Return ore rate prediction method, return ore rate control method, method for producing sintered ore, creating method of return ore rate prediction model and device for controlling return ore rate

Abstract

To provide a return ore rate prediction method that accurately predicts a return ore rate and a method for creating a return ore rate prediction model.SOLUTION: A method for predicting return rate is a prediction method for the return ore rate, which is a ratio of returned ore that has been screened by a screening machine to sintered ore in a producing process that includes a firing process to sinter raw materials using firing equipment, and a screening process to sort the sintered ore generated in the firing process using the screening machine. This method includes a prediction step that predicts the return ore rate based on at least one operating parameter of the firing process and an operating parameter of the screening process.SELECTED DRAWING: Figure 7

Description

Technical Field

[0001] The present disclosure relates to a return ore rate prediction method, a return ore rate control method, a sintered ore manufacturing method, a return ore rate prediction model generation method, and a return ore rate control device.

Background Art

[0002] In the iron and steel industry, sintered ore obtained by mixing fine iron ore, fine limestone, fine coke, etc. and sintering them is used as a charging material for blast furnaces.

[0003] FIG. 1 is a diagram showing an overview of a sintered ore manufacturing facility 1 used as a charging material for a blast furnace. The raw materials for sintered ore are stored in storage tanks, and the raw materials mixed at a predetermined blending ratio are mixed with water in a mixer and processed into granulated products having a larger particle size than the original raw materials. The granulated products are charged into a sintering machine 4, ignited in an ignition furnace, and then a combustion reaction gradually proceeds layer by layer from above to below by air suction from below. The sintered ore after firing is discharged from the sintering machine 4, crushed by a crusher 5, and then sent to a cooler 6. After being cooled by air in the duct of the cooler 6, it is sorted by a sieve, and the ones with a larger particle size are sent to the blast furnace as good products ("finished product 15" in FIG. 1). The sintered ore with a small particle size (for example, less than 5 mm in particle size) becomes return ore 16 and is put back into the storage tank again as a raw material.

[0004] Here, the return ore rate is defined as the ratio of return ore in the sintered ore after firing. That is, the return ore rate is the ratio of the sintered ore sieved out using a sieve to the sintered ore after firing. As a method for reducing the return ore rate and improving the yield, it is effective to promote combustion in order to prevent the sintered ore from passing through the sintering machine without being sintered. For example, it is effective to increase the blending ratio of the coagulant as a heat source, increase the blending ratio of quicklime acting as a binder during granulation for improving ventilation, and increase the ratio of the upper-layer fine coke to make it easier to ignite in the ignition furnace. Also, it is effective to lower the pallet speed to ensure the firing time. On the other hand, since there is a trade-off relationship between the reduction of the return ore rate and productivity, it is necessary to set an appropriate target value according to the operating conditions and control the return ore rate.

[0005] The return ore rate is measured in the process after screening by a sieve. Therefore, it takes time from the start of firing until the return ore rate is measured, and for example, it takes about two hours. Therefore, even if the return ore rate increases, it is detected, for example, two hours later, and the action for correction is delayed by two hours. The same applies when the return ore rate decreases, and the action for correction is delayed.

[0006] In order to detect and control fluctuations in the return ore rate at an early stage, it is necessary to predict the future return ore rate and take actions in advance. For example, Patent Document 1 discloses a method for predicting the amount of return ore from the content rate of calcium ferrite, slag content rate, pore size distribution index, and porosity, which are sintered ore constituent minerals. Patent Document 2 discloses a method for generating a database that accumulates operation condition data required for sintering treatment including the return ore occurrence ratio for each past case, and calculating the return ore occurrence ratio that conforms to the required condition data from the generated database using the method of local regression to predict the amount of return ore.

Prior Art Documents

Patent Documents

[0007]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0008] Here, the above prior art has the following problems. The method of Patent Document 1 requires a sampling test for physical property identification and cannot continuously predict the return ore rate. Therefore, it is difficult to detect fluctuations in the return ore rate at an early stage. The method of Patent Document 2 is effective in that it accumulates in a database the relationship between the operating condition data obtained from past operating results and the return ore rate, and searches for past operating results corresponding to the required condition data to detect the return ore rate at an early stage. The method of Patent Document 2 uses, as the operating condition data obtained from past operating results, the operating condition data related to firing equipment such as a sintering machine and a cooler 6, and associates it with the actual value of the return ore rate. Here, the state of the screen for screening sintered ore changes over time due to wear or clogging. Since the method of Patent Document 2 does not consider the influence of the change over time of the screen state on the return ore rate, there is room for improvement in improving the prediction accuracy of the return ore rate.

[0009] The present disclosure has been made to solve the above problems. An object of the present disclosure is to provide a return ore rate prediction method and a method for generating a return ore rate prediction model that accurately predict the return ore rate in a manufacturing process having a firing process of firing a sintering raw material using firing equipment and a screening process of screening the sintered ore using a screening machine. Another object of the present disclosure is to provide a return ore rate control method and a return ore rate control device that improve the ratio of the sintered ore that becomes a finished product and improve the yield of the sintered ore. Still another object of the present disclosure is to provide a method for manufacturing a sintered ore that produces a sintered ore with a good yield.

Means for Solving the Problems

[0010] (1) The return ore rate prediction method according to an embodiment of the present disclosure is a return ore rate prediction method for predicting a return ore rate, which is a ratio of return ore screened out by the screening machine to the sintered ore in a manufacturing process having a firing process of firing a sintering raw material using firing equipment and a screening process of screening the sintered ore generated in the firing process using a screening machine, A prediction step of predicting the return ore rate based on at least one of the operating parameters of the firing process and the operating parameters of the screening process is included.

[0011] (2) As one embodiment of the present disclosure, in (1), The prediction step uses a return ore rate prediction model generated by machine learning that includes, as input data, at least one of the operating parameters of the firing process and the operating parameters of the screening process, and outputs the return ore rate as output data.

[0012] (3) As one embodiment of the present disclosure, in (2), The manufacturing process includes a blending process of blending raw materials and a granulation process of granulating the raw materials blended in the blending process to generate the sintering raw materials. The input data includes any one of at least one of the operating parameters of the blending process and at least one of the operating parameters of the granulation process.

[0013] (4) As one embodiment of the present disclosure, in any one of (1) to (3), The operating parameters of the screening process include data related to the screening accuracy calculated using the measured value of the weight of less than the reference particle size contained in at least one sample of the sintered ore, finished product, and return ore with respect to the reference particle size corresponding to the preset screen opening.

[0014] (5) As one embodiment of the present disclosure, in any one of (1) to (4), The prediction step predicts the return ore rate by associating the operating parameters of the firing process with the operating parameters of the screening process obtained most recently from among the operating parameters of the screening process obtained in the past, based on the time point when the operating parameters of the firing process are acquired.

[0015] (6) The return ore rate control method according to one embodiment of the present disclosure is An operation amount calculation step of calculating an operation amount of an operation variable for the firing process is included so that a deviation between the return ore rate predicted using any one of the return ore rate prediction methods (1) to (5) and a target value of the return ore rate set in advance is reduced.

[0016] (7) The method for manufacturing sintered ore according to an embodiment of the present disclosure uses the operation amount of the operation variable calculated using the return ore rate control method (6) to manufacture sintered ore.

[0017] (8) The method for generating a return ore rate prediction model according to an embodiment of the present disclosure is a method for generating a return ore rate prediction model that predicts a return ore rate, which is a ratio of return ore sieved out by the screening machine to the sintered ore, in a manufacturing process having a firing process of firing a sintering raw material using a firing facility and a screening process of performing screening of the sintered ore generated in the firing process using a screening machine, and generates a return ore rate prediction model by machine learning using a plurality of learning data, where the input performance data includes at least one of the operation parameters of the firing process and the operation parameters of the screening process, and the return ore rate corresponding to the input performance data is used as output performance data.

[0018] (9) The return ore rate control device according to an embodiment of the present disclosure is a return ore rate control device that controls a return ore rate, which is a ratio of return ore sieved out by the screening machine to the sintered ore, in a manufacturing process having a firing process of firing a sintering raw material using a firing facility and a screening process of performing screening of the sintered ore generated in the firing process using a screening machine, and includes an acquisition unit that acquires at least one of the operation parameters of the firing process and the operation parameters of the screening process. A return ore rate prediction unit that predicts the return ore rate using a return ore rate prediction model generated by machine learning, the input data including at least one of the operating parameters of the obtained firing process and the operating parameters of the screening process, and the return ore rate being output data. An operation amount calculation unit that calculates an operation amount of an operation variable for the firing process so that a deviation between the predicted return ore rate and a target value of the return ore rate set in advance is reduced.

Advantages of the Invention

[0019] According to the present disclosure, it is possible to provide a return ore rate prediction method for accurately predicting the return ore rate and a method for generating a return ore rate prediction model. Further, according to the present disclosure, it is possible to provide a return ore rate control method and a return ore rate control device for improving the ratio of sintered ore as a finished product and improving the yield of sintered ore. Further, according to the present disclosure, it is possible to provide a method for manufacturing sintered ore for manufacturing sintered ore with good yield.

Brief Description of the Drawings

[0020]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

DETAILED DESCRIPTION OF THE INVENTION

[0021] The return ore rate prediction method according to this embodiment generally includes a prediction step of predicting the return ore rate based on at least one of the operating parameters of the firing process and the operating parameters of the screening process. In the prior art, no consideration was given to the change in the substantial reference particle size due to the wear or clogging of the screen 75 (see FIG. 3) used. In contrast, in the return ore rate prediction method according to this embodiment, by predicting using the operating parameters of the screening process, the return ore rate that affects the production efficiency and cost can be predicted while also reflecting the change over time in the state of the screen 75. Hereinafter, with reference to the drawings, a return ore rate prediction method, a return ore rate control method, a sintered ore manufacturing method, a return ore rate prediction model generation method, and a return ore rate control device 90 (see FIG. 8) according to an embodiment of the present disclosure will be described.

[0022] (Manufacturing Equipment) FIG. 1 is a diagram showing an overview of a sintered ore manufacturing facility 1 for manufacturing the charging raw materials of a blast furnace. The sintered ore manufacturing facility 1 includes a blending tank 2, a granulator 3, a sintering machine 4, a crusher 5, a cooler 6, and a screening machine 7. The blending tank 2 is composed of a plurality of storage tanks, namely surge hoppers 21. The components of the raw material 11 of the sintered ore are stored in each of the plurality of surge hoppers 21.

[0023] Examples of the components of raw material 11 include iron-containing raw materials containing various iron ores such as hematite and magnetite, CaO-containing raw materials containing limestone and quicklime, MgO-containing raw materials containing dolomite and refined nickel slag, etc. The components of raw material 11 further include carbonaceous materials (solid fuels) such as pulverized coke and anthracite. Here, the iron-containing raw material may include dust generated in the steelworks, such as dust that rises in the steelworks. Also, the return ore 16 that has been sorted by the screening machine 7 and has not become the finished product 15 is stored in the surge hopper 21 and becomes the raw material 11 again.

[0024] The raw material 11 stored in the blending tank 2 has a predetermined amount cut out and is sent to the granulator 3 by the blending raw material conveyor 22. The granulator 3 is, for example, a drum mixer. An appropriate amount of water is added to the raw material 11 fed into the granulator 3, and it is granulated into pseudo-particles with an average particle size of 3.0 to 6.0 mm, for example. A plurality of granulators 3 may be used, and instead of the drum mixer, a pelletizer granulator or the like may be used. The sintering raw material 12 granulated by the granulator 3 is conveyed to the sintering machine 4 by the sintering raw material conveyor 31.

[0025] The sintering machine 4 is, for example, a down-draft Dwight-Lloyd type sintering machine. The sintering machine 4 has a sintering raw material charging device 41, an endless moving pallet 42, an ignition furnace 43, a gaseous fuel supply device 44, and a wind box (wind box 45). The sintering raw material 12 is supplied to the sintering raw material charging device 41 from the sintering raw material conveyor 31. The sintering raw material 12 is charged onto the endless moving pallet 42 from the sintering raw material charging device 41. Then, the endless moving pallet 42 conveys the sintering raw material 12 in one direction (hereinafter also referred to as the conveying direction). The ignition furnace 43 is provided above the pallet 42 to sinter the sintering raw material 12. The wind box 45 is provided below the pallet 42 to generate an air flow from above to below by suction.

[0026] The sintering raw material charging device 41 includes a feeding hopper, a drum feeder, a charging gate, and a chute. The sintering raw material 12 is supplied to the feeding hopper, cut out by the drum feeder, slides down on the chute, and is charged into the pallet 42 that constitutes the pallet trolley. The charged sintering raw material 12 on the pallet 42 forms a charged layer of the sintering raw material 12.

[0027] The ignition furnace 43 ignites (ignites) the surface of the charged layer to cause a sintering reaction from the upper surface to the lower surface of the charged layer. The ignition furnace 43 is covered by a cover and is provided with a burner for igniting the surface of the charged layer. The burner injects a flame toward the upper surface of the charged layer charged into the ignition furnace 43. Thereby, the carbon material on the upper surface of the charged layer ignites. When the carbon material ignites on the upper surface of the charged layer, the carbon material in the charged layer burns sequentially downward. When the carbon material burns in the charged layer, a molten zone of the sintering raw material 12 is formed, and the sintering reaction proceeds downward in the charged layer. The charged layer is continuously conveyed by the pallet 42. The upper surface of the charged layer conveyed to the position of the burner in the ignition furnace 43 is continuously ignited, and as the pallet 42 advances, the sintering reaction proceeds from the upper surface to the lower surface of the charged layer, and a sintered cake 13 obtained by sintering the sintering raw material 12 is formed.

[0028] Below the pallet 42 passing through the ignition furnace 43, an air box 45 for generating an air flow from above to below in the thickness direction of the charged layer is provided. The air box 45 is connected to the main duct 46. By the exhaust fan 47 sucking the gas in the main duct 46, the air box 45 generates an air flow from above to below in the thickness direction of the charged layer. The exhaust fan 47 is, for example, a blower or a pump. A dust collector 48 may be provided between the main duct 46 and the exhaust fan 47 to remove dust and the like contained in the gas exhausted to the main duct 46. The exhaust fan 47 is connected to a chimney 49. The gas exhausted from the main duct 46 is discharged after harmful substances are removed.

[0029] By providing a wind box 45 below the pallet 42 passing through the ignition furnace 43, the combustion reaction of the carbonaceous material in the charging layer is promoted downward, and the sintered cake 13 can be efficiently formed.

[0030] A gaseous fuel supply device 44 may be provided on the downstream side of the ignition furnace 43 in the advancing direction of the pallet 42. By supplying gaseous fuel into the hood installed above the charging layer, the temperature of the upper surface of the charging layer, which is difficult to maintain in a high-temperature state, can be kept high. However, the sintered ore production facility 1 may be configured not to include the gaseous fuel supply device 44.

[0031] When the sintered cake 13 is formed by the sintering machine 4, the sintered cake 13 is sent to a crusher (grinder 5). Specifically, the sintered cake 13 falls at the inclined portion in the advancing direction of the pallet 42. A guide portion (crushing guide) for receiving the falling sintered cake 13 is provided, and the sintered cake 13 is guided from the guide portion to the grinder 5. The grinder 5 is, for example, a roll crusher.

[0032] The sintered cake 13 is crushed to approximately 300 mm or less by the grinder 5 to become sintered ore. However, since the sintered ore formed by the grinder 5 is in a high-temperature state (for example, 500 to 700 °C), it is air-cooled by a cooler (cooling machine 6) to 100 °C or less.

[0033] The cooler 6 uses, for example, a forced ventilation circular cooler or a water-sealed circular cooler. The forced ventilation circular cooler pushes air using a blower against the trough car that conveys the sintered ore on a circular track. The water-sealed circular cooler can reduce air leakage by realizing the seal between the trough car that conveys the sintered ore on a circular track and the air blowing section with a water seal. The sintered ore formed by the crusher 5 accumulates on the trough car. A ventilation opening is provided below the trough car, and the air sent by the blower passes through the gaps of the sintered ore accumulated on the trough car from the ventilation opening, thereby cooling the sintered ore. At this time, since the cooling air that has passed through the sintered ore has become high temperature by heat exchange, it is recovered by waste heat recovery equipment such as a separately provided boiler and used as steam. Therefore, in order to improve the waste heat recovery efficiency, the cooler 6 is provided with a hood so as to surround the trough car. That is, the blower, the trough car, and the waste heat recovery equipment are sealed so that air leakage is reduced.

[0034] As described above, since the sintering machine 4, the crusher 5, and the cooler 6 constitute a series of equipment for producing sintered ore from the sintering raw material 12 granulated by the granulator 3, in this embodiment, these are referred to as firing equipment (see FIG. 4).

[0035] The sintered ore cooled by the cooler 6 is sent to the screening machine 7. The screening machine 7 has the opening size (hole size) of the sieve 75 used in advance set. The opening size of the sieve 75 determines the particle size of the sintered ore to be sorted (referred to as the reference particle size). That is, the sintered ore that is sorted by the sieve 75 of the screening machine 7 and has a particle size less than the reference particle size is sieved out and becomes the undersize sintered ore, and the sintered ore with a particle size greater than or equal to the reference particle size remains on the sieve and becomes the oversize sintered ore. Then, the sintered ore with a particle size greater than or equal to the reference particle size sorted by the screening machine 7 becomes the finished product 15, and the sintered ore with a particle size less than the reference particle size becomes the returned ore 16. For example, the reference particle size is set to 3 - 5 mm by the sieve 75 used in the screening machine 7. The finished product 15 sorted by the screening machine 7 is charged into the blast furnace as a blast furnace raw material. On the other hand, the returned ore 16 is conveyed to the blending tank 2 and used as a raw material 11 for sintered ore.

[0036] In this embodiment, the return ore rate is defined as the ratio (weight ratio) of the return ore 16 selected by the screening machine 7 with respect to the sintered ore produced by the firing facility. That is, the smaller the return ore rate, the larger the proportion of the finished product 15 in the sintered ore. As a result, more sintered ore with a predetermined particle size can be supplied to the blast furnace, and the yield of the manufacturing process is improved. On the other hand, when the return ore rate is large, the proportion of the finished product 15 in the sintered ore becomes small. Therefore, the amount of sintered ore with a predetermined particle size that can be supplied to the blast furnace decreases, and it becomes necessary to perform firing again. As a result, the yield of the manufacturing process decreases, and the manufacturing cost of the finished product 15 increases. Further, when the return ore rate increases, the amount of sintered ore returned to the blending tank 2 increases, so that the fluctuation of the storage level of the surge hopper 21 for storing the return ore 16 becomes large. As a result, there is a problem that the blending ratio of the raw material 11 is restricted, and the strength of the finished product 15 is likely to fluctuate.

[0037] FIG. 2 shows a schematic configuration of the screening machine 7. The screening machine 7 shown in FIG. 2 includes a preliminary screen 71, a final screen 72, and a screening grinder 73. The preliminary screen 71 is a screen 75 for performing a pretreatment of selecting large sintered ore. The mesh size of the preliminary screen 71 is set to, for example, 200 mm. The sintered ore on the screen of the preliminary screen 71 is crushed by the screening grinder 73 into sintered ore of 200 mm or less. Then, the sintered ore sieved by the preliminary screen 71 and the sintered ore crushed by the screening grinder 73 are selected by the final screen 72. The mesh size of the final screen 72 corresponds to the reference particle size so as to select the finished product 15 and the return ore 16. The sintered ore on the screen of the final screen 72 becomes the finished product 15, and the sintered ore under the screen becomes the return ore 16. Here, a plurality of stages of screens 75 may be used in the screening machine 7 in order from the screen 75 with a large mesh size, and the finished product 15 may be selected step by step. In that case, as the screen 75 used in the final stage, the one with a mesh size corresponding to the reference particle size is used.

[0038] FIG. 3 shows an example of a sieve 75 used in the screening machine 7. The sieve 75 shown in FIG. 3 is formed by forming a plurality of perforations 77 in a steel plate 76. As the steel plate 76, for example, a steel plate with a thickness of 4.5 to 6.5 mm is used. The size of the steel plate 76 is, for example, 1200 mm in length × 2350 mm in width. The perforations 77 are formed so that the sintered ore having a particle size equal to or larger than the reference particle size remains on the sieve, and the sintered ore having a particle size smaller than the reference particle size is sieved out. The shape of the perforations 77 is generally a long hole. In this case, the short side length of the long hole is formed to match the reference particle size, and the long side length is formed to be larger than the reference particle size. As an example, a sieve 75 with a short side length of 5 mm, a long side length of 25 mm, and an aperture ratio, which is the area ratio of the perforations 77 to the area of the steel plate 76, of 35% is used. Thereby, the sintered ore having a particle size smaller than the reference particle size passes through the perforations 77 and becomes the undersize sintered ore. However, depending on the shape of the sintered ore, even if the average particle size converted from the volume of the sintered ore is equal to or larger than the reference particle size, it may pass through the perforations 77 and become the undersize sintered ore. In any case, the sintered ore passing through the perforations 77 formed in the sieve 75 becomes the return ore 16, and the sintered ore that did not pass through the perforations 77 becomes the finished product 15.

[0039] In the manufacturing process, the weight M1 of the finished product 15 conveyed per unit time is measured, and the weight M2 of the return ore 16 conveyed per unit time is measured. The weight M1 of the finished product 15 is measured by a weight sensor disposed on the conveyor that conveys the finished product 15. The weight M2 of the return ore 16 is measured by a weight sensor disposed on the conveyor that conveys the return ore 16. The return ore rate (R) as an actual result in the manufacturing process is specified using the following formula (1).

[0040] R = M2 / (M1 + M2) Formula (1)

[0041] Further, a weight sensor may be arranged on a conveyor that conveys the sintered ore to the sieve 7 to measure the weight M1 + M2 of the sintered ore conveyed per unit time, and the return ore rate R may be specified using Equation (1). The weight sensor may be any sensor that can measure the weight of the object being conveyed by the conveyor. For example, a Merrick type or load cell type weight sensor can be used. However, the weights of the finished product 15 and the return ore 16 may be measured by a method of estimating from an image obtained by imaging the sintered ore being conveyed on the conveyor (see, for example, Japanese Patent Application Laid-Open No. 2002-139312). Further, as another method, the weights of the finished product 15 and the return ore 16 may be measured by a method using three-dimensional scanning with a laser beam (see, for example, Japanese Patent Application Laid-Open No. 2001-304813). Since these weight measurement methods can be executed online, the actual performance data of the return ore rate can be continuously acquired. Therefore, the actual performance data of the return ore rate can be continuously acquired, for example, at time intervals of about 1 second to 5 minutes.

[0042] The sintered ore production facility 1 is controlled in operation by a control computer 74 that controls the production process of the sintered ore. The control computer 74 may be a configuration included in the sintered ore production facility 1. The control computer 74 operates each facility constituting the sintered ore production facility 1 in response to an instruction from a host computer 70 that gives a production instruction to the sintered ore production facility 1. The control computer 74 is communicably connected to each facility so as to collect the operation data acquired by the blending tank 2, the granulator 3, the sintering machine 4, the crusher 5, the cooler 6, and the sieve 7.

[0043] (Manufacturing Process and Operating Parameters) Figure 4 shows the flow of the sintered ore manufacturing process. The manufacturing process shown in Figure 4 is composed of a blending process for blending raw material 11, a granulation process for generating sintering raw material 12, a firing process for generating sintered ore, and a screening process for sorting into finished product 15 and return ore 16. Further, the firing process includes a sintering process for generating sinter cake 13 from sintering raw material 12, a pulverizing process for pulverizing sinter cake 13, and a cooling process for cooling the pulverized sinter cake 13 to generate sintered ore. In the present embodiment, the operating conditions specified in each process constituting the above manufacturing process are referred to as operating parameters. The operating parameters in the present embodiment mean the operating conditions of each process that can affect the yield of sintered ore, which is the ratio of the finished product 15 contained in the sintered ore, and the return ore rate.

[0044] In the manufacturing process shown in Figure 4, first, the raw material 11 stored in the blending tank 2 is blended at a predetermined blending ratio (blending process). In the blending process, raw material brands indicating the origin of raw materials such as iron-containing raw materials, CaO-containing raw materials, and MgO-containing raw materials are specified, and these are blended at a preset blending ratio. In the blending process, the blending ratio of the carbonaceous material, which is a coagulant, and the blending ratio of the return ore 16 stored in the surge hopper 21 are specified. Since the brand and blending ratio of the raw material 11 blended in the blending process affect the strength of the sintered ore produced in the sintering process, there is a correlation with the particle size distribution of the sintered ore. Therefore, as the operating parameters of the blending process, at least one of the raw material brand, the blending ratio of the coagulant, the blending ratio of quicklime, and the blending ratio of iron ore may be used. Further, as the operating parameter of the blending process, the moisture content (raw material moisture content) in the raw material 11 may be specified. Here, these operating parameters of the blending process may be referred to as data related to the raw material 11 as operating conditions derived from the raw material 11.

[0045] The raw material 11 compounded by the compounding process is granulated using the granulator 3 to produce the sintering raw material 12 (granulation process). In the granulation process, the watering flow rate added to the raw material 11 is set, and granulation is performed by adding a predetermined amount of moisture. Also, the particle size of the pseudo-particles granulated in the granulation process may be measured. For example, the average particle size of the sintering raw material 12 may be specified. Since these affect the strength of the sintered ore produced in the firing process, there is a correlation with the particle size distribution of the sintered ore. Therefore, as the operating parameters of the granulation process, the amount of moisture added in the granulation process or the average particle size of the sintering raw material 12 can be used.

[0046] The sintering raw material 12 produced by the granulation process is sent to the firing facility and the firing process is executed. The firing process includes a sintering process executed by the sintering machine 4, a grinding process executed by the grinder 5, and a cooling process executed by the cooler 6.

[0047] In the sintering process, the sintering raw material charging device 41 of the sintering machine 4 forms a charging layer in which the sintering raw material 12 is supplied to the pallet 42. Then, by igniting the surface of the charging layer with the ignition furnace 43, a sintering reaction is caused in the charging layer, and while the pallet 42 conveys the charging layer, a sintering reaction is sequentially caused from the upper surface of the charging layer downward. As the fuel gas of the burner of the ignition furnace 43, one or more fuels selected from coal gas, city gas, natural gas, methane gas, ethane gas, propane gas, and shale gas are used. In the sintering process, the air box 45 connected to the main duct 46 generates an air flow from above to below in the thickness direction of the charging layer, thereby promoting the sintering reaction of the charging layer. The sintering raw material 12 is fired by the sintering process to produce the sintered cake 13.

[0048] In the sintering reaction during the sintering process, the thickness of the charged layer, the conveying speed of the charged layer by the pallet 42, and the flow rate of the exhaust gas by the wind box 45 have an impact. Therefore, these data can be used as operating parameters of the sintering process, but are not limited thereto. For example, the operating parameters of the sintering process may use the concentration of the components of the exhaust gas discharged by the exhaust fan 47 via the wind box 45. As a specific example, NO x Concentration (exhaust gas NO x Concentration), O2 concentration (exhaust gas O2 concentration), SO x Concentration (exhaust gas SO x Concentration), CO concentration (exhaust gas CO concentration), CO2 concentration (exhaust gas CO2 concentration), and temperature (exhaust gas temperature) may be used. The gas components contained in the exhaust gas change depending on the combustion state of the burner of the ignition furnace 43 or the state of the sintering reaction occurring in the charged layer. For example, when there is insufficient heat, the CO partial pressure decreases and the NO x Concentration of the exhaust gas increases. That is, the increase in the NO x Concentration corresponds to the phenomenon of insufficient heat in the sintering process. That is, the NO x Concentration, O2 concentration, SO x Concentration, CO concentration, CO2 concentration, and exhaust gas temperature are correlated with the sintering reaction occurring in the charged layer and affect the strength of the sintered ore. Therefore, these may be used as operating parameters of the sintering process. Here, the components and temperature of the exhaust gas discharged from the ignition furnace 43 are specified using a gas concentration meter or thermometer arranged in the main duct 46 or the exhaust fan 47.

[0049] The crushing process crushes the sintered cake 13 produced by the sintering process with the crusher 5. As the operating parameter of the crushing process, the current value or power of the power source driving the crusher 5 may be used. Since the amount of work imparted to the sintered cake 13 by the power of the crusher 5 changes, it affects the particle size distribution of the sintered ore.

[0050] The cooling process cools the sinter cake 13 crushed by the crushing process by air cooling to produce sintered ore. As the operating parameter of the cooling process, the duct internal pressure or the blowing pressure measured in the duct of the cooler 6 can be used. The cooling rate of the sinter cake 13 in the cooling process changes due to the duct internal pressure or the blowing pressure, which may affect the magnitude of the thermal stress. In addition, sintered ore with high strength often has a small porosity inside. When the porosity is small, the air permeability inside the sintered ore decreases, so the duct internal pressure in the cooling process increases. On the other hand, when the crushed fine sintered ore accumulates on the trough trolley, the ventilation in the cooler 6 is inhibited, so the duct internal pressure increases. In any case, there is a relationship between the duct internal pressure and the strength of the sintered ore, and a correlation can be seen with the particle size distribution of the sintered ore. The duct internal pressure and the blowing pressure are measured by, for example, a pressure gauge provided in the cooler. A strain gauge type, a metal gauge type, a semiconductor gauge type, or a semiconductor diaphragm type pressure gauge can be used.

[0051] The screening process uses a screen 7 to perform screening of the sinter produced in the firing process. The screening process sorts the sinter produced by the firing process into finished product 15 and returned ore 16. The finished product 15 is charged into the blast furnace as a blast furnace raw material. On the other hand, the returned ore 16 is conveyed to the blending tank 2 and used again as the raw material 11. In the screening process, a reference particle size for sorting the finished product 15 and the returned ore 16 is set. And as the final screen 72, the mesh opening of the screen 75 is set so that the sinter with a particle size equal to or larger than the reference particle size becomes the oversize sinter on the screen, and the sinter with a particle size smaller than the reference particle size is sieved out. As in the example of the screen 75 shown in FIG. 3, the mesh opening of the final screen 72 may be set so that the short side length of the perforated portion 77 matches the reference particle size. As operating parameters of the screening process, data regarding the reference particle size and the mesh opening set for the final screen 72 can be used. The reference particle size directly affects the ratio of the returned ore 16 to the sinter. The data regarding the mesh opening set for the final screen 72 is related to the dimensions or shape of the perforated portion 77 and may be specified by the short side length and the long side length. This is because the ratio of the sinter sieved out by the final screen 72 changes depending on the dimensions or shape of the perforated portion 77. Also, as an operating parameter of the screening process, the power of the electric motor that drives the screening crusher 73 shown in FIG. 2 may be used. This is because the particle size distribution of the sinter supplied to the final screen 72 changes depending on the operating conditions of crushing the large sinter remaining on the screen by the preceding screen 71 with the screening crusher 73, which affects the yield of the sinter.

[0052] On the one hand, as the operating parameters of the screening process, it is preferable to use data related to the screening accuracy calculated using the measured value of the weight below the reference particle size contained in at least one sample of sinter, finished product 15, and returned ore 16 with respect to the reference particle size. Here, the reference particle size is a particle size determined to correspond to the mesh opening of the preset screen 75. The data related to the screening accuracy is information related to the sorting accuracy of the finished product 15 and the returned ore 16 by the screening machine 7. The final screen 72 of the screening machine 7, during the period of use in the screening machine 7 after the new perforation part 77 is formed, due to the wear of the perforation part 77 of the steel plate 76 or the clogging of the sinter in the perforation part 77, the ratio of the sinter sieved out by the final screen 72 changes. In particular, the sinter that has undergone the firing process is relatively hard and has corners in its shape. Therefore, when the sinter is applied to the screen 75, abrasive wear progresses and the substantial dimensions or shape of the perforation part 77 change. Also, in the screening process, when the sinter having a particle size similar to the reference particle size clogs the perforation part 77 of the final screen 72, the size of the perforation part 77 may be substantially reduced, thereby changing the ratio of the sinter sieved out by the final screen 72. The data related to the screening accuracy is data related to the change in the state of the screen 75 over time as described above.

[0053] More specifically, the finished product 15 and the return ore 16 in the manufacturing process are separated by the screening machine 7 into oversize sinter ore and undersize sinter ore. Each of the oversize sinter ore and the undersize sinter ore is conveyed by a dedicated conveyor. The oversize sinter ore is supplied to the blast furnace, and the undersize sinter ore is returned to the blending tank 2. Therefore, when the screening accuracy of the screen 75 used in the screening machine 7 changes, the sinter ore with a particle size less than the reference particle size is contained in the finished product 15. Similarly, the sinter ore with a particle size greater than the reference particle size is contained in the return ore 16. However, even if the finished product 15 contains a certain amount of sinter ore with a particle size less than the reference particle size, no quality problems will immediately occur in the operation of the blast furnace. Similarly, even if the return ore 16 contains a certain amount of sinter ore with a particle size greater than the reference particle size, no quality problems will immediately occur in the granulation process or the firing process. Therefore, it is not necessary for the screening machine 7 to always accurately screen the sinter ore based on the reference particle size. However, since the return ore rate generated in the manufacturing process has a great influence on the production capacity of the sinter ore manufacturing facility 1 and the blast furnace and affects the manufacturing cost of the finished product 15, it is necessary to predict the return ore rate on the premise that the screening accuracy by the screening machine 7 changes.

[0054] For example, in the prior art such as the method disclosed in Patent Document 2, the relationship between the operation condition data obtained from the past operation results and the return ore rate is stored in the database. Therefore, it can be said that the influence of the change in the screening accuracy by the screening machine 7 is reflected in the return ore rate accumulated in the database. However, when the operation conditions of the firing process change, the return ore rate is affected in a relatively short time span (for example, 1 to 3 hours), while the data related to the screen accuracy changes in a relatively long time span (for example, 6 to 48 hours). The relatively long time span is the time span corresponding to the wear or clogging of the screen 75. In addition, the database formed as described above includes both the return ore rate obtained in a state of high screen accuracy and the return ore rate obtained in a state of reduced screen accuracy. Therefore, there is room for improvement in predicting the return ore rate using such a database. In the present embodiment, including the operation parameters of the screening process as explanatory variables for predicting the return ore rate is to improve the prediction accuracy of the return ore rate for such problems.

[0055] Next, data related to the screening accuracy in this embodiment will be specifically described. The data related to the screening accuracy is calculated using the measured value of the weight of particles smaller than the reference particle size contained in at least one of the sintered ore, finished product 15, and returned ore 16 with respect to the reference particle size corresponding to the preset opening size of the sieve 75. The reference particle size is the particle size of the sintered ore that serves as a reference for separating the finished product 15 and the returned ore 16 in the screening process. In contrast, as the opening size of the sieve 75, the shape and size of the perforated portion 77 of the final sieve 72 are set. And the weight of the sintered ore smaller than the reference particle size contained in the sintered ore can be measured by collecting a sample (referred to as a pre-screening sample) from the sintered ore supplied to the screening machine 7 and measuring the weight P0 of the sintered ore smaller than the reference particle size contained in the pre-screening sample by offline measurement. Also, a sample (referred to as an over-screen sample) can be collected from the finished product 15 selected by the screening machine 7, and the weight P1 of particles smaller than the reference particle size contained in the over-screen sample can be measured by offline measurement. Similarly, a sample (referred to as an under-screen sample) can be collected from the returned ore 16 selected by the screening machine 7, and the weight P2 of particles smaller than the reference particle size contained in the under-screen sample can be measured by offline measurement. In this embodiment, data related to the screening accuracy is calculated from at least one of the weights P0, P1, and P2 measured offline in this way.

[0056] When using the weight P0 of the sintered ore smaller than the reference particle size contained in the pre-screening sample as the data related to the screening accuracy, if the weight of the pre-screening sample is M0, the theoretically calculated return ore rate (referred to as the reference return ore rate) RI0 separated by the screening machine 7 is calculated by P0 / M0. In contrast, referring to the actual performance data R0 of the return ore rate at the time of obtaining the pre-screening sample, the screening accuracy variable S can be calculated by R0 / RI0. The screening accuracy variable S calculated in this way represents the ratio of the actual return ore rate to the theoretically calculated return ore rate separated by the screening machine 7. Therefore, it can be used as data related to the accuracy of separating the finished product 15 and the returned ore 16 by the screening machine 7.

[0057] Also, as data regarding screening accuracy, when using the weight P1 of sinter ore with a particle size smaller than the reference particle size contained in the oversize sample, the weight of the sample before screening is calculated by M1 / (1 - R1). Here, the weight of the oversize sample is M1, and the actual performance data of the return ore rate at the time of obtaining the oversize sample is R1. Therefore, the theoretically selected return ore rate RI1 by the screening machine 7 is calculated by P1×(1 - R1) / M1, and the screening accuracy variable S can be calculated by R1 / RI1. Thus, it can represent the ratio of the actual return ore rate to the theoretically selected return ore rate by the screening machine 7, and can be used as relevant data regarding the separation of the finished product 15 and the return ore 16 by the screening machine 7.

[0058] Furthermore, as data regarding screening accuracy, when using the weight P2 of sinter ore with a particle size smaller than the reference particle size contained in the undersize sample, the weight of the sample before screening is calculated by M2 / R2. Here, the weight of the undersize sample is M2, and the actual performance data of the return ore rate at the time of obtaining the undersize sample is R2. Therefore, the theoretically selected return ore rate RI2 by the screening machine 7 is calculated by P2×R1 / M2, and the screening accuracy variable S can be calculated by R2 / RI2. Thus, it can represent the ratio of the actual return ore rate to the theoretically selected return ore rate by the screening machine 7, and can be used as data regarding the accuracy of the separation of the finished product 15 and the return ore 16 by the screening machine 7.

[0059] As described above, data regarding the screening accuracy calculated using the measured value of the weight less than the reference particle size contained in at least one sample among the sintered ore, finished product 15, and returned ore 16 with respect to the reference particle size corresponding to the preset mesh opening of the screen 75 is used. By this, data regarding the sorting accuracy of the finished product 15 and the returned ore 16 by the screening machine 7 can be specified, and by using the specified data, the prediction accuracy of the return ore rate can be improved. Here, the data regarding the screening accuracy is not limited to the above. For example, the same effect can be obtained by using the measured value of the weight greater than or equal to the reference particle size contained in at least one sample among the sintered ore, finished product 15, and returned ore 16 with respect to the reference particle size corresponding to the preset mesh opening of the screen 75. That is, as the data regarding the screening accuracy, data calculated with the relationship between the theoretically measured return ore rate and the actually measured return ore rate data sorted by the screening machine 7 as a variable may be used.

[0060] The data regarding the screening accuracy needs to be specified by offline measurement. That is, the measured value of the weight less than the reference particle size contained in at least one sample among the sintered ore, finished product 15, and returned ore 16 is obtained by offline measurement. For the measurement of the weight less than the reference particle size contained in the sample, using a screen 75 whose accuracy has been verified, the sintered ore less than the reference particle size and the sintered ore greater than or equal to the reference particle size are sorted, and the weight of the sintered ore sorted to be less than the reference particle size is measured. However, the measurement of the weight less than the reference particle size contained in the sample is not limited to using the screen 75. By measuring the particle size distribution (grain size distribution) of the sample using a screen 75 having a plurality of mesh openings, the weight ratio less than the reference particle size contained in the sample may be specified. Further, the weight less than the reference particle size contained in the sample may be calculated based on the specified weight ratio. For the measurement of the particle size distribution, methods such as laser diffraction / scattering particle size distribution measurement or image processing method may be applied, for example.

[0061] Data on the screening accuracy is identified by off-line measurement. Therefore, it is difficult to identify data on the screening accuracy in a short time period. Data on the screening accuracy can be acquired, for example, at a frequency of once every 30 to 90 minutes. Accordingly, data on the screening accuracy may be acquired at any time during the operation of the manufacturing process, and the latest measurement data may be used as the data on the current screening accuracy. This is because data on the screening accuracy changes over a relatively long time span such as wear or clogging of the screen 75, and it can be estimated that there will be no rapid change until the next off-line measurement. Also, data on the screening accuracy may store the data measured in the past including the latest measurement data, and use the average value of the weights less than the reference particle size obtained by multiple off-line measurements. In this case, a reference period of 1 to 2 days may be set, and a moving average may be calculated for the data within the reference period including the latest measurement data, and the calculated moving average may be used as the data on the screening accuracy. Data on the screening accuracy changes over a relatively long time span such as wear or clogging of the screen 75, and is preferable in that it can remove measurement noise for each off-line measurement.

[0062] (Generation of Return Ore Rate Prediction Model) Hereinafter, a return ore rate prediction model generation unit 80 that executes the method for generating a return ore rate prediction model according to the present disclosure will be described. The return ore rate prediction model generation unit 80 includes, in the sintered ore manufacturing facility 1, at least one of the operation parameters of the firing process acquired by the control computer 74 and the operation parameters of the screening process in the input performance data. Then, the return ore rate prediction model generation unit 80 generates a return ore rate prediction model that predicts the return ore rate by machine learning using a plurality of learning data with the actual performance data of the return ore rate as the output performance data. Further, the return ore rate prediction model preferably includes, as input performance data, at least one of the actual performance data of at least one of the operation parameters of the blending process and at least one of the actual performance data of the operation parameters of the granulation process.

[0063] FIG. 5 is a diagram for explaining the configuration of the return ore rate prediction model generation unit 80. As shown in FIG. 5, the return ore rate prediction model generation unit 80 includes a database unit 81 and a machine learning unit 82. The database unit 81 accumulates the operation parameters of the firing process, the operation parameters of the screening process, and the actual data of the return ore rate acquired by the control computer 74. Further, if necessary, the actual data of the operation parameters of the blending process or the granulation process may be accumulated. The actual data accumulated in the database unit 81 may appropriately acquire the information stored in the control computer 74 or the upper computer 70 that gives a production instruction to the control computer 74. However, the actual data accumulated in the database unit 81 may be acquired via an operation data server 94 (see FIG. 8) that is communicably connected to the control computer 74 or the upper computer 70.

[0064] A data acquisition unit 83 may be provided in the return ore rate prediction model generation unit 80 to collect the actual data accumulated in the database unit 81. After the actual data is temporarily stored in the data acquisition unit 83 and a data set associating a plurality of types of actual data is generated, it may be accumulated in the database unit 81.

[0065] The data acquisition unit 83 preferably associates a plurality of types of performance data in consideration of the delay time based on the timing at which the actual performance data of the return ore rate can be specified. Here, the timing at which the return ore rate can be specified is, for example, when the screening machine 7 performs the process of sorting the sintered raw material 12 into the finished product 15 and the return ore 16 after firing by the sintering machine 4. The actual performance data of the operating parameters that affect the particle size distribution of the sintered ore at the time of this determination is acquired in consideration of the above delay time. That is, in the manufacturing process of sintered ore, each process of the raw material 11 blending process, granulation process, sintering process, pulverization process, cooling process, and screening process is executed in sequence. Therefore, there is a delay time of about 3 hours from the execution of the blending process to the screening process where the actual performance data of the return ore rate is specified. Also, there is a delay time of about 2 hours from the execution of the sintering process to the screening process. Further, there is a delay time of about 1 hour from the execution of the cooling process to the screening process. Therefore, the data acquisition unit 83 associates the actual performance data of the return ore rate acquired in the screening process with the operating parameters at the time when each process is executed. Since a data set in which the actual performance data of the operating conditions in the manufacturing process and the actual performance data of the return ore rate are appropriately associated is generated, the prediction accuracy of the return ore rate prediction model can be improved. Here, the delay time may be specified using the information stored in the control computer 74 that controls each process or the upper computer 70 that gives a manufacturing instruction to the control computer 74.

[0066] On the one hand, when using data related to screening accuracy as an operating parameter of the screening process, the data related to screening accuracy is specified by off-line measurement. Therefore, the data acquisition unit 83 may associate the latest data related to screening accuracy based on the timing at which the actual data of the return ore rate can be specified from the data related to screening accuracy acquired in the past with the actual data of the return ore rate. Since the change in the data related to screening accuracy is small compared to the cycle (e.g., 1 to 5 minutes) in which the actual data of the return ore rate can be acquired, the influence on the prediction accuracy of the return ore rate is small. Also, the data related to screening accuracy may be associated with the actual data of the return ore rate acquired at the current time by calculating the moving average of the data within a reference period (e.g., 1 to 2 days) including the latest measurement data. Here, since both the actual data of the return ore rate and the operating parameters of the screening process are acquired in the screening process, there is no need to consider the above-mentioned delay time.

[0067] The return ore rate prediction model generation unit 80 can be provided in the control computer 74 that controls the sintered ore manufacturing facility 1. Also, the return ore rate prediction model generation unit 80 may be provided in the upper computer 70 that gives a production instruction to the control computer 74, or may be provided in an independent computer that can communicate with other devices. Further, the return ore rate prediction model generation unit 80 may be configured in a device separate from the database unit 81 using a device capable of receiving the data set stored in the database unit 81.

[0068] The database unit 81 preferably stores operation data of the sintered ore manufacturing process for six months or more, more preferably one year or more, and even more preferably two years or more. Further, the database unit 81 stores 30,000 or more data sets, preferably 200,000 or more, and even more preferably 500,000 or more. The data stored in the database unit 81 may be screened as necessary. Also, the data sets stored in the database unit 81 may be appropriately updated within a certain upper limit of the number of data sets. Further, the data sets stored in the database unit 81 may be updated with the latest operation data obtained over a certain period.

[0069] The machine learning unit 82 includes, as input performance data, at least one of the operation parameters of the firing process acquired by the control computer 74 and the operation parameters of the screening process, using the data sets stored in the database unit 81. Further, the machine learning unit 82 generates a return ore rate prediction model by machine learning using a plurality of learning data, with the return ore rate corresponding to the input performance data as output performance data. The return ore rate prediction model may include at least one of the operation parameters of the blending process as input performance data. Also, the return ore rate prediction model may include at least one of the operation parameters of the granulation process as input performance data. The machine learning for generating the return ore rate prediction model only needs to obtain a practically sufficient prediction accuracy for the return ore rate and is not limited to a specific method. For example, a neural network (including deep learning or convolutional neural network, etc.), decision tree learning, GBDT (Gradient Boosting Decision Tree), random forest, support vector regression, etc. may be used. Also, an ensemble model combining a plurality of models may be used. Further, a classification model such as the k-nearest neighbor method or logistic regression may be used.

[0070] For example, a return ore rate prediction model can be generated by machine learning using a general neural network as shown in FIG. 6. The reference numerals L1, L2, and L3 in FIG. 6 indicate the input layer, the intermediate layer, and the output layer, respectively. In particular, when using deep learning, other operating parameters having a correlation with the return ore rate can be freely selected as inputs without considering the problem of multicollinearity, so that the prediction accuracy of the return ore rate can be improved. For example, a neural network with one intermediate layer and four nodes each can be used, and ReLU (Rectified Linear Unit) can be used as the activation function.

[0071] The machine learning unit 82 may improve the prediction accuracy of the return ore rate by dividing the data set stored in the database unit 81 into training data and test data and performing learning. For example, the machine learning unit 82 learns the weight coefficients of the neural network using the training data, and generates a return ore rate prediction model while appropriately changing the structure of the neural network (the number of intermediate layers and the number of nodes) so that the correct answer rate of the return ore rate in the test data becomes high. The error backpropagation method can be used to update the weight coefficients. Here, the return ore rate prediction model may be updated to a new model by re-learning, for example, every six months or every year. This is because the more data is stored in the database unit 81, the more accurate the prediction of the return ore rate becomes, and by updating the return ore rate prediction model based on the latest data, a return ore rate prediction model reflecting the change in the operating conditions of the sinter ore manufacturing facility 1 can be generated.

[0072] As the operating parameters of the firing process used for the input of the return ore rate prediction model, at least any one of the operating parameters selected from the operating parameters of the sintering process, the operating parameters of the grinding process, and the operating parameters of the cooling process may be used.

[0073] As the operating parameters of the sintering process, the thickness of the charging layer, the conveying speed of the charging layer by the pallet 42, and the flow rate of the exhaust gas by the wind box 45 can be used. As the operating parameters of the sintering process, the NO of the exhaust gas discharged by the exhaust fan 47 through the wind box 45x concentration, O2 concentration, SO x Concentration, CO concentration, CO2 concentration, and exhaust gas temperature may be used. The operating parameters of the sintering process affect the sintering reaction occurring in the charging layer and are correlated with the strength of the sintered ore. Therefore, the operating parameters of the sintering process affect the particle size distribution of the sintered ore in the crushing process or the cooling process.

[0074] As the operating parameter of the crushing process, the power of the motor driving the crusher 5 may be used. This is because it affects the particle size distribution of the crushed sintered ore.

[0075] As the operating parameter of the cooling process, the duct internal pressure and the blowing pressure measured in the duct of the cooler 6 can be used. This is because the duct internal pressure or the blowing pressure changes according to the particle size distribution of the crushed sintered cake 13, which affects the particle size distribution of the sintered ore.

[0076] On the other hand, the operating parameters of the screening process used as the input of the return ore rate prediction model may include the reference particle size, data related to the mesh opening set for the final screen 72, and the power of the motor driving the screening crusher 73. Also, as the operating parameter of the screening process, it is preferable to use data related to the screening accuracy. Since the data related to the screening accuracy reflects the change in the state of the screen 75 over time, the prediction accuracy of the return ore rate can be improved.

[0077] Furthermore, when using the operating parameters of the blending process as the input of the return ore rate prediction model, as the operating parameters of the blending process, it is advisable to use the raw material brand, the blending ratio of the binder, the blending ratio of quicklime, the blending ratio of iron ore, and the raw material moisture content. Also, when using the operating parameters of the granulation process as the input of the return ore rate prediction model, as the operating parameters of the granulation process, it is advisable to use the amount of water added in the granulation process or the average particle size of the sintering raw material 12. This is because the blending state or granulation state of the sintering raw material 12 affects the sintering reaction in the sintering machine 4 and the particle size distribution of the sintered ore in the crushing process or the cooling process.

[0078] (Prediction method of return ore rate) The prediction method of the return ore rate according to the present disclosure includes a prediction step of predicting the return ore rate based on at least one of the operating parameters of the firing process and the operating parameters of the screening process in the sintered ore manufacturing facility 1.

[0079] For the prediction of the return ore rate, the operating parameters of the firing process and the operating parameters of the screening process are respectively classified, and the actual values (for example, average values) of the return ore rate belonging to the corresponding classifications are stored as data in table form. During the operation of the sintered ore manufacturing facility 1, the actual data of the operating parameters of the firing process and the actual data of the operating parameters of the screening process are acquired from the control computer 74. Then, the return ore rate corresponding to the classification of the above table may be output as the prediction result of the return ore rate. Also, using the data accumulated in the database unit 81, the return ore rate may be predicted by a locally weighted regression method using the actual data of the operating parameters of the firing process and the actual data of the operating parameters of the screening process acquired from the control computer 74.

[0080] On the other hand, in the sintered ore manufacturing facility 1, it is preferable to predict the return ore rate using the above return ore rate prediction model in the prediction method of the return ore rate according to the present disclosure.

[0081] FIG. 7 is a diagram for explaining the configuration of a return ore rate prediction unit 84 that predicts the return ore rate. The return ore rate prediction unit 84 is configured to be able to communicate with a return ore rate prediction model generation unit 80 so as to acquire the return ore rate prediction model generated by the return ore rate prediction model generation unit 80. The return ore rate prediction unit 84 can be provided in the control computer 74 that controls the sintered ore manufacturing facility 1. Also, the return ore rate prediction unit 84 may be provided in an independent computer that can communicate with other devices.

[0082] During the operation of the sintered ore production facility 1, the return ore rate prediction unit 84 acquires the actual data of the operation parameters of the firing process and the actual data of the operation parameters of the screening process from the control computer 74, inputs them into the return ore rate prediction model, and outputs the prediction result of the return ore rate. The return ore rate prediction unit 84 may, if necessary, input at least one of the operation parameters of the blending process or the granulation process from the control computer 74 into the return ore rate prediction model and output the prediction result of the return ore rate.

[0083] In this case, it is preferable to provide the acquisition unit 85 and associate the operation parameters in each of the blending process, granulation process, sintering process, pulverization process, and cooling process. That is, among the operation parameters of the firing process, the timing at which the actual data of the final process closest to the screening process is acquired is used as a reference. And for the actual data of the operation parameters in the upstream process, it is advisable to associate multiple types of actual data considering the delay time until the final process. Thereby, when the final process is being executed at the current time, the return ore rate when the screening process is later executed can be accurately predicted.

[0084] On the other hand, when using data related to the screening accuracy as the operation parameter of the screening process, the timing at which the acquisition unit 85 acquires the operation parameter of the firing process may be used as a reference. And the operation parameter of the screening process acquired most recently from among the operation parameters of the screening process acquired in the past may be associated with the operation parameter of the firing process. The data related to the screening accuracy is intermittently specified by off-line measurement. Therefore, until the data related to the screening accuracy is updated to the latest information, it is preferable for the acquisition unit 85 to store the data related to the screening accuracy acquired most recently and associate it with the operation parameter of the firing process specified at the current time.

[0085] The prediction of the return ore rate is executed, for example, at a pitch of 1 to 5 minutes because the control computer 74 continuously acquires the operation result data of the manufacturing process. The prediction result of the return ore rate may be output to a monitor screen provided in the control room of the sintered ore manufacturing facility 1. The operator of the sintered ore manufacturing facility 1 can appropriately change the operation conditions of the sintered ore manufacturing facility 1 to appropriate ones according to the prediction result of the return ore rate displayed on the monitor screen.

[0086] (Control method of manufacturing facility and manufacturing method of sintered ore) The return ore rate control method according to the present disclosure includes an operation amount calculation step. The operation amount calculation step calculates the operation amount of the operation variable for the firing process so that the deviation between the return ore rate predicted using the above return ore rate prediction method and the target value of the return ore rate set in advance is reduced. Further, the manufacturing method of sintered ore according to the present disclosure manufactures sintered ore using the operation amount of the operation variable calculated using the operation amount calculation step.

[0087] FIG. 8 is a diagram showing a configuration example of the return ore rate control device 90 according to the present embodiment. As shown in FIG. 8, the return ore rate control device 90 includes a storage unit 91, an acquisition unit 85, a return ore rate prediction unit 84, an operation amount calculation unit 92, and an output unit 93. The return ore rate control device 90 acquires an actual value and a target value of the return ore rate in the manufacturing process executed by the sintered ore manufacturing facility 1 from the operation data server 94. The actual value may include measured values of operation parameters of the sintered ore manufacturing facility 1 and current operation variables. The operation data server 94 can communicate with the return ore rate control device 90 via a network and may be realized by, for example, the control computer 74. The network is, for example, the Internet. The return ore rate control device 90 executes the above-described process, that is, the process of predicting the return ore rate using the return ore rate prediction model and obtaining the operation amount of the operation variable of the firing process so that the future return ore rate is maintained near the target value. Further, in the present embodiment, the return ore rate control device 90 has a function of outputting the operation amount of the operation variable to the facility by the output unit 93 or presenting it as a guidance operation amount. When the output unit 93 presents the guidance operation amount, the return ore rate control device 90 functions as an operation guidance device. The display unit 95 displays the guidance operation amount output from the return ore rate control device 90 (operation guidance device). The return ore rate control device 90 may be configured by a computer different from the operation data server 94 (for example, a process computer that manages the operation of the sintering machine 4 or a sintering operation guidance server). The display unit 95 may be a display device such as a liquid crystal display (Liquid Crystal Display) or an organic EL panel (Organic Electro-Luminescence Panel). Further, the display unit 95 may be realized by a display of a terminal device such as a smartphone or a tablet. The terminal device can communicate with the return ore rate control device 90 via a network. A sintering operation guidance system may be configured by a sintering operation guidance server having the function of the return ore rate control device 90 and a terminal device having the function of the display unit 95. The sintering operation guidance server and the terminal device may be in the same place (for example, in the same factory) or may be physically separated. The sintering operation guidance system may further include the operation data server 94.

[0088] The return ore rate prediction model may have a return ore rate prediction model generation unit 80 provided in the return ore rate control device 90, and be generated thereby and stored in the storage unit 91. Alternatively, the return ore rate prediction model generation unit 80 may be provided in another computer, and the return ore rate prediction model generated by the other computer may be stored in the storage unit 91.

[0089] The components of the return ore rate control device 90 will be described below. The storage unit 91 stores the return ore rate prediction model. The storage unit 91 also stores programs and data related to return ore rate control. The storage unit 91 may store the acquired actual value and target value. The storage unit 91 may store various information obtained by processes for return ore rate control. The storage unit 91 may include any storage device such as a semiconductor storage device, an optical storage device, and a magnetic storage device. The semiconductor storage device may include, for example, a semiconductor memory. The storage unit 91 may include a plurality of types of storage devices.

[0090] The acquisition unit 85 acquires the target value of the return ore rate and the input data of the return ore rate prediction model including the operation parameters of the firing process and the operation parameters of the screening process, taking into account the delay time based on the timing at which the return ore rate can be specified.

[0091] The return ore rate prediction unit 84 predicts the return ore rate using the input data and the return ore rate prediction model.

[0092] The operation amount calculation unit 92 calculates the operation amount of the operation variable for the firing process so as to reduce the deviation between the predicted value and the target value of the return ore rate.

[0093] The output unit 93 outputs the calculated operation amount of the operation variable to the control computer 74 or presents it to the display unit 95 as a guidance operation amount.

[0094] When the operation amount of the manipulated variable is output from the output unit 93 to the control computer 74, the manipulated variable may be automatically updated by the control computer 74. That is, the return ore rate control method according to the present embodiment can be executed as a part of the method for manufacturing sintered ore. Further, the operator may change the operating conditions of the sintered ore manufacturing facility 1 based on the guidance operation amount shown on the display unit 95. Such operation guidance for the sintered ore manufacturing facility 1 can be executed as a part of the manufacturing method for manufacturing sintered ore.

[0095] The return ore rate control device 90 can be realized by, for example, a computer as described above. The computer includes, for example, a memory, a hard disk drive (storage device), a CPU (processing device), and the like. The program can be stored in the hard disk drive and read from the hard disk drive into the memory when executed by the CPU. Also, data during processing is stored in the memory and stored in the HDD if necessary. The storage unit 91 may be realized by, for example, a storage device. The acquisition unit 85, the return ore rate prediction unit 84, the operation amount calculation unit 92, and the output unit 93 may be realized by, for example, a CPU that reads and executes a program.

[0096] FIG. 9 is a flowchart showing the return ore rate control method according to the present embodiment. The return ore rate control device 90 may calculate the operation amount of the manipulated variable and output it as the guidance operation amount according to the flowchart shown in FIG. 9. The return ore rate control method shown in FIG. 9 is also an operation guidance method and can be executed as a part of the method for manufacturing sintered ore.

[0097] The acquisition unit 85 acquires the actual value and the target value (step S1, acquisition step). In the present embodiment, the actual value includes the actual data of the operation parameters of the firing process and the actual data of the operation parameters of the screening process. Here, the actual data in the flowchart of FIG. 9 is data including the operation parameters including real time and the operation parameters with a time difference considering the above-mentioned delay time. Further, the actual data of the operation parameters of the blending process or the actual data of the operation parameters of the granulation process may be included. The return ore rate prediction unit 84 predicts the return ore rate using the input data and the return ore rate prediction model (step S2, prediction step). The return ore rate is the output data of the return ore rate prediction model. The operation amount calculation unit 92 calculates the operation amount of the operation variable for the firing process so as to reduce the deviation between the predicted value and the target value of the return ore rate (step S3, operation amount calculation step). The output unit 93 outputs the calculated operation variable (step S4).

[0098] As an action for reducing the return ore rate, it is effective to promote combustion in order to prevent the sintered ore from passing through the sintering machine 4 without being fired. For example, it is conceivable to increase the blending ratio of the coagulant which is a heat source. Also, increasing the blending ratio of quicklime acting as a binder during granulation for improving ventilation, increasing the ratio of powdered coke in the upper layer to make it easier to ignite in the ignition furnace 43, and ensuring the firing time by lowering the pallet speed are also effective. As an action for improving productivity when the return ore rate is low, it is effective to lower the blending ratio of the coagulant, lower the blending ratio of quicklime, reduce the ratio of powdered coke in the upper layer, and increase the pallet speed.

[0099] (Example) As an example of the present disclosure, an example in which the return ore rate is predicted using the return ore rate prediction model in the sintered ore production facility 1 will be described below.

[0100] In the embodiment, the actual data of the return ore rate, the actual data of the operating parameters of the screening process, and the actual data of the operating parameters of the sintering process obtained considering the delay time were associated with each other in a database and stored as learning data. The learning data was accumulated by manufacturing sintered ore in the sintered ore manufacturing facility 1 for six months. Then, using the learning data stored in the database, the operating parameters of the sintering process and the operating parameters of the screening process were used as input actual data. Also, the return ore rate corresponding to the input actual data was used as output actual data, and a return ore rate prediction model was generated by machine learning using the neural network method.

[0101] As the operating parameters of the sintering process for the operating parameters of the sintering process used as the input of the return ore rate prediction model, NO x concentration, CO concentration, and the flow rate of the exhaust gas were used. NO x The concentration and CO concentration are the NO of the exhaust gas discharged by the exhaust fan 47 through the wind box 45. x concentration and CO concentration. Also, the flow rate of the exhaust gas is the flow rate of the exhaust gas by the wind box 45. In addition, as the operating parameters of the sintering process, the duct internal pressure measured in the duct of the cooler 6 as the operating parameters of the cooling process was used.

[0102] The operating parameters of the screening process were used by measuring the weight P0 of sintered ore less than the reference particle size (5 mm) contained in the pre-screening sample by offline measurement as data related to the screening accuracy, calculating the screening accuracy variable S using the reference return ore rate RI0, and using this. Here, the data related to the screening accuracy was specified at a frequency of once every 30 minutes while the sintered ore manufacturing facility 1 was manufacturing sintered ore.

[0103] In the embodiment, using the return ore rate prediction model generated in this way, the return ore rate of sintered ore was predicted for one month after obtaining the learning data. The prediction of the return ore rate was executed at a frequency of once per minute, and the return ore rate 1.5 hours after the firing process was executed was predicted. As a result, as shown in FIG. 10, the mean absolute error (MAE) between the predicted result and the actual result of the return ore rate was 1.0%. On the other hand, as a comparative example, using the above learning data, a prediction model was generated with the operating parameters of the firing process as input and without using the operating parameters of the screening process as input, and the prediction accuracy of the return ore rate was evaluated. As a result, as shown in FIG. 10, the mean absolute error (MAE) between the predicted result and the actual result of the return ore rate in the comparative example was 1.3%. Thereby, the return ore rate prediction model using the operating parameters of the screening process has improved prediction accuracy of the return ore rate compared to the prediction model using only the operating parameters of the firing process.

[0104] As described above, according to the present disclosure, in a manufacturing process having a firing process and a screening process, a return ore rate prediction method and a method for generating a return ore rate prediction model that accurately predict the return ore rate by reflecting the temporal change in the state of the screen 75 can be provided. Further, a return ore rate control method and a return ore rate control device 90 that improve the ratio of sintered ore that becomes the finished product 15 and improve the yield of sintered ore can be provided. Further, a method for manufacturing sintered ore that produces sintered ore with a good yield can be provided.

[0105] The embodiments according to the present disclosure have been described based on the drawings and examples, but it should be noted that those skilled in the art can easily make various modifications or corrections based on the present disclosure. Therefore, it should be noted that these modifications or corrections are included in the scope of the present disclosure. For example, the functions included in each component or each step can be rearranged so as not to be logically contradictory, and a plurality of components or steps can be combined into one or divided. The embodiments according to the present disclosure can also be realized as a program executed by a processor provided in the device or a storage medium storing the program. It should be understood that these are also included in the scope of the present disclosure.

[0106] The configuration of the return ore rate control device 90 shown in FIG. 8 is an example. The return ore rate control device 90 does not necessarily include all of the components shown in FIG. 8. Further, the return ore rate control device 90 may include components other than those shown in FIG. 8. For example, the return ore rate control device 90 may be configured to further include a display unit 95.

Explanation of Signs

[0107] 1 Sinter production facility 2 Blending tank 3 Granulator 4 Sintering machine 5 Crusher 6 Cooler 7 Sieve 11 Raw material 12 Sintering raw material 13 Sinter cake 15 Finished product 16 Return ore 21 Surge hopper 22 Blending raw material conveyor 31 Sintering raw material conveyor 41 Sintering raw material charging device 42 Pallet 43 Ignition furnace 44 Gas fuel supply device 45 Wind box 46 Main duct 47 Exhaust fan 48 Dust collector 49 Chimney 70 Host computer 71 Pre-screen 72 Final screen 73 Screening crusher 74 Control computer 75 Screen 76 Steel plate 77 Perforated part 80 Return ore rate prediction model generation unit 81 Database unit 82 Machine learning unit 83 Data acquisition unit 84 Return ore rate prediction unit 85 Acquisition unit 90 Return ore rate control device 91 Memory unit 92 Operation amount calculation unit 93 Output unit 94 Operation data server 95 Display unit

Claims

1. A return ore rate prediction method for predicting the ratio of return ore to sintered ore that is screened out by a screening machine in a manufacturing process having a firing process for firing a sintering raw material using a firing facility and a screening process for performing screening of the sintered ore generated in the firing process using a screening machine, comprising: a prediction step of predicting the return ore rate based on at least one of the operating parameters of the firing process and the operating parameters of the screening process.

2. The prediction step includes using a return ore rate prediction model generated by machine learning, which includes, as input data, at least one of the operating parameters of the firing process and the operating parameters of the screening process, and outputs the return ore rate as output data. The return ore rate prediction method according to claim 1.

3. The manufacturing process includes a blending process for blending raw materials and a granulation process for granulating the raw materials blended in the blending process to generate the sintering raw material. The input data includes any one of at least one of the operating parameters of the blending process and at least one of the operating parameters of the granulation process. The return ore rate prediction method according to claim 2.

4. The operating parameters of the screening process include data related to screening accuracy calculated using measured values of the weight of particles smaller than a reference particle size included in at least one sample of the sintered ore, finished products, and return ore with respect to a reference particle size corresponding to a preset screen aperture. The return ore rate prediction method according to any one of claims 1 to 3.

5. The prediction step predicts the return ore rate by associating the operating parameters of the firing process with the operating parameters of the screening process obtained most recently from among the operating parameters of the screening process obtained in the past, based on the time point when the operating parameters of the firing process are acquired. The return ore rate prediction method according to any one of claims 1 to 3.

6. An operation amount calculation step of calculating an operation amount of an operation variable for the firing process so that a deviation between the return ore rate predicted using the return ore rate prediction method according to any one of claims 1 to 3 and a target value of the preset return ore rate is reduced. A return ore rate control method.

7. A method for manufacturing sintered ore, which uses the operation amount of the operation variable calculated by using the return ore rate control method according to claim 6 to manufacture sintered ore.

8. A method for generating a return ore rate prediction model that predicts the return ore rate, which is the ratio of the return ore screened out by the screening machine to the sintered ore, in a manufacturing process having a firing process of firing sintering raw materials using a firing facility and a screening process of performing screening of the sintered ore generated in the firing process using a screening machine, A method for generating a return ore rate prediction model, which generates a return ore rate prediction model by machine learning using a plurality of learning data, where the input performance data includes at least one of the operation parameters of the firing process and the operation parameters of the screening process, and the return ore rate corresponding to the input performance data is used as the output performance data.

9. A return ore rate control device that controls the return ore rate, which is the ratio of the return ore screened out by the screening machine to the sintered ore, in a manufacturing process having a firing process of firing sintering raw materials using a firing facility and a screening process of performing screening of the sintered ore generated in the firing process using a screening machine, An acquisition unit that acquires at least one of the operation parameters of the firing process and the operation parameters of the screening process; A return ore rate prediction unit that predicts the return ore rate by using a return ore rate prediction model generated by machine learning, where the input data includes at least one of the acquired operation parameters of the firing process and the operation parameters of the screening process, and the return ore rate is used as the output data; A return ore rate control device including an operation amount calculation unit that calculates the operation amount of the operation variable for the firing process so as to reduce the deviation between the predicted return ore rate and a preset target value of the return ore rate.

Images