Material design device, material design method, and program

Abstract

The present invention performs a local search that maintains initial value dependency and is less likely to fall into a local solution. A material design device includes: a peripheral composition generation unit configured to generate a peripheral composition obtained by varying each factor included in a composition candidate of a target substance within a predetermined exploration range; a characteristic prediction section configured to predict a material characteristic of the target substance based on the peripheral composition; a composition candidate updating unit configured to update the composition candidate with a peripheral composition when there is the peripheral composition that improves the material characteristics; a search range expanding part configured to expand the search range when there is no peripheral composition that improves material characteristics; and a factor exchange unit configured to exchange the factors constituting the candidates when the search range exceeds a predetermined maximum search range.

Description

Technical Field

[0001] The present disclosure relates to a material design device, a material design method, and a program. Background Art

[0002] In recent years, technologies that use machine learning and other methods to improve the efficiency of material development have been applied. In material development using machine learning and other methods, the optimal solution for characteristic values ​​is efficiently searched by combining large-scale exploration and local exploration.

[0003] For example, patent document 1 discloses an invention in which a Bayesian model is generated by modeling the problem of exploring a combination of multiple parameter values ​​that gives an optimal value of a characteristic value related to a target substance, and the Bayesian model is used to explore the combination in an exploration space narrower than the full exploration space based on all combinations of the multiple parameter values.

[0004] Patent Document 1: Japanese Patent Application Publication No. 2020-187642 Summary of the invention

[0005] <Problems to be Solved by the Invention>

[0006] However, in the previous technology, there is a problem that the factor exchange is not performed in the local exploration. If the factor exchange is not performed, it may be trapped in the local solution depending on the initial value in the local exploration. On the other hand, if the factor exchange is performed randomly, the material properties that were originally better at the initial value may be lost.

[0007] One aspect of the present disclosure aims to perform local exploration that maintains initial value dependence and at the same time is not prone to falling into local solutions.

[0008] <Methods used to solve the problem>

[0009] The present disclosure includes the following configurations.

[0010] [1] A material design device, comprising:

[0011] a peripheral composition generating unit configured to generate a peripheral composition obtained by varying each factor included in the composition candidate of the target substance within a predetermined search range;

[0012] a property prediction unit configured to predict a material property of the target substance based on the surrounding composition;

[0013] a composition candidate updating unit configured to update the composition candidate using the surrounding composition when there is the surrounding composition that improves the material property;

[0014] a search range expansion unit configured to expand the search range when the peripheral composition that improves the material property does not exist; and

[0015] The factor exchanging unit is configured to exchange the factors of the constituent candidates when the search range exceeds a predetermined maximum search range.

[0016] [2] The material design device according to [1] above, wherein:

[0017] The property prediction unit is configured to predict the material property using a prediction model that uses the material composition of the target substance as an explanatory variable and the material property as a target variable,

[0018] The factor exchanging unit is configured to exchange the factors constituting the candidate based on the contribution of each factor in the prediction model.

[0019] [3] The material design device according to [2] above, wherein:

[0020] The characteristic prediction unit is configured to predict a plurality of the material characteristics.

[0021] The composition candidate updating unit is configured to update the composition candidate when the first material characteristic is maintained and the second material characteristic is improved.

[0022] [4] The material design device according to [3] above, wherein:

[0023] The factor exchange unit is configured to exchange a factor having a high contribution to the second material characteristic with a factor having a low contribution to the first material characteristic.

[0024] [5] The material design device according to [4] above, further comprising:

[0025] The initial composition determination unit is configured to determine, as the composition candidate, the material composition having the better first material characteristic among the material compositions generated randomly or with a predetermined step size from the entire material composition of the target substance.

[0026] [6] The material design device according to [4] or [5] above, wherein:

[0027] The factor exchange unit is configured to exchange factors of the same type.

[0028] [7] The material design device according to any one of [1] to [6] above, wherein:

[0029] The composition candidate updating unit is configured to output the composition candidate when the composition candidate is updated a predetermined number of times.

[0030] [8] The material design device according to any one of [1] to [7] above, wherein:

[0031] The property prediction unit is configured to predict the material property for the surrounding composition having a similarity with the composition candidate that is greater than or equal to a predetermined threshold.

[0032] [9] A material design method, wherein a computer executes the following steps:

[0033] generating a peripheral composition obtained by varying each factor included in the composition candidate of the target substance within a predetermined search range;

[0034] predicting material properties of the target substance based on the surrounding composition;

[0035] When the peripheral composition that improves the material property exists, updating the composition candidate using the peripheral composition;

[0036] In the absence of the surrounding composition that improves the material properties, expanding the exploration range; and

[0037] When the search range exceeds a prescribed maximum search range, the factors constituting the candidates are exchanged.

[0038]

[10] A program for causing a computer to execute the following steps:

[0039] generating a peripheral composition obtained by varying each factor included in the composition candidate of the target substance within a predetermined search range;

[0040] predicting material properties of the target substance based on the surrounding composition;

[0041] When the peripheral composition that improves the material property exists, updating the composition candidate using the peripheral composition;

[0042] In the absence of the surrounding composition that improves the material properties, expanding the exploration range; and

[0043] When the search range exceeds a prescribed maximum search range, the factors constituting the candidates are exchanged.

[0044] <Effects of the Invention>

[0045] According to one aspect of the present invention, local exploration can be performed which maintains initial value dependence and is not prone to falling into a local solution. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1This is a block diagram showing an example of the overall configuration of the design support system.

[0047] Figure 2 This is a block diagram showing an example of the hardware configuration of a computer.

[0048] Figure 3 This is a block diagram showing an example of the functional configuration of the design support system.

[0049] Figure 4 This is a flowchart showing an example of the processing steps of the material design method.

[0050] Figure 5 (A) is a diagram showing an example of the contribution to the adhesive force. Figure 5 (B) is a diagram showing an example of contribution to strength.

[0051] Figure 6 (A) is a diagram showing an example of initial values ​​of composition candidates. Figure 6 (B) is a diagram showing an example of composition candidates after the first factor exchange. Figure 6 (C) is a diagram showing an example of composition candidates after the second factor exchange.

[0052] Figure 7 (A) is a diagram showing an example of the first local exploration. Figure 7 (B) is a diagram showing an example of the first local exploration. Figure 7 (C) is a diagram showing an example of the second local exploration. Figure 7 (D) is a diagram showing an example of the second local exploration.

[0053] Figure 8 (A) is a diagram showing an example of the x-th local exploration. Figure 8 (B) is a diagram showing an example of the x-th local exploration. Figure 8 (C) is a diagram showing an example of the x-th local exploration. Figure 8 (D) is a diagram showing an example of the x-th local exploration.

[0054] Fig. 9 (A) is a diagram showing an example of the result of the first local exploration. Fig. 9 (B) is a diagram showing an example of the result of the second local exploration. Fig. 9 (C) is a diagram showing an example of the result of the third local exploration. Fig. 9 (D) is a diagram showing an example of the result of the first local exploration after factor exchange.

[0055] Fig. 9 (E) is a diagram showing an example of the result of the second local exploration after factor exchange. Fig. 9(F) is a diagram showing an example of the result of the third local exploration after factor exchange.

[0056] Fig.10 (A) is a diagram showing an example of the result of the first local exploration after the first update. Fig.10 (B) is a diagram showing an example of the result of the first local exploration after the second update. Fig.10 (C) is a diagram showing an example of the result of the first local exploration after the third update.

[0057] Fig.11 (A) is a schematic diagram showing an example of a graph in which search results from the first search to the last search are summarized. Fig.11 (B) is a diagram showing an example of a graph of material properties of peripheral components in the first search and the final search. DETAILED DESCRIPTION

[0058] Hereinafter, embodiments of the present invention will be described with reference to the accompanying drawings. Note that in the specification and the drawings, components having substantially the same functional configuration are given the same reference numerals to omit repeated descriptions.

[0059] [Implementation Method]

[0060] In the past, material development was carried out based on the experience and intuition of technicians. In recent years, technologies that improve the efficiency of material development by using machine learning and other methods have been applied. In material development using machine learning, for example, large-scale exploration based on machine learning is carried out. Large-scale exploration based on machine learning can explore design conditions of materials including areas that humans have not anticipated.

[0061] In large-scale exploration based on machine learning, sometimes design conditions are obtained that can achieve the target value of a certain material property A, but cannot meet the standards of other material properties B that have a trade-off relationship with material property A. In this case, since the target value of material property A has been achieved, it is efficient if the design condition can be used to advance material design.

[0062] In this case, for example, it is possible to consider performing exploration based on machine learning again. However, in the re-exploration based on machine learning, it is not possible to consider the design conditions that have achieved the target value for at least material property A. In addition, even if it is predicted that better material properties can be obtained through exploration based on machine learning alone, it is not guaranteed that it is a material composition that can be actually achieved.

[0063] In addition, for example, it is possible to consider using the obtained design conditions as initial values ​​and perform exploration through experimental planning. Alternatively, it is possible to consider having a technician adjust (tune) the obtained design conditions. However, in methods based on experimental planning and adjustment, there are problems such as an increase in the number of experiments and difficulty in considering factors other than the factors included in the initial values.

[0064] In machine learning-based exploration, large-scale exploration and local exploration are sometimes combined to efficiently perform exploration. However, in conventional techniques, since factor exchange is not performed during exploration, it is possible to fall into a local solution depending on the initial value. Even if factor exchange in mutation based on genetic algorithms is combined, factor exchange irrelevant to the material properties to be optimized may be performed, which may deteriorate the material properties.

[0065] One embodiment of the present disclosure is a design support system that supports the design of a substance manufactured using a plurality of substances. Hereinafter, the substance to be designed is referred to as a "target substance", and the substance used to manufacture the target substance is referred to as a "material substance". The design support system in this embodiment is an information processing system that explores the material composition of the target substance that optimizes the desired material properties according to the user's instructions, and prompts the user with the exploration results.

[0066] As an example, the target substance in this embodiment is a composite material such as a semiconductor material, which is composed of a plurality of resins, additives and / or fillers. Therefore, as an example, the material substance in this embodiment is a resin, an additive and / or a filler. An example of a semiconductor material is a resist material, an adhesive, an adhesive, a sealing material, etc. The target substance and the material substance are not limited thereto, and this embodiment can be applied to any substance as long as it is a substance made using a plurality of substances.

[0067] The design support system of this embodiment uses the material composition of the target material selected by the user as the initial value of the composition candidate, and repeatedly performs local exploration in the surrounding area of ​​the composition candidate to explore the composition candidate of the target material that can obtain better material properties. In the local exploration, multiple surrounding compositions are generated by changing each factor of the composition candidate within a specified exploration range. If there is a surrounding composition whose material properties predicted based on each surrounding composition are better than the target value specified by the user, the composition candidate is updated using the surrounding composition.

[0068] When there is no peripheral component that can improve material properties within a certain exploration range, the design support system of this embodiment expands the exploration range and performs local exploration again. In addition, when the exploration range exceeds the predetermined maximum exploration range, the factors that constitute the candidate are exchanged and local exploration is performed again. In the factor exchange, the factor that has a large impact on the material properties to be improved is exchanged for the factor that has a small impact on the material properties to be maintained.

[0069] Therefore, according to the design support system of this embodiment, it is possible to obtain an exploration result that is not much different from the initial value of the constituent candidate and maintains the initial value dependency. In addition, according to the design support system of this embodiment, since the factor exchange is performed when there is no update of the constituent candidate within the specified exploration range, it is possible to perform an exploration that is not prone to falling into a local solution.

[0070] <Overall Configuration of Design Support System>

[0071] Reference Figure 1 The overall configuration of the design support system of the present embodiment will be described. Figure 1 This is a block diagram showing an example of the overall configuration of the design support system according to the present embodiment.

[0072] like Figure 1 As shown, the design support system 1 of this embodiment includes a material design apparatus 10 and a user terminal 20. The material design apparatus 10 and the user terminal 20 are connected to each other via a communication network N1 such as a LAN (Local Area Network) or the Internet so as to be capable of data communication.

[0073] The material design device 10 is an information processing device such as a personal computer, a workstation, or a server, and explores the material composition of a target substance that can obtain better material properties in response to a request from a user terminal 20. The material design device 10 receives initial composition data related to the material composition of the target substance from the user terminal 20. The material design device 10 explores the material composition of the target substance based on the initial composition data, and sends exploration result data representing the material composition of the target substance obtained as a result of the exploration to the user terminal 20.

[0074] The user terminal 20 is an information processing terminal such as a personal computer, a tablet terminal, or a smart phone operated by a user. The user terminal 20 generates initial composition data according to the user's operation and sends it to the material design device 10. The user terminal 20 displays the exploration result to the user based on the exploration result data received from the material design device 10.

[0075] Notice, Figure 1The overall configuration of the design support system 1 shown is an example, and various system configurations may be used depending on the use and purpose. For example, the material design device 10 may be implemented by multiple computers or as a cloud computing service. In addition, for example, the design support system 1 may also be implemented by an independent information processing device that has the functions that the material design device 10 and the user terminal 20 should each have.

[0076] <Hardware Configuration of Design Support System>

[0077] Reference Figure 2 The hardware configuration of the design support system 1 of the present embodiment will be described.

[0078] 《Computer Hardware Configuration》

[0079] The material design device 10 and the user terminal 20 of the present embodiment are realized by, for example, a computer. Figure 2 This is a block diagram showing an example of the hardware configuration of the computer 500 according to the present embodiment.

[0080] like Figure 2 As shown, the computer 500 includes a CPU (central processing unit) 501, a ROM (read only memory) 502, a RAM (random access memory) 503, an HDD (hard disk drive) 504, an input device 505, a display device 506, a communication I / F (interface) 507, and an external I / F 508. The CPU 501, the ROM 502, and the RAM 503 form a so-called computer. The various hardware of the computer 500 are connected to each other via a bus 509. Note that the input device 505 and the display device 506 can be connected to the external I / F 508 to be used.

[0081] The CPU 501 is a computing device that realizes the control and functions of the entire computer 500 by reading programs or data from a storage device such as the ROM 502 or the HDD 504 to the RAM 503 and executing processing.

[0082] ROM502 is an example of a nonvolatile semiconductor memory (storage device) that can retain programs and data even when the power is turned off. ROM502 is used as a main storage device that stores various programs, data, and the like required for CPU501 to execute various programs installed in HDD504. Specifically, ROM502 stores data such as boot programs such as BIOS (Basic Input / Output System) and EFI (Extensible Firmware Interface) executed when the computer 500 is started, OS (Operating System) settings, network settings, and the like.

[0083] RAM 503 is an example of a volatile semiconductor memory (storage device) that deletes programs and data when the power is turned off. RAM 503 is, for example, DRAM (dynamic random access memory), SRAM (static random access memory), etc. RAM 503 provides a work area that is expanded when CPU 501 executes various programs installed in HDD 504.

[0084] The HDD 504 is an example of a nonvolatile storage device that stores programs and data. The programs and data stored in the HDD 504 include an OS, which is basic software for controlling the entire computer 500, and applications that provide various functions on the OS. Note that the computer 500 may use a storage device (e.g., an SSD: solid state drive, etc.) that uses a flash memory as a storage medium instead of the HDD 504.

[0085] The input device 505 is a touch panel, operation keys or buttons, a keyboard or a mouse used by a user to input various signals, a microphone for inputting audio data such as sound, or the like.

[0086] The display device 506 is constituted by a display such as liquid crystal or organic EL (electroluminescence) for displaying a screen, a speaker for outputting audio data such as sound, and the like.

[0087] The communication I / F 507 is an interface for connecting the computer 500 to a communication network and performing data communication.

[0088] The external I / F 508 is an interface with an external device. The external device includes a driver device 510 and the like.

[0089] The drive device 510 is a device for setting the recording medium 511. The recording medium 511 mentioned here includes a medium that records information optically, electrically or magnetically, such as a CD-ROM, a floppy disk, a magneto-optical disk, etc. In addition, the recording medium 511 may include a semiconductor memory that records information electrically, such as a ROM, a flash memory, etc. Therefore, the computer 500 can read and / or write to the recording medium 511 via the external I / F 508.

[0090] Note that various programs installed on HDD 504 are installed, for example, by setting distributed recording medium 511 in drive device 510 connected to external I / F 508 and reading various programs recorded on recording medium 511 through drive device 510. Alternatively, various programs installed on HDD 504 may be installed by downloading from a network other than the communication network via communication I / F 507.

[0091] <Functional configuration of design support system>

[0092] Will refer to Figure 3The functional configuration of the design support system according to the present embodiment is explained. Figure 3 This is a block diagram showing an example of the functional configuration of the design support system 1 according to the present embodiment.

[0093] 《Functional configuration of material design device》

[0094] like Figure 3 As shown, the material design device 10 of this embodiment includes a model storage unit 100, an initial composition determination unit 101, a peripheral composition generation unit 102, a characteristic prediction unit 103, an exploration range expansion unit 104, a factor exchange unit 105, a composition candidate update unit 106, an exploration range reduction unit 107 and a result output unit 108.

[0095] The initial composition determination unit 101, the peripheral composition generation unit 102, the characteristic prediction unit 103, the search range expansion unit 104, the factor exchange unit 105, the composition candidate update unit 106, the search range reduction unit 107 and the result output unit 108 are executed by the CPU 501. Figure 2 The model storage unit 100 is implemented by the processing of the program expanded on the RAM 503 by the HDD 504 shown in FIG. Figure 2 The HDD 504 shown is implemented.

[0096] The model storage unit 100 stores a learned property prediction model. The property prediction model is a machine learning model that uses the factors included in the material composition of the target substance as explanatory variables and the material properties of the target substance as target variables. An example of a machine learning model is a regression model. The structure of the machine learning model can be, for example, a random forest or a deep neural network.

[0097] The initial composition determination unit 101 receives initial composition data generated according to the user's operation from the user terminal 20. The initial composition data of this embodiment includes the material composition of the target substance and the target value of the material property. The initial composition determination unit 101 determines the material composition of the target substance included in the initial composition data as the initial value of the composition candidate.

[0098] For the material composition included in the initial composition data, a material composition with a better measured value or predicted value of the desired material property can be selected from the experimental results or the results of large-scale exploration based on machine learning. In addition, a material composition with a better measured value of the desired material property can be selected from the experimental results using the material composition explored by large-scale exploration. If a material composition that can obtain good material properties is selected as an initial value in the experimental results, it can be expected to obtain a composition candidate with a high possibility of realization.

[0099] Note that whether the material properties are good refers to whether the properties that a product manufactured using the target substance should have are excellent. In the case where the greater the value of the material property, the better the product properties, it can be considered that the greater the value of the material property, the better. On the other hand, in the case where the smaller the value of the material property, the better the product properties, it can be considered that the smaller the value of the material property, the better.

[0100] As an example, the experimental results using the material composition explored through large-scale exploration can be obtained as follows. First, a set of material compositions is generated from the entire material composition of the target substance, in which each factor varies with a specified variation range and within the entire range. Then, a specified number of material compositions are randomly sampled from the set of material compositions. Next, experiments are conducted on the sampled material compositions to measure multiple material properties including the desired material properties. Based on the experimental results obtained in this way, a material composition with better desired material properties can be selected as the material composition included in the initial composition data. A specified number of material compositions can be sampled from the set of material compositions at a specified step size.

[0101] The target values ​​of the material properties included in the initial composition data can be arbitrarily determined by the user, and the best material properties obtained through experimental results or large-scale exploration, or material properties that are further improved from the best material properties, can also be set as target values.

[0102] The peripheral composition generation unit 102 generates a plurality of peripheral compositions obtained by varying each factor included in the composition candidate of the target substance within a predetermined search range. Specifically, the peripheral composition generation unit 102 generates a predetermined number of peripheral compositions by randomly or with a predetermined step length sampling a predetermined number of material compositions from a set of material compositions obtained by varying the blending amount of each factor included in the composition candidate within a predetermined range of variation and within a predetermined search range.

[0103] The characteristic prediction unit 103 predicts a plurality of material characteristics based on each peripheral composition generated by the peripheral composition generation unit 102. The plurality of material characteristics include material characteristics that are to be maintained (hereinafter also referred to as "maintained characteristics") and material characteristics that are to be improved (hereinafter also referred to as "improvement target characteristics") among the material characteristics based on the initial composition. The characteristic prediction unit 103 calculates the plurality of material characteristics by inputting each factor included in the peripheral composition into the learned characteristic prediction model stored in the model storage unit 100.

[0104] The property prediction unit 103 may predict material properties only for the surrounding compositions whose similarity to the composition candidate is greater than or equal to a predetermined threshold. Regarding the similarity in this embodiment, for example, for the composition candidate and the surrounding composition, a vector with the amount of each factor as an element may be generated, and the distance between the vectors may be calculated using an arbitrary distance scale. An example of the distance scale in this embodiment is the Euclidean distance.

[0105] When there is no surrounding composition in which the predicted value of the material property predicted by the property prediction unit 103 is better than the target value of the material property, the search range expansion unit 104 expands the search range.

[0106] When the search range is greater than or equal to a predetermined maximum search range, the factor exchange unit 105 exchanges factors constituting the candidate. An example of the maximum search range in this embodiment is a range expanded a predetermined number of times by the search range expansion unit 104. When the factors are exchanged, the factor exchange unit 105 restores the search range to the initial state.

[0107] The factor exchange unit 105 exchanges factors based on the contribution of each factor in the learned characteristic prediction model stored in the model storage unit 100. Specifically, the factor exchange unit 105 exchanges factors with high contribution to the improved characteristic in the initial values ​​of the constituent candidates for factors with low contribution to the maintained characteristic.

[0108] When there is a surrounding composition whose predicted value of the material property predicted by the property prediction unit 103 is better than the target value of the material property, the composition candidate update unit 106 updates the composition candidate using the surrounding composition. When the maintained characteristic of the surrounding composition maintains the target value and the improved characteristic of the surrounding composition is improved than the target value, the composition candidate update unit 106 uses the surrounding composition as a new composition candidate. When the composition candidate is updated, the composition candidate update unit 106 updates the target value of the material property with the predicted value of the material property based on the new composition candidate.

[0109] Note that maintenance of material properties means that there is no significant difference in the material properties between the original material property value and the new material property value. Improvement of material properties means that the new material property value is better than the original material property value.

[0110] When the factors of the composition candidates are exchanged by the factor exchanging unit 105 or when the composition candidates are updated by the composition candidate updating unit 106 , the search range narrowing unit 107 narrows the search range to the initial state.

[0111] When the specified termination condition is met, the result output unit 108 generates exploration result data and sends it to the user terminal 20. An example of the termination condition in this embodiment is when the composition candidate is updated a specified number of times. Another example of the termination condition in this embodiment is when the predicted value based on the material properties of the composition candidate reaches the target value specified by the user.

[0112] The exploration result data in this embodiment includes the current composition candidate of the target substance and the predicted values ​​of multiple material properties based on the composition candidate. The exploration result data in this embodiment may also include historical information of the composition candidate updated during the exploration. The historical information of the composition candidate may include, for example, a list of each composition candidate updated during the exploration and a list of predicted values ​​of multiple material properties based on each composition candidate.

[0113] <<Functional configuration of user terminal 20>>

[0114] like Figure 3 As shown, the user terminal 20 of this embodiment includes an initial composition input unit 201 and a result display unit 202 .

[0115] The initial composition input unit 201 and the result display unit 202 are configured by causing the CPU 501 to execute Figure 2 The HDD 504 shown is implemented by processing of a program expanded on the RAM 503 .

[0116] The initial composition input unit 201 receives input of initial composition data according to a user's operation and transmits the received initial composition data to the material design device 10 .

[0117] The result display unit 202 receives the search result data from the material design device 10. The result display unit 202 displays the composition candidates of the target substance and the material properties of the composition candidates on the display device 506 based on the received search result data.

[0118] <Processing steps of the design support system>

[0119] Will refer to Figure 4 The processing steps of the material design method executed by the design support system 1 of the present embodiment will be described. Figure 4 This is a flowchart showing an example of the processing procedure of the material design method according to the present embodiment.

[0120] In step S1 , the initial composition input unit 201 of the user terminal 20 receives input of initial composition data including material composition of the target substance and target values ​​of material properties according to user operation. Then, the initial composition input unit 201 transmits the received initial composition data to the material design device 10 .

[0121] In the material design device 10, the initial composition determination unit 101 receives the initial composition data from the user terminal 20. Then, the initial composition determination unit 101 determines the material composition of the target substance included in the initial composition data as the initial value of the composition candidate. Then, the initial composition determination unit 101 transmits the initial value of the composition candidate and the target value of the material property to the peripheral composition generation unit 102.

[0122] In step S2, the peripheral composition generation unit 102 of the material design device 10 receives the initial value of the composition candidate and the target value of the material property from the initial composition determination unit 101. Next, the peripheral composition generation unit 102 generates a plurality of peripheral compositions obtained by varying the factors included in the composition candidate of the target material within a predetermined search range. Then, the peripheral composition generation unit 102 transmits the generated plurality of peripheral compositions to the property prediction unit 103.

[0123] As an example, the search range in this embodiment is set to a range of ±5% in the initial state. The search range is expanded to a range of ±10% in the first expansion. Furthermore, the search range is expanded to a range of ±15% in the second expansion.

[0124] In step S3, the characteristic prediction unit 103 of the material design device 10 receives a plurality of surrounding components from the surrounding component generation unit 102. Next, the characteristic prediction unit 103 calculates similarities between the component candidates and each surrounding component, and deletes surrounding components whose similarities are smaller than a predetermined threshold.

[0125] The characteristic prediction unit 103 reads the learned characteristic prediction model from the model storage unit 100. Next, the characteristic prediction unit 103 predicts a plurality of material characteristics including the maintained characteristics and the improved characteristics by inputting each factor included in each peripheral composition into the learned characteristic prediction model.

[0126] In step S4, the characteristic prediction unit 103 of the material design device 10 determines whether there is a peripheral composition whose predicted value of the material characteristic is better than the target value of the material characteristic among the peripheral compositions whose material characteristics are predicted. When there is a peripheral composition whose material characteristics are improved ("Yes"), the characteristic prediction unit 103 causes the process to enter step S9. On the other hand, when there is no peripheral composition for improving the material characteristics ("No"), the characteristic prediction unit 103 causes the process to enter step S5.

[0127] In step S5, the exploration range expansion unit 104 of the material design device 10 determines whether the number of expansions of the exploration range is greater than or equal to a predetermined threshold value. When the number of expansions is greater than or equal to the predetermined threshold value ("Yes"), the exploration range expansion unit 104 causes the process to proceed to step S7. On the other hand, when the number of expansions is less than the predetermined threshold value ("No"), the exploration range expansion unit 104 causes the process to proceed to step S6.

[0128] In this embodiment, the maximum number of expansions is 2. That is, when the current search range is greater than or equal to ±15%, the search range expansion unit 104 advances the process to step S7. When the current search range is less than ±15% (±5% or ±10%), the search range expansion unit 104 advances the process to step S6.

[0129] In step S6, the search range expansion unit 104 of the material design device 10 expands the search range to ±10% if the current search range is ±5% and to ±15% if the current search range is ±10%.

[0130] In step S7, the factor exchange unit 105 of the material design device 10 reads the learned characteristic prediction model from the model storage unit 100. Next, the factor exchange unit 105 obtains the contribution of each factor from the learned characteristic prediction model. The contribution is an indicator indicating the magnitude of the influence of each factor on model learning. The factor exchange unit 105 obtains the contribution of each factor for each material characteristic predicted by the characteristic prediction unit 103.

[0131] Here, we will refer to Figure 5 The contribution degree in this embodiment will be described. Figure 5 This is a diagram showing an example of the contribution degree in this embodiment.

[0132] Figure 5 It is assumed that the target substance is used as an adhesive and that the strength is to be improved while maintaining the adhesive force based on the initial value of the composition candidate. Figure 5 (A) is a diagram showing an example of the contribution to the adhesive force. Figure 5 (B) is a diagram showing an example of contribution to strength.

[0133] like Figure 5 As shown, the contribution of each factor is a combination of the factor and the contribution, obtained for each material property. Figure 5 The example is an example of a contribution obtained from a feature prediction model installed by a random forest, but as long as an indicator indicating the influence on learning or prediction can be obtained or derived from a learned model, the indicator can be used as a contribution.

[0134] The factor exchange unit 105 exchanges the factors constituting the candidate according to the contribution of each factor. Specifically, the factor exchange unit 105 exchanges the factors with a high contribution to the improved characteristic with the factors with a low contribution to the maintained characteristic. Note that the factor exchange unit 105 exchanges the factors of the same material. For example, when the factor with a high contribution to the improved characteristic is an additive, the factor with a low contribution to the maintained characteristic is selected from the additive.

[0135] Here, we will refer to Figure 6 The factor exchange in this embodiment is described. Figure 6 This is a diagram showing an example of composition candidates before and after factor exchange in this embodiment.

[0136] Figure 6 (A) is a diagram showing an example of initial values ​​of composition candidates. Figure 6 (B) is a diagram showing an example of composition candidates after the first factor exchange. Figure 6 (C) is a diagram showing an example of composition candidates after the second factor exchange.

[0137] like Figure 6 As shown, the minimum and maximum values ​​are set for the factors after the factor exchange. The minimum and maximum values ​​represent the minimum and maximum amounts of the material substance that can be mixed in the manufacture of the target substance. The minimum and maximum values ​​are pre-set based on other experimental results using the material substance.

[0138] exist Figure 6 In the example of the first factor exchange shown in (B), it can be seen that "Additive 5" is exchanged for "Additive 6". Figure 6 Among the material substances included in the composition candidates shown in (A), "additive 5" is Figure 5 The contribution to strength shown in (B) is the highest. Figure 6 Among the additives not included in the composition candidates shown in (A), "Additive 6" is Figure 5 The contribution of (A) to adhesion is the lowest. Figure 6 In the example of factor exchange shown in (A), "Additive 5" is exchanged for "Additive 6".

[0139] Note that factor exchange is set to always be performed based on the initial value of the composition candidate. That is, even if multiple factor exchanges are performed, no further factor exchange is performed from the composition candidate after the factor exchange. Therefore, the material composition obtained as a result of the exploration is exchanged with the initial value of the composition candidate by only one factor, so that the initial value dependency can be maintained.

[0140] exist Figure 6 In the example of the second factor exchange shown in (C), it can be seen that "resin 5" is exchanged for "resin 4". This is because Figure 6 Among the material substances included in the composition candidates shown in (A), Figure 5 The contribution to strength shown in (B) is the highest for "resin 5" except for "additive 5", while Figure 6 Among the resins not included in the composition candidates shown in (A), Figure 5The lowest contribution to the adhesive force shown in (A) is "Resin 4". At this time, it can be seen that "Additive 5" and "Additive 6" exchanged in the first factor exchange have returned to the initial values.

[0141] Will return Figure 4 In step S8, the search range reduction unit 107 of the material design device 10 reduces the search range to the initial state. If the current search range exceeds ±5% (±10% or ±15%), the search range reduction unit 107 reduces the search range to ±5%.

[0142] In step S9, when there is a peripheral composition whose predicted value of material properties is better than the target value of material properties, the composition candidate updating unit 106 of the material design device 10 updates the composition candidate using the peripheral composition. When there are multiple peripheral compositions whose predicted values ​​of material properties are better than the target value of material properties, the composition candidate updating unit 106 updates the composition candidate using the peripheral composition with the best material properties among these peripheral compositions. Next, the composition candidate updating unit 106 updates the target value of material properties using the predicted value of material properties based on the new composition candidate.

[0143] In step S10, the composition candidate update unit 106 of the material design device 10 determines whether the number of updates of the composition candidates is greater than or equal to a predetermined threshold. When the number of updates is greater than or equal to the predetermined threshold ("Yes"), the composition candidate update unit 106 causes the process to proceed to step S11. On the other hand, when the number of updates is less than the predetermined threshold ("No"), the composition candidate update unit 106 returns the process to step S8.

[0144] In this embodiment, an example of the prescribed number of times is 5 times. If the composition candidates are updated too many times, the search may be performed in an area far from the initial value of the composition candidates. As a result, problems such as loss of initial value dependence and output of unrealistic material composition may occur. Therefore, in this embodiment, the number of updates of the composition candidates is limited, and the search is terminated if the prescribed number of updates is performed.

[0145] In step S11 , the result output unit 108 of the material design device 10 generates search result data including composition candidates of the target substance and predicted values ​​of material properties based on the composition candidates, and then transmits the search result data to the user terminal 20 .

[0146] In the user terminal 20, the result display unit 202 receives the search result data from the material design device 10. The result display unit 202 displays the composition candidates of the target substance and the material properties of the composition candidates on the display device 506 based on the received search result data.

[0147] Here, we will refer to Figure 7 and Figure 8 The search process of this embodiment is described. Figure 7 and Figure 8 This is a diagram showing an example of the search process.

[0148] Figure 7 The plane shown is a graph of the distribution of the material composition. In fact, it should be expressed in a space with the same dimension as the number of factors included in the material composition, but for the sake of convenience, it is shown in a two-dimensional plane using arbitrary two factors x1 and x2.

[0149] Figure 7 (A) indicates that in the first local search, the local search is performed within the search range 311 of ±5% from the initial value 301 of the self-composition candidate. Figure 7 (B) indicates that, as a result of the first local exploration, a new composition candidate 302 is explored within the exploration range 311 of ±5%.

[0150] Figure 7 (C) indicates that in the second local exploration, the local exploration is performed within the exploration range 312 of ±5% from the new composition candidate 302 . Figure 7 (D) indicates that, as a result of the second local exploration, a new composition candidate 303 is explored within the exploration range 312 of ±5%.

[0151] Figure 8 (A) indicates that in the x-th local exploration, the component candidates are not updated within the exploration range 331 of ±5% from the component candidate 321, and the local exploration is performed within the exploration range 332 of ±10%. Figure 8 (B) indicates that, as a result of the x-th local exploration, a new composition candidate 322 is explored within the exploration range 332 of ±10%.

[0152] Figure 8 (C) indicates that in the xth local exploration, the composition candidate is not updated within the exploration range 332 of ±10% from the composition candidate 321, and the local exploration is performed within the exploration range 333 of ±15%, and as a result, a new composition candidate 323 is explored.

[0153] Figure 8 (D) indicates that in the xth local exploration, the constituent candidates are not updated within the exploration range 333 of ±15% from the constituent candidate 321, and the factor x1 is replaced by the factor x3. At this time, the exploration range is narrowed to the exploration range 334 of ±5% from the constituent candidate 321, and the local exploration is performed within the exploration range 334.

[0154] <Example>

[0155] One embodiment of the present disclosure is an example of searching for a material composition capable of maintaining adhesive force and obtaining a strength that reaches a target value in the material design of an adhesive by using the material design device 10 of the present embodiment. Figures 9 to 11 This is a diagram showing an example of the results of searching for material composition in this example.

[0156] In this embodiment, the material composition that obtains an adhesive force of 4.4 and a strength of 52 is selected as the initial value of the composition candidate. In contrast, the target value of the adhesive force is set to 4.6, and the target value of the strength is set to 64. Fig. 9 and Fig.10 The figures shown show the coordinates and material properties of the surrounding components generated by one local exploration in the exploration space. Fig. 9 and Fig.10 In each of the figures shown, the left vertical and horizontal axes represent material composition, and the right vertical axis represents a color bar, which represents the strength of the property to be improved by color. In each figure, the initial value is represented by a white hollow asterisk, and the optimal value is represented by a black solid asterisk.

[0157] Fig. 9 (A) is the result of the first local exploration. In the first local exploration, the peripheral composition was generated within the range of ±5%. The optimal values ​​were 4.42 for adhesion and 58.784 for strength. Since the material properties did not meet the target values, the composition candidates were not updated.

[0158] Fig. 9 (B) is the result of the second local exploration. In the second local exploration, the peripheral composition was generated within the range of ±10%. The optimal values ​​were 4.41 for adhesion and 61.683 for strength. Since the material properties did not meet the target values, the composition candidates were not updated.

[0159] Fig. 9 (C) is the result of the third local exploration. In the third local exploration, peripheral compositions were generated within the range of ±15%. The optimal values ​​were 4.22 for adhesion and 61.734 for strength. Since the material properties did not meet the target values, the composition candidates were not updated.

[0160] Fig. 9 (D) is the result of the first local exploration after the factor exchange. In the first local exploration after the factor exchange, the peripheral composition was generated within the range of ±5%. The optimal values ​​were 4.45 for adhesion and 58.933 for strength. Since the material properties did not meet the target values, the composition candidates were not updated.

[0161] Fig. 9(E) is the result of the second local exploration after the factor exchange. In the second local exploration after the factor exchange, the peripheral composition was generated within the range of ±10%. The optimal values ​​were 4.38 for adhesion and 62.191 for strength. Since the material properties did not meet the target values, the composition candidates were not updated.

[0162] Fig. 9 (F) is the result of the third local exploration after the factor exchange. In the third local exploration after the factor exchange, the peripheral composition was generated within the range of ±15%. The optimal values ​​were 4.53 for adhesion and 64.024 for strength. Since the material properties met the target values, the composition candidates and target values ​​were updated using the material composition that obtained the optimal value.

[0163] Fig.10 (A) is the result of the first local exploration after the first update. In the first local exploration after the first update, the peripheral composition was generated within the range of ±5%. The optimal values ​​were adhesion 4.55 and strength 65.430. Since the material properties met the target values, the second update was performed using the material composition that obtained the optimal value.

[0164] Fig.10 (B) is the result of the first local exploration after the second update. In the first local exploration after the second update, the peripheral composition was generated within the range of ±5%. The optimal values ​​were 4.56 for adhesion and 65.437 for strength. Since the material properties met the target values, the third update was performed using the material composition that obtained the optimal value.

[0165] Fig.10 (C) is the result of the first local exploration after the third update. In the first local exploration after the third update, the peripheral composition was generated within the range of ±5%. The optimal values ​​were 4.59 for adhesion and 65.549 for strength. Since the material properties met the target values, the fourth update was performed using the material composition that obtained the optimal value.

[0166] Fig.11 This is an example showing the results of searching for material composition in this embodiment. Fig.11 (A) is a summary of Fig. 9 (A) to Fig.10 (C) shows a graph of the exploration results from the first local exploration (hereinafter referred to as the "initial exploration") to the first local exploration after the third update (hereinafter referred to as the "final exploration"). Fig.11 The graph shown in (A) shows the initial value in the first exploration, the optimal value in the third exploration after factor exchange, and the optimal value in the final exploration. Fig.11As shown in (A), it can be seen that in the search for obtaining a material composition that reaches the target value, the search was conducted in a relatively close area.

[0167] Fig.11 (B) is a graph showing the material properties of the peripheral composition in the initial exploration and the material properties of the peripheral composition in the final exploration, with the horizontal axis representing the adhesive force and the vertical axis representing the strength. Fig.11 As shown in (B), it can be seen that in the final exploration, exploration was conducted in an area where the adhesion and strength were higher than those in the initial exploration.

[0168] The above results show that the material design device 10 according to the present embodiment can efficiently search for a material composition that can obtain better material properties.

[0169] <Effects of Implementation Methods>

[0170] The design support system of this embodiment updates the composition candidate using the peripheral composition when there is a peripheral composition that improves the material properties in the peripheral composition obtained by changing each factor included in the composition candidate of the target substance within a prescribed exploration range, and expands the exploration range when there is no peripheral composition that improves the material properties. At this time, when the exploration range exceeds the prescribed range, the factors of the composition candidate are exchanged. Therefore, according to the design support system of this embodiment, it is possible to perform local exploration that maintains the initial value dependency and is not easily trapped in a local solution.

[0171] In particular, in the design support system of the present embodiment, in the local exploration, the composition candidates are updated using the surrounding composition that maintains the material characteristics to be maintained and improves the material characteristics to be improved, and in the factor exchange, the factor with a high contribution to the material characteristics to be improved is exchanged for a factor with a low contribution to the material characteristics to be maintained. Therefore, according to the design support system of the present embodiment, it is possible to explore a material composition that takes into account a plurality of material characteristics in a trade-off relationship and improves the material characteristics at the same time.

[0172] In addition, the design support system of the present embodiment ends the exploration after updating the composition candidate a predetermined number of times, and outputs the composition candidate as the exploration result. In addition, the design support system of the present embodiment does not explore the surrounding composition whose similarity with the composition candidate is less than a predetermined threshold. Therefore, according to the design support system of the present embodiment, the composition candidate is not explored in the area far from the initial value of the composition candidate, so that the initial value dependency can be maintained.

[0173] [Replenish]

[0174] Each function of the above-described embodiments can be implemented by one or more processing circuits. Here, the "processing circuit" in this specification includes a processor programmed to perform each function by software, such as a processor implemented by an electronic circuit, an ASIC (Application Specific Integrated Circuit) designed to perform the above functions, a DSP (Digital Signal Processor), an FPGA (Field Programmable Gate Array), a conventional circuit module, and other devices.

[0175] As mentioned above, although embodiment of this invention was demonstrated in detail, this invention is not limited to these embodiment, Various deformation|transformation or change is possible within the range of the summary of this invention described in a claim.

[0176] This application claims the priority of Japanese Invention Patent Application No. 2022-130049 filed with the Japan Patent Office on August 17, 2022, the entire contents of which are incorporated herein by reference.

[0177] Description of Reference Numerals

[0178] 1 Design support system

[0179] 10 Material Design Device

[0180] 101 Initial Composition Decision Department

[0181] 102 Peripheral component generation unit

[0182] 103 Characteristics Prediction Department

[0183] 104 Exploration Range Expansion Department

[0184] 105 Factor Exchange Department

[0185] 106 Formation Candidate Update Department

[0186] 107 Exploration scope reduction

[0187] 108 Result output unit

[0188] 20 User Terminal

[0189] 201 Initial composition input unit

[0190] 202 result display unit.

Claims

1. A material design device, comprising: a peripheral composition generating unit configured to generate a peripheral composition obtained by varying each factor included in the composition candidate of the target substance within a predetermined search range; a property prediction unit configured to predict a material property of the target substance based on the surrounding composition; a composition candidate updating unit configured to update the composition candidate using the surrounding composition when there is the surrounding composition that improves the material property; a search range expansion unit configured to expand the search range when the peripheral composition that improves the material property does not exist; as well as The factor exchanging unit is configured to exchange the factors of the constituent candidates when the search range exceeds a predetermined maximum search range.

2. The material design device according to claim 1, wherein: The property prediction unit is configured to predict the material property using a prediction model that uses the material composition of the target substance as an explanatory variable and the material property as a target variable, The factor exchanging unit is configured to exchange the factors constituting the candidate based on the contribution of each factor in the prediction model.

3. The material design device according to claim 2, wherein: The characteristic prediction unit is configured to predict a plurality of the material characteristics, The composition candidate updating unit is configured to update the composition candidate when the first material characteristic is maintained and the second material characteristic is improved.

4. The material design device according to claim 3, wherein: The factor exchange unit is configured to exchange a factor having a high contribution to the second material characteristic with a factor having a low contribution to the first material characteristic.

5. The material design device according to claim 4, further comprising: The initial composition determination unit is configured to determine, as the composition candidate, the material composition having the better first material characteristic among the material compositions generated randomly or with a predetermined step size from the entire material composition of the target substance.

6. The material design device according to claim 4 or 5, wherein: The factor exchange unit is configured to exchange factors of the same type.

7. The material design device according to any one of claims 1 to 6, wherein: The composition candidate updating unit is configured to output the composition candidate when the composition candidate is updated a predetermined number of times.

8. The material design device according to any one of claims 1 to 7, wherein: The property prediction unit is configured to predict the material property for the surrounding composition having a similarity with the composition candidate that is greater than or equal to a predetermined threshold.

9. A material design method, comprising the steps of: generating a peripheral composition obtained by varying each factor included in the composition candidate of the target substance within a predetermined search range; predicting material properties of the target substance based on the surrounding composition; When the peripheral composition that improves the material property exists, updating the composition candidate using the peripheral composition; When the surrounding composition that improves the material properties does not exist, expanding the exploration range; as well as When the search range exceeds a prescribed maximum search range, the factors constituting the candidates are exchanged.

10. A program for causing a computer to execute the following steps: generating a peripheral composition obtained by varying each factor included in the composition candidate of the target substance within a predetermined search range; predicting material properties of the target substance based on the surrounding composition; When the peripheral composition that improves the material property exists, updating the composition candidate using the peripheral composition; When the surrounding composition that improves the material properties does not exist, expanding the exploration range; as well as When the search range exceeds a prescribed maximum search range, the factors constituting the candidates are exchanged.

Images