JP7633917B2 - System and program storage medium - Google Patents

Description

The present invention relates to a system and a program storage medium.

Patent Document 1, in relation to a method for authenticating a labeled polymer, describes the following: "A method for authenticating that a test polymer is a labeled polymer, in which the labeled polymer includes a substrate polymer, a compound containing a forensic authentication marker, and a dynamic response authentication marker, the forensic authentication marker being present in the labeled polymer in an amount sufficient to be detected by a forensic analysis technique, and the dynamic response authentication marker being present in the labeled polymer in an amount sufficient to be detected by a dynamic response analysis technique, comprising testing the test polymer for the forensic authentication marker using a forensic analysis technique, testing the test polymer for the dynamic response authentication marker using a dynamic response analysis technique, and authenticating that the test polymer is a labeled polymer if the forensic authentication marker and the dynamic response authentication marker are detected."

Special Publication No. 2007-514259

In recent years, as society shifts from the traditional linear economy of mass consumption and mass waste to a circular economy, there has been an increasing social demand for the use of recycled materials that are made by recovering and regenerating materials from waste. As a result, when the price of virgin materials falls, problems have arisen where material manufacturers falsely sell virgin materials as recycled materials. When material fraud occurs in this way, there is a risk that manufacturers may be held liable for material fraud for products that they market as using recycled materials.

The authentication method described in Patent Document 1 involves the material manufacturer adding an authentication marker to a labeled polymer and analyzing the test polymer for the authentication marker, and does not allow the manufacturer to determine the authenticity of the material. In addition, the technology in this document requires the authentication marker to be added to the material in advance, and requires additional analysis separate from the product manufacturing process, resulting in associated costs.

The present invention was made in consideration of the above problems, and aims to determine the authenticity of materials and authenticate the materials using sensor information obtained during the manufacturing process without the need for additional inspection processes or the addition of authentication markers to the materials.

The present application includes multiple means for solving at least part of the above problems, examples of which are as follows: A system according to one aspect of the present invention for solving the above problems is a system having one or more processors and one or more storage media, in which the processor learns measurement values obtained from multiple sensors installed in manufacturing equipment for products using raw materials, generates a trained classification model that outputs information used to determine the authenticity of raw materials, stores the trained classification model in the storage medium, inputs the measurement values of uncertified raw materials into the trained classification model, and certifies the raw materials when the uncertified raw materials are determined to be authentic.

According to the present invention, the authenticity of materials can be determined and materials authenticated using sensor information obtained during the manufacturing process without the need for additional inspection steps or the addition of authentication markers to the materials.

FIG. 2 is a schematic configuration diagram showing an example of a functional configuration of the system according to the first embodiment. FIG. 2 is a flow diagram showing an example of a manufacturing process executed in the system according to the first embodiment. FIG. 11 is a flow diagram illustrating an example of a process for generating a trained classification model. FIG. 11 is a flow diagram showing an example of a material lot authentication process. FIG. 13 is a diagram showing an outline of an experimental example for verifying the effects of the first embodiment. FIG. 13 is a diagram showing an example of an experimental result. FIG. 13 is a diagram showing an example of an experimental result. FIG. 2 illustrates an example of a hardware configuration of the system. FIG. 11 is a diagram illustrating an example of a hardware configuration of a system according to a second embodiment. FIG. 13 is a diagram illustrating an example of a hardware configuration of a system according to a third embodiment.

Below, each embodiment of the present invention will be described with reference to the drawings.

First Embodiment
1 is a schematic diagram showing an example of the functional configuration of a system 1000 according to this embodiment. As shown in the figure, the system 1000 has a plurality of subsystems, and the system 1000 as a whole performs authentication of materials (raw materials) and manufacture of products.

Specifically, the system 1000 has the following subsystems: production management system 100, manufacturing supervision system 200, material authentication system 300, and manufacturing execution system 400. Furthermore, each subsystem works in conjunction with the others to execute various processes.

In addition to the subsystems, the system 1000 also has functional units (not shown). Specifically, the system 1000 has an input reception unit that receives instructions and information from a user via an input device provided in the system 1000. The system 1000 also has an output display unit that generates display information (screen information) to be displayed on an output device (display) provided in the system 1000.

The system 1000 also has a storage unit that stores various information. Specifically, the storage unit has an authentication information DB 320, a material fingerprint information DB 340, and a classification model DB 370, which are used in the material authentication system 300. The storage unit also has a manufacturing process DB 420 that stores a learning manufacturing process 421, a trial manufacturing process 422, and a mass production manufacturing process 423, which are used in the manufacturing execution system 400.

The system 1000 also has a communication unit that communicates information with an external device (e.g., the manufacturing equipment 500). Specifically, the communication unit is connected to the external device so as to be able to communicate with each other via a communication path CN2, such as a public network such as the Internet, a LAN (Local Area Network), or a WAN (Wide Area Network).

The input reception unit, output display unit, memory unit, and communication unit work together with the functional units of each subsystem to carry out various processes.

Next, we will explain the functional configuration of the production management system 100.

The production management system 100 functions as a processing unit that executes various processes related to the management of production plans. Specifically, the production management system 100 has a production plan management unit 110. The production plan management unit 110 is a functional unit that generates a production plan that includes production specifications, quantities, timing, etc., in accordance with the order status and inventory status.

Next, we will explain the functional configuration of the manufacturing management system 200.

The manufacturing supervision system 200 functions as a processing unit that performs supervised processing related to manufacturing, such as issuing a predetermined type of manufacturing instruction to the manufacturing execution system 400. Specifically, the manufacturing supervision system 200 has a manufacturing condition determination unit 210, a material authentication acquisition unit 220, and a manufacturing instruction unit 230.

The manufacturing condition determination unit 210 is a functional unit that determines manufacturing conditions using the production plan generated by the production plan management unit 110. The manufacturing conditions include, for example, information that identifies the manufacturing equipment 500 (equipment within the factory) for the product, information that identifies the model number and lot of the material used in production, and the quantity and production time of the product to be produced.

Here, the material model number is identification information (e.g., a number) defined by the material manufacturer to identify the characteristics of the material. Also, a material lot is the smallest unit of production or delivery defined by the material manufacturer for materials with the same model number.

The material authentication acquisition unit 220 is a functional unit that acquires material authentication information from the material authentication system 300. Specifically, the material authentication acquisition unit 220 acquires material authentication information from the material authentication system 300 by sending basic information used for material authentication determined by the manufacturing condition determination unit 210 to the material authentication system 300. The basic information required for material authentication includes, for example, the model number of the material and the lot of the material.

When the material authentication acquisition unit 220 acquires information (authentication information) from the material authentication system 300 indicating that the lot of material used in production has been certified, it requests the production instruction unit 230 to mass-produce the product. When the material authentication acquisition unit 220 acquires information from the material authentication system 300 indicating that the lot of material used is not certified, it requests the production instruction unit 230 to perform trial production for material authentication. When the material authentication acquisition unit 220 acquires information from the material authentication system 300 indicating that no trained classification model exists that corresponds to the model number of the material used, it requests the production instruction unit 230 to perform training production to generate such a classification model.

Note that trial production refers to production using uncertified material lots. Training production is production to obtain information to be used in supervised learning to generate a trained classification model, and refers to production using materials that are known in advance, i.e., multiple material model numbers and lots that are known to be true.

The manufacturing instruction unit 230 is a functional unit that instructs the manufacturing execution system 400 to execute manufacturing. Specifically, the manufacturing instruction unit 230 outputs a manufacturing instruction for either mass production, trial production, or learning production to the manufacturing execution unit 410 based on a request from the material authentication acquisition unit 220. In addition to the manufacturing instruction, the manufacturing instruction unit 230 also outputs the manufacturing conditions determined by the manufacturing condition determination unit 210, such as the model number and lot of the material used in manufacturing, to the manufacturing execution unit 410.

Next, we will explain the functional configuration of the material authentication system 300.

The material authentication system 300 functions as a processing unit that performs various processes related to material authentication. Specifically, the material authentication system 300 has a material authentication issuing unit 310, a material fingerprint information generating unit 330, a dimensionality reduction unit 350, a classification model generating unit 360, a material authenticity determining unit 380, and a material verifying unit 390.

The material authentication issuing unit 310 is a functional unit that issues authentication information when the material for which authentication information has been requested has already been authenticated. Specifically, when the material authentication issuing unit 310 receives a request to acquire authentication information including the model number and lot of a material from the material authentication acquisition unit 220, it refers to the authentication information DB 320, and outputs the authentication information to the material authentication acquisition unit 220 if the lot of the material has already been authenticated.

In addition, the material certification issuing unit 310 refers to the classification model DB 370 when the material lot is uncertified, and if there is no trained classification model corresponding to the material model number related to the acquisition request, outputs information indicating this to the material certification acquisition unit 220. In addition, the material certification issuing unit 310 refers to the authentication information DB 320 and the classification model DB 370, and if there is a trained classification model corresponding to the material model number related to the acquisition request but the material lot is uncertified, outputs information indicating this to the material certification acquisition unit 220.

When the material authentication issuing unit 310 acquires information indicating that the material lot is genuine from the material authenticity determining unit 380 or the material verifying unit 390, it newly authenticates the material lot and stores the authentication information in the authentication information DB 320. In this case, the material authentication issuing unit 310 also issues authentication information for the model number and lot of the newly authenticated material and outputs it to the material authentication acquiring unit 220.

The material fingerprint information generation unit 330 is a functional unit that acquires values output from the various sensors 520 included in the manufacturing equipment 500 and product quality as measured values during trial manufacturing or learning manufacturing, and generates in-process material fingerprint information. The material fingerprint information generation unit 330 also stores the generated material fingerprint information in the material fingerprint information DB 340. For the measured values acquired in the case of learning manufacturing, the material fingerprint information generation unit 330 associates them with information indicating that the material is genuine and stores them in the material fingerprint information DB 340.

Here, material fingerprint information refers to multiple feature amounts correlated with the physical properties of a material obtained from the measurements of multiple sensors 520. The material authenticity determination unit 380 and material verification unit 390 can determine the authenticity of a material or verify its authenticity with high accuracy by utilizing multiple feature amounts obtained from multiple sensors 520, rather than just product quality or a single feature amount. Details of an experimental example showing this effect will be described later.

In addition, when the material fingerprint information generation unit 330 obtains a judgment result or a verification result regarding the authenticity of a material from the material authenticity judgment unit 380 or the material verification unit 390, it registers this in association with the corresponding material in the material fingerprint information DB 340.

The dimension reduction unit 350 is a functional unit that generates dimension-reduced features that reduce the number of dimensions (compress) so that the material fingerprint information can be input to a trained classification model. When handling multidimensional features of in-process material fingerprint information, dimensional reduction makes it easier to determine whether the information is true or false. Dimension reduction methods include, for example, UMAP (Uniform Manifold Approximation and Projection) or t-SNE (t-distributed Stochastic Neighbor Embedding), and the dimension reduction unit 350 uses one of these methods.

In addition, in the case of learning manufacturing, the dimension reduction unit 350 outputs the dimension reduction feature and information indicating that the material is genuine to the classification model generation unit 360 in association with each other. In addition, in the case of trial manufacturing, the dimension reduction unit 350 outputs the dimension reduction feature to the material genuineness determination unit 380.

The classification model generation unit 360 is a functional unit that generates a trained classification model for materials. Specifically, the classification model generation unit 360 learns dimensionality reduction features of materials that are known in advance, i.e., multiple materials that are known to be true, and generates a trained classification model that corresponds to the model number of the materials. The classification model generation unit 360 also stores the generated trained classification model in the classification model DB 370 in association with the model number of the material. Note that as the trained classification model, a supervised classification model such as logistics regression or a support vector machine (SVM) may be used.

The classification model generation unit 360 also performs active learning for uncertified material lots obtained in trial manufacturing by associating the dimension reduction features input from the dimension reduction unit 350 with the determination results or verification results of the authenticity of materials input from the material authenticity determination unit 380 or the material verification unit 390, and re-learning the trained classification model.

The material authenticity determination unit 380 is a functional unit that determines the authenticity of materials. Specifically, the material authenticity determination unit 380 acquires dimension reduction features based on material fingerprint information of an unauthenticated material lot obtained in trial production from the dimension reduction unit 350. In addition, the material authenticity determination unit 380 inputs the acquired dimension reduction features into a trained classification model corresponding to the model number of the material, thereby obtaining output information used to determine the authenticity of the material.

When the judgment result based on the output information from the trained classification model is true, the material authenticity judgment unit 380 associates the judgment result (true) with the material model number and lot and outputs it to the authentication information issuing unit. On the other hand, when the judgment result is false, the material authenticity judgment unit 380 requests the material verification unit 390 to perform additional verification of the unauthenticated material lot used in the trial manufacturing. When the judgment result is true, the material authenticity judgment unit 380 outputs the judgment result to the classification model generation unit 360 and the material fingerprint information generation unit 330.

The material verification unit 390 is a functional unit that performs additional verification of materials. Specifically, in response to a request for additional verification from the material authenticity determination unit 380, the material verification unit 390 verifies the authenticity of materials using any analysis technique.

The analytical technique may be any of a variety of techniques, including spectroscopic analysis such as FT-IR, NIR, or Raman spectroscopy, thermal analysis such as TG-DTA or DSC, or mass analysis such as GC-MS, and the material verification unit 390 will select an appropriate analytical technique from these.

If the verification result is true, the material verification unit 390 associates the verification result (true) with the material model number and lot and outputs it to the material authentication issuing unit 310. On the other hand, if the verification result is false, the material verification unit 390 outputs information recommending that the material lot be changed to the material authentication issuing unit 310. The material verification unit 390 also outputs the verification result to the classification model generation unit 360 and the material fingerprint information generation unit 330.

Next, we will explain the functional configuration of the manufacturing execution system 400.

The manufacturing execution system 400 functions as a processing unit that manufactures products by executing a predetermined manufacturing process based on instructions from the manufacturing supervision system 200. Specifically, the manufacturing execution system 400 has a manufacturing execution unit 410 and a quality inspection unit 430.

The manufacturing execution unit 410 is a functional unit that manufactures products by executing a specified manufacturing process based on instructions from the manufacturing supervision system 200. Specifically, when the manufacturing execution unit 410 receives a request for learning manufacturing from the manufacturing instruction unit 230, it acquires the learning manufacturing process 421 from the manufacturing process DB 420. In addition, the manufacturing execution unit 410 outputs an instruction to execute manufacturing using the learning manufacturing process 421 to the manufacturing equipment 510 using the model numbers and lots of materials that are known in advance, i.e., multiple materials that are known to be true.

In addition, when a trial production is requested by the production instruction unit 230, the production execution unit 410 retrieves a trial production process 422 from the production process DB 420. In addition, the production execution unit 410 outputs an instruction to execute production by the trial production process 422 to the production equipment 510 using an uncertified material lot.

In addition, when mass production is requested by the manufacturing instruction unit 230, the manufacturing execution unit 410 acquires the mass production process 423 from the manufacturing process DB 420. In addition, the manufacturing execution unit 410 outputs an instruction to execute manufacturing by the mass production process 423 to the manufacturing equipment 510 using the model number and lot of the authenticated material.

The quality inspection unit 430 is a functional unit that inspects the quality of the product. Specifically, the quality inspection unit 430 inspects the quality of the product based on, for example, shape characteristics, mechanical and optical characteristics, etc.

The quality inspection unit 430 also associates the model number and lot of the material that underwent quality inspection with the quality inspection results and outputs them to the material fingerprint information generation unit 330 of the material authentication system 300.

The functional configuration of each subsystem has been explained above.

Next, we will explain the various types of information stored in the memory unit of this system 1000.

The authentication information DB 320 used in the material authentication system 300 stores, in association with each other, information identifying the material model number, information identifying the lot, and information indicating whether the corresponding lot has been authenticated.

In addition, the material fingerprint information DB340 stores, in association with each other, information indicating features extracted from the measured values, information identifying the model number and lot of the material, and information indicating the authenticity of the material.

In addition, the classification model DB370 stores, for example, material model numbers and corresponding trained classification models in association with each other.

In addition, the manufacturing process DB 420 used in the manufacturing execution system 400 stores various manufacturing processes for carrying out learning, trial, and mass production manufacturing.

The memory unit also stores specific information such as the order status and inventory status of the product, and the operating status of the manufacturing equipment 500 at the manufacturing plant.

The communication unit of the system 1000 communicates information between the material authentication system 300 and the manufacturing equipment 500. The communication unit also communicates information between the manufacturing execution system 400 and the manufacturing equipment 500.

The manufacturing facility 500 includes manufacturing equipment 510 that manufactures products based on the manufacturing process output from the manufacturing execution unit 410, and a number of various sensors 520 that acquire measurement values by measuring physical quantities during the manufacturing process. The sensors 520 also output the measurement values to the material fingerprint information generation unit 330 of the material authentication system 300.

The above describes the general configuration of the system 1000 according to this embodiment.

[Operation Description]
2 is a flow diagram showing an example of a manufacturing process executed by the present system 1000. This process is started when the input receiving unit receives an instruction to execute the manufacturing process from a user via an input device, for example.

When processing begins, the production management system 100 determines a production plan (step S10). Specifically, the production plan management unit 110 of the production management system 100 acquires predetermined information such as order status and inventory status from the storage unit. In addition, the production plan management unit 110 generates screen information for displaying the information acquired from the storage unit via the output display unit, and displays this on an output device (display).

The user refers to the displayed order status and inventory status, and inputs the appropriate production use, quantity, and timing via the input device. When the production plan management unit 110 acquires the input information via the input reception unit, it determines the production plan based on a specified algorithm.

The production plan management unit 110 may automatically determine the production plan without relying on input from the user by using a mathematical planning model and its algorithm for optimizing the entire logistics. Such a mathematical planning model and algorithm may be stored in the storage unit in advance.

Next, the manufacturing condition determination unit 210 of the manufacturing supervision system 200 determines the manufacturing conditions (step S20). Specifically, the manufacturing condition determination unit 210 acquires the production plan and the operating status of the manufacturing equipment 500 from the storage unit. In addition, the manufacturing condition determination unit 210 generates screen information for displaying the information acquired from the storage unit via the output display unit, and displays this on the output device (display).

The user refers to the displayed production plan and the operating status of the manufacturing equipment 500, and inputs manufacturing conditions related to an appropriate combination of manufacturing equipment 500 and materials via the input device. When the manufacturing condition determination unit 210 acquires the input information via the input reception unit, it determines the manufacturing conditions based on a predetermined algorithm. In addition, the manufacturing condition determination unit 210 outputs the manufacturing conditions, including the model number and lot of the material used to manufacture the product, to the material authentication acquisition unit 220.

The manufacturing condition determination unit 210 may automatically determine the manufacturing conditions without relying on input from the user by using a mathematical programming model and its algorithm for optimizing production efficiency. Furthermore, such a mathematical programming model and algorithm may be stored in the storage unit in advance.

Next, the material authentication acquisition unit 220 makes a request to acquire authentication information for the material included in the manufacturing conditions (step S30). Specifically, the material authentication acquisition unit 220 outputs a request to acquire authentication information for the material, including the model number and lot of the material, to the material authentication issuer 310.

Next, the material authentication issuing unit 310 determines whether or not a trained classification model exists for the material related to the acquisition request (step S40). Specifically, the material authentication issuing unit 310 refers to the authentication information DB 320, and if the material lot related to the authentication information acquisition request is uncertified, refers to the classification model DB 370. Furthermore, if a trained classification model corresponding to the model number of the material related to the acquisition request is registered in the classification model DB 370, the material authentication issuing unit 310 determines that a trained classification model corresponding to the model number of the material related to the acquisition request exists.

If it is determined that a trained classification model exists (Yes in step S40), the material certification issuing unit 310 proceeds to step S50. On the other hand, if it is determined that a trained classification model does not exist (No in step S40), the material certification issuing unit 310 outputs information indicating that such a classification model does not exist to the material certification acquisition unit 220, and proceeds to step S90. In step S90, a process for generating a trained classification model is executed. The process for generating a trained classification model will be described in detail later.

If it is determined that a trained classification model exists (Yes in step S40), the material authentication issuing unit 310 proceeds to step S50, where it determines whether authentication information for the material lot to be used exists. Specifically, the material authentication issuing unit 310 refers to the authentication information DB 320 and checks whether authentication information for the material lot related to the acquisition request is registered. Furthermore, if authentication information for the material lot is registered, the material authentication issuing unit 310 determines that authentication information for the material lot related to the acquisition request exists.

If it is determined that authentication information for the material lot exists (Yes in step S50), the material authentication issuing unit 310 issues authentication information for the model number and lot of the material related to the acquisition request. The material authentication issuing unit 310 also outputs the authentication information to the material authentication acquisition unit 220 and proceeds to step S60. On the other hand, if it is determined that authentication information for the material lot does not exist (No in step S50), the material authentication issuing unit 310 outputs information indicating that the material lot related to the acquisition request has not been authenticated to the material authentication acquisition unit 220 and proceeds to step S80. In step S80, authentication processing for the material lot is executed. Details of the authentication processing for the material lot will be described later.

In step S60, which is reached when it is determined that authentication information for the material lot exists (Yes in step S50), the material authentication acquisition unit 220 acquires the material model number and lot authentication information included in the manufacturing conditions from the authentication information issuing unit. In addition, upon acquiring the authentication information, the material authentication acquisition unit 220 outputs a request to execute mass production to the manufacturing instruction unit 230, and the process proceeds to step S70. Note that such a request includes information such as the production plan and manufacturing conditions.

In step S70, the manufacturing instruction unit 230 generates mass production instruction information using the production plan and manufacturing conditions, and outputs this to the manufacturing execution system 400. Furthermore, when the manufacturing execution unit 410 of the manufacturing execution system 400 acquires the mass production instruction, it acquires the mass production process 423 from the manufacturing process DB 420, and outputs the mass production process 423 to the manufacturing equipment 500 via the communication unit, thereby executing mass production of products according to the production plan and manufacturing conditions. Furthermore, when the manufacturing execution unit 410 outputs the mass production process 423 to the manufacturing equipment 500, the processing of this flow ends.

Next, we will explain the details of the process of generating a trained classification model (step S90).

Figure 3 is a flow diagram showing an example of the process of generating a trained classification model.

In step S91, the manufacturing control system 200, the manufacturing execution system 400, and the material authentication system 300 cooperate to execute learning manufacturing and generate a learned classification model. Specifically, when the material authentication acquisition unit 220 of the manufacturing control system 200 acquires information indicating that there is no learned classification model corresponding to the model number of the material related to the authentication information acquisition request, it requests the manufacturing instruction unit 230 to execute learning manufacturing to generate such a classification model.

In addition, the manufacturing instruction unit 230 of the manufacturing supervision system 200 outputs manufacturing instructions for learning manufacturing, including the model number and lot of the material used in manufacturing, to the manufacturing execution unit 410 based on a request from the material authentication acquisition unit 220.

When the manufacturing execution unit 410 obtains manufacturing instructions for learning manufacturing from the manufacturing instruction unit 230, it obtains the learning manufacturing process 421 from the manufacturing process DB 420. The manufacturing execution unit 410 also executes learning manufacturing by outputting to the manufacturing equipment 510 an instruction to execute manufacturing using the learning manufacturing process 421 using the model numbers and lots of materials that are known in advance, i.e., multiple materials that are known to be true.

Next, the material fingerprint information generating unit 330 of the material authentication system 300 generates material fingerprint information using the measurement values in the learning production (step S92). Specifically, the material fingerprint information generating unit 330 acquires the measurement values (sensor values and product quality) output from the multiple various sensors 520 and the quality inspection unit 430 in the manufacturing equipment 500 in the learning production. The material fingerprint information generating unit 330 also extracts multiple feature quantities that correlate with the material properties from the acquired measurement values, and generates in-process material fingerprint information. The material fingerprint information generating unit 330 also associates the generated material fingerprint information with the authenticity of the material (in this case, authenticity) and stores it in the material fingerprint information DB 340.

Next, the dimension reduction unit 350 generates dimension-reduced features (step S93). Specifically, the dimension reduction unit 350 uses a predetermined method such as UMAP to acquire material fingerprint information and information on the authenticity of the material (in this case, authenticity) from the material fingerprint information DB 340, and uses this information to generate dimension-reduced features with a reduced number of dimensions that can be input to the trained classification model.

Next, the classification model generation unit 360 generates a trained classification model (step S94). Specifically, the classification model generation unit 360 uses a predetermined supervised classification model to learn dimensionality reduction features, thereby generating a trained classification model corresponding to the model number of the material in the training manufacturing.

Next, the classification model generation unit 360 registers the generated trained classification model in the classification model DB 370 (step S95), and ends the processing of this flow. After ending this flow, the classification model generation unit 360 returns the processing to step S30 in FIG. 2.

The above explains the process of generating a trained classification model.

Next, we will explain the details of the material lot authentication process (step S80).

Figure 4 is a flow diagram showing an example of material lot authentication processing.

In step S81, the manufacturing control system 200, the manufacturing execution system 400, and the material authentication system 300 cooperate to carry out trial manufacturing and authenticate the material lot. Specifically, when the material authentication acquisition unit 220 of the manufacturing control system 200 acquires information indicating that the material lot related to the authentication information acquisition request is uncertified, it requests the manufacturing instruction unit 230 to carry out trial manufacturing to authenticate the material lot.

In addition, the manufacturing instruction unit 230 of the manufacturing supervision system 200 outputs manufacturing instructions regarding trial manufacturing, including the model number and lot of the material to be used in manufacturing, to the manufacturing execution unit 410 based on a request from the material authentication acquisition unit 220.

When the manufacturing execution unit 410 obtains a manufacturing instruction for trial manufacturing from the manufacturing instruction unit 230, it obtains the trial manufacturing process 422 from the manufacturing process DB 420. The manufacturing execution unit 410 also executes trial manufacturing by outputting an instruction to execute manufacturing by the trial manufacturing process 422 to the manufacturing equipment 510 using the model number and lot of the material to be authenticated.

Next, the material fingerprint information generation unit 330 of the material authentication system 300 generates material fingerprint information using the measurement values in the trial production (step S82). Specifically, the material fingerprint information generation unit 330 acquires the measurement values (sensor values and product quality) output from the multiple various sensors 520 and the quality inspection unit 430 in the manufacturing equipment 500 during the trial production. The material fingerprint information generation unit 330 also extracts multiple feature amounts that correlate with the material properties from the acquired measurement values, and generates in-process material fingerprint information. The material fingerprint information generation unit 330 also stores the generated material fingerprint information in the material fingerprint information DB 340.

Next, the dimension reduction unit 350 generates dimension-reduced features (step S83). Specifically, the dimension reduction unit 350 acquires the material fingerprint information from the material fingerprint information DB 340, and uses a predetermined method such as UMAP to generate dimension-reduced features with a reduced number of dimensions that can be input to the trained classification model.

Next, the material authenticity determination unit 380 determines whether the result of the authenticity determination by the trained classification model is true or not (step S84). Specifically, the material authenticity determination unit 380 acquires a trained classification model corresponding to the model number of the material lot to be authenticated from the classification model DB 370. In addition, the material authenticity determination unit 380 inputs the generated dimensionality reduction features into the trained classification model to acquire output information to be used for determining the authenticity of the material, and uses this to determine the authenticity of the material.

If it is determined that the authenticity determination result is true (Yes in step S84), the material authenticity determination unit 380 outputs to the material authentication issuing unit 310 that the authenticity determination result of the material lot is true. The material authenticity determination unit 380 also outputs the determination result to the classification model generation unit 360 and the material fingerprint information generation unit 330, and the process proceeds to step S85. The material fingerprint information generation unit 330 registers the determination result in association with the material fingerprint information of the corresponding material lot in the material fingerprint information DB 340.

On the other hand, if the result of the authenticity determination is determined to be not authentic (No in step S084), the material authenticity determination unit 380 requests additional verification from the material verification unit 390 and transitions to step S87.

Next, in step S85, which is performed if it is determined that the result of the authenticity determination is true (Yes in step S84), the material authentication issuing unit 310 authenticates the material lot related to the trial production. Specifically, when the material authentication issuing unit 310 obtains information indicating that the result of the authenticity determination is true from the material authenticity determination unit 380, it authenticates the material lot related to the trial production and registers this in the authentication information DB 320. The material authentication issuing unit 310 issues authentication information and outputs it to the material authentication acquisition unit 220, and the process proceeds to step S86.

The classification model generation unit 360 also performs re-learning (active learning) of the trained classification model using the authenticity determination result of the material by the material authenticity determination unit 380, the dimensionality reduction feature used in the authenticity determination, and the trained classification model of the corresponding material model number (step S86). The classification model generation unit 360 also uses the re-learned trained classification model to update the corresponding trained classification model registered in the classification model DB 370. After performing this process, the classification model generation unit 360 ends the process of this flow. After ending this flow, the classification model generation unit 360 returns the process to step S30 in FIG. 2.

In addition, in step S87, which is performed when it is determined that the result of the authenticity determination described above is not authentic (No in step S84), the material verification unit 390 determines whether the result of the authenticity determination by additional verification is authentic. Specifically, the material verification unit 390 performs additional verification of the material lot using any analysis technique such as FT-IR, and determines its authenticity.

If the result of the authenticity determination is determined to be authentic (Yes in step S87), the material verification unit 390 proceeds to step S85. Note that the process of step S85 is the same as described above, so a detailed description will be omitted.

On the other hand, if it is determined that the result of the true/false judgment is not true (No in step S87), the material verification unit 390 displays information that the verification result is false and that it is recommended to change the material lot on the output device (display) via the output display unit (step S88), and the process proceeds to step S86. In step S86, the trained classification model is retrained (active learning) using the fact that the judgment result was false, the dimensionality reduction features used in the true/false judgment, and the trained classification model for the corresponding material model number.

In this case, the material verification unit 390 also outputs information indicating that the judgment result was false to the classification model generation unit 360 and the material fingerprint information generation unit 330. When the material fingerprint information generation unit 330 obtains such a judgment result, it registers it in the material fingerprint information DB 340 in association with the material fingerprint information of the corresponding material lot.

The above explains the material lot authentication process.

Next, we will use experimental examples to explain the effectiveness of the system according to this embodiment, which can determine the authenticity of materials with high accuracy by utilizing multiple feature amounts obtained from multiple sensors 520 to authenticate materials.

Figure 5 is a diagram showing an outline of an experimental example for verifying the effect of this embodiment. The experimental example is an experimental example in an injection molding process for manufacturing a resin product (e.g., a product using plastic resin). The illustrated mold 600 has a structure in which resin is flowed into a product forming section in the mold 600 from a runner 650 mounted in the mold 600 using a five-point pin gate method. In the experimental example, multiple pressure sensors and resin temperature sensors are arranged in a sensor arrangement section 660 provided on the runner 650 and a sensor arrangement section 610 provided on the product forming section. Measurement values of the time changes in pressure and temperature were obtained from these sensors.

In addition, in this experimental example, the peak pressure, the maximum differential value of the pressure, the integral value of the pressure, and the maximum differential value of the temperature, as well as the product weight as an indicator of product quality, were obtained as measured values, and the feature quantities of these multiple measured values were generated as in-process material fingerprint information.

Polypropylene was used as the material for manufacturing. As the real material for learning, one model number of recycled polypropylene derived from discarded home appliances was used, and 15 different material lots of that model number were used for manufacturing. As fake materials for learning, one lot of virgin material of each of eight model numbers and three lots of recycled material of each of two model numbers were used for manufacturing.

We also used UMAP as the dimensionality reduction model and SVM as the classification model.

Figure 6 is a scatter plot showing the experimental results of constructing a classification model using SVM using only two parameters, the pressure integral value and product weight obtained from a sensor installed on the runner 650. As shown in the figure, a negative correlation is observed between the pressure integral value shown on the horizontal axis and the product weight shown on the vertical axis, and a tendency is observed for them to be plotted at different positions depending on the differences in material properties. However, it can be seen that some false actual measurement results are included in the true classification area. From these experimental results, it can be concluded that when using a trained classification model constructed based on a small number of parameters, it is difficult to perform high-precision classification for materials with similar properties.

Figure 7 is a scatter plot showing the experimental results of generating 20-dimensional features as in-process material fingerprint information, performing 2-dimensional dimensionality reduction, and then constructing a trained classification model. As shown in the figure, the horizontal axis corresponds to dimensionality reduction feature 1, and the vertical axis corresponds to dimensionality reduction feature 2. In addition, the 20-dimensional features are generated based on the types of features in the aforementioned measured values (four types in this example) and the number of sensors that measured each feature (e.g., five sensors in this example). Note that features measured using different types of sensors or features measured at different positions can each be treated as mutually different features.

In this experimental example, the recycled polypropylene used as the true material is a mixture of various materials and materials with usage histories, resulting in large variations between material lots. As a result, a wide distribution was observed depending on the material lot in the obtained scatter plot. Furthermore, compared to Figure 6, no false actual measurement results were included in the true classification area of the classification obtained by SVM, indicating that the trueness of materials can be classified with higher accuracy. This is thought to be because by using multiple feature values obtained from multiple sensors as in-process material fingerprint information, dimensionality-reduced feature values that reflect even slight differences in material properties were obtained.

As described above, this system can determine the authenticity of materials and authenticate the materials using sensor information obtained during the manufacturing process without the need for additional inspection processes or the addition of authentication markers to the materials. Furthermore, this system can manufacture products using only authenticated materials.

In particular, this system can determine authenticity and authenticate materials based on measurements obtained from multiple sensors during the manufacturing process, without the need for additional inspection steps or the addition of authentication markers. In addition, by using measurements obtained from multiple sensors as in-process material fingerprint information for the material, it becomes possible to determine authenticity with higher accuracy than if it were determined based only on limited parameters such as material quality and molded product quality.

Furthermore, this system makes it possible to determine the authenticity of materials with large variations, such as recycled materials. This allows the manufacturing manufacturer to independently authenticate the materials, even if the materials manufacturer counterfeits the materials. Therefore, by using certified materials for manufacturing, the manufacturing manufacturer can prevent the risks associated with the discovery of counterfeit materials.

In addition, according to this system, if an uncertified material is determined to be fake, additional verification analysis is performed and the trained classification model is retrained using the dimensionality-reduced features associated with the authenticity determination result, thereby improving the accuracy of the trained classification model.

In addition, according to this system, even if an uncertified material is judged to be genuine by the trained classification model, the accuracy of the trained classification model can be improved by re-training using the dimensionality reduction features associated with the true judgment result.

Next, we will explain the hardware configuration of the system 1000 according to this embodiment.

Figure 8 shows an example of the hardware configuration of the present system 1000. The present system 1000 can be realized, for example, using one computer 700 (information processing device).

Specifically, the computer 700 includes an arithmetic unit 710, a memory 720, a storage device 730, an input device 740, an output device 750, a communication device 760, and a media interface 770, and these devices are electrically interconnected by a communication path CN1. Note that the communication path CN1 is, for example, an internal bus or a LAN (Local Area Network).

The computing device 710 is composed of, for example, a processor (including a microprocessor, a CPU (Central Processing Unit), a GPU (Graphics Processing Unit), or other semiconductor device).

Memory 720 is a non-volatile or volatile storage medium such as RAM (Random Access Memory) or ROM (Read Only Memory).

The storage device 730 is a device that stores computer programs 731 that realize the functional parts of each subsystem, various information used in the processing of the entire system 1000 or each subsystem, etc. The storage device 730 is a rewritable storage medium such as a flash memory, a hard disk, or an SSD (Solid State Drive).

The input device 740 is a device through which a user (e.g., an operator using the system) inputs information and instructions to the computer 700. Examples of the input device 740 include a keyboard, a touch panel, a pointing device such as a mouse, and a voice input device such as a microphone. The output device 750 (display device) is a device that outputs (displays) information generated by the computer 700 or information obtained from an external device. Examples of the output device 750 include a display, a printer, and a voice synthesizer.

The communication device 760 is a device that allows communication between an external information processing device and the computer 700 via the communication path CN2. Examples of external information processing devices include another computer 700 and an external storage device 800 (not shown). Examples of the external storage device 800 include a USB (Universal Serial Bus) memory, a memory card, and a hard disk.

The computer 700 can read information (e.g., unique information and production performance information used by the computer 700 of the present system 1000) and computer programs stored in the external storage device 800. The computer 700 can also transmit all or part of the computer program 731 and information stored in the storage device 730 to the external storage device 800 for storage therein.

The medium interface 770 is a device that reads and writes to an external recording medium 810 (e.g., a USB memory, a memory card, or a DVD). The calculator 700 can transfer a computer program 731 and information from the external recording medium 810 to the storage device 730, and can also transfer all or part of the computer program 731 and information stored in the storage device 730 to the external recording medium 810 for storage therein.

The above describes the hardware configuration of the system 1000 according to this embodiment.

The processing units of the system 1000, i.e., the functional units of each subsystem, are realized by a program 731 that causes the computing device 710 to perform processing. This program is stored, for example, in the storage device 730 or the external storage device 800, and is loaded, for example, into the memory 720 when the program 731 is executed, and is executed by the computing device 710.

The input reception unit is realized by the input device 740. The output display unit is realized by the output device 750. The storage unit is realized by the memory 720, the storage device 730, or the external storage device 800, or a combination of these. The communication unit is realized by the communication device 760.

Furthermore, each configuration, function, processing unit, processing means, etc. of the present system 1000 may be realized in part or in whole by hardware, for example by designing it as an integrated circuit. Furthermore, the present system 1000 may also realize each function in part or in whole by software, or by a combination of software and hardware. Furthermore, the present system 1000 may use hardware having fixed circuits, or may use hardware having at least some of the circuits that are changeable.

As mentioned above, each subsystem can also be implemented as a system by having a user (operator) implement some or all of the functions without using dedicated software or hardware.

In this embodiment, the system 1000 is realized from one computer 700, but the present invention is not limited to this. It is also possible to configure one or more systems by linking multiple computers 700. The configuration of such a system will be described in detail in the second and third embodiments below.

Second Embodiment
In the above-described first embodiment, a case has been described in which the system 1000 is realized by one computer 700, but the present invention is not limited to such a configuration.

Figure 9 is a diagram showing an example of the hardware configuration of the system 1000 according to the second embodiment. As shown in the figure, the system 1000 according to the second embodiment includes a computer 700a (hereinafter sometimes referred to as the material authentication side computer 700a) that realizes the material authentication system 300, and a plurality of computers 700b (hereinafter sometimes referred to as the user side computers 700b) that realize the other subsystems, the production management system 100, the manufacturing supervision system 200, and the manufacturing execution system 400, each of which is connected to be able to communicate with each other via a communication path CN2. Note that there may be only one user side computer 700b. Note that the same components are given the same reference numerals and detailed explanations are omitted.

Even with this system configuration, the authenticity of materials can be determined and the materials authenticated using sensor information obtained during the manufacturing process, without the need for additional inspection processes or the addition of authentication markers to the materials. In particular, in this system, the material authentication computer is placed on the cloud, and each user computer requests material authentication from the material authentication computer via communication path CN2, thereby achieving the same effect as in the first embodiment.

Furthermore, with this configuration, the material authentication computer (material authentication system) receives material authentication requests from multiple user computers, which allows the trained classification model to perform more re-learning, thereby further improving the accuracy of determining whether the material is genuine or counterfeit. Furthermore, the user computer executes product manufacture when it obtains authentication information using a trained classification model with improved accuracy in determining whether the material is genuine or counterfeit, allowing the manufacturer to manufacture the materials used with high reliability.

Third Embodiment
In the second embodiment described above, the computer 700a which realizes the material authentication system 300 and the multiple computers 700b which realize the other subsystems are connected to each other via the communication path CN2 so that they can communicate with each other. However, the present invention is not limited to such a configuration.

Figure 10 is a diagram showing an example of the hardware configuration of a system 1000 according to the third embodiment. As shown in the figure, in the system 1000 according to the third embodiment, each subsystem is realized by a separate computer 700o to 700r, which are communicatively interconnected via a communication path CN2. As shown in the figure, in this embodiment, there may be multiple computers 700r that realize the manufacturing execution system 400. Note that the same components are denoted by the same reference numerals and detailed explanations are omitted.

Even with this configuration, the same effects as in the first embodiment can be obtained. In particular, when there are multiple computers 700r that realize the manufacturing execution system 400, even if each is located in a distributed manufacturing plant, the processes executed by each computer 700 of the manufacturing execution system 400 can be consolidated and managed by a common production management system 100, manufacturing supervision system 200, and material authentication system 300.

The computer 700 in the system of the present invention can also be provided as a computer (distribution server) that distributes a program 731 for realizing an arbitrary subsystem among subsystems such as the production management system 100, the manufacturing supervision system 200, the material authentication system 300, and the manufacturing execution system 400 to other computers 700. For example, when the present system 1000 receives a request to acquire the program 731 for realizing an arbitrary subsystem from an external computer 700 (not shown), it can transmit the computer program 731 for realizing each subsystem stored in the memory unit to the external computer 700 via the communication unit.

In the above embodiment, the measurements obtained during trial production are used to determine whether the material is genuine or not by the material genuineness determination unit 380, and the dimension reduction features used in the genuineness determination are used to determine whether the material is genuine. The present invention is not limited to this. For example, the system 1000 may input dimension reduction features generated based on measurements obtained during mass production and information indicating that the material is genuine into the trained classification model of the corresponding material model number, thereby repeatedly retraining the trained classification model during mass production.

The present invention is not limited to the above-mentioned embodiments and modifications, but includes various modifications within the scope of the same technical idea. For example, the above-mentioned embodiments have been described in detail to clearly explain the present invention, and are not necessarily limited to those having all of the configurations described. It is also possible to replace part of the configuration of one embodiment with the configuration of another embodiment, and it is also possible to add the configuration of another embodiment to the configuration of one embodiment. It is also possible to add, delete, or replace part of the configuration of each embodiment with other configurations.

In addition, in the above explanation, the control lines and information lines are those that are considered necessary for the explanation, and do not necessarily show all the control lines and information lines in the product. In reality, it can be assumed that almost all components are interconnected.

1000...System, 100...Production management system, 110...Production planning management department, 200...Manufacturing management system, 210...Manufacturing condition determination department, 220...Material authentication acquisition department, 230...Manufacturing instruction department, 300...Material authentication system, 310...Material authentication issuing department, 320...Authentication information DB, 330...Material fingerprint information generation department, 340...Material fingerprint information DB, 350...Dimension reduction department, 360...Classification model generation department, 370...Classification model DB, 380...Material authenticity determination department, 390...Material verification department, 400...Manufacturing Execution system, 410... manufacturing execution unit, 420... manufacturing process DB, 430... quality inspection unit, 500... manufacturing equipment, 510... manufacturing device, 520... sensor, 700... computer, 710... arithmetic unit, 720... memory, 730... storage device, 731... computer program that realizes each functional unit of the subsystem, 740... input device, 750... output device, 760... communication device, 770... media interface, 800... external storage device, 810... external recording medium, CN1, CN2... communication path

Claims (11)

1. A system having one or more processors and one or more storage media, comprising:
The processor,
generating a trained classification model that learns measurement values obtained from a plurality of sensors installed in a manufacturing facility for a product using raw materials and outputs information used to determine the authenticity of the raw materials, and storing the trained classification model in the storage medium;
inputting the measurements of uncertified raw materials into the trained classification model;
If the uncertified raw material is determined to be genuine, the system authenticates the raw material. 2. The system of claim 1,
The processor,
generating in-process fingerprint information including a plurality of feature quantities extracted from the measurements in a training production using a plurality of raw materials that are known to be true;
generating dimension-reduced features by reducing the number of dimensions of the features in the in-process fingerprint information;
A system characterized by generating a trained classification model corresponding to the characteristics of the raw material by training the dimensionality-reduced features into a predetermined type of classification model. 3. The system of claim 2,
The feature amount includes:
The system includes at least one of the following: peak pressure, maximum differential of pressure, integral of pressure, and maximum differential of temperature. 2. The system of claim 1,
The processor,
If the uncertified raw materials are determined to be fake, additional verification analysis will be conducted regarding their authenticity;
and authenticating the uncertified raw material if the result of the verification analysis is determined to be true. 3. The system of claim 2,
The processor,
A system characterized by updating the trained classification model by re-training the trained classification model using the truth/falsehood determination result, the dimensionality reduction features, and the trained classification model corresponding to the raw materials. 2. The system of claim 1,
The processor,
Request the acquisition of said certification for the raw materials used in production,
If the raw material is not certified, a trial production for certification is requested, and information corresponding to the determination result of the authenticity of the raw material based on the measurement value obtained by the trial production is obtained;
If the raw materials are certified, the system performs manufacturing of the product using the certified raw materials. 7. The system of claim 6,
The processor,
The system is characterized in that, if the authenticity determination result based on the measurement values obtained by the trial production is false, information indicating that the raw material for which the certification was requested is fake is displayed. 7. The system of claim 6,
The processor,
A system characterized by updating the trained classification model by re-training the trained classification model using the measurement values obtained during the manufacture of the product using the authenticated raw materials, information indicating that the materials are genuine, and the trained classification model corresponding to the model number of the authenticated raw materials. 2. The system of claim 1,
The system is characterized in that the raw materials are recycled materials. 2. The system of claim 1,
The raw material is a plastic material,
the manufacturing process for producing said product is an injection molding process;
The system, wherein the measurements include measurements obtained from a plurality of sensors mounted within at least a mold through which the plastic material flows. A program storage medium storing a program executed in a system having one or more processors and one or more storage media,
The processor,
generating a trained classification model that learns measurement values obtained from a plurality of sensors installed in a manufacturing facility for a product using raw materials and outputs information used to determine the authenticity of the raw materials, and storing the trained classification model in the storage medium;
inputting the measurements of uncertified raw materials into the trained classification model;
A program storage medium storing a program for executing a process of authenticating the uncertified raw material when the uncertified raw material is determined to be genuine.

Images