Search method for raw material compounds for paints

Abstract

To provide a method for searching for raw material compounds for paints, which can search for raw material compounds for paints quickly, accurately and simply. [Solution] A method for searching for paint raw material compounds, comprising: a selection step of selecting candidate paint raw material compounds using a judgment model obtained by machine learning; and a confirmation step of conducting experiments using the candidate paint raw material compounds selected in the selection step to confirm their properties, wherein the paint raw material compounds are one or more selected from the group consisting of curing agents, curing agent components, film-forming resins, additives, pigments, and solvents.

Description

[Technical Field]

[0001] The present invention relates to a method for searching for raw material compounds for paints. [Background technology]

[0002] When developing paints, selecting the right raw materials for the paint is crucial to creating a paint with the desired properties. Until now, the selection of raw materials for paints has depended on the developer's knowledge based on data on the compositions and properties of paints that have been created. However, as the requirements for paint properties have become more sophisticated and complex, and the number of raw materials for paints has increased, many experiments are required to arrive at a paint with the desired properties. This has resulted in increased time, effort, cost, and waste generation during paint development.

[0003] In light of this situation, the use of materials informatics (MI) is being considered for the design and development of new materials. MI combines experimental science, theoretical science, computational science, and information (data-driven) science in the design and development of new materials, making it possible to design and develop new materials efficiently and with high quality, and has been used in a variety of research and development projects.

[0004] The following techniques have been known so far for developing new materials. Patent Document 1 discloses a materials design device for designing target materials, including materials consisting of multiple compositions or materials manufactured by combining multiple manufacturing conditions. This device uses a trained model that has been acquired by machine learning to determine the correspondence between input information, including the design conditions of the target material, and output information, including material property values. Patent Document 2 discloses a materials development method using materials informatics to search for a target material, in which the properties, composition, and / or process of the material are predicted based on data including first data indicating at least one of composition data, process data, and property data for each sample, and second data indicating structural data including specific feature amounts for the sample. Patent Document 3 discloses a method for searching for novolac phenolic resins, which is executed by an information processing device, and includes the steps of: generating a plurality of prediction models corresponding to a plurality of objective variables using performance data related to novolac phenolic resins; and searching for novolac phenolic resins having a desired balance of physical properties by inverse analysis using the plurality of prediction models, wherein the performance data includes a polymer composition, a structural formula, a reaction solvent, and reaction parameters related to the novolac phenolic resin, and the objective variables include developability, heat resistance, and molecular weight. [Prior art documents] [Patent documents]

[0005] [Patent Document 1] International Publication No. 2020 / 090848 [Patent Document 2] Japanese Patent Application Publication No. 2023-24492 [Patent Document 3] Japanese Patent Publication No. 2023-90563 Summary of the Invention [Problem to be solved by the invention]

[0006] Until now, the search for paint raw material compounds has involved selecting candidate compounds that could become paint raw material compounds based on past experimental data and researchers' knowledge, then producing paint using the candidate compounds, evaluating them, and deciding whether to accept or reject them.However, because paints contain a wide variety of ingredients and the composition ratios vary widely, searching for the optimal paint raw material compound from among the wide variety of paint raw material compounds requires a huge amount of time, money, and experiments, and improvements were needed. None of Patent Documents 1 to 3 relates to a search method for raw material compounds for paints, and even if they were used to search for raw material compounds for paints, the accuracy of the search would not be satisfactory.

[0007] The problem to be solved by the present invention is to provide a method for searching for compounds for use as raw materials for paints, which can search for raw materials for paints quickly, accurately, and simply. [Means for solving the problem]

[0008] As a result of intensive research to solve the above problems, the inventors have found that the above problems can be solved by a method for searching for raw material compounds for paints that includes specific steps, and have thus completed the present invention. That is, the present invention relates to the following items 1 to 9. [Item 1] A method for searching for paint raw material compounds, comprising: a selection step of selecting candidate paint raw material compounds using a judgment model obtained by machine learning; and a confirmation step of conducting experiments using the candidate paint raw material compounds selected in the selection step to confirm their properties, wherein the paint raw material compounds are one or more selected from the group consisting of curing agents, curing agent components, film-forming resins, additives, pigments, and solvents. [Item 2] The method for searching for paint raw material compounds according to Item 1, wherein the selection step includes a virtual library creation step of creating a focused library for searching for candidate paint raw material compounds as the virtual library. [Item 3] The method for searching for paint raw material compounds according to Item 1 or 2, wherein the selection process includes a virtual library creation stage, a judgment model creation stage, a candidate compound judgment stage for paint raw material compounds, a judgment model suitability judgment stage, and a virtual library suitability judgment stage. [Item 4] The method for searching for a paint raw material compound according to any one of Items 1 to 3, wherein the selection step comprises a step of acquiring a response variable from data related to the paint raw material compound, a step of acquiring explanatory variables by analyzing the data related to the paint raw material compound, and a step of performing machine learning using the response variable and the explanatory variables to create a determination model that predicts the response variable from the explanatory variables. [Item 5] The method for searching for a paint raw material compound according to any one of Items 1 to 4, wherein the selection step involves creating a determination model by machine learning using one or more different algorithms, and selecting and using one or more models with high regression indices or classification indices or consensus models from the created determination models. [Item 6] The method for searching for a paint raw material compound according to any one of Items 1 to 5, wherein the selection step includes a step of using a regression model or a classification model using molecular descriptors as a judgment model to judge whether or not the candidate paint raw material compound satisfies a judgment criterion. [Item 7] A method for searching for paint raw material compounds according to any one of Items 1 to 6, characterized in that, if the paint raw material compound confirmed in the confirmation step does not satisfy the judgment criterion, the suitability of the judgment model is judged, and based on the judgment result, the process returns to the judgment model creation stage, and then the suitability of the virtual library is judged, and based on the judgment result, the process returns to the virtual library creation stage. [Item 8] The method for searching for a raw material compound for a paint according to any one of Items 1 to 7, wherein the raw material compound for a paint is a raw material compound for an electrodeposition paint. [Item 9] The method for searching for a raw material compound for a coating material according to any one of Items 1 to 8, wherein the raw material compound for a coating material is one or more of an ultraviolet absorber, a curing catalyst, or a blocking agent for an isocyanate compound. [Effects of the Invention]

[0009] The present invention provides a method for searching for compounds for use as raw materials for paints, which allows for rapid, accurate and simple searching for raw material compounds for paints. [Brief explanation of the drawings]

[0010] [Figure 1] FIG. 1 is a flow chart of a method for searching for raw material compounds for paint according to one embodiment of the present invention. DETAILED DESCRIPTION OF THE INVENTION

[0011] Hereinafter, embodiments of the present invention will be described in detail with reference to the drawings. It should be understood that the present invention is not limited to the following embodiments, but also includes various modifications that are implemented within the scope of the present invention.

[0012] Fig. 1 is a flow diagram of a method for searching for paint raw material compounds according to one embodiment of the present invention. The method for searching for paint raw material compounds shown in Fig. 1 comprises a selection step of selecting candidate paint raw material compounds and a confirmation step of conducting experiments to confirm the selected candidate paint raw material compounds.

[0013] The selection step is a step of selecting candidate compounds for the paint raw material compounds to proceed to the confirmation step from the virtual library using the created judgment model. In the flow diagram of the method for searching for raw material compounds for paints shown in FIG. 1, after the start, the method proceeds to a virtual library creation stage and a judgment model creation stage, where a virtual library and a judgment model are created. After the virtual library is created, the process proceeds to a virtual library compatibility assessment step as necessary to determine the compatibility of the virtual library. If the virtual library compatibility assessment step fails, the process returns to the virtual library creation step, and if the virtual library passes, the process proceeds to a candidate compound assessment step for paint raw material compounds. After the judgment model is created, the process proceeds to the judgment model suitability assessment stage as necessary to determine the suitability of the judgment model. If the judgment model suitability assessment stage fails, the process returns to the judgment model creation stage, and if it passes, the process proceeds to the candidate compound assessment stage for paint raw material compounds.

[0014] After the virtual library and the judgment model are completed, in the candidate compound judgment stage, candidate compounds for paint raw material compounds to proceed to the confirmation step are judged from the virtual library using the created judgment model. In the candidate compound evaluation stage, if many of the candidate paint raw material compounds are judged to be passable, or if no passable judgment is made, they are deemed to have failed and proceed to the virtual library compatibility evaluation stage and / or the judgment model compatibility evaluation stage.

[0015] In the virtual library compatibility assessment stage, the compatibility of the virtual library is assessed, if necessary, based on the information obtained in the candidate compound assessment stage. If the compatibility is unsuccessful, the process proceeds to the virtual library creation stage, where the virtual library is recreated and modified. In the judgment model suitability assessment stage, the suitability of the judgment model is assessed, if necessary, based on the information obtained in the candidate compound assessment stage. If the model fails, the process proceeds to the judgment model creation stage, where the judgment model is recreated and corrected. If a candidate compound fails in the candidate compound evaluation stage, at least one of the virtual library and the evaluation model will be recreated and modified through the virtual library compatibility evaluation stage and / or the evaluation model compatibility evaluation stage, taking into account the cause of the failure in the candidate compound evaluation stage.

[0016] The selection process is ultimately completed when an appropriate number of paint raw material compounds are judged to be acceptable in the candidate compound evaluation stage.

[0017] The confirmation process is sequentially comprised of a candidate compound acquisition stage for acquiring the candidate compound selected in the selection process, a candidate compound experiment stage, and a candidate compound final judgment stage. If the candidate compound is judged as pass in the candidate compound final judgment stage, the search for paint raw material compounds ends, and if the candidate compound is judged as fail, the process proceeds to a virtual library compatibility judgment stage and / or a judgment model compatibility judgment stage in the selection process.

[0018] In the virtual library compatibility assessment stage, the compatibility of the virtual library is assessed, if necessary, based on the information obtained in the candidate compound assessment stage. If the compatibility is unsuccessful, the process proceeds to the virtual library creation stage, where the virtual library is recreated and modified. In the judgment model suitability assessment stage, the suitability of the judgment model is assessed, if necessary, based on the information obtained in the candidate compound assessment stage. If the model fails, the process proceeds to the judgment model creation stage, where the judgment model is recreated and corrected. If a candidate compound fails in the final evaluation stage, at least one of the virtual library and the evaluation model will be recreated and modified through the virtual library compatibility evaluation stage and / or the evaluation model compatibility evaluation stage, taking into account the cause of the failure in the final evaluation stage of the candidate compound.

[0019] The method for searching for paint raw material compounds of the present invention increases the accuracy of the search by conducting experiments in the confirmation step, and if the confirmation step is failed, the failed experimental data can be used to repeat the search for paint raw material compounds again, thereby increasing the accuracy of prediction. This will be explained in detail below.

[0020] [Raw material compounds for paints] In the method for searching for a raw material compound for a paint of the present invention, the raw material compound for a paint is a raw material compound used in a wide variety of paints. Examples of the paints include vehicle paints, architectural paints, household paints, industrial paints, structural paints, ship bottom paints, conductive paints, etc. Examples of the paints include top coat paints, intermediate paints, base paints, electrodeposition paints, primers, and rust-preventive paints. In the method for searching for a raw material compound for a paint of the present invention, the raw material compound for a paint is preferably a raw material compound for an electrodeposition paint.

[0021] In the method for searching for a paint raw material compound of the present invention, the paint raw material compound is any compound used in constructing a paint. For example, it is one or more compounds selected from the group consisting of a curing agent, a curing agent component, a coating film-forming resin, an additive, a pigment, and a solvent. Examples of curing agent components include a blocking agent for curing agent reactivity. Examples of additives include one or more of an ultraviolet absorber, a light stabilizer, a thickener, an antifoaming agent, a surface conditioner, an anti-settling agent, a rust inhibitor, a chelating agent, a dehydrating agent, a neutralizing agent, a plasticizer, a filler, a pigment, and a curing catalyst. In the method for searching for a raw material compound for a coating material of the present invention, the raw material compound for a coating material is preferably one or more of an ultraviolet absorber, a curing catalyst, or a blocking agent for an isocyanate compound.

[0022] [Selection process] In the method for searching for paint raw material compounds of the present invention, the selection step is a step of selecting candidate paint raw material compounds from the created virtual library using a determination model obtained by machine learning. The selection process includes at least a virtual library creation stage, a judgment model creation stage, a virtual library compatibility determination stage, a judgment model compatibility determination stage, and a candidate compound determination stage for paint raw material compounds. The selection step may include a step of selecting candidate compounds for paint raw material compounds using a regression model or a classification model using molecular descriptors. The step of selecting candidate compounds for paint raw material compounds using a regression model or a classification model using molecular descriptors may include, for example, a candidate compound determination step of using a determination model that is a regression model or a classification model using molecular descriptors to select candidate compounds for paint raw material compounds.

[0023] <Virtual library creation stage> The virtual library creation stage is a stage in which a focused library for searching for candidate compounds for paint raw material compounds is created as a virtual library. A focused library is a library of compounds having partial structures or molecular structures that can be expected to have specific effects or functions as paint. If there are partial structures that are considered preferable as paint raw material compounds or partial structures that should be avoided, it is preferable to create a focused library that reflects that information in the virtual library creation stage. When creating a virtual library, candidate compounds and novel compounds can be selected based on empirical rules. When creating a virtual library, candidate compounds can be selected by searching a database or the like. As the database, in addition to an in-house database, a public database such as PubChem, ZINC, ChemSpider, Dortmund Data Bank, or SPRESI database is preferable.

[0024] In the selection process, the virtual library creation stage may be set before the candidate compound determination stage. The virtual library creation stage may be set before the judgment model creation stage, after the judgment model creation stage, or in parallel with the judgment model creation stage. One embodiment shown in Figure 1 is an embodiment in which the virtual library creation stage is set in parallel with the judgment model creation stage.

[0025] <Decision model creation stage> The determination model creation stage is a stage in which, when selecting a candidate compound for a paint raw material compound, a determination model is created to determine whether or not the selected candidate compound should proceed to the confirmation step. The determination model is preferably a determination model that is obtained by performing machine learning using a dependent variable and an explanatory variable, and that predicts the dependent variable from the explanatory variable. The determination model creation stage may include one or more of the following phases: a phase in which a response variable is obtained from data related to a paint raw material compound; and a phase in which an explanatory variable is obtained by analyzing data related to a paint raw material compound.

[0026] In the method for searching for a raw material compound for a paint of the present invention, the objective variable is a variable for evaluating whether or not the raw material compound for a paint constitutes a paint that fulfills the intended purpose (satisfies the desired characteristics). Examples of the objective variables include one or more of the following: properties of a coating film (gel fraction, elongation, gloss, haze, light transmittance, storage modulus, loss tangent, solar reflectance, ultraviolet transmittance, heat resistance, water resistance, color tone, water resistance, acid resistance, chemical resistance, weather resistance, surface roughness, gloss, luster, smoothness, adhesion, contact angle, scratch resistance, electrical resistance, stain resistance, light reflectance, flip-flop property, micro-glossiness, coating film appearance, standards related to coating films, etc.); properties of a paint (storage stability, low-temperature curing property, blocking agent dissociation temperature, viscosity, dispersibility, defoaming property, foaming property, color, viscoelasticity, stringiness, sagging property, surface tension, solids concentration, pH, repelling property, standards related to paints, etc.); and properties of paint constituents (constituent atoms, environmental properties, heat resistance, reactivity, toxicity, cost, availability, laws and regulations, safety, etc.).

[0027] In the method for searching for paint raw material compounds of the present invention, explanatory variables are variables that affect one or more of the paint raw material compounds, the properties of the paint using the same, or the objective variable. Examples of explanatory variables include the type of paint constituent component, the content ratio of the paint constituent component, the properties of the paint constituent component (constituent atoms, environmental properties, heat resistance, reactivity, toxicity, safety, cost, availability, legal regulations, molecular weight, surface tension, contact angle, solid content, pH, viscosity, SP value, solubility, vapor pressure, evaporation rate, particle size, etc.), paint physical properties (storage stability, low-temperature curing property, blocking agent dissociation temperature, viscosity, dispersibility, defoaming property, foaming property, color, viscoelasticity, stringiness, sagging property, surface tension, solid content, pH, repelling property, paint specifications, etc.), paint manufacturing conditions (stirring conditions, manufacturing temperature, etc.), etc. ), coating conditions (temperature, humidity, film thickness, drying temperature, etc.), paint use (electrodeposition paint, primer paint, top coat paint, base coat paint, glitter paint, primer, etc.), paint film properties (film thickness, layer structure, elongation, gloss, haze, light transmittance, storage modulus, loss tangent, solar reflectance, ultraviolet transmittance, heat resistance, water resistance, color tone, water resistance, acid resistance, chemical resistance, weather resistance, surface roughness, gloss, luster, smoothness, adhesion, contact angle, scratch resistance, electrical resistance, stain resistance, light reflectance, flip-flop properties, micro-glossiness, paint film appearance, paint film standards, etc.).

[0028] Molecular descriptors can be used as explanatory variables, which allows for the quantification of explanatory variables and is advantageous when generating a decision model using machine learning.

[0029] Molecular descriptors are numerical representations of molecular structures and consist of experimentally measured values, 0-dimensional descriptors, 1-dimensional descriptors, 2-dimensional descriptors, 3-dimensional descriptors, and 4-dimensional descriptors.

[0030] The experimentally measured values ​​are descriptors that are regarded as universal properties, such as log P, refractive index, dipole moment, polarity, molecular weight, and NMR shift.

[0031] The zero-dimensional descriptor is a descriptor that can be considered as a constituent descriptor or a count descriptor, and examples thereof include the number of atom types, bond patterns, the number of atoms such as C, H, O, N, and halogens, the number of heavy atoms, the number of bonds, the number of rotatable bonds, and the number of multiple bonds.

[0032] One-dimensional descriptors are descriptors that are positioned as fragment lists or fingerprints, and examples thereof include the number of hydrogen bond donors or acceptors, the number of rings, the number of functional groups, and fingerprints (structure keys, hashed fingerprints).

[0033] Two-dimensional descriptors are descriptors that are positioned as topological indices, and examples include the topological index (an index that quantifies the connectivity pattern based on an adjacency matrix) and the Wiener index (the sum of the shortest distances connecting atoms within a molecule).

[0034] Three-dimensional descriptors are descriptors that can be classified as geometric descriptors or quantum chemical descriptors, and examples thereof include geometric descriptors, polar surface areas, 3D-MoRSE descriptors (3D molecular representations based on electron diffraction), WHIM descriptors (calculated from atomic coordinates of molecular structures), GETAWAY descriptors (calculated by calculating molecular influence matrices in combination with geometric atomic distances), HOMO / LUMO, charge, dipole moment, polarizability, binding energy, reaction heat, activation barriers, IR frequency and intensity, and NMR chemical shifts.

[0035] Four-dimensional descriptors are positioned as interaction descriptors, and examples include the GRID method (exhaustive conformational search and narrowing down of active conformation candidates), the CoMFA method (a regression analysis method for predicting interactions between proteins and small molecules), and Volsurf (calculation from the interaction energy between a molecule and atoms placed at its surrounding grid points). Among these molecular descriptors, those obtained by calculation can be obtained from compound information in SMILES notation obtained from the above-mentioned public databases using publicly known and publicly used RDKit, Mordred, or commercially available programs such as alvaDesc and Codessa. When using the results of quantum chemical calculations as molecular descriptors, programs such as Gaussian, GAMESS, Jaguar, Q-Chem, and Spartan can also be used.

[0036] It is not easy to directly link the structure and composition of a material with its physical properties and characteristics. For this reason, molecular descriptors, which are proxy indices that govern physical properties, are often used to clarify the relationship between the structure and composition of the material and the proxy indices before predicting physical properties. Utilizing, examining, and exploring these proxy indices is important for successfully creating a judgment model. Machine learning involves extracting single or multiple proxy indices (molecular descriptors) that correlate with the physical property values ​​of the material, which are the target variables. Creating a more accurate machine learning model leads to the selection of better candidate compounds.

[0037] In the determination model creation stage, the algorithm used for machine learning to create the determination model is not particularly limited, regardless of whether it is regression or classification, as long as it is a machine learning algorithm capable of creating a determination model. Examples of regression algorithms that can be used include CatBoost regression, gradient boosting regression, extratree regression, random forest regression, XGBoost regression, LiteGBM regression, Adaboost regression, linear regression, ridge regression, least-angle regression, Bayesian linear regression, decision tree regression, elastic net regression, lasso regression, Huber regression, orthogonal matching pursuit, K-nearest neighbor regression, lasso least-angle regression, PA, automatic relevance determination, RANSAC, Theilsen regression, kernel ridge regression, support vector machine regression, multilayer perceptron, partial least squares regression, Gaussian process regression, etc., or a consensus model using one or more algorithms. The consensus regression model is a model that calculates the average value of models using the regression algorithms with the highest regression index (1 or higher). Classification algorithms that can be used include random forest classification, LiteGBM classification, XGBoost classification, CatBoost classification, ExtraTree classification, gradient boosting classification, decision tree classification, Adaboost classification, K-nearest neighbor classification, logistic regression, linear discriminant analysis, ridge classification, support vector machine (linear kernel), naive Bayes, quadratic discriminant analysis, etc., or a consensus model using one or more algorithms. A classification consensus model is a model that calculates a majority vote of models using one or more of the classification algorithms with the highest classification index. Regardless of whether it is regression or classification, it is preferable to optimize hyperparameters for each algorithm. Regression indices that can be used include the coefficient of determination (R2), mean absolute error (MAE), mean squared error (MSE), root mean squared error (RMSE), and root mean squared logarithmic error (RMSLE).Classification indices that can be used include accuracy, area under the curve (AUC), recall, precision, F1 (F1-score), kappa coefficient, and Matthews correlation coefficient (MCC).

[0038] When creating a decision model, an appropriate machine learning framework / library can be used depending on the data scale, execution environment, etc. The machine learning framework / library may be one that has been developed, or one or more available ones can be used. Examples of machine learning framework / libraries that can be used include one or more of TensorFlow, PyTorch, Keras, Caffe, Microsoft Cognitive Toolkit, MXNet, Chainer, Scikit-learn, OpenCV, spaCy, Caret, and PyCaret.

[0039] In machine learning, input of the target variable and explanatory variables can be performed by transmitting data via wired, wireless, or a combination of these communication means, or via a recording medium. Examples of input via a communication means include a combination of one or more of various communication networks such as a LAN (local area network), a WAN (wide area network), the Internet, and a telephone network. Input via a recording medium can be performed by reading data from a recording medium such as a magnetic recording medium, an optical recording medium, or a paper recording medium using an appropriate reading means.

[0040] In machine learning, the reliability of the objective variable and explanatory variables can be increased by improving consistency (consistency of labeling standards), reproducibility (highly reproducible labeling for the same data), and transparency (clarity of the data creation process and the basis for labeling).

[0041] In machine learning, it is preferable to preprocess data related to the objective variable and explanatory variables. Preprocessing can include processing such as filling in missing data, filtering such as removing outliers, encoding to convert qualitative variables (character data) into quantitative variables (numeric data), normalization / standardization of data, and arithmetic processing (logarithmic processing, exponentiation, etc.). This allows the data to be commonly applicable to various machine learning algorithms and can also be specialized for specific machine learning algorithms.

[0042] In the selection process, a decision model can be created by machine learning using one or more different algorithms. In the selection step, one or more models with high regression indices or classification evaluation indices or consensus models can be selected and used from the created determination models. In the selection process, a judgment model is created by machine learning using one or more different algorithms, and one or more of the created judgment models with high regression indexes or classification indexes or consensus models can be selected and used. In the selection process, the judgment model creation stage may be set before the candidate compound judgment stage. The judgment model creation stage may be set before the virtual library creation stage, after the virtual library creation stage, or in parallel with the virtual library creation stage. One embodiment shown in Figure 1 is an embodiment in which the judgment model creation stage is set in parallel with the virtual library creation stage.

[0043] <Virtual library compatibility assessment stage> The virtual library suitability assessment stage is a stage in which, if a candidate compound fails the candidate compound assessment stage in the selection process or the candidate compound final assessment stage in the confirmation process, it is judged whether it is suitable as a focused library (virtual library) to which the assessment model is applied, based on previous knowledge and newly obtained knowledge.Since no candidate compound satisfying the initial properties can be obtained using a virtual library, the virtual library creation stage is a very important step.

[0044] In the selection process, the virtual library compatibility determination stage may be performed after the virtual library creation stage, and may be performed before or after the judgment model creation stage. Alternatively, the virtual library compatibility determination stage may be performed in parallel with the judgment model creation stage.

[0045] <Decision model compatibility assessment stage> The judgment model suitability assessment stage is a stage in which, when a candidate compound fails the candidate compound assessment stage in the selection process or the candidate compound final assessment stage in the confirmation process, the accuracy of the judgment model is examined and it is determined whether or not the judgment model needs to be recreated (improved).

[0046] In the candidate compound determination stage in the selection process or the candidate compound final determination stage in the confirmation process, explanatory variables for each candidate compound are obtained. In particular, in the candidate compound final determination stage in the confirmation process, the correspondence between the explanatory variables and the objective variable is obtained. In the judgment model suitability assessment stage, the necessity of recreating the judgment model (improving the judgment model) is considered based on these explanatory variables or the correspondence between the explanatory variables and the objective variable.

[0047] In the method for searching for raw material compounds for paints of the present invention, if a candidate compound fails the final judgment stage in the confirmation process, the judgment model can be repeatedly recreated (improved) to further improve the predictive accuracy of the judgment model.

[0048] In the judgment model compatibility assessment stage, if it is determined that the recreation of the judgment model is unnecessary, or if it is determined that the recreation of the judgment model is necessary and the judgment model is created, the process proceeds to the virtual library compatibility assessment and creation stage, and if it is determined that the recreation of the virtual library is necessary, a focused library of a new candidate compound is created and the process proceeds to the candidate compound assessment stage, and if it is determined that the recreation of the virtual library is unnecessary, the process proceeds to the candidate compound assessment stage. Note that, when returning to the virtual library creation stage, it is preferable to create a library using a method other than the library creation method used previously (for example, a compound having a different partial structure, etc.). If it is necessary to recreate the judgment model, an experiment may be conducted using the failed candidate compounds to obtain the objective variables, and the obtained objective variables may be used to recreate the judgment model (improve the judgment model).

[0049] If it is determined that the judgment model needs to be recreated in the judgment model suitability assessment stage, the process returns to the judgment model creation stage, where machine learning is performed again to recreate the judgment model (improve the judgment model). Furthermore, if a candidate compound fails the final judgment stage in the confirmation process, it has been experimentally proven that it does not exhibit the desired properties. Therefore, by adding the obtained correspondence between the explanatory variables and the target variables to the data used to create the judgment model and performing machine learning, it is possible to improve the trained judgment model and increase the accuracy of the judgment model.

[0050] <Candidate compound evaluation stage> The candidate compound judgment step is a step of predicting whether or not a candidate compound in the virtual library satisfies the desired properties using the judgment model created in the judgment model creation step. In the candidate compound determination stage, in addition to determination using a determination model, other considerations such as availability, ease of acquisition, suitability for synthesis, price, environmental characteristics, and safety can be used in the determination. The determination model used in the candidate compound determination step is preferably a machine learning model that includes explanatory variables as inputs and outputs a response variable. In the candidate compound determination stage, it is preferable to predict the objective variable by inputting the explanatory variables of the candidate compound selected from the virtual library into a determination model, and to pass the candidate compound that satisfies the objective variable threshold, and to reject the candidate compound that does not satisfy the objective variable threshold. In the candidate compound determination step, a large number of candidate compounds can be easily determined by using electronic devices such as calculators and computers. If there are a large number of candidate compounds that satisfy the desired threshold, the number may be narrowed down by changing the threshold.

[0051] If the candidate compound passes the candidate compound evaluation stage, it will proceed to the confirmation step. If the candidate compound fails the candidate compound evaluation stage, it will proceed to the evaluation model suitability evaluation stage, and then to the virtual library suitability evaluation stage.

[0052] [Confirmation process] In the method for searching for raw material compounds for paints of the present invention, the confirmation step includes, in order, a candidate compound obtaining step for obtaining the candidate compound selected in the selection step, a candidate compound experiment step, and a candidate compound final judgment step.

[0053] <Candidate compound acquisition stage> The candidate compound obtaining step is a step of specifically obtaining the candidate compounds for the paint raw material compounds selected in the selection step. The candidate compounds can be obtained by one or more of the following means: commercial acquisition based on a catalog or the like; synthesis by referring to literature or the like; and the like.

[0054] <Candidate compound experimental stage> The candidate compound experiment stage is a stage in which, after specifically obtaining the candidate compounds of the paint raw material compounds selected in the selection step, a paint is prepared using the candidate compounds of the paint raw material compounds, and data such as the properties of the candidate compounds, the properties of the paint film, etc. are obtained. In the candidate compound experiment stage, specific numerical values ​​of the objective variables are obtained.

[0055] <Final evaluation stage of candidate compounds> The final candidate compound evaluation stage is a stage in which the candidate compounds selected in the selection stage are obtained and subjected to experiments to confirm the objective variables obtained, and to determine whether or not they can be used as raw material compounds for paints. In the final candidate compound evaluation stage, the usability can be determined based on the objective variables, which are the characteristic values ​​obtained in the candidate compound experiment stage.

[0056] If the candidate compound is judged to be acceptable in the final evaluation stage, the search for raw material compounds for paints will be completed. If the candidate compound fails the final evaluation stage because it does not satisfy the evaluation criteria, the selection process proceeds to the evaluation model suitability evaluation stage, where the suitability of the evaluation model is evaluated, and based on the evaluation results, the process proceeds to the evaluation model creation stage, and then to the virtual library suitability evaluation stage, where the suitability of the virtual library is evaluated, and based on the evaluation results, the process proceeds to the virtual library creation stage.

[0057] In the final stage of determining the candidate compounds, electronic calculators, computers, etc. can be used. When creating and recreating a virtual library, candidate compounds can be selected based on empirical rules. When creating or recreating a virtual library, candidate compounds can be selected by searching a database or the like. When selecting candidate compounds by searching, a step of searching for candidate compounds of raw material compounds for paints using a machine learning model using molecular descriptors may be included. The virtual library creation step may identify compounds with novel structures. In the virtual library creation and virtual library re-creation stages, explanatory variables of the selected candidate compounds are obtained.

[0058] [Applications of the method for searching for raw material compounds for paints] The method for searching for paint raw material compounds of the present invention makes it possible to predict the state of a paint to be produced and a coating film produced from the paint when a specific paint raw material compound is used as a paint raw material, and therefore can be suitably used in applications where it is necessary to produce paints and coating films with desired properties or to predict changes in the properties of the produced paints and coating films. For example, it can be used to speed up the development of paints and coatings that meet required properties, and to predict the functions and properties of coatings, making it useful when developing and proposing paints.

[0059] [System for searching for raw material compounds for paints] The system for performing the method for searching for paint raw material compounds of the present invention is a system equipped with a judgment model for judging whether a paint raw material compound satisfies predetermined properties. The judgment model is generated by machine learning using data in which a target variable and an explanatory variable are associated with each other.

[0060] An example of a system for performing the method for searching for paint raw material compounds is one configured within a computer. Specifically, the system includes a decision model (machine learning model), a target variable prediction unit, a virtual library database, a reception unit, a display unit, and a recording unit. The database may be an external database provided via a communication line. If necessary, the system may be connectable to an external user terminal (one or more of a personal computer, a mobile phone (smartphone), a personal digital assistant, etc.) via the communication line. The display unit may be connectable to a virtual space, and may be configured to be able to display, in the virtual space, a predicted state of a paint or coating film obtained using the paint raw material compounds. External users may be charged at least in part for their use of the system or for some of their operations when using the system. [Example]

[0061] [Search for blocking agents for blocked polyisocyanates] An example will be shown in which a blocking agent for a blocked polyisocyanate used in a thermosetting coating was searched for using the method for searching for a raw material compound for a coating of the present invention.

[0062] Using data from past in-house experiments and data from publicly available literature, a dataset was created in which the curing start temperature and curing end temperature of the blocking agent in blocked polyisocyanate were used as objective variables and the various molecular descriptors of the blocking agent were used as explanatory variables. Next, machine learning was performed using all applicable algorithms, and ultimately, random forest, which had the smallest mean absolute error (MAE), was used as the algorithm. After optimizing the hyperparameters, a decision model (machine learning model) was created in which the curing start temperature and curing end temperature of the blocking agent in blocked polyisocyanate were used as objective variables.

[0063] A database (PubChem: approximately 120 million compounds) was searched for partial structures extracted from blocking agents that have been used to date, and 1,428 compounds were selected as a virtual library after filtering by molecular weight and number of heteroatoms. Reluctant structures were then removed, and 882 compounds were selected. The random forest model described above was used to determine which 882 selected compounds passed the test, with blocked polyisocyanates having curing start and end temperatures of 140°C or less. Nine compounds were selected.

[0064] Experiments were conducted on the nine selected compounds to confirm their properties, but all of them failed. The judgment model was deemed to be suitable, but the virtual library was deemed to be unsuitable, so the virtual library was recreated.

[0065] A search was performed using a database (PubChem) with a different partial structure as a search key, filtering by molecular weight and structural complexity to select 8,128 compounds to create a virtual library. Next, reluctant structures were removed, and 7,238 compounds were selected. Of the 7,238 selected compounds, a random forest model was used to determine whether they passed the test if their curing start temperature was 90°C or less and their curing end temperature was 130°C or less, and 668 compounds were selected. Of the 668 selected compounds, 38 were selected based on availability, and four were selected from the selected 38 compounds based on price and suitability for synthesis.

[0066] The four selected compounds were actually obtained and experiments were carried out to measure the curing initiation temperature and curing completion temperature of the blocking agent when they were used to form blocked polyisocyanates. Ultimately, two compounds were identified: N-methylaniline (curing start temperature 122°C, curing completion temperature 134°C) and tetrahydroquinoline (curing start temperature 109°C, curing completion temperature 125°C). Conventional blocking agents have a curing start temperature of 151°C and a curing completion temperature of 168°C.

[0067] Blocked polyisocyanates were synthesized using N-methylaniline or tetrahydroquinoline as a blocking agent, and coating materials were prepared using these. Each of the resulting paints was applied to a chemically treated cold-rolled steel plate as the substrate, baked at 130°C for 20 minutes to harden, and the storage stability and finish were confirmed to be equivalent to those of conventional paints.

Claims

1. a selection step of selecting candidate compounds for paint raw material compounds using a determination model obtained by machine learning; and a confirmation step of conducting experiments to confirm the properties of the candidate compounds of the paint raw material compounds selected in the selection step; A method for searching for a raw material compound for a paint, comprising: The selection step is performed by a computer, The selection step includes a step of determining whether or not a candidate compound for a paint raw material compound satisfies a determination criterion, the selection step includes at least a virtual library creation step, a judgment model creation step, a virtual library compatibility determination step, a judgment model compatibility determination step, and a candidate compound determination step for a paint raw material compound, the determination model creation step is a step of creating a determination model that predicts the dependent variable from the explanatory variable by performing machine learning using the dependent variable and the explanatory variable; the objective variable is one or more of a coating film characteristic, a paint characteristic, and a paint component characteristic; The explanatory variables are one or more of the following: type of paint constituent component, content ratio of paint constituent component, characteristics of paint constituent component, physical properties of paint, manufacturing conditions of paint, painting conditions, use of paint, characteristics of paint film, and molecular descriptor; The coating material compound is at least one selected from the group consisting of a curing agent, a curing agent component, a coating film-forming resin, an additive, a pigment, and a solvent. The method for searching for raw material compounds for paint.

2. 2. The method for searching for paint raw material compounds according to claim 1, wherein the selection step comprises a virtual library creation step of creating a focused library for searching candidate paint raw material compounds as the virtual library.

3. 3. The method for searching for a paint raw material compound according to claim 1 or 2, wherein the selection step comprises a step of acquiring a response variable from data related to the paint raw material compound, a step of acquiring explanatory variables by analyzing the data related to the paint raw material compound, and a step of performing machine learning using the response variable and the explanatory variables to create a determination model that predicts the response variable from the explanatory variables.

4. 3. The method for searching for paint raw material compounds according to claim 1 or 2, wherein the selection step creates a determination model by machine learning using one or more different algorithms, and selects and uses one or more models having high regression indices or classification indices or consensus models from the created determination models.

5. 3. The method for searching for paint raw material compounds according to claim 1 or 2, wherein the selection step includes a step of using a regression model or a classification model using molecular descriptors as a judgment model to judge whether or not the candidate paint raw material compounds satisfy a judgment criterion.

6. 3. The method for searching for paint raw material compounds according to claim 1 or 2, characterized in that, if the paint raw material compound confirmed in the confirmation step does not satisfy the judgment criterion, the suitability of the judgment model is judged, and based on the judgment result, the process returns to the judgment model creation stage, and then the suitability of the virtual library is judged, and based on the judgment result, the process returns to the virtual library creation stage.

7. 3. The method for searching for a raw material compound for a paint according to claim 1, wherein the raw material compound for a paint is a raw material compound for an electrodeposition paint.

8. 3. The method for searching for a raw material compound for a paint according to claim 1 or 2, wherein the raw material compound for a paint is one or more of an ultraviolet absorber, a curing catalyst, or a blocking agent for an isocyanate compound.

Images