Formulation planning support method

Abstract

To estimate the content of components in molten steel produced in a furnace operation in which molten steel is produced by melting a raw material in which a plurality of scrap brands are blended in an arbitrary blending ratio, and to support planning of blending of a plurality of scrap brands.SOLUTION: In a furnace operation for generating molten steel by melting a raw material in which a plurality of scrap brands are blended in an arbitrary blending ratio, the blending plan support method supports the planning of a blending plan in which a blending ratio of a plurality of scrap brands is designated. The blending plan support method generates an estimation model for outputting the content of a predetermined component contained in molten steel generated from a raw material in which a plurality of scrap brands are blended in an arbitrary blending ratio based on a plurality of actual data relating the blending results of a plurality of scrap brands to the component content results of molten steel in past furnace operations. The estimation model is used to output an estimated value of the content of a predetermined component contained in molten steel generated from a raw material based on a blending plan in which a blending ratio of a plurality of scrap brands is specified.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present invention relates to a blending plan support method for assisting in formulating a blending plan specifying the blending ratios of a plurality of scrap grades in a furnace operation for producing molten steel by melting a raw material in which a plurality of scrap grades are blended at an arbitrary blending ratio.

Background Art

[0002] The production of steel is roughly classified into the blast furnace method and the electric furnace method. In the steelmaking process of the electric furnace method, a plurality of scrap grades mixed in predetermined amounts based on a blending plan are charged into an electric furnace as raw materials and melted to obtain molten steel. Examples of scrap grades include shredder, heavy, new cut, C press, Darai powder, pig iron, in-plant generated scraps, etc. Scrap grades may be managed in association with the source of acquisition, such as "shredder supplied by Company A".

[0003] The components in the molten steel are generally measured after melting the scrap for each charge. Here, when trump elements such as copper are mixed into the molten steel, it becomes a factor for generating defective products such as cracks during the production of steel plates. However, since the measurement of the trump elements contained in the molten steel is at the timing after melting the scrap, if the molten steel contains trump elements exceeding the standard values preset for each steel grade, the steel produced in that charge must be shredded. Therefore, when formulating a blending plan for scrap grades, when specifying the blending ratios of a plurality of scrap grades, a technique for estimating how much trump element is contained in the charge based on the blending plan is important. <~

[0004] For example, Patent Document 1 discloses a management system for scrap inventory information. Patent Document 1 discloses a business process for performing a quality inspection of a product, changing the scrap grade supplied to the electric furnace when the trump element contained in the product is greater than or equal to a predetermined value, and identifying the supplier of the scrap that caused the increase in the trump element concentration based on the shipping form information obtained at the time of receipt.

[0005] In addition, Patent Document 2 discloses a method for measuring the elemental content of scrap. In the method of Patent Document 2, an element not contained in the scrap is used as a marker element. First, the marker element is added to the molten metal and its concentration is measured. Then, the scrap is charged into the molten metal and dissolved, and in that state, the concentrations of the marker element and the specific element in the molten metal are measured. Then, the weight of the molten metal increased by the dissolution of the scrap is determined from the change in the concentration of the marker element in the molten metal before and after the dissolution of the scrap, and the content of the specific element in the scrap is determined from the change in the weight of the molten metal and the change in the concentration of the specific element before and after the dissolution of the scrap.

Prior Art Documents

Patent Documents

[0006]

Patent Document 1

Patent Document 2

Non-Patent Documents

[0007]

Non-Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0008] However, the method described in Patent Document 1 aims to identify specific scraps that led to an increase in trump element concentration, and is not a technique for estimating the trump element concentration of each scrap type currently held in inventory. Furthermore, the method described in Patent Document 2 involves dissolving one component from a given scrap type and measuring the trump element concentration. For this reason, estimating the average trump element concentration for each scrap type consisting of a large number of components would require conducting many tests, making it an impractical method from the standpoint of manufacturing costs and test time constraints.

[0009] Therefore, the present invention has been made in view of the above problems, and the object of the present invention is to provide a scrap blending plan support method that can estimate the component content in the molten steel produced in furnace operation, in which molten steel is produced by melting raw materials blended with multiple scrap types in arbitrary proportions, and support the formulation of a blending plan for multiple scrap types. [Means for solving the problem]

[0010] To solve the above problems, according to one aspect of the present invention, in furnace operation to produce molten steel by melting a raw material in which multiple scrap materials are blended in arbitrary proportions, a blending plan support method that supports the formulation of a blending plan specifying the blending proportions of the multiple scrap materials, comprising: an estimation model generation step of generating an estimation model that outputs the content of a predetermined component contained in molten steel produced from a raw material in which multiple scrap materials are blended in arbitrary proportions, based on multiple performance data that associate the blending performance, which is the blending proportion of the multiple scrap materials, with the component content performance of the molten steel in past furnace operations; and a content estimation step of outputting an estimated value of the content of a predetermined component contained in molten steel produced from a raw material based on a blending plan specifying the blending proportions of the multiple scrap materials, using the estimation model. A formulation planning support method is provided, which includes this.

[0011] Furthermore, the content estimation step may include a similar performance data extraction step, which extracts performance data from past furnace operations that include blending results similar to the blending plan, and a similar performance reflection step, which outputs an estimated value of the content of a predetermined component contained in the molten steel produced from the raw materials based on the blending plan, based on the performance data including blending results similar to the blending plan extracted in the similar performance data extraction step and the estimation model.

[0012] The content estimation step may output information regarding the probability distribution of the content of a predetermined component in the molten steel produced from the raw materials based on the blending plan.

[0013] Furthermore, the content estimation step may also include a fit calculation step that outputs the probability that the estimated content of a predetermined component in the molten steel produced from the raw materials based on the blending plan falls outside the acceptable content range, based on the information regarding the probability distribution and a preset acceptable content range.

[0014] Furthermore, the blending plan support method may include a blending plan adjustment step, which involves adjusting the blending plan by evaluating the relationship between the estimated content of a predetermined component in the molten steel produced from the raw materials based on the blending plan and the cost of the raw materials based on the blending plan. [Effects of the Invention]

[0015] As described above, according to the present invention, in furnace operation in which molten steel is produced by melting raw materials in which multiple scrap materials are blended in any blending ratio, it is possible to estimate the component content in the molten steel produced and to support the formulation of a blending plan for multiple scrap materials. [Brief explanation of the drawing]

[0016] [Figure 1] This is an explanatory diagram showing an overview of a scrap blending plan support method according to one embodiment of the present invention. [Figure 2] This is a block diagram showing one example configuration of a formulation planning support device according to the same embodiment. [Figure 3] This is a block diagram showing another configuration example of the formulation planning support device according to the same embodiment. [Figure 4] This is a flowchart showing the process for estimating the component concentration in molten steel in the blending plan support method according to the same embodiment. [Figure 5] This is a flowchart showing the processing of the estimation model in the formulation planning support method according to the same embodiment. [Figure 6] This is a flowchart showing the formulation planning support process in the formulation planning support method according to the same embodiment. [Figure 7] This is an explanatory diagram showing an example of probability distribution information. [Figure 8] This block diagram shows an example of the hardware configuration of an information processing device that performs scrap metal blending planning support. [Modes for carrying out the invention]

[0017] Preferred embodiments of the present invention will be described in detail below with reference to the attached drawings. In this specification and drawings, components having substantially the same functional configuration are denoted by the same reference numerals, and redundant descriptions will be omitted.

[0018] [1. Overview] First, an overview of the scrap blending plan support method according to one embodiment of the present invention will be described based on Figure 1. Figure 1 is an explanatory diagram showing an overview of the scrap blending plan support method according to this embodiment.

[0019] The scrap blending plan support method according to this embodiment is a method for supporting the formulation of a blending plan that specifies the blending ratios of multiple scrap materials in furnace operations that produce molten steel by melting raw materials blended with multiple scrap materials in arbitrary blending ratios. In the scrap blending plan support method according to this embodiment, first, an estimation model is generated that outputs the content of a predetermined component contained in molten steel produced from raw materials blended with multiple scrap materials in arbitrary blending ratios, based on multiple historical data that associate blending results (blending results) of multiple scrap materials in past furnace operations with the component content results. Then, using the generated estimation model, an estimated value of the content of a predetermined component contained in molten steel produced from raw materials based on a blending plan that specifies the blending ratios of multiple scrap materials is output.

[0020] In the following explanation, we will use the example of a case where the raw materials for obtaining molten steel consist only of multiple scrap types, but the raw materials may also include materials other than scrap, such as molten iron. In that case, if the content of a predetermined component in the raw materials other than scrap, such as molten iron, is unknown, it may be treated as one of the scrap types. On the other hand, if the content of a predetermined component in the raw materials other than scrap is known, the content of that predetermined component in the raw materials other than scrap can be subtracted from the actual component content of the molten steel, and it can be treated in the same way as when only multiple scrap types are blended. Furthermore, in the following explanation, we will mainly use trump elements as the predetermined component, but the predetermined component is not limited to trump elements, and may include other elements such as sulfur (S) or phosphorus (P).

[0021] As shown in Figure 1, past furnace operation results are linked to the blending ratio x_km of multiple scrap materials and the actual component content y_kn of the molten steel produced by melting raw materials blended with these scrap materials, resulting in multiple performance data points. Performance data is acquired for each charge.

[0022] In the blending ratio x_km of scrap materials, k is the charge number (k=1,2,…,s), m is a unique number for each scrap material (m=1,2,…,p), and the scrap material is represented by xm. For example, blending ratio x_11 represents the blending ratio of scrap material x1 in charge No. 1. The sum of the blending ratios x_km of scrap materials added in one charge is 1 (i.e., Σ m (x_km)=1.0(k=1,2,…,s)).

[0023] Furthermore, in the molten steel component content data y_kn, k is the charge number (k=1,2,…,s), n is a unique number for each component y (n=1,2,…,q), and the components are represented by yn. For example, the component content data y_11 represents the content of component 1 in the molten steel produced in charge No. 1.

[0024] Using such multiple historical data, an estimation model is generated that takes the blending ratios of multiple scrap materials as input and outputs the component content of various components contained in molten steel produced from raw materials blended with those scrap materials in arbitrary proportions. The type and format of the estimation model are not particularly limited, but machine learning models such as Long Short-Term Memory (LSTM) or multiple regression models may be used.

[0025] Then, by inputting a blending plan specifying the blending ratios of multiple scrap materials into the estimation model, an estimated value of the content of a predetermined component in the molten steel produced based on that blending plan can be obtained. The blending plan represents the blending ratio X_m (m=1,2,…,p) for scrap material xm in a future charge. The estimation model outputs an estimated value Y_n (n=1,2,…,q) of the content of component yn in the molten steel relative to the blending plan, based on the input blending plan (blending ratio X_m).

[0026] When formulating a blending plan, it is not practical to melt large quantities of scrap materials that exist in inventory and estimate the amount of trump elements and other components contained in each scrap material, as described in Patent Document 2 above. Therefore, in this embodiment, an estimation model is generated that estimates the component content of molten steel produced from raw materials blended with multiple scrap materials in arbitrary proportions, based on the blending ratios and component content after melting of each scrap material in multiple past charges. By using the generated estimation model, it is possible to estimate the component content in the molten steel produced based on the blending plan without having to newly perform the melting of large quantities of scrap materials. As a result, it is possible to predict how much trump elements will be contained in the molten steel produced, and it is possible to support workers in formulating blending plans for scrap materials, such as avoiding blending scrap materials in which the trump element content exceeds the standard value.

[0027] In this invention, the blending ratio x_km of multiple scrap materials may be expressed as the weight ratio of each scrap material to the total scrap weight added in one charge, or as the weight of each scrap material added in one charge. When the blending ratio x_km of multiple scrap materials is expressed as the weight ratio of each scrap material to the total scrap weight added in one charge, the actual component content y_kn of the molten steel is expressed as component concentration (mass%). When the blending ratio x_km of multiple scrap materials is expressed as the weight of each scrap material added in one charge, the actual component content y_kn of the molten steel is expressed as component weight (kg or ton). When the actual component content y_kn of the molten steel is expressed as component weight, as in the latter case, it can be converted to component concentration by dividing the weight of each component by the weight of the molten steel produced.

[0028] Furthermore, the blending plan (blending ratio X_m) input to the estimation model, and the component content Y_n output from the estimation model, will be in the same units as the actual data (actual blending ratio x_km and actual component content y_kn of molten steel) used when generating the estimation model.

[0029] The following describes in detail the scrap pulp mixing plan support method according to this embodiment.

[0030] [2. Formulation Planning Support Device] First, an example configuration of the blending plan support device 100 that performs the scrap blending plan support method according to this embodiment will be described based on Figures 2 and 3. Figure 2 is a block diagram showing an example configuration of the blending plan support device 100 according to this embodiment. Figure 3 is a block diagram showing another example configuration of the blending plan support device 100 according to this embodiment.

[0031] As shown in Figure 2, the formulation planning support device 100 according to this embodiment includes a performance data acquisition unit 110, an estimation model generation unit 120, and a content estimation unit 140. The formulation planning support device 100 may also include a formulation planning adjustment unit 180. Furthermore, the formulation planning support device 100 may include a storage unit 170, an input unit 130, and an output unit 150, either internally within the formulation planning support device 100 or as external devices (separate units) connected to the formulation planning support device 100.

[0032] The performance data acquisition unit 110 acquires performance data from past furnace operations. The performance data acquisition unit 110 acquires multiple performance data sets from the storage unit 170 (described later), which stores performance data from past furnace operations, or from external devices connected to the performance data acquisition unit 110, associating the blending ratio x_km of multiple scrap materials with the actual component content y_kn of molten steel produced by melting raw materials blended with multiple scrap materials based on the said blending ratio. The performance data acquisition unit 110 only needs to acquire the number of performance data sets necessary for generating the estimation model; for example, it may acquire the most recent 100 to 1000 performance data sets. The performance data acquisition unit 110 outputs the acquired performance data to the estimation model generation unit 120. If the acquired performance data is not stored in the storage unit 170, the performance data acquisition unit 110 may record it in the storage unit 170.

[0033] The estimation model generation unit 120 generates an estimation model based on multiple historical data, which outputs the content of predetermined components contained in molten steel produced from raw materials based on the blending ratios of multiple scrap brands input. The estimation model generation unit 120 generates, for example, a machine learning model such as Long Short-Term Memory (LSTM) or a multiple regression model as the estimation model. The estimation model generation unit 120 may also include a brand-specific component concentration estimation unit 121 and a brand-specific component concentration acquisition unit 123. Details of the estimation model generation process by the estimation model generation unit 120 will be described later. The estimation model generation unit 120 outputs the generated estimation model to the content estimation unit 140.

[0034] The input unit 130 is an interface that receives input information. The input unit 130 is, for example, a device for an operator to input information, such as a keyboard, mouse, or touch panel. For example, if an operator uses the input unit 130 to specify the blending ratios of multiple scrap materials in a future charge, the specified blending ratios of multiple scrap materials are input into the blending plan support device 100 as a blending plan.

[0035] The content estimation unit 140 uses an estimation model to output the content of a predetermined component contained in molten steel produced from raw materials based on a blending plan that specifies the blending ratios of multiple scrap materials. The content estimation unit 140 inputs the blending plan input from the input unit 130 into the estimation model generated by the estimation model generation unit 120, and obtains the content of the component contained in the molten steel produced based on the blending plan from the estimation model. The content estimation unit 140 outputs the component content obtained from the estimation model as an estimated value to the output unit 150.

[0036] In another embodiment, the content estimation unit 140 may include a similar performance data extraction unit 141 and a similar performance data reflection unit 143, as shown in Figure 3.

[0037] The Similar Performance Data Extraction Unit 141 extracts performance data from past furnace operations that includes blending results with blending ratios similar to the blending plan. The Similar Performance Data Extraction Unit 141 refers to the storage unit 170, etc., which stores the performance data acquired by the Performance Data Acquisition Unit 110, and extracts performance data that includes blending results with blending ratios similar to the blending plan input from the Input Unit 130. The Similar Performance Data Extraction Unit 141 determines the similarity between the blending plan and the blending results based on the similarity of the blending ratios of the scrap materials and the temporal distance of the performance data. Details of the Similar Performance Data Extraction Process by the Similar Performance Data Extraction Unit 141 will be described later. The Similar Performance Data Extraction Unit 141 outputs one or more performance data extracted from the storage unit 170 to the Similar Performance Reflection Unit 143.

[0038] The Similar Performance Reflection Unit 143 inputs performance data, including those similar to the blending plan extracted by the Similar Performance Data Extraction Unit 141, into the estimation model, thereby outputting an estimated value of the content of a predetermined component in the molten steel produced from the raw materials based on the blending plan. In other words, the Similar Performance Reflection Unit 143 does not randomly use past performance data, but rather selects blending performance data similar to the blending plan and inputs it into the estimation model.

[0039] The reason why the content estimation unit 140 includes a similar performance data extraction unit 141 and a similar performance reflection unit 143 is as follows: The estimation model generated by the estimation model generation unit 120 is generated based on all past performance data, that is, past performance data within a time-unconstrained range. However, in actual furnace operation, even if the furnace operating conditions are not changed, gradual fluctuations (trends in performance data) occur in the values ​​of the performance data over time. Therefore, when trying to estimate the content of a predetermined component contained in molten steel with higher accuracy using the estimation model, it may be better to strongly reflect the trend of the performance data, such as using the most recent performance data, which is closer to the current furnace operation, rather than using all past performance data without time constraints. Therefore, the similar performance reflection unit 143 acquires performance data from operations close to the current furnace operation from the similar performance data extraction unit 141 and inputs such performance data into the estimation model, thereby reflecting the recent trend of operations in the estimation of the content of a predetermined component contained in molten steel produced from raw materials based on the blending plan.

[0040] The estimated content of predetermined components in the molten steel produced from the raw materials based on the blending plan, obtained in this way, is output to the output unit 150.

[0041] In the above explanation, it has been assumed that the estimated values ​​of the content of predetermined components in the molten steel produced from raw materials based on the blending plan, output from the content estimation unit 140, have a single value for each component. However, estimated values ​​inherently involve variability corresponding to estimation errors. Therefore, the content estimation unit 140 may obtain and output information regarding the probability distribution of the content of predetermined components in the molten steel produced from raw materials based on the blending plan (hereinafter also referred to as "probability distribution information"). Probability distribution information includes, for example, the probability distribution (probability density) of component content, the 95% confidence interval for component content, etc. The content estimation unit 140 may output the probability distribution information to the blending plan adjustment unit 180, which will be described later.

[0042] Furthermore, this probability distribution information may be added to the calculation results of the estimated content of a predetermined component in molten steel, obtained from the estimation model generated by the estimation model generation unit 120, so as to give it a certain probability range. Alternatively, when generating the estimation model in the estimation model generation unit 120, various types of models may be generated, and the estimation model itself may be given a tolerance range by allowing a range in the coefficients of the generated estimation model, etc., so that the calculation results of these models are included in the results of the generated estimation model. In this case, as a result, the calculation results of the estimated content of a predetermined component in molten steel using the estimation model will have a certain probability range.

[0043] A more concrete example is when using a multiple regression model as the estimation model. In this case, the estimation model is obtained by regression based on multiple historical data used to generate the estimation model. For example, if the least squares method is used for regression, the estimated component content calculated from the estimation model will be obtained as a single regression line. However, because the data is variable, not all of the multiple historical data lie on that single regression line; rather, the multiple historical data are scattered around that regression line.

[0044] Given this, it is natural to consider that the method of fitting the regression line is not determined by a single slope obtained by the least squares method, but rather that it can take various slopes within a range that does not deviate excessively from multiple actual data. In other words, it is natural to consider that the slope of the regression line has a range (a probability distribution) corresponding to the variability of multiple actual data, rather than being determined by a single value obtained by the least squares method. In fact, it is known that the estimated values ​​of partial regression coefficients obtained by multiple regression follow a normal distribution. Since the partial regression coefficients representing the content of a certain component for each scrap material have a probability distribution, as a result, the estimated value of the content of a given component in the molten steel produced by blending these multiple scrap materials also has a probability distribution. The method for calculating the probability distribution of this estimated component content will be described later.

[0045] Furthermore, the content estimation unit 140 may output the probability that the estimated content of the predetermined component falls outside the acceptable content range, based on information regarding the probability distribution and a preset acceptable content range. As described above, since the estimated content of a component can be assigned a probability, it is naturally possible to know the probability that the content of the component of interest falls outside the preferred range based on that probability. In actual furnace operation, there are times when it is necessary to know the specific value of the content of the component of interest, but on the other hand, there are also times when it is necessary to know whether or not it is acceptable to proceed with furnace operation as per the corresponding blending plan. For this reason, knowing the probability that the component content falls outside a predetermined range is preferable because it allows for a determination of whether or not it is acceptable to proceed with furnace operation as per the corresponding blending plan.

[0046] The output unit 150 is an interface that outputs information from the blending plan support device 100. The output unit 150 is a device for presenting the information output from the blending plan support device 100 to the operator, and is an output device such as a display or printer. For example, when the content estimation unit 140 inputs the content of components contained in the molten steel produced based on the blending plan, the output unit 150 outputs the content of components in the molten steel estimated by the estimation model. Also, when the blending plan adjustment unit 180 inputs the optimal blending plan for a scrap material, for example, the output unit 150 outputs the optimal blending plan.

[0047] The memory unit 170 is a memory unit that stores the performance data acquired by the performance data acquisition unit 110. The memory unit 170 may store all of the performance data acquired by the performance data acquisition unit 110, or it may store performance data for the most recent predetermined period. In addition, the memory unit 170 can store programs and data used in the blending plan support device 100, and can also store the cost (unit price) of each scrap material.

[0048] The blending plan adjustment unit 180 adjusts the blending plan by evaluating the relationship between the estimated content of predetermined components in the molten steel produced from the raw materials based on the blending plan and the cost of the raw materials based on the blending plan. That is, the blending plan adjustment unit 180 determines the optimal blending plan based on the estimated content of components obtained by the content estimation unit 140 and, as appropriate, probability distribution information. For example, the blending plan adjustment unit 180 creates an evaluation function from a function based on the estimated content of predetermined components in the molten steel produced from the raw materials based on the blending plan and a function based on the cost of the raw materials (i.e., the blended scrap grades) based on the blending plan.

[0049] Generally, the estimated content of a given component in molten steel produced from raw materials based on a blending plan often yields more favorable values ​​if higher-quality scrap materials are selected. On the other hand, selecting higher-quality scrap materials leads to increased costs for purchasing them. Therefore, the function based on the estimated content of a given component in molten steel produced from raw materials based on a blending plan and the function based on the cost of the raw materials (i.e., the blended scrap materials) based on the blending plan are mutually exclusive. One function acts as a constraint on the other, preventing the overall evaluation function from diverging and causing it to settle at a constant value.

[0050] Therefore, using the created evaluation function, the blending plan that yields a better value from the evaluation function is evaluated as the optimal blending plan. Then, by applying this optimal blending plan as the blending plan to be used in future reactor operations, the reactor operations can be adjusted. Details of the blending plan support processing by the content estimation unit 140 and the blending plan adjustment unit 180 will be described later. The blending plan adjustment unit 180 outputs the obtained optimal blending plan to the output unit 150.

[0051] The above describes one example configuration of the formulation planning support device 100 according to this embodiment. In the present invention, the formulation planning support device 100 only needs to include at least a performance data acquisition unit 110, an estimation model generation unit 120, and a content estimation unit 140. Furthermore, the formulation planning support device 100 shown in Figures 2 and 3 includes a performance data acquisition unit 110, an estimation model generation unit 120, an input unit 130, a content estimation unit 140, an output unit 150, a storage unit 170, and a formulation planning adjustment unit 180 in a single device, but the present invention is not limited to this example. For example, the formulation planning support device 100 may be constructed by providing each functional unit in multiple devices, such as constructing the estimation model generation unit 120 and other functional units in separate devices.

[0052] [3. Methods to support formulation planning] Next, the blending plan support method according to this embodiment will be described. In the following description, as an example, the blending ratio x_km of multiple scrap types will be expressed as the weight ratio of each scrap type to the total scrap weight added in one charge, and the actual component content y_kn of the molten steel will be expressed as the component concentration (mass%).

[0053] [3-1. Estimation of component content in molten steel produced based on the blending plan] First, the component concentration estimation process in the molten steel in the blending plan support method according to this embodiment will be explained based on Figures 2 to 4. In the component concentration estimation process, the component content in the molten steel produced based on the blending plan is estimated. Figure 4 is a flowchart showing the component concentration estimation process in the molten steel in the blending plan support method according to this embodiment.

[0054] In the process of estimating the component concentration in molten steel, first, as shown in Figure 4, the performance data acquisition unit 110 acquires performance data from past furnace operations (S100). The performance data acquisition unit 110 acquires multiple performance data sets that associate the blending ratio x_km of multiple scrap materials with the performance y_kn of the component content of the molten steel produced by blending and melting multiple scrap materials based on that blending ratio. The performance data acquisition unit 110 outputs the acquired performance data to the estimation model generation unit 120.

[0055] Next, the estimation model generation unit 120 generates an estimation model (S110) that outputs the content of a predetermined component in the molten steel produced based on the mixing ratio of multiple scrap grades input, based on the mixing ratio of multiple scrap grades input. The estimation model generation unit 120 generates, for example, a machine learning model such as Long Short-Term Memory (LSTM) or a multiple regression model as the estimation model.

[0056] For example, when generating a multiple regression model as an estimation model, the estimation model generation unit 120 generates the multiple regression model using the brand-specific component concentration estimation unit 121 and the brand-specific component concentration acquisition unit 123, as shown in Figures 2 and 3.

[0057] The brand-specific component concentration estimation unit 121 inputs the actual data obtained in step S100 into the following formula (1) and performs fitting to obtain the partial regression coefficient b ij We find x. i is the actual blending ratio of scrap metal brands included in the performance data (i=1,2,…,p), y j The actual component content included in the performance data (j=1,2,…,q), and the partial regression coefficient b. ij is the concentration of component j contained in scrap material i. p is the number of charges in the historical data used to fit the multiple regression model, and q is the number of scrap material types included in the historical data. The magnitude of p can be, for example, around 100 to 1000.

[0058]

number

[0059] Once the component concentration estimation unit 121 for each brand has finished fitting the actual data to the above equation (1), the partial regression coefficient b ij The decision will be made.

[0060] The component concentration acquisition unit 123 for each brand determines the partial regression coefficient b ijObtain it and obtain a multiple regression model represented by the following formula (2). Here, X i is the blending ratio of scrap stocks (i = 1, 2, …, p) based on the blending plan, and Y j is the estimated value of the component content (j = 1, 2, …, q).

[0061] In the above formula (1), since the partial regression coefficient b ij corresponds to the concentration of the components contained in each scrap stock, in the above fitting, it is desirable to impose a constraint (non - negative constraint) of b ij ≧0 for all i and j. Also, even when x i is defined as the blending weight of the scrap stocks included in the actual data instead of the blending ratio of the scrap stocks included in the actual data, it is similarly desirable to impose a non - negative constraint.

[0062] Also, as described above, the partial regression coefficient b ij is not obtained as a single value obtained by fitting with the least - squares method, but may be obtained in a form (probability distribution, probability density) with a probability width corresponding to the variation in fitting.

[0063]

Equation

[0064] If the blending plan X i of the scrap stocks is input into the above formula (2), the estimated value Y i of the component content in the molten steel generated based on the blending plan X j can be obtained. When the value of the partial regression coefficient b ij is determined and appropriate actual data values (x i , y j ) are input, the above formula (2) that enables the estimated value Y j of the component content in the molten steel to be obtained is referred to as an estimation model. The estimation model generation unit 120 outputs the generated estimation model (for example, the above formula (2)) to the content estimation unit 140.

[0065] Then, the content estimation unit 140 uses the estimation model generated in step S110 to estimate the component content Y in the molten steel produced from the raw materials based on the blending plan, which specifies the blending ratio of multiple scrap materials. j The formula X is calculated (S120). For example, when the estimation model is represented by the multiple regression model of equation (2) above, the content estimation unit 140 calculates the formula X input from the input unit 130. i Input the above formula (2) into the formula (2), and the output is the formulation plan X. i Estimated component content Y in molten steel produced from raw materials based on the above. j To obtain.

[0066] Subsequently, the content estimation unit 140 calculates the estimated component content Y obtained from the estimation model. j The output is sent to the output unit 150 (S130). The content estimation unit 140 may output estimated values ​​of the content of all components contained in the molten steel, or it may output estimated values ​​of the content of only specific components such as trump elements.

[0067] The above describes the process for estimating component concentrations in molten steel in the blending plan support method according to this embodiment. In this process, an estimation model is generated that outputs the content of predetermined components in molten steel produced from raw materials blended with multiple scrap materials in specific blending ratios, based on past performance data. Using this estimation model, an estimated value of the component content in the molten steel produced based on the blending plan is obtained. This allows the operator to estimate the component content in the molten steel produced based on the blending plan for a future charge without having to melt a large amount of scrap material. As a result, it is possible to predict how much trump element will be contained in the molten steel produced, and it is possible to support the operator in formulating a scrap material blending plan, such as avoiding blending scrap materials in which the trump element content exceeds the standard value.

[0068] [3-2. Improving the accuracy of the estimation model] To improve the accuracy of the estimation model shown in Figure 4, the estimation model and actual data including actual blending results similar to the blending plan may be used to obtain an estimate of the component content in the molten steel produced based on the blending plan. The process for improving the accuracy of the estimation model will be described below with reference to Figure 5. Figure 5 is a flowchart showing the processing of the estimation model in the blending plan support method according to this embodiment.

[0069] First, as shown in Figure 5, the performance data acquisition unit 110 acquires performance data from past furnace operations (S200). The performance data acquisition unit 110 outputs the acquired performance data to the estimation model generation unit 120 and also records it in the storage unit 170. Next, the estimation model generation unit 120 generates an estimation model (S210) that outputs the content of predetermined components in the molten steel produced from raw materials blended according to the blending ratios of multiple scrap brands input, based on the multiple performance data acquired in step S200. The processing in step S210 can be performed in the same way as in step S110 in Figure 4.

[0070] Furthermore, the similar performance data extraction unit 141 extracts performance data from past furnace operations that includes blending results similar to the blending plan (S220). The similar performance data extraction unit 141 refers to the storage unit 170, etc., which stores the performance data acquired by the performance data acquisition unit 110, and extracts performance data that includes blending results similar to the blending plan entered from the input unit 130.

[0071] The similar performance data extraction unit 141 determines the similarity between the blending plan and the actual blending results based on the similarity of the blending ratios of the scrap materials and the time distance of the actual data.

[0072] The similarity of the blending ratios of scrap materials is, for example, the blending plan {X i} and the actual blending ratio {x iThis can also be expressed by the magnitude of the vector distance (difference) between the two scrap materials. The smaller the vector distance (difference), the higher the similarity in the blending ratios of the scrap materials. The temporal distance of the actual data can be expressed, for example, by the time from when the scrap materials were blended to the present, or by the charge number k. The charge numbers are assigned starting from the most recent charge, k=1, 2, ... The smaller the temporal distance, the higher the temporal similarity.

[0073] Similarity can be measured, for example, by equation (3) below. The first term of equation (3) represents the magnitude of the distance (difference) between the vectors, and the second term represents the temporal distance. Charges with high similarity can be extracted by searching for conditions that make equation (3) small.

[0074]

number

[0075] Here, h is the current charge number, and p and q are weighting coefficients. Note that there are various ways to define equation (3). For example, {X i} and {x i As for how to determine the magnitude of the vector distance (difference) with}, you may use the L2 norm instead of the L1 norm shown in the first term of equation (3). Alternatively, you may assign weights to each scrap material, taking into account the relative amounts of components to be removed, such as playing card elements, contained in each scrap material.

[0076] The similar performance data extraction unit 141 extracts, for example, performance data that has the minimum cumulative value of the similarity of the blending ratio of scrap materials and the time distance of the performance data, and similar performance data {x ei ,y ej Extracted as}. Then, the similar performance reflection unit 143 uses the coefficient b obtained from the above formula (2). ij Using (i.e., using an estimation model), and further, similar performance data (x ei ,y ejBy using ) and transforming the estimation model as shown in equation (4) below, the predicted value of the component content in molten steel {Y j Calculate} (S230).

[0077]

number

[0078] In a simple multiple regression model like the one in equation (2) above, the accurate partial regression coefficient b ij To determine this, it is necessary to trace back a large number of charge counts. As a result, component fluctuations within each scrap issue that may occur within the time span of the historical data used to generate the multiple regression model tend to be ignored, and the time-averaged partial regression coefficients are obtained. In particular, similar blending ratios in past historical data {x ki If multiple instances of} exist, multicollinearity necessitates tracing back through more charge counts.

[0079] Therefore, in order to predict the ingredient content, taking into account the recent trends in ingredients, within the time span of actual data, the formulation plan {X i Actual blending ratio close to {x ei Search for performance data for}. Then, formulate a blending plan {X i} and a similar blending ratio in actual results {x ei The regression coefficient b is applied only to the difference between}. ij The component content is calculated by multiplying by the actual component content of similar data {y ej By adding this to the formula, the estimation model will reflect actual blending ratios (similar performance data) that are close to the planned blending ratio. The reason for considering the time distance of the performance data is to predict the ingredient content while taking into account the trends of the ingredients as recently as possible.

[0080] The number of similar historical data points extracted in step S220 may be one or multiple. Since historical data includes reactor operation trends that differ from the current ones, reflecting more similar historical data in the estimation model is expected to improve the accuracy of the estimation model compared to extracting only one similar historical data point and performing estimation with the estimation model.

[0081] When performing estimation using an estimation model that reflects multiple similar performance data, the similar performance data extraction unit 141 first extracts N (N>1) performance data {x} in order from those with the smallest cumulative value of the similarity between the blending ratio of scrap material brands and the time distance of the performance data. ei_N ,y ej_N Extracts}. Then, in step S230, the similar performance reflection unit 143 extracts similar performance data {x ei_N ,y ej_N For each of these, use equation (4) above to predict the component content Y in the molten steel. j After calculating the predicted values ​​of the component content in N molten steel {Y j We take the weighted average of} and predict the final component content {Y j Let's assume this. Alternatively, we can take a weighted average by assigning larger weights to the data with the smallest cumulative value of the similarity in the blending ratio of scrap materials and the time distance of the actual data, or we can take a weighted average by giving equal weights to N similar actual data. The number N of similar actual data to extract can be set appropriately according to the total number of actual data.

[0082] Thus, the similar performance data extraction unit 141 outputs one or more performance data extracted from the storage unit 170 to the similar performance data reflection unit 143, and the similar performance data reflection unit 143 uses an estimation model that reflects the similar performance data to predict the final component content {Y j Outputs}.

[0083] Subsequently, the similar performance reflection unit 143 calculates the estimated component content Y obtained from the estimation model. j This is output to the output unit 150 (S240). Step S240 can be performed in the same way as step S130 in Figure 4.

[0084] The above describes the process for improving the accuracy of the estimation model in the blending plan support method according to this embodiment. In this process, actual data (similar actual data) including blending results similar to the blending plan are extracted from past furnace operations. Then, using the estimation model that reflects the extracted similar actual data, estimated values ​​of the content of predetermined components in the molten steel generated based on the blending plan are output. This makes it possible to improve the estimation accuracy of the estimated values ​​of component content in molten steel by the estimation model.

[0085] [3-3. Support processing for formulation planning] As described above, the blending plan support device 100 according to this embodiment can obtain estimated values ​​of the component content in molten steel produced based on the blending plan using an estimation model. Furthermore, it may also provide support information for formulating a suitable blending plan or present a suitable blending plan. The blending plan support process will be described below with reference to Figures 6 and 7. Figure 6 is a flowchart showing the blending plan support process in the blending plan support method according to this embodiment. Figure 7 is an explanatory diagram showing an example of probability distribution information.

[0086] First, as shown in Figure 6, the performance data acquisition unit 110 acquires performance data from past furnace operations (S300). The performance data acquisition unit 110 outputs the acquired performance data to the estimation model generation unit 120. Next, the estimation model generation unit 120 generates an estimation model (S310) that outputs the content of a predetermined component in the molten steel produced based on the raw materials blended at a given blending ratio, based on the blending ratio of multiple scrap materials input. The processing in steps S300 to S310 can be performed in the same way as steps S100 to S110 in Figure 4.

[0087] The blending plan adjustment unit 180 then generates an evaluation function from a function that takes as input the estimated content of components contained in the molten steel produced from the raw materials based on the blending plan, calculated using an estimation model, and a function that takes as input the cost of the raw materials based on the blending plan (i.e., the cost of the scrap materials blended based on the blending plan) (S320).

[0088] For example, the evaluation function Ev can be expressed as shown in equation (5) below, using the estimation model Es, the function f, the function g, and the weight coefficients α and β. For example, as the function f, the estimated value Y of the component content in the molten steel generated based on the candidate blending plan is obtained using the estimation model generated in step S310. j (=Es(x i , y i A function can be used to evaluate its desirability. Alternatively, for example, function g can be a function (a linear expression for the blending ratio) that multiplies the unit price and blending ratio of each scrap material included in the blending plan and then multiplies them together.

[0089]

number

[0090] The form of the evaluation function is not limited to the form shown in equation (5), but generally, the desirability of the estimated component content (i.e., the first term on the right side of equation (5)) becomes more favorable when high-grade scrap is selected, while the cost of the raw materials (i.e., the second term on the right side of equation (5)) becomes higher and less favorable when high-grade scrap is selected. For this reason, it is preferable that the evaluation function be a function that can evaluate that these two factors are inversely related and are in balance with each other.

[0091] The blending plan adjustment unit 180 then adjusts the blending plan by evaluating it using the evaluation function (S330). This adjustment is performed by inputting actual data corresponding to various blending plan candidates into the evaluation function and checking the output value of the evaluation function, and determining that a blending plan whose output value satisfies predetermined conditions is a preferable blending plan.

[0092] The output value of the evaluation function can be verified by checking whether the output value of the evaluation function corresponding to the blending plan takes the maximum (or minimum) value among the various possible blending plan candidates. Alternatively, the output value of the evaluation function can be verified by checking whether the output value of the evaluation function corresponding to the blending plan falls within a predetermined range, or by ranking the plans that fall within the predetermined range.

[0093] In other words, preparing a large amount of actual data equivalent to a blending plan and inputting it into an evaluation function in a brute-force manner to evaluate the evaluation function is often undesirable due to the large computational load. On the other hand, even if it is not the optimal blending plan, knowing a blending plan that is reasonably desirable is often sufficient for reactor operation. Therefore, instead of evaluating the evaluation function in a brute-force manner, if the value of the evaluation function falls within a predetermined range, it may be considered that a reasonable blending plan has been found and the calculation is stopped at that point. Thus, a blending plan in which the estimated value of the component content falls within the desired range, and other factors such as the raw material cost of the scrap material also fall within the desirable range {X i The system may also search for} and determine whether the blending plan has been adjusted based on whether the output value of the evaluation function falls within a predetermined range.

[0094] As an example of determining whether adjustments have been made based on whether the output value of the evaluation function falls within a predetermined range, when an upper limit of the component range is set, the margin for the desired component range can be defined, for example, as shown in equation (6) below.

[0095]

number

[0096] Here, V is a set of components for which an upper limit has been set on the component range, U j Its upper limit, c j This is the weighting coefficient.

[0097] Alternatively, one could design an evaluation function that incorporates both the margin for the desired component range and the cost of each scrap material, and then pursue a balance between the component margin and cost.

[0098] As described above, the cost of the scrap materials blended according to the blending plan is g(x i , y i ) is a linear equation representing the blending ratio of each scrap material. Also, Y is the estimated value of the component content in the molten steel produced based on the blending plan. j This also becomes a linear formula for the mixing ratio. Thus, the cost of scrap materials g(x i , y i ) and estimated content Y j If the constraints and the objective function to be maximized (or minimized) can all be expressed as linear equations, then the optimal blending plan can be found within the framework of a known linear programming problem. In this case, since the optimal blending plan can be obtained directly by defining the constraints and objective function, it is preferable to obtain the optimal blending plan with less computational load, without having to evaluate the evaluation function by brute force or by searching for blending plans that fall within a desirable range.

[0099] The blending plan adjusted by the blending plan adjustment unit 180 is then output to the output unit 150 as a suitable blending plan.

[0100] In the above explanation, the component content Y in molten steel for the mix design and candidate mix design is j While estimation has been performed assuming that it takes a single value, the content estimation unit 140 estimates the component content Y jIn addition, it is possible to output information (probability distribution information) regarding the probability distribution of the content of predetermined components in the molten steel produced based on the mixing plan to the output unit 150 and utilize that information. The probability distribution information is, for example, the estimated value Y of the component content as shown in Figure 7. j It may also be a probability distribution, and the estimated value Y of the component content. j This is the 95% confidence interval for [the given value].

[0101] When a multiple regression model is used as the estimation model, the content estimation unit 140 estimates the component content Y j The t-distribution can be used as the probability distribution. The expected value of the prediction by the prediction model is Y. j Assuming a hat symbol (^), equation (7) below follows a t-distribution with sp-1 degrees of freedom (Student's t-distribution). Here, s is the number of historical data points used to generate the estimation model, and p is the number of scrapped stocks. Also, D0 is the Mahalanobis general distance, and V e is the residual variance in analysis of variance (see Non-Patent Document 1). As s approaches infinity, the t-distribution coincides with the normal distribution.

[0102]

number

[0103] When probability distribution information is input from the content estimation unit 140, the formulation plan adjustment unit 180 estimates the component content Y based on the probability distribution information. j The probability that the content falls outside the permissible range can be determined, and this probability information can be used for conformity assessment.

[0104] Specifically, the formulation planning adjustment unit 180 first estimates the component content Y based on probability distribution information and a preset allowable content range. j The probability that the content falls outside the acceptable range is determined. The estimated component content Y is obtained by inputting the preferred formulation plan obtained in step S330 into the estimation model. j This is the expected value, and furthermore, the estimated value Y of the component content. j If the probability distribution is obtained, the estimated value of the component Y jThe probability that the content falls outside the acceptable range can be determined. For example, the estimated value Y of the component content shown in Figure 7 is used as probability distribution information. j Assume that a probability distribution is obtained. In this case, when the upper limit of the allowable component content in the charge based on the blending plan is given by c, the estimated value of the component content Y j The probability that the content falls outside the permissible range is represented by the area of ​​the shaded region in Figure 7.

[0105] Estimated value Y of component content in a trade-off relationship i An optimal formulation plan may be sought by maximizing (or minimizing) an evaluation function corresponding to equation (5), which is obtained by weighting and summing two factors: the probability of the content falling outside the acceptable range and the cost.

[0106] Note that the estimated value of the component content is Y. j Instead of the probability that it falls outside the acceptable content range, use the estimated value Y of the component content. j When incorporating the probability (pass rate) that falls within the acceptable content range into the evaluation function, it is advisable to formulate the evaluation function such that the pass rate to be maximized and the cost of the scrap material to be minimized have different signs.

[0107] In the above explanation, only the cost of scrap materials was considered as a factor that has a trade-off relationship with quality (ingredient content). However, other factors such as the inventory level of scrap materials may also be added in addition to the cost of scrap raw materials. For example, if the inventory level of scrap materials is added as a factor to the objective function, it is conceivable to set the objective function so that a penalty is imposed if scrap materials with low current inventory levels (or rare materials) are blended.

[0108] The above describes the formulation planning support process in the formulation planning support method according to this embodiment.

[0109] In the above explanation, the estimation model was generated by performing a multiple regression process that fitted multiple historical data to equation (1) each time an estimation was performed. In other words, in the multiple regression process, for each new formulation plan, a certain number of historical data from previous charges were extracted, and a new estimation model was created (i.e., the partial regression coefficients were estimated). In short, the estimation model used in the previous estimation was not reused.

[0110] On the other hand, the method for generating the estimation model is not limited to the example above; the estimation model used in the previous estimation may also be utilized.

[0111] For example, using a Bayesian estimation framework, models can be generated sequentially as follows: In the brand-specific component concentration estimation unit 121, the probability distribution of brand-specific component concentrations in the estimation model generated to estimate the components of the previous charge is stored in the storage unit 170 as a prior distribution. After the charge is dissolved, the likelihood for the above prior distribution is calculated for the actual component measurement results, and the posterior distribution is obtained by multiplying it by the above prior distribution. This posterior distribution can be used as the probability distribution of brand-specific component concentrations for the model used to estimate the components of the next charge. This is also used as the prior distribution of brand-specific component concentrations for the estimation model used to estimate the components of the charge after that, and is therefore stored in the storage unit 170. By repeating the above process for each charge, the probability distribution of brand-specific component concentrations can be updated sequentially each time the component measurement results for each charge are obtained.

[0112] Furthermore, as a method for calculating the posterior distribution by multiplying the prior distribution by the likelihood, as described above, a known Markov chain Monte Carlo method can be used, for example. Also, in the first cycle of the above update cycle, it is not possible to refer to the probability distribution of component concentrations for each scrap material in past models, but an appropriate prior distribution can be used as the initial value. This initial value may be a non-negative uniform distribution, or the concentration distribution of each scrap material may be set based on past knowledge.

[0113] As described above, by using the model generated to estimate the components of past charges, it is possible to more accurately consider the time-series changes in the component concentrations of each brand, thereby obtaining a highly accurate model.

[0114] In the Bayesian estimation described above, the probability distribution of component concentrations by brand, generated for estimating the components of the previous charge, was used as the prior distribution for the probability distribution of component concentrations by brand in the model used to estimate the components of the next charge. This prior distribution can also utilize prior knowledge other than the brand of scrap being input. For example, external information such as the packaging of the scrap is known to correlate with quality, such as the amount of playing card elements. If the external appearance of the scrap is photographed separately with a camera when it arrives, the predicted range of component concentrations (probability distribution) from that image information can be used as prior knowledge. Specifically, the probability distribution obtained by multiplying the probability distribution of component concentrations by brand in the model generated to estimate the components of the previous charge by the probability distribution of component concentrations predicted from the above image information can be used as the prior distribution for the probability distribution of component concentrations by brand in the model used to estimate the components of the next charge.

[0115] [4. Hardware Configuration] Based on Figure 8, the hardware configuration of the formulation planning support device 100 according to this embodiment will be described. Figure 8 is a block diagram showing an example of the hardware configuration of the information processing device 900 that performs formulation planning support according to this embodiment.

[0116] The information processing device 900 includes a processor (CPU 901 in Figure 8), a ROM 903, and a RAM 905. The information processing device 900 also includes a bus 907, an input interface 909, an output interface 911, a storage device 913, a drive 915, a connection port 917, and a communication device 919.

[0117] The CPU 901 functions as both an arithmetic processing unit and a control unit. The CPU 901 controls the overall operation or a part of it within the information processing unit 900 according to various programs recorded in the ROM 903, RAM 905, storage device 913, or removable recording medium 925. The ROM 903 stores programs or arithmetic parameters used by the CPU 901. The RAM 905 temporarily stores programs used by the CPU 901, or parameters that change as appropriate during program execution. These are interconnected by a bus 907, which is composed of an internal bus such as the CPU bus.

[0118] Bus 907 is connected to external buses such as the PCI (Peripheral Component Interconnect / Interface) bus and PCI Express® via a bridge.

[0119] The input interface 909 is an interface that receives input from an input device 921, which is a means of operation operated by the user, such as a mouse, keyboard, touch panel, button, switch, and lever. The input interface 909 is configured, for example, as an input control circuit that generates an input signal based on information entered by the user using the input device 921 and outputs it to the CPU 901. The input device 921 may be, for example, a remote control device using infrared or other radio waves, or an external device 927 such as a PDA that is compatible with the operation of the information processing device 900. The user of the information processing device 900 can operate the input device 921 to input various data to the information processing device 900 or instruct it to perform processing operations.

[0120] The output I / F 911 is an interface that outputs the input information to an output device 923 that can visually or audibly notify the user. The output device 923 may be, for example, a display device such as a CRT display, liquid crystal display, plasma display, EL display, or lamp. Alternatively, the output device 923 may be an audio output device such as a speaker or headphones, or a printer, mobile communication terminal, or facsimile. The output I / F 911 instructs the output device 923 to output, for example, the processing results obtained from various processes performed by the information processing device 900. Specifically, the output I / F 911 instructs the display device to display the processing results from the information processing device 900 as text or an image. The output I / F 911 also instructs the audio output device to convert an audio signal, such as audio data that has been instructed to be played, into an analog signal and output it.

[0121] The storage device 913 is one of the storage units of the information processing device 900 and is a device for storing data. The storage device 913 is composed of, for example, a magnetic storage device such as an HDD (Hard Disk Drive), a semiconductor storage device such as an SSD (Solid State Drive), an optical storage device, or a magneto-optical storage device. The storage device 913 stores programs executed by the CPU 901, various data generated by the execution of programs, and various data acquired from external sources.

[0122] The drive 915 is a reader / writer for recording media and is either built into or external to the information processing device 900. The drive 915 reads information recorded on the installed removable recording media 925 and outputs it to the RAM 905. The drive 915 can also write information to the installed removable recording media 925. The removable recording media 925 is, for example, a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory. Specifically, the removable recording media 925 may be CD media, DVD media, Blu-ray® media, CompactFlash® (CF), flash memory, SD memory card (Secure Digital memory card), etc. Alternatively, the removable recording media 925 may be, for example, an IC card (Integrated Circuit card) equipped with a contactless IC chip or an electronic device.

[0123] The connection port 917 is a port for directly connecting equipment to the information processing device 900. The connection port 917 can be, for example, a USB (Universal Serial Bus) port, an eSATA (external Serial Advanced Technology Attachment), or a SAS (Serial Attached SCSI (Small Computer System Interface)) port. The information processing device 900 can directly acquire various data from or provide various data to external devices 927 connected to the connection port 917. For example, an alarm notification device such as a rotating light for notifying alarm information may be connected via the connection port 917. Alternatively, a NAS (Network Attached Storage) may be connected as the external device 927 and used as a storage device.

[0124] The communication device 919 is a communication interface composed of, for example, a communication device for connecting to the communication network 929. The communication device 919 is, for example, a communication card for wired or wireless LAN (Local Area Network), Bluetooth (registered trademark), or WUSB (Wireless USB). Alternatively, the communication device 919 may be a router for optical communication, a router for ADSL (Asymmetric Digital Subscriber Line), or a modem for various types of communication. The communication device 919 can, for example, send and receive signals to and from the Internet or other communication devices in accordance with a predetermined protocol such as TCP / IP. For example, a computer for operating the information processing device 900 can also be connected via the communication device 919. The communication network 929 connected to the communication device 919 is composed of a network connected by wire or wireless means. For example, the communication network 929 is the Internet, a home LAN, infrared communication, radio wave communication, or satellite communication.

[0125] The above describes an example of the hardware configuration of the information processing device 900. Each of the above-mentioned components may be made using general-purpose materials, or each component may be made using hardware specialized for its function. The hardware configuration of the information processing device 900 can be appropriately changed depending on the level of technology at the time of implementing this embodiment.

[0126] Although preferred embodiments of the present invention have been described in detail above with reference to the attached drawings, the present invention is not limited to these examples. It is clear to any person with ordinary skill in the art to which the present invention belongs that various modifications or alterations can be conceived within the scope of the technical idea described in the claims, and these are also understood to fall within the technical scope of the present invention. [Explanation of Symbols]

[0127] 100 Formulation Planning Support Device 110 Performance Data Acquisition Department 120 Estimation Model Generation Unit 121 Estimation section for component concentration by brand 123 Brand-Specific Ingredient Concentration Acquisition Section 130 Input section 140 Content estimation part 141 Similar Performance Data Extraction Unit 143 Similar Performance Reflection Section 150 Output section 170 Storage section 180 Formulation Planning and Adjustment Department 900 Information Processing Equipment 907 Bus 913 Storage device 915 Drive 917 Connection Ports 919 Communication equipment 921 Input device 923 Output device 925 Removable recording media 927 External equipment 929 Communications Network

Claims

1. A blending plan support method for furnace operation that produces molten steel by melting raw materials in which multiple scrap materials are blended in arbitrary proportions, which supports the formulation of a blending plan that specifies the blending proportions of the multiple scrap materials, An estimation model generation step generates an estimation model that outputs the content of a predetermined component in molten steel produced from raw materials blended with the aforementioned scrap materials in arbitrary proportions, based on multiple historical data that associate the blending ratio of the aforementioned scrap materials in past furnace operations with the component content of the molten steel. A content estimation step involves using the estimation model to output an estimated value of the content of a predetermined component in molten steel produced from raw materials based on a blending plan that specifies the blending ratio of the multiple scrap materials, Includes, The aforementioned estimation model generation step is: A method for supporting blending planning, which generates the estimation model using the probability distribution of component concentrations for each of the multiple scrap brands obtained based on the multiple performance data mentioned above.

2. The estimation model generation step is: The blending plan support method according to claim 1, wherein the probability distribution of the component concentrations for each brand is updated sequentially each time the component content data for the molten steel is obtained, thereby generating the estimation model.

3. A blending plan support method for assisting in the formulation of a blending plan that specifies the blending ratios of the multiple scrap materials in a furnace operation that produces molten steel by melting a raw material in which multiple scrap materials are blended in arbitrary proportions, An estimation model generation step generates an estimation model that outputs the content of a predetermined component in molten steel produced from raw materials blended with the aforementioned scrap materials in arbitrary proportions, based on multiple historical data that associate the blending ratio of the aforementioned scrap materials in past furnace operations with the component content of the molten steel. A content estimation step involves using the estimation model to output an estimated value of the content of a predetermined component in molten steel produced from raw materials based on a blending plan that specifies the blending ratio of the multiple scrap materials, Includes, The aforementioned content estimation step is, A similar performance data extraction step involves extracting performance data from past reactor operations that includes performance data similar to the aforementioned blending plan, and A similar performance data extraction step outputs an estimated value of the content of a predetermined component in the molten steel produced from the raw materials based on the blending plan, based on the estimation model modified using the performance data extracted in the similar performance data extraction step. Includes, The aforementioned step of extracting similar performance data is: A blending plan support method that determines the similarity between the blending plan and the actual blending results based on the similarity of the blending ratios of scrap materials and the time distance of the actual data, and extracts the actual data including the blending results that are determined to be similar to the blending plan.

4. A blending plan support method for assisting in the formulation of a blending plan that specifies the blending ratios of the multiple scrap materials in a furnace operation that produces molten steel by melting a raw material in which multiple scrap materials are blended in arbitrary proportions, An estimation model generation step generates an estimation model that outputs the content of a predetermined component in molten steel produced from raw materials blended with the aforementioned scrap materials in arbitrary proportions, based on multiple historical data that associate the blending ratio of the aforementioned scrap materials in past furnace operations with the component content of the molten steel. A content estimation step involves using the estimation model to output an estimated value of the content of a predetermined component in molten steel produced from raw materials based on a blending plan that specifies the blending ratio of the multiple scrap materials, Includes, The aforementioned content estimation step is, The fitting calculation step includes calculating the probability that the estimated content of a predetermined component in the molten steel produced from the raw materials based on the aforementioned blending plan falls outside the acceptable content range, based on information regarding the probability distribution of the content of a predetermined component in the molten steel produced from the raw materials based on the aforementioned blending plan and a preset acceptable content range. A blending plan support method further includes a blending plan adjustment step, which involves adjusting the blending plan by evaluating the relationship between the probability calculated in the suitability calculation step and the cost of raw materials based on the blending plan.

Images