Information processor, prediction method, prediction program, and method of manufacturing molded article

Abstract

To achieve an information processor that accurately predicts whether quality of a target molded article deviates a specification.SOLUTION: An information processor (2) comprises: a condition acquisition section (201) for acquiring manufacturing conditions including a blending ratio condition indicating a ratio when a plurality of raw materials are blended, and a physical property condition indicating physical properties of the plurality of raw materials; a quality prediction section (202) for predicting quality of a molded article using a prediction model; and a quality correction section (203) for correcting quality of the molded article using a correction model.SELECTED DRAWING: Figure 1

Description

[Technical Field]

[0001] The present invention relates to an information processing device, a prediction method, a prediction program, and a method for manufacturing a molded product. [Background technology]

[0002] A method for molding a fiber-reinforced polyphenylene sulfide resin composition by blending glass fiber and various additives with polyphenylene sulfide resin is known. While the blending ratio and other parameters of such fiber-reinforced polyphenylene sulfide resin compositions have traditionally been determined by trial and error, the invention described in Patent Document 1 uses a learning model to efficiently determine the blending ratio and other parameters.

[0003] A technique is known in which additives are added to a polymer extracted from a microorganism, the mixture is kneaded, and the mixture is processed into pellets (for example, Patent Document 2). Patent Document 2 describes that a composition obtained from kenaf fiber and poly(3-hydroxyalkanoate) is processed into pellets using an extruder molding machine. [Prior art documents] [Patent documents]

[0004] [Patent Document 1] Japanese Patent Application Publication No. 2023-51839 [Patent Document 2] International Publication No. 2005 / 054366 Summary of the Invention [Problem to be solved by the invention]

[0005] When using a single learning model, as in the invention described in Patent Document 1, data covering a wide variety of variables related to raw materials and processing is required to generate the learning model, and the amount of data required is large in both the number of items and the number of cases. It can be difficult to prepare a large amount of data due to constraints such as manufacturing sites and production plans, and there is room for improvement.

[0006] When the culture lot of the polymer is changed, the physical properties also change, and there is a risk that the pellets may deviate from the quality standard due to the difference in physical properties. This point is not mentioned in Patent Document 2, and there is room for improvement.

[0007] An object of one aspect of the present invention is to realize an information processing device that accurately predicts whether the quality of a target molded product deviates from a standard. [Means for solving the problem]

[0008] In order to solve the above-mentioned problems, an information processing device according to one embodiment of the present invention includes: a condition acquisition unit that acquires manufacturing conditions for a molded product produced by blending at least two or more raw materials extracted from microorganisms, the manufacturing conditions including blending ratio conditions indicating the ratio when blending the multiple raw materials and physical property conditions indicating the physical properties of the multiple raw materials; a quality prediction unit that predicts the quality of the molded product using a prediction model in which the manufacturing conditions acquired by the condition acquisition unit are explanatory variables and the quality of the molded product is a target variable; and a quality correction unit that corrects the quality of the molded product predicted by the quality prediction unit using a correction model in which the quality conditions indicating the quality of the molded product output from the prediction model and the processing conditions used to manufacture the molded product are explanatory variables and the quality of the molded product is a target variable.

[0009] In order to solve the above-mentioned problems, a prediction method according to one embodiment of the present invention is a prediction method executed by an information processing device, and includes: a condition acquisition step of acquiring manufacturing conditions for a molded product produced by blending at least two or more raw materials extracted from microorganisms, the manufacturing conditions including blending ratio conditions indicating the ratio when blending the multiple raw materials and physical property conditions indicating the physical properties of the multiple raw materials; a quality prediction step of predicting the quality of the molded product using a prediction model in which the manufacturing conditions acquired in the condition acquisition step are used as explanatory variables and the quality of the molded product is used as a target variable; and a quality correction step of correcting the quality of the molded product predicted in the quality prediction step using a correction model in which the quality conditions indicating the quality of the molded product output from the prediction model and the processing conditions used to manufacture the molded product are used as explanatory variables and the quality of the molded product is used as a target variable.

[0010] The information processing device according to each aspect of the present invention may be realized by a computer. In this case, a control program that causes the computer to operate as each unit (software element) of the information processing device to realize the information processing device on the computer, and a computer-readable recording medium on which the control program is recorded, also fall within the scope of the present invention. In addition, such a control program may be stored in a cloud server and provided so as to be downloadable from the cloud server. [Effects of the Invention]

[0011] According to one aspect of the present invention, it is possible to accurately predict whether the quality of a target molded product deviates from a standard. [Brief explanation of the drawings]

[0012] [Figure 1] 1 is a block diagram illustrating an example of the configuration of an information processing device 1 and an information processing device 2 according to a first embodiment of the present invention. [Figure 2] FIG. 2 is a diagram illustrating an example of the flow of various data. [Figure 3] 10 is a flowchart showing an example of a method for generating a prediction model 211 and a correction model 212. [Figure 4] 1 is a flowchart illustrating an example of a pellet quality prediction method and a pellet quality correction method. [Figure 5] 1 is a graph showing the relationship between the actually measured value and the predicted value of the MFR potential. [Figure 6] 1 is a graph showing the relationship between resin temperature and the ratio of MFR to MFR potential. [Figure 7] 1 is a graph showing the relationship between the actual measured value and the predicted value of MFR. [Figure 8] 10 is a block diagram illustrating an example of the configuration of an information processing device 1 and an information processing device 2 according to a second embodiment of the present invention. FIG. [Figure 9] 10 is a flowchart illustrating an example of an optimization method. DETAILED DESCRIPTION OF THE INVENTION

[0013] [Embodiment 1] Hereinafter, a first embodiment of the present invention will be described in detail with reference to the drawings. In the description of the drawings, the same parts are given the same reference numerals and the description will be omitted.

[0014] (System Overview) An overview of the system in this embodiment will be described. As an example, the system in this embodiment is a system used to predict whether the quality of a target molded product deviates from a standard. The "molded product" referred to here is not particularly limited, but is, for example, pellets used in the manufacture of plastic products. Such a system includes information processing devices 1 and 2 shown in FIG. 1.

[0015] The information processing device 1 generates a prediction model and a correction model through machine learning using training data, and the information processing device 2 uses the prediction model and correction model to predict whether the quality of the molded product deviates from the standard based on the prediction data.

[0016] (Configuration examples of information processing device 1 and information processing device 2) FIG. 1 is a block diagram showing an example of the configuration of an information processing device 1 and an information processing device 2. As shown in FIG.

[0017] 1, the information processing device 1 includes a control unit 10 that controls each unit of the information processing device 1, and a storage unit 11 that stores various data used by the information processing device 1. The information processing device 1 also includes a communication unit 12 that enables the information processing device 1 to communicate with other devices, an input unit 13 that accepts various data input to the information processing device 1, and an output unit 14 that enables the information processing device 1 to output various data. The control unit 10 also includes a training data acquisition unit 101 and a learning unit 102. The storage unit 11 stores training data 111.

[0018] The information processing device 1 may be configured as a general-purpose computer. When the information processing device 1 is configured as a general-purpose computer, the control unit 10 may be configured to include a processor, a read-only memory (ROM), a random access memory (RAM), and the like. The processor can read programs from the ROM and execute various programs using the RAM as a work area. The storage unit 11 may be configured as a hard disk drive (HDD) or a solid state drive (SSD). The storage unit 11 does not necessarily have to be provided in the information processing device 1. In this case, for example, the storage unit 11 may be provided in an external server. Such an external server may be a virtual server running on a hypervisor or a cloud server. When the storage unit 11 is provided in an external server, the information processing device 1 can access the storage unit 11 via a communication network such as a wireless local area network (LAN). As described above, this embodiment includes both a case where the information processing device 1 is physically connected to the storage unit 11 and a case where the information processing device 1 is not physically connected to the storage unit 11.

[0019] The input unit 13 is typically a keyboard or a mouse, but is not limited to these, and may be any configuration that allows various data to be input. For example, the input unit 13 may be a microphone, a touch panel, etc. The output unit 14 is typically a display, but is not limited to these, and may be any configuration that allows various data to be output. For example, the output unit 14 may be a speaker, a warning light, etc.

[0020] The training data 111 stored in the storage unit 11 will be described. The training data 111 is labeled data used in machine learning. Machine learning is a technique for learning features contained in input data and generating a "model" that predicts results corresponding to newly input data. A prediction model 211 and a correction model 212, which will be described later, are trained models generated by machine learning.

[0021] The training data 111 is generated by associating input data with correct answer data and stored in the storage unit 11. An example of the "input data" here is data indicating the manufacturing conditions of a molded product manufactured by blending multiple raw materials. As mentioned above, an example of a "molded product" is pellets used in the manufacture of plastic products. While raw materials used in the manufacture of such pellets include polymers and additives, the following description will mainly focus on polymers. Therefore, the "manufacturing conditions" here may also be the manufacturing conditions of pellets manufactured by blending at least multiple polymers.

[0022] A specific example of the "production conditions" will be described. The production conditions may include blending ratio conditions indicating the blending ratio of multiple polymers and physical property conditions indicating the physical properties of the multiple polymers. The type of polymer is not particularly limited, but one example may be a polyhydroxyalkanoate resin. Polyhydroxyalkanoate resins have excellent seawater degradability and are therefore attracting attention as materials that can solve environmental problems caused by discarded plastics. Such polymers are produced by a microbial fermentation process. All of the polymers to be blended may be polyhydroxyalkanoate resins, or at least one of the polymers to be blended may be a polyhydroxyalkanoate resin.

[0023] In the compounding process of this embodiment, predetermined additives are added to a polymer extracted from a microorganism, which is then melted and kneaded and processed into pellets. The raw polymers are classified into predetermined types based on their molecular weight structure and functional group composition, and pellets are typically produced by blending two to four types of polymers. The melting properties of the pellets and their physical properties after molding can be manipulated by changing the types and blending ratios of the polymers blended. The produced pellets are classified into predetermined grades depending on their intended use. For each grade, a standard recipe is established that defines the types and blending ratios of polymers to reproduce quality.

[0024] As described above, the production process of the polymers used to manufacture pellets is based on biochemical processes, and therefore, differences in physical properties may occur between culture lots of the same type of polymer. Therefore, even if the polymers are blended according to a standard recipe, the desired quality may not be obtained, i.e., the quality may deviate from the specifications. Therefore, in this embodiment, the quality of the pellets is predicted based on the blending ratio conditions and physical property conditions of the polymers.

[0025] Next, an example of "physical property conditions" will be explained. Physical property conditions may include molecular weight data, thermal stability, thermal decomposition temperature, melt flow rate (MFR), etc. of each polymer to be blended. Molecular weight data may include number average molecular weight (Mn), weight average molecular weight (Mw), molecular weight distribution (Mw / Mn), etc. Melt flow rate is a well-known index for evaluating the fluidity of resins. "Thermal stability" here refers to the degree of molecular weight reduction resulting from treatment at a specified temperature and pressure for a specified time, and is expressed in units of %.

[0026] The "physical property conditions" may also include the content of repeating units of a predetermined functional group possessed by each polymer to be blended. It is sufficient for the "physical property conditions" to include at least one of these pieces of data. Of course, the "physical property conditions" may also include all of these pieces of data.

[0027] Next, an example of "correct data" will be described. Correct data is data that indicates the quality of pellets manufactured according to the blending ratio conditions and physical property conditions, using "blending ratio conditions" and "physical property conditions" as input data. Hereinafter, "data indicating the quality of pellets manufactured according to the blending ratio conditions and physical property conditions" may be referred to as the MFR potential. As will be described in detail later, in this embodiment, two indicators, "MFR potential" and "MFR," are used to indicate "pellet quality." The definition of the MFR potential is not limited to the above definition. For example, the MFR potential may be defined as the MFR that is not affected by changes or fluctuations in processing conditions, such as when all samples are manufactured under the same conditions. According to this definition, the relative differences in MFR potential between samples can be explained solely by the blending ratio conditions and physical property conditions.

[0028] The training data 111 is generated by associating the above-mentioned input data with the correct answer data, and is stored in the storage unit 11.

[0029] The training data acquisition unit 101 refers to the storage unit 11 to acquire training data 111. The training data acquisition unit 101 outputs the acquired training data 111 to the learning unit .

[0030] The learning unit 102 generates a prediction model 211 for predicting the MFR potential of pellets from the prediction data through machine learning using the training data 111 acquired from the training data acquisition unit 101. The prediction model 211 is a computational model that uses the blending ratio conditions of multiple polymers and the physical property conditions of multiple polymers as explanatory variables and the MFR potential of pellets as a response variable. The machine learning algorithm is not particularly limited. For example, the learning unit 102 may generate the prediction model 211 using a neural network, or may generate the prediction model 211 using regression analysis, random forest, or the like.

[0031] The prediction model 211 generated by the learning unit 102 is transmitted to the information processing device 2 and stored in the storage unit 21 of the information processing device 2. Note that the prediction model 211 generated by the learning unit 102 may be stored in the storage unit 11.

[0032] Next, the information processing device 2 will be described. The information processing device 2 includes a control unit 20 that controls each unit of the information processing device 2, and a storage unit 21 that stores various data used by the information processing device 2. The information processing device 2 also includes a communication unit 22 that enables the information processing device 2 to communicate with other devices, an input unit 23 that accepts various data input to the information processing device 2, and an output unit 24 that enables the information processing device 2 to output various data. The control unit 20 also includes a condition acquisition unit 201, a quality prediction unit 202, a quality correction unit 203, and an output control unit 205. The storage unit 21 stores a prediction model 211 and a correction model 212.

[0033] The information processing device 2 may also be configured as a general-purpose computer, similar to the information processing device 1. When the information processing device 2 is configured as a general-purpose computer, the control unit 20, the storage unit 21, the communication unit 22, the input unit 23, and the output unit 24 may have the same configurations as the control unit 10, the storage unit 11, the communication unit 12, the input unit 13, and the output unit 14 of the information processing device 1. In other words, the storage unit 21 may also be provided in an external server, rather than in the information processing device 2. This embodiment includes both a case where the information processing device 2 is physically connected to the storage unit 21 and a case where the information processing device 2 is not physically connected to the storage unit 21.

[0034] The condition acquisition unit 201 acquires blending ratio conditions for a plurality of polymers and physical property conditions for a plurality of polymers. The method of acquiring the blending ratio conditions and physical property conditions is not particularly limited, and for example, the condition acquisition unit 201 may acquire data input into the input unit 23 by an operator engaged in pellet production as the blending ratio conditions and physical property conditions. The condition acquisition unit 201 outputs the blending ratio conditions and physical property conditions to the quality prediction unit 202. The blending ratio conditions and physical property conditions acquired by the condition acquisition unit 201 can be said to be data for predicting the MFR potential of the pellets.

[0035] The quality prediction unit 202 predicts the MFR potential of the pellets from the blending ratio conditions and physical property conditions, which are prediction data, using a prediction model 211 generated by learning the relationship between the blending ratio conditions, physical property conditions, and the MFR potential of the pellets. Specifically, the quality prediction unit 202 can predict the MFR potential of the pellets by inputting the blending ratio conditions and physical property conditions acquired by the condition acquisition unit 201 into the prediction model 211. The MFR potential of the pellets predicted by the quality prediction unit 202 is output to the quality correction unit 203.

[0036] The quality correction unit 203 corrects the MFR potential of the pellet predicted by the quality prediction unit 202 using a correction model 212. The "correction model 212" will be described. Like the prediction model 211, the correction model 212 is also a model generated by machine learning. However, the input data and correct data of the correction model 212 are different from those of the prediction model 211.

[0037] The input data for generating the correction model 212 includes the MFR potential output from the prediction model 211 and the processing conditions for producing pellets. The "processing conditions" may include the resin temperature at the tip of the extruder, the resin pressure at the tip of the extruder, the feed rate of the raw material to the extruder, the screw rotation speed of the extruder, etc. The feed rate here refers to the amount of raw material fed to the extruder per unit time.

[0038] The correct data used to generate the correction model 212 is data that indicates the quality of pellets manufactured according to the MFR potential and processing conditions, using the "MFR potential" and "processing conditions" as input data. Hereinafter, the "data that indicates the quality of pellets manufactured according to the MFR potential and processing conditions" may be referred to as MFR.

[0039] A data set generated by associating such input data with corrective data serves as training data for generating the corrected model 212. That is, in this embodiment, the training data 111 includes two data sets. One is a data set for generating the prediction model 211, and the other is a data set for generating the corrected model 212.

[0040] The training data acquisition unit 101 may acquire a necessary data set from the training data 111 according to the model to be generated, and output the data set to the learning unit 102. This allows the learning unit 102 to generate both the prediction model 211 and the correction model 212 according to the data set.

[0041] The correction model 212 is a calculation model that uses the MFR potential and processing conditions as explanatory variables and the MFR of the pellets as a target variable. The machine learning algorithm used to generate the correction model 212 is not particularly limited, as is the case when generating the prediction model 211. The learning unit 102 may generate the correction model 212 using a neural network, or may generate the correction model 212 using regression analysis, random forest, or the like. The processing conditions may also be acquired by the condition acquisition unit 201.

[0042] The quality correction unit 203 can correct the MFR potential by inputting the MFR potential predicted by the quality prediction unit 202 and the processing conditions into the correction model 212 generated in this manner. In this embodiment, "correcting the MFR potential of the pellet" is synonymous with "predicting the MFR of the pellet." The MFR potential of the pellet may be rephrased as data indicating the quality of the pellet before correction, and the MFR of the pellet may be rephrased as data indicating the quality of the pellet after correction.

[0043] The output control unit 205 causes the output unit 24 to output the prediction result obtained by the quality correction unit 203. The output mode is arbitrary, and for example, the output control unit 205 may cause the prediction result to be output in at least one of display output, audio output, and print output. Furthermore, the device that outputs the prediction result may be an external device to the information processing device 2.

[0044] Furthermore, the output control unit 205 may output information according to the prediction result by the quality correction unit 203. For example, when the prediction result by the quality correction unit 203 indicates a sign of deviation from the standard, the output control unit 205 may output a notification urging correction of the blending ratio of the polymers.

[0045] As described above, the information processing device 2 includes a condition acquisition unit 201 that acquires manufacturing conditions for a molded product produced by blending at least two or more raw materials extracted from microorganisms, including blending ratio conditions indicating the ratio when blending the multiple raw materials and physical property conditions indicating the physical properties of the multiple raw materials; a quality prediction unit 202 that predicts the quality of the molded product using a prediction model in which the manufacturing conditions acquired by the condition acquisition unit 201 are used as explanatory variables and the quality of the molded product is used as a target variable; and a quality correction unit 203 that corrects the quality of the molded product predicted by the quality prediction unit 202 using a correction model in which the quality conditions indicating the quality of the molded product output from the prediction model and the processing conditions used to manufacture the molded product are used as explanatory variables and the quality of the molded product is used as a target variable.

[0046] (Example of data flow) Next, an example of the data flow when predicting the quality of pellets will be described with reference to FIG.

[0047] The prescription data 26 shown in FIG. 2 includes at least the blending ratio conditions of multiple polymers specified in the standard prescription for each grade. The raw material data 27 includes at least the physical property conditions for each polymer. The process data 28 includes at least the processing conditions for producing pellets. The prescription data 26 and raw material data 27 acquired by the condition acquisition unit 201 are input to a prediction model 211, and the MFR potential is output from the prediction model 211. The prescription data 26 and raw material data 27 are explanatory variables in the prediction model 211, and the MFR potential is a target variable in the prediction model 211.

[0048] The MFR potential output from the prediction model 211 and the process data 28 are input to the correction model 212, and the MFR is output from the correction model 212. The MFR potential and the process data 28 are explanatory variables in the correction model 212, and the MFR is a response variable in the correction model 212.

[0049] The prescription data 26 and raw material data 27 used to generate the prediction model 211 and the MFR potential and process data 28 used to generate the correction model 212 may be acquired in different facilities. For example, the prescription data 26 and raw material data 27 may be acquired in research facilities, while the MFR potential and process data 28 may be acquired in manufacturing facilities. This makes it possible to acquire data for model generation without affecting, for example, the pellet production plan. In the following, it is assumed that the prescription data 26, raw material data 27, and MFR potential are acquired in research facilities, and the process data 28 and MFR are acquired in manufacturing facilities. However, as long as conditions such as not interfering with the pellet production plan are met, the prescription data 26, raw material data 27, MFR potential, process data 28, and MFR may be acquired in the same facility (e.g., manufacturing facility).

[0050] Note that prescription data 26 and raw material data 27, and MFR potential and process data 28 are acquired in different facilities only in the process of generating training data 111 for model generation. That is, in a scene where generation of prediction model 211 and correction model 212 is completed and these models are utilized in an actual manufacturing site, prescription data 26, raw material data 27, MFR, and process data 28 are acquired in manufacturing facilities.

[0051] 2, this embodiment employs a two-stage prediction process in which output from a prediction model 211 is input to a correction model 212. In machine learning, a large number of variables are required to generate a computation model that performs quality prediction as in this embodiment, but by employing such a two-stage prediction process, fewer variables are required for each model, contributing to a reduction in the amount of required data.

[0052] The structure of the training dataset is not particularly limited. For example, the structure of the training dataset may be a structure in which all variables are directly input as a single variable. This structure will be specifically described. Assume that the recipe data 26 includes the blending ratios of three polymers, for example, the blending ratio of polymer A, the blending ratio of polymer B, and the blending ratio of polymer C. Assume also that the raw material data 27 includes molecular weight data of polymer A, the thermal stability of polymer A, the melt flow rate of polymer A, molecular weight data of polymer B, the thermal stability of polymer B, the melt flow rate of polymer B, molecular weight data of polymer C, the thermal stability of polymer C, and the melt flow rate of polymer C. Assume also that the process data 28 includes the resin temperature, resin pressure, the feed rate to the extruder, and the screw rotation speed of the extruder. In this case, the structure of the training dataset may be a structure in which these 16 variables are directly input as a single variable.

[0053] In addition, the structure of the training dataset may be a structure in which arithmetic operations and calculations such as exponentiation are performed on multiple variables of different categories, such as prescription data 26 and raw material data 27, or multiple variables of the same category, and the solutions are input as representative variables.

[0054] (Method of generating prediction model 211 and correction model 212) An example of a method for generating the prediction model 211 and the correction model 212 executed by the information processing device 1 will be described with reference to Fig. 3. Fig. 3 is a flowchart showing an example of a method for generating the prediction model 211 and the correction model 212.

[0055] In step S101, the training data acquisition unit 101 acquires training data 111 by referring to the storage unit 11. The training data 111 includes a dataset for generating a prediction model 211 and a dataset for generating a correction model 212. The dataset for generating the prediction model 211 is a dataset generated by associating recipe data 26 and raw material data 27 acquired in research equipment with corrective data. The recipe data 26 includes blending ratio conditions for multiple polymers, and the raw material data 27 includes physical property conditions for each polymer. The dataset for generating the correction model 212 is a dataset generated by associating MFR potential and process data 28 acquired in manufacturing equipment with corrective data. The process data 28 includes processing conditions for producing pellets. The MFR potential may be acquired using the same recipe in research equipment, or the predicted MFR potential output from the prediction model 211 may be used.

[0056] The process proceeds to step S102, where the learning unit 102 generates a prediction model 211 and a correction model 212 through machine learning using the training data 111 acquired in the process of S101. The prediction model 211 and the correction model 212 generated by the learning unit 102 are transmitted to the information processing device 2 and stored in the storage unit 21 of the information processing device 2. This completes the process of FIG. 3.

[0057] (Quality prediction method and quality correction method) Next, an example of a pellet quality prediction method and a pellet quality correction method executed by the information processing device 2 will be described with reference to Fig. 4. Fig. 4 is a flowchart showing an example of a pellet quality prediction method and a pellet quality correction method.

[0058] In step S201, the condition acquisition unit 201 acquires prediction data to be input to the prediction model 211. The prediction data includes at least the blending ratio conditions of multiple polymers and the physical property conditions for each polymer. The condition acquisition unit 201 acquires the prediction data, for example, when the operator is about to start producing pellets or while the operator is currently producing pellets. The condition acquisition unit 201 also acquires processing conditions.

[0059] The process proceeds to step S202, where the quality prediction unit 202 inputs the prediction data acquired in the process of step S201 into the prediction model 211 to predict the MFR potential of the pellet.

[0060] The process proceeds to step S203, where the quality correction unit 203 corrects the MFR potential of the pellet by inputting the MFR potential predicted in the process of step S202 and the processing conditions acquired in the process of step S201 into the correction model 212. This correction gives the MFR of the pellet.

[0061] The process proceeds to step S204, where the output control unit 205 outputs the MFR of the pellet obtained in the process of step S203 to the output unit 24. This completes the process of FIG.

[0062] Next, the MFR potential predicted by the prediction model 211 will be described with reference to Fig. 5. Fig. 5 is a graph showing the relationship between the measured value and the predicted value of the MFR potential. The X axis of Fig. 5 shows the measured value of the MFR potential, and the Y axis shows the predicted value of the MFR potential.

[0063] As shown in Fig. 5, the measured values of the MFR potential and the predicted values of the MFR potential fit roughly on a straight line, which indicates that the prediction model 211 makes predictions with high accuracy.

[0064] Next, referring to FIG. 6, the relationship between the resin temperature at the tip of the extruder and the ratio of MFR to MFR potential will be described. The X-axis in FIG. 6 represents the resin temperature at the tip of the extruder, and the Y-axis represents the ratio of MFR to MFR potential for each pellet production lot. As described above, the polymer in this embodiment is produced by a microbial fermentation process, and therefore the physical properties may vary for each culture lot. In addition, the pellets actually produced are significantly affected by processing conditions, and the MFR may not match the MFR potential. For example, pellets with a high MFR may be produced even with a recipe with a low MFR potential. Therefore, simply predicting the MFR potential is insufficient; the MFR potential must be corrected. Therefore, in this embodiment, the quality prediction unit 202 predicts the MFR potential using a prediction model 211, and the quality correction unit 203 corrects the MFR potential using a correction model 212.

[0065] An example of a correction calculation using the correction model 212 is to predict the MFR by correcting the predicted value of the MFR potential shown in FIG. 5 based on the relational expression of three variables, resin temperature, MFR potential, and MFR, shown in FIG. 6. The calculation results are shown in FIG. 7. FIG. 7 is a graph showing the relationship between the actual measured value and the predicted value of the MFR for each culture lot. The X-axis of FIG. 7 represents the actual measured value of the MFR, and the Y-axis represents the predicted value of the MFR predicted by the correction model 212. As shown in FIG. 7, the actual measured value of the MFR and the predicted value of the MFR fit roughly on a straight line. This indicates that the prediction by the correction model 212 is performed with high accuracy.

[0066] (Action and effect) As described above, according to the first embodiment, the following advantageous effects can be obtained.

[0067] In this embodiment, the blending ratio conditions of multiple polymers and the physical property conditions for each polymer are input into a prediction model 211 to predict the MFR potential of the pellets. In addition, in this embodiment, the MFR potential predicted by the prediction model 211 and the processing conditions for producing the pellets are input into a correction model 212 to correct the MFR potential of the pellets and predict the MFR of the pellets. With this configuration, by using a learning model in which input data is previously associated with correct data, it becomes possible to accurately predict whether the quality of the pellets deviates from the specifications simply by inputting the blending ratio conditions, physical property conditions, and processing conditions.

[0068] Furthermore, in this embodiment, a two-stage prediction process is adopted in which output from the prediction model 211 is input to the correction model 212. In machine learning, a large number of variables are required to generate a computational model that performs quality prediction, as in this embodiment. However, by adopting such a two-stage prediction process, fewer variables are required for each model, contributing to a reduction in the amount of required data. Furthermore, generating models with fewer variables also contributes to improved flexibility and adaptability in model management.

[0069] In addition, in this embodiment, the prescription data 26, raw material data 27, and MFR potential used to generate the prediction model 211, and the process data 28 and MFR used to generate the correction model 212 are acquired by different equipment. In this way, by using multiple pieces of equipment, it is possible to acquire the data necessary for model generation without affecting the pellet production plan.

[0070] Furthermore, the operator can also perform a simulation of pellet production by inputting various conditions and having the quality prediction unit 202 and the quality correction unit 203 make predictions. In this way, the information processing device 2 can also be used as an auxiliary tool for determining the conditions for pellet production.

[0071] [Embodiment 2] Next, a second embodiment of the present invention will be described. For ease of explanation, the same reference numerals will be used to designate components having the same functions as those described in the first embodiment, and the description thereof will not be repeated.

[0072] 8 is a schematic configuration diagram showing an example of an information processing device 1 and an information processing device 2 according to embodiment 2. The information processing device 1 has the same configuration as in embodiment 1, and therefore a description thereof will be omitted. Embodiment 2 differs from embodiment 1 in that the control unit 20 of the information processing device 2 further includes an optimization calculation unit 204.

[0073] An example of the operation of the information processing device 2 according to the second embodiment will be described with reference to Fig. 9. However, since the processes of steps S301 to S303 and 305 in Fig. 9 are similar to the processes of steps S201 to S204 in Fig. 4, the description will be omitted where appropriate. Here, the optimization calculation executed by the optimization calculation unit 204 in step S304 will be mainly described.

[0074] In step S304, the optimization calculation unit 204 performs optimization calculations of the blending ratio conditions and processing conditions based on the physical properties of each culture lot so that the MFR predicted by the correction model 212 falls within the target range. A well-known optimization method may be used for the optimization calculations. Examples of well-known optimization methods include grid search, random search, genetic algorithms, and Bayesian optimization. An example of the "target range" here is the range of specifications required for each grade of pellets, which is set in advance.

[0075] Here, an optimization method using grid search will be described as an example. The optimization calculation unit 204 comprehensively calculates candidate blending ratios from the range of possible blending ratios for each polymer, and selects a blending ratio that ensures that the MFR predicted by the correction model 212 falls within the target range.

[0076] A specific example of grid search will be explained. Here, it is assumed that there are four types of polymers to be blended (polymer A, polymer B, polymer C, and polymer D). Furthermore, it is assumed that the variables in the grid search are the blending ratio of polymer A (%), blending ratio of polymer B (%), blending ratio of polymer C (%), blending ratio of polymer D (%), resin temperature at the end of the extruder (°C), and resin pressure at the end of the extruder (MPa).

[0077] Regarding the controllability of these variables, the blend ratio of polymer A, the blend ratio of polymer C, the blend ratio of polymer D, and the resin temperature are variable, while the blend ratio of polymer B and the resin pressure are fixed. As the reference values for the variables that are variable parameters, the blend ratio of polymer A is assumed to be 30, the blend ratio of polymer C is assumed to be 15, the blend ratio of polymer D is assumed to be 15, and the resin temperature is assumed to be 185. The limit range for the variables that are variable parameters is assumed to be ±10 for the blend ratio of polymer A, the blend ratio of polymer C, the blend ratio of polymer D, and the resin temperature.

[0078] Assume that the following three conditions are set as constraints on variable manipulation: (1) Condition 1: Polymer A blend ratio + Polymer C blend ratio = X (2) Condition 2: X≦60 (3) Condition 3: Polymer D blend ratio = 60-X

[0079] For example, when optimizing three parameters (the blending ratio of polymer A, the blending ratio of polymer B, and the resin temperature) under the conditions set as described above, the optimization calculation unit 204 selects a grid point in a three-dimensional space in which these three parameters form orthogonal coordinate axes, at which the MFR predicted by the correction model 212 approaches a target value. This allows the blending ratio and resin temperature to be optimized according to the physical property conditions of each polymer. The target value here refers to any value within a target range. While the example described here illustrates the optimization calculation unit 204 optimizing the blending ratio and the resin temperature, this is not limiting, and the optimization calculation unit 204 may optimize only the blending ratio. Furthermore, the optimization calculation unit 204 may also optimize one or any combination of processing conditions from among the blending ratio, resin temperature, resin pressure, extruder feed rate, and extruder screw rotation speed.

[0080] The process proceeds to step S305, where the output control unit 205 outputs the blending ratio conditions and processing conditions optimized in the process of step S304 to the output unit 24. This allows the operator to grasp and set the optimal blending ratio conditions and processing conditions each time the culture lot of the polymer to be blended is changed. This allows the quality of the pellets to be reproduced, and the desired quality to be obtained, even if the culture lot of the polymer to be blended is changed.

[0081] As described above, the production process of the polymer used to manufacture pellets relies on biochemical processes, and therefore, even for the same type of polymer, differences in physical properties may occur between culture lots. For this reason, in conventional manufacturing sites, production was stopped each time the culture lot was changed, and trial and error adjustments were made while changing the blending ratio conditions and processing conditions, until the quality met the standard. This conventional method did not allow for continuous production, resulting in significant losses of time, raw materials, and personnel. In this regard, according to the present embodiment, even when the culture lot was changed, optimal blending ratio conditions and processing conditions were obtained through optimization calculations, making it possible to efficiently produce high-quality pellets without having to stop production and make adjustments.

[0082] Note that the processing flow in the flowcharts shown in Figures 3, 4, and 9 is an example, and steps may be deleted, new steps may be added, or the processing order may be changed within the scope of the main idea.

[0083] [Modification] The execution entity of each process described in each of the above-mentioned embodiments is arbitrary and is not limited to the above-mentioned examples. In other words, functions similar to those of information processing devices 1 and 2 can be realized by a plurality of information processing devices that can communicate with each other. For example, each process shown in FIGS. 1 and 8 may be shared and executed by a plurality of information processing devices. Furthermore, an information processing device that combines the functions of information processing device 1 and information processing device 2 is also included in the scope of the present invention.

[0084] The manufacturing method of the molded article described in each embodiment can also be expressed as follows: That is, the manufacturing method of the molded article containing the polyhydroxyalkanoate resin may include a step of blending a plurality of raw materials under blending ratio conditions optimized by the optimization calculation executed by the information processing device 2, and a step of processing the molded article under processing conditions optimized by the optimization calculation executed by the information processing device 2.

[0085] [Software implementation example] The functions of the information processing devices 1 to 2 (hereinafter simply referred to as "devices") can be realized by a program that causes a computer to function as the device, and a program that causes a computer to function as each control block of the device (particularly each part included in the control unit 10 and control unit 20).

[0086] In this case, the device includes a computer having at least one control device (e.g., a processor) and at least one storage device (e.g., a memory) as hardware for executing the program. The computer executes the program to realize each function described in each embodiment.

[0087] The program may be non-transitory and may be recorded on one or more computer-readable recording media. The recording media may or may not be included in the device. In the latter case, the program may be supplied to the device via any wired or wireless transmission medium.

[0088] In addition, some or all of the functions of each control block can be realized by a logic circuit. For example, an integrated circuit in which a logic circuit that functions as each control block is formed is also included in the scope of the present invention. In addition, the functions of each control block can also be realized by, for example, a quantum computer.

[0089] Furthermore, each process described in each embodiment may be executed by AI (Artificial Intelligence). In this case, the AI may run on the above-mentioned device or on another device (for example, an edge computer or a cloud server).

[0090] The present invention is not limited to the above-described embodiments, and various modifications are possible within the scope of the claims. Embodiments obtained by appropriately combining the technical means disclosed in different embodiments are also included in the technical scope of the present invention. [Explanation of symbols]

[0091] 2. Information processing equipment 201 Condition Acquisition Unit 202 Quality Prediction Department 203 Quality Correction Department 204 Optimization Calculation Unit 211 Predictive Model 212 Correction Model

Claims

1. a condition acquisition unit that acquires manufacturing conditions for a molded product that is manufactured by blending at least two or more raw materials extracted from microorganisms, the manufacturing conditions including blending ratio conditions that indicate the ratio when blending the multiple raw materials and physical property conditions that indicate the physical properties of the multiple raw materials; a quality prediction unit that predicts the quality of the molded product using a prediction model that uses the manufacturing conditions acquired by the condition acquisition unit as explanatory variables and the quality of the molded product as a target variable; a quality correction unit that corrects the quality of the molded product predicted by the quality prediction unit using a correction model that uses quality conditions indicating the quality of the molded product output from the prediction model and processing conditions for manufacturing the molded product as explanatory variables and the quality of the molded product as a target variable. Information processing device.

2. further comprising an optimization calculation unit that optimizes the blending ratio conditions and the processing conditions so that the quality corrected by the quality correction unit falls within a preset target range; The information processing device according to claim 1 .

3. the raw material is a polymer, The physical property conditions include at least one of molecular weight data of the polymer, thermal stability of the polymer, thermal decomposition temperature of the polymer, melt flow rate of the polymer, and content of repeating units of a predetermined functional group contained in the polymer.

3. The information processing device according to claim 1 or 2.

4. The processing conditions include at least one of a resin temperature, a resin pressure, a feeding rate into an extruder, and a screw rotation speed of the extruder.

3. The information processing device according to claim 1 or 2.

5. A prediction method executed by an information processing device, A condition acquisition step of acquiring manufacturing conditions for a molded product manufactured by blending at least two or more raw materials extracted from microorganisms, the manufacturing conditions including blending ratio conditions indicating the ratio when blending the multiple raw materials and physical property conditions indicating the physical properties of the multiple raw materials; a quality prediction step of predicting the quality of the molded product using a prediction model in which the manufacturing conditions acquired in the condition acquisition step are used as explanatory variables and the quality of the molded product is used as a target variable; a quality correction step of correcting the quality of the molded product predicted in the quality prediction step by using a correction model in which quality conditions indicating the quality of the molded product output from the prediction model and processing conditions for manufacturing the molded product are used as explanatory variables and the quality of the molded product is used as a target variable. Forecasting methods.

6. 2. A prediction program for causing a computer to function as the information processing device according to claim 1, the prediction program causing a computer to function as the condition acquisition unit, the quality prediction unit, and the quality correction unit.

7. A method for producing a molded article containing a polyhydroxyalkanoate resin, comprising: blending the plurality of ingredients under blending ratio conditions optimized by the optimization calculation executed by the information processing device according to claim 2; and processing the molded product under processing conditions optimized by the optimization calculation executed by the information processing device. Manufacturing method of molded products.

Images