Blast volume control system and blast volume determination program

Abstract

To realize a tapping amount extremely close to a target tapping amount by controlling a blast volume.SOLUTION: A blast volume control system comprises: model storage means (43) for storing a learned model (M) generated by mechanically learning the influence of the increase / decrease of a tapping amount relative to the increase / decrease of a blast volume; blast volume calculation means (41) for calculating an optimum blast volume for realizing a target tapping amount based on operation data obtained during the continuous operation of a cupola; and control means (32) for controlling a blast machine so that the blast volume reaches the optimum blast volume. The blast volume calculation means includes: prediction processing means (44) for inputting plural blast volume candidates one by one to the learnt model together with condition data including a blast volume and a tapping amount in an immediate time, thus calculating a tapping predicted amount per blast volume candidate; and determination means (46) for extracting a tapping predicted amount close to the target tapping amount from the plural tapping predicted amounts, and specifying a corresponding blast volume candidate to determine the optimum blast volume.SELECTED DRAWING: Figure 4

Description

[Technical Field]

[0001] The present invention relates to an air blast rate control system and an air blast rate determination program, and more particularly to an air blast rate control system and an air blast rate determination program for controlling the rate of hot air sent to a cupola. [Background technology]

[0002] Cupolas (melting furnaces) are used to melt iron sources such as scrap iron and pig iron. In cupolas, tuyere holes are installed at the bottom of the furnace wall facing the bed coke layer, which is formed by packing coke up to a certain height from the bottom of the furnace. Hot air (air) is blown in through the holes to combust the coke. This combustion heat melts the iron sources, such as scrap iron and pig iron, charged above the bed coke layer, and the molten iron is extracted from a tap hole installed at the bottom of the furnace.

[0003] For example, as disclosed in Japanese Patent Laid-Open Publication No. 9-125123 (Patent Document 1), in order to make the melting capacity variable in cupola operation, it is common to increase or decrease the coke ratio (the weight ratio of added coke to charged base metal) or to increase or decrease the amount of air sent from the tuyere via an air box.

[0004] The document discloses that when various conditions, such as melting capacity, coke ratio, and tapping temperature, which are varied according to the properties of the metal to be melted, are input into a digital indicating controller, a programmable controller calculates the amount of blast air corresponding to the melting capacity of the molten metal based on these data, and a volume of air corresponding to this calculation is supplied to the wind box from the blower pipe, while at the same time, an appropriate amount of pure oxygen to be enriched in the air is calculated, and an appropriate amount of pure oxygen within the range of 1% to 20% is supplied to the blower pipe from an oxygen tank, and the oxygen concentration is varied and supplied to the wind box as blast air.The document also discloses that the amount of molten metal tapped from the siphon section of the furnace body is measured, and the difference from the set melting capacity is calculated, and if the amount of molten metal tapped is insufficient, the amount of oxygen enrichment is increased to satisfy the set conditions. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] Japanese Patent Application Publication No. 9-125123 Summary of the Invention [Problem to be solved by the invention]

[0006] Increasing the amount of air blown to the cupola promotes combustion in the furnace and increases the amount of molten metal tapped, but it is difficult to achieve the target amount of molten metal tapped simply by mechanically managing the blower.

[0007] In addition, in Patent Document 1, the amount of air blown is kept constant based on the planned amount of hot water used, and the amount of oxygen enrichment is increased or decreased to adjust the amount of hot water discharged, but there is also the problem that using oxygen enrichment leads to increased costs.

[0008] The present invention has been made to solve the above-mentioned problems, and its purpose is to provide a blast volume control system that can achieve an iron production rate that is as close as possible to the target iron production rate by controlling the blast volume (only).

[0009] Another object of the present invention is to provide a blast rate determination program capable of calculating an optimum blast rate that can achieve a pig iron production rate as close as possible to a target pig iron production rate. [Means for solving the problem]

[0010] A blast rate control system according to one aspect of the present invention controls the rate at which hot air is blown to a cupola, and includes: model storage means for storing a "trained model" generated by machine learning of the effect of increases or decreases in blast rate on increases or decreases in pig iron production rate using operational history data including the blast rate to the cupola and the pig iron production rate from the cupola; blast rate calculation means for calculating an optimal blast rate for achieving a target pig iron production rate at regular intervals based on operational data obtained during continuous cupola operation; and control means for controlling a blower so that the blast rate to the cupola is the optimal blast rate calculated by the blast rate calculation means. The "regular interval" corresponds, for example, to the monitoring interval of the cupola's operating status.

[0011] The blast volume calculation means includes a prediction processing means that calculates the predicted iron production amount for each blast volume candidate by inputting each of the multiple blast volume candidates into a trained model together with condition data including the blast volume and iron production amount at the immediately preceding first time (one or more monitoring times), and a determination means that extracts a predicted iron production amount that is closest to the target iron production amount from the multiple predicted iron production amounts obtained by the prediction processing means, and identifies the blast volume candidate that corresponds to the extracted predicted iron production amount, thereby determining the optimal blast volume for this time.

[0012] Preferably, the blast volume control system further includes an acquisition means for acquiring the amount of molten metal held in the holding furnace located downstream of the cupola, and a calculation means for calculating a target iron production volume based on the amount of molten metal held by the acquisition means and a predetermined planned amount of molten metal to be used.

[0013] The calculation means preferably calculates the target tapping rate so that the amount of molten metal held in the holding furnace ultimately approaches a predetermined target amount of stored molten metal within a preset time range. The "time range" is set, for example, to a range of 1 hour to 3 hours.

[0014] Preferably, the condition data further includes the amount of oxygen enrichment at the first time and the current time, and the condition data further includes the amount of material charged into the cupola at a second time before the first time.

[0015] A blast rate determination program according to another aspect of the present invention is a program for determining the optimal blast rate of hot air to a cupola, and causes a computer to execute the following determination means: a step of extracting condition data including the blast rate and iron production rate for one or more immediately preceding monitoring periods from operating data obtained during continuous operation of the cupola; a step of calculating a predicted iron production rate for each blast rate candidate by inputting multiple blast rate candidates together with the condition data into a trained model generated by machine learning the effect of increases and decreases in iron production rate on increases and decreases in blast rate; and a step of determining the current optimal blast rate by extracting a predicted iron production rate closest to the target iron production rate from the multiple predicted iron production rates and identifying a blast rate candidate corresponding to the extracted predicted iron production rate. [Effects of the Invention]

[0016] According to the present invention, a "trained model" generated by machine learning the effect of increases or decreases in blast rate on increases or decreases in iron production rate using historical operating data including the blast rate to the cupola and the iron production rate from the cupola can be used to calculate an optimal blast rate for achieving a target iron production rate. Therefore, by controlling only the blast rate, it is possible to achieve an iron production rate that is as close as possible to the target iron production rate. [Brief explanation of the drawings]

[0017] [Figure 1] FIG. 1 is a diagram illustrating a manufacturing process of ductile cast iron. [Figure 2] 1 is a schematic diagram showing a general configuration of an air flow control system according to an embodiment of the present invention; [Figure 3] 1 is a conceptual diagram schematically illustrating a method for determining an air flow rate by an air flow rate determination device according to an embodiment of the present invention. [Figure 4] FIG. 1A is a block diagram showing the functional configuration of an airflow volume determination device according to an embodiment of the present invention, and FIG. 1B is a block diagram showing the functional configuration of a learning device according to an embodiment of the present invention. [Figure 5] FIG. 3 is a diagram showing types of data included in driving data in the embodiment of the present invention. [Figure 6] FIG. 1 is a diagram schematically illustrating the relationship between input data and output data to a trained model in an embodiment of the present invention. [Figure 7] 4 is a flowchart showing an air flow rate determination process executed by the air flow rate determination device according to the embodiment of the present invention. [Figure 8] FIG. 2 is an explanatory diagram of a method for calculating a target iron production rate in an embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0018] DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS An embodiment of the present invention will be described in detail with reference to the drawings. In the drawings, the same or corresponding parts are designated by the same reference numerals and description thereof will not be repeated.

[0019] <Outline of the cast iron manufacturing process> First, an outline of the manufacturing process of cast iron will be described. In this embodiment, the manufacturing process of ductile cast iron will be described as an example. Fig. 1 is a diagram showing a schematic diagram of the manufacturing process of ductile cast iron.

[0020] Ductile cast iron is typically produced through the following steps in order: melting process P1, desulfurization process P2, molten metal holding process P3, molten metal composition analysis process P4, composition adjustment process P5, spheroidizing process P6, slag removal process P7, and casting process P8.

[0021] In the melting step P1, materials such as pig iron, scrap iron, and steel are melted in a cupola 11. The cupola 11 is a melting furnace that uses coke as a heat source. In the desulfurization step P2, the molten metal produced in the cupola 11 is poured into a desulfurization ladle 12, and the sulfur content in the molten metal in the ladle 12 is removed. Note that the desulfurization step P2 may be omitted.

[0022] In the molten metal holding process P3, the molten metal tapped from the cupola 11 is poured into a low-frequency furnace (hereinafter referred to as the "holding furnace") 13 for holding the molten metal, where it is held. In the molten metal composition analysis process P4, the composition of the molten metal in the holding furnace 13 is analyzed by a QV (countback) analyzer 14. After the composition analysis, the molten metal in the holding furnace 13 is poured into a dispensing ladle 15 of a predetermined capacity (for example, 6 tons). The ladle 15 is transported from the upstream side to a molten metal receiving position near the holding furnace 13 by a transport device (not shown). The ladle 15 is then transported to a cored wire device 21.

[0023] In the composition adjustment process P5, the composition of the molten metal is adjusted based on the analysis results from the QV analyzer 14. Specifically, the analysis results from the QV analyzer 14 are input to the composition adjuster 16, which then adjusts the amounts of elemental components such as carbon (C) and silicon (Si) to be added to the molten metal in the ladle 15.

[0024] In the spheroidizing treatment step P6, the graphite spheroidizing treatment device 20 performs a treatment of spheroidizing the graphite in the molten metal by adding magnesium into the ladle 15. The graphite spheroidizing treatment device 20 typically includes a cored wire device 21 that supplies the ladle 15 with a cored wire 22 configured such that granular magnesium alloy is enclosed in an iron outer shell.

[0025] In the slag removal process P7, the spheroidized ladle 15A is removed from the cored wire device 21, and slag is removed from the molten metal in the ladle 15A. The slag-removed ladle 15B is transported to the casting process P8 by a transport means such as a crane. In the casting process P8, the molten metal in the ladle 15B is poured into a mold, and ductile cast iron is cast.

[0026] The melting capacity of the cupola 11 in the melting step P1 varies depending on the amount of air (hot air) blown into the furnace by the blower. However, it is difficult to achieve the target pig iron production rate simply by mechanically managing the blower during continuous operation of the cupola 11. Therefore, the main feature of the blast rate control system according to this embodiment is that it is possible to calculate the blast rate appropriate for achieving the target pig iron production rate (hereinafter referred to as the "optimal blast rate") by using a "trained model" generated by machine learning the effect of increases or decreases in blast rate on increases or decreases in pig iron production rate using operational history data including the blast rate to the cupola 11 and the pig iron production rate from the cupola 11.

[0027] An air flow control system according to an embodiment of the present invention will be described in detail below.

[0028] <About the airflow control system> (Schematic configuration) 2 is a schematic diagram showing the general configuration of an air blast control system SYS according to an embodiment of the present invention. The air blast control system SYS automatically controls (adjusts) the amount of hot air sent to the cupola 11 during continuous operation of the cupola 11.

[0029] First, a brief description will be given of the cupola 11. The cupola 11 includes a furnace body 110 having a material inlet 11a at its upper end and a tapping port 11b at its lower end, and tuyere ports 11c are provided radially on the peripheral wall of the furnace body 110. The tuyere ports 11c are connected via branch pipes 112 to an annular hot blast piping 111 that surrounds the peripheral wall of the furnace body 110, and air (hot blast) sent from the blower 31 into the hot blast piping 111 is blown into the furnace body 110 through the tuyere ports 11c. A bed coke layer (not shown) is formed within the furnace body 110 of the cupola 11, and additional material is charged onto this bed coke layer through the material inlet 11a. The coke is combusted by hot air blown in from the tuyere 11c, and the heat of combustion melts the iron source, such as iron scrap or pig iron, charged on top of the bed coke layer, and the molten iron (molten metal) is taken out from the tap hole 11b at the bottom. During operation of the cupola 11, additional coke is charged into the material charging port 11a in addition to the iron source. Generally, the additional coke and the additional iron source are layered alternately at a predetermined ratio.

[0030] The air blowing volume control system SYS of this embodiment includes a controller (control means) 32 that controls the blower 31, an iron tapping volume meter (iron tapping volume measuring means) 33 that measures the amount of iron tapped from the tapping port 11b, an air blowing volume determination device 40 that determines the optimal air blowing volume for the blower 31, and a learning device 50 that performs machine learning to learn the effect of increases and decreases in the air blowing volume on increases and decreases in the amount of iron tapped during continuous operation of the cupola 11.

[0031] The blast rate determination device 40 is an information processing device including a processor and memory, and is connected to the iron tapping rate meter 33 by wire or wirelessly. Fig. 3 is a conceptual diagram that schematically shows a method for determining the blast rate by the blast rate determination device 40. As shown in Fig. 3, the blast rate determination device 40 inputs input data (condition data) extracted from operating data (raw data) obtained during continuous operation of the cupola 11 and a target iron tapping rate based on the planned molten metal consumption rate into an optimization program based on the trained model M, and determines the optimal blast rate.

[0032] The learning device 50 shown in Fig. 2 is also an information processing device including a processor and a memory, and is connected to the tapping rate meter 33 by wire or wirelessly. The learning device 50 generates a trained model M by machine learning the effect of increases and decreases in the tapping rate on increases and decreases in the blast rate using operation history data obtained during continuous operation of the cupola 11. Specifically, the trained model M is generated by learning the correlation between the blast rate and the tapping rate at the same monitoring time, taking into account the molten metal conditions at the previous (immediately preceding) monitoring time.

[0033] The air supply volume determination device 40 and the learning device 50 are connected via a network or the like. The air supply volume determination device 40 and the learning device 50 may be configured to receive operating data from a monitoring device installed in a factory. Although the air supply volume determination device 40 and the learning device 50 are shown as separate devices in FIG. 2, the air supply volume determination device 40 and the learning device 50 may be implemented by a common information processing device.

[0034] (Functional configuration of the airflow volume determination device) 4A is a functional block diagram showing the functional configuration of airflow rate determination device 40. Airflow rate determination device 40 includes a processor 41 such as a CPU (Central Processing Unit), an input unit 42, and a model storage unit 43.

[0035] The input unit 42 inputs operating data (raw data) indicating the operating status of the cupola 11 for each monitoring time (time) during continuous operation of the cupola 11. The operating data input by the input unit 42 includes multiple types of data used as input data to the trained model M.

[0036] Specifically, as shown in FIG. 5 , the operating data includes the "airflow rate (airflow speed)" per unit time of the blower 31, the "iron tapping rate (iron tapping speed)" per unit time measured by the iron tapping rate meter 33, the "oxygen enrichment amount" per unit time added to the airflow path from the blower 31 to the tuyere 11c, and the "material input amount" added to the material input port 11a. The material input amount includes the weight of each iron source material and / or the amount of coke. Note that during operation of the present system SYS, the "airflow rate" included in the operating data may correspond to the optimal airflow rate determined by a determination unit 46, which will be described later.

[0037] The model storage unit 43 stores the trained model M. The trained model M is a prediction model of the amount of hot metal produced by the learning device 50. The model storage unit 43 may be realized by a cloud server.

[0038] The processor 41 functions as a blast rate calculation means, and calculates the optimal blast rate for achieving a target iron production rate at regular time intervals (for example, every 10 minutes) based on operating data obtained during continuous operation of the cupola 11. The processor 41 includes, as its functions, a prediction processing unit 44, a calculation unit 45, and a determination unit 46. That is, the functions of the prediction processing unit 44, the calculation unit 45, and the determination unit 46 are realized by the processor 41 executing software (optimization program) stored in memory.

[0039] During continuous operation of the cupola 11, the prediction processor 44 inputs each of the multiple blast rate candidates into the trained model M together with predetermined condition data, thereby calculating the predicted iron production rate for each blast rate candidate. The "multiple blast rate candidates" may be values ​​individually set by the user, or may be any value included in an appropriate range between an upper limit and a lower limit. Alternatively, the "multiple blast rate candidates" may include the immediately preceding blast rate and a blast rate that is a predetermined amount greater or less than that blast rate, or may simply include the blast rates of the immediately preceding multiple blasts.

[0040] The "condition data" includes at least the blast rate and the iron production rate at the immediately preceding monitoring time (hereinafter referred to as the "first time"). It also includes the oxygen enrichment amount at the first time and the current time, and the material input amount at the monitoring time before the first time (hereinafter referred to as the "second time"). The monitoring time interval may be the same as the calculation timing of the optimal blast rate (for example, 10 minutes).

[0041] The "first time" includes at least one monitoring time, and preferably includes multiple monitoring times. For example, when determining the tapping rate y at monitoring time t8 in Fig. 5, the first time includes at least the most recent monitoring time t7. In this embodiment, as an example, the first time includes three monitoring times t5 to t7.

[0042] The "second time period" also includes at least one monitoring period, and preferably includes multiple monitoring periods. In this embodiment, as an example, it includes three monitoring periods t1 to t3. The first time period and the second time period may be continuous. Also, the first time period and the second time period may partially overlap. The number of monitoring periods included in each of the first time period and the second time period is, for example, 2 or more and 5 or less.

[0043] When determining the amount of pig iron tapped y at monitoring time t8 in Figure 5, the "condition data" includes the blast volume and amount of pig iron tapped for each of monitoring times t5 to t7, the amount of oxygen enrichment for each of monitoring times t5 to t8, and the integrated values ​​for each type of material for monitoring times t1 to t3, as shown by the bold frames in Figure 5.

[0044] The prediction processing unit 44 fixes the condition data, inputs the blast rate candidates at the current monitoring time t8 one by one into the trained model M, and outputs multiple predicted iron production rates. Figure 6 is a diagram schematically showing the relationship between input data and output data in the trained model M.

[0045] If a certain monitoring time (e.g., t8 in FIG. 5) is defined as the target prediction time, the input data includes a candidate blast rate D1 at the target prediction time ("x" in FIG. 5), the blast rate D2, the iron tapping rate D3, and the oxygen enrichment amount D4 for each of the three monitoring times immediately preceding the target prediction time (e.g., t5 to t7 in FIG. 5), and the oxygen enrichment amount D4A at the target prediction time. The input data also includes an accumulated weight D5 for each iron source type and an accumulated weight D6 of coke for multiple monitoring times (e.g., t1 to t3 in FIG. 5) a predetermined time before the target prediction time.

[0046] The output data is the pig iron production rate at the target prediction time ("y" in Figure 5). By including in the input data the blast rate D2, pig iron production rate D3, and oxygen enrichment rate D4 for each monitoring time immediately prior to the target prediction time, as well as the oxygen enrichment rate D4A at the target prediction time, the pig iron production rate from the cupola 11 can be appropriately predicted. Furthermore, since the iron source and coke introduced through the material inlet 11a affect the pig iron production rate with a delay compared to the introduction time, the accuracy of the prediction of the pig iron production rate can be improved by including in the input data the amount of materials introduced at a monitoring time some time before the target prediction time.

[0047] The calculation unit 45 calculates the target tapping rate based on the amount of molten metal held in the holding furnace 13 located downstream of the cupola 11 and a predetermined planned amount of molten metal to be used. The amount of molten metal held in the holding furnace 13 can be measured or calculated (predicted) by a known method, for example.

[0048] The determination unit 46 extracts, from the multiple predicted iron production rates obtained by the prediction processing unit 44, a predicted iron production rate that is close to the target iron production rate calculated by the calculation unit 45. Then, the determination unit 46 determines the optimal blast rate for this time by identifying a blast rate candidate that corresponds to the extracted predicted iron production rate.

[0049] The optimum air flow rate determined by the determination unit 46 is output to the controller 32. As a result, the controller 32 controls the operation of the blower 31 so that the air flow rate (air flow rate of the blower 31) from the tuyere 11c into the furnace body 110 becomes the optimum air flow rate. This makes it possible to automatically adjust the amount of molten metal tapped from the tapping port 11b to an amount close to the target amount of molten metal tapped.

[0050] At least one of the prediction processing unit 44, the calculation unit 45, and the determination unit 46 shown in FIG. 4(A) may be realized by hardware.

[0051] (Functional configuration of the learning device) 4(B) is a functional block diagram showing the functional configuration of the learning device 50. The learning device 50 includes a processor 51 such as a CPU, a history information storage unit 52, and a model storage unit 55.

[0052] The history information storage unit 52 stores, in chronological order, the past operation data (raw data) of the cupola 11 as operation history data. The operation history data includes the same types of data as in Fig. 5 and is stored in association with the monitoring time.

[0053] The processor 51 performs machine learning of the effect of an increase or decrease in the blast rate on an increase or decrease in the pig iron tapping rate, using the operation history data stored in the history information storage unit 52. The processor 51 includes, as its functions, a model generation unit 53 and an evaluation unit 54. That is, the functions of the model generation unit 53 and the evaluation unit 54 are realized by the processor 51 executing software stored in the memory.

[0054] The model generation unit 53 uses the operation history data as training data to generate a prediction model 55A for predicting the amount of pig iron produced from the cupola 11. The model generation unit 53 performs machine learning using the input data and output data shown in Fig. 6 as explanatory variables and objective variables of the prediction model 55A, respectively. In this embodiment, the prediction model 55A is generated by machine learning the correlations between the degree of increase or decrease in the blast rate, the degree of increase or decrease in the amount of oxygen enrichment, and the type and amount of charged materials, and the degree of increase or decrease in the amount of pig iron produced.

[0055] As an algorithm used for machine learning, for example, (multiple) regression analysis may be adopted, but other methods such as neural networks, support vector machines, decision trees, and k-NN may also be adopted.

[0056] The evaluation unit 54 evaluates the accuracy of the prediction model 55A generated by the model generation unit 53 and outputs the evaluation result to the model generation unit 53. The model generation unit 53 appropriately corrects the prediction model 55A so as to reduce the error between the iron production rate predicted by the prediction model 55A (predicted iron production rate) and the actual measured value.

[0057] Prediction model 55A generated by model generation unit 53 is stored as trained model M in model storage unit 43 in air flow rate determination device 40. Note that air flow rate determination device 40 may have the functionality of learning device 50, and in that case, air flow rate determination device 40 may be able to update the prediction model as needed.

[0058] (Method for determining airflow volume) 7 is a flowchart showing the air blast rate determination process executed by the air blast rate determination device 40. This process is executed by the processor 41 at regular intervals (for example, every 10 minutes) during continuous operation of the cupola 11. This process is realized by the processor 41 reading and executing a program stored in memory (including a downloaded program).

[0059] 7, first, the processor 41 extracts predetermined condition data from the operation data of the cupola 11 obtained via the input unit 42 (step S1). The operation data input from time to time by the input unit 42 is temporarily stored in, for example, a memory accessible by the processor 41.

[0060] The predetermined condition data is as described above and includes the blast volume D2, the iron production volume D3, the oxygen enrichment volume D4 during the immediately preceding multiple monitoring times (first time) shown in Figure 6, the oxygen enrichment volume D4A at the prediction target time, the cumulative weight D5 of each iron source type during the multiple monitoring times (second time) before the predetermined time, and the cumulative weight D6 of coke.

[0061] Next, the calculation unit 45 calculates the target pig iron productivity (step S3). The calculation unit 45 acquires the amount of molten metal held in the holding furnace 13, and calculates the target pig iron productivity based on the acquired amount of molten metal held and a preset planned amount of molten metal to be used. Note that this calculation process does not need to be performed each time the optimal blast rate is determined. A preferred method for calculating the target pig iron productivity will be described later.

[0062] The prediction processor 44 then calculates the predicted iron production rate for each of the preset blast rate candidates (step S5). That is, the blast rate D1 at the prediction target time shown in Fig. 6 is used as one of the candidate values, and the candidate value and the above-mentioned condition data are input to the trained model M to determine multiple predicted iron production rates for each of the blast rate candidates. Note that this process may be performed before step S3.

[0063] Next, the determination unit 46 extracts one predicted iron productivity rate that is close to the target iron productivity rate from the multiple predicted iron productivity rates calculated in step S3, and identifies a blast rate candidate that corresponds to the extracted predicted iron productivity rate (step S7). That is, from the multiple blast rate candidates, one blast rate candidate that has obtained a predicted iron productivity rate that is close to the target iron productivity rate is identified.

[0064] The determination unit 46 determines the current optimal airflow rate based on the identified airflow rate candidate (step S9). Typically, the identified airflow rate candidate is determined as the optimal airflow rate. Note that the specification is not limited to determining the identified airflow rate candidate as the optimal airflow rate as is, and the optimal airflow rate may be determined using a weighting coefficient or the like.

[0065] When the process of step S9 is completed, the process returns to step S1 and the above process is repeated.

[0066] (Calculation method for target iron production rate) Figure 8 is an explanatory diagram of a method for calculating the target iron tapping rate. Figure 8(A) illustrates the planned amount of molten metal to be used per hour. Figure 8(B) illustrates the lower and upper limits of the amount of molten metal stored in the holding furnace 13, the target amount of molten metal stored in the holding furnace 13, and the lower and upper limits of the tapping rate from the cupola 11 (the amount of iron tapped per unit time).

[0067] As shown in the graph of Fig. 8(C), the calculation unit 45 calculates the target tapping rate so that the final tapping rate will be close to the target molten metal storage rate in the holding furnace 13 within a preset time range (for example, 2 hours) while still maintaining the lower limit and upper limit of the molten metal storage rate in the holding furnace 13. In Fig. 8(C), if the target tapping rate at 11:00 is, for example, 17.5 tons / h, the molten metal storage rate in the holding furnace 13 after 2 hours can be set to the target molten metal storage rate (for example, 25 tons).

[0068] In this way, by calculating the target iron production rate after predicting the change in the amount of molten metal stored in the holding furnace 13 over a relatively long time range, it is possible to reduce the fluctuation range of the target iron production rate, thereby preventing the operating state of the cupola 11 from becoming unstable.

[0069] (Action and effect) As explained above, the blast rate determination device 40 can determine (calculate) a blast rate (optimum blast rate) suitable for achieving the target iron production rate by using the trained model M generated by machine learning the effect of increases and decreases in the iron production rate on increases and decreases in the blast rate. Therefore, according to the blast rate control system SYS of this embodiment, by controlling (only) the blast rate blown into the furnace body 110 from the tuyere 11c, it is possible to achieve an iron production rate as close as possible to the target iron production rate.

[0070] Furthermore, since the target iron production rate is set within a relatively long time range, the fluctuation of the optimum blast rate can be reduced, thereby suppressing changes in the tapped iron temperature (temperature drop) caused by sudden increases or decreases in the blast rate.

[0071] In addition, since the amount of iron tapped from the cupola 11 can be controlled simply by controlling the blast rate, the amount of oxygen enrichment can be minimized, thereby reducing the running costs of the cupola 11.

[0072] The airflow rate determination method (calculation method) executed by processor 41 of airflow rate determination device 40 can also be provided as a program. Similarly, the learning method executed by learning device 50 can also be provided as a program. Such a program can be provided by being recorded on an optical medium such as a CD-ROM (Compact Disc-ROM) or a computer-readable non-transitory recording medium such as a memory card. The program can also be provided by downloading it over a network.

[0073] The program according to the present invention may execute processing by calling necessary modules in a predetermined sequence at a predetermined timing among program modules provided as part of a computer's operating system (OS). In this case, the program itself does not include the modules, and executes processing in cooperation with the OS. Programs that do not include such modules may also be included in the program according to the present invention.

[0074] Furthermore, the program according to the present invention may be provided as a part of another program. In this case, the program itself does not include the modules included in the other program, and executes processing in cooperation with the other program. Such a program incorporated in another program may also be included in the program according to the present invention.

[0075] The embodiments disclosed herein should be considered to be illustrative in all respects and not restrictive. The scope of the present invention is defined by the claims, not by the above description, and is intended to include all modifications within the meaning and scope of the claims. [Explanation of symbols]

[0076] 11 Cupola, 13 Holding furnace, 31 Blower, 32 Controller, 33 Iron tapping rate meter, 40 Blast rate determination device, 41, 51 Processor, 42 Input unit, 43, 55 Model memory unit, 44 Prediction processing unit, 45 Calculation unit, 46 Determination unit, 50 Learning device, 52 History information memory unit, 53 Model generation unit, 54 Evaluation unit, 55A Prediction model, M Trained model, SYS Blast rate control system.

Claims

1. A blowing amount control system for controlling the amount of hot air blown to a cupola, a model storage means for storing a trained model generated by machine learning the influence of an increase or decrease in the blast rate on an increase or decrease in the blast rate using operation history data including the blast rate to the cupola and the blast rate from the cupola; a blast rate calculation means for calculating an optimum blast rate for achieving a target iron production rate at regular time intervals based on operational data obtained during continuous operation of the cupola; a control means for controlling the blower so that the amount of air blown to the cupola becomes the optimum amount of air blown calculated by the air blowing amount calculation means, The airflow calculation means a prediction processing means for inputting a plurality of blast rate candidates one by one into the trained model together with condition data including the blast rate and the iron production rate at the immediately preceding first time period, thereby calculating a predicted iron production rate for each of the blast rate candidates; a determination means for determining the optimal blast volume for this time by extracting a predicted blast volume that is closest to the target blast volume from the multiple predicted blast volumes obtained by the prediction processing means, and identifying the candidate blast volume that corresponds to the extracted predicted blast volume.

2. an acquisition means for acquiring the amount of molten metal held in a holding furnace located downstream of the cupola; The airflow control system according to claim 1, further comprising a calculation means for calculating the target iron production rate based on the molten metal retention rate acquired by the acquisition means and a preset planned amount of molten metal to be used.

3. 3. The airflow control system according to claim 2, wherein the calculation means calculates the target iron production rate so that the amount of molten metal held in the holding furnace ultimately approaches a predetermined target amount of stored molten metal within a preset time range.

4. The airflow control system according to claim 1 , wherein the condition data further includes an oxygen enrichment amount for the first time period and a current time period.

5. The airflow control system according to claim 1 , wherein the condition data further includes an amount of material input into the cupola at a second time that is earlier than the first time.

6. A program for determining an optimum amount of hot air to be blown into a cupola, extracting condition data including the blast rate and the iron production rate for one or more immediately preceding monitoring periods from operational data obtained during continuous operation of the cupola; a step of inputting a plurality of blast rate candidates one by one together with the condition data into a trained model generated by machine learning the influence of an increase or decrease in the blast rate on an increase or decrease in the iron production rate, thereby obtaining a predicted iron production rate for each of the blast rate candidates; A blast volume determination program that causes a computer to execute a step of determining the optimal blast volume for this time and a determination means, by extracting a predicted blast volume that is closest to the target blast volume from a plurality of predicted blast volumes and identifying the candidate blast volume that corresponds to the extracted predicted blast volume.

Images