Reproduction condition specifying device and reproduction condition specifying method

Abstract

To provide technology of identifying regeneration conditions for retaining the amount of a binder eluted from regenerated casting sand at a target value.SOLUTION: A regeneration condition identification device (1) comprises one or plural processors practicing a regeneration condition identification step of, from the amount of a binder eluted from casting sand regenerated by a sand regeneration system, identifying regeneration conditions in the sand regeneration system.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to an apparatus and method for managing the sand properties of foundry sand reclaimed by a sand reclamation system. [Background technology]

[0002] There are known techniques for managing the properties of molding sand used in molding molds. For example, Patent Document 1 describes a system that utilizes measurements from a molding sand testing device to retrieve and display data from a knowledge base that describes molding sand management methods that have traditionally been determined empirically by sand management engineers.

[0003] Also, Patent Document 2 describes a computer-based system for sand optimization aimed at reducing casting rejects in foundries (casting plants). In the system described in Patent Document 2, a processor module calculates correlations between parameter values, parameters, rejects, and rejection types, such as the pH value of sand, and generates a normative solution. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 07-051791 (published on February 28, 1995) [Patent Document 2] Special Publication No. 2016-508882 (published on March 24, 2016) Summary of the Invention [Problem to be solved by the invention]

[0005] However, the quality of a mold made from recycled foundry sand varies depending on the sand properties of the foundry sand (particularly, the amount of binder eluted from the foundry sand). Therefore, it is important to stabilize the sand properties of the recycled foundry sand, particularly the amount of binder eluted. In order to stabilize the amount of binder eluted, it is necessary to appropriately set the regeneration conditions (such as plant control parameters) in the foundry sand regeneration process. However, how the regeneration conditions should be set depends in part on the experience of the manager of the regeneration process, and it has been difficult to appropriately set the regeneration conditions to stabilize the amount of binder eluted. Even with the techniques described in the above-mentioned Patent Documents 1 and 2, it has been difficult to appropriately maintain the amount of binder eluted from the recycled foundry sand.

[0006] One aspect of the present invention aims to realize a technology for specifying regeneration conditions for maintaining the amount of binder leaching from foundry sand regenerated by a sand reclamation system at a target value. [Means for solving the problem]

[0007] In order to solve the above problems, a reproduction condition specifying device according to an aspect of the present invention includes one or more processors that execute a reproduction condition specifying step. Also, a reproduction condition specifying method according to an aspect of the present invention includes a reproduction condition specifying step.

[0008] In the regeneration condition specifying device and the regeneration condition specifying method, the regeneration condition specifying step is performed by the processor detecting the amount of leaching material dissolved from the foundry sand regenerated by the sand reclamation system. Binder This is a step of determining the regeneration conditions in the sand reclamation system based on the amount of sand. Effect of the Invention

[0009] According to one aspect of the present invention, it is possible to specify regeneration conditions for maintaining the amount of binder leaching from foundry sand regenerated by a sand reclamation system at a target value. [Brief description of the drawings]

[0010] [Figure 1] 1 is a diagram showing a configuration of a sand management system according to a first embodiment of the present invention. [Diagram 2] FIG. 2 is a block diagram showing the configuration of a sand reclamation system included in the sand management system of FIG. 1. [Diagram 3] 2 is a block diagram showing the configuration of a regeneration condition specifying device included in the sand management system of FIG. 1. [Figure 4] 4 is a flowchart showing the flow of a reproduction condition specifying method implemented by the reproduction condition specifying device of FIG. 3. [Diagram 5] FIG. 1 is a diagram illustrating a genetic algorithm. [Figure 6] 10 is a flowchart showing a flow of a relational expression specifying step executed by the regeneration condition specifying device. [Figure 7] 10 is a flowchart showing the flow of processing performed by a reproduction condition specifying device. [Figure 8] FIG. 13 is a diagram illustrating an example of the contents of a crossover. [Figure 9] FIG. 1 illustrates the contents of a subtree mutation. [Figure 10] FIG. 1 illustrates the content of a hoist mutation. [Figure 11] FIG. 1 illustrates the contents of point mutations. [Figure 12] FIG. 11 is a diagram showing the configuration of a sand management system according to a second embodiment of the present invention. [Figure 13] 13 is a flowchart showing a flow of a reclamation condition specifying method implemented by a reclamation condition specifying device included in the sand management system of FIG. 12. [Figure 14] FIG. 1 illustrates an example of a trained model and a genetic algorithm. [Figure 15] FIG. 13 is a block diagram showing the configuration of a machine learning device included in the sand management system of FIG. 12. [Figure 16] 16 is a flowchart showing the flow of a machine learning method implemented by the machine learning device of FIG. 15 . DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0011] [First embodiment] (Sand management system) The configuration of a sand management system S1 according to a first embodiment of the present invention will be described with reference to Fig. 1. Fig. 1 is a block diagram showing the configuration of the sand management system S1.

[0012] The sand management system S1 is a system for managing molding sand used in molding a casting mold. As shown in FIG. 1, the sand management system S1 includes a casting system S11, a sand reclamation system S12, a reclamation condition specifying device 1, and a data logger 2.

[0013] The casting system S11 is a system for casting using molding sand (a mixture of recycled sand and new sand, which will be described later). The casting system S11 adds additives including a binder to the molding sand, kneads the mixture, and fills a molding flask with the kneaded molding sand to create a mold. The binder is, for example, an inorganic material such as water glass. The mold to be created is, for example, a main mold or a core. The casting system S11 pours molten metal into the mold and then cools the molten metal poured into the mold. The cooled molten metal solidifies inside the mold to become a casting. The casting system S11 applies vibration to the mold to dismantle the mold and turn it into a block of sand, and removes the casting obtained by cooling.

[0014] The sand reclamation system S12 is a system that reclaims foundry sand from the lump of sand obtained by the casting system S11. The configuration of the sand reclamation system S12 will be described later with reference to the accompanying drawings.

[0015] The regeneration condition identification device 1 is a device for implementing the regeneration condition identification method M1. The regeneration condition identification method M1 includes a relational expression identification step of identifying a nonlinear relational expression representing the relationship between the regeneration condition data and the binder data by using a nonlinear regression algorithm, with reference to a set of regeneration condition data indicating the regeneration conditions in the sand reclamation system S12 and binder data indicating the amount of binder eluted from the casting sand regenerated by the sand reclamation system S12. The regeneration condition identification method M1 also includes a regeneration condition identification step of identifying the regeneration condition data by calculation using the binder data and the identified relational expression. The configuration of the regeneration condition identification device 1 and the flow of the regeneration condition identification method M1 will be described in detail below with reference to the drawings.

[0016] The binder data is data representing the amount of binder eluted from the recycled foundry sand. The binder data includes, for example, data representing at least one of the amount of eluted binder and the amount of residual binder. The amount of eluted binder refers to the amount of binder eluted from the foundry sand at the temperature at which the foundry sand is kneaded. The amount of residual binder refers to the amount of binder eluted from the foundry sand at a predetermined temperature higher than the temperature at which the foundry sand is kneaded. The amount of eluted binder or the amount of residual binder is represented, for example, by pH, conductivity, or strength.

[0017] The reclamation condition data is data that indicates the reclamation conditions in the sand reclamation system S12. Details of the reclamation conditions will be described later with reference to the drawings.

[0018] The data logger 2 collects binder data and regeneration condition data, and provides the collected binder data and regeneration condition data to the regeneration condition identifying device 1. The data logger 2 can be configured, for example, by a PLC (Programmable Logic Controller) or an IPC (Industrial PC).

[0019] (Sand Reclamation System) 2 is a block diagram showing the configuration of the sand reclamation system S12. As shown in FIG. 2, the sand reclamation system S12 includes a crusher 21, a fine powder extractor 22, a dust collector 23, a fluidized bed baking device 24, a sand cooler 25, a sand polisher 26, a dust collector 27, a sand classifier 28, and a dust collector 29.

[0020] The crusher 21 crushes the lumps of sand obtained in the casting system S11 into sand grains. The collection of sand grains obtained by crushing with the crusher 21 contains not only the foundry sand grains to be recycled, but also fine powders other than foundry sand, such as silica and water glass.

[0021] The fine powder removing device 22 separates fine powder other than foundry sand (silica, water glass, etc.) from the aggregate of sand grains obtained by the crusher 21. The fine powder removing device 22 is equipped with a blower 221 that blows air to the sand grains. The fine powder is collected in a dust collector 23 by the air blown by the blower 221. The fine powder removing device 22 measures the air volume in the fine powder removing device 22 and outputs the measured value to a data logger 2.

[0022] The dust collector 23 collects the fine powder removed from the sand grains by the fine powder removal device 22. The dust collector 23 measures the air volume in the dust collector 23 and outputs the measured value to the data logger 2.

[0023] The fluidized bed incinerator 24 is an incineration device that heats the granular foundry sand from which fine powder has been removed. The fluidized bed incinerator 24 transports the foundry sand from the fine powder removal device 22 by suction or pressure, and heats the transported foundry sand. The water of crystallization contained in the water glass crystals evaporates as a result of heating. As the water of crystallization evaporates, voids are created in the water glass crystals. This makes the binders such as water glass contained in the foundry sand brittle, making the sand easier to crush.

[0024] A preset amount of foundry sand (hereinafter also referred to as the "foundry sand input amount") is input into the fluidized bed baking device 24. Here, the foundry sand input amount refers to, for example, the weight or volume of foundry sand input into the fluidized bed baking device 24 per unit time (in the case of continuous processing) or per processing (in the case of batch processing). The fluidized bed baking device 24 provides the set foundry sand input amount to the data logger 2.

[0025] In addition, the fluidized roasting device 24 measures the heating temperature of the fluidized roasting device 24 and outputs the measured value to the data logger 2.

[0026] The sand cooling device 25 cools the heated foundry sand. The sand cooling device 25 is, for example, an air-cooled cooling device that cools the foundry sand by taking in outside air. The sand cooling device 25 reduces the temperature of the foundry sand from about 400 degrees to a temperature suitable for the next process (for example, 80 degrees or less).

[0027] The sand polishing device 26 is a device that collides granular molding sand with each other to remove binder residue such as water glass from the molding sand and to make the molding sand spherical. The sand polishing device 26 transports molding sand from the sand cooling device 25 by suction or pressure. The sand polishing device 26 also rotates a rotating drum (not shown). The molding sand is stirred up by centrifugal force generated by the rotation of the rotating drum, and the molding sand collides with each other to be polished, and the residue is removed. The molding sand obtained by the sand polishing device 26, i.e., the molding sand after the residue attached to the surface has been removed, is hereinafter also referred to as "reclaimed sand". The sand polishing device 26 measures the polishing current value and outputs the measured value to the data logger 2.

[0028] The dust collector 27 collects light debris such as water glass peeled off from the casting sand by the sand polishing device 26. The dust collector 27 measures the air volume in the dust collector 27 and outputs the measured value to the data logger 2.

[0029] The sand classifier 28 separates the recycled sand from residues such as water glass that have been peeled off from the recycled sand. The sand classifier 28 also adds new sand, i.e., unused molding sand, to the recycled sand from which the residues have been separated. The mixture of the recycled sand and new sand is used as molding sand in the casting system S11.

[0030] The sand classifier 28 includes a blower 281 that blows air to the recycled sand. Residue is collected in a dust collector 29 by the air blown by the blower 281. The sand classifier 28 measures the air volume in the sand classifier 28 and outputs the measured value to the data logger 2.

[0031] A preset amount of recycled sand (hereinafter also referred to as the “recycled sand input amount”) is input to the sand classifier 28. Here, the recycled sand input amount refers to, for example, the weight or volume of recycled sand input to the sand classifier 28 per unit time (in the case of continuous processing) or per processing (in the case of batch processing). The sand classifier 28 provides the set recycled sand input amount to the data logger 2.

[0032] A preset amount (hereinafter also referred to as the “amount of new sand charged”) of new sand is charged into the sand classifier 28. Here, the amount of new sand charged refers to, for example, the weight or volume of new sand charged into the sand classifier 28 per unit time (in the case of continuous processing) or per processing (in the case of batch processing). The sand classifier 28 provides the set amount of new sand charged to the data logger 2.

[0033] The reclamation condition data collected by the data logger 2 from the sand reclamation system S12 includes some or all of the following data: Air volume in the fine powder removal device 22 Air volume in dust collector 23 Amount of casting sand charged into the fluidized bed roasting device 24 Heating temperature of fluidized bed roaster 24 · Polishing current value of sand polishing device 26 Dust collector 27 air volume Amount of recycled sand fed to the sand classifier 28 - New sand input to sand classifier 28 In other words, the regeneration condition data collected by the data logger 2 includes data indicating at least any of the air volume in the fine powder removal device 22 and the dust collector 23, the amount of foundry sand fed into the fluidized bed roasting device 24 and the heating temperature, the polishing current value of the sand polishing device 26 and the air volume for removing residues from the foundry sand, and the amount of regenerated sand and the amount of new sand fed when mixing the regenerated sand regenerated by the sand reclamation system S12 with new sand.

[0034] (Configuration of the playback condition specifying device) The configuration of the reproduction condition identifying device 1 will be described with reference to Fig. 3. Fig. 3 is a block diagram showing the configuration of the reproduction condition identifying device 1.

[0035] The playback condition specifying device 1 is realized using a general-purpose computer, and includes a processor 11, a primary memory 12, a secondary memory 13, an input / output interface 14, a communication interface 15, and a bus 16. The processor 11, the primary memory 12, the secondary memory 13, the input / output interface 14, and the communication interface 15 are connected to each other via the bus 16.

[0036] The secondary memory 13 stores the regeneration condition specification program P1, the binder data, and the regeneration condition data. The processor 11 expands the regeneration condition specification program P1, the binder data, and the regeneration condition data stored in the secondary memory 13 onto the primary memory 12. The processor 11 then executes each step included in the regeneration condition specification method M1 according to the instructions included in the regeneration condition specification program P1 expanded onto the primary memory 12. The binder data and the regeneration condition data expanded onto the primary memory 12 are used when the processor 11 executes the relational equation specification step M11 (described later) of the regeneration condition specification method M1. The regeneration condition specification program P1 being stored in the secondary memory 13 refers to the source code, or an executable file obtained by compiling the source code, being stored in the secondary memory 13.

[0037] Examples of devices that can be used as the processor 11 include a central processing unit (CPU), a graphic processing unit (GPU), a digital signal processor (DSP), a micro processing unit (MPU), a floating point number processing unit (FPU), a physics processing unit (PPU), a microcontroller, or a combination of these. The processor 11 is also called a "computing device."

[0038] Moreover, a device that can be used as the primary memory 12 can be, for example, a semiconductor RAM (Random Access Memory). The primary memory 12 is sometimes called a "main memory device". Moreover, a device that can be used as the secondary memory 13 can be, for example, a flash memory, a hard disk drive (HDD), a solid state drive (SSD), an optical disk drive (ODD), a floppy (registered trademark) disk drive (FDD), or a combination thereof. The secondary memory 13 is sometimes called an "auxiliary memory device". The secondary memory 13 may be built in the playback condition identification device 1, or may be built in another computer (for example, a computer constituting a cloud server) connected to the playback condition identification device 1 via the input / output interface 14 or the communication interface 15. In this embodiment, the storage in the playback condition identification device 1 is realized by two memories (the primary memory 12 and the secondary memory 13), but is not limited to this. In other words, the storage in the playback condition identification device 1 may be realized by one memory. In this case, for example, one storage area of ​​the memory may be used as the primary memory 12 and another storage area of ​​the memory may be used as the secondary memory 13.

[0039] An input device and / or an output device are connected to the input / output interface 14. Examples of the input / output interface 14 include interfaces such as Universal Serial Bus (USB), Advanced Technology Attachment (ATA), Small Computer System Interface (SCSI), and Peripheral Component Interconnect (PCI). An example of an input device connected to the input / output interface 14 is a data logger 2. Data acquired in the playback condition identification method M1 is input to the playback condition identification device 1 via the data logger 2 and stored in the primary memory 12. Examples of input devices connected to the input / output interface 14 include a keyboard, a mouse, a touchpad, a microphone, or a combination of these. Examples of output devices connected to the input / output interface 14 include a display, a projector, a printer, a speaker, a headphone, or a combination of these. Information provided to a user in the playback condition identification method M1 is output from the playback condition identification device 1 via these output devices. The playback condition identification device 1 may each have a built-in keyboard that functions as an input device and a built-in display that functions as an output device, like a laptop computer. Alternatively, the reproduction condition identifying device 1 may have a built-in touch panel that functions as both an input device and an output device, like a tablet computer.

[0040] The communication interface 15 is connected to other computers by wire or wirelessly via a network. Examples of the communication interface 15 include interfaces such as Ethernet (registered trademark) and Wi-Fi (registered trademark). Available networks include a personal area network (PAN), a local area network (LAN), a campus area network (CAN), a metropolitan area network (MAN), a wide area network (WAN), a global area network (GAN), or an internetwork including these networks. The internetwork may be an intranet, an extranet, or the Internet. Data (e.g., binder data, regeneration condition data) acquired by the regeneration condition identification device 1 from other computers (e.g., data logger 2) in the regeneration condition identification method M1, and data provided by the regeneration condition identification device 1 to other computers in the regeneration condition identification method M1 are transmitted and received via these networks.

[0041] In this embodiment, a configuration is adopted in which the reproduction condition identification method M1 is executed using a single processor (processor 11), but the present invention is not limited to this. That is, a configuration may be adopted in which the reproduction condition identification method M1 is executed using a plurality of processors. In this case, the plurality of processors that execute the reproduction condition identification method M1 in cooperation with each other may be provided in a single computer and configured to be able to communicate with each other via a bus, or may be provided in a distributed manner in a plurality of computers and configured to be able to communicate with each other via a network. As an example, a mode in which a processor built in a computer that constitutes a cloud server and a processor built in a computer owned by a user of the cloud server execute the reproduction condition identification method M1 in cooperation with each other may be considered.

[0042] In addition, in this embodiment, the binder data and the regeneration condition data are stored in a memory (secondary memory 13) built in the same computer as the processor (processor 11) that executes the regeneration condition identification method M1, but the present invention is not limited to this. That is, a configuration may be adopted in which the binder data and the regeneration condition data are stored in a memory built in a computer different from the processor that executes the regeneration condition identification method M1. In this case, the computer having the memory that stores the binder data and the regeneration condition data is configured to be able to communicate with the computer having the processor that executes the regeneration condition identification method M1 via a network. As an example, a mode in which the binder data and the regeneration condition data are stored in a memory built in a computer constituting a cloud server, and a processor built in a computer owned by a user of the cloud server executes the regeneration condition identification method M1, etc., can be considered.

[0043] In addition, in this embodiment, a configuration is adopted in which the binder data and the regeneration condition data are stored in a single memory (secondary memory 13), but the present invention is not limited to this. That is, a configuration in which the binder data and the regeneration condition data are distributed and stored in a plurality of memories may be adopted. In this case, the plurality of memories that store the binder data and the regeneration condition data may be provided in a single computer (which may or may not be a computer with a processor that executes the regeneration condition identification method M1 built in), or may be distributed and provided in a plurality of computers (which may or may not include a computer with a processor that executes the regeneration condition identification method M1). As an example, a configuration in which the binder data and the regeneration condition data are distributed and stored in memories built in each of a plurality of computers that constitute a cloud server can be considered.

[0044] (Method of measuring binder data) An example of a measurement method for measuring binder data of foundry sand, which is a mixture of recycled sand and new sand generated by the sand reclamation system S12, will be described.

[0045] The measurement of the binder data may be performed by a user such as an administrator of the sand reclamation system S12, or a device for measuring the binder data may execute the measurement process of the binder data. When the measurement is performed by a user, the user collects a portion of the recycled molding sand and pours a predetermined amount of the collected molding sand into a measuring cup with an electrode filled with temperature-controlled pure water. The user stirs the poured molding sand in the pure water for a predetermined time to dissolve the binder, and measures the pH. Alternatively, the user applies a voltage to the electrode to measure the conductivity. The user inputs the binder data representing the measured pH or conductivity into the data logger 2.

[0046] When the measuring device measures the binder data, the measuring device (not shown) collects a portion of the recycled foundry sand and places a predetermined amount of the collected foundry sand in a measuring cup equipped with electrodes filled with temperature-controlled pure water. The measuring device stirs the placed foundry sand in the pure water for a predetermined time to dissolve the binder and measures the pH. Alternatively, the measuring device applies a voltage to the electrodes to measure the conductivity. The measuring device outputs binder data indicating the measured pH or conductivity to the data logger 2.

[0047] Another method for measuring binder data will be described. A portion of the recycled molding sand is collected, and a specified amount of the collected molding sand is placed in a measuring cup filled with temperature-controlled pure water. The measuring device stirs the placed molding sand in the pure water for a specified period of time to dissolve the binder. Then, a test piece is molded from this molding sand, and the strength of the test piece is measured. The strength of this test piece is Binder It becomes data. Binder The data may be input to the data logger 2 by a user as described above, or may be output to the data logger 2 by a measurement device.

[0048] (Flow of how to identify playback conditions) The flow of the reproduction condition specification method M1 will be described with reference to Fig. 4. Fig. 4 is a flowchart showing the flow of the reproduction condition specification method M1.

[0049] The regeneration condition specifying method M1 includes a relational equation specifying step M11, a relational equation output step M12, a determination step M13, a regeneration condition specifying step M14, and a control step M15.

[0050] The relational equation specifying step M11 is a step in which the processor 11 refers to a set of reclamation condition data x1, x2, ..., xn indicating the reclamation conditions in the sand reclamation system S12 and binder data y indicating the amount of binder eluted from the foundry sand reclaimed by the sand reclamation system S12, and specifies a nonlinear relational equation expressing the relationship between the reclamation condition data and the binder data using a nonlinear regression algorithm. In this embodiment, the processor 11 specifies the relational equation y=f(x1, x2, ..., xn) using a genetic algorithm. A specific example of the relational equation specifying step M11 will be described later with reference to different drawings.

[0051] The relational equation output step M12 is a step in which the processor 11 outputs the relational equation y=f(x1, x2, ..., xn) identified in the relational equation identification step M11. In this embodiment, the processor 11 outputs (displays) the relational equation y=f(x1, x2, ..., xn) on a display. At this time, the processor 11 may display a graph representing the relational equation on the display.

[0052] The user visually checks the relational expression y=f(x1, x2, ..., xn) output on the display and judges whether the relational expression y=f(x1, x2, ..., xn) identified by the playback condition identifying device 1 is an appropriate relational expression. After completing this judgment, the user performs a user operation to input the judgment result to the playback condition identifying device 1.

[0053] In the determination step M13, the processor 11 determines whether the relational equation y=f(x1, x2, ..., xn) identified in the relational equation identification step M11 is an appropriate relational equation in response to the user operation described above. If it is determined in the determination step M13 that the relational equation is an appropriate relational equation, the processor 11 executes the regeneration condition identification step M14 described below. On the other hand, if it is determined in the determination step M13 that the relational equation is not an appropriate relational equation, the processor 11 executes the above-mentioned processes from the relational equation identification step M11 onwards again. When executing the processes from the relational equation identification step M11 onwards again, the processor 11 may change the set of binder data and regeneration condition data used in the relational equation identification step M11, or change parameters used in the genetic algorithm.

[0054] The regeneration condition specifying step M14 is a step for specifying what values ​​the regeneration condition data x1, x2, ..., xn should be set to in order to make the binder data y coincide with the target value, using the relational expression y=f(x1, x2, ..., xn) specified in the relational expression specifying step M11. That is, in this embodiment, the processor 11 specifies the regeneration condition data by calculation using the binder data and the relational expression.

[0055] In this embodiment, the processor 11 obtains an equation with the regeneration condition data x1, x2, ..., xn as unknowns by substituting the target value of the binder data y into the relational expression y=f(x1, x2, ..., xn). The processor 11 then obtains the set values ​​of the regeneration condition data x1, x2, ..., xn by solving this equation. Note that if the number of unknown regeneration condition data x1, x2, ..., xn is two or more, the solution of the above equation cannot be uniquely determined. In this case, the processor 31 sets at least one of the solutions of the above equation as the set value of the regeneration condition data x1, x2, ..., xn.

[0056] In the control step M15, the processor 11 controls the sand reclamation system S12 so that the value of the reclamation condition data becomes the set value specified in the reclamation condition specification step M14. That is, in this embodiment, the processor 31 controls the sand reclamation system S12 using the reclamation condition data specified in the reclamation condition specification step M14.

[0057] In this embodiment, the processor 11 controls the fine powder removal device 22, the sand polishing device 26, and the sand classifier 28 so that the air volume becomes the set value specified in the regeneration condition specification step M14. The processor 31 also controls the fluidized bed baking device 24 so that the amount of foundry sand charged and the heating temperature become the set values ​​specified in the regeneration condition specification step M14. The processor 31 also controls the sand polishing device 26 so that the polishing current value becomes the set value specified in the regeneration condition specification step M14. The processor 31 also controls the sand classifier 28 so that the amount of reclaimed sand charged and the amount of new sand charged become the set values ​​specified in the regeneration condition specification step M14.

[0058] In this embodiment, a genetic algorithm is used as an algorithm for specifying a nonlinear relational expression that expresses the relationship between the binder data y and the regeneration condition data x1, x2, ..., xn. However, the present invention is not limited to this. In other words, a nonlinear regression algorithm other than the genetic algorithm, such as logistic regression, may be used as an algorithm for specifying a nonlinear relational expression that expresses the relationship between the binder data y and the regeneration condition data x1, x2, ..., xn.

[0059] (Specific example of relational expression identification step) A specific example of the relational equation specifying step M11 included in the reproduction condition specifying method M1 will be described with reference to Fig. 5 and Fig. 6. Fig. 5 is a diagram illustrating a genetic algorithm GA. Fig. 6 is a flowchart illustrating a flow of the relational equation specifying step M11 executed by the processor 11. In the example of Fig. 5, the genetic algorithm GA includes a first generation G1 to a fourth generation G4.

[0060] In the relational equation specifying step M11 according to this specific example, the relational equation y=f(x1, x2, ..., xn) is specified using a genetic algorithm. Here, the genetic algorithm refers to an algorithm in which a plurality of individuals i, each of which represents a solution candidate by genes, are prepared, and an individual i with high fitness Di is preferentially selected, and a solution is searched for by repeating operations such as crossover and mutation. In this embodiment, the individual i is a nonlinear relational equation expressed in a tree structure, and the operators and arguments included in the relational equation are represented by the nodes of the tree. The fitness Di is given by the fitness function.

[0061] The processor 11 executes the relational equation specification step M11 using a predetermined module (hereinafter referred to as "A module"). The A module is a module that executes a genetic algorithm. In the A module, the processor 31 starts by creating a population of simple random equations that represent the relationships between known independent variables and their dependent variable targets to predict new data. The processor 31 then evolves the population to generate the next generation of the population by selecting the fittest individuals from the population to be subjected to genetic manipulation. Through the above operations, the relational equation that best represents the above relationships is specified.

[0062] In this specific example, a module that executes genetic programming is used as module A. Genetic programming is an extension of genetic algorithms, and uses a tree structure as an expression of a gene type. Note that the flow of the relational expression specification step M11 shown in FIG. 6 is an example, and the method of specifying a relational expression using a genetic algorithm GA is not limited to the method shown in FIG. 6. Various other methods can be adopted as the method of specifying a relational expression using a genetic algorithm GA.

[0063] In step M121, the processor 31 acquires the binder data and the regeneration condition data. In this operation example, the processor 31 acquires the binder data and the regeneration condition data by reading the binder data and the regeneration condition data stored in the secondary memory 33.

[0064] In step M122, the processor 11 acquires parameters (hereinafter referred to as "individual parameters") used in the genetic algorithm GA. The individual parameters include, for example, the number of individuals to be generated N, the tournament size Nt, the crossover probability Pc, the mutation probability Pms, the number of evolutionary generations Ng, the operator Oj used in the syntax tree, the maximum depth d of the syntax tree, and the event occurrence probabilities Pk1 to Pk5. The values ​​of each individual parameter are input to the reproduction condition specifying device 1 by, for example, the user.

[0065] The number of generated individuals N represents the number of individuals i to be included in the set. The tournament size Nt is the number of individuals i to be randomly selected from the current generation set. The mutation probability Pms is the probability that a gene will mutate. Operators Oi used in the syntax tree are, for example, Max, Min, sqrt (root), log (natural logarithm), +, -, ×, ÷, sin (radians), cos (radians), tan (radians), abs, neg, and inv. Max is an operator that selects the maximum value. Min is an operator that selects the minimum value. neg is an operator that makes the sign negative. inv is an operator that sets arguments close to zero to 0.

[0066] Event occurrence probabilities Pk1 to Pk5 are probabilities that operations m1 to m5 are selected as operations for evolving the next-generation set. Processor 11 evolves the next-generation set by one of operations m1 to m5. The sum of event occurrence probabilities Pk1 to Pk5 is 1. As an example, the values ​​of event occurrence probabilities Pk1, Pk2, Pk3, Pk4, and Pk5 are "0.1", "0.2", "0.3", "0.4", and "0.1", respectively. Operations m1 to m5 will be described later with reference to different drawings.

[0067] In step M123, processor 11 randomly generates N individuals i based on specified individual parameters (the operator Oi to be used in the syntax tree, the maximum depth d of the syntax tree, etc.) and generates a set of N individuals i that will become the initial current generation.

[0068] In step M124, the processor 11 calculates the fitness Di of each individual i included in the set of the current generation. The fitness Di is given by a fitness function.

[0069] In step M125, processor 11 randomly selects individuals i of the tournament size Nt from the set of the current generation, selects the individual i with the highest fitness Di among them, and adds it to the set of the next generation. The individual i selected in step M125, i.e., the individual i added to the set of the next generation, is also called the "winner tree."

[0070] Processor 11 repeats the process of step M125 until the number of individuals in the next generation becomes N, the same as the current generation, that is, while the number of individuals in the next generation has not reached N (step M126; NO). When the number of individuals in the next generation reaches N (step M126; YES), processor 11 executes the process of step M127.

[0071] In step M127, the processor 11 executes a process of evolving the next generation set. Details of step M127 will be described later with reference to another drawing.

[0072] In step M130, the processor 11 overwrites the next-generation set to the current-generation set. In step M131, the processor 11 determines whether the number of evolutionary generations Ng has been reached. If the number of evolutionary generations Ng has not been reached (step M131; NO), the processor 11 returns to the processing of step M124. On the other hand, if the number of evolutionary generations Ng has been reached (step M131; YES), the processor 11 proceeds to the processing of step M132.

[0073] In step M132, the processor 11 identifies the individual i having the highest fitness Di from among the individuals i included in the set of the current generation. Through the above process, the processor 11 identifies a nonlinear relational expression that expresses the relationship between the binder data and the regeneration condition data.

[0074] FIG. 7 is a flow chart illustrating the flow of step M127 executed by the processor 11. In step M201, the processor 11 selects one of the operations m1 to m5 based on the event occurrence probabilities Pk1 to Pk5 set by the user. If the operation m1 is selected (step M201; "operation m1"), the processor 11 proceeds to the process of step M202. If the operation m2 is selected (step M201; "operation m2"), the processor 11 proceeds to the process of step M211. If the operation m3 is selected (step M201; "operation m3"), the processor 11 proceeds to the process of step M221. If the operation m4 is selected (step M201; "operation m4"), the processor 11 proceeds to the process of step M231. If the operation m5 is selected (step M201; "operation m5"), the processor 11 ends the process.

[0075] Operation m1 is a crossover. Crossover is a method of mixing genetic material between individuals. In the case of crossover, in step M202, processor 11 randomly selects subtrees included in each winning tree for the winning trees included in the next generation set.

[0076] In step M203, the processor 11 generates a next-generation set for donors. The processing content of step M203 is the same as the processing content of steps M123 to M125 in FIG. 6. That is, the processor 11 first randomly generates N individuals i based on individual parameters designated by the user, and generates a set of N individuals i (hereinafter referred to as a "donor set"). Next, the processor 11 calculates the fitness Di of each of the individuals i included in the donor set. Next, the processor 11 randomly extracts individuals i in a number equal to the tournament size Nt from the donor set, selects the individual i with the highest fitness Di among them, and adds it to the donor set of the next generation. The individual i selected by this processing is also referred to as a "donor tree". The processor 11 repeats the donor tree selection processing until the number of donor trees of the next generation reaches N.

[0077] In step M204, processor 11 randomly selects a subtree included in the donor tree (hereinafter referred to as a "donor subtree").

[0078] In step M205, processor 11 replaces the subtrees in the winning tree. In this embodiment, processor 11 removes the subtree selected in step M202 from the winning tree, and transplants the donor subtree selected in step M203 into the location where the subtree was. That is, processor 11 replaces the subtree included in the winning tree with the donor subtree. The winning tree with the replaced subtree becomes the next generation descendant (individual).

[0079] Operation m2 is an operation for mutating a subtree. By mutating a subtree, extinct functions and operators can be reintroduced into the population to maintain diversity. In this case, in step M211, processor 11 randomly selects a subtree to be included in the winning tree.

[0080] In step M212, processor 11 randomly generates a subtree. In step M213, processor 11 replaces the subtrees in the winning tree. In this embodiment, processor 11 removes the subtree selected in step M202 from the winning tree, and transplants the subtree generated in step M212 into the location of the removed subtree. That is, processor 11 replaces the subtree included in the winning tree with the subtree generated in step M212. The winning tree with the replaced subtree becomes the next generation descendant (individual).

[0081] Operation m3 is hoist mutation. Hoist mutation is a mutation operation that fights tree bloat. In step M221, the processor randomly selects a subtree to be included in the winning tree. In step M222, processor 11 randomly selects a subtree to be included in the subtree selected in step M221.

[0082] In step M223, processor 11 rolls up the subtree selected in step M222 to the position of the original subtree (the subtree selected in step M221). The winning tree into which this subtree has been rolled up becomes the next generation's descendant (individual).

[0083] Operation m4 is point mutation. Point mutation is an operation for reintroducing extinct relations and operators into a population in order to maintain diversity. In step M231, processor 11 randomly selects a node of the winning tree. In step M232, processor 11 replaces the node selected in step M231 with another node. As a result, the relational expression represented by the winning tree is replaced with another relational expression that requires the same number of arguments as the original node. The winning tree obtained by the replacement becomes the descendant (individual) of the next generation.

[0084] Operation m5 is regeneration, where the winning tree is replicated and included in the next generation without modification.

[0085] 8 to 11 are diagrams illustrating the contents of operations performed on the winning tree. Fig. 8 is a diagram illustrating the contents of operation m1 (crossover). In the example of Fig. 8, the subtree tr111 of the winning tree tr11 is replaced with the subtree tr121 of the donor tree tr12 to become the winning tree tr13. The winning tree tr13 becomes the descendant (individual) of the next generation.

[0086] Fig. 9 is a diagram illustrating the contents of operation m2 (subtree mutation). In the example of Fig. 9, subtree tr111 of winner tree tr11 is replaced with subtree tr22 to become winner tree tr23. Winner tree tr23 becomes a descendant (individual) of the next generation.

[0087] Figure 10 is a diagram illustrating the contents of operation m3 (hoist mutation). In the example of Figure 10, subtree tr1121 of winner tree tr11 is rolled up to the position of subtree tr112 to become winner tree tr31. Winner tree tr31 becomes a descendant (individual) of the next generation.

[0088] Fig. 11 is a diagram illustrating the contents of operation m4 (point mutation). In the example of Fig. 11, node n21 and node n34 included in the winning tree tr11 are replaced with node n421 and node n434, resulting in a winning tree tr41. The winning tree tr41 becomes a descendant (individual) of the next generation.

[0089] As described above, according to this embodiment, the reclamation condition specifying device 1 specifies a nonlinear relational expression that expresses the relationship between the binder data and the reclamation condition data by using a nonlinear regression algorithm, and specifies the reclamation condition data by substituting the target value of the binder data into the relational expression and performing a calculation. This makes it possible to specify the reclamation condition data for stabilizing the sand properties of the foundry sand to be reclaimed by the sand reclamation system S12.

[0090] [Second embodiment] (Sand management system configuration) The configuration of a sand management system S2 according to a second embodiment of the present invention will be described with reference to Fig. 12. Fig. 12 is a block diagram showing the configuration of the sand management system S2. For convenience of explanation, the same reference numerals are given to members having the same functions as those described in the above embodiment, and the explanations thereof will not be repeated.

[0091] The sand management system S2 includes a casting system S11, a sand reclamation system S12, a reclamation condition identification device 1B, a data logger 2, and a machine learning device 5. Of these, the casting system S11, the sand reclamation system S12, and the data logger 2 have the same configuration as in the first embodiment described above.

[0092] The machine learning device 5 is a device for implementing the machine learning method M2. The machine learning method M2 is a method for constructing a learning data set DS (a set of binder data and regeneration condition data) using data provided from the data logger 2, and constructing a trained model LM1 by machine learning (supervised learning) using the learning data set. As the trained model LM1, for example, an algorithm such as a neural network model such as a convolutional neural network or a recurrent neural network, a regression model such as linear regression, or a tree model such as a regression tree can be used. The configuration of the machine learning device 5 and the flow of the machine learning method M2 will be described in detail below with reference to the drawings.

[0093] The trained model LM1 is a trained model that machine-learns the correlation between input data and output data. The input of the trained model LM1 includes the above-mentioned regeneration condition data x1, x2, ..., xn. The output of the trained model LM1 includes the above-mentioned binder data y.

[0094] [Reproduction condition identification device] The reproduction condition identifying device 1B includes a data collecting device 111. The data collecting device 111 temporarily holds input data to the learned model LM1 and output data from the learned model LM1, and inputs the held input data and output data in a nonlinear regression algorithm (for example, a genetic algorithm GA) at the next stage in a synchronized manner. The data collecting device 111 is, for example, a buffer equipped with an information storage means (for example, a secondary memory 13) such as a semiconductor memory or a hard disk, and an input / output means. Note that, although the example of FIG. 12 illustrates a configuration in which the data collecting device 111 is included in the reproduction condition identifying device 1B, the data collecting device 111 may be configured as a device separate from the reproduction condition identifying device 1B. The data collecting device 111 may be configured as a single piece of hardware, or may be configured as hardware and software.

[0095] When the data collection device 111 is configured as a device separate from the reproduction condition identification device 1B, the data collection device 111 includes, for example, a memory (not shown) and a processor (not shown).

[0096] The regeneration condition identification device 1B executes the regeneration condition identification method MB1. The flow of the regeneration condition identification method MB1 will be described later in detail with reference to the drawings. In addition to the regeneration condition identification program P2, the binder data, and the regeneration condition data, the secondary memory 13 of the regeneration condition identification device 1B stores the trained model LM1. The processor 11 expands the trained model LM1 stored in the secondary memory 13 onto the primary memory 12. The trained model LM1 expanded onto the primary memory 12 is used when the processor 11 executes the estimation step MB11 (described later) of the regeneration condition identification method MB1. Note that the trained model LM1 being stored in the secondary memory 13 means that parameters defining the trained model LM1 are stored in the secondary memory 13.

[0097] [Flow of the method for identifying regeneration conditions] The flow of the reproduction condition specification method MB1 will be described with reference to Fig. 13. Fig. 13 is a flowchart showing the flow of the reproduction condition specification method MB1. The reproduction condition specification method MB1 includes an estimation step MB11, a relational equation specification step MB12, a relational equation output step M12, a judgment step M13, a reproduction condition specification step M14, and a control step M15. Among these steps, the processes of the relational equation output step M12, the judgment step M13, the reproduction condition specification step M14, and the control step M15 are the same as the processes of the respective steps described in Fig. 4 in the above-mentioned first embodiment.

[0098] The estimation step MB11 is a step in which the processor 11 estimates output data from input data using the trained model LM1. In the estimation step MB11, the processor 11 acquires regeneration condition data collected by the data logger 2, and inputs the input data including the acquired regeneration condition data x1, x2, ..., xn to the trained model LM1 to acquire output data. The output data output by the trained model LM1 includes the above-mentioned binder data. The input data input to the trained model LM1 and the output data output from the trained model LM1 are temporarily stored in the data collection device 111.

[0099] The relational equation specification step MB12 is a step in which the processor 11 specifies, by a genetic algorithm GA, a nonlinear relational equation that expresses the relationship between the regeneration condition data input to the trained model LM1 and the binder data estimated in the estimation step MB11. The data collection device 111 inputs the temporarily stored input data and output data of the trained model LM1 in time synchronization with the genetic algorithm GA. In this way, the data collection device 111 acts as a kind of buffer that collects the data of the trained model LM1 (in certain information units) and inputs it to the next stage of the genetic algorithm GA.

[0100] FIG. 14 is a diagram illustrating a trained model LM1 and a genetic algorithm GA. The trained model LM1 is a deep neural network composed of multiple layers. The trained model LM1 includes multiple layers including an input layer LX to which input data is input, a hidden layer LY, and an output layer LZ to which output data is output. Each layer has a structure in which multiple nodes are connected by edges.

[0101] Each layer has a function called an activation function, and edges can have weights. The output value of each node is calculated from the output value of the node in the previous layer that is connected to that node. In other words, the output value of each node is calculated from the output value of the node in the previous layer, the weight value of the connecting edge, and the activation function of the layer.

[0102] 14, the input layer LX includes nodes X1, X2, X3, and X4. That is, in this example, the input data input to the trained model LM1 includes output values ​​x1, x2, x3, and x4 of the nodes X1, X2, X3, and X4.

[0103] The hidden layer LY includes nodes Y1, Y2, and Y3. The output values ​​y1, y2, and y3 of the nodes Y1, Y2, and Y3 are calculated from the output values ​​x1, x2, x3, and x4 of the nodes of the input layer LX, which is the layer before the hidden layer LY.

[0104] The output layer LZ includes a node Z1. The output value z1 of the node Z1 is calculated from the output values ​​y1, y2, y3, and y4 of the nodes in the hidden layer LY, which is the layer before the output layer LZ. That is, in this example, the output data of the trained model LM1 is the output value z1 of the node Z1.

[0105] The data collection device 111 temporarily holds input data to the trained model LM1 and output data from the trained model LM1, and inputs the held data to the next stage of the genetic algorithm GA.

[0106] The method of identifying the relational equation in the relational equation identifying step MB12 is the same as the method of identifying the relational equation in the relational equation identifying step M11 shown in the first embodiment described above. After identifying the relational equation, the processor 11 executes the relational equation output step M12 and the determination step M13. The processor 11 then proceeds to the regeneration condition identifying step M14, where the processor 11 identifies regeneration condition data by substituting the binder data into the identified relational equation and performing a calculation. In the control step M15, the processor 11 controls the sand reclamation system S12 using the identified regeneration condition data.

[0107] According to this embodiment, the reproduction condition identification device 1 estimates the output of the trained model LM1 using a nonlinear relational expression identified by the genetic algorithm GA, and uses the estimated output of the trained model LM1 to: BinderThe reproduction condition for maintaining the data at a target value is specified. That is, the reproduction condition specifying device 1 can estimate the reproduction condition from the input data without using the learned model LM1.

[0108] In this embodiment, the data collection device 111 buffers the input data and output data of the trained model LM1 and inputs them to the genetic algorithm GA in a time-synchronized manner. Since the output from the trained model LM1 and the input to the genetic algorithm GA can be synchronized in time, it is possible to predict the time until the next-stage genetic algorithm GA outputs the final solution.

[0109] In this embodiment, the data collection device 111 acts as a kind of buffer that collects data of the learned model LM1 (in certain information units) and inputs it to the next stage of the genetic algorithm GA. This makes it possible to reduce wasted memory space and wasted calculation processing in the next stage of the genetic algorithm GA, and to improve the efficiency of the generation calculation processing of the genetic algorithm.

[0110] [Configuration of machine learning device] The configuration of the machine learning device 5 will be described with reference to Fig. 15. Fig. 15 is a block diagram showing the configuration of the machine learning device 5.

[0111] The machine learning device 5 is realized using a general-purpose computer, and includes a processor 51, a primary memory 52, a secondary memory 53, an input / output interface 54, a communication interface 55, and a bus 56. The processor 51, the primary memory 52, the secondary memory 53, the input / output interface 54, and the communication interface 55 are connected to each other via the bus 56.

[0112] The secondary memory 53 stores a machine learning program P5 and a training data set DS. The training data set DS is a collection of teacher data DS1, DS2, etc. The teacher data DS1, DS2, etc. are a set of binder data and regeneration condition data. The processor 51 expands the machine learning program P5 stored in the secondary memory 53 on the primary memory 52. ​​The processor 51 executes each step included in the machine learning method M2 according to the instructions included in the machine learning program P5 expanded on the primary memory 52. ​​The training data set DS stored in the secondary memory 53 is constructed in a training data set construction step M21 (described later) of the machine learning method M2, and is used in a trained model construction step M22 (described later) of the machine learning method M2. In addition, the trained model LM1 constructed in the trained model construction step M22 of the machine learning method M2 is also stored in the secondary memory 53. Note that the machine learning program P5 being stored in the secondary memory 53 refers to a source code or an executable file obtained by compiling the source code being stored in the secondary memory 53. In addition, the trained model LM1 being stored in the secondary memory 53 means that parameters defining the trained model LM1 are stored in the secondary memory 53.

[0113] Examples of devices that can be used as the processor 51 include a central processing unit (CPU), a graphic processing unit (GPU), a digital signal processor (DSP), a micro processing unit (MPU), a floating point number processing unit (FPU), a physics processing unit (PPU), a microcontroller, or a combination of these. The processor 51 is also called a "computing device."

[0114] Moreover, examples of devices that can be used as the primary memory 52 include, for example, semiconductor RAM (Random Access Memory). The primary memory 52 is sometimes called a "main storage device". Examples of devices that can be used as the secondary memory 53 include, for example, a flash memory, a hard disk drive (HDD), a solid state drive (SSD), an optical disk drive (ODD), a floppy disk drive (FDD), or a combination of these. The secondary memory 53 is sometimes called an "auxiliary storage device". The secondary memory 53 may be built in the machine learning device 5, or may be built in another computer (for example, a computer constituting a cloud server) connected to the machine learning device 5 via the input / output interface 54 or the communication interface 55. In this embodiment, the storage in the machine learning device 5 is realized by two memories (the primary memory 52 and the secondary memory 53), but is not limited to this. That is, the storage in the machine learning device 5 may be realized by one memory. In this case, for example, a certain storage area of ​​the memory may be used as the primary memory 52, and another storage area of ​​the memory may be used as the secondary memory 53.

[0115] An input device and / or an output device are connected to the input / output interface 54. Examples of the input / output interface 54 include interfaces such as Universal Serial Bus (USB), Advanced Technology Attachment (ATA), Small Computer System Interface (SCSI), and Peripheral Component Interconnect (PCI). An example of an input device connected to the input / output interface 54 is a data logger 2. Data acquired in the machine learning method M2 is input to the machine learning device 5 via the data logger 2 and stored in the primary memory 52. ​​Examples of input devices connected to the input / output interface 54 include a keyboard, a mouse, a touchpad, a microphone, or a combination thereof. Data acquired from a user in the machine learning method M2 is input to the machine learning device 5 via these input devices and stored in the primary memory 52. ​​Examples of output devices connected to the input / output interface 54 include a display, a projector, a printer, a speaker, a headphone, or a combination thereof. Information provided to a user in the machine learning method M2 is output from the machine learning device 5 via these output devices. The machine learning device 5 may incorporate a keyboard that functions as an input device and a display that functions as an output device, like a laptop computer, or may incorporate a touch panel that functions as both an input device and an output device, like a tablet computer.

[0116] The communication interface 55 is connected to other computers via a network in a wired or wireless manner. Examples of the communication interface 55 include interfaces such as Ethernet (registered trademark) and Wi-Fi (registered trademark). Available networks include a personal area network (PAN), a local area network (LAN), a campus area network (CAN), a metropolitan area network (MAN), a wide area network (WAN), a global area network (GAN), or an internetwork including these networks. The internetwork may be an intranet, an extranet, or the Internet. Data (e.g., the trained model LM1) provided by the machine learning device 5 to other computers (e.g., the reproduction condition identification device 1) is transmitted and received via these networks.

[0117] In this embodiment, a configuration is adopted in which the machine learning method M2 is executed using a single processor (processor 51), but the present invention is not limited to this. That is, a configuration may be adopted in which the machine learning method M2 is executed using multiple processors. In this case, the multiple processors that execute the machine learning method M2 in cooperation with each other may be provided in a single computer and configured to be able to communicate with each other via a bus, or may be provided in a distributed manner in multiple computers and configured to be able to communicate with each other via a network. As an example, a processor built in a computer that constitutes a cloud server and a processor built in a computer owned by a user of the cloud server may cooperate to execute the machine learning method M2.

[0118] In addition, in this embodiment, a configuration is adopted in which the training dataset DS is stored in a memory (secondary memory 53) built into the same computer as the processor (processor 51) that executes the machine learning method M2, but the present invention is not limited to this. That is, a configuration may be adopted in which the training dataset DS is stored in a memory built into a computer different from the processor that executes the machine learning method M2. In this case, the computer with the built-in memory for storing the training dataset DS is configured to be able to communicate with the computer with the built-in processor that executes the machine learning method M2 via a network. As an example, a mode in which the training dataset DS is stored in a memory built into a computer constituting a cloud server, and a processor built into a computer owned by a user of the cloud server executes the machine learning method M2, etc., can be considered.

[0119] In addition, in this embodiment, a configuration in which the training dataset DS is stored in a single memory (secondary memory 53) is adopted, but the present invention is not limited to this. That is, a configuration in which the training dataset DS is distributed and stored in multiple memories may be adopted. In this case, the multiple memories that store the training dataset DS may be provided in a single computer (which may or may not be a computer with a built-in processor that executes the machine learning method M2), or may be distributed and provided in multiple computers (which may or may not include a computer with a built-in processor that executes the machine learning method M2). As an example, a configuration in which the training dataset DS is distributed and stored in memories built into each of the multiple computers that constitute the cloud server is conceivable.

[0120] In addition, in the present embodiment, a configuration is adopted in which the playback condition identification method M1 and the machine learning method M2 are executed using different processors (processor 11 and processor 51), but the present invention is not limited to this. That is, the playback condition identification method M1 and the machine learning method M2 may be executed using the same processor. In this case, by executing the machine learning method M2, the trained model LM1 is stored in a memory built into the same computer as this processor. Then, when executing the playback condition identification method M1, this processor uses the trained model LM1 stored in this memory to identify the relational expression and the playback conditions.

[0121] [Machine learning method flow] The flow of the machine learning method M2 will be described with reference to Fig. 16. Fig. 16 is a flowchart showing the flow of the machine learning method M2.

[0122] The machine learning method M2 includes a learning dataset construction step M21 and a trained model construction step M22.

[0123] The learning dataset construction step M21 is a step in which the processor 51 constructs a learning dataset DS, which is a collection of teacher data DS1, DS2, .... Each teacher data DSi (i = 1, 2, ...) includes binder data and regeneration condition data. The binder data included in the teacher data DSi is the same as the input data input to the trained model LM1. In the learning dataset construction step M21, the processor 51 acquires this data in the same manner as the regeneration condition identification device 1. The teacher data DSi also includes regeneration condition data. In the learning dataset construction step M21, the processor 11 associates the binder data with the regeneration condition data and stores them in the secondary memory 13. The processor 11 repeats the above process to construct the learning dataset DS.

[0124] The trained model construction step M22 is a step in which the processor 51 constructs the trained model LM1. In the trained model construction step M22, the processor 51 constructs the trained model LM1 by supervised learning using the training data set DS. Then, the processor 11 stores the constructed trained model LM1 in the secondary memory 53.

[0125] 〔summary〕 The regeneration condition specifying device according to the first aspect is a device for determining whether or not a substance is dissolved from casting sand regenerated by a sand reclamation system. Binder The method includes one or more processors that execute a regeneration condition determination step of determining regeneration conditions in the sand reclamation system from the amount of sand.

[0126] According to the above configuration, the reclamation condition specifying device specifies reclamation condition data for maintaining the sand properties of the foundry sand reclaimed by the sand reclamation system at target values.

[0127] The regeneration condition specifying device according to the second aspect has the following features in addition to the features of the regeneration condition specifying device according to the first aspect. That is, in the regeneration condition specifying device according to the second aspect, the processor further executes a relational equation specifying step of specifying a nonlinear relational equation expressing the relationship between the regeneration condition data and the binder data by using a nonlinear regression algorithm, with reference to a set of the regeneration condition data and binder data expressing the amount of binder eluted from the casting sand regenerated by the sand reclamation system, and in the regeneration condition specifying step, specifies the regeneration condition data by calculation using the binder data and the relational equation.

[0128] According to the above configuration, the reclamation condition specifying device specifies reclamation condition data for maintaining the sand properties of the foundry sand reclaimed by the sand reclamation system at target values.

[0129] The regeneration condition determination device according to the third aspect has the following features in addition to the features of the regeneration condition determination device according to the first or second aspect. That is, in the regeneration condition determination device according to the third aspect, the processor executes a control step of controlling the sand reclamation system using the regeneration condition data determined in the regeneration condition determination step.

[0130] According to the above configuration, the reclamation condition identification device can stabilize the sand properties of the foundry sand reclaimed by the sand reclamation system by controlling the sand reclamation system using the identified reclamation condition data.

[0131] The regeneration condition specifying device according to the fourth aspect has the following features in addition to the features of the regeneration condition specifying device according to the second aspect. That is, in the regeneration condition specifying device according to the fourth aspect, the binder data includes data indicating at least one of an amount of eluted binder, which indicates the amount of binder eluted from the molding sand at a temperature when the molding sand is kneaded, and an amount of residual binder, which indicates the amount of binder eluted from the molding sand at a temperature higher than the temperature.

[0132] According to the above configuration, the regeneration condition specifying device can stabilize the amount of dissolved binder or the amount of residual binder in the foundry sand regenerated by the sand reclamation system.

[0133] The regeneration condition specifying device according to the fifth aspect has the following features in addition to the features of the regeneration condition specifying device according to the second aspect. That is, in the regeneration condition specifying device according to the fifth aspect, the regeneration condition data includes data indicating at least one of the air volume for separating the foundry sand from fine powder in the sand reclamation system, the amount of the foundry sand fed into an incineration device that heats the foundry sand and the heating temperature, the polishing current value of a sand polishing device that polishes the foundry sand, and the air volume for removing residues from the foundry sand, as well as the amount of the regenerated sand and the amount of the new sand fed when mixing the regenerated sand regenerated by the sand reclamation system with new sand.

[0134] According to the above configuration, the regeneration condition identification device can identify at least one of the following regeneration conditions for maintaining the sand properties of the foundry sand at target values: the air volume for classifying foundry sand and fine powder in the sand reclamation system, the amount of foundry sand fed into the incineration device that heats the foundry sand, and the heating temperature, the polishing current value of the sand polishing device that polishes the foundry sand, the air volume for removing residues from the foundry sand, and the amount of regenerated sand and the amount of new sand regenerated by the sand reclamation system.

[0135] The regeneration condition determining device according to aspect 6 has the following features in addition to the features of the regeneration condition determining device according to aspect 2 or 5. That is, in the regeneration condition determining device according to aspect 6, the processor determines the nonlinear relational expression by a genetic algorithm in the relational expression determining step.

[0136] According to the above configuration, the reclamation condition specifying device can specify reclamation condition data for maintaining the sand properties of the foundry sand reclaimed by the sand reclamation system at target values.

[0137] The regeneration condition identifying device according to aspect 7 has the following features in addition to the features of the regeneration condition identifying device according to aspect 2, 5 or 6. That is, in the regeneration condition identifying device according to aspect 7, the processor further executes an estimation step of estimating the regeneration condition data from the binder data by using a trained model that has been machine-learned to learn the correlation between the binder data and the regeneration condition data, and in the relational expression identifying step, the processor identifies the nonlinear relational expression that expresses the relationship between the binder data input to the trained model and the regeneration condition data estimated in the estimation step.

[0138] According to the above configuration, the regeneration condition identification device can identify regeneration condition data for maintaining the sand properties of the foundry sand regenerated by the sand reclamation system at target values ​​without using a learned model.

[0139] In the method for identifying regeneration conditions according to an eighth aspect, one or more processors are configured to determine whether or not a certain amount of a substance is dissolved from foundry sand regenerated by a sand reclamation system. Binder and a regeneration condition determination step for determining the regeneration conditions in the sand reclamation system from the amount of sand.

[0140] According to the above configuration, it is possible to identify reclamation condition data for stabilizing the sand properties of the foundry sand reclaimed by the sand reclamation system.

[0141] [Additional Note 1] In each of the above-mentioned embodiments, in the regeneration condition specification step M14, the processor 11 obtains an equation in which the regeneration condition data x1, x2, ..., xn are unknowns by substituting the target value of the binder data y into the relational equation y=f(x1, x2, ..., xn). Then, the processor 31 obtains the set values ​​of the regeneration condition data x1, x2, ..., xn by solving this equation. The data to be substituted into the relational equation is not limited to those shown in the above-mentioned embodiments. For example, in addition to the target value of the binder data y, some set values ​​of the regeneration condition data x1, x2, ..., xn may be substituted into the above-mentioned relational equation. In this case, the processor 31 obtains the remaining set values ​​of the regeneration condition data x1, x2, ..., xn by solving the equation obtained by substituting each value into the relational equation.

[0142] [Additional Note 2] In the above-described second embodiment, a trained model trained by machine learning through supervised learning is used as the trained model LM1, but a trained model trained by machine learning through unsupervised learning may be used. For the trained model trained by machine learning through unsupervised learning, the processor 11 also identifies a nonlinear relational expression that expresses the relationship between the input and output of the trained model by a genetic algorithm.

[0143] [Additional Note 3] In the above-described second embodiment, the machine learning device 5 constructs the trained model LM1, but the trained model LM1 may be constructed in advance by another device other than the machine learning device 5. In this case, the reproduction condition identification device 1 performs the above-described reproduction condition identification method MB1 using the trained model constructed in advance by the other device.

[0144] [Additional Note 4] In the above-mentioned second embodiment, a trained model LM1 is used in which regeneration condition data x1, x2, ..., xn are input and binder data y is output, but the trained model according to the present invention is not limited to that shown in the above-mentioned embodiment. The trained model may be, for example, a trained model in which a part of the regeneration condition data and binder data are input and any of the regeneration condition data is output. The trained model may also be, for example, a trained model in which binder data is input and regeneration condition data is output.

[0145] [Additional Note 5] In the second embodiment described above, the processor 11 specifies a nonlinear relational equation representing the relationship between input data and output data of a trained model by a genetic algorithm that sets the nonlinear relational equation as a candidate solution, but the method of specifying the relational equation representing the relationship between input data and output data is not limited to the method described in the second embodiment described above. For example, the processor 11 may specify the first relational equation and the second relational equation by a genetic algorithm that sets as candidate solutions a nonlinear first relational equation representing the relationship between a numerical value representing the input data and each output value of some or all nodes belonging to a hidden layer, and a nonlinear second relational equation representing the relationship between each output value of the node and a numerical value representing the output data. In this case, the processor 11 specifies a nonlinear relational equation representing the relationship between a numerical value representing the input data and a numerical value representing the output data by, for example, solving a simultaneous equation of the specified first relational equation and the specified second relational equation.

[0146] In addition, the algorithm for identifying a nonlinear relational expression that represents the relationship between input data and output data of the trained model is not limited to a genetic algorithm, and may be another nonlinear algorithm. As an example, the processor 11 may identify a nonlinear relational expression that represents the relationship between input data and output data using a Monte Carlo method. In this case, as an example, the processor 11 creates multiple relational expressions by randomly selecting the length of the relational expression and elements of the relational expression (variables, operators, etc.), and selects the relational expression with the smallest error from the multiple relational expressions created, thereby constructing a relational expression that derives output data from input data.

[0147] In the second embodiment described above, the processor 11 identifies a nonlinear relational expression that represents the relationship between the input and output of the trained model LM1 by nonlinear regression, and Binder The reproduction condition was specified by substituting the amount of the reproduction condition into the learning model LM1 and performing a calculation. The method of specifying the reproduction condition is not limited to the method described in the above embodiment. Binder The reproduction conditions represented by the reproduction condition data obtained by inputting data may be specified.

[0148] [Additional Note 6] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Other embodiments obtained by appropriately combining the technical means disclosed in the above-described embodiments are also included in the technical scope of the present invention. [Explanation of symbols]

[0149] 1, 1B Playback condition identification device 2 Data logger 5. Machine learning equipment 11, 51 processor 12, 52 Primary memory 13, 53 Secondary memory 14, 54 Input / Output Interface 15, 55 Communication Interface 16, 56 Bus How to specify M1, MB1 playback conditions M11, MB12 relational equation specification step M12 Equation Output Step M13 Judgment step M14 Playback condition specification step M15 Control Step M2 Machine Learning Method M21 Training Dataset Construction Steps M22 Trained model construction steps MB11 Estimation Step S1, S2 Sand Management System S11 Casting System S12 Sand Reclamation System

Claims

1. A regeneration condition determination step is provided for determining regeneration conditions in the sand reclamation system based on the amount of binder dissolved from the molding sand regenerated by the sand reclamation system, The processor, and further executing a relational equation determination step of determining, by using a nonlinear regression algorithm, a nonlinear relational equation that represents the relationship between the regeneration condition data and the binder data, by referring to a set of regeneration condition data indicating the regeneration conditions and binder data indicating the amount of binder eluted from the foundry sand regenerated by the sand reclamation system; In the regeneration condition specifying step, the regeneration condition data is specified by a calculation using the binder data and the relational expression. A reproduction condition specifying device characterized by:

2. The processor executes a control step of controlling the sand reclamation system using the reclamation condition data identified in the reclamation condition identification step.

2. The reproduction condition specifying device according to claim 1.

3. The binder data includes data indicating at least one of an amount of binder eluted from the molding sand at a temperature when the molding sand is kneaded and an amount of binder eluted from the molding sand at a temperature higher than the temperature.

3. The reproduction condition specifying device according to claim 1 or 2.

4. The regeneration condition data includes data indicating at least any one of the following: the air volume for separating the foundry sand from the fine powder in the sand reclamation system, the amount of the foundry sand fed into an incineration device that heats the foundry sand, and the heating temperature, the polishing current value of a sand polishing device that polishes the foundry sand, and the air volume for removing residues from the foundry sand, as well as the amount of the regenerated sand and the amount of the new sand fed when mixing the regenerated sand regenerated by the sand reclamation system with new sand.

4. The reproduction condition specifying device according to claim 1, wherein the reproduction condition specifying device is a reproduction condition specifying unit.

5. The processor, in the relational expression specifying step, specifies the nonlinear relational expression by a genetic algorithm.

5. The reproduction condition specifying device according to claim 1, wherein the reproduction condition specifying device is a reproduction condition specifying unit.

6. The processor further executes an estimation step of estimating the binder data from the regeneration condition data by using a trained model that has been machine-learned to learn a correlation between the binder data and the regeneration condition data; In the relational expression identifying step, the processor identifies the nonlinear relational expression that represents a relationship between the regeneration condition data input to the trained model and the binder data estimated in the estimation step.

6. The reproduction condition specifying device according to claim 1, wherein the reproduction condition specifying device is a reproduction condition specifying unit.

7. A regeneration condition determination step in which one or more processors determine the regeneration conditions in the sand reclamation system based on the amount of binder dissolved from the foundry sand regenerated by the sand reclamation system; A relational equation determination step in which the processor refers to a set of regeneration condition data indicating the regeneration conditions and binder data indicating the amount of binder eluted from the foundry sand regenerated by the sand reclamation system, and determines a nonlinear relational equation that indicates the relationship between the regeneration condition data and the binder data using a nonlinear regression algorithm; Including, The processor, in the regeneration condition specifying step, specifies the regeneration condition data by a calculation using the binder data and the relational expression. A reproduction condition specifying method comprising the steps of:

Images