Method for controlling return ore rate, method for manufacturing sintered ore, and return ore rate control device

Abstract

Provided is a return ore rate control method for controlling a sintering process executed in a facility that includes a sintering machine and a cooler, on the basis of a predicted value of a return ore rate predicted using a return ore rate prediction model in which at least data related to the cooler is used as an explanatory variable and the return ore rate is used as an objective variable, the return ore rate control method comprising: an acquisition step (S1) for acquiring a target value of the return ore rate and input data of the return ore rate prediction model including data related to the cooler, in consideration of a delay time based on a timing at which the return ore rate can be identified; a return ore rate prediction step (S2) for predicting the return ore rate using the input data and the return ore rate prediction model; and an operation amount calculation step (S3) for calculating an operation amount of an operation variable for the sintering process so as to reduce the deviation between the predicted value and the target value of the return ore rate.

Description

Return ore rate control method, sintered ore manufacturing method, and return ore rate control device

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

[0002] In the steel industry, the quality of iron ore has declined over many years of mining. As a result, the proportion of fine ore with a high fineness that has been dressed at the mine site is increasing, and the sintering process, in which the fine ore is solidified to produce sintered ore before being charged into a blast furnace, is becoming increasingly important.

[0003] Figure 1 shows an overview of the sintering process. Sinter is produced by burning small-particle iron ore using the heat generated by the combustion of a coagulant. If powdered iron ore were directly charged into a sintering machine, poor ventilation would inhibit the combustion reaction. Therefore, in the granulation process, iron ore is mixed with water in a mixer along with other raw materials such as quicklime and coke to produce granules with larger particle sizes than the original raw materials. The granules are then charged into the sintering machine, ignited in an ignition furnace, and the combustion reaction proceeds gradually from top to bottom in layers due to air suction from below. After firing, the sintered ore is discharged from the sintering machine, crushed in a crusher, and then sent to a cooler. After being cooled by air in the cooler duct, it is sorted through a sieve, and the larger particles are sent to the blast furnace as non-defective sinter ("product" in Figure 1). Smaller sinter (e.g., particles 4 mm or less) is returned to the sintering machine as return ore.

[0004] Here, the return ore ratio is defined as the proportion of return ore in the sintered ore after firing. Effective methods for reducing the return ore ratio and improving yield include promoting combustion to prevent sintered ore from passing through the sinter machine unfired. For example, increasing the blending ratio of the coagulant (heat source), increasing the blending ratio of quicklime (which acts as a binder during granulation) to improve permeability, and increasing the proportion of coke fines in the upper layer to facilitate ignition in the ignition furnace are effective. Reducing the pallet speed is also effective in ensuring sufficient firing time. However, there is a trade-off between reducing the return ore ratio and productivity: increasing the blending ratio of coagulant, quicklime, and coke fines increases production costs, while reducing the pallet speed reduces production volume. Therefore, it is necessary to control the return ore ratio by setting an appropriate target value depending on the operating conditions.

[0005] The return ore rate is measured in a process after sorting by sieving. Therefore, it takes time from the start of firing until the return ore rate is measured, for example, about two hours. Therefore, even if the return ore rate increases, it will not be detected until, for example, two hours later, and corrective action will be delayed by two hours. The same applies if the return ore rate decreases, and corrective action will be 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 action in advance. For example, Patent Document 1 discloses a method for predicting the return ore rate from the calcium ferrite content, slag content, pore size distribution index, and porosity, which are minerals constituting sintered ore.

[0007] Japanese Unexamined Patent Publication No. 7-11349

[0008] The method of Patent Document 1 requires sampling tests to identify physical properties, and is therefore unable to continuously predict the return ore rate, making it impossible to detect fluctuations in the return ore rate at an early stage.

[0009] An object of the present disclosure is to provide a return ore rate control method, a sintered ore manufacturing method, and a return ore rate control device that make it possible to predict the return ore rate of sintered ore and control it to a target value.

[0010] (1) A return ore rate control method according to an embodiment of the present disclosure is a method for controlling a sintering process executed in equipment including a sintering machine and a cooler, based on a predicted value of the return ore rate predicted using a return ore rate prediction model in which data related to at least the cooler is used as an explanatory variable and the return ore rate is used as a target variable, and the method includes: an acquisition step of acquiring a target value of the return ore rate and input data for the return ore rate prediction model, the input data including data related to the cooler, taking into consideration a delay time based on a timing at which the return ore rate can be identified; a return ore rate prediction step of predicting the return ore rate using the input data and the return ore rate prediction model; and an operation amount calculation step of calculating an operation amount of an operation variable for the sintering process so as to reduce a deviation between the predicted value and the target value of the return ore rate.

[0011] (2) As one embodiment of the present disclosure, in (1), the data related to the cooler includes at least one of a duct pressure and a blowing air pressure of the cooler.

[0012] (3) As one embodiment of the present disclosure, in (1) or (2), the input data includes data related to baking.

[0013] (4) As an embodiment of the present disclosure, in (3), the data regarding the firing is exhaust gas NO ｘ Concentration, exhaust gas O ２ The parameters include at least one of the concentration, the exhaust gas temperature, and the pallet speed.

[0014] (5) As one embodiment of the present disclosure, in any one of (1) to (4), the input data includes data related to raw materials to be charged into the sintering machine.

[0015] (6) As an embodiment of the present disclosure, in (5), the data on the raw materials includes at least one of a raw material brand, raw material moisture content, a water spray flow rate in a granulation mixer, a mixing ratio of a coagulant, a mixing ratio of quicklime, and a mixing ratio of iron ore.

[0016] (7) As an embodiment of the present disclosure, in any one of (1) to (6), the return ore rate prediction model is generated by machine learning using learning data in which data corresponding to the explanatory variables extracted from performance data in the sintering process and the return ore rate extracted from the performance data are associated with each other in consideration of the delay time.

[0017] (8) A method for producing sintered ore according to an embodiment of the present disclosure produces sintered ore using an manipulated variable calculated by any one of the return ore rate control methods (1) to (7).

[0018] (9) A return ore rate control device according to an embodiment of the present disclosure is a return ore rate control device that controls a sintering process executed in equipment including a sintering machine and the cooler, based on a predicted value of the return ore rate predicted using a return ore rate prediction model that uses data related to at least the cooler as an explanatory variable and the return ore rate as a target variable, and includes: an acquisition unit that acquires input data for the return ore rate prediction model that includes a target value of the return ore rate and data related to the cooler, taking into account a delay time based on a timing at which the return ore rate can be identified; a return ore rate prediction unit that predicts the return ore rate using the input data and the return ore rate prediction model; and an operation amount calculation unit that calculates an operation amount of an operation variable for the sintering process so as to reduce a deviation between the predicted value and the target value of the return ore rate.

[0019] (10) As an embodiment of the present disclosure, in (9), the method further includes a learning unit that evaluates an error of the return ore rate prediction model and performs additional learning or re-learning of the return ore rate prediction model in accordance with a predetermined standard.

[0020] According to the present disclosure, it is possible to provide a return ore rate control method, a sintered ore manufacturing method, and a return ore rate control device that make it possible to predict the return ore rate of sintered ore and control it to a target value.

[0021] FIG. 1 is a diagram showing an overview of a sintering process. FIG. 2 is a diagram showing errors in predicted values ​​of a return ore rate prediction model according to the number of explanatory variables. FIG. 3 is a diagram showing an example of the configuration of a return ore rate control device according to an embodiment. FIG. 4 is a flowchart showing a return ore rate control method according to an embodiment. FIG. 5 is a diagram showing results of Example 1. FIG. 6 is a diagram showing results of Example 2.

[0022] A return ore rate control method, a sintered ore manufacturing method, and a return ore rate control device according to an embodiment of the present disclosure will be described below with reference to the drawings. The return ore rate control method according to this embodiment, in summary, predicts the return ore rate using a machine learning model in which the pressure inside the cooler duct is included as an explanatory variable. Then, by calculating and outputting manipulated variables based on the predicted values ​​so as to reduce fluctuations in the return ore rate, it is possible to control the return ore rate to a target value or to implement appropriate operational actions.

[0023] The return ore ratio control method according to this embodiment utilizes the relationship that, as the return ore ratio increases in the duct of a cooler downstream of a sintering machine (i.e., the particle size of the sintered ore is finer), ventilation becomes worse, resulting in a decrease in the air suction pressure (increase in negative pressure) or an increase in the air blowing pressure when cooling the sintered ore in the cooler. In other words, at least "data related to the cooler" is used as an explanatory variable of the return ore ratio prediction model. The data related to the cooler includes at least one of the pressure inside the duct of the cooler and the air blowing pressure. The pressure inside the duct and the air blowing pressure are measured, for example, by a sensor provided in the cooler. In this embodiment, the data related to the cooler is described as the pressure inside the duct. The sensor for measuring the pressure inside the duct may be any sensor capable of measuring pressure, and for example, a strain gauge type, metal gauge type, semiconductor gauge type, or semiconductor diaphragm type pressure gauge may be used.

[0024] In the return ore rate control method according to this embodiment, a return ore rate prediction model is used to predict the return ore rate. The return ore rate prediction model uses at least data related to the cooler as explanatory variables and the return ore rate as a response variable. Then, based on the predicted value of the return ore rate, the sintering process executed in the equipment including the sinter machine and the cooler can be controlled. For example, the equipment can automatically or via an operator's operation take operational action to reduce fluctuations in the return ore rate during the sintering process.

[0025] Here, the return ore rate prediction model includes data related to the cooler as explanatory variables, but other operation factors may be used in combination. The explanatory variables of the return ore rate prediction model (i.e., input data for the return ore rate prediction model) may include "data related to calcination." The data related to calcination may include exhaust gas NO ｘ Concentration, exhaust gas O ２ The exhaust gas NO concentration may include at least one of exhaust gas temperature and pallet speed. ｘ Concentration, exhaust gas O ２ The NO concentration and exhaust gas temperature are the NO concentration and NO2 concentration of the exhaust gas generated during firing. ｘ Concentration, O ２ The pallet speed is the speed at which the pallet transports the raw material (pseudo particles) charged from the surge hopper. For example, if there is a lack of heat, the CO partial pressure will decrease and the NO in the exhaust gas will ｘ As the concentration increases, NO ｘ As the concentration increases, the return rate increases. ｘ It is possible to use concentration as an explanatory variable as well.

[0026] Furthermore, the input data for the return ore ratio prediction model may include "data related to raw materials" to be charged into the sinter machine. The data related to raw materials may include at least one of the raw material brand, raw material moisture content, water spray flow rate in the granulation mixer, coagulant blending ratio, quicklime blending ratio, and iron ore blending ratio. The raw material brand and raw material moisture content are the brand of iron ore and the amount of moisture in the raw material, respectively.

[0027] The return ore rate prediction model is not limited to a specific one as long as it is configured to obtain a response variable (return ore rate) from explanatory variables. The return ore rate prediction model may be, for example, a physical model or a machine learning model. In the present embodiment, the return ore rate prediction model is generated by machine learning. As a machine learning method, linear regression, neural network, decision tree, GBDT (Gradient Boosting Decision Tree), random forest, etc. can be used, and is not particularly limited. The return ore rate prediction model is generated, for example, using actual data (past measured values, set values, etc.) in the sintering process before the return ore rate is predicted.

[0028] FIG. 2 shows an example of evaluation of a return ore rate prediction model generated by a linear regression technique. The number (type) of explanatory variables in machine learning was gradually increased, and the error between the predicted value of the return ore rate by each machine learning model and the actual return ore rate was measured. For example, the prediction error by a return ore rate prediction model in which the only explanatory variable is the duct pressure is 1.4%. Here, the error was evaluated using the mean absolute error (MAE). In contrast, for example, when the explanatory variables are the duct pressure and flue gas NO , the error is 1.4%. ｘ Concentration, exhaust gas O ２ The prediction error using the return ore rate prediction model, which is the concentration and exhaust gas temperature, is 1.0%. As such, a tendency was observed for the error to decrease as the number of explanatory variables increased. Here, Figure 2 shows an example of the order in which explanatory variables are added. For example, the coagulation agent blending ratio may be added after the duct pressure as an explanatory variable. Regardless of the addition order in Figure 2, a tendency was observed for the error to decrease as the number of explanatory variables increased.

[0029] As mentioned above, various operational factors (operational variables for the sintering process) can be used as input data for the return ore rate prediction model, but the time it takes for the operational variables to affect the return ore rate varies. For example, the blending ratio of quicklime, which is data related to raw materials, is located upstream of the sintering process, so it can affect the return ore rate after three hours. For example, the NOx content of exhaust gas, which is data related to firing, can affect the return ore rate after three hours. ｘThe concentration is located midstream in the sintering process, so it can affect the return ore rate after two hours. For example, the duct pressure, which is data related to the cooler, is located downstream in the sintering process, so it can affect the return ore rate after one hour. Furthermore, the time until the return ore rate is affected can also vary depending on the pallet speed. Therefore, for accurate prediction, it is preferable that the input data for the return ore rate prediction model be acquired taking into account a delay time based on the timing at which the return ore rate can be determined. Here, the timing at which the return ore rate can be determined is, for example, when the process of sieving the sintered ore after cooling in the cooler and determining whether the particle size is below the standard is executed. The operating variables that affected the sintered ore at this determination are acquired taking into account the delay time. Similarly, it is preferable that the return ore rate prediction model be generated by machine learning using training data in which data corresponding to explanatory variables extracted from actual data in the sintering process and the return ore rate extracted from the actual data are matched taking into account the delay time.

[0030] A process for predicting the return ore rate is executed using the input data and the return ore rate prediction model as described above. Furthermore, manipulated variables for the manipulated variables of the sintering process are calculated so as to reduce the deviation between the predicted value of the return ore rate and the target value. In calculating the manipulated variables, a machine learning model using the manipulated variables as explanatory variables may be constructed, and the manipulated variables that make the predicted value of the return ore rate coincide with the target value may be calculated by inverse analysis.

[0031] Effective measures to reduce the return ore ratio include promoting combustion to prevent sintered ore from passing through the sinter machine unburned. For example, increasing the blending ratio of the agglomerating agent, which is a heat source, is considered. Other effective measures include increasing the blending ratio of quicklime, which acts as a binder during granulation to improve aeration, increasing the proportion of coke fines in the upper layer to facilitate ignition in the ignition furnace, and ensuring firing time by reducing the pallet speed. Effective measures to improve productivity when the return ore ratio is low include decreasing the blending ratio of the agglomerating agent, decreasing the blending ratio of quicklime, reducing the proportion of coke fines in the upper layer, and increasing the pallet speed. As mentioned above, there is a trade-off between reducing the return ore ratio and productivity. Increasing the blending ratios of agglomerating agent, quicklime, and coke fines increases production costs, while reducing the pallet speed reduces production volume. Therefore, the target value of the return ore ratio is preferably determined from the perspective of production cost and production volume, and it is preferable to maintain the return ore ratio at the target value. Here, the corrective action may be to determine a range of the manipulated variable's manipulated amount per operation and perform the operation within that range so as to prevent excessive operation (excessive manipulated amount).

[0032] Here, the calculated manipulated variables may be output so that they can be reflected by a process computer that manages the sintering process. Here, the output of the manipulated variables includes output as operation guidance for an operator who operates the sintering machine. In other words, the manipulated variables may be output so that appropriate manipulated variables for the manipulated variables are reflected in the sintering process through the judgment of the operator. The information output as operation guidance includes at least the optimal manipulated variables for the manipulated variables, and may be displayed on a display that can be seen by the operator.

[0033] FIG. 3 is a diagram showing an example of the configuration of the return ore rate control device 10 according to an embodiment. As shown in FIG. 3 , the return ore rate control device 10 includes a storage unit 11, an acquisition unit 12, a return ore rate prediction unit 13, an operation variable calculation unit 14, and an output unit 15. The return ore rate control device 10 may further include a learning unit 16. The return ore rate control device 10 acquires actual values ​​and target values ​​of the sintering process performed in the equipment including the sinter machine and the cooler from an operation data server 60. The actual values ​​may include various measured values ​​indicating the operational status and current operation variables. The target value is a target value for the return ore rate. The operation data server 60 is capable of communicating with the return ore rate control device 10 via a network and may be realized, for example, by a computer that manages the production of sintered ore. The network is, for example, the Internet. The return ore rate control device 10 performs the above-described process, i.e., predicts the return ore rate using the return ore rate prediction model, and performs a process of determining the operation variables for operation variables such as the duct pressure so that the future return ore rate is maintained near the target value. In this embodiment, the return ore rate control device 10 has a function of outputting the manipulated variable of the operation variable to the equipment or a function of presenting it as a guidance manipulated variable via the output unit 15. When the output unit 15 presents the guidance manipulated variable, the return ore rate control device 10 functions as an operation guidance device. The display unit 30 displays the guidance manipulated variable output from the return ore rate control device 10 (operation guidance device). The return ore rate control device 10 may be configured as a computer separate from the operation data server 60 (e.g., a process computer that manages the operation of the sintering machine or a sintering operation guidance server). The display unit 30 may be a display device such as a liquid crystal display (LCD) or an organic electroluminescence panel (OLED). The display unit 30 may also be realized by the display of a terminal device such as a smartphone or tablet. The terminal device is capable of communicating with the return ore rate control device 10 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 10 and a terminal device having the function of the display unit 30. The sintering operation guidance server and the terminal device may be located in the same place (for example, in the same factory) or may be physically separated from each other.The sintering operation guidance system may further include an operation data server 60 .

[0034] Here, the return ore rate prediction model may be generated by the return ore rate control device 10 and stored in the storage unit 11, or may be generated by another computer and stored in the storage unit 11. When the return ore rate control device 10 generates the return ore rate prediction model, it may further include a model generation unit that generates the return ore rate prediction model by the above-mentioned method and stores it in the storage unit 11. Here, the learning unit 16 may also function as the model generation unit.

[0035] The components of the return ore rate control device 10 are described below. The memory unit 11 stores a return ore rate prediction model. The memory unit 11 also stores programs and data related to return ore rate control. The memory unit 11 may store acquired actual values ​​and target values. The memory unit 11 may store various information obtained in the process for return ore rate control. The memory unit 11 may include any memory device, such as a semiconductor memory device, an optical memory device, or a magnetic memory device. The semiconductor memory device may include, for example, a semiconductor memory. The memory unit 11 may include multiple types of memory devices.

[0036] The acquisition unit 12 acquires input data for the return ore rate prediction model, including the target value of the return ore rate and data related to the cooler, taking into consideration a delay time based on the timing at which the return ore rate can be identified.

[0037] The return ore rate prediction unit 13 predicts the return ore rate using the input data and the return ore rate prediction model.

[0038] The manipulated variable calculation unit 14 calculates manipulated variables for the sintering process so as to reduce the deviation between the predicted value and the target value of the return ore rate.

[0039] The output unit 15 outputs the calculated manipulated variable to the equipment, or displays it on the display unit 30 as a guidance manipulated variable.

[0040] When the manipulated variable of the manipulated variable is output from the output unit 15 to the equipment for producing sintered ore, the equipment may automatically update the manipulated variable. That is, the return ore rate control method according to this embodiment may be executed as part of a production method for producing sintered ore. Furthermore, the operator may change the operating conditions of the sintering machine based on the guidance manipulated variable displayed on the display unit 30. Such operation guidance for the sintering machine may be executed as part of a production method for producing sintered ore.

[0041] The learning unit 16 evaluates the error of the return ore rate prediction model and performs additional learning or re-learning of the return ore rate prediction model in accordance with a predetermined criterion. The predetermined criterion is not limited to a specific value, and can be, for example, a threshold value described below. Additional learning is additional machine learning performed on the return ore rate prediction model using new added learning data. Re-learning is re-machine learning performed on the return ore rate prediction model using learning data used in previous machine learning (hereinafter referred to as "previously used learning data"). The return ore rate prediction model is updated by additional learning or re-learning. The learning unit 16 performs additional learning or re-learning of the return ore rate prediction model based on the evaluation, thereby maintaining the accuracy of the return ore rate prediction model. Furthermore, even if the return ore rate changes due to a change in the brand of raw material used, for example, the need to update the return ore rate prediction model can be determined by error evaluation, and the return ore rate prediction model can be updated.

[0042] The error (prediction error) of the return ore rate prediction model is the difference between the actual value and the predicted value of the return ore rate, and can be evaluated when the actual value of the return ore rate is obtained. The prediction error may be the mean absolute error as described above, or another index may be used. The learning unit 16 evaluates the performance of the return ore rate prediction model by monitoring the prediction error. When the prediction error becomes large due to recent data, the learning unit 16 can maintain the prediction accuracy of the return ore rate prediction model by additionally learning or relearning the prediction model using the recent data.

[0043] The learning unit 16 may, for example, monitor the standard deviation and average error of the prediction errors and set thresholds for each of the standard deviation and average error. The learning unit 16 may perform additional learning or relearning when the average error of the prediction errors for a predetermined period from the evaluation point in time (hereinafter, the "predetermined most recent past") exceeds the threshold, or when the standard deviation of the prediction errors for the predetermined most recent past exceeds the threshold. That is, the learning unit 16 may set the above threshold as a predetermined criterion and update the return ore ratio prediction model through additional learning using data from a period in which at least one of the standard deviation and average error of the prediction errors exceeds the threshold. Alternatively, the learning unit 16 may update the return ore ratio prediction model through relearning by adding data from a period in which at least one of the standard deviation and average error of the prediction errors exceeds the threshold to previously used learning data. Here, the evaluation period for the prediction error is set to a period including at least one cycle of data from when the return ore is discharged from the surge hopper to the sinter machine, sintering is performed, and a portion of the return ore is newly charged into the surge hopper as return ore. Furthermore, when a long-term evaluation is performed, the evaluation period for the prediction error may be set, for example, in units of one month.

[0044] As described above, the return ore rate control device 10 may be realized, for example, by a computer. The computer may include, for example, a memory, a hard disk drive (storage device), and a CPU (processing device). The program may be stored in the hard disk drive, and when executed by the CPU, it is read from the hard disk drive to the memory. Data during processing is stored in the memory, and, if necessary, stored in a hard disk drive (HDD). The memory unit 11 may be realized, for example, by a storage device. The acquisition unit 12, the return ore rate prediction unit 13, the operation amount calculation unit 14, the output unit 15, and the learning unit 16 may be realized, for example, by a CPU that reads and executes a program.

[0045] Fig. 4 is a flowchart showing a method for controlling the return ore rate according to one embodiment. The return ore rate control device 10 may calculate manipulated variables and output the manipulated variables as guidance manipulated variables according to the flowchart shown in Fig. 4. The return ore rate control method shown in Fig. 4 is also an operation guidance method, and may be executed as part of a method for producing sintered ore.

[0046] The acquisition unit 12 acquires actual values ​​and target values ​​(step S1, acquisition step). In this embodiment, the actual values ​​include data related to the cooler, specifically, the pressure inside the duct. The return ore rate prediction unit 13 predicts the return ore rate using the input data and a return ore rate prediction model (step S2, return ore rate prediction step). The manipulated variable calculation unit 14 calculates manipulated variables for the sintering process so as to reduce the deviation between the predicted value and the target value of the return ore rate (step S3, manipulated variable calculation step). The output unit 15 outputs the calculated manipulated variables for the manipulated variables (step S4).

[0047] Example 1 A specific example (Example) in which the optimal manipulated variable is determined by the above-described control method will be described below. In Example 1, in a sintering line, the return ore rate two hours into the future was predicted using a return ore rate prediction model, which is a linear model with the pressure inside the cooler duct as an explanatory variable. Assuming that operation was being performed with the return ore rate at the target value, an action to increase quicklime was taken when the predicted value of the return ore rate two hours into the future was higher than 15%. Furthermore, an action to decrease quicklime was taken when the predicted value of the return ore rate two hours into the future was less than 15% below the target value. The manipulated amount of quicklime per operation was fixed at 0.1%, and an action was taken every hour. As a result, as shown in FIG. 5 , the difference between the target and actual return ore rates (MAE: Mean Absolute Error) was 1.5% after 100 hours of operation. As a comparative example, in conventional operation, the difference between the target and actual return ore rates (MAE: Mean Absolute Error) was 3.1% after 100 hours of operation. In conventional operations, actions were taken based on the current measured return ore rate without using a model to obtain a predicted value. Here, it takes about one hour for the cooler duct pressure to affect the return ore rate. Therefore, when predicting the return ore rate two hours in the future as in this embodiment, the duct pressure one hour later from the time of prediction is required, but the prediction can be performed assuming that the duct pressure measured at the time of prediction will be maintained. This is because the duct pressure does not change frequently.

[0048] Example 2 In Example 2, the return ore ratio in a sintering line was predicted two hours in advance using a return ore ratio prediction model, which was a linear model with the cooler duct pressure and the quicklime blending ratio as explanatory variables. Unlike Example 1, a return ore ratio prediction model with additional explanatory variables was used. Other conditions were the same as in Example 1. As a result, as shown in FIG. 6 , the difference between the target and actual return ore ratios (MAE: Mean Absolute Error) after 100 hours of operation was 1.3%. It was demonstrated that the above-described control method makes it possible to predict the return ore ratio of sintered ore with high accuracy and control it to the target value. Here, it takes about one hour for the cooler duct pressure to affect the return ore ratio. Furthermore, it takes about three hours for the quicklime blending ratio to affect the return ore ratio. Therefore, when predicting the return ore ratio two hours in advance as in this example, data on the quicklime blending ratio one hour prior to the prediction time is used. Furthermore, as in Example 1, prediction can be performed assuming that the duct pressure is maintained at the measurement value at the time of prediction.

[0049] As described above, the return ore rate control method, sintered ore manufacturing method, and return ore rate control device 10 according to the present embodiment make it possible to predict the return ore rate of sintered ore and control it to a target value. Furthermore, the return ore rate control method, sintered ore manufacturing method, and return ore rate control device 10 according to the present embodiment can continuously predict the return ore rate and detect fluctuations early. As a result, the return ore rate can be maintained near the target value, and the yield in sintered ore production can be improved.

[0050] Although embodiments of the present disclosure have been described based on the drawings and examples, it should be noted that those skilled in the art would easily be able to make various modifications or alterations based on the present disclosure. Therefore, it should be noted that these modifications and alterations are included within the scope of the present disclosure. For example, functions included in each component or step can be rearranged so as not to cause logical inconsistencies, and multiple components or steps can be combined or divided into one. The embodiments of the present disclosure can also be realized as a program executed by a processor included in an apparatus or a storage medium on which a program is recorded. It should be understood that these are also included within the scope of the present disclosure.

[0051] The configuration of the return ore rate control device 10 shown in Fig. 3 is one example. The return ore rate control device 10 does not need to include all of the components shown in Fig. 3. Furthermore, the return ore rate control device 10 may include components other than those shown in Fig. 3. For example, the return ore rate control device 10 may be configured to further include a display unit 30.

[0052] REFERENCE SIGNS LIST 10 Return ore rate control device 11 Memory unit 12 Acquisition unit 13 Return ore rate prediction unit 14 Operation amount calculation unit 15 Output unit 16 Learning unit 30 Display unit 60 Operation data server

Claims

1. A return ore rate control method for controlling a sintering process performed in equipment including a sinter machine and the cooler, based on a predicted value of the return ore rate predicted using a return ore rate prediction model in which data related to at least the cooler is used as an explanatory variable and the return ore rate is used as a response variable, the method comprising: an acquisition step of acquiring input data for the return ore rate prediction model, the input data including a target value of the return ore rate and data related to the cooler, taking into account a delay time based on a timing at which the return ore rate can be identified; a return ore rate prediction step of predicting the return ore rate using the input data and the return ore rate prediction model; and an operation amount calculation step of calculating an operation amount of an operation variable for the sintering process so as to reduce a deviation between the predicted value of the return ore rate and the target value.

2. The method for controlling the return ore rate according to claim 1, wherein the data relating to the cooler includes at least one of the duct pressure and the blowing pressure of the cooler.

3. The method for controlling the return ore rate according to claim 1 or 2, wherein the input data includes data relating to firing.

4. The data on the firing is based on exhaust gas NO ｘ Concentration, exhaust gas O ２ The method for controlling a return ore rate according to claim 3, wherein the method includes at least one of concentration, exhaust gas temperature, and pallet speed.

5. A method for controlling a return ore rate according to any one of claims 1 to 4, wherein the input data includes data relating to raw materials to be charged into the sintering machine.

6. The method for controlling the return ore rate according to claim 5, wherein the data relating to the raw materials includes at least one of the raw material brand, raw material moisture content, water spray flow rate in the granulation mixer, coagulant blending ratio, quicklime blending ratio, and iron ore blending ratio.

7. A method for controlling a return ore rate according to any one of claims 1 to 6, wherein the return ore rate prediction model is generated by machine learning using learning data in which data corresponding to the explanatory variables extracted from performance data in the sintering process and the return ore rate extracted from the performance data are matched in consideration of the delay time.

8. A method for producing sintered ore, which produces sintered ore using the manipulated variable calculated by the method for controlling the return ore rate according to any one of claims 1 to 7.

9. A return ore rate control device that controls a sintering process performed in equipment including a sinter machine and the cooler, based on a predicted value of the return ore rate predicted using a return ore rate prediction model that uses data related to at least the cooler as explanatory variables and the return ore rate as a response variable, the return ore rate control device comprising: an acquisition unit that acquires input data for the return ore rate prediction model, including a target value of the return ore rate and data related to the cooler, taking into account a delay time based on a timing at which the return ore rate can be identified; a return ore rate prediction unit that predicts the return ore rate using the input data and the return ore rate prediction model; and an operation amount calculation unit that calculates an operation amount of an operation variable for the sintering process so as to reduce a deviation between the predicted value of the return ore rate and the target value.

10. A return ore rate control device according to claim 9, further comprising a learning unit that evaluates an error in the return ore rate prediction model and performs additional learning or re-learning of the return ore rate prediction model in accordance with a predetermined standard.

Images