Method for producing mixed material using material composition search method

Abstract

To provide a method for producing a mixed material, capable of reducing the time and effort of creating the Ising model for solving the optimization problem of a material composition.SOLUTION: A method for producing a mixed material comprises: calculating a relation between explanatory variables and the physical properties of the mixed material by a machine learning model described using one or more explanatory variables of a function converted into the Ising model and learned for predicting the physical properties of the mixed material; determining the optimum value and allowable variation width of explanatory variables for target physical properties on the basis of a relation between the explanatory variables and the physical properties of the mixed material; outputting the determined optimum value of the explanatory variables as the target value of the explanatory variables in the function and outputting the weighting coefficient of the explanatory variables in the function on the basis of the allowable variation width of the determined explanatory variables; converting a function obtained by substituting the target value of the determined explanatory variables and the weighting coefficient of the explanatory variables into the Ising model; and calculating the optimal solution of a material composition gradually approximating the target value using the Ising model to produce the mixed material using the optimal solution of the material composition gradually approximating to the target value.SELECTED DRAWING: Figure 4

Description

[Technical field]

[0001] The present disclosure relates to a method for producing mixed materials using a material composition exploration method. [Background technology]

[0002] For example, combinatorial optimization problems exist, in which the optimal combination is selected from among various combinations of elements, such as searching for a material composition with optimal physical properties. As the number of elements increases, the number of combinations of combinatorial optimization problems increases explosively, and they may not be solvable within a realistic time frame. For example, when creating a mixed material by combining 100 types of materials in 1% increments, the number of combinations is 5 × 10 58 It becomes.

[0003] An annealing machine using the Ising model has been proposed as an architecture specialized for solving such combinatorial optimization problems. Annealing machines can efficiently solve combinatorial optimization problems converted into the Ising model.

[0004] Conventionally, there is known a technique for optimizing the thermophysical properties of a refrigerant mixture by using a computer architecture specialized for combinatorial optimization problems (see, for example, Non-Patent Document 1). [Prior art documents] [Non-patent literature]

[0005] [Non-Patent Document 1] "Optimization of Thermophysical Properties of Refrigerant Mixtures Using Digital Annealer", Proceedings of the 2019 Thermal Engineering Conference of the Japan Society of Mechanical Engineers, No. 19-303 [2019.10.12-13, Nagoya] Summary of the Invention [Problem to be solved by the invention]

[0006] For example, a user who wants to solve a combinatorial optimization problem of a material composition that asymptotically approaches (approximates) a target physical property value using an annealing machine needs to prepare a function to convert to an Ising model. The function to convert to an Ising model is written to include a set of explanatory variables, and must satisfy the constraint that the set of variables at the minimum value is the optimization result (optimum mixed material composition). In addition, the function to convert to an Ising model includes arbitrary constants such as the target values ​​of the explanatory variables and weighting coefficients, and these constants need to be adjusted.

[0007] As such, when using an annealing machine to solve a combinatorial optimization problem of material compositions that asymptotically (approximately) reach target physical property values, there was a problem in that it was time-consuming to create a function to convert into an Ising model.

[0008] An object of the present disclosure is to provide a method for generating a mixed material using a material composition searching method that can reduce the effort required for creating an Ising model to allow an annealing-based computing device to solve a combinatorial optimization problem of material compositions that asymptotically approach target physical properties. [Means for solving the problem]

[0009] The present disclosure has the following configuration.

[0010] [1] An information processing device that supports the creation of an Ising model for allowing an annealing-based computing device to solve a combinatorial optimization problem of a material composition that asymptotically approaches a target property, comprising: A calculation unit that calculates a relationship between the one or more explanatory variables and the physical properties of the mixed material by a trained machine learning model for predicting the physical properties of the mixed material, the relationship being described using one or more explanatory variables of a function that converts into the Ising model; A determination unit that determines optimal values ​​and allowable change ranges of the one or more explanatory variables for the target property based on a relationship between the one or more explanatory variables and the property of the mixed material; an output unit that outputs the determined optimal values ​​of the one or more explanatory variables as target values ​​of the one or more explanatory variables in the function, and outputs weighting coefficients of the one or more explanatory variables in the function based on the determined allowable change ranges of the one or more explanatory variables; 13. An information processing device comprising:

[0011] [2] The one or more explanatory variables are characteristics of the mixed material that can be described by a weighted average of the proportions of the material composition. The information processing device according to [1],

[0012] [3] The output unit outputs a smaller weighting coefficient for the explanatory variable having a larger allowable change range, and outputs a larger weighting coefficient for the explanatory variable having a smaller allowable change range. The information processing device according to [1] or [2],

[0013] [4] The determination unit determines an allowable change range for each of the one or more explanatory variables based on a threshold value for allowable deviation of the target property. The information processing device according to any one of [1] to [3],

[0014] [5] The machine learning model has learned the relationship between the characteristics of the mixed material, which can be described by a weighted average of the proportions of the material composition, and the physical properties of the mixed material, using experimental data. An information processing device according to any one of [1] to [4],

[0015] [6] a conversion unit that converts the function into which the target values ​​of the one or more explanatory variables and the weighting coefficients of the one or more explanatory variables output by the output unit are substituted into the Ising model; The information processing device according to any one of [1] to [5], further comprising:

[0016] [7] An information processing system having an annealing-based computing device using an Ising model and an information processing device that supports the creation of an Ising model for causing the computing device to solve a combinatorial optimization problem of a material composition that asymptotically approaches a target property, A calculation unit that calculates a relationship between the one or more explanatory variables and the physical properties of the mixed material by a trained machine learning model for predicting the physical properties of the mixed material, the relationship being described using one or more explanatory variables of a function that converts into the Ising model; A determination unit that determines optimal values ​​and allowable change ranges of the one or more explanatory variables for the target property based on a relationship between the one or more explanatory variables and the property of the mixed material; an output unit that outputs the determined optimal values ​​of the one or more explanatory variables as target values ​​of the one or more explanatory variables in the function, and outputs weighting coefficients of the one or more explanatory variables in the function based on the determined allowable change ranges of the one or more explanatory variables; a conversion unit that converts the function into which the target values ​​of the one or more explanatory variables and the weighting coefficients of the one or more explanatory variables output by the output unit are substituted into the Ising model; an optimal solution calculation unit that calculates an optimal solution of a material composition that asymptotically approaches the target value by using the Ising model; a display unit that displays an optimal solution of the material composition that approaches the target value; An information processing system comprising:

[0017] [8] An information processing device that supports the creation of an Ising model for allowing an annealing-based computing device to solve a combinatorial optimization problem of a material composition that asymptotically approaches a target property, A calculation unit that calculates a relationship between the one or more explanatory variables and the physical properties of the mixed material using a trained machine learning model for predicting the physical properties of the mixed material, the relationship being described using one or more explanatory variables of a function that converts into the Ising model; A determination unit that determines optimal values ​​and allowable change ranges of the one or more explanatory variables for the target property based on a relationship between the one or more explanatory variables and the property of the mixed material; an output unit that outputs the determined optimal values ​​of the one or more explanatory variables as target values ​​of the one or more explanatory variables in the function, and outputs weighting coefficients of the one or more explanatory variables in the function based on the determined allowable change ranges of the one or more explanatory variables; A program to function as a

[0018] [9] A material composition search method for an information processing system having an annealing-based computing device using an Ising model and an information processing device that supports the creation of an Ising model for causing the computing device to solve a combinatorial optimization problem of a material composition that asymptotically approaches a target property, comprising: Calculating a relationship between the one or more explanatory variables and the physical properties of the mixed material using a trained machine learning model for predicting the physical properties of the mixed material, the model being described using one or more explanatory variables of a function that converts into the Ising model; determining optimal values ​​and allowable variation ranges of the one or more explanatory variables for the target property based on a relationship between the one or more explanatory variables and the physical property of the mixed material; outputting the determined optimal values ​​of the one or more explanatory variables as target values ​​of the one or more explanatory variables in the function, and outputting weighting coefficients of the one or more explanatory variables in the function based on the determined allowable change ranges of the one or more explanatory variables; converting the function into which the target values ​​of the one or more explanatory variables output by the output unit and the weighting coefficients of the one or more explanatory variables are substituted, into the Ising model; Calculating an optimal solution for a material composition that approaches the target value using the Ising model; displaying an optimal solution for the material composition that asymptotically approaches the target value; A material composition exploration method comprising: Effect of the Invention

[0019] According to the present disclosure, it is possible to provide a method for generating a mixed material using a material composition searching method that can reduce the effort required for creating an Ising model for having an annealing-based computing device solve a combinatorial optimization problem of a material composition that asymptotically approaches target physical properties. [Brief description of the drawings]

[0020] [Figure 1] 1 is a configuration diagram of an example of an information processing system according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a diagram illustrating an example of a hardware configuration of a computer according to the present embodiment. [Diagram 3] 1 is a configuration diagram of an example of an information processing system according to an embodiment of the present invention. [Figure 4] 1 is a flowchart showing an example of a processing procedure of a material composition exploration method using the information processing system according to the present embodiment. [Diagram 5] FIG. 11 is an explanatory diagram of an example of a function for converting into an Ising model. [Figure 6A] FIG. 1 is an explanatory diagram of an example of a problem to be solved as a combinatorial optimization problem. [Figure 6B] FIG. 1 is an explanatory diagram of an example of a problem to be solved as a combinatorial optimization problem. [Figure 6C] FIG. 1 is an explanatory diagram of an example of a problem to be solved as a combinatorial optimization problem. [Figure 7] FIG. 1 is a diagram illustrating the collaboration between combinatorial optimization and machine learning. [Figure 8] FIG. 2 is a diagram illustrating an example of material information. [Figure 9] FIG. 13 is a diagram showing an example of experimental data. [Figure 10] FIG. 13 is a diagram showing an example of the relationship between the calculated explanatory variable Xi and the variable Z. [Figure 11] FIG. 13 is a diagram illustrating an example of an optimal value and an allowable change range of an explanatory variable Xi. [Figure 12] FIG. 11 is an explanatory diagram of an example of a process for determining a weighting coefficient Ai of an explanatory variable Xi. [Figure 13] 1 is a diagram illustrating an example of the relationship between a threshold C for the allowable deviation from the minimum variable Z and the shape of the explanatory variable domain of a mixed material asymptotically approaching a target property. FIG. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0021] Next, an embodiment of the present invention will be described in detail, but the present invention is not limited to the following embodiment.

[0022] <System configuration> Fig. 1 is a configuration diagram of an example of an information processing system according to this embodiment. The information processing system 1 shown in Fig. 1 has an annealing-based computing device 10 and an information processing device 12. The annealing-based computing device 10 and the information processing device 12 are connected to each other so as to be able to communicate data with each other via a communication network 18 such as a local area network (LAN) or the Internet.

[0023] The annealing-based computing device 10 is an example of a device that uses an Ising model to solve combinatorial optimization problems. The annealing-based computing device 10 is, for example, an annealing machine that uses an Ising model. The annealing-based computing device 10 may be realized by a quantum computer, or may be realized by a Digital Annealer (registered trademark), which is a computer architecture that realizes the annealing method using digital circuits.

[0024] An annealing machine solves a combinatorial optimization problem that has been reduced to an Ising model by the convergence behavior of the Ising model. The Ising model is a statistical mechanics model that represents the behavior of magnetic materials. The Ising model has the property that the spin state of the magnetic material is updated so that the energy (Hamiltonian) is minimized due to the interaction between the spins, and the energy is ultimately minimized. An annealing machine reduces a combinatorial optimization problem to an Ising model and finds the state that minimizes the energy, thereby obtaining that state as the optimal solution for the combinatorial optimization problem.

[0025] The information processing device 12 is a device operated by a user, such as a PC, a tablet terminal, or a smartphone. The information processing device 12 receives an input of information required for causing the annealing-based computing device 10 to solve a combinatorial optimization problem reduced to an Ising model from a user, and causes the annealing-based computing device 10 to solve the Ising model.

[0026] The information required for the annealing computing device 10 to solve the combinatorial optimization problem reduced to an Ising model includes information on a function for converting into an Ising model. The information processing device 12 supports the user in creating a function for converting into an Ising model as described below, thereby supporting the creation of an Ising model.

[0027] Furthermore, the information processing device 12 receives the optimal solution to the combinatorial optimization problem solved by the annealing-based computing device 10, and outputs the optimal solution so that it can be confirmed by a user, for example by displaying it on a display device.

[0028] The information processing system 1 in FIG. 1 is just an example, and the information processing system 1 may be configured such that a user accesses and uses the information processing device 12 from a user terminal (not shown) connected to the information processing device 12 via the communication network 18.

[0029] The annealing method computing device 10 may also be realized as a cloud computing service. For example, the annealing method computing device 10 may be available by calling an API (Application Programming Interface) via a communication network 18.

[0030] Furthermore, the annealing method computing device 10 is not limited to being realized as a cloud computing service, but may be realized on-premise or operated by another company. The annealing method computing device 10 may be realized by multiple computers.

[0031] In addition, in a form in which a user accesses and uses the information processing device 12, the information processing device 12 may be realized as a cloud computing service, may be realized on-premise, may be operated by another company, or may be realized by a plurality of computers. It goes without saying that there are various system configuration examples of the information processing system 1 in FIG. 1 according to the use and purpose.

[0032] <Hardware configuration> The information processing device 12 in FIG. 1 is realized by, for example, a computer 500 having a hardware configuration shown in FIG.

[0033] Fig. 2 is a hardware configuration diagram of an example of a computer according to this embodiment. The computer 500 in Fig. 2 includes an input device 501, a display device 502, an external I / F 503, a RAM 504, a ROM 505, a CPU 506, a communication I / F 507, and a HDD 508, which are all connected to each other via a bus B. The input device 501 and the display device 502 may be connected to each other for use.

[0034] The input device 501 is a touch panel, operation keys or buttons, a keyboard, a mouse, etc., used by the user to input various signals. The display device 502 is composed of a display such as a liquid crystal or organic EL display for displaying a screen, a speaker for outputting sound data such as voice and sound, etc. The communication I / F 507 is an interface for the computer 500 to perform data communication.

[0035] The HDD 508 is an example of a non-volatile storage device that stores programs and data. The stored programs and data include an OS, which is basic software that controls the entire computer 500, and applications that provide various functions on the OS. Note that the computer 500 may use a drive device (such as a solid-state drive (SSD)) that uses a flash memory as a storage medium instead of the HDD 508.

[0036] The external I / F 503 is an interface with an external device. The external device may include a recording medium 503a. This allows the computer 500 to read and / or write data from and to the recording medium 503a via the external I / F 503. The recording medium 503a may include a flexible disk, a CD, a DVD, an SD memory card, a USB memory, and the like.

[0037] The ROM 505 is an example of a non-volatile semiconductor memory (storage device) that can retain programs and data even when the power is turned off. The ROM 505 stores programs and data such as BIOS, OS settings, and network settings that are executed when the computer 500 is started. The RAM 504 is an example of a volatile semiconductor memory (storage device) that temporarily retains programs and data.

[0038] The CPU 506 is a calculation device that reads out programs and data from storage devices such as the ROM 505 and the HDD 508 onto the RAM 504 and executes processing to realize the overall control and functions of the computer 500. The information processing device 12 according to this embodiment can realize various functions as described below. Note that a description of the hardware configuration of the annealing-based calculation device 10 will be omitted.

[0039] <Configuration> The configuration of an information processing system 1 according to this embodiment will be described. Fig. 3 is a configuration diagram of an example of an information processing system according to this embodiment. Note that in the configuration diagram of Fig. 3, parts that are not necessary for the description of this embodiment are omitted as appropriate.

[0040] 3 includes a call receiving unit 20 and an optimal solution calculation unit 22. The information processing device 12 includes an input receiving unit 30, a calculation unit 32, a determination unit 34, an output unit 36, a conversion unit 38, a linking unit 40, a display unit 42, an experimental data storage unit 50, a formula storage unit 52, and a material information storage unit 54.

[0041] The input receiving unit 30 is an input interface that receives an input of information required for the annealing-based computing device 10 to solve the combinatorial optimization problem reduced to an Ising model from a user. The information required for the annealing-based computing device 10 to solve the combinatorial optimization problem reduced to an Ising model includes information on a function for converting to an Ising model. The function for converting to an Ising model is written to include one or more explanatory variables (group of explanatory variables).

[0042] The calculation unit 32 calculates the relationship between one or more explanatory variables and the physical properties of the mixed material using a trained machine learning model described using one or more explanatory variables of a function that converts to an Ising model. In addition, the explanatory variables used to describe the machine learning model are preferably characteristics of the mixed material that can be described by a weighted average of the ratio of the material composition. The machine learning model is assumed to have learned the relationship between the characteristics of the mixed material that can be described by a weighted average of the ratio of the material composition and the physical properties of the mixed material using experimental data.

[0043] The determination unit 34 determines optimal values ​​and allowable change ranges of one or more explanatory variables for the target properties based on the relationship between the one or more explanatory variables calculated by the calculation unit 32 and the properties of the mixed material, as described below.

[0044] The output unit 36 ​​outputs the optimal values ​​of the one or more explanatory variables determined by the determination unit 34 as target values ​​of the one or more explanatory variables of the function for converting into an Ising model. The output unit 36 ​​also outputs weighting coefficients of the one or more explanatory variables of the function for converting into an Ising model, based on the allowable change range of the one or more explanatory variables determined by the determination unit 34, as described below.

[0045] The conversion unit 38 converts the function into which the target values ​​of one or more explanatory variables and the weighting coefficients of one or more explanatory variables output by the output unit 36 ​​are substituted, into an Ising model in a data format that can be used by the annealing-based computing device 10.

[0046] The collaboration unit 40 transmits the Ising model converted by the conversion unit 38 to the annealing-based calculation device 10. The collaboration unit 40 also receives the optimal solution calculated by the annealing-based calculation device 10.

[0047] The display unit 42 displays the optimum solution received by the cooperation unit 40 on the display device 502 for the user to confirm. The optimum solution displayed on the display device 502 is displayed as, for example, a mixing ratio of the mixed materials, which is easy for the user to understand.

[0048] The experimental data storage unit 50 stores experimental data used for learning the machine learning model. The mathematical formula storage unit 52 stores a function for converting to an Ising model and a machine learning model. The material information storage unit 54 stores material information such as the characteristics of the material.

[0049] The call receiving unit 20 receives a call from the information processing device 12 and receives an Ising model converted into a usable data format from the information processing device 12. The optimal solution calculation unit 22 searches for an optimal solution of the mixture ratio of the mixed materials where the physical properties asymptotically approach the target values ​​by determining a state where the energy (Hamiltonian) of the Ising model received by the call receiving unit 20 is minimized. The call receiving unit 20 transmits the optimal solution thus found to the information processing device 12.

[0050] 3 is just an example, and various configurations of the information processing system 1 according to this embodiment are possible.

[0051] <Processing> 4 is a flowchart showing an example of a processing procedure of a material composition searching method using the information processing system according to the present embodiment. In step S100, a user creates a function for converting to an Ising model. The function for converting to an Ising model has a configuration as shown in FIG. 5, for example.

[0052] FIG. 5 is an explanatory diagram of an example of a function for converting into an Ising model. The function for converting into an Ising model shown in FIG. 5 includes one or more explanatory variables X i Contains the values ​​(i=1, 2, …). X ibest is the explanatory variable X i The target value of explanatory variable X i is a property of a mixed material, such as HSP value or molecular weight, that describes the weighted average of the percentages of the material components.

[0053] In Figure 5, the terms other than the last one are the explanatory variables X i is the target value X ibestThe closer it is to , the smaller the value becomes. The mixture ratio constraint term is a term that becomes "0" when the sum of the mixture ratios of the materials is "100%". In addition, in the function for converting to the Ising model in Figure 5, A i is the explanatory variable X i B is the weighting coefficient for the mixture ratio constraint term.

[0054] The annealing calculation device 10 obtains the state in which the function in FIG. 5 is minimized as the optimization result. i is the target value X ibest The optimum material composition is calculated as the state where the sum of the material mixture ratios is as close as possible to "100%."

[0055] In this embodiment, the target value X ibest and weighting factor A i By outputting the above to the information processing device 12, the creation of a function for converting into an Ising model is supported.

[0056] 6A to 6C are explanatory diagrams of an example of a problem to be solved as a combinatorial optimization problem. 0 This indicates that the goal is to create a mixture of solvents that asymptotically approaches (approximates)

[0057] The average molecular weight M of the mixed solvent obtained by mixing the solvent groups can be calculated by the formula shown in FIG. 6B. In addition, when the average molecular weight M of the mixed solvent obtained by mixing the solvent groups is M 0 The function that becomes smaller as the mixed solvent approaches asymptotically can be described, for example, as shown in FIG. 6C. 0 is an example of a property of a mixed solvent that can be described by a weighted average of the mixing ratios of the mixed solvents.

[0058] Returning to FIG. 4, in step S102, the user selects an explanatory variable X i A machine learning model for predicting the physical property Y of the mixed material is created, which is described using the following formula: The process of step S102 will be described with reference to FIG.

[0059] Figure 7 is a diagram explaining the link between combinatorial optimization and machine learning. Figure 7 shows an example of solving the optimal mixture ratio of monomer types to create a polymer with physical property Y as a combinatorial optimization problem.

[0060] The function for converting into an Ising model for solving the combinatorial optimization problem in Figure 7 contains a group of explanatory variables X that can be described by a weighted average of the mixing ratio of monomer types, as explained using Figure 5 etc. Therefore, the relationship between the group of explanatory variables X of a polymer and the physical property Y of the polymer can be obtained by creating a machine learning model for predicting the physical property of the polymer described using the group of explanatory variables X of the polymer and having it learn from experimental data.

[0061] In this manner, in this embodiment, the explanatory variable set X i By using a machine learning model that has learned the relationship between the explanatory variable group Y and the physical property Y of the polymer, the region of the explanatory variable group in the explanatory variable space for obtaining a polymer with the target physical property can be obtained as described below. By obtaining the region of the explanatory variable group Y for obtaining a polymer with the target physical property in the explanatory variable space of FIG. 7, this embodiment can obtain the explanatory variable group X shown in FIG. i The optimal value and allowable change range of are determined as described below, and the explanatory variable X i Target value X ibest and weighting factor A i Used for output of.

[0062] Returning to Fig. 4, in step S104, the user trains the machine learning model created in step S102 with experimental data. There are no restrictions on the type of machine learning model, and LASSO or random forest can be used. For example, material information as shown in Fig. 8 and experimental data as shown in Fig. 9 are used to train the machine learning model.

[0063] FIG. 8 is a diagram showing an example of material information. In the material information in FIG. 8, characteristic values ​​are set for each material. FIG. 9 is a diagram showing an example of experimental data. In the experimental data in FIG. 9, the compounding ratio of the mixed materials and the physical property values ​​of the mixed materials mixed at that compounding ratio are set. Explanatory variables i can be calculated based on the material information in FIG. 8 and the compounding ratio of the mixed materials shown in FIG.

[0064] 4, in step S106, the user uses the machine learning model trained in step S104 to create a variable Z that outputs a smaller value for a mixed material that approaches the target property. The calculation unit 32 of the information processing device 12 determines grid points for performing a comprehensive calculation in the explanatory variable space, for example, and performs a comprehensive calculation of the variable Z to generate an explanatory variable X as shown in FIG. i Calculate the relationship between and variable Z.

[0065] Figure 10 shows the calculated explanatory variable X i This is an example diagram showing the relationship between the variable Z and the variable Z. The smaller the value of the variable Z, the closer the mixed material is to the target physical property. Therefore, the explanatory variable X corresponding to the variable Z with the smallest value shown in FIG. i But the explanatory variable X i In addition, by setting a threshold value for the allowable deviation from the minimum variable Z value (for example, allowing a deviation of c% from the variable Z value), multiple variable Z values ​​within the allowable deviation threshold from the minimum variable Z value can be selected.

[0066] Returning to FIG. 4, in step S108, the decision unit 34 selects the explanatory variable X when the variable Z is at its minimum. i The value of the explanatory variable X i The output unit 36 ​​outputs the determined explanatory variable X i The optimal value of the explanatory variable X i Target value X ibest The output is as follows:

[0067] In step S110, the decision unit 34 selects a plurality of variable Z values ​​within a threshold of the allowable deviation from the minimum variable Z value, and calculates an explanatory variable X iUsing the values ​​of i Determine the allowable variation range.

[0068] Figure 11 shows the explanatory variable X i FIG. 11 is a diagram illustrating an example of the optimal value and the allowable change range of explanatory variable X i Here is a two-dimensional example. Explanatory variable X 1 The optimal value of is X 1best The explanatory variable X 2 The optimal value of is X 2best The value is:

[0069] The elliptical area shown in Figure 11 is the explanatory variable X of multiple variables Z within the threshold of the deviation allowance from the minimum variable Z value. 1 and explanatory variable X 2 The elliptical region shown in Fig. 11 is the explanatory variable region of the mixed material that asymptotically approaches the target physical property. Note that Fig. 11 shows an example in which the explanatory variable region of the mixed material that asymptotically approaches the target physical property is an ellipse, but it is not limited to an ellipse.

[0070] The determination unit 34 determines the explanatory variable X from the elliptical region shown in FIG. 1 and explanatory variable X 2 For example, the allowable range of change for each explanatory variable X 1 The allowable variation range of is the distance of the minor axis of the ellipse shown in FIG. 11. Also, for example, the explanatory variable X 2 The allowable variation range is the distance of the major axis of the ellipse shown in FIG.

[0071] Returning to FIG. 4, in step S112, the output unit 36 ​​outputs the determined explanatory variable X i From the allowable change range for each explanatory variable X i Weighting factor A i is determined and outputted, for example, by a method as shown in FIG.

[0072] Figure 12 shows the explanatory variable X i Weighting factor A i FIG. 1 is an explanatory diagram of an example of a process for determining an explanatory variable X i Weighting factor A i is the axial distance r of the ellipse iIn this case, 1 / r i In the example of Figure 12, the explanatory variable X 1 Weighting factor A 1 is 1 / r 1 In the example of Figure 12, the explanatory variable X 2 Weighting factor A 2 is 1 / r 2 It becomes.

[0073] For example, in the explanatory variable region (ellipse) of the mixed material that asymptotically approaches the target property shown in Figure 12, the explanatory variable X 1 The sensitivity of the explanatory variable X 2 Since the sensitivity of i The importance of explanatory variable X 1 > Explanatory variable X 2 Weighting factor A 1 >Weighting factor A 2 In the example of Figure 12, the explanatory variable X i The weighting factor A i Explanatory variable X with small allowable variation (small variance) i The weighting factor A i is large and is output.

[0074] The shape of the explanatory variable region of the mixed material asymptotically approaching the target property shown in Fig. 12 changes depending on the threshold value of the allowable deviation from the value of the minimum variable Z. Fig. 13 is a diagram illustrating an example of the relationship between the threshold value C of the allowable deviation from the minimum variable Z and the shape of the explanatory variable region of the mixed material asymptotically approaching the target property.

[0075] If the threshold C for the allowable deviation from the minimum variable Z is too large, the shape of the explanatory variable region of the mixed material approaching the target property becomes too large, as shown in the graph on the left side of Figure 13, and the explanatory variable X for the target property becomes too large. 1 and explanatory variable X 2 In addition, if the threshold C for the allowable deviation from the minimum variable Z is too small, the shape of the explanatory variable region of the mixed material approaching the target property becomes too small, as shown in the graph on the right side of Figure 13, and the explanatory variable X for the target property becomes too small. 1and explanatory variable X 2 There is a risk that the sensitivity of the

[0076] Therefore, the threshold C for the allowable deviation from the minimum variable Z is determined by the explanatory variable X, as shown in the middle graph of Figure 13. i It is preferable to set the threshold value C so that the difference between the minimum variable Z can be distinguished. The threshold value C for the allowable deviation from the minimum variable Z may be adjusted by the user with reference to the shape of the explanatory variable domain of the mixed material that asymptotically approaches the target physical property, or may be adjusted when it is determined that the optimization result by the annealing-based calculation device 10 is not sufficient.

[0077] Returning to FIG. 4, in step S114, the conversion unit 38 converts the explanatory variable X i Target value X ibest and the explanatory variable X output in step S112 i Weighting factor A i In step S116, the conversion unit 38 substitutes the explanatory variable X i Target value X ibest and explanatory variable X i Weighting factor A i The function into which and are substituted is converted into the Ising model (the matrix elements of the Ising model formula are calculated).

[0078] The technology for converting a function into a quadratic unconstrained binary optimization (QUBO) format for an evaluation function, or into an Ising model in a data format that can be used by the annealing-based computing device 10, is provided as a WebAPI, etc., and is an existing technology.

[0079] For example, the transformation unit 38 is an explanatory variable X i Target value X ibest and explanatory variable X i Weighting factor A i Expand the function into which and are substituted, and obtain the matrix element Q of the Ising model. i,j The linking unit 40 calculates the matrix element Q i,j is transmitted to the annealing calculation device 10 as a parameter of the Ising model.

[0080] In step S118, the optimal solution calculation unit 22 of the annealing-based computing device 10 that has received the parameters of the Ising model searches for an optimal solution of the mixture ratio of the mixed materials that minimizes the function of Fig. 5. The call receiving unit 20 transmits information indicating the optimal solution thus searched for to the information processing device 12.

[0081] In step S120, the display unit 42 converts the information (bit information of the annealing-based calculation device 10) received by the collaboration unit 40 as the optimal solution into information such as a mixing ratio of the mixed materials that is easy for the user to understand, and outputs the information. For example, the display unit 42 displays the names of the materials contained in the mixed solvent of the optimal solution and the compounding ratio of the materials.

[0082] In this embodiment, the target value X of the explanatory variable, which the user had to input in order to have the annealing machine solve the combinatorial optimization problem reduced to the Ising model, is ibest and weighting factor A i You can omit entering

[0083] The mixture ratio of the mixed material searched as the optimal solution is used to generate the mixed material by specifying the materials to be mixed and their mixture ratio. For example, it can be used to control a mixed material generation device. In addition, the physical properties of the mixed material generated by the mixed material generation device can be evaluated by an evaluation device.

[0084] Therefore, the information of the mixed material searched for as the optimal solution can be compared with the physical properties of the mixed material generated by the mixed material generation device by specifying the mixed material information, and the accuracy of the search for the optimal solution can be improved by feeding back the results of this comparison.

[0085] As described above, the information processing system 1 according to this embodiment can reduce the effort required to create an Ising model for making the annealing-based computing device 10 solve a combinatorial optimization problem of material compositions that asymptotically approach target properties.

[0086] Although the present embodiment has been described above, it will be understood that various changes in form and details can be made without departing from the spirit and scope of the claims.

[0087] Although the present invention has been described above based on the embodiments, the present invention is not limited to the above embodiments and various modifications are possible within the scope of the claims. This application claims priority to basic application No. 2022-031084 filed on March 1, 2022 with the Japan Patent Office, the entire contents of which are incorporated herein by reference. [Explanation of symbols]

[0088] 1. Information Processing Systems 10 Annealing-based computing device 12 Information processing equipment 18 Communication Networks 20 Call Reception Department 22 Optimal solution calculation section 30 Input reception section 32 Calculation section 34 Decision Section 36 Output section 38 Conversion section 40 Collaboration Department 42 Display section 50 Experimental data storage unit 52 Formula memory section 54 Material information storage section

Claims

[Claim 1] A method for generating a mixed material using a material composition search method of an information processing system having an annealing type computing device using an Ising model and an information processing device that supports the creation of an Ising model for causing the computing device to solve a combinatorial optimization problem of a material composition that asymptotically approaches a target property, comprising: Calculating a relationship between the one or more explanatory variables and the physical properties of the mixed material using a trained machine learning model for predicting the physical properties of the mixed material, the relationship being described using one or more explanatory variables of a function that converts into the Ising model; determining optimal values ​​and allowable variation ranges of the one or more explanatory variables for the target property based on a relationship between the one or more explanatory variables and the physical property of the mixed material; outputting the determined optimal values ​​of the one or more explanatory variables as target values ​​of the one or more explanatory variables in the function, and outputting weighting coefficients of the one or more explanatory variables in the function based on the determined allowable change ranges of the one or more explanatory variables; converting the function into which the output target values ​​of the one or more explanatory variables and weighting coefficients of the one or more explanatory variables are substituted into the Ising model; Calculating an optimal solution for a material composition that approaches the target value using the Ising model; generating the mixed material using an optimal solution of the material composition that asymptotically approaches the target value; Methods for producing mixed materials.

Images