Method for estimating the properties of composite materials

Abstract

To estimate the properties of composite materials composed of multiple materials in a simple and highly accurate manner. [Solution] A method for estimating the target properties of a composite material composed of multiple materials includes a first step, a second step, a third step, and a fourth step. In the first step, at least two materials are selected as multiple target materials from a group of materials containing the multiple materials as components in the correct proportions. In the second step, multiple material data for the multiple target materials and multiple mixing ratios for the multiple target materials are obtained. In the third step, a first estimation model calculates mixing parameters that express the closeness between the multiple material data from first information including the multiple material data. In the fourth step, a second estimation model calculates estimated values ​​of the target properties of the composite material composed of the multiple target materials at the multiple mixing ratios from second information including the multiple material data, the multiple mixing ratios, and the mixing parameters.

Description

Technical Field

[0001] The present invention relates to a method for estimating the properties of composite materials.

Background Art

[0002] In the conventional development process of new materials, it is premised on the trial and error of developers, inevitably requiring a huge amount of time and cost, and there are major problems in terms of efficiency. In contrast, in recent years, a process using machine learning has attracted attention as a development process for new materials. In the process using machine learning, by using a machine learning model that has learned trends and patterns from past data, estimation and classification can be performed on new data, so the examination range of new materials can be narrowed in advance. Therefore, in the process using machine learning, it becomes possible to efficiently reach a new material having the characteristics required by the developer.

[0003] As an example, Patent Documents 1 to 3 disclose a method for estimating the properties of a composite material by inputting the properties and blending ratios of each material constituting the composite material, and in some cases, parameters related to manufacturing conditions into a machine learning model.

[0004] Specifically, Patent Document 1 discloses a technique for preparing experimental condition information representing experimental conditions in which structural information representing the structure of each material constituting a composite material is arranged, and using experimental information in which the abundance ratio of the structural information of each material constituting the composite material and the mixing ratio of each material in the composite material match to estimate the performance value of the composite material.

[0005] Patent Document 2 discloses a technique for obtaining predicted values of a plurality of physical property types of a polymer by a machine learning model for each physical property type based on monomer information indicating a plurality of monomers constituting the polymer and their content ratios, calculating a deviation value with respect to the monomer information from the predicted values of the plurality of physical property types, and determining the monomers used in the synthesis of the polymer and their content ratios based on the deviation value.

[0006] Patent Document 3 discloses a technique for estimating the value of a physical quantity (dependent variable) for a composite material by using a machine learning model and using not only the composite characteristic value calculated based on the characteristic value and mixing ratio of each material constituting the composite material, but also the mixing information of each material constituting the composite material as input parameters (explanatory variables) for the approximation function. [Prior art documents] [Patent Documents]

[0007] [Patent Document 1] Re-tabled publication No. 2019-198644 [Patent Document 2] Japanese Patent Publication No. 2021-196710 [Patent Document 3] Japanese Patent Publication No. 2024-035516 [Overview of the project] [Problems that the invention aims to solve]

[0008] However, the technology described in Patent Document 1 requires the prior preparation of experimental condition information, including the relative abundance of structural information for each material constituting the composite material, making it impossible to efficiently search for new composite materials.

[0009] In the technology described in Patent Document 2, when there are few types of monomers constituting the polymer, the information of each monomer is easily reflected in the model. However, as the number of types of monomers constituting the polymer increases and the molar ratio of each monomer decreases, the information of each monomer becomes less easily reflected in the model. In order to construct a model that accurately reflects the information of each monomer even when there are many types of monomers constituting the polymer, it is necessary to prepare a vast amount of data and train the model with this data.

[0010] The technology described in Patent Document 3 allows for estimation that reflects each characteristic value when the number of constituent materials in the composite material is small, but as the number of constituent materials in the composite material increases, estimation that reflects each characteristic value becomes more difficult. In order to enable highly accurate estimation even when the number of constituent materials in the composite material is large, it is necessary to prepare a large amount of data and train a model with this data.

[0011] The objective of this invention is to easily and accurately estimate the properties of composite materials composed of multiple materials. [Means for solving the problem]

[0012] A method for estimating the desired properties of a composite material composed of multiple materials according to one embodiment of the present invention includes a first step, a second step, a third step, and a fourth step. In the first step described above, at least two materials are selected as multiple target materials from a group of materials that contain the above-mentioned multiple materials in the correct proportions. In the second step described above, multiple material data for the multiple target materials and multiple blending ratios for the multiple target materials are obtained. In the third step described above, a mixing parameter is calculated from the first information, which includes the multiple material data, using the first estimation model described below, which represents the closeness between the multiple material data. In the fourth step described above, the estimated value of the target properties of the composite material composed of the multiple target materials with the multiple mixing ratios is calculated using the second estimation model described below, based on the second information including the multiple material data, the multiple mixing ratios, and the mixing parameters. First estimation model: A machine learning model trained to produce the above mixed parameter output in response to the above first information input. Second estimation model: A machine learning model trained to output the above estimated value in response to the input of the second piece of information.

[0013] An information processing device according to one embodiment of the present invention estimates the target properties of a composite material composed of multiple materials. The above-mentioned information processing device comprises a processing unit that executes a first step, a second step, a third step, and a fourth step. In the first step described above, at least two materials are selected as multiple target materials from a group of materials that contain the above-mentioned multiple materials in the correct proportions. In the second step described above, multiple material data for the multiple target materials and multiple blending ratios for the multiple target materials are obtained. In the third step described above, a mixing parameter is calculated from the first information, which includes the multiple material data, using the first estimation model described below, which represents the closeness between the multiple material data. In the fourth step described above, the estimated value of the target properties of the composite material composed of the multiple target materials with the multiple mixing ratios is calculated using the second estimation model described below, based on the second information including the multiple material data, the multiple mixing ratios, and the mixing parameters. First estimation model: A machine learning model trained to produce the above mixed parameter output in response to the above first information input. Second estimation model: A machine learning model trained to output the above estimated value in response to the input of the second piece of information.

[0014] A program according to one embodiment of the present invention is configured to estimate the desired properties of a composite material composed of multiple materials. The above program causes the information processing device to execute the first, second, third, and fourth steps. In the first step described above, at least two materials are selected as multiple target materials from a group of materials that contain the above-mentioned multiple materials in the correct proportions. In the second step described above, multiple material data for the multiple target materials and multiple blending ratios for the multiple target materials are obtained. In the third step described above, a mixing parameter is calculated from the first information, which includes the multiple material data, using the first estimation model described below, which represents the closeness between the multiple material data. In the fourth step, an estimated value of a target property of a composite material composed of the plurality of target materials with the plurality of blending ratios is calculated from second information including the plurality of material data, the plurality of blending ratios, and the mixing parameters by the following second estimation model. First Estimation Model: A machine learning model trained to output the mixing parameters for an input of the first information. Second Estimation Model: A machine learning model trained to output the estimated value for an input of the second information.

[0015] A recording medium according to an aspect of the present invention stores a program that causes an information processing apparatus to execute a first step, a second step, a third step, and a fourth step. In the first step, at least two materials included in a material group that contains the plurality of materials without excess or deficiency as components are selected as a plurality of target materials. In the second step, a plurality of material data of the plurality of target materials and a plurality of blending ratios of the plurality of target materials are acquired. In the third step, a mixing parameter representing proximity between the plurality of material data is calculated from first information including the plurality of material data by the following first estimation model. In the fourth step, an estimated value of a target property of a composite material composed of the plurality of target materials with the plurality of blending ratios is calculated from second information including the plurality of material data, the plurality of blending ratios, and the mixing parameters by the following second estimation model. First Estimation Model: A machine learning model trained to output the mixing parameters for an input of the first information. Second Estimation Model: A machine learning model trained to output the estimated value for an input of the second information.

Advantageous Effects of the Invention

[0016] According to the present invention, the properties of a composite material composed of a plurality of materials can be estimated simply and with high accuracy.

Brief Description of the Drawings

[0017] [Figure 1] This flowchart shows a method for estimating the properties of a composite material according to one embodiment of the present invention. [Figure 2] This is a diagram showing the group of materials that make up a composite material. [Figure 3] This diagram schematically shows an example of the first processing flow. [Figure 4] This is a diagram showing a group of materials, including intermediate composite materials. [Figure 5] This figure shows the material group at the end of the estimation method described above. [Figure 6] This diagram schematically shows one example of the processing flow for the estimation method described above. [Figure 7] This diagram schematically shows one example of the processing flow for the estimation method described above. [Figure 8] This diagram schematically shows one example of the processing flow for the estimation method described above. [Figure 9] This is a block diagram showing an information processing device capable of implementing the above estimation method. [Modes for carrying out the invention]

[0018] The present invention is a method for estimating the desired properties of a composite material composed of multiple materials, The first step involves selecting at least two materials from a group of materials containing the above-mentioned multiple materials in the correct proportions as components, as multiple target materials. A second step involves obtaining multiple material data for the multiple target materials mentioned above, and multiple blending ratios for the multiple target materials mentioned above. A third step involves calculating a mixing parameter that expresses the closeness between the multiple material data sets using the following first estimation model, based on the first information including the multiple material data sets described above. A fourth step involves calculating an estimated value of the target properties of a composite material composed of the above-mentioned multiple target materials at the above-mentioned multiple mixing ratios, using the second information including the above-mentioned multiple material data, above-mentioned multiple mixing ratios, and above-mentioned mixing parameters, using the second estimation model described below. This is a method for estimating the desired properties of composite materials that include [specific material]. First estimation model: A machine learning model trained to produce the above mixed parameter output in response to the above first information input. Second estimation model: A machine learning model trained to output the above estimated value in response to the input of the second piece of information.

[0019] An estimation method according to one embodiment of the present invention is configured to efficiently estimate the properties of a composite material composed of multiple materials in specific mixing ratios by calculation using an information processing device. In the estimation method according to this embodiment, the properties of a new composite material can be estimated solely by calculation, without experimental evaluation. Therefore, according to the estimation method according to this embodiment, it is possible to search for new composite materials with excellent properties with little effort and cost.

[0020] The estimation method according to this embodiment estimates at least one property of a composite material using the material data of each of the multiple materials constituting the composite material and the mixing ratio of the multiple materials.

[0021] In the estimation method according to this embodiment, "target characteristic" refers to at least one physical property of the composite material that is estimated in the estimation method according to this embodiment.

[0022] In this embodiment, "material data" refers to data prepared for each material constituting the composite material, consisting of at least one known property of each material constituting the composite material. The material data includes at least the property to be estimated by the estimation method according to this embodiment for the composite material, i.e., the target property, and may also include any other type of property. The material data is preferably expressed as one of the following: a real number, a categorical value, or a probability density function. The mixing ratio is typically expressed as a mass ratio (mass%). The estimation method according to this embodiment calculates estimated values ​​for all properties included in the material data for the composite material and obtains the target property to be estimated for the composite material as part or all of the estimated values.

[0023] In the estimation method according to this embodiment, it is possible to directly estimate the target properties of a composite material using the material data and mixing ratios of all the materials constituting the composite material. However, it is also possible to estimate the material data of an intermediate composite material composed of only some of the materials using the material data and mixing ratios of only some of the materials constituting the composite material. In this case, the estimation method according to this embodiment can ultimately estimate the target properties of a composite material composed of all the materials by repeatedly estimating the material data of the intermediate composite material.

[0024] [Method for estimating the properties of composite materials] (Brief explanation) The following describes the method for estimating the target properties of a composite material according to this embodiment. In the estimation method according to this embodiment, the composite material for which the target properties are to be estimated is referred to as "composite material α". The method for estimating the target properties of composite material α according to this embodiment includes the first step S01, the second step S02, the third step S03, the fourth step S04, and the fifth step S05 shown in Figure 1.

[0025] In the following, we will explain a method for estimating the target properties of composite material α according to this embodiment, using a specific example where composite material α is a mixed oil composed of materials a, b, c, d, e, and f in specific blending ratios, and the oil-water interface tension, which is one of the physical properties of composite material α, is used as the target property. Figure 2 shows a group of materials Gr consisting of six oil-soluble materials a to f of different types. All of the materials a to f constituting the group of materials Gr are existing materials, and their material data, including oil-water interface tension, is known.

[0026] The material data for materials a to f should include at least the oil-water interface tension for composite material α, which is the target of the estimation of the desired properties according to the estimation method of this embodiment, but may also include properties other than oil-water interface tension. Examples of properties other than oil-water interface tension that can be included in the material data include Hansen solubility parameters (δD, δP, δH, δTotal), LogP, specific gravity, refractive index, and molecular fingerprint.

[0027] In the estimation method according to this embodiment, the target properties of a composite material α of one composition are estimated by a series of processing flows P consisting of a first step S01, a second step S02, a third step S03, and a fourth step S04. Therefore, one processing flow P can estimate the physical properties of one intermediate composite material composed of some of the materials a to f that make up the composite material α. In the specific example shown below, the material data of the intermediate composite material is estimated by repeating the estimation of material data of the intermediate composite material through multiple processing flows P, thereby ultimately estimating the target properties of the composite material α composed of all of materials a to f.

[0028] (Step 1 S01: Select multiple target materials from the material group) In the first step S01, multiple target materials are selected from the material group Gr. In the estimation method according to this embodiment, in the third step S03 and the fourth step S04, processing is performed to estimate the properties of a composite material (intermediate composite material or composite material α) composed of the target materials selected in the first step S01. In the first step S01 of the first processing flow P1 in the specific example, two or more materials are selected as target materials from materials a to f that constitute the material group Gr.

[0029] The number of target materials selected from the material group Gr in the first step S01 may be two or more, but increasing the number increases the computational load in the third step S03 and the fourth step S04, so it is preferable to have two or three, and more preferable to have two. In the specific example, the number of target materials selected in the first step S01 of all processing flows P is set to two. Therefore, in the first processing flow P1 in the specific example, two materials a and b are selected as target materials, and the properties of the intermediate composite material A composed of target materials a and b are estimated, as shown in Figure 3.

[0030] (Step 2 S02: Obtain material data and mixing ratios) In the second step S02, the material data and mixing ratios of the target materials selected in the first step S01 are obtained. The material data of the target materials can be obtained from existing databases, etc. The mixing ratio of the target materials can be obtained as the mixing ratio of only the target materials calculated from the material mixing ratios of composite material α. In the first processing flow P1 in the specific example, the material data and mixing ratios of target materials a and b are obtained.

[0031] (Step 3 S03: Calculate mixing parameters) In the third step S03, a mixing parameter is calculated by the first estimation model from the first information, which includes the material data of each target material obtained in the second step S02. The mixing parameter is calculated once for all combinations of target materials and represents the closeness of the material data between the target materials. In the first processing flow P1 in the specific example, the mixing parameter is calculated from the first information, which includes the material data of target materials a and b. It is preferable to express the mixing parameter as a non-negative real number.

[0032] The parameters calculated in the third step S03 are indicators that represent the closeness between material data, and essentially do not require the blending ratio. For this reason, it is preferable that the mixing parameters calculated in the third step S03 are parameters that are determined without considering the blending ratio of the target materials. In other words, the first information may include data other than the material data of the target materials, but it is preferable that it does not include the blending ratio of the target materials.

[0033] The first estimation model is a machine learning model trained to output mixed parameters in response to the input of first information. In other words, the first estimation model is a model that has learned the trends and patterns in the relationship between first information and experimentally obtained mixed parameters for various different combinations of materials.

[0034] In the present invention, it is preferable that the first estimation model does not use the blending ratio of the target materials to calculate the mixing parameters.

[0035] For the first estimation model, regression models are preferable as the learning method. Suitable examples of regression models include multiple regression analysis, LASSO regression, Ridge regression, support vector machine regression, decision trees, random forest regression, Bayesian linear regression, neural network models, boosting ensembles, and ensemble models composed of one or more of these models.

[0036] (Step 4, S04: Calculate estimated values ​​for the target properties of the composite material) In the fourth step S04, an estimated value of the target properties of the composite material composed of the target materials at the specified mixing ratio is calculated by a second estimation model from second information, which includes the material data and mixing ratio of each target material obtained in the second step S02, and the mixing parameters calculated in the third step S03. The second information may include data other than the material data and mixing ratio of the target materials and the mixing parameters. In the fourth step S04, an estimated value of the target properties of the composite material is calculated from second information, which includes the material data and mixing ratio of the target materials and the mixing parameters. In the first processing flow P1 in the specific example, the estimated value X of the target properties of intermediate composite material A is calculated from second information, which includes the material data and mixing ratio of target materials a and b and the mixing parameters. A Calculate.

[0037] The second estimation model is a machine learning model trained to output estimated material data for composite materials in response to the input of second information. In other words, the second estimation model is a model that has learned the trends and patterns in the relationship between second information for various different combinations of target materials and experimentally obtained material data for composite materials. A regression model is preferable as the training method for the second estimation model, and examples of models that can be used include multiple regression analysis, LASSO regression, Ridge regression, support vector machine regression, decision trees, random forest regression, Bayesian linear regression, neural network models, boosting ensembles, and ensemble models composed of one or more of these models.

[0038] In this embodiment, it is preferable that the fourth step S04 includes a step to update the material group Gr after calculating the estimated value of the target characteristic. That is, this step involves deleting the target material selected in the first step S01 from the material group Gr and adding the material (intermediate composite material or composite material α) for which the estimated value of material data has been calculated to the material group Gr. In the first processing flow P1 in the specific example, as shown in Figure 4, target materials a and b are deleted from the material group Gr and intermediate composite material A is added to the material group Gr. At this time, the material data of intermediate composite material A is the estimated value X of the target characteristic mentioned above. A Thus, by updating the material group Gr for each processing flow P, the material group Gr always contains all the components constituting the composite material α in the correct proportions, either as multiple materials or as intermediate composite materials.

[0039] (5th step S05: Judgment) In the method for estimating the target properties of a composite material according to this embodiment, it is preferable to have a fifth step S05 in which it is determined whether the material group Gr includes two or more materials, including an intermediate composite material and a material not selected as the target material. Specifically, in the fifth step S05, if the estimated material data of composite material α has already been calculated in the fourth step S04 of the processing flow P and there is a step to update the material group Gr, then as shown in Figure 5, the material group Gr will only contain composite material α. However, in the fifth step S05, it is determined whether the material group Gr contains composite material α or not.

[0040] In the fifth step S05, if it is determined that the material group Gr contains two or more materials, it is preferable to have a processing flow that repeats the first step S01. In the first processing flow P1 in the specific example, as shown in Figure 4, the material group Gr contains intermediate composite material A, material c, material d, and material e, so the process proceeds to the first step S01 of the second processing flow P2. On the other hand, if the material group Gr does not contain two or more materials, that is, if only composite material α remains, the estimated target properties have been obtained, and the method for estimating the target properties of the composite material according to this embodiment is terminated.

[0041] (Processing flow P for the second and subsequent times) In the method for estimating the properties of composite material α, there are various patterns of combinations of target materials that can be selected in the first step S01 of the second and subsequent processing flows P. If the target materials include an intermediate composite material, the second and subsequent processing flows P can be executed in the same way as the first processing flow P by using the estimated material data of the intermediate composite material calculated in the fourth step S04 of the previous processing flow P as the material data of the intermediate composite material.

[0042] In other words, in this embodiment, it is preferable to include an operation in which an intermediate composite material, which is a composite material whose target properties have already been calculated, is selected as one of a plurality of target materials in the first step.

[0043] Specifically, in this embodiment, it is preferable to include an operation to select at least two materials as multiple target materials in the first step, which are included in a group of materials consisting of at least one intermediate composite material whose target properties have already been calculated, and a material that has not been selected as one of the multiple target materials from among the multiple materials.

[0044] When selecting at least one intermediate composite material, which is a composite material for which the desired properties have already been calculated, as one of the multiple target materials in the first step, it is preferable to obtain the material data of at least one intermediate composite material as an estimated value of the material data calculated by the second estimation model in the second step, and to obtain the desired properties of a composite material composed of multiple materials by repeating a series of processing flows including the first, second, third, and fourth steps until the material group no longer contains two or more materials.

[0045] Figures 6 and 7 are schematic diagrams illustrating multiple processing flows P, including the first processing flow P1 described above. In the examples shown in Figures 6 and 7, two target materials are selected from the material group Gr in the first step S01 of all processing flows P. Note that in Figures 6 and 7, each intermediate composite material A, B, C, and D is independent; that is, intermediate composite materials with the same name do not necessarily have the same configuration.

[0046] Let's explain the example shown in Figure 6. In the second processing flow P2, intermediate composite materials A and c are selected as target materials, and estimated material data for intermediate composite material B is calculated. In the third processing flow P3, intermediate composite materials B and d are selected as target materials, and estimated material data for intermediate composite material C is calculated. In the fourth processing flow P4, intermediate composite materials C and e are selected as target materials, and estimated material data for intermediate composite material D is calculated. In the fifth processing flow P5, intermediate composite materials D and f are selected as target materials, and the target properties of composite material α are calculated.

[0047] Let's explain the example shown in Figure 7. In the second processing flow P2, materials c and d are selected as target materials, and estimated material data values ​​for intermediate composite material B are calculated. In the third processing flow P3, materials e and f are selected as target materials, and estimated material data values ​​for intermediate composite material C are calculated. In the fourth processing flow P4, intermediate composite materials A and B are selected as target materials, and estimated material data values ​​for intermediate composite material D are calculated. In the fifth processing flow P5, intermediate composite materials C and D are selected as target materials, and the target properties of composite material α are calculated.

[0048] In this embodiment, if N is the number of materials constituting the composite material for which the target properties are to be estimated, it is preferable to repeat the processing flow P N-1 times. If N is the number of target materials constituting the composite material for which the target properties are to be estimated, repeating the processing flow P N-1 times reduces the computational load and allows for efficient acquisition of estimated values ​​of the target properties of the composite material. Note that the examples shown in Figures 6 and 7 are schematic diagrams of an example in which the processing flow P is repeated 5 times when the number of target materials constituting the composite material for which the target properties are to be estimated is 6.

[0049] (Other examples) In the example above, the number of target materials selected in the first step S01 of each processing flow P was set to 2. However, multiple processing flows P may include processing flows P in which the number of target materials selected in the first step S01 is 3 or more. Figure 8 is a schematic diagram illustrating multiple processing flows P, including a processing flow P in which the number of target materials selected in the first step S01 is 3.

[0050] Let's explain the example shown in Figure 8. In the first processing flow P1, materials a, b, and c are selected as target materials, and estimated material data values ​​for intermediate composite material A are calculated. In the second processing flow P2, intermediate composite materials A, d, and e are selected as target materials, and estimated material data values ​​for intermediate composite material B are calculated. In the third processing flow P3, intermediate composite materials B and f are selected as target materials, and estimated values ​​of the target properties of composite material α are calculated.

[0051] [Information Processing Device] The information processing device according to this embodiment is an information processing device that estimates the target properties of a composite material composed of multiple materials. The information processing device comprises a processing unit that executes a first step, a second step, a third step, and a fourth step. In the first step described above, at least two materials are selected as multiple target materials from a group of materials that contain the above-mentioned multiple materials in the correct proportions. In the second step described above, multiple material data for the multiple target materials and multiple blending ratios for the multiple target materials are obtained. In the third step described above, a mixing parameter is calculated from the first information, which includes the multiple material data, using the first estimation model described below, which represents the closeness between the multiple material data. In the fourth step described above, the estimated value of the target properties of the composite material composed of the multiple target materials with the multiple mixing ratios is calculated using the second estimation model described below, based on the second information including the multiple material data, the multiple mixing ratios, and the mixing parameters. First estimation model: A machine learning model trained to produce the above mixed parameter output in response to the above first information input. Second estimation model: A machine learning model trained to output the above estimated value in response to the input of the second piece of information.

[0052] The method for estimating the target properties of a composite material according to the above embodiment can be implemented, for example, using the information processing device 100 shown in Figure 9.

[0053] The information processing device 100 according to this embodiment can be configured with various types of computers, and may be configured with a single computer or with a combination of two or more computers. The information processing device 100 includes an input unit 101, an output unit 102, a processing unit 103, and a storage unit 104.

[0054] The input unit 101 is configured as a user interface and is used, for example, to input the target material selected in the first step S01. The output unit 102 is configured to output information to a display device such as a display, and can display, for example, the estimated value of the composite material data calculated in the fourth step S04. The processing unit 103 is configured to execute all or part of the processing included in the composite material property estimation method according to the above embodiment, according to a program. The storage unit 104 is configured to store various types of data and stores, for example, a program to cause the processing unit 103 to execute each process, and various types of data necessary for each process performed by the processing unit 103.

[0055] Furthermore, the information processing device 100 may have configurations other than those described above. For example, it may include a communication unit to obtain data from a database on the cloud via communication. Also, the information processing device 100 only needs to include a processing unit 103, and may not include at least one of the other configurations.

[0056] [program] The program according to this embodiment is configured to estimate the desired properties of a composite material composed of multiple materials. The above program causes the information processing device to execute the first, second, third, and fourth steps. In the first step described above, at least two materials are selected as multiple target materials from a group of materials that contain the above-mentioned multiple materials in the correct proportions. In the second step described above, multiple material data for the multiple target materials and multiple blending ratios for the multiple target materials are obtained. In the third step described above, a mixing parameter is calculated from the first information, which includes the multiple material data, using the first estimation model described below, which represents the closeness between the multiple material data. In the fourth step described above, the estimated value of the target properties of the composite material composed of the multiple target materials with the multiple mixing ratios is calculated using the second estimation model described below, based on the second information including the multiple material data, the multiple mixing ratios, and the mixing parameters. First estimation model: A machine learning model trained to produce the above mixed parameter output in response to the above first information input. Second estimation model: A machine learning model trained to output the above estimated value in response to the input of the second piece of information.

[0057] [Recording medium] The recording medium according to this embodiment is a recording medium on which the program according to the above embodiment is stored.

[0058] In other words, the recording medium according to this embodiment stores a program that causes an information processing device to execute a first step, a second step, a third step, and a fourth step. In the first step described above, at least two materials are selected as multiple target materials from a group of materials that contain the above-mentioned multiple materials in the correct proportions. In the second step described above, multiple material data for the multiple target materials and multiple blending ratios for the multiple target materials are obtained. In the third step described above, a mixing parameter is calculated from the first information, which includes the multiple material data, using the first estimation model described below, which represents the closeness between the multiple material data. In the fourth step described above, the estimated value of the target properties of the composite material composed of the multiple target materials with the multiple mixing ratios is calculated using the second estimation model described below, based on the second information including the multiple material data, the multiple mixing ratios, and the mixing parameters. First estimation model: A machine learning model trained to produce the above mixed parameter output in response to the above first information input. Second estimation model: A machine learning model trained to output the above estimated value in response to the input of the second piece of information.

[0059] Specific examples of recording media storing the program of the present invention include magnetic disks, magneto-optical disks, magnetic tapes, flash memory, and cloud storage on the internet.

[0060] [Other embodiments] Although embodiments of the present invention have been described above, the present invention is not limited to the embodiments described above, and various modifications can be made without departing from the spirit of the present invention.

[0061] For example, the properties estimated by the estimation method according to this embodiment, i.e., the target properties, are not limited to oil-water interface tension and can be arbitrarily determined. Suitable examples of target properties to be estimated in this embodiment include one or more selected from oil-water interface tension, surface tension, critical micelle concentration, solubility, emulsifying power, foaming properties, contact angle, viscosity, dielectric constant, conductivity, boiling point, freezing point, elastic modulus, thermal expansion coefficient, glass transition temperature, and impact strength. Furthermore, the material data acquired in the second step S02 can be appropriately determined so as to accurately estimate the properties that are the target of estimation by the estimation method according to this embodiment.

[0062] Furthermore, the composite material α whose properties are to be estimated using the estimation method according to this embodiment is not limited to mixed oils, but can be arbitrarily determined. Examples of such composite materials include mixed oils, aqueous solutions, surfactants, polymers, inorganic materials, and inorganic-organic hybrid materials. Moreover, the materials constituting the composite material α whose properties are to be estimated using the estimation method according to this embodiment are not limited to specific materials. Examples of such materials include oil-soluble liquids, water-soluble liquids, surfactants, monomers, polymers, and inorganic materials.

[0063] In addition, in the estimation method according to this embodiment, it is not necessary to repeat the processing flow P. In other words, in the estimation method according to this embodiment, when selecting all materials constituting the material group Gr as the target material in the first step S01, the process may be completed in one processing flow P, especially if the composite material α is composed of only two materials, and the material group Gr contains only two materials in the first place.

[0064] [Examples] The present invention will be described in detail below based on examples. However, the present invention is not limited in any way by the following examples.

[0065] (Example 1) In Example 1, the oil-water interface tension of composite material α1, composed of three types of materials a1, b1, and c1 shown in Table 1, was estimated. Materials a1, b1, and c1 are all oil-soluble materials, and composite material α1 is a mixed oil. Table 1 lists the CAS registry number, oil-water interface tension, and blending ratio for materials a1, b1, and c1 that constitute composite material α1.

[0066] [Table 1]

[0067] The first processing flow P1 in Example 1 will be described. In the first step S01, materials a1 and b1 were selected. In the second step S02, material data such as specific gravity and molecular weight were obtained for materials a1 and b1, in addition to the oil-water interface tension shown in Table 1. Also in the second step S02, a blending ratio of 50 mass% for each of materials a1 and b1 was obtained. In the third step S03, the mixing parameter was calculated from the material data of materials a1 and b1 using the first estimation model. The mixing parameter calculated in the third step S03 was 2.1. In the fourth step S04, the material data of intermediate composite material A1, composed of 50 mass% of material a1 and 50 mass% of material b1, was estimated using the second estimation model based on the material data and blending ratio of materials a1 and b1, as well as the mixing parameter. The estimated oil-water interface tension of intermediate composite material A1, calculated in the fourth step S04, was 20.1 mN / m.

[0068] The second processing flow P2 in Example 1 will now be described. In the first step S01, intermediate composite material A1 and material c1 were selected. In the second step S02, the estimated material data calculated in the fourth step S04 of the first processing flow P1 was obtained as material data for intermediate composite material A1. In addition, in the second step S02, material data such as specific gravity and molecular weight were obtained for material c1, in addition to the oil-water interface tension shown in Table 1. Furthermore, in the second step S02, a blending ratio of 60 mass% was obtained for intermediate composite material A1 and a blending ratio of 40 mass% was obtained for material c1. In the third step S03, the mixing parameter was calculated from the material data of intermediate composite material A1 and material c1 using the first estimation model. The mixing parameter calculated in the third step S03 was 3.0. In the fourth step, S04, the material data and mixing ratios of intermediate composite material A1 and material c1, as well as the mixing parameters, were used to calculate the estimated material data of composite material α1, which is composed of 60 mass% intermediate composite material A1 and 40 mass% material c1, using the second estimation model. The estimated oil-water interface tension of composite material α1 calculated in the fourth step, S04, was 14.6 mN / m.

[0069] In response, composite material α1 was actually fabricated, and its oil-water interface tension at 25°C was measured. The oil-water interface tension was measured using a contact angle meter DM-501 (manufactured by Kyowa Interface Science Co., Ltd.) by the pendant drop method. As a result, the measured value of the oil-water interface tension of composite material α1 was 14.9 mN / m, which was very close to the value of the target characteristic, which is the estimated value of the oil-water interface tension of composite material α1 according to Example 1. This confirmed that the estimation method according to Example 1 can estimate the oil-water interface tension of composite material α1 with high accuracy.

[0070] (Example 2) In Example 2, the oil-water interface tension of composite material α2, composed of seven types of materials a2, b2, c2, d2, e2, f2, and g2 as shown in Table 2, was estimated. Materials a2, b2, c2, d2, e2, f2, and g2 are all oil-soluble materials, and composite material α2 is a mixed oil. Table 2 lists the CAS registry number, oil-water interface tension, and blending ratio for materials a2, b2, c2, d2, e2, f2, and g2 that make up composite material α2.

[0071] [Table 2]

[0072] In Example 2, six processing flows P were performed. In the first processing flow P1, materials a2 and b2 were selected as target materials, and estimated material data values ​​for intermediate composite material A2 were calculated. In the second processing flow P2, intermediate composite materials A2 and c2 were selected as target materials, and estimated material data values ​​for intermediate composite material B2 were calculated. In the third processing flow P3, intermediate composite materials B2 and d2 were selected as target materials, and estimated material data values ​​for intermediate composite material C2 were calculated. In the fourth processing flow P4, intermediate composite materials C2 and e3 were selected as target materials, and estimated material data values ​​for intermediate composite material D2 were calculated. In the fifth processing flow P5, intermediate composite materials D2 and f2 were selected as target materials, and estimated material data values ​​for intermediate composite material E2 were calculated. In the sixth processing flow P6, intermediate composite materials E2 and g2 were selected as target materials, and estimated material data values ​​for composite material α2 were calculated.

[0073] In Example 2, the estimated oil-water interface tension of composite material α2 calculated in the fourth step S04 of the sixth processing flow P6 was 20.9 mN / m. In contrast, composite material α2 was actually fabricated, and its oil-water interface tension at 25°C was measured. As a result, the measured value of the oil-water interface tension of composite material α2 was 20.8 mN / m, which was very close to the value of the target characteristic, which is the estimated oil-water interface tension of composite material α2 in Example 2. This confirmed that the estimation method in Example 2 can estimate the oil-water interface tension of composite material α2 with high accuracy. [Explanation of Symbols]

[0074] a,b,c,d,e,f...material A,B,C,D…Intermediate composite material α…Composite material Gr…Material group 100... Information Processing Device 101...Input section 102...Output section 103... Processing Unit 104...Storage section

Claims

1. A method for estimating the desired properties of a composite material composed of multiple materials, A first step is to select at least two materials from a group of materials containing the aforementioned multiple materials in the correct proportions as components, as multiple target materials. A second step involves obtaining multiple material data for the multiple target materials and multiple blending ratios for the multiple target materials. A third step involves calculating a mixing parameter that expresses the closeness between the multiple material data sets using the following first estimation model, based on the first information including the multiple material data sets. A fourth step involves calculating an estimated value of the target properties of a composite material composed of the multiple target materials at the multiple mixing ratios, using the second information including the multiple material data, the multiple mixing ratios, and the mixing parameters, based on the second estimation model described below. A method for estimating the desired properties of a composite material containing [a specific element]. First estimation model: A machine learning model trained to output the mixed parameters in response to the input of the first information. Second estimation model: A machine learning model trained to output the estimated value in response to the input of the second piece of information.

2. The process includes selecting at least two materials from the group of materials, which consists of at least one intermediate composite material whose target properties have already been calculated, and materials from the plurality of materials that have not been selected as the plurality of target materials, as the plurality of target materials in the first step. A method for estimating the desired properties of a composite material according to claim 1.

3. When the number of the aforementioned materials is N, the series of processing flows including the first, second, third, and fourth steps are repeated N-1 times. A method for estimating the desired properties of a composite material according to claim 2.

4. The first estimation model does not use the plurality of blending ratios to calculate the blending parameters. A method for estimating the desired properties of a composite material according to claim 1 or 2.

5. The aforementioned mixture parameter is expressed as a non-negative real number. A method for estimating the desired properties of a composite material according to claim 1 or 2.

6. The aforementioned multiple material data are expressed as one of the following: a real number, a categorical value, or a probability density function. A method for estimating the desired properties of a composite material according to claim 1 or 2.

7. The aforementioned target characteristic is one or more selected from oil-water interfacial tension, surface tension, viscosity, critical micelle concentration, solubility, emulsifying power, foaming ability, contact angle, viscosity, dielectric constant, conductivity, boiling point, freezing point, elastic modulus, thermal expansion coefficient, glass transition temperature, and impact strength. A method for estimating the desired properties of a composite material according to claim 1 or 2.

8. An information processing device for estimating the target properties of a composite material composed of multiple materials, The information processing device performs a first step of selecting at least two materials from a group of materials containing the aforementioned plurality of materials as components in the correct proportions, A second step involves obtaining multiple material data for the multiple target materials and multiple blending ratios for the multiple target materials. A third step involves calculating a mixing parameter that expresses the closeness between the multiple material data sets using the following first estimation model, based on the first information including the multiple material data sets. A fourth step involves calculating an estimated value of the target properties of a composite material composed of the multiple target materials at the multiple mixing ratios, using the second information including the multiple material data, the multiple mixing ratios, and the mixing parameters, based on the second estimation model described below. It comprises a processing unit that performs the following: Information processing device. First estimation model: A machine learning model trained to output the mixed parameters in response to the input of the first information. Second estimation model: A machine learning model trained to output the estimated value in response to the input of the second piece of information.

9. A program for estimating the desired properties of a composite material composed of multiple materials, A first step is to select at least two materials from a group of materials containing the aforementioned multiple materials in the correct proportions as components, as multiple target materials. A second step involves obtaining multiple material data for the multiple target materials and multiple blending ratios for the multiple target materials. A third step involves calculating a mixing parameter that expresses the closeness between the multiple material data sets using the following first estimation model, based on the first information including the multiple material data sets. A fourth step involves calculating an estimated value of the target properties of a composite material composed of the multiple target materials at the multiple mixing ratios, using the second information including the multiple material data, the multiple mixing ratios, and the mixing parameters, based on the second estimation model described below. A program that causes an information processing device to execute. First estimation model: A machine learning model trained to output the mixed parameters in response to the input of the first information. Second estimation model: A machine learning model trained to output the estimated value in response to the input of the second piece of information.

10. A recording medium for storing a program for estimating the target properties of a composite material composed of multiple materials, The aforementioned program, A first step is to select at least two materials from a group of materials containing the aforementioned multiple materials in the correct proportions as components, as multiple target materials. A second step involves obtaining multiple material data for the multiple target materials and multiple blending ratios for the multiple target materials. A third step involves calculating a mixing parameter that expresses the closeness between the multiple material data sets using the following first estimation model, based on the first information including the multiple material data sets. A fourth step involves calculating an estimated value of the target properties of a composite material composed of the multiple target materials at the multiple mixing ratios, using the second information including the multiple material data, the multiple mixing ratios, and the mixing parameters, based on the second estimation model described below. To have the information processing device execute it Recording medium. First estimation model: A machine learning model trained to output the mixed parameters in response to the input of the first information. Second estimation model: A machine learning model trained to output the estimated value in response to the input of the second piece of information.

Images