Content ratio measurement device, content ratio measurement method, and content ratio measurement system

Abstract

The present invention measures the content ratios of constituent materials in a product in a non-destructive manner.　This content ratio measurement device measures the content ratios of constituent materials constituting a product, the content ratio measurement device comprising: a content ratio estimation unit that estimates the content ratios of constituent materials constituting a target product for which the content ratios are to be measured, by inputting—to an estimation model that has been trained using fluorescent fingerprint data, measured with a fluorometer with respect to a training product for which the true values of the content ratios are known, and the true values of the content ratios of the training product, the fluorescent fingerprint data consisting of fluorescent intensities corresponding to combinations of excitation wavelengths and fluorescent wavelengths—the fluorescent fingerprint data measured with the fluorometer with respect to the target product; and a pass / fail determination unit that determines whether or not the estimated content ratios of the constituent materials constituting the target product satisfy design standards of the target product.

Description

Content Ratio Measuring Device, Content Ratio Measuring Method, and Content Ratio Measuring System

[0001] The present invention relates to a content ratio measuring device, a content ratio measuring method, and a content ratio measuring system. The present invention claims the priority of Japanese Patent Application No. 2025-002728 filed on January 8, 2025, and for designated countries where incorporation by reference is permitted, the contents described in that application are incorporated herein by reference.

[0002] In the transition to a recycling-based society, there is a movement to promote the use of recycled materials in plastic products and synthetic resin products. Therefore, in order to prove that each manufacturer is using recycled materials in an environmentally conscious manner and is not involved in greenwashing, etc., the development of technologies that can nondestructively measure the presence or absence of recycled materials in products and their content ratios is underway.

[0003] As technologies for nondestructively measuring the content ratio of a predetermined material in a product, for example, a method using near-infrared spectroscopy and a method using fluorescence fingerprint spectroscopy are known. However, in the method using near-infrared spectroscopy, when the target product is black, the irradiated near-infrared light is absorbed, so there are difficulties in measurement accuracy. On the other hand, in the method using fluorescence fingerprint spectroscopy, measurement can be performed even when the target product is black.

[0004] Regarding the method using fluorescence fingerprint spectroscopy, for example, Patent Document 1 describes "a method for determining the content ratio of oil A and / or oil B in a determination target oil composition by comparing fluorescence fingerprints obtained by exciting two or more selected from oil A, oil B, and an oil composition containing oil A and oil B using a spectrofluorometer at all or part of excitation wavelengths of 250 to 700 nm, measuring all or part of fluorescence wavelengths of 250 to 800 nm, selecting one or more peaks characteristic of oil A and / or oil B, and measuring the fluorescence intensity of the determination target oil composition at the wavelengths (excitation wavelength, fluorescence wavelength) of one or more of the peaks using a spectrofluorometer."

[0005] Japanese Unexamined Patent Application Publication No. 2018-136151

[0006] The technology described in Patent Document 1 can determine the content ratio of different oils A and B in an oil composition. However, it cannot measure the content of recycled materials, etc., in resin-molded products.

[0007] This invention has been made in view of the above points, and aims to enable non-destructive measurement of the content of constituent materials in the product.

[0008] This application includes several means to solve at least some of the above problems, and some examples are as follows.

[0009] To solve the above problems, a content measurement device according to one aspect of the present invention is a content measurement device for measuring the content of constituent materials of a product, comprising: a content estimation unit that estimates the content of the constituent materials of a target product by inputting the fluorescence fingerprint data, measured using a fluorescence photometer on a target product whose content is to be measured, to an estimation model learned using fluorescence fingerprint data, which consists of fluorescence intensity corresponding to a combination of excitation wavelength and fluorescence wavelength, measured using a fluorescence photometer on a learning product for which the true value of the content is known; and a pass / fail determination unit that determines whether the estimated content of the constituent materials of the target product satisfies the design specifications of the target product.

[0010] According to the present invention, it is possible to non-destructively measure the content of constituent materials in the product.

[0011] Other issues, configurations, and effects not mentioned above will be clarified by the following description of the embodiments.

[0012] Figure 1 shows an example of the relationship between the recycled material content and fluorescence intensity in a product. Figure 2 shows an example of the configuration of a content measurement system according to one embodiment of the present invention. Figure 3 shows an example of a design information TBL (table). Figure 4 shows an example of a training dataset. Figure 5 is a flowchart for explaining an example of regression model learning processing. Figure 6 is a flowchart for explaining an example of preprocessing for fluorescence fingerprint data. Figure 7 is a diagram for explaining the removal of unnecessary regions. Figure 8 is a diagram for explaining the one-dimensionalization of data. Figure 9 is a diagram for explaining the effect of feature extraction processing. Figure 10 is a diagram for explaining the effect of logarithmic transformation on fluorescence intensity features. Figure 11 is a flowchart for explaining an example of recycled material content measurement processing for mass-produced products using the content measurement system. Figure 12 shows an example of the UI screen display. Figure 13 shows an example of the UI screen display. Figure 14 shows an example of the UI screen display. Figure 15 shows an example of the UI screen display.

[0013] One embodiment of the present invention will be described below with reference to the drawings. The embodiments are illustrative examples for explaining the present invention, and have been omitted and simplified as appropriate for clarity of explanation. The present invention can be implemented in various other forms. Unless otherwise specified, each component may be singular or plural. The position, size, shape, and range of each component shown in the drawings may not represent the actual position, size, shape, and range in order to facilitate understanding of the invention. In all drawings used to explain the embodiments, the same reference numerals are used for identical members as a general rule, and repeated explanations are omitted. In addition, in the following embodiments, the components (including element steps, etc.) are not necessarily essential unless specifically stated or considered to be clearly essential in principle. Also, when referring to "consisting of A," "having A," or "including A," other elements are not excluded unless specifically stated to refer only to that element. Similarly, in the following embodiments, when referring to the shape, positional relationship, etc. of components, etc., it includes those that are substantially similar or similar to that shape, etc., unless specifically stated or considered to be clearly not so in principle. Furthermore, "acquisition" shall, as a concrete example, include at least generating, calculating, or receiving from an external source by the subject.

[0014] <Outline of the Content Measurement System According to One Embodiment of the Precedent> First, an overview of the content measurement system according to one embodiment of the present invention will be described. This content measurement system performs fluorescence fingerprint analysis on a product manufactured by injection molding using virgin material and recycled material as constituent materials, and measures the content of recycled material constituting the product without destroying the product. This product corresponds to the product of the present invention.

[0015] Here, fluorescence fingerprint analysis refers to the process of using a fluorophotometer to irradiate a product with excitation light while changing the wavelength (excitation wavelength), measuring the wavelength and intensity of fluorescence emitted from the product in response to the excitation light, and obtaining and analyzing fluorescence fingerprint data consisting of the combination of excitation wavelength and fluorescence wavelength, and fluorescence intensity. First, fluorescence fingerprint analysis is used to obtain fluorescence fingerprint data corresponding to products for which the true value of recycled material content is known (hereinafter referred to as "training products"), and a regression model is trained using machine learning with the true value of recycled material content and the fluorescence fingerprint data. Next, fluorescence fingerprint data corresponding to target products for which the recycled material content is to be measured is obtained, and the recycled material content in the target product is estimated from the fluorescence fingerprint data using the regression model. The regression model corresponds to the estimation model of the present invention.

[0016] The principle behind fluorescence generation in response to excitation light irradiation of a product is that electrons in the fluorescent substance within the additive (antioxidant) added to the virgin material transition from the ground state to the excited state upon excitation light, and release energy as fluorescence when they return from the excited state to the ground state. Examples of fluorescent substances include organic molecules such as aromatic rings and π conjugate electrons.

[0017] Therefore, if the recycled material content in a product changes, the virgin material content also changes, which in turn changes the concentration of fluorescent substances in the virgin material, and thus the fluorescent fingerprint data changes. By analyzing this fluorescent fingerprint data, the recycled material content can be estimated.

[0018] Figure 1 shows the relationship between fluorescence wavelength and intensity when products with different recycled material content are irradiated with a predetermined excitation wavelength (250 nm). In this figure, the horizontal axis represents fluorescence wavelength, and the vertical axis represents fluorescence intensity. In this figure, it can be seen that the peak fluorescence intensity occurs around a fluorescence wavelength of 300 nm. It can also be seen that the fluorescence intensity is strongest when the recycled material content is 0% (when the virgin material content is 100%), and weakest when the recycled material content is 100% (when the virgin material content is 0%).

[0019] Furthermore, the relationship between the concentration of a fluorescent substance and its fluorescence intensity is known to be given by the following equation (1) based on the Lambert-Beer law: I = I ０ × (1-e －εｃＬ ) ... (1) Here, I is fluorescence intensity, I ０ ε is the excitation light intensity, ε is the molar extinction coefficient, c is the fluorescence concentration, and L is the optical path length (the distance the excitation light travels through the product).

[0020] From equation (1), it can be seen that the relationship between the fluorescent substance concentration c and the fluorescence intensity I is nonlinear. If a regression model is trained using data that remains nonlinear, the number of data points will increase, leading to overfitting and a decrease in the accuracy of the regression model's estimation of the recycled material content. From equation (1), it can be seen that if the fluorescence intensity I is logarithmically transformed, the relationship with the fluorescent substance concentration c will approach linearity. Therefore, in this embodiment, the logarithmically transformed fluorescence intensity I is used to train the regression model.

[0021] <Example of the configuration of the content measurement system 100 according to one embodiment of the present invention> Figure 2 shows an example of the configuration of the content measurement system 100 according to one embodiment of the present invention. The content measurement system 100 comprises a production management device 10, production equipment 20, and a content measurement device 30.

[0022] The production management device 10 is for controlling the production equipment 20. The production management device 10 consists of a general-purpose computer, such as a personal computer or a server computer. This computer includes a processor such as a CPU (Central Processing Unit), memory such as DRAM (Dynamic Random Access Memory), storage such as an HDD (Hard Disk Drive) or SSD (Solid State Drive), input devices such as a keyboard, mouse, and media drive, output devices such as a display, and communication modules such as an Ethernet® card or Wi-Fi® adapter.

[0023] The production management device 10 may be implemented using one physical or logical computer, or using two or more physical or logical computers. The two or more physical or logical computers may be distributed and located on a network N.

[0024] The production management device 10 includes a processing unit 11, a storage unit 12, and a communication unit 13.

[0025] The processing unit 11 is implemented by the processor of the computer that makes up the production management device 10. The processing unit 11 controls the entire production management device 10. The processing unit 11 (processor) implements the production instruction unit 111 as a functional block by executing a program (not shown) stored in the memory unit 12.

[0026] The production instruction unit 111 refers to the design information TBL (table) 121 (described later) in the storage unit 12 and instructs the production equipment 20 to manufacture the product by specifying the model numbers and proportions (contents) of virgin material and recycled material, respectively, as materials for the product.

[0027] The memory unit 12 is implemented by the memory and storage of the computer that constitutes the production management device 10. The memory unit 12 stores design information TBL 121 and production performance information 122.

[0028] Figure 3 shows an example of design information TBL121. Design information TBL121 contains information representing the design standard, and records the model numbers, percentages [%], and tolerance ranges [%] of the virgin and recycled materials used as materials for each product produced in the production facility 20, corresponding to the product ID. For example, for the production of product ID: A, it is recorded that 30% of virgin material with model number VA01 and 70% of recycled material with model number RE02 are used, and the tolerance range for the tolerance range is ±10%. In this case, if the percentage of virgin material is between 20% and 40%, it is within the tolerance range for the tolerance range, and if the percentage of recycled material is between 60% and 80%, it is within the tolerance range for the tolerance range.

[0029] Return to Figure 2. The production performance information 122 records information representing the production performance of products at the production equipment 20. Note that the storage unit 12 may store other information and data.

[0030] The communication unit 13 is implemented by a communication module of the computer that constitutes the production management device 10. The communication unit 13 connects to the content measurement device 30 via the network N and communicates various information and data. The network N is a two-way communication network such as the Internet.

[0031] The production equipment 20 includes an injection molding machine 21 and a fluorophotometer 22. The injection molding machine 21 manufactures products by injection molding using materials (virgin material and recycled material) instructed by the production instruction unit 111. The fluorophotometer 22 irradiates the products manufactured by the injection molding machine 21 (learning products and target products) with excitation light and measures the fluorescence generated from the products accordingly to acquire fluorescence fingerprint data. The fluorophotometer 22 outputs the generated fluorescence fingerprint data to the content measurement device 30 via the network N.

[0032] The content measurement device 30 trains a regression model using fluorescent fingerprint data corresponding to the training product. The content measurement device 30 also estimates the recycled material content in the target product using the regression model based on the fluorescent fingerprint data of the target product.

[0033] The content measurement device 30, like the production management device 10, consists of a general-purpose computer. The content measurement device 30 may be implemented using one physical or logical computer, or using two or more physical or logical computers. These two or more physical or logical computers may be distributed and located on a network N.

[0034] The content measurement device 30 comprises a processing unit 31, a storage unit 32, a communication unit 33, a display unit 34, and an input unit 35.

[0035] The processing unit 31 is implemented by the computer processor that makes up the content measurement device 30. The processing unit 31 controls the entire content measurement device 30. The processing unit (processor) 31 implements the following functional blocks by executing the program 321 stored in the memory unit 32: the UI (user interface) control unit 311, the preprocessing unit 312, the learning data extraction unit 313, the regression model learning unit 314, the material abnormality determination unit 315, the content estimation unit 316, and the pass / fail determination unit 317.

[0036] The UI control unit 311 displays the UI screen on the display unit 34. The UI control unit 311 also accepts user input using the UI screen and the input unit 35.

[0037] The preprocessing unit 312 acquires fluorescent fingerprint data corresponding to the learning product and the target product, respectively, from the fluorescence photometer 22 of the production equipment 20 via the communication unit 33 and the network N. The preprocessing unit 312 then performs preprocessing on the acquired fluorescent fingerprint data. Details of the preprocessing will be described later.

[0038] The training data extraction unit 313 extracts fluorescent fingerprint data to be used for training the regression model from pre-processed fluorescent fingerprint data corresponding to the training product, based on the user's selection, as training data.

[0039] The regression model learning unit 314 learns a regression model that takes the fluorescent fingerprint data of the target product as input and outputs the recycled material content of the target product, using machine learning with training data corresponding to the training product and the true value of the recycled material content in the training product.

[0040] The material anomaly determination unit 315 calculates the degree of deviation (e.g., k-neighbor distance) between the fluorescent fingerprint data corresponding to the learning product and the fluorescent fingerprint data corresponding to the target product, and determines that there is a material anomaly if the degree of deviation is greater than or equal to a predetermined threshold. Note that evaluation values ​​other than the k-neighbor distance may be used as the degree of deviation. Here, a material anomaly refers to a condition that does not require measuring the recycled material content, for example, when the materials used in the target product (virgin material and recycled material) are of a different type than the standard.

[0041] The content rate estimation unit 316 estimates the content rate of the recycled material in the target product by inputting the fluorescence fingerprint data corresponding to the target product into the regression model.

[0042] The pass / fail determination unit 317 refers to the design information TBL121 to obtain the ratio of the recycled material (content rate of the recycled material) of the target product and the allowable ratio of the standard width, and determines whether the content rate of the recycled material estimated for the target product satisfies the allowable ratio of the standard width of the target product.

[0043] The storage unit 32 is realized by the memory and storage of the computer constituting the content rate measuring device 30. In the storage unit 32, a program 321, a learning data set 322, preprocessing parameters 323, and a regression model 324 are stored. Note that other information and data may be stored in the storage unit 32.

[0044] The program 321 is a program for realizing each functional block in the processing unit (processor) 31. In the learning data set 322, fluorescence fingerprint data corresponding to the learning product used for machine learning of the regression model, such as fluorescence intensity for a combination of excitation wavelength and fluorescence wavelength, is recorded in advance.

[0045] FIG. 4 shows an example of the learning data set 322. In the learning data set 322, fluorescence fingerprint data composed of fluorescence intensity for a combination of excitation wavelength and fluorescence wavelength and the true value of the content rate of the recycled material in the learning product are recorded in advance in association with the product ID of the learning product. Note that in the learning data set 322, even if the product ID is the same, for example, fluorescence fingerprint data corresponding to a plurality of learning products with different production lots and the true value of the content rate of the recycled material are recorded, and they are not necessarily the same. Also, the true value of the content rate of the recycled material in the learning product adopts, for example, a value accurately measured by destructive measurement of the learning product.

[0046] Returning to FIG. 2. The preprocessing parameters 323 are information representing the type of preprocessing executed in the preprocessing unit 312 and its parameters, and are recorded by the preprocessing unit 312. The regression model 324 is learned and recorded by the regression model learning unit 314.

[0047] The communication unit 33 is implemented by a communication module of the computer that makes up the content measurement device 30. The communication unit 33 connects to the production management device 10 and the fluorescence photometer 22 of the production equipment 20 via the network N and communicates various information and data.

[0048] The display unit 34 is implemented by the output device of the computer that constitutes the content measurement device 30. The display unit 34 displays a UI screen. The input unit 35 is implemented by the input device of the computer that constitutes the content measurement device 30. The input unit 35 accepts various operation inputs from the user.

[0049] <About the regression model learning process using the content measurement device 30> Figure 5 is a flowchart illustrating an example of the regression model learning process using the content measurement device 30.

[0050] The regression model learning process is initiated, for example, when the user inputs various information on the regression model learning instruction screen 1000 (Figure 12) displayed on the display unit 34 in response to a predetermined operation from the user and operates the "Start Learning" button 1008. First, the UI control unit 311 of the content measurement device 30 accepts user input (specifically, input of a product ID) on the regression model learning instruction screen 1000 (Figure 12) to select the learning product to be trained on the regression model (Step S1).

[0051] Next, the preprocessing unit 312 obtains the fluorescent fingerprint data and the true value of the recycled material content corresponding to the learning product having the product ID entered by the user from the learning dataset 322 in the storage unit 32 (step S2). Here, there are multiple learning products with the same product ID, and even if they are the same product, they may include products with different manufacturing lots, resulting in different fluorescent fingerprint data or different true values ​​of recycled material content.

[0052] Next, the preprocessing unit 312 performs preprocessing on the fluorescent fingerprint data corresponding to the learning product acquired in step S2 (step S3).

[0053] Figure 6 is a flowchart illustrating an example of the preprocessing in step S3.

[0054] First, the preprocessing unit 312 removes unwanted regions from the fluorescence fingerprint data (step S11). Next, the preprocessing unit 312 converts the fluorescence fingerprint data into one-dimensional data for each excitation wavelength at predetermined intervals (for example, 50 nm intervals) (step S12).

[0055] Figure 7 is a diagram illustrating the process in step S11. The two-dimensional histogram shown on the left side of the figure is an Excitation Emission Matrix (EEM) in which the fluorescence fingerprint data is represented with fluorescence wavelength (EM) on the horizontal axis, excitation wavelength (EX) on the vertical axis, and fluorescence intensity at the pixel value.

[0056] In step S11, as shown from the left to the right side of the figure, data from the triangular A region located in the upper left of the EEM, which corresponds to the non-fluorescent component (excitation wavelength > fluorescent component), data from the band-shaped B region passing near the lower left and upper right vertices of the EEM, which corresponds to scattered light, and data from the triangular C region located in the lower right of the EEM, which also corresponds to scattered light, are removed from the fluorescence fingerprint data, which includes data corresponding to the entire EEM region. This reduces the amount of fluorescence fingerprint data, thereby reducing the amount of computation required in subsequent steps.

[0057] Figure 7B is a diagram illustrating the process in step S12. In step S12, the fluorescence fingerprint data from which unnecessary regions were removed in step S11 is converted into one-dimensional data in which fluorescence intensity is associated with fluorescence wavelength for each excitation wavelength at predetermined intervals (converted to one dimension).

[0058] Returning to Figure 6, the preprocessing unit 312 performs feature extraction processing on the one-dimensional fluorescence fingerprint data corresponding to each of the multiple training products having the same product ID, thereby removing noise components and reducing the variability of the one-dimensional fluorescence fingerprint data corresponding to each of the multiple training products (step S13).

[0059] The feature extraction process shall include at least one of the following: standardization, centering, smoothing, differentiation, baseline processing, SNV (standard normal validate) processing, and SG (Savitzky-Golay) processing. Multiple processes may be combined. Furthermore, the feature extraction process may be performed continuously on the fluorescence wavelength while scanning the excitation wavelength, or conversely, continuously on the excitation wavelength while scanning the fluorescence wavelength. Additionally, the process may be performed on the entire one-dimensional fluorescence fingerprint data at once, or it may be limited to a portion of the one-dimensional fluorescence fingerprint data.

[0060] The user can select the type of processing (such as standardization) to be performed as feature extraction processing and the parameters for each processing, but these may also be fixed in advance. In step S13, the preprocessing unit 312 stores the type of processing (such as standardization) and parameters of the processing performed as feature extraction processing as preprocessing parameters 323 in the storage unit 32.

[0061] Figure 9 illustrates the effect of the feature extraction process. The left side of the figure shows the one-dimensional fluorescence fingerprint data before the feature extraction process, with the horizontal axis representing fluorescence wavelength and the vertical axis representing fluorescence intensity. Each of the curves corresponds to multiple training products with the same product ID. The right side of the figure shows the one-dimensional fluorescence fingerprint data after the feature extraction process, with the horizontal axis representing fluorescence wavelength and the vertical axis representing fluorescence intensity features. From this figure, it can be seen that the feature extraction process can converge the variability in fluorescence fingerprint data that may occur among multiple training products with the same product ID. However, as can be seen from the values ​​on the vertical axis on the right side of the figure, the fluorescence intensity features extracted by the feature extraction process can be negative.

[0062] Returning to Figure 6, the preprocessing unit 312 then performs a logarithmic transformation on the fluorescence intensity features extracted by the feature extraction process so that the relationship between the fluorescence intensity features and the concentration of the fluorescent substance approaches linearity (step S14). However, as mentioned above, the fluorescence intensity features can be negative values, so a normal logarithmic transformation cannot be applied. Therefore, in step S14, a logarithmic transformation that can handle negative values ​​is adopted. Specifically, the Yeo-Johnson transformation, Box-Cox transformation, logarithmic symmetric transformation, or bilogarithmic transformation can be adopted. This concludes the explanation of the preprocessing of the fluorescence fingerprint data of the learning product in step S3.

[0063] Returning to Figure 5, the UI control unit 311 receives the user's selection of training data to be used for training the regression model (specifically, a combination of excitation wavelength and fluorescence wavelength) on the regression model training instruction screen 1000, and the training data extraction unit 313 extracts the data corresponding to the user's selection from the pre-processed fluorescence fingerprint data as training data (combination of excitation wavelength and fluorescence wavelength, fluorescence intensity feature quantities) (step S4).

[0064] The execution order of the training data extraction process in step S4 and the logarithmic transformation process in step S14 described above may be reversed. In that case, the logarithmic transformation process only needs to be performed on the training data, thus reducing the amount of computation required for the logarithmic transformation process.

[0065] Next, the regression model learning unit 314 uses the training data extracted in step S4 and the true value of the recycled material content obtained in step S2 to train a regression model using an arbitrary machine learning method (step S5), and saves it as the regression model 324 in the storage unit 32 (step S6). This concludes the explanation of the regression model learning process.

[0066] This regression model learning process reduces subsequent computation by preprocessing the fingerprint fluorescence data to remove data corresponding to unnecessary regions in the EEM. Furthermore, since the fluorescence intensity is logarithmically transformed, the estimation accuracy of the regression model can be improved compared to cases where logarithmic transformation is not performed.

[0067] Figure 10 illustrates the difference between the estimated recycled material content using a regression model generated with logarithmically transformed fluorescence intensity features and the estimated recycled material content using a regression model generated with untransformed fluorescence intensity features. In this figure, the horizontal axis represents the true recycled material content, and the vertical axis represents the estimated recycled material content. The circular dots represent the estimated recycled material content when using a regression model generated with untransformed fluorescence intensity features, while the diamond-shaped dots represent the estimated recycled material content when using a regression model generated with logarithmically transformed fluorescence intensity features.

[0068] In the figure, if the horizontal axis is X and the vertical axis is Y, then the closer the estimated value is to the line Y = X, the higher the estimation accuracy of the regression model. From the figure, it can be seen that the diamond-shaped point group corresponding to the case where logarithmic transformation is performed is closer to the line Y = X than the circular point group corresponding to the case where logarithmic transformation is not performed. Therefore, it can be seen that the regression model generated using the fluorescence intensity feature after logarithmic transformation has higher estimation accuracy for the recycled material mixing ratio.

[0069] <Regarding the process of measuring the recycled material content of mass-produced products using the content measurement system 100> Figure 11 is a flowchart illustrating an example of the process of measuring the recycled material content of mass-produced products using the content measurement system 100.

[0070] The process of measuring the recycled material content is initiated, for example, in response to a predetermined operation by the user. First, the UI control unit 311 of the content measuring device 30 displays a UI screen (not shown) on the display unit 34, and accepts the user's operation to select the target product on the UI screen (an operation to enter the product ID of a product in mass production) (step S21).

[0071] Next, the pass / fail determination unit 317 refers to the design information TBL121 stored in the storage unit 12 of the production management device 10 via the communication unit 33 and the network N to obtain the recycled material content and the allowable width ratio specified for the target product selected in step S21 (step S22).

[0072] Next, the preprocessing unit 312 requests the production instruction unit 111 of the production management device 10, via the communication unit 33 and the network N, to extract a sample of the product currently being mass-produced in the production equipment 20 (the target product having the product ID selected in step S21) and to measure the fluorescence fingerprint data of the sample using the fluorescence photometer 22. Then, the preprocessing unit 312 acquires the fluorescence fingerprint data of the target product sample measured by the fluorescence photometer 22 (step S23).

[0073] Next, the preprocessing unit 312 retrieves from the storage unit 32 the preprocessing parameters 323 (saved in step S13 in Figure 6) related to the preprocessing performed on the fluorescent fingerprint data of the learning product during the learning of the regression model (step S24).

[0074] Next, the preprocessing unit 312 performs the same preprocessing on the fluorescent fingerprint data of the target product sample as the preprocessing performed on the fluorescent fingerprint data of the learning product (Figure 6), according to the preprocessing parameters 323 acquired in step S23 (step S25).

[0075] Next, the material anomaly determination unit 315 calculates the degree of discrepancy (e.g., k-neighbor distance) between the pre-processed fluorescent fingerprint data corresponding to the training product used to train the regression model and the pre-processed fluorescent fingerprint data corresponding to the sample of the target product, and compares the degree of discrepancy with a predetermined threshold to determine whether or not there is a material anomaly (step S26).

[0076] If the degree of deviation is greater than or equal to a predetermined threshold and it is determined that there is a material abnormality (YES in step S26), then the UI control unit 311 displays an alert on the measurement result screen 1100 (Figure 13), which is a UI screen, indicating that a material abnormality has occurred (step S30).

[0077] Conversely, if in step S26 the degree of deviation is smaller than a predetermined threshold and it is determined that there is no material abnormality (NO in step S26), then the content estimation unit 316 reads the regression model 324 from the storage unit 32 and inputs the pre-processed fluorescent fingerprint data corresponding to the target product into the regression model 324 to estimate the recycled material content in the target product (step S27).

[0078] Next, the pass / fail determination unit 317 determines whether the estimated recycled material content for the target product meets the standard width tolerance ratio, based on the recycled material content and standard width tolerance ratio of the target product obtained in step S22 (step S28). If it is determined that the estimated recycled material content for the target product meets the standard width tolerance ratio (YES in step S28), the UI control unit 311 then displays on the measurement result screen 1100 (Figure 14) as a UI screen that the recycled material content of the target product meets the standard (step S29).

[0079] Conversely, if it is determined that the estimated recycled material content for the target product does not meet the standard width tolerance (NO in step S28), the UI control unit 311 then displays an alert on the measurement result screen 1100 (Figure 15) as a UI screen indicating that the recycled material content is abnormal (step S30). This concludes the explanation of the recycled material content measurement process.

[0080] The recycled material content measurement process allows for the measurement of recycled material content without damaging products in mass production. The system can notify users whether the measured recycled material content meets the required standards. Furthermore, if any material defects are found in the product, a warning can be issued.

[0081] <Example of UI screen display> Next, we will explain an example of the UI screen display.

[0082] Figure 12 shows an example of the display of the regression model learning instruction screen 1000 as a UI screen displayed on the display unit 34 of the content measurement device 30.

[0083] The regression model training instruction screen 1000 includes an input field 1001 for entering the product ID of the training product to be trained on the regression model using fluorescent fingerprint data, a selection field 1002 for selecting the type of preprocessing, an input field 1003 for entering the parameters of the selected preprocessing, a selection field 1004 for selecting whether or not to perform logarithmic transformation, and a selection field 1005 for selecting training data. The regression model training instruction screen 1000 also includes a display area 1006 for displaying the selected training data, a selection field 1007 for selecting the method to use when training the regression model, and a "Start Training" button 1008 for instructing the start of training the regression model.

[0084] On the regression model learning instruction screen 1000 displayed on the display unit 34, the user can select the product to be trained on the regression model, set the details of the preprocessing, select the training data, and select the machine learning method for the regression model.

[0085] Figures 13 to 15 show examples of the measurement result screen 1100, which is a UI screen displayed on the display unit 34 of the content measurement device 30, and show the measurement results for four samples (molding numbers 1 to 4) of the target product.

[0086] The measurement results screen 1100 is provided with display areas 1101 to 1104. The display areas show the design information of the target product (the model numbers, proportions, and tolerance ranges of the virgin and recycled materials used in the product). Display area 1102 shows the degree of deviation between the pre-treated fluorescent fingerprint data corresponding to the training product used to train the regression model and the pre-treated fluorescent fingerprint data corresponding to the sample of the target product. Display area 1103 shows the estimated recycled material content. Display area 1104 shows the status of the target product during production.

[0087] Here, the state of the product in production refers to one of the following states: a material defect has occurred, the recycled material content is within specifications, or the recycled material content is outside specifications. Figure 13 shows an example of a display corresponding to a material defect, Figure 14 shows an example of a display corresponding to a recycled material content within specifications, and Figure 15 shows an example of a display corresponding to a recycled material content outside specifications.

[0088] The user can understand the status of the target product during production by viewing the measurement results screen 1100 displayed on the display unit 34.

[0089] The present invention is not limited to the embodiments described above, and various modifications are possible. For example, the embodiments described above are described in detail to make the present invention easier to understand, and are not necessarily limited to those having all the configurations described. Furthermore, it is possible to replace or add to the configurations of one embodiment with those of another embodiment.

[0090] 10...Production management device, 11...Processing unit, 111...Production instruction unit, 12...Storage unit, 121...Design information TBL, 122...Production performance information, 13...Communication unit, 21...Injection molding machine, 22...Fluorescence photometer, 30...Content measurement device, 31...Processing unit, 311...UI control unit, 312...Preprocessing unit, 313...Learning data extraction unit, 314...Regression model learning unit, 315...Material abnormality determination unit, 316...Content estimation unit, 317...Pass / fail determination unit, 32...Storage unit, 321...Program, 322...Learning dataset, 323...Preprocessing parameters, 324...Regression model, 33...Communication unit, 34...Display unit, 35...Input unit, 100...Content measurement system, 1000...Regression model learning instruction screen, 1100...Measurement result screen

Claims

1. A content measurement device for measuring the content of constituent materials constituting a product, comprising: a content estimation unit that estimates the content of constituent materials constituting a target product by inputting the fluorescence fingerprint data measured using the fluorescence photometer for a target product whose content is to be measured to an estimation model learned using fluorescence fingerprint data consisting of fluorescence intensity corresponding to a combination of excitation wavelength and fluorescence wavelength measured using a fluorescence photometer for a learning product whose true value of content is known; and fluorescence fingerprint data measured using the fluorescence photometer for a target product whose content is to be measured; and a pass / fail determination unit that determines whether the estimated content of constituent materials constituting the target product satisfies the design specifications of the target product.

2. A content measurement device according to claim 1, comprising a model learning unit that learns the estimation model using the fluorescence fingerprint data measured with respect to the learning product using the fluorescence photometer and the true value of the content of the learning product.

3. A content measurement device according to claim 2, comprising: a fluorescent fingerprint data corresponding to the training product used for training the estimation model, and a preprocessing unit that performs preprocessing on the fluorescent fingerprint data corresponding to the target product.

4. A content measurement device according to claim 3, wherein the preprocessing unit performs at least one of the following processes: a process to remove unwanted regions from the fluorescence fingerprint data by removing non-fluorescent components and data corresponding to scattered light; a process to convert the fluorescence fingerprint data into a one-dimensional object for each excitation wavelength; a feature extraction process to extract feature quantities from the fluorescence fingerprint data; and a logarithmic transformation process to logarithmically transform the feature quantities of the fluorescence intensity of the fluorescence fingerprint data.

5. A content measurement device according to claim 4, wherein the feature extraction process includes at least one of the following processes: standardization process, centering process, smoothing process, differential process, baseline process, SNV (standard normal validate) process, and SG (Savitzky-Golay) process.

6. A content measurement device according to claim 4, wherein the logarithmic transformation process is one of Yeo-Johnson transformation, Box-Cox transformation, logarithmic transformation, or double logarithmic transformation.

7. A content measurement device according to claim 1, comprising a UI control unit that displays the determination result by the pass / fail determination unit on a UI screen.

8. A content measurement device according to claim 7, comprising: a material abnormality determination unit that calculates the degree of deviation between the fluorescent fingerprint data corresponding to the training product used to train the estimation model and the fluorescent fingerprint data corresponding to the target product, and determines a material abnormality in the target product based on the degree of deviation, wherein the UI control unit displays the determination result by the material abnormality determination unit on the UI screen.

9. A content measuring device according to claim 1, wherein the product is formed by an injection molding machine using virgin material and recycled material as materials, and the content estimating unit is a content measuring device that estimates the content of the recycled material constituting the target product.

10. A method for measuring the content of constituent materials constituting a product using a content measuring device, comprising: a content estimation step of estimating the content of constituent materials constituting a target product by inputting the fluorescence fingerprint data, measured using a fluorescence photometer for a target product whose content is to be measured, to an estimation model learned using fluorescence fingerprint data consisting of fluorescence intensity corresponding to a combination of excitation wavelength and fluorescence wavelength, measured using a fluorescence photometer for a learning product whose true value of content is known; and a pass / fail determination step of determining whether the estimated content of constituent materials constituting the target product satisfies the design specifications of the target product.

11. A content measurement system comprising: production equipment including an injection molding machine and a fluorophotometer; a production management device for controlling the production equipment; and a content measurement device for measuring the content of constituent materials constituting a product produced by the injection molding machine, wherein the content measurement device comprises: a content estimation unit that estimates the content of constituent materials constituting a target product by inputting the fluorescence fingerprint data measured using the fluorophotometer for a target product whose content is to be measured to an estimation model learned using fluorescence fingerprint data consisting of fluorescence intensity corresponding to a combination of excitation wavelength and fluorescence wavelength measured using a fluorophotometer for a learning product whose true value of content is known, and the true value of the content of the learning product; and a pass / fail determination unit that determines whether the estimated content of the constituent materials constituting the target product satisfies the design specifications of the target product.

Images