JP7302134B2 - Physical property information estimation method, physical property information estimation model generation method, physical property information estimation device, physical property information estimation model generation device, physical property information estimation program, and physical property information estimation model generation program - Google Patents

Description

The present disclosure relates to a physical property information estimation method, a physical property information estimation model generation method, a physical property information estimation device, a physical property information estimation model generation device, a physical property information estimation program, and a physical property information estimation model generation program.

Adhesives characterized by containing a liquid epoxy resin composition are conventionally known (see, for example, Patent Document 1). The adhesive is disclosed to contain at least (a) an epoxy resin and (b) two or more latent curing agents.

JP 2013-100525 A

By the way, when manufacturing an adhesive, what kind of composition should be blended to obtain an adhesive having desired physical properties is an important concern for adhesive manufacturers. Therefore, it is desirable to clarify the relationship between the formulation of the adhesive composition and the physical properties of the adhesive.

However, in the technique of Patent Document 1, only an adhesive containing an epoxy resin and two or more latent curing agents is disclosed. There is no disclosure as to whether an adhesive is available. Therefore, the technique disclosed in Patent Document 1 has a problem that the physical properties of the adhesive cannot be estimated from the formulation of the adhesive.

The subject of the present disclosure is a physical property information estimating method, a physical property information estimating model generating method, a physical property information estimating device, a physical property information estimating model generating device, and a physical property information estimating program capable of estimating the physical properties of the adhesive from the compounding of the adhesive. , and to provide a physical property information estimation model generation program.

The method for estimating physical property information of the present disclosure is for estimating the physical property information of the adhesive from the blending information representing the blending ratio of the composition of the adhesive to be estimated. This is a method of estimating physical property information in which a computer executes a process of inputting information to a learned model and estimating physical property information of the adhesive corresponding to the input compounding information.

Further, the physical property information estimating method of the present disclosure generates a plurality of the compounding information by changing the compounding information of the reference adhesive, and each of the generated plurality of the compounding information is applied to the learned model. to estimate the physical property information corresponding to each of the plurality of the compounding information, select the physical property information that satisfies a predetermined criterion from the estimated plurality of the physical property information, and select the physical property Information and the combination information corresponding to the selected physical property information can be output.

In addition, the physical property information estimation method of the present disclosure randomly generates a plurality of the combination information, and inputs each of the generated plurality of the combination information to the learned model, so that the plurality of the combination information and selecting the physical property information that satisfies a predetermined criterion from a plurality of the estimated physical property information, and selecting the physical property information and the physical property information corresponding to the selected physical property information Combination information can be output.

Further, the physical property information can be moisture permeability, glass transition temperature, or adhesive strength of the adhesive.

In addition, the learned model includes a first learned model for estimating the moisture permeability as the physical property information from the compounding information, and a first learned model for estimating the glass transition temperature as the physical property information from the compounding information. 2, and a third learned model for estimating adhesive strength as the physical property information from the combination information and at least one of the moisture permeability and the glass transition temperature, and the physical property information. When estimating the adhesive strength as, the glass transition temperature and the compounding information estimated by the moisture permeability or the second learned model estimated by the first learned model, the third learning It is possible to estimate the adhesive strength of the adhesive corresponding to the input formulation information by inputting to the finished model.

In addition, the physical property information estimation model generation method of the present disclosure is based on learning data representing a combination of learning compounding information representing the blending ratio of the composition of the learning adhesive and the physical property information of the learning adhesive. A physical property information estimation model generation method in which a computer executes processing for generating a trained model for estimating physical property information of an adhesive from compounding information representing the compounding ratio of the composition of the adhesive.

Further, the physical property information estimation device of the present disclosure includes an information acquisition unit that acquires composition information representing the composition ratio of the composition of the adhesive to be estimated, and the composition information acquired by the information acquisition unit. an estimating unit for inputting into a learned model for estimating the physical property information of the adhesive from the compounding information representing the compounding ratio of the composition, and estimating the physical property information of the adhesive corresponding to the input compounding information; , is a physical property information estimation device including.

In addition, the physical property information estimation model generation device of the present disclosure is based on learning data representing a combination of learning compounding information representing the blending ratio of the composition of the learning adhesive and the physical property information of the learning adhesive. and a physical property information estimation model generation device including a learning unit for generating a learned model for estimating physical property information of the adhesive from compounding information representing the compounding ratio of the composition of the adhesive.

Further, the physical property information estimation program of the present disclosure estimates the physical property information of the adhesive from the formulation information representing the blending ratio of the composition of the adhesive to be estimated. A physical property information estimating program for causing a computer to execute a process of inputting to a learned model for and estimating physical property information of the adhesive corresponding to the input compounding information.

In addition, the physical property information estimation model generation program of the present disclosure is based on learning data representing a combination of learning compounding information representing the blending ratio of the composition of the learning adhesive and the physical property information of the learning adhesive. is a physical property information estimation model generation program for causing a computer to execute a process of generating a trained model for estimating physical property information of an adhesive from compounding information representing a compounding ratio of a composition of the adhesive.

According to the physical property information estimation model generation method, the physical property information estimation model generation device, and the physical property information estimation model generation program of the present disclosure, a model for estimating the physical properties of the adhesive can be obtained from the formulation of the adhesive. The effect is obtained.

According to the physical property information estimating method, the physical property information estimating device, and the physical property information estimating program of the present disclosure, it is possible to obtain an effect that the physical properties of the adhesive can be estimated from the compounding of the adhesive.

BRIEF DESCRIPTION OF THE DRAWINGS It is the schematic which shows an example of a structure of the physical-property information estimation system of this embodiment. It is a schematic block diagram of a computer functioning as a physical property information estimation device. FIG. 4 is an explanatory diagram for explaining learning data according to the embodiment; It is a figure which shows an example of the learning processing routine performed by a physical property information estimation apparatus. It is a figure which shows an example of the estimation processing routine performed by a physical property information estimation apparatus. FIG. 2 shows the results of Example 1; FIG. 2 shows the results of Example 1; FIG. 2 shows the results of Example 1; FIG. 2 shows the results of Example 1; FIG. 2 shows the results of Example 1; FIG. 2 shows the results of Example 1; FIG. 2 shows the results of Example 1; FIG. 10 is a diagram showing the results of Example 2; FIG. 10 is a diagram showing the results of Example 2;

The present embodiment will be described in detail below.

<System configuration of physical property information estimation system according to embodiment>

FIG. 1 shows an example of the configuration of a physical property information estimation system 10 according to the first embodiment. The physical property information estimating system 10 of this embodiment can be functionally represented by a configuration including an input device 12, a physical property information estimating device 14, and a display device 40, as shown in FIG.

The input device 12 accepts information input from the outside. The input device 12 is implemented by, for example, a keyboard, a mouse, or an input/output device that receives input from an external device.

The physical property information estimation device 14 can be configured by a computer including a CPU, a RAM, and a ROM storing programs and various data for executing each processing routine to be described later.

For example, the physical property information estimation device 14 can be realized by the computer 50 shown in FIG. The computer 50 includes a CPU 51 , a memory 52 as a temporary storage area, and a non-volatile storage section 53 . The computer 50 also includes an input/output interface (I/F) 54 to which an input/output device (not shown) is connected, and a read/write (R/W) unit 55 for controlling reading and writing of data to and from a recording medium. Prepare. The computer 50 also has a network I/F 56 connected to a network such as the Internet. The CPU 51 , memory 52 , storage unit 53 , input/output I/F 54 , R/W unit 55 and network I/F 56 are connected to each other via a bus 57 .

The storage unit 53 can be implemented by a hard disk drive (HDD), solid state drive (SSD), flash memory, or the like. A program for causing the computer 50 to function is stored in the storage unit 53 as a storage medium. The CPU 51 reads out the program from the storage unit 53, develops it in the memory 52, and sequentially executes the processes of the program.

Functionally, the physical property information estimation device 14 includes an information acquisition unit 16, a data storage unit 18, a learning unit 20, a trained model storage unit 22, and a data generation unit 24, as shown in FIG. , an estimation unit 26 and a selection unit 28 .

Information output from the physical property information estimation device 14 is displayed on the display device 40 .

The physical property information estimation device 14 of the present embodiment estimates physical property information of the adhesive from compounding information representing the compounding ratio of the composition of the adhesive. Specifically, the physical property information estimating device 14 of the present embodiment is a learned model for estimating physical property information of the adhesive from compounding information representing the compounding ratio of the composition of the adhesive based on learning data. to generate Then, the physical property information estimation device 14 uses the learned model to estimate the physical property information of the adhesive to be estimated.

The physical properties of the adhesive differ depending on the formulation of the composition contained in the adhesive. Here, what kind of composition should be blended to produce an adhesive having desired physical properties is an important matter of concern for adhesive manufacturers.

For example, the user may want to know the blending ratio of the adhesive so that the moisture permeability of the adhesive is 50 or less and the glass transition temperature Tg due to UV and heat is around 100°C. There is also a demand for increasing the adhesive strength of the adhesive as much as possible.

In this case, for example, in order to obtain an adhesive having desired physical properties, an experiment is conducted to actually produce an adhesive after changing the formulation of the adhesive composition, and the physical properties of the adhesive are measured. There is a way. When adopting this method, it is necessary to repeat experiments until an adhesive having desired physical properties is obtained.

However, there is a problem that the number of possible combinations of composition formulations is enormous, and many experiments must be performed until an adhesive having desired physical properties is obtained.

Therefore, the physical property information estimating device 14 of the present embodiment generates a learned model for estimating the physical properties of the adhesive from the compounding of the adhesive by machine learning. Then, the physical property information estimation device 14 inputs compounding information representing the compounding of the adhesive to be estimated to the learned model, and estimates the physical properties of the adhesive. As a result, the physical properties of the adhesive can be inferred from the formulation of the adhesive without actually manufacturing the adhesive.

A specific description will be given below.

The input device 12 receives input of learning data. The learning data of the present embodiment is data representing a combination of learning compounding information representing the compounding ratio of the composition of the learning adhesive and physical property information of the learning adhesive. The physical property information is, for example, moisture permeability, glass transition temperature, or adhesive strength of the adhesive.

In addition, the input device 12 receives composition information representing the composition ratio of the adhesive for which physical property information is to be estimated (hereinafter simply referred to as “composition information to be estimated”).

The information acquisition unit 16 of the physical property information estimation device 14 acquires learning data received by the input device 12 . The information acquisition unit 16 then stores the learning data in the data storage unit 18 . The information acquisition unit 16 also acquires the combination information to be inferred received by the input device 12 .

The data storage unit 18 stores learning data acquired by the information acquisition unit 16 . FIG. 3 shows an example of learning data stored in the data storage unit 18. As shown in FIG. In the example shown in FIG. 3, multiple pieces of learning data are stored in a table format. For example, in the table shown in FIG. 3 , the learning data with “ID” of “00001”, when the blending ratio of composition a is “AAA1” and the blending ratio of composition b is “BBB1”, It represents that the moisture permeability, which is an example of physical property information, was "P1", the glass transition temperature was "P2", and the adhesive strength was "P3". Note that in the present embodiment, some of the plurality of learning data stored in the data storage unit 18 are used for learning processing by the learning unit 20, and other data among the plurality of learning data are tested. used as data for For this reason, the learning data used for the learning process by the learning unit 20 is hereinafter referred to as a "learning sample". Further, learning data used as test data is hereinafter referred to as "test sample".

The learning unit 20 generates a learned model for estimating the physical property information of the adhesive from the compounding information of the adhesive, based on a plurality of learning samples.

Specifically, the learning unit 20 obtains the correct learning physical property information corresponding to the learning compounding information from among the plurality of learning samples, and the adhesive property information obtained by inputting the learning compounding information to the learned model. The learned model is trained so that the difference between the physical property information of the agent is small.

For example, when “AAA1” and “BBB1” are input to the trained model as the composition information of the composition of the learning sample shown in FIG. 3, “P1′” is output as the moisture permeability. In this case, the learned model is learned so that the difference between the correct moisture permeability values “P1” and “P1′” corresponding to “AAA1” and “BBB1” as the compounding information becomes smaller.

The learning unit 20 of the present embodiment is a learned model (hereinafter referred to as simply called "trained SVM"). Specifically, through learning processing by the learning unit 20, SVM parameters suitable for estimating the physical property information of the adhesive are obtained. Then, the learning unit 20 stores the generated learned SVM in the learned model storage unit 22 . The learned SVM of the present embodiment is a combination of parameters and a model formula based on SVM.

The learned model storage unit 22 stores the learned SVM generated by the learning unit 20 . In this embodiment, a case of generating a learned SVM for estimating moisture permeability, a learned SVM for estimating glass transition temperature, and a learned SVM for estimating adhesive strength will be described as an example.

Next, the process of estimating the physical property information of the adhesive will be described.

When estimating the physical property information of the adhesive, processing is performed by the data generating section 24, the estimating section 26, and the selecting section 28. FIG.

First, the data generating unit 24 generates a plurality of pieces of compounding information by changing the compounding information of the adhesive that serves as a reference.

When estimating physical property information of an adhesive with a new composition, the problem is how to determine the composition. Therefore, in the present embodiment, a plurality of pieces of combination information are generated by changing the combination information of the adhesive that serves as a reference. Then, in the present embodiment, a plurality of generated compounding information are input to the learned SVM, and physical property information corresponding to each of the plurality of compounding information is estimated. Then, in the present embodiment, combination information that satisfies a predetermined standard is selected from among the plurality of generated combination information.

The estimating unit 26 inputs each of the plurality of compounding information generated by the data generating unit 24 to the learned SVM stored in the learned model storage unit 22, thereby calculating physical properties corresponding to each of the plurality of compounding information. Guess information. As physical property information, for example, moisture permeability and glass transition temperature are estimated.

Specifically, the estimating unit 26 estimates the moisture permeability by inputting each of the plurality of combination information to the learned SVM for estimating the moisture permeability stored in the learned model storage unit 22 . Further, the estimation unit 26 estimates the glass transition temperature by inputting each of the plurality of pieces of compounding information to the learned SVM for estimating the glass transition temperature stored in the learned model storage unit 22 . In addition, the estimating unit 26 estimates the adhesive strength by inputting each of the plurality of compounding information to the learned SVM for estimating the adhesive strength stored in the learned model storage unit 22 .

The selection unit 28 selects physical property information that satisfies a predetermined criterion from the plurality of pieces of physical property information estimated by the estimation unit 26 . For example, the selection unit 28 selects those having a moisture permeability of 60 or less from a plurality of moisture permeability estimated as physical property information. Further, the selection unit 28 selects those having a predetermined value or higher from among the plurality of glass transition temperatures estimated as the physical property information. Further, the selection unit 28 selects adhesive strengths equal to or greater than a predetermined value from a plurality of adhesive strengths estimated as physical property information. The predetermined criteria are preset by the user.

Then, the selection unit 28 outputs the selected physical property information and the combination information corresponding to the selected physical property information.

The display device 40 displays the physical property information and the composition information output by the estimation unit 26 as a result.

As a result, compounding information that satisfies a predetermined standard is selected from a plurality of compounding information generated from a reference adhesive, and the user can obtain compounding information of an adhesive having physical properties desired by the user.

<Action of physical property information estimation device>

Next, the operation of the physical property information estimation device 14 will be described. When each piece of learning data representing a combination of learning compounding information and learning physical property information is input to the physical property information estimation device 14 , the information acquisition unit 16 stores the learning data in the data storage unit 18 . Then, when receiving the instruction signal for the learning process, the physical property information estimation device 14 executes the learning process routine shown in FIG.

<Learning processing routine>

In step S<b>100 , the learning unit 20 reads multiple learning samples from the learning data stored in the data storage unit 18 .

In step S102, the learning unit 20 learns an SVM, which is an example of a learning model, based on the learning samples read out in step S100, and obtains a trained SVM.

In step S<b>104 , the learning unit 20 stores the learned SVM learned in step S<b>102 in the learned model storage unit 22 .

In step S106, the learning unit 20 reads a plurality of test samples from the learning data stored in the data storage unit 18, and uses the plurality of test samples to evaluate the performance of the learned SVM.

In step S108, the learning unit 20 determines whether or not the performance of the SVM that has been learned in step S106 is equal to or greater than a predetermined threshold. If the performance is equal to or above the predetermined threshold, the learning processing routine is terminated. If the performance is less than the threshold, return to step S100. The performance of the learned SVM is determined, for example, by whether or not the error between the correct value and the estimated value is equal to or less than a predetermined threshold.

Next, when the physical property information estimating device 14 receives the input of the reference adhesive composition information, it executes the physical property information estimating processing routine shown in FIG.

<Physical Property Information Prediction Processing Routine>

In step S200, the information acquisition unit 16 acquires the blending information of the adhesive that serves as a reference.

In step S202, the data generation unit 24 generates a plurality of pieces of combination information by changing the reference adhesive combination information acquired in step S200.

In step S<b>204 , the estimation unit 26 reads the learned SVM stored in the learned model storage unit 22 .

In step S206, the estimation unit 26 inputs each of the plurality of mixture information generated in step S202 to the learned SVM read out in step S204, thereby corresponding to each of the plurality of mixture information. Infer physical information.

In step S208, the selection unit 28 selects physical property information that satisfies a predetermined criterion from the plurality of physical property information estimated in step S206. For example, physical property information with a water vapor permeability of 60 or less is selected from a plurality of water vapor permeability estimated as physical property information.

In step S210, the selection unit 28 outputs the physical property information selected in step S208 and the combination information corresponding to the selected physical property information as a result.

The display device 40 displays the physical property information and the composition information output by the estimation unit 26 as a result.

The user confirms the results displayed on the display device 40 . Then, the user refers to the compounding information displayed on the display device 40 to manufacture a new adhesive.

As described above, the physical property information estimation device 14 according to the present embodiment obtains the composition information representing the composition ratio of the adhesive composition to be estimated from the composition information representing the composition ratio of the adhesive composition. Input to the learned SVM for estimating the physical property information of the adhesive, and guess the physical property information of the adhesive corresponding to the input compounding information. Thereby, physical property information of the adhesive can be inferred from the compounding information of the adhesive.

Further, the physical property information estimation device 14 according to the present embodiment generates a plurality of pieces of composition information by changing the composition information of the adhesive that serves as a reference. Then, the physical property information estimating device 14 inputs each of the plurality of generated compounding information to the learned SVM, thereby estimating the physical property information corresponding to each of the plurality of compounding information. Then, the physical property information estimation device 14 selects physical property information that satisfies a predetermined criterion from a plurality of estimated physical property information, and outputs the selected physical property information and the compounding information corresponding to the selected physical property information. This provides information on what formulation should be used to manufacture an adhesive having desired physical properties.

Further, the physical property information estimating device 14 according to the present embodiment uses learning data representing a combination of learning compounding information representing the compounding ratio of the composition of the learning adhesive and the physical property information of the learning adhesive. Based on this, a learned SVM is generated for estimating the physical property information of the adhesive from the compounding information representing the compounding ratio of the composition of the adhesive. As a result, it is possible to obtain a learned SVM for estimating the physical property information of the adhesive from the compounding information of the adhesive.

[Second embodiment]

Next, a second embodiment will be described. The physical property information estimating system of the second embodiment differs from the first embodiment in that the physical property information is estimated by selecting compounding information as an explanatory variable.

When a user mixes compositions to produce an adhesive, there are cases where he wants to know information such as which composition works on which physical property. In addition, even if the physical property information of the adhesive is estimated using a composition that does not affect the desired physical properties, it is expected that accurate results cannot be obtained.

Therefore, the physical property information estimating device 14 of the second embodiment identifies a composition having a high correlation with the physical property information. Specifically, the physical property information estimation device 14 of the present embodiment uses a genetic algorithm, which is a known technique, to set various compositions as explanatory variables, and selects a composition suitable for estimation from among them. Identify. A genetic algorithm is an algorithm for optimization that imitates the heredity of organisms.

First, the learning section 20 reads out a plurality of learning samples stored in the data storage section 18 . Then, the learning unit 20 sets a specific composition from each composition in the learning compounding information of the plurality of learning samples. Setting a specific composition by the learning unit 20 corresponds to generating a new generation of individuals in the genetic algorithm.

One individual in the genetic algorithm of this embodiment is represented by the vector shown in the following formula (1). Each element of the vector below indicates whether or not the composition of each composition is used as an explanatory variable.

[0, 1, 0, 0, ...]
(1)

For example, the first element “0” in the vector [0, 1, 0, 0, . Also, the second element "1" in the vector [0, 1, 0, 0, ...] in the above formula (1) indicates that the blend of composition b is used as an explanatory variable.

The learning unit 20 sets a plurality of individuals, and sets the formulation of the composition corresponding to the individuals as learning formulation information used for learning.

Based on a plurality of learning samples, the learning unit 20 obtains the adhesive composition information X based on the learning composition information set according to the individual and the correct physical property information corresponding to the learning composition information. Generate a trained model for estimating physical property information Y from

Note that the learning unit 20 of the second embodiment is a learned model ( hereinafter simply referred to as “learned PLS”). Specifically, partial least-squares regression provides parameters suitable for estimating adhesive physical property information. Then, the learning unit 20 stores the generated learned PLS in the learned model storage unit 22 . The learned PLS of the present embodiment is a combination of parameters obtained by partial least squares regression and a model formula by PLS.

Then, the learning unit 20 sets the combination information according to the individual for the plurality of test samples stored in the data storage unit 18 . That is, which composition of the composition information is to be used is set according to the individual.

Then, the learning unit 20 outputs the correct physical property information of the adhesive of the test sample and the physical property information of the adhesive output when the combination information set according to the individual for the test sample is input to the learned PLS. Calculate the difference between In the present embodiment, an example will be described in which correlation is used as the difference between the correct adhesive physical property information and the adhesive physical property information output from the learned PLS. It can be said that the higher the correlation, the more accurately the learned PLS can estimate the physical property information.

Therefore, in the present embodiment, the setting of the learning compounding information used for learning, the learning of the learned PLS, and the calculation of the correlation are repeated until the correlation calculated by the learning unit 20 satisfies a predetermined criterion. Specifically, each process is repeated until the correlation reaches or exceeds a predetermined threshold (0.9, for example).

It should be noted that the setting of a new vector in the above formula (1) in the iterative processing of this embodiment corresponds to the generation of a new generation of individuals in the genetic algorithm. When the learning unit 20 generates new generation individuals, the new generation individuals are generated by mutation and crossover of the previous generation individuals, and new vectors are set. Calculation of the correlation in the iterative processing of this embodiment corresponds to calculation of the evaluation function in the genetic algorithm.

When the correlation value R ² between the correct value of the physical property information and the estimated value of the physical property information output from the learned PLS becomes equal to or greater than a predetermined threshold by the above repeated processing, the composition set by the learning unit 20 The combination of materials is an effective composition for inferring physical property information of the adhesive. For this reason, the learning unit 20 sets a combination of compositions used as an explanatory variable when the correlation is equal to or greater than a predetermined threshold as a combination of compositions used when estimating the physical property information of the adhesive. As a result, it is specified which of the compositions of the adhesive is to be used for estimating the physical property information.

As described above, the physical property information estimating device 14 according to the present embodiment is set so as to use a specific composition composition as an explanatory variable from the composition information of the adhesive composition. Thereby, it is possible to specify a composition for accurately estimating physical property information from the compounding information of the adhesive.

[Third embodiment]

Next, a third embodiment will be described. The physical property information estimation system of the third embodiment differs from the first and second embodiments in that physical property information is estimated by two-step estimation processing.

As described in the first embodiment, examples of physical property information of the adhesive include moisture permeability, glass transition temperature, and adhesive strength.

It is believed that each of moisture permeability, glass transition temperature, and adhesive strength is correlated with each other. For this reason, rather than simply estimating physical property information from the compounding information of the adhesive, for example, after estimating the moisture permeability and the glass transition temperature, using those moisture permeability and the glass transition temperature, the adhesive strength can be estimated. It is expected that a good estimation result of

Therefore, in this embodiment, after estimating the moisture permeability and the glass transition temperature of the physical property information, the adhesion strength of the physical property information is estimated using the moisture permeability or the glass transition temperature.

A trained SVM, which is an example of a trained model of the third embodiment, includes a first trained SVM, a second trained SVM, and a third trained SVM.

The first learned SVM estimates the moisture permeability as physical property information from the compounding information of the adhesive. The second learned SVM estimates the glass transition temperature as physical property information from the compounding information of the adhesive.

The third learned SVM obtains adhesion as physical property information from at least one of the compounding information of the adhesive and the moisture permeability estimated by the first learned SVM and the glass transition temperature estimated by the second learned SVM. Guess strength.

Therefore, when estimating the adhesive strength as physical property information, the estimating unit 26 of the third embodiment uses the moisture permeability estimated by the first learned SVM or the glass transition estimated by the second learned SVM. The temperature and the compounding information of the adhesive to be inferred are input to the third learned SVM. Then, the estimation unit 26 of the third embodiment estimates the adhesive strength of the adhesive corresponding to the input compounding information.

The third learned SVM is learned based on learning data in which learning composition information, learning moisture permeability or learning glass transition temperature, and learning physical property information are combined.

As described above, the physical property information estimation device 14 according to the third embodiment, when estimating the adhesive strength as physical property information, the moisture permeability or The glass transition temperature of the adhesive to be inferred, which is estimated by the second learned SVM, and the composition information of the adhesive to be estimated are input to the third learned SVM, and correspond to the input composition information. Estimate the bond strength of the adhesive to be used. As a result, the physical property information of the adhesive can be accurately estimated by the two-step estimation process.

In the first and third embodiments, the case of using SVM as a model was explained, and in the second embodiment, the case of using PLS, which is a kind of linear modeling, was explained as an example. However, it is not limited to this. For example, models for estimating the physical property information of the adhesive from the compounding information of the adhesive include multiple regression analysis, SVM (support vector machine), LASSO (least absolute shrinkage and selection operator), ridge regression, elastic net , Gaussian processes, decision trees, random forests, gradient boosting, or neural nets.

In addition, in the above-described first embodiment, the case of generating a plurality of pieces of compounding information by changing the compounding information of the adhesive used as a reference has been described as an example, but the present invention is not limited to this. For example, a plurality of pieces of combination information may be randomly generated.

Next, each example is shown. Two examples are described below.

[Example 1]

The results of Example 1 are shown in FIG. In addition, Example 1 was performed under the following conditions.

Objective variable: Glass transition temperature Tg and moisture permeability Explanatory variable: Mixing information of adhesive composition Model: Linear support vector machine (hereinafter simply referred to as linear SVM)

The linear SVM model was verified by 10fold cross-validation. A trained SVM for estimating moisture permeability and a trained SVM for estimating glass transition temperature were generated using the above learning data.

The graph on the left side of Figure 6 shows the actual moisture permeability (labeled "Actual Y" in Figure 6) and the moisture permeability estimated by the trained SVM for moisture permeability estimation (labeled "Estimated Y in CV" in Figure 6). notation). The unit of moisture permeability is [g/m2·24hr]. The coefficient of determination between the actual and estimated moisture permeability is R2 = 0.61, and the mean absolute error (MAE) between the actual and estimated moisture permeability is MAE = was 19.5.

The graph on the right side of FIG. 6 shows the actual glass transition temperature (denoted as “Actual Y” in FIG. 6) and the glass transition temperature estimated by the trained SVM for estimating the glass transition temperature (“Estimated Y in CV”). The unit of the glass transition temperature is [°C]. Note that the coefficient of determination between the actual and estimated glass transition temperatures, R2=0.61, and the mean absolute error between the actual and estimated glass transition temperatures, MAE=7.9. rice field.

In addition, in the graph of FIG. 6, outliers are removed. As shown in FIG. 6, it can be seen that the estimated values and the actual values match, and a highly accurate trained SVM is obtained.

Next, FIG. 7 and FIG. 8 show results of standard regression coefficients in trained SVM. a, b, c, ..., A, B, C and a', b', c', ... A', B', C' shown in FIGS. represents each composition that causes The values in Figures 7 and 8 are the standard regression coefficient values for each composition. 7 and 8 show the standard regression coefficients of the above 30 compositions with large absolute values of the standard regression coefficients.

Next, as in the first embodiment, a plurality of pieces of compounding information were generated by changing the compounding information of the adhesive composition.

FIG. 9 shows a diagram showing how a plurality of pieces of compounding information are generated by variously changing the compounding information of the adhesive. FIG. 9 is a diagram of compounding information, which is high-dimensional data, compressed into two-dimensional data by a known technique, the t-SNE method. The horizontal axis tSNE1 in FIG. 9 represents the first axis obtained by the t-SNE method, and the vertical axis tSNE2 represents the second axis obtained by the t-SNE method. X in FIG. 9 represents a composition sample group in which a part of the composition of the standard adhesive was changed.

As shown in FIG. 9, it can be seen that there are samples close to the target estimated values in the compounding information that is close to the actual experimental data.

A plurality of compounding information obtained as described above are input to the learned SVM for estimating moisture permeability and the learned SVM for estimating glass transition temperature, and the moisture permeability as physical property information corresponding to the compounding information is input. and the glass transition temperature were estimated.

FIG. 10 shows the results of water vapor permeability estimated by the learned SVM for water vapor permeability estimation. FIG. 10 is a table showing the moisture permeability estimated from the newly generated compounding information and the moisture permeability of the adhesive manufactured based on the compounding information.

For each of "NEW-1", "NEW-2", "NEW-3" and "NEW-4", the adhesive corresponding to the compounding information obtained by changing the compounding information of the standard adhesive is used. represents. As shown in FIG. 10, it can be seen that the moisture permeability is estimated with high accuracy. In addition, the estimation error of moisture permeability is 20% or less for any adhesive.

FIG. 11 shows the result of the glass transition temperature estimated by the trained SVM for glass transition temperature estimation. FIG. 11 is a table showing the glass transition temperature Tg estimated from the newly generated compounding information and the glass transition temperature Tg of the adhesive manufactured based on the compounding information.

For each of "NEW-1", "NEW-2", "NEW-3" and "NEW-4", the adhesive corresponding to the compounding information obtained by changing the compounding information of the standard adhesive is used. represents. As shown in FIG. 11, it can be seen that the glass transition temperature Tg is estimated with high accuracy.

Further, FIG. 12 shows the results of an experiment aimed at lowering the moisture permeability while maintaining the same level of adhesive strength as the current product.

Each of "NEW-1" and "NEW-2" shown in FIG. 12 represents an adhesive corresponding to compounding information obtained by changing the compounding information of the reference adhesive "existing product X". there is It can be seen that the moisture permeability of these "NEW-1" and "NEW-2" is lower than that of the existing product X, as shown on the right side of FIG. Also, although not shown, it has been confirmed that each of "NEW-1" and "NEW-2" is near the target value of 100°C. For this reason, it can be said that the moisture permeability can be lowered while maintaining the same level of strength as the current product.

[Example 2]

Next, the results of Example 2 are shown. In addition, Example 2 was performed under the following conditions.

In Example 1 above, physical property information was estimated by one-step estimation. However, for the adhesive strength, the accuracy of the one-step estimation is low. Therefore, similarly to the third embodiment, after estimating the moisture permeability and the glass transition temperature, the adhesive strength was estimated in consideration of these values.

In addition, Example 2 was performed under the following conditions.

Objective variable: Adhesive strength Explanatory variable: Adhesive composition formulation, estimated moisture permeability and estimated glass transition temperature Method: Generate learned PLS using genetic algorithm and PLS, leave-one Evaluate the model with -out

The results of Example 2 are shown in FIG. The graph on the left side of Fig. 13 shows the relationship between the actual adhesive strength (labeled "Actual Y" in Figure 13) and the adhesive strength estimated by the learned Random Forest (labeled "Calculated Y" in Figure 13). represents Note that the coefficient of determination between the actual bond strength and the estimated bond strength is R ² =0.28, and the mean absolute error (MAE) between the actual bond strength and the estimated bond strength is MAE. = 4.94.

The graph on the right side of FIG. 13 is an estimation result obtained by a two-stage estimation in which the estimated glass transition temperature and the estimated moisture permeability are added as explanatory variables to estimate the adhesive strength. The graph on the right side of FIG. 13 shows the relationship between the actual adhesive strength (denoted as “Actual Y” in FIG. 13) and the adhesive strength estimated by the learned PLS (denoted as “Estimated Y in CV” in FIG. 13). represents Note that the coefficient of determination between the actual and estimated bond strengths was R ² =0.49 and the mean absolute error between the actual and estimated bond strengths MAE=3.93.

From FIG. 13, it can be seen that two-stage estimation yields more accurate results than one-stage estimation.

Also, FIG. 14 shows the results of comparison between the one-stage estimation and the two-stage estimation. It can be seen that two-stage estimation yielded more accurate results than one-stage estimation for those with high adhesive strength as targeted (NEW-1, NEW-2).

The present invention is not limited to the above-described embodiments, and various modifications and applications are possible without departing from the gist of the present invention.

For example, the case of estimating moisture permeability, glass transition temperature, and adhesive strength as physical property information has been described as an example, but the present invention is not limited to this, and other physical properties may be estimated.

Further, in the present embodiment, the case where the physical property information estimation device 14 executes both the learning process and the physical property information estimation process has been described as an example, but the present invention is not limited to this. For example, the system may be configured by a learning device that executes the learning process and a physical property information estimation device that executes the physical property information estimation process.

Further, in the specification of the present application, an embodiment in which the program is pre-installed has been described, but it is also possible to store the program in a computer-readable recording medium and provide it.

10 physical property information estimation system 12 input device 14 physical property information estimation device 16 information acquisition unit 18 data storage unit 20 learning unit 22 trained model storage unit 24 data generation unit 26 estimation unit 28 selection unit 40 display device 50 computer

Claims (9)

Moisture permeability, glass transition temperature, or adhesive strength as physical property information of the adhesive is estimated from formulation information representing the blending ratio of the composition of the adhesive to be inferred. input to a trained model for, and estimate the moisture permeability, glass transition temperature, or adhesive strength as physical property information of the adhesive corresponding to the input compounding information;
A physical property information estimation method in which processing is performed by a computer. generating a plurality of the compounding information by changing the compounding information of the reference adhesive;
estimating the physical property information corresponding to each of the plurality of blending information by inputting each of the plurality of blending information generated into the learned model;
Selecting the physical property information that satisfies a predetermined criterion from the plurality of inferred physical property information, and outputting the selected physical property information and the combination information corresponding to the selected physical property information;
The physical property information estimation method according to claim 1. randomly generating a plurality of the compounding information;
estimating the physical property information corresponding to each of the plurality of blending information by inputting each of the plurality of blending information generated into the learned model;
Selecting the physical property information that satisfies a predetermined criterion from the plurality of inferred physical property information, and outputting the selected physical property information and the combination information corresponding to the selected physical property information;
The physical property information estimation method according to claim 1. inputting the compounding information representing the compounding ratio of the adhesive composition to be inferred into a learned model for estimating the physical property information of the adhesive from the compounding information representing the compounding ratio of the adhesive composition, and inputting A physical property information estimation method in which a computer executes processing for estimating physical property information of the adhesive corresponding to the compounding information obtained,
The trained model is
a first trained model for estimating moisture permeability as the physical property information from the combination information;
a second trained model for estimating the glass transition temperature as the physical property information from the compounding information;
and a third trained model for estimating the adhesive strength as the physical property information from the compounding information and at least one of the moisture permeability and the glass transition temperature. when
The moisture permeability estimated by the first trained model or the glass transition temperature and the composition information estimated by the second trained model are input to the third trained model, and the inputted estimating the adhesive strength of the adhesive corresponding to formulation information;
Physical property information estimation method. Based on learning data representing a combination of learning compounding information representing the blending ratio of the composition of the learning adhesive and moisture permeability, glass transition temperature or adhesive strength as physical property information of the learning adhesive, Generating a trained model for estimating moisture permeability, glass transition temperature, or adhesive strength as physical property information of the adhesive from compounding information representing the compounding ratio of the composition of the adhesive,
A physical property information inference model generation method in which processing is executed by a computer. an information acquiring unit that acquires compounding information representing the compounding ratio of the adhesive composition to be inferred;
The compounding information acquired by the information acquiring unit is learned for estimating the moisture permeability, the glass transition temperature, or the adhesive strength as the physical property information of the adhesive from the compounding information representing the compounding ratio of the composition of the adhesive. an estimating unit for inputting to a model and estimating the moisture permeability, glass transition temperature or adhesive strength as physical property information of the adhesive corresponding to the input compounding information;
Physical property information estimation device including. Based on learning data representing a combination of learning compounding information representing the blending ratio of the composition of the learning adhesive and moisture permeability, glass transition temperature or adhesive strength as physical property information of the learning adhesive, Physical property information estimation model generation including a learning unit for generating a trained model for estimating moisture permeability, glass transition temperature, or adhesive strength as physical property information of the adhesive from composition information representing the composition ratio of the adhesive composition Device. Moisture permeability, glass transition temperature, or adhesive strength as physical property information of the adhesive is estimated from formulation information representing the blending ratio of the composition of the adhesive to be inferred. input to a trained model for, and estimate the moisture permeability, glass transition temperature, or adhesive strength as physical property information of the adhesive corresponding to the input compounding information;
A physical property information estimation program for causing a computer to execute processing. Based on learning data representing a combination of learning compounding information representing the blending ratio of the composition of the learning adhesive and moisture permeability, glass transition temperature or adhesive strength as physical property information of the learning adhesive, Generating a trained model for estimating moisture permeability, glass transition temperature, or adhesive strength as physical property information of the adhesive from compounding information representing the compounding ratio of the composition of the adhesive,
A physical property information inference model generation program for causing a computer to execute processing.

Images