Physical property information estimation method, physical property information estimation model generation method, device, and program

Abstract

To provide a method for inferring property information of a resin composition produced by a screw extruder from information on a manufacturing condition, a method for generating a property information estimation model, an apparatus, and a program for inferring property information of the resin composition.SOLUTION: A guessing unit 24 of the property information guessing device 14 inputs information about manufacturing conditions for manufacturing a resin composition from a thermoplastic resin using a screw extruder equipped with a screw to a learned model for guessing property information of the resin composition from the manufacturing condition information for manufacturing the resin composition, and guesses the property information of the resin composition corresponding to the input manufacturing condition information.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present disclosure relates to a physical property information estimation method, a physical property information estimation model generating method, an apparatus, and a program. [Background technology]

[0002] Conventionally, a simulation device for a screw-type extruder is known (for example, see Patent Document 1). This simulation device calculates the distribution of the filling rate, pressure, temperature, solid phase occupancy, and residence time inside the extruder based on the device configuration and operating conditions of the extruder and the resin properties.

[0003] Also, a simulation device is known that efficiently realizes low-dimensional flow analysis and high-dimensional flow analysis under the operating conditions of the same kneading device (for example, see Patent Document 2). This simulation device is an extrusion molding machine simulation device that performs flow analysis of a material, and performs low-dimensional flow analysis of the material in a calculation target field of the kneading device based on setting information including the physical properties of the material and configuration data and operating conditions of the kneading device that kneads the material. Furthermore, after or prior to the low-dimensional flow analysis, this simulation device accepts the selection of a target region to be the target of high-dimensional flow analysis in the calculation target field, extracts physical quantities of the material related to the target region from the results of the low-dimensional flow analysis, and performs high-dimensional flow analysis of the material in the target region based on the extracted physical quantities and the setting information. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] JP 2011-173276 A [Patent Document 2] JP 2016-88056 A Summary of the Invention [Problem to be solved by the invention]

[0005] When a resin composition is produced from a thermoplastic resin using a screw extruder, the physical properties of the finished product of the resin composition vary depending on the production conditions used to produce the resin composition. For this reason, it is desirable to obtain information on what production conditions should be set to obtain a resin composition having desired physical properties.

[0006] However, Patent Document 1 and Patent Document 2 do not disclose estimation of physical property information of a resin composition when the resin composition is produced from a thermoplastic resin by a screw extruder. Therefore, the techniques disclosed in Patent Document 1 and Patent Document 2 have a problem in that it is not possible to estimate physical property information of a resin composition when the resin composition is produced by a screw extruder.

[0007] An object of the present disclosure is to provide a physical property information estimation method, a physical property information estimation model generation method, an apparatus, and a program that can estimate physical property information of a resin composition produced by a screw extruder from production condition information of the resin composition. [Means for solving the problem]

[0008] The physical property information estimation method disclosed herein is a physical property information estimation method in which a computer executes a process in which production condition information for producing a resin composition from a thermoplastic resin using a screw extruder equipped with a screw is input into a trained model for estimating physical property information of the resin composition from the production condition information for producing a resin composition from a thermoplastic resin using a screw extruder equipped with a screw, and physical property information of the resin composition corresponding to the inputted production condition information is estimated.

[0009] In addition, the manufacturing condition information of the present disclosure may include at least one of information regarding the screw configuration of the screw extruder, information regarding additives added to the resin composition, discharge volume information of the screw extruder, screw rotation speed information of the screw extruder, and temperature information when the resin composition is manufactured.

[0010] In addition, the trained model of the present disclosure may include a first trained model that estimates temperature information when the resin composition is produced from at least one of information on the screw configuration of the screw extruder, information on the output rate of the screw extruder, and information on the screw rotation speed of the screw extruder, which are information on the production conditions of the resin composition using a screw extruder, and a second trained model that estimates physical property information of the resin composition from the temperature information output from the first trained model and at least one of information on the screw configuration of the screw extruder, information on the output rate of the screw extruder, and information on the screw rotation speed of the screw extruder, which are information on the production conditions of the resin composition using a screw extruder.

[0011] Moreover, the information regarding the additives added to the resin composition of the present disclosure may be information regarding a filler.

[0012] In addition, the manufacturing condition information of the present disclosure is information regarding the screw configuration of the screw extruder, information regarding additives added to the resin composition, extrusion volume information of the screw extruder, and screw rotation speed information of the screw extruder, and a plurality of pieces of manufacturing condition information having different information regarding the screw configuration of the screw extruder, information regarding additives added to the resin composition, extrusion volume information of the screw extruder, and screw rotation speed information of the screw extruder can be generated, and each of the plurality of generated manufacturing condition information can be input into the trained model to infer the physical property information corresponding to each of the plurality of pieces of manufacturing condition information.

[0013] In addition, the physical property information of the present disclosure may include any one of the flexural modulus, tensile strength, tensile elongation, flexural strength, Izod impact strength, Charpy impact strength, and compressive yield stress of the resin composition.

[0014] In addition, the trained model of the present disclosure may be a model based on Partial Least Squares regression (PLS), a support vector machine model, a random forest model, a Gaussian process model, or a neural network model.

[0015] The physical property information estimation model generation method disclosed herein is a physical property information estimation model generation method in which a computer executes a process to generate a trained model for estimating physical property information of a resin composition from manufacturing condition information of a resin composition produced from a thermoplastic resin using a screw extruder equipped with a screw, based on training data representing a combination of manufacturing condition information of a training resin composition produced from a thermoplastic resin using a screw extruder equipped with a screw and physical property information of the training resin composition.

[0016] The physical property information estimation device disclosed herein is a physical property information estimation device that includes an estimation unit that inputs manufacturing condition information for producing a resin composition from a thermoplastic resin using a screw extruder equipped with a screw into a trained model for estimating physical property information of the resin composition from the manufacturing condition information for producing a resin composition from a thermoplastic resin using a screw extruder equipped with a screw, and estimates physical property information of the resin composition that corresponds to the input manufacturing condition information.

[0017] The physical property information estimation model generating device disclosed herein is a physical property information estimation model generating device that includes a generation unit that generates a trained model for estimating physical property information of a resin composition from manufacturing condition information of a resin composition produced from a thermoplastic resin using a screw extruder equipped with a screw, based on training data representing a combination of manufacturing condition information of a training resin composition produced from a thermoplastic resin using a screw extruder equipped with a screw and physical property information of the training resin composition.

[0018] The physical property information prediction program disclosed herein is a physical property information prediction program for causing a computer to execute a process of inputting manufacturing condition information for producing a resin composition from a thermoplastic resin using a screw extruder equipped with a screw into a trained model for predicting physical property information of the resin composition from the manufacturing condition information for producing a resin composition from a thermoplastic resin using a screw extruder equipped with a screw, and predicting physical property information of the resin composition corresponding to the input manufacturing condition information.

[0019] The physical property information estimation model generation program disclosed herein is a physical property information estimation model generation program for causing a computer to execute a process to generate a trained model for estimating physical property information of a resin composition from manufacturing condition information of a resin composition produced from a thermoplastic resin using a screw extruder equipped with a screw, based on training data representing a combination of manufacturing condition information of a training resin composition produced from a thermoplastic resin using a screw extruder equipped with a screw and physical property information of the training resin composition. Effect of the Invention

[0020] The physical property information estimation method, device, and program disclosed herein have the effect of making it possible to estimate physical property information of a resin composition from production condition information of the resin composition produced by a screw extruder.

[0021] The physical property information estimation model generating method, apparatus, and program disclosed herein have the advantage of being able to obtain a trained model for estimating physical property information of a resin composition from production condition information of the resin composition produced by a screw extruder. [Brief description of the drawings]

[0022] [Figure 1] 1 is a schematic diagram showing an example of the configuration of a physical property information estimation system according to an embodiment of the present invention. [Diagram 2] FIG. 2 is a schematic block diagram of a computer that functions as a physical property information estimation device. [Diagram 3]FIG. 2 is an explanatory diagram for explaining a screw extruder. [Figure 4] FIG. 2 is an explanatory diagram for explaining learning data according to the embodiment; [Diagram 5] FIG. 4 is a diagram showing an example of a learning process routine executed by the physical property information estimation device. [Figure 6] FIG. 2 is a diagram showing an example of a physical property information estimation processing routine executed by the physical property information estimation device. [Figure 7] FIG. 11 is an explanatory diagram for explaining learning data according to the second embodiment. [Figure 8] FIG. 1 shows the results of Example 1 and Example 2. [Figure 9] FIG. 1 shows the results of Examples 3 and 4. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0023] Each embodiment will be described in detail below.

[0024] [First embodiment]

[0025] <System configuration of the physical property information estimation system according to the first embodiment>

[0026] Fig. 1 shows an example of the configuration of a physical property information estimation system 10 according to the first embodiment. Functionally, the physical property information estimation system 10 of the first embodiment can be represented as a configuration including an input device 12, a physical property information estimation device 14, and a display device 30, as shown in Fig. 1. The physical property information estimation system 10 of the present embodiment estimates physical property information of a resin composition produced from a thermoplastic resin by a screw extruder.

[0027] The input device 12 receives information input from the outside and is realized by, for example, a keyboard, a mouse, or an input / output device that receives input from an external device.

[0028] The physical property information estimation device 14 can be configured by a computer including a CPU, a RAM, and a ROM that stores programs for executing each processing routine described below and various data.

[0029] For example, the physical property information estimation device 14 can be realized by a computer 50 shown in Fig. 2. The computer 50 includes a CPU 51, a memory 52 as a temporary storage area, and a non-volatile storage unit 53. The computer 50 also includes an input / output interface (I / F) 54 to which an input / output device or the like (not shown) is connected, and a read / write (R / W) unit 55 that controls reading and writing of data from and to a recording medium. The computer 50 also includes a network I / F 56 that is connected to a network such as the Internet. The CPU 51, the memory 52, the storage unit 53, the input / output I / F 54, the R / W unit 55, and the network I / F 56 are connected to one another via a bus 57.

[0030] The storage unit 53 can be realized by a hard disk drive (HDD), a solid state drive (SSD), a flash memory, etc. The storage unit 53 as a storage medium stores a program for operating the computer 50. The CPU 51 reads the program from the storage unit 53, expands it in the memory 52, and sequentially executes the processes contained in the program.

[0031] As shown in FIG. 1, the physical property information estimation device 14 functionally comprises an information acquisition unit 16, a learning data storage unit 18, a generation unit 20, a learned model storage unit 22, and an estimation unit 24.

[0032] The display device 30 displays the information output from the physical property information estimation device 14 .

[0033] The physical property information estimation device 14 of this embodiment estimates physical property information of a resin composition from information on production conditions of the resin composition using a screw extruder.

[0034] FIG. 3 is an explanatory diagram for explaining a screw extruder. As shown in FIG. 3, a thermoplastic resin, a filler, an additive, etc. are fed into the screw extruder Se to produce a resin composition. The thermoplastic resin, the filler, the additive, etc. fed into the screw extruder Se are kneaded by the screws Sc installed at various positions of the screw extruder Se to produce a resin composition. Note that the term "resin composition" does not only refer to the resin composition itself, but also includes a molded article or a molded body obtained from the resin composition.

[0035] As the screw extruder, a single screw extruder, a twin screw extruder, a multi-screw extruder, etc. can be used. In the case of a twin screw extruder, the rotation direction and combination of the screws can be either co-rotating or counter-rotating, or intermeshing or non-intermeshing. The size of the screw extruder, i.e., the screw diameter and length, can be appropriately selected depending on the amount to be processed.

[0036] The screw refers to a screw that can knead materials by shearing action. Examples of screw elements having kneading ability include a forward kneading disk, a forward mixing screw, a forward rotor screw, a reverse kneading disk, a neutral kneading disk, a reverse mixing screw, or a reverse rotor screw.

[0037] The thermoplastic resin may be one or more resins or copolymer resins selected from the group consisting of polypropylene, polyethylene, polyoxymethylene, polyamide, polycarbonate, polyvinyl chloride, acrylonitrile-butadiene-styrene copolymer resin (ABS resin), acrylonitrile-styrene copolymer resin (AS resin), polyester, and thermoplastic elastomer.

[0038] 3, for example, the screw Sc includes a segmented screw piece P and a screw shaft S. There are various screw configurations, and the physical properties of the resin composition vary depending on the combination of screws installed in the screw extruder.

[0039] In addition, the physical properties of the resin composition vary depending on the temperature of the resin composition when it is produced. For example, as shown in Fig. 3, the physical properties of the resin composition finally produced vary depending on the values ​​of temperatures T1, T2, T3, and T4 at certain points 1, 2, 3, and 4 of the screw extruder.

[0040] Therefore, the physical property information estimation device 14 of the present embodiment generates a trained model for estimating the physical property information of a resin composition from manufacturing condition information for manufacturing a resin composition from a thermoplastic resin using a screw extruder equipped with a screw. Then, the physical property information estimation device 14 of the present embodiment estimates the physical property information of the resin composition using the trained model.

[0041] The specific details will be explained below.

[0042] The input device 12 accepts input of learning data. The learning data in this embodiment is data representing a combination of manufacturing condition information of a learning resin composition produced from a thermoplastic resin using a screw extruder equipped with a screw, and physical property information of the learning resin composition.

[0043] The manufacturing condition information of the resin composition for learning includes information on the screw configuration of the screw extruder, information on additives added to the resin composition, information on the discharge amount of the screw extruder, information on the screw rotation speed of the screw extruder, and information on the temperature when the resin composition is manufactured. In addition, the information on the screw configuration of the screw extruder includes information on the screw shape and screw angle. The information on the screw configuration is expressed, for example, by the screw configuration X1,j (j=1 to an integer, the screw shape and angle).

[0044] The input device 12 also receives manufacturing condition information of a resin composition for which physical property information is to be inferred (hereinafter simply referred to as "manufacturing condition information to be inferred"). The manufacturing condition information to be inferred includes information on the screw configuration of the screw extruder, information on additives added to the resin composition, information on the discharge rate of the screw extruder, information on the screw rotation speed of the screw extruder, and information on the temperature when the resin composition is produced. Furthermore, the information on the screw configuration of the screw extruder includes information on the screw shape and screw angle.

[0045] In addition, the information on the additives added to the resin composition of the present embodiment is described as information on a filler. Note that the temperature information in the manufacturing condition information to be estimated is an actual measured value.

[0046] The information acquiring unit 16 of the physical property information estimation device 14 acquires the learning data accepted by the input device 12. Then, the information acquiring unit 16 stores the learning data in the learning data storage unit 18. In addition, the information acquiring unit 16 acquires manufacturing condition information of the estimation target accepted by the input device 12.

[0047] The learning data storage unit 18 stores the learning data acquired by the information acquisition unit 16. FIG. 4 shows an example of the learning data stored in the learning data storage unit 18. In the example shown in FIG. 4, a plurality of learning data are stored in a table format. For example, the learning data with "ID" of "00001" in the table shown in FIG. 4 indicates that the bending modulus, which is an example of physical property information of the generated resin composition, is "EEE1" when the filler configuration is "AAA1", the screw configuration is "BBB1", the discharge amount of the screw extruder is "CCC1", the rotation speed of the screw extruder is "DDD1", the temperature 1 of a certain point 1 of the screw extruder is "t1", the temperature 2 of another point 2 is "t2", the temperature 3 of another point 3 is "t3", and the temperature 4 of another point 4 is "t4". Based on the plurality of learning data, a trained model, which will be described later, is generated.

[0048] Although FIG. 4 shows a case where the physical property information is the flexural modulus, the physical property information of the resin composition is tensile strength (MPa), tensile elongation (%), flexural strength (MPa), flexural modulus (%), Izod impact strength (KJ / m 2 ), Charpy impact strength (KJ / m 2 ), and compressive yield stress (MPa).

[0049] For example, the information on the filler configuration is shown in Table 1 below. In the table below, "sample code" is identification information for identifying the filler.

[0050] [Table 1]

[0051] When producing the resin composition using a screw extruder, in addition to the talc and mica shown in Table 1 above, glass fiber, clay, silica, alumina, kaolin, quartz, graphite, and the like may be further used.

[0052] The filler may be an inorganic filler, such as, for example, an inorganic compound such as mica, talc, silica, alumina, magnesium hydroxide, barium sulfate, calcium carbonate, titanium oxide, zinc oxide, antimony trioxide, magnesium oxide, kaolinite, basic magnesium carbonate, sericite, montmorolinite, kaolinite, or bentonite.

[0053] The shape of the filler is not particularly limited, and may be any shape. Examples of the shape of the filler include columnar fillers, plate-shaped fillers, spherical or flat fillers, particulate fillers, fibrous fillers, and scale-shaped (layered) fillers. Examples of particulate fillers include glass (e.g., glass beads), calcium carbonate, silica, magnesium carbonate, magnesium oxide, alumina, barium sulfate, carbon black, graphite, ferrite, aluminum hydroxide, magnesium hydroxide, antimony oxide, titanium oxide, and zinc oxide. Examples of plate-shaped fillers include talc, kaolin, mica, and montmorillonite. Examples of fibrous fillers include glass (e.g., glass fiber), carbon fiber, metal fiber, wollastonite, calcium titanate, xonotlite, and basic magnesium sulfate. Examples of scaly fillers include talc, mica, and kaolin.

[0054] The aspect ratios of the fillers shown in Table 1 above were determined by the following measurement method.

[0055] [Particle size D50L (average particle size measured by laser diffraction scattering method)]

[0056] An aqueous dispersion with a filler concentration of 0.1% by mass was prepared, and the particle size distribution (volume basis) was measured using a laser diffraction particle size distribution analyzer (Shimadzu Corporation, product name "SALD-7000") in accordance with JIS Z8825: 2013. In the obtained particle size distribution, the particle size at which the cumulative volume was 50% was defined as D50L.

[0057] [Particle size D50S (average particle size measured by liquid-phase gravity sedimentation method)]

[0058] An aqueous dispersion with a filler concentration of 0.1% by mass was prepared, and the particle size distribution (volume basis) was measured using an X-ray sedimentation particle size distribution analyzer (Micromeritics Japan LLC, product name "SediGraphIII5120") in accordance with ISO13317-3: 2001. In the obtained particle size distribution, the particle size at which the cumulative volume was 50% was defined as D50S.

[0059] Aspect Ratio

[0060] The aspect ratio of the filler was calculated using the above D50L and D50S according to the formula described in the following literature (Pabst, W., Berthold, Ch., Part. Part. Syst. Char., 24, 458-463 (2007); Pabst, W., Berthold, Ch., Gregorova, E., J. Eur. Ceram. Soc., 27, 1759-1762 (2007)).

[0061] The generation unit 20 generates a trained model for inferring physical property information Y of a resin composition from manufacturing condition information X for manufacturing the resin composition, based on multiple learning data stored in the learning data memory unit 18.

[0062] The generation unit 20 of this embodiment generates a trained model (hereinafter simply referred to as a "trained Gaussian process model") for estimating physical property information Y of a resin composition from manufacturing condition information X by a screw extruder using a Gaussian process, which is a type of nonlinear modeling (see, for example, reference (CE Rasmussen, CKI Williams, "Gaussean Process for Machine Learning", The MIT Press, Cambridge, MA, USA, (2006))). Specifically, parameters suitable for estimating physical property information are obtained by the Gaussian process. Then, the generation unit 20 stores the generated trained Gaussian process model in the trained model storage unit 22. The trained Gaussian process model of this embodiment is a combination of parameters obtained by the Gaussian process and a model formula by the Gaussian process.

[0063] The trained model storage unit 22 stores the trained Gaussian process model generated by the generation unit 20.

[0064] The estimation unit 24 acquires the manufacturing condition information of the estimation target acquired by the information acquisition unit 16. Then, based on the manufacturing condition information of the estimation target, the estimation unit 24 generates a plurality of pieces of manufacturing condition information with different information on the screw configuration, information on the filler, information on the discharge amount of the screw extruder, information on the screw rotation speed of the screw extruder, and temperature information. For example, the estimation unit 24 generates 1 million samples as the manufacturing condition information.

[0065] Furthermore, the estimation unit 24 reads out the trained Gaussian process model stored in the trained model storage unit 22. Then, the estimation unit 24 inputs each of the generated pieces of manufacturing condition information to the trained Gaussian process model, and estimates physical property information of the resin composition corresponding to each of the pieces of manufacturing condition information.

[0066] Specifically, when manufacturing condition information to be estimated is input to the trained Gaussian process model, physical property information of the resin composition manufactured under manufacturing conditions corresponding to the input manufacturing condition information is output. More specifically, the trained Gaussian process model obtains a distribution of estimated values ​​(e.g., the average, standard deviation, and attainment probability of the estimated values) for each of a plurality of pieces of manufacturing condition information.

[0067] In this way, by preparing training data in which training production condition information and training physical property information are associated when producing a resin composition using a screw extruder, and generating a trained Gaussian process model using the training data, it is possible to obtain a model for inferring physical property information of a resin composition produced by a screw extruder from the training condition information of the resin composition produced by the screw extruder.

[0068] In addition, by generating a plurality of different pieces of manufacturing condition information from the manufacturing condition information to be estimated, and estimating the physical property information of the resin composition for the plurality of pieces of manufacturing condition information using a trained Gaussian process model, it is possible to obtain an approximate distribution of estimated values, which makes it possible to predict what physical properties of a resin composition will be produced under what manufacturing conditions.

[0069] The display device 30 displays the physical property information estimated by the estimation unit 24 as a result.

[0070] <Functions of the physical property information estimation device>

[0071] Next, a description will be given of the operation of the physical property information estimation device 14. When each piece of learning data representing a combination of learning manufacturing condition information and learning physical property information is input to the physical property information estimation device 14, the information acquisition unit 16 stores the learning data in the learning data storage unit 18. Then, when the physical property information estimation device 14 receives a learning process instruction signal, it executes the learning process routine shown in FIG.

[0072] <Learning processing routine>

[0073] In step S100, the generation unit 20 acquires each piece of learning data stored in the learning data storage unit .

[0074] In step S102, the generation unit 20 generates a Gaussian process model for estimating physical property information from manufacturing condition information based on each of the learning data acquired in step S100. Then, the generation unit 20 acquires a trained Gaussian process model.

[0075] In step S104, the generation unit 20 stores the trained Gaussian process model generated in step S102 in the trained model storage unit 22, and ends the learning processing routine.

[0076] Next, when the physical property information estimation device 14 receives the input of the manufacturing condition information to be estimated, it executes a physical property information estimation process routine shown in FIG.

[0077] <Physical property information estimation processing routine>

[0078] In step S200, the information acquiring unit 16 acquires manufacturing condition information of the estimation target.

[0079] In step S202, the estimation unit 24 generates multiple pieces of manufacturing condition information with different information related to the screw configuration, information related to the filler, information related to the discharge rate of the screw extruder, information related to the screw rotation speed of the screw extruder, and temperature information based on the manufacturing condition information of the estimation target acquired in step S200.

[0080] In step S204, the estimation unit 24 reads out the trained Gaussian process model stored in the trained model storage unit 22.

[0081] In step S206, the estimation unit 24 inputs the multiple pieces of manufacturing condition information generated in step S202 to the trained Gaussian process model read out in step S204, and estimates physical property information.

[0082] In step S208, the estimation unit 24 outputs the physical property information of the resin composition estimated in step S206 as a result.

[0083] As described above, the physical property information estimation device 14 according to the first embodiment inputs manufacturing condition information for producing a resin composition from a thermoplastic resin using a screw extruder equipped with a screw to a trained model for estimating physical property information of the resin composition from the manufacturing condition information for producing a resin composition from a thermoplastic resin using a screw extruder equipped with a screw, and estimates physical property information of the resin composition corresponding to the input manufacturing condition information. This makes it possible to estimate physical property information of a resin composition from manufacturing condition information of a resin composition produced by a screw extruder.

[0084] Moreover, according to the physical property information estimation device 14 of the first embodiment, a plurality of different pieces of manufacturing condition information are generated from the manufacturing condition information to be estimated, and the physical property information of the resin composition for the plurality of pieces of manufacturing condition information is estimated using a trained Gaussian process model, thereby obtaining an approximate distribution of estimated values. This makes it possible to predict what kind of physical properties a resin composition will have under what manufacturing conditions.

[0085] Furthermore, the physical property information estimation device 14 according to the first embodiment generates a trained model for estimating the physical property information of a resin composition from the manufacturing condition information of a resin composition produced from a thermoplastic resin using a screw extruder equipped with a screw, based on training data representing a combination of manufacturing condition information of the resin composition produced from a thermoplastic resin using a screw extruder equipped with a screw and physical property information of the resin composition for training. This makes it possible to obtain a trained model for estimating the physical property information of a resin composition from the manufacturing condition information of the resin composition produced by a screw extruder equipped with a screw.

[0086] The physical property information estimation device 14 according to the first embodiment has been described using a Gaussian process, which is a type of nonlinear modeling, as a model, but is not limited thereto. For example, a model for estimating physical property information of a resin composition from manufacturing condition information of the resin composition may be a support vector machine model, a model based on random forest (for example, see reference (DS Palmer, NM O'Boyle, RC Glen, JBO Mitchell, "Random Forest Models To Predict Aqueous Solubility", J. Chem. Inf. Model. 47 (2007) p.150-158.)), or a neural network model.

[0087] [Second embodiment]

[0088] Next, a second embodiment will be described. The physical property information estimation system of the second embodiment is different from the first embodiment in that it estimates temperature information during the production of a resin composition, and then estimates physical property information of the resin composition from the estimated temperature information, information on the screw configuration, information on the discharge rate of the screw extruder, and information on the screw rotation speed of the screw extruder.

[0089] In addition, since the configuration of the physical property information estimation system according to the second embodiment is similar to that of the first embodiment, the same reference numerals are used and the description thereof is omitted.

[0090] In the learning data storage unit 18 of the second embodiment, learning data acquired by the information acquisition unit 16 is further stored.

[0091] FIG. 7 shows an example of the learning data stored in the learning data storage unit 18 of the second embodiment. In addition to the learning data of the first embodiment, the learning data storage unit 18 of the second embodiment further stores learning data as shown in FIG. 7. For example, the learning data with "ID" of "00001" in the table shown in FIG. 7 indicates that the temperature 1 at a certain point 1 of the screw extruder is "t1", the temperature 2 at another point 2 is "t2", the temperature 3 at another point 3 is "t3", and the temperature 4 at another point 4 is "t4" when the filler configuration is "AAA1", the screw configuration is "BBB1", the discharge amount of the screw extruder is "CCC1", and the rotation speed of the screw extruder is "DDD1". Based on the learning data in which the manufacturing condition information of the learning resin composition and the temperature information when the learning resin composition is manufactured under the manufacturing conditions are associated with each other as shown in FIG. 7, a first trained model to be described later is generated.

[0092] In the first embodiment, a trained Gaussian process model for estimating physical property information of a resin composition from manufacturing condition information for manufacturing the resin composition was generated, and when manufacturing condition information to be estimated was input to the trained Gaussian process model, physical property information of the resin composition manufactured under the manufacturing conditions of the input manufacturing condition information to be estimated was output. Note that, in the first embodiment, the temperature information of the manufacturing condition information in this case was an actual measurement value.

[0093] In contrast, the physical property information estimation device of the second embodiment first estimates temperature information using a first trained model. This first trained model is a model for estimating temperature information when the resin composition is produced from information on the screw configuration, information on the discharge rate of the screw extruder, and information on the screw rotation speed of the screw extruder.

[0094] The physical property information estimation device of the second embodiment estimates the physical property information of the resin composition from the temperature information output from the first trained model, the information on the screw configuration, the discharge amount information of the screw extruder, and the screw rotation speed information of the screw extruder, using the trained Gaussian process model of the first embodiment, which is the second trained model. This model corresponds to the trained Gaussian process model of the first embodiment.

[0095] Specifically, first, the generation unit 20 of the second embodiment generates a first trained model for estimating temperature information when the resin composition is produced from information on the screw configuration, the discharge volume information of the screw extruder, and the screw rotation speed information of the screw extruder, based on the training data stored in the training data memory unit 18.

[0096] More specifically, the generation unit 20 of the second embodiment generates a trained model (hereinafter simply referred to as "trained PLS") for estimating temperature information when the resin composition is produced from information on the production conditions of the resin composition, such as information on the screw configuration, information on the discharge rate of the screw extruder, and information on the screw rotation speed of the screw extruder, by partial least squares regression (PLS). Parameters suitable for estimating temperature information are obtained by partial least squares regression. Then, the generation unit 20 stores the generated trained PLS in the trained model storage unit 22. The trained PLS of this embodiment is a combination of parameters obtained by partial least squares regression and a model formula by PLS. The trained PLS is an example of a first trained model.

[0097] Similarly to the first embodiment, the generating unit 20 of the second embodiment generates a trained Gaussian process model for estimating physical property information of the resin composition from temperature information, information on the screw configuration, discharge rate information of the screw extruder, and screw rotation speed information of the screw extruder, based on the training data stored in the training data storage unit 18. The trained Gaussian process model is an example of a second trained model.

[0098] In the trained model storage unit 22 of the second embodiment, a trained PLS and a trained Gaussian process model are stored.

[0099] The estimation unit 24 of the second embodiment acquires manufacturing condition information of the estimation target acquired by the information acquisition unit 16. Note that the manufacturing condition information of the estimation target of the second embodiment does not include temperature information.

[0100] Then, the estimation unit 24 generates a plurality of pieces of manufacturing condition information, which are different in information on the screw configuration, information on the filler, information on the discharge rate of the screw extruder, and information on the screw rotation speed of the screw extruder, based on the manufacturing condition information to be estimated. For example, the estimation unit 24 generates 1 million samples as the manufacturing condition information.

[0101] Next, the estimation unit 24 reads out the trained PLS stored in the trained model storage unit 22. Then, the estimation unit 24 inputs the generated multiple pieces of manufacturing condition information to the trained PLS, and outputs an estimated value of temperature information corresponding to each of the multiple pieces of manufacturing condition information.

[0102] Next, the estimation unit 24 reads out the trained Gaussian process model stored in the trained model storage unit 22. Then, the estimation unit 24 inputs the estimated value of the temperature information output from the trained PLS and the manufacturing condition information to the trained Gaussian process model to obtain an estimated value of the physical property information of the resin composition.

[0103] As described above, the physical property information estimation device according to the second embodiment uses the first trained model to estimate temperature information when the resin composition is produced from information on the screw configuration, the discharge amount information of the screw extruder, and the screw rotation speed information of the screw extruder, which are information on the production conditions of the resin composition by the screw extruder. Then, the physical property information estimation device according to the second embodiment uses the second trained model to estimate physical property information of the resin composition from the temperature information output from the first trained model, information on the screw configuration, the discharge amount information of the screw extruder, and the screw rotation speed information of the screw extruder. This makes it possible to estimate physical property information of the resin composition produced under the conditions of the production condition information to be estimated, without actually measuring the temperature information.

[0104] In addition, the physical property information estimation device of the second embodiment has been described as an example in which a first trained model is generated by partial least squares regression, which is a type of linear modeling, and a second trained model is generated by a Gaussian process, which is a type of nonlinear modeling, but this is not limited to this and any model may be used. EXAMPLES

[0105] Next, examples of each embodiment will be described. Four examples will be described below.

[0106] [Example 1]

[0107] The results of Example 1 are shown on the left side of Fig. 8. The plot points shown in Fig. 8 are plot points that represent the relationship between the predicted value Y' of the physical property estimated by the trained Gaussian process model (when information on the screw configuration, information on the additives added to the resin composition, information on the discharge rate of the screw extruder, and information on the screw rotation speed of the screw extruder are used as the production condition information) and the actual measured value Y.

[0108] [Example 2]

[0109] The results of Example 2 are shown on the right side of Fig. 8. The plot points shown in Fig. 8 are plot points that represent the relationship between the predicted value Y' of the physical property estimated by a trained random forest model (when information on the screw configuration, information on additives added to the resin composition, information on the discharge rate of the screw extruder, and information on the screw rotation speed of the screw extruder are used as production condition information) and the actual measured value Y.

[0110] [Example 3]

[0111] The results of Example 3 are shown on the left side of Fig. 9. The plot points shown in Fig. 9 are plot points that represent the relationship between the predicted value Y' of the physical property estimated by a trained Gaussian process model (when the temperature information during the production of the resin composition, the information on the screw configuration, the information on the additives added to the resin composition, the discharge amount information of the screw extruder, and the screw rotation speed information of the screw extruder are used as the production condition information) and the actual measured value Y. Note that the temperature information used in this case was predicted by the trained PLS of the second embodiment.

[0112] [Example 4]

[0113] The results of Example 4 are shown on the right side of Fig. 9. The plot points shown in Fig. 9 are plot points that represent the relationship between the predicted value Y' of the physical property estimated by a trained random forest model (when the temperature information during the production of the resin composition, the information on the screw configuration, the information on the additives added to the resin composition, the discharge amount information of the screw extruder, and the screw rotation speed information of the screw extruder are used as the production condition information) and the actual measured value Y. Note that the temperature information used in this case was predicted by the trained PLS of the second embodiment.

[0114] Next, the above results are summarized in Table 2 below.

[0115] [Table 2]

[0116] The "predicted value Y'" and the "actual value Y" shown in Table 2 represent the predicted value and the actual value for a certain production condition information. As shown in Table 2, in all the examples, it can be seen that the physical property information of the resin composition is accurately estimated from the production condition information of the resin composition by the screw extruder.

[0117] The present invention is not limited to the above-described embodiment, and various modifications and applications are possible without departing from the spirit and scope of the present invention.

[0118] For example, the screw extruder may in principle have only one screw, and one or more thermometers are arbitrarily installed at the screw installation locations as the installation locations of the thermometers used when collecting learning data. Also, for example, when four screws are installed in a screw extruder, one thermometer may be used, but if the tip of the screw is the upstream side from the direction in which the resin flows and the direction in which the resin comes out of the extruder is the downstream side, it is better to install the thermometer on the upstream side, and it is preferable to install multiple thermometers.

[0119] In addition, the resin composition produced by the screw extruder may further contain other components within a range that does not impair the effects of the present invention. Examples of other components include antioxidants (phenols, amines, sulfurs, phosphorus, etc.), light stabilizers (benzotriazoles, triazines, benzophenones, benzoates, hindered amines, oxanilides, etc.), other polymers (polyolefins, olefin copolymers such as ethylene-propylene copolymers, ethylene-1-butene copolymers, olefin copolymers such as propylene-1-butene copolymers, polystyrene, polyamide, polycarbonate, polyacetal, polysulfone, polyphenylene oxide, fluororesins, silicone resins, LCP, etc.), flame retardants (bromine-based, chlorine-based, phosphorus-based, antimony-based, inorganic, etc.), release agents, flow improvers, fluorescent whitening agents, plasticizers, thickeners, antistatic agents, pigments, crystal nucleating agents, etc.

[0120] Furthermore, although the present specification has been described as an embodiment in which the program is pre-installed, the program may also be provided by being stored on a computer-readable recording medium. [Explanation of symbols]

[0121] 10. Physical property information estimation system 12 Input Devices 14 Physical property information estimation device 16 Information acquisition department 18 Learning data storage unit 20 Generation part 22 Trained model memory 24 Guessing part 30 Display device 50 Computer

Claims

1. inputting production condition information for producing a resin composition from a thermoplastic resin using a screw extruder equipped with a screw into a trained model for predicting physical property information of the resin composition from the production condition information for producing a resin composition from a thermoplastic resin using a screw extruder equipped with a screw, and predicting physical property information of the resin composition corresponding to the input production condition information; A physical property information estimation method in which processing is executed by a computer, The manufacturing condition information includes information on additives added to the resin composition, information on a screw configuration of the screw extruder, information on a discharge amount of the screw extruder, and information on a screw rotation speed of the screw extruder, The trained model is A first trained model that estimates temperature information when the resin composition is produced from information on additives added to the resin composition, information on the screw configuration of the screw extruder, information on the discharge rate of the screw extruder, and information on the screw rotation speed of the screw extruder; A second trained model that estimates physical property information of the resin composition from the temperature information output from the first trained model, information on additives added to the resin composition, information on the screw configuration of the screw extruder, information on the discharge rate of the screw extruder, and information on the screw rotation speed of the screw extruder, The physical property information includes any one of the flexural modulus, tensile strength, tensile elongation, flexural strength, Izod impact strength, Charpy impact strength, and compressive yield stress of the resin composition. Method for estimating physical property information.

2. The information on the additives added to the resin composition is information on a filler, The physical property information estimation method according to claim 1 .

3. The manufacturing condition information includes information on the screw configuration of the screw extruder, information on additives to be added to the composition, information on the discharge rate of the screw extruder, and information on the screw rotation speed of the screw extruder; Generate a plurality of pieces of manufacturing condition information, each of which has different information on a screw configuration of the screw extruder, information on additives to be added to the resin composition, information on a discharge amount of the screw extruder, and information on a screw rotation speed of the screw extruder; Inputting each of the plurality of pieces of generated manufacturing condition information into the trained model, and inferring the physical property information corresponding to each of the plurality of pieces of manufacturing condition information. The physical property information estimation method according to claim 1 .

4. The trained model is a model based on PLS (Partial Least Squares regression), A port vector machine model, a random forest model, a Gaussian process model, or a neural network model. The physical property information estimation method according to any one of claims 1 to 3.

5. A trained model is generated for inferring physical property information of a resin composition from manufacturing condition information of a resin composition produced from a thermoplastic resin of which physical property information is to be inferred, based on training data representing a combination of manufacturing condition information of the resin composition produced from a thermoplastic resin using a screw extruder equipped with a screw and physical property information of the resin composition. A physical property information estimation model generating method in which a computer executes a process, The manufacturing condition information of the resin composition for which the manufacturing condition information and physical property information of the resin composition for learning are to be inferred includes information on additives added to the resin composition, information on the screw configuration of the screw extruder, information on the discharge amount of the screw extruder, and information on the screw rotation speed of the screw extruder, The trained model is A first trained model that estimates temperature information when the resin composition is produced from information on additives added to the resin composition, information on the screw configuration of the screw extruder, information on the discharge rate of the screw extruder, and information on the screw rotation speed of the screw extruder; A second trained model that estimates physical property information of the resin composition from the temperature information output from the first trained model, information on additives added to the resin composition, information on the screw configuration of the screw extruder, information on the discharge rate of the screw extruder, and information on the screw rotation speed of the screw extruder, The physical property information includes any one of the flexural modulus, tensile strength, tensile elongation, flexural strength, Izod impact strength, Charpy impact strength, and compressive yield stress of the resin composition. A method for generating models to estimate physical property information.

6. A prediction unit that inputs production condition information for producing a resin composition from a thermoplastic resin using a screw extruder equipped with a screw into a trained model for predicting physical property information of the resin composition from the production condition information for producing a resin composition from a thermoplastic resin using a screw extruder equipped with a screw, and predicts physical property information of the resin composition corresponding to the input production condition information. A physical property information estimation device comprising: The manufacturing condition information includes information on additives added to the resin composition, information on a screw configuration of the screw extruder, information on a discharge amount of the screw extruder, and information on a screw rotation speed of the screw extruder, The trained model is A first trained model that estimates temperature information when the resin composition is produced from information on additives added to the resin composition, information on the screw configuration of the screw extruder, information on the discharge rate of the screw extruder, and information on the screw rotation speed of the screw extruder; A second trained model that estimates physical property information of the resin composition from the temperature information output from the first trained model, information on additives added to the resin composition, information on the screw configuration of the screw extruder, information on the discharge rate of the screw extruder, and information on the screw rotation speed of the screw extruder, The physical property information includes any one of the flexural modulus, tensile strength, tensile elongation, flexural strength, Izod impact strength, Charpy impact strength, and compressive yield stress of the resin composition. Physical property information estimation device.

7. A generation unit that generates a trained model for estimating physical property information of a resin composition from manufacturing condition information of a resin composition for which physical property information is to be estimated, based on training data representing a combination of manufacturing condition information of a training resin composition produced from a thermoplastic resin using a screw extruder equipped with a screw and physical property information of the training resin composition. A physical property information estimation model generating device comprising: The manufacturing condition information of the resin composition for which the manufacturing condition information and physical property information of the resin composition for learning are to be inferred includes information on additives added to the resin composition, information on the screw configuration of the screw extruder, information on the discharge amount of the screw extruder, and information on the screw rotation speed of the screw extruder, The trained model is A first trained model that estimates temperature information when the resin composition is produced from information on additives added to the resin composition, information on the screw configuration of the screw extruder, information on the discharge rate of the screw extruder, and information on the screw rotation speed of the screw extruder; A second trained model that estimates physical property information of the resin composition from the temperature information output from the first trained model, information on additives added to the resin composition, and at least one of information on the screw configuration of the screw extruder, information on the discharge rate of the screw extruder, and information on the screw rotation speed of the screw extruder. The physical property information includes any one of the flexural modulus, tensile strength, tensile elongation, flexural strength, Izod impact strength, Charpy impact strength, and compressive yield stress of the resin composition. A device for generating models to estimate physical property information.

8. inputting production condition information for producing a resin composition from a thermoplastic resin using a screw extruder equipped with a screw into a trained model for predicting physical property information of the resin composition from the production condition information for producing a resin composition from a thermoplastic resin using a screw extruder equipped with a screw, and predicting physical property information of the resin composition corresponding to the input production condition information; A physical property information estimation program for causing a computer to execute a process, The manufacturing condition information includes information on additives added to the resin composition, information on a screw configuration of the screw extruder, information on a discharge amount of the screw extruder, and information on a screw rotation speed of the screw extruder, The trained model is A first trained model that estimates temperature information when the resin composition is produced from information on additives added to the resin composition, information on the screw configuration of the screw extruder, information on the discharge rate of the screw extruder, and information on the screw rotation speed of the screw extruder; A second trained model that estimates physical property information of the resin composition from the temperature information output from the first trained model, information on additives added to the resin composition, information on the screw configuration of the screw extruder, information on the discharge rate of the screw extruder, and information on the screw rotation speed of the screw extruder, The physical property information includes any one of the flexural modulus, tensile strength, tensile elongation, flexural strength, Izod impact strength, Charpy impact strength, and compressive yield stress of the resin composition. A program for predicting physical property information.

9. A trained model is generated for inferring physical property information of a resin composition from manufacturing condition information of a resin composition for training, the resin composition being produced from a thermoplastic resin using a screw extruder equipped with a screw, based on training data representing a combination of manufacturing condition information of the resin composition for training and physical property information of the resin composition for training, the physical property information of which is to be inferred. A physical property information estimation model generation program for causing a computer to execute a process, The manufacturing condition information of the resin composition for which the manufacturing condition information and physical property information of the resin composition for learning are to be inferred includes information on additives added to the resin composition, information on the screw configuration of the screw extruder, information on the discharge amount of the screw extruder, and information on the screw rotation speed of the screw extruder, The trained model is A first trained model that estimates temperature information when the resin composition is produced from information on additives added to the resin composition, information on the screw configuration of the screw extruder, information on the discharge rate of the screw extruder, and information on the screw rotation speed of the screw extruder; A second trained model that estimates physical property information of the resin composition from the temperature information output from the first trained model, information on additives added to the resin composition, information on the screw configuration of the screw extruder, information on the discharge rate of the screw extruder, and information on the screw rotation speed of the screw extruder, The physical property information includes any one of the flexural modulus, tensile strength, tensile elongation, flexural strength, Izod impact strength, Charpy impact strength, and compressive yield stress of the resin composition. A program for generating models to predict physical property information.

Images