Property prediction model generation method, property prediction model generation apparatus, property prediction method, property prediction apparatus, property prediction program, composition prediction method, composition prediction apparatus, and composition prediction program

Abstract

To provide a technique for standardizing processes of arithmetic processing for generating features, thereby improving the convenience for a user.SOLUTION: A property prediction model generation method includes: a data mart generation step of generating a data mart which includes the amount of composition of a plurality of materials for use in generating rubber composition, property values of the materials, and physical properties of the rubber composition; a base computation generation step of generating a base computation table in which materials and arithmetic processing for the materials are described; a base computation step which calculates features based on the arithmetic processing in the base computation table; a training data generation step of generating a training data mart which is composed of at least one of the features calculated in the base computation step, and data included in the data mart; and a training step of training a learning-type arithmetic model configured to predict physical properties of rubber composition, based on the training data mart.SELECTED DRAWING: Figure 6

Description

Technical Field

[0001] The present invention relates to a method for generating a physical property prediction model in a rubber compound, a physical property prediction model generation device, a physical property prediction method, a physical property prediction device, a physical property prediction program, a formulation prediction method, a formulation prediction device, and a formulation prediction program.

Background Art

[0002] For example, a rubber compound used for a tire or the like mounted on a vehicle is a composite material in which a polymer, which is a rubber material as a main raw material, is added with a reinforcing agent and various chemicals. Rubber compounds are formulated based on various materials according to their uses, and those having different physical properties have been developed.

[0003] Patent Document 1 discloses a conventional physical property prediction system. The physical property prediction system includes a prediction device, an information terminal, and a database. The performance prediction unit of the prediction device predicts the value of the physical property of the rubber to be predicted based on the material information regarding a plurality of materials constituting the rubber to be predicted. The display processing unit of the information terminal associates the predicted value predicted by the performance prediction unit with the feature amount of a specific material among the plurality of materials and displays the same on the display unit.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Summary of the Invention

Problems to be Solved by the Invention

[0005] When extracting the feature amounts of the materials used in the formulation of a rubber compound to predict the physical properties, the feature amounts have been conventionally derived based on experience and knowledge, and there has been a problem that the prediction of the physical properties is complicated. Further, when generating a large number of feature amounts for predicting the physical properties, a high-level data processing ability regarding the materials is required, and there has been a concern that the convenience in performing the physical property prediction is reduced.

[0006] The present invention has been made in view of such circumstances, and an object thereof is to provide a physical property prediction model generation method, a physical property prediction model generation apparatus, a physical property prediction method, a physical property prediction apparatus, a physical property prediction program, a formulation prediction method, a formulation prediction apparatus, and a formulation prediction program that can share the process of arithmetic processing for generating feature quantities and improve user convenience.

Means for Solving the Problems

[0007] One aspect of the present invention is a physical property prediction model generation method. The physical property prediction model generation method includes a data mart generation step of generating a data mart including the blending amounts of a plurality of materials used in the production of a rubber compound, the property values of the materials, and the physical properties of the rubber compound, a base operation generation step of generating a base operation table describing the materials and the arithmetic processing for the materials, a base operation step of calculating feature quantities based on the arithmetic processing by the base operation table, a learning data generation step of generating a learning data mart including at least one of the feature quantities calculated by the base operation step and the data included in the data mart, and a learning step of training a learning-type arithmetic model for predicting the physical properties of a rubber compound based on the learning data mart.

[0008] Another aspect of the present invention is a physical property prediction model generation apparatus. The physical property prediction model generation apparatus includes a data mart generation unit that generates a data mart including the blending amounts of a plurality of materials used in the production of a rubber compound, the property values of the materials, and the physical properties of the rubber compound, a base operation unit that generates a base operation table describing the materials and the arithmetic processing for the materials and calculates feature quantities based on the arithmetic processing by the base operation table, a learning data generation unit that generates a learning data mart by adding the feature quantities calculated by the base operation unit to the data mart, and a learning processing unit that trains a learning-type arithmetic model for predicting the physical properties of a rubber compound based on the learning data mart.

[0009] Another aspect of the present invention is a physical property prediction method. The physical property prediction method includes a data mart generation step of generating a data mart including compounding amounts of a plurality of materials used in the production of a rubber compound and property values of the materials, a base operation generation step of generating a base operation table describing the materials and arithmetic processing for the materials, a base operation step of calculating feature amounts based on the arithmetic processing by the base operation table, and a physical property prediction step of inputting at least one of the feature amounts calculated by the base operation step and the data included in the data mart into a learned arithmetic model to predict the physical properties of the rubber compound.

[0010] Another aspect of the present invention is a physical property prediction device. The physical property prediction device includes a data mart generation unit that generates a data mart including compounding amounts of a plurality of materials used in the production of a rubber compound and property values of the materials, a base operation unit that generates a base operation table describing the materials and arithmetic processing for the materials and calculates feature amounts based on the arithmetic processing by the base operation table, and a physical property prediction processing unit that inputs at least one of the feature amounts calculated by the base operation unit and the data included in the data mart into a learned arithmetic model to predict the physical properties of the rubber compound.

[0011] Another aspect of the present invention is a physical property prediction program. The physical property prediction program causes a computer to execute a data mart generation step of generating a data mart including compounding amounts of a plurality of materials used in the production of a rubber compound and property values of the materials, a base operation generation step of generating a base operation table describing the materials and arithmetic processing for the materials, a base operation step of calculating feature amounts based on the arithmetic processing by the base operation table, and a physical property prediction step of inputting at least one of the feature amounts calculated by the base operation step and the data included in the data mart into a learned arithmetic model to predict the physical properties of the rubber compound.

[0012] Another aspect of the present invention is a compounding prediction method. The compounding prediction method includes a base operation generation step of generating a base operation table that describes materials and arithmetic processing for the materials, a base operation step of calculating feature quantities based on the arithmetic processing by the base operation table, and using at least one of the feature quantities calculated in the base operation step as an explanatory variable and applying an optimization method based on a target physical property and constraint conditions regarding the compounding amount of the materials and the property values of the materials in a rubber compound to predict the compounding of the rubber compound in a compounding prediction step.

[0013] Another aspect of the present invention is a compounding prediction device. The compounding prediction device includes a base operation unit that generates a base operation table that describes materials and arithmetic processing for the materials and calculates feature quantities based on the arithmetic processing by the base operation table, and a compounding prediction processing unit that uses at least one of the feature quantities calculated by the base operation unit as an explanatory variable and applies an optimization method based on a target physical property and constraint conditions regarding the compounding amount of the materials and the property values of the materials in a rubber compound to predict the compounding of the rubber compound.

[0014] Another aspect of the present invention is a compounding prediction program. The compounding prediction program causes a computer to execute a base operation generation step of generating a base operation table that describes materials and arithmetic processing for the materials, a base operation step of calculating feature quantities based on the arithmetic processing by the base operation table, and a compounding prediction step of using at least one of the feature quantities calculated in the base operation step as an explanatory variable and applying an optimization method based on a target physical property and constraint conditions regarding the compounding amount of the materials and the property values of the materials in a rubber compound to predict the compounding of the rubber compound.

Advantages of the Invention

[0015] According to the present invention, the process of arithmetic processing for generating feature quantities can be shared, improving the convenience for users.

Brief Description of the Drawings

[0016]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Figure 12

MODE FOR CARRYING OUT THE INVENTION

[0017] Hereinafter, the present invention will be described based on preferred embodiments with reference to FIGS. 1 to 12. The same or equivalent components and members shown in each drawing are denoted by the same reference numerals, and redundant explanations will be omitted as appropriate. Also, the dimensions of the members in each drawing are appropriately enlarged or reduced for easy understanding. Further, some of the members that are not important for explaining the embodiments in each drawing are omitted from the display.

[0018] (Embodiment 1) FIG. 1 is a block diagram showing the functional configuration of a physical property prediction model generation device 100 according to Embodiment 1. The physical property prediction model generation device 100 includes a storage unit 10, an operation unit 31, a display unit 32, and an arithmetic processing unit 40, and generates an arithmetic model for predicting the physical properties of a rubber compound used in, for example, a tire or the like. The physical property prediction model generation device 100 is an information processing device such as a PC (personal computer). Each part in the physical property prediction model generation device 100 can be realized by an electronic processing circuit or mechanical parts composed of electronic elements including a computer CPU in terms of hardware, and can be realized by a computer program or the like in terms of software. Here, however, functional blocks realized by the cooperation of these are depicted. Therefore, it is understood by those skilled in the art that these functional blocks can be realized in various forms by a combination of hardware and software.

[0019] The storage unit 10 is a storage device composed of, for example, an SSD (Solid State Drive), a hard disk, a CD-ROM, a DVD, or the like. The storage unit 10 stores a data mart 11 used for predicting the physical properties of a rubber compound, a table 21 used for calculations, a computer program executed by the arithmetic processing unit 40, and data used for the execution of the computer program.

[0020] The data mart 11 includes formulation data 12, physical property data 13, raw material data 14, and processing condition data 15 corresponding to each of a plurality of rubber formulations. In the data mart 11, for example, identification information is given to each of the plurality of rubber formulations by means of a serial number or the like, and the formulation, physical properties, property values of raw materials, and processing conditions corresponding to each identification information are included in the formulation data 12, physical property data 13, raw material data 14, and processing condition data 15.

[0021] The formulation data 12 is a data group indicating the formulation amounts of each material in a plurality of rubber formulations. The formulation data 12 includes the names and formulation amounts of rubber materials, reinforcing agents, and various chemicals as the main components corresponding to each of the plurality of rubber formulations. The rubber materials are, for example, natural rubber (NR), butadiene rubber (BR), styrene-butadiene rubber (SBR), isobutylene-isoprene-styrene rubber (IIR), and the like. The reinforcing materials (fillers) are, for example, carbon black, silica, and the like. The various chemicals are, for example, sulfur, wax, and the like. Note that the formulation amounts are represented by the volume, weight, and ratios of the materials.

[0022] The physical property data 13 is a data group such as hardness Hs, tensile stress S, loss tangent tanδ, abrasion resistance, rebound resilience, etc. in each of the plurality of rubber formulations. The raw material data 14 is a data group such as the property values and categorical variables of each material used in each of the plurality of rubber formulations. The property values are, for example, DBP (Dibutyl Phthalate) absorption amount, carbon surface area, etc., and the categorical variables are, for example, ASTM grade of carbon, etc.

[0023] The processing condition data 15 is a data group including conditions such as time, temperature, the name of the machine during processing, and the name of the processing method in each processing process such as kneading, molding, and vulcanization during the production of each of the plurality of rubber compounds. The data may be in the form of numerical variables or categorical variables. The data mart 11 may acquire the formulation data 12, physical property data 13, raw material data 14, and processing condition data 15 of the plurality of rubber compounds from the external device 90. The external device 90 is an information processing device such as a PC, and is assumed to have a storage device for storing the formulation data 12, physical property data 13, raw material data 14, and processing condition data 15 of the plurality of rubber compounds.

[0024] Table 21 includes a base operation table 22, a material group table 23, a material category table 24, an analysis group table 25, an analysis category table 26, and a composite operation table 27 used in the arithmetic processing unit 40. These tables will be described later.

[0025] The operation unit 31 has an operable input device such as a touch panel, a switch, a keyboard, and a mouse device, and receives a user's operation input. The operation unit 31 receives a user's operation input in the generation of the physical property prediction model of the rubber compound, the physical property prediction, and the formulation prediction.

[0026] The display unit 32 has a display device such as a liquid crystal display, and displays a screen for receiving various data and a user's operation input in the process of generating the physical property prediction model of the rubber compound. The display unit 32 also displays a screen for receiving various data and a user's operation input in the physical property prediction and formulation prediction of the rubber compound.

[0027] The arithmetic processing unit 40 includes a data mart generation unit 41, a base arithmetic unit 42, a composite arithmetic unit 43, a learning data generation unit 44, a physical property prediction processing unit 45, and a learning processing unit 46. The arithmetic processing unit 40 is an electronic circuit that executes arithmetic processing such as a CPU, and functions by reading and executing a computer program and data stored in the storage unit 10. The data mart generation unit 41, the base arithmetic unit 42, the composite arithmetic unit 43, the learning data generation unit 44, the physical property prediction processing unit 45, and the learning processing unit 46 in the arithmetic processing unit 40 may be constructed as a plurality of program modules formed by a computer program.

[0028] The data mart generation unit 41 acquires the formulation data 12, physical property data 13, raw material data 14, and processing condition data 15 of a plurality of rubber formulations from the external device 90 and generates a data mart 11. Further, the data mart generation unit 41 may set, as the data mart 11, all or part of a database of the formulations, physical properties, property values of raw materials, and categorical variables of a plurality of rubber formulations, and processing conditions, etc., stored in the storage unit 10 in advance for use in generating a physical property prediction model. The data mart generation unit 41 may receive an operation input of a user who designates a rubber formulation to be included in the data mart 11 by the operation unit 31 and generate the data mart 11.

[0029] The base arithmetic unit 42 generates a base arithmetic table 22 that describes the materials used in the formulation of the rubber formulation and the arithmetic processing for the materials. The base arithmetic unit 42 calculates feature amounts based on the arithmetic processing by the generated base arithmetic table 22.

[0030] Figure 2(a) is a chart showing an example of the base arithmetic table 22, and Figure 2(b) is a chart showing an example of the feature amounts calculated by the base arithmetic table 22. As shown in Figure 2(a), the base arithmetic table 22 is composed of an action number, a material group, a numerical analysis group, and action contents.

[0031] In the base operation table 22 generated by the base operation unit 42, for example, for action number A1, material group G1 is specified to perform an operation of calculating the weighted average value W_avg based on the blending amount as the numerical analysis group SG1 and the action content. The action content for action number A2 etc. is an operation of calculating the total value SUM by addition. The action content for action number A4 is an operation of clustering by principal component analysis PCA and extracting the principal components. The action content in the base operation table 22 includes special numerical analysis operation processes such as arithmetic operations, weighted average, principal component analysis, and t-distribution type probabilistic neighborhood embedding method (t-SNE).

[0032] The base operation unit 42 calculates the feature amounts shown in Fig. 2(b) by executing the operation process according to the base operation table 22 shown in Fig. 2(a). Fig. 3(a) is a chart showing an example of the material group table 23 that specifies the correspondence between the material group and the material target number, and Fig. 3(b) is a chart showing an example of the material category table 24 that specifies the correspondence between the material target number and the material category. Fig. 4(a) is a chart showing an example of the analysis group table 25 that specifies the correspondence between the numerical analysis group and the analysis target number, and Fig. 4(b) is a chart showing an example of the analysis category table 26 that specifies the correspondence between the analysis target number and the numerical analysis category.

[0033] As shown in FIG. 3(a), material target number M1 is associated with material group G1. For material group G4, material target numbers M4 and M5 are associated. As shown in FIG. 3(b), a material classification and a material category are associated with the material target number. For example, for material target number M1, carbon (material category) of a reinforcing material (material classification) is associated. For material target number M2, carbon and silica of the reinforcing material are associated. Oil of a softening agent is associated with material target number M3, SBR of rubber is associated with material target number M4, and BR of rubber is associated with material target number M5. In addition to the material classification and material category shown in FIG. 3(b), further subdivided classifications and categories may be provided. Also, the material classification and material category may be identified by symbols, and a master table of materials corresponding to each symbol may be provided.

[0034] Also, as shown in FIG. 4(a), analysis target number B1 is associated with numerical analysis group SG1, and analysis target number B2 is associated with numerical analysis group SG2. As shown in FIG. 4(b), for analysis target number B1, as a numerical analysis category, the connection of carbon particles and the DBP absorption amount indicating the degree of aggregation are associated. For analysis target number B2, as numerical analysis categories, the cis-1,4 bond amount (denoted as Cis%) of the butadiene unit part in SBR or BR, the trans-1,4 bond amount (denoted as Trans%) of the butadiene unit part in SBR or BR, the styrene content (denoted as St%) in SBR, and the 1,2-vinyl bond amount (denoted as Vinyl%) of the butadiene unit part in SBR or BR are associated.

[0035] Regarding the arithmetic processing in the base arithmetic unit 42, the arithmetic processing will be described using the action number A1 as an example. Returning to Fig. 2(a), in the material group G1 for the action number A1, the material target number M1 (see Fig. 3(a)) is specified, and the carbon of the reinforcing material (see Fig. 3(b)) corresponds. In the numerical analysis group SG1 for the action number A1, the analysis target number B1 (see Fig. 4(a)) is specified, and the DBP absorption amount (see Fig. 4(b)) corresponds. The action content for the action number A1 is to calculate the weighted average value W_avg based on the compounding amount. As shown in Fig. 2(b), the base arithmetic unit 42 calculates, as a feature quantity, the weighted average value W_avg based on the compounding amount with respect to the DBP absorption amount of various carbons used in the compounding of the rubber compound by the arithmetic operation of the action number A1.

[0036] The compounding amounts of various carbons are included in the compounding data 12 of the data mart 11. The DBP absorption amounts of various carbons are included in the raw material data 14 of the data mart 11. The base arithmetic unit 42 reads the necessary data from the data mart 11 and executes the arithmetic processing in the base arithmetic table.

[0037] Based on the materials used in formulating the rubber compound and the user input specifying the arithmetic processing for the materials, the base arithmetic unit 42 generates a base arithmetic table 22 that describes the arithmetic processing and stores it in the storage unit 10. Further, based on the user input specifying a material group or the like, the base arithmetic unit 42 generates a material group table 23, a material category table 24, an analysis group table 25, and an analysis category table 26, and stores them in the storage unit 10. The base arithmetic table 22, the material group table 23, the material category table 24, the analysis group table 25, and the analysis category table 26 may be generated based on user input by the operation unit 31, or may be generated using each of the tables prepared in advance. Also, the administrator of the physical property prediction model generation device 100 may prepare in advance an electronic file in which each table such as the base arithmetic table 22 is set, and store it in the storage unit 10. Note that the workers who perform the processing related to the physical property prediction model, including the administrator of the physical property prediction model generation device 100, may be positioned as users.

[0038] When the user newly adds a feature amount, the base arithmetic unit 42 sequentially adds a new action number to the base arithmetic table 22, and updates the base arithmetic table 22 based on the user input specifying the material group, the numerical analysis group, and the action content. The base arithmetic unit 42 updates the material group table 23, the material category table 24, the analysis group table 25, and the analysis category table 26 based on the user input specifying the information related to the new action number.

[0039] The material group table 23 and the material category table 24 function as auxiliary tables that assist the information on the materials in the base arithmetic table 22. Also, the analysis group table 25 and the analysis category table 26 function as auxiliary tables that assist the information on the content of the numerical analysis.

[0040] The composite operation unit 43 selects at least two of the feature amounts calculated by the base operation unit 42 and generates a composite operation table 27 that describes the operation process. The composite operation unit 43 calculates a feature amount based on the operation process by the generated composite operation table 27. The composite operation unit 43 selects at least two of the feature amounts calculated by the base operation unit 42 and generates a composite operation table 27 that describes the operation process based on a user input that designates the operation process, and stores it in the storage unit 10. The composite operation table 27 may be generated based on a user input by the operation unit 31, or may be generated using a previously prepared table. Also, an administrator of the physical property prediction model generation device 100 may prepare in advance an electronic file in which the composite operation table 27 is set and store it in the storage unit 10.

[0041] FIG. 5(a) is a chart showing an example of the composite operation table 27, and FIG. 5(b) is a chart showing an example of the feature amount calculated by the composite operation table 27. As shown in FIG. 5(a), the composite operation table 27 is composed of an action number, an action number in the base operation table 22, and action contents.

[0042] In the composite operation table 27 generated by the composite operation unit 43, for example, for the action number AM1, the action numbers A2 and A3 in the base operation table 22 and an operation by subtraction are specified as the action contents. The action contents in the composite operation table 27 are arithmetic operations on at least two of the feature amounts calculated by the base operation table 22.

[0043] The composite operation unit 43 calculates the feature amount shown in FIG. 5(b) by executing the operation process by the composite operation table 27 shown in FIG. 5(a).

[0044] As shown in Fig. 5(b), by action number AM1, a value obtained by subtracting the total oil amount from the total filler amount is obtained as a feature amount. The value obtained by subtracting the total oil amount from the total filler amount tends to have a high correlation with the hardness of the rubber. By action number AM2, a value obtained by dividing the total oil amount by the total filler amount is obtained as a feature amount. The value obtained by dividing the total oil amount by the total filler amount tends to have a high correlation with the hardness of the rubber. Generally, the hardness of rubber is considered to be determined by the balance between oil and filler.

[0045] By action number AM3, a value obtained by adding the total filler amount and the total oil amount is obtained as a feature amount. The value obtained by adding the total filler amount and the total oil amount tends to have a high correlation with the fuel consumption performance. Generally, the fuel consumption performance is determined by the energy loss component in the rubber compound, and oil and filler are considered to be energy loss components.

[0046] By action number AM4, a value obtained by multiplying the total filler amount and the DBP absorption amount of carbon is obtained as a feature amount. The value obtained by multiplying the total filler amount and the DBP absorption amount of carbon tends to have a high correlation with the wear resistance performance. Generally, the wear resistance performance depends on the reinforcing property of the filler, and it is considered that the reinforcing property of the filler can be represented by multiplying the total filler amount and the DBP absorption amount of carbon.

[0047] The learning data generation unit 44 generates a learning data mart composed of at least one of the feature amounts calculated by the base operation unit 42 and the feature amounts calculated by the composite operation unit 43, and the data included in the data mart 11. The learning data mart may include at least one of the feature amounts calculated by the base operation unit 42 and may not include the feature amounts calculated by the composite operation unit 43. The learning data mart may not include the feature amounts calculated by the base operation unit 42 and may include at least one of the feature amounts calculated by the composite operation unit 43. The learning data mart may include at least one of the feature amounts calculated by the base operation unit 42 and may include at least one of the feature amounts calculated by the composite operation unit 43.

[0048] The learning data generation unit 44 generates a learning data mart including some or all of the plurality of rubber formulations included in the data mart 11. The rubber formulations included in the learning data mart may be preset or may be selected by accepting user input at the operation unit 31. Further, the learning data generation unit 44 generates a learning data mart including some or all of the formulation data 12 and the physical property data 13. The learning data mart may include some or all of the raw material data 14. The learning data mart may include some or all of the processing condition data 15.

[0049] The learning data generation unit 44 may include preset data among the formulation data 12, the physical property data 13, the raw material data 14, and the processing condition data 15 in the learning data mart. Also, the learning data generation unit 44 may accept user input at the operation unit 31 and select the data to be included in the learning data mart among the formulation data 12, the physical property data 13, the raw material data 14, and the processing condition data 15.

[0050] The physical property prediction processing unit 45 inputs data other than the physical property data 13 in the learning data mart generated by the learning data generation unit 44 as explanatory variables into the learning-based arithmetic model 45a, and predicts the physical properties of the rubber compound. In the physical property prediction processing unit 45, the explanatory variables that are input data to the arithmetic model 45a are the compounding data 12, the feature amounts calculated by the base arithmetic unit 42 and the composite arithmetic unit 43, the raw material data 14, and the processing condition data 15. The arithmetic model 45a is a learning-based model of the random forest type, which is an ensemble learning algorithm using decision trees.

[0051] The learning processing unit 46 learns the arithmetic model 45a based on the learning data mart generated by the learning data generation unit 44. The learning processing unit 46 compares the estimated value of the physical property of the rubber compound as the output data from the arithmetic model 45a with the known physical property data 13 as teacher data, and learns the arithmetic model 45a. Note that the arithmetic model 45a may be, for example, a generalized linear regression model, a DNN (Deep Neural Network) model, a gradient boosting type decision tree model, or the like. Also, known verification methods such as random data sampling and cross-validation can be used for the verification of the arithmetic model 45a.

[0052] Next, the operation of the physical property prediction model generation device 100 will be described. FIG. 6 is a flowchart showing the procedure of the generation process of the arithmetic model 45a. The data mart generation unit 41 of the arithmetic processing unit 40 generates a data mart 11 used for learning the arithmetic model 45a (S1). The data mart 11 includes at least the compounding data 12, the physical property data 13, and the raw material data 14. Also, the data mart 11 may include the processing condition data 15 in addition to the compounding data 12, the physical property data 13, and the raw material data 14.

[0053] The base calculation unit 42 generates a base calculation table 22 that describes the materials of the rubber compound and the calculation process for the materials (S2). In step S2, the base calculation unit 42 generates the material group table 23, the material category table 24, the analysis group table 25, and the analysis category table 26 together with the generation of the base calculation table 22.

[0054] The base calculation unit 42 calculates a feature amount based on the calculation process by the base calculation table 22 (S3). In step S3, the base calculation unit 42 refers to the material group table 23, the material category table 24, the analysis group table 25, and the analysis category table 26, and executes the calculation process of the base calculation table 22.

[0055] The composite calculation unit 43 selects at least two of the feature amounts calculated by the base calculation unit 42 and generates a composite calculation table 27 that describes the calculation process (S4). The composite calculation unit 43 calculates a feature amount based on the calculation process by the base calculation table 22 (S5).

[0056] The learning data generation unit 44 generates a learning data mart composed of at least one of the feature amounts calculated by the base calculation unit 42 and the composite calculation unit 43, and the data included in the data mart 11 (S6). The learning data mart includes the compounding data 12, the physical property data 13, and the raw material data 14. The learning data mart may also include the processing condition data 15.

[0057] The learning processing unit 46 learns a learning-based calculation model 45a for predicting the physical properties of the rubber compound based on the learning data mart (S7), and ends the process. The learning processing unit 46 compares the estimated value of the physical properties of the rubber compound calculated by the calculation model 45a with the known physical property data 13 as the teacher data, and learns the calculation model 45a.

[0058] Figure 7 is a flowchart showing the procedure of the calculation process of the feature amount based on the base operation table 22. The base operation unit 42 selects one unexecuted action number in the base operation table 22 (S11). The base operation unit 42 refers to the material group table 23 and the material category table 24 based on the material group at the selected action number, and extracts data from the blending data 12 (S12).

[0059] The base operation unit 42 determines whether there is a numerical analysis group corresponding to the action number in the base operation table 22 (S13). If it is determined in step S13 that there is no numerical analysis group (S13: NO), the base operation unit 42 executes the action content in the base operation table 22 to calculate the feature amount (S14).

[0060] If it is determined in step S13 that there is a numerical analysis group (S13: YES), the base operation unit 42 refers to the analysis group table 25 and the analysis category table 26, and extracts data such as property values from the raw material data 14 (S15). After step S15, the base operation unit 42 transfers to step S14, and executes the action content in the base operation table 22 to calculate the feature amount.

[0061] After step S14, the base operation unit 42 determines whether it has been executed for all action numbers (S16). If it is determined in step S16 that it has not been executed for all action numbers (S16: NO), the base operation unit 42 returns to step S11 to repeat the process. If it is determined in step S16 that it has been executed for all action numbers (S16: YES), the base operation unit 42 ends the process.

[0062] FIG. 8 is a flowchart showing the procedure of the calculation process of the feature amount based on the composite operation table 27. The composite operation unit 43 selects one unexecuted action number in the composite operation table 27 (S21). The composite operation unit 43 extracts the feature amount of the action number of the base operation table 22 corresponding to the selected action number (S22). The composite operation unit 43 executes the action content in the composite operation table 27 on the feature amount extracted in step S22 to calculate the feature amount (S23).

[0063] After step S23, the composite operation unit 43 determines whether it has been executed for all action numbers (S24). In step S24, if it is determined that it has not been executed for all action numbers (S24: NO), the composite operation unit 43 returns to step S21 and repeats the process. In step S24, if it is determined that it has been executed for all action numbers (S24: YES), the composite operation unit 43 ends the process.

[0064] The physical property prediction model generation method in this embodiment generates a data mart 11 including the compounding amounts of a plurality of materials used in the generation of the rubber compound, the property values of the materials, and the physical properties of the rubber compound in the data mart generation step. The physical property prediction model generation method generates a base operation table 22 that describes the materials and the operation processes for the materials in the base operation generation step. The physical property prediction model generation method calculates the feature amount based on the operation process in the base operation table 22 in the base operation step.

[0065] The physical property prediction model generation method generates a learning data mart composed of at least one of the feature amounts calculated in the base operation step and the data included in the data mart 11 in the learning data generation step. The physical property prediction model generation method learns a learning-type operation model 45a for predicting the physical properties of the rubber compound based on the learning data mart in the learning step. According to this physical property prediction model generation method, the convenience of the user can be improved by sharing the process of the operation process for generating the feature amount.

[0066] In the physical property prediction model generation method, the base operation table 22 includes at least one of the four arithmetic operations. Thereby, in the physical property prediction model generation method, by specifying the four arithmetic operations in the base operation table 22, the feature amount in the rubber compound can be generated.

[0067] In the physical property prediction model generation method, the data mart 11 includes data on processing conditions in the rubber compound. Thereby, in the physical property prediction model generation method, the operation model 45a can be learned while reflecting the influence of the processing condition data 15.

[0068] The physical property prediction model generation method further generates a composite operation table 27 that describes the operation process by selecting at least two of the feature amounts calculated in the base operation step by a composite operation generation step. The physical property prediction model generation method calculates the feature amount based on the operation process in the composite operation table 27 by the composite operation step. The physical property prediction model generation method adds at least one of the feature amounts calculated in the composite operation step to the learning data mart by a learning data generation step. Thereby, in the physical property prediction model generation method, the operation process of the feature amount calculated in a composite manner can be made common, and the convenience for the user can be further improved.

[0069] In the physical property prediction model generation method, the composite operation table 27 includes at least one of the four arithmetic operations. Thereby, in the physical property prediction model generation method, by specifying the four arithmetic operations in the composite operation table 27, a composite feature amount in the rubber compound can be generated.

[0070] In the physical property prediction model generation method, the base operation table 22 executes an operation within a material group including a plurality of materials. Thereby, in the physical property prediction model generation method, for example, when a plurality of carbon materials are compounded, the feature amount can be generated in consideration of the total compounding amount and property values of the carbon materials.

[0071] The action content specified in the base operation table 22 is not limited to arithmetic operations, and may include numerical analysis such as the above-mentioned DBP absorption amount, Cis%, Trans%, St%, Vinyl%, etc.

[0072] Also, in the physical property prediction model generation method, the calculation model 45a can use, for example, a random forest model, a gradient boosting decision tree model, and a generalized linear regression model. Thereby, the physical property prediction model generation method can construct a calculation model that utilizes the characteristics of each model.

[0073] (Embodiment 2) FIG. 9 is a block diagram showing the functional configuration of the physical property prediction device 110 according to Embodiment 2. The physical property prediction device 110 has a configuration excluding the learning data generation unit 44 and the learning processing unit 46 from the physical property prediction model generation device 100 in Embodiment 1. Further, as the calculation model 45a in the physical property prediction processing unit 45 of the physical property prediction device 110, a calculation model that has been learned by the physical property prediction model generation device 100 is used.

[0074] The data mart generation unit 41 of the physical property prediction device 110 generates a data mart 11 composed of the formulation data 12, raw material data 14, and processing condition data 15 of a new rubber formulation for predicting physical properties. Since the data mart 11 generated here does not learn the calculation model 45a, it only needs to have the formulation data 12, raw material data 14, and processing condition data 15, and does not need to include the physical property data 13.

[0075] Similar to Embodiment 1, the base calculation unit 42 of the physical property prediction device 110 generates the base operation table 22 and calculates the feature amount based on the operation process using the base operation table 22. During the operation process using the base operation table 22, the material group table 23, the material category table 24, the analysis group table 25, and the analysis category table 26 are referred to. Similar to Embodiment 1, the composite calculation unit 43 generates the composite operation table 27 and calculates the feature amount based on the operation process using the composite operation table 27.

[0076] The physical property prediction processing unit 45 uses, as explanatory variables, at least one of the feature quantities calculated for the new rubber compound by the base calculation unit 42 and the composite calculation unit 43, as well as the compounding data 12, raw material data 14, and processing condition data 15 of the new rubber compound. The physical property prediction processing unit 45 inputs the data used as explanatory variables into the learned calculation model 45a and predicts the physical properties.

[0077] The explanatory variables used by the physical property prediction processing unit 45 as inputs to the learned calculation model 45a during physical property prediction are equivalent to the explanatory variables used during the learning of the calculation model 45a in the above-described Embodiment 1. The physical property prediction processing unit 45 may store, in the storage unit 10 as physical property data 13, the physical properties predicted for the new rubber compound by the calculation model 45a.

[0078] Next, the operation of the physical property prediction apparatus 110 will be described. FIG. 10 is a flowchart showing the procedure of physical property prediction processing. The data mart generation unit 41 of the arithmetic processing unit 40 generates a data mart 11 related to the new rubber compound (S31). The processes from step S32 to step S35 are equivalent to the processes from step S2 to step S5 in FIG. 6, and the description thereof is omitted for the sake of simplicity. Also, the calculation processes in step S33 and step S35 are equivalent to the calculation processes described with reference to FIGS. 7 and 8.

[0079] The physical property prediction processing unit 45 predicts the physical properties of the new rubber compound using the feature quantities calculated in step S33 and step S35 by the calculation model 45a (S36), and ends the processing.

[0080] The physical property prediction method in this embodiment generates a data mart 11 including the compounding amounts of a plurality of materials and the property values of the materials used in the production of a rubber compound in a data mart generation step. The physical property prediction method generates a base operation table 22 describing the materials and the operation processing for the materials in a base operation generation step. The physical property prediction method calculates a feature amount based on the operation processing in the base operation table 22 in a base operation step.

[0081] The physical property prediction method predicts the physical properties of the rubber compound by inputting at least one of the feature amounts calculated in the base operation step and the data included in the data mart 11 into a learned operation model 45a in a physical property prediction step. According to this physical property prediction method, the convenience of the user can be improved by sharing the process of the operation processing for generating the feature amount.

[0082] The physical property prediction method further generates a composite operation table 27 describing the operation processing by selecting at least two of the feature amounts calculated in the base operation step in a composite operation generation step. The physical property prediction method calculates a feature amount based on the operation processing in the composite operation table 27 in a composite operation step. The physical property prediction method predicts the physical properties of the rubber compound by inputting at least one of the feature amounts calculated in the composite operation step and the data included in the data mart 11 into a learned operation model 45a in a physical property prediction step. Thereby, the physical property prediction method can share the process of the operation processing of the feature amount calculated in a composite manner and further improve the convenience of the user.

[0083] (Embodiment 3) FIG. 11 is a block diagram showing the functional configuration of a compounding prediction device 120 according to Embodiment 3. The compounding prediction device 120 includes a compounding prediction processing unit 47. The compounding prediction processing unit 47 has a learned operation model 45a by a physical property prediction model generation device 100.

[0084] The data mart generation unit 41 in the compounding prediction device 120 generates a data mart 11 based on the constraint conditions regarding the rubber compound for predicting the compounding. The data mart 11 includes the constraint conditions regarding the compounding data 12, raw material data 14, and processing condition data 15 for the rubber compound.

[0085] Similar to Embodiment 1, the base calculation unit 42 has a function of generating a base calculation table 22 and calculating feature amounts based on the calculation process using the base calculation table 22. During the calculation process using the base calculation table 22, the material group table 23, material category table 24, analysis group table 25, and analysis category table 26 are referred to. Similar to Embodiment 1, the composite calculation unit 43 has a function of generating a composite calculation table 27 and calculating feature amounts based on the calculation process using the composite calculation table 27.

[0086] The compounding prediction processing unit 47 uses the calculation model 45a to predict the compounding of the rubber compound as an inverse problem by means of a genetic algorithm based on the target physical properties, the constraint conditions regarding the compounding amounts of materials in the rubber compound, raw material data, and processing condition data. Also, the compounding prediction processing unit 47 may predict the compounding of the rubber compound as an inverse problem using a known optimization method, for example, using the gradient descent method or the like. The compounding prediction processing unit 47 may predict a rubber compound that satisfies the physical properties and the compounding unit price by adding the target compounding unit price in addition to the target physical properties. In this case, the calculation model 45a outputs the compounding unit price, and the compounding unit price as physical property data is pre-learned as teacher data. For example, by including the unit price of each material in the raw material data 14 and taking a weighted average using, for example, the compounding amount of the material and the unit price of the material, the compounding unit price can be calculated and registered in the physical property data 13, so that the compounding unit price can be treated in the same way as the physical properties.

[0087] The compounding prediction processing unit 47 may add the compounding data, raw material data, and processing condition data of the rubber compound assumed in the prediction process to the data mart 11, calculate the feature amounts by the base calculation unit 42 and the composite calculation unit 43, and reflect them in the compounding prediction.

[0088] FIG. 12 is a flowchart showing the procedure of the compounding prediction process. The compounding prediction unit 47 acquires the physical properties of the target rubber compound (S41), and acquires the constraint conditions of the rubber compound (S42). The compounding prediction unit 47 uses the learned arithmetic model 45a to predict the compounding of the rubber compound as an inverse problem by an optimization method based on the target physical properties and the constraint conditions regarding the compounding amounts of the materials and the property values of the materials in the rubber compound (S43), and ends the process. Note that the constraint conditions may include processing conditions. Further, the compounding prediction unit 47 may be configured to predict at least any one of the property values and category variables of the raw materials and the processing conditions, in addition to the compounding (material ratio), for the rubber compound having the target physical properties.

[0089] In the compounding prediction method according to the present embodiment, a base operation table 22 describing materials and arithmetic processing for the materials is generated in the base operation generation step. The compounding prediction method calculates feature amounts based on the arithmetic processing in the base operation table 22 in the base operation step. The compounding prediction method predicts the compounding of the rubber compound by an optimization method based on the target physical properties and the constraint conditions regarding the compounding amounts of the materials and the property values of the materials in the rubber compound, using an arithmetic model learned by adding at least one of the feature amounts calculated in the base operation step as an explanatory variable. According to this compounding prediction method, the process of the arithmetic processing for generating the feature amounts is shared to improve the convenience for the user, and the compounding of the rubber compound that can obtain the target physical properties can be predicted.

[0090] The compounding prediction method further generates a compound operation table 27 that describes an operation process by selecting at least two of the feature quantities calculated in the base operation step through a compound operation generation step. The compounding prediction method calculates a feature quantity based on the operation process in the compound operation table 27 through a compound operation step. The compounding prediction method predicts the compounding of a rubber compound using an operation model that is learned by adding at least one of the feature quantities calculated in the compound operation step as an explanatory variable in a compounding prediction step. Thereby, the compounding prediction method can improve the convenience for the user by sharing the process of the operation process of the feature quantity calculated in a complex manner.

[0091] Generalizing the technical idea embodied by the above embodiments, it can be said that the technical idea described in the following items is included.

[0092] The first item is a physical property prediction model generation method including a data mart generation step of generating a data mart including compounding amounts of a plurality of materials used in the production of a rubber compound, property values of the materials, and physical properties of the rubber compound, a base operation generation step of generating a base operation table that describes the materials and operation processes for the materials, a base operation step of calculating a feature quantity based on the operation process by the base operation table, a learning data generation step of generating a learning data mart including at least one of the feature quantities calculated by the base operation step and data included in the data mart, and a learning step of training a learning-type operation model for predicting the physical properties of the rubber compound based on the learning data mart.

[0093] The second item is the physical property prediction model generation method according to the first item, wherein the base operation table includes at least one of the four arithmetic operations.

[0094] The third item is the physical property prediction model generation method according to the first or second item, wherein the data mart includes data on processing conditions in the rubber compound.

[0095] The fourth item further includes a composite operation generation step of generating a composite operation table that selects at least two of the feature amounts calculated in the base operation step and describes the operation process, and a composite operation step of calculating a feature amount based on the operation process by the composite operation table. The learning data generation step is the physical property prediction model generation method according to any one of Items 1 to 3, which adds at least one of the feature amounts calculated in the composite operation step to the learning data mart.

[0096] The fifth item is the physical property prediction model generation method according to Item 4, wherein the composite operation table includes at least one of the four arithmetic operations.

[0097] The sixth item is the physical property prediction model generation method according to any one of Items 1 to 5, wherein the base operation table executes an operation within a material group including a plurality of materials.

[0098] The seventh item is a physical property prediction model generation apparatus including a data mart generation unit that generates a data mart including compounding amounts of a plurality of materials used in the production of a rubber compound, property values of the materials, and physical properties of the rubber compound, a base operation unit that generates a base operation table that describes the materials and the operation process for the materials, and calculates a feature amount based on the operation process by the base operation table, a learning data generation unit that adds the feature amount calculated by the base operation unit to the data mart to generate a learning data mart, and a learning processing unit that learns a learning-type operation model for predicting the physical properties of the rubber compound based on the learning data mart.

[0099] Item 8 is a physical property prediction method comprising: a data mart generation step of generating a data mart including compounding amounts of a plurality of materials used in the production of a rubber compound and property values of said materials; a base operation generation step of generating a base operation table describing said materials and arithmetic processing on said materials; a base operation step of calculating feature quantities based on the arithmetic processing by said base operation table; and a physical property prediction step of inputting at least one of the feature quantities calculated by said base operation step and data included in said data mart into a learned arithmetic model to predict physical properties of the rubber compound.

[0100] Item 9 is a physical property prediction apparatus comprising: a data mart generation unit that generates a data mart including compounding amounts of a plurality of materials used in the production of a rubber compound and property values of said materials; a base operation unit that generates a base operation table describing said materials and arithmetic processing on said materials and calculates feature quantities based on the arithmetic processing by said base operation table; and a physical property prediction processing unit that inputs at least one of the feature quantities calculated by said base operation unit and data included in said data mart into a learned arithmetic model to predict physical properties of the rubber compound.

[0101] Item 10 is a physical property prediction program for causing a computer to execute: a data mart generation step of generating a data mart including compounding amounts of a plurality of materials used in the production of a rubber compound and property values of said materials; a base operation generation step of generating a base operation table describing said materials and arithmetic processing on said materials; a base operation step of calculating feature quantities based on the arithmetic processing by said base operation table; and a physical property prediction step of inputting at least one of the feature quantities calculated by said base operation step and data included in said data mart into a learned arithmetic model to predict physical properties of the rubber compound.

[0102] Item 11 is a compounding prediction method comprising: a base operation generation step of generating a base operation table that describes materials and arithmetic processing on the materials; a base operation step of calculating feature amounts based on the arithmetic processing by the base operation table; and a compounding prediction step of predicting the compounding of a rubber compound by an optimization method based on a target physical property, and constraint conditions regarding the compounding amount of the materials and the property values of the materials in the rubber compound, using an arithmetic model learned by adding at least one of the feature amounts calculated by the base operation step as an explanatory variable.

[0103] Item 12 is a compounding prediction method comprising: a base operation unit that generates a base operation table that describes materials and arithmetic processing on the materials, and calculates feature amounts based on the arithmetic processing by the base operation table; and a compounding prediction processing unit that predicts the compounding of a rubber compound by an optimization method based on a target physical property, and constraint conditions regarding the compounding amount of the materials and the property values of the materials in the rubber compound, using an arithmetic model learned by adding at least one of the feature amounts calculated by the base operation unit as an explanatory variable.

[0104] Item 13 is a compounding prediction program for causing a computer to execute: a base operation generation step of generating a base operation table that describes materials and arithmetic processing on the materials; a base operation step of calculating feature amounts based on the arithmetic processing by the base operation table; and a compounding prediction step of predicting the compounding of a rubber compound by an optimization method based on a target physical property, and constraint conditions regarding the compounding amount of the materials and the property values of the materials in the rubber compound, using an arithmetic model learned by adding at least one of the feature amounts calculated by the base operation step as an explanatory variable.

[0105] The above has been described based on the embodiments of the present invention. These embodiments are examples, and it is understood by those skilled in the art that various modifications and changes are possible within the scope of the claims of the present invention, and such modifications and changes are also within the scope of the claims of the present invention. Therefore, the descriptions and drawings in this specification should be treated as illustrative rather than restrictive.

Explanation of Reference Numerals

[0106] 11 Data mart, 22 Base operation table, 27 Composite operation table, 41 Data mart generation unit, 42 Base operation unit, 43 Composite operation unit, 44 Learning data generation unit, 45 Physical property prediction processing unit, 45a Operation model, 46 Learning processing unit, 47 Blending prediction processing unit, 100 Physical property prediction model generation device, 110 Physical property prediction device, 120 Blending prediction device.

Claims

1. A data mart generation step of generating a data mart including the blending amounts of a plurality of materials used in the production of a rubber compound, the property values of the materials, and the physical properties of the rubber compound; A base operation generation step of generating a base operation table describing the materials and the arithmetic processing for the materials; A base operation step of calculating feature quantities based on the arithmetic processing by the base operation table; A learning data generation step of generating a learning data mart composed of at least one of the feature quantities calculated by the base operation step and the data included in the data mart; A learning step of training a learning-based arithmetic model for predicting the physical properties of the rubber compound based on the learning data mart; A method for generating a physical property prediction model comprising the above.

2. The method for generating a physical property prediction model according to claim 1, wherein the base operation table includes at least one of the four arithmetic operations.

3. The method for generating a physical property prediction model according to claim 1, wherein the data mart includes data on processing conditions in the rubber compound.

4. A composite operation generation step of selecting at least two of the feature quantities calculated by the base operation step and generating a composite operation table describing the arithmetic processing; A composite operation step of calculating feature quantities based on the arithmetic processing by the composite operation table, further comprising: The learning data generation step of adding at least one of the feature quantities calculated by the composite operation step to the learning data mart in the method for generating a physical property prediction model according to claim 1.

5. The method for generating a physical property prediction model according to claim 4, wherein the composite operation table includes at least one of the four arithmetic operations.

6. The method for generating a physical property prediction model according to any one of claims 1 to 5, wherein the base operation table executes operations within a material group including a plurality of materials.

7. A data mart generation unit that generates a data mart including the blending amounts of a plurality of materials used in the production of a rubber compound, the property values of the materials, and the physical properties of the rubber compound; A base operation unit that generates a base operation table describing the materials and the arithmetic processing for the materials, and calculates feature quantities based on the arithmetic processing by the base operation table; A learning data generation unit that adds the feature quantities calculated by the base operation unit to the data mart to generate a learning data mart; A learning processing unit that trains a learning-based arithmetic model for predicting the physical properties of a rubber compound based on the learning data mart; A physical property prediction model generation device comprising the same. **Claim 8** A data mart generation step of generating a data mart including the compounding amounts of a plurality of materials used in the production of a rubber compound and the property values of the materials; A base operation generation step of generating a base operation table describing the materials and the arithmetic processing for the materials; A base operation step of calculating feature amounts based on the arithmetic processing by the base operation table; A physical property prediction step of inputting at least one of the feature amounts calculated in the base operation step and the data included in the data mart into a trained arithmetic model to predict the physical properties of the rubber compound; A physical property prediction method comprising the same. **Claim 9** A data mart generation unit that generates a data mart including the compounding amounts of a plurality of materials used in the production of a rubber compound and the property values of the materials; A base operation unit that generates a base operation table describing the materials and the arithmetic processing for the materials, and calculates feature amounts based on the arithmetic processing by the base operation table; A physical property prediction processing unit that inputs at least one of the feature amounts calculated by the base operation unit and the data included in the data mart into a trained arithmetic model to predict the physical properties of the rubber compound; A physical property prediction device comprising the same. **Claim 10** A data mart generation step of generating a data mart including the compounding amounts of a plurality of materials used in the production of a rubber compound and the property values of the materials; A base operation generation step of generating a base operation table describing the materials and the arithmetic processing for the materials; A base operation step of calculating feature amounts based on the arithmetic processing by the base operation table; A physical property prediction step of inputting at least one of the feature amounts calculated in the base operation step and the data included in the data mart into a trained arithmetic model to predict the physical properties of the rubber compound; A physical property prediction program for causing a computer to execute the same. **Claim 11** A base operation generation step of generating a base operation table describing the materials and the arithmetic processing for the materials; A base operation step of calculating feature amounts based on the arithmetic processing by the base operation table; Using an operation model learned by adding at least one of the feature quantities calculated in the base operation step to an explanatory variable, based on the target physical properties and the constraint conditions regarding the compounding amounts of the materials and the property values of the materials in the rubber compound, a compounding prediction step for predicting the compounding of the rubber compound by an optimization method; A compounding prediction method comprising the above.

12. A base operation unit that generates a base operation table describing the materials and the operation processing for the materials, and calculates feature quantities based on the operation processing by the base operation table; A compounding prediction processing unit that uses an operation model learned by adding at least one of the feature quantities calculated by the base operation unit to an explanatory variable, and predicts the compounding of the rubber compound by an optimization method based on the target physical properties and the constraint conditions regarding the compounding amounts of the materials and the property values of the materials in the rubber compound; A compounding prediction apparatus comprising the above.

13. A base operation generation step for generating a base operation table describing the materials and the operation processing for the materials; A base operation step for calculating feature quantities based on the operation processing by the base operation table; A compounding prediction step for predicting the compounding of the rubber compound by an optimization method using an operation model learned by adding at least one of the feature quantities calculated in the base operation step to an explanatory variable, based on the target physical properties and the constraint conditions regarding the compounding amounts of the materials and the property values of the materials in the rubber compound; A compounding prediction program for causing a computer to execute the above.

Images