Method for estimating chemical composition of scrap materials

Abstract

To provide a composition estimation method capable of efficiently manufacturing a metal product having a desired chemical composition using a scrap material as a raw material, a scrap material management method, a composition estimation program, a composition estimation device, a fusion furnace facility, a scrapyard, and a trained model generation method.SOLUTION: A composition estimation method for estimating a chemical composition of a scrap material 16, comprises: a pre-processing step S2 of creating an image group consisting of all or some of a plurality of divided images 18b obtained by dividing an image 18 obtained by photographing a predetermined scrap material 16 into predetermined image sizes; and a chemical composition output step S4 of using the divided image 18b as an input image for each divided image 18b included in the image group, and outputting a chemical composition corresponding to the predetermined scrap material 16 in the input image using a trained model 19.SELECTED DRAWING: Figure 7

Description

[Technical Field]

[0001] The present invention relates to a composition estimation method for estimating the chemical composition of scrap material. [Background technology]

[0002] There is known a technique for obtaining desired metal products by supplying various raw materials, including scrap material, to a melting furnace and melting them (see, for example, Patent Document 1). There is also known a technique for estimating the grade of scrap material from an image of the scrap material (see, for example, Patent Document 2). [Prior art documents] [Patent documents]

[0003] [Patent Document 1] Japanese Patent Publication No. 2020-185573 [Patent Document 2] Patent No. 7036296 Summary of the Invention [Problem to be solved by the invention]

[0004] If the chemical composition of scrap material can be estimated before it is fed into a melting furnace, metal products with the desired chemical composition can be efficiently produced.

[0005] Therefore, an object of the present invention is to provide a component estimation method, a scrap material management method, a component estimation program, a component estimation device, melting furnace equipment, a scrap storage area, and a method for generating a trained model that can enable the efficient production of metal products having desired chemical components using scrap material as a raw material. [Means for solving the problem]

[0006] One aspect of the present invention is as follows.

[0007] [1] 1. A composition estimation method for estimating chemical composition of scrap material, comprising: a pre-processing step of creating an image group consisting of all or part of a plurality of divided images obtained by dividing an image of a predetermined scrap material into predetermined image sizes; a chemical component output step of outputting, for each of the divided images included in the image group, chemical components corresponding to the predetermined scrap material in the input image using the divided image as an input image and a trained model; A component estimation method having

[0008] [2] The component estimation method according to [1], wherein the predetermined image size has all sides of 2000 mm or less in real scale.

[0009] [3] The chemical component output step includes a product category information output step for outputting, for each of the divided images included in the image group, product category information corresponding to the specified scrap material in the input image using a trained model, the product category information obtained via the product category information output step as input, and outputting the chemical components corresponding to the input using a database. The component estimation method described in [1] or [2].

[0010] [4] The component estimation method according to [3], wherein in the product category information output step, one product category is output as the product category information for the input image.

[0011] [5] The component estimation method according to any one of [1] to [4], further comprising a post-processing step of estimating the chemical components of the original image using the chemical components output for each of the input images by the chemical component output step.

[0012] [6] For the scrap material to be managed, the chemical composition is estimated by the composition estimation method described in any one of [1] to [5], A scrap material management method, wherein if the chemical components estimated by the component estimation method do not satisfy predetermined conditions, a portion of the scrap material to be managed is removed based on the estimated chemical components, or another scrap material is added to the scrap material to be managed.

[0013] [7] A melting furnace operation method comprising a scrap material supply step of supplying part or all of scrap material, the chemical composition of which has been estimated in advance by the component estimation method according to any one of [1] to [5], to a melting furnace based on the chemical composition estimated by the component estimation method.

[0014] [8] A melting furnace operating method comprising a scrap material supply step of supplying to the melting furnace part or all of the scrap material to be managed by the scrap material management method described in [6].

[0015] [9] A component estimation program that causes a computer to execute the component estimation method according to any one of [1] to [5].

[0016]

[10] A component estimation device for estimating chemical components of scrap material, comprising: a pre-processing unit that creates an image group consisting of all or part of a plurality of divided images obtained by dividing an image of a predetermined scrap material into predetermined image sizes; a chemical component output unit that uses the divided image included in the image group as an input image and outputs chemical components corresponding to the predetermined scrap material in the input image using a trained model; A component estimation device having the following.

[0017]

[11]

[10] A melting furnace facility having the component estimation device according to the present invention and a melting furnace.

[0018]

[12]

[10] A scrapyard having a component estimation device according to the present invention.

[0019]

[13] A method for generating a trained model that takes an image of a scrap material as input and outputs product category information of the scrap material, A method for generating a trained model, which generates the trained model by learning using training data including a combination of a product category and a training image containing scrap material that falls into the product category, and extended training data generated by resizing the training image by enlarging or reducing it, moving it, rotating it, partially cropping or flipping it, or any combination of these.

[0020]

[14] A trained model generated by the trained model generation method described in

[13] . [Effects of the Invention]

[0021] According to the present invention, it is possible to provide a component estimation method, a scrap material management method, a component estimation program, a component estimation device, a melting furnace facility, a scrap storage area, and a method for generating a trained model, which can enable the efficient production of metal products having desired chemical components using scrap material as a raw material. [Brief explanation of the drawings]

[0022] [Figure 1] 1 is a block diagram showing a component estimation device according to an embodiment of the present invention; [Figure 2] FIG. 2 is a schematic diagram showing a trained model used in the component estimation device shown in FIG. 1. [Figure 3] FIG. 2 is a schematic diagram showing an example of an image obtained by photographing a predetermined scrap material. [Figure 4] 4 is a schematic diagram showing an example of a background-removed image formed by a preprocessing unit from the image shown in FIG. 3. FIG. [Figure 5] FIG. 2 is a schematic diagram showing an example of a melting furnace facility having the component estimation device shown in FIG. 1. [Figure 6] FIG. 6 is a schematic diagram showing a state in which a scrap material is being photographed while being moved in the melting furnace facility shown in FIG. 5. [Figure 7]2 is a flow chart showing an example of a melting furnace operation method using the component estimation device shown in FIG. 1. [Figure 8] 2 is a flow chart showing a first specific example of a melting furnace operation method using the component estimation device shown in FIG. 1. [Figure 9] 1. FIG. 4 is a flowchart showing a second specific example of a melting furnace operation method using the component estimation device shown in FIG. [Figure 10] FIG. 2 is an explanatory diagram illustrating an example of a database in the chemical component output unit shown in FIG. [Figure 11] FIG. 3 is an explanatory diagram illustrating an example of training data for generating the trained model shown in FIG. 2. [Figure 12] FIG. 12 is an explanatory diagram illustrating an example of extended training data generated from the training data shown in FIG. DETAILED DESCRIPTION OF THE INVENTION

[0023] Hereinafter, an embodiment of the present invention will be described with reference to the drawings.

[0024] As shown in Fig. 1, in one embodiment of the present invention, a component estimation device 1 has a pre-processing unit 2, a chemical component output unit 4 (including a product category information output unit 3), a post-processing unit 5, a photographing device 6, and a notification unit 7. The pre-processing unit 2, the chemical component output unit 4, and the post-processing unit 5 are each configured as part of a processing unit 9 of a computer 8. The computer 8 is configured by a central processing unit 10, a main memory device 11, and an auxiliary memory device 12. The photographing device 6, an input device 13, and an output device 14 are communicatively connected to the computer 8 by wire or wirelessly. The notification unit 7 is configured by the output device 14.

[0025] The central processing unit 10 is configured, for example, by a CPU (Central Processing Unit). The central processing unit 10 may also be configured by an accelerator with multiple calculation cores, such as a GPU (Graphics Processing Unit), for high-speed calculations. The main storage device 11 is configured, for example, by memory. The auxiliary storage device 12 is configured, for example, by an SSD (Solid State Drive) or an HDD (Hard Disk Drive). The photographing device 6 is configured, for example, by a camera. The input device 13 is configured, for example, by a keyboard, mouse, or touch panel. The output device 14 is configured, for example, by a display or speaker.

[0026] The component estimation device 1 of this embodiment estimates the chemical components of scrap material 16 in order to efficiently operate a melting furnace 15 for producing iron-based metal products such as steel (see FIG. 5). Note that the component estimation device 1 of this embodiment can also be applied to the operation of a melting furnace 15 for producing metal products other than iron-based metal products. In this embodiment, an electric furnace having electrodes 17 for melting the scrap material 16 is used as the melting furnace 15, but a melting furnace 15 other than an electric furnace may also be used. The photographing device 6 photographs a predetermined scrap material 16 and generates an image 18 of the scrap material 16 (see FIG. 3). The auxiliary storage device 12 stores the image 18 generated by the photographing device 6, data obtained by arithmetic processing by the processing unit 9, and the like. The central processing unit 10 processes the component estimation program.

[0027] To reduce the cost of molten iron in electric furnaces, it is necessary to utilize various iron-based scrap materials16, including low-grade iron-based scrap materials16 (referred to as low-grade scrap materials) such as H2 and H3. Low-grade scrap materials are often collected from the market, such as automobiles, construction waste, and electrical appliances. The chemical composition of the iron contained in the low-grade scrap materials is unknown, and the chemical composition of the low-grade scrap materials is not consistent. Low-grade scrap materials may also contain metals other than iron-based scrap materials16, such as electric motors and electrical wires, whose primary component is not iron (Fe). Copper (Cu), in particular, is difficult to remove using current steelmaking methods. If copper is present in the molten iron, it must be diluted to a specified level with other Cu-free iron-based scrap materials16 (referred to as high-grade scrap materials).

[0028] If the chemical composition of the molten iron-based scrap material 16 is determined only by measured values, a large amount of additional material must be added if the chemical composition significantly deviates from the target chemical composition, resulting in increased costs for the additional material. Furthermore, if the molten iron-based scrap material contains a large amount of undesirable chemical components, such as Cu, it must be diluted with high-grade scrap material, which requires the addition of a large amount of high-grade scrap material. Because melting a large amount of high-grade scrap material takes a long time, the time required to produce molten metal with the desired chemical composition increases, significantly reducing the operational efficiency of the melting furnace 15. To reduce the cost of additional material and improve the operational efficiency of the melting furnace 15, it is preferable to estimate the chemical composition of the scrap material 16 to be supplied to the melting furnace 15 in advance and adjust the supply of the material based on the estimated value.

[0029] For this reason, the component estimation program of this embodiment causes the computer 8 to execute the photographing step S1, the pre-processing step S2, the chemical component output step S4 (including the product category information output step S3), the post-processing step S5, and the notification step S6 (see FIG. 7 ). Therefore, in the component estimation method of this embodiment, the photographing step S1, the pre-processing step S2, the chemical component output step S4 (including the product category information output step S3), the post-processing step S5, and the notification step S6 are executed in this order. Furthermore, in the melting furnace operation method of this embodiment, the photographing step S1, the pre-processing step S2, the chemical component output step S4 (including the product category information output step S3), the post-processing step S5, the notification step S6, the scrap material inspection step S7, and the scrap material supply step S8 are executed in this order (see FIG. 7 ). The component estimation program of this embodiment can be modified to any configuration that causes the computer 8 to execute the pre-processing step S2 and the chemical component output step S4 in this order. Accordingly, the component estimation method and the melting furnace operation method can be modified accordingly.

[0030] In the photographing step S1, the processing unit 9 photographs a predetermined scrap material 16 using the photographing device 6 based on settings or instructions via the input device 13, and generates an image 18 of the scrap material 16 (see FIG. 3). In the photographing step S1, the image 18 is taken with the photographing device 6, such as a still camera or video camera, so that the collection of scrap material 16 is captured.

[0031] The pre-processing step S2 includes an image division step. In the image division step, the pre-processing unit 2 divides an image 18 (referred to as an entire image 18a) obtained by photographing the predetermined scrap material 16 into a predetermined image size, generates a plurality of divided images 18 (referred to as divided images 18b), and creates an image group consisting of all or part of the plurality of divided images 18b. In this case, for example, it is preferable to divide the entire image 18a into a grid pattern. For example, it is more preferable to divide the divided image 18b so that each divided image 18b is a parallelogram (including a rectangle, a diamond, and a square). In this case, each divided image 18b has four sides. It is even more preferable to divide the divided image 18b so that each divided image 18b is a rectangle (including a square), as shown in FIG. 3.

[0032] Furthermore, it is preferable that the divided images 18b include as few types of scrap material 16 as possible. In other words, it is preferable that the divided images 18b be divided at a predetermined image size such that each divided image 18b includes a size that allows the product category of the scrap material 16 to be identified. Furthermore, it is more preferable that the divided images 18b be divided at a predetermined image size such that each divided image 18b includes a single type of scrap material 16. Therefore, since scrap material 16 is often between 600 mm and 2000 mm, it is preferable that the image size of each divided image 18b be set so that all sides of the divided image 18b are 2000 mm or less in actual scale. Although scrap material 16 has various shapes, most are larger than a few millimeters. Therefore, it is more preferable that the image size of each divided image 18b be set so that all sides of the divided image 18b are between 10 mm and 2000 mm. Furthermore, taking into consideration the installation distance of the camera, it is more preferable to set all sides of one divided image 18b to be 100 mm or more and 2000 mm or less. For example, if the entire image 18a is divided so that one divided image 18b is a parallelogram, this means that the sizes of all four sides of the divided image 18b satisfy the above conditions. The image division step may be performed only when the image 18 contains a certain amount or more of scrap material 16 (for example, a certain number or more).

[0033] The preprocessing step S2 may include a background separation step. In the background separation step, the preprocessing unit 2 separates the background, i.e., the portion other than the scrap material 16, from the image 18 to generate an image 18 from which the background has been at least partially removed (see FIG. 4). The background separation step may be performed only if the image 18 contains a certain amount of background. For example, image segmentation may be performed using an input image 18 containing the scrap material 16, in which pre-trained image segmentation is performed using annotation images that label pixels representing the background and pixels representing the scrap material 16 in the image 18. In the background separation step, if the hue of the image clearly differs from that of the scrap material 16, the image may be identified by color and removed. For example, if the sky is shown in the background, the scrap material 16 is often gray or brown, so the segmented image 18b, which contains a large amount of blue, may be removed as the background. Furthermore, because the scrap material 16 is located vertically below the image 18, the segmented images 18b at the top of the image 18 and with a different hue than the bottom of the image 18 may be removed as background. A machine learning model trained on the scrap material 16 and background may be used to identify the background, and segmented images 18b other than the scrap material 16 may be removed as background. It is preferable to flag the segmented images 18b to be removed so that they can be identified in future processing. The background separation step can be performed after the image segmentation step. In this case, the preprocessing unit 2 creates an image group consisting of a portion of the multiple segmented images 18b obtained by dividing the image 18 into predetermined image sizes, i.e., the segmented images 18b other than the background portion. The background separation step can also be performed before the image segmentation step. In this case, the preprocessing unit 2 creates an image group consisting of all or a portion of the multiple segmented images 18b obtained by dividing the image 18 after background separation into predetermined image sizes.

[0034] In the product category information output step S3, the product category information output unit 3 receives an image 18 obtained by photographing a predetermined scrap material 16 as an input, and outputs product category information 20 of the scrap material 16 using a trained model 19 (see FIG. 2 ). In the product category information output step S3 of this embodiment, for each divided image 18b included in the image group created by the pre-processing unit 2, the product category information output unit 3 receives the divided image 18b as an input image, and uses the trained model 19 to output product category information 20 corresponding to the predetermined scrap material 16 in the input image.

[0035] As shown in FIG. 2, the trained model 19 is a trained machine learning model that receives an image 18 of scrap material 16 as input and outputs product category information 20 for the scrap material 16. The trained model 19 is generated by machine learning using a combination of the image 18 of scrap material 16 and the product category information 20 corresponding to the scrap material 16 as training data. For example, the trained model 19 may be configured using a convolutional neural network or the like, and may be generated by prior training using training data that includes a combination of a product category and a training image containing scrap material 16 corresponding to the product category, as shown in FIG. 11 (i.e., an image 18 containing scrap material 16 belonging to a specific product category is used as input and the product category is used as output). Training to obtain the trained model 19 may use augmented training data generated by augmenting the training image by, for example, resizing (enlarging or reducing), moving, rotating, cropping, or flipping, or any combination thereof, as shown in FIG. 12. It is preferable to prepare training data at least 10 times the number of parameters of the machine learning system being used. However, preparing such a large amount of training data is often physically difficult. Therefore, it is preferable to use augmented training data, which is training data obtained by data augmentation of training data. Data augmentation involves randomly combining the enlargement, reduction, translation, rotation, partial cropping, and inversion of the training image, as shown in Figure 12. Using data augmentation techniques such as enlargement, reduction, translation, rotation, partial cropping, and inversion enables training using training images that are closer to the actual image of the scrap material 16. Multiple enlargement, reduction, translation, rotation, partial cropping, and inversion may be selected. Because scrap material 16 is photographed in various positions, augmentation can improve estimation accuracy. It is preferable that the machine learning model include multiple neural networks or one or more decision trees for multi-class classification. The teacher image may correspond to a product category, show scrap material 16 corresponding to that product category, and the number of pixels of the teacher image that show the corresponding product category may be greater than the number of pixels that show other scrap material 16.The training data used for training the machine learning model may include training data in which a background image, which is an image other than the scrap material 16, is associated with a category other than the scrap material.

[0036] Scrap materials 16 come in a variety of colors and shapes. Considering the combinations of overlapping scrap materials 16, the amount of training data required is nearly infinite, making it generally difficult to create a machine learning model that operates reliably. However, as in the present embodiment, by segmenting the image 18 and generating segmented images 18b that preferably capture a single type of scrap material 16, the training images of a single type of scrap material 16 are sufficient for the training data, significantly reducing the amount of training data required. Furthermore, scrap materials 16 may be randomly positioned and photographed from various angles and distances. Applying data augmentation enables efficient training of a machine learning model for estimating the product category of scrap materials 16 with a small amount of data. It is more effective to include not only scrap materials 16 but also images of background materials other than scrap materials 16 in the training data. In the preprocessing step S2, segmented images 18b that clearly do not contain scrap materials 16 may be flagged for removal. However, even in this case, segmented images 18b may contain images with a large amount of background. Therefore, by learning the background image, it becomes possible to distinguish the divided images 18b other than the scrap material 16 from the divided images 18b, thereby improving the accuracy of the estimated chemical components.

[0037] The product category information 20 indicates the product category of the original product of the scrap material 16 included in the image 18, which has various shapes such as steel beams, bars, and wires. The product categories can be obtained by classifying various products that can become the scrap material 16 into several categories in advance. For example, product categories may be classified into automobiles, building materials, electrical appliances, and the like. The product categories may be further subdivided. For example, steel frames and rebars may be used instead of building materials, and refrigerators and microwaves may be used instead of electrical appliances. If the image 18 contains multiple scrap materials 16, the product category information 20 corresponding to the image 18 may indicate multiple product categories. In this case, the product category information 20 may indicate, for example, 60% automobiles, 20% building materials, and 10% electrical appliances, as shown in FIG. 2 . In this case, the numerical values ​​indicating the proportions of each product category included in the product category information 20 (such as "60%" for automobiles) may indicate the likelihood (accuracy) that the image 18 corresponds to the product category. In this case, for example, the trained model 19 uses a softmax function to output the estimated probability of each product category for the segmented image 18b. The product category information output step S3 may output one product category as the product category information 20 for the input image. In other words, the product category information 20 may be information indicating only the specific product category to which the image 18 is most likely to correspond. In this case, for example, the product category information output step S3 estimates that the product category with the highest probability for the segmented image 18b is the product category of the segmented image 18b. The product category information 20 may indicate a category other than the product category. Such a category may include, for example, the background (parts of the image 18 other than the scrap material 16).

[0038] In this embodiment, in the chemical component output step S4, the chemical component output unit 4 receives, as input, the product category information 20 obtained via the product category information output unit 3, for each segmented image 18b included in the image group, and outputs chemical components corresponding to the input. In the chemical component output step S4 of this embodiment, the chemical component output unit 4 receives, as input, the product category information 20 obtained via the product category information output step S3, and outputs chemical components corresponding to the input, for example, using a database such as that shown in FIG. 10. More specifically, for each piece of product category information 20 output by the product category information output unit 3, the product category information 20 is received as input, and the chemical components corresponding to the input are output using the database. In other words, the chemical component output unit 4 outputs chemical components corresponding to each segmented image 18b, for each piece of product category information 20 corresponding to that segmented image 18b.

[0039] Scrap material 16 often contains metal materials with similar chemical compositions if they belong to the same product category. Therefore, a table (database) that identifies the chemical compositions for each product category is created in advance. For example, automobiles can be identified as having A% iron and B% copper, while building materials can be identified as having C% iron and D% copper. By comparing the table, the chemical compositions corresponding to each segmented image 18b can be output. In this case, for example, the chemical composition output step S4 may predict the chemical composition of image 18 by taking a weighted average of the chemical compositions in the chemical composition table associated with the product category, using the estimated probability of the product category as a weight. For example, if the product category information 20 of image 18 input to the chemical composition output unit 4 is 50% automobiles and 50% building materials, the chemical compositions output by the chemical composition output unit 4 will be (A+C) / 2% iron and (B+D) / 2% copper.

[0040] In post-processing step S5, the post-processing unit 5 averages the chemical components output by the chemical component output unit 4. The averaging may be performed using an arithmetic mean or a harmonic mean, or, for example, a weighted mean, in which weights are assigned to the segmented images 18b. The post-processing unit 5 estimates the chemical components of the original image 18 using the chemical components output for each input image in the chemical component output step S4. That is, the post-processing unit 5 outputs the chemical components corresponding to the entire image 18a (i.e., the chemical components of the scrap material 16 included in the entire image 18a) as estimated values ​​through the averaging process. The chemical component output unit 4 may flag the segmented images 18b whose product category is determined to be background by the chemical component output unit 4 in the chemical component output step S4, and the post-processing unit 5 may exclude the flagged segmented images 18b from the averaging process in post-processing step S5. If the total weight of the scrap material 16 is known, the estimated chemical components may be multiplied by the total weight to output the estimated weight of each chemical component.

[0041] In this way, by performing the photographing step S1, the pre-processing step S2, the chemical composition output step S4 (which, as mentioned above, includes the product category information output step S3 in this embodiment), and the post-processing step S5 in this order, it is possible to estimate the chemical composition of a given scrap material 16. Therefore, by performing the scrap material supply step S8 (see FIG. 7) of supplying the scrap material 16 to the melting furnace 15 based on the results of this estimation, it is possible to realize efficient production of metal products. The scrap material supply step S8 will be described in detail later.

[0042] After the product category information output step S3, the processing unit 9 may be provided with a product category information averaging processing unit that executes a product category information averaging processing step that obtains product category information 20 corresponding to the entire image 18a by averaging the product category information 20 output for each divided image 18b, and after the product category information averaging processing step, in a chemical component output step S4, the chemical component output unit 4 may input the product category information 20 corresponding to the entire image 18a and output the chemical components corresponding to the input. With such a configuration and method, the chemical components of a given scrap material 16 can also be estimated.

[0043] In the notification step S6, when the product category information output unit 3 outputs product category information 20 including a predetermined product category, the notification unit 7 issues a notification requesting an inspection of the photographed predetermined scrap material 16. The notification can be made, for example, by an image or sound via the output device 14. According to the notification step S6, for example, when the product category information output unit 3 outputs product category information 20 including a predetermined product category (e.g., a product category including electric motors that are likely to contain a large amount of copper), which is undesirable for supplying to the melting furnace 15 when manufacturing iron-based metal products, in the scrap material inspection step S7, a person or machine, for example, can inspect the photographed predetermined scrap material 16 before supplying it to the melting furnace 15 based on the notification and remove scrap material 16 belonging to the predetermined product category from the scrap material 16 as necessary. Therefore, the notification step S6 can enable more efficient production of metal products.

[0044] The chemical component output unit 4 may set an inspection flag on the divided image 18b on which the product category information 20 as described above has been output by the chemical component output unit 4 in the chemical component output step S4, and may notify the divided image 18b with the inspection flag set in the notification step S6. Such notification allows the inspection to be easily performed in the inspection step.

[0045] If scrap material 16 containing undesirable chemical components is removed by the notification step S6 and the scrap material inspection step S7, the need to dilute the molten metal with additional high-quality scrap material 16 can be reduced, thereby reducing production costs. Also, the melting time of the additional high-quality scrap material 16 can be reduced, thereby reducing operation time.

[0046] It is also possible to omit the notification step S6 (notification unit 7) and the scrap material inspection step S7. It is also possible to omit the pre-processing step S2 (pre-processing unit 2) and the post-processing step S5 (post-processing unit 5). The processing may be started by a person taking a photograph using the photographing device 6 and inputting the obtained image 18 into the computer 8.

[0047] The chemical components estimated in the post-processing step S5 can be output to the output device 14 for notification to a person, for example, or can be stored in the auxiliary storage device 12. Furthermore, if an inspection flag is attached, the notification unit 7 of the output device 14 can notify the user that an inspection is necessary. The divided image 18b and the corresponding product category information 20, inspection flag, and exclusion flag may be stored in the auxiliary storage device 12 and output to the output device 14 or the like when necessary in accordance with an instruction from the input device 13.

[0048] In the scrap material supply step S8, for example, a person or a machine supplies the scrap material 16 to the melting furnace 15 based on the estimation results of the chemical composition of the scrap material 16 using the component estimation method of this embodiment (i.e., the estimated value of the chemical composition of the scrap material 16 included in the overall image 18a, or the estimated chemical composition of the scrap material 16 included in the overall image 18a). By supplying the scrap material 16 whose approximate chemical composition is known in advance to the melting furnace 15, metal products can be manufactured efficiently. In other words, the melting furnace operation method of this embodiment includes a scrap material supply step in which part or all of the scrap material 16 whose chemical composition has been estimated in advance using the component estimation method of this embodiment is supplied to the melting furnace based on the chemical composition estimated by the component estimation method.

[0049] As an alternative to the melting furnace operation method, a scrap material 16 management method, which will be described later, can be used. That is, a melting furnace operation method according to another embodiment includes a scrap material supply step in which some or all of the scrap material 16 to be managed by the scrap material management method, which will be described later, is supplied to the melting furnace. In this case, the chemical composition of the scrap material 16 is managed in advance according to predetermined conditions, and the scrap material 16 can be supplied to the melting furnace 15 without further chemical composition estimation, thereby enabling efficient production of metal products. In this case, multiple portions of the managed scrap material 16 can be prepared and further mixed before supply. Furthermore, the chemical composition of the mixed scrap material 16 can be estimated again using the composition estimation method of this embodiment before supplying it to the melting furnace.

[0050] Furthermore, the chemical composition of the scrap material 16 may be managed based on the chemical composition of the scrap material 16 estimated in advance by the component estimation method of this embodiment. In this case, the scrap material management method estimates the chemical composition of the scrap material 16 to be managed by the component estimation method of this embodiment, and if the chemical composition estimated by the component estimation method does not satisfy a predetermined condition, the scrap material 16 is partially removed from the scrap material 16 to be managed, or another scrap material 16 is added to the scrap material to be managed, based on the estimated chemical composition. The predetermined condition is, for example, set in advance.

[0051] When a portion of a scrap pile 21 (see Figure 5 ) containing piled scrap material 16 is photographed, the estimated chemical composition of the scrap material 16 obtained from the captured image 18 may be used as a representative value of the chemical composition of the scrap pile 21. In this case, the representative value is used as an estimated value of the chemical composition of any scrap material 16 removed from the scrap pile 21. The weight of the scrap material 16 removed from the scrap pile 21 is measured using a weighing device, and the weight is multiplied by the representative value of the chemical composition to estimate the amount of chemical composition of the scrap material 16 to be supplied to the melting furnace 15. Based on the estimated chemical composition, the difference from the target chemical composition of the molten metal is confirmed, and the scrap pile 21 from which the next scrap material 16 will be removed and the amount of scrap material 16 to be removed are determined. This allows for the production of metal products with the target composition in a short time. Adjusting the amount of auxiliary materials, such as ore, can further improve the molten metal's chemical composition, thereby shortening the production time and improving the quality of the metal products. If molten metal with the desired chemical composition can be produced in a short time, the energy cost for operating the melting furnace 15 can be reduced, and therefore the cost of metal products can be reduced.

[0052] In a melting furnace facility 25 having a scrap storage area 24 equipped with a component estimation device 1 and a melting furnace 15, when an image 18 is acquired by photographing scrap material 16 being transported by a transport device such as a lifting magnet 22, a dolly 23, or a hopper, as shown in FIG. 6, the chemical components obtained from the image 18 may be used as a representative value of the chemical components of the scrap material 16 being transported. After measuring the weight of the scrap material 16 being transported, the weight is multiplied by the representative value of the chemical components to estimate the amount of chemical components of the scrap material 16 being transported and supplied to the melting furnace 15. Weighing each transport device enables more accurate estimation of the chemical components than for each scrap pile 21.

[0053] The camera device 6 is preferably positioned so as to capture images of the scrap material 16 while it is moving. For example, the camera device 6 captures images of the scrap material 16 while it is moving, such as when the scrap material 16 is being moved by the lifting magnet 22, when the conveying device is tilted to transfer the scrap material 16 to another conveying device, or when the scrap material 16 is being dropped from the conveying device into the melting furnace 15 and is being lowered, to generate images 18. By capturing images of the scrap material 16 while it is moving, it is possible to capture images of the scrap material 16 in a more fragmented state, thereby improving the accuracy of the estimation of product category information 20 estimated from the resulting images 18. While only the surface of a mass of scrap material 16 (such as the scrap pile 21) can be captured when the scrap material 16 is stationary, capturing images of the interior of the mass of scrap material 16 makes it easier to capture images of the interior of the mass of scrap material 16, thereby improving the accuracy of the estimation of product category information 20.

[0054] The melting furnace operating method of this embodiment may be performed, for example, as shown in a first specific example in FIG. 8 . In the first specific example, first, conditions such as the target chemical composition of the metal product (molten metal) are set, and then the photographing device 6 photographs an image 18 including the scrap material 16 (photographing step S1). Then, the preprocessing unit 2 segments the image 18 to create segmented images 18b (preprocessing step S2). Then, the product category information output unit 3 of the chemical composition output unit 4 outputs product category information 20 for each segmented image 18b using the trained model 19 (product category information output step S3). Then, the chemical composition output unit 4 outputs chemical compositions for each segmented image 18b from the product category information 20 output by the product category information output unit 3 using a table (chemical composition output step S4). Then, the chemical composition output unit 4 assigns an exclusion flag to segmented images 18b that the product category information output unit 3 estimates to be other than the scrap material 16 (background). Then, the post-processing unit 5 averages the chemical components of the segmented images 18b that are not flagged for exclusion, and outputs the averaged values ​​as estimated values ​​of the chemical components of the entire image 18a (post-processing step S5). Then, in accordance with the condition settings, the chemical components are output to the output device 14 as needed. Then, in accordance with the condition settings, the processing results by the processing unit 9 are stored in the auxiliary storage device 12 as needed. Then, when it is OK to end the processing (for example, when a molten metal with the target composition has been obtained), the processing is ended.

[0055] The melting furnace operating method of this embodiment may also be performed, for example, as shown in a second specific example in FIG. 9 . In the second specific example, first, conditions such as the target chemical composition of the metal product (molten metal) are set, and then the photographing device 6 photographs an image 18 including the scrap material 16 (photographing step S1). Then, the preprocessing unit 2 segments the image 18 to create segmented images 18b (preprocessing step S2). Then, the product category information output unit 3 of the chemical composition output unit 4 outputs product category information 20 for each segmented image 18b using the trained model 19 (product category information output step S3). Then, the chemical composition output unit 4 outputs chemical compositions for each segmented image 18b from the product category information 20 output by the product category information output unit 3 using a table (chemical composition output step S4). Then, the chemical composition output unit 4 assigns an exclusion flag to segmented images 18b that the product category information output unit 3 estimates to be other than the scrap material 16 (background). In the second specific example, if there is a segmented image 18b to which product category information 20 including a product category that may contain a large amount of chemical components that are undesirable for supply to the melting furnace 15 has been output by the chemical component output unit 4, an inspection flag is attached to the segmented image 18b. Then, the post-processing unit 5 averages the chemical components of the segmented images 18b that have not been flagged for exclusion, and outputs the averaged value as an estimated value of the chemical components of the entire image 18a (post-processing step S5). If there is an inspection flag, the inspection flag is output to the output device 14. Then, in accordance with the condition settings, the chemical components are output to the output device 14 as needed. Then, in accordance with the condition settings, the processing results by the processing unit 9 are stored in the auxiliary storage device 12 as needed. Then, if it is OK to end the processing (for example, if a molten metal with the target components has been obtained), the processing is ended.

[0056] In the embodiments described above, the trained model 19 outputs product category information 20 of the scrap material 16, but the present invention is not limited thereto. For example, the trained model 19 may be a trained machine learning model that receives an image 18 of the scrap material 16 as input and outputs chemical components corresponding to the product category information 20 of the scrap material 16. That is, in the above embodiments, the trained model 19 may have the combined function of outputting product category information 20 and serving as a database that receives the output product category information 20 as input and outputs chemical components corresponding to the input. In this case, the trained model 19 is trained by machine learning using a combination of the image 18 of the scrap material 16 and the chemical components corresponding to the scrap material 16 as training data.

[0057] 7, the flow of the component estimation method in this case does not include the product category information output step S3. In other words, the method has a chemical component output step S4 that does not include the product category information output step S3. In this case, in the chemical component output step S4, the chemical component output unit 4 uses the divided image 18b included in the image group as an input image and outputs the chemical components corresponding to the input image using the trained model 19.

[0058] In this case, the component estimation program of this embodiment causes the computer 8 to execute a chemical component output step S4 that does not include the product category information output step S3.

[0059] To reduce CO2 emissions, the use of electric furnaces, which are said to emit about one-quarter the amount of CO2 emitted by blast furnaces, is expected to expand. The primary raw materials used in electric furnaces are waste materials from steel mills and various iron-based scraps from automobiles, construction, and electrical appliances. Because the composition of the molten steel tapped from an electric furnace is directly related to the quality of the final product, controlling the composition of the molten metal produced in the electric furnace is important in electric furnace operation. In many cases, electric furnace operation involves measuring the chemical composition of the molten metal and adding necessary additives, such as coke, to achieve the desired composition. Repeated detection of the chemical composition of the molten metal and the addition of additives are required to achieve the desired composition, which is a labor-intensive process. According to the present embodiment described above, the chemical composition of the scrap material input into an electric furnace can be estimated in advance, enabling the composition of the molten metal to be adjusted in a short time, thereby significantly reducing the workload of such electric furnaces. Furthermore, according to this embodiment, the operation time of the electric furnace can be shortened, the chemical composition of the molten metal in the electric furnace can be stabilized, thereby improving the quality, and the production cost of the molten metal produced in the electric furnace can be reduced.

[0060] The present invention is not limited to the above-described embodiment, and various modifications can be made without departing from the spirit and scope of the present invention. [Explanation of symbols]

[0061] 1 Component Estimator 2 Pretreatment section 3 Product category information output section 4 Chemical component output section 5 Post-processing section 6. Imaging equipment 7. Information Department 8. Computers 9 Processing section 10 central processing unit 11 Main memory 12 Auxiliary storage 13 Input Devices 14 Output Devices 15 Melting furnace 16 Scrap materials 17 electrodes 18 images 18a Full image 18b Split image 19 Pre-trained models 20 Product Category Information 21 Scrap Mountain 22 Lifting Magnet 23 Cart 24 Scrap Yard 25 Melting furnace equipment S1 Shooting Steps S2 Pretreatment step S3 Product category information output step S4 Chemical composition output step S5 Post-processing step S6 Notification step S7 Scrap material inspection step S8 Scrap material supply step

Claims

1. 1. A composition estimation method for estimating chemical composition of scrap material, comprising: a pre-processing step of creating an image group consisting of all or part of a plurality of divided images obtained by dividing an image of a predetermined scrap material into predetermined image sizes; a chemical component output step of outputting, for each of the divided images included in the image group, chemical components corresponding to the predetermined scrap material in the input image using the divided image as an input image and a trained model; and The chemical component output step includes a product category information output step of using, for each of the divided images included in the image group, the divided image as an input image, and outputting product category information corresponding to the predetermined scrap material in the input image using the trained model, and using the product category information obtained through the product category information output step as an input, outputting chemical components corresponding to the input using a database; The product category information is information indicating a category of an original product of the scrap material having a shape included in the segmented image, A component estimation method in which the trained model is generated by learning using training data including a combination of a product category and a training image containing scrap material that falls into the product category.

2. The component estimation method according to claim 1 , wherein the predetermined image size has all sides of 100 mm or more and 2000 mm or less in real scale.

3. 2. The component estimation method according to claim 1, wherein the product category information output step outputs one product category as the product category information for the input image.

4. 2. The component estimation method according to claim 1, further comprising a post-processing step of estimating chemical components of an original image using the chemical components output for each of the input images by the chemical component output step.

5. The chemical composition of a scrap material to be managed is estimated by the composition estimation method according to any one of claims 1 to 4, A scrap material management method, wherein if the chemical components estimated by the component estimation method do not satisfy predetermined conditions, a portion of the scrap material to be managed is removed based on the estimated chemical components, or another scrap material is added to the scrap material to be managed.

6. 5. A melting furnace operation method comprising a scrap material supply step of supplying part or all of scrap material, the chemical composition of which has been estimated in advance by the component estimation method according to any one of claims 1 to 4, to a melting furnace based on the chemical composition estimated by said component estimation method.

7. A melting furnace operating method comprising a scrap material supply step of supplying a part or all of the scrap material to be managed by the scrap material management method according to claim 5 to the melting furnace.

8. A component estimation program that causes a computer to execute the component estimation method according to any one of claims 1 to 4.

9. A component estimation device for estimating chemical components of scrap material, comprising: a pre-processing unit that creates an image group consisting of all or part of a plurality of divided images obtained by dividing an image of a predetermined scrap material into predetermined image sizes; a chemical component output unit that uses the divided image included in the image group as an input image and outputs chemical components corresponding to the predetermined scrap material in the input image using a trained model, The chemical component output unit has a product category information output unit that uses, for each of the divided images included in the image group, the divided image as an input image and outputs product category information corresponding to the predetermined scrap material in the input image using the trained model, and uses the product category information obtained via the product category information output unit as an input and outputs chemical components corresponding to the input using a database; The product category information is information indicating a category of an original product of the scrap material having a shape included in the segmented image, A component estimation device in which the trained model is generated by learning using training data including combinations of product categories and training images containing scrap material that falls into the product categories.

10. A melting furnace facility comprising the component estimation device according to claim 9 and a melting furnace.

11. A scrap yard having the composition estimation device according to claim 9.

12. A method for generating a trained model in which a divided image obtained by dividing an image of a predetermined scrap material into a predetermined image size is input, and product category information of the scrap material is output, The product category information is information indicating a category of an original product of the scrap material having a shape included in the segmented image, A method for generating a trained model, which generates the trained model by learning using training data including a combination of a product category and a training image containing scrap material that falls into the product category, and extended training data generated by resizing the training image by enlarging or reducing it, moving it, rotating it, partially cropping or flipping it, or any combination of these.

13. The image processing system is generated by machine learning using training data including combinations of product categories and training images including scrap materials corresponding to the product categories, A computer is caused to function so that divided images obtained by dividing an image of a predetermined scrap material into predetermined image sizes are input, and product category information of the scrap material is output; The product category information is a trained model that indicates the category of the original product of the scrap material having the shape included in the segmented image.

Images