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

Abstract

A return ore control device (10) that is to be used in a sintering production apparatus and that performs control related to return ore, the return ore control device including: an acquisition unit (12) that acquires, as input data, data related to operating conditions in the sintering production apparatus; a return ore hopper level prediction unit (13) that predicts a return ore hopper level after a predetermined time set in advance, based on the acquired input data; an operation amount calculation unit (14) that calculates an operation amount of an operation variable selected from the operating conditions and charging conditions of raw materials, based on the predicted return ore hopper level and a target range of return ore hopper level set in advance; and an output unit (15) that outputs the calculated operation amount to the sintering production apparatus or present the calculated operation amount as a guidance operation amount.

Description

TECHNICAL FIELD

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

[0002] In the ironmaking industry, the quality of iron ore has been declining due to years of mining. Accordingly, the use ratio of pulverized ore with a high pulverization ratio that has undergone ore dressing at a mine site has been increasing, thereby raising the importance of a sintering process of producing sintered ore by binding pulverized ore before it is charged into a blast furnace.

[0003] FIG. 1 is a view illustrating an overview of a sintering process. Sintered ore is produced by firing small particle-sized iron ore using combustion heat of cokes so that it can be used as a source of iron in a blast furnace. Raw materials for sintered ore are stored in a hopper and cut out therefrom. If pulverized iron ore is directly charged into a sintering machine, combustion reaction is weakened due to poor air permeability. For this reason, in a granulation process, iron ore is mixed with water, together with other materials, such as quicklime or coke, in a granulation mixer and processed into granulated materials having a larger particle size than the original raw materials. The granulated materials are charged into the sintering machine and ignited in an ignition furnace. Then, combustion reaction gradually proceeds in a layered manner from top to bottom due to air suction from below. The sintered ore after calcination is discharged from the sintering machine, crushed by a crusher, and sent to a cooler. After being cooled by air in a duct of the cooler, it is sorted by a sieve, and larger particle-sized sintered ore is regarded as a good product (indicated as "product" in FIG. 1) and sent to the blast furnace. Sintered ore having a small particle size (for example, a particle size of 4 mm or less) is returned as return ore to a return ore hopper, cut out from the return ore hopper, and fed into the sintering machine again.

[0004] Here, a returned ore rate is defined as the proportion of return ore to sintered ore after calcination. To reduce the returned ore rate and improve the yield, it is effective to promote combustion so as to prevent sintered ore from passing through the sintering machine without being calcined. For example, it is effective to increase the blend ratio of cokes as heat sources, to improve air permeability by increasing the blend ratio of quicklime, which acts as a binder during granulation, and to increase the proportion of pulverized coke on a top layer to facilitate ignition in the ignition furnace. Reducing pallet speed to secure calcination time is also effective. On the other hand, there is a trade-off relationship between reducing the returned ore rate and productivity. That is, increasing the blend ratios of cokes, quicklime, and pulverized coke raises production cost, and reducing pallet speed decreases productivity. It is therefore necessary to set an appropriate target value and control the returned ore rate depending on operating conditions.

[0005] The returned ore rate is measured in a process after sorting using a sieve. Accordingly, it takes time from the start of calcination until the returned ore rate is measured, and approximately two hours are required in one example. Furthermore, it takes additional time (for example, approximately 0.5 hours) until the return ore is fed into the return ore hopper. Thus, even when the returned ore rate increases, detection occurs for example two hours later, delaying corrective actions by two hours. The same applies when the returned ore rate decreases, and corrective actions are delayed.

[0006] To detect and control fluctuations in the returned ore rate at an early stage, it is necessary to predict the future returned ore rate and take action in advance. For example, Patent Literature (PTL) 1 describes a method of predicting the amount of returned ore based on the content of calcium ferrite, which is a constituent mineral of sintered ore, the content of slag, pore size distribution index, and porosity.CITATION LISTPatent Literature

[0007] PTL 1: JP H07-011349 ASUMMARY(Technical Problems)

[0008] Here, the method of PTL 1 requires an extraction test for identifying physical properties and cannot continuously predict the returned ore rate. It is therefore not possible to detect fluctuations in the returned ore rate at an early stage.

[0009] Furthermore, due to variations in raw material components or calcination conditions, for example, the returned ore rate fluctuates with a relatively short cycle of approximately one hour. Accordingly, in a case in which an operation, such as increasing the blend ratio of quicklime, is performed so as to reduce the returned ore rate, for example at the timing when the returned ore rate temporarily increases, it may result in an excessive increase in the blend ratio of quicklime, thereby increasing production cost. Similarly, in a case in which an operation, such as decreasing the blend ratio of quicklime, is performed so as to increase the returned ore rate, for example at the timing when the returned ore rate temporarily decreases, it may result in an excessive decrease in the blend ratio of quicklime, thereby increasing the returned ore rate and reducing productivity. Moreover, in a case in which an attempt is made to correct fluctuations in the returned ore rate with a one-hour cycle, it leads to frequent operational actions, which may destabilize the furnace condition. In the following, such operational actions that respond to temporary increases or decreases in the returned ore rate and excessively increase or decrease the blend ratio of quicklime are referred to as "excess actions."

[0010] In view of such circumstances, an objective of the present disclosure is to provide a return ore control device, a return ore control method, and a sintered ore production method that can reduce excess actions and prevent an increase in production cost.(Solution to Problem)

[0011] As a result of intensive studies by the present inventors to solve the above problems, it has been found that controlling a return ore hopper level, instead of directly controlling the returned ore rate, is effective in preventing excess actions. Because the returned ore rate in response to fluctuations in the returned ore rate is accumulated and reflected in the return ore hopper level, the returned ore rate can be indirectly managed, by controlling the return ore hopper level within a target range. Furthermore, because the fluctuations in the amount of returned ore are evaluated based on an accumulated value, the influence of temporary increases or decreases in the returned ore rate can be reduced. (1) A return ore control device according to an embodiment of the present disclosure is a return ore control device that is to be used in a sintering production apparatus and that performs control related to return ore, the return ore control device including: an acquisition unit configured to acquire, as input data, data related to operating conditions in the sintering production apparatus; a return ore hopper level prediction unit configured to predict a return ore hopper level after a predetermined time set in advance, based on the acquired input data; an operation amount calculation unit configured to calculate an operation amount of an operation variable selected from the operating conditions and charging conditions of raw materials, based on the predicted return ore hopper level and a target range of return ore hopper level set in advance; and an output unit configured to output the calculated operation amount to the sintering production apparatus or present the calculated operation amount as a guidance operation amount. (2) As an embodiment of the present disclosure, in (1), the input data includes a measured value of return ore hopper level and a returned ore cutout amount, and further includes at least one of raw material grade, raw material moisture, sprinkling flow rate in a granulation mixer, blend ratio of cokes, blend ratio of quicklime, blend ratio of iron ore, blend ratio of return ore, exhaust gas NO x concentration, exhaust gas CO concentration, exhaust gas CO 2 concentration, exhaust gas O 2 concentration, exhaust gas temperature, cooler blast pressure, pallet speed, layer thickness, and production amount. (3) As an embodiment of the present disclosure, in (1) or (2), the return ore hopper level prediction unit is configured to predict the return ore hopper level using a prediction model, and the prediction model is generated by machine learning using training data obtained by associating data corresponding to the input data extracted from actual data in a sintering process, and the return ore hopper level extracted from the actual data, with consideration given to a delay time. (4) A return ore control method according to an embodiment of the present disclosure is a return ore control method that is to be used in a sintering production apparatus and that performs control related to return ore, the return ore control method including: an acquisition step of acquiring, as input data, data related to operating conditions in the sintering production apparatus; a return ore hopper level prediction step of predicting a return ore hopper level after a predetermined time set in advance, based on the acquired input data; an operation amount calculation step of calculating an operation amount of an operation variable selected from the operating conditions and charging conditions of raw materials, based on the predicted return ore hopper level and a target range of return ore hopper level set in advance; and an output step of outputting the calculated operation amount to the sintering production apparatus or presenting the calculated operation amount as a guidance operation amount. (5) A sintered ore production method according to an embodiment of the present disclosure includes producing sintered ore using the operation amount output to the sintering production apparatus by the return ore control method according to (4). (Advantageous Effect)

[0012] According to the present disclosure, the return ore control device, the return ore control method, and the sintered ore production method that can reduce excess actions and prevent an increase in production cost can be provided.BRIEF DESCRIPTION OF THE DRAWINGS

[0013] In the accompanying drawings: FIG. 1 is a view illustrating an overview of a sintering process; FIG. 2 is a view illustrating errors of predicted values of a prediction model according to the number of explanatory variables; FIG. 3 is a view illustrating an example configuration of a return ore control device according to an embodiment of the present disclosure; FIG. 4 is a flowchart illustrating a return ore control method according to an embodiment of the present disclosure; FIG. 5 is a view illustrating results of Example 1; and FIG. 6 is a view illustrating results of Example 2. DETAILED DESCRIPTION

[0014] Hereinafter, a return ore control device, a return ore control method, and a sintered ore production method according to an embodiment of the present disclosure will be described with reference to the drawings. In an outline, the return ore control method according to the present embodiment predicts the return ore hopper level in real time using a machine learning model. The machine learning model is a trained model generated through machine learning. Based on a predicted value, operation amounts are calculated and output so that the return ore hopper level falls within a target range, thereby allowing the return ore hopper level to be controlled within an appropriate range or allowing appropriate operational actions to be implemented.

[0015] In the return ore control method according to the present embodiment, the return ore hopper level is predicted using a prediction model. As illustrated in FIG. 1, the return ore hopper level refers to the amount of returned ore returned and fed into a return ore hopper, and may be expressed, for example, as a percentage (%) with respect to the maximum amount of the return ore hopper as in the present embodiment. The return ore hopper level is not limited to a percentage, and it may be expressed, for example, as a height (height level) from a bottom in the return ore hopper, a volume of the returned ore, or the like.

[0016] The return ore hopper level is estimated at each control cycle of return ore hopper level, by predicting an increase / decrease amount of return ore hopper level at a future time set in advance, based on the current return ore hopper level. The increase / decrease amount can be predicted using the following formula (1), based on a predicted value Y p of returned ore generation from the current time to the future time set in advance and a returned ore cutout amount u from the hopper. H p t + k = H t + Y p t + k − u t + k

[0017] Here, H (t) is a measured value of return ore hopper level at time t. When it is assumed that time t is the current time, H (t) corresponds to the current return ore hopper level. H p (t + k) is a predicted value of return ore hopper level at future time t + k, which is time k ahead of time t, wherein time k is set in advance. Y p (t + k) is a predicted value of returned ore generation until future time t + k, which is time k ahead of time t, wherein time k is set in advance. u (t + k) is the returned ore cutout amount from the return ore hopper until future time t + k, which is time k ahead of time t, wherein time k is set in advance. Here, the returned ore cutout amount u (t + k) from the return ore hopper may be calculated by assuming that an actual value of returned ore cutout amount at time t is maintained. The returned ore cutout amount is not frequently modified or changed. It is therefore often acceptable to assume that the current actual value is maintained without significant degradation in prediction accuracy. Even when the returned ore cutout amount is modified or changed, the modification or the change is reflected in the next control cycle after it is modified or changed and does not have a significant influence on the prediction accuracy.

[0018] The prediction model may be used to calculate the predicted value Y p (t + k), which is returned ore generation from time t to future time t + k. Here, in the following, Y p (t + k) may be omitted and referred to simply as the "predicted value of returned ore generation." Furthermore, time k is determined in accordance with an apparatus configuration or material transport time. For example, time k can be set by considering the time from the start of calcination until the returned ore rate is measured (for example, two hours), or the time until the return ore is fed into the return ore hopper (for example, 0.5 hours).

[0019] As input data to the prediction model, data related to operating conditions in an apparatus (sintering production apparatus), which includes a sintering machine, the return ore hopper, or the like are input. In the prediction model, these input data are used as explanatory variables, and returned ore generation from time t to time t + k is used as a target variable. Based on a predicted value of returned ore generation, the return ore hopper level at the prediction time, and a returned ore cutout amount from the return ore hopper, a predicted value of return ore hopper level is obtained. Based on the predicted value of return ore hopper level, a sintering process to be executed in the sintering production apparatus can be controlled. For example, the sintering production apparatus may implement operational actions automatically or through operator operations so that the return ore hopper level falls within a target range set in advance in the sintering process.

[0020] The input data for the prediction model of returned ore generation may include at least one of raw material grade [id], raw material moisture [%], sprinkling flow rate in a granulation mixer [ton / hr], blend ratio of cokes [%], blend ratio of quicklime [%], blend ratio of iron ore [%], blend ratio of return ore [%], exhaust gas NO x concentration [%], exhaust gas CO concentration [%], exhaust gas CO 2 concentration [%], exhaust gas O 2 concentration [%], exhaust gas temperature [°C], cooler blast pressure [kPa], pallet speed [m / min], layer thickness [mm], and production amount [ton / hr]. For example, the input data may include all of these features, and may further include other operational factors. Here, raw material grade and raw material moisture respectively indicate the grade of iron ore in raw materials and the moisture content in the raw materials. Raw material grades are classified into a plurality of groups based on raw material component compositions, assigned a label corresponding to each group, and treated as weight data for those labels. Furthermore, raw material moisture is indicated as a weight ratio of the water content in the raw materials. As described above, iron ore is mixed with water, together with other raw materials, such as quicklime or coke, in the granulation mixer, and sprinkling flow rate indicates the flow rate of water sprinkled during mixing. Blend ratio of the cokes, blend ratio of quicklime, blend ratio of iron ore, and blend ratio of return ore respectively represent the blend ratios of cokes, quicklime, iron ore, and return ore in the raw materials for the sintered ore. Exhaust gas NO x concentration, exhaust gas CO concentration, exhaust gas CO 2 concentration, exhaust gas O 2 concentration, exhaust gas temperature indicate the concentrations and the temperature of respective components in the exhaust gas generated during calcination. Cooler blast pressure indicates the pressure of air sent by the cooler during cooling of the sintered ore. Pallet speed indicates the transport speed of the pallet transporting the raw materials during calcination. Layer thickness indicates the thickness of raw material layer on the pallet. Moreover, production amount indicates the amount of sintered ore produced. Production amount may specifically be determined based on the weight and the transport speed measured by a Merrick type conveyor scale or the like installed on an exit side of the sintering machine.

[0021] Here, among the aforementioned features, cooler blast pressure, exhaust gas NO x concentration, blend ratio of return ore, and production amount are particularly important. The input data are preferably configured to include at least cooler blast pressure, exhaust gas NO x concentration, blend ratio of return ore, and production amount. As the returned ore rate increases, the particle size of the sintered ore becomes finer, and permeability deteriorates, and when the sintered ore is cooled in the cooler, air suction pressure decreases (negative pressure increases). Accordingly, the higher the cooler blast pressure, the more the return ore hopper level increases. Here, a sensor for measuring cooler blast pressure may be any device capable of measuring pressure, and for example, a strain gauge-type, a metal gauge-type, a semiconductor gauge-type, or a semiconductor diaphragm-type pressure gauge can be used. Furthermore, when heat is insufficient, the CO partial pressure decreases, and the exhaust gas NO x concentration increases. Accordingly, as the exhaust gas NO x concentration increases, the returned ore rate increases, and the return ore hopper level increases. Likewise, as the production amount increases, the amount of returned ore increases, and the return ore hopper level increases. Conversely, as the blend ratio of return ore increases, the returned ore cutout amount increases, thus causing the return ore hopper level to decrease.

[0022] The prediction model is not limited to any specific type, provided it is configured to obtain a target variable (returned ore generation) from explanatory variables. The prediction model may, for example, be a physical model or a machine learning model. In the present embodiment, the prediction model is generated by machine learning. As machine learning techniques, linear regression, neural network, decision tree, Gradient Boosting Decision Tree (GBDT), random forest, or the like may be used, without particular limitation. Here, the prediction model may also be a model that directly predicts the return ore hopper level. In this case, in addition to the explanatory variables for the prediction model of returned ore generation, measured values of return ore hopper level and returned ore cutout amounts are required as input data to construct the trained model. Furthermore, by using the hierarchical Bayesian estimation technique, a model may be constructed that estimates the future return ore hopper level based on a measured value of return ore hopper level using a returned ore cutout amount and the explanatory variables for the prediction model of returned ore generation. The prediction model is generated, before performing prediction of the return ore hopper level, using, for example, actual data (past measured values, past set values, or the like) in a sintering process.

[0023] FIG. 2 illustrates an example of evaluating prediction models generated by the neural network method. The number (types) of explanatory variables in the machine learning model is increased, and for each machine learning model, the error between a predicted value of returned ore generation and actual returned ore generation is measured. For example, the prediction error of a prediction model with cooler blast pressure as the sole explanatory variable is 7.3 %. Here, the error is evaluated using the Mean Absolute Error (MAE). In contrast, for a prediction model wherein the explanatory variables are cooler blast pressure, exhaust gas NO x concentration, production amount, and blend ratio of return ore, the prediction error is 3.2 %. Thus, there is a tendency for the error to decrease as the number of explanatory variables increases. FIG. 2 here illustrates one example of the order in which explanatory variables are added, and for example, blend ratio of return ore may be added after cooler blast pressure as explanatory variables. Regardless of the addition order illustrated in FIG. 2, there is a tendency for the error to decrease as the number of explanatory variables increases.

[0024] Although various operational factors (operation variables for the sintering process) can be used as input data for the prediction model, the time until each operation variable influences the return ore hopper level, that is, the time until it affects the returned ore generation fed into the return ore hopper, differs. For example, blend ratio of quicklime, which corresponds to data related to raw materials, is located upstream of the sintering process, and may influence the return ore hopper level 3.5 hours later. For example, exhaust gas NO x concentration, which corresponds to data related to calcination, is located midstream of the sintering process, and may influence the return ore hopper level 2.5 hours later. For example, cooler blast pressure, which corresponds to data related to the cooler, is located downstream of the sintering process, and may influence the return ore hopper level 1.5 hours later. Moreover, the time until influencing the returned ore generation may also vary depending on the pallet speed. Accordingly, for accurate prediction, the input data for the prediction model is preferably acquired, with consideration given to a delay time based on the timing when the return ore hopper level can be identified. Here, the timing when the return ore hopper level can be identified is, for example, the time when, after cooling in the cooler and classifying with a sieve to determine that the particle size is below the standard, processing is executed to transport and feed the return ore into the return ore hopper. Operation variables that affect the sintered ore at that specific timing are acquired, with consideration given to the aforementioned delay time. Similarly, the prediction model is preferably generated by machine learning using training data obtained by associating data corresponding to explanatory variables (input data) extracted from actual data in the sintering process and return ore hopper levels extracted from the actual data while considering a delay time.

[0025] Using the input data that include the measured values of return ore hopper level and the returned ore cutout amounts and the prediction model as described above, processing of predicting the return ore hopper level is executed. In order to ensure that corrective actions are not delayed, the return ore hopper level after a predetermined time set in advance (for example, 2.5 hours) is predicted. Furthermore, so that a predicted value of return ore hopper level falls within the target range, operation amounts of operation variables for the sintering process are calculated at each control cycle set separately (for example, every hour). In a case in which the return ore hopper level is higher than the upper limit of the target range, an action to lower the returned ore rate is performed, in a case in which it is lower than the lower limit of the target range, an action to raise the returned ore rate is performed, and otherwise, actions are deferred, thereby preventing excess action. In calculating the operation amounts, a machine learning model that uses operation amounts as explanatory variables may be constructed, and operation amounts that make the predicted value of return ore hopper level match a target value (for example, the median value of the target range) may be calculated by inverse analysis.

[0026] As an action to lower the return ore hopper level, it is effective to promote combustion so as to prevent sintered ore from passing through the sintering machine without being calcined. For example, increasing the blend ratio of cokes as heat sources is an option. Furthermore, increasing the blend ratio of quicklime, which acts as a binder during granulation, so as to improve permeability, increasing the proportion of pulverized coke in the upper layer so as to facilitate ignition in the ignition furnace, and decreasing pallet speed so as to secure calcination time are also effective. As actions to raise the return ore hopper level, decreasing the blend ratio of cokes, decreasing the blend ratio of quicklime, decreasing the proportion of pulverized coke in the upper layer, and increasing the pallet speed are effective. Here, as described above, reduction of the return ore hopper level and productivity are in a trade-off relationship. That is, increasing the blend ratios of cokes, quicklime, or pulverized coke raises production cost, and reducing the pallet speed decreases the production amount. Accordingly, the target range of return ore hopper level is preferably determined in view of production cost and production amount, and it is preferable that the return ore hopper level be maintained within the target range. Here, to prevent excessive operations (excessive operation amounts) in corrective actions, a range of operation amounts of operation variables per operation may be determined so that operations are performed within that range.

[0027] Here, the calculated operation amounts of the operation valuables may be output so that a process computer that manages the sintering process can reflect them. Here, the output of operation amounts includes output as operating guidance to an operator who operates the sintering machine. That is, the operation amounts may be output so that appropriate operation amounts of operation variables are reflected through the judgment of the operator in the sintering process. The information output as the operating guidance includes at least the calculated operation amounts of operation variables, and may be displayed on a display that the operator can view.

[0028] FIG. 3 is a view illustrating a configuration of a return ore control device 10 according to the present embodiment. The return ore control device 10 is used in the sintering production apparatus and performs control related to return ore, including the aforementioned return ore control method. As illustrated in FIG. 3, the return ore control device 10 includes a storage unit 11, an acquisition unit 12, a return ore hopper level prediction unit 13, an operation amount calculation unit 14, and an output unit 15. The return ore control device 10 acquires, from an operation data server 60, data related to operating conditions in the sintering production apparatus, that is, the aforementioned input data. The input data includes, in addition to the measured values of return ore hopper level and the returned ore cutout amounts from the return ore hopper, actual values of features, such as cooler blast pressure, exhaust gas NO x concentration, blend ratio of return ore, or production amount. The actual values include measured values and set values of operation variables. Furthermore, the return ore control device 10 may acquire the target range of return ore hopper level from the operation data server 60. The operation data server 60 can communicate with the return ore control device 10 via a network, and may be realized, for example, by a computer that manages the production of sintered ore. The network may be, for example, the Internet. The return ore control device 10 executes the aforementioned processing, that is, the processing of predicting the return ore hopper level using the prediction model and obtaining operation amounts of operation variables, such as cooler blast pressure, so that the future return ore hopper level is maintained within the target range. Further, in the present embodiment, the return ore control device 10 includes the function, performed by the output unit 15, of outputting the operation amounts to the sintering production apparatus or presenting them as guidance operation amounts. In a case in which the output unit 15 presents guidance operation amounts, the return ore control device 10 functions as an operating guidance device. A display unit 30 displays the guidance operation amounts output from the return ore control device 10 (operating guidance device). The return ore control device 10 may be configured by a computer (for example, a process computer that manages the operation of the sintering machine or a sintering operation guidance server) that is separate from the operation data server 60. The display unit 30 may be a display device, such as a Liquid Crystal Display (LCD) or an Organic Electro-Luminescence Panel (OLED). The display unit 30 may also be implemented by the display of a terminal device, such as a smartphone or a tablet. The terminal device can communicate with the return ore control device 10 via a network. A sintering operation guidance system may be configured by the sintering operation guidance server having the functions of the return ore control device 10 and the terminal device having the functions of the display unit 30. The sintering operation guidance server and the terminal device may be located in the same place (for example, within the same plant) or may be arranged physically away. The sintering operation guidance system may be configured to further include the operation data server 60.

[0029] Here, the prediction model may be generated by the return ore control device 10 and stored in the storage unit 11, or it may be generated by another computer and stored in the storage unit 11. In a case in which the return ore control device 10 generates the prediction model, it may further include a model generation unit that generates the prediction model using the aforementioned techniques and stores it in the storage unit 11.

[0030] In the following, the components of the return ore control device 10 will be described. The storage unit 11 stores the prediction model. The storage unit 11 also stores programs and data related to return ore hopper level control. The storage unit 11 may store the acquired input data. The storage unit 11 may store the target range of return ore hopper level. The storage unit 11 may store various types of information obtained by processing for return ore hopper level control. The storage 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, semiconductor memory. The storage unit 11 may include a plurality of types of memory devices.

[0031] The acquisition unit 12 acquires the input data. The acquisition unit 12 preferably acquires features of the input data, with consideration given to a delay time based on the timing when the return ore hopper level can be identified.

[0032] The return ore hopper level prediction unit 13 predicts the return ore hopper level after a predetermined time set in advance, based on the acquired input data. The prediction model is used for predicting the return ore hopper level.

[0033] The operation amount calculation unit 14 calculates operation amounts selected from the operating conditions and raw material charging conditions, based on the predicted return ore hopper level and the target range of return ore hopper level set in advance. The raw material charging conditions include, for example, blend ratios of cokes, quicklime, iron ore, or return ore.

[0034] The output unit 15 outputs the calculated operation amounts to the sintering production apparatus or presents them to the display unit 30 as guidance operation amounts.

[0035] In a case in which the operation amounts of operation variables are output from the output unit 15 to the sintering production apparatus, the sintering production apparatus may automatically update the operation variables to the output operation amounts and produce sintered ore. That is, the return ore control method according to the present embodiment may be executed as part of a production method of producing sintered ore. The operator may also change the operating conditions of the sintering machine, based on the guidance operation amounts displayed on the display unit 30. In a case in which the return ore hopper level is likely to exceed the target range set in advance, the operator may perform an action to lower the returned ore rate and thereby lower the return ore hopper level. In a case in which the return ore hopper level falls below the target range, the operator may perform an action to raise the returned ore rate and thereby raise the return ore hopper level. Such operating guidance for the sintering machine may be executed as part of the production method of producing sintered ore.

[0036] The return ore control device 10 can be implemented, for example, by a computer as mentioned above. The computer may include, for example, a memory, a hard disk drive (storage device), a Central Processing Unit (CPU), or the like. Programs may be stored on the hard disk drive and, when executed by the CPU, are read into memory from the hard disk drive. Data during processing may be stored in memory and, if necessary, in the Hard Disk Drive (HDD). The storage unit 11 may be implemented, for example, by a storage device. The acquisition unit 12, the return ore hopper level prediction unit 13, the operation amount calculation unit 14, and the output unit 15 may be realized by a CPU that reads and executes programs, for example.

[0037] FIG. 4 is a flowchart illustrating the return ore control method according to the present embodiment. The return ore control device 10 may calculate operation amounts of operation variables and output them as guidance operation amounts in accordance with the flowchart of FIG. 4. The return ore control method illustrated in FIG. 4 is also an operating guidance method and may be executed as part of the sintered ore production method.

[0038] The acquisition unit 12 acquires input data (Step S1, acquisition step). The return ore hopper level prediction unit 13 predicts the return ore hopper level after the predetermined time using the input data and the prediction model (Step S2, return ore hopper level prediction step). The operation amount calculation unit 14 calculates operation amounts of operation variables so that the predicted value of return ore hopper level falls within the target range (Step S3, operation amount calculation step). The output unit 15 outputs the calculated operation amounts of operation variables (Step S4, output step).EXAMPLES(Example 1)

[0039] Concrete example (Examples) in which an optimal operation amount of operation variable is determined by the aforementioned control method will be described below. In Example 1, on the sintering line, returned ore generation 2.5 hours ahead was predicted using a neural network machine learning model, and the return ore hopper level was predicted based on a measured value of return ore hopper level and a returned ore cutout amount. The input data were raw material grade, raw material moisture, sprinkling flow rate in the granulation mixer, blend ratio of cokes, blend ratio of quicklime, blend ratio of iron ore, blend ratio of return ore, exhaust gas NO x concentration, exhaust gas CO concentration, exhaust gas CO 2 concentration, exhaust gas O 2 concentration, exhaust gas temperature, cooler blast pressure, pallet speed, layer thickness, and production amount. The target range of the return ore hopper level was set to 40 % to 60 %. When the return ore hopper level fell below 40 %, an action to reduce quicklime was performed. When the return ore hopper level exceeded 60 %, an action to increase quicklime was performed. The operation amount of quicklime was fixed at 0.1 % per action, and the operation (action) was performed every hour. As illustrated in FIG. 5, under operation with conventional control (direct control of returned ore rate), usage rates of quicklime (blend ratio of quicklime) averaged 1.5 % over 100 hours of operation. By applying the method according to the present embodiment, usage rates of quicklime averaged 1.3 % over 100 hours. In the present Example, the method according to the present embodiment did not cause excessive increases in the blend ratio of quicklime. That is, without causing excess action, the blend ratio of quicklime was restrained, and an increase in production cost was prevented.(Example 2)

[0040] In Example 2, on the sintering line, returned ore generation 2.5 hours ahead was predicted using the neural network machine learning model, and the return ore hopper level was predicted based on a measured value of return ore hopper level and a returned ore cutout amount. The input data were the same as in Example 1. The target range of the return ore hopper level was set to 40 % to 60 %. When the return ore hopper level fell below 40 %, an action to reduce quicklime was performed. When the return ore hopper level exceeded 60 %, an action to increase quicklime was performed. The operation amount of quicklime was determined by inverse analysis, based on calculation results of the machine learning model. Although the inverse analysis is not limited to a particular method, in the present Example, calculation was performed a plurality of times with varied input data, and an optimal operation amount of quicklime was searched using dichotomy. The operation (action) was performed every hour. As illustrated in FIG. 6, under operation with conventional control (direct control of returned ore rate), usage rates of quicklime (blend ratio of quicklime) averaged 1.5 % over 100 hours. By applying the method according to the present embodiment, usage rates of quicklime averaged 1.2 % over 100 hours. In the present Example, the method according to the present embodiment did not cause excessive increases in the blend ratio of quicklime. That is, without causing excess action, the blend ratio of quicklime was restrained, and an increase in production cost was prevented.

[0041] As described above, the return ore control device 10, the return ore control method, and the sintered ore production method according to the present embodiment can reduce excess actions and prevent an increase in production cost, by controlling the return ore hopper level, as is clear from the Examples on the above.

[0042] While an embodiment according to the present disclosure has been described with reference to the drawings and examples, it is to be noted that various revisions and modifications may be implemented by those skilled in the art based on the present disclosure. Accordingly, such revisions and modifications are included within the scope of the present disclosure. For example, functions or the like included in each component, each step, or the like can be rearranged without logical inconsistency, and a plurality of components, steps, or the like can be combined together or divided. An embodiment according to the present disclosure can also be implemented as a program executed by a processor included in the apparatus, or as a storage medium in which the program is recorded. It is be understood that these are also included in the scope of the present disclosure.

[0043] The configuration of the return ore control device 10 illustrated in FIG. 3 is one example. The return ore control device 10 does not need to include all the components of FIG. 3. Furthermore, the return ore control device 10 may include components other than those of FIG. 3. For example, the return ore control device 10 may further include the display unit 30.

[0044] Moreover, in the above embodiment, it is described that the prediction model may be a "prediction model of returned ore generation (first prediction model)" or a "model that directly predicts the return ore hopper level (second prediction model)." Here, both models may be considered to have no substantial difference in terms of obtaining a predicted value of return ore hopper level. The second prediction model, however, is constructed further using measured values of return ore hopper level and returned ore cutout amounts as input data as described above, and it can be provided with dynamic characteristics that include a delay response in the behavior of the return ore hopper level. For example, in a case in which there is a temporary return ore buffer in the actual transport path of return ore, the construction of a model that includes a delay response may be expected to further improve prediction accuracy.REFERENCE SIGNS LIST

[0045] 10Return ore control device 11Storage unit 12Acquisition unit 13Return ore hopper level prediction unit 14Operation amount calculation unit 15Output unit 30Display unit 60Operation data server

Claims

1. A return ore control device that is to be used in a sintering production apparatus and that performs control related to return ore, the return ore control device comprising: an acquisition unit configured to acquire, as input data, data related to operating conditions in the sintering production apparatus; a return ore hopper level prediction unit configured to predict a return ore hopper level after a predetermined time set in advance, based on the acquired input data; an operation amount calculation unit configured to calculate an operation amount of an operation variable selected from the operating conditions and charging conditions of raw materials, based on the predicted return ore hopper level and a target range of return ore hopper level set in advance; and an output unit configured to output the calculated operation amount to the sintering production apparatus or present the calculated operation amount as a guidance operation amount.

2. The return ore control device according to claim 1, wherein the input data includes a measured value of return ore hopper level and a returned ore cutout amount, and further includes at least one of raw material grade, raw material moisture, sprinkling flow rate in a granulation mixer, blend ratio of cokes, blend ratio of quicklime, blend ratio of iron ore, blend ratio of return ore, exhaust gas NOx concentration, exhaust gas CO concentration, exhaust gas CO2 concentration, exhaust gas O2 concentration, exhaust gas temperature, cooler blast pressure, pallet speed, layer thickness, and production amount.

3. The return ore control device according to claim 1 or 2, wherein the return ore hopper level prediction unit is configured to predict the return ore hopper level using a prediction model, and the prediction model is generated by machine learning using training data obtained by associating data corresponding to the input data extracted from actual data in a sintering process, and the return ore hopper level extracted from the actual data, with consideration given to a delay time.

4. A return ore control method that is to be used in a sintering production apparatus and that performs control related to return ore, the return ore control method comprising: an acquisition step of acquiring, as input data, data related to operating conditions in the sintering production apparatus; a return ore hopper level prediction step of predicting a return ore hopper level after a predetermined time set in advance, based on the acquired input data; an operation amount calculation step of calculating an operation amount of an operation variable selected from the operating conditions and charging conditions of raw materials, based on the predicted return ore hopper level and a target range of return ore hopper level set in advance; and an output step of outputting the calculated operation amount to the sintering production apparatus or presenting the calculated operation amount as a guidance operation amount.

5. A sintered ore production method comprising producing sintered ore using the operation amount output to the sintering production apparatus by the return ore control method according to claim 4.

Images