Parameter calculation device, parameter calculation program, and parameter calculation method

Abstract

To achieve accuracy enhancement of a simulation considering a condition specific to a production line of a steel material.SOLUTION: A parameter calculation device (100) includes: a simulator acquisition unit (41) for acquiring a discrete event simulator (S); an achievement data acquisition unit (42) for acquiring achievement data (D); an evaluation function value calculation unit (43) for calculating an evaluation function value on the basis of the discrete event simulator (S) and the achievement data (D); an acquired function value calculation unit (44) for calculating an acquired function value on the basis of the evaluation function value; and a parameter calculation unit (45) for calculating an optimum value of the parameter by Bayesian optimization using the acquired function value.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] The present invention relates to a parameter calculation device, a parameter calculation program, and a parameter calculation method.

Background Art

[0002] In order to efficiently manufacture a large number of steel materials, it is important to balance the capabilities of each facility in the production line so as to avoid logistics problems. Examples of such logistics problems include, for example, the problem that the storage capacity of the storage area where intermediate products (such as coils) before becoming steel materials are temporarily placed becomes insufficient and the previous process has to be stopped, and the problem that the subsequent process cannot be started because there are no intermediate products in the storage area. In order to avoid such problems, it is desirable to predict as accurately as possible the change in the inventory quantity of intermediate products in the storage area before starting production and to take measures to prevent logistics problems from occurring. Techniques for simulating the change in the inventory quantity of intermediate products in the storage area have been conventionally known. For example, in Patent Document 1, input information including information on the quantity of intermediate products and route information between storage areas is input, a mathematical model based on the input information is created, and the change in the inventory quantity of intermediate products in the storage area is output based on the mathematical model. A method for simulating the change in inventory quantity is described.

Prior Art Documents

Patent Documents

[0003]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0004] However, the prior art as described above treats the amount of intermediate products as a continuous amount based on weight. For this reason, the prior art has a problem that it is impossible to perform a simulation considering the conditions specific to the steel manufacturing line (for example, the number of intermediate products that can be transported at one time by the transport equipment is limited, and the transport equipment may be used for transporting other intermediate products and may not be able to transport the target intermediate product immediately). One aspect of the present invention aims to improve the accuracy of simulation considering the conditions specific to the steel manufacturing line.

Means for Solving the Problems

[0005] In order to solve the above problems, a parameter calculation device according to one aspect of the present invention is a parameter calculation device that calculates an optimal value of a parameter used in a discrete event simulator that predicts the transition of the number of stays of the intermediate product in a place where the intermediate product before becoming the steel material is temporarily placed, which is provided in the middle of the steel manufacturing line. The parameter calculation device includes a simulator acquisition unit that acquires the discrete event simulator, a performance data acquisition unit that acquires performance data indicating the actual transition of the number of stays when steel materials were manufactured in the past on the manufacturing line, an evaluation function value calculation unit that calculates an evaluation function value representing the relationship between one of a plurality of candidate values that can be the optimal value and the prediction accuracy of the discrete event simulator based on the discrete event simulator and the performance data, an acquisition function value calculation unit that calculates an acquisition function value based on the evaluation function value, and a parameter calculation unit that calculates the optimal value by Bayesian optimization using the acquisition function value.

[0006] The parameter calculation device according to each aspect of the present invention may be realized by a computer. In this case, a parameter calculation program of the parameter calculation device that realizes the parameter calculation device by operating the computer as each part (software element) provided in the parameter calculation device, and a computer-readable recording medium on which it is recorded also fall within the scope of the present invention.

[0007] Further, a parameter calculation method according to one aspect of the present invention is a parameter calculation method for calculating, using a computer, an optimal value of a parameter used in a discrete event simulator that predicts the transition of the number of intermediate products in a storage area provided in the middle of a steel material production line, where the intermediate products before becoming the steel material are temporarily placed. The method includes: a step of acquiring the discrete event simulator; a step of acquiring performance data indicating the actual transition of the number of stays when steel materials were produced in the past on the production line; a step of calculating an evaluation function value representing the relationship between one of a plurality of candidate values that can be the optimal value and the prediction accuracy of the discrete event simulator based on the discrete event simulator and the performance data; a step of calculating an acquisition function value based on the evaluation function value; and a step of calculating the optimal value by Bayesian optimization using the acquisition function value.

Effects of the Invention

[0008] According to one aspect of the present invention, it is possible to improve the accuracy of simulation considering the conditions specific to the steel material production line.

Brief Description of the Drawings

[0009]

Figure 1

Figure 2

Figure 3

Modes for Carrying Out the Invention

[0010] Hereinafter, an embodiment of the present invention will be described in detail.

[0011] [Discrete Event Simulator] First, an example of a discrete event simulator S (hereinafter referred to as simulator S) to which the parameters calculated by the parameter calculation device 100 (hereinafter referred to as calculation device 100) according to the present embodiment are applied will be described. FIG. 1 is a block diagram showing the configuration of simulator S.

[0012] To solve the problems of conventional simulators that treat intermediate products as continuous quantities, the simulator S used here is for predicting the transition of the number of intermediate products in a yard provided in the middle of a steel manufacturing line. The simulator S according to the present embodiment predicts the transition of the number of intermediate products by simulating the flow of intermediate products in the steel manufacturing line. Note that the simulator S may be a general one created using discrete event simulation technology.

[0013] The yard provided in the middle of the steel manufacturing line is a place where intermediate products are temporarily placed.

[0014] Intermediate products are those before becoming steel (final products). Intermediate products include, for example, steel slabs (slabs, blooms, billets) after continuous casting, coils after hot rolling and before cold rolling, coils after cold rolling and before surface treatment (such as plating), etc. Intermediate products can be counted by the number. That is, the number of intermediate products in residence is a discrete quantity.

[0015] The simulator S according to the present embodiment simulates the flow of intermediate products in a part of the manufacturing line. Specifically, as shown in FIG. 1, the simulator S according to the present embodiment simulates the flow of intermediate products from before the start of processing in the first factory 1 responsible for one process of steel manufacturing, through the intermediate yard 3, and until after the end of processing in the second factory 2 responsible for the next process. Note that the simulator S may simulate the flow of intermediate products among three or more factories.

[0016] The first factory 1 is any one of a casting factory, a hot rolling factory, a cold rolling factory, and an annealing factory.

[0017] The second factory 2 is a factory responsible for the next process after leaving the first factory 1 among the hot rolling factory, cold rolling factory, annealing factory, and continuous hot dip galvanizing factory.

[0018] The production line to be simulated includes at least one device and a storage area. The device according to this embodiment includes a processing device and a transfer device. That is, for the simulator according to this embodiment, the positions, operations, etc. of the processing device, storage area, and transfer device are set.

[0019] (Processing device) The processing device performs a predetermined process on the intermediate product. The processing device according to this embodiment includes a first processing device 11 and a second processing device 21.

[0020] The first processing device 11 is installed in the first factory 1. And the first processing device 11 performs a process corresponding to one step (continuous casting, hot rolling, cold rolling, or annealing) on the intermediate product.

[0021] The second processing device 21 is installed in the second factory 2. And the second processing device 21 performs a process corresponding to the next step (hot rolling, cold rolling, annealing, or plating) on the intermediate product.

[0022] (Storage area) The storage area is where the intermediate product is temporarily placed. The storage area according to this embodiment includes a first front storage area 12, a first front buffer 13, a first rear storage area 14, a first rear buffer 15, a second front storage area 22, a second front buffer 23, a second rear storage area 24, a second rear buffer 25, a temporary storage area 31, a first pallet 32, and a second pallet 33. Note that the storage area may not include the temporary storage area 31.

[0023] The first front storage area 12 is provided on the front (entrance) side of the first processing device 11 in the first factory 1.

[0024] The first front buffer 13 is provided between the first processing device 11 and the first front storage area 12. The intermediate products placed on the first front buffer 13 are sequentially set in the first processing device 11.

[0025] The first rear storage area 14 is provided on the rear (exit) side of the first processing device 11.

[0026] The first rear buffer 15 is provided between the first processing device 11 and the first rear storage area 14. The intermediate products processed by the first processing device 11 are sequentially placed on the first rear buffer 15.

[0027] The second front storage area 22 is provided on the front side of the second processing device 21 in the second factory 2.

[0028] The second front buffer 23 is provided between the second processing device 21 and the second front storage area 22. The intermediate products placed on the second front buffer 23 are sequentially set in the second processing device 21.

[0029] The second rear storage area 24 is provided on the rear side of the second processing device 21.

[0030] The second rear buffer 25 is provided between the second processing device 21 and the second rear storage area 24. The intermediate products processed by the second processing device 21 are sequentially placed on the second rear buffer 25.

[0031] The temporary storage area 31 is provided in the intermediate yard 3 (between the first factory 1 and the second factory 2).

[0032] The first pallet 32 is moved by a conveying device. The stationary first pallet 32 is adjacent to the entrance side of the first rear storage area 14 or the temporary storage area 31.

[0033] The second pallet 33 is moved by a conveying device. The stationary second pallet 33 is adjacent to the exit side of the temporary storage area 31 or the second front storage area 22.

[0034] (Conveying device) The conveying device conveys intermediate products. The conveying device according to this embodiment includes a crane and a carrier 35.

[0035] The crane lifts and conveys intermediate products. The crane according to this embodiment includes a first front crane 16, a first rear crane 17, a second front crane 26, a second rear crane 27, and a temporary storage crane 34.

[0036] The first front crane 16 is provided on the front side of the first processing device 11. A job of conveying the intermediate product placed on the first front storage area 12 to the first front buffer 13 is set for the first front crane 16.

[0037] The first rear crane 17 is provided on the rear side of the first processing device 11. A plurality of types of jobs are set for the first rear crane 17. Specifically, there are two types of jobs: a job of conveying the intermediate product placed on the first rear buffer 15 to the first rear storage area 14, and a job of conveying the intermediate product placed on the first rear storage area 14 to the first pallet 32. The first rear crane 17 cannot execute the other job while executing one job.

[0038] The second front crane 26 is provided on the front side of the second processing device 21. A plurality of types of jobs are set for the second front crane 26. Specifically, there are two types of jobs: a job of conveying the intermediate product placed on the first pallet 32 or the second pallet 33 to the second front storage area 22, and a job of conveying the intermediate product placed on the second front storage area 22 to the second front buffer 23. The second front crane 26 cannot execute the other job while executing one job.

[0039] The second rear crane 27 is provided on the rear side of the second processing device 21. A job of conveying the intermediate product placed on the second rear buffer 25 to the second rear storage area 24 is set for the second rear crane 27.

[0040] The temporary crane 34 is provided at the temporary storage area 31. Multiple types of jobs are set for the temporary crane 34. Specifically, there are two types of jobs: a job of transporting the intermediate product placed on the first pallet 32 to the temporary storage area 31, and a job of transporting the intermediate product placed in the temporary storage area 31 to the second pallet 33. The temporary crane 34 cannot execute the other job while executing one job.

[0041] Note that one type or three or more types of jobs may be set for at least any one of the first rear crane 17, the second front crane 26, and the temporary crane 34. Also, for at least one of the first front crane 16 and the second rear crane 27, two or more types of jobs may be set.

[0042] The carrier 35 loads and transports the intermediate product. The carrier 35 according to this embodiment loads and transports the first pallet 32 or the second pallet 33 on which at least one intermediate product is placed. Multiple transport destinations from the storage area are set for the carrier 35. Specifically, there are two, namely the second front storage area 22 and the temporary storage area 31. Note that one or three or more transport destinations may be set for the carrier 35. Also, the carrier may be divided into a first carrier that transports the first pallet and a second pallet that transports the second pallet.

[0043] By using the simulator S described above, it is possible to perform a simulation that takes into account the conditions specific to the steel manufacturing line (not present in manufacturing lines for fluids, etc.) (for example, transporting individual intermediate products with a transport device, there is a limit to the number of intermediate products that the transport device can carry at one time, there may be a case where the transport device is not available when you want to transport, etc.) and is in line with the actual steel manufacturing process.

[0044] Note that the simulator S is also used when the calculation device 100 calculates the optimal value of the parameter. Details of how the calculation device 100 uses the simulator S will be described later.

[0045] [Configuration of calculation device 100] Next, the configuration of the calculation device 100 will be described. FIG. 2 is a block diagram showing the functional configuration of the calculation device 100.

[0046] The calculation device 100 is for calculating the optimal values of the parameters used in the simulator S. As shown in FIG. 2, the calculation device 100 includes a control unit 4 and a storage unit 5. The calculation device 100 according to the present embodiment further includes an operation unit 6 and an output unit 7.

[0047] The storage unit 5 is composed of a non-volatile memory. The storage unit 5 according to the present embodiment stores the simulator S, the evaluation function F1, and the acquisition function F2. Details of the evaluation function F1 and the acquisition function F2 will be described later. Also, the storage unit 5 according to the present embodiment can store the performance data D and the search assistance information I. Details of the performance data D and the search assistance information I will be described later. Note that the calculation device 100 may be configured to receive at least any one of the simulator S, the performance data D, various mathematical formulas, and the search assistance information I from another device. In that case, the storage unit 5 does not have to store these.

[0048] The operation unit 6 is operated by the user. The operation unit 6 includes a keyboard, a mouse, a touch panel, and the like. The operation unit 6 inputs information corresponding to the operation made by the user to the control unit 4.

[0049] The output unit 7 outputs the calculation result of the control unit 4. The output unit 7 includes a display device for displaying the calculation result, a printer for printing the calculation result, a communication module for transmitting the data of the calculation result to another device, and the like.

[0050] The control unit 4 includes a simulator acquisition unit 41, a performance data acquisition unit 42, an evaluation function value calculation unit 43, an acquisition function value calculation unit 44, and a parameter calculation unit 45. Also, the control unit 4 according to the present embodiment further includes a condition setting unit 46, a search assistance information acquisition unit 47, and a search assistance information creation unit 48.

[0051] (Simulator Acquisition Unit) The simulator acquisition unit 41 acquires the above-mentioned simulator S. The simulator acquisition unit 41 according to the present embodiment is configured to read the simulator S from the storage unit 5. Note that the simulator acquisition unit 41 may be configured to read the simulator S from a medium inserted into the calculation device 100. It may also be configured to receive from another device via a communication module.

[0052] (Condition Setting Unit) The condition setting unit 46 sets simulation conditions including the transport capacity of the transport equipment and the storage capacity of the storage area in the simulator S. The condition setting unit 46 according to the present embodiment sets the simulation conditions input by the operation made by the user on the operation unit 6. The condition setting unit 46 according to the present embodiment also sets the processing capacity of the processing device as a simulation condition. The condition setting unit 46 according to the present embodiment sets the processing capacities of all the processing devices (the first processing device 11 and the second processing device 21). As shown in Table 1 below, the processing capacity includes the number of processable items and the processing time. The number of processable items is the number of intermediate products that the first processing device 11 and the second processing device 21 can process in one process. The processing time is the time required for one process. The processing time is a value that varies according to the coil attributes (size, quality, etc.).

[0053]

Table 1

[0054] Also, the condition setting unit 46 according to the present embodiment sets the transport capacities of all the transport devices (the first front crane 16, the first rear crane 17, the second front crane 26, the second rear crane 27, the temporary crane 34, and the carrier 35). As shown in Table 2 below, the transport capacity includes the number of transportable items and the transport time. The number of transportable items is the number of objects (intermediate products, the first pallet 32, the second pallet 33) that each transport device can transport simultaneously in one transport. The transport time is the time required for one transport.

[0055]

Table 2

[0056] As described above, two jobs are respectively set for the first rear crane 17 and the second front crane 26 of the simulator S. Therefore, the condition setting unit 46 according to the present embodiment sets job selection conditions for the first rear crane 17 and the second front crane 26 respectively.

[0057] For the first rear crane 17, job selection conditions are set as follows, for example. · When the number of coils in the first rear buffer 15 ≧ the storage capacity of the first rear buffer 15 - α books, prioritize the transfer from the first rear buffer 15 to the first rear storage area 14. · When the number of coils in the first rear buffer 15 < the storage capacity of the first rear buffer 15 - α books, prioritize the transfer from the first rear storage area 14 to the first pallet 32.

[0058] Also, for the second front crane 26, job selection conditions are set as follows, for example. · When the number of coils in the second front buffer 23 ≧ the storage capacity of the second front buffer 23 - β books, prioritize the transfer from the first pallet 32 or the second pallet 33 to the second front storage area 22. · When the number of coils in the second front buffer 23 < the storage capacity of the second front buffer 23 - β books, prioritize the transfer from the second front storage area 22 to the second front buffer 23.

[0059] α and β in the conditions are parameters for which the optimal values will be calculated hereafter. Also, the optimal values of the parameters are criteria (equivalent to the operator's judgment) when the device selects which job to execute. In this way, even when simulating the number of in-process products in a production line where the device cannot execute other jobs while executing a certain job (for example, two jobs of transporting from the previous process to the storage area and from the storage area to the subsequent process are set, and the other transportation cannot be performed while one transportation is in progress), the transition of the number of in-process products in the storage area can be predicted more accurately.

[0060] Also, the carrier 35 of the simulator S can transport a plurality of in-process products at a time. For this reason, the condition setting unit 46 according to the present embodiment sets a transport start condition for the carrier 35. The transport start condition is, for example, as follows. · Start transporting when the in-process products are loaded on the first pallet 32 or the second pallet 33 by a predetermined weight (for example, 135t). · When a predetermined time (for example, 1 hour) has elapsed since the first in-process product was loaded on the first pallet 32 or the second pallet 33, start transporting even if the total weight of the in-process products has not reached the predetermined weight.

[0061] In this way, even when simulating the number of in-process products in a production line where there are multiple transport patterns (the number of in-process products transported at a time) of the carrier 35, the transition of the number of in-process products in the storage area can be predicted more accurately.

[0062] Also, two transport destinations are set for the carrier 35 of the simulator S. For this reason, the condition setting unit 46 according to the present embodiment sets a transport destination selection condition for the carrier 35. The transport destination selection condition is, for example, as follows. · When there is space in the second front storage area 22, transport to the second front storage area 22. · When the second front storage area 22 is full, transport to the temporary storage area 31.

[0063] Note that the above conveyance destination selection conditions may be as follows. · When the storage capacity of the second front storage area 22 ≥ γ pieces, convey to the second front storage area 22. · When the storage capacity of the second front storage area 22 < γ pieces, convey to the temporary storage area 31.

[0064] In this case, γ in the conditions is one of the parameters for which the optimal value will be calculated from now on. Also, the optimal value of the parameter is a criterion when selecting to which conveyance destination the carrier 35 will convey. By doing so, even when using the simulator S in which the conveyance route of the intermediate product in the manufacturing line branches, it is possible to more accurately predict the transition of the number of intermediate products staying in the storage area.

[0065] In addition, the condition setting unit 46 according to the present embodiment sets the storage capacities of all storage areas (the first front storage area 12, the first front buffer 13, the first rear storage area 14, the first rear buffer 15, the second front storage area 22, the second front buffer 23, the second rear storage area 24, the second rear buffer 25, the temporary storage area 31, the first pallet 32, and the second pallet 33). As shown in Table 3 below, the condition setting unit 46 according to the present embodiment sets the storage capacities of the first pallet 32 and the second pallet 33 in weight (t), and sets the storage capacities of the other storage areas in number.

[0066]

Table 3

[0067] By setting the simulation conditions that the condition setting unit 46 has described above, it is possible to perform simulation using a general-purpose simulator S for which the simulation conditions are not set (the simulator S can be customized in the device). Also, by doing so, even when a simulator S for which the simulation conditions are set in advance is acquired, it is possible to adjust the simulation conditions later.

[0068] Here, the case of setting job selection conditions for the first rear crane 17 and the second front crane 26 has been exemplified. However, the condition setting unit 46 may be configured to set job selection conditions for either the first rear crane 17 or the second front crane 26 for other devices. Further, the condition setting unit 46 may be configured to set job selection conditions for cranes other than the first rear crane 17 and the second front crane 26. Also, here, the case of setting destination selection conditions for the carrier 35 has been exemplified, but the destination selection conditions may be set for the crane. Moreover, here, the case of setting simulation conditions based on the user's input (manually) has been exemplified, but the condition setting unit 46 may be configured to automatically set the simulation conditions.

[0069] (Actual result data acquisition unit) The actual result data acquisition unit 42 acquires actual result data D. The actual result data D relates to the actual transition of the number of stays when steel materials were manufactured in the production line in the past.

[0070] (Evaluation function value calculation unit) The evaluation function value calculation unit 43 calculates an evaluation function value based on the simulator S and the actual performance data D. The evaluation function value represents the relationship between one of a plurality of candidate values that can be an optimal value and the prediction accuracy of the simulator S. As described above, the simulator S according to the present embodiment has a plurality (n) of types of parameters for which an optimal value is to be calculated. Therefore, the evaluation function value according to the present embodiment represents the relationship between one of a plurality of combinations (hereinafter, combinations of candidate values) obtained by selecting one candidate value from each parameter and the prediction accuracy of the simulator S. The number of the combinations of candidate values can be calculated by (the number of candidate values that the first type of parameter can take) × (the number of candidate values that the second type of parameter can take) × ··· × (the number of candidate values that the nth type of parameter can take). The evaluation function value calculation unit 43 according to the present embodiment calculates information regarding the difference between the actual value of the residence number obtained from the actual performance data D and the predicted value of the residence number predicted by the simulator S as the evaluation function value J(x). Specifically, Equation (1) stored in the storage unit is acquired, and the predicted value and the actual value are substituted into Equation (1) for calculation.

[0071] [Equation]

[0072] Here, x is a combination of candidate values [α, β ···], and ŷ t is the predicted value of the residence number at time t, and y t is the actual value. In this way, the evaluation function value can be easily calculated based on the actual performance data D and the predicted value obtained by a simple method of periodically counting the residence number of the intermediate products in the storage location.

[0073] Note that the evaluation function value is not limited to that of the above embodiment as long as it relates to the difference between the predicted value and the actual value. For example, the evaluation function value calculation unit 43 may calculate, as the evaluation function value, information regarding the difference between the actual value at the time when the intermediate product reaches a predetermined state obtained from the performance data D and the predicted value at the time predicted by the simulator S. The time when the intermediate product reaches the predetermined state includes the time when the processing by the processing device starts, the time when the processing ends (including the time when the final processing ends and the steel material (finished product) is obtained), the time when it is placed on the storage yard, and the like.

[0074] Also, generally, a plurality of intermediate products are manufactured. For example, when the predicted values and the measured values for each intermediate product at the time when the processing by the first processing device 11 is completed, as shown in Table 4 below, are obtained, the sum of the absolute values of the time differences of each intermediate product (|3| + |-2| + |3| = 8) may be used as the evaluation function value, or the time difference (3) of the last intermediate product C may be used as the evaluation function value.

[0075]

Table 4

[0076] The difference between the actual value and the predicted value of the time has a correlation with the number of intermediate products staying in the storage yard. Therefore, in this way, for example, even when the number of intermediate products staying in the storage yard cannot be directly counted, the evaluation function value can be calculated.

[0077] (Search assistance information acquisition unit) The search assistance information acquisition unit 47 acquires search assistance information I. The search assistance information I is related to the search for the optimal values of parameters when steel materials were manufactured in the production line in the past. For example, when there are two types of parameters α and β for which the optimal values are to be calculated, the search assistance information I becomes a numerical group arranged in a matrix of the number of candidate values that α can take × the number of candidate values that β can take. Each candidate value constituting the numerical group indicates that the higher the value is closer to 1, the higher the priority of the search. Note that the search assistance information I may not be the above-mentioned candidate value group, but a tendency readable from the candidate value group or the memory of a person (such as an operator, a manager of the production line, etc.) (for example, an empirical rule that the flow of intermediate products seems to be good when α = 4).

[0078] (Acquisition function value calculation unit) The acquisition function value calculation unit 44 calculates an acquisition function value based on the evaluation function value. The acquisition function value calculation unit 44 according to the present embodiment calculates, as the acquisition function value, the expected value of the amount that can be improved from the minimum value of the evaluation function values obtained so far (Expected Improvement (EI)). Specifically, the following Equation 2 stored in the storage unit is acquired, and the necessary numerical values are substituted into the Equation 2 for calculation. Further, the acquisition function value calculation unit 44 according to the present embodiment calculates a plurality of acquisition function values corresponding to one combination of candidate values as one acquisition function value group.

[0079] [Equation]

[0080] Here, x is a combination of candidate values [α, β ···], μ(x) and σ(x) are the mean value and the standard deviation calculated by a Gaussian process regression model constructed from the observed values obtained so far, φ is the standard Gaussian function, and Φ is its cumulative density function. Note that the acquisition function value is not limited to that of the above embodiment as long as it can be an index indicating the balance between the region where the optimal value is likely to exist and the region with high uncertainty.

[0081] When the search assistance information I has been obtained in advance, the acquisition function value calculation unit 44 calculates the acquisition function value based on the evaluation function value and the search assistance information I. Specifically, the acquisition function value calculation unit 44 calculates the acquisition function value using, for example, the following equation 3 instead of the above equation 2.

[0082] [Equation]

[0083] Here, Prio(x) is the numerical value corresponding to x in the search assistance information I. That is, this equation 3 is obtained by multiplying the above equation 2 by one of the numerical value groups constituting the search assistance information I. In this way, based on the search assistance information I regarding the search for the optimum value, it is possible to narrow down the search candidates (combinations of candidate values) with high priority, so that the optimum value of the parameter can be calculated more quickly.

[0084] (Parameter calculation unit) The parameter calculation unit 45 calculates the optimum value of the parameter by Bayesian optimization using the acquisition function value. The parameter calculation unit 45 according to the present embodiment repeats a set of a series of operations (the operation of causing the evaluation function value calculation unit 43 to calculate the evaluation function value corresponding to the maximum (any only for the first time) acquisition function value, the operation of causing the acquisition function value calculation unit 44 to calculate a group of acquisition function values, and the operation of selecting the maximum acquisition function value from the group of acquisition function values) a predetermined number of times or until the selected maximum acquisition function value becomes equal to or less than a predetermined reference value.

[0085] (Search assistance information creation unit) The search assistance information creation unit 48 calculates search assistance information I. The search assistance information creation unit 48 according to the present embodiment first calculates a new evaluation function value corresponding to a combination of candidate values not calculated by the evaluation function value calculation unit 43 by Gaussian process regression using the calculated evaluation function values. Further, the search assistance information creation unit 48 according to the present embodiment normalizes an evaluation function value group including the calculated new evaluation function value and the evaluation function value calculated by the evaluation function value calculation unit 43 to 0-1 so that the error becomes smaller as it approaches 1. Further, the search assistance information creation unit 48 according to the present embodiment stores the normalized evaluation function value group in the storage unit 5 as new search assistance information I.

[0086] In the manufacturing line, the operator's judgment rules are periodically changed. Therefore, it is necessary to calculate the optimal value of the parameter using the calculation device 100 in accordance with the change of the judgment rule. The new search assistance information I calculated by the search assistance information creation unit 48 can be used as the search assistance information I acquired by the search assistance information acquisition unit 47 when calculating the optimal value of the parameter after the next time.

[0087] [Flow of Parameter Calculation Method] Next, a parameter calculation method (hereinafter, calculation method) using the calculation device 100 will be described. FIG. 3 is a flowchart showing the flow of the calculation method.

[0088] First, when the user performs a predetermined operation (for example, power-on operation, predetermined start operation, etc.) on the calculation device 100, as shown in FIG. 3, the simulator acquisition unit 41 acquires the simulator S (step S1).

[0089] In the calculation method according to the present embodiment, after acquiring the simulator S, the calculation device 100 receives an input of simulation conditions. When the simulation conditions are input by the user, the condition setting unit 46 sets the input simulation conditions (step S2). Note that when the simulator acquisition unit 41 acquires a simulator S in which the simulation conditions are set in advance, this step 2 may be omitted.

[0090] After setting the simulation conditions, the control unit 4 acquires the evaluation function F1 (Equation 1) (step S3). In the calculation method according to the present embodiment, the data of the evaluation function F1 is read from the storage unit 5.

[0091] In the calculation method according to the present embodiment, after acquiring the evaluation function F1, the control unit 4 determines whether or not the search assistance information I exists (step S4). Specifically, the control unit 4 determines whether, for example, it is stored in the storage unit 5 of the calculation device 100.

[0092] When it is determined that the search assistance information I exists (step S4: Yes), the search assistance information acquisition unit 47 acquires the search assistance information I (step S5). On the other hand, when it is determined that the search assistance information I does not exist, step S5 is skipped.

[0093] After acquiring the search assistance information I or skipping step S5, the control unit 4 acquires the acquisition function F2 (step S6). In the calculation method according to the present embodiment, the data of the acquisition function F2 is read from the storage unit 5. When the search assistance information I is acquired in step S5, the control unit 4 acquires the acquisition function F2 of Equation 3. On the other hand, when the search assistance information I is not acquired in step S5, the control unit 4 acquires the acquisition function F2 of Equation 2.

[0094] After acquiring the acquisition function F2, the performance data acquisition unit 42 acquires the performance data D (step S7). In the calculation method according to the present embodiment, the performance data D is read from the storage unit 5.

[0095] Here, the case where the evaluation function F1, the search assistance information I, the acquisition function F2, and the performance data D are acquired in this order is exemplified. However, if these data are ready by the time step S8 starts, the acquisition order is arbitrary.

[0096] After obtaining the evaluation function F1, the search assistance information I, the acquisition function F2, and the performance data D, the evaluation function value calculation unit 43 calculates an initial evaluation function value based on any combination of candidate values (step S8). The determination of the combination of candidate values may be made by the user or automatically by the calculation device 100.

[0097] After calculating the evaluation function value, the acquisition function value calculation unit 44 calculates an acquisition function value group based on the evaluation function value (step S9).

[0098] After generating the acquisition function value group, the parameter calculation unit 45 calculates the optimal value of the parameter (step S10). Specifically, the parameter calculation unit 45 first selects the maximum acquisition function value from the acquisition function value group (step S11). After selecting the maximum acquisition function value, the parameter calculation unit 45 determines whether the selected maximum acquisition function value is less than or equal to a reference value (step S12). Here, if it is determined that it is not less than or equal to the reference value, the evaluation function value calculation unit 43 calculates a new evaluation function value based on the combination of candidate values corresponding to the maximum acquisition function value (step S13). Then, the acquisition function value calculation unit 44 calculates the acquisition function value group based on the new evaluation function value (step S9). Through such a process, steps S9, S11 to S13 are repeated until the maximum acquisition function value among those less than or equal to the reference value is found. On the other hand, if it is determined in step S12 that it is less than or equal to the reference value, the parameter calculation unit 45 outputs the combination of candidate values corresponding to the selected maximum acquisition function value to the output unit as the optimal value (step S14). Note that in step S12, instead of determining whether the maximum acquisition function value is less than or equal to the reference value, it may be determined whether steps S13, S9, and S10 have been repeated a predetermined number of times.

[0099] After calculating the optimal value of the parameter, the search assistance information creation unit 48 calculates the search assistance information I (step S15).

[0100] [Function and effect] In order to improve the prediction accuracy of the simulator S for predicting the discrete quantity, it is necessary to accurately set the simulation conditions. However, various judgments that affect the number of intermediate products staying in the storage area (corresponding to the optimal values of the parameters in the simulator S) had some ambiguous parts depending on the operator's experience. For this reason, it is difficult to accurately set all the simulation conditions. However, according to the calculation device according to the present embodiment, based on the actual performance data D, it is possible to obtain the optimal values of the parameters that are closer to the operator's judgment. Also, when the number of types of parameters and the number of candidate values that the parameters can take are small, it is not impossible to check all combinations of the candidate values one by one by brute force, but this would take too much time. In particular, as the number of types of parameters increases, the combinations of candidate values increase exponentially, so it becomes impossible to finish the check within a realistic time. However, the calculation device 100 according to the present embodiment predicts the optimal value using Bayesian optimization. Therefore, according to the calculation device 100 according to the present embodiment, it is possible to more quickly obtain the combination of candidate values that becomes the optimal value (closer to the operator's judgment). Also, according to the calculation device 100 according to the present embodiment, even when the number of types of parameters is large, it is possible to obtain the combination of candidate values that becomes the optimal value within a realistic time.

[0101] Also, if the simulation conditions are set using this optimal value, it is possible to realize an improvement in the accuracy of the simulation considering the conditions specific to the steel manufacturing line. Also, if the obtained optimal value is left as equivalent to the know-how of a skilled operator's judgment, an inexperienced operator can use it in the absence of a skilled operator, etc. Also, according to such a configuration, by improving the accuracy of the simulator S, the operating efficiency of each device in the production line is improved, and the amount of energy consumed can be reduced. Thereby, it can contribute to the achievement of the sustainable development goals (SDGs).

Example

[0102] An example of the present invention will be described below.

[0103] First, the following job selection conditions (judgment rules) were set for the first rear crane 17 of the above simulator S (manufacturing line). · When the number of coils in the first rear buffer 15 ≥ the storage capacity of the first rear buffer 15 - α pieces, the conveyance from the first rear buffer 15 to the first rear storage area 14 is prioritized. · When the number of coils in the first rear buffer 15 < the storage capacity of the first rear buffer 15 - α pieces, the conveyance from the first rear storage area 14 to the first pallet 32 is prioritized.

[0104] Then, α in the job selection conditions was set as one of the parameters for which the optimum value is to be calculated. The storage capacity of the first rear buffer 15 is 5 pieces, and since the above case classification does not hold for α = 5, the possible values of α are five values: 0, 1, 2, 3, and 4.

[0105] Also, the following job selection conditions (judgment rules) were set for the second front crane 26. · When the number of coils in the second front buffer 23 ≥ the storage capacity of the second front buffer 23 - β pieces, the conveyance from the first pallet 32 or the second pallet 33 to the second front storage area 22 is prioritized. · When the number of coils in the second front buffer 23 < the storage capacity of the second front buffer 23 - β pieces, the conveyance from the second front storage area 22 to the second front buffer 23 is prioritized.

[0106] Then, β in the job selection conditions was set as another parameter for which the optimum value is to be calculated. The storage capacity of the second front buffer 23 is 5 pieces, and since the above case classification does not hold for β = 5, the possible values of β are five values: 0, 1, 2, 3, and 4.

[0107] Also, in this embodiment, the search auxiliary information I was acquired in advance. Since α and β according to this embodiment can each take five values, the search auxiliary information I is a numerical group in a 5×5 = 25 matrix form as shown in Table 5. Also, in this embodiment, the reference value serving as the criterion for ending the search was set to 3.

[0108]

Table 5

[0109] Next, the combination of candidate values (α, β) of the simulator S was set to (0, 0) and (4, 4), and the first simulation was performed to calculate the predicted value of the number of coils in residence. Also, the (α, β) of the judgment rule of the production line was set to (0, 0) and (4, 4), and the first line operation was performed to obtain the measured value of the number of coils in residence. When J(x) was calculated based on the predicted value and the measured value, J(0, 0) = 110.62 and J(4, 4) = 25.22 were obtained.

[0110] Next, based on the smallest of the obtained J(x) (J(4, 4)) and the acquisition function F2, the first acquisition function value group as shown in Table 6 was calculated. In this embodiment, the acquisition function value group was calculated taking into account the above search assistance information I. Among the obtained acquisition function value groups, EI(4, 3) = 6.67 was the maximum acquisition function value.

[0111] [Table 6]

[0112] Next, the (α, β) of the simulator S was set to (4, 3), and the second simulation was performed to calculate the predicted value of the number of coils in residence. Also, the (α, β) of the judgment rule of the production line was set to (4, 3), and the second line operation was performed to obtain the measured value of the number of coils in residence. When J(x) was calculated based on the predicted value and the measured value, J(4, 3) = 20.98 was obtained.

[0113] Next, based on the obtained J(4, 3), the acquisition function F2, and the search assistance information I, the second acquisition function value group as shown in Table 7 was calculated. Among the obtained acquisition function value groups, EI(4, 0) = 3.70 was the maximum acquisition function value.

[0114] [Table 7]

[0115] Next, the combination of candidate values (α, β) of the simulator S was set to (4, 0), and the third simulation was performed to calculate the predicted value of the number of coils staying. Also, the combination of (α, β) of the determination rule of the production line was set to (4, 0), and the third line operation was performed to obtain the measured value of the number of coils staying. When J(x) was calculated based on the predicted value and the measured value, J(4, 0) = 20.89 was obtained.

[0116] Next, based on the obtained J(4, 0), the acquisition function F2, and the search assistance information I, the third acquisition function value group as shown in Table 8 was calculated. Among the obtained acquisition function value groups, EI(4, 1) = 2.83 was the maximum acquisition function value. At this time, since EI(4, 1) was less than the reference value of 3, the search was completed here. As reference information, when searching with all combinations of candidate values, it was confirmed that the evaluation function value J(x) was minimized when (α, β) was set to (4, 0). In this way, it was possible to reach the optimal value with fewer times (3 times in this embodiment) than simulating and operating the line for all combinations of candidate values (25 cases).

[0117]

Table 8

[0118] Next, new search assistance information I was calculated. By Gaussian process regression using the already calculated J(0, 0), J(4, 4), J(4, 3), J(4, 0), J(x) was calculated for combinations of other candidate values, and an evaluation function value group as shown in Table 9 was obtained. Then, the evaluation function value group shown in Table 9 was normalized to 1 - 0, and search assistance information I as shown in Table 10 was obtained.

[0119]

Table 9

[0120]

Table 10

[0121] Table 9 shows that when exploring the range of α = 4 and β = 1 to 3 in the calculation of the next optimal value, it is possible to explore efficiently.

[0122] 〔Example of Software Implementation〕 The functions of the above calculation device 100 can be a parameter calculation program for causing a computer to function as the calculation device 100, and can be realized by a parameter calculation program for causing a computer to function as each control block (particularly each part included in the control unit 4) of the calculation device 100. In this case, the above calculation device 100 includes a computer having at least one control device (for example, the control unit 4) and at least one storage device (for example, a memory) as hardware for executing the above program. By executing the above parameter calculation program with this control device and storage device, each function described in the above embodiments is realized. The above parameter calculation program may be recorded on one or more computer-readable recording media, not temporarily. This recording medium may or may not be provided in the above calculation device 100. In the latter case, the above parameter calculation program may be supplied to the above device via any wired or wireless transmission medium.

[0123] Also, part or all of the functions of the above control blocks can also be realized by a logic circuit. For example, an integrated circuit in which a logic circuit functioning as each of the above control blocks is formed is also included in the scope of the present invention. In addition to this, for example, it is also possible to realize the functions of the above control blocks by a quantum computer.

[0124] Also, each process described in the above embodiments may be executed by AI (Artificial Intelligence). In this case, the AI may operate in the above calculation device, or may operate in another device (for example, an edge computer or a cloud server, etc.).

[0125] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope shown in the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention.

Explanation of Signs

[0126] S Discrete event simulator 1 First factory 11 First processing device 12 First front yard 13 First front buffer 14 First rear yard 15 First rear buffer 16 First front crane 17 First rear crane 2 Second factory 21 Second processing device 22 Second front yard 23 Second front buffer 24 Second rear yard 25 Second rear buffer 26 Second front crane 27 Second rear crane 3 Intermediate yard 31 Temporary yard 32 First pallet 33 Second pallet 34 Temporary crane 35 Carrier 100 Parameter calculation device 4 Control unit 41 Simulator acquisition unit 42 Performance data acquisition unit 43 Evaluation function value calculation unit 44 Acquisition function value calculation unit 45 Parameter calculation unit 46 Condition setting unit 47 Search assistance information acquisition unit 5 Storage unit D Performance data F1 Evaluation function F2 Acquisition function I Exploration Assistance Information 6 Operation Unit 7 Output Unit

Claims

1. A parameter calculation device that calculates an optimal value of a parameter used in a discrete event simulator for predicting the transition of the number of intermediate products in a storage area where the intermediate products before becoming the steel material are temporarily placed, provided in the middle of a steel material manufacturing line, comprising: a simulator acquisition unit that acquires the discrete event simulator; a performance data acquisition unit that acquires performance data indicating the actual transition of the number of stays when steel materials were manufactured in the past on the manufacturing line; an evaluation function value calculation unit that calculates an evaluation function value representing the relationship between one of a plurality of candidate values that can be the optimal value and the prediction accuracy of the discrete event simulator based on the discrete event simulator and the performance data; an acquisition function value calculation unit that calculates an acquisition function value based on the evaluation function value; a parameter calculation unit that calculates the optimal value by Bayesian optimization using the acquisition function value. A parameter calculation device having the above components.

2. The evaluation function value calculation unit calculates, as the evaluation function value, information regarding the difference between the actual value of the number of stays obtained from the performance data and the predicted value of the number of stays predicted by the discrete event simulator. The parameter calculation device according to Claim 1.

3. The evaluation function value calculation unit calculates, as the evaluation function value, information regarding the difference between the actual value of the time when the intermediate product reaches a predetermined state obtained from the performance data and the predicted value of the time predicted by the discrete event simulator. The parameter calculation device according to Claim 1.

4. The parameter calculation device further includes a search auxiliary information acquisition unit that acquires search auxiliary information regarding the search for the parameter when steel materials were manufactured in the past on the manufacturing line, and the acquisition function value calculation unit calculates the acquisition function value based on the evaluation function value and the search auxiliary information. The parameter calculation device according to any one of Claims 1 to 3.

5. A plurality of types of jobs are set for at least one device included in the manufacturing line, and the parameter is a criterion when selecting which job the device executes. The parameter calculation device according to any one of Claims 1 to 4.

6. The device includes a transfer device that transfers the intermediate product, and a plurality of transfer destinations from the storage area are set for the transfer device, and the parameter is a criterion when selecting which transfer destination the transfer device transfers to. The parameter calculation device according to Claim 5.

7. The parameter calculation device according to claim 1, further comprising a condition setting unit that sets simulation conditions including the conveyance capacity of the conveyance device and the storage capacity of the storage site in the discrete event simulator.

8. A parameter calculation program for causing a computer to function as the parameter calculation device according to claim 1, A parameter calculation program for causing a computer to function as the simulator acquisition unit, the performance data acquisition unit, the evaluation function value calculation unit, the acquisition function value calculation unit, and the parameter calculation unit.

9. A parameter calculation method for calculating, using a computer, an optimum value of a parameter used in a discrete event simulator that predicts a change in the number of intermediate products in a storage site where an intermediate product before becoming the steel material is temporarily placed, provided in the middle of a steel material manufacturing line, a step of acquiring the discrete event simulator; a step of acquiring performance data indicating an actual change in the number of stays when steel materials were manufactured in the past on the manufacturing line; a step of calculating an evaluation function value representing a relationship between one of a plurality of candidate values that can be the optimum value and the prediction accuracy of the discrete event simulator based on the discrete event simulator and the performance data; a step of calculating an acquisition function value based on the evaluation function value; and a step of calculating the optimum value by Bayesian optimization using the acquisition function value.

Images