Learning model generation method, program, storage medium, and learned model

Abstract

To provide a new learning model generation method on manufacture of an impregnated body by impregnating a fluoropolymer aqueous dispersion liquid in a glass cloth, and the like.SOLUTION: A learning model generation method for generating a learning model in which an impregnated body is evaluated using a computer comprises an acquisition step, a learning step, and a generation step. In the acquisition step, a computer acquires information including at least glass cloth information, dispersion liquid information, and evaluation of the impregnated body as teacher data. In the learning step, the computer learns on the basis of a plurality of pieces of teacher data acquired by the acquisition step. In the generation step, the computer generates a learning model on the basis of results learned in the learning step. The learning model outputs an evaluation with input information, which is unknown information different from teacher data, as input. The input information includes at least glass cloth information and dispersion liquid information.SELECTED DRAWING: Figure 1

Description

Technical Field

[0001] The present disclosure relates to a learning model generation method, a program, a storage medium storing the program, and a learned model.

Background Art

[0002] Patent Document 1 discloses a polytetrafluoroethylene film-like material having an impregnated body and a film obtained by impregnating the impregnated body with a specific polytetrafluoroethylene aqueous dispersion composition.

[0003] Patent Document 2 discloses an optimization analysis device and a storage medium storing an optimization analysis program.

Prior Art Documents

Patent Documents

[0004]

Patent Document 1

Patent Document 2

Summary of the Invention

Problems to be Solved by the Invention

[0005] An object of the present disclosure is to provide a novel learning model generation method, a program, a storage medium storing the program, and a learned model related to the production of an impregnated body obtained by impregnating a glass cloth with a fluoropolymer aqueous dispersion.

Means for Solving the Problems

[0006] The present disclosure is a learning model generation method for generating a learning model that determines, using a computer, an evaluation of an impregnated body obtained by impregnating a glass cloth with a fluoropolymer aqueous dispersion, An acquisition step (S12) in which the computer acquires, as teacher data, information including at least glass cloth information which is the information of the glass cloth, dispersion liquid information which is the information of the fluoropolymer aqueous dispersion, and the evaluation of the impregnated body; A learning step (S15) in which the computer learns based on a plurality of pieces of the teacher data acquired in the acquisition step (S12); A generation step (S16) in which the computer generates the learning model based on the result learned in the learning step (S15); comprising; The learning model takes, as input, input information which is unknown information different from the teacher data, and outputs the evaluation; The input information is information including at least the glass cloth information and the dispersion liquid information. Relates to a learning model generation method.

[0007] The present disclosure is a learning model generation method for generating a learning model that determines, using a computer, optimal dispersion liquid information for obtaining an evaluation of a target impregnated body, An acquisition step (S12) in which the computer acquires, as teacher data, information including at least glass cloth information which is the information of the glass cloth, dispersion liquid information which is the information of the fluoropolymer aqueous dispersion to be impregnated into the glass cloth, and the evaluation of the impregnated body obtained by impregnating the glass cloth with the dispersion liquid; A learning step (S15) in which the computer learns based on a plurality of pieces of the teacher data acquired in the acquisition step (S12); A generation step (S16) in which the computer generates a learning model based on the result learned in the learning step (S15); comprising; The learning model takes, as input, input information which is unknown information different from the teacher data, and outputs optimal dispersion liquid information for obtaining an evaluation of a target impregnated body; The input information is information including at least the glass cloth information and the information of the evaluation. Also relates to a learning model generation method.

[0008] The present disclosure is a learning model generation method for generating a learning model that determines, using a computer, optimal glass cloth information for obtaining an evaluation of a target impregnated body, comprising: an acquisition step (S12) in which a computer acquires, as teacher data, information including at least dispersion liquid information which is information on a fluoropolymer aqueous dispersion liquid, glass cloth information which is information on a glass cloth to be impregnated with the dispersion liquid, and an evaluation of an impregnated body obtained by impregnating the glass cloth with the dispersion liquid; a learning step (S15) in which the computer learns based on a plurality of pieces of the teacher data acquired in the acquisition step (S12); a generation step (S16) in which the computer generates a learning model based on the result of learning in the learning step (S15); and the learning model takes, as input, input information which is unknown information different from the teacher data, and outputs optimal glass cloth information for obtaining an evaluation of a target impregnated body, wherein the input information is information including at least the dispersion liquid information and the evaluation information. It also relates to a learning model generation method.

[0009] The teacher data further includes processing condition information which is information on processing conditions for obtaining the impregnated body, and preferably, the input information further includes the processing condition information.

[0010] The present disclosure is a learning model generation method for generating a learning model that determines, using a computer, optimal processing condition information for obtaining an evaluation of a target impregnated body, At least, glass cloth information which is information of the glass cloth, dispersion liquid information which is information of the fluoropolymer aqueous dispersion to be impregnated into the glass cloth, processing condition information which is information of processing conditions for obtaining an impregnated body by impregnating the glass cloth with the dispersion liquid, and evaluation of the impregnated body obtained by impregnating the glass cloth with the dispersion liquid. An acquisition step (S12) in which a computer acquires the information including these as teacher data. A learning step (S15) in which the computer learns based on a plurality of pieces of the teacher data acquired in the acquisition step (S12). A generation step (S16) in which the computer generates a learning model based on the result learned in the learning step (S15). Comprising The learning model takes, as input, input information which is unknown information different from the teacher data, and outputs optimal processing condition information for obtaining an evaluation of a target impregnated body. The input information is information including at least the glass cloth information, the dispersion liquid information, and the evaluation information. Also relates to a learning model generation method.

[0011] It is preferable that the learning step (S15) performs learning by regression analysis and / or ensemble learning in which a plurality of regression analyses are combined.

[0012] The present disclosure is a program in which a computer determines an evaluation of an impregnated body obtained by impregnating a glass cloth with a fluoropolymer aqueous dispersion using a learning model, An input step (S22) in which the computer is input with input information. A determination step (S23) in which the computer determines the evaluation. An output step (S24) in which the computer outputs the evaluation determined in the determination step (S23). Comprising The above learning model learns, using information including at least the glass cloth information which is information about the above glass cloth, the dispersion liquid information which is information about the above fluoropolymer aqueous dispersion liquid, and the above evaluation, as teacher data. The above input information is information including at least the above glass cloth information and the above dispersion liquid information, and is unknown information different from the above teacher data. It also relates to the program.

[0013] The present disclosure is a program in which a computer determines dispersion liquid information which is information about an optimal fluoropolymer aqueous dispersion liquid for obtaining an evaluation of a target impregnated body, using a learning model. An input step (S22) in which the above computer inputs input information. A determination step (S23) in which the above computer determines the optimal above dispersion liquid information. An output step (S24) in which the above computer outputs the optimal above dispersion liquid information determined in the above determination step (S23). It comprises: The above learning model learns, using information including at least the glass cloth information which is information about the above glass cloth, the dispersion liquid information which is information about the above fluoropolymer aqueous dispersion liquid to be impregnated into the above glass cloth, and the evaluation of the impregnated body obtained by impregnating the above glass cloth with the above fluoropolymer aqueous dispersion liquid, as teacher data. The above input information is information including at least the above glass cloth information and the above evaluation information, and is unknown information different from the above teacher data. It also relates to the program.

[0014] The present disclosure is a program in which a computer determines glass cloth information which is information about an optimal glass cloth for obtaining an evaluation of a target impregnated body, using a learning model. An input step (S22) in which the above computer inputs input information. A determination step (S23) in which the above computer determines the optimal above glass cloth information. An output step (S24) in which the computer outputs the optimal glass cloth information determined in the determination step (S23); comprising The learning model learns, as teacher data, information including at least dispersion liquid information that is information on the fluoropolymer aqueous dispersion, glass cloth information that is information on the glass cloth to be impregnated with the dispersion liquid, and an evaluation of an impregnated body obtained by impregnating the glass cloth with the dispersion liquid. The input information is information including at least the dispersion liquid information and the evaluation information, and is unknown information different from the teacher data. It also relates to a program.

[0015] The teacher data further includes processing condition information that is information on processing conditions for obtaining the impregnated body. The input information preferably further includes the processing condition information.

[0016] The present disclosure is a program in which a computer determines, using a learning model, processing condition information that is information on optimal processing conditions for obtaining an evaluation of a target impregnated body, an input step (S22) in which the computer inputs input information; a determination step (S23) in which the computer determines the optimal processing condition information; an output step (S24) in which the computer outputs the optimal processing condition information determined in the determination step (S23); comprising The learning model learns, as teacher data, information including at least glass cloth information that is information on the glass cloth, dispersion liquid information that is information on the fluoropolymer aqueous dispersion to be impregnated into the glass cloth, processing condition information that is information on processing conditions for obtaining an impregnated body by impregnating the glass cloth with the dispersion liquid, and an evaluation of the impregnated body obtained by impregnating the glass cloth with the dispersion liquid. The above input information is information including at least the above glass cloth information, the above dispersion information, and the above evaluation information, and is unknown information different from the above teacher data. It also relates to a program.

[0017] The above processing condition information preferably includes information related to at least one selected from the group consisting of the solid content concentration during processing of the above dispersion, the surface tension during processing, the additives during processing, the number of impregnations and the impregnation amount, and the heating conditions.

[0018] The above evaluation preferably includes information related to at least one selected from the group consisting of the crack limit film thickness, color, peelability, bending life, surface roughness, electrical properties, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, and surface resistivity of the above impregnated body.

[0019] The above glass cloth information preferably includes information related to at least one selected from the group consisting of the weaving and knitting method, thickness, count, density, twist, filament diameter, weight, type, laminate structure, and tensile strength of the above glass cloth.

[0020] The above dispersion information preferably includes at least one type of information selected from the group consisting of information related to the fluoropolymer contained in the above dispersion, information related to components other than the above fluoropolymer contained in the above dispersion, and information related to the properties of the above dispersion.

[0021] The present disclosure also relates to a storage medium storing any of the above programs.

[0022] The present disclosure is a learned model for causing a computer to function so as to perform an operation based on the weighting coefficients of the neural network on the glass cloth information and the dispersion information input to the input layer of the neural network, and output an evaluation of the impregnated body from the output layer of the neural network, The above weighting coefficients are obtained by learning using at least the above glass cloth information, the above dispersion information, and the above evaluation as teacher data. The above glass cloth information is information about the above glass cloth, The above dispersion liquid information is information about the aqueous dispersion of fluoropolymer to be impregnated into the above glass cloth, The above evaluation is an evaluation of the impregnated body obtained by impregnating the above glass cloth with the above dispersion liquid, It also relates to the learned model.

[0023] The present disclosure is a learned model for causing a computer to function so as to perform an operation based on the weighting coefficient of the neural network on the glass cloth information and the evaluation information input to the input layer of the neural network, and output optimal dispersion liquid information for obtaining a target evaluation from the output layer of the neural network, The above weighting coefficient is obtained by learning using at least the above glass cloth information, the above dispersion liquid information, and the above evaluation as teacher data, The above glass cloth information is information about the above glass cloth, The above dispersion liquid information is information about the aqueous dispersion of fluoropolymer to be impregnated into the above glass cloth, The above evaluation is an evaluation of the impregnated body obtained by impregnating the above glass cloth with the above dispersion liquid, It also relates to the learned model.

[0024] The present disclosure is a learned model for causing a computer to function so as to perform an operation based on the weighting coefficient of the neural network on the dispersion liquid information and the evaluation information input to the input layer of the neural network, and output optimal glass cloth information for obtaining a target evaluation from the output layer of the neural network, The above weighting coefficient is obtained by learning using at least the above glass cloth information, the above dispersion liquid information, and the above evaluation as teacher data, The above dispersion liquid information is information about the aqueous dispersion of fluoropolymer, The above glass cloth information is information about the glass cloth to be impregnated with the above dispersion liquid, The above evaluation is an evaluation of an impregnated body obtained by impregnating the above glass cloth with the above dispersion liquid. It also relates to a learned model.

[0025] The above teacher data further includes processing condition information which is information on processing conditions for obtaining the above impregnated body. It is preferable that the above processing condition information is further input to the above input layer.

[0026] The present disclosure is a learned model for causing a computer to function so as to perform an operation based on the weighting coefficients of the neural network on the glass cloth information, dispersion liquid information, and evaluation information input to the input layer of the neural network, and output optimal processing condition information for obtaining a target evaluation from the output layer of the neural network, The above weighting coefficients are obtained by learning using at least the above glass cloth information, the above dispersion liquid information, the above processing condition information, and the above evaluation as teacher data. The above glass cloth information is information on the above glass cloth. The above dispersion liquid information is information on the fluoropolymer aqueous dispersion liquid to be impregnated into the above glass cloth. The above processing condition information is information on processing conditions for obtaining an impregnated body by impregnating the above glass cloth with the above dispersion liquid. The above evaluation is an evaluation of an impregnated body obtained by impregnating the above glass cloth with the above dispersion liquid. It also relates to a learned model.

Advantages of the Invention

[0027] According to the present disclosure, it is possible to provide a novel learning model generation method, program, storage medium storing the program, and learned model for manufacturing an impregnated body obtained by impregnating a glass cloth with a fluoropolymer aqueous dispersion liquid.

Brief Description of the Drawings

[0028]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Figure 7

Figure 8

Figure 9

Figure 10

Figure 11

Mode for Carrying Out the Invention

[0029] Hereinafter, a learning model according to an embodiment of the present disclosure will be described. Note that the following embodiments are specific examples and do not limit the technical scope, and can be appropriately changed without departing from the gist.

[0030] (1) Overview FIG. 1 is a diagram showing the configuration of the learning model generation device. FIG. 2 is a diagram showing the configuration of the user device. The learning model is generated by a learning model generation device 10 which is one or more computers acquiring and learning training data. The generated learning model is implemented as a so-called learned model on general-purpose computers or terminals, or downloaded as a program or the like, or distributed in a state stored in a storage medium, and is used in a user device 20 which is one or more computers.

[0031] The learning model can output correct answers for unknown information different from the teacher data. Furthermore, the learning model can be updated so that correct answers are output for various input data.

[0032] (2) Configuration of the learning model generation device 10 The learning model generation device 10 generates a learning model used in the user device 20 described later. The learning model generation device 10 is a device having the functions of a so-called computer. The learning model generation device 10 may include a communication interface such as a NIC and a DMA controller, and may be capable of communicating with the user device 20 etc. via a network. Although the learning model generation device 10 shown in FIG. 1 is illustrated as one device, it is preferable that the learning model generation device 10 supports cloud computing. For this reason, the hardware configuration of the learning model generation device 10 does not need to be housed in one housing or provided as an integrated device. For example, it is configured by dynamically connecting and disconnecting the resources of the hard learning model generation device 10 according to the load. The learning model generation device 10 has a control unit 11 and a storage unit 14.

[0033] (2-1) Control unit 11 The control unit 11 is, for example, a CPU, and controls the entire learning model generation device 10. The control unit 11 causes each functional unit described later to function appropriately, and executes the learning model generation program 15 stored in advance in the storage unit 14. The control unit 11 has functional units such as an acquisition unit 12 and a learning unit 13.

[0034] Among the control unit 11, the acquisition unit 12 acquires the teacher data input to the learning model generation device 10, and stores the acquired teacher data in the database 16 constructed in the storage unit 14. The teacher data may be directly input to the learning model generation device 10 by the person using the learning model generation device 10, or may be acquired from other devices etc. via a network.

[0035] The method for acquiring the teacher data by the acquisition unit 12 is not particularly limited. The teacher data is information for generating a learning model that achieves the learning objective. Here, the learning objective is any one of outputting an evaluation of an impregnated body obtained by impregnating a fluoropolymer aqueous dispersion into a glass cloth, outputting optimal dispersion liquid information for obtaining an evaluation of a target impregnated body, outputting optimal glass cloth information for obtaining an evaluation of a target impregnated body, and outputting optimal processing condition information for obtaining an evaluation of a target impregnated body. Details will be described later.

[0036] The learning unit 13 extracts a learning data set from the teacher data stored in the storage unit 14 and automatically performs machine learning. The learning data set is a set of data for which the correct answer for the input is known. The learning data set extracted from the teacher data varies depending on the learning objective. By performing learning by the learning unit 13, a learning model is generated.

[0037] (2-2) Machine learning The method of machine learning performed by the learning unit 13 is not particularly limited as long as it is supervised learning using the learning data set. Examples of models or algorithms used in supervised learning include regression analysis, decision trees, support vector machines, neural networks, ensemble learning, random forests, and the like. Also, class classification may be performed in advance, and then supervised learning may be performed for each class. The class classification at that time may be either supervised or unsupervised.

[0038] Regression analysis includes, for example, linear regression analysis, multiple regression analysis, and logistic regression analysis. Regression analysis is a method of fitting a model between input data (explanatory variables) and learning data (objective variables) using the least squares method or the like. The dimension of the explanatory variables is 1 in linear regression analysis and 2 or more in multiple regression analysis. In logistic regression analysis, a logistic function (sigmoid function) is used as the model. Also, when the dimension of the explanatory variables is large, it is preferable to perform dimension compression such as principal component regression analysis or partial least squares regression analysis and then perform regression analysis.

[0039] A decision tree is a model for generating a complex discrimination boundary by combining multiple discriminators. Details of the decision tree will be described later.

[0040] A support vector machine is an algorithm for generating a linear discrimination function for two classes. Details of the support vector machine will be described later.

[0041] A neural network models a network formed by connecting neurons in the human nervous system with synapses. In a narrow sense, a neural network means a multi-layer perceptron using the error backpropagation method. Representative neural networks include convolutional neural networks (CNNs) and recurrent neural networks (RNNs). A CNN is a type of feedforward neural network that is not fully connected (the connections are sparse). Details of the neural network will be described later.

[0042] Ensemble learning is a technique for improving discrimination performance by combining multiple models. Techniques used in ensemble learning include, for example, bagging, boosting, and random forests. Bagging is a technique that trains multiple models using bootstrap samples of the training data and determines the evaluation of new input data by majority vote among the multiple models. Boosting is a technique that weights the training data according to the learning results of bagging and trains the misclassified training data more intensively than the correctly classified training data. Random forest is a technique for generating a group of decision trees (random forest) consisting of multiple decision trees with low correlation when using a decision tree as a model. Details of the random forest will be described later.

[0043] (2-2-1) Decision Tree A decision tree is a model for obtaining a complex discrimination boundary (such as a non-linear discrimination function) by combining a plurality of discriminators. A discriminator is, for example, a rule regarding the magnitude relationship between the value of a certain feature axis and a threshold value. As a method for constructing a decision tree from learning data, for example, there is a divide-and-conquer method that repeatedly obtains a rule (discriminator) for dividing the feature space into two parts. FIG. 3 is an example of a decision tree constructed by the divide-and-conquer method. FIG. 4 shows the feature space divided by the decision tree of FIG. 3. In FIG. 4, the learning data is indicated by white circles or black circles, and by the decision tree shown in FIG. 3, each learning data is classified into the class of white circles or the class of black circles. FIG. 3 shows nodes numbered from 1 to 11 and links connecting the nodes with labels of Yes or No. In FIG. 3, the terminal nodes (leaf nodes) are indicated by squares, and the non-terminal nodes (root node and internal nodes) are indicated by circles. The terminal nodes are the nodes numbered from 6 to 11, and the non-terminal nodes are the nodes numbered from 1 to 5. Each terminal node shows a white circle or a black circle representing the learning data. Each non-terminal node is attached with a discriminator. The discriminator is a rule for judging the magnitude relationship between the values of the feature axes x1 and x2 and the thresholds a to e. The label attached to the link indicates the judgment result of the discriminator. In FIG. 4, the discriminator is indicated by a dotted line, and the region divided by the discriminator is attached with the number of the corresponding node.

[0044] In the process of constructing an appropriate decision tree by the divide-and-conquer method, it is necessary to consider the following three points (a) to (c). (a) Selection of the feature axis and threshold value for constructing the discriminator. (b) Decision of the terminal nodes. For example, the number of classes to which the learning data included in one terminal node belongs. Or the selection of how far to perform pruning of the decision tree (obtaining the same subtree for the root node). (c) Assignment of classes by majority vote for the terminal nodes.

[0045] For the decision tree learning method, for example, CART, ID3, and C4.5 are used. As shown in FIGS. 3 and 4, CART is a method of generating a binary tree as a decision tree by dividing the feature space into two for each feature axis at each node other than the terminal node.

[0046] In learning using a decision tree, in order to improve the discrimination performance of learning data, it is important to divide the feature space at an optimal division candidate point at a non-terminal node. As a parameter for evaluating the division candidate point of the feature space, an evaluation function called impurity may be used. As a function I(t) representing the impurity of node t, for example, the parameters represented by the following formulas (1-1) to (1-3) are used. K is the number of classes. (a) Error rate at node t

Number

Number

Number

[0047] (2-2-2) Support Vector Machine A support vector machine (SVM) is an algorithm for finding a two-class linear discrimination function that realizes the maximum margin. Figure 5 is a diagram for explaining the SVM. The two-class linear discrimination function represents discrimination hyperplanes P1 and P2, which are hyperplanes for linearly separating the learning data of two classes C1 and C2 in the feature space shown in Figure 5. In Figure 5, the learning data of class C1 is shown as circles, and the learning data of class C2 is shown as squares. The margin of the discrimination hyperplane is the distance between the learning data closest to the discrimination hyperplane and the discrimination hyperplane. In Figure 5, the margin d1 of the discrimination hyperplane P1 and the margin d2 of the discrimination hyperplane P2 are shown. In SVM, an optimal discrimination hyperplane P1, which is a discrimination hyperplane with the maximum margin, is obtained. The minimum value d1 of the distance between the learning data of one class C1 and the optimal discrimination hyperplane P1 is equal to the minimum value d2 of the distance between the learning data of the other class C2 and the optimal discrimination hyperplane P2.

[0048] In Figure 5, the learning data set D L used for supervised learning of a two-class problem is represented by the following equation (2-1).

Equation

[0049] In Figure 5, the normalized linear discrimination function that holds for all learning data x i is represented by the following two equations (2-2) and (2-3). w is a coefficient vector, and b is a bias. [Mathematics] These two equations are represented by the following single equation (2-4). [Mathematics]

[0050] When the discrimination hyperplanes P1 and P2 are represented by the following equation (2-5), the margin d is represented by equation (2-6). [Mathematics] In equation (2-6), ρ(w) represents the minimum value of the difference in the lengths obtained by projecting the learning data x of each of the classes C1 and C2 onto the normal vector w of the discrimination hyperplanes P1 and P2. The terms "min" and "max" in equation (2-6) are the points indicated by the signs "min" and "max" in Fig. 5, respectively. In Fig. 5, the optimal discrimination hyperplane is the discrimination hyperplane P1 at which the margin d is maximized. i Fig. 5 represents a feature space in which the learning data of two classes are linearly separable. Fig. 6 represents a feature space similar to Fig. 5, in which the learning data of two classes are linearly inseparable. When the learning data of two classes are linearly inseparable, the following equation (2-7) obtained by introducing the slack variable ξ

[0051] i can be used. [Mathematics] The slack variable ξ i is used only during learning and takes a value of 0 or greater. Fig. 6 shows the discrimination hyperplane P3, the margin boundaries B1 and B2, and the margin d3. The equation of the discrimination hyperplane P3 is the same as equation (2-5). The margin boundaries B1 and B2 are hyperplanes at a distance of the margin d3 from the discrimination hyperplane P3.

[0052] The slack variable ξ i ​When it is 0, Equation (2-7) is equivalent to Equation (2-4). At this time, as shown by the white circles or squares in Fig. 6, the training data x that satisfies Equation (2-7) i is correctly identified within the margin d3. At this time, the training data x i the distance between and the discrimination hyperplane P3 is greater than or equal to the margin d3.

[0053] Slack variable ξ i When it is greater than 0 and less than or equal to 1, as shown by the hatched circles or squares in Fig. 6, the training data x that satisfies Equation (2-7) i exceeds the margin boundaries B1, B2 but does not exceed the discrimination hyperplane P3 and is correctly identified. At this time, the training data x i the distance between and the discrimination hyperplane P3 is less than the margin d3.

[0054] Slack variable ξ i When it is greater than 1, as shown by the black circles or squares in Fig. 6, the training data x that satisfies Equation (2-7) i exceeds the discrimination hyperplane P3 and is misrecognized.

[0055] In this way, by using Equation (2-7) with the slack variable ξ i introduced, even when the two-class training data is not linearly separable, the training data x i can be discriminated.

[0056] From the above description, the sum of the slack variables ξ i for all the training data x i represents the upper limit of the number of misrecognized training data x i Here, the evaluation function L p is defined by the following Equation (2-8).

Equation

[0057] (2-2-3) Neural Network Figure 7 is a schematic diagram of the model of a neuron in a neural network. Figure 8 is a schematic diagram of a three-layer neural network composed by combining the neurons shown in Figure 7. As shown in Figure 7, a neuron outputs an output \(y\) for a plurality of inputs \(x\) (inputs \(x1, x2, x3\) in Figure 7). To each input \(x\) (inputs \(x1, x2, x3\) in Figure 7), a corresponding weight \(w\) (weights \(w1, w2, w3\) in Figure 7) is multiplied. The neuron outputs the output \(y\) using the following Equation (3-1).

Equation

[0058] In the three-layer neural network shown in Figure 8, a plurality of input vectors \(x\) (input vectors \(x1, x2, x3\) in Figure 8) are input from the input side (the left side of Figure 8), and a plurality of output vectors \(y\) (output vectors \(y1, y2, y3\) in Figure 8) are output from the output side (the right side of Figure 8). This neural network is composed of three layers \(L1, L2, L3\).

[0059] In the first layer \(L1\), the input vectors \(x1, x2, x3\) are input by being multiplied by corresponding weights to each of the three neurons \(N11, N12, N13\). In Figure 8, these weights are collectively denoted as \(W1\). The neurons \(N11, N12, N13\) each output feature vectors \(z11, z12, z13\).

[0060] In the second layer L2, the feature vectors z11, z12, z13 are input to the respective two neurons N21, N22 after being multiplied by corresponding weights. In FIG. 8, these weights are collectively denoted as W2. The neurons N21, N22 output the feature vectors z21, z22 respectively.

[0061] In the third layer L3, the feature vectors z21, z22 are input to the respective three neurons N31, N32, N33 after being multiplied by corresponding weights. In FIG. 8, these weights are collectively denoted as W3. The neurons N31, N32, N33 output the output vectors y1, y2, y3 respectively.

[0062] The operation of the neural network has a learning mode and a prediction mode. In the learning mode, the weights W1, W2, W3 are learned using a learning dataset. In the prediction mode, predictions such as identification are made using the parameters of the learned weights W1, W2, W3.

[0063] The weights W1, W2, W3 can be learned, for example, by the error backpropagation method. In this case, information regarding the error is transmitted from the output side towards the input side, that is, from the right side towards the left side in FIG. 8. The error backpropagation method is a technique for learning by adjusting the weights W1, W2, W3 so as to reduce the difference between the output y when the input x is input and the true output y (teacher data) in each neuron. Examples of techniques for optimizing the weights include general techniques such as stochastic gradient descent, RMSprop, and Adamax.

[0064] The neural network can be configured to have more than three layers. The machine learning technique using a neural network with four or more layers is known as deep learning.

[0065] (2-2-4) Random Forest Random forest is a type of ensemble learning, which is a technique for enhancing discrimination performance by combining multiple decision trees. In learning using a random forest, a group of multiple decision trees with low correlation (random forest) is generated. The following algorithms are used for the generation and discrimination of the random forest. (A) Repeat the following from m = 1 to M. (a) Generate m bootstrap samples Z m from N d-dimensional learning data. (b) Using Z m as learning data, divide each node t in the following procedure to generate m decision trees. (i) Randomly select d' features from d features. (d' < d) (ii) From the selected d' features, find the feature and split point (threshold) that give the optimal split of the learning data. (iii) Split node t into two at the obtained split point. (B) Output a random forest consisting of m decision trees. (C) For the input data, obtain the discrimination results of each decision tree in the random forest. The discrimination result of the random forest is determined by a majority vote of the discrimination results of each decision tree.

[0066] In learning using a random forest, the correlation between decision trees can be reduced by randomly selecting a predetermined number of features used for discrimination at each non-terminal node of the decision tree.

[0067] (2-3) Memory unit 14 The storage unit 14 shown in FIG. 1 is an example of a recording medium, and is configured by, for example, a flash memory, a RAM, an HDD, or the like. In the storage unit 14, a learning model generation program 15 executed in the control unit 11 is stored in advance. A database 16 is constructed in the storage unit 14, and a plurality of teacher data acquired by the acquisition unit 12 are stored and appropriately managed respectively. The database 16 stores a plurality of teacher data, as shown in FIG. 9, for example. Note that FIG. 9 shows a part of the teacher data stored in the database 16. In addition to the teacher data, the storage unit 14 may store information for generating a learning model, such as a learning data set and inspection data.

[0068] (3) Teacher data It has been found that there is a correlation among glass cloth information, dispersion liquid information, and evaluation. Therefore, the teacher data acquired for generating a learning model includes at least glass cloth information, dispersion liquid information, and evaluation information as shown below. Further, it is preferable to include processing condition information from the viewpoint of improving the accuracy of the output value. Of course, the teacher data may include information other than the information shown below. It is assumed that the database 16 of the storage unit 14 in the present disclosure stores a plurality of teacher data including the information shown below.

[0069] (3-1) Glass cloth information The glass cloth information is information on the glass cloth that is the object to be impregnated with the fluoropolymer aqueous dispersion. The above glass cloth is a woven or knitted fabric of glass fibers, which is made by weaving or knitting glass fibers. The above glass cloth may be a single woven or knitted fabric, or may be a laminate of a plurality of glass cloths. Examples of the above glass cloth information include information on the weaving / knitting method, thickness, count, density, twist, filament diameter, weight, type, laminate structure, tensile strength, etc. of the above glass cloth.

[0070] Examples of the weaving method (way of weaving) when the glass cloth is a fabric include plain weave, twill weave, nanako weave, damask weave, twill weave, intertwined weave, triaxial weave, horizontal stripe weave, etc. Examples of the knitting method (way of knitting) when it is a knitted fabric include horizontal knitting such as plain knitting, rubber knitting, and pearl knitting, vertical knitting such as single denier knitting, single cord knitting, and double knitting, lace knitting, float knitting, pile knitting, etc. The above glass cloth is preferably a fabric of glass fibers, and the information on the above weaving and knitting methods is preferably information on the way of weaving.

[0071] The above count is the count of the warp and weft that make up the above glass cloth, and is represented by, for example, the mass per 1000 m.

[0072] The above density is the density of the warp and weft that make up the above glass cloth, and is represented by, for example, the number per unit length.

[0073] The above twist is the presence or absence of twist and the way of twisting of the warp and weft that make up the above glass cloth. Examples of the above way of twisting include S twist, Z twist, etc. Also, it can be expressed by the number of single fibers to be twisted and the number of twisted yarns to be bundled. For example, when 3 twisted yarns obtained by twisting 4 single fibers are bundled, the way of twisting is expressed as 4 / 3.

[0074] The above filament diameter is the average diameter of the single fiber that makes up the above glass cloth.

[0075] The above type is the type (composition) of the glass fiber that makes up the above glass cloth, and examples include E glass, D glass, S glass, C glass, H glass, R glass, T glass, NE glass, L glass, etc.

[0076] The above laminated structure is the presence or absence of lamination of the above glass cloth and the number of layers when laminated, etc.

[0077] The above tensile strength is the tensile strength in the warp direction, weft direction, etc. of the above glass cloth.

[0078] The above glass cloth information preferably includes information regarding at least one selected from the group consisting of a weaving and knitting method, thickness, count, density, twist, filament diameter, weight, type, laminate structure, and tensile strength, and more preferably includes information regarding a weaving and knitting method, thickness, count, density, twist, filament diameter, weight, type, laminate structure, and tensile strength. Since these pieces of information have a strong correlation particularly with the evaluation of the impregnated body, by using these pieces of information, a more accurate output can be obtained. Note that the teacher data in FIG. 9 includes the above items which are glass cloth information, but some illustrations are omitted.

[0079] (3-2) Dispersion liquid information The dispersion liquid information is information regarding the fluoropolymer aqueous dispersion liquid to be impregnated into the glass cloth. The above dispersion liquid information can include, for example, information regarding the fluoropolymer contained in the above dispersion liquid.

[0080] As the above fluoropolymer, a fluororesin is preferable, and examples thereof include polytetrafluoroethylene [PTFE], tetrafluoroethylene [TFE] / perfluoro(alkyl vinyl ether) [PAVE] copolymer [PFA], TFE / hexafluoropropylene [HFP] copolymer [FEP], ethylene [Et] / TFE copolymer [ETFE], Et / TFE / HFP copolymer [EFEP], polychlorotrifluoroethylene [PCTFE], chlorotrifluoroethylene [CTFE] / TFE copolymer, CTFE / TFE / PAVE copolymer, Et / CTFE copolymer, polyvinyl fluoride [PVF], polyvinylidene fluoride [PVdF], vinylidene fluoride [VdF] / TFE copolymer, VdF / HFP copolymer, VdF / TFE / HFP copolymer, VdF / HFP / (meth)acrylic acid copolymer, VdF / CTFE copolymer, VdF / pentafluoropropylene copolymer, VdF / PAVE / TFE copolymer, and the like. As the above fluoropolymer, PTFE is preferred. Information on the above fluoropolymer is preferably information on PTFE.

[0081] The above PTFE may be a TFE homopolymer consisting only of tetrafluoroethylene (TFE) or a modified PTFE. The modified PTFE contains TFE units and modified monomer units based on a modified monomer copolymerizable with TFE. The above modified monomer units mean a part of the molecular structure of PTFE derived from the modified monomer.

[0082] The above modified monomer is not particularly limited as long as it can copolymerize with TFE. For example, perfluoroolefins such as hexafluoropropylene [HFP]; perhaloolefins such as chlorotrifluoroethylene [CTFE]; hydrogen-containing fluoroolefins such as trifluoroethylene and vinylidene fluoride [VDF]; perfluorovinyl ether; perfluoroallyl ether; (perfluoroalkyl)ethylene; ethylene; fluorine-containing vinyl ether having a nitrile group, etc. may be mentioned. Also, the modified monomer used may be one kind or a plurality of kinds.

[0083] The above perfluorovinyl ether is not particularly limited. For example, the following general formula (1) CF2=CF-ORf 1 (1) (In the formula, Rf 1 represents a perfluoro organic group.) Perfluoro unsaturated compounds represented by the like may be mentioned. In this specification, the above "perfluoro organic group" means an organic group in which all hydrogen atoms bonded to carbon atoms are substituted by fluorine atoms. The above perfluoro organic group may have an ether oxygen.

[0084] As the above perfluorovinyl ether, for example, in the above general formula (1), Rf 1Examples of the perfluoro(alkyl vinyl ether) [PAVE] in which the perfluoroalkyl group has 1 to 10 carbon atoms include those in which the number of carbon atoms of the perfluoroalkyl group is preferably 1 to 5.

[0085] Examples of the perfluoroalkyl group in the above PAVE include a perfluoromethyl group, a perfluoroethyl group, a perfluoropropyl group, a perfluorobutyl group, a perfluoropentyl group, a perfluorohexyl group, and the like. Examples of the above PAVE include perfluoro(methyl vinyl ether) [PMVE], perfluoro(ethyl vinyl ether) [PEVE], perfluoro(propyl vinyl ether) [PPVE], and the like.

[0086] Examples of the above perfluorovinyl ether further include those in which, in the general formula (1), Rf 1 is a perfluoro(alkoxyalkyl) group having 4 to 9 carbon atoms, Rf 1 is a group represented by the following formula:

Chemical formula

Chemical formula

[0087] Examples of the above perfluoroallyl ether include the following general formula (1-A): CF2=CF-CF2-ORf 1 (1-A) (wherein Rf 1 represents a perfluoro organic group). Examples thereof include perfluoro unsaturated compounds represented by the formula.

[0088] The perfluoroalkyl ethylene is not particularly limited, and examples thereof include perfluorobutyl ethylene (PFBE), perfluorohexyl ethylene (PFHE), and the like. Examples of the fluorine-containing vinyl ether having a nitrile group include CF2=CFORf 2 CN (wherein Rf 2 represents an alkylene group having 2 to 7 carbon atoms in which an oxygen atom may be inserted between two carbon atoms). Examples thereof include fluorine-containing vinyl ethers represented by the formula.

[0089] In the above modified PTFE, the content of the polymerization unit based on the modified monomer (modified monomer unit) is usually in the range of 0.001 to 2.0% by mass, preferably 0.001 to 1.0% by mass, more preferably 0.001 to 0.5% by mass.

[0090] The information on the above PTFE can include information on the presence or absence of modification in the above PTFE, the type of modified monomer, the content of the modified monomer unit, and the like.

[0091] The above PTFE may have a core-shell structure. Examples of the PTFE having a core-shell structure include modified PTFE containing a core of high molecular weight PTFE and a shell of lower molecular weight PTFE or modified PTFE in the particles. Examples of such modified PTFE include PTFE described in JP-T-2005-527652.

[0092] The above core-shell structure can take the following structures. Core: TFE homopolymer Shell: TFE homopolymer Core: Modified PTFE Shell: TFE homopolymer Core: Modified PTFE Shell: Modified PTFE Core: TFE homopolymer Shell: Modified PTFE Core: Low molecular weight PTFE Shell: High molecular weight PTFE Core: High molecular weight PTFE Shell: Low molecular weight PTFE Note that the "low molecular weight PTFE" and "high molecular weight PTFE" in the above core-shell structure are classifications based on the relative molecular weights of the core and the shell, and correspond to the PTFE constituting the particle core part and the particle shell part described in, for example, International Publication No. 2006 / 054612.

[0093] The above core or the above shell can also have a structure of two or more layers. For example, it may be PTFE having a three-layer structure including a core central part of modified PTFE, an outer core part of a TFE homopolymer, and a shell of modified PTFE. Examples of such a fluoropolymer having a three-layer structure include PTFE described in International Publication No. 2006 / 054612.

[0094] The information on the above PTFE can also include information regarding the presence or absence of a core-shell structure in the above PTFE, the type of core-shell structure, the composition and ratio of the core and the shell, etc.

[0095] The information on the above PTFE can also include information regarding the average primary particle diameter, particle size distribution, particle shape, etc. of the above PTFE. The above average primary particle diameter can be indirectly determined from the above transmittance based on, for example, a calibration curve between the transmittance of transmitted light of 550 nm per unit length of a PTFE aqueous dispersion adjusted to a solid content concentration of 0.02 mass% and the average primary particle diameter determined from a transmission electron micrograph. The measurement of the above transmittance can be performed using, for example, a dynamic light scattering measurement device. The above average primary particle diameter can also be measured by the dynamic light scattering method. Although the specific method is not particularly limited, for example, an aqueous dispersion of a fluoropolymer adjusted to a fluoropolymer solid content concentration of 1.0 mass% is prepared, and using ELSZ-1000S (manufactured by Otsuka Electronics Co., Ltd.), with the refractive index of the solvent (water) being 1.3328 and the viscosity of the solvent (water) being 0.8878 mPa·s, the measurement can be performed at 25°C and integrated 70 times.

[0096] The above particle size distribution can be measured, for example, by observing a PTFE aqueous dispersion diluted so that the solid content concentration is about 1% by mass with a scanning electron microscope (SEM) and performing image processing on 100 or more randomly extracted particles.

[0097] The above particle shape can be confirmed, for example, using a scanning electron microscope. Alternatively, the aspect ratio of the above PTFE particles may be used as information on the above particle shape. The above aspect ratio can be calculated, for example, by measuring the major axis and minor axis from a photograph of the particles taken using a scanning electron microscope.

[0098] The information on the above PTFE can also include information on the melting point, standard specific gravity (SSG), extrusion pressure, etc. of the above PTFE.

[0099] The above melting point is the temperature corresponding to the maximum value in the melting heat curve when the temperature is raised at a rate of 10°C / min using a differential scanning calorimeter [DSC]. The above melting point may be the first melting point of PTFE. The above first melting point is the temperature corresponding to the maximum value in the melting heat curve when a differential scanning calorimeter [DSC] is used to raise the temperature of PTFE without a heating history at a temperature of 300°C or higher at a rate of 10°C / min. The first melting point of the above PTFE is, for example, 333 to 347°C.

[0100] The above SSG can be measured, for example, using a sample molded in accordance with ASTM D 4895-89 and by the water displacement method in accordance with ASTM D-792.

[0101] The above extrusion pressure can be measured, for example, by the following method at a predetermined reduction ratio (compression ratio). A predetermined amount of hydrocarbon oil Isopar-G (manufactured by Exxon Mobil), which is an extrusion aid, is added to the PTFE powder obtained from the above dispersion, uniformly mixed in a sealed container, and aged at room temperature (25 ± 2°C) for 1 hour. Next, the above mixture is filled into the cylinder of an extruder (equipped with a die of a predetermined reduction ratio) conforming to ASTM D 4895, held at room temperature for 1 minute, and then immediately a predetermined load is applied to the piston inserted into the cylinder, and it is immediately extruded from the orifice at a predetermined ram speed at room temperature. The value obtained by dividing the load (N) at the point when the pressure reaches an equilibrium state during the extrusion operation by the cross-sectional area of the cylinder is defined as the extrusion pressure (MPa).

[0102] Note that the PTFE in the present disclosure may have non-melt secondary processability. The above non-melt secondary processability means the property that the melt flow rate cannot be measured at a temperature higher than the crystallization melting point in accordance with ASTM D-1238 and D-2116.

[0103] The information about the above PTFE is preferably includes information on at least one selected from the group consisting of the presence or absence of modification in the above PTFE, the type of modified monomer, the content of the modified monomer unit, the presence or absence of a core-shell structure, the type of core-shell structure, the composition and ratio of the core and shell, the melting point, the standard specific gravity (SSG), and the extrusion pressure. more preferably includes information on the presence or absence of modification in the above PTFE, the type of modified monomer, the content of the modified monomer unit, the presence or absence of a core-shell structure, the type of core-shell structure, the composition and ratio of the core and shell, the melting point, the standard specific gravity (SSG), and the extrusion pressure. Since these information have a strong correlation with the evaluation of the impregnated body, more accurate output can be obtained by using these information.

[0104] The above fluoropolymer may be a fluororesin that can be melt-processed. In this specification, being melt-processable means that it is possible to melt and process the polymer using conventional processing equipment such as an extruder and an injection molding machine. The above melt-processable fluororesin usually has a melt flow rate (MFR) of 0.1 to 100 g / 10 min. The MFR is obtained, according to ASTM D1238, using a melt indexer at a measurement temperature determined according to the type of fluoropolymer (for example, 372 °C for PFA and FEP, 297 °C for ETFE), a load (for example, 5 kg for PFA, FEP, and ETFE), and is the mass (g / 10 min) of the polymer flowing out per 10 minutes from a nozzle with an inner diameter of 2 mm and a length of 8 mm.

[0105] Examples of the above melt-processable fluororesin include the above-mentioned PFA, FEP, ETFE, EFEP, PCTFE, PVdF, etc.

[0106] When the above fluoropolymer is a melt-processable fluororesin, the information on the above fluoropolymer can include information on the monomer composition, melting point, MFR, average primary particle diameter, particle size distribution, particle shape, etc. of the above fluoropolymer.

[0107] The above dispersion liquid information can also include information on components other than the above fluoropolymer contained in the above dispersion liquid.

[0108] The above aqueous dispersion of fluoropolymer contains at least an aqueous medium in addition to the above fluoropolymer. The above aqueous medium may be water alone or a mixed solvent of water and an organic solvent. The information on components other than the above fluoropolymer can include information on the type and content of the above aqueous medium.

[0109] The above aqueous dispersion of fluoropolymer may further contain a surfactant. Examples of the above surfactant include nonionic surfactants, cationic surfactants, anionic surfactants, amphoteric surfactants, etc. The above surfactant may be a fluorine-based surfactant or a non-fluorine-based surfactant. Examples of the fluorosurfactant include fluorine-containing carboxylic acid surfactants, fluorine-containing sulfonic acid surfactants, and the like. The use of the fluorosurfactant affects the wettability of the dispersion liquid. The information on components other than the above PTFE may also include the information on the type and content of the surfactant.

[0110] The above fluoropolymer aqueous dispersion may further contain various additives. Examples of the additives include viscosity modifiers, defoamers, preservatives, pH adjusters, film-forming improvers, pigments, fillers, desiccants, leveling agents, anti-splashing agents, stabilizers, and the like.

[0111] Examples of the viscosity modifier include anionic surfactants, methyl cellulose, alumina sol, polyvinyl alcohol, carboxylated vinyl polymers, and the like. Examples of the anionic surfactant include alkyl sulfonates, alkyl sulfates, alkyl aryl sulfates and their salts; aliphatic (carboxylic) acids and their salts; alkyl phosphate esters, alkyl aryl phosphate esters or their salts; and the like. Examples of the defoamer include nonpolar solvents such as toluene, xylene, and hydrocarbon systems having 9 to 11 carbon atoms, silicone oil, and the like. Examples of the preservative include isothiazolone-based, azole-based, pronopol, chlorothalonil, methylsulfonyltetrachloropyridine, carbendazim, fluoropholbet, disodium diacetate, diiodomethylparatolylsulfone, and the like. Examples of the pH adjuster include ammonia, aqueous ammonia, sodium hydroxide, potassium hydroxide, lithium hydroxide, ammonium carbonate, ammonium citrate, and the like. Examples of the stabilizer include water-soluble polymer compounds. For example, polyethylene oxide (dispersion stabilizer), polyethylene glycol (dispersion stabilizer), polyvinyl pyrrolidone (dispersion stabilizer), phenolic resin, urea resin, epoxy resin, melamine resin, polyester resin, polyether resin, acrylic silicone resin, silicone resin, silicone polyester resin, polyurethane resin, etc. Examples of the film-forming improver include acrylic resin, silicone surfactant, nonionic surfactant with a decomposition residue of 5% or more at 380 °C, drying retardant, fluorosurfactant, etc. Examples of the drying retardant include solvents having a boiling point of about 200 to 300 °C, especially water-soluble solvents. The pigment is not particularly limited, and examples include titanium oxide, carbon black, and red iron oxide. And other known pigments. The filler is not particularly limited, and examples include talc, mica, clay, glass flakes, glass beads, etc. Examples of the desiccant include cobalt oxide, etc.

[0112] The information on components other than the fluoropolymer may also include information on the types and contents of the above additives. Among them, it is preferable to include information on the types and contents of the viscosity modifier, defoamer, preservative, pH adjuster, and stabilizer.

[0113] The information on components other than the fluoropolymer Preferably includes at least one selected from the group consisting of information on the type and content of the aqueous medium, the type and content of the surfactant, and the type and content of the additives in the dispersion. More preferably, it includes at least one selected from the group consisting of information on the type and content of the aqueous medium, the type and content of the surfactant, and the types and contents of the viscosity modifier, defoamer, preservative, pH adjuster, and stabilizer in the dispersion. It is more preferable that the information includes the type and content of the aqueous medium, the type and content of the surfactant, and the type and content of the viscosity modifier, defoamer, preservative, pH adjuster, and stabilizer in the above dispersion liquid. Since these pieces of information have a strong correlation with the evaluation of the impregnated body, more accurate output can be obtained by using these pieces of information.

[0114] The above dispersion liquid information may also include information regarding the properties of the above dispersion liquid. The information regarding the properties of the above dispersion liquid may include information regarding the appearance, solid content concentration, specific gravity, viscosity, pH, viscosity-temperature transition, surface tension, etc. of the above dispersion liquid.

[0115] Examples of the information regarding the appearance of the above dispersion liquid include information such as the color tone, uniformity, and presence or absence of precipitation of the above dispersion liquid. The information regarding the appearance can be obtained, for example, by visually observing the above dispersion liquid. The results of visual observation may be digitized.

[0116] The above solid content concentration is, for example, obtained by taking the ratio of the mass of the residue obtained by drying 10 g of the above dispersion liquid at a temperature of 380°C for 45 minutes to the mass of the dispersion liquid before drying as a percentage.

[0117] The above specific gravity is, for example, a value measured in accordance with JIS K6893.

[0118] The above viscosity can be measured, for example, at a temperature of 25°C in accordance with JIS K6893 using a B-type rotational viscometer.

[0119] The above pH is, for example, a value measured in accordance with JIS K6893.

[0120] The above viscosity temperature transition (VTT) represents the viscosity-temperature dependence of the dispersion. The VTT can be obtained, for example, by heating the sample between 20°C and 50°C and measuring the viscosity of the dispersion while checking the viscosity every 1°C. The VTT point is the temperature at which the viscosity reaches the same value again as when measured at 20°C. This means that the viscosity of the sample was measured at 20°C. Then, the sample is heated above 20°C to lower the viscosity. At a specific temperature, the viscosity of the dispersion increases again. The VTT is the temperature at which the viscosity of the sample rises again to the value it had at 20°C.

[0121] The above surface tension can be measured, for example, by a surface tensiometer.

[0122] The information regarding the properties of the above dispersion preferably includes information on at least one selected from the group consisting of the appearance, solid content concentration, specific gravity, viscosity, pH, viscosity temperature transition, and surface tension of the above dispersion, and more preferably includes information on the appearance, solid content concentration, specific gravity, viscosity, pH, viscosity temperature transition, and surface tension of the above dispersion. Since these information have a particularly strong correlation with the evaluation of the impregnated body, higher-precision output can be obtained by using these information.

[0123] The above dispersion information preferably includes at least one type of information selected from the group consisting of information on the above fluoropolymer, information on components other than the above fluoropolymer, and information on the properties of the above dispersion, and more preferably includes information on the above fluoropolymer, information on components other than the above fluoropolymer, and information on the properties of the above dispersion. The preferred items of each of these information are as described above.

[0124] The above dispersion information may of course include information other than the above. Note that the teacher data in FIG. 9 includes each of the above items which are dispersion information, but some illustrations are omitted.

[0125] (3-3) Evaluation The evaluation is information on an impregnated body obtained by impregnating a fluoropolymer aqueous dispersion into a glass cloth. The impregnated body may have a glass cloth and a fluoropolymer film covering the glass cloth. The evaluation can include information such as crack-limiting film thickness, color tone, peelability, flex life, surface roughness, electrical properties, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, surface resistivity, etc.

[0126] The crack-limiting film thickness is the maximum film thickness at which cracks do not occur in the film. However, since it cannot be directly measured during impregnation of the glass cloth, it is measured, for example, by the following method. A fluoropolymer aqueous dispersion (5 ml) is dropped onto an aluminum plate of 20 cm × 10 cm × 1.5 mm, and using a coating applicator (manufactured by Yasuda Seiki Co., Ltd.), it is applied so that the film thickness continuously changes between more than 0 μm and 200 μm or less. After drying the coating film at 100 °C for 10 minutes and then baking it at 380 °C for 15 minutes, the maximum film thickness at which cracks do not occur is measured and taken as the crack-limiting film thickness.

[0127] The color tone can be examined by visually observing the fluoropolymer film in the impregnated body. The visual results may be quantified.

[0128] The peelability is indicated, for example, by the peel strength of the film measured in accordance with ASTM D4851 (Adhesion Test of Coatings to Fabrics).

[0129] Examples of the flex life include the MIT flex life. The MIT flex life is indicated by, for example, the number of double folds required to break a sample of the impregnated body, which is determined using a standard MIT folding durability tester described in ASTM D2176. The MIT flex life is preferably determined for two directions, longitudinal and transverse, of the impregnated body. Also, the test may be carried out multiple times and the results averaged.

[0130] The surface roughness can be measured, for example, using a surface roughness measuring instrument in accordance with JIS B 0601.

[0131] Examples of the electrical properties include relative permittivity, dielectric loss tangent, etc. The relative permittivity and dielectric loss tangent can be measured, for example, by the SPDR (split post dielectric resonator) method at a predetermined temperature, humidity, and frequency. The electrical properties may be measured for a laminate of a plurality of the impregnated bodies.

[0132] The weather resistance is indicated, for example, by the gloss retention rate and tensile strength retention rate after performing a weather resistance test for a predetermined time using an accelerated weather resistance test device.

[0133] The tensile strength can be measured, for example, in accordance with JIS L 1096 (cut strip method).

[0134] The tear strength can be measured, for example, in accordance with JIS L 1096 (trapezoid method).

[0135] The breakdown voltage can be measured, for example, in accordance with JIS C 2110-1.

[0136] The volume resistivity and surface resistivity can be measured, for example, in accordance with JIS K 6911.

[0137] The items for evaluation may be selected according to the intended use of the impregnated body. Examples of the uses include roofing materials (tent membranes) for membrane structure buildings, conveyor belts, high-frequency printed circuit boards, packings, etc. Among them, roofing materials and high-frequency printed circuit boards are preferred.

[0138] The above evaluation preferably includes information regarding at least one selected from the group consisting of the crack-limiting film thickness, color tone, peelability, flexural life, surface roughness, electrical properties, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body, more preferably includes information regarding at least one selected from the group consisting of the crack-limiting film thickness, color tone, peelability, flexural life, surface roughness, and electrical properties of the impregnated body, still more preferably includes information regarding the crack-limiting film thickness, color tone, peelability, flexural life, surface roughness, and electrical properties of the impregnated body, and particularly preferably includes information regarding the crack-limiting film thickness, color tone, peelability, flexural life, surface roughness, electrical properties, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body. The above evaluation also preferably includes information regarding the crack-limiting film thickness, color tone, peelability, flexural life, and surface roughness of the impregnated body.

[0139] Note that the teacher data in FIG. 9 includes each of the above items for evaluation, but some illustrations are omitted.

[0140] (3-4) Processing condition information The processing condition information is information on the processing conditions for obtaining an impregnated body by impregnating a glass cloth with an aqueous dispersion of a fluoropolymer. By impregnating the glass cloth with the above dispersion and further performing heating (such as drying, firing, etc.), the above impregnated body can be obtained. The impregnation can be performed by a method such as dipping the glass cloth in the above dispersion. The above processing condition information can include information such as the solid content concentration during processing, surface tension during processing, additives during processing, number of impregnations and impregnation amount, and heating conditions of the above dispersion.

[0141] The solid content concentration during processing is the solid content concentration of the above dispersion during processing and can affect handleability, etc. The solid content concentration during processing is, for example, obtained by determining the ratio of the mass of the residue obtained by drying 10 g of the above dispersion at a temperature of 380°C for 45 minutes to the mass of the dispersion before drying as a percentage.

[0142] The surface tension during processing is the surface tension of the dispersion liquid during processing and can be measured, for example, using a surface tensiometer.

[0143] The additive during processing is a component added during the processing of the dispersion liquid, and examples thereof include surfactants. The surfactant can affect the ease of overcoating the dispersion liquid during processing. The information on the additive during processing can include information on the type and content of the additive during processing.

[0144] Examples of the heating conditions include drying conditions and firing conditions.

[0145] Examples of the drying conditions include the drying temperature, drying time, heating rate, ventilation conditions, etc. in the drying process described above. Examples of the ventilation conditions include the air flow rate inside the heating furnace. When cooling once after drying, it is preferable to also consider the cooling rate.

[0146] Examples of the firing conditions include the firing temperature, firing time, heating rate, cooling rate, ventilation conditions, etc. in the firing process described above. Examples of the ventilation conditions include the air flow rate inside the heating furnace.

[0147] The processing condition information preferably includes information on at least one selected from the group consisting of the solid content concentration during processing, the surface tension during processing, the additive during processing, the number of impregnations and the impregnation amount, and the heating conditions of the dispersion liquid. More preferably, it includes information on at least one selected from the group consisting of the solid content concentration during processing, the surface tension during processing, the additive during processing, the number of impregnations and the impregnation amount, the drying conditions, and the firing conditions of the dispersion liquid. Even more preferably, it includes information on the impregnation amount, the drying conditions, and the firing conditions of the dispersion liquid. The processing condition information also preferably includes information regarding at least one selected from the group consisting of the solid content concentration during processing of the dispersion liquid, the surface tension during processing, additives during processing, the number of impregnation times and the impregnation amount, the drying temperature, the drying time, the heating rate, and the ventilation conditions in the drying process, and the firing temperature, the firing time, the heating rate, the cooling rate, and the ventilation conditions in the firing process. More preferably, it includes information regarding the solid content concentration during processing of the dispersion liquid, the surface tension during processing, additives during processing, the number of impregnation times and the impregnation amount, the drying temperature, the drying time, the heating rate, and the ventilation conditions in the drying process, and the firing temperature, the firing time, the heating rate, the cooling rate, and the ventilation conditions in the firing process. The processing condition information also preferably includes information regarding at least one selected from the group consisting of the solid content concentration during processing of the dispersion liquid, the surface tension during processing, additives during processing, the number of impregnation times and the impregnation amount, the drying temperature, the drying time, the firing temperature, the firing time, the heating rate, the cooling rate, and the ventilation conditions. Since these pieces of information have a particularly strong correlation with the evaluation of the impregnated body, higher-precision output can be obtained by using these pieces of information.

[0148] Note that the teacher data in FIG. 9 includes each of the above items which are processing condition information, but some illustrations are omitted.

[0149] (4) Operation of the learning model generation device 10 The following describes the outline of the operation of the learning model generation device 10 with reference to FIG. 10. First, in step S11, the learning model generation device 10 activates the learning model generation program 15 stored in the storage unit 14. Thereby, the learning model generation device 10 operates based on the learning model generation program 15 and starts generating a learning model.

[0150] In step S12, based on the learning model generation program 15, the acquisition unit 12 acquires a plurality of teacher data.

[0151] In step S13, the acquisition unit 12 stores a plurality of teacher data in the database 16 constructed in the storage unit 14. The storage unit 14 stores and appropriately manages a plurality of teacher data.

[0152] In step S14, the learning unit 13 extracts a learning data set from the teacher data stored in the storage unit 14. The extracted A data set is determined according to the learning purpose of the learning model generated by the learning model generation device 10. The data set is based on the teacher data.

[0153] In step S15, the learning unit 13 performs learning based on the extracted plurality of data sets.

[0154] In step S16, based on the result of learning by the learning unit 13 in step S15, a learning model corresponding to the learning purpose is generated.

[0155] The operation of the learning model generation device 10 ends here. Note that the order of the operation of the learning model generation device 10 and the like can be changed as appropriate. The generated learning model is implemented on a general-purpose computer or terminal, or downloaded as software or an application, or distributed in a state stored in a storage medium, etc., and is used.

[0156] (5) Configuration of the user device 20 FIG. 2 shows the configuration of the user device 20 used by the user in this embodiment. Here, the user is a person who inputs some information to the user device 20 or outputs some information. The user device 20 uses the learning model generated by the learning model generation device 10.

[0157] The user device 20 is a device having computer functions. The user device 20 may include a communication interface such as a NIC and a DMA controller, and may be capable of communicating with the learning model generation device 10 etc. via a network. Although the user device 20 shown in FIG. 2 is illustrated as one device, it is preferable that the user device 20 supports cloud computing. For this reason, the hardware configuration of the user device 20 does not need to be housed in one housing or provided as an integrated device. For example, it is configured by dynamically connecting and disconnecting the resources of the user device 20 that are hard according to the load.

[0158] The user device 20 has, for example, an input unit 24, an output unit 25, a control unit 21, and a storage unit 26.

[0159] (5-1) Input unit 24 The input unit 24 is, for example, a keyboard, a touch panel, a mouse, etc. The user can input information to the user device 20 via the input unit 24.

[0160] (5-2) Output unit 25 The output unit 25 is, for example, a display, a printer, etc. The output unit 25 can output the result of analysis using the learning model by the user device 20.

[0161] (5-3) Control unit 21 The control unit 21 is, for example, a CPU, and executes the control of the entire user device 20. The control unit 21 has functional units such as an analysis unit 22 and an update unit 23.

[0162] The analysis unit 22 of the control unit 21 analyzes the input information input via the input unit 24 using a learning model as a program stored in advance in the storage unit 26. The analysis performed by the analysis unit 22 is preferably performed using the machine learning method described above, but is not limited thereto. By using the learned learning model in the learning model generation device 10, the analysis unit 22 can output the correct answer even for unknown input information.

[0163] The update unit 23 updates the learning model stored in the storage unit 26 to an optimal state in order to obtain a high-quality learning model. For example, in a neural network, the update unit 23 optimizes the weighting between neurons in each layer.

[0164] (5-4) Storage unit 26 The storage unit 26 is an example of a recording medium and is composed of, for example, a flash memory, RAM, HDD, etc. The storage unit 26 stores in advance a learning model executed in the control unit 21. A plurality of teacher data are stored in the database 27 in the storage unit 26 and are appropriately managed. Note that other information such as a learning dataset may be stored in the storage unit 26. The teacher data stored in the storage unit 26 are information such as the glass cloth information, dispersion liquid information, evaluation, and processing condition information described above.

[0165] (6) Operation of the user device 20 The outline of the operation of the user device 20 will be described below with reference to FIG. 11. Here, the user device 20 is in a state where the learning model generated in the learning model generation device 10 is stored in the storage unit 26.

[0166] First, in step S21, the user device 20 activates the learning model stored in the storage unit 26. The user device 20 operates based on the learning model.

[0167] In step S22, the user who uses the user device 20 inputs input information via the input unit 24. The input information input via the input unit 24 is sent to the control unit 21.

[0168] In step S23, the analysis unit 22 of the control unit 21 receives input information from the input unit 24, performs analysis, and determines the information to be output by the output unit. The information determined by the analysis unit 22 is sent to the output unit 25.

[0169] In step S24, the output unit 25 outputs the result information received from the analysis unit 22.

[0170] In step S25, the update unit 23 updates the learning model to an optimal state based on the input information, the result information, and the like.

[0171] The operation of the user device 20 ends here. Note that the order of operations of the user device 20 can be changed as appropriate.

[0172] (7) Specific example Hereinafter, a specific example using the above-described learning model generation device 10 and user device 20 will be described.

[0173] (7-1) Crack limit film thickness learning model Here, a crack limit film thickness learning model that outputs the crack limit film thickness of the impregnated body will be described.

[0174] (7-1-1) Crack limit film thickness learning model generation device 10 To generate a crack limit film thickness learning model, the crack limit film thickness learning model generation device 10 includes at least information regarding the weaving and knitting method, thickness, count, density, twist, filament diameter, weight, type, laminate structure, and tensile strength of the glass cloth, information regarding the presence or absence of modification in PTFE contained in the PTFE aqueous dispersion, the type of modified monomer, the content of the modified monomer unit, the presence or absence of a core-shell structure, the type of core-shell structure, the composition and ratio of the core and shell, the average primary particle diameter, the particle size distribution, the particle shape, the melting point, the standard specific gravity (SSG), and the extrusion pressure, Information regarding the type and content of the aqueous medium contained in the PTFE aqueous dispersion, the type and content of the surfactant, and the type and content of the additive, Information regarding the appearance, solid content concentration, specific gravity, viscosity, pH, viscosity-temperature transition, and surface tension of the PTFE aqueous dispersion, Information regarding the solid content concentration during processing of the PTFE aqueous dispersion, the surface tension during processing, additives during processing, number of impregnations and impregnation amount, drying temperature, drying time, firing temperature, firing time, heating rate, cooling rate, and ventilation conditions, and, Critical crack film thickness information, A plurality of teacher data including the above must be acquired. Note that the critical crack film thickness learning model generation device 10 may acquire other information.

[0175] The critical crack film thickness learning model generation device 10 performs learning based on the acquired teacher data, Glass cloth information including information regarding the weaving and knitting method, thickness, count, density, twist, filament diameter, weight, type, laminate structure, and tensile strength of the glass cloth, and, Information regarding the presence or absence of modification in the PTFE contained in the PTFE aqueous dispersion, the type of modified monomer, the content of the modified monomer unit, the presence or absence of a core-shell structure, the type of core-shell structure, the composition and ratio of the core and shell, average primary particle diameter, particle size distribution, particle shape, melting point, standard specific gravity (SSG), and extrusion pressure, information regarding the type and content of the aqueous medium contained in the PTFE aqueous dispersion, the type and content of the surfactant, and the type and content of the additive, and, information regarding the appearance, solid content concentration, specific gravity, viscosity, pH, viscosity-temperature transition, and surface tension of the PTFE aqueous dispersion, i.e., dispersion liquid information, Processing condition information including information regarding the solid content concentration during processing of the PTFE aqueous dispersion, the surface tension during processing, additives during processing, number of impregnations and impregnation amount, drying temperature, drying time, firing temperature, firing time, heating rate, cooling rate, and ventilation conditions, Using the above as input and the critical crack film thickness information as output, it is possible to generate a critical crack film thickness learning model.

[0176] (7-1-2) User device 20 using the critical crack film thickness learning model The user device 20 is a device capable of using a crack limit film thickness learning model. The user who uses the user device 20 glass cloth information including information on the weaving and knitting method, thickness, count, density, twist, filament diameter, weight, type, laminated structure, and tensile strength of the glass cloth, and dispersion liquid information including information on the presence or absence of modification in PTFE contained in the PTFE aqueous dispersion, the type of modified monomer, the content of the modified monomer unit, the presence or absence of a core-shell structure, the type of core-shell structure, the composition and ratio of the core and shell, the average primary particle diameter, the particle size distribution, the particle shape, the melting point, the standard specific gravity (SSG), and the extrusion pressure, information on the type and content of the aqueous medium contained in the PTFE aqueous dispersion, the type and content of the surfactant, and the type and content of the additive, and information on the appearance, solid content concentration, specific gravity, viscosity, pH, viscosity-temperature transition, and surface tension of the PTFE aqueous dispersion, and processing condition information including information on the solid content concentration during processing of the PTFE aqueous dispersion, the surface tension during processing, the additives during processing, the number of impregnation times and the impregnation amount, the drying temperature, the drying time, the firing temperature, the firing time, the heating rate, the cooling rate, and the ventilation conditions, and are input into the user device 20. The user device 20 uses the crack limit film thickness learning model to determine crack limit film thickness information. The output unit 25 outputs the determined crack limit film thickness information.

[0177] (7-2) Other evaluation learning models Here, other evaluation learning models that output evaluations of the impregnated body other than the crack limit film thickness (color and taste of the impregnated body, peelability, bending life, surface roughness, electrical characteristics, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, or surface resistivity) will be described.

[0178] (7-2-1) Other evaluation learning model generation device 10 In order to generate other evaluation learning models, the other evaluation learning model generation device 10 includes at least information on the weaving and knitting method, thickness, count, density, twist, filament diameter, weight, type, laminated structure, and tensile strength of the glass cloth, Information regarding the presence or absence of modification in PTFE contained in the PTFE aqueous dispersion, the type of modified monomer, the content of the modified monomer unit, the presence or absence of a core-shell structure, the type of core-shell structure, the composition and ratio of the core and shell, the average primary particle diameter, the particle size distribution, the particle shape, the melting point, the standard specific gravity (SSG), and the extrusion pressure, Information regarding the type and content of the aqueous medium contained in the PTFE aqueous dispersion, the type and content of the surfactant, and the type and content of the additive, Information regarding the appearance, solid content concentration, specific gravity, viscosity, pH, viscosity-temperature transition, and surface tension of the PTFE aqueous dispersion, Information regarding the solid content concentration during processing of the PTFE aqueous dispersion, the surface tension during processing, additives during processing, the number of impregnations and the impregnation amount, the drying temperature, the drying time, the firing temperature, the firing time, the heating rate, the cooling rate, and the ventilation conditions, and Information regarding the color and taste, peelability, flex life, surface roughness, electrical properties, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, or surface resistivity of the impregnated body, A plurality of teacher data including the above must be acquired. Note that the other evaluation learning model generation device 10 may acquire other information.

[0179] The other evaluation learning model generation device 10 performs learning based on the acquired teacher data, Glass cloth information including information regarding the weaving and knitting method, thickness, count, density, twist, filament diameter, weight, type, laminate structure, and tensile strength of the glass cloth, and Information regarding the presence or absence of modification in PTFE contained in the PTFE aqueous dispersion, the type of modified monomer, the content of the modified monomer unit, the presence or absence of a core-shell structure, the type of core-shell structure, the composition and ratio of the core and shell, the average primary particle diameter, the particle size distribution, the particle shape, the melting point, the standard specific gravity (SSG), and the extrusion pressure, information regarding the type and content of the aqueous medium contained in the PTFE aqueous dispersion, the type and content of the surfactant, and the type and content of the additive, and information regarding the appearance, solid content concentration, specific gravity, viscosity, pH, viscosity-temperature transition, and surface tension of the PTFE aqueous dispersion, i.e., dispersion liquid information, Processing condition information including information on the solid content concentration, surface tension during processing, additives during processing, number of impregnations and impregnation amount, drying temperature, drying time, firing temperature, firing time, heating rate, cooling rate, and ventilation conditions of the PTFE aqueous dispersion, and using this as input, it is possible to generate another evaluation learning model that outputs information on the color, peelability, flexural life, surface roughness, electrical properties, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, or surface resistivity of the impregnated body.

[0180] (7-2-2) User device 20 using another evaluation learning model The user device 20 is a device that can utilize another evaluation learning model. The user who uses the user device 20 inputs glass cloth information including information on the weaving and knitting method, thickness, count, density, twist, filament diameter, weight, type, laminate structure, and tensile strength of the glass cloth, and information on the presence or absence of modification in the PTFE contained in the PTFE aqueous dispersion, type of modified monomer, content of modified monomer unit, presence or absence of core-shell structure, type of core-shell structure, composition and ratio of core and shell, average primary particle diameter, particle size distribution, particle shape, melting point, standard specific gravity (SSG), and extrusion pressure, information on the type and content of the aqueous medium contained in the PTFE aqueous dispersion, type and content of surfactant, and type and content of additives, and dispersion information including information on the appearance, solid content concentration, specific gravity, viscosity, pH, viscosity-temperature transition, and surface tension of the PTFE aqueous dispersion, and processing condition information including information on the solid content concentration, surface tension during processing, additives during processing, number of impregnations and impregnation amount, drying temperature, drying time, firing temperature, firing time, heating rate, cooling rate, and ventilation conditions of the PTFE aqueous dispersion, and inputs it into the user device 20. The user device 20 uses another evaluation learning model to determine information on the color, peelability, flexural life, surface roughness, electrical properties, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, or surface resistivity of the impregnated body. The output unit 25 outputs the determined information on the color, peelability, flexural life, surface roughness, or electrical properties of the impregnated body.

[0181] (7-3) Dispersion learning model Here, a dispersion learning model that outputs an optimal PTFE aqueous dispersion will be described.

[0182] (7-3-1) Dispersion learning model generation device 10 In order to generate a dispersion learning model, the dispersion learning model generation device 10 includes at least information regarding the weaving method, thickness, count, density, twist, filament diameter, weight, type, laminate structure, and tensile strength of the glass cloth, information regarding the presence or absence of modification in the PTFE contained in the PTFE aqueous dispersion, the type of modified monomer, the content of the modified monomer unit, the presence or absence of a core-shell structure, the type of core-shell structure, the composition and ratio of the core and shell, the average primary particle diameter, the particle size distribution, the particle shape, the melting point, the standard specific gravity (SSG), and the extrusion pressure, information regarding the type and content of the aqueous medium contained in the PTFE aqueous dispersion, the type and content of the surfactant, and the type and content of the additive, information regarding the appearance, solid content concentration, specific gravity, viscosity, pH, viscosity-temperature transition, and surface tension of the PTFE aqueous dispersion, information regarding the solid content concentration during processing, the surface tension during processing, the additives during processing, the number of impregnations and the impregnation amount, the drying temperature, the drying time, the firing temperature, the firing time, the heating rate, the cooling rate, and the ventilation conditions of the PTFE aqueous dispersion, and information regarding the crack limit film thickness, color tone, peelability, flexural life, surface roughness, electrical properties, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body, and must acquire a plurality of teacher data including the above. Note that the dispersion learning model generation device 10 may acquire other information.

[0183] The dispersion learning model generation device 10 performs learning based on the acquired teacher data, glass cloth information including information regarding the weaving method, thickness, count, density, twist, filament diameter, weight, type, laminate structure, and tensile strength of the glass cloth, and Processing condition information including information on the solid content concentration, surface tension during processing, additives during processing, number of impregnations and impregnation amount, drying temperature, drying time, firing temperature, firing time, heating rate, cooling rate, and ventilation conditions of the PTFE aqueous dispersion, and Evaluation information including information on the crack limit film thickness, color tone, peelability, flex life, surface roughness, electrical properties, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body, and Using these as inputs and outputting the optimal dispersion liquid information for obtaining the target evaluation of the impregnated body, it is possible to generate a dispersion liquid learning model.

[0184] (7-3-2) User device 20 using the dispersion liquid learning model The user device 20 is a device capable of using the dispersion liquid learning model. The user who uses the user device 20 Inputs glass cloth information including information on the weaving and knitting method, thickness, count, density, twist, filament diameter, weight, type, laminated structure, and tensile strength of the glass cloth, and Processing condition information including information on the solid content concentration, surface tension during processing, additives during processing, number of impregnations and impregnation amount, drying temperature, drying time, firing temperature, firing time, heating rate, cooling rate, and ventilation conditions of the PTFE aqueous dispersion, and Evaluation information including information on the crack limit film thickness, color tone, peelability, flex life, surface roughness, electrical properties, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body to the user device 20. The user device 20 uses the dispersion liquid learning model to determine the optimal dispersion liquid information for obtaining the target evaluation of the impregnated body. The output unit 25 outputs the determined dispersion liquid information.

[0185] (7-4) Glass cloth learning model Here, a glass cloth learning model that outputs an optimal glass cloth will be described.

[0186] (7-4-1) Glass cloth learning model generation device 10 In order to generate a glass cloth learning model, the glass cloth learning model generation device 10 includes at least Information regarding the weaving and knitting method, thickness, count, density, twist, filament diameter, weight, type, laminate structure, and tensile strength of the glass cloth, Information regarding the presence or absence of modification in PTFE contained in the PTFE aqueous dispersion, type of modified monomer, content of modified monomer unit, presence or absence of core-shell structure, type of core-shell structure, composition and ratio of core and shell, average primary particle diameter, particle size distribution, particle shape, melting point, standard specific gravity (SSG), and extrusion pressure, Information regarding the type and content of the aqueous medium contained in the PTFE aqueous dispersion, type and content of surfactant, and type and content of additive, Information regarding the appearance, solid content concentration, specific gravity, viscosity, pH, viscosity-temperature transition, and surface tension of the PTFE aqueous dispersion, Information regarding the solid content concentration during processing, surface tension during processing, additives during processing, number of impregnations and impregnation amount, drying temperature, drying time, firing temperature, firing time, heating rate, cooling rate, and ventilation conditions of the PTFE aqueous dispersion, and Information regarding the crack-limiting film thickness, color tone, peelability, flexural life, surface roughness, electrical properties, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body, A plurality of teacher data including the above must be acquired. Note that the glass cloth learning model generation device 10 may acquire other information.

[0187] The glass cloth learning model generation device 10 performs learning based on the acquired teacher data, Dispersion information including information regarding the presence or absence of modification in PTFE contained in the PTFE aqueous dispersion, type of modified monomer, content of modified monomer unit, presence or absence of core-shell structure, type of core-shell structure, composition and ratio of core and shell, average primary particle diameter, particle size distribution, particle shape, melting point, standard specific gravity (SSG), and extrusion pressure, information regarding the type and content of the aqueous medium contained in the PTFE aqueous dispersion, type and content of surfactant, and type and content of additive, and information regarding the appearance, solid content concentration, specific gravity, viscosity, pH, viscosity-temperature transition, and surface tension of the PTFE aqueous dispersion, and Processing condition information including information on the solid content concentration, surface tension during processing, additives during processing, number of impregnations and impregnation amount, drying temperature, drying time, firing temperature, firing time, heating rate, cooling rate, and ventilation conditions of the PTFE aqueous dispersion, and Evaluation information including information on the crack limit film thickness, color tone, peelability, flexural life, surface roughness, electrical properties, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body, and Taking these as inputs and outputting the optimal glass cloth information for obtaining the target evaluation of the impregnated body, it is possible to generate a glass cloth learning model.

[0188] (7-4-2) User device 20 using the glass cloth learning model The user device 20 is a device capable of utilizing the glass cloth learning model. The user who uses the user device 20 Information on the presence or absence of modification in the PTFE contained in the PTFE aqueous dispersion, the type of modified monomer, the content of the modified monomer unit, the presence or absence of a core-shell structure, the type of core-shell structure, the composition and ratio of the core and shell, the average primary particle diameter, particle size distribution, particle shape, melting point, standard specific gravity (SSG), and extrusion pressure, information on the type and content of the aqueous medium contained in the PTFE aqueous dispersion, the type and content of the surfactant, and the type and content of the additive, and Processing condition information including information on the solid content concentration, surface tension during processing, additives during processing, number of impregnations and impregnation amount, drying temperature, drying time, firing temperature, firing time, heating rate, cooling rate, and ventilation conditions of the PTFE aqueous dispersion, and Evaluation information including information on the crack limit film thickness, color tone, peelability, flexural life, surface roughness, electrical properties, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body, and are input into the user device 20. The user device 20 determines the optimal glass cloth information for obtaining the target evaluation of the impregnated body using the glass cloth learning model. The output unit 25 outputs the determined glass cloth information.

[0189] (7-5)Processing condition learning model Here, a processing condition learning model that outputs optimal processing conditions will be described.

[0190] (7-5-1)Processing condition learning model generation device 10 In order to generate a processing condition learning model, the processing condition learning model generation device 10 includes at least information regarding the weaving method, thickness, count, density, twist, filament diameter, weight, type, laminate structure, and tensile strength of the glass cloth, information regarding the presence or absence of modification in the PTFE contained in the PTFE aqueous dispersion, the type of modified monomer, the content of the modified monomer unit, the presence or absence of a core-shell structure, the type of core-shell structure, the composition and ratio of the core and shell, the average primary particle diameter, the particle size distribution, the particle shape, the melting point, the standard specific gravity (SSG), and the extrusion pressure, information regarding the type and content of the aqueous medium contained in the PTFE aqueous dispersion, the type and content of the surfactant, and the type and content of the additive, information regarding the appearance, solid content concentration, specific gravity, viscosity, pH, viscosity-temperature transition, and surface tension of the PTFE aqueous dispersion, information regarding the solid content concentration during processing, the surface tension during processing, the additives during processing, the number of impregnation times and the impregnation amount, the drying temperature, the drying time, the firing temperature, the firing time, the heating rate, the cooling rate, and the ventilation conditions of the PTFE aqueous dispersion, and information regarding the crack limit film thickness, color tone, peelability, bending life, surface roughness, electrical properties, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body, must acquire a plurality of teacher data including the above. Note that the processing condition learning model generation device 10 may acquire other information.

[0191] The processing condition learning model generation device 10 performs learning based on the acquired teacher data, glass cloth information including information regarding the weaving method, thickness, count, density, twist, filament diameter, weight, type, laminate structure, and tensile strength of the glass cloth, and Dispersion information including information on the presence or absence of modification in the PTFE contained in the PTFE aqueous dispersion, the type of modified monomer, the content of the modified monomer unit, the presence or absence of a core-shell structure, the type of core-shell structure, the composition and ratio of the core and shell, the average primary particle diameter, the particle size distribution, the particle shape, the melting point, the standard specific gravity (SSG), and the extrusion pressure, information on the type and content of the aqueous medium contained in the PTFE aqueous dispersion, the type and content of the surfactant, and the type and content of the additive, and information on the appearance, solid content concentration, specific gravity, viscosity, pH, viscosity-temperature transition, and surface tension of the PTFE aqueous dispersion, Evaluation information including information on the crack limit film thickness, color tone, peelability, bending life, surface roughness, electrical properties, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body, It is possible to generate a processing condition learning model that takes the above as input and outputs the optimal processing condition information for obtaining the target evaluation of the impregnated body.

[0192] (7-5-2) User device 20 using the processing condition learning model The user device 20 is a device that can utilize the processing condition learning model. The user who uses the user device 20 Glass cloth information including information on the weaving and knitting method, thickness, count, density, twist, filament diameter, weight, type, laminate structure, and tensile strength of the glass cloth, Dispersion information including information on the presence or absence of modification in the PTFE contained in the PTFE aqueous dispersion, the type of modified monomer, the content of the modified monomer unit, the presence or absence of a core-shell structure, the type of core-shell structure, the composition and ratio of the core and shell, the average primary particle diameter, the particle size distribution, the particle shape, the melting point, the standard specific gravity (SSG), and the extrusion pressure, information on the type and content of the aqueous medium contained in the PTFE aqueous dispersion, the type and content of the surfactant, and the type and content of the additive, and information on the appearance, solid content concentration, specific gravity, viscosity, pH, viscosity-temperature transition, and surface tension of the PTFE aqueous dispersion, Evaluation information including information on the crack limit film thickness, color tone, peelability, bending life, surface roughness, electrical properties, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body, Input it into the user device 20. The user device 20 determines the optimal processing condition information for obtaining the target impregnated body evaluation using the processing condition learning model. The output unit 25 outputs the determined processing condition information.

[0193] (8) Features (8-1) The learning model generation method of the present embodiment is a learning model generation method for generating a learning model that determines, using a computer, the evaluation of an impregnated body obtained by impregnating a fluoropolymer aqueous dispersion into a glass cloth. The learning model generation method includes an acquisition step S12, a learning step S15, and a generation step S16. In the acquisition step S12, the computer acquires teacher data. The teacher data includes glass cloth information, dispersion liquid information, and the evaluation of the impregnated body. The glass cloth information is information about the glass cloth. The dispersion liquid information is information about the fluoropolymer aqueous dispersion. In the learning step S15, the computer learns based on a plurality of teacher data acquired in the acquisition step S12. In the generation step S16, the computer generates a learning model based on the result learned in the learning step S15. The learning model takes input information as input and outputs an evaluation. The input information is unknown information different from the teacher data. The input information is information including at least glass cloth information and dispersion liquid information.

[0194] Furthermore, as described above, a learning model obtained by learning glass cloth information, dispersion liquid information, and evaluation as teacher data is used in the computer as a program to determine the evaluation. The learning model includes an input step S22, a determination step S23, and an output step S24. In the input step S22, input information that is information including glass cloth information and dispersion liquid information and is unknown information different from the teacher data is input. In the determination step S23, the evaluation is determined using the learning model. In the output step S24, the evaluation determined in the determination step S23 is output.

[0195] Conventionally, the evaluation of an impregnated body obtained by impregnating a glass cloth with a fluoropolymer aqueous dispersion has been carried out at the site by testing each of various combinations of glass cloths and fluoropolymer aqueous dispersions. Such a conventional evaluation method requires a lot of time and processes for evaluation, and improvement of the evaluation method has been demanded. Also, as shown in Patent Document 2 (International Publication No. 99 / 07543), programs using neural networks and the like have been designed in different fields to output optimal combinations, but in the special field of impregnated bodies using fluoropolymer aqueous dispersions, programs using neural networks and the like have not been designed. The learning model generated by the learning model generation method of the present embodiment can be evaluated using a computer. It is possible to reduce a lot of time and processes that were conventionally required. Furthermore, by reducing the processes, it is also possible to reduce the number of personnel required for evaluation, and the cost for evaluation can also be reduced.

[0196] (8-2) The learning model generation method of the present embodiment is a method for generating a learning model that determines an optimal fluoropolymer aqueous dispersion for obtaining an evaluation of a target impregnated body using a computer. It includes an acquisition step S12, a learning step S15, and a generation step S16. In the acquisition step S12, the computer acquires teacher data. The teacher data includes glass cloth information, dispersion liquid information, and an evaluation. The glass cloth information is information about the glass cloth. The dispersion liquid information is information about the fluoropolymer aqueous dispersion. The evaluation is an evaluation of an impregnated body obtained by impregnating a glass cloth with a fluoropolymer aqueous dispersion. In the learning step S15, the computer learns based on a plurality of teacher data acquired in the acquisition step S12. The generation step S16 generates a learning model based on the result of learning in the learning step S15. The learning model takes input information as input and outputs dispersion liquid information. The input information is unknown information different from the teacher data. The input information is information including at least glass cloth information and evaluation information.

[0197] Furthermore, as described above, a learning model that has learned glass cloth information, dispersion liquid information, and evaluation as teacher data is used as a program in a computer to determine the dispersion liquid information. The program includes an input step S22, a determination step S23, and an output step S24. The input step S22 receives input information that is information including glass cloth information and evaluation information and is unknown information different from the teacher data. The determination step S23 determines optimal dispersion liquid information for obtaining an evaluation of a target impregnated body using the learning model. The output step S24 outputs the dispersion liquid information determined in the determination step S23.

[0198] In the conventional evaluation method, when an evaluation of a target impregnated body cannot be obtained, further research and improvement had to be carried out to find an optimal fluoropolymer aqueous dispersion, which required a lot of time and processes. The learning model generated by the learning model generation method of the present embodiment can determine an optimal fluoropolymer aqueous dispersion for obtaining an evaluation of a target impregnated body using a computer. As a result, it is possible to reduce the time, processes, personnel, costs, etc. for selecting an optimal fluoropolymer aqueous dispersion.

[0199] (8-3) The method for generating a learning model according to this embodiment is a method for generating a learning model that determines, using a computer, an optimal glass cloth for obtaining an evaluation of a target impregnated body. The method includes an acquisition step S12, a learning step S15, and a generation step S16. In the acquisition step S12, the computer acquires teacher data. The teacher data includes dispersion liquid information, glass cloth information, and an evaluation. The dispersion liquid information is information on an aqueous fluoropolymer dispersion. The glass cloth information is information on a glass cloth. The evaluation is an evaluation of an impregnated body obtained by impregnating a glass cloth with an aqueous fluoropolymer dispersion. In the learning step S15, the computer learns based on a plurality of pieces of teacher data acquired in the acquisition step S12. The generation step S16 generates a learning model based on the result of learning in the learning step S15. The learning model takes input information as input and outputs glass cloth information. The input information is unknown information different from the teacher data. The input information is information including at least dispersion liquid information and evaluation information.

[0200] Furthermore, as described above, a learning model obtained by learning dispersion liquid information, glass cloth information, and an evaluation as teacher data is used as a program in a computer to determine glass cloth information. The program includes an input step S22, a determination step S23, and an output step S24. The input step S22 receives input information that is information including dispersion liquid information and evaluation information and is unknown information different from the teacher data. The determination step S23 determines optimal glass cloth information for obtaining an evaluation of a target impregnated body using the learning model. The output step S24 outputs the glass cloth information determined in the determination step S23.

[0201] In the conventional evaluation method, when an evaluation of a target impregnated body cannot be obtained, further research and improvement had to be carried out to find an optimal glass cloth, which required a lot of time and processes. The learning model generated by the learning model generation method of the present embodiment can determine, using a computer, an optimal glass cloth for obtaining a target impregnated body evaluation. This makes it possible to reduce the time, processes, personnel, costs, etc. for selecting the optimal glass cloth.

[0202] (8-4) In the learning model generation method and program of (8-1) to (8-3) described above, the teacher data preferably further includes processing condition information which is information on processing conditions for obtaining the impregnated body. In this aspect, the input information preferably further includes the processing condition information. The teacher data preferably includes information on many items, and the larger the number of teacher data, the better. This makes it possible to obtain a more accurate output.

[0203] (8-5) The learning model generation method of the present embodiment is a learning model generation method for determining, using a computer, an optimal processing condition for evaluating a target impregnated body. It includes an acquisition step S12, a learning step S15, and a generation step S16. In the acquisition step S12, the computer acquires teacher data. The teacher data includes glass cloth information, dispersion liquid information, processing condition information, and an evaluation. The glass cloth information is information on the glass cloth. The dispersion liquid information is information on the fluoropolymer aqueous dispersion. The processing condition information is information on the processing conditions for impregnating the glass cloth with the fluoropolymer aqueous dispersion to obtain an impregnated body. The evaluation is an evaluation of the impregnated body obtained by impregnating the glass cloth with the fluoropolymer aqueous dispersion. In the learning step S15, the computer learns based on a plurality of teacher data acquired in the acquisition step S12. The generation step S16 is for the computer to generate a learning model based on the result of learning in the learning step S15. The learning model takes input information as input and outputs processing condition information. The input information is unknown information different from the teacher data. The input information is information including at least glass cloth information, dispersion liquid information, and evaluation information.

[0204] Furthermore, as described above, a learning model obtained by learning glass cloth information, dispersion liquid information, processing condition information, and evaluation as teacher data is used as a program in a computer to determine the processing condition information. The program includes an input step S22, a determination step S23, and an output step S24. The input step S22 receives input information that is information including glass cloth information, dispersion liquid information, and evaluation information and is unknown information different from the teacher data. The determination step S23 determines the optimal processing condition information for obtaining an evaluation of the target impregnated body using the learning model. The output step S24 outputs the processing condition information determined in the determination step S23.

[0205] In the conventional evaluation method, when the evaluation of the target impregnated body cannot be obtained, further research and improvement must be carried out to find the optimal processing conditions, which requires a lot of time and processes. The learning model generated by the learning model generation method of the present embodiment can determine the optimal processing conditions for obtaining an evaluation of the target impregnated body using a computer. As a result, it is possible to reduce the time, processes, personnel, costs, etc. for selecting the optimal processing conditions.

[0206] (8-6) The learning step S15 of the learning model generation method of the present embodiment is preferably performed by regression analysis and / or ensemble learning combining a plurality of regression analyses.

[0207] The evaluation of the learning model as a program of the present embodiment preferably includes information regarding at least one selected from the group consisting of the crack limit film thickness, color tone, peelability, bending life, surface roughness, electrical characteristics, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body. The glass cloth information preferably includes information regarding at least one selected from the group consisting of the weaving and knitting method, thickness, count, density, twist, filament diameter, weight, type, laminate structure, and tensile strength of the glass cloth. The dispersion information preferably includes at least one type of information selected from the group consisting of information regarding the fluoropolymer contained in the dispersion, information regarding components other than the fluoropolymer contained in the dispersion, and information regarding the properties of the dispersion. The processing condition information preferably includes information regarding at least one type selected from the group consisting of the solid content concentration during processing of the dispersion, the surface tension during processing, additives during processing, the number of impregnation times and the impregnation amount, and the heating conditions. Since these glass cloth information, dispersion information, and processing condition information have a strong correlation with the above evaluation items for the impregnated body, by using this information, a more accurate output can be obtained.

[0208] (8-7) The learning model as the program of the present embodiment may be distributed via a storage medium storing the program.

[0209] (8-8) The learned model of the present embodiment is a learned model learned in the learning model generation method. The learned model of the present embodiment is a learned model for causing a computer to function so as to perform an operation based on the weighting coefficients of the neural network on the glass cloth information, which is the information of the glass cloth input to the input layer of the neural network, and the dispersion information, which is the information of the aqueous dispersion of the fluoropolymer to be impregnated into the glass cloth, and output an evaluation of the impregnated body obtained by impregnating the glass cloth with the dispersion from the output layer of the neural network. The weighting coefficients are obtained by learning using at least the glass cloth information, the dispersion information, and the evaluation as teacher data.

[0210] (8-9) The learned model of the present embodiment is a learned model for causing a computer to function so as to perform an operation based on the weighting coefficients of a neural network on the glass cloth information, which is the information of the glass cloth input to the input layer of the neural network, and the information of the evaluation of the impregnated body obtained by impregnating the glass cloth with a fluoropolymer aqueous dispersion, and output optimal dispersion liquid information for obtaining a target evaluation from the output layer of the neural network. The weighting coefficients are obtained by learning using at least the glass cloth information, the dispersion liquid information, which is the information of the fluoropolymer aqueous dispersion to be impregnated into the glass cloth, and the evaluation of the impregnated body as teacher data.

[0211] (8-10) The learned model of the present embodiment is a learned model for causing a computer to function so as to perform an operation based on the weighting coefficients of a neural network on the dispersion liquid information, which is the information of the fluoropolymer aqueous dispersion to be impregnated into the glass cloth, and the information of the evaluation of the impregnated body obtained by impregnating the glass cloth with a fluoropolymer aqueous dispersion, and output optimal glass cloth information for obtaining a target evaluation from the output layer of the neural network. The weighting coefficients are obtained by learning using at least the dispersion liquid information, the glass cloth information, which is the information of the glass cloth to be impregnated with the fluoropolymer aqueous dispersion, and the evaluation of the impregnated body as teacher data.

[0212] (8-11) In the learned model of (8-8) to (8-10) described above, it is preferable that the teacher data further includes processing condition information, which is information on processing conditions for obtaining the impregnated body. In this aspect, it is preferable that the processing condition information is further input to the input layer.

[0213] (8-12) The learned model of the present embodiment is a learned model for causing a computer to function such that, for glass cloth information which is information of a glass cloth input to an input layer of a neural network, dispersion liquid information which is information of an aqueous dispersion of a fluoropolymer to be impregnated into the glass cloth, and information on evaluation of an impregnated body obtained by impregnating the glass cloth with the aqueous dispersion of the fluoropolymer, an operation based on the weighting coefficients of the neural network is performed, and optimum processing condition information for obtaining a target evaluation is output from an output layer of the neural network. The weighting coefficients are obtained by learning using, as teacher data, at least the glass cloth information, the dispersion liquid information, processing condition information which is information on processing conditions for obtaining an impregnated body by impregnating the glass cloth with the aqueous dispersion of the fluoropolymer, and the evaluation of the impregnated body.

[0214] (8-13) In each of the above-described embodiments, the evaluation as input information or output information preferably includes information regarding at least one selected from the group consisting of the crack-limiting film thickness, color tone, peelability, flex life, surface roughness, electrical properties, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body, more preferably includes information regarding at least one selected from the group consisting of the crack-limiting film thickness, color tone, peelability, flex life, surface roughness, and electrical properties of the impregnated body, still more preferably includes information regarding the crack-limiting film thickness, color tone, peelability, flex life, surface roughness, and electrical properties of the impregnated body, and particularly preferably includes information regarding the crack-limiting film thickness, color tone, peelability, flex life, surface roughness, electrical properties, weather resistance, tensile strength, tear strength, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body. The evaluation as input information or output information preferably also includes information regarding the crack-limiting film thickness, color tone, peelability, flex life, and surface roughness of the impregnated body. The glass cloth information as input information or output information preferably includes information regarding at least one selected from the group consisting of the weaving and knitting method, thickness, count, density, twist, filament diameter, weight, type, laminate structure, and tensile strength of the above glass cloth, and more preferably includes information regarding the weaving and knitting method, thickness, count, density, twist, filament diameter, weight, type, laminate structure, and tensile strength. The dispersion liquid information as input information or output information preferably includes at least one type of information selected from the group consisting of information regarding the fluoropolymer contained in the above dispersion liquid, information regarding components other than the fluoropolymer contained in the above dispersion liquid, and information regarding the properties of the above dispersion liquid, and more preferably includes information regarding the fluoropolymer contained in the above dispersion liquid, information regarding components other than the fluoropolymer contained in the above dispersion liquid, and information regarding the properties of the above dispersion liquid. The preferred items of each of these information are as described above. The processing condition information as input information or output information preferably includes information regarding at least one selected from the group consisting of the solid content concentration during processing, surface tension during processing, additives during processing, number of impregnations and impregnation amount, and heating conditions of the above dispersion liquid, and preferably includes information regarding at least one selected from the group consisting of the solid content concentration during processing, surface tension during processing, additives during processing, number of impregnations and impregnation amount, drying conditions, and firing conditions of the above dispersion liquid, and more preferably includes information regarding the impregnation amount, drying conditions, and firing conditions of the above dispersion liquid. The processing condition information as input information or output information also preferably includes information regarding at least one selected from the group consisting of the solid content concentration during processing, surface tension during processing, additives during processing, number of impregnations and impregnation amount, drying temperature, drying time, heating rate, and ventilation conditions in the drying process, and firing temperature, firing time, heating rate, cooling rate, and ventilation conditions in the firing process of the above dispersion liquid, and more preferably includes information regarding the solid content concentration during processing, surface tension during processing, additives during processing, number of impregnations and impregnation amount, drying temperature, drying time, heating rate, and ventilation conditions in the drying process, and firing temperature, firing time, heating rate, cooling rate, and ventilation conditions in the firing process of the above dispersion liquid. The processing condition information as input information or output information preferably also includes information regarding at least one selected from the group consisting of the solid content concentration during processing of the dispersion liquid, the surface tension during processing, additives during processing, the number of impregnations and the impregnation amount, the drying temperature, the drying time, the firing temperature, the firing time, the heating rate, the cooling rate, and the ventilation conditions.

[0215] (9) As described above, the embodiments of the present disclosure have been explained, and it will be understood that various changes in form and details are possible without departing from the spirit and scope of the present disclosure described in the claims.

Explanation of Reference Numerals

[0216] S12 Acquisition Step S15 Learning Step S16 Generation Step S22 Input Step S23 Determination Step S24 Output Step

Claims

1. A learning model generation method for generating a learning model that determines, using a computer, an evaluation of an impregnated body obtained by impregnating a glass cloth with a fluoropolymer aqueous dispersion, comprising: an acquisition step (S12) in which the computer acquires, as teacher data, information including at least glass cloth information which is information on the glass cloth, dispersion liquid information which is information on the fluoropolymer aqueous dispersion, and the evaluation of the impregnated body; a learning step (S15) in which the computer learns based on the plurality of pieces of teacher data acquired in the acquisition step (S12); a generation step (S16) in which the computer generates the learning model based on the result of learning in the learning step (S15); wherein the learning model takes, as input, input information which is unknown information different from the teacher data, and outputs the evaluation; the input information is information including at least the glass cloth information and the dispersion liquid information; the evaluation includes information regarding at least one selected from the group consisting of electrical characteristics, weather resistance, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body; a learning model generation method.

2. A learning model generation method for generating a learning model that determines, using a computer, optimal dispersion liquid information for obtaining an evaluation of a target impregnated body, comprising: an acquisition step (S12) in which the computer acquires, as teacher data, information including at least glass cloth information which is information on the glass cloth, dispersion liquid information which is information on the fluoropolymer aqueous dispersion to be impregnated into the glass cloth, and an evaluation of the impregnated body obtained by impregnating the glass cloth with the dispersion liquid; a learning step (S15) in which the computer learns based on the plurality of pieces of teacher data acquired in the acquisition step (S12); a generation step (S16) in which the computer generates a learning model based on the result of learning in the learning step (S15); wherein the learning model takes, as input, input information which is unknown information different from the teacher data, and outputs optimal dispersion liquid information for obtaining an evaluation of a target impregnated body; the input information is information including at least the glass cloth information and information on the evaluation; the evaluation includes information regarding at least one selected from the group consisting of electrical characteristics, weather resistance, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body; a learning model generation method.

3. A learning model generation method for generating a learning model that determines optimal glass cloth information for obtaining an evaluation of a target impregnated body using a computer, comprising: an acquisition step (S12) in which a computer acquires, as teacher data, information including at least dispersion liquid information that is information on a fluoropolymer aqueous dispersion, glass cloth information that is information on a glass cloth to be impregnated with the dispersion liquid, and an evaluation of an impregnated body obtained by impregnating the glass cloth with the dispersion liquid; a learning step (S15) in which the computer learns based on a plurality of pieces of the teacher data acquired in the acquisition step (S12); a generation step (S16) in which the computer generates a learning model based on the result learned in the learning step (S15); and comprising: the learning model takes, as input, input information that is unknown information different from the teacher data, and outputs optimal glass cloth information for obtaining an evaluation of a target impregnated body; the input information is information including at least the dispersion liquid information and the evaluation information; the evaluation includes information regarding at least one selected from the group consisting of electrical properties, weather resistance, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body; A learning model generation method.

4. The teacher data further includes processing condition information that is information on processing conditions for obtaining the impregnated body; the input information further includes the processing condition information; The learning model generation method according to any one of Claims 1 to 3.

5. A learning model generation method for generating a learning model that determines optimal processing condition information for obtaining an evaluation of a target impregnated body using a computer, comprising: an acquisition step (S12) in which a computer acquires, as teacher data, information including at least glass cloth information that is information on a glass cloth, dispersion liquid information that is information on a fluoropolymer aqueous dispersion to be impregnated into the glass cloth, processing condition information that is information on processing conditions for obtaining an impregnated body by impregnating the glass cloth with the dispersion liquid, and an evaluation of an impregnated body obtained by impregnating the glass cloth with the dispersion liquid; a learning step (S15) in which the computer learns based on a plurality of pieces of the teacher data acquired in the acquisition step (S12); a generation step (S16) in which the computer generates a learning model based on the result learned in the learning step (S15); and comprising: The learning model takes input information, which is unknown information different from the teacher data, and outputs optimal processing condition information for obtaining an evaluation of a target impregnated body. The input information is information including at least the glass cloth information, the dispersion liquid information, and the evaluation information. The evaluation includes information regarding at least one selected from the group consisting of the electrical properties, weather resistance, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body. Learning model generation method.

6. In the learning step (S15), learning is performed by regression analysis and / or ensemble learning combining a plurality of regression analyses. The learning model generation method according to any one of Claims 1 to 5.

7. A program in which a computer uses a learning model to determine an evaluation of an impregnated body obtained by impregnating a glass cloth with a fluoropolymer aqueous dispersion, an input step (S22) in which the computer inputs input information, a determination step (S23) in which the computer determines the evaluation, and an output step (S24) in which the computer outputs the evaluation determined in the determination step (S23), comprising: The learning model learns, as teacher data, information including at least glass cloth information which is information of the glass cloth, dispersion liquid information which is information of the fluoropolymer aqueous dispersion, and the evaluation. The input information is information including at least the glass cloth information and the dispersion liquid information, which is unknown information different from the teacher data. The evaluation includes information regarding at least one selected from the group consisting of the electrical properties, weather resistance, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body. Program.

8. A program in which a computer uses a learning model to determine dispersion liquid information, which is information of an optimal fluoropolymer aqueous dispersion for obtaining an evaluation of a target impregnated body, an input step (S22) in which the computer inputs input information, a determination step (S23) in which the computer determines the optimal dispersion liquid information, and an output step (S24) in which the computer outputs the optimal dispersion liquid information determined in the determination step (S23), comprising: The learning model learns, as teacher data, information including at least glass cloth information which is information of the glass cloth, dispersion liquid information which is information of the fluoropolymer aqueous dispersion to be impregnated into the glass cloth, and evaluation of an impregnated body obtained by impregnating the glass cloth with the fluoropolymer aqueous dispersion. The input information is information including at least the glass cloth information and the evaluation information, which is unknown information different from the teacher data. The evaluation includes information regarding at least one selected from the group consisting of electrical characteristics, weather resistance, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body. Program.

9. A program in which a computer uses a learning model to determine glass cloth information which is optimal glass cloth information for obtaining an evaluation of a target impregnated body, an input step (S22) in which the computer inputs input information, a determination step (S23) in which the computer determines the optimal glass cloth information, an output step (S24) in which the computer outputs the optimal glass cloth information determined in the determination step (S23), comprising: The learning model learns, as teacher data, information including at least dispersion liquid information which is information of the fluoropolymer aqueous dispersion, glass cloth information which is information of the glass cloth to be impregnated with the dispersion liquid, and evaluation of an impregnated body obtained by impregnating the glass cloth with the dispersion liquid. The input information is information including at least the dispersion liquid information and the evaluation information, which is unknown information different from the teacher data. The evaluation includes information regarding at least one selected from the group consisting of electrical characteristics, weather resistance, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body. Program.

10. The teacher data further includes processing condition information which is information of processing conditions for obtaining the impregnated body. The input information further includes the processing condition information. The program according to any one of Claims 7 to 9.

11. A program in which a computer uses a learning model to determine processing condition information which is optimal processing condition information for obtaining an evaluation of a target impregnated body, an input step (S22) in which the computer inputs input information, a determination step (S23) in which the computer determines the optimal processing condition information, An output step (S24) in which the computer outputs the optimal processing condition information determined in the determination step (S23); comprising; The learning model learns, as teacher data, information including at least glass cloth information which is information of a glass cloth, dispersion liquid information which is information of a fluoropolymer aqueous dispersion to be impregnated into the glass cloth, processing condition information which is information of processing conditions for obtaining an impregnated body by impregnating the glass cloth with the dispersion liquid, and evaluation of the impregnated body obtained by impregnating the glass cloth with the dispersion liquid; The input information is information including at least the glass cloth information, the dispersion liquid information, and information of the evaluation, and is unknown information different from the teacher data; The evaluation includes information regarding at least one selected from the group consisting of electrical characteristics, weather resistance, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body; Program.

12. The processing condition information includes information regarding at least one selected from the group consisting of the solid content concentration during processing of the dispersion liquid, the surface tension during processing, additives during processing, the number of impregnation times and the impregnation amount, and heating conditions; The program according to claim 10 or 11.

13. The glass cloth information includes information regarding at least one selected from the group consisting of the weaving and knitting method of the glass cloth, thickness, count, density, twist, filament diameter, weight, type, laminate structure, and tensile strength; The program according to any one of claims 7 to 12.

14. The dispersion liquid information includes at least one type of information selected from the group consisting of information regarding the fluoropolymer contained in the dispersion liquid, information regarding components other than the fluoropolymer contained in the dispersion liquid, and information regarding the properties of the dispersion liquid; The program according to any one of claims 7 to 13.

15. A storage medium storing the program according to any one of claims 7 to 14.

16. A learned model for causing a computer to function such that, for the glass cloth information and the dispersion liquid information input to the input layer of a neural network, an operation based on the weight coefficients of the neural network is performed, and an evaluation of the impregnated body is output from the output layer of the neural network, wherein the weight coefficients are obtained by learning using at least the glass cloth information, the dispersion liquid information, and the evaluation as teacher data. The glass cloth information is information on a glass cloth, The dispersion liquid information is information on an aqueous dispersion of a fluoropolymer to be impregnated into the glass cloth, The evaluation is an evaluation of an impregnated body obtained by impregnating the glass cloth with the dispersion liquid, and includes information regarding at least one selected from the group consisting of the electrical properties, weather resistance, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body. A learned model.

17. A learned model for causing a computer to function so as to perform an operation based on the weighting coefficients of the neural network on the glass cloth information input to the input layer of the neural network and the evaluation information, and output optimal dispersion liquid information for obtaining a target evaluation from the output layer of the neural network, The weighting coefficients are obtained by learning using at least the glass cloth information, the dispersion liquid information, and the evaluation as teacher data, The glass cloth information is information on a glass cloth, The dispersion liquid information is information on an aqueous dispersion of a fluoropolymer to be impregnated into the glass cloth, The evaluation is an evaluation of an impregnated body obtained by impregnating the glass cloth with the dispersion liquid, and includes information regarding at least one selected from the group consisting of the electrical properties, weather resistance, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body. A learned model.

18. A learned model for causing a computer to function so as to perform an operation based on the weighting coefficients of the neural network on the dispersion liquid information input to the input layer of the neural network and the evaluation information, and output optimal glass cloth information for obtaining a target evaluation from the output layer of the neural network, The weighting coefficients are obtained by learning using at least the glass cloth information, the dispersion liquid information, and the evaluation as teacher data, The dispersion liquid information is information on an aqueous dispersion of a fluoropolymer, The glass cloth information is information on a glass cloth to be impregnated with the dispersion liquid, The evaluation is an evaluation of an impregnated body obtained by impregnating the glass cloth with the dispersion liquid, and includes information regarding at least one selected from the group consisting of the electrical properties, weather resistance, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body. A learned model.

19. The teacher data further includes processing condition information which is information on processing conditions for obtaining the impregnated body. The processing condition information is further input to the input layer. The learned model according to any one of claims 16 to 18.

20. A learned model for causing a computer to function so as to perform an operation based on the weighting coefficients of the neural network on the glass cloth information, the dispersion liquid information, and the evaluation information input to the input layer of the neural network, and output optimal processing condition information for obtaining a target evaluation from the output layer of the neural network, The weighting coefficients are obtained by learning using at least the glass cloth information, the dispersion liquid information, the processing condition information, and the evaluation as teacher data. The glass cloth information is information on a glass cloth. The dispersion liquid information is information on an aqueous dispersion of a fluoropolymer to be impregnated into the glass cloth. The processing condition information is information on processing conditions for obtaining an impregnated body by impregnating the glass cloth with the dispersion liquid. The evaluation is an evaluation of the impregnated body obtained by impregnating the glass cloth with the dispersion liquid, and includes information on at least one selected from the group consisting of the electrical characteristics, weather resistance, breakdown voltage, volume resistivity, and surface resistivity of the impregnated body. Learned model.

Images