Physical property prediction device, inference device, machine learning device, physical property prediction method, inference method, and machine learning method

Abstract

To provide a physical property prediction device capable of appropriately predicting physical property values of a molded body from characteristics of a raw material.SOLUTION: The information processing device 5 functioning as the physical property prediction device includes a data acquisition unit 5020 that acquires input data and a data generation unit 5021 that generates output data for the input data by inputting the input data acquired by the data acquisition unit 5020 to the learning model 16. The learning model 16 is a learned model in which a correlation between input data and output data is learned by machine learning, the input data includes molecular chain information 11 related to a molecular chain of a polymer compound that is a raw material of a molded article and crystallinity information 12 related to a crystalline portion of the polymer compound, and the output data includes physical property information 14 indicating a physical property value of a molded article obtained by melt-molding a solid raw material.SELECTED DRAWING: Figure 7

Description

[Technical Field]

[0001] The present invention relates to a property prediction device, an inference device, a machine learning device, a property prediction method, an inference method, and a machine learning method. [Background technology]

[0002] BACKGROUND ART Conventionally, molded articles of various shapes have been produced by melt molding polymer compounds such as polyethylene as raw materials (see, for example, Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Application Laid-Open No. 2014-231231 Summary of the Invention [Problem to be solved by the invention]

[0004] There are a wide variety of polymer compounds used as raw materials for molded bodies, and the characteristics of the polymer compounds vary depending on the type of raw material. Therefore, it has been difficult to predict the physical properties of molded bodies obtained by melt molding of the raw materials.

[0005] The present invention has been made to solve the above-mentioned problems, and aims to provide a physical property prediction device, an inference device, a machine learning device, a physical property prediction method, an inference method, and a machine learning method that enable appropriate prediction of the physical property values ​​of a molded body from the characteristics of the raw material. [Means for solving the problem]

[0006] In order to achieve the above object, a physical property prediction apparatus according to one aspect of the present invention includes: a data acquisition unit for acquiring input data; a data generation unit that generates output data for the input data by inputting the input data acquired by the data acquisition unit into a learning model; The learning model is a trained model that has been trained by machine learning to determine the correlation between the input data and the output data; The input data is molecular chain information on the molecular chains of the polymer compound that is the raw material for the molded body; and crystallinity information regarding the crystalline portion of the polymer compound; The output data is The information includes physical property information indicating the physical property values ​​of the molded body obtained by melt molding the solid raw material. [Effects of the Invention]

[0007] According to a physical property prediction device of one aspect of the present invention, by inputting input data including molecular chain information about the molecular chains of a polymer compound and crystallinity information about the crystalline portion of the polymer compound into a learning model, output data including physical property information indicating the physical property values ​​of a molded product is generated for the input data. Therefore, the physical property values ​​of a molded product can be appropriately predicted from the characteristics of the raw materials.

[0008] Problems, configurations, and effects other than those described above will become apparent from the detailed description of the invention that follows. [Brief explanation of the drawings]

[0009] [Figure 1] 1 is an overall schematic diagram showing an example of a physical property prediction system 1. FIG. [Figure 2] FIG. 2 is a block diagram showing an example of an information processing device 5. [Figure 3] FIG. 5 is a data configuration diagram showing an example of an information management database 510. [Figure 4A] FIG. 2 is a data structure diagram showing an example of molecular chain information 11 and crystallinity information 12. [Figure 4B] 1 is a data structure diagram showing an example of raw material information 10, process information 13, and physical property information 14. FIG. [Figure 5] FIG. 5 is a functional explanatory diagram showing an example of a learning function 501. [Figure 6A]FIG. 1 is a diagram showing first learning data 15-1 and a first learning model 16-1. [Figure 6B] FIG. 10 is a diagram showing second learning data 15-2 and a second learning model 16-2. [Figure 6C] FIG. 10 is a diagram showing third learning data 15-3 and a third learning model 16-3. [Figure 6D] FIG. 10 is a diagram showing fourth learning data 15-4 and a fourth learning model 16-4. [Figure 6E] FIG. 10 is a diagram showing fifth learning data 15-5 and a fifth learning model 16-5. [Figure 6F] FIG. 10 is a diagram showing sixth learning data 15-6 and a sixth learning model 16-6. [Figure 6G] FIG. 10 is a diagram showing seventh learning data 15-7 and a seventh learning model 16-7. [Figure 6H] FIG. 10 is a diagram showing eighth learning data 15-8 and an eighth learning model 16-8. [Figure 6I] FIG. 10 is a diagram showing ninth learning data 15-9 and a ninth learning model 16-9. [Figure 6J] FIG. 16 is a diagram showing tenth learning data 15-10 and a tenth learning model 16-10. [Figure 6K] FIG. 11 is a diagram showing eleventh learning data 15-11 and an eleventh learning model 16-11. [Figure 7] FIG. 5 is a functional explanatory diagram showing an example of a property prediction function 502. [Figure 8] FIG. 9 is a hardware configuration diagram showing an example of a computer 900. [Figure 9] 10 is a flowchart showing an example of a machine learning method by a learning function 501. [Figure 10] 10 is a flowchart showing an example of a property prediction method by the property prediction function 502. [Figure 11] FIG. 2 is a screen configuration diagram showing an example of a property prediction input screen 17. [Figure 12] FIG. 2 is a screen configuration diagram showing an example of a physical property value output screen 18. DETAILED DESCRIPTION OF THE INVENTION

[0010] Hereinafter, an embodiment for carrying out the present invention will be described with reference to the drawings. The scope necessary for the explanation to achieve the object of the present invention will be schematically shown, and the scope necessary for explaining the relevant part of the present invention will be mainly explained, and the parts that are omitted from the explanation will be based on publicly known techniques.

[0011] 1 is an overall schematic diagram showing an example of a physical property prediction system 1. The physical property prediction system 1 is a system for predicting the physical property values ​​of a molded body obtained by melt molding a solid raw material.

[0012] The raw material is solid and has been processed into, for example, pellets. The raw material is a polymer compound, and examples thereof include thermoplastic resin materials such as polyolefin resins such as polyethylene and polypropylene, polyester resins such as polyethylene terephthalate, polystyrene resins, polyamide resins, acrylic resins, polyvinyl chloride resins, polyacrylonitrile resins, polyvinyl alcohol resins, and biodegradable resins. The raw material contains a large number of molecular chains of the polymer compound, and has crystalline portions in which the molecular chains are regularly arranged and amorphous portions in which the molecular chains are irregularly arranged. In this embodiment, the raw material is mainly polyethylene.

[0013] The raw materials may be a mixture of multiple types of raw materials in a predetermined blending ratio, or may be recycled materials that have been recycled. Various pieces of information about the raw materials are recorded as raw material information 10 and used in the physical property prediction system 1.

[0014] The molded article is a molded piece having a predetermined shape or a film having a predetermined thickness. The molded piece is, for example, rod-shaped or plate-shaped, and is mainly molded by injection molding or compression molding. The shape of the molded piece is preferably a shape suitable for analysis or testing, but it may also be molded by blow molding or the like, adopting a shape similar to that of the final product, such as a bottle shape. The film is, for example, membrane-shaped, and is mainly molded by extrusion molding. Furthermore, the molded piece or film may be molded by multiple molding processes.

[0015] The physical property prediction system 1 includes an analysis device 2, a molding device 3, a test device 4, an information processing device 5, and a user terminal device 6. Each of the devices 2 to 6 is configured, for example, as a general-purpose or dedicated computer (see FIG. 8 described later), and is connected to a wired or wireless network 7 so that various data can be transmitted and received between them. Note that the number of the devices 2 to 6 and the connection configuration of the network 7 are not limited to the example in FIG. 1 and may be changed as appropriate.

[0016] The analyzer 2 is a device that analyzes raw materials. A plurality of types of analyzers 2 (2A to 2C) that use different analysis methods are used as the analyzer 2. The plurality of types of analyzers 2 used include a melt analyzer 2A that performs melt analysis of raw materials, a solution analyzer 2B that performs solution analysis of raw materials, and a solid analyzer 2C that performs solid analysis of raw materials.

[0017] The melt analyzer 2A is an apparatus that analyzes the characteristics of raw materials in a molten state. Examples of the melt analyzer 2A include a melt flow rate (MFR) analyzer, a viscometer, and a differential scanning calorimetry (DSC) analyzer. The solution analyzer 2B is an apparatus that analyzes the characteristics of raw materials using a solution in which the raw materials are dissolved in a solvent. Examples of the solution analyzer 2B include a gel permeation chromatography (GPC) analyzer and a nuclear magnetic resonance (NMR) analyzer. The solid analyzer 2C is an apparatus that analyzes the characteristics of raw materials in a solid state. Examples of the solid analyzer 2C include a dry densitometer and a density gradient tube analyzer.

[0018] The analysis results obtained by the analysis device 2 are recorded as molecular chain information 11 relating to the molecular chains of the polymer compound and crystallinity information 12 relating to the crystalline portion of the polymer compound, and are used in the property prediction system 1. The analysis conditions for analyzing raw materials are set in accordance with standards such as JIS, and the analysis conditions may be recorded in the molecular chain information 11 and the crystallinity information 12.

[0019] The molding device 3 is a device that performs a molding process in which raw materials are melt-molded into a molded body. A plurality of types of molding devices 3 that perform different molding processes are used. The plurality of types of molding devices 3 that are used include a molded piece molding device 3A that melt-moldes raw materials into molded pieces and a film molding device 3B that melt-moldes raw materials into a film.

[0020] Various information related to the molding process by the molding device 3 is recorded as process information 13 and used in the physical property prediction system 1. In this case, various information related to the molding process by the molded piece molding device 3A is recorded as molded piece process information 13A. Various information related to the molding process by the film molding device 3B is recorded as film process information 13B.

[0021] The test equipment 4 is used to test the durability and heat resistance of molded articles. Multiple types of test equipment 4 with different test methods are used. The multiple types of test equipment 4 include a molded piece test equipment 4A for testing molded pieces and a film test equipment 4B for testing films. The molded piece test equipment 4A tests molded pieces for environmental stress crack resistance (ESCR), flexural modulus, tensile modulus, Charpy impact strength, tensile elongation at break, Vicat softening temperature, strength characteristics, content characteristics, usability, sealability, etc. The film test equipment 4B tests films for impact burst strength, tensile impact strength, tensile modulus, heat shrinkage, tear strength, loop stiffness, drop characteristics, openability, storage stability, usability, line suitability, etc.

[0022] Various information regarding the test results obtained by the testing device 4 is recorded as physical property information 14 indicating the physical property values ​​of the molded body, and is used in the physical property prediction system 1. In this case, various information regarding the test results of the molded piece obtained by the molded piece testing device 4A is recorded as molded piece physical property information 14A. Various information regarding the test results of the film obtained by the film molding device 3B is recorded as film physical property information 14B. Note that the test conditions for testing the molded body are set in accordance with standards such as JIS, and the test conditions may be recorded in the molded piece physical property information 14A and the molded piece physical property information 14B.

[0023] The information processing device 5 includes an information management database 510 that can register various pieces of information 10 to 14 (raw material information 10, molecular chain information 11, crystallinity information 12, process information 13, and physical property information 14) obtained in the past when raw materials were analyzed and molded bodies obtained by melt molding the raw materials were tested, in association with each other. The information processing device 5 then generates a learning model 16 by performing machine learning based on the various pieces of information 10 to 14 registered in the information management database 510.

[0024] In addition, the information processing device 5 uses a learning model 16 to predict, as physical property information 14, the physical property values ​​of a new molded body obtained by melt-molding the new raw material, based on input data including raw material information 10, molecular chain information 11, and crystallinity information 12 regarding the new raw material to be predicted, and process information 13 when melt-molding the new raw material.

[0025] The user terminal device 6 is a device used by users such as operators who perform analysis, melt molding, and testing, and administrators of the physical property prediction system 1. Various information management programs, a web browser, and the like are installed on the user terminal device 6, and the user terminal device 6 accepts various input operations and outputs various information via a display screen or voice. The user terminal device 6 is used, for example, for users to input raw material information 10 as input data and to output physical property information 14 (output data) which is the prediction result of the information processing device 5.

[0026] (Configuration of information processing device 5) 2 is a block diagram showing an example of the information processing device 5. The information processing device 5 includes a control unit 50 configured with a processor or the like, a storage unit 51 configured with an HDD, an SSD, a memory or the like, a communication unit 52 which is a communication interface with the network 7 and external devices, an input unit 53 configured with a keyboard, a mouse or the like, and a display unit 54 configured with a display or the like. Note that the input unit 53 and the display unit 54 may be omitted.

[0027] The storage unit 51 stores an information management database 510, a trained model management database 511, and an information processing program 512, as well as an operating system, other programs, data, etc.

[0028] The control unit 50 executes an information processing program 512 stored in the storage unit 51 to realize an information management function 500, a learning function 501, and a physical property prediction function 502. The control unit 50 includes, as the units that realize the learning function 501, a learning data acquisition unit 5010 and a machine learning unit 5011. The control unit 50 includes, as the units that realize the physical property prediction function 502, a data acquisition unit 5020, a data generation unit 5021, and an output processing unit 5022.

[0029] The data configuration of each of the functions 500 to 502 and each of the databases 510 and 511 will be described below.

[0030] (Information management function 500) The control unit 50 of the information processing device 5 implements the information management function 500 using the information management database 510. The information processing device 5 is provided with the information management function 500, so that the information processing device 5 operates as an information management device.

[0031] Fig. 3 is a data structure diagram showing an example of the information management database 510. Fig. 4A is a data structure diagram showing an example of molecular chain information 11 and crystallinity information 12. Fig. 4B is a data structure diagram showing an example of raw material information 10, process information 13, and physical property information 14.

[0032] The information management database 510 has multiple records for each management ID for associating various pieces of information 10 to 14 handled by the physical property prediction system 1. The management ID is information (product number, model number, etc.) for identifying raw materials and molded bodies. Each record has fields in which raw material information 10, molecular chain information 11, crystallinity information 12, process information 13, and physical property information 14 can be registered, for example.

[0033] For example, when the information management function 500 receives raw material information 10 from the user terminal device 6, molecular chain information 11 and crystallinity information 12 from the analysis device 2, process information 13 from the molding device 3, or physical property information 14 from the testing device 4, the information management function 500 registers the various information 10 to 14 in the information management database 510. Furthermore, when the information management function 500 receives physical property information 14 as a processing result of the physical property prediction function 502, the information management function 500 registers the physical property information 14 in the information management database 510.

[0034] The various pieces of information 10 to 14 registered in the information management database 510 can be referenced from the user terminal device 6. Furthermore, editing operations such as adding, deleting, and modifying the various pieces of information 10 to 14 can be performed on a display screen displayed by an information management program or a web browser on the user terminal device 6. When the information management function 500 receives the results of the editing operations of the various pieces of information 10 to 14 from the user terminal device 6, it registers the results of the operations in the information management database 510. If the analytical device 2, the molding device 3, and the testing device 4 do not have a communication function, the user can use the user terminal device 6 to perform editing operations to add molecular chain information 11, crystallinity information 12, process information 13, and physical property information 14, and the information management function 500 can register the results of the editing operations in the information management database 510 when it receives the results of the editing operations from the user terminal device 6.

[0035] As shown in FIGS. 3 and 4B, the raw material information 10 includes information on the type of raw material, information on the blending ratio of the raw materials, information on the degree of deterioration of the raw materials, etc. The information on the type of raw material specifies the raw material name and grade name. For example, if the type of polyethylene is to be registered, the raw material name is selected from multiple options: HDPE, LDPE, and LLDPE. The information on the blending ratio of the raw materials is the blending ratio of each raw material contained in a mixed material in which multiple types of raw materials are mixed. The information on the degree of deterioration of the raw material indicates the degree of deterioration of the recycled material and is specified, for example, by the amount of chemiluminescence light emission or a carbonyl group-derived peak measured by a Fourier transform infrared spectrophotometer (FT-IR). Note that when the raw material is a mixed material in which multiple types of raw materials are mixed, the type of raw material and the degree of deterioration may be specified for each type of raw material contained in the mixed material.

[0036] As shown in Figures 3 and 4A, the molecular chain information 11 includes melt analysis molecular chain information 11A, which indicates information obtained by performing melt analysis on molecular chains on raw materials using melt analysis apparatus 2A, and solution analysis molecular chain information 11B, which indicates information obtained by performing solution analysis on molecular chains on raw materials using solution analysis apparatus 2B.

[0037] The melt analysis molecular chain information 11A is information on melt viscosity, etc. The information on melt viscosity is, for example, the analysis result by an MFR analyzer or the analysis result by a viscometer, for example, viscosity value for each shear rate.

[0038] The solution analysis molecular chain information 11B includes information on molecular weight, information on molecular weight distribution, information on branching structure, etc. The information on molecular weight and molecular weight distribution is the analysis result obtained by a GPC analyzer. The information on branching structure is the analysis result obtained by an NMR analyzer.

[0039] As shown in Figures 3 and 4A, the crystallinity information 12 includes melt analysis crystallinity information 12A, which indicates information obtained by performing melt analysis on the crystalline portion of the raw material using melt analysis apparatus 2A, and solid analysis crystallinity information 12B, which indicates information obtained by performing solid analysis on the crystalline portion of the raw material using solid analysis apparatus 2C.

[0040] The melting analysis crystallinity information 12A is information on the degree of crystallinity, the crystallization rate, the crystal thickness, the melting point, etc. The information on the degree of crystallinity, the crystallization rate, the crystal thickness, or the melting point is the analysis result obtained by a DSC analyzer.

[0041] The solid analysis crystallinity information 12B is information about density, etc. The information about density is the result of analysis using a dry densitometer or a density gradient tube analyzer.

[0042] As shown in Figures 3 and 4B, the process information 13 includes molded piece process information 13A related to the molding process when raw materials are melt-molded into molded pieces using a molded piece molding device 3A, and film process information 13B related to the molding process when raw materials are melt-molded into a film using a film molding device 3B.

[0043] The piece process information 13A is the setting conditions of each part when the piece forming device 3A performs the forming process. The film process information 13B is the setting conditions of each part when the film forming device 3B performs the forming process.

[0044] As shown in Figures 3 and 4B, the physical property information 14 includes molded piece physical property information 14A that indicates the physical property values ​​of the molded piece when the raw material is melt-molded into a molded piece, and film physical property information 14B that indicates the physical property values ​​of the film when the raw material is melt-molded into a film.

[0045] Molded piece physical property information 14A is obtained as a test result of molded piece testing device 4A or as a predicted result by learning model 16. Film physical property information 14B is obtained as a test result of film testing device 4B or as a predicted result by learning model 16.

[0046] Molded piece property information 14A includes environmental stress crack resistance, flexural modulus, tensile modulus, Charpy impact strength, tensile elongation at break, Vicat softening temperature, strength characteristics, content characteristics, usability, sealability, etc. Film property information 14B includes impact burst strength, tensile impact strength, tensile modulus, heat shrinkage, tear strength, loop stiffness, drop characteristics, openability, storage stability, usability, line suitability, etc.

[0047] Each value included in the various information 10-14 may be registered as a numerical value, as a stage value selected from a plurality of stage values, or as an option selected from a plurality of options. In this case, each value included in the various information 10-14 may be a specific value or a value having a specific range such as an upper limit or a lower limit. Furthermore, when the raw material is a mixture of multiple types of raw materials, molecular chain information 11 and crystallinity information 12 regarding each raw material may be registered for each type of raw material included in the mixture, or molecular chain information 11 and crystallinity information 12 regarding the mixture may be registered.

[0048] (Learning function 501) Fig. 5 is a functional explanatory diagram showing an example of the learning function 501. Figs. 6A to 6K are diagrams showing first to eleventh learning data 15-1 to 15-11 and first to eleventh learning models 16-1 to 16-11, respectively. The learning data acquisition unit 5010 and the machine learning unit 5011 of the information processing device 5 realize the learning function 501 using an information management database 510 and a trained model management database 511 (trained model storage unit). By providing the learning function 501 in the information processing device 5, the information processing device 5 operates as a machine learning device.

[0049] The learning model 16 stored in the learned model management database 511 is a learned model that has been trained by machine learning to determine the correlation between input data including at least molecular chain information 11 and crystallinity information 12 and output data including physical property information 14, as shown in Figure 5. That is, when input data is input, the learning model 16 outputs output data for the input data. Note that the input data may further include raw material information 10 and process information 13 in addition to the molecular chain information 11 and crystallinity information 12, as shown in Figure 5, or may further include either the raw material information 10 or the process information 13.

[0050] The raw material information 10 included in the input data is at least one of information on the type of raw material, information on the blending ratio of the raw materials, and information on the deterioration degree of the raw materials. The details of the raw material information 10 included in the input data are the same as those shown in FIG. 4B.

[0051] The molecular chain information 11 included in the input data is at least one of melt analysis molecular chain information 11A and solution analysis molecular chain information 11B. In this case, the melt analysis molecular chain information 11A is information about melt viscosity. Furthermore, the solution analysis molecular chain information 11B is at least one of information about molecular weight, information about molecular weight distribution, information about branching structure, etc. Details of the melt analysis molecular chain information 11A and solution analysis molecular chain information 11B included in the input data are the same as those shown in FIG. 4A.

[0052] The crystallinity information 12 included in the input data is at least one of melt analysis crystallinity information 12A and solid analysis crystallinity information 12B. In this case, the melt analysis crystallinity information 12A is information regarding the crystallinity, crystallization rate, crystal thickness, or melting point. Furthermore, the solid analysis crystallinity information 12B is information regarding density. Details of the melt analysis crystallinity information 12A and solid analysis crystallinity information 12B included in the input data are the same as those shown in FIG. 4A.

[0053] In addition, when the raw material is a mixture of multiple types of raw materials, the input data may include molecular chain information 11 and crystallinity information 12 regarding each raw material contained in the mixture, or may include molecular chain information 11 and crystallinity information 12 regarding the mixture.

[0054] The process information 13 included in the input data is at least one of molded piece process information 13A and film process information 13B. In this case, the molded piece process information 13A is the setting conditions of each part when the molded piece molding device 3A performs the molding process. Also, the film process information 13B is the setting conditions of each part when the film molding device 3B performs the molding process. Details of the molded piece process information 13A and film process information 13B included in the input data are the same as those shown in FIG. 4B.

[0055] The physical property information 14 included in the output data is at least one of molded piece physical property information 14A and film physical property information 14B. The molded piece physical property information 14A is at least one of environmental stress crack resistance, flexural modulus, tensile modulus, Charpy impact strength, tensile elongation at break, Vicat softening temperature, strength characteristics, content characteristics, usability, and sealability. The film physical property information 14B is at least one of impact burst strength, tensile impact strength, tensile modulus, thermal shrinkage, tear strength, loop stiffness, drop characteristics, openability, storage stability, usability, and line suitability.

[0056] In this embodiment, the learned model management database 511 stores, as specific examples of the learning model 16, the first to eleventh learning models 16-1 to 16-11 that have been machine-learned using the first to eleventh learning data 15-1 to 15-11 shown in Figures 6A to 6K.

[0057] The first learning model 16-1 takes solution analysis molecular chain information 11B, melt analysis crystallinity information 12A, and solid analysis crystallinity information 12B as input data, and outputs environmental stress crack resistance (molded piece physical property information 14A) as output data.

[0058] The second learning model 16-2 takes melt analysis molecular chain information 11A, solution analysis molecular chain information 11B, melt analysis crystallinity information 12A, and solid analysis crystallinity information 12B as input data and outputs bending modulus (molded piece physical property information 14A) as output data.

[0059] The third learning model 16-3 receives melt analysis molecular chain information 11A, solution analysis molecular chain information 11B, and solid analysis crystallinity information 12B as input data, and outputs Charpy impact strength (molded piece physical property information 14A) as output data.

[0060] The fourth learning model 16-4 takes melt analysis molecular chain information 11A, solution analysis molecular chain information 11B, melt analysis crystallinity information 12A, and solid analysis crystallinity information 12B as input data and outputs tensile breaking elongation (molded piece physical property information 14A) as output data.

[0061] The fifth learning model 16-5 takes the solution analysis molecular chain information 11B, the melt analysis crystallinity information 12A, and the solid analysis crystallinity information 12B as input data, and outputs the Vicat softening temperature (molded piece physical property information 14A) as output data.

[0062] The sixth learning model 16-6 takes melt analysis molecular chain information 11A, solution analysis molecular chain information 11B, melt analysis crystallinity information 12A, and solid analysis crystallinity information 12B as input data and outputs impact burst strength (film physical property information 14B) as output data.

[0063] The seventh learning model 16-7 takes melt analysis molecular chain information 11A, solution analysis molecular chain information 11B, melt analysis crystallinity information 12A, and solid analysis crystallinity information 12B as input data and outputs tensile impact strength (film physical property information 14B) as output data.

[0064] The eighth learning model 16-8 takes melt analysis molecular chain information 11A, solution analysis molecular chain information 11B, melt analysis crystallinity information 12A, and solid analysis crystallinity information 12B as input data and outputs tear strength (film physical property information 14B) as output data.

[0065] The ninth learning model 16-9 receives the solution analysis molecular chain information 11B, the melt analysis crystallinity information 12A, and the solid analysis crystallinity information 12B as input data, and outputs the loop stiffness (film physical property information 14B) as output data.

[0066] The tenth learning model 16-10 takes as input data at least one of melt analysis molecular chain information 11A, solution analysis molecular chain information 11B, melt analysis crystallinity information 12A, and solid analysis crystallinity information 12B, and outputs as output data molded piece physical property information 14A, which is at least one of strength characteristics, content characteristics, usability, and sealability.

[0067] The 11th learning model 16-11 takes as input data at least one of melt analysis molecular chain information 11A, solution analysis molecular chain information 11B, melt analysis crystallinity information 12A, and solid analysis crystallinity information 12B, and outputs as output data film physical property information 14B, which is at least one of drop characteristics, openability, storage properties, usability, and line suitability.

[0068] The learning model 16 (first to eleventh learning models 16-1 to 16-11) employs, for example, a neural network structure, and includes an input layer 160, an intermediate layer 161, and an output layer 162, as shown in FIG. 5. Synapses (not shown) that connect the neurons are laid between the layers, and each synapse is associated with a weight. A weight parameter group consisting of the weights of each synapse is adjusted by a machine learning algorithm such as backpropagation.

[0069] The input layer 160 has neurons whose number corresponds to the input data, and each value is input to each neuron. The output layer 162 has neurons whose number corresponds to the output data, and the output data is output as the inference result.

[0070] The training data acquisition unit 5010 acquires a plurality of sets of training data 15 (first to eleventh training data 15-1 to 15-11) each consisting of input data including at least molecular chain information 11 and crystallinity information 12 and output data including physical property information 14. The training data 15 (first to eleventh training data 15-1 to 15-11) are data used as teacher data (training data), verification data, and test data in supervised learning. In addition, the output data included in the training data 15 (first to eleventh training data 15-1 to 15-11) are data used as correct answer labels in supervised learning.

[0071] For example, the learning data acquisition unit 5010 acquires the learning data 15 (first to eleventh learning data 15-1 to 15-11) by referring to various pieces of information 10 to 14 registered in the information management database 510 or by receiving an input operation from the user terminal device 6. Note that if information corresponding to the learning data 15 (first to eleventh learning data 15-1 to 15-11) is stored in an external system (such as a raw material management system or a molded body management system), the learning data acquisition unit 5010 may acquire the learning data 15 (first to eleventh learning data 15-1 to 15-11) from the external system.

[0072] The machine learning unit 5011 performs machine learning for each learning model 16 (first to eleventh learning models 16-1 to 16-11) using multiple sets of learning data 15 (learning data 15-1 to 15-11) acquired by the learning data acquisition unit 5010. Then, the machine learning unit 5011 generates trained learning models 16 (first to eleventh learning models 16-1 to 16-11) by having each learning model 16 (first to eleventh learning models 16-1 to 16-11) learn the correlation between input data and output data.

[0073] When performing machine learning, the machine learning unit 5011 can employ any method, such as online learning, batch learning, mini-batch learning, etc. Furthermore, the machine learning unit 5011 may perform predetermined pre-processing on input data input to each learning model 16 (first to eleventh learning models 16-1 to 16-11), or may perform predetermined post-processing on output data output from each learning model 16 (first to eleventh learning models 16-1 to 16-11).

[0074] The trained model management database 511 stores trained learning models 16 (first to eleventh learning models 16-1 to 16-11) (specifically, adjusted weight parameter groups) generated by the machine learning unit 5011. The trained learning models 16 (first to eleventh learning models 16-1 to 16-11) stored in the trained model management database 511 may be provided to other systems via the network 7, a recording medium, etc.

[0075] In this embodiment, the data configurations of the training data 15 (first to eleventh training data 15-1 to 15-11) and the training models 16 (first to eleventh training models 16-1 to 16-11) have been described as shown in FIG. 5 and FIG. 6A to FIG. 6K. However, a plurality of data configurations with different conditions, such as different machine learning techniques, different input data, or different output data, may be employed. In this case, the training data acquisition unit 5010 acquires a plurality of types of training data corresponding to a plurality of data configurations with different conditions, and the machine learning unit 5011 performs machine learning using each of the training data, and the trained training models 16 are stored in the trained model management database 511. For example, other examples of the learning model 16 include a learning model for outputting the tensile modulus of a molded piece (molded piece physical property information 14A) using molecular chain information 11 and crystallinity information 12 as input data, a learning model for outputting the tensile modulus of a film (film physical property information 14B) using molecular chain information 11 and crystallinity information 12 as input data, and a learning model for outputting the thermal shrinkage rate of a film (film physical property information 14B) using molecular chain information 11 and crystallinity information 12 as input data.

[0076] (Physical property prediction function 502) 7 is a functional explanatory diagram showing an example of the physical property prediction function 502. The data acquisition unit 5020, data generation unit 5021, and output processing unit 5022 of the information processing device 5 realize the physical property prediction function 502 using the trained model management database 511. By including the physical property prediction function 502 in the information processing device 5, the information processing device 5 operates as a physical property prediction device.

[0077] The data acquisition unit 5020 acquires input data for a new raw material to be predicted, including at least molecular chain information 11 and crystallinity information 12. At this time, the data acquisition unit 5020 acquires the input data according to the data configuration of the learning model 16. Therefore, if the input data further includes at least one of raw material information 10 and process information 13, the data acquisition unit 5020 acquires at least one of the raw material information 10 and process information 13 included in the input data.

[0078] For example, the data acquisition unit 5020 acquires input data by receiving molecular chain information 11 and crystallinity information 12 from the analysis device 2, receiving process information 13 from the molding device 3, referring to various types of information 10 to 13 registered in the information management database 510, or accepting input operations for various types of information 10 to 13 from the user terminal device 6.

[0079] The data generation unit 5021 inputs the input data acquired by the data acquisition unit 5020 into the learning model 16, thereby generating output data for the input data.

[0080] The learning models 16 (first to eleventh learning models 16-1 to 16-11) used by the data generation unit 5021 are the learned learning models 16 (first to eleventh learning models 16-1 to 16-11) stored in the learned model management database 511. When the first to eleventh learning models 16-1 to 16-11 shown in FIGS. 6A to 6K are stored in the learned model management database 511 as the learning models 16, the data generation unit 5021 can use the first to eleventh learning models 16-1 to 16-11 selectively or in parallel. In this case, the data generation unit 5021 may input input data corresponding to the data configuration of the input data in the first to eleventh learning models 16-1 to 16-11 to each of the first to eleventh learning models 16-1 to 16-11.

[0081] The data generation unit 5021 may perform predetermined pre-processing on input data to be input to the learning models 16 (the first to eleventh learning models 16-1 to 16-11). The data generation unit 5021 may also perform predetermined post-processing on output data output from the learning models 16 (the first to eleventh learning models 16-1 to 16-11).

[0082] The output processing unit 5022 performs output processing for outputting the output data (physical property information 14) generated by the data generation unit 5021. For example, as the output processing, the output processing unit 5022 may transmit display information for displaying the physical property information 14 on the user terminal device 6 to the user terminal device 6, or may register the physical property information 14 in the information management database 510.

[0083] 8 is a hardware configuration diagram showing an example of a computer 900 constituting each device. Each of the devices 2 to 6 in the property prediction system 1 is constituted by a general-purpose or dedicated computer 900.

[0084] 8, the computer 900 includes, as its main components, a bus 910, a processor 912, a memory 914, an input device 916, an output device 917, a display device 918, a storage device 920, a communication I / F (interface) unit 922, an external device I / F unit 924, an I / O (input / output) device I / F unit 926, and a media input / output unit 928. Note that the above components may be omitted as appropriate depending on the application of the computer 900.

[0085] The processor 912 is composed of one or more arithmetic processing devices (such as a CPU (Central Processing Unit), MPU (Micro-Processing Unit), DSP (Digital Signal Processor), GPU (Graphics Processing Unit), or NPU (Neural Processing Unit)), and operates as a control unit that controls the entire computer 900. The memory 914 stores various data and programs 930, and is composed of, for example, a volatile memory (such as a DRAM or SRAM) that functions as a main memory, a non-volatile memory (ROM), a flash memory, etc.

[0086] The input device 916 is composed of, for example, a keyboard, a mouse, a numeric keypad, an electronic pen, etc., and functions as an input unit. The output device 917 is composed of, for example, a sound (audio) output device, a vibration device, etc., and functions as an output unit. The display device 918 is composed of, for example, a liquid crystal display, an organic EL display, electronic paper, a projector, etc., and functions as an output unit. The input device 916 and the display device 918 may be integrated into one device, such as a touch panel display. The storage device 920 is composed of, for example, an HDD, an SSD, etc., and functions as a storage unit. The storage device 920 stores various data necessary for executing the operating system and the program 930.

[0087] The communication I / F unit 922 is connected to a network 940 such as the Internet or an intranet (which may be the same as network 7 in FIG. 1) via a wired or wireless connection and functions as a communication unit that transmits and receives data to and from other computers in accordance with a predetermined communication protocol. The external device I / F unit 924 is connected to an external device 950 such as a camera, printer, scanner, or reader / writer via a wired or wireless connection and functions as a communication unit that transmits and receives data to and from the external device 950 in accordance with a predetermined communication protocol. The I / O device I / F unit 926 is connected to an I / O device 960 such as various sensors and actuators and functions as a communication unit that transmits and receives various signals and data, such as detection signals from sensors and control signals to actuators, to and from the I / O device 960. The media input / output unit 928 is composed of a drive device such as a DVD drive or CD drive, a memory card slot, and a USB connector, and reads and writes data from and to media (non-transitory storage media) 970 such as a DVD, CD, memory card, or USB memory.

[0088] In the computer 900 having the above configuration, the processor 912 loads a program 930 stored in the storage device 920 into the memory 914, executes the program, and controls each unit of the computer 900 via the bus 910. The program 930 may be stored in the memory 914 instead of the storage device 920. The program 930 may be recorded on the medium 970 in an installable file format or an executable file format and provided to the computer 900 via the media input / output unit 928. The program 930 may be provided to the computer 900 by being downloaded via the communication I / F unit 922 over the network 940. Furthermore, the computer 900 may implement various functions realized by the processor 912 executing the program 930 using hardware such as an FPGA (Field-Programmable Gate Array) or an ASIC (Application Specific Integrated Circuit).

[0089] The computer 900 is an electronic device of any type, such as a desktop computer or a portable computer. The computer 900 may be a client computer, a server computer, a cloud computer, or an embedded computer such as a control panel or a controller (including a microcomputer, a programmable logic controller, or a sequencer).

[0090] (Operation of the physical property prediction system 1) Hereinafter, as the operation of the property prediction system 1, the learning function 501 and the property prediction function 502 realized by the information processing device 5 will be described.

[0091] (machine learning methods) 9 is a flowchart showing an example of a machine learning method using the learning function 501. The following describes a case where a learning model 16 (one of the first to eleventh learning models 16-1 to 16-11) is generated using multiple sets of learning data 15 (for example, one of the learning data 15-1 to 15-11 shown in FIGS. 6A to 6K).

[0092] First, in step S100, the learning data acquisition unit 5010 acquires a desired number of learning data 15 as a preliminary preparation for starting machine learning, and temporarily stores the acquired learning data 15 in the memory unit 51.

[0093] Next, in step S110, in order to start machine learning, the machine learning unit 5011 prepares a pre-learning learning model 16. The pre-learning learning model 16 prepared here is configured, for example, by a neural network model, and the weight of each synapse is set to an initial value.

[0094] Next, in step S120, the machine learning unit 5011 acquires, for example, one set of learning data 15 at random from the multiple sets of learning data 15 stored in the storage unit 51.

[0095] Next, in step S130, the machine learning unit 5011 inputs the molecular chain information 11 and the crystallinity information 12 (input data) contained in one set of training data 15 to the input layer 160 of the prepared training model 16 before training (or during training). As a result, the output layer 162 of the training model 16 outputs the physical property information 14 (output data) as an inference result, and this output data has been generated by the training model 16 before training (or during training). Therefore, in the state before training (or during training), the output data output as an inference result indicates information different from the physical property information 14 (correct label) contained in the training data 15.

[0096] Next, in step S140, the machine learning unit 5011 performs machine learning by comparing the physical property information 14 (correct label) included in the set of learning data 15 acquired in step S120 with the physical property information 14 (output data) output as an inference result from the output layer 162 in step S130, and performing a process of adjusting the weight of each synapse (backpropagation).In this way, the machine learning unit 5011 causes the learning model 16 to learn the correlation between the molecular chain information 11 and the crystallinity information 12 and the physical property information 14.

[0097] Next, in step S150, the machine learning unit 5011 determines whether a predetermined learning termination condition has been met, for example, based on the evaluation value of an error function based on the physical property information 14 (correct label) included in the learning data 15 and the physical property information 14 (output data) output as an inference result, or the remaining number of unlearned learning data 15 stored in the memory unit 51.

[0098] In step S150, if the machine learning unit 5011 determines that the learning termination condition is not satisfied and that machine learning should be continued (No in step S150), the process returns to step S120, and the processes of steps S120 to S140 are performed multiple times on the learning model 16 being trained using untrained training data 15. On the other hand, in step S150, if the machine learning unit 5011 determines that the learning termination condition is satisfied and that machine learning should be terminated (Yes in step S150), the process proceeds to step S160.

[0099] Then, in step S160, the machine learning unit 5011 stores the trained learning model 16 (adjusted weight parameter group) generated by adjusting the weights associated with each synapse in the trained model management database 511, and ends the series of machine learning methods shown in Fig. 9. In the machine learning method, step S100 corresponds to the training data acquisition step, steps S110 to S150 correspond to the machine learning step, and step S160 corresponds to the trained model storage step.

[0100] As described above, the learning function 501 and machine learning method according to this embodiment can provide a learning model 16 that generates output data including physical property information 14 in response to input data including molecular chain information 11 and crystallinity information 12.

[0101] (Physical property prediction method) 10 is a flowchart showing an example of a physical property prediction method using the physical property prediction function 502. In the following description, it is assumed that the learned model management database 511 stores the first to eleventh learned models 16-1 to 16-11 shown in FIGS. 6A to 6K as the learned models 16 learned by the learning function 501.

[0102] First, in step S200, the user terminal device 6 executes the information management program and displays the property prediction input screen 17 in cooperation with the property prediction function 502 of the information processing device 5, and receives an input operation from the user regarding a new raw material to be predicted on the property prediction input screen 17. Then, the user terminal device 6 transmits input operation information indicating the result of the input operation by the user to the information processing device 5. Note that the user terminal device 6 may also receive an input operation to specify at least one of the raw material information 10 and the process information 13, or to specify a physical property value to be predicted, along with the molecular chain information 11 and the crystallinity information 12, and transmit this to the information processing device 5.

[0103] 11 is a screen configuration diagram showing an example of the physical property prediction input screen 17. The physical property prediction input screen 17 includes a raw material specification field 170 for specifying various information related to a new raw material to be predicted, and a physical property value specification field 171 for specifying the physical property value of the prediction target.

[0104] In the ingredient specification field 170, a new ingredient to be predicted is specified, for example, by specifying a management ID that has been registered in the information management database 510. Also, various information 10 to 13 may be specified on an information setting screen that is displayed by pressing a setting button, or various information 10 to 13 that has been registered in the past may be specified.

[0105] The physical property values ​​to be predicted can be specified using check boxes in the physical property value specification field 171. The example in Fig. 11 shows a case where "flexural modulus" and "tensile breaking elongation" are specified to be predicted as the physical property values ​​of a molded piece obtained by melt molding a raw material assigned the management ID "ID03" specified in the raw material specification field 170.

[0106] Next, in step S210, upon receiving the input operation information transmitted in step S200, the data acquisition unit 5020 acquires input data to be input to the learning model 16 based on the input operation information. In the example of the property prediction input screen 17 shown in FIG. 11, the management ID "ID03" is specified, so the data acquisition unit 5020 acquires, for example, the molecular chain information 11 and the crystallinity information 12 associated with the management ID "ID03" in the information management database 510.

[0107] Next, in step S220, the data generation unit 5021 inputs the input data (molecular chain information 11, crystallinity information 12, etc.) acquired in step S210 into the learning model 16, thereby generating output data (physical property information 14) for the input data. In the example of the physical property prediction input screen 17 shown in FIG. 11, "flexural modulus" and "tensile breaking elongation" are specified as the physical property values ​​to be predicted. Therefore, the data generation unit 5021 selects a second learning model 16-2 whose output data includes "flexural modulus" and a fourth learning model 16-4 whose output data includes "tensile breaking elongation." Then, the data generation unit 5021 predicts "flexural modulus" and "tensile breaking elongation" based on the physical property information 14 output as output data by inputting the molecular chain information 11 and crystallinity information 12 of the raw material to be predicted specified via the physical property prediction input screen 17 into the second learning model 16-2 and the fourth learning model 16-4, respectively.

[0108] Next, in step S230, the output processing unit 5022 executes output processing to output the physical property information 14 as output data generated in step S220. For example, when display information for displaying the physical property information 14 is transmitted to the user terminal device 6, in step S240, the user terminal device 6 displays the physical property value output screen 18 based on the display information, thereby presenting the physical property information 14 to the user.

[0109] 12 is a screen configuration diagram showing an example of the physical property output screen 18. The physical property output screen 18 includes an input data display field 180 that displays input data based on an input operation input on the physical property prediction input screen 17, and a physical property information display field 181 that displays the physical property information 14 as output data generated by the data generation unit 5021.

[0110] The physical property value output screen 18 shown in FIG. 12 displays the predicted results of "flexural modulus" and "tensile breaking elongation" as the physical property information 14 as output data generated by the data generating unit 5021.

[0111] In this manner, the series of steps in the physical property prediction method shown in Fig. 10 is completed. In the above physical property prediction method, step S210 corresponds to a data acquisition step, step S220 corresponds to a data generation step, and step S230 corresponds to an output processing step. Note that the series of steps in the physical property prediction method can be executed at any timing as long as the information processing device 5 can acquire input data.

[0112] As described above, according to the physical property prediction function 502 and physical property prediction method of the information processing device 5 of this embodiment, by inputting input data including molecular chain information 11 and crystallinity information 12 into the learning model 16, output data including physical property information 14 indicating the physical property values ​​of the molded body for the input data is generated. Therefore, the physical property values ​​of the molded body can be appropriately predicted from the characteristics of the raw materials.

[0113] (Other embodiments) The present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the spirit and scope of the present invention, all of which are included in the technical concept of the present invention.

[0114] In the above embodiment, the multiple functions of the information processing device 5 are described as being realized by one device, but each function may be distributed among multiple devices (computers) and thereby realized by multiple devices. For example, an information management device that realizes the information management function 500, a machine learning device that realizes the learning function 501, and a physical property prediction device that realizes the physical property prediction function 502 may be configured as different devices.

[0115] In the above embodiment, the information processing device 5 is described as having a communication unit 52 and transmitting and receiving various data between the information processing device 5 and the user terminal device 6 via the communication unit 52, but the information processing device 5 may also operate as a standalone device.

[0116] In the above embodiment, the information processing device 5 operates according to the flowcharts shown in Figures 9 and 10, but some of the steps (units) may be omitted or other steps may be added. In this case, the omitted steps (units) may be executed by an external system.

[0117] In the above embodiment, a case has been described in which the learning function 501 of the information processing device 5 generates the first to eleventh learning models 16-1 to 16-11, and the physical property prediction function 502 generates output data (physical property information 14) from input data (molecular chain information 11, crystallinity information 12, etc.) using the first to eleventh learning models 16-1 to 16-11. In contrast, the learning model 16 generated by the learning function 501 and used by the physical property prediction function 502 may be any one of the first to eleventh learning models 16-1 to 16-11, or any combination of two or more of them.

[0118] In the above embodiment, a case has been described in which a neural network is employed as the learning model 16 that realizes machine learning by the learning function 501, but other machine learning models may also be employed. Examples of other machine learning models include tree types such as decision trees and regression trees, ensemble learning such as bagging and boosting, neural network types (including deep learning) such as recurrent neural networks, convolutional neural networks and LSTM, clustering types such as hierarchical clustering, non-hierarchical clustering, k-nearest neighbors and k-means, multivariate analyses such as principal component analysis, factor analysis and logistic regression, and support vector machines.

[0119] In the above embodiment, a case has been described in which the physical property prediction function 502 of the information processing device 5 generates output data (physical property information 14) from input data (molecular chain information 11, crystallinity information 12, etc.) using the learning model 16. In contrast, the physical property prediction function 502 may generate output data (physical property information 14) from input data (molecular chain information 11, crystallinity information 12, etc.) using known information 10 to 14 registered in the information management database 510 (storage unit) without using the learning model 16.

[0120] Specifically, the data generation unit 5021 searches the information management database 510, which stores a plurality of pieces of known molecular chain information 11 and known crystallinity information 12 associated with each other, and known physical property information 14 indicating the physical property values ​​of a molded body obtained by melt-molding a raw material having a molecular chain indicated by the known molecular chain information 11 and a crystalline portion indicated by the known crystallinity information 12, for known molecular chain information 11 and known crystallinity information 12 that match or are similar to the molecular chain information 11 and known crystallinity information 12 included in the input data acquired by the data acquisition unit 5020, and generates output data including the known physical property information 14 based on the known physical property information 14 associated with the known molecular chain information 11 and known crystallinity information 12 extracted as the search result.

[0121] A search method for searching for known molecular chain information 11 and known crystallinity information 12 that match or are similar to known molecular chain information 11 and known crystallinity information 12 included in the input data may be, for example, to convert the molecular chain information 11 and the crystallinity information 12 into feature quantities (e.g., feature vectors, etc.), compare the feature quantities, and extract, as search results, known molecular chain information 11 and known crystallinity information 12 that satisfy a predetermined degree of match or similarity with the input data. In this case, the input data may include known raw material information 10 or known process information 13 in addition to the known molecular chain information 11 and known crystallinity information 12.

[0122] The process by the data generation unit 5021 of inputting input data into the learning model 16 to generate output data (model input process) and the process by which the data generation unit 5021 searches for known information 10-14 registered in the information management database 510 based on the input data to generate output data (database search process) may be performed selectively according to predetermined selection conditions or in parallel. For example, the data generation unit 5021 may generate output data by the database search process if the number of pieces of data registered in the information management database 510 exceeds a predetermined reference value, or may generate output data by the model input process if this is not the case. The data generation unit 5021 may also search for known molecular chain information 11 and known crystallinity information 12 registered in the information management database 510 based on the input data, and if the degree of match or similarity to the known molecular chain information 11 and known crystallinity information 12 extracted as the search results exceeds a predetermined reference value, generate output data by the database search process, or may generate output data by the model input process if this is not the case.

[0123] (Inference device, inference method and inference program) The present invention can be provided not only in an embodiment in which the information processing device 5 (information processing method or information processing program) according to the above embodiment functions as a physical property prediction device, but also in an embodiment in which an inference device (inference method or inference program) is used to infer physical property information 14. In this case, the inference device (inference method or inference program) can include a memory and a processor, and the processor executes a series of processes. The series of processes includes a data acquisition process (data acquisition step) for acquiring input data including molecular chain information 11 and crystallinity information 12, and an inference process (inference step) for inferring output data including physical property information 14 from the input data acquired by the data acquisition process.

[0124] Various aspects of the present disclosure are summarized below as appendices.

[0125] (Appendix 1) a data acquisition unit for acquiring input data; a data generation unit that generates output data for the input data by inputting the input data acquired by the data acquisition unit into a learning model; The learning model is a trained model that has been trained by machine learning to determine the correlation between the input data and the output data; The input data is molecular chain information on the molecular chains of the polymer compound that is the raw material for the molded body; and crystallinity information regarding the crystalline portion of the polymer compound; The output data is The solid raw material is melt-molded to form the molded body, and the molded body includes physical property information indicating the physical property values ​​of the molded body obtained by melt-molding the solid raw material. Physical property prediction device.

[0126] (Appendix 2) The molecular chain information included in the input data is melt analysis molecular chain information indicating information obtained by performing melt analysis on the molecular chain of the solid raw material; and At least one of solution analysis molecular chain information indicating information obtained by performing solution analysis on the molecular chain of the solid raw material. 2. The physical property prediction device according to claim 1.

[0127] (Appendix 3) The melt analysis molecular chain information is Information about melt viscosity, The solution analysis molecular chain information is Information about molecular weight, Information about molecular weight distribution, and At least one of the information about the branching structure, 3. The physical property prediction device according to claim 2.

[0128] (Appendix 4) The crystallinity information included in the input data is Melt analysis crystallinity information indicating information obtained by performing melt analysis on the crystalline portion of the solid raw material; and At least one of solid analysis crystallinity information indicating information obtained by performing solid analysis on the crystalline portion of the solid raw material. 4. The physical property prediction device according to claim 1, wherein the physical property prediction device is a physical property prediction device.

[0129] (Appendix 5) The melting analysis crystallinity information is Information about the degree of crystallinity, the rate of crystallization, the crystal thickness or the melting point. The solid analysis crystallography information is This is information about density, 5. The physical property prediction device according to claim 4.

[0130] (Appendix 6) The input data is further comprising ingredient information regarding the solid ingredient; 6. A physical property prediction device according to any one of Supplementary Note 1 to Supplementary Note 5.

[0131] (Appendix 7) The raw material information included in the input data is information regarding the type of said raw material; Information regarding the blending ratio of the raw materials, and At least one piece of information relating to the deterioration degree of the raw material; 7. The physical property prediction device according to claim 6.

[0132] (Appendix 8) The input data is Further includes process information regarding a molding process when the solid raw material is melt-molded. 8. The physical property prediction device according to claim 1, wherein the physical property prediction device is a physical property prediction device.

[0133] (Appendix 9) The process information included in the input data is Molded piece process information regarding a molding process when the raw material is melt-molded into a molded piece having a predetermined shape as the molded body; and At least one piece of film process information relating to a molding process when the raw material is melt-molded into a film having a predetermined thickness as the molded body. 9. The physical property prediction device according to claim 8.

[0134] (Appendix 10) The physical property information included in the output data is The molded piece physical property information indicates the physical property values ​​when the raw material is melt-molded into a molded piece having a predetermined shape as the molded body. 10. The physical property prediction device according to any one of Supplementary Note 1 to Supplementary Note 9.

[0135] (Appendix 11) The molded piece physical property information is Environmental stress crack resistance, flexural modulus, tensile modulus, Charpy impact strength, Tensile elongation at break, and Vicat softening temperature, 11. The physical property prediction device according to claim 10.

[0136] (Appendix 12) The molded piece physical property information is Strength properties, Contents characteristics, Usability, and At least one of hermeticity, 11. The physical property prediction device according to claim 10.

[0137] (Appendix 13) The physical property information included in the output data is The film physical property information indicates the physical property values ​​when the raw material is melt-molded into a film having a predetermined thickness as the molded body. 13. The physical property prediction device according to claim 1.

[0138] (Appendix 14) The film physical property information is Impact burst strength, tensile impact strength, tensile modulus, Heat shrinkage rate, Tear strength, and At least one of the loop stiffness 14. The physical property prediction device according to claim 13.

[0139] (Appendix 15) The film physical property information is Drop characteristics, unsealability, Preservability, Usability, and At least one of the line qualifications, 14. The physical property prediction device according to claim 13.

[0140] (Appendix 16) The data generation unit a storage unit storing a plurality of pieces of known molecular chain information and known crystallinity information, in association with each other, and known physical property information indicating physical property values ​​of the molded body obtained by melt molding the raw material having the molecular chain indicated by the known molecular chain information and the crystalline portion indicated by the known crystallinity information, searching for known molecular chain information and known crystallinity information that matches or is similar to the molecular chain information and the crystallinity information included in the input data acquired by the data acquisition unit, and generating the output data based on the known physical property information associated with the known molecular chain information and the known crystallinity information extracted as a search result; 16. The physical property prediction device according to any one of Supplementary Note 1 to Supplementary Note 15. [Explanation of symbols]

[0141] 1...Physical property prediction system, 2...Analysis equipment, 2A...Melting analysis equipment, 2B...Solution analysis equipment, 2C...solid analysis device, 3...forming device, 3A...molded piece forming device, 3B...film forming device, 4...Testing apparatus, 4A...Molded piece testing apparatus, 4B...Film testing apparatus, 5...information processing device, 6...user terminal device 10...raw material information, 11...molecular chain information, 11A...melt analysis molecular chain information, 11B...solution analysis molecular chain information, 12...crystallization information, 12A...melting analysis crystallization information, 12B...Solid analysis crystallinity information, 13...Process information, 13A...Molded piece process information, 13B...film process information, 14...physical property information, 14A...molded piece physical property information, 14B...Film property information, 15, 15-1 to 15-11...Learning data 16, 16-1~16-11...Learning model, 50...control unit, 51...storage unit, 52...communication unit, 53...input unit, 54...display unit, 500...information management function, 501...learning function, 502...physical property prediction function, 510...Information management database, 511...Model management database, 512...information processing program, 5010...Learning data acquisition unit, 5011...Machine learning unit, 5020: data acquisition unit, 5021: data generation unit, 5022: output processing unit

Claims

1. a data acquisition unit for acquiring input data; a data generation unit that generates output data for the input data by inputting the input data acquired by the data acquisition unit into a learning model; The learning model is a trained model that has been trained by machine learning to determine the correlation between the input data and the output data; The input data is molecular chain information on the molecular chains of the polymer compound that is the raw material for the molded body; and crystallinity information regarding the crystalline portion of the polymer compound; The output data is The solid raw material is melt-molded to form the molded body, and the molded body includes physical property information indicating the physical property values ​​of the molded body obtained by melt-molding the solid raw material. Physical property prediction device.

2. The molecular chain information included in the input data is melt analysis molecular chain information indicating information obtained by performing melt analysis on the molecular chain of the solid raw material; and at least one of solution analysis molecular chain information indicating information obtained by performing solution analysis on the molecular chain of the solid raw material; The physical property prediction device according to claim 1 .

3. The melt analysis molecular chain information is Information about melt viscosity, The solution analysis molecular chain information is Information about molecular weight, Information about the molecular weight distribution, and At least one of information about a branched structure, The physical property prediction device according to claim 2 .

4. The crystallinity information included in the input data is Melt analysis crystallinity information indicating information obtained by performing melt analysis on the crystalline portion of the solid raw material; and At least one of solid analysis crystallinity information indicating information obtained by performing solid analysis on the crystalline portion of the solid raw material, The physical property prediction device according to claim 1 .

5. The melting analysis crystallinity information is Information about the degree of crystallinity, the rate of crystallization, the crystal thickness, or the melting point, The solid analysis crystallography information is This is information about density, The physical property prediction device according to claim 4 .

6. The input data is further comprising ingredient information regarding the solid ingredient; The physical property prediction device according to claim 1 .

7. The raw material information included in the input data is information regarding the type of said raw material; Information regarding the blending ratio of the raw materials, and At least one piece of information relating to the deterioration degree of the raw material; The physical property prediction device according to claim 6 .

8. The input data is Further includes process information regarding a molding process when the solid raw material is melt-molded. The physical property prediction device according to claim 1 .

9. The process information included in the input data is Molded piece process information regarding a molding process when the raw material is melt-molded into a molded piece having a predetermined shape as the molded body; and At least one of film process information relating to a molding process when the raw material is melt-molded into a film having a predetermined thickness as the molded body, That is, The physical property prediction device according to claim 8 .

10. The physical property information included in the output data is The molded piece physical property information indicates the physical property values ​​when the raw material is melt-molded into a molded piece having a predetermined shape as the molded body. The physical property prediction device according to claim 1 .

11. The molded piece physical property information is Environmental stress crack resistance, flexural modulus, tensile modulus, Charpy impact strength, Tensile elongation at break, and Vicat softening temperature, The physical property prediction device according to claim 10.

12. The molded piece physical property information is Strength properties, Contents characteristics, Usability, and At least one of hermeticity, The physical property prediction device according to claim 10.

13. The physical property information included in the output data is The film physical property information indicates the physical property values ​​when the raw material is melt-molded into a film having a predetermined thickness as the molded body. The physical property prediction device according to claim 1 .

14. The film physical property information is Impact burst strength, tensile impact strength, tensile modulus, Heat shrinkage rate, Tear strength, and At least one of the loop stiffness The physical property prediction device according to claim 13.

15. The film physical property information is Drop characteristics, unsealability, Preservability, Usability, and At least one of the line qualifications, The physical property prediction device according to claim 13.

16. The data generation unit a storage unit storing a plurality of pieces of known molecular chain information and known crystallinity information, in association with each other, and known physical property information indicating physical property values ​​of the molded body obtained by melt molding the raw material having the molecular chain indicated by the known molecular chain information and the crystalline portion indicated by the known crystallinity information, searching for known molecular chain information and known crystallinity information that matches or is similar to the molecular chain information and the crystallinity information included in the input data acquired by the data acquisition unit, and generating the output data based on the known physical property information associated with the known molecular chain information and the known crystallinity information extracted as a search result; The physical property prediction device according to claim 1 .

17. An inference device comprising a memory and a processor, The processor: a data acquisition process for acquiring input data including molecular chain information on molecular chains of a polymer compound that is a raw material for a molded body and crystallinity information on a crystalline portion of the polymer compound; When the input data is acquired by the data acquisition process, an inference process is executed to infer output data including physical property information indicating physical property values ​​of the molded body obtained by melt molding the solid raw material from the input data. Reasoning device.

18. a learning data acquisition unit that acquires multiple sets of learning data each consisting of input data and output data; a machine learning unit that uses the plurality of sets of learning data acquired by the learning data acquisition unit to train a learning model by machine learning to learn a correlation between the input data and the output data; a learned model storage unit that stores the learned model in which the correlation has been learned by the machine learning unit, The input data is molecular chain information on the molecular chains of the polymer compound that is the raw material for the molded body; and crystallinity information regarding the crystalline portion of the polymer compound; The output data is The solid raw material is melt-molded to form the molded body, and the molded body includes physical property information indicating the physical property values ​​of the molded body obtained by melt-molding the solid raw material. Machine learning device.

19. 1. A computer-implemented method for predicting physical properties, comprising: a data acquisition step of acquiring input data; a data generation step of inputting the input data acquired by the data acquisition step into a learning model to generate output data for the input data; The learning model is a trained model that has been trained by machine learning to determine the correlation between the input data and the output data; The input data is molecular chain information on the molecular chains of the polymer compound that is the raw material for the molded body; and crystallinity information regarding the crystalline portion of the polymer compound; The output data is The solid raw material is melt-molded to form the molded body, and the molded body includes physical property information indicating the physical property values ​​of the molded body obtained by melt-molding the solid raw material. Physical property prediction methods.

20. An inference method executed by an inference device having a memory and a processor, The processor: a data acquisition process for acquiring input data including molecular chain information on molecular chains of a polymer compound that is a raw material for a molded body and crystallinity information on a crystalline portion of the polymer compound; When the input data is acquired by the data acquisition process, an inference process is executed to infer output data including physical property information indicating physical property values ​​of the molded body obtained by melt molding the solid raw material from the input data. Reasoning method.

21. 1. A computer-implemented machine learning method comprising: a learning data acquisition step of acquiring a plurality of sets of learning data each composed of input data and output data; a machine learning step of causing a learning model to learn a correlation between the input data and the output data by machine learning using the plurality of sets of learning data acquired by the learning data acquisition step; a trained model storage step of storing the trained model, which has learned the correlation through the machine learning step, in a trained model storage unit; The input data is molecular chain information on the molecular chains of the polymer compound that is the raw material for the molded body; and crystallinity information regarding the crystalline portion of the polymer compound; The output data is The solid raw material is melt-molded to form the molded body, and the molded body includes physical property information indicating the physical property values ​​of the molded body obtained by melt-molding the solid raw material. Machine learning methods.

Images