System, information processing method, machine learning device, information processing device, program, and learned model

Abstract

To provide a system, information processing method, a machine learning device, an information processing device, a program, and a learned model for outputting a blending amount or optical characteristics of a reproduced cyclic olefin resin in a short time with high precision.SOLUTION: A system for processing information related to a resin composition including a reproduced cyclic olefin resin, comprises: a storage part which stores relation information obtained by associating blending information indicating a blending amount of the reproduced cyclic olefin resin in the resin composition with optical characteristic information indicating optical characteristics of the reproduced cyclic olefin resin; and an arithmetic part which outputs information associated by at least one piece of information and the relation information on the basis of the relation information, to an input of the at least one piece of information out of the blending information and the optical characteristic information.SELECTED DRAWING: Figure 1

Description

[Technical field]

[0001] The present invention relates to a system, an information processing method, a machine learning device, an information processing device, a program, and a trained model. [Background technology]

[0002] Conventionally, resin molding is performed using recycled cyclic olefin resin obtained by recycling cyclic olefin resin. However, since the recycled cyclic olefin resin obtained by recycling may contain a wide variety of mixtures, the physical properties (optical properties, etc.) of the resin molded product may be deteriorated when the recycled cyclic olefin resin is used. Therefore, it has been proposed to estimate the physical properties of the resin molded product by simulation when resin molding is performed using the recycled cyclic olefin resin (Patent Document 1). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] JP 2020-87446 A Summary of the Invention [Problem to be solved by the invention]

[0004] However, in the technology described in Patent Document 1, the simulation analysis process takes a long time, and the low accuracy of the simulation analysis may result in a deterioration of the physical properties of the resin molded product. In particular, the optical properties of the resin molded product are greatly affected by the blending amount of the recycled cyclic olefin resin.

[0005] The present invention has been made in consideration of these circumstances, and one of its objectives is to provide a system, information processing method, machine learning device, information processing device, program, and trained model for outputting the compounding amount or optical properties of recycled cyclic olefin resin with high accuracy and in a short period of time. [Means for solving the problem]

[0006] [1] A system for processing information relating to a resin composition containing recycled cyclic olefin resin, comprising: a memory unit that stores relationship information correlating formulation information indicating the amount of recycled cyclic olefin resin in the resin composition with optical property information indicating the optical properties of the recycled cyclic olefin resin; and a calculation unit that, in response to input of at least one piece of information among the formulation information and the optical property information, outputs information correlated by the relationship information with the at least one piece of information.

[0007] [2] The system described in [1], wherein the calculation unit outputs the optical property information based on the relationship information in response to the input of the blending information.

[0008] [3] The system described in [2] further comprises a learning unit that performs machine learning using the blending information as an input variable and the optical property information as an output variable to generate a trained model as the relationship information, and the calculation unit outputs the optical property information based on the trained model in response to the input of the blending information.

[0009] [4] A system as described in [3], comprising a candidate generation unit that generates multiple candidates for the blending amount, and an extraction unit that extracts at least one candidate from the multiple candidates, wherein the calculation unit outputs the optical property information for each of the multiple candidates based on the trained model in response to input of candidate blending information indicating the candidate, and the extraction unit extracts at least one candidate from the multiple candidates based on the optical property information for each of the multiple candidates.

[0010] [5] The system described in [1], wherein the calculation unit outputs the blending information based on the relationship information in response to input of the optical property information.

[0011] [6] The system according to any one of [1] to [5], wherein the optical characteristic information includes at least one of haze, light transmittance, refractive index, birefringence, and yellow index.

[0012] [7] An information processing method in which a system for processing information relating to a resin composition containing recycled cyclic olefin resin reads from a memory unit relationship information correlating formulation information indicating the amount of recycled cyclic olefin resin in the resin composition with optical property information indicating the optical properties of the recycled cyclic olefin resin, and in response to input of at least one piece of information among the formulation information and the optical property information, outputs information correlated by the relationship information with the at least one piece of information.

[0013] [8] A machine learning device that processes information about a resin composition containing a recycled cyclic olefin resin, the machine learning device comprising: a learning unit that performs machine learning using formulation information indicating the amount of the recycled cyclic olefin resin in the resin composition as an input variable and optical property information indicating the optical properties of the recycled cyclic olefin resin as an output variable, to generate a trained model.

[0014] [9] An information processing device that processes information related to a resin composition containing recycled cyclic olefin resin, comprising: a calculation unit that reads out a trained model generated by machine learning using formulation information indicating the amount of the recycled cyclic olefin resin in the resin composition as an input variable and optical property information indicating the optical properties of the recycled cyclic olefin resin as an output variable; and outputs the optical property information based on the trained model in response to input of the formulation information.

[0015]

[10] A program for causing a computer that processes information relating to a resin composition containing a recycled cyclic olefin resin to perform machine learning using formulation information indicating the amount of the recycled cyclic olefin resin in the resin composition as an input variable and optical property information indicating the optical properties of the recycled cyclic olefin resin as an output variable, thereby generating a trained model.

[0016]

[11] A program that causes a computer to process information related to a resin composition containing a recycled cyclic olefin resin to read out a trained model generated by machine learning using formulation information indicating the amount of the recycled cyclic olefin resin in the resin composition as an input variable and optical property information indicating the optical properties of the recycled cyclic olefin resin as an output variable, and causes the computer to output the optical property information based on the trained model in response to the input of the formulation information.

[0017]

[12] A trained model for causing a computer to function so as to output information relating to a resin composition containing a recycled cyclic olefin resin, the trained model comprising: machine learning performed using formulation information indicating the amount of the recycled cyclic olefin resin in the resin composition as input variables; and optical property information indicating the optical properties of the recycled cyclic olefin resin as output variables, to update the learning variables; and performing calculations based on the updated learning variables for the input formulation information, thereby outputting the optical property information. Effect of the Invention

[0018] According to one aspect of the present invention, it is possible to provide a system, an information processing method, a machine learning device, an information processing device, a program, and a trained model for outputting the blending amount or optical properties of recycled cyclic olefin resin with high accuracy and in a short period of time. [Brief description of the drawings]

[0019] [Figure 1] 1 is a diagram illustrating an example of a system configuration according to a first embodiment. [Diagram 2] A figure showing an example of the configuration of a trained model M. [Diagram 3] FIG. 2 is a diagram illustrating an example of processing performed by a calculation unit 140. [Figure 4] FIG. 1 is a schematic diagram showing a part of a neural network used in the trained model M. [Diagram 5]FIG. 13 is a diagram showing an example of teacher data D2. [Figure 6] FIG. 13 is a diagram showing an example of candidate data D1. [Figure 7] 13 is a diagram showing an example of candidate data D1 to which optical property information and a quality score are added. FIG. [Figure 8] 13 is a flowchart illustrating an example of a learning stage process. [Figure 9] 13 is a flowchart illustrating an example of a process in an operation stage. [Figure 10] FIG. 11 is a diagram showing an example of processing performed by a calculation unit 140 according to the second embodiment. [Figure 11] FIG. 11 is a diagram showing an example of processing performed by a calculation unit 140 according to the third embodiment. [Figure 12] FIG. 13 is a diagram showing an example of processing performed by a calculation unit 140 according to the fourth embodiment. [Figure 13] FIG. 13 is a diagram showing a configuration of an information processing device 100A according to a fifth embodiment. DETAILED DESCRIPTION OF THE PREFERRED EMBODIMENTS

[0020] Hereinafter, the system, information processing method, machine learning device, information processing device, program, and trained model of the present invention will be described with reference to the drawings.

[0021] First Embodiment [System Configuration] FIG. 1 is a diagram showing an example of a system configuration according to the first embodiment. The system 1 is a system for processing information related to a resin composition containing a recycled cyclic olefin resin. The recycled cyclic olefin resin is a resin obtained by recycling a cyclic olefin resin. Specific examples of cyclic olefin resins will be described later. The system 1 includes an information processing device 100 and a machine learning device 200. The information processing device 100 communicates with the machine learning device 200 via a network NW. The network NW includes, for example, the Internet, a LAN (Local Area Network), a wireless base station, a provider device, and the like.

[0022] The information processing device 100, details of which will be described later, is a computer that outputs information about recycled cyclic olefin resins using a trained model M. The information processing device 100 includes a communication unit 110, an input unit 120, a display unit 130, a calculation unit 140, a candidate generation unit 150, an extraction unit 160, and a storage unit 170.

[0023] The communication unit 110 is a communication interface for connecting to the network NW. The communication unit 110 is, for example, a network interface card.

[0024] The input unit 120 is an input device such as a keyboard, a pointing device, etc. The display unit 130 is a display device such as a liquid crystal display device.

[0025] The calculation unit 140, the candidate generation unit 150, and the extraction unit 160 are realized by, for example, a hardware processor such as a CPU executing a program (software). Some or all of these components may be realized by hardware (including circuitry) such as a large-scale integration (LSI), an application-specific integrated circuit (ASIC), a field-programmable gate array (FPGA), or a graphics processing unit (GPU), or may be realized by a combination of software and hardware. The program may be stored in advance in a storage device (a storage device having a non-transient storage medium) such as a hard disk drive (HDD) or a flash memory, or may be stored in a removable storage medium (a non-transient storage medium) such as a DVD or a CD-ROM, and installed in the storage device by mounting the storage medium in a drive device.

[0026] The storage unit 170 is a HDD, a flash memory, a RAM (Random Access Memory), etc. The storage unit 170 may be a NAS (Network Attached Storage) device that the information processing device 100 can access via the network NW. The storage unit 170 stores information such as the trained model M and the candidate data D1.

[0027] The machine learning device 200, which will be described in detail later, is a computer that generates a trained model M by learning information related to recycled cyclic olefin resins. The machine learning device 200 includes a communication unit 210, a learning unit 220, and a storage unit 270.

[0028] The communication unit 210 is a communication interface for connecting to the network NW. The communication unit 210 is, for example, a network interface card.

[0029] The learning unit 220 is realized, for example, by a hardware processor such as a CPU executing a program (software). The learning unit 220 may be realized by hardware (including circuitry) such as an LSI, ASIC, FPGA, or GPU, or may be realized by a combination of software and hardware. The program may be stored in advance in a storage device (a storage device having a non-transient storage medium) such as an HDD or flash memory, or may be stored in a removable storage medium (non-transient storage medium) such as a DVD or CD-ROM, and installed in the storage device by mounting the storage medium in a drive device.

[0030] The storage unit 270 is a HDD, a flash memory, a RAM, etc. The storage unit 270 may be a NAS device that can be accessed by the machine learning device 200 via the network NW. The storage unit 270 stores information such as the teacher data D2.

[0031] Although details will be described later, the learning unit 220 generates a trained model M by performing machine learning using the teacher data D2, and controls the communication unit 210 to transmit the trained model M to the information processing device 100. The information processing device 100 stores the trained model M received from the machine learning device 200 in the storage unit 170.

[0032] The trained model M is an example of relational information in which blending information indicating the blending amount of recycled cyclic olefin resin in the resin composition is related to optical property information indicating the optical properties of the recycled cyclic olefin resin. In response to an input of at least one piece of information out of blending information and optical property information, the calculation unit 140 of the information processing device 100 outputs, based on the trained model M, at least one piece of information and information related by the trained model M. The process performed by the calculation unit 140 will be described below.

[0033] [Trained model] Fig. 2 is a diagram showing an example of the configuration of a trained model M. The trained model M includes an input layer L1, an intermediate layer L2, and an output layer L3. In Fig. 2, the trained model M includes one intermediate layer L2, but may include multiple intermediate layers L2.

[0034] The calculation unit 140 reads out the trained model M from the storage unit 170. When the calculation unit 140 inputs an input variable to the input layer L1 of the trained model M, an output variable is output from the output layer L3 of the trained model M.

[0035] FIG. 3 is a diagram showing an example of processing performed by the calculation unit 140. As shown in FIG. 3, in this embodiment, an example will be described in which the input variables input to the trained model M are blending information, and the output variables output from the trained model M are optical property information. The blending information is information indicating the blending amount of the recycled cyclic olefin resin in the resin composition. For example, the blending amount may be the ratio of the weight of the recycled cyclic olefin resin to the weight of the entire resin composition. In addition, the optical property information may include at least one of the haze, light transmittance, refractive index, birefringence, and yellow index of the resin composition.

[0036] The calculation unit 140 reads out the trained model M from the storage unit 170, and acquires the combination information (an example of an input variable) input by the user using the input unit 120. Then, the calculation unit 140 outputs optical property information (an example of an output variable) based on the trained model M (an example of relationship information) in response to the input of the combination information (an example of an input variable). The calculation unit 140 may display the output optical property information on the display unit 130.

[0037] Figure 4 is a schematic diagram of a part of the neural network used in trained model M. x indicates the value of the unit (node) in each layer. The value of each unit is calculated using a specified function by multiplying the output of the unit in the previous layer by a weight w and adding them together. These can be expressed as the following equation.

[0038]

number

[0039] Here, b is a bias term. The function f may be a ReLU or sigmoid function. The calculation of the above formula (1) is repeated from the input layer L1 to the output layer L3, and the output variables are output from the trained model M.

[0040] The learning unit 220 generates a trained model M by performing machine learning using the teacher data D2. "Learning" means adjusting the weights w and bias terms b to appropriate values. Learning is performed by a stochastic gradient descent method or the like. That is, random values ​​are given to the weights w and bias terms b of each layer, and the teacher data D2 is given to the input layer L1. In the early stages of learning, the output value output from the output layer L3 is an incorrect value, but in the case of the teacher data D2, the value (target value) that is originally desired to be output is known, so the weights and the like are updated in reverse order from the output layer L3 to the input layer L1 so that the deviation (error) between the output value and the target value becomes small. This is called error backpropagation, and the error between the output value and the target value is expressed by squared error, cross entropy, or the like. At this time, if the error is expressed by a differentiable function, the amount of adjustment (gradient) to reduce the error can be calculated.

[0041] FIG. 5 is a diagram showing an example of teacher data D2. In teacher data D2, for example, blending information and optical property information are associated with a teacher data ID. The teacher data ID is identification information for identifying teacher data. As described above, the blending information is information indicating the blending amount of recycled cyclic olefin resin in the resin composition. For example, the blending amount may be the ratio of the weight of recycled cyclic olefin resin to the weight of the entire resin composition. In addition, the optical property information may include at least one of the haze, light transmittance, refractive index, birefringence, and yellow index of the resin composition.

[0042] The processing of the system 1 of this embodiment includes processing in a learning stage and processing in an operation stage. In the learning stage, the learning unit 220 of the machine learning device 200 generates a trained model M by performing machine learning using the teacher data D2. That is, the learning unit 220 generates the trained model M by performing machine learning using the blending information as an input variable and the optical property information as an output variable. The learning unit 220 controls the communication unit 210 to transmit the trained model M to the information processing device 100. The information processing device 100 stores the trained model M received from the machine learning device 200 in the storage unit 170.

[0043] Also, in the operation stage, the calculation unit 140 of the information processing device 100 outputs optical property information using the trained model M generated in the learning stage. Specifically, the calculation unit 140 acquires blending information input by a user using the input unit 120. Then, the calculation unit 140 outputs optical property information based on the trained model M in response to the input of the blending information.

[0044] [Candidate extraction process] FIG. 6 is a diagram showing an example of candidate data D1. The candidate data D1 is data showing a list of candidates for blending information. For example, the candidate data D1 is data in which blending information is associated with a candidate data ID. The candidate data ID is identification information for identifying the candidate data. As described above, the blending information is information showing the blending amount of recycled cyclic olefin resin in the resin composition.

[0045] The candidate generating unit 150 of the information processing device 100 generates the candidate data D1. For example, the candidate generating unit 150 may generate the candidate data D1 based on a plurality of blend amounts input by the user by operating the input unit 120.

[0046] The extraction unit 160 extracts at least one candidate from a plurality of combination information candidates included in the candidate data D1. For example, the extraction unit 160 may extract combination information having the highest quality score, which will be described later, from a plurality of combination information candidates included in the candidate data D1.

[0047] Specifically, the extraction unit 160 outputs the candidate data D1 shown in Fig. 6 to the calculation unit 140. For each of the multiple candidates included in the candidate data D1, the calculation unit 140 outputs optical property information based on the trained model M in response to the input of candidate blending information indicating the candidate. The extraction unit 160 calculates a quality score based on the optical property information for each of the multiple candidates output from the calculation unit 140. The quality score is a score that indicates a higher value as the quality of the resin composition based on the optical property information (haze, light transmittance, refractive index, birefringence, yellow index, etc.) is better.

[0048] Fig. 7 is a diagram showing an example of candidate data D1 to which optical property information and a quality score are added. As described above, the optical property information in Fig. 7 is information output from the calculation unit 140, and the quality score in Fig. 7 is a score calculated by the extraction unit 160.

[0049] For example, the extraction unit 160 may extract the combination information with the highest quality score from a plurality of combination information candidates included in the candidate data D1. The extraction unit 160 may also extract combination information with a quality score higher than a predetermined threshold from a plurality of combination information candidates included in the candidate data D1. The extraction unit 160 may also display the extracted combination information on the display unit 130. This allows the user to easily grasp the combination information of a high-quality resin composition.

[0050] [Flowchart of the learning stage process] 8 is a flowchart showing an example of processing in the learning stage. For example, the processing according to this flowchart is executed by the computer of the machine learning device 200 executing a program.

[0051] First, the learning unit 220 of the machine learning device 200 reads out the teacher data D2 from the storage unit 270 (S11). As shown in FIG. 5, the teacher data D2 includes a teacher data ID, combination information, and optical property information.

[0052] Next, the learning unit 220 performs machine learning on the teacher data D2 to generate a trained model M (S12). Specifically, the learning unit 220 performs machine learning on the training data D2 using blending information indicating the blending amount of the recycled cyclic olefin resin in the resin composition as an input variable and optical property information indicating the optical properties of the recycled cyclic olefin resin as an output variable to generate the trained model M.

[0053] After that, the learning unit 220 controls the communication unit 210 to transmit the trained model M to the information processing device 100 (S13). The information processing device 100 stores the trained model M received from the machine learning device 200 in the storage unit 170. This ends the processing according to this flowchart.

[0054] As described above, the learning unit 220 of the machine learning device 200 of this embodiment performs machine learning using blending information indicating the blending amount of recycled cyclic olefin resin in the resin composition as an input variable and optical property information indicating the optical properties of the recycled cyclic olefin resin as an output variable, thereby generating a trained model M. This allows the machine learning device 200 of this embodiment to generate a trained model M for outputting highly accurate optical property information.

[0055] [Operational Phase Processing Flowchart] 9 is a flowchart showing an example of a process in the operation stage. For example, the process according to this flowchart is executed by the computer of the information processing device 100 executing a program.

[0056] First, the calculation unit 140 of the information processing device 100 reads out the trained model M from the storage unit 170 (S21). The trained model M is a neural network model including an input layer L1, an intermediate layer L2, and an output layer L3, as shown in FIG.

[0057] Next, the calculation unit 140 acquires the combination information (S22). For example, the calculation unit 140 may acquire the combination information input by the user using the input unit 120.

[0058] Thereafter, the calculation unit 140 outputs optical property information based on the trained model M in response to the input of the blending information (S23). For example, when the calculation unit 140 inputs blending information (an example of an input variable) to the input layer L1 of the trained model M, optical property information (an example of an output variable) is output from the output layer L3 of the trained model M. The calculation unit 140 outputs the optical property information acquired using the trained model M. For example, the calculation unit 140 may display the optical property information on the display unit 130 by outputting the optical property information to the display unit 130. This ends the processing according to this flowchart.

[0059] As described above, the calculation unit 140 of the information processing device 100 of this embodiment reads out the trained model M generated by machine learning using the blending information indicating the blending amount of recycled cyclic olefin resin in the resin composition as an input variable and the optical property information indicating the optical properties of the recycled cyclic olefin resin as an output variable, and outputs optical property information in response to the input of the blending information based on the trained model M. This allows the information processing device 100 of this embodiment to output highly accurate optical property information based on the blending information.

[0060] The trained model M of this embodiment is a model for causing a computer to function to output information about a resin composition containing recycled cyclic olefin resin. The trained model M is a model for causing a computer to function to update the learning variables by performing machine learning using compounding information indicating the amount of recycled cyclic olefin resin in the resin composition as an input variable and optical property information indicating the optical properties of the recycled cyclic olefin resin as an output variable, and to perform calculations based on the updated learning variables for the input compounding information and output optical property information. As a result, the trained model M of the embodiment can output the compounding amount or optical properties of the recycled cyclic olefin resin with high accuracy and in a short time.

[0061] [Cyclic olefin resin] The cyclic olefin resin in this embodiment will be described in detail below. The cyclic olefin resin may be a cyclic olefin resin (A) shown below. For example, the cyclic olefin resin (A) may contain at least one selected from a copolymer of ethylene or α-olefin and a cyclic olefin, and a ring-opening polymer of a cyclic olefin. The cyclic olefin resin (A) in this embodiment may be used alone or in combination of two or more.

[0062] The cyclic olefin resin (A) is available as a commercially available product, and examples thereof include products manufactured by Zeon Corporation under the trade names "Zeonex (registered trademark)" and "Zeonor (registered trademark)", products manufactured by JSR Corporation under the trade name "ARTON (registered trademark)", products manufactured by Mitsui Chemicals, Inc. under the trade name "APPEL (registered trademark)", products manufactured by TOPAS Advanced Polymers, and the like.

[0063] The resin composition containing the recycled cyclic olefin resin according to the present embodiment (hereinafter, the recycled cyclic olefin resin composition) contains the recycled cyclic olefin resin (A1) and may contain components other than the recycled cyclic olefin resin (A1) as necessary. The recycled cyclic olefin resin (A1) is a cyclic olefin resin that is recovered and reused by recycling.

[0064] Examples of the recovered cyclic olefin resin (A1) that can be used include injection-molded products such as optical components described below, sprues and runners generated during molding of injection-molded products, waste materials generated in molded products after extrusion molding, and waste materials generated after cutting extrusion-molded products into desired shapes.

[0065] These molded bodies, sprues, runners, waste materials, etc. may be used as they are, or may be crushed to an appropriate size before use. When crushed, the diameter after crushing is preferably 20 mm or less, more preferably 15 mm or less, and even more preferably 10 mm or less. If the diameter after crushing is within this range, melt kneading is easy. A crusher or the like can be used for crushing.

[0066] In this specification, the recycled cyclic olefin resin composition according to the present embodiment is also called a recycled cyclic olefin resin composition when it contains only the recycled cyclic olefin resin (A1). The recycled cyclic olefin resin composition according to the present embodiment may also contain unused cyclic olefin resin (A).

[0067] In this specification, the recovered cyclic olefin resin (A1) and the unused cyclic olefin resin (A) are collectively referred to as "cyclic olefin resin (A)". From the viewpoint of further improving heat resistance while maintaining a good balance of the performance of transparency and refractive index of the obtained molded body and optical parts, and from the viewpoint of further improving moldability, the cyclic olefin resin (A) may contain a cyclic olefin copolymer (P1) having at least one olefin-derived repeating unit (a) represented by the following formula (I) and at least one cyclic olefin monomer-derived repeating unit (b) selected from the group consisting of the repeating unit (AA) represented by the following formula (II), the repeating unit (AB) represented by the following formula (III), and the repeating unit (AC) represented by the following formula (IV).

[0068] [ka]

[0069] In the formula (I), R 300 represents a hydrogen atom or a linear or branched hydrocarbon group having 1 to 29 carbon atoms.

[0070] [ka]

[0071] In the formula (II), u is 0 or 1, v is 0 or a positive integer, preferably an integer of 0 to 2, more preferably 0 or 1, w is 0 or 1, R 61 ~R 78 And R a1 and R b1may be the same or different, and each is a hydrogen atom, a halogen atom, an alkyl group having 1 to 20 carbon atoms, a halogenated alkyl group having 1 to 20 carbon atoms, a cycloalkyl group having 3 to 15 carbon atoms, or an aromatic hydrocarbon group having 6 to 20 carbon atoms; R 75 ~R 78 may be bonded to each other to form a monocyclic or polycyclic ring.

[0072] [ka]

[0073] In the formula (III), x and d are 0 or an integer of 1 or more, preferably an integer of 0 or more and 2 or less, more preferably 0 or 1, y and z are 0, 1, or 2, and R 81 ~R 99 may be the same or different, and each is a hydrogen atom, a halogen atom, an aliphatic hydrocarbon group which is an alkyl group having 1 to 20 carbon atoms or a cycloalkyl group having 3 to 15 carbon atoms, an aromatic hydrocarbon group having 6 to 20 carbon atoms, or an alkoxy group; R 89 and R 90 and the carbon atom to which R is bonded. 93 or the carbon atom to which R 91 may be bonded directly or via an alkylene group having 1 to 3 carbon atoms, and when y=z=0, R 95 and R 92 or R 95 and R 99 may be bonded to each other to form a monocyclic or polycyclic aromatic ring.

[0074] [ka]

[0075] In the formula (IV), R 100 and R 101 may be the same or different and represent a hydrogen atom or a hydrocarbon group having 1 to 5 carbon atoms, and f is an integer satisfying 1≦f≦18.

[0076] The olefin monomer, which is one of the copolymerization raw materials for the cyclic olefin copolymer (P1), undergoes addition copolymerization to form the repeating unit (a) represented by the above formula (I). Specifically, the olefin monomer represented by the following formula (Ia) corresponding to the above formula (I) is used.

[0077] [ka]

[0078] In the formula (Ia), R 300 represents a hydrogen atom or a linear or branched hydrocarbon group having 1 to 29 carbon atoms. Examples of the olefin monomer represented by the formula (Ia) include ethylene, propylene, 1-butene, 1-pentene, 1-hexene, 3-methyl-1-butene, 3-methyl-1-pentene, 3-ethyl-1-pentene, 4-methyl-1-pentene, 4-methyl-1-hexene, 4,4-dimethyl-1-hexene, 4,4-dimethyl-1-pentene, 4-ethyl-1-hexene, 3-ethyl-1-hexene, 1-octene, 1-decene, 1-dodecene, 1-tetradecene, 1-hexadecene, 1-octadecene, and 1-eicosene. Two or more types of olefin monomers represented by the formula (Ia) may be used.

[0079] When the total repeating units constituting the cyclic olefin copolymer (P1) is taken as 100 mol %, the proportion of the olefin-derived repeating units (a) is preferably 5 mol % or more and 95 mol % or less, more preferably 20 mol % or more and 90 mol % or less, even more preferably 40 mol % or more and 85 mol % or less, and even more preferably 50 mol % or more and 80 mol % or less.

[0080] The proportion of repeating units (a) derived from olefins is as follows: 13 It can be measured by C-NMR.

[0081] The cyclic olefin monomer, which is one of the copolymerization raw materials of the cyclic olefin copolymer (P1), undergoes addition copolymerization to form the repeating unit (b) derived from the cyclic olefin monomer represented by the formula (II), the formula (III) or the formula (IV). Specifically, the cyclic olefin monomers represented by the formulas (IIa), (IIIa) and (IVa) corresponding to the formulas (II), (III) and (IV), respectively, are used.

[0082] [ka]

[0083] In the formula (IIa), u is 0 or 1, v is 0 or a positive integer, preferably an integer of 0 to 2, more preferably 0 or 1, w is 0 or 1, R 61 ~R 78 And R a1 and R b1 may be the same or different, and each is a hydrogen atom, a halogen atom, an alkyl group having 1 to 20 carbon atoms, a halogenated alkyl group having 1 to 20 carbon atoms, a cycloalkyl group having 3 to 15 carbon atoms, or an aromatic hydrocarbon group having 6 to 20 carbon atoms; R 75 ~R 78 may be bonded to each other to form a monocyclic or polycyclic ring.

[0084] [ka]

[0085] In the formula (IIIa), x and d are 0 or an integer of 1 or more, preferably an integer of 0 or more and 2 or less, more preferably 0 or 1, y and z are 0, 1, or 2, and R 81 ~R 99 may be the same or different, and each is a hydrogen atom, a halogen atom, an aliphatic hydrocarbon group which is an alkyl group having 1 to 20 carbon atoms or a cycloalkyl group having 3 to 15 carbon atoms, an aromatic hydrocarbon group having 6 to 20 carbon atoms, or an alkoxy group; R 89 and R90 and the carbon atom to which R is bonded. 93 or the carbon atom to which R 91 may be bonded directly or via an alkylene group having 1 to 3 carbon atoms, and when y=z=0, R 95 and R 92 or R 95 and R 99 may be bonded to each other to form a monocyclic or polycyclic aromatic ring.

[0086] [ka]

[0087] In the formula (IVa), R 100 and R 101 may be the same or different and represent a hydrogen atom or a hydrocarbon group having 1 to 5 carbon atoms, and f is an integer satisfying 1≦f≦18.

[0088] By using the olefin monomer represented by the above-mentioned formula (Ia) or the cyclic olefin monomer represented by the formula (IIa), (IIIa) or (IVa) as the copolymerization component, the solubility of the cyclic olefin resin (A) in a solvent is further improved, resulting in good moldability and improved product yield.

[0089] Specific examples of the cyclic olefin monomer represented by formula (IIa), (IIIa) or (IVa) that can be used include the compounds described in paragraphs 0037 to 0063 of WO 2006 / 118261.

[0090] Specifically, bicyclo-2-heptene derivatives (bicyclohept-2-ene derivatives), tricyclo-3-decene derivatives, tricyclo-3-undecene derivatives, tetracyclo-3-dodecene derivatives, pentacyclo-4-pentadecene derivatives, pentacyclopentadecadiene derivatives, pentacyclo-3-pentadecene derivatives, pentacyclo-4-hexadecene derivatives, pentacyclo-3-hexadecene derivatives, hexacyclo-4-heptadecene derivatives, and heptacyclo-5-eicosene derivatives. conductors, heptacyclo-4-eicosene derivatives, heptacyclo-5-heneicosene derivatives, octacyclo-5-docosene derivatives, nonacyclo-5-pentacosene derivatives, nonacyclo-6-hexacosene derivatives, cyclopentadiene-acenaphthylene adducts, 1,4-methano-1,4,4a,9a-tetrahydrofluorene derivatives, 1,4-methano-1,4,4a,5,10,10a-hexahydroanthracene derivatives, and cycloalkylene derivatives having 3 to 20 carbon atoms.

[0091] As the cyclic olefin monomer represented by the formula (IIa), bicyclo[2.2.1]-2-heptene (also referred to as "norbornene" in this specification) and tetracyclo[4.4.0.1 2,5 .1 7,10 ]-3-dodecene (also referred to as "tetracyclododecene" in this specification).

[0092] When the total number of repeating units constituting the cyclic olefin copolymer (P1) is taken as 100 mol %, the proportion of repeating units (b) derived from cyclic olefin monomers is preferably 5 mol % or more and 95 mol % or less, more preferably 10 mol % or more and 80 mol % or less, even more preferably 15 mol % or more and 60 mol % or less, and even more preferably 20 mol % or more and 50 mol % or less.

[0093] The copolymer type of the cyclic olefin copolymer (P1) is not particularly limited, and examples thereof include random copolymers and block copolymers.

[0094] Cyclic olefin copolymers (P1) include ethylene and tetracyclo[4.4.0.1 2,5 .1 7,10 ]-3-dodecene, random copolymers of ethylene and bicyclo[2.2.1]-2-heptene, and random copolymers of ethylene and tetracyclo[4.4.0.1 2,5 .1 7,10 The copolymer may be at least one selected from the group consisting of random copolymers of 1-dodecene and benzonorbornadiene.

[0095] The cyclic olefin copolymer (P1) may be used alone or in combination of two or more kinds.

[0096] The cyclic olefin copolymer (P1) can be produced by appropriately selecting the conditions according to the methods described in, for example, JP-A-60-168708, JP-A-61-120816, JP-A-61-115912, JP-A-61-115916, JP-A-61-271308, JP-A-61-272216, JP-A-62-252406, JP-A-62-252407, etc.

[0097] In addition, from the viewpoint of further improving heat resistance while maintaining a good balance of the transparency and refractive index of the obtained molded body and optical component, and from the viewpoint of further improving moldability, the cyclic olefin resin (A) may contain a cyclic olefin copolymer (P2) having a repeating unit (AA) represented by the formula (II) and a structural unit (C) derived from a cyclic olefin having an aromatic ring, wherein the repeating unit (AA) does not contain an aromatic ring, and the cyclic olefin having an aromatic ring contains at least one selected from the group consisting of a compound represented by the following formula (C-1), a compound represented by the following formula (C-2), and a compound represented by the following formula (C-3).

[0098] [ka]

[0099] In the formula (C-1), n ​​and q each independently represent 0, 1, or 2; 1 ~R 17 each independently represents a hydrogen atom, a halogen atom other than a fluorine atom, or a hydrocarbon group having 1 to 20 carbon atoms which may be substituted with a halogen atom other than a fluorine atom; R 10 ~R 17 One of them is a bond, and when q=0, R 10 and R 11 , R 11 and R 12 , R 12 and R 13 , R 13 and R 14 , R 14 and R 15 , R 15 and R 10 may be bonded to each other to form a monocyclic or polycyclic ring, and when q=1 or 2, R 10 and R 11 , R 11 and R 17 , R 17 and R 17 , R 17 and R 12 , R 12 and R 13 , R 13 and R 14 , R 14 and R 15 , R 15 and R 16 , R 16 and R 16 , R 16 and R 10 may be bonded to each other to form a monocycle or polycycle, the monocycle or polycycle may have a double bond, or the monocycle or polycycle may be an aromatic ring.

[0100] [ka]

[0101] In the formula (C-2), n and m are each independently 0, 1, or 2, q is 1, 2, or 3, and R 18 ~R 31each independently represents a hydrogen atom, a halogen atom other than a fluorine atom, or a hydrocarbon group having 1 to 20 carbon atoms which may be substituted with a halogen atom other than a fluorine atom, and when q=1, R 28 and R 29 , R 29 and R 30 , R 30 and R 31 may be bonded to each other to form a monocyclic or polycyclic ring, and when q=2 or 3, R 28 and R 28 , R 28 and R 29 , R 29 and R 30 , R 30 and R 31 , R 31 and R 31 may be bonded to each other to form a monocycle or polycycle, the monocycle or polycycle may have a double bond, and the monocycle or polycycle may be an aromatic ring.

[0102] [ka]

[0103] In the formula (C-3), q is 1, 2 or 3; R 32 ~R 39 each independently represents a hydrogen atom, a halogen atom other than a fluorine atom, or a hydrocarbon group having 1 to 20 carbon atoms which may be substituted with a halogen atom other than a fluorine atom, and when q=1, R 36 and R 37 , R 37 and R 38 , R 38 and R 39 may be bonded to each other to form a monocyclic or polycyclic ring, and when q=2 or 3, R 36 and R 36 , R 36 and R 37 , R 37 and R 38 , R 38 and R 39 , R 39 and R 39may be bonded to each other to form a monocycle or polycycle, the monocycle or polycycle may have a double bond, and the monocycle or polycycle may be an aromatic ring.

[0104] When the total content of the repeating unit (AA) and the structural unit (C) in the cyclic olefin copolymer (P2) is taken as 100 mol%, the content of the structural unit (C) in the cyclic olefin copolymer (P2) is preferably 5 mol% or more and 95 mol% or less, more preferably 5 mol% or more and 70 mol% or less, and even more preferably 5 mol% or more and 50 mol% or less.

[0105] In this embodiment, the contents of the repeating unit (AA) and the structural unit (C) are, for example, 1 H-NMR or 13 It can be measured by C-NMR.

[0106] The copolymer type of the cyclic olefin copolymer (P2) is not particularly limited, and examples thereof include random copolymers and block copolymers.

[0107] The cyclic olefin copolymer (P2) can be produced by appropriately selecting the conditions according to the methods described in, for example, JP-A-60-168708, JP-A-61-120816, JP-A-61-115912, JP-A-61-115916, JP-A-61-271308, JP-A-61-272216, JP-A-62-252406, JP-A-62-252407, JP-A-2007-314806, JP-A-2010-241932, etc.

[0108] [Other ingredients] In addition to the recycled cyclic olefin resin (A1) and the virgin cyclic olefin resin (A), the recycled cyclic olefin resin composition of this embodiment may contain other known optional components within a range that does not impair the good physical properties of the recycled cyclic olefin resin composition of the present invention.

[0109] Examples of other components include antioxidants, resins other than the cyclic olefin resin (A) (polycarbonate resins, polyester resins, acrylic resins), weather stabilizers, heat stabilizers, antistatic agents, flame retardants, slip agents, antiblocking agents, antifogging agents, lubricants, dyes, pigments, natural oils, synthetic oils, waxes, organic or inorganic fillers, etc. The blending ratio of other components is an appropriate amount.

[0110] <Second embodiment> In the first embodiment described above, the calculation unit 140 outputs combination information based on a trained model M (an example of related information) in response to an input of optical property information. In contrast, in the second embodiment, the calculation unit 140 outputs combination information based on a trained model M (an example of related information) in response to an input of optical property information. Details of the second embodiment will be described below.

[0111] Fig. 10 is a diagram showing an example of processing performed by the calculation unit 140 according to the second embodiment. As shown in Fig. 10, in the second embodiment, an example will be described in which the input variable input to the trained model M is optical property information, and the output variable output from the trained model M is blending information.

[0112] In the learning stage of the second embodiment, the learning unit 220 of the machine learning device 200 generates a trained model M by performing machine learning using the teacher data D2. That is, the learning unit 220 generates the trained model M by performing machine learning using the optical property information as an input variable and the blending information as an output variable. The learning unit 220 controls the communication unit 210 to transmit the trained model M to the information processing device 100. The information processing device 100 stores the trained model M received from the machine learning device 200 in the storage unit 170.

[0113] In the operation stage of the second embodiment, the calculation unit 140 of the information processing device 100 acquires optical property information input by a user using the input unit 120. Thereafter, the calculation unit 140 outputs blending information based on the trained model M in response to the input of the optical property information. For example, when the calculation unit 140 inputs optical property information (an example of an input variable) to the input layer L1 of the trained model M, blending information is output from the output layer L3 of the trained model M. The calculation unit 140 outputs the blending information acquired using the trained model M.

[0114] As described above, the calculation unit 140 of the information processing device 100 of this embodiment reads out the trained model M generated by machine learning using the optical property information as an input variable and the blending information as an output variable, and outputs blending information in response to the input of the optical property information based on the trained model M. This allows the information processing device 100 of this embodiment to output highly accurate blending information based on the optical property information.

[0115] <Third embodiment> In the third embodiment, the calculation unit 140 outputs the second optical property information based on the trained model M (an example of the relationship information) in response to the input of the blending information and the first optical property information. The details of the third embodiment will be described below.

[0116] FIG. 11 is a diagram showing an example of processing performed by the calculation unit 140 according to the third embodiment. As shown in FIG. 11, in the third embodiment, an example will be described in which the input variables input to the learned model M are blending information and first optical property information, and the output variables output from the learned model M are second optical property information. The first optical property information and the second optical property information are different information from each other. In this embodiment, as an example, the first optical property information is assumed to be haze and light transmittance, and the second optical property information is assumed to be refractive index, birefringence, and yellow index. Note that the first optical property information and the second optical property information may be a combination of information other than these.

[0117] In the learning stage of the third embodiment, the learning unit 220 of the machine learning device 200 generates a learned model M by performing machine learning using the teacher data D2. That is, the learning unit 220 generates the learned model M by performing machine learning using the blending information and the first optical property information (haze and light transmittance) as input variables and the second optical property information (refractive index, birefringence, and yellow index) as output variables. The learning unit 220 controls the communication unit 210 to transmit the learned model M to the information processing device 100. The information processing device 100 stores the learned model M received from the machine learning device 200 in the storage unit 170.

[0118] In the operation stage of the third embodiment, the calculation unit 140 of the information processing device 100 acquires the blending information and the first optical property information (haze and light transmittance) input by the user using the input unit 120. After that, the calculation unit 140 outputs the second optical property information (refractive index, birefringence, and yellow index) based on the trained model M in response to the input of the blending information and the first optical property information (haze and light transmittance). For example, when the calculation unit 140 inputs the blending information and the first optical property information (an example of an input variable) to the input layer L1 of the trained model M, the second optical property information is output from the output layer L3 of the trained model M. The calculation unit 140 outputs the second optical property information acquired using the trained model M.

[0119] As described above, the calculation unit 140 of the information processing device 100 of this embodiment reads out the trained model M generated by machine learning with the combination information and the first optical property information as input variables and the second optical property information as output variables, and outputs the second optical property information in response to the input of the combination information and the first optical property information based on the trained model M. This allows the information processing device 100 of this embodiment to output highly accurate second optical property information based on the combination information and the first optical property information.

[0120] <Fourth embodiment> In the fourth embodiment, the calculation unit 140 outputs combination information and first optical property information based on a trained model M (an example of relationship information) in response to an input of second optical property information. Details of the fourth embodiment will be described below.

[0121] FIG. 12 is a diagram showing an example of processing performed by the calculation unit 140 according to the fourth embodiment. As shown in FIG. 12, in the fourth embodiment, an example will be described in which the input variables input to the learned model M are the second optical property information, and the output variables output from the learned model M are the blending information and the first optical property information. The first optical property information and the second optical property information are different information from each other. As in the third embodiment described above, in this embodiment, as an example, the first optical property information is assumed to be haze and light transmittance, and the second optical property information is assumed to be refractive index, birefringence, and yellow index. Note that the first optical property information and the second optical property information may be a combination of information other than these.

[0122] In the learning stage of the fourth embodiment, the learning unit 220 of the machine learning device 200 generates a learned model M by performing machine learning using the teacher data D2. That is, the learning unit 220 generates the learned model M by performing machine learning using the second optical property information (refractive index, birefringence, and yellow index) as input variables and the blending information and the first optical property information (haze and light transmittance) as output variables. The learning unit 220 controls the communication unit 210 to transmit the learned model M to the information processing device 100. The information processing device 100 stores the learned model M received from the machine learning device 200 in the storage unit 170.

[0123] In the operation stage of the fourth embodiment, the calculation unit 140 of the information processing device 100 acquires the second optical property information (refractive index, birefringence, and yellow index) input by the user using the input unit 120. After that, the calculation unit 140 outputs the blending information and the first optical property information (haze and light transmittance) based on the trained model M in response to the input of the second optical property information (refractive index, birefringence, and yellow index). For example, when the calculation unit 140 inputs the second optical property information (an example of an input variable) to the input layer L1 of the trained model M, the blending information and the first optical property information are output from the output layer L3 of the trained model M. The calculation unit 140 outputs the blending information and the first optical property information acquired using the trained model M.

[0124] As described above, the calculation unit 140 of the information processing device 100 of this embodiment reads out the trained model M generated by machine learning with the second optical property information as an input variable and the blending information and the first optical property information as output variables, and outputs the blending information and the first optical property information in response to the input of the second optical property information based on the trained model M. This allows the information processing device 100 of this embodiment to output highly accurate blending information and first optical property information based on the second optical property information.

[0125] <Fifth embodiment> In the first embodiment described above, the information processing device 100 and the machine learning device 200 are different devices. In contrast, in the fifth embodiment, the functions of the machine learning device 200 are incorporated into the information processing device 100, so that an information processing device 100A in which the information processing device 100 and the machine learning device 200 are integrated is provided in the system 1. Details of the fifth embodiment will be described below.

[0126] Fig. 13 is a diagram showing the configuration of an information processing device 100A according to the fifth embodiment. As shown in Fig. 13, the information processing device 100A includes a learning unit 220. Furthermore, a storage unit 170 of the information processing device 100A stores teacher data D2.

[0127] In the learning stage of the fifth embodiment, the learning unit 220 of the information processing device 100A reads the teacher data D2 from the storage unit 170 and performs machine learning using the teacher data D2 to generate a trained model M. The learning unit 220 stores the generated trained model M in the storage unit 170.

[0128] In the operation stage of the fifth embodiment, the calculation unit 140 of the information processing device 100A reads out the trained model M from the storage unit 170. The calculation unit 140 also acquires input variables (e.g., blending information) input by a user using the input unit 120. After that, the calculation unit 140 outputs output variables (e.g., optical property information) based on the trained model M in response to the input of the input variables (e.g., blending information).

[0129] As described above, the information processing device 100A of this embodiment has both the functions of the information processing device 100 and the machine learning device 200 shown in Fig. 1. This allows the information processing device 100A of this embodiment to execute both learning stage processing and operation stage processing.

[0130] The system 1 of the first to fifth embodiments has been described above. The system 1 of these embodiments is a system for processing information related to a resin composition containing a recycled cyclic olefin resin. The system 1 includes a storage unit 170 and a calculation unit 140. The storage unit 170 stores relationship information in which blending information indicating the blending amount of the recycled cyclic olefin resin in the resin composition and optical property information indicating the optical properties of the recycled cyclic olefin resin are associated with each other. The calculation unit 140 outputs information associated with at least one of the information from among the blending information and the optical property information based on the relationship information in response to an input of at least one of the information from among the blending information and the optical property information. This allows the system 1 of the embodiment to output the blending amount or optical property of the recycled cyclic olefin resin with high accuracy and in a short time.

[0131] In the above embodiment, the trained model M has been described as an example of the relationship information, but the present invention is not limited to this. For example, the relationship information may be a table in which input variables and output variables are associated with each other.

[0132] The above describes the form for carrying out the present invention using an embodiment, but the present invention is not limited to such an embodiment, and various modifications and substitutions can be made within the scope that does not deviate from the gist of the present invention. [Explanation of symbols]

[0133] 1 System 100 Information processing device 110 Communications Department 120 Input section 130 Display section 140 Arithmetic section 150 Candidate generation section 160 Extraction part 170 Storage section 200 Machine Learning Device 210 Communications Department 220 Learning Department 270 Storage section D1 Candidate Data D2 Teacher data M trained models

Claims

1. A system for processing information regarding a resin composition containing recycled cyclic olefin resin, comprising: A storage unit that stores relationship information that correlates blending information indicating the blending amount of the recycled cyclic olefin resin in the resin composition with optical property information indicating the optical properties of the recycled cyclic olefin resin; A calculation unit that outputs, in response to an input of at least one piece of information among the blending information and the optical property information, the at least one piece of information and information related to the at least one piece of information by the relationship information based on the relationship information; A system comprising:

2. The calculation unit outputs the optical property information based on the relationship information in response to the input of the blending information. The system of claim 1 .

3. A learning unit that performs machine learning using the blending information as an input variable and the optical property information as an output variable to generate a learned model as the relationship information, The calculation unit outputs the optical property information based on the trained model in response to the input of the blending information. The system of claim 2.

4. A candidate generating unit that generates a plurality of candidates for the blending amount; An extraction unit that extracts at least one candidate from the plurality of candidates, The calculation unit outputs the optical property information based on the trained model in response to input of candidate blending information indicating each of the plurality of candidates, The extraction unit extracts at least one candidate from the plurality of candidates based on the optical characteristic information for each of the plurality of candidates. The system of claim 3.

5. The calculation unit outputs the blending information based on the relationship information in response to the input of the optical property information. The system of claim 1 .

6. The optical characteristic information includes at least one of haze, light transmittance, refractive index, birefringence, and yellow index. A system according to any one of claims 1 to 5.

7. A system for processing information regarding a resin composition containing recycled cyclic olefin resin, Reading from a storage unit relationship information correlating blending information indicating the blending amount of the recycled cyclic olefin resin in the resin composition with optical property information indicating the optical properties of the recycled cyclic olefin resin; In response to an input of at least one of the blending information and the optical property information, the at least one of the blending information and the optical property information is output based on the relationship information. Information processing methods.

8. A machine learning device that processes information about a resin composition containing recycled cyclic olefin resin, A learning unit that performs machine learning using blending information indicating the blending amount of the recycled cyclic olefin resin in the resin composition as an input variable and optical property information indicating the optical properties of the recycled cyclic olefin resin as an output variable to generate a trained model; A machine learning device comprising:

9. An information processing device that processes information regarding a resin composition containing a recycled cyclic olefin resin, a calculation unit that reads out a trained model generated by machine learning using blending information indicating the blending amount of the recycled cyclic olefin resin in the resin composition as an input variable and optical property information indicating the optical properties of the recycled cyclic olefin resin as an output variable, and outputs the optical property information based on the trained model in response to the input of the blending information; An information processing device comprising:

10. A computer that processes information regarding a resin composition containing a recycled cyclic olefin resin A machine learning is performed using blending information indicating the blending amount of the recycled cyclic olefin resin in the resin composition as an input variable and optical property information indicating the optical properties of the recycled cyclic olefin resin as an output variable to generate a trained model. program.

11. A computer that processes information regarding a resin composition containing a recycled cyclic olefin resin A trained model is read out by performing machine learning using blending information indicating the blending amount of the recycled cyclic olefin resin in the resin composition as an input variable and optical property information indicating the optical properties of the recycled cyclic olefin resin as an output variable; In response to the input of the blending information, the optical property information is output based on the trained model. program.

12. A trained model for causing a computer to function to output information about a resin composition containing a recycled cyclic olefin resin, Machine learning is performed using blending information indicating the blending amount of the recycled cyclic olefin resin in the resin composition as an input variable and optical property information indicating the optical properties of the recycled cyclic olefin resin as an output variable, thereby updating the learning variables; Performing a calculation based on the updated learning variables for the input blending information, and outputting the optical property information; A trained model for causing the computer to function.

Images