Apparatus and method for predicting physical properties of resin composition

Abstract

To provide a device and a method for predicting a physical property of a composite material which can accurately predict a physical property of a composite material.SOLUTION: A device 1 for predicting a physical property of a resin composition includes: a regression model creation processing unit 22 for mechanically learning a relation between explanatory variable data and physical property data 14 including information on a physical property of a prediction object, and creating a regression model 7 indicating a correlation between the explanatory variable data and the physical property data 14; a physical property prediction processing unit 23 for predicting the physical property of the prediction object using the regression model 7; database 6 including blending amount data 11 including at least a base material and information on each blending amount of a plurality of addition materials, and physical property data 14 including at least information on the physical property of the prediction object; and a data extraction processing unit 21 for extracting the explanatory variable data from the database 6, according to physical property predicted by the physical property prediction processing unit.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to resin composition The present invention relates to an apparatus and method for predicting physical properties. [Background technology]

[0002] In recent years, methods and devices for designing materials using the results of machine learning have been proposed (see, for example, Patent Document 1). By using such methods and devices, it becomes possible to predict physical properties according to, for example, the blending amounts of raw materials, shortening the material development period and reducing development costs. [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2019-82790 A Summary of the Invention [Problem to be solved by the invention]

[0004] by the way, Resin composition used as a coating material for electric wires Things When designing, As physical properties Elongation and tensile strength Degree In some cases, it is desirable to predict the temperature accurately. Ma In addition, flame retardants, flame retardant assistants, antioxidants, copper damage inhibitors, lubricants, colorants, crosslinking assistants Agents etc. Many materials such as resin composition In this case, resin composition When predicting the physical properties of a material, simply using the blend amounts of all materials may not provide sufficient prediction accuracy.

[0005] Therefore, the present invention provides resin composition The physical properties of resin composition The present invention aims to provide an apparatus and method for predicting physical properties. [Means for solving the problem]

[0006] In order to solve the above-mentioned problems, the present invention provides a physical property prediction device for predicting the properties of a composite material manufactured using a base material and a plurality of additive materials that are added to the base material to impart functions to the base material, the device comprising: a regression model creation processing unit that performs machine learning to learn the relationship between explanatory variable data and physical property data including information on the physical properties of a target to be predicted, and creates a regression model that represents the correlation between the explanatory variable data and the physical property data; a physical property prediction processing unit that predicts the physical properties of the target to be predicted using the regression model; a database that includes blending amount data including information on the blending amounts of at least the base material and the plurality of additive materials, and physical property data including at least information on the physical properties of the target to be predicted; and a data extraction processing unit that extracts the explanatory variable data from the database according to the physical properties predicted by the physical property prediction processing unit, wherein the data extraction processing unit selects and discards information related to the physical property to be predicted from the information on each blending amount included in the blending amount data, and performs data extraction such that the acquired first information is included in the explanatory variable data.

[0007] Further, in order to solve the above-mentioned problems, the present invention provides a method for predicting the properties of a composite material produced using a base material and a plurality of additive materials that are added to the base material in order to impart functions to the base material, the method comprising: a regression model creation process for performing machine learning on the relationship between explanatory variable data and physical property data including information on the physical properties of a prediction target, and creating a regression model that expresses the correlation between the explanatory variable data and the physical property data; and a physical property prediction process for predicting the physical properties of the prediction target using the regression model. The present invention provides a method for predicting physical properties of a composite material, comprising: a data acquisition processing step of registering in a database blending amount data including information on the blending amounts of at least the base material and the plurality of additive materials, and physical property data including at least information on the physical property to be predicted, and a data extraction processing step of extracting the explanatory variable data from the database according to the physical property to be predicted by the physical property prediction processing, wherein the data extraction processing step selects and discards first information related to the physical property to be predicted from the information on the blending amounts included in the blending amount data, and performs data extraction so that the acquired first information is included in the explanatory variable data. Effect of the Invention

[0008] According to the present invention, resin composition The physical properties of resin composition It is possible to provide a property prediction device and method. [Brief description of the drawings]

[0009] [Figure 1] 1 is a schematic diagram of a resin composition property prediction device according to an embodiment of the present invention. FIG. [Diagram 2] FIG. 11 is a diagram illustrating an example of learning data. [Diagram 3] FIG. 11 is a diagram illustrating a data extraction process. [Figure 4] FIG. 1A is a diagram for explaining a regression model creation process, and FIG. 1B is a diagram for explaining a property prediction process. [Diagram 5] FIG. 1 is a flow diagram of a method for predicting physical properties of a resin composition according to one embodiment of the present invention. [Figure 6] 13A is a flow diagram of the data acquisition process, and FIG. 13B is a flow diagram of the data extraction process. [Figure 7] FIG. 1A is a flow diagram of a regression model creation process, and FIG. [Figure 8] FIG. 2A is a diagram showing data used as learning data in an example of the present invention, and FIG. 2B is a diagram showing the average value of the mean absolute error rate obtained by the study. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0010] [Embodiment Mode] Hereinafter, an embodiment of the present invention will be described with reference to the accompanying drawings.

[0011] (Schematic configuration of resin composition property prediction device 1) 1 is a schematic diagram of a resin composition property prediction device 1 according to the present embodiment (hereinafter, simply referred to as the property prediction device 1). The property prediction device 1 is a device that predicts the properties of a resin composition produced using a base polymer, a flame retardant, a flame retardant assistant, and other materials.

[0012] Here, the case will be described where the resin composition to be predicted for physical properties is a non-halogen resin composition that is a coating material for electric wires and is crosslinked by electron beam irradiation. In this embodiment, the physical property to be predicted is the initial elongation or tensile strength of the resin composition to be predicted. Note that "initial" here refers to a state in which the resin composition has not deteriorated.

[0013] 1, the property prediction apparatus 1 includes a control unit 2, a storage unit 3, a display unit 4, and an input device 5. In this embodiment, the property prediction apparatus 1 is configured by a personal computer.

[0014] The control unit 2 is realized by appropriately combining a calculation element such as a CPU, a memory, an interface, software, a storage device, etc. In this embodiment, the control unit 2 has a data acquisition processing unit 20, a data extraction processing unit 21, a regression model creation processing unit 22, a physical property prediction processing unit 23, and a predicted physical property presentation processing unit 24. Details of each unit will be described later.

[0015] The storage unit 3 is realized by a predetermined storage area of ​​a memory or a storage device. The display unit 4 is, for example, a liquid crystal display, and the input device 5 is, for example, a keyboard or a mouse. The display unit 4 may be configured as a touch panel, and the display unit 4 may also function as the input device 5. The display unit 4 and the input device 5 may be configured separately from the physical property prediction device 1, and may be communicated by wireless communication or the like. The display device 4 or the input device 5 may be configured to be capable of communicating with the physical property prediction device 1. In this case, the display device 4 or the input device 5 may be configured as a mobile terminal such as a tablet or a smartphone.

[0016] (About Database 6) Next, the database 6 will be described. The database 6 is a database that integrates all information including information used for machine learning. FIG. 2 is a diagram showing an example of the database 6. Note that FIG. 2 shows the concept of the database 6, and does not describe actual experimental data, etc. As shown in FIG. 2, the database 6 includes at least blend amount data 11, irradiation amount data 13, and physical property data 14. In the illustrated example, the database 6 includes ID data (ID) 12 for identifying the resin composition. Note that the database 6 may include information other than these, and may further include, for example, data representing the manufacturing conditions of the resin composition (so-called process data) and data representing the state of the structure of the resin composition (so-called structure data).

[0017] The blending amount data 11 includes information on the blending amount of each material constituting the resin composition. More specifically, the blending amount data 11 includes base polymer blending amount data 11a including information on the blending amount of the base polymer, filler blending amount data 11b including information on the blending amount of the filler (i.e., flame retardant and flame retardant assistant), and other material blending amount data 11c including information on the blending amount of each other material. In this embodiment, the other materials include an antioxidant, a copper damage inhibitor, a lubricant, a crosslinking assistant, and a colorant. Note that the other materials may include other materials such as a crosslinking agent and a stabilizer.

[0018] In this embodiment, since the resin composition crosslinked by electron beam irradiation is the object of predicting the physical properties, the database 6 has irradiation amount data 13 including information on the irradiation amount of the electron beam during crosslinking. The irradiation amount data 13 is data representing the manufacturing conditions, and can be said to be a part of so-called process data. Note that, in the case of a resin composition that is not crosslinked by electron beam irradiation, In this case, the dose data 13 can be omitted.

[0019] In this embodiment, since the physical property to be predicted is elongation or tensile strength, the physical property data 14 includes at least the elongation or tensile strength of the resin composition to be predicted. Note that the physical property data 14 may include, in addition to the initial elongation and tensile strength, degradation characteristics of elongation and tensile strength under specific environments. In the illustrated example, the elongation and tensile strength upon heat aging are shown as degradation characteristics, but are not limited thereto, and may include degradation characteristics under various environments such as oil resistance, fuel resistance, and cold resistance.

[0020] (Data acquisition processing unit 20) The data acquisition processing unit 20 performs a data acquisition process (see FIG. 6(a)) in which various data input from the outside are registered in the database 6 and stored in the storage unit 3. The input of various data may be performed by the input device 5, or may be performed by communication from an external device (wired communication, wireless communication, communication via a network, etc.), or may be performed via a medium such as a USB memory. The data acquisition processing unit 20 may also be configured to actively acquire data, for example, by transmitting a signal requesting data to an external device.

[0021] (Data extraction processing unit 21) The data extraction processing unit 21 performs a data extraction process to extract explanatory variable data 61 used as explanatory variables during machine learning and dependent variable data 62 used as dependent variables during machine learning from a database 6 containing various data according to the physical properties to be predicted (physical properties predicted by the physical property prediction processing unit 23) to generate learning data 6a (see Figures 3 and 6(b)).

[0022] When the physical property predicted by the physical property prediction processing unit 23 is the initial elongation or tensile strength of the prediction target, the data extraction processing unit 21 extracts data so that the explanatory variable data 61 includes at least information on the respective blend amounts of the base polymer, the flame retardant, and the flame retardant aid (i.e., the base polymer blend amount data 11a and the filler blend amount data 11b). In this embodiment, when the physical property predicted by the physical property prediction processing unit 23 is the initial elongation or tensile strength of the prediction target, the data extraction processing unit 21 extracts data so that the explanatory variable data 61 does not include information on the respective blend amounts of the antioxidant, the copper inhibitor, and the lubricant.

[0023] This is because prediction accuracy can be improved by using information on the amounts of the base polymer, flame retardant, and flame retardant aid, which are relatively large, as explanatory variables. The antioxidant, copper inhibitor, and lubricant are used in small amounts and the amount of each agent fluctuates little, so they have little effect on the initial elongation or tensile strength. If information on the amounts of these agents is included in the explanatory variables, it may become noise during machine learning and reduce prediction accuracy. Therefore, when the property predicted by the property prediction processing unit 23 is the initial elongation or tensile strength of the object to be predicted, it is possible to improve prediction accuracy by omitting information on the amounts of the antioxidant, copper inhibitor, and lubricant from the explanatory variables.

[0024] In this embodiment, since the prediction target is a halogen-free resin composition cross-linked by electron beam irradiation, the data extraction processing unit 21 is configured to extract data such that the explanatory variable data 61 includes the irradiation dose data 13. Furthermore, a cross-linking assistant is included as an additional material in a halogen-free resin composition cross-linked by electron beam irradiation, and since this cross-linking assistant may affect elongation and tensile strength, the data extraction processing unit 21 extracts data such that the explanatory variable data 61 includes information on the blending amount of the cross-linking assistant.

[0025] However, it is not essential to include the irradiation dose data 13 and information on the blending amount of the cross-linking auxiliary in the explanatory variable data 61. For example, even when a halogen-free resin composition that is cross-linked by electron beam irradiation is the prediction target, if the irradiation amount of the electron beam is always constant (or there is almost no variation in the irradiation amount), the irradiation dose data 13 may be omitted from the explanatory variable data 61. Furthermore, if the blending amount of the cross-linking auxiliary added is very small or the variation in the blending amount of the cross-linking auxiliary is very small, information on the blending amount of the cross-linking auxiliary may also be omitted from the explanatory variable data 61.

[0026] If the resin composition to be predicted is one that is not crosslinked by electron beam irradiation, then it is of course possible to omit the irradiation dose data 13 and the information on the blending amount of the crosslinking aid. Note that, for example, if the resin composition to be predicted is one that is crosslinked by chemical crosslinking, then the explanatory variable data 61 may contain information on the blending amount of the crosslinking agent, and other data (process data, organizational data, etc.) may be added to the explanatory variable data 61 as necessary.

[0027] In this embodiment, the explanatory variable data 61 further includes information on the blending amount of the colorant, but the information on the blending amount of the colorant can be omitted. Also, as described above, even when the prediction target is a halogen-free resin composition that is crosslinked by electron beam irradiation, the blending amount of the crosslinking aid can be omitted from the explanatory variable data 61. In other words, the explanatory variable data 61 needs to include at least information on the blending amounts of the base polymer, the flame retardant, and the flame retardant aid (base polymer blending amount data 11a and filler blending amount data 11b), and does not need to include information on the blending amounts of other materials (other material blending amount data 11c).

[0028] 3, for example, when the physical property predicted by the physical property prediction processing unit 23 is the "initial elongation" to be predicted, the data extraction processing unit 21 extracts data so as to include the base polymer blend amount data 11a, the filler blend amount data 11b, the information on the blend amounts of the crosslinking assistant and the colorant in the other material blend amount data 11c, and the irradiation dose data 13 in the explanatory variable data 61, and excludes the information on the blend amounts of the antioxidant, the copper damage inhibitor, and the lubricant in the other material blend amount data 11c. Then, only the data on the initial elongation to be predicted is extracted as the objective variable data 62 from the physical property data 14.

[0029] When the physical property to be predicted is not the initial elongation or tensile strength, the data extraction processing unit 21 may include information on the blending amount of any one of the antioxidant, copper inhibitor, and lubricant in the explanatory variable data 61. For example, when the physical property to be predicted is the deterioration characteristic of the elongation or tensile strength of the resin composition under a specific environment (elongation or tensile strength after deterioration, or the rate of decrease in elongation or tensile strength after deterioration relative to the initial elongation or tensile strength, etc.), the influence of other materials may be large. In such a case, the explanatory variable data 61 may include information on the blending amount of one or more of the antioxidant, copper inhibitor, and lubricant, in addition to information on the blending amounts of the base polymer, flame retardant, and flame retardant assistant.

[0030] (Regression model creation processing unit 22) The regression model creation processing unit 22 performs machine learning using the learning data 6a (explanatory variable data 61 and dependent variable data 62) extracted by the data extraction processing unit 21, and performs a regression model creation process (see Figure 7(a)) to create a regression model 7.

[0031] As shown in FIG. 4(a), explanatory variable data 61 and target variable data 62 (the same as the “learning data 6a” shown on the right side of FIG. 3) extracted by the data extraction processing unit 21 are input to the regression model creation processing unit 22.

[0032] The regression model creation processing unit 22 includes software such as a learning algorithm for learning by itself, by machine learning, the correlation between each parameter of the input explanatory variable data 61 and the parameters of the response variable data 62. The learning algorithm is not particularly limited, and a known learning algorithm can be used, for example, so-called deep learning using a neural network having three or more layers can be used. What the regression model creation processing unit 22 learns corresponds to a model structure that represents the correlation between the explanatory variable data 61 and the response variable data 62.

[0033] When the physical property to be predicted is initial elongation or tensile strength, the regression model creation processing unit 22 creates a regression model 7 that represents the correlation between the base polymer blend amount data 11a, the filler blend amount data 11b, the information on the blend amounts of crosslinking aid and colorant in the other material blend amount data 11c, and the irradiation dose data 13 input as explanatory variable data 61, and the physical property data 14 (initial elongation or tensile strength) input as objective variable data 62.

[0034] More specifically, the regression model creation processing unit 22 repeatedly executes learning based on a data set including explanatory variable data 61 and response variable data 62, based on the input learning data 6a, and automatically interprets the correlation between the two. Note that the correlation is unknown at the start of learning, but as learning progresses, the correlation between the explanatory variable data 61 and the response variable data 62 is gradually interpreted, and by using the regression model 7, which is a trained model obtained as a result, it becomes possible to interpret the correlation between the explanatory variable data 61 and the response variable data 62.

[0035] The regression model creation processing unit 22 stores the created regression model 7 in the storage unit 3. In this embodiment, the regression model creation processing unit 22 updates the regression model 7 every time the learning data 6a is updated. However, this is not limiting, and for example, when executing a physical property prediction process described later, the regression model 7 may be updated by learning the updates of the learning data 6a collectively.

[0036] (Physical property prediction processing unit 23) The physical property prediction processing unit 23 performs a physical property prediction process (see FIG. 7(b)) that predicts the physical property data 14 according to the prediction source data 8, which is the source of the prediction, using the regression model 7. Hereinafter, the blending amount data 11 used as the prediction source data 8 is referred to as prediction source blending amount data 11d. The physical property prediction processing unit 23 stores the predicted physical property data 14 as prediction data 9 in the storage unit 3. Hereinafter, the predicted physical property data 14 is referred to as predicted physical property data 14a.

[0037] As shown in FIG. 4(b), in the physical property prediction process, the regression model 7 and the prediction source data 8 are input to the physical property prediction processing unit 23. Although not shown, in this embodiment, the prediction source data 8 is used as the prediction source data 8 in addition to the prediction source blending amount data 11d. The physical property prediction processing unit 23 obtains the physical property data 14 corresponding to the prediction source data 8 using the regression model 7, and stores the obtained physical property data 14 (predicted physical property data 14a) in the storage unit 3 as the prediction data 9. The prediction data 9 obtained here represents the predicted physical property (for example, initial elongation or tensile strength) of the resin composition when the resin composition is manufactured using the material and the irradiation amount of the prediction source data 8.

[0038] (Predicted physical properties display processing unit 24) The predicted physical property display processing unit 24 performs a predicted physical property display process to display the predicted data 9. In the predicted physical property display process, for example, the predicted data 9 is displayed on the display device 4. Note that the predicted physical property display process may be configured to also display data other than the predicted data 9, for example, items used as the explanatory variable data 61, etc.

[0039] (Method for predicting physical properties of resin composition) (Main routine) Fig. 5 is a flow diagram of a method for predicting physical properties of a resin composition according to the present embodiment. In Fig. 5, solid arrows indicate a flow of control, and dashed arrows indicate input and output of signals and data.

[0040] 5, in the method for predicting physical properties of a resin composition according to the present embodiment, first, in step S1, the control unit 2 determines whether new data has been input. If the determination in step S1 is NO (N), the process proceeds to step S5. If the determination in step S1 is YES (Y), the process proceeds to step S2, where data acquisition processing is performed.

[0041] In the data acquisition process of step S2, as shown in Fig. 6(a), the data acquisition processing unit 20 receives various data (mixture amount data 11, irradiation amount data 13, and physical property data 14) (step S21), registers the received various data in the database 6, and stores them in the storage unit 3 (step S22). Then, the process returns.

[0042] After the data acquisition process of step S2, a data extraction process is performed in step S3. In the data extraction process of step S3, as shown in FIG. 6(b), the data acquisition processing unit 20 judges whether the physical property to be predicted is initial elongation or tensile strength (step S31). The physical property to be predicted may be specified, for example, by the input device 5. If the result of the judgment in step S31 is YES (Y), in step S32, the data acquisition processing unit 20 extracts the base polymer blend amount data 11a, the filler blend amount data 11b, the blend amounts of the crosslinking assistant and the colorant in the other material blend amount data 11c, and the irradiation amount data 13 from the database 6 as explanatory variable data 61. At this time, the blend amounts of the antioxidant, the copper damage inhibitor, and the lubricant in the other material blend amount data 11c are excluded. Then, the process proceeds to step S34.

[0043] If the result of the determination in step S31 is NO (N), in step S33, the data acquisition processing unit 20 extracts explanatory variable data 61 from the database 6 according to the physical property of the prediction target. For example, when the physical property of the prediction target is the deterioration characteristic of elongation or tensile strength, at least the antioxidant, copper inhibitor, and lubricant in the other material blending amount data 11c are extracted as the explanatory variable data 61. The extraction may be performed so that each of the explanatory variable data 61 and the objective variable data 62 includes one blending amount. Then, the process proceeds to step S34. In step S34, the data acquisition processing unit 20 extracts the physical property data 14 of the prediction target from the database 6 as the objective variable data 62. Although not shown, the extracted explanatory variable data 61 and the objective variable data 62 may be associated with each other to generate learning data 6a, which may be stored in the storage unit 3. Then, the process returns and proceeds to step S4 in FIG. 5.

[0044] In step S4, a regression model creation process is performed. In the regression model creation process, as shown in Fig. 7(a), first, in step S41, the regression model creation processing unit 22 uses the unlearned learning data 6a (the explanatory variable data 61 and the target variable data 62 extracted in step S3) for machine learning to update the regression model 7. Note that in step S41, if the regression model 7 has not been created, a new regression model 7 is created. Thereafter, in step S42, the updated (or created) regression model 7 is stored in the storage unit 3, and the process returns.

[0045] When predicting the physical properties of a resin composition, the prediction source data 8 is inputted by the input device 5 or the like (step S10). Note that the data to be the prediction source data 8 (blending amount data 11 and irradiation dose data 13) may be inputted in advance to the resin composition physical property prediction device 1, and the input device 5 may be configured to select the data to be used as the prediction source data 8.

[0046] In step S5, the control unit 2 judges whether or not the prediction base data 8 has been input. If the judgment in step S5 is NO, the process returns (returns to step S1). If the judgment in step S5 is YES, the process proceeds to step S6.

[0047] In step S6, a physical property prediction process is performed. In the physical property prediction process, as shown in Fig. 7(b), first, in step S61, the physical property prediction processing unit 23 predicts physical property data 14 (predicted physical property data 14a) corresponding to the prediction source data 8 using the regression model 7, and sets it as predicted data 9. Then, in step S62, the obtained predicted data 9 is stored in the storage unit 3. Then, the process returns and proceeds to step S7 in Fig. 5.

[0048] In step S7, a predicted physical property display process is performed. In the predicted physical property display process, for example, the predicted data 9 predicted in step S6 is displayed on the display 4, thereby displaying the predicted data 9. Then, the process returns (returns to step S1).

[0049] (Consideration of explanatory variables suitable for predicting initial elongation or tensile strength) A study was conducted on explanatory variables suitable for predicting initial elongation or tensile strength. In the study, as shown in FIG. 8(a), data including 17 types of base polymers, 17 types of fillers (flame retardants and flame retardant assistants), 2 types of antioxidants, 1 type of copper inhibitor, 5 types of lubricants, 1 type of colorant, 1 type of crosslinking assistant, and irradiation dose data was used as learning data 6a, and the change in prediction accuracy of the regression model 7 when the explanatory variable data 61 used was reduced by cross-validation was examined. In the study, 70% of the total data was divided into learning data 6a and the remaining 30% into test data, and the regression model 7 was created using the divided learning data 6a, and the test data was evaluated with the created regression model 7, and the mean absolute error (MAPE) was calculated. This division of data, creation of the regression model 7, and evaluation were repeated 30 times, and the average value of the mean absolute error (MAPE) was calculated. It can be said that the smaller the mean absolute error, the more excellent the regression model 7 with high prediction accuracy has been created. The physical properties to be predicted were initial elongation and tensile strength.

[0050] The following examples were examined: Example 1, in which all blending amount information was used as explanatory variable data 61; Example 2, in which some (three out of five) of the antioxidants, copper inhibitors, and lubricants were omitted from the explanatory variable data 61; and Example 3, in which all of the antioxidants, copper inhibitors, and lubricants were omitted from the explanatory variable data 61. In FIG. 8(a), data excluded from the explanatory variable data 61 in Example 2 is shown as Data Set 1, and data excluded from the explanatory variable data 61 in Example 3 is shown as Data Set 2. The average values ​​of the mean absolute error rates obtained by the examination are all shown in FIG. 8(b).

[0051] 8(b), compared with Examples 1 and 2, Example 3 had the smallest average absolute error rate, and it was found that a regression model 7 with high prediction accuracy was obtained. From this result, it was confirmed that when the physical property to be predicted is initial elongation or tensile strength, the prediction accuracy can be improved by omitting the antioxidant and copper inhibitor from the explanatory variable data 61, and the prediction accuracy can be further improved by omitting the antioxidant, copper inhibitor, and lubricant from the explanatory variable data 61.

[0052] (Functions and Effects of the Embodiments) As described above, the physical property prediction device 1 according to this embodiment includes a regression model creation processing unit 22 that performs machine learning on the relationship between the explanatory variable data 61 and the physical property data 14 and creates a regression model 7 that represents the correlation between the explanatory variable data 61 and the physical property data 14, and a physical property prediction processing unit 23 that predicts the physical properties of the prediction target using the regression model 7. When the physical property predicted by the physical property prediction processing unit 23 is the initial elongation or tensile strength of the prediction target, the explanatory variable data 61 includes information on the respective blend amounts of the base polymer, the flame retardant, and the flame retardant auxiliary.

[0053] The prediction accuracy can be improved by using information on the amounts of base polymer, flame retardant, and flame retardant aid, which are mixed in relatively large amounts, as explanatory variables. Furthermore, by excluding information on the amounts of antioxidant, copper inhibitor, and lubricant mixed in from the explanatory variable data 61, it becomes possible to predict the initial elongation or tensile strength with higher accuracy.

[0054] (Summary of the embodiment) Next, the technical ideas grasped from the above-described embodiment will be described by using the reference numerals and the like in the embodiment. However, the reference numerals and the like in the following description do not limit the components in the claims to the members and the like specifically shown in the embodiment.

[0055] [1] A physical property prediction device (1) for predicting physical properties of a resin composition produced using a base polymer, a flame retardant, a flame retardant auxiliary, and other materials, the device comprising: a regression model creation processing unit (22) that performs machine learning to learn a relationship between explanatory variable data (61) and physical property data (14) including information on physical properties of a target to be predicted, and creates a regression model (7) that expresses a correlation between the explanatory variable data (61) and the physical property data (14); and a physical property prediction processing unit (23) that predicts the physical properties of the target to be predicted using the regression model (7). When the physical property predicted by the physical property prediction processing unit (23) is an initial elongation or tensile strength of the target to be predicted, the explanatory variable data (61) includes information on the respective blending amounts of the base polymer, the flame retardant, and the flame retardant auxiliary. The physical property prediction device (1) for predicting physical properties of a resin composition.

[0056] [2] The device (1) for predicting physical properties of a resin composition according to [1], wherein, in a case where the other materials include one or more of an antioxidant and a copper inhibitor, and the physical property predicted by the physical property prediction processing unit (23) is an initial elongation or tensile strength of a prediction target, the explanatory variable data (61) does not include information on the respective blending amounts of the antioxidant and the copper inhibitor.

[0057] [3] The device (1) for predicting physical properties of a resin composition described in [1], in which the other materials include one or more of an antioxidant, a copper inhibitor, and a lubricant, and the physical property predicted by the physical property prediction processing unit (23) is the initial elongation or tensile strength of the prediction target, the explanatory variable data (61) does not include information on the respective blending amounts of the antioxidant, the copper inhibitor, and the lubricant.

[0058] [4] The device (1) for predicting physical properties of a resin composition according to [3], comprising: a database including blending amount data (11) including information on the blending amounts of at least the base polymer, the flame retardant, the flame retardant auxiliary, and the other materials, and physical property data (14) including at least information on a physical property to be predicted; and a data extraction processing unit (21) for extracting the explanatory variable data (61) from the database (6) in accordance with a physical property to be predicted by the physical property prediction processing unit (23), wherein, when the physical property predicted by the physical property prediction processing unit (23) is an initial elongation or tensile strength of the prediction target, the data extraction processing unit (21) performs data extraction so that the explanatory variable data includes at least information on the blending amounts of the base polymer, the flame retardant, and the flame retardant auxiliary, and does not include information on the antioxidant, the copper inhibitor, and the lubricant.

[0059] [5] The device (1) for predicting physical properties of a resin composition according to [1], wherein the resin composition to be predicted is a halogen-free resin composition that is crosslinked by electron beam irradiation, and the explanatory variable data includes information on the amount of electron beam irradiation during crosslinking.

[0060] [6] The device (1) for predicting a physical property of a resin composition according to [5], wherein, in a case where the other materials include a crosslinking auxiliary, and the physical property predicted by the physical property prediction processing unit (23) is an initial elongation or tensile strength of a prediction target, the explanatory variable data (61) includes information on each blend amount of the crosslinking auxiliary.

[0061] [7] A method for predicting the physical properties of a resin composition produced using a base polymer, a flame retardant, a flame retardant auxiliary, and other materials, comprising: a regression model creation process for learning the relationship between explanatory variable data (61) and physical property data (14) including information on the physical properties of a target to be predicted by machine learning, and creating a regression model (7) that expresses the correlation between the explanatory variable data (61) and the physical property data (14); and a physical property prediction process for predicting the physical properties of the target to be predicted using the regression model (7). When the physical property predicted in the physical property prediction process is the initial elongation or tensile strength of the target to be predicted, the explanatory variable data (61) includes information on the respective blending amounts of the base polymer, the flame retardant, and the flame retardant auxiliary.

[0062] (Additional Note) Although the embodiment of the present invention has been described above, the invention according to the claims is not limited to the embodiment described above. It should be noted that not all of the combinations of features described in the embodiment are essential to the means for solving the problems of the invention. The present invention can be modified appropriately without departing from the spirit of the invention. [Explanation of symbols]

[0063] 1...Resin composition property prediction device (property prediction device) 2...Control section 3...Storage section 6. Database 6a…Learning data 7. Regression Model 8…Source data for prediction 9. Prediction data 11...Composition data 11a...Base polymer blend data 11b…Filler content data 11c…Other material content data 14...Physical property data 20...Data acquisition processing section 21...Data extraction processing section 22...Regression model creation processing section 23...Physical property prediction processing section 24…Predicted physical properties display processing unit 61...Explanatory variable data 62…Objective variable data

Claims

1. A property prediction device for predicting the properties of a resin composition produced using a base polymer, a flame retardant, a flame retardant assistant, and other materials, comprising: a regression model creation processing unit that performs machine learning on the relationship between explanatory variable data and physical property data including information on the physical properties of a prediction target, and creates a regression model that represents the correlation between the explanatory variable data and the physical property data; a property prediction processing unit that predicts a property of a prediction target using the regression model; A database including blending amount data including information on the blending amounts of at least the base polymer, the flame retardant, the flame retardant auxiliary, and the other materials, and physical property data including information on at least the physical properties of a prediction target; a data extraction processing unit that extracts the explanatory variable data from the database in accordance with the physical property predicted by the physical property prediction processing unit, The data extraction processing unit selects information related to the predicted physical property from the information of each blend amount included in the blend amount data, and performs data extraction so that the acquired first information is included in the explanatory variable data. A device for predicting the physical properties of resin compositions.

2. The database further includes at least one of data representing manufacturing conditions of the resin composition or data representing a state of the structure of the resin composition; the data extraction processing unit selects information related to the physical property to be predicted from among the data representing the manufacturing conditions and the data representing the state of the structure, and performs data extraction such that the acquired second information is further included in the explanatory variable data. The apparatus for predicting physical properties of a resin composition according to claim 1 .

3. When the physical property predicted by the physical property prediction processing unit is the initial elongation or tensile strength of the resin composition, the data extraction processing unit extracts at least extracting data so that information on the blending amounts of the flame retardant and the flame retardant auxiliary is included in the explanatory variable data as the first information; The apparatus for predicting physical properties of a resin composition according to claim 1 .

4. The resin composition to be predicted is a halogen-free resin composition produced by crosslinking through electron beam irradiation, the database includes information on the amount of electron beam irradiation during the crosslinking as data representing the production conditions, When the physical property predicted by the physical property prediction processing unit is the initial elongation or tensile strength of the non-halogen-based resin composition, the data extraction processing unit performs data extraction so that at least information on the blending amounts of the base polymer, the flame retardant, and the flame retardant auxiliary is included as the first information, and information on the irradiation amount of the electron beam during the crosslinking is included as the second information in the explanatory variable data. The apparatus for predicting physical properties of a resin composition according to claim 2.

5. A method for predicting physical properties of a resin composition produced using a base polymer, a flame retardant, a flame retardant assistant, and other materials, comprising: a regression model creation process in which a relationship between explanatory variable data and physical property data including information on the physical properties of a target object is learned by machine learning, and a regression model is created that represents a correlation between the explanatory variable data and the physical property data; A property prediction process for predicting a property of a prediction target using the regression model; a data acquisition process for registering in a database blending amount data including information on the blending amounts of at least the base polymer, the flame retardant, the flame retardant auxiliary, and the other materials, and physical property data including information on at least the physical properties of a prediction target; a data extraction process step of extracting the explanatory variable data from the database in accordance with the physical properties predicted by the physical property prediction process, The data extraction process selects and discards first information related to the predicted physical property from the information on each blend amount included in the blend amount data, and performs data extraction so that the acquired first information is included in the explanatory variable data. A method for predicting physical properties of a resin composition.

Images