediction model 613. When the temperature drop prediction model 613 uses a random forest model similar to the energization heating prediction model 612, the procedure of the model generation process is the same as that of the energization heating prediction model 612. The control unit 10 acquires a plurality of input data (problem data) at the time of temperature drop prediction described above and corresponding answer data (measured temperature drop value) from the temperature training data 623 (see FIG. 12) in the training data DB 62 (step S6131). The control unit 10 generates a temperature drop prediction model 613 that outputs the molten steel temperature drop value during vacuum degassing using the acquired input data at the time of temperature drop prediction and the measured temperature drop value (step S6132).

[0069] The control unit 10 sequentially inputs the input data at the time of temperature drop prediction to the temperature drop prediction model 613 and sequentially outputs the molten steel temperature drop value. The control unit 10 compares the output molten steel temperature drop value with the measured temperature drop value, which is the teacher data included in the temperature training data 623, and optimizes various parameters of the intermediate layer so that the two are approximated to generate the temperature drop prediction model 613.

[0070] The control unit 10 stores the generated temperature drop prediction model 613 in the learning model management DB 63 of the mass storage unit 60 (step S6133) and ends a series of processes. Specifically, the control unit 10 assigns a model ID to the generated temperature drop prediction model 613, and stores the file of the temperature drop prediction model 613 and the generation date and time as one record in the learning model management DB 63 in association with the assigned model ID.

[0071] Note that in this embodiment, the storage unit 20 and the mass storage unit 60 may be configured as an integrated storage device. Further, the mass storage unit 60 may be configured by a plurality of storage devices. Furthermore, the mass storage unit 60 may be an external storage device connected to the information processing device 100. Note that the information processing device 100 may be configured by a single computer, may be configured by a plurality of distributed computers, or may be executed in a distributed manner by virtual machines.

[0072] The information processing apparatus 100 according to this embodiment having such a configuration performs scheduling processing for a series of operations in the refining process as described above using the prediction results (output data) output from the prediction models 611 to 613.

[0073] FIG. 16 is a flowchart for explaining the scheduling processing in the information processing apparatus 100 according to this embodiment. Hereinafter, an example of the scheduling processing will be described.

[0074] First, the information shown in FIGS. 5 and 6 as setting information necessary for various calculations and the initial components obtained by collecting and analyzing molten steel at the site are input (step S000). An example of the initial components is shown in FIG. 17, and the same components as those whose target ranges are defined in FIG. 6 are analyzed.

[0075] Next, the following calculations are performed. First calculation: Calculation related to the oxygen heating operation (step S100), Second calculation: Calculation related to the slag-forming agent addition operation (step S200), Third calculation: Calculation related to the change in the components of molten steel by the vacuum degassing treatment (step S300), Fourth calculation: Calculation related to the weight and number of additions of alloy iron to be added (step S400), Fifth calculation: Calculation related to the scrap addition (step S500), Sixth calculation: Calculation related to the electric heating to compensate for the decrease in the temperature of molten steel due to the addition of alloy iron and scrap (step S600), Seventh calculation: Calculation related to the temperature of molten steel that decreases during the vacuum degassing treatment (step S700), Eighth calculation: Calculation related to the electric heating performed on the molten steel whose temperature has decreased after the vacuum degassing treatment (step S800)

[0076] The first calculation (S100) is a calculation related to oxygen heating. As described above, the refining operation starts from tapping the molten steel melted in the electric furnace process into a container called a ladle. At this time, since the heat of the molten steel is taken away by the ladle, the temperature of the molten steel drops rapidly. The operation to compensate for this temperature drop in a short time is the oxygen heating operation. As a calculation flow, using the operation condition information obtained in S000, the temperature rise value of the molten steel by the oxygen heat treatment is calculated. Hereinafter, the case of adding Si will be taken as an example for explanation.

[0077] Based on conventional empirical values, when 1% of Si by weight ratio to the molten steel oxidizes, the corresponding required oxygen amount is 8.2 Nm 3 / 1000 kg, and the temperature of the molten steel rising at this time is determined to be 20°C. On the other hand, according to FIG. 5, the oxygen amount used in this embodiment is 100 Nm 3 , and since the weight of the molten steel is 82,800 kg, the oxygen amount is 100 Nm 3 / 82,800 kg in terms of the weight ratio to the molten steel. Here, if the Si amount corresponding to the oxygen amount of 100 Nm 3 / 82,800 kg is set as X (%), the following equation holds. 1 (%): 8.2 (Nm 3 / 1000 kg) = X (%): 100 (Nm 3 / 82,800 kg) From this equation, X = 0.147 (%), and when using 100 Nm 3 of oxygen for 82,800 kg of molten steel, the Si that needs to be added is 82,800 (kg) × 0.147 (%) ≈ 122 (kg) and is obtained.

[0078] Also, the temperature rise of the molten steel is Temperature rise of molten steel = 20 (°C / Si: 1%) × 0.147 (Si: %) = 29.4 (°C) It is calculated that the temperature rise value of the molten steel by the oxygen heating treatment is 29.4 °C. For the convenience of explanation below, it is set to 29 °C. As a result of these, the temperature of the molten steel was 1552 °C at the initial stage of refining, but it is calculated that it rises by 29 °C due to the oxygen heating operation and becomes 1581 °C. Also, in this embodiment, the required time for this oxygen heating operation is fixed at 2 minutes.

[0079] The second calculation (S200) is a calculation related to the slag-forming agent addition operation. The slag-forming agent has two types, CaO as the main component and two others. When these dissolve, a slag layer is formed on the molten steel as described above, and it becomes the heat source of Joule heat in the electric heating. In this slag-forming agent addition operation, not only does the required time actually change depending on the added weight, but the temperature drop of the molten steel also changes. However, considering that these effects are small, the required time in this embodiment is fixed at 2 minutes, and it is assumed that neither the weight nor the temperature of the molten steel changes. For convenience, it is referred to as the second calculation.

[0080] Next, following the actual operation at the site, it is a calculation to determine whether to add scrap of the same component as the target when considering the initial amount of molten steel and adding scrap if there is a shortage. Here, in the actual operation, it is necessary to judge the necessity of scrap addition in consideration of the weight increase due to the addition of alloy iron performed in a later process. Specifically, the judgment of whether there is a shortage of the amount of molten steel is made based on the following formula (S300). Shortage of molten steel amount = Target weight - (Initial weight + Alloy iron addition weight) In this formula, if the "shortage of molten steel amount" is a positive number, this value is the scrap addition weight, and if the "shortage of molten steel amount" is a negative number, scrap addition is not required.

[0081] The weight of alloy iron added is calculated taking into account the compositional fluctuations due to vacuum degassing treatment. Therefore, as shown in Fig. 1, the actual operation is carried out in the order of scrap addition, heating by energization, alloy iron addition, heating by energization, vacuum degassing treatment, etc. In this embodiment, however, the calculations are performed in the order of calculation related to the compositional change of the molten steel due to vacuum degassing treatment (the third calculation), calculation related to the weight and number of additions of the alloy iron to be added (the fourth calculation), calculation related to scrap addition (the fifth calculation), etc. From these relationships, the third calculation / calculation related to the compositional fluctuations due to vacuum degassing treatment and the fourth calculation / calculation related to the alloy iron addition treatment will be described below.

[0082] Fig. 18 is a flowchart showing an overview of the fourth calculation / calculation related to alloy iron addition. Based on the target composition of the molten steel and the compositional change value due to vacuum degassing treatment, the compositional difference between the initial composition and the target composition is calculated (step S401). From the calculated compositional difference and the initial weight of the molten steel, the weight (pure component amount) to be added for each component is calculated (step S402). The weight of the alloy iron to be added is calculated according to the order of the priority of the alloy iron given for each component so that the total weight of the main components contained in the alloy iron to be added satisfies the compositional difference (step S403). At this time, it is confirmed that the other components do not deviate from the target range due to the increase of the other components other than the main components.

[0083] For each of groups 1 to 3, the total weight of the alloy iron to be added is calculated (step S404). For each of groups 1 to 3, the number of additions N of the alloy iron is calculated so that the weight of one addition of the alloy iron does not exceed 4000 kg (step S405).

[0084] The number of additions N is determined by summing up the weights of the alloy iron added for adjusting the same component in the order of the highest priority, and determining the number of additions so that the total becomes a maximum of 4000 kg at that time. For example, when there are two types of alloy iron added for the same component, with priority 1: 2500 kg and priority 2: 3500 kg, the following two additions are made. First addition: Priority 1: 2500 kg + Priority 2: 1500 kg Second addition: Priority 2: 2000 kg

[0085] The following describes the third and fourth calculations in accordance with this embodiment. The calculation of component variations by the third calculation / vacuum degassing treatment is performed using the component variation prediction model 611 shown in FIG. 4. First, the addition weight of alloy iron is calculated for Si. As shown in FIG. 7, Si belongs to one group in the addition of alloy iron, so the addition weight of alloy iron is calculated first. In this embodiment, although the specific values are not shown for both the specific target component range shown in FIG. 6 and the initial components shown in FIG. 17, the initial component of Si is below the target lower limit regardless of the refining charge and is almost 0%. The reason for this is that a large amount of oxygen is used during scrap melting in the electric furnace, and the oxidized Si is absorbed into the electric furnace slag as SiO2 and separated from the molten steel when tapping into the ladle.

[0086] Furthermore, as shown in FIG. 10, Si decreases by 0.153% by performing vacuum degassing treatment, and this value is the result of the third calculation calculated by the above-described component variation prediction model 611. And the Si addition (%) is shown by the following formula. Si addition (%) = target value - initial component + decrease of 0.153 (%) in vacuum degassing treatment For example, when the initial component is 0%, the target value is 0.04%, and the decrease in vacuum degassing treatment is 0.153%, the amount of Si to be added is Si amount (%) = 0.04 (%) - 0 (%) + 0.153 (%) = 0.193 (%) Moreover, since this Si amount (%) is the ratio with respect to the initial weight of 82,800 kg, the addition weight is calculated by the following formula. Addition weight = initial weight × amount to be added (%) ÷ content (%) contained in alloy iron

[0087] In this embodiment, since the upper limit value of Al is provided in FIG. 6, Al cannot be used during the refining of this steel grade. Therefore, it is a Si-deoxidized steel in which deoxidation is carried out with Si. Accordingly, the alloy iron used for Si adjustment is defined as low-carbon Fe-Si according to FIG. 7. Furthermore, since the Si content contained in the low-carbon Fe-Si is 75.2 - 75.3% (average 75.25%), Weight of low-carbon Fe-Si to be added (kg) = 82800 (kg) × 0.193 (%) ÷ 75.25 (%) = 212 (kg) ≈ 220 (kg) It becomes as follows.

[0088] Next, calculate the weight of alloy iron added for Ni and Mo, which belong to the same group as Si. Ni and Mo are not shown in FIG. 10, which means that there is no change in the component amount before and after the vacuum degassing treatment. Also, from FIG. 7, when adding Ni, the alloy iron is Ni briquette, and its content is 99.8%. When adding Mo, the alloy iron is Fe-Mo, and its content is 61.0 - 62.5% (average 61.75%). From these conditions, the weight of alloy iron added to adjust Ni and Mo, similar to Si, Weight of alloy iron added = Initial weight × Amount to be added (%) ÷ Content contained in the alloy iron (%) is calculated by, and as a result, Addition weight of Ni briquette: 400 kg Addition weight of Fe-Mo: 120 kg is calculated as such. In this way, the weights of alloy iron added for Si, Ni, and Mo are calculated respectively, and it is calculated that a total addition of 740 kg (= 220 + 400 + 120) is required for the whole group.

[0089] Next, an example of the calculation for Mn belonging to Group 2 in FIG. 7 is illustrated. Mn has the initial component XXXX in FIG. 17, but this value is lower than the target value shown in FIG. 6, and the reason for this is the same as that of Si, which is due to the use of oxygen in the electric furnace. Also, similar to Si, by performing the vacuum degassing treatment to be carried out later, as shown in FIG. 10, it increases by 0.01%. Therefore, the addition (%) of Mn is shown by the following formula, Mn addition (%) = target value - initial composition - increase during vacuum degassing treatment (0.01%) Here, when the initial composition is 0.09%, the target value is 0.28%, and the increase during vacuum degassing treatment is 0.01%, the Mn to be added is 0.28 - 0.09 - 0.01 = 0.18 (%). Furthermore, since this Mn amount (%) is the ratio with respect to the initial weight of 82,800 kg, the added weight is calculated by the following formula. Added weight (kg) = initial weight (kg) × addition amount (%) ÷ content (%) contained in ferroalloy As an example, when adding 0.18% of Mn to an initial weight of 82,800 kg, if the Mn amount contained in the ferroalloy is 100%, the required added weight is 82,800 (kg) × 0.18 (%) ÷ 100 (%) ≈ 150 (kg) That is.

[0090] On the other hand, when adding Mn, as shown in Fig. 7, the selection priority order is: 1st / Si-Mn, 2nd / high-carbon Fe-Mn, 3rd / low-carbon Fe-Mn, and the Mn amount (%) contained in each and the amount of impurity components other than Mn are determined. Therefore, when adding 150 kg of pure Mn, first consider adding Si-Mn with the highest priority. The Si-Mn with the highest priority has a Mn content of 78.0 - 79.6% (average 78.8%). When trying to add the above 150 kg of Mn only with Si-Mn, the added weight is Added weight (kg) = 150 (kg) kg ÷ 78.8 (%) = 190 (kg) That is.

[0091] However, since Si-Mn only contains 10.9 - 13.1% Si (average 12.0%), the 190 kg of Si-Mn contains 23 kg of Si, and this weight will increase the Si content of the molten steel by 0.03%. Here, since the Si content has already reached the target value due to the addition of 1 group / low-carbon Fe-Si, the Si-Mn with the highest priority for Mn addition cannot be used, and instead, it is necessary to add high-carbon Fe-Mn or low-carbon Fe-Mn with a low Si content. Since the Mn contained in high-carbon Fe-Mn is 74.2 - 76.2% (average 75.2%) from Figure 7, about 200 kg of high-carbon Fe-Mn is required to add 150 kg of pure Mn. However, since the components other than Mn are sufficiently low, the influence on the components other than Mn is not significantly problematic.

[0092] In this way, although the priority order of alloy irons used for each component is determined, due to the balance with other components contained in each alloy iron, there are restrictions on the usable weight. However, the method for calculating the weight of the alloy iron to be added is based on the change amount of the components of the molten steel by vacuum degassing treatment, the initial components, the target components, the weight of the molten steel, the content of each alloy iron, and the content of other components, in the same way as explained for Si, Ni, Mo, and Mn above. Addition amount (%) = Target value - Initial component ± Variation in vacuum degassing treatment Addition weight (kg) = Initial weight (kg) × Addition amount (%) ÷ Component amount (%) contained in the alloy iron Based on this, the calculation is performed for each necessary component. As a result, the alloy iron to be added is as shown in Figure 19, and the fourth calculation is completed. At this time, in this embodiment, the total of the alloy iron to be added is calculated to be 2920 kg.

[0093] After performing the calculations up to this point (after calculating the added weight of alloy iron), the calculation related to the fifth calculation / scrap addition becomes possible. FIG. 20 is a flowchart showing an overview of the calculation related to the fifth calculation / scrap addition. First, it is determined whether the sum of the initial weight and the added weight of alloy iron is less than the target weight (step S501). If the former is greater than or equal to the latter (S501: NO), scrap addition is unnecessary and the process ends. On the other hand, if the former is less than the latter (S501: YES), since scrap addition is necessary, the added weight of scrap is calculated according to the following formula (step S502). Added weight of scrap (kg) =(Target weight - (Initial weight + Added weight of alloy iron)) (kg)

[0094] The number of scrap addition times N is calculated so that the added weight of scrap per addition does not exceed 4000 kg (step S503). Specifically, the number of alloy iron addition times N is determined as follows. Number of scrap addition times N = [Added weight of scrap (kg) ÷ 4000 (kg)] + 1

[0095] Next, based on the determined number of addition times N, the added weight for each time is calculated (step S504). Specifically, the added weight is determined according to the number of addition times N as follows. · When N = 1 Added weight for the first time (kg) = Added weight of scrap (kg) · When N ≥ 2 Added weight for the 1st to (N - 1)th times (kg) = 4000 (kg) Added weight for the Nth time (kg) = Total added weight of scrap (kg) - 4000 × (N - 1) (kg)

[0096] In this embodiment, the molten steel with an initial weight of 82,800 kg further increases in weight due to the addition of a total of 2,920 kg of alloy iron for composition adjustment. As a result, since it exceeds the target weight of 79,700 kg (S501: NO), it is determined that scrap addition is unnecessary. On the other hand, if scrap addition is required, S502 to S504 described above are executed to calculate the number of scrap additions and the added weight.

[0097] Calculations are performed as described above. Calculations related to the change in the composition of the molten steel by the third calculation / vacuum degassing treatment shown in FIG. 16, the fourth calculation / weight of the alloy iron to be added, and the fifth calculation / determination of the necessity of scrap addition are carried out. In addition to the calculations related to scrap addition corresponding to the shortage of the weight of the molten steel, the number of additions and the added weight are calculated for the alloy iron addition performed for adjusting the chemical composition.

[0098] Since the fifth calculation is completed in this way, summarizing the scheduling of the refining process at this time results in FIG. 21. The temperature drops caused by the three alloy iron additions are calculated by the following calculations respectively. ·Temperature drop for the first time Ratio to molten steel weight = 740 (kg) ÷ 82,800 (kg) × 100 = 0.89 (%) Temperature drop of molten steel (°C) = 0.89 (%) × 20 (°C / %) ≈ 18 (°C) ·Temperature drop for the second time Ratio to molten steel weight = 2,080 (kg) ÷ {82,800 (kg) + 740 added for the first time (kg)} × 100 = 2.49 (%) Temperature drop of molten steel (°C) = 2.49 (%) × 20 (°C / %) ≈ 50 (°C) ·Temperature drop for the third time Ratio to molten steel weight = 100 (kg) ÷ {82,800 (kg) + 740 added for the first time (kg) + 2,080 added for the second time (kg)} × 100 = 0.12 (%) Temperature drop of molten steel (°C) = 0.12 (%) × 20 (°C / %) ≈ 2 (°C)

[0099] Next, as the sixth calculation shown in FIG. 16, a calculation related to energization heating is performed to compensate for the decrease in the molten steel temperature due to the addition of alloy iron and scrap (step S600). FIG. 22 shows a flowchart in the sixth calculation. First, 1 is set to the variable i (step S601).

[0100] The sixth calculation is performed when alloy iron and scrap are added. Among the total N times of alloy iron addition, while changing the variable i from 1 to N, calculations related to alloy iron addition for N times are performed. In this embodiment, since the alloy iron is added 3 times, N = 3. Also, in this embodiment, since the weight of the molten steel is ensured and scrap addition is not performed, energization heating before scrap addition is not performed. Further, in this embodiment, when the oxygen heating is completed, the temperature of the molten steel is 1581 °C and the weight is 82,800 kg, and the above-mentioned three times of alloy iron addition are performed on the molten steel in this state.

[0101] First, the target temperature of the molten steel at each time is calculated. This calculates the target temperature of energization each time from the temperature at the end of oxygen heating as the start temperature to the target temperature of 1600 °C before vacuum degassing treatment. Here, the target temperature is calculated without considering the temperature drop of the molten steel due to the addition of scrap or alloy iron. Also, in this embodiment, initially, with i = the first time, it is raised from the current start temperature of 1581 °C to 1600 °C in three energizations. However, after that, when i = the second time, the number of available energizations becomes 2 remaining, and when i = the third time, the number of available energizations becomes 1 remaining.

[0102] Therefore, the target temperature after the i-th addition is, Target temperature after the i-th alloy iron addition (°C) = 1581+(1600 - 1581)÷(N-(i - 1))×i It can be expressed as, and specifically, the results in the following table are calculated (step S602).

[0103]

Table 1

[0104] On the other hand, although the amount of temperature decrease of the molten steel due to the addition of alloy iron has been obtained by the above-described calculation, even if such a temperature decrease occurs, it is necessary to perform preheating by energization in advance so that the temperature of the molten steel does not fall below the allowable minimum temperature of 1540°C. Specifically, it is necessary to calculate what temperature the molten steel will reach with the first addition of alloy iron to determine whether preheating by energization is required before the first addition of alloy iron. In this embodiment, the temperature of the molten steel is 1581°C after oxygen heating. When the first addition of alloy iron is performed here, the temperature drops by 18°C. Therefore, Molten steel temperature (°C) after the first addition of alloy iron = Temperature before the first addition of alloy (°C) - Temperature decrease due to the first addition of alloy (°C) = 1581 (°C) - 18 (°C) = 1563 (°C) > Allowable minimum temperature of 1540 (°C) From the above calculation, since the temperature after addition is higher than the allowable minimum temperature, it is determined that preheating by energization is not required before the first addition (step S603: YES).

[0105] On the other hand, if it is determined in S603 that preheating by energization is necessary (step S603: NO), it is necessary to give the molten steel a temperature lower than the minimum allowable temperature of 1540°C in advance by adding alloy iron. Therefore, after giving the target temperature by the following formula, preheating calculation is performed by the preheating prediction model 612 to raise the temperature of the molten steel before adding alloy iron (step S604). Target temperature (°C) = Current temperature + Temperature lower than 1540 (°C) due to the addition of alloy iron = Current temperature + 1540 (°C) - (Current temperature - Temperature decrease due to addition) = 1540 (°C) + Temperature decrease due to addition After performing preheating by energization as necessary in this way, the first addition of alloy iron is performed. In step S605, the temperature of the molten steel after addition is obtained.

[0106] Next, it is determined whether the second addition of alloy iron can be performed at the current temperature (step S606). In this embodiment, Molten steel temperature - Temperature drop due to the second addition (°C) = 1563 - 50 = 1513 <1594 (Target temperature after the second addition) Therefore, it is determined that energization heating is required before the second addition (S606: NO).

[0107] Thus, energization heating is required before the second addition, but it is necessary to determine the target temperature for energization. First, in step S601, the target temperature after the first addition was obtained as 1587°C. However, since this temperature does not consider the temperature drop due to the addition of ferroalloy, it is necessary to determine whether it is appropriate as the target temperature for energization heating. Therefore, assuming heating up to 1587°C and considering the case where the second addition is performed at this temperature, the molten steel temperature is Molten steel temperature (°C) = 1587 - 50 = 1537 <1540 It can be seen that the temperature of 1587°C is too low as the energization target temperature, and a higher temperature is required (S607: NO).

[0108] Therefore, in the same way as S604, a new energization heating target temperature is calculated as follows (step S608). Target temperature (°C) = Current temperature + Temperature below 1540°C due to the addition of ferroalloy = Current temperature + (1540 - (Current temperature - Temperature drop due to the addition of ferroalloy)) = 1540 + Temperature drop due to the addition of ferroalloy By this calculation, 1590 (= 1540 + 50) (°C) becomes the new energization heating target temperature.

[0109] When the calculation for i = 1st time is completed in this way (S610: YES), i is incremented by 1 (step S611), and then the calculation for i = 2nd time is performed. Similar to the above case, the target temperature after the second addition is calculated (S602). Since the first addition has already been completed, the remaining number of additions is 2. The current start temperature of 1581°C is raised to 1600°C in 3 energizations. Since the calculation is first performed with i = 1st time, Target temperature (°C) after the i-th addition of alloy iron = 1581 + (1600 - 1581) ÷ N × i It can be expressed as (S602).

[0110] And then 1593 - 50 = 1543 > 1540 (S603: YES) It becomes The molten steel temperature = 1543 °C (S605).

[0111] Next, it is determined whether energized heating is required before the next addition (step S606). Molten steel temperature - Temperature drop due to the third addition (°C) = 1543 - 2 = 1541 < 1591 (Target temperature after the second addition) Based on this, it is determined that energized heating is required before the third addition (S606: NO).

[0112] In this way, energized heating is required before the third addition, but it is necessary to determine the target temperature for energization. Initially, the target temperature after the second addition was obtained as 1591 °C in S501. However, since this temperature does not consider the temperature drop due to the addition of alloy iron, it is necessary to determine whether it is appropriate as the target temperature for energized heating. Therefore, considering the case where the temperature is raised to 1591 °C and the third addition is performed at this temperature, the molten steel temperature is Molten steel temperature (°C) = 1591 - 2 = 1589 > 1540 Based on this, it can be seen that the temperature of 1591 °C is sufficiently high as the energization target temperature, and it is not necessary to change the energization target temperature (S607: YES).

[0113] According to this judgment, since it is not necessary to change the energized heating target temperature, it becomes the initially calculated 1591 °C. Then, the energized heating before the third addition is calculated using the energized heating prediction model 612 (step S609). The calculation result at this time is that the required time is 30 minutes, there is overshoot, and the temperature after heating is 1594 °C.

[0114] In this way, the calculation for the second time (i = 2) is completed. After going through the loop of S610 - S611, next, the calculation for the third time (i = 3) is performed. Similar to the above, the target temperature after the third addition is calculated (S602). Since the second addition has already been completed, the remaining number of additions is one. It is raised from the start temperature of 1581 °C to 1600 °C with one power-on. However, as it is calculated as the third time (i = 3), according to the following formula, Target temperature (°C) after the i-th addition of alloy iron = 1581 + (1600 - 1581) ÷ N × i the target temperature after the third addition of alloy iron is 1600 °C (S602) according to the above formula.

[0115] Also, since the temperature drop during the third addition of alloy iron is 2 °C, 1594 - 2 = 1592 > 1540 (S603: YES) resulting in the molten steel temperature = 1594 °C (step S605).

[0116] Furthermore, in step S405, since the fourth addition of alloy iron is not performed, the temperature drop is regarded as zero °C, molten steel temperature = 1592 - 0 = 1592 < 1600 it can be seen that it is not necessary to change the power-on target temperature (step S605: NO). Based on this judgment, 1600 °C becomes the power-on heating target temperature, and the calculation of power-on heating is performed (step S608). The calculation result at this time is that the required time is 10 minutes, there is overshoot, and the temperature after heating is 1605 °C.

[0117] Summarizing the scheduling of the refining process when the sixth calculation is completed as above, it becomes as shown in Fig. 23. In other words, Fig. 23 is a chart obtained by adding the results of the sixth calculation to the table in Fig. 21.

[0118] After all the alloy iron additions and the pre-vacuum degassing energized heating are completed, next, as the seventh calculation, a calculation related to the molten steel temperature drop during the vacuum degassing treatment is performed (step S700 in FIG. 16). As described above, when the vacuum degassing treatment is performed, the temperature of the molten steel drops. Therefore, after the vacuum degassing treatment, energized heating is performed. However, depending on the degree of temperature drop, the required time for energized heating changes, which affects the scheduling of the refining process.

[0119] The seventh calculation is performed using the temperature drop prediction model 613 shown in FIG. 4. In this embodiment, based on the various conditions of the refining process, it is calculated by the temperature drop prediction model 613 that the temperature of the molten steel drops by 28 °C during the vacuum degassing treatment. That is, in this embodiment, it is understood that from the temperature of the molten steel (1605 °C) before the vacuum degassing treatment when all the alloy iron additions are completed, the temperature of the molten steel drops by 28 °C due to the vacuum degassing treatment and becomes 1577 °C.

[0120] When the seventh calculation is completed, subsequently, as the eighth calculation, a calculation related to the energized heating to be performed on the molten steel whose temperature has dropped due to the vacuum degassing treatment is performed (step S800 in FIG. 16).

[0121] The eighth calculation is performed using the energization heating prediction model 612 shown in FIG. 4. As described above, during vacuum degassing, the molten steel temperature drops by 28°C. Therefore, in order to compensate for the temperature drop of the molten steel, it is necessary to energize and heat the molten steel after the completion of vacuum degassing. Similar to the calculation of the energization heating time after addition described above, the energization heating time after vacuum degassing is calculated using the energization heating prediction model 612. Using the energization heating prediction model 612, the temperature of the molten steel is predicted every 5 minutes, and the predicted temperature of the molten steel is compared with the target temperature at the end of refining (1650°C shown in FIG. 5). The time when the predicted temperature of the molten steel is closest to such a target temperature is taken as the energization heating time after vacuum degassing. Specifically, when the energization time is 50 minutes, the temperature of the molten steel output from the energization heating prediction model 612 is 1647°C, and when the energization time is 55 minutes, the temperature of the molten steel output from the energization heating prediction model 612 is 1655°C. In this case, "50 minutes" corresponding to 1647°C, which is closest to the target temperature at the end of refining (1650°C), is taken as the energization heating time after vacuum degassing.

[0122] The results of performing the first to eighth calculations as described above are shown in FIG. 24. This FIG. 24 is a chart exemplifying the result of the scheduling process performed by the information processing apparatus 100 of the present embodiment. The scheduling result in FIG. 24 includes a work name column, a required time column, a work content column, a molten steel weight column, and a molten steel temperature column. Each work in the refining process is shown in the work name column. In the required time column, the required time calculated as described above and the predetermined required time are shown for each work. The detailed work content in each work is shown in the work content column. The weight of the molten steel before and after each work is shown in the molten steel weight column. The temperature of the molten steel before and after each work is shown in the molten steel temperature column.

[0123] The present invention is not limited to the above description. For example, the control unit 10 may be configured to display the result of the scheduling process shown in FIG. 24 on the display unit 40.

[0124] As described above, by performing a series of complex operations in the refining process using the result scheduled by the information processing apparatus 100 of the present embodiment (hereinafter referred to as a work schedule), it is possible to support the work of workers in all steps of the refining process, suppress variations in the work of each worker, and stabilize the quality of the product.

[0125] Also, as described above, since the information processing apparatus 100 of the present embodiment uses the machine-learned prediction models 611 to 613 to create a work schedule, a work schedule can be created based on a highly accurate prediction.

[0126] Furthermore, as described above, when the information processing apparatus 100 of the present embodiment calculates the addition weight of alloy iron, it uses the amount of component variation after degassing output by the component variation prediction model 611. That is, when calculating the addition weight of alloy iron, since the variation in the component amount due to the reduction reaction that occurs during vacuum degassing is considered, a more accurate result can be obtained.

[0127] Then, as described above, the information processing apparatus 100 of the present embodiment predicts the temperature of the molten steel while extending the energization time at a predetermined time interval, and compares each predicted temperature of the molten steel with the target temperature, thereby calculating the energization heating time. Therefore, the information processing apparatus 100 of the present embodiment can more quickly reduce the burden related to the calculation process of the energization heating time.

[0128] In the above, the case where the component variation prediction model 611 uses elastic net regression has been described as an example, but it is not limited thereto (the refining process is a reduction process that reduces the oxygen concentration in the molten steel while using Si or Al as a deoxidizing agent, and in this case, since the prediction of the influence of the oxidation-reduction reaction is less affected by other external influences, more accurate results can be obtained with elastic net regression than with random forest regression, and elastic net regression was selected). For example, ridge regression or Lasso regression may be used.

[0129] In the above description, the energization heating prediction model 612 and the temperature drop prediction model 613 were described by taking the case of using a random forest model as an example, but the present invention is not limited thereto (since the change in the molten steel temperature is also affected by parameters such as the heat storage amount of the ladle that are difficult to measure and are not included in the input conditions, a higher-precision result can be obtained by random forest regression than by elastic net regression, and random forest regression was selected). For example, a logistic regression model, a support vector machine, etc. may be used.

[0130] In the above description, the case where the training (learning) of the prediction models 611 to 613 is performed by the information processing apparatus 100 was described by taking it as an example, but the present invention is not limited thereto, and it may be executed by other computers. That is, the learned prediction models 611 to 613 learned by other computers may be deployed to the information processing apparatus 100.

[0131] It should be considered that all the embodiments disclosed this time are illustrative in all respects and not restrictive. The scope of the present invention is shown not by the above description but by the claims, and it is intended that all modifications within the meaning and scope equivalent to the claims are included.

[0132] The matters described in each embodiment can be combined with each other. Also, the independent claims and dependent claims described in the claims can be combined with each other in all possible combinations regardless of the citation format. Further, although the claims use a format (multi-claim format) in which claims that cite two or more other claims are described, the present invention is not limited thereto.

Explanation of Reference Numerals

[0133] 10 Control unit 20 Storage unit 62 Training data DB 63 Learning model management DB 100 Information processing apparatus 611 Component variation prediction model (first learning model) 612 Electric current application heat generation prediction model (second learning model) 613 Temperature drop prediction model (third learning model)

Claims

1. The first calculation related to the oxygen heating operation for increasing the molten steel temperature at the initial stage of refining, the second calculation related to the slag-forming agent addition operation, the third calculation related to the component change of the molten steel by vacuum degassing treatment, the fourth calculation related to the addition of alloy iron, the fifth calculation related to the addition of scrap, the sixth calculation related to the electric heating for compensating the decrease in the molten steel temperature accompanying the addition of alloy iron and scrap, the seventh calculation related to the molten steel temperature decreasing during vacuum degassing treatment, the eighth calculation related to the electric heating for compensating the decrease in the molten steel temperature by vacuum degassing treatment are executed, the required time necessary for the oxygen heating operation, slag-forming agent addition operation, scrap addition operation, alloy iron addition operation, vacuum degassing treatment operation, and electric heating operation is obtained, the addition weight of scrap, the type, addition weight, and addition frequency of the alloy iron to be added are obtained, and a process for scheduling the entire refining process is a program to be executed by a computer.

2. In the first calculation, the program according to claim 1, which calculates the molten steel temperature increased by the oxygen heating operation from the initial weight of the molten steel and the oxygen consumption.

3. In the second calculation, the program according to claim 1, which gives the required time for the operation of adding a slag-forming agent to the molten steel.

4. In the third calculation, first input data including the molten steel component amount at the initial stage of refining, the molten steel weight at the initial stage of refining, the slag-forming agent consumption, and the vacuum degassing conditions is acquired, and the component fluctuation amount after vacuum degassing treatment is output when the first input data is input by inputting the acquired first input data into a first learning model learned to output the component fluctuation amount after vacuum degassing treatment. The program according to claim 1.

5. In the fourth calculation, the addition weight of the alloy iron is calculated based on the target amount of the target component in the molten steel, the molten steel component amount at the initial stage of refining, and the calculated component fluctuation amount after vacuum degassing treatment, and the addition weight and addition frequency at each addition are calculated so as not to exceed the maximum allowable addition weight per addition. The program according to claim 4.

6. In the fifth calculation, the program according to claim 5, which calculates the addition weight of the scrap based on the molten steel weight at the initial stage of refining, the target weight of the molten steel after the end of refining, and the calculated addition weight of the alloy iron.

7. In the sixth calculation, a decrease value of the molten steel temperature is calculated based on the molten steel temperature at the initial stage of refining, the added weight of the ferroalloy, and the added weight of the scrap, and it is determined whether energization heating is necessary so as not to fall below the minimum allowable temperature. When energization heating is necessary, after setting the energization target temperature, the second input data including information identifying the ladle furnace, the molten steel temperature before energization, and the energization target temperature is acquired, and the second input data is input to the second learning model trained to output the rising temperature of the molten steel during energization, and the energization time is obtained using the rising temperature during energization obtained by inputting the acquired second input data. The program according to claim 1.

8. In the seventh calculation, the third input data including information identifying the ladle furnace, the molten steel weight and temperature at the initial stage of refining, the target temperature at the start of vacuum degassing treatment, the target temperature at the end of refining, and the amount of slag-forming agent used is acquired, and the third input data is input to the third learning model trained to output the decrease value of the molten steel temperature due to vacuum degassing treatment when the third input data is input, and the decrease value of the molten steel temperature due to vacuum degassing is obtained. The program according to claim 1.

9. In the eighth calculation, the second input data including information identifying the ladle furnace, the molten steel temperature before energization, and the energization target temperature is acquired, and the second input data is input to the second learning model trained to output the rising temperature of the molten steel during energization when the second input data is input, and the energization time is obtained using the rising temperature during energization obtained by inputting the acquired second input data. The program according to claim 1.

10. In the scheduling, the required time for the oxygen heating operation and the slag-forming agent addition operation determined in advance, the addition operation time of the ferroalloy calculated using the number of ferroalloy additions obtained in the fourth calculation, the addition operation time of the scrap calculated using the number of scrap additions obtained in the fifth calculation, the energization time required for energization heating obtained in the sixth calculation, the predetermined vacuum degassing treatment time, and the energization time required for energization heating after vacuum degassing obtained in the eighth calculation are combined in time series, In the third calculation, the change in the component amount of the molten steel before and after vacuum degassing treatment is predicted, in the fourth calculation, the weight and number of additions of the ferroalloy to be added are calculated, and in the fifth calculation, the weight and number of additions of the scrap to be added are calculated, The program according to any one of claims 1 to 9, which is capable of predicting in advance the detailed content of the refining process operation and assisting the operator.

11. Comprising a control unit, the control unit The first calculation related to the oxygen heating operation for raising the molten steel temperature at the initial stage of refining, The second calculation related to the slag-forming agent addition operation, The third calculation related to the component change of the molten steel by vacuum degassing treatment, The fourth calculation related to the addition of alloy iron, The fifth calculation related to the addition of scrap, The sixth calculation related to the electric heating for compensating the decrease in the molten steel temperature accompanying the addition of alloy iron and scrap, The seventh calculation related to the molten steel temperature decreasing in the vacuum degassing treatment, The eighth calculation related to the electric heating for compensating the decrease in the molten steel temperature by the vacuum degassing treatment are executed, An information processing device that obtains the required time required for the oxygen heating operation, slag-forming agent addition operation, scrap addition operation, alloy iron addition operation, vacuum degassing treatment operation, and electric heating operation, obtains the addition weight of scrap, the type, addition weight, and addition times of the alloy iron to be added, and schedules the entire refining process.

12. The first calculation related to the oxygen heating operation for raising the molten steel temperature at the initial stage of refining, The second calculation related to the slag-forming agent addition operation, The third calculation related to the component change of the molten steel by vacuum degassing treatment, The fourth calculation related to the addition of alloy iron, The fifth calculation related to the addition of scrap, The sixth calculation related to the electric heating for compensating the decrease in the molten steel temperature accompanying the addition of alloy iron and scrap, The seventh calculation related to the molten steel temperature decreasing in the vacuum degassing treatment, The eighth calculation related to the electric heating for compensating the decrease in the molten steel temperature by the vacuum degassing treatment are executed, An information processing method in which a computer executes a process of obtaining the required time required for the oxygen heating operation, slag-forming agent addition operation, scrap addition operation, alloy iron addition operation, vacuum degassing treatment operation, and electric heating operation, obtaining the addition weight of scrap, the type, addition weight, and addition times of the alloy iron to be added, and scheduling the entire refining process.

Images