WO2020170849A1 - 操業結果予測方法、学習モデルの学習方法、操業結果予測装置および学習モデルの学習装置 - Google Patents
操業結果予測方法、学習モデルの学習方法、操業結果予測装置および学習モデルの学習装置 Download PDFInfo
- Publication number
- WO2020170849A1 WO2020170849A1 PCT/JP2020/004738 JP2020004738W WO2020170849A1 WO 2020170849 A1 WO2020170849 A1 WO 2020170849A1 JP 2020004738 W JP2020004738 W JP 2020004738W WO 2020170849 A1 WO2020170849 A1 WO 2020170849A1
- Authority
- WO
- WIPO (PCT)
- Prior art keywords
- learning
- learning model
- operation result
- model
- operating conditions
- Prior art date
Links
- 238000000034 method Methods 0.000 title claims abstract description 139
- 238000004519 manufacturing process Methods 0.000 claims abstract description 44
- 238000007664 blowing Methods 0.000 claims description 66
- QVGXLLKOCUKJST-UHFFFAOYSA-N atomic oxygen Chemical compound [O] QVGXLLKOCUKJST-UHFFFAOYSA-N 0.000 claims description 37
- 229910052760 oxygen Inorganic materials 0.000 claims description 37
- 239000001301 oxygen Substances 0.000 claims description 37
- 229910000831 Steel Inorganic materials 0.000 claims description 29
- 239000010959 steel Substances 0.000 claims description 29
- 239000007789 gas Substances 0.000 claims description 25
- 239000002436 steel type Substances 0.000 claims description 24
- 238000004364 calculation method Methods 0.000 description 6
- 238000012417 linear regression Methods 0.000 description 6
- 238000012795 verification Methods 0.000 description 6
- 229910000655 Killed steel Inorganic materials 0.000 description 4
- 238000005261 decarburization Methods 0.000 description 4
- 238000007670 refining Methods 0.000 description 4
- 238000000611 regression analysis Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 3
- 238000010438 heat treatment Methods 0.000 description 3
- 229910052751 metal Inorganic materials 0.000 description 3
- 239000002184 metal Substances 0.000 description 3
- 229910001220 stainless steel Inorganic materials 0.000 description 3
- 229910000975 Carbon steel Inorganic materials 0.000 description 2
- XEEYBQQBJWHFJM-UHFFFAOYSA-N Iron Chemical compound [Fe] XEEYBQQBJWHFJM-UHFFFAOYSA-N 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 2
- 239000010962 carbon steel Substances 0.000 description 2
- 230000000052 comparative effect Effects 0.000 description 2
- 230000002596 correlated effect Effects 0.000 description 2
- 238000006477 desulfuration reaction Methods 0.000 description 2
- 230000023556 desulfurization Effects 0.000 description 2
- 238000007637 random forest analysis Methods 0.000 description 2
- 239000010935 stainless steel Substances 0.000 description 2
- 238000009628 steelmaking Methods 0.000 description 2
- 238000003860 storage Methods 0.000 description 2
- 229910000669 Chrome steel Inorganic materials 0.000 description 1
- 230000001276 controlling effect Effects 0.000 description 1
- 239000000498 cooling water Substances 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000010891 electric arc Methods 0.000 description 1
- 230000010365 information processing Effects 0.000 description 1
- 229910052742 iron Inorganic materials 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000002844 melting Methods 0.000 description 1
- 230000008018 melting Effects 0.000 description 1
- 239000000203 mixture Substances 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 239000002994 raw material Substances 0.000 description 1
- 230000000630 rising effect Effects 0.000 description 1
- 238000010079 rubber tapping Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
Images
Classifications
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F27—FURNACES; KILNS; OVENS; RETORTS
- F27B—FURNACES, KILNS, OVENS, OR RETORTS IN GENERAL; OPEN SINTERING OR LIKE APPARATUS
- F27B3/00—Hearth-type furnaces, e.g. of reverberatory type; Tank furnaces
- F27B3/10—Details, accessories, or equipment peculiar to hearth-type furnaces
- F27B3/28—Arrangement of controlling, monitoring, alarm or the like devices
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F27—FURNACES; KILNS; OVENS; RETORTS
- F27D—DETAILS OR ACCESSORIES OF FURNACES, KILNS, OVENS, OR RETORTS, IN SO FAR AS THEY ARE OF KINDS OCCURRING IN MORE THAN ONE KIND OF FURNACE
- F27D19/00—Arrangements of controlling devices
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F27—FURNACES; KILNS; OVENS; RETORTS
- F27D—DETAILS OR ACCESSORIES OF FURNACES, KILNS, OVENS, OR RETORTS, IN SO FAR AS THEY ARE OF KINDS OCCURRING IN MORE THAN ONE KIND OF FURNACE
- F27D21/00—Arrangements of monitoring devices; Arrangements of safety devices
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F27—FURNACES; KILNS; OVENS; RETORTS
- F27D—DETAILS OR ACCESSORIES OF FURNACES, KILNS, OVENS, OR RETORTS, IN SO FAR AS THEY ARE OF KINDS OCCURRING IN MORE THAN ONE KIND OF FURNACE
- F27D19/00—Arrangements of controlling devices
- F27D2019/0003—Monitoring the temperature or a characteristic of the charge and using it as a controlling value
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F27—FURNACES; KILNS; OVENS; RETORTS
- F27D—DETAILS OR ACCESSORIES OF FURNACES, KILNS, OVENS, OR RETORTS, IN SO FAR AS THEY ARE OF KINDS OCCURRING IN MORE THAN ONE KIND OF FURNACE
- F27D19/00—Arrangements of controlling devices
- F27D2019/0006—Monitoring the characteristics (composition, quantities, temperature, pressure) of at least one of the gases of the kiln atmosphere and using it as a controlling value
- F27D2019/0009—Monitoring the pressure in an enclosure or kiln zone
-
- F—MECHANICAL ENGINEERING; LIGHTING; HEATING; WEAPONS; BLASTING
- F27—FURNACES; KILNS; OVENS; RETORTS
- F27D—DETAILS OR ACCESSORIES OF FURNACES, KILNS, OVENS, OR RETORTS, IN SO FAR AS THEY ARE OF KINDS OCCURRING IN MORE THAN ONE KIND OF FURNACE
- F27D19/00—Arrangements of controlling devices
- F27D2019/0096—Arrangements of controlling devices involving simulation means, e.g. of the treating or charging step
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N20/00—Machine learning
- G06N20/20—Ensemble learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/06—Resources, workflows, human or project management; Enterprise or organisation planning; Enterprise or organisation modelling
- G06Q10/063—Operations research, analysis or management
- G06Q10/0637—Strategic management or analysis, e.g. setting a goal or target of an organisation; Planning actions based on goals; Analysis or evaluation of effectiveness of goals
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/04—Manufacturing
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Definitions
- the present invention relates to an operation result prediction method, a learning model learning method, an operation result prediction device, and a learning model learning device.
- Patent Documents 1 and 2 disclose a method for determining the amount of input electric power required for an arc furnace used for melting raw materials such as scrap, adjusting the components, and raising the temperature.
- the required amount of input power is calculated from the component concentration of the molten steel, and in addition, when the temperature of the molten steel needs to be raised, the amount of power generated by the difference from the reference tapping temperature is calculated. It is moderate. However, there are cases where the target amount of temperature rise of molten steel is unknown in advance, and in that case, it is necessary to predict the amount of electric power from other variables.
- the present invention has been made in view of the above, and an operation result prediction method, a learning model learning method, an operation result prediction device, and a learning model learning device capable of highly accurately predicting an operation result of an industrial process.
- the purpose is to provide.
- the operation result prediction method is an operation result prediction method for predicting the operation result of the industrial process from a plurality of operating conditions of the industrial process, Model selection for selecting a specific learning model from a plurality of learning models according to whether or not to use a specific key operating condition among the plurality of operating conditions as an explanatory variable when predicting an operating result And a prediction step of predicting an operation result based on the learning model selected in the model selection step.
- the plurality of learning models include an operation condition that is the specific key, and is classified by an operation condition related to the operation condition that is the specific key.
- a first learning model created from untrained first teacher data and a first learning model that does not include the specific key operating conditions and is classified by operating conditions related to the specific key operating conditions A second learning model for creating two teacher data, wherein the model selecting step selects the first learning model when the operating condition serving as the specific key is used as the explanatory variable, When the specific key operating condition is not used as an explanatory variable, the second learning model is selected.
- the industrial process is an arc process in a steel mill
- the operation result is the power used in batch operation of the arc process
- the plurality of operations The conditions include at least the temperature increase target amount, the scheduled processing time, and the steel type information.
- the industrial process is a converter process in a steel mill
- the operation result is a gas generated in batch operation of the converter process
- a learning model learning method is a learning model learning method used when predicting an operation result of an industrial process from a plurality of operating conditions of the industrial process.
- a first teacher data that includes an operating condition that is a specific key among the plurality of operating conditions and that is not classified by an operating condition that is related to the operating condition that is the specific key is created.
- Data creating step, and a second data creating step of creating second teacher data that is classified by operating conditions that do not include the specific key operating conditions and that are related to the specific key operating conditions
- the industrial process is an arc process in a steel mill
- the operation result is the power used in batch operation of the arc process
- the plurality of The operating conditions include at least the temperature increase target amount, the scheduled processing time, and the steel type information.
- the industrial process is a converter process in a steel mill
- the operation result is a gas generated in batch operation of the converter process
- the plurality of operating conditions include at least a planned blowing oxygen amount, a planned treatment time, and a blowing mode.
- the operation result prediction device is an operation result prediction device that predicts the operation result of the industrial process from a plurality of operating conditions of the industrial process, Model selection for selecting a specific learning model from a plurality of learning models according to whether or not to use a specific key operating condition among the plurality of operating conditions as an explanatory variable when predicting an operating result And a predicting unit that predicts the operation result based on the learning model selected by the model selecting unit.
- a learning model learning device is a learning model learning device used when predicting an operation result of an industrial process from a plurality of operating conditions of the industrial process.
- a first teacher data that includes an operating condition that is a specific key among the plurality of operating conditions and that is not classified by an operating condition that is related to the operating condition that is the specific key is created.
- Data creating means, and second data creating means for creating second teacher data that is classified by operating conditions that do not include the specific key operating conditions and that are related to the specific key operating conditions.
- a first model creating means for creating a first learning model for predicting an operation result of the industrial process based on the first teacher data, and the industrial process based on the second teacher data.
- Second model creating means for creating a second learning model for predicting the operation result of.
- a learning model with high prediction accuracy is selected and predicted according to an operating condition that can be grasped in advance. It can be predicted with high accuracy.
- FIG. 1 is a block diagram showing a schematic configuration of an operation result prediction device and a learning device according to an embodiment of the present invention.
- FIG. 2 is a flowchart showing the flow of the learning method of the learning model according to the embodiment of the present invention.
- FIG. 3 is a flowchart showing details of the data creation step in the learning method of the learning model according to the first embodiment of the present invention.
- FIG. 4 is a flowchart showing the flow of the operation result prediction method according to the first embodiment of the present invention.
- FIG. 5 is a flowchart showing details of the data creation step in the learning method of the learning model according to the second embodiment of the present invention.
- FIG. 6 is a flowchart showing the flow of the operation result prediction method according to the second embodiment of the present invention.
- FIG. 1 is a block diagram showing a schematic configuration of an operation result prediction device and a learning device according to an embodiment of the present invention.
- FIG. 2 is a flowchart showing the flow of the learning method of the learning model according
- FIG. 7 is an example of the operation result prediction device and the learning device according to the first embodiment of the present invention, and is a graph showing the verification accuracy of the predicted power consumption.
- FIG. 8 is an example of the operation result prediction device and the learning device according to the second embodiment of the present invention, and is a graph showing the verification accuracy of the predicted generated gas.
- learning device An operation result prediction method, a learning model learning method, an operation result prediction device, and a learning model learning device (hereinafter, referred to as “learning device”) according to an embodiment of the present invention will be described with reference to the drawings.
- the operation result prediction device is a device that predicts the operation result of the industrial process from a plurality of operating conditions of the industrial process.
- an example will be described in which, in an arc process in a steel mill, the power consumption in batch operation of the arc process is predicted from a plurality of operating conditions.
- the arc process is a process that is performed subsequent to the primary refining process of molten steel, and is a process of adjusting the components such as desulfurization by heating the molten steel by arc discharge.
- the temperature is increased in consideration of the temperature variation of the molten steel received from the previous process, and the target component value differs depending on the steel type and the component adjustment differs, so the processing time required for these may differ.
- This arc process is a process that consumes electric power particularly in the steel making process, and the electric power consumption varies depending on the temperature rising process and the component adjustment process. Therefore, it is required to accurately predict the used electric power.
- a plurality of operating conditions when planning an arc process include, for example, a target temperature increase amount, a scheduled processing time, a processing waiting time, and steel grade information.
- the temperature increase target amount of these operating conditions there are cases where there is data and cases where there is no data depending on the operating conditions. That is, when the pre-process (primary refining process) is appropriately performed, the arc process plan can be drafted based on the result of the pre-process, so that it is possible to grasp the target heating amount. is there.
- the previous process is not properly performed, it is difficult to grasp the target temperature increase amount because the arc process cannot be planned based on the result of the previous process.
- the operation result prediction device 1 is realized by a general-purpose information processing device such as a personal computer or a workstation, and includes an input unit 10, a database (DB) 20, a calculation unit 30, and a display unit 40. I have it.
- DB database
- the input unit 10 is an input means for the arithmetic unit 30, and is realized by an input device such as a keyboard, a mouse pointer, or a ten-key pad.
- the database 20 stores past operation data (actual data) in the arc process.
- the arithmetic unit 30 is realized by, for example, a processor including a CPU (Central Processing Unit) and the like, and a memory (main storage unit) including a RAM (Random Access Memory) and a ROM (Read Only Memory).
- the arithmetic unit 30 realizes a function that matches a predetermined purpose by loading and executing a program in a work area of a main storage unit and controlling each component or the like through the execution of the program.
- the arithmetic unit 30 functions as a data creation unit (data creation unit) 31, a model creation unit (model creation unit) 32, a model selection unit (model selection unit) 33, and a prediction unit (prediction unit) 34 through the execution of the program. Function.
- the learning device according to the present embodiment is realized by the configuration of the operation result prediction device 1 excluding the model selection unit 33 and the prediction unit 34.
- the data creation unit 31 creates teacher data used in the model creation in the model creation unit 32 based on the past operation data stored in the database 20.
- the data creation unit 31 specifically creates two types of teacher data composed of first teacher data and second teacher data.
- the teacher data used for learning be created from operation data whose equipment conditions are as close as possible. Therefore, it is desirable that the data creation unit 31 creates the teacher data based on the operation data within a limited period (for example, the past two months).
- the first teaching data is teaching data that includes a particular key operating condition among a plurality of operating conditions of the arc process and is not classified by the operating condition related to the particular key operating condition. Is shown.
- the above-mentioned "operating condition that is a specific key” refers to the target temperature increase amount in the arc process.
- the first teacher data is, for example, as shown in Table 1 below, a plurality of operating conditions (a target temperature increase amount and a planned processing time), and electric power used when an arc process is performed under the plurality of operating conditions.
- the data is a combination of
- "steel number" in the same table indicates a processing number in the arc process.
- the second teaching data is teaching data that does not include a specific key operating condition among the plurality of operating conditions of the arc process, and is classified by operating conditions related to the specific key operating condition. Is shown.
- the above-mentioned "operating condition related to operating condition that is a specific key” indicates steel type information. Examples of the steel type of the steel type information include carbon steel, Al-killed steel, high-tensile steel, Si-killed steel, high-chrome steel, stainless steel and the like.
- the model creating unit 32 creates a first learning model by learning the first teacher data (see Table 1) created by the data creating unit 31.
- the model creating unit 32 also creates a second learning model by learning the second teacher data (see Table 2) created by the data creating unit 31.
- the model creation unit 32 uses regression analysis as a learning method. Further, as a method of regression analysis, for example, a least squares method which is a kind of linear regression, a partial least squares method or linear regression with regularization, or a kind of regression tree, random forest, gradient boosting, or nonlinear Neural networks, support vector regression, etc., which are a type of regression, can be used.
- a least squares method which is a kind of linear regression, a partial least squares method or linear regression with regularization, or a kind of regression tree, random forest, gradient boosting, or nonlinear Neural networks, support vector regression, etc., which are a type of regression, can be used.
- the model selection unit 33 determines whether or not to use a specific key operating condition (a target temperature increase amount) among a plurality of operating conditions as an explanatory variable when predicting the operating result (power consumption) of the arc process. Then, a specific learning model is selected from a plurality of learning models. That is, the model selection unit 33 selects the first learning model when using a specific key operating condition as an explanatory variable, and selects the second learning model when not using a specific key operating condition as an explanatory variable. Select the learning model of.
- a specific key operating condition a target temperature increase amount
- the prediction unit 34 predicts the operation result (power consumption) of the arc process based on the learning model selected by the model selection unit 33. Specifically, the prediction unit 34 calculates the predicted value of the power consumption by inputting the scheduled value of the explanatory variable (for example, the target temperature increase amount or the scheduled processing time).
- the scheduled value of the explanatory variable for example, the target temperature increase amount or the scheduled processing time.
- the learning method of the learning model according to this embodiment will be described with reference to FIGS. 2 and 3.
- the learning method of the learning model is implemented mainly by the data creation unit 31 and the model creation unit 32 of the calculation unit 30.
- the data creation unit 31 reads the operation data necessary for creating teacher data from the database 20 (step S1, data reading step). Then, the data creation unit 31 creates teacher data based on the read operation data (step S2, data creation step). Subsequently, the model creating unit 32 creates a learning model by learning the teacher data (step S3, model creating step).
- step S2 the above-described data creation step (step S2) is specifically carried out according to the flow shown in FIG.
- the data creation unit 31 determines whether or not the operation temperature data read from the database 20 includes the target temperature increase amount (step S21). When it is determined that the target temperature increase amount is included (Yes in step S21), the data creation unit 31 selects the target temperature increase amount and the scheduled processing time as explanatory variables (step S22), and As shown, teacher data (first teacher data, see Table 1) that is not classified by steel type information is created (step S23).
- the data creation unit 31 selects the scheduled processing time as the explanatory variable (step S24). ), teacher data (second teacher data, see Table 2) classified by the steel type information as shown in Table 2 is created (step S25). The presence/absence of the temperature increase target amount in step S21 is determined by a flag such as NaN.
- the operation result prediction method according to this embodiment will be described with reference to FIG.
- the operation result prediction method is implemented mainly by the model selection unit 33 and the prediction unit 34 of the calculation unit 30.
- the model selection unit 33 determines whether or not the temperature increase target value is included in the explanatory variables when predicting the operation result (power consumption) of the arc process (step S41, determination step). When it is determined that the temperature increase target amount is included (Yes in step S41), the model selection unit 33 selects a learning model (first learning model) that is not classified by the steel type information (step S42, model). (Selection step), the power consumption is predicted by inputting the temperature increase target amount and the scheduled processing time into the learning model (step S43, prediction step).
- step S41 when it is determined that the temperature increase target value is not included in the explanatory variables when predicting the operation result (power consumption) of the arc process (No in step S41), the model selection unit 33 is classified by the steel type information.
- step S44 model selecting step
- step S45 prediction step
- the prediction accuracy is high in accordance with the operation conditions (explanatory variables) having a correlation that can be grasped in advance. Since the learning model is selected and the prediction is performed, the operation result of the industrial process can be predicted with high accuracy. For example, when the operation result prediction device and the operation result prediction method are applied to an arc process in a steel mill, if the temperature increase target amount, which is a key operation condition, cannot be used as an explanatory variable, teacher data classified by steel type information ( A learning model (second learning model) created based on the second teacher data) is selected, and the power consumption is predicted. As a result, it is possible to highly accurately predict the power used in the arc process.
- a plurality of learning models are created according to operating conditions that can be grasped in advance.
- the operation result of the industrial process can be predicted with high accuracy according to the operating condition that can be grasped in advance.
- the configurations of the operation result prediction device and the learning device according to the second embodiment of the present invention are the same as those in FIG.
- the operation result prediction device is a device that predicts the operation result of the industrial process from a plurality of operating conditions of the industrial process.
- an example will be described in which a generated gas in a batch operation of a converter process is predicted from a plurality of operating conditions in a converter process in a steel mill.
- the converter process is a primary refining process in which iron ore is melted in a blast furnace and hot metal that has been subjected to pretreatment such as desulfurization is transferred to molten steel. In this process, oxygen is blown to perform decarburization and dephosphorization refining. Even with the same treatment time, decarburization and dephosphorization produce larger amounts of gas during decarburization. Even with the same decarburization, the amount of gas generated differs because the treatment differs depending on the blowing mode.
- the converter process is a process in which gas is generated particularly in the steelmaking process, and in order to efficiently generate power using this generated gas, it is required to accurately predict the generated gas. For that purpose, it is necessary to consider the processing contents as operating conditions.
- a plurality of operating conditions when planning a converter process include, for example, a planned blowing oxygen amount, a planned treatment time, and a blowing form.
- a planned blowing oxygen amount in these operating conditions, there are cases where there is data and cases where there is no data depending on the operating conditions. That is, the blowing oxygen amount is calculated and determined immediately before the converter blowing and during the converter blowing. They are called static controls and dynamic controls. Therefore, it is difficult to grasp the blowing oxygen amount of the next blowing until the previous blowing of the converter is completed.
- the first teacher data used in the data creation unit 31 includes an operating condition that is a specific key among a plurality of operating conditions of the converter process, and is classified by an operating condition that is related to the operating condition that is the specific key. It indicates that the teacher data has not been created.
- the above-mentioned "operating condition which is a specific key” indicates the planned blowing oxygen amount in the converter process.
- the first teacher data is, for example, as shown in Table 3 below, a plurality of operating conditions (scheduled blowing oxygen amount and scheduled processing time), and gas generated when the converter process is carried out under the plurality of operating conditions.
- the data is a combination of and.
- "steel number" in the table indicates a processing number in the converter process.
- the second teacher data used by the data creation unit 31 is a plurality of operating conditions of the converter process, does not include an operating condition that is a specific key, and is based on an operating condition that is related to the operating condition that is the specific key. This indicates the classified teacher data.
- the above-mentioned “operating condition related to the operating condition that is a specific key” indicates the blowing mode in the present embodiment. This blowing mode is assigned according to the content of the hot metal pretreatment.
- the second teacher data is, for example, as shown in Table 4 below, the combination of the operating condition (scheduled processing time) and the gas generated when the converter process is carried out under the operating condition for each blowing mode. It is classified data.
- the model creating unit 32 creates a first learning model by learning the first teacher data (see Table 3) created by the data creating unit 31.
- the model creating unit 32 also creates a second learning model by learning the second teacher data (see Table 4) created by the data creating unit 31.
- the model creation unit 32 uses regression analysis as a learning method. Further, as a method of regression analysis, for example, a least squares method which is a kind of linear regression, a partial least squares method or linear regression with regularization, or a kind of regression tree, random forest, gradient boosting, or nonlinear Neural networks, support vector regression, etc., which are a type of regression, can be used.
- a least squares method which is a kind of linear regression, a partial least squares method or linear regression with regularization, or a kind of regression tree, random forest, gradient boosting, or nonlinear Neural networks, support vector regression, etc., which are a type of regression, can be used.
- a specific learning model is selected from a plurality of learning models. That is, the model selection unit 33 selects the first learning model when using a specific key operating condition as an explanatory variable, and selects the second learning model when not using a specific key operating condition as an explanatory variable. Select the learning model of.
- the prediction unit 34 predicts the operation result (generated gas) of the converter process based on the learning model selected by the model selection unit 33. Specifically, the prediction unit 34 calculates the predicted value of the generated gas by inputting the scheduled value of the explanatory variable (for example, the scheduled blowing oxygen amount or the scheduled processing time).
- the learning method of the learning model according to this embodiment will be described with reference to FIGS. 2 and 5.
- the learning method of the learning model is implemented mainly by the data creation unit 31 and the model creation unit 32 of the calculation unit 30.
- As a learning method of the learning model the processes of steps S1 to S3 of FIG. 2 are performed.
- step S2 the above-described data creation step (step S2) is specifically carried out according to the flow shown in FIG.
- the data creation unit 31 determines whether or not the planned blowing oxygen amount is included in the operation data read from the database 20 (step S51).
- the data creation unit 31 selects the planned blowing oxygen amount and the planned processing time as explanatory variables (step S52), and the above table is displayed.
- teacher data first teacher data, see Table 3 that is not classified in the blowing mode is created (step S53).
- the data creation unit 31 selects the planned processing time as the explanatory variable (step S51).
- the teacher data second teacher data, see Table 4 classified in the blowing mode is created as shown in Table 4 (step S55).
- the presence/absence of the planned blowing oxygen amount in step S51 is determined by a flag such as NaN.
- the operation result prediction method according to this embodiment will be described with reference to FIG.
- the operation result prediction method is implemented mainly by the model selection unit 33 and the prediction unit 34 of the calculation unit 30.
- the model selection unit 33 determines whether or not the expected blowing oxygen amount is included in the explanatory variables when predicting the operation result (generated gas) of the converter process (step S61, determination step). When it is determined that the planned blowing oxygen amount is included (Yes in step S61), the model selection unit 33 selects a learning model (first learning model) that is not classified in the blowing mode (step S62). , Model selection step), the planned blowing oxygen amount and the planned processing time are input to the learning model to predict the generated gas (step S63, prediction step).
- the model selection unit 33 causes the blowing mode.
- the generated gas is predicted by selecting the learning model (second learning model) classified in (step S64, model selecting step) and inputting the scheduled processing time into the learning model (step S65, predicting step). ..
- the prediction accuracy is high in accordance with the operation conditions (explanatory variables) having a correlation that can be grasped in advance. Since the learning model is selected and the prediction is performed, the operation result of the industrial process can be predicted with high accuracy. For example, when the operation result prediction device and the operation result prediction method are applied to a converter process in a steel mill, if the planned blowing oxygen amount, which is a key operating condition, cannot be used as an explanatory variable, teachers classified by blowing form A learning model (second learning model) created based on the data (second teacher data) is selected, and the generated gas is predicted. Thereby, the gas generated in the converter process can be predicted with high accuracy.
- the planned blowing oxygen amount which is a key operating condition
- a plurality of learning models are created according to operating conditions that can be grasped in advance.
- the operation result of the industrial process can be predicted with high accuracy according to the operating condition that can be grasped in advance.
- a learning model in which teacher data that does not include the temperature increase target amount as an explanatory variable and is not classified by the steel type information is learned
- each learning model was created based on the operation data for the past two months.
- Lasso regression which is a type of linear regression with regularization
- the target variable was the amount of electric power used for each steel number
- the explanatory variable candidates were the temperature increase target amount for each steel number, the planned processing time, the processing waiting time, and the steel type information.
- the steel type information is assigned from among carbon steel, Al killed steel, high strength steel, Si killed steel, high chromium steel, and stainless steel.
- an ideal situation where the predicted value is 100% is assumed, and the actual value is used for each variable instead of the planned value.
- Fig. 7 shows the comparison results of the errors obtained in the accuracy verification.
- the error as the accuracy index was calculated by using RMSE (Root Mean Square Error) as 100 when only the planned processing time was used and the target heating amount and steel type information were not used.
- RMSE Root Mean Square Error
- the prediction result of the learning model C including the temperature raising target value in the explanatory variable is the best. Since it is clear that the temperature increase target value has a high correlation with the electric power used, this is a natural result. On the other hand, in the case where the temperature rise target value cannot be used, the prediction result of the learning model A that is not classified by steel type has a prediction error accuracy of about 30% worse than the prediction result of the learning model C. Has become. On the other hand, when the temperature raising target value cannot be used, the learning model B in which the data is classified by the steel type information that is considered to be closely related to the temperature raising target value and learning is performed without using the temperature raising target value. In the prediction result of, the accuracy of the prediction error is improved by 10% or more with respect to the prediction result of the learning model A.
- the learning model in which the target temperature increase amount is simply omitted is used. If the prediction error with different learning models classified by explanatory variables (here, steel type information) that are assumed to be related to the target temperature rise is evaluated, and the prediction error is expected to improve for this, by using the learning model classified by the steel type information, the prediction error of the learning model can be improved.
- explanatory variable here, the target temperature increase amount
- steel type information the learning model classified by the steel type information
- D A learning model in which teacher data that does not include the planned blowing oxygen amount as an explanatory variable and is not classified by the blowing form is learned
- E Does not include the planned blowing oxygen amount as an explanatory variable and depends on the blowing form
- F A learning model (first learning model) in which the teacher data (first teacher data) that includes the planned blowing oxygen amount as an explanatory variable and is not classified by the blowing mode is learned.
- each learning model was created based on the operation data of the past 300 blowing.
- Lasso regression which is a type of linear regression with regularization
- the objective variable is the generated gas amount of one blowing
- the candidates of the explanatory variables are the planned blowing oxygen amount of one blowing, the scheduled processing time, and the blowing mode.
- the blowing form was assigned according to the content of the hot metal pretreatment. Further, in the accuracy verification of each learning model, an ideal situation where the predicted value is 100% is assumed, and the actual value is used for each variable instead of the planned value.
- Fig. 8 shows the comparison results of the errors obtained in the accuracy verification.
- the error used as the accuracy index was calculated by using only the planned treatment time and RMSE when the planned blowing oxygen amount and the blowing mode were not used as 100.
- the prediction result of the learning model F including the planned blowing oxygen amount in the explanatory variable is the best. It is clear that the planned amount of blown oxygen has a high correlation with the generated gas, which is a natural result. On the other hand, in the case where the planned blowing oxygen amount cannot be used, the prediction result of the learning model D that is not classified by the blowing pattern has a prediction error accuracy of 19% with respect to the prediction result of the learning model F. It is a bad result. On the other hand, when the planned blowing oxygen amount cannot be used, the data is categorized by the blowing pattern that is considered to be closely related to the planned blowing oxygen amount, and the learning is performed without using the planned blowing oxygen amount. In the prediction result of the learning model E, the accuracy of the prediction error is improved by about 12% with respect to the prediction result of the learning model D.
- the explanatory variable here, the planned blowing oxygen amount
- the learning model in which the planned blowing oxygen amount is simply omitted is used. Evaluate the prediction error with different learning models classified by the explanatory variable (here, blowing pattern) that is assumed to be related to the planned blowing oxygen amount, rather than using it, and if the prediction error is improved If it is assumed, it is possible to improve the prediction error of the learning model by using the learning model classified in the blowing mode.
- the operation result predicting method, the learning model learning method, the operation result predicting apparatus, and the learning model learning apparatus according to the present invention have been specifically described above with reference to modes and embodiments for carrying out the invention.
- the gist is not limited to these descriptions, and should be broadly construed based on the claims. Further, it goes without saying that various changes and modifications based on these descriptions are also included in the gist of the present invention.
- Operation result prediction device 10
- Input unit 20
- Database (DB) 30 arithmetic unit 31
- data creation unit data creation means
- model creation unit model creation means
- model selection unit model selection means
- Predictor Predictor
Landscapes
- Engineering & Computer Science (AREA)
- General Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Economics (AREA)
- General Physics & Mathematics (AREA)
- Strategic Management (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- Human Resources & Organizations (AREA)
- Mathematical Physics (AREA)
- Operations Research (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Development Economics (AREA)
- Artificial Intelligence (AREA)
- Game Theory and Decision Science (AREA)
- Computing Systems (AREA)
- Medical Informatics (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Evolutionary Computation (AREA)
- Quality & Reliability (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Feedback Control In General (AREA)
- Carbon Steel Or Casting Steel Manufacturing (AREA)
- Waste-Gas Treatment And Other Accessory Devices For Furnaces (AREA)
- Vertical, Hearth, Or Arc Furnaces (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Testing And Monitoring For Control Systems (AREA)
Abstract
Description
(操業結果予測装置)
本発明の実施形態1に係る操業結果予測装置および学習装置の構成について、図1を参照しながら説明する。操業結果予測装置は、工業プロセスの複数の操業条件から当該工業プロセスの操業結果を予測する装置である。本実施形態では、製鉄所内のアークプロセスにおいて、複数の操業条件から、アークプロセスのバッチ操業における使用電力を予測する場合の例について説明する。
本実施形態に係る学習モデルの学習方法について、図2および図3を参照しながら説明する。なお、学習モデルの学習方法は、演算部30のデータ作成部31およびモデル作成部32が主体となって実施される。
本実施形態に係る操業結果予測方法について、図4を参照しながら説明する。なお、操業結果予測方法は、演算部30のモデル選択部33および予測部34が主体となって実施される。
(操業結果予測装置)
本発明の実施形態2に係る操業結果予測装置および学習装置の構成は、図1と同様である。操業結果予測装置は、工業プロセスの複数の操業条件から当該工業プロセスの操業結果を予測する装置である。本実施形態では、製鉄所内の転炉プロセスにおいて、複数の操業条件から、転炉プロセスのバッチ操業における発生ガスを予測する場合の例について説明する。
本実施形態に係る学習モデルの学習方法について、図2および図5を参照しながら説明する。なお、学習モデルの学習方法は、演算部30のデータ作成部31およびモデル作成部32が主体となって実施される。学習モデルの学習方法は、図2のステップS1~S3の処理を行う。
本実施形態に係る操業結果予測方法について、図6を参照しながら説明する。なお、操業結果予測方法は、演算部30のモデル選択部33および予測部34が主体となって実施される。
B:昇温目標量を説明変数として含まず、かつ鋼種情報によって分類された教師データ(第二の教師データ)を学習させた学習モデル(第二の学習モデル)
C:昇温目標量を説明変数として含み、かつ鋼種情報によって分類されていない教師データ(第一の教師データ)を学習させた学習モデル(第一の学習モデル)
E:予定吹錬酸素量を説明変数として含まず、かつ吹錬形態によって分類された教師データ(第二の教師データ)を学習させた学習モデル(第二の学習モデル)
F:予定吹錬酸素量を説明変数として含み、かつ吹錬形態によって分類されていない教師データ(第一の教師データ)を学習させた学習モデル(第一の学習モデル)
10 入力部
20 データベース(DB)
30 演算部
31 データ作成部(データ作成手段)
32 モデル作成部(モデル作成手段)
33 モデル選択部(モデル選択手段)
34 予測部(予測手段)
40 表示部
Claims (9)
- 工業プロセスの複数の操業条件から前記工業プロセスの操業結果を予測する操業結果予測方法であって、
前記操業結果を予測する際の説明変数として、前記複数の操業条件のうち特定のキーとなる操業条件を用いるか否かに応じて、複数の学習モデルの中から特定の学習モデルを選択するモデル選択ステップと、
前記モデル選択ステップで選択した学習モデルに基づいて、操業結果を予測する予測ステップと、
を含む操業結果予測方法。 - 前記複数の学習モデルは、
前記特定のキーとなる操業条件を含み、かつ前記特定のキーとなる操業条件に関連した操業条件によって分類されていない第一の教師データから作成された第一の学習モデルと、
前記特定のキーとなる操業条件を含まず、かつ前記特定のキーとなる操業条件に関連した操業条件によって分類された第二の教師データを作成する第二の学習モデルと、
を含み、
前記モデル選択ステップは、
前記説明変数として前記特定のキーとなる操業条件を用いる場合は前記第一の学習モデルを選択し、
前記説明変数として前記特定のキーとなる操業条件を用いない場合は前記第二の学習モデルを選択する、
請求項1に記載の操業結果予測方法。 - 前記工業プロセスは、製鉄所内のアークプロセスであり、
前記操業結果は、前記アークプロセスのバッチ操業における使用電力であり、
前記複数の操業条件は、少なくとも昇温目標量、予定処理時間および鋼種情報を含む、
請求項1または請求項2に記載の操業結果予測方法。 - 前記工業プロセスは、製鉄所内の転炉プロセスであり、
前記操業結果は、前記転炉プロセスのバッチ操業における発生ガスであり、
前記複数の操業条件は、少なくとも予定吹錬酸素量、予定処理時間および吹錬形態を含む、
請求項1または請求項2に記載の操業結果予測方法。 - 工業プロセスの複数の操業条件から前記工業プロセスの操業結果を予測する際に用いる学習モデルの学習方法であって、
前記複数の操業条件のうち特定のキーとなる操業条件を含み、かつ前記特定のキーとなる操業条件に関連した操業条件によって分類されていない第一の教師データを作成する第一のデータ作成ステップと、
前記特定のキーとなる操業条件を含まず、かつ前記特定のキーとなる操業条件に関連した操業条件によって分類された第二の教師データを作成する第二のデータ作成ステップと、
前記第一の教師データに基づいて、前記工業プロセスの操業結果を予測する第一の学習モデルを作成する第一のモデル作成ステップと、
前記第二の教師データに基づいて、前記工業プロセスの操業結果を予測する第二の学習モデルを作成する第二のモデル作成ステップと、
を含む学習モデルの学習方法。 - 前記工業プロセスは、製鉄所内のアークプロセスであり、
前記操業結果は、前記アークプロセスのバッチ操業における使用電力であり、
前記複数の操業条件は、少なくとも昇温目標量、予定処理時間および鋼種情報を含む、
請求項5に記載の学習モデルの学習方法。 - 前記工業プロセスは、製鉄所内の転炉プロセスであり、
前記操業結果は、前記転炉プロセスのバッチ操業における発生ガスであり、
前記複数の操業条件は、少なくとも予定吹錬酸素量、予定処理時間および吹錬形態を含む、
請求項5に記載の学習モデルの学習方法。 - 工業プロセスの複数の操業条件から前記工業プロセスの操業結果を予測する操業結果予測装置であって、
前記操業結果を予測する際の説明変数として、前記複数の操業条件のうち特定のキーとなる操業条件を用いるか否かに応じて、複数の学習モデルの中から特定の学習モデルを選択するモデル選択手段と、
前記モデル選択手段で選択した学習モデルに基づいて、操業結果を予測する予測手段と、
を備える操業結果予測装置。 - 工業プロセスの複数の操業条件から前記工業プロセスの操業結果を予測する際に用いる学習モデルの学習装置であって、
前記複数の操業条件のうち特定のキーとなる操業条件を含み、かつ前記特定のキーとなる操業条件に関連した操業条件によって分類されていない第一の教師データを作成する第一のデータ作成手段と、
前記特定のキーとなる操業条件を含まず、かつ前記特定のキーとなる操業条件に関連した操業条件によって分類された第二の教師データを作成する第二のデータ作成手段と、
前記第一の教師データに基づいて、前記工業プロセスの操業結果を予測する第一の学習モデルを作成する第一のモデル作成手段と、
前記第二の教師データに基づいて、前記工業プロセスの操業結果を予測する第二の学習モデルを作成する第二のモデル作成手段と、
を備える学習モデルの学習装置。
Priority Applications (5)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
KR1020217026129A KR102579633B1 (ko) | 2019-02-19 | 2020-02-07 | 조업 결과 예측 방법, 학습 모델의 학습 방법, 조업 결과 예측 장치 및 학습 모델의 학습 장치 |
BR112021016447A BR112021016447A2 (pt) | 2019-02-19 | 2020-02-07 | Método de previsão de resultado de operação, método de treinamento de modelo de aprendizagem, dispositivo de previsão de resultado de operação e dispositivo de treinamento para modelo de aprendizagem |
JP2020546514A JP6954479B2 (ja) | 2019-02-19 | 2020-02-07 | 操業結果予測方法、学習モデルの学習方法、操業結果予測装置および学習モデルの学習装置 |
EP20759314.6A EP3929516A4 (en) | 2019-02-19 | 2020-02-07 | METHOD FOR PREDICTING OPERATING RESULTS, METHOD FOR LEARNING A LEARNING MODEL, DEVICE FOR PREDICTING OPERATING RESULTS, AND DEVICE FOR LEARNING A LEARNING MODEL |
CN202080014961.1A CN113454413B (zh) | 2019-02-19 | 2020-02-07 | 操作结果预测方法、学习模型的学习方法、操作结果预测装置及学习模型的学习装置 |
Applications Claiming Priority (2)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
JP2019-027509 | 2019-02-19 | ||
JP2019027509 | 2019-02-19 |
Publications (1)
Publication Number | Publication Date |
---|---|
WO2020170849A1 true WO2020170849A1 (ja) | 2020-08-27 |
Family
ID=72143889
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
PCT/JP2020/004738 WO2020170849A1 (ja) | 2019-02-19 | 2020-02-07 | 操業結果予測方法、学習モデルの学習方法、操業結果予測装置および学習モデルの学習装置 |
Country Status (6)
Country | Link |
---|---|
EP (1) | EP3929516A4 (ja) |
JP (1) | JP6954479B2 (ja) |
KR (1) | KR102579633B1 (ja) |
CN (1) | CN113454413B (ja) |
BR (1) | BR112021016447A2 (ja) |
WO (1) | WO2020170849A1 (ja) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2022059753A1 (ja) * | 2020-09-18 | 2022-03-24 | 株式会社Uacj | 溶解炉のエネルギー効率を予測する学習済み予測モデルを生成する方法、溶解炉のエネルギー効率を予測する方法、およびコンピュータプログラム |
Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH06307766A (ja) | 1993-04-21 | 1994-11-01 | Nippon Steel Corp | 製鋼用アーク炉の電力投入制御方法 |
JP2011256407A (ja) | 2010-06-04 | 2011-12-22 | Nippon Steel Engineering Co Ltd | 製鋼用アーク炉の電力投入制御方法 |
JP2012004181A (ja) * | 2010-06-14 | 2012-01-05 | Sharp Corp | 特性予測装置、特性予測方法、特性予測プログラムおよびプログラム記録媒体 |
JP2018045559A (ja) * | 2016-09-16 | 2018-03-22 | 富士通株式会社 | 情報処理装置、情報処理方法およびプログラム |
JP2018147280A (ja) * | 2017-03-07 | 2018-09-20 | 株式会社日立ソリューションズ | データ分析装置及びデータ分析方法 |
Family Cites Families (12)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPS5161408A (en) * | 1974-11-26 | 1976-05-28 | Nippon Kokan Kk | Seikoyoaakurono yorakuhanteiho |
JPS637766A (ja) * | 1986-06-26 | 1988-01-13 | Toyohiko Nishimura | 焼きそば等の麺類寿司及びその製造方法 |
JP3645306B2 (ja) * | 1995-03-31 | 2005-05-11 | 日新製鋼株式会社 | 電気炉設備 |
CN102373310B (zh) * | 2011-10-19 | 2013-06-12 | 北京科技大学 | 一种指导转炉补吹过程操作的方法 |
DE102012224184A1 (de) * | 2012-12-21 | 2014-06-26 | Sms Siemag Ag | Verfahren zur Vorhersage, Steuerung und/oder Regelung von Stahlwerksprozessen |
CN103882176B (zh) * | 2014-03-25 | 2015-09-30 | 东北大学 | 一种基于数据驱动的转炉炼钢过程在线动态控制方法 |
US20180181875A1 (en) * | 2014-03-28 | 2018-06-28 | Nec Corporation | Model selection system, model selection method, and storage medium on which program is stored |
JP6856023B2 (ja) * | 2015-09-30 | 2021-04-07 | 日本電気株式会社 | 最適化システム、最適化方法および最適化プログラム |
CN105807741B (zh) * | 2016-03-09 | 2018-08-07 | 北京科技大学 | 一种工业生产流程预测方法 |
CN105714014A (zh) * | 2016-03-30 | 2016-06-29 | 本钢板材股份有限公司 | 转炉氧枪/加料/温度制度综合简化模型系统及操作方法 |
JP6573035B2 (ja) * | 2016-07-14 | 2019-09-11 | 日本製鉄株式会社 | 溶鋼中りん濃度推定方法及び転炉吹錬制御装置 |
JP6897260B2 (ja) * | 2017-04-14 | 2021-06-30 | 日本製鉄株式会社 | 溶鋼中りん濃度推定方法、転炉吹錬制御装置、プログラム及び記録媒体 |
-
2020
- 2020-02-07 WO PCT/JP2020/004738 patent/WO2020170849A1/ja unknown
- 2020-02-07 BR BR112021016447A patent/BR112021016447A2/pt unknown
- 2020-02-07 KR KR1020217026129A patent/KR102579633B1/ko active IP Right Grant
- 2020-02-07 EP EP20759314.6A patent/EP3929516A4/en active Pending
- 2020-02-07 JP JP2020546514A patent/JP6954479B2/ja active Active
- 2020-02-07 CN CN202080014961.1A patent/CN113454413B/zh active Active
Patent Citations (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
JPH06307766A (ja) | 1993-04-21 | 1994-11-01 | Nippon Steel Corp | 製鋼用アーク炉の電力投入制御方法 |
JP2011256407A (ja) | 2010-06-04 | 2011-12-22 | Nippon Steel Engineering Co Ltd | 製鋼用アーク炉の電力投入制御方法 |
JP2012004181A (ja) * | 2010-06-14 | 2012-01-05 | Sharp Corp | 特性予測装置、特性予測方法、特性予測プログラムおよびプログラム記録媒体 |
JP2018045559A (ja) * | 2016-09-16 | 2018-03-22 | 富士通株式会社 | 情報処理装置、情報処理方法およびプログラム |
JP2018147280A (ja) * | 2017-03-07 | 2018-09-20 | 株式会社日立ソリューションズ | データ分析装置及びデータ分析方法 |
Cited By (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2022059753A1 (ja) * | 2020-09-18 | 2022-03-24 | 株式会社Uacj | 溶解炉のエネルギー効率を予測する学習済み予測モデルを生成する方法、溶解炉のエネルギー効率を予測する方法、およびコンピュータプログラム |
JP7519453B2 (ja) | 2020-09-18 | 2024-07-19 | 株式会社Uacj | 溶解炉のエネルギー効率を予測する学習済み予測モデルを生成する方法、溶解炉のエネルギー効率を予測する方法、およびコンピュータプログラム |
Also Published As
Publication number | Publication date |
---|---|
BR112021016447A2 (pt) | 2021-11-09 |
CN113454413A (zh) | 2021-09-28 |
JPWO2020170849A1 (ja) | 2021-03-11 |
EP3929516A4 (en) | 2022-04-20 |
CN113454413B (zh) | 2023-06-27 |
KR102579633B1 (ko) | 2023-09-15 |
JP6954479B2 (ja) | 2021-10-27 |
KR20210116578A (ko) | 2021-09-27 |
EP3929516A1 (en) | 2021-12-29 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Liu et al. | A dynamic analytics method based on multistage modeling for a BOF steelmaking process | |
CN109447346B (zh) | 基于灰色预测与神经网络组合模型转炉氧耗量预测方法 | |
Matino et al. | Forecasting blast furnace gas production and demand through echo state neural network-based models: Pave the way to off-gas optimized management | |
WO2020152750A1 (ja) | 金属材料の設計支援方法、予測モデルの生成方法、金属材料の製造方法、及び設計支援装置 | |
Feng et al. | End temperature prediction of molten steel in LF based on CBR–BBN | |
Matino et al. | Quantification of energy and environmental impacts in uncommon electric steelmaking scenarios to improve process sustainability | |
CN109754137B (zh) | 确定钢铁材料产品的质量和/或等级的方法及设备 | |
JP4093934B2 (ja) | モデルパラメータ決定方法及びそのプログラム,モデル予測方法及びそのプログラム | |
WO2020170849A1 (ja) | 操業結果予測方法、学習モデルの学習方法、操業結果予測装置および学習モデルの学習装置 | |
CN111020118B (zh) | 一种基于粒子群优化案例推理的rh终点温度预测方法 | |
Liu et al. | A novel dynamic operation optimization method based on multiobjective deep reinforcement learning for steelmaking process | |
Zhang | Optimal control problem of converter steelmaking production process based on operation optimization method | |
Spirin et al. | Complex of model systems for supporting decisions made in managing blast-furnace smelting technology | |
Ruuska et al. | Mass-balance based multivariate modelling of basic oxygen furnace used in steel industry | |
Howard et al. | Identifying Best Practice Melting Patterns in Induction Furnaces: A Data-Driven Approach Using Time Series K-Means Clustering and Multi-Criteria Decision Making | |
Singh et al. | Correlation and Prediction of Molten Steel Temperature in Steel Melting Shop Using Reliable Machine Learning (RML) Approach | |
Sorsa et al. | Data-driven multivariate analysis of basic oxygen furnace used in steel industry | |
JP2012167365A (ja) | 生石灰濃度予測装置及び吹錬制御方法 | |
JP2008007828A (ja) | 脱燐制御方法および装置 | |
JP2004059954A (ja) | 転炉吹錬制御方法 | |
Tsai et al. | The exponential grey forecasting model for CO2 emissions in Taiwan | |
JP7485241B1 (ja) | エネルギー運用支援システム、情報処理装置およびエネルギー運用支援方法 | |
JP2004059955A (ja) | 転炉吹錬制御方法 | |
Bigeev et al. | Adapting a mathematical model of the end of the blow of a converter heat to existing conditions in the oxygen-converter shop at the Magnitogorsk Metallurgical Combine | |
Saxena | Energy Management and Monitoring Systems |
Legal Events
Date | Code | Title | Description |
---|---|---|---|
ENP | Entry into the national phase |
Ref document number: 2020546514 Country of ref document: JP Kind code of ref document: A |
|
121 | Ep: the epo has been informed by wipo that ep was designated in this application |
Ref document number: 20759314 Country of ref document: EP Kind code of ref document: A1 |
|
ENP | Entry into the national phase |
Ref document number: 20217026129 Country of ref document: KR Kind code of ref document: A |
|
NENP | Non-entry into the national phase |
Ref country code: DE |
|
REG | Reference to national code |
Ref country code: BR Ref legal event code: B01A Ref document number: 112021016447 Country of ref document: BR |
|
ENP | Entry into the national phase |
Ref document number: 2020759314 Country of ref document: EP Effective date: 20210920 |
|
ENP | Entry into the national phase |
Ref document number: 112021016447 Country of ref document: BR Kind code of ref document: A2 Effective date: 20210819 |