Prediction system, prediction device, and prediction method

Abstract

To provide a forecasting system capable of forecasting a reaction ratio between molecules forming a resin.SOLUTION: A forecasting system 100 comprises: a forecasting device 20 for forecasting a reaction ratio between reaction groups in reaction of reaction groups included in a monomer which is a molecule forming a resin. The forecasting device 20 comprises: a DB 222 which associates a first index related to activation energy defined by a formula of Arrhenius of the monomer, with a second index related to a frequency factor defined by the formula of Arrhenius of the monomer, and the reaction ratio; and a forecasting unit 214 for forecasting the reaction ratio from the first index and second index of the monomer determined from the molecule information related to the monomer, and the DB 222.SELECTED DRAWING: Figure 1

Description

Technical Field

[0004] , ,

[0001] The present disclosure relates to a prediction system, a prediction device, and a prediction method.

Background Art

[0002] The properties of a resin are changed by the reaction rates of reactions such as the reaction for synthesizing the resin from monomers as raw material molecules, the formation and cleavage reactions of bonds that bind additives to the resin, and the like. Therefore, in the material design of a resin, it is preferable to predict the reaction rate of the monomers to be used. Further, even for monomers that form the same type of resin (for example, polyimide), there can be many types and bonding positions of substituents. When the structures are different, the ease of progress of the reaction may change. Therefore, it is preferable to conduct experimental studies every time the composition is changed and evaluate the progress and performance of the reaction.

[0003] For evaluation, when formulating the composition and process of a resin, the reaction rate can be evaluated by actually manufacturing the resin through various processes. At this time, by fitting using the Arrhenius equation or the like, trial and error for exploring the optimal composition and process can be carried out. However, with such a method, the experimental volume of the operator and the cost related to obtaining materials are large.

[0004] In recent years, in material design, the utilization of digital technologies such as machine learning using computer simulations, databases, etc. has been progressing. In the abstract of Patent Document 1, it is described that "the reaction mechanism generation method includes a step of performing molecular dynamics calculations for each atom constituting each molecule in the reaction system at each time step, a step of identifying the reaction molecules and generated molecules that contributed to the chemical reaction when a chemical reaction occurred in the reaction system before and after the time step, a step of constructing an elementary reaction composed of the related reaction molecules and generated molecules based on the atomic relevance between the reaction molecules and the generated molecules, and a step of calculating the reaction rate constant of the constructed elementary reaction."

Prior Art Documents

[0005] [Patent Document 1] International Publication No. 2016 / 133002 [Overview of the Initiative] [Problems that the invention aims to solve]

[0006] The technology described in Patent Document 1 is a technology for calculating reaction rate constants, and is not a technology for predicting the reaction rate between molecules during the reaction (polymerization) of the molecules that form the resin. The problem that this disclosure aims to solve is to provide a prediction system, prediction device, and prediction method that can predict the reaction rate between molecules that form a resin. [Means for solving the problem]

[0007] The prediction system of this disclosure includes a prediction device that predicts the reaction rate between reactive groups during a reaction between reactive groups contained in a monomer, which is a molecule that forms a resin, and the prediction device associates a first index of the monomer with the activation energy defined by the Arrhenius equation, a second index of the monomer with the frequency factor defined by the Arrhenius equation, and the reaction rate. Furthermore, it includes a predictive model constructed by machine learning with the first and second indicators as features and the response rate as the target variable. The system includes a database, a first index and a second index of the monomer determined from molecular information relating to the monomer, and a prediction unit that predicts the reaction rate from the database. The first indicator includes at least one of the activation energy of the monomer or the heat of reaction when the resin is produced from the monomer; the second indicator includes at least one of the structure of the monomer corresponding to the molecular information, a molecular property determined from the electronic state of the monomer corresponding to the molecular information and affecting the reactivity with other monomers, at least one of the attractive force or motion acting between at least one of the monomers or between the reactive groups contained in the monomer that react with other monomers, or steric hindrance to the reactive group from other monomers; and the prediction unit predicts the reaction rate as the value output by substituting the first and second indicators into the prediction model. Other solutions will be described later in the descriptions of embodiments for carrying out the invention. [Effects of the Invention]

[0008] According to this disclosure, a prediction system, prediction device, and prediction method can be provided that can predict the reaction rate between molecules forming a resin. [Brief explanation of the drawing]

[0009] [Figure 1] This is a block diagram of the prediction device in this disclosure. [Figure 2] This is a schematic diagram of an input screen according to one embodiment. [Figure 3] This is a block diagram showing the hardware configuration of a prediction device according to one embodiment. [Figure 4] This is a schematic diagram showing a hierarchical structure for extracting features related to frequency factors according to one embodiment. [Figure 5] This is a schematic diagram of a molecular property database according to one embodiment. [Figure 6] This is a flowchart showing a method for constructing a molecular properties database according to one embodiment. [Figure 7] This is a schematic diagram of a second structural database according to one embodiment. [Figure 8] This is a schematic diagram showing a hierarchical structure for extracting feature quantities related to activation energy, according to one embodiment. [Figure 9] This is a schematic diagram of a reaction pathway database according to one embodiment. [Figure 10] This is a flowchart showing a method for constructing a reaction pathway database according to one embodiment. [Figure 11] This is a schematic diagram of a first structural database according to one embodiment. [Figure 12] This is a schematic diagram of a predictive model database according to one embodiment. [Figure 13] This graph compares predicted values ​​with actual values ​​when a prediction model is created in one embodiment. [Figure 14] This flowchart shows the prediction method used in this disclosure. [Modes for carrying out the invention]

[0010] Hereinafter, embodiments for implementing the present disclosure will be described with reference to the drawings (referred to as embodiments). In the description of one of the following embodiments, descriptions of other embodiments applicable to one embodiment will be made as appropriate. The present disclosure is not limited to one of the following embodiments, and different embodiments can be combined or arbitrarily modified within a range that does not significantly impair the effects of the present disclosure. Also, the same members will be denoted by the same reference numerals, and overlapping descriptions will be omitted. Furthermore, those having the same function will be given the same name. The illustrated content is merely schematic, and for the convenience of illustration, it may be changed from the actual configuration within a range that does not significantly impair the effects of the present disclosure, or the illustration of some members may be omitted or deformed between the drawings. Also, in the same embodiment, it is not necessarily required to include all configurations.

[0011] FIG. 1 is a block diagram of a prediction system 100 of the present disclosure. The prediction system 100 includes an input device 10, a prediction device 20, and an output device 30. The prediction system 100 and the prediction device 20 predict the reaction rate between reactive groups during the reaction of reactive groups contained in monomers, which are molecules forming a resin. The prediction is executed using, for example, a database of parameters that vary depending on the raw materials, such as activation energy and frequency factor (both described later). The molecule referred to here may be an oligomer in which a plurality of monomers are bonded, in addition to the monomer (raw material) that is a molecule forming the resin. The molecular information is information regarding the monomer, and for example, it is the type of monomer (molecular formula, structural formula, etc.).

[0012] As described above, the reaction rate is the reaction rate (the ratio of reaction) between reactive groups during the reaction of reactive groups contained in the monomer. The reactive group is a reactive functional group (reaction site) involved in the resin formation reaction (polymerization), and the functional group includes double bonds, triple bonds, and the like. In the present disclosure, the reaction rate is defined by the following formula. Reaction rate (%) = (Number of reacted reactive groups / Number of reactive groups at the start of reaction) × 100

[0013] For example, when polyethylene (resin) is produced from ethylene (raw material, monomer), the double bond (reactive group) constituting ethylene changes to a single bond and ethylene molecules polymerize with each other. Therefore, when there are 10 ethylene molecules, the number of reactive groups at the start of the reaction is 10. And, for example, when 9 double bonds have reacted, the number of reacted functional groups is 9, so the reaction rate is (9 / 10)×100 = 90%. Note that the polymerization is not limited to the polymerization that occurs by the change of a double bond or a triple bond to a single bond or the like, and for example, any polymerization such as condensation polymerization or ring-opening polymerization may be used.

[0014] The input device 10, the prediction device 20, and the output device 30 are connected to each other. The prediction device 20 (specifically, the reception unit 23 and the output unit 24) is connected to an external server 50 via a network 40. The prediction device 20 may have a function of inputting and outputting from the server 50 and the external network 40. In this case, instead of the input device 10 and the output device 30, input devices and output devices (both not shown) arranged at a remote location may be used. The information in the storage unit 22 may be updated based on information obtained from the outside via the server 50. However, the prediction device 20 does not have to be connected to the network 40 and the server 50.

[0015] The input device 10 includes, for example, at least one of a mouse, a keyboard, a touch panel, a display, a monitor, etc., all not shown. The predicted reaction rate or the like is displayed on the output device 30. The output device 30 is, for example, a display device such as a display, a printer, or the like.

[0016] FIG. 2 is a schematic diagram of an input screen 101 according to an embodiment. The input screen 101 is displayed on a display device (not shown) such as a display, a monitor, etc. The input screen 101 is provided with an input field 102 for inputting input information such as raw materials. The above molecular information is input into the input field 102.

[0017] The molecular information preferably has a format that allows for the determination of the three-dimensional structure of the monomer. In the example shown in Figure 2, SMILES is used as the chemical structure of the monomer, which is the molecular information. Instead of input field 102, a file representing the three-dimensional structure of the monomer in coordinates may be entered as the molecular information. The molecular information entered may be one or more, depending on the monomers that form the resin. In the example in Figure 2, multiple molecular information entries are entered (monomer 1, monomer 2, etc.).

[0018] Figure 3 is a block diagram showing the hardware configuration of a prediction device 20 according to one embodiment. The prediction device 20 comprises a calculation unit 21, a storage unit 22, and an output device 30. The calculation unit 21 includes a CPU 2100, a GPU 2101, RAM 2102, etc. The storage unit 22 includes an HDD 2200 (or SSD), ROM 2201, etc. The output device 30 includes an I / F 3000, etc. The prediction device 20 is realized when a predetermined program stored in the HDD 2200, ROM 2201, etc., is loaded into the RAM 2102 and executed by the CPU 2100.

[0019] Returning to Figure 1, the prediction device 20 comprises a calculation unit 21, a storage unit 22, a reception unit 23, and an output unit 24. The reception unit 23 is connected to the input device 10 and receives molecular information about monomers, which are molecules that form the resin. The output unit 24 is connected to the output device 30 and outputs prediction results such as reaction rates to the output device 30.

[0020] For the sake of explanation, regarding the arithmetic unit 21 and the storage unit 22, we will first explain the DB222 provided in the storage unit 22, and then, within the explanation of the storage unit 22, we will explain the various components that make up the arithmetic unit 21 as appropriate. DB is an abbreviation for database, and the same applies hereafter.

[0021] DB222 is a database provided in the prediction device 20 that associates a first indicator (described later), a second indicator (described later), and the reaction rate. These are further associated with monomer identification information corresponding to molecular information. As will be described in detail later, in the example of this disclosure, the molecular properties DB223 associates the identification information (e.g., heading 604 (Figure 5)) with the second indicator. Also, in the reaction pathway DB224, the identification information (e.g., heading 704 (Figure 9)) is associated with the first indicator and the reaction rate. In the example of this disclosure, DB222 includes the molecular properties DB223, the reaction pathway DB224, the prediction model DB225, the first structure DB226, and the second structure DB227. In addition, although not shown in the figures, the memory unit 22 stores a calculation execution program.

[0022] Molecular Properties DB223 is a database that links monomer types with their molecular properties. Molecular properties are those that affect the reactivity with other reactive groups (which may be monomers, oligomers, or other molecules). Molecular Properties DB223 (Second Indicator Database) includes molecular properties (second indicator) related to the frequency factor of monomers. The frequency factor is defined by the Arrhenius equation.

[0023] The reaction rate can be derived by differentiating the reaction rate with respect to time, and the Arrhenius equation can be used to calculate the reaction rate. Therefore, this disclosure refers to the Arrhenius equation. The Arrhenius equation uses multiple variables and defines a frequency factor (κ) and an activation energy (Ea). The frequency factor and activation energy vary depending on the monomer or its combination. The frequency factor is the frequency at which the reacting monomers collide, and the activation energy is the energy barrier during the reaction. The feature includes at least the frequency factor and an index related to the activation energy.

[0024] Molecular properties DB223 (an example of DB222) includes headings 604 (Figure 5; an example of the above identification information) relating to at least one of the following: reactive groups or resins, which are products of reactions between reactive groups. Including headings 604 makes it easier to search for at least one of the reactive groups or resins. Furthermore, monomers can be obtained from molecular properties DB223 when constructing the prediction model described later. Note that although headings relating to the type of resin are not shown in Figure 5, headings relating to both the type of reactive group and the type of resin may be included.

[0025] Heading 604 includes information that can be used to classify or label each monomer, reactive group, etc. Specifically, this includes, for example, a molecular identification number, molecular name, functional groups involved in bond formation and / or cleavage, molecular SMILES notation, molecular CAS number, etc.

[0026] Heading 604 may be provided in addition to, or in place of, Molecular Properties DB223, in at least one of the reaction pathway DB224 or prediction model DB225 described below.

[0027] Figure 4 is a schematic diagram showing a hierarchical structure for extracting features related to frequency factors according to one embodiment. As shown at the top of Figure 4, the chemical reaction, which is the synthesis reaction of resin, is governed by both the frequency factor and the activation energy. Therefore, prediction based on these two indicators is preferable for predicting the reaction rate. Of the two indicators, the frequency factor is explained in Figure 4, and the activation energy is explained in Figure 8, which will be described later.

[0028] The features related to the frequency factor are features that can reproduce the phenomenon from the collision of reactive groups until they are temporarily stabilized, and are organized by the multilayer structure shown in Figure 4. Figure 4 illustrates, for example, phenomenon 601, factor 602, and feature 603. Phenomenon 601 is the phenomenon that occurs before the collision of reactive groups. Factor 602 is a factor that affects the likelihood of phenomenon 601 occurring. Feature 603 is a molecular property that represents factor 602. Phenomenon 601, factor 602, and feature 603 influence each other, and together they can represent the frequency factor. Therefore, feature 603 is included in the molecular property DB223 (Figure 1) as an example of the second index (information related to the frequency factor of monomers).

[0029] Phenomenon 601, shown in the top level, is a chronological list of phenomena occurring up to the collision of reactive groups in monomers, described from left to right. Phenomenon 601 includes phenomena such as compatibility with the solvent, ease of approach between molecules (e.g., at least one of monomers or oligomers), ease of approach between reactive groups, and stabilization after the reactive groups have approached each other. First, the monomers disperse in the solution due to their compatibility with the solvent. However, this process may be omitted if the monomers are mixed directly. Next, the monomers, or oligomers which are reaction products of monomers, move and diffuse, bringing them closer together. Subsequently, the reactive groups of the monomers or oligomers approach each other, collide, and the state of proximity between the reactive groups is maintained until the reaction (bond formation, dissociation, etc.) occurs.

[0030] Furthermore, the top level also represents the process by which the resin undergoes structural changes during the reaction, altering the reaction mode. For example, initially, monomers and oligomers are dispersed, and the process of them approaching each other (the left half of phenomenon 601) is the rate-determining step. However, as the reaction progresses, the oligomers become a mass of resin. As a result, the unreacted reactive groups approach each other due to the thermal motion of the surrounding structure (the right side of phenomenon 601), leading to the reaction between the reactive groups.

[0031] Factor 602, shown in the middle level, is a compilation of factors influencing the likelihood of the occurrence of the phenomenon 601, as organized in Phenomenon 601. Factor 602 includes, for example, attractive forces arising from intermolecular forces, Coulomb forces, etc., and molecular motion. These are involved in the approach of molecules in Phenomenon 601. Factor 602 includes steric hindrance, flexibility of molecular structure, attractive forces arising from intermolecular forces, Coulomb forces, etc. These are involved in the approach of reactive groups in Phenomenon 601. Factor 602 includes attractive forces arising from intermolecular forces, Coulomb forces, etc., the adsorption state of reactive groups, and the ease of electron transfer. These are involved in the stabilization after the reactive groups approach each other in Phenomenon 601. The adsorption state here refers to a state in which reactive groups are close together. It is thought that the attractive forces between reactive groups and the stability of the electron system energy in the adsorption state influence this stability. Therefore, as a molecular property that expresses the stability of the electron system energy, the HOMO / LUMO, which relates to the ease of electron transfer, is used in the characteristic quantity 603 described later.

[0032] The 603 features shown at the lowest level of the hierarchy are each second indicators related to the frequency factor, and are features that can represent the frequency factor. Feature 603 are molecular properties that influence the reactions between molecules during reactions between reactive groups contained in a molecule.

[0033] Feature quantities 603 include, for example, HSP (Hansen solubility parameter), logP, and desolvation energy. These are involved in the compatibility with the solvent in phenomenon 601. Feature quantities 603 include, for example, dipole moment and polarizability, which affect the attractive term. These are involved in the intermolecular forces that form the attractive force that contributes to the ease with which molecules approach each other in phenomenon 601. Feature quantities 603 include, for example, atomic charge and molecular charge. These are involved in the attractive force that contributes to the ease with which molecules approach each other in phenomenon 601. Feature quantities 603 include, for example, molecular weight, which affects mobility. This is involved in molecular mobility, which contributes to the ease with which molecules approach each other in phenomenon 601.

[0034] Feature 603 includes, for example, the integrated value of the radial distribution function. This relates to the steric hindrance around the reactive group in factor 602. Steric hindrance is defined as the reduction in reactivity caused by the proximity of other atoms within the molecule to a reactive group. Therefore, steric hindrance can be represented by the radial distribution function around the atom at the reaction center. The radial distribution function is a function of distance r from the origin, representing the distribution (probability of existence) of atoms around any atom in the reactive group, or the centroid of the reactive group, with the origin being the origin. The function itself can be used as a feature, but for simplicity, the integrated value (area) with respect to the distance from the central atom can also be used.

[0035] Feature quantities 603 include, for example, the number of aromatic bonds, the number of aromatic rings, and the number of rotatable bonds. These are related to the flexibility of the molecular structure (mobility of reactive groups) of factor 602. Feature quantities 603 include the number of hydrogen bond donors and hydrogen bond acceptors as attractive terms between reactive groups. These are related to intermolecular forces (hydrogen bonds, Coulomb forces, etc.) that are involved in the attraction between reactive groups in phenomenon 601 and the stabilization after they approach each other. Feature quantities 603 include the charge of the reactive groups as an attractive term between reactive groups. This is related to the Coulomb forces that are involved in the attraction between reactive groups in phenomenon 601 and the stabilization after they approach each other. Feature quantities 603 include, for example, HOMO / LUMO. This is related to the ease of electron transfer in factor 602.

[0036] As described above, phenomenon 601, factor 602, and feature 603 are interrelated. Therefore, for example, the approach of molecules in phenomenon 601 can be expressed by attractive forces (intermolecular forces, Coulomb forces, etc.) and molecular motion in factor 602, as described above. Furthermore, the attractive forces and molecular motion in factor 602 can be expressed by the dipole moment, polarizability, and molecular charge in feature 603. The same applies to the other phenomena 601, factors 602, and feature 603 in Figure 4.

[0037] In actual operation, features other than those shown in Figure 4 can be used depending on the reaction mechanism. However, it is important to systematically organize the phenomena through hierarchical structuring. By changing to a more appropriate reaction mechanism according to the reaction system, and consequently adding or deleting at least one of the 603 features, the accuracy of the prediction model can be expected to improve.

[0038] Figure 5 is a schematic diagram of a molecular properties DB223 according to one embodiment. The molecular properties DB223 includes feature quantities 603 determined based on the reasons explained with reference to Figure 4 above. The structure of the molecular properties DB223 is not limited to a tabular format as shown in Figure 5, but may also be in a graph format including correlation coefficients between feature quantities 603, causality, data generation procedure nodes, etc. Furthermore, the molecular properties DB223 includes the above-mentioned headings 604.

[0039] As described above, the molecular properties DB223 includes a second index related to the frequency factor. The second index includes at least one of the following: the structure of the monomer corresponding to the molecular information, or the molecular properties determined from the electronic state of the monomer corresponding to the molecular information (e.g., feature quantity 603 above). The molecular properties referred to here are those that affect the reactivity with other monomers. Figure 5 includes both of these as an example. By including at least one of these, the frequency factor corresponding to the monomer can be determined. In the example shown in Figure 5, the monomer structure is included as SMILES, and the molecular properties determined from the electronic state of the monomer are included as HOMO / LUMO.

[0040] The second indicator includes at least one of the attractive forces or motions acting between monomers or between reactive groups, or at least one of the steric hindrance of a reactive group to other monomers. Reactive groups are groups contained in monomers that react with other monomers. By including at least one of these, the frequency factors corresponding to monomers can be determined. In the example shown in Figure 5, the attractive forces acting between monomers are, for example, dipole moment, polarizability, molecular charge, and atomic charge. The motion (molecular mobility) acting between monomers is, for example, molecular weight. The attractive forces acting between reactive groups are the number of hydrogen bond acceptors and hydrogen bond donors. The motion acting between reactive groups is, for example, the number of rotatable bonds.

[0041] Figure 6 is a flowchart illustrating a method for constructing a molecular properties DB223 according to one embodiment. The construction of the molecular properties DB223 can be performed by the second construction unit 216 (Figure 1) using a program that can be used for molecular simulations, generation of molecular descriptors, etc. For example, various software can be used depending on the type of feature, such as the Python library RDKit, the quantum chemistry calculation software Gauusian and GAMESS, and the molecular dynamics calculation software LAMMPS. However, the second construction unit 216 is not required, and a pre-constructed molecular properties DB223 may be used.

[0042] The second construction unit 216 organizes the phenomena 601 that occur during the reaction between reactive groups in chronological order, for example, as shown in phenomenon 601 in Figure 4 above. Furthermore, the second construction unit 216 identifies factors 602 that influence phenomenon 601, for example, as shown in factor 602 in Figure 4 above. As a result, the second construction unit 216 constructs a molecular property DB 223 (an example of a second index DB) that includes characteristic quantities 603 (an example of a second index) as molecular properties that express factor 602. In this way, the chemical reactions of the resin can be systematically understood and the second index can be determined. The specific method of constructing the molecular property DB 223 will be explained below with reference to Figure 6.

[0043] Step S1: First, the second construction unit 216 sets the functional groups (reacting groups) involved in the reaction. The items to be set are the type and arrangement of the functional groups. The type of functional group is determined by calling it from a database (not shown) of functional groups stored in the memory unit 22 (Figure 1). The type of functional group is any functional group involved in the formation of the resin, such as an amino group, acid anhydride, isocyanate group, etc. The number of functional groups is arbitrary, but one or more is preferred, and two or three is more preferred. Multiple types of functional groups may also be selected. The arrangement of the functional groups is the bonding site in the basic skeleton in step S2 described below.

[0044] Step S2: The second construction unit 216 obtains the basic framework of the resin formed by the monomers from a database (not shown) of basic frameworks stored in the memory unit 22. The basic framework of the molecule is, for example, a benzene ring, an alkyl group, etc. The number of benzene rings, the number of carbon atoms in the alkyl group, the bond order, etc., are all arbitrary.

[0045] Step S3: The second construction unit 216 places the determined type of functional group at the determined location on the basic resin skeleton determined in Step S2, based on the type and arrangement of the functional group determined in Step S1. The arrangement is performed by simulation. This allows the reactive functional groups (reactive groups) that form the resin to be placed on the basic resin skeleton, and the molecular structure of the resin is obtained.

[0046] To ensure that all placed functional groups participate in the reaction, the placement of functional groups should be as follows: First, functional groups that form the resin are, in principle, placed at the ends of the molecular structure. However, if a functional group can participate in the reaction even if it is not at the end of the molecule, it can be inserted between the basic skeletons. Such functional groups include, for example, amino groups. Since amino groups can be reactive even if they are secondary or tertiary, they may be inserted between the basic skeletons. Furthermore, to prevent reactive groups from losing their reactivity, it is preferable that functional groups do not bond to each other within the same molecule.

[0047] Step S4: The second construction unit 216 determines the type and number of substituents based on the second structure DB227 (Figure 7) of the substituents stored in the memory unit 22 (Figure 3). In this disclosure, substituents mean molecular structures that do not participate in the resin formation reaction. Substituents can be any substituents generally found in organic compounds, such as methyl groups, ethyl groups, ethers, hydroxyl groups, etc. Multiple types of substituents can be used. However, substituents are of a different type from the functional groups that participate in the reaction. Also, substituents do not, in principle, bond to functional groups. This is to prevent functional groups from losing their reactivity or changing into other functional groups due to bonding with substituents.

[0048] Figure 7 is a schematic diagram of the second structure DB227 according to one embodiment. The second structure DB227 is a database that stores information such as the molecular skeleton, functional groups, substituents, the name and structure of the resin to be produced. The reactive groups, the basic skeleton of the resin, and substituents used in steps S1 to S4 of Figure 6 are obtained from the second structure DB227. The second structure DB227 consists of an ID assigned to each data item, the type of structure (basic skeleton, substituents, etc.), the name of the structure, and a notation that can generate the three-dimensional structure (SMILES is shown in Figure 7). Note that a separate database may be created for each type of structure.

[0049] Step S5: Returning to Figure 6, the second construction unit 216 determines the placement locations of the substituents based on the type and number of substituents determined in step S4.

[0050] Step S6: The second construction unit 216 obtains the molecular structure based on the results of steps S1 to S5. Any format can be used to obtain the molecular structure, as long as it is a format that can represent molecular structures of two dimensions or more. For example, SMILES notation, a file containing the coordinates of the molecule, etc.

[0051] Step S7: The second construction unit 216 calculates the molecular descriptors of the substituents based on the acquired molecular structure. This allows for the determination of two-dimensional molecular properties. The molecular properties determined in Step S7 include, for example, molecular weight, hydrogen bond donor, number of hydrogen bond acceptor, logP, molecular charge, aromaticity, number of rotatable bonds, etc. Step S8: The second construction unit 216 randomly generates multiple three-dimensional structures of the resin based on the molecular structure obtained in step S6.

[0052] Step S9: The second construction unit 216 optimizes the structure of each molecular structure generated in Step S8 using molecular force fields. At this time, the second construction unit 216 obtains the energy value of each molecular structure. Structural optimization can be performed by molecular simulations such as molecular mechanics calculations. Step S10: The second construction unit 216 compares the energy values ​​for each molecular structure obtained in Step S9 and selects the structure with the lowest energy as the most stable structure. Step S11: The second construction unit 216 further optimizes the structure of the most stable structure obtained in step S10 using quantum chemical calculations.

[0053] Step S12: The second construction unit 216 obtains the vibrational frequencies for the structurally optimized structure. If only positive vibrational frequencies are obtained, then that structure is the most stable structure among all the structures the molecule can take. Therefore, the second construction unit 216 selects it as the most stable structure and executes the next step S13. On the other hand, if imaginary vibrational frequencies are included, then that structure is not the true most stable structure. Therefore, the second construction unit 216 returns to step S8 once again to search for the most stable structure.

[0054] Step S13: The second construction unit 216 obtains molecular properties such as the charge of the reactive groups, dipole moment, and polarizability, which can be determined by quantum chemical calculations, for the most stable structure obtained in Step S11. Step S14: The second construction unit 216 calculates the radial distribution function around the reactive group for the most stable structure obtained in step S11. The calculation of the radial distribution function is as described above with reference to Figure 4. Step S15: The second construction unit 216 integrates the radial distribution function obtained in step S14 and obtains the integrated value. Step S16: The second construction unit 216 stores all the data obtained from the above calculations in the molecular properties DB 223 (Figure 1) of the storage unit 22 (Figure 1). At this time, the second construction unit 216 also assigns new IDs, etc. The stored data includes, for example, molecular properties (feature quantity 603 (Figure 4)), but may also include frequency factors, etc.

[0055] Step S17: The second constructor 216 determines whether there are any unevaluated substituent locations. If there are unevaluated locations, the second constructor 216 repeats steps S5 onwards; otherwise, it proceeds to step S18. Step S18: The second constructor 216 determines whether there are any unevaluated substituent types. If there are unevaluated types, the second constructor 216 repeats steps S4 onwards; otherwise, it proceeds to step S19. Step S19: The second constructor 216 determines whether or not there are any unevaluated functional groups. If there are any unevaluated functional groups, the second constructor 216 repeats steps S3 onwards; otherwise, it proceeds to step S20. Step S20: The second construction unit 216 determines whether there are any unevaluated basic frameworks. If there are unevaluated basic frameworks, the second construction unit 216 repeats steps S2 onward; otherwise, it terminates the flow.

[0056] In the example shown in Figure 6, molecular mechanics calculations were performed in step S9, and quantum chemical calculations were performed in step S11. However, the calculations for performing molecular simulations are not limited to these examples. That is, the second construction unit 216 can construct the molecular properties DB223 from molecular information by performing at least one of the following calculations: monomer structure-based descriptor calculations, quantum chemical calculations, molecular mechanics calculations, molecular dynamics calculations, first-principles calculations, and first-principles molecular dynamics calculations. The molecular properties DB223 can be constructed by performing at least one of these calculations.

[0057] Furthermore, the second construction unit 216 can construct the molecular properties DB 223 for the molecular structure of the resin obtained in step S6 by at least one of calculating molecular descriptors or performing a molecular simulation. The molecular structure referred to here is the molecular structure of the resin obtained in step S6 by arranging the reactive functional groups that form the resin on the basic skeleton of the resin. The molecular properties DB 223 can be constructed by at least one of these calculations. In the example of this disclosure, molecular descriptors are calculated in step S7 and molecular simulations are performed in step S9.

[0058] Returning to Figure 1, the reaction pathway DB224 is a database that includes a first index and a predetermined reaction rate 703 (Figure 9) for each of the multiple monomer combinations. The first index is an index relating to the activation energy of the monomer. The activation energy is defined by the Arrhenius equation. The reaction rate includes at least one of the measured value or the calculated value. The calculated value can be obtained, for example, by molecular simulation such as quantum chemical calculations. As will be described in detail later, the reaction rate prediction is performed by the prediction unit 214 using the reaction rates included in the reaction pathway DB224.

[0059] Figure 8 is a schematic diagram showing a hierarchical structure for extracting feature quantities related to activation energy according to one embodiment. Similar to the frequency factors described above, chemical reactions are also governed by activation energy. Therefore, as shown in Figure 8, similar to the frequency factors, feature quantities can be extracted for activation energy by organizing the phenomena through hierarchical structures. Phenomenon 701 is a phenomenon that occurs during monomer reactions, and feature quantity 702 is a feature quantity related to the ease of reaction. Phenomenon 701 and feature quantity 702 influence each other, and together they can represent the activation energy. Accordingly, feature quantity 702 is included in the reaction pathway DB224 (Figure 1) as the first index (information related to the activation energy of monomers).

[0060] Chemical reactions involve both forward and reverse reactions proceeding simultaneously. Furthermore, chemical reactions can proceed in multiple steps, such as through an intermediate. Therefore, phenomenon 701 includes, for example, forward reaction (1), forward reaction (2), reverse reaction (1), and reverse reaction (2). In this example, forward reaction (1) and reverse reaction (1) proceed simultaneously to produce an intermediate from a monomer, and then forward reaction (2) and reverse reaction (2) proceed simultaneously from the intermediate to produce a resin.

[0061] The ease with which a forward reaction occurs is influenced by the activation energy, while the ease with which a reverse reaction occurs is influenced by the activation energy, reaction heat, etc. Therefore, feature quantity 702 includes the activation energy (1) that influences the forward reaction (1) and the reverse reaction (1), and the activation energy (2) that influences the forward reaction (2) and the reverse reaction (2). Furthermore, feature quantity 702 includes the reaction heat (1) that influences the reverse reaction (1), and the reaction heat (2) that influences the reverse reaction (2). Note that feature quantities related to activation energy (e.g., activation energy (1), activation energy (2)) may be calculated values ​​using molecular simulations such as quantum chemical calculations or first-principles calculations, or they may be literature values.

[0062] In the examples of this disclosure, though not limited to these examples, the activation energy is used as a feature for each reaction. For example, if activation energy (1) and activation energy (2) are included, it is preferable that activation energy (1) and activation energy (2) be treated as independent feature quantities. This makes it easier to represent chemical reactions using activation energy. Furthermore, in another embodiment, although not limited to this example, only the highest activation energy among the activation energies of each reaction can be used as a feature. This is because the reaction with the highest activation energy becomes the rate-determining step. As a result, even in the case of multi-stage reactions, there is only one feature related to the activation energy, which reduces the load on learning.

[0063] Figure 9 is a schematic diagram of a reaction pathway DB224 according to one embodiment. In the example in Figure 9, the reaction pathway DB224 is in tabular format, but it may also be in graph format. The reaction pathway DB224 includes the above-mentioned feature quantity 702 as the first indicator. The feature quantity 702 (first indicator) includes at least one of the following: the activation energy of the monomer, or the reaction heat when the resin is produced from the monomer. Since these affect the chemical reaction when the resin is produced from the monomer as described above, the reaction pathway DB224 can be constructed by including at least one of these.

[0064] Reaction pathway DB224 includes reaction rate 703 for systems with known reaction rates. Reaction rate 703 is used as training data when creating the predictive model (described later).

[0065] The reaction pathway DB224 includes a heading 704 (an example of the identification information described above), similar to heading 604 (Figure 5). Heading 704 relates to at least one of the following: a reactive group, or a resin which is a product of the reaction between reactive groups. Data can be obtained from the reaction pathway DB224 when creating the prediction model described later. As an example, Figure 9 shows amines and acid anhydrides as reactive groups, p-PDA as a monomer reactant (1), and PMDA as a monomer reactant (2). By including heading 704, it is possible to easily extract at least one of the target reactive group or resin.

[0066] Furthermore, heading 704 includes at least one of the following: a classification of combinations of multiple monomers, or labelable information. Heading 704 includes information about each molecule, such as a molecular identification number, molecular name, functional groups involved in bond formation and cleavage, molecular SMILES notation, and molecular CAS number, as well as the type of resin (e.g., polyimide).

[0067] Figure 10 is a flowchart illustrating a method for constructing a reaction pathway DB224 according to one embodiment. The flowchart in Figure 10 shows a method for determining a first indicator (such as activation energy) for a resin that will serve as training data for the prediction model described later, assuming that the reaction rate has been obtained in advance. The construction of the reaction pathway DB224 can be performed, for example, by a first construction unit 215 (Figure 1). However, the first construction unit 215 is not required, and a pre-constructed reaction pathway DB224 may be used.

[0068] The first construction unit 215 organizes the phenomena 701 (Figure 8) that occur during the reaction of reactive groups in chronological order and evaluates the likelihood of phenomena 701 occurring, thereby constructing a reaction pathway DB 224 (Figure 1; an example of a first indicator database). In this way, the resin formation reaction can be systematically understood and the first indicator can be determined. The specific method of constructing the reaction pathway DB 224 will be explained below with reference to Figure 10.

[0069] Step S31: The first construction unit 215 sets the product for a resin whose monomer reaction rate is known. The product refers to a repeating unit, which is a structure formed by the reaction of two monomer molecules.

[0070] Step S32: The first construction unit 215 generates a three-dimensional structure based on the repeating unit information obtained in step S31, and then obtains the most stable structure. The method for obtaining the most stable structure is the same as the method described in step S10, etc., of the molecular properties DB223 described above. The most stable structure obtained here is the final state when monomers react with each other to produce a resin.

[0071] Step S33: The first construction unit 215 determines the reactants based on the information of the repeating units obtained in step S31 (such as the type of resin containing the repeating units). Reactants refer to the monomer molecules that generate the repeating units and their molecular structures. The functional groups (reactants) involved in the reaction are determined from the type of resin containing the repeating units, and the reactants are determined based on these.

[0072] Figure 11 is a schematic diagram of the first structure DB226 according to one embodiment. The first structure DB226 is a database that stores information such as the molecular skeleton, functional groups, substituents, the name and structure of the resin to be produced. In step S33 (Figure 10), the functional groups involved in the reaction are retrieved from the first structure DB226. The first structure DB226 consists of an ID assigned to each data, a notation that can generate the three-dimensional structure of the structure (SMILES are shown in Figure 11), the names and SMILES of each reactive group in a multi-step reaction, etc. Note that a separate DB may be created for each type of structure.

[0073] Step S34: Returning to Figure 10, the first construction unit 215 generates a three-dimensional structure based on the monomer molecule (reactant) information obtained in step S33, and then obtains the most stable structure. The method for obtaining the most stable structure is the same as the method described in step S10 of the molecular properties DB223 above.

[0074] Step S35: The first construction unit 215 generates multiple structures by bringing the functional groups (reacting groups) involved in the reaction closer together, based on the most stable structure obtained in step S34. All of the generated structures are then optimized using molecular simulations such as molecular mechanics calculations. The structure with the lowest energy is selected from the energies obtained for each structure, thereby creating the pre-reaction structure (initial state; initial structure). The first construction unit 215 further optimizes the created structure using quantum chemical calculations to obtain energy values.

[0075] In this way, the first construction unit 215 determines the three-dimensional structure of a resin with a known reaction rate and the monomers that form the resin. Then, the first construction unit 215 constructs the reaction pathway DB 224 by generating multiple three-dimensional structures in which the reactive groups in the monomers whose reaction rates are to be predicted are brought closer together and performing structural optimization. This makes it possible to create the structure before the reaction.

[0076] Step S36: The first construction unit 215 obtains the reaction heat by calculating the difference between the energy obtained in step S35 (energy in the initial structure) and the energy obtained in step S31 (energy in the final structure).

[0077] Step S37: The first construction unit 215 obtains the activation energy and reaction pathway based on the initial state structure obtained in step S35 and the final state structure obtained in step S31. This acquisition can be performed by the reaction pathway determination unit 212 (Figure 1). The reaction pathway determination unit 212 determines the structure of the resin from the monomer and determines the reaction pathway from the monomer to the resin by molecular simulation. By determining the reaction pathway using the reaction pathway determination unit 212, the activation energy and reaction heat can be determined.

[0078] The specific determination method will now be explained. The reaction pathway determination unit 212 determines the functional groups (reactive groups) involved in the reaction from the molecular information. Based on the combination of functional groups, the reaction pathway determination unit 212 determines the type of resin (e.g., polyimide), thereby determining the molecular structure of the resin. Then, using the method described above, the reaction pathway determination unit 212 determines the initial state structure based on the molecular information and the final state structure based on the molecular structure of the resin. From the initial and final state structures, the reaction pathway determination unit 212 can calculate the activation energy and reaction heat by performing a reaction pathway search using molecular simulations such as quantum chemical calculations and first-principles calculations. For example, the reaction pathway search method can be the GRRM (Global Reaction Route Mapping) method, the NEB (Nudged Elastic Band) method, the String method, or methods developed from these methods.

[0079] Step S38: The first construction unit 215 stores all the data obtained above in the reaction pathway DB 224 (Figure 1) of the storage unit 22 (Figure 1). At this time, the first construction unit 215 also assigns new IDs, etc. The stored data includes, for example, activation energy (1), activation energy (2), reaction heat (1), reaction heat (2) (feature quantity 702 (Figure 8)), etc.

[0080] If the reaction is a multi-step reaction, the first constructing unit 215 repeats steps S33 to S38 as many times as there are steps in the reaction. For example, if the reaction is two-step, in which an intermediate is formed from reactants (monomers) and a product is formed from the intermediate, the first constructing unit 215 processes the reactants as the initial state and the intermediate as the final state in the first reaction. In the second reaction, the first constructing unit 215 processes the intermediate as the initial state and the product as the final state.

[0081] Thus, when the reaction from monomer to resin proceeds through intermediates and multiple reaction steps, the first construction unit 215 constructs the reaction pathway DB224 in multiple stages: from the molecule to the intermediate, and from the intermediate to the resin. This suppresses the complexity of the phenomenon caused by the involvement of multiple reactions and simplifies the calculations.

[0082] Returning to Figure 1, the prediction model DB225 (an example of DB222) is a database containing a prediction model that predicts reaction rates. By using the prediction model, the reaction rates of monomers that produce resin can be predicted from molecular information entered by the user. The prediction model DB225 further includes the type of resin. Examples of resin types include polyimide and epoxy resin.

[0083] Figure 12 is a schematic diagram of the prediction model DB225 according to one embodiment. In the prediction model DB225, the prediction model is included in DB222 for each type of resin (reaction system). In the example shown in Figure 12, for example, when producing imide resin, the prediction model is a function that includes a second indicator such as the radial distribution function and a first indicator such as the activation energy as variables.

[0084] The prediction model is constructed by the prediction model construction unit 213 (Figure 1). The prediction model construction unit 213 constructs the prediction model using machine learning with the first and second indicators as features and the reaction rate as the target variable. Of the features used in machine learning, the feature related to the first indicator is feature 702 (Figure 8), and the feature related to the second indicator is feature 603 (Figure 4). By constructing a prediction model using machine learning, the reaction rate can be predicted by inputting the first and second indicators into the prediction model. The first indicator is included in the reaction pathway DB 224 (Figure 1), for example, and includes, for example, activation energy and reaction heat. The second indicator is included in the molecular properties DB 223, for example, and includes, for example, molecular properties related to frequency factors. Any algorithm can be used for machine learning. Supervised learning is preferred for machine learning, but unsupervised learning or reinforcement learning may also be used.

[0085] The prediction model building unit 213 preferably builds a different prediction model for each resin. Since the reaction differs depending on the resin, this allows for the construction of a more accurate prediction model.

[0086] The reaction rates used in machine learning include, for example, measured values ​​obtained by actually reacting monomers with each other. Including measured values ​​allows the predicted reaction rate to be brought closer to the actual reaction rate, thereby improving prediction accuracy.

[0087] The reaction rates used in machine learning include, for example, calculated values ​​from molecular simulations assuming the reaction of monomers. These molecular simulation calculations possess a certain degree of accuracy. Therefore, by doing so, a large number of calculation values ​​can be prepared through molecular simulations, and a predictive model can be constructed using these numerous values. This improves the prediction accuracy of the predictive model.

[0088] The molecular properties (second index) used in machine learning can be obtained by the molecular property determination unit 211 (Figure 1). The molecular property determination unit 211 calculates the molecular properties used in machine learning based on molecular information (molecular name, structural formula, molecular coordinate data, etc.) as identification information to identify monomer molecules. The calculation can be performed by molecular simulations such as descriptor calculations, quantum chemical calculations, and molecular dynamics calculations. Molecular simulations can be performed by executing any program.

[0089] Furthermore, the activation energy (first indicator) and reaction rate used in machine learning can be obtained by the reaction pathway determination unit 212 (Figure 1).

[0090] Figure 13 is a graph comparing predicted and measured values ​​when a prediction model was created in one embodiment. In Figure 13, the horizontal axis (x-axis) represents the reaction rate (predicted value), the vertical axis (y-axis) represents the reaction rate (measured value), and the line L represents the graph y=x. The graph in Figure 13 plots the reaction rate (measured value) when resin is actually created using monomers in the input molecular information, and the reaction rate (predicted value) predicted using the same molecular information and the prediction model. As shown in Figure 13, each plot is located near the line L, which is the graph y=x. Therefore, the predicted values ​​are close to the measured values, indicating that the prediction accuracy of the prediction model is high.

[0091] Returning to Figure 1, the calculation unit 21 comprises a molecular property determination unit 211, a reaction pathway determination unit 212, a prediction model construction unit 213, a prediction unit 214, a first construction unit 215, and a second construction unit 216. Of these, the molecular property determination unit 211, the reaction pathway determination unit 212, the prediction model construction unit 213, the first construction unit 215, and the second construction unit 216 have been explained with reference to Figure 12 above, so their explanation will be omitted here.

[0092] The prediction unit 214 predicts the reaction rate from the first and second indicators of the monomer determined from molecular information about the monomer, and from DB 222. The first indicator is an indicator related to the activation energy, as described above, and is determined by the reaction pathway determination unit 212. The reaction pathway determination unit 212 (first determination unit) determines the first indicator to be used by the prediction unit 214 by obtaining information corresponding to the molecular information received by the reception unit 23 from DB 222 (specifically, the reaction pathway DB 224). By including the reaction pathway determination unit 212, information used to predict the reaction rate can be obtained from the input molecular information.

[0093] The reaction pathway determination unit 212 (first determination unit) determines the structure of the resin from the monomer using molecular information, and determines the reaction pathway from the monomer to the resin through molecular simulation. This allows the reaction pathway to be determined.

[0094] The second indicator is a molecular property, which is an indicator related to the frequency factor as described above, and is determined by the molecular property determination unit 211. The molecular property determination unit 211 (second determination unit) determines the second indicator to be used by the prediction unit 214 by obtaining information corresponding to the molecular information received by the reception unit 23 from DB 222 (specifically, molecular property DB 223). By providing the molecular property determination unit 211, information used to predict the reaction rate can be obtained from the input molecular information. The prediction unit 214 predicts the reaction rate by substituting the obtained first and second indicators into the prediction model in the prediction model DB 225.

[0095] Figure 14 is a flowchart of the prediction method of this disclosure. The flow shown in Figure 14 can be executed by the prediction device 20 (Figure 1). Therefore, the explanation of Figure 14 will be given with reference to Figure 1 and other figures as appropriate. The reaction rates predicted by the prediction method and prediction device 20 of this disclosure predict the reaction rates for monomers whose reaction rates are unknown in a reaction system for which the above prediction model has already been constructed.

[0096] Step S51: The user inputs molecular information about the monomers to be used as raw materials through the input device 10. Step S52: The molecular property determination unit 211 and the reaction pathway determination unit 212 determine the functional groups (reacting groups) involved in the reaction based on molecular information (names of the reacting molecules, structural formulas, structural information, etc.). Step S53: The molecular property determination unit 211 and the reaction pathway determination unit 212 determine the type of resin (reaction system) based on the determined reaction groups. At this time, the molecular property determination unit 211 and the reaction pathway determination unit 212 retrieve the type of resin from the first structure DB226 (Figure 11) and the second structure DB227 (Figure 7) as appropriate. Step S54: The molecular property determination unit 211 determines the molecular properties as a second indicator based on the molecular information. This determination can be performed by executing steps S5 onwards, for example, in the flowchart described with reference to Figure 6 above.

[0097] Step S55: The reaction pathway determination unit 212 obtains the most stable structure of the resin product based on the monomer structure in the molecular information input in step S51 and the type of resin obtained in step S53. The acquisition of the most stable structure can be carried out in the same manner as described with reference to Figures 6 and 10 above. The structure of the resin product is the structure after the reaction of one monomer molecule each (two molecules if there is only one type of monomer), similar to when constructing the reaction pathway DB224.

[0098] Step S56: The reaction pathway determination unit 212 obtains the reaction heat and activation energy as first indicators based on the molecular properties and most stable structure obtained in steps S54 and S55. This acquisition can be performed by executing, for example, steps S5 onwards, as shown in the flowchart described with reference to Figure 10 above.

[0099] Step S57: The prediction unit 214 predicts the reaction rate by substituting the molecular properties, reaction heat, and activation energy determined in steps S54 and S56 into the prediction model corresponding to the type of resin determined in step S53. Step S57 is a prediction step in which the reaction rate is predicted from the first and second indicators of the monomer, which are molecules that form the resin and were received in step S51, and DB222. DB222 is a database that associates the first indicator of the monomer, which is a molecule that forms the resin, the second indicator of the monomer, and the reaction rate, as described above. Step S58: The prediction unit 214 outputs the predicted response rate using the output device 30.

[0100] As described above, the prediction device 20 and prediction method of this disclosure perform two main steps. In the first step, a prediction model is constructed by machine learning using information on multiple monomers with known reaction rates as training data. In the second step, feature quantities are determined by molecular simulation (molecular calculation) from molecular information such as the molecular structure and structural formula of the monomers, and the reaction rate is predicted. However, a pre-constructed prediction model may be used. The prediction model may also be updated as appropriate.

[0101] Furthermore, according to the prediction device 20 and the prediction method of this disclosure, the reaction rate between reactive groups when they react with each other to produce a resin can be predicted.

[0102] Furthermore, in the example provided in this disclosure, predicting the reaction rate only requires inputting molecular information about the candidate monomers. Therefore, users do not necessarily need to have expertise in resin materials. As a result, the design of resin materials, which was previously a highly specialized task requiring experts, can be efficiently carried out by workers without specialized expertise.

[0103] Furthermore, the number of monomers used in the reaction (polymerization) can be predicted from the predicted reaction rate. This allows for the prediction of the number-average molecular weight of the resulting polymer chain, and from the number-average molecular weight, the average length of the molecular chain can be determined. The number-average molecular weight can be used when modeling the three-dimensional structure of the resin in molecular simulations.

[0104] Conventionally, when modeling the three-dimensional structure of resins, the length of molecular chains was determined based on the experience and know-how of the operator. Furthermore, resin structures were sometimes fabricated by simulating the reaction that synthesizes the resin using molecular simulation. Of these, the former method had the problem of reduced accuracy because the reaction rate was not reflected in the model. On the other hand, the latter method had the problem of requiring computational resources and time in the process of synthesizing the resin using molecular simulation. However, according to this disclosure, it is possible to model resin structures with higher accuracy in a shorter time. [Explanation of Symbols]

[0105] 10 Input devices 100 Prediction Systems 101 Input Screen 102 Input fields 20 Prediction device 21 Arithmetic section 211 Molecular property determination section (second determination section) 212 Reaction pathway determination unit (first determination unit) 213 Predictive Model Construction Department 214 Prediction Section 215 First Construction Section 216 Second Construction Section 22 Memory section 222 DB (Database) 223 Molecular Properties Database (Database, Second Indicator Database) 224 Reaction Pathway Database (Database, Primary Indicator Database) 225 Predictive Model Database 226. First Structure DB (Database) 227 Second Structure DB (Database) 23 Reception Department 24 Output section 30 Output device 601 Phenomenon 602 factors 603 Features (Second Indicator) 604 Heading (Identification Information) 701 Phenomenon 702 Features (First Metric) 703 Response rate 704 Heading (Identification Information)

Claims

1. It is equipped with a predictive device that predicts the reaction rate between reactive groups during the reaction of reactive groups contained in monomers, which are molecules that form resins. The prediction device is A database including a predictive model constructed by machine learning, relating a first index relating to the activation energy of the monomer as defined by the Arrhenius equation, a second index relating to the frequency factor of the monomer as defined by the Arrhenius equation, and the reaction rate, and using the first and second indexes as features and the reaction rate as the target variable, The system includes a prediction unit that predicts the reaction rate from the first and second indices of the monomer determined from molecular information relating to the monomer, and the database. The first indicator is, The activation energy of the monomer, or The heat of reaction when the resin is produced from the monomer, Including at least one of the following, The second indicator mentioned above is, The structure of the monomer corresponding to the aforementioned molecular information, Molecular properties, which are determined from the electronic state of the monomer corresponding to the molecular information and which affect the reactivity with other monomers, At least one of the attractive force or motion acting between the monomers themselves, or between the reactive groups contained in the monomers that react with other monomers, Steric hindrance to the other monomers for the reactant group, Includes at least one of the following: The prediction unit predicts the response rate as the value output by substituting the first indicator and the second indicator into the prediction model. A prediction system characterized by the following features.

2. A prediction model construction unit is provided, which constructs the prediction model by machine learning with the first indicator and the second indicator as features and the response rate as the target variable. The prediction system according to claim 1.

3. The reaction rate includes measured values ​​obtained by actually reacting the monomers together. The prediction system according to claim 2, characterized in that it is as described above.

4. The reaction rate includes calculated values ​​obtained from molecular simulations assuming that the monomers react with each other. The prediction system according to claim 2, characterized in that it is as described above.

5. The prediction model construction unit constructs a different prediction model for each resin. The prediction system according to claim 2, characterized in that it is as described above.

6. The prediction unit includes a second determination unit that determines the second indicator used by the prediction unit by obtaining information corresponding to the molecular information from the database. The prediction system according to claim 1 or 2, characterized in that it is the same as described above.

7. The prediction unit includes a first determination unit that determines the first index used by the prediction unit by obtaining information corresponding to the molecular information from the database. The prediction system according to claim 1 or 2, characterized in that it is the same as described above.

8. The first determination unit determines the structure of the resin from the monomer and determines the reaction pathway from the monomer to the resin by molecular simulation. The prediction system according to feature 7.

9. The aforementioned database includes a second indicator database which includes the second indicator, The second construction unit organizes the phenomena that occur during the reaction between the aforementioned reactive groups in chronological order, identifies the factors that influence the aforementioned phenomena, and constructs a second indicator database that includes the second indicator as a molecular property that expresses the aforementioned factors. The prediction system according to claim 1 or 2, characterized in that it is the same as described above.

10. The second construction unit constructs the second index database from the molecular information by performing at least one of the following calculations: descriptor calculation based on the monomer structure, quantum chemical calculation, molecular mechanics calculation, molecular dynamics calculation, first-principles calculation, and first-principles molecular dynamics calculation. The prediction system according to feature 9.

11. The second construction unit constructs the second index database for the molecular structure of the resin obtained by arranging the reactive functional groups that form the resin on the basic skeleton of the resin, by at least one of calculating molecular descriptors or performing molecular simulations. The prediction system according to feature 9.

12. The database includes identification information relating to at least one of the types of the reactive group or the resin, which is a product of the reaction between the reactive groups. The prediction system according to claim 1 or 2, characterized in that it is the same as described above.

13. The database includes a first indicator database which includes the first indicator, The system includes a first construction unit that constructs the first indicator database by organizing the phenomena that occur during the reaction between the aforementioned reactive groups in chronological order and evaluating the likelihood of these phenomena occurring. The prediction system according to claim 1 or 2, characterized in that it is the same as described above.

14. The first construction unit determines the three-dimensional structure of the resin whose reaction rate is known and the monomers that form the resin, and constructs the first index database by generating multiple three-dimensional structures in which the reactive groups in the monomers whose reaction rate is to be predicted are brought closer together and performing structural optimization. The prediction system according to claim 13, characterized in that it is the same as described above.

15. If the reaction to produce the resin from the monomer proceeds through an intermediate and multiple reaction steps, the first construction unit constructs the first index database in multiple steps, from the molecule to the intermediate and from the intermediate to the resin. The prediction system according to claim 13, characterized in that it is the same as described above.

16. A database including a predictive model constructed by machine learning, which associates a first index relating to the activation energy of a monomer, a molecule that forms a resin, as defined by the Arrhenius equation; a second index relating to the frequency factor of the monomer, as defined by the Arrhenius equation; and the reaction rate between reactive groups during their reaction with each other in the monomer; and uses the first and second indices as features and the reaction rate as the target variable. The system includes a prediction unit that predicts the reaction rate from the first and second indices of the monomer determined from molecular information relating to the monomer, and the database. The first indicator is, The activation energy of the monomer, or The heat of reaction when the resin is produced from the monomer, Including at least one of the following, The second indicator mentioned above is, The structure of the monomer corresponding to the aforementioned molecular information, Molecular properties, which are determined from the electronic state of the monomer corresponding to the molecular information and which affect the reactivity with other monomers, At least one of the attractive force or motion acting between the monomers themselves, or between the reactive groups contained in the monomers that react with other monomers, Steric hindrance to the other monomers for the reactant group, Includes at least one of the following: The prediction unit predicts the response rate as the value output by substituting the first indicator and the second indicator into the prediction model. A prediction device characterized by the following features.

17. A database including a predictive model constructed by machine learning, which associates a first index relating to the activation energy of a monomer, a molecule that forms a resin, as defined by the Arrhenius equation; a second index relating to the frequency factor of the monomer, as defined by the Arrhenius equation; and the reaction rate between the reactive groups contained in the monomer during their reaction; and uses the first and second indexes as features and the reaction rate as the target variable. The first and second indicators of the monomer determined from molecular information relating to the monomer, The process includes a prediction step of predicting the reaction rate, The first indicator is, The activation energy of the monomer, or The heat of reaction when the resin is produced from the monomer, Including at least one of the following, The second indicator mentioned above is, The structure of the monomer corresponding to the aforementioned molecular information, Molecular properties, which are determined from the electronic state of the monomer corresponding to the molecular information and which affect the reactivity with other monomers, At least one of the attractive force or motion acting between the monomers themselves, or between the reactive groups contained in the monomers that react with other monomers, Steric hindrance to the other monomers for the reactant group, Includes at least one of the following: In the prediction step, the values ​​output by substituting the first indicator and the second indicator into the prediction model are predicted as the response rate. A prediction method characterized by the following features.

Images