Estimation method, estimation device, and program

Abstract

To provide an estimation method, a model generation method, an estimation device, a model generation device, and a program capable of estimating characteristics of a resin composition with high accuracy.SOLUTION: An estimation method is executed by one or more computers. The estimation method according to the present embodiment includes an estimation step S11. In the estimation step S11, one or more characteristic values of a resin composition obtained in accordance with mixing data are estimated using the mixing data and feature quantities. The mixing data represent mixing ratios of a plurality of materials including a resin and a filler. The feature quantities are the feature quantities that relate to at least one of a shape of the filler, an inter-particle distance of the filler in the resin composition calculated using the mixing data, and a modulus of elasticity of the resin composition calculated using the mixing data.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an estimation method, a model generation method, an estimation apparatus, a model generation apparatus, and a program. [Background technology]

[0002] When attempting to obtain a resin composition with desired properties, the effort required for experimentation can be reduced if it is possible to accurately estimate what properties of the resin composition can be obtained based on the material formulation. Techniques have been developed to use computers to estimate the properties of a resin composition obtained from information about its formulation.

[0003] Patent Document 1 describes a process in which a computer inputs formulation information into a trained model and estimates the physical properties of the adhesive corresponding to the input formulation information. [Prior art documents] [Patent Documents]

[0004] [Patent Document 1] Japanese Patent Publication No. 2021-26478 [Overview of the Initiative] [Problems that the invention aims to solve]

[0005] However, there was room for improvement in the accuracy of the estimation when predicting the properties of the resin composition.

[0006] One aspect of the present invention provides a technology that enables the estimation of the properties of a resin composition with high accuracy. [Means for solving the problem]

[0007] According to one embodiment of the present invention, the following estimation method, model generation method, estimation apparatus, model generation apparatus, and program are provided.

[0008] 1. An estimation method executed by one or more computers, comprising: an estimation step of estimating one or more characteristic values of a resin composition obtained according to the compounding data, using the compounding data indicating the compounding ratio of each of a plurality of materials including a resin and a filler, and a feature quantity; wherein the feature quantity relates to at least any one of the shape of the filler, the inter-particle distance of the filler in the resin composition calculated using the compounding data, and the elastic modulus of the resin composition calculated using the compounding data; Estimation method. 2. The estimation method according to 1., further comprising a feature quantity specifying step of specifying the feature quantity using the compounding data; Estimation method. 3. The estimation method according to 1. or 2., wherein the feature quantity includes a value obtained by multiplying the particle size of the filler by the specific surface area of the filler; Estimation method. 4. The estimation method according to any one of 1. to 3., wherein the one or more characteristic values include one or more of specific gravity, water absorption rate, diffusion coefficient, volume resistivity, electrical conductivity, surface resistance, refractive index, relative permittivity, total light transmittance, glass transition temperature, linear expansion rate, volume expansion rate, thermal conductivity, thermal diffusivity, specific heat, die shear strength, peel strength, flexural strength, tensile strength, compressive strength, dynamic elastic modulus, tensile elastic modulus, flexural elastic modulus, elongation rate, and viscosity; Estimation method. 5. The estimation method according to any one of 1. to 4., further comprising a clustering step of clustering a plurality of the compounding data into a plurality of clusters; Estimation method. 6. The estimation method according to 5., further comprising a representative compounding data extraction step of extracting one or more of the compounding data from each of the plurality of clusters; Estimation method. 7. The estimation method according to any one of 1. to 6., In the estimation step, a plurality of characteristic values of the resin composition are estimated, and further includes a preference degree calculation step of calculating a preference degree for the blending data by using the plurality of characteristic values and weight coefficients for each of the plurality of characteristic values. Estimation method. 8. In the estimation method according to 7., further includes a display data output step of outputting display data for displaying an input screen for the weight coefficient, and a weight coefficient acquisition step of acquiring the weight coefficient input on the input screen, In the preference degree calculation step, the preference degree is calculated by using the weight coefficient acquired in the weight coefficient acquisition step. Estimation method. 9. In the estimation method according to 7. or 8., further includes a blending output step of outputting the blending data and the calculated preference degree. Estimation method. 10. In the estimation method according to 7. or 8., in the estimation step, for each of the plurality of blending data, the plurality of characteristic values are estimated, and further includes a blending output step of outputting the plurality of blending data in an order based on the preference degree. Estimation method. 11. In the estimation method according to 7. or 8., in the estimation step, for each of the plurality of blending data, the plurality of characteristic values are estimated, and further includes a blending output step of outputting one or more of the blending data extracted from the plurality of blending data based on the preference degree. Estimation method. 12. In the estimation method according to any one of 1. to 11., in the estimation step, the one or more characteristic values are estimated by using an estimation model in which machine learning has been performed. Estimation method. 13. The system includes an estimation unit that estimates one or more characteristic values ​​of a resin composition obtained according to the formulation data, using formulation data indicating the blending ratio of each of several materials including resin and filler, and characteristic quantities. The aforementioned feature quantity relates to at least one of the following: the shape of the filler, the interparticle distance of the filler in the resin composition calculated using the formulation data, and the elastic modulus of the resin composition calculated using the formulation data. Estimation device. 14. Computers, It functions as an estimation unit that estimates one or more characteristic values ​​of a resin composition obtained according to the formulation data, using formulation data indicating the mixing ratio of each of several materials including resin and filler, and characteristic quantities. The aforementioned feature quantity relates to at least one of the following: the shape of the filler, the interparticle distance of the filler in the resin composition calculated using the formulation data, and the elastic modulus of the resin composition calculated using the formulation data. program. 15. A model generation method performed by one or more computers, The process includes a model generation step in which machine learning is performed to generate an estimation model using training data that includes formulation data showing the mixing ratio of each of several materials including resins and fillers, feature quantities, and one or more characteristic values ​​of the resin composition obtained according to the formulation data. The aforementioned feature quantity relates to at least one of the following: the shape of the filler, the interparticle distance of the filler in the resin composition calculated using the formulation data, and the elastic modulus of the resin composition calculated using the formulation data. Model generation method. 16. The system includes a model generation unit that generates an estimation model by performing machine learning using training data which includes formulation data indicating the blending ratio of each of several materials including resin and filler, feature quantities, and one or more characteristic values ​​of the resin composition obtained according to the formulation data. The aforementioned feature quantity relates to at least one of the following: the shape of the filler, the interparticle distance of the filler in the resin composition calculated using the formulation data, and the elastic modulus of the resin composition calculated using the formulation data. Model generation device. 17. Computers, This system functions as a model generation unit that generates an estimation model by performing machine learning using training data that includes formulation data showing the mixing ratio of each of several materials, including resins and fillers, feature quantities, and one or more characteristic values ​​of the resin composition obtained according to that formulation data. The aforementioned feature quantity relates to at least one of the following: the shape of the filler, the interparticle distance of the filler in the resin composition calculated using the formulation data, and the elastic modulus of the resin composition calculated using the formulation data. program. [Effects of the Invention]

[0009] According to one aspect of the present invention, a technology is provided that can estimate the properties of a resin composition with high accuracy. [Brief explanation of the drawing]

[0010] [Figure 1] This is a flowchart outlining the estimation method according to the first embodiment. [Figure 2] This is a block diagram illustrating the overview of the estimation device according to the first embodiment. [Figure 3] This is a block diagram illustrating the functional configuration of the estimation device according to the first embodiment. [Figure 4] This is a flowchart illustrating the flow of the estimation method according to the first embodiment. [Figure 5] This diagram illustrates a computer used to implement an estimation device. [Figure 6] This diagram illustrates the structure of the formulation data. [Figure 7] This diagram illustrates an example of the estimation model used by the estimation unit. [Figure 8]This is a flowchart outlining the model generation method according to the first embodiment. [Figure 9] This is a block diagram illustrating the functional configuration of a model generation device according to the first embodiment. [Figure 10] This figure illustrates the functional configuration of the estimation device according to the second embodiment. [Figure 11] This is a flowchart illustrating the flow of the estimation method according to the second embodiment. [Figure 12] This diagram illustrates the structure of constraint information. [Figure 13] This is a diagram illustrating the structure of request information. [Figure 14] This figure illustrates the functional configuration of the estimation device according to the third embodiment. [Figure 15] This is a flowchart illustrating the flow of the estimation method according to the third embodiment. [Figure 16] This figure illustrates the functional configuration of the estimation device according to the fourth embodiment. [Figure 17] This is a flowchart illustrating the flow of the estimation method according to the fourth embodiment. [Figure 18] This diagram illustrates the structure of weight information. [Figure 19] This is a flowchart showing a modified example of the estimation device according to the fourth embodiment. [Modes for carrying out the invention]

[0011] Embodiments of the present invention will be described below with reference to the drawings. In all drawings, similar components are denoted by the same reference numerals, and their descriptions are omitted as appropriate.

[0012] (First Embodiment) Figure 1 is a flowchart illustrating the overview of the estimation method according to the first embodiment. The estimation method according to this embodiment is performed by one or more computers. The estimation method according to this embodiment includes an estimation step S11. In estimation step S11, one or more characteristic values ​​of the resin composition obtained according to the formulation data are estimated using formulation data and feature quantities. The formulation data is data indicating the mixing ratio of each of a plurality of materials, including resin and filler. The feature quantities are feature quantities relating to at least one of the following: the shape of the filler, the interparticle distance of the filler in the resin composition calculated using the formulation data, and the elastic modulus of the resin composition calculated using the formulation data.

[0013] Figure 2 is a block diagram illustrating the overview of the estimation device 10 according to this embodiment. The estimation device 10 according to this embodiment includes an estimation unit 130. The estimation unit 130 uses formulation data and feature quantities to estimate one or more characteristic values ​​of a resin composition obtained according to the formulation data. The formulation data is data indicating the mixing ratio of each of a plurality of materials, including resin and fillers. The feature quantities are feature quantities relating to at least one of the following: the shape of the filler, the interparticle distance of the filler in the resin composition calculated using the formulation data, and the elastic modulus of the resin composition calculated using the formulation data.

[0014] The estimation method according to this embodiment can be performed by the estimation device 10.

[0015] One example is designing a compound to obtain a composite material in which fillers are dispersed in a resin matrix for a specific application. In such compound design, it is necessary to determine the combination and mixing ratio of multiple materials while matching multiple physical properties to the required characteristics. In particular, when compound design is performed including trace amounts of additives, the number of materials increases, and the cost and time required for compound design increase. Therefore, there is a need for technology that can reduce the effort required for compound design.

[0016] Firstly, from one perspective, accurately estimating the properties of the resin composition obtained with each formulation makes it easier to narrow down the candidates to a formulation that will produce a resin composition with the required properties.

[0017] In particular, when a resin composition contains fillers, the properties of that resin composition strongly depend on the filler content and the characteristics of the fillers. Furthermore, the shapes of fillers that can be used as materials vary, and even if the material is the same, using fillers with different shapes can change the properties. Therefore, by using feature quantities that quantitatively represent the characteristics related to the shape of the fillers, the accuracy of estimating the properties of the resulting resin composition (generalization performance) can be improved. In addition, by using feature quantities related to the interparticle distance of fillers in the resin composition and the theoretical elastic modulus of the resin composition containing fillers, the accuracy of estimating the properties of the resin composition (generalization performance) can be improved.

[0018] Figure 3 is a block diagram illustrating the functional configuration of the estimation device 10 according to this embodiment. In the example in Figure 3, the estimation device 10 further comprises a feature identification unit 120 and a material storage unit 121. Figure 4 is a flowchart illustrating the flow of the estimation method according to this embodiment. In the example in Figure 4, the estimation method according to this embodiment further includes a feature identification step S10. In the feature identification step S10, the feature identification unit 120 identifies features using the blending data.

[0019] The hardware configuration of the estimation device 10 is described below. Each functional component of the estimation device 10 (feature identification unit 120 and estimation unit 130) may be implemented by hardware that realizes each functional component (e.g., hardwired electronic circuits), or by a combination of hardware and software (e.g., a combination of electronic circuits and a program that controls them). The case in which each functional component of the estimation device 10 is implemented by a combination of hardware and software will be further explained below.

[0020] Figure 5 illustrates a computer 1000 for implementing the estimation device 10. Computer 1000 is any computer. For example, computer 1000 could be an SoC (System on Chip), a Personal Computer (PC), a server machine, a tablet terminal, or a smartphone. Computer 1000 may be a dedicated computer designed to implement the estimation device 10, or it may be a general-purpose computer. Furthermore, the estimation device 10 may be implemented by a single computer 1000, or by a combination of multiple computers 1000.

[0021] Computer 1000 includes a bus 1020, a processor 1040, memory 1060, a storage device 1080, an input / output interface 1100, and a network interface 1120. Bus 1020 is a data transmission path for the processor 1040, memory 1060, storage device 1080, input / output interface 1100, and network interface 1120 to send and receive data to and from each other. However, the method of connecting the processor 1040 and the other components is not limited to bus connection. The processor 1040 is a variety of processor such as a CPU (Central Processing Unit), GPU (Graphics Processing Unit), or FPGA (Field-Programmable Gate Array). Memory 1060 is a main memory device implemented using RAM (Random Access Memory), etc. Storage device 1080 is an auxiliary storage device implemented using a hard disk, SSD (Solid State Drive), memory card, or ROM (Read Only Memory), etc.

[0022] The input / output interface 1100 is an interface for connecting the computer 1000 with input / output devices. For example, input devices such as a keyboard and output devices such as a display are connected to the input / output interface 1100. The method by which the input / output interface 1100 connects to the input and output devices may be wireless or wired.

[0023] The network interface 1120 is an interface for connecting the computer 1000 to a network. This network is, for example, a LAN (Local Area Network) or a WAN (Wide Area Network). The method by which the network interface 1120 connects to the network may be wireless or wired.

[0024] The storage device 1080 stores program modules that realize each functional component of the estimation device 10. The processor 1040 reads each of these program modules into the memory 1060 and executes them to realize the function corresponding to each program module.

[0025] The resin composition for which the estimation unit 130 estimates characteristic values ​​(hereinafter also referred to as the "target resin composition") is not particularly limited, but may be, for example, a paste-like composition or a liquid composition. The application of the target resin composition is not particularly limited, but may be, for example, a paste for die bonding, a composition for insulating casting of ignition coils, a composition for insulating sealing of relays (automotive, communications, power, industrial, etc.), or a composition for insulating adhesive of electronic products.

[0026] The one or more characteristic values ​​estimated by the estimation unit 130 are not particularly limited, but the one or more characteristic values ​​estimated by the estimation unit 130 include, for example, one or more of the following: specific gravity, water absorption rate, volume resistivity, electrical conductivity, surface resistance, refractive index, relative permittivity, total light transmittance, glass transition temperature, linear expansion coefficient, volume expansion coefficient, thermal conductivity, thermal diffusivity, specific heat, die shear strength, peel strength, bending strength, tensile strength, compressive strength, dynamic modulus of elasticity, tensile modulus of elasticity, bending modulus of elasticity, elongation, and viscosity. The number of characteristic values ​​estimated by the estimation unit 130 are not particularly limited, and may be one, two or more, or three or more.

[0027] Figure 6 illustrates the structure of formulation data. The formulation data shows the mixing ratio of each of several materials. The several materials whose mixing ratios are shown in the formulation data (hereinafter also referred to as the "material group") include at least resins and fillers. The material group may include several different resins. The material group may also include several different fillers. The materials included in the material group may be identified by part numbers or the like. The mixing ratio of each material shown in the formulation data may, for example, be a value that indicates the volume ratio or mass ratio of the materials in the target resin composition.

[0028] The resin is not particularly limited, but for example, the material group may include one or more resins selected from the group consisting of epoxy resins, phenolic resins, melamine resins, unsaturated polyester resins, divinylbenzene polymers, divinylbenzene-styrene copolymers, divinylbenzene-acrylic acid ester copolymers, diallyl phthalate polymers, triallyl isocyanurate polymers, and benzoguanamine polymers.

[0029] The fillers are not particularly limited, but for example, the material group may include one or more fillers selected from the group consisting of conductive particles and non-conductive particles. Examples of conductive particles include silver, gold, copper, silica, alumina, and organic fillers. Examples of non-conductive particles include glass. The material group may include one or more fillers selected from the group consisting of silver, gold, copper, silica, alumina, organic fillers, and glass.

[0030] Other materials may include one or more selected from the group consisting of curing agents, coupling agents, curing accelerators, diluents, radical initiators, and stress reducers. The materials may or may not contain solvents. Furthermore, the materials may include compositions as materials.

[0031] As described above, the feature quantities used by the estimation unit 130 according to this embodiment relate to at least one of the following: the shape of the filler, the interparticle distance of the filler within the target resin composition, and the elastic modulus of the target resin composition. The feature quantities may be scalars or vectors combining multiple numerical values.

[0032] Examples of filler shape-related features are described below. Filler shape-related features include, for example, a shape factor. The shape factor is a value obtained by multiplying the filler particle size by the filler's specific surface area. As the filler particle size, a particle size D50 at a passing mass percentage of 50% can be used.

[0033] However, the feature quantities related to the shape of the filler are not limited to the shape factor. The feature quantities related to the shape of the filler may include one or more of the following at a transmission mass percentage of 50%: particle size D50, shape factor, equivalent circle diameter, major axis diameter, minor axis diameter, maximum major axis, particle perimeter, equivalent spherical volume, circularity, perimeter envelope, and aspect ratio. The equivalent circle diameter, major axis diameter, minor axis diameter, maximum major axis, particle perimeter, equivalent spherical volume, circularity, perimeter envelope, and aspect ratio are shape characteristic values ​​obtained by analyzing images of the particles. Specifically, one or more of the peak value, mean value, and median value in a set of shape characteristic values ​​obtained by imaging the filler and analyzing images of a predetermined number of filler particles can be included as feature quantities.

[0034] The meaning of each shape characteristic value is explained below. The equivalent circle diameter is the diameter of a circle with the same area as the imaged particle. The minor axis diameter is the smallest distance between parallel lines when the particle in the image is enclosed by two sets of parallel lines tangent to the particle. The direction perpendicular to the parallel lines when this smallest distance is taken is called the minor axis. The major axis diameter is the length of the particle in the direction perpendicular to the minor axis (major axis). The maximum major axis is the maximum distance between two points on the contour of the particle in the image. The particle perimeter is the length of the circumference of the particle. The equivalent spherical volume is a value derived by (π × equivalent circle diameter cubed) ÷ 6. The circularity is the ratio of the circumference of a circle with the same area as the imaged particle to the particle perimeter. For example, if the imaged particle is a perfect circle, the circularity is 1. The perimeter envelope is a value obtained by dividing the convex hull perimeter by the actual perimeter of the particle. For example, the smoother the contour of the imaged particle, the closer the perimeter envelope is to 1. The aspect ratio is the ratio of the major axis diameter to the minor axis diameter of a particle.

[0035] This section describes an example of a feature quantity related to the interparticle distance of fillers within a target resin composition. In a resin composition, the resin and fillers typically have different properties. For example, the thermal conductivity of the resin is low, while that of the filler is high. Consequently, heat is conducted within the resin composition by passing through the fillers. Therefore, by using the thickness of the resin between fillers, i.e., the interparticle distance, as a feature quantity, the accuracy of the thermal conductivity estimation (generalization performance) can be improved. A similar approach can be generalized and applied to the estimation of other physical properties such as how force is transmitted and how electricity is transmitted.

[0036] The interparticle distance of fillers within the target resin composition is, for example, the particle size d of the filler. p It can be calculated based on the mixing ratio of the filler. Specifically, the interparticle distance h can be calculated using the following formula (1) described in "Basic Principles of Dispersion of Fine Particles in Liquid" (Hidehiro Kamiya, Journal of the Color Materials Association, 2013, Vol. 86, No. 1, pp. 26-30). F is the volume concentration of the filler relative to the target resin composition, and can be calculated using the mixing ratio of the filler. The particle size d of the filler pFor example, a particle size D50 at a passing mass percentage of 50% can be used.

[0037]

number

[0038] This section describes examples of feature quantities related to the elastic modulus of a target resin composition. The elastic modulus of a resin composition depends on the materials of the fillers and resins, as well as the mixing ratio of the fillers and resins. Furthermore, the way force is transmitted changes depending on the shape of the fillers, which can alter the overall elastic modulus of the resin composition. Therefore, the elastic modulus can be said to quantitatively reflect the materials of the fillers and resins, the mixing ratio of the fillers and resins, and the shape of the fillers, and it is effective to use the elastic modulus as a feature quantity when estimating other physical properties. Thus, by using the theoretically calculated elastic modulus as a feature quantity, the estimation accuracy (generalization performance) of the properties of the target resin composition can be improved. Examples of elastic modulus include dynamic elastic modulus, tensile elastic modulus, and flexural elastic modulus.

[0039] The theoretical modulus of elasticity of the target resin composition can be calculated using, for example, information on the modulus of elasticity of the filler, the modulus of elasticity of the resin, the mixing ratio of the filler and the resin, and the shape of the filler. Specifically, for example, the theoretical modulus of elasticity E of the target resin composition can be calculated based on the following equations (2) to (4) described in "Mechanical Properties of High Aspect Ratio Filler-Filled Polymer Nanocomposites" (Seira Moriya (Morimune), Journal of the Adhesion Society of Japan, 2017, Vol. 53, No. 10, pp. 348-354). c It is possible to calculate this.

[0040]

number

number

number

[0041] Incidentally, V f is the volume filling ratio (vol%) of the filler, E m is the elastic modulus of the resin, and E f is the elastic modulus of the filler. When the filler is fibrous, ξ = 2×(l f / t f ), and when the filler is plate-shaped, ξ = (2 / 3)×(l f / t f ). When the filler is spherical or cubic, ξ = 2. Here, l f is the major axis diameter of the filler, and t f is the diameter of the filler when the filler is fibrous, or the thickness of the filler when the filler is plate-shaped.

[0042] Incidentally, even when estimating the elastic modulus of the target resin composition, the estimation accuracy can be improved by using the theoretical elastic modulus as a characteristic quantity.

[0043] The material storage unit 121 stores information on the materials included in the material group in advance. In the feature quantity identification step S10, the feature quantity identification unit 120 can identify the feature quantity using the blending data and the information on the materials stored in the material storage unit 121. The material storage unit 121 may be provided inside the estimation device 10 or outside the estimation device 10. When the material storage unit 121 is provided inside the estimation device 10, the material storage unit 121 is realized by using, for example, the storage device 1080 of the computer 1000 that realizes the estimation device 10.

[0044] In the estimation step S10, the estimation unit 130 can estimate one or more characteristic values using the estimation model on which machine learning has been performed.

[0045] Figure 7 is an example of the estimation model 131 used by the estimation unit 130. The estimation model 131 accepts at least formulation data and feature quantities as input. The estimation model 131 then outputs estimated characteristic values ​​(hereinafter also referred to as "estimated characteristic values") of the target resin composition corresponding to the formulation data. The target resin composition corresponding to a given formulation data refers to a resin composition obtained by mixing the materials included in the material group according to the formulation ratio shown in that formulation data.

[0046] When the estimation unit 130 estimates multiple characteristic values ​​for a single blending data set, the estimation unit 130 may use an estimation model 131 prepared for each characteristic value, or it may use an estimation model 131 capable of outputting multiple estimated characteristic values.

[0047] The estimation model 131 can accurately estimate the characteristic values ​​of the target resin composition by using the features described above.

[0048] Referring to Figures 3 and 4, the flow of the estimation method according to this embodiment will be described in detail below.

[0049] In one example, the feature identification unit 120 of the estimation device 10 acquires blending data. For example, the feature identification unit 120 can acquire blending data by reading blending data that has been previously stored in a storage unit accessible from the feature identification unit 120. This storage unit may be located inside the estimation device 10 or outside the estimation device 10. If this storage unit is located inside the estimation device 10, it may be implemented using, for example, the storage device 1080 of the computer 1000 that implements the estimation device 10. Alternatively, the feature identification unit 120 may acquire blending data from a device other than the estimation device 10, or from other functional components within the estimation device 10. Furthermore, the user may input blending data into the estimation device 10 using an input device, and the feature identification unit 120 may acquire the blending data by receiving the input.

[0050] In the feature identification step S10, the feature identification unit 120 identifies the materials to be blended into the target resin composition (i.e., materials whose blending ratio is not zero) based on the acquired blending data. The feature identification unit 120 then reads information about the identified materials from the material storage unit 121 and uses that information to identify the features. Here, the feature identification unit 120 further identifies the features using the blending ratio of each material as needed.

[0051] For example, the feature identification unit 120 can read shape coefficients, equivalent circle diameter, or major axis diameter from the material storage unit 121 and use them as feature quantities related to the shape of the filler. In addition, the feature identification unit 120 can read the particle size d of the filler from the material storage unit 121. p The system reads the data and calculates the interparticle distance based on the read information and the filler mixing ratio. The feature identification unit 120 also reads the elastic modulus of the filler, the elastic modulus of the resin, and the aspect ratio of the filler from the material storage unit 121, and calculates the theoretical elastic modulus using the read information and the mixing ratio of the filler and resin.

[0052] Furthermore, when multiple fillers, each made of different materials, are blended into the target resin composition, feature quantities can be calculated for all fillers included in the target resin composition based on the blending ratio of the multiple fillers. Specifically, for example, the feature quantity identification unit 120 can calculate a weighted average of the feature quantities related to the shape of each of the multiple fillers (e.g., equivalent diameter of a circle) based on the blending ratio of the multiple fillers, and include this as a feature quantity related to the shape of the entire filler (e.g., equivalent diameter of a circle), which can then be included in the feature quantities used for estimation.

[0053] The estimation unit 130 acquires the blending data and the features generated by the feature identification unit 120 by relating them to each other.

[0054] In estimation step S11, the estimation unit 130 can read and use the estimation model 131 that has been pre-held in a storage unit accessible from the estimation unit 130. This storage unit may be located inside the estimation device 10 or outside the estimation device 10. If this storage unit is located inside the estimation device 10, it may be implemented, for example, using the storage device 1080 of the computer 1000 that implements the estimation device 10.

[0055] The estimation unit 130 inputs the blending data and features into the estimation model 131 and outputs estimated characteristic values ​​from the estimation model 131. The estimation unit 130 outputs one or more obtained estimated characteristic values ​​as estimation results for the blending data. The estimation unit 130 may also output the blending data and the estimation results for that blending data in association.

[0056] In this embodiment, the estimation unit 130 may acquire multiple blending data and estimate one or more characteristic values ​​for each of these blending data. In that case, the estimation unit 130 may output the estimation results for the multiple blending data one by one, or it may output them all together.

[0057] The estimation unit 130 stores the estimation results in a storage unit accessible from the estimation unit 130, for example. The storage unit may be located within the estimation device 10 or outside the estimation device 10. If the storage unit is located within the estimation device 10, it may be implemented using, for example, the storage device 1080 of the computer 1000 that implements the estimation device 10.

[0058] As another example, the estimation unit 130 may output the estimation results to other functional components within the estimation device 10, or to a device other than the estimation device 10. Alternatively, the estimation unit 130 may display the estimation results on a display connected to the computer 1000 that implements the estimation device 10.

[0059] Although the example of the feature identification unit 120 identifying the feature quantities has been described above, the estimation device 10 does not necessarily have to have the feature identification unit 120 and the material storage unit 121. In that case, the estimation unit 130 can acquire the feature quantities by reading the formulation data and feature quantities that have been previously stored in a storage unit accessible from the estimation unit 130. This storage unit may be located inside the estimation device 10 or outside the estimation device 10. If this storage unit is located inside the estimation device 10, it may be implemented using, for example, the storage device 1080 of the computer 1000 that implements the estimation device 10. Alternatively, the estimation unit 130 may acquire the formulation data and feature quantities from a device other than the estimation device 10, or from other functional components within the estimation device 10. Furthermore, the user may input the formulation data and feature quantities into the estimation device 10 using an input device, and the estimation unit 130 may acquire the information by receiving the input.

[0060] Figure 8 is a flowchart illustrating the overview of the model generation method according to this embodiment. The model generation method according to this embodiment is executed by one or more computers. The model generation method according to this embodiment includes a model generation step S21. In the model generation step S21, an estimated model 131 is generated by performing machine learning using training data. The training data includes formulation data, feature quantities, and one or more characteristic values ​​of the resin composition obtained according to the formulation data. The formulation data indicates the mixing ratio of each of a plurality of materials including resin and filler. The feature quantities are feature quantities relating to at least one of the following: the shape of the filler, the interparticle distance of the filler in the resin composition calculated using the formulation data, and the elastic modulus of the resin composition calculated using the formulation data.

[0061] The estimated model generated by the model generation method according to this embodiment can be used as the estimated model 131 in the estimation device 10.

[0062] Figure 9 is a block diagram illustrating the functional configuration of the model generation device 20 according to this embodiment. The model generation device 20 according to this embodiment includes a model generation unit 210. The model generation unit 210 generates an estimated model by performing machine learning using training data. The training data includes formulation data, feature quantities, and one or more characteristic values ​​of the resin composition obtained according to the formulation data. The formulation data indicates the mixing ratio of each of a plurality of materials, including resin and filler. The feature quantities are feature quantities relating to at least one of the following: the shape of the filler, the interparticle distance of the filler in the resin composition calculated using the formulation data, and the elastic modulus of the resin composition calculated using the formulation data.

[0063] The model generation method according to this embodiment can be executed by the model generation device 20.

[0064] According to the model generation method and model generation apparatus 20 of this embodiment, an estimated model 131 is obtained that can estimate properties with high accuracy using feature quantities relating to at least one of the filler shape, the interparticle distance of the filler, and the elastic modulus.

[0065] The hardware configuration of the computer implementing the model generation device 20 according to this embodiment is shown, for example, in Figure 5, similar to the estimation device 10. However, the storage device 1080 of the computer 1000 implementing the model generation device 20 according to this embodiment stores program modules that implement each functional component (model generation unit 210) of the model generation device 20. The processor 1040 of the computer 1000 implementing the model generation device 20 implements the functions corresponding to each program module by reading these program modules into the memory 1060 and executing them. Furthermore, the model generation device 20 may be implemented by a single computer 1000, or by a combination of multiple computers 1000. One or more computers 1000 for implementing the model generation device 20 may also serve as one or more computers 1000 for implementing the estimation device 10.

[0066] The formulation data and features included in the training data are as described above.

[0067] One or more characteristic values ​​included in the training data (hereinafter also referred to as "ground truth characteristic values") are used as ground truth data in machine learning. An example of a ground truth characteristic value is as exemplified by the one or more characteristic values ​​estimated by the estimation unit 130. A ground truth characteristic value can be obtained, for example, by measuring the properties of a resin composition prepared in advance according to formulation data. A resin composition prepared according to formulation data means a resin composition obtained by mixing the materials included in the material group according to the formulation ratio shown in the formulation data.

[0068] The flow of the model generation method according to this embodiment will be described in detail below. In one example, the model generation unit 210 of the model generation device 20 acquires training data. For example, the model generation unit 210 can acquire training data by reading training data that has been previously stored in a storage unit accessible from the model generation unit 210. This storage unit may be located inside the model generation device 20 or outside the model generation device 20. If this storage unit is located inside the model generation device 20, it is implemented, for example, using the storage device 1080 of the computer 1000 that implements the model generation device 20. Alternatively, the model generation unit 210 may acquire training data from a device other than the model generation device 20, or from other functional components within the model generation device 20.

[0069] As another example, the model generation unit 210 may acquire a set of combination data and ground truth characteristic values, and then identify features from the combination data in the same manner as the feature identification unit 120. The model generation unit 210 can generate training data by combining the identified features, the combination data, and the ground truth characteristic values. The model generation unit 210 can obtain the information necessary for generating training data by reading it from a storage unit accessible from the model generation unit 210 as described above.

[0070] The model generation unit 210 can perform machine learning using existing technologies. Specifically, the model generation unit 210 inputs the compound data and features included in the training data into the estimation model 131. Then, it updates several parameters for configuring the estimation model 131 so that the difference between the estimated characteristic values ​​output from the estimation model 131 and the correct characteristic values ​​included in the training data becomes smaller.

[0071] The model generation unit 210 similarly acquires multiple training data sets and repeats parameter updates until the termination condition is met. The termination condition is, for example, that a predetermined number of parameter updates have been performed, or that the difference between the estimated characteristic value and the correct characteristic value included in the training data falls below a predetermined threshold.

[0072] The model generation unit 210 can store the trained estimated model 131 in a storage unit accessible from the model generation unit 210. This storage unit may be located inside the model generation device 20 or outside the model generation device 20. If this storage unit is located inside the model generation device 20, it may be implemented, for example, using the storage device 1080 of the computer 1000 that implements the model generation device 20.

[0073] According to this embodiment, highly accurate property estimation is possible by using feature quantities relating to at least one of the following: the shape of the filler, the interparticle distance of the filler in the resin composition calculated using the compounding data, and the elastic modulus of the resin composition calculated using the compounding data.

[0074] (Second embodiment) Figure 10 is a diagram illustrating the functional configuration of the estimation device 10 according to the second embodiment. The estimation device 10 according to this embodiment is the same as the estimation device 10 according to the first embodiment, except that it further comprises a blending data generation unit 110 and an extraction unit 150.

[0075] Figure 11 is a flowchart illustrating the flow of the estimation method according to this embodiment. The estimation method according to the second embodiment includes a blending data generation step S30, a feature identification step S31, an estimation step S32, and an extraction step S33. The feature identification step S31 and the estimation step S32 according to the second embodiment are the same as the feature identification step S10 and the estimation step S11 according to the first embodiment, respectively.

[0076] The estimation method according to the second embodiment is the same as the estimation method according to the first embodiment, except that it includes a blending data generation step S30 and an extraction step S33. In the blending data generation step S30, the blending data generation unit 110 acquires constraint information regarding the blend and generates a plurality of blending data based on the constraint information.

[0077] In the estimation step S32 of this embodiment, the estimation unit 130 acquires a plurality of blending data generated by the blending data generation unit 110 and estimates one or more characteristic values ​​for each blending data, as described in the first embodiment. In the extraction step S33, the extraction unit 150 extracts blending data from the plurality of blending data in which the estimated one or more characteristic values ​​satisfy the required characteristics. The estimation device 10 and estimation method according to this embodiment will be described in detail below.

[0078] The hardware configuration of the computer implementing the estimation device 10 according to this embodiment is shown, for example, in Figure 5, similar to the estimation device 10. However, the storage device 1080 of the computer 1000 implementing the estimation device 10 according to this embodiment further stores program modules that realize the functions of the blending data generation unit 110 and the extraction unit 150.

[0079] Figure 12 illustrates the structure of constraint information. Constraint information indicates constraints on the composition. For example, constraint information indicates the range of permissible mixing ratios for each material included in the material group.

[0080] In the formulation data generation step S30, the formulation data generation unit 110 of the estimation device 10 can acquire constraint condition information by reading it from a storage unit that is accessible from the formulation data generation unit 110 in advance. This storage unit may be located inside the estimation device 10 or outside the estimation device 10. If this storage unit is located inside the estimation device 10, it may be implemented using, for example, the storage device 1080 of the computer 1000 that implements the estimation device 10. Alternatively, the formulation data generation unit 110 may acquire constraint condition information from a device other than the estimation device 10, or from other functional components within the estimation device 10. Furthermore, the user may input constraint condition information into the estimation device 10 using an input device, and the formulation data generation unit 110 may acquire the information by receiving the input constraint condition information.

[0081] The formulation data generation unit 110 generates formulation data that satisfies the constraints indicated in the acquired constraint information. That is, for each material included in the material group, the formulation data generation unit 110 identifies the blending ratio within the range indicated in the constraints. In this way, the formulation data generation unit 110 generates multiple formulation data. The number of formulation data generated by the formulation data generation unit 110 may be predetermined or may be input to the estimation device 10 by the user. For example, the formulation data generation unit 110 comprehensively generates multiple formulation data while determining the blending ratio of each material at intervals corresponding to the number of formulation data to be generated.

[0082] In the feature identification step S31, the feature identification unit 120 acquires multiple blending data generated by the blending data generation unit 110. Then, in the same manner as described in the first embodiment, it identifies features for each of the multiple blending data. The feature identification unit 120 associates the features identified based on each blending data with that blending data.

[0083] In estimation step S32, the estimation unit 130 inputs each of the multiple sets of blending data and features into the estimation model 131 and outputs one or more estimated characteristic values. In this way, one or more estimated characteristic values ​​are obtained for each of the multiple sets of blending data. The estimation unit 130 associates one or more estimated characteristic values ​​with each set of blending data.

[0084] In extraction step S33, the extraction unit 150 obtains multiple sets from the estimation unit 130, each set consisting of a combination of related blending data and one or more estimated characteristic values.

[0085] Furthermore, the extraction unit 150 acquires requirement information that indicates the required characteristics. Figure 13 is a diagram illustrating the structure of the requirement information. The requirement information indicates the conditions that each of the one or more estimated characteristic values ​​must satisfy.

[0086] The extraction unit 150 can acquire the request information, for example, by reading the request information that has been previously stored in a storage unit accessible from the extraction unit 150. This storage unit may be located inside the estimation device 10 or outside the estimation device 10. If this storage unit is located inside the estimation device 10, it may be implemented using, for example, the storage device 1080 of the computer 1000 that implements the estimation device 10. Alternatively, the extraction unit 150 may acquire the request information from a device other than the estimation device 10, or from other functional components within the estimation device 10. Furthermore, the user may input the request information into the estimation device 10 using an input device, and the extraction unit 150 may acquire the request information by receiving the input.

[0087] The extraction unit 150 determines, for each of the multiple sets of blending data and one or more estimated characteristic values ​​obtained from the estimation unit 130, whether or not the one or more estimated characteristic values ​​satisfy all the conditions indicated in the request information. Then, the extraction unit 150 extracts one or more sets from the multiple sets in which the one or more estimated characteristic values ​​satisfy all the conditions indicated in the request information.

[0088] The extraction unit 150 outputs one or more extracted sets. The extraction unit 150 stores the one or more extracted sets in a storage unit accessible from the extraction unit 150, for example. The storage unit may be located inside the estimation device 10 or outside the estimation device 10. If the storage unit is located inside the estimation device 10, it may be implemented using, for example, the storage device 1080 of the computer 1000 that implements the estimation device 10.

[0089] As another example, the extraction unit 150 may output one or more extracted sets to other functional components within the estimation device 10, or to a device other than the estimation device 10. Alternatively, the extraction unit 150 may display one or more extracted sets on a display connected to the computer 1000 that implements the estimation device 10.

[0090] Furthermore, if there is no set among the multiple sets in which one or more estimated characteristic values ​​satisfy all the conditions indicated in the request information, the extraction unit 150 may output information indicating that no sets were extracted.

[0091] According to this embodiment, the same functions and effects as in the first embodiment can be obtained. In addition, according to this embodiment, formulation data that conforms to the constraints and requirements for the characteristics of the formulation can be obtained.

[0092] (Third embodiment) Figure 14 is a diagram illustrating the functional configuration of the estimation device 10 according to the third embodiment. The estimation device 10 according to this embodiment is the same as the estimation device 10 according to at least one of the first and second embodiments, except that it further comprises a clustering unit 170.

[0093] Figure 15 is a flowchart illustrating the flow of the estimation method according to this embodiment. In the example in Figure 15, the estimation method according to this embodiment includes a blending data generation step S40, a feature identification step S41, an estimation step S42, an extraction step S43, a clustering step S44, and a representative blending data extraction step S45. The blending data generation step S40, the feature identification step S41, the estimation step S42, and the extraction step S43 according to this embodiment are the same as the blending data generation step S30, the feature identification step S31, the estimation step S32, and the extraction step S33 according to the second embodiment, respectively.

[0094] The estimation method according to the third embodiment is the same as the estimation method according to at least one of the first and second embodiments, except that it includes a clustering step S44 and a representative blend data extraction step S45. In the clustering step S44, the clustering unit 170 clusters a plurality of blend data into a plurality of clusters. In the representative blend data extraction step S45, the clustering unit 170 extracts one or more blend data from each of the plurality of clusters. The estimation device 10 and estimation method according to this embodiment will be described in detail below.

[0095] For example, a prototype experiment may be conducted to determine whether the resin composition obtained with the formulation shown in the formulation data actually possesses the properties predicted for the set (formulation data and one or more estimated characteristic values) output from the estimation device 10. If the extraction unit 150 extracts a large number of sets, conducting a prototype experiment for all of them would require considerable effort.

[0096] In contrast, in the estimation device 10 according to this embodiment, the clustering unit 170 clusters multiple blending data, classifying the blending data into multiple clusters based on the similarity between the blending data. Then, one or more blending data are extracted from each of the multiple clusters as representative blending data. By conducting prototype experiments with the representative blending data extracted in this way, the desired blending data can be searched with less effort than if prototype experiments were conducted for all blending data. Furthermore, because representative blending data is extracted from each cluster, a diverse range of blending data can be used as the subject of prototype experiments without being biased towards blending data of the same type.

[0097] The hardware configuration of the computer implementing the estimation device 10 according to this embodiment is shown, for example, in Figure 5, similar to the estimation device 10. However, the storage device 1080 of the computer 1000 implementing the estimation device 10 according to this embodiment further stores program modules that realize the functions of the clustering unit 170.

[0098] In the clustering step S44, the clustering unit 170 acquires multiple sets extracted by the extraction unit 150. Then, it extracts blended data from each of the multiple sets. By performing clustering on the set of multiple extracted blended data, multiple clusters are generated. Clustering can be performed by executing an existing clustering process, such as the Diana method, Ward's method, group average method, shortest distance method, centroid method, weighted average method, median method, k-means method, k-means++ method, x-means method, k-medoids method, hidden Markov model, DBSCAN, fuzzy C-means method, or NMF. Each of the multiple clusters contains two or more blended data.

[0099] In the representative blend data extraction step S45, the clustering unit 170 extracts one or more blend data from each of the multiple clusters as representative blend data. For example, the clustering unit 170 extracts a predetermined number of blend data from each cluster as representative blend data. The clustering unit 170 may extract one blend data from each of the multiple clusters, or two or more from each cluster. Alternatively, the clustering unit 170 may extract a predetermined proportion of blend data to the number of blend data included in the cluster. In this way, the clustering unit 170 can extract multiple blend data as representative blend data.

[0100] If the number of combination data points included in a cluster is greater than the number of combination data points to be extracted from that cluster, the clustering unit 170 extracts some of the multiple combination data points included in each cluster as representative combination data, and does not extract the rest. In this way, it is possible to extract fewer representative combination data points than the number of sets extracted by the extraction unit 150.

[0101] The clustering unit 170 outputs multiple representative blend data (i.e., multiple extracted blend data). The clustering unit 170 stores the extracted multiple representative blend data in a storage unit accessible from the clustering unit 170, for example. The storage unit may be located inside the estimation device 10 or outside the estimation device 10. If the storage unit is located inside the estimation device 10, this storage unit is implemented, for example, using the storage device 1080 of the computer 1000 that implements the estimation device 10.

[0102] As another example, the clustering unit 170 may output the extracted representative blend data to other functional components within the estimation device 10, or to a device other than the estimation device 10. Alternatively, the clustering unit 170 may display the extracted representative blend data on a display connected to the computer 1000 that implements the estimation device 10.

[0103] The clustering unit 170 may further output one or more estimated characteristic values ​​associated with each representative blend data.

[0104] However, the clustering unit 170 according to this embodiment may output the clustering results in the manner described above without performing the representative blending data extraction step S45. In that case, the clustering unit 170 outputs multiple blending data in a state that allows the cluster to which each blending data belongs to to be identified. In that case, the user can select blending data from each cluster.

[0105] The estimation device 10 according to this embodiment does not necessarily have to include a blending data generation unit 110. Also, the estimation method according to this embodiment does not necessarily have to include the blending data generation step S40. In this case, the estimation unit 130 acquires a plurality of pre-prepared blending data and estimates characteristic values, similar to the method described in the first embodiment. Also, the estimation method according to this embodiment does not necessarily have to include the feature identification step S41. In this case, the estimation unit 130 acquires a plurality of pre-prepared blending data and feature quantities and estimates characteristic values, similar to the method described in the first embodiment.

[0106] According to this embodiment, the same functions and effects as in the first embodiment can be obtained. In addition, according to this embodiment, the clustering unit 170 clusters multiple formulation data into multiple clusters. Therefore, it is possible to search for desired formulation data while grasping the overall picture with less effort than when confirmation experiments are performed for all formulation data.

[0107] (Fourth embodiment) Figure 16 is a diagram illustrating the functional configuration of the estimation device 10 according to the fourth embodiment. The estimation device 10 according to this embodiment is the same as the estimation device 10 according to at least one of the first to third embodiments, except for the points described below.

[0108] Figure 17 is a flowchart illustrating the flow of the estimation method according to this embodiment. In the example in Figure 17, the estimation method according to this embodiment includes a blending data generation step S50, a feature identification step S51, an estimation step S52, an extraction step S53, a clustering step S54, and a representative blending data extraction step S55, a preference calculation step S56, and a blending output step S57. The blending data generation step S50, the feature identification step S51, the estimation step S52, the extraction step S53, the clustering step S54, and the representative blending data extraction step S55 according to this embodiment are the same as the blending data generation step S40, the feature identification step S41, the estimation step S42, the extraction step S43, the clustering step S44, and the representative blending data extraction step S45 according to the third embodiment.

[0109] The estimation method according to the fourth embodiment is the same as the estimation method according to at least one of the first to third embodiments, except for the points described below.

[0110] In the estimation step S52 of this embodiment, the estimation unit 130 estimates multiple characteristic values ​​of the resin composition. Then, in the preference calculation step S56, the preference calculation unit 190 calculates the preference for the formulation data using the multiple characteristic values ​​and the weight coefficients for each of the multiple characteristic values. Once the preference is calculated, it becomes easier to understand which formulation data is particularly preferable, taking into account the importance of each characteristic value. The estimation device 10 and estimation method according to this embodiment will be described in detail below.

[0111] The hardware configuration of the computer implementing the estimation device 10 according to this embodiment is shown in Figure 5, for example, similar to the estimation device 10. However, the storage device 1080 of the computer 1000 implementing the estimation device 10 according to this embodiment further stores a program module that implements the functions of the preference calculation unit 190.

[0112] The preference calculation unit 190 according to this embodiment acquires a plurality of representative blend data extracted by the clustering unit 170. The preference calculation unit 190 further acquires a plurality of estimated characteristic values ​​associated with each of the plurality of representative blend data.

[0113] Furthermore, the preference calculation unit 190 acquires weight information that indicates weight coefficients for each of the multiple characteristic values. Figure 18 is a diagram illustrating the configuration of the weight information. The weight information includes, for example, weight coefficients for viscosity, weight coefficients for thermal conductivity, and so on.

[0114] The preference calculation unit 190 can obtain weight information, for example, by reading it from a storage unit that is accessible from the preference calculation unit 190 in advance. This storage unit may be located inside the estimation device 10 or outside the estimation device 10. If this storage unit is located inside the estimation device 10, it may be implemented using, for example, the storage device 1080 of the computer 1000 that implements the estimation device 10. Alternatively, the preference calculation unit 190 may obtain the weight information from a device other than the estimation device 10, or from other functional components within the estimation device 10.

[0115] As another example, the user may input weight coefficients to the estimation device 10 using an input device, and the preference calculation unit 190 may receive and obtain the input weight coefficients.

[0116] Figure 19 is a flowchart showing a modified example of the estimation device 10 according to this embodiment. Figure 19 shows an example in which a user inputs weight coefficients to the estimation device 10. The estimation method according to this modified example further includes a display data output step S561 and a weight coefficient acquisition step S562. In the display data output step S561, the preference calculation unit 190 outputs display data to display the weight coefficient input screen. Then, the weight coefficient input screen is displayed on the display connected to 10 using this display data. The user inputs the weight coefficients for each characteristic value on the input screen. In the weight coefficient acquisition step S562, the preference calculation unit 190 acquires the weight coefficients entered on the input screen. Then, in the preference calculation step S56, the preference calculation unit 190 calculates the preference using the weight coefficients acquired in the weight coefficient acquisition step S562. In this way, the user of the estimation device 10 can arbitrarily set the importance of each characteristic value and reflect it in the preference.

[0117] In the preference calculation step S56 of this embodiment, the preference calculation unit 190 calculates the preference A using, for example, the following formula (5). Here, i is the characteristic value p i This is a number that identifies each item (viscosity, thermal conductivity, etc.), w i is the characteristic value p i This is the weighting coefficient for P i The standardized characteristic value p i That is. Also, Σw i The value = 1 holds true.

[0118]

number

[0119] Methods for standardizing characteristic values ​​are not particularly limited, but one example is Z-scaling. Specifically, characteristic value p i The mean value μ for the following: i and standard deviation σ i Using P i =(p i -μ i ) / σ i Using the formula, the standardized Pi It is possible to find this.

[0120] The preference calculation unit 190 uses the average value μ i and standard deviation σ i Each of the multiple characteristic values ​​p estimated by the estimation unit 130 in estimation step S52 is represented as follows: i The mean and standard deviation calculated using the population can be used. Alternatively, the preference calculation unit 190 can use the mean μ i and standard deviation σ i Alternatively, the mean and standard deviation calculated from multiple characteristic values ​​stored in an existing database may be used as the population. The existing database may include measured values ​​as characteristic values. Alternatively, the preference calculation unit 190 may use the mean μ i and standard deviation σ i A predetermined value may be used.

[0121] The preference calculation unit 190 outputs at least one blending data. The destination of the blending data output from the preference calculation unit 190 is not particularly limited, but for example, the preference calculation unit 190 stores the blending data in a storage unit accessible from the preference calculation unit 190. The storage unit may be located inside the estimation device 10 or outside the estimation device 10. If the storage unit is located inside the estimation device 10, this storage unit is implemented, for example, using the storage device 1080 of the computer 1000 that implements the estimation device 10.

[0122] As another example, the preference calculation unit 190 may output the blending data to other functional components within the estimation device 10, or to a device other than the estimation device 10. Alternatively, the preference calculation unit 190 may display the blending data on a display connected to the computer 1000 that implements the estimation device 10.

[0123] Further explanation will be given regarding an example of how to output data from the preference calculation unit 190.

[0124] In the first example, in the blending output step S57, the preference calculation unit 190 outputs the blending data and the calculated preference. The preference calculation unit 190 may further output the blending data along with one or more corresponding estimated characteristic values. This allows the user to understand how promising the outputted blending data is. The preference calculation unit 190 may output one set of blending data and preference, or multiple sets of blending data and preference.

[0125] In the second example, in estimation step S52, the estimation unit 130 estimates multiple characteristic values ​​for each of the multiple blending data, and in blending output step S57, the preference calculation unit 190 outputs the multiple blending data in order based on preference. The preference calculation unit 190 can, for example, output a table showing the multiple blending data arranged in order based on preference. When displaying multiple blending data on a display, the preference calculation unit 190 may display the multiple blending data one by one in order based on preference. In this way, the user can easily grasp the ranking of the multiple blending data.

[0126] In the third example, in estimation step S52, the estimation unit 130 estimates multiple characteristic values ​​for each of the multiple blending data, and in blending output step S57, the preference calculation unit 190 outputs one or more blending data extracted from the multiple blending data based on preference. For example, the preference calculation unit 190 extracts and outputs a predetermined number (e.g., the top three blending data) of blending data with high preference from the multiple blending data. Alternatively, the preference calculation unit 190 may extract and output blending data whose preference is above a predetermined threshold. Alternatively, the preference calculation unit 190 extracts and outputs blending data with high preference in a predetermined proportion (e.g., the top 10% of blending data) from the multiple blending data. In this way, the user can narrow down and check only the promising blending data.

[0127] The preference calculation unit 190 may also output data in a manner that combines the first to third examples described above.

[0128] Referring to FIGS. 17 and 19, an example in which both clustering and calculation of preference are performed has been described above, but the calculation of preference may be performed without performing clustering. That is, the preference calculation step S56 or the display data output step S561 may be performed after the extraction step S53. In that case, the preference calculation unit 190 calculates the preference for the formulation data extracted in the extraction step S53.

[0129] Further, the estimation device 10 according to the present embodiment may not include the formulation data generation unit 110. Further, the estimation method according to the present embodiment may not include the formulation data generation step S50. In this case, the estimation unit 130 acquires a plurality of pre-prepared formulation data and estimates characteristic values, as described in the first embodiment. Further, the estimation method according to the present embodiment may not include the feature quantity identification step S51. In this case, the estimation unit 130 acquires a plurality of pre-prepared formulation data and feature quantities and estimates characteristic values, as described in the first embodiment.

[0130] As described above, embodiments of the present invention have been described with reference to the drawings, but these are examples of the present invention, and various configurations other than those described above can also be adopted.

[0131] In addition, in the plurality of flowcharts used in the above description, a plurality of steps (processes) are described in order, but the execution order of the steps executed in each embodiment is not limited to the described order. In each embodiment, the order of the illustrated steps can be changed within a range that does not substantially affect the content. In addition, the above-described embodiments can be combined within a range where the contents do not conflict.

Description of Reference Numerals

[0132] 10 Estimation device 20 Model generation device 110 Formulation data generation unit 120 Feature quantity identification unit 121 Material storage unit 130 Estimation unit 131 Estimated Models 150 Extraction part 170 Clustering section 190 Preference calculation part 210 Model Generation Unit 1000 calculator 1020 Bus 1040 processor 1060 memory 1080 Storage Devices 1100 Input / Output Interface 1120 Network Interface

Claims

1. An estimation method performed by one or more computers, The process includes an estimation step in which formulation data indicating the mixing ratio of each of several materials, including resin and filler, and feature quantities are input into a machine learning-based estimation model, and one or more characteristic values ​​of the resin composition obtained according to the formulation data are output from the estimation model, thereby estimating the one or more characteristic values. The aforementioned characteristic quantity includes at least one of the following: the interparticle distance of the filler in the resin composition calculated using the formulation data, and a value obtained by multiplying the particle size of the filler by the specific surface area of ​​the filler. The one or more characteristic values ​​mentioned above include one or more of the following: specific gravity, water absorption rate, diffusion coefficient, volume resistivity, electrical conductivity, surface resistance, refractive index, relative permittivity, total light transmittance, glass transition temperature, linear expansion coefficient, volume expansion coefficient, thermal conductivity, thermal diffusivity, specific heat, die shear strength, peel strength, flexural strength, tensile strength, compressive strength, dynamic modulus, tensile modulus, flexural modulus, elongation, and viscosity. Estimation method.

2. An estimation method performed by one or more computers, An estimation step in which formulation data indicating the blending ratio of each of several materials including resin and filler, and feature quantities are input into an estimation model that has been subjected to machine learning, and one or more characteristic values ​​of the resin composition obtained according to the formulation data are output from the estimation model, thereby estimating the one or more characteristic values. A clustering step of clustering multiple sets of aforementioned formulation data into multiple clusters, A representative blending data extraction step in which one or more of the blending data are extracted from each of the aforementioned clusters as representative blending data, The step includes outputting the extracted representative blend data, The aforementioned feature quantity relates to at least one of the following: the shape of the filler, the interparticle distance of the filler in the resin composition calculated using the formulation data, and the elastic modulus of the resin composition calculated using the formulation data. The one or more characteristic values ​​mentioned above include one or more of the following: specific gravity, water absorption rate, diffusion coefficient, volume resistivity, electrical conductivity, surface resistance, refractive index, relative permittivity, total light transmittance, glass transition temperature, linear expansion coefficient, volume expansion coefficient, thermal conductivity, thermal diffusivity, specific heat, die shear strength, peel strength, flexural strength, tensile strength, compressive strength, dynamic modulus, tensile modulus, flexural modulus, elongation, and viscosity. Estimation method.

3. An estimation method performed by one or more computers, An estimation step in which formulation data indicating the mixing ratio of each of several materials including resin and filler, and feature quantities are input into an estimation model that has been subjected to machine learning, and several characteristic values ​​of the resin composition obtained according to the formulation data are output from the estimation model, thereby estimating the several characteristic values, The process includes a preference calculation step that calculates the preference for the blending data using the plurality of characteristic values ​​and a weight coefficient for each of the plurality of characteristic values, The aforementioned feature quantity relates to at least one of the following: the shape of the filler, the interparticle distance of the filler in the resin composition calculated using the formulation data, and the elastic modulus of the resin composition calculated using the formulation data. The aforementioned multiple characteristic values ​​include one or more of the following: specific gravity, water absorption rate, diffusion coefficient, volume resistivity, electrical conductivity, surface resistance, refractive index, relative permittivity, total light transmittance, glass transition temperature, linear expansion coefficient, volume expansion coefficient, thermal conductivity, thermal diffusivity, specific heat, die shear strength, peel strength, flexural strength, tensile strength, compressive strength, dynamic modulus, tensile modulus, flexural modulus, elongation, and viscosity. Estimation method.

4. In the estimation method described in claim 3, A display data output step that outputs display data for displaying the input screen for the weight coefficients, The process further includes a weight coefficient acquisition step, which involves acquiring the weight coefficient entered on the input screen, In the preference calculation step, the preference is calculated using the weight coefficients obtained in the weight coefficient acquisition step. Estimation method.

5. In the estimation method described in claim 3, The step further includes outputting the aforementioned blending data and the calculated preference score. Estimation method.

6. In the estimation method described in claim 3, In the estimation step, the multiple characteristic values ​​are estimated for each of the multiple blending data, The step further includes outputting the plurality of formulation data in an order based on the preference. Estimation method.

7. In the estimation method described in claim 3, In the estimation step, the multiple characteristic values ​​are estimated for each of the multiple blending data, The system further comprises a formulation output step that outputs one or more formulation data extracted from the plurality of formulation data based on the preference. Estimation method.

8. The system includes an estimation unit that inputs formulation data indicating the blending ratio of each of several materials, including resin and filler, and feature quantities, into a machine learning-based estimation model, and outputs one or more characteristic values ​​of the resin composition obtained according to the formulation data from the estimation model, thereby estimating one or more characteristic values. The aforementioned characteristic quantity includes at least one of the following: the interparticle distance of the filler in the resin composition calculated using the formulation data, and a value obtained by multiplying the particle size of the filler by the specific surface area of ​​the filler. The one or more characteristic values ​​mentioned above include one or more of the following: specific gravity, water absorption rate, diffusion coefficient, volume resistivity, electrical conductivity, surface resistance, refractive index, relative permittivity, total light transmittance, glass transition temperature, linear expansion coefficient, volume expansion coefficient, thermal conductivity, thermal diffusivity, specific heat, die shear strength, peel strength, flexural strength, tensile strength, compressive strength, dynamic modulus, tensile modulus, flexural modulus, elongation, and viscosity. Estimation device.

9. An estimation unit that inputs formulation data indicating the blending ratio of each of several materials including resin and filler, and feature quantities, into an estimation model that has undergone machine learning, and outputs one or more characteristic values ​​of the resin composition obtained according to the formulation data from the estimation model, thereby estimating one or more characteristic values. The system includes a clustering unit that clusters multiple sets of aforementioned formulation data into multiple clusters, The clustering unit, From each of the aforementioned clusters, one or more of the aforementioned blending data are extracted as representative blending data. The extracted representative blend data is output, The aforementioned feature quantity relates to at least one of the following: the shape of the filler, the interparticle distance of the filler in the resin composition calculated using the formulation data, and the elastic modulus of the resin composition calculated using the formulation data. The one or more characteristic values ​​mentioned above include one or more of the following: specific gravity, water absorption rate, diffusion coefficient, volume resistivity, electrical conductivity, surface resistance, refractive index, relative permittivity, total light transmittance, glass transition temperature, linear expansion coefficient, volume expansion coefficient, thermal conductivity, thermal diffusivity, specific heat, die shear strength, peel strength, flexural strength, tensile strength, compressive strength, dynamic modulus, tensile modulus, flexural modulus, elongation, and viscosity. Estimation device.

10. An estimation unit that inputs formulation data indicating the blending ratio of each of several materials including resin and filler, and feature quantities, into an estimation model that has undergone machine learning, and outputs several characteristic values ​​of the resin composition obtained according to the formulation data from the estimation model, thereby estimating the several characteristic values. The system includes a preference calculation unit that calculates a preference for the blending data using the plurality of characteristic values ​​and a weight coefficient for each of the plurality of characteristic values, The aforementioned feature quantity relates to at least one of the following: the shape of the filler, the interparticle distance of the filler in the resin composition calculated using the formulation data, and the elastic modulus of the resin composition calculated using the formulation data. The aforementioned multiple characteristic values ​​include one or more of the following: specific gravity, water absorption rate, diffusion coefficient, volume resistivity, electrical conductivity, surface resistance, refractive index, relative permittivity, total light transmittance, glass transition temperature, linear expansion coefficient, volume expansion coefficient, thermal conductivity, thermal diffusivity, specific heat, die shear strength, peel strength, flexural strength, tensile strength, compressive strength, dynamic modulus, tensile modulus, flexural modulus, elongation, and viscosity. Estimation device.

11. Computers, Formulation data indicating the mixing ratio of each of several materials, including resin and filler, and feature quantities are input into a machine learning estimation model, and one or more characteristic values ​​of the resin composition obtained according to the formulation data are output from the estimation model, thereby functioning as an estimation unit that estimates the one or more characteristic values. The aforementioned characteristic quantity includes at least one of the following: the interparticle distance of the filler in the resin composition calculated using the formulation data, and a value obtained by multiplying the particle size of the filler by the specific surface area of ​​the filler. The one or more characteristic values ​​mentioned above include one or more of the following: specific gravity, water absorption rate, diffusion coefficient, volume resistivity, electrical conductivity, surface resistance, refractive index, relative permittivity, total light transmittance, glass transition temperature, linear expansion coefficient, volume expansion coefficient, thermal conductivity, thermal diffusivity, specific heat, die shear strength, peel strength, flexural strength, tensile strength, compressive strength, dynamic modulus, tensile modulus, flexural modulus, elongation, and viscosity. program.

12. Computers, An estimation unit that inputs formulation data indicating the blending ratio of each of several materials including resin and filler, and feature quantities, into an estimation model that has been subjected to machine learning, and outputs one or more characteristic values ​​of the resin composition obtained according to the formulation data from the estimation model, thereby estimating one or more characteristic values, and This unit functions as a clustering unit that clusters multiple sets of the aforementioned formulation data into multiple clusters. The clustering unit, From each of the aforementioned clusters, one or more of the aforementioned blending data are extracted as representative blending data. The extracted representative blend data is output, The aforementioned feature quantity relates to at least one of the following: the shape of the filler, the interparticle distance of the filler in the resin composition calculated using the formulation data, and the elastic modulus of the resin composition calculated using the formulation data. The one or more characteristic values ​​mentioned above include one or more of the following: specific gravity, water absorption rate, diffusion coefficient, volume resistivity, electrical conductivity, surface resistance, refractive index, relative permittivity, total light transmittance, glass transition temperature, linear expansion coefficient, volume expansion coefficient, thermal conductivity, thermal diffusivity, specific heat, die shear strength, peel strength, flexural strength, tensile strength, compressive strength, dynamic modulus, tensile modulus, flexural modulus, elongation, and viscosity. program.

13. Computers, An estimation unit that inputs formulation data indicating the blending ratio of each of several materials including resin and filler, and feature quantities, into an estimation model that has been subjected to machine learning, and outputs several characteristic values ​​of the resin composition obtained according to the formulation data from the estimation model, thereby estimating the several characteristic values, and The unit functions as a preference calculation unit that calculates the preference for the blending data using the aforementioned multiple characteristic values ​​and weight coefficients for each of the aforementioned multiple characteristic values. The aforementioned feature quantity relates to at least one of the following: the shape of the filler, the interparticle distance of the filler in the resin composition calculated using the formulation data, and the elastic modulus of the resin composition calculated using the formulation data. The aforementioned multiple characteristic values ​​include one or more of the following: specific gravity, water absorption rate, diffusion coefficient, volume resistivity, electrical conductivity, surface resistance, refractive index, relative permittivity, total light transmittance, glass transition temperature, linear expansion coefficient, volume expansion coefficient, thermal conductivity, thermal diffusivity, specific heat, die shear strength, peel strength, flexural strength, tensile strength, compressive strength, dynamic modulus, tensile modulus, flexural modulus, elongation, and viscosity. program.

Images