Physical property prediction device, learned model, method for predicting physical property, and program

Abstract

To improve the accuracy of predicting the physical property of a laminate body with a fluorine resin layer.SOLUTION: A physical property prediction device includes: a basic information acquisition unit for acquiring basic information as information on a condition of generating a prediction target laminate body as a laminate body with a fluorine resin layer as a target of prediction; and a prediction unit for predicting physical property information of the prediction target laminate body from basic information of the prediction target laminate body as a laminate body with a fluorine resin layer as a target of prediction on the basis of a learned model which has learned the relation between the basic information of the learning target laminate body as a laminate body with a fluorine resin layer as a learning target and the physical information as information showing the physical property of the learning target laminate body.SELECTED DRAWING: Figure 2

Description

[Technical field]

[0001] The present invention relates to a property prediction device, a trained model, a property prediction method, and a program. [Background technology]

[0002] In order to efficiently search for laminates with a fluororesin layer obtained by coating a substrate with a fluoropolymer dispersion, prediction of physical properties using machine learning has been carried out (for example, see Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Patent Publication No. 2022-79265 Summary of the Invention [Problem to be solved by the invention]

[0004] However, the prediction accuracy of the physical properties of the fluororesin layer-attached laminate has not always been high, and therefore there is a demand for improving the prediction accuracy of the physical properties of the fluororesin layer-attached laminate.

[0005] The present invention has been made in consideration of the above points, and provides a property prediction device, a trained model, a property prediction method, and a program that can improve the prediction accuracy of the physical properties of a laminate with a fluororesin layer. [Means for solving the problem]

[0006] In order to solve the above problems, the present invention has the following configuration. [1] A physical property prediction device comprising: a basic information acquisition unit that acquires basic information, which is information regarding the conditions for generating a prediction target laminate, which is a fluororesin layer-attached laminate to be predicted; and a prediction unit that predicts the physical property information of the prediction target laminate, which is a fluororesin layer-attached laminate to be predicted, from the basic information of the prediction target laminate, which is a fluororesin layer-attached laminate to be predicted, based on a trained model in which the relationship between the basic information of the training target laminate, which is a fluororesin layer-attached laminate to be learned, and physical property information, which is information indicating the physical properties of the training target laminate, is learned.

[0007] [2] The property prediction device according to [1], wherein the basic information includes dispersion information which is information about a dispersion used in producing the fluororesin layer-attached laminate, and substrate information which is information about a substrate used in producing the fluororesin layer-attached laminate.

[0008] [3] The property prediction device according to [2], wherein the dispersion information includes information about a fluoropolymer contained in the dispersion.

[0009] [4] The physical property prediction device according to [2], wherein the fluoropolymer contained in the dispersion is a tetrafluoroethylene-based polymer.

[0010] [5] The physical property prediction device described in [1], wherein the physical property information includes any one of the dielectric properties, peel strength, surface roughness, color, tensile strength / elongation, CTE, and thermal expansion coefficient of the laminate to be predicted. [6] The property prediction device described in [1], wherein the trained model is a trained model trained based on regression analysis, class classification, or ensemble learning that combines multiple regression analyses or class classifications.

[0011] [7] The property prediction device described in [1], wherein the trained model is a trained model trained based on Gaussian process regression.

[0012] [8] A trained model in which the relationship between basic information of a training subject laminate, which is a laminate with a fluororesin layer, and physical property information, which is information indicating the physical properties of the training subject laminate, has been trained.

[0013] [9] A method for predicting physical properties, comprising: acquiring basic information which is information relating to the conditions for producing a prediction target laminate, which is a fluororesin layer-attached laminate to be predicted; and predicting physical property information of the prediction target laminate, which is a fluororesin layer-attached laminate to be predicted, from the basic information of the prediction target laminate, which is a fluororesin layer-attached laminate to be predicted, based on a trained model in which the relationship between the basic information of the training target laminate, which is a fluororesin layer-attached laminate to be learned, and physical property information which is information indicating the physical properties of the training target laminate, is learned.

[0014]

[10] A program for causing a computer to obtain basic information which is information regarding the conditions for producing a prediction target laminate which is a fluororesin layer-attached laminate to be predicted, and for predicting physical property information of the prediction target laminate which is a fluororesin layer-attached laminate to be predicted from the basic information of the prediction target laminate which is a fluororesin layer-attached laminate to be predicted, based on a trained model in which the relationship between the basic information of the training target laminate which is a fluororesin layer-attached laminate to be learned and physical property information which is information indicating the physical properties of the training target laminate has been learned. Effect of the Invention

[0015] According to the present invention, it is possible to improve the prediction accuracy of the physical properties of a laminate with a fluororesin layer. [Brief description of the drawings]

[0016] [Figure 1] FIG. 1 is a diagram illustrating an example of a configuration of a property prediction system according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a diagram illustrating an example of a functional configuration of a property prediction apparatus according to an embodiment of the present invention. [Diagram 3] FIG. 11 is a diagram illustrating an example of a learning process according to an embodiment of the present invention. [Figure 4] FIG. 11 is a diagram illustrating an example of a prediction process according to an embodiment of the present invention. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0017] (Embodiment) Hereinafter, an embodiment of the present invention will be described in detail with reference to the drawings. Fig. 1 is a diagram showing an example of a configuration of a property prediction system 1 according to the present embodiment. The property prediction system 1 includes a property prediction device 2, an input device 3, a display device 4, and a data server 5.

[0018] The physical property prediction device 2 predicts the physical properties of a fluororesin layer-attached laminate obtained by coating a substrate with a fluoropolymer dispersion, based on machine learning, using information on the dispersion used to produce the fluororesin layer-attached laminate (hereinafter referred to as "dispersion information") and information on the substrate (hereinafter referred to as "substrate information") as explanatory variables. The physical property prediction device 2 is, for example, an information processing device such as a personal computer (PC).

[0019] The input device 3 inputs various data to the property prediction device 2. The input device 3 is, for example, a keyboard, a mouse, or a touch panel.

[0020] The display device 4 displays various data used for processing by the property prediction device 2 and the prediction results by the property prediction device 2. The display device 4 is, for example, a display.

[0021] The data server 5 stores various data used for information processing by the property prediction device 2. The data server 5 supplies the various data to the property prediction device 2. The various data includes basic property information data A1.

[0022] The basic physical property information data A1 is a set of information on the conditions for producing the fluororesin layer-attached laminate (hereinafter referred to as "basic information") and physical property information of the fluororesin layer-attached laminate produced under the conditions indicated by the basic information. The basic information includes, for example, dispersion information and substrate information.

[0023] The dispersion information includes some or all of the following information: information on the fluoropolymer used in the dispersion, information on the dispersant used, information on the dispersion medium used, etc. The information on the fluoropolymer used in the dispersion may include the type of fluoropolymer and the particle size of the fluoropolymer. Specific examples of the type of fluoropolymer include tetrafluoroethylene-based polymers. Specific examples of the type of tetrafluoroethylene-based polymers include PTFE (Poly Tetra Fluoro Etylene), PFA (Perfluoroalkoxy alkane), FEP (Fluorinated Ethylene Propylene), ETFE (Ethylene Tetra Fluoro Ethylene), etc.

[0024] More specifically, tetrafluoroethylene-based polymers described in WO 2022 / 050253 and WO 2019 / 230569, fluorine-containing polymers described in JP 2022-140517 A, and more specifically, PFAs described in WO 2021 / 039735 and WO 2016 / 017801 can be mentioned. Suitable embodiments of the fluoropolymer are the same as those in the above-mentioned documents. As the fluoropolymer, PFA is preferred, and it is preferable to have an oxygen-containing polar group. Among them, a carbonyl group-containing group is preferred. However, other fluoropolymers such as PTFE can also be used as appropriate.

[0025] The information on the fluoropolymer includes particle size, shape, melting point, thermal decomposition temperature, water content, etc. The information on the dispersant includes dispersant components, cloud point, thermal decomposition temperature, surface tension, main chain structure, side chain structure, solubility parameter, etc. The information on the dispersion medium includes information indicating the type of the dispersion medium. Specific examples of the type of dispersion medium include water and organic solvents. Information indicating the type of organic solvent may be used as the information indicating the type of dispersion medium. Examples of the dispersion medium include liquid dispersion mediums described in International Publication No. 2022 / 050253.

[0026] In addition, when components other than fluoropolymers, dispersants, and dispersion media are included, information on organic components other than fluoropolymers and inorganic fillers may be used. Representative examples of the organic components include acrylic polymers, aromatic polymers, alicyclic polymers, and buffer materials, and the information on the organic components may include the composition species, composition ratio, particle size, melting point, solubility, glass transition temperature, and thermal decomposition temperature constituting the organic components. Representative examples of the inorganic components include silica, boron nitride, aluminum oxide, aluminum nitride, barium titanate, and silicon nitride, and the information on the inorganic components may include particle size, shape, and surface treatment components. Examples of organic components and inorganic fillers other than fluoropolymers include other resins and inorganic fillers described in International Publication No. 2022 / 050253. Examples of inorganic fillers include inorganic fillers described in JP-A-2022-140517.

[0027] Specific examples of the dispersion liquid include, but are not limited to, the dispersion liquids described in the examples of WO 2016 / 017801.

[0028] The substrate information includes a part or all of information such as information on the type of substrate and information on the characteristics of the substrate. Examples of substrate types include heat-resistant resins and metal foils. Specific examples of heat-resistant resins include polyimide, liquid crystal polyester, PTFE, PEEK (Poly Ether Ether Ketone), and PPS (Poly Phenylene Sulfide). Examples of polyimide substrates include aromatic polyimides described in International Publication No. 2021 / 039735. Specific examples of metal foils include copper foil, copper alloy foil, aluminum foil, aluminum alloy foil, SUS foil, titanium foil, titanium alloy foil, nickel foil, and nickel alloy foil. Examples of metal foils include those described in International Publication No. 2019 / 230569, and copper foil is preferable among them. Specific examples of information indicating the characteristics of the substrate include the surface roughness of the substrate, the elastic modulus of the substrate, the CTE (Coefficient of Thermal Expansion) of the substrate, the film thickness, the shrinkage rate, the glass transition temperature, the degree of hardening, the elongation, the viscoelastic properties, the dielectric breakdown strength, the surface treatment components, and the water absorption rate.

[0029] The basic information may include information on the preparation of the dispersion. The information on the preparation may include, for example, information on the device and auxiliary parts used for stirring, the stirring temperature, the stirring time, the stirring speed, the dispersion medium temperature, the stirring position, the filtration diameter, the type of the filtration filter, and the like. Specific examples of the device used for stirring include a disperser, a bead mill, a planetary mixer, and a rotation-revolution mixer. The basic information may include information on polymerization. Specific examples of the information on polymerization include information indicating whether the dispersion is a suspension or an emulsion.

[0030] The basic information may include information on the conditions when coating the substrate with the dispersion. Specific examples of such information include information on the conditions when coating the substrate with the dispersion, such as the filtration filter type, filtration filter diameter, liquid supply speed, coating gap, coater type, etc., information on the conditions when drying the coated substrate, such as the coating transport speed, drying temperature, air supply conditions, exhaust conditions, furnace length, etc., and information on the conditions when baking the coated substrate, such as the baking transport speed, furnace length, tension, baking temperature, atmosphere, etc.

[0031] The physical property information may include any one or all of the dielectric properties, peel strength, surface roughness, color, tensile strength and elongation, CTE, and thermal expansion coefficient of the fluororesin layer-attached laminate.

[0032] The physical property prediction device 2 and the data server 5 communicate with each other. The communication may be wireless communication or wired communication. The data server 5 is, for example, a database.

[0033] [Functional configuration of property prediction device 2] 2 is a diagram showing an example of a functional configuration of the physical property prediction apparatus 2 according to the present embodiment. The physical property prediction apparatus 2 includes a control unit 20 and a storage unit .

[0034] The control unit 20 includes, for example, a CPU (Central Processing Unit) and performs various calculations and information transfer. The control unit 20 includes a basic physical property information data acquisition unit 200, a teacher data generation unit 201, a learning unit 202, a basic information acquisition unit 203, a prediction unit 204, and an output unit 205. Each of these functional units is realized, for example, by the CPU loading a program read from a ROM (Read Only Memory) into a RAM (Random Access Memory) and executing processing according to the program. The ROM and RAM are included in the storage unit 21.

[0035] The basic physical property information data acquiring unit 200 acquires the basic physical property information data A1. In this embodiment, the basic physical property information data acquiring unit 200 acquires the basic physical property information data A1 from the data server 5.

[0036] The teacher data generating unit 201 generates teacher data B1 from the basic physical property information data A1. The teacher data generating unit 201 generates teacher data used in the learning unit 202 by performing so-called pre-processing on information included in the basic physical property information data. The teacher data generating unit 201 may perform processing such as logarithmic conversion and scaling on basic information and physical property information expressed by numerical values, for example. It is desirable that the type of pre-processing that the teacher data generating unit 201 performs on what information is determined in advance. Here, in this embodiment, the number of the basic physical property information data A1 is, for example, about several thousand. In this embodiment, about several thousand is any number between one thousand and ten thousand. The number of the basic physical property information data A1 may be about several thousand or less, or may be about several thousand or more.

[0037] The training data B1 includes basic information about the learning target laminate and physical property information about the learning target laminate. In the training data B1, the basic information and the physical property information are associated with each other. The learning target laminate is a fluororesin layer-attached laminate to be trained. The basic information (information about the conditions for producing the fluororesin layer-attached laminate) corresponds to an explanatory variable in machine learning. The physical property information corresponds to a target variable in machine learning.

[0038] Here, the fluororesin layer-attached laminate to be studied is obtained by coating a substrate with a fluoropolymer dispersion. Therefore, information on the dispersion and substrate used in the fluororesin layer-attached laminate has a significant effect on the physical properties of the fluororesin layer-attached laminate. Therefore, information on the fluoropolymer dispersion (dispersion information) and information on the substrate (substrate information) are correlated with the physical property information of the fluororesin layer-attached laminate.

[0039] More specifically, it is as follows. Among the dispersion information, information on the fluoropolymer used in the dispersion is correlated with the physical property information of the laminate. This is because the fluoropolymer used acts as a part of the laminate, and the state and characteristics of the fluoropolymer affect the physical properties of the laminate. Among the dispersion information, information on the dispersant is correlated with the physical property information of the laminate. This is because the dispersant used affects the state in which the fluoropolymer is dispersed. If a dispersion in a state in which it is not sufficiently dispersed is used, it may have an adverse effect on the physical properties of the laminate. Among the dispersion information, information on the dispersion medium is correlated with the physical property information of the laminate. This is because the dispersion medium used affects the state in which the fluoropolymer is dispersed. If a dispersion in a state in which it is not sufficiently dispersed is used, it may have an adverse effect on the physical properties of the laminate.

[0040] Among the substrate information, information on the type of substrate correlates with the physical property information of the laminate. This is because the substrate used acts as part of the laminate, and in particular the type of substrate affects the physical properties of the laminate. Among the substrate information, information on the characteristics of the substrate correlates with the physical property information of the laminate. This is because the substrate used acts as part of the laminate, and in particular its characteristics (e.g. surface characteristics, etc.) affect the physical properties of the laminate.

[0041] The learning unit 202 generates a trained model C1 by performing machine learning using the teacher data B1. Here, the learning unit 202 performs machine learning using basic information about the training target laminate contained in the teacher data B1 as explanatory variables and physical property information about the training target laminate as objective variables. The learning unit 202 stores the generated trained model C1 in the storage unit 21. The trained model C1 is a model in which the relationship between basic information and physical property information about the training target laminate has been learned. The training target laminate is a laminate with a fluororesin layer that is a training target.

[0042] The machine learning performed by the learning unit 202 is, for example, regression analysis, class classification, or ensemble learning. The ensemble learning is, for example, ensemble learning that combines multiple regression analyses or class classifications. In these cases, the trained model C1 is a trained model trained based on regression analysis, class classification, or ensemble learning that combines multiple regression analyses or class classifications.

[0043] As another example, the machine learning performed by the learning unit 202 may be machine learning based on Gaussian process regression. In this case, the trained model C1 is a trained model trained based on Gaussian process regression.

[0044] The machine learning performed by the learning unit 202 may be other types of machine learning, such as machine learning using a neural network, such as deep learning. However, deep learning requires a larger amount of teacher data than machine learning such as regression analysis, classification, ensemble learning combining multiple regression analyses or classifications, or machine learning based on Gaussian process regression, and therefore it is preferable to use the above-mentioned machine learning for several thousand pieces of teacher data B1 as in this embodiment.

[0045] The basic information acquisition unit 203 acquires basic information D1 about the prediction target laminate. The prediction target laminate is a fluororesin layer-attached laminate whose physical properties are to be predicted. In this embodiment, the basic information acquisition unit 203 acquires the basic information D1 from the input device 3. Here, acquiring the basic information D1 from the input device 3 refers to acquiring the basic information D1 input by a user from the input device 3, or acquiring data indicating the basic information from a predetermined information processing device or recording medium.

[0046] The prediction unit 204 performs preprocessing on the acquired basic information similar to the preprocessing performed by the teacher data generation unit 201. The prediction unit 204 predicts physical property information F1 for the prediction target laminate from the basic information data D1 on which the preprocessing has been performed, based on the trained model C1. Here, the type of physical property information F1 predicted by the prediction unit 204 is the same as the type of physical property information included in the teacher data B1 used by the learning unit 202 for machine learning.

[0047] The output unit 205 outputs the physical property information F1 predicted by the prediction unit 204. The output unit 205, for example, causes the display device 4 to display the physical property information. The output unit 205 may store the physical property information in an external server such as the data server 5 or in the storage unit 21.

[0048] The storage unit 21 stores various information. For example, the storage unit 21 stores a trained model C1. The storage unit 21 stores basic physical property information data A1 acquired by the basic physical property information data acquisition unit 200, teacher data B1 generated by the teacher data generation unit 201, basic information D1 acquired by the basic information acquisition unit 203, and basic information D1 preprocessed by the prediction unit 204. The storage unit 21 is configured using a storage device such as a magnetic hard disk device or a semiconductor storage device.

[0049] The processes performed by the property prediction device 2 include a learning process and a prediction process. Here, the learning process and the prediction process will be described with reference to Figs. 3 and 4.

[0050] [Learning process] FIG. 3 is a diagram showing an example of a learning process according to this embodiment. The learning process is a process for generating a learned model C1, and is executed by the control unit 20. The learning process is executed, for example, at a time prior to the prediction process. Note that the learning process may be performed using data combining preprocessed basic information data D1 and physical property information F1 for the prediction target laminate on which the prediction process has been performed, as new teacher data. In other words, the learned model C1 may be updated based on the prediction result of the physical property information F1.

[0051] Step S10: The basic physical property information data acquisition unit 200 acquires the basic physical property information data A1.

[0052] Step S20: The teacher data generating unit 201 generates teacher data B1 from the basic physical property information data A1 acquired by the basic physical property information data acquiring unit 200.

[0053] Step S30: The learning unit 202 performs machine learning using the teacher data B1 generated by the teacher data generating unit 201.

[0054] Step S40: The learning unit 202 generates a trained model C1 as a result of machine learning. The learning unit 202 stores the generated trained model C1 in the storage unit 21. With this, the control unit 20 ends the learning process.

[0055] [Prediction processing] 4 is a diagram showing an example of the prediction process according to the present embodiment. The prediction process is a process for predicting physical property information F1 for a laminate to be predicted, and is executed by the control unit 20.

[0056] Step S110: The basic information acquisition unit 203 acquires basic information D1 about the prediction target laminate.

[0057] Step S120: The prediction unit 204 performs a predetermined pre-processing on the basic information D1 acquired by the basic information acquisition unit 203.

[0058] Step S130: The prediction unit 204 predicts physical property information F1 for the prediction target laminate from the basic information data D1 on which preprocessing has been performed, based on the trained model C1. When the trained model C1 receives the basic information data D1 on which preprocessing has been performed, the trained model C1 outputs the physical property information F1. The prediction unit 204 predicts the physical property information F1 by inputting the basic information data D1 on which preprocessing has been performed into the trained model C1 and having the trained model C1 output the physical property information F1.

[0059] Step S150: The output unit 205 outputs the physical property information F1 predicted by the prediction unit 204. With this, the control unit 20 ends the prediction process.

[0060] In the present embodiment, an example in which the functional unit for performing the learning process is provided in the physical property prediction device 2 has been described, but the present invention is not limited to this. The functional unit for performing the learning process is the basic physical property information data acquisition unit 200, the teacher data generation unit 201, and the learning unit 202. The functional unit for performing the learning process may be omitted from the configuration of the physical property prediction device 2. In this case, the learned model C1 is generated in advance by a learning model generation device that is an information processing device separate from the physical property prediction device 2. The learning model generation device is, for example, a PC or a server. The physical property prediction device 2 acquires the previously generated learned model C1 from the learning model generation device and uses it in the prediction process.

[0061] In the present embodiment, an example in which the basic physical property information data A1 is stored in the data server 5 has been described, but the present invention is not limited to this. The basic physical property information data A1 may be stored in a storage unit 21 included in the physical property prediction device 2. In this case, the data server 5 may be omitted from the configuration of the physical property prediction system 1.

[0062] In the present embodiment, an example in which the physical property prediction device 2 is a PC has been described, but the present invention is not limited to this. The physical property prediction device 2 may be composed of one or more servers. The one or more servers may be constructed on the Internet as cloud servers. When the physical property prediction device 2 is composed of one or more servers, each functional unit of the control unit 20 and the storage unit 21 are distributed among the one or more servers.

[0063] As described above, the property prediction device 2 according to this embodiment includes the prediction unit 204. The prediction unit 204 predicts physical property information F1 for the learning target laminate, which is the laminate to be learned, from the basic information D1 acquired by the basic information acquisition unit 203 based on a trained model C1 in which the relationship between basic information and physical property information, which is information indicating the physical properties of the learning target laminate, is learned. With this configuration, the physical property prediction device 2 of this embodiment can predict physical property information F1 for a fluororesin layered laminate from basic information D1 based on the learned model C1, thereby improving the prediction accuracy of the physical properties of the fluororesin layered laminate.

[0064] In addition, in the physical property prediction device 2 of this embodiment, the trained model C1 may be a trained model trained based on regression analysis, class classification, or ensemble learning that combines multiple regression analyses or class classifications. With this configuration, in the physical property prediction device 2 of this embodiment, regression analysis, class classification, or ensemble learning that combines multiple regression analyses or class classifications requires a smaller amount of data than deep learning, etc., so that it is possible to improve the prediction accuracy of the physical properties of the fluororesin layer-attached laminate even when learning is performed with a smaller amount of data than deep learning, etc.

[0065] In addition, in the property prediction device 2 according to this embodiment, the trained model C1 may be a trained model trained based on Gaussian process regression. With this configuration, in the physical property prediction device 2 of this embodiment, Gaussian process regression requires a smaller amount of data than deep learning, etc., so that it is possible to improve the prediction accuracy of the physical properties of the fluororesin layer-attached laminate even when learning is performed with a smaller amount of data than deep learning, etc.

[0066] Furthermore, in the physical property prediction device 2 according to this embodiment, the physical property information may include one or more of dielectric properties, peel strength, surface roughness, color, tensile strength / elongation, CTE, and thermal expansion coefficient. With this configuration, the physical property prediction device 2 of this embodiment can predict one or more of the dielectric properties, peel strength, surface roughness, color, tensile strength / elongation, CTE, and thermal expansion coefficient for a laminate with a fluororesin layer based on the trained model C1, thereby improving the prediction accuracy of one or more of the dielectric properties, peel strength, surface roughness, color, tensile strength / elongation, CTE, and thermal expansion coefficient.

[0067] In addition, a part of the physical property prediction device 2 in the above-mentioned embodiment, for example, the basic physical property information data acquisition unit 200, the teacher data generation unit 201, the learning unit 202, the basic information acquisition unit 203, the prediction unit 204, and the output unit 205 may be realized by a computer. In that case, a program for realizing this control function may be recorded in a computer-readable recording medium, and the program recorded in the recording medium may be read into a computer system and executed to realize the control function. Note that the "computer system" here refers to a computer system built into the physical property prediction device 2, and includes hardware such as an operating system (OS) and peripheral devices. In addition, the "computer-readable recording medium" refers to a portable medium such as a flexible disk, a magneto-optical disk, a ROM (Read Only Memory), a CD-ROM (Compact Disc-Read Only Memory), and a storage device such as a hard disk built into the computer system. Furthermore, the term "computer-readable recording medium" may include a medium that dynamically stores a program for a short period of time, such as a communication line when transmitting a program via a network such as the Internet or a communication line such as a telephone line, and a medium that stores a program for a certain period of time, such as a volatile memory inside a computer system that serves as a server or client in such a case. The above program may be one that realizes part of the above-mentioned functions, or may be one that can realize the above-mentioned functions in combination with a program already recorded in the computer system. In addition, a part or all of the physical property prediction device 2 in the above-mentioned embodiment may be realized as an integrated circuit such as an LSI (Large Scale Integration). Each functional block of the physical property prediction device 2 may be individually processed, or a part or all of the functional blocks may be integrated into a processor. The integrated circuit method is not limited to LSI, and may be realized by a dedicated circuit or a general-purpose processor. In addition, when an integrated circuit technology that replaces LSI appears due to the progress of semiconductor technology, an integrated circuit based on that technology may be used.

[0068] Although one embodiment of the present invention has been described in detail above with reference to the drawings, the specific configuration is not limited to the above, and various design modifications and the like are possible within the scope that does not deviate from the gist of the present invention. [Explanation of symbols]

[0069] 2...physical property prediction device, 204...prediction unit, C1...trained model, F1...physical property information

Claims

1. a basic information acquiring unit for acquiring basic information which is information on the conditions for producing a prediction target laminate which is a fluororesin layer-attached laminate to be predicted; a prediction unit that predicts physical property information of a prediction target laminate, which is a fluororesin layer-attached laminate to be learned, from basic information of the prediction target laminate, which is a prediction target fluororesin layer-attached laminate, based on a trained model in which a relationship between basic information of the learning target laminate, which is a learning target fluororesin layer-attached laminate, and physical property information, which is information indicating the physical properties of the learning target laminate, has been learned; A physical property prediction device comprising:

2. The basic information includes dispersion information, which is information about a dispersion used in producing the fluororesin layer-attached laminate, and substrate information, which is information about a substrate used in producing the fluororesin layer-attached laminate. The physical property prediction device according to claim 1 .

3. The dispersion information includes information about the fluoropolymer contained in the dispersion. The physical property prediction device according to claim 2 .

4. The physical property prediction device according to claim 2 , wherein the fluoropolymer contained in the dispersion liquid is a tetrafluoroethylene-based polymer.

5. The physical property prediction device according to claim 1 , wherein the physical property information includes any one of a dielectric property, a peel strength, a surface roughness, a color, a tensile strength / elongation, a CTE, and a thermal expansion coefficient of the laminate to be predicted.

6. The trained model is a trained model trained based on regression analysis, class classification, or ensemble learning that combines multiple regression analyses or class classifications. The physical property prediction device according to claim 1 .

7. The physical property prediction device according to claim 1 , wherein the trained model is a trained model trained based on Gaussian process regression.

8. A relationship between basic information of the learning target laminate, which is a laminate with a fluororesin layer, and physical property information, which is information showing the physical properties of the learning target laminate, is learned. Trained model.

9. Acquire basic information, which is information on the conditions for producing a prediction target laminate, which is a prediction target fluororesin layer-attached laminate, predicting physical property information of a prediction target laminate, which is a fluororesin layer-attached laminate to be predicted, from basic information of a prediction target laminate, which is a prediction target fluororesin layer-attached laminate, based on a trained model in which a relationship between basic information of a training target laminate, which is a training target fluororesin layer-attached laminate, and physical property information, which is information indicating the physical properties of the training target laminate, has been trained; Methods for predicting physical properties.

10. On the computer, Acquire basic information which is information on the conditions for producing a prediction target laminate, which is a prediction target fluororesin layer-attached laminate; predicting physical property information of a prediction target laminate, which is a fluororesin layer-attached laminate to be predicted, from basic information of a prediction target laminate, which is a prediction target fluororesin layer-attached laminate, based on a trained model in which a relationship between basic information of a training target laminate, which is a training target fluororesin layer-attached laminate, and physical property information, which is information indicating the physical properties of the training target laminate, has been trained; Program for.

Images