Apparatus and method for assigning a material value score to printed circuit board waste or a portion thereof, and system for sorting printed circuit board waste

Abstract

To provide an apparatus and a method for assigning a material value score to a printed circuit board waste material or a portion thereof, and a system for sorting the printed circuit board waste material.SOLUTION: An apparatus 100 comprises a dual energy or spectral x-ray image acquisition unit 120 and a material value score determination unit 130. The apparatus assigns a material value score 110 to a printed circuit board waste material 200 or a portion 210 thereof. The material value score determination unit 130 which is a machine learning module determines the material value score 110 by applying a dual energy or spectral x-ray image 122 of the printed circuit board waste material 200 or the portion 210 thereof to the machine learning module.SELECTED DRAWING: Figure 2

Description

Technical Field

[0001] Each embodiment according to the present invention is an apparatus and method for assigning a material value score to printed circuit board waste or a portion thereof, and a system for sorting printed circuit board waste, for example, applying a component detection method assisted by machine learning to an X-ray image and performing recycling based on, for example, material value estimation.

Background Art

[0002] Electronic devices and electronic apparatuses play a major role in daily life. However, in order to achieve the goal of a circular economy, the recycling of waste electrical and electronic equipment (WEEE) has become important.

[0003] Currently, in the process of recycling printed circuit board waste (WPCB) from waste electrical and electronic equipment (WEEE), the estimation of the value of the entire WPCB before actual metallurgical recovery is only approximate. Especially for the value of large batches, there is no choice but to estimate based on a few random samples and visual inspection by skilled human operators, which is time-consuming and error-prone. Since neither the seller nor the buyer of WPCB can grasp the total value of the WPCB batch, it is difficult for both parties to set prices. At present, no commercially available system for estimating the value of WPCB with sufficient speed and accuracy is known.

[0004] There are initial scientific publications regarding the detection of components on a PCB [1] and the value estimation of WPCB [2]. However, both are based on visual images of (W)PCB. When components are mounted on both sides of (W)PCB, the information obtained by this method is only insufficient. This is because only one side can be analyzed simultaneously as long as an image based on visible light is used. Also, even for electronic components with similar appearances such as integrated circuits, the internal materials may be different. Such limitations apply to any other type of image that generates a surface image without penetrating the object (reflected light image generation). For example, but not limited to, infrared, ultraviolet, and terahertz.

[0005] Therefore, it is desirable to provide a concept that provides a better compromise between speed and accuracy in determining the value of WPCBs.

[0006] The above problems are achieved by the subject matter of the independent claims of the present application.

[0007] Further embodiments according to the present invention are defined by the subject matter of the dependent claims of the present application.

SUMMARY OF THE INVENTION

[0008] In one aspect of the present invention, the inventor of the present application has noticed that one problem when attempting to recycle printed circuit boards stems from the fact that the value of waste printed circuit boards (WPCBs) is often determined by humans. According to a first aspect of the present application, the above problem is solved by using an apparatus configured to automatically analyze WPCBs. This can minimize errors and improve the speed of analysis. Furthermore, the inventor has found that it is advantageous to determine a material value score associated with each PCB. This material value score makes it possible to determine whether sufficient valuable materials, particularly valuable metals, can be recovered from the WPCBs. Thereby, the accuracy in the value evaluation of WPCBs targeted for recycling can be improved.

[0009] Therefore, according to this aspect of the present application, one embodiment relates to an apparatus for assigning a material value score to printed circuit board waste or a portion thereof. The material value score may depend on, for example, the content of one or more predetermined materials within or on the WPCB. The material value score may indicate the monetary value of the WPCB. The one or more predetermined materials may be one or more materials to be recovered from the printed circuit board waste. The predetermined materials may be predefined, for example, by the user of the apparatus. Thus, it becomes possible to read from the material value score whether it is cost-effective to recover one or more predetermined materials from the WPCB. The content of the one or more predetermined materials may be determined based on elemental analysis of representative components in the WPCB. The elemental analysis may be performed using inductively coupled plasma optical emission spectrometry (ICP-OES).

[0010] According to one embodiment, the apparatus is configured to determine a material value score and is configured to perform the determination based on a dual-energy or spectral X-ray image of printed circuit board waste or a portion thereof. Here, the dual-energy or spectral X-ray image may include a derived image thereof, for example, a basic material decomposition image, or an image obtained by decomposing into an effective atomic number and a surface density. The apparatus may be configured to perform the determination of the material value score, for example, by detecting components on the WPCB from the dual-energy or spectral X-ray image and determining the number of components for each component type. As an optional form, the apparatus may be configured to derive the size and / or weight of the components on the WPCB from the dual-energy or spectral X-ray image. The apparatus calculates, for example, the amount of a predetermined material included in or recoverable from the WPCB based on information obtained from the dual-energy or spectral X-ray image, and determines the material value score based on the calculated amount of the predetermined material. The apparatus includes, for example, a dual-energy or spectral X-ray unit configured to acquire a dual-energy or spectral X-ray image of printed circuit board waste or a portion thereof, or the apparatus is configured to receive the dual-energy or spectral X-ray image from an external dual-energy or spectral X-ray unit. The use of the dual-energy or spectral X-ray image in this way is based on the finding that the components on the WPCB are shown in the dual-energy or spectral X-ray image regardless of the orientation of the WPCB. The advantages include that information about the components on both the front and back sides of the WPCB can be obtained in the dual-energy or spectral X-ray image. Since there are no hidden components, it is possible to improve the evaluation accuracy of the WPCB to be recycled by analyzing all the components on the WPCB. In this way, the value of the WPCB can be determined more simply and efficiently. Furthermore, the inventor has found that dual-energy or spectral X-ray is an extremely robust imaging technology for applications in environments with a lot of dust, sand, and dirt. Therefore, it is possible to distinguish each component arranged on the WPCB even if dust, sand, or dirt adheres to it.This enables high accuracy to be obtained and also makes it possible to assign a meaningful material value score to the WPCB.

[0011] According to one embodiment, the apparatus is configured to perform a determination by subjecting a dual - energy or spectral X - ray image to a machine - learning module. This is based on the idea that high component detection performance can be achieved by the machine - learning module. To date, there is no freely accessible annotated PCB and X - ray image data of its respective components. The inventor has found that an efficient and accurate machine - learning module can be achieved, for example, by selecting and annotating representative components to be recycled from the WPCB according to the value of the components targeted for the recycling process. In addition, the inventor has noticed that the processing speed required for analyzing the value of the WPCB can be achieved by using the machine - learning module.

[0012] According to one embodiment, the apparatus is configured to detect components of printed circuit board waste, such as electronic components like integrated circuits (ICs), tantalum capacitors, ball grid arrays (BGAs), pin grid arrays (PGAs), connectors, etc., from a dual - energy or spectral X - ray image.

[0013] The apparatus may be configured to obtain information regarding the average material composition for each component type. For example, an external device may analyze a plurality of components for each component type to determine the average content of each material in the plurality of components of each component type for one or more materials. The average content may be represented by the average molar concentration or average mass of each material per component or per mass of the component of each component type. The analysis of the materials of the components can be performed using inductively coupled plasma optical emission spectrometry (ICP - OES). The apparatus can receive information regarding the average content as information regarding the average material composition from an external device. Alternatively, the apparatus may be provided with a database indicating the average content of one or more materials in the components of each component type as information regarding the average material composition for one or more component types.

[0014] According to one embodiment, the apparatus may be configured to perform the determination of the material value score for each component type and for each material to be recovered from the WPCB (e.g., for at least one component type and one material) by the following steps: multiplying the number of components of each component type by the average mass of each material per component of each component type to obtain the amount·mass of each material in each component type, and determining the value of the amount·mass of each material.

[0015] In addition, the apparatus may be configured to obtain the material value score by obtaining the sum of all values, or to obtain the material value score by obtaining the sum of all values associated with each material for each material. The material to be recovered may be predefined by the user of the apparatus. Accordingly, the material value score depends on the material of interest. The material value score of the WPCB may depend on the material selected for determination. When there are two or more materials, the material value score may indicate either individual values, or a single value that is the sum or weighted sum of all values, for each of the two or more materials. The determination of the value of the amount·mass of each material may depend on the rate·price of the material. The apparatus may be configured to update the rate·price of the material, for example, daily.

[0016] According to one embodiment, the apparatus is configured to estimate the weight and / or size of the entire WPCB, and / or its components, such as the type, weight, and (optionally) size of electronic components, from the dual energy or the spectral X-ray image. The information regarding the weight and size of the entire WPCB makes it possible to determine the content of valuable materials present in the WPCB. When determining the content of valuable materials, the apparatus may be configured to consider the weight and size of the entire WPCB in combination with the information regarding the type, weight, and (optionally) size of its components. In particular, the information regarding the weight and / or size of its components is advantageous in determining the content of valuable materials.

[0017] According to one embodiment, the apparatus may be configured to perform the determination of the material value score for each component type and for each material to be recovered from the WPCB (e.g., for at least one component type and one material) by the following steps: for each component, multiply the weight of each component by the average mass of each material per unit mass of each component type to obtain the component-specific quantity / mass of each material for each component, sum up the component-specific quantities / masses of each material across all components to obtain the overall quantity / mass of each material, and determine the value of the overall quantity / mass of each material.

[0018] In addition, the apparatus may be configured to obtain the material value score by summing all the values, or for each material, obtain the material value score by summing all the values associated with each material. The materials to be recovered may be predefined by the user of the apparatus. Thus, the material value score depends on the materials of interest. The material value score of the WPCB may depend on the materials selected for determination. When there are two or more materials, the material value score may indicate either individual values or the sum of all values as a single value for each of the two or more materials. The determination of the value of the quantity / mass of each material may depend on the rate / price of the material. The apparatus may be configured to update the rate / price of the material, for example, daily.

[0019] According to one embodiment, the apparatus is configured to determine the component material value based on the type, weight, and / or size associated with each component, e.g., an electronic component. The component material value depends on the content of one or more predetermined materials in each component. One or more predetermined materials may be materials to be recovered from the WPCB. The component material value may indicate the monetary value of each component. If a component or one or more components in a component become an obstacle to one of the subsequent steps of the recycling process, the material value may be negative.

[0020] According to one embodiment, the apparatus is configured to execute determination of a material value score based on the component material value. For example, the apparatus is configured to obtain a single value of the WPCB by obtaining the sum of all component material values.

[0021] According to one embodiment, the apparatus is configured to obtain a material value score by simultaneously analyzing both sides of the printed circuit board waste. In this way, it becomes possible to efficiently analyze double-sided WPCBs, and it becomes possible to avoid overlooking downward-facing components, i.e., components facing the conveyor belt or support pads. Due to this feature, it is no longer necessary to separately evaluate each side of the WPCB, and it becomes possible to detect all components regardless of whether they are facing the camera.

[0022] A further embodiment relates to a method for assigning a material value score to a printed circuit board waste or a part thereof. The method is based on the same considerations as the above-described apparatus. Additionally, the method can be completed by all the features and functions described with respect to the apparatus.

[0023] A further embodiment relates to a system for sorting printed circuit board waste. The system includes the above-described apparatus for assigning a material value score to a printed circuit board waste or a part thereof. Additionally, the system includes a sorting apparatus for classifying a plurality of printed circuit board wastes into two or more grades for the printed circuit board waste according to the material value scores assigned to each of the plurality of printed circuit board wastes by the above-described apparatus.

[0024] A further embodiment relates to a method for sorting printed circuit board waste. The method comprises the above-described method for assigning a material value score to printed circuit board waste or a portion thereof. Additionally, the method comprises the step of classifying a plurality of printed circuit board wastes into two or more grades for the printed circuit board waste according to the material value scores assigned to each of the plurality of printed circuit board wastes by the above-described method. The method is based on the same considerations as the above-described system. Additionally, the method can be completed by all the features and functions described with respect to the system.

Brief Description of the Drawings

[0025] The drawings are not necessarily to scale and generally emphasize explaining the principles of the present invention. In the following description, various embodiments of the present invention will be described with reference to the following drawings.

[0026]

Figure 1

Figure 2

Figure 3

Figure 4

Figure 5

Figure 6

Mode for Carrying Out the Invention

[0027] [Detailed Description of the Embodiment] In the following description, elements that are equal or equivalent, or elements having equal or equivalent functions, are assigned equal reference numerals or equivalent reference numerals even if they appear in different figures.

[0028] In the following description, numerous details are set forth in order to provide a more thorough explanation of each embodiment of the present invention. However, it will be apparent to those skilled in the art that each embodiment of the present invention can be practiced without these specific details. On the other hand, well-known structures and devices are shown in block diagram form rather than in detail in order to avoid obscuring the description of each embodiment of the present invention. Also, the features of the different embodiments described in this specification can be combined with each other unless otherwise stated.

[0029] FIG. 1 shows an apparatus 100 for assigning a material value score 110, such as 110a or 110b, to a WPCB 200 or a portion 210 thereof. The material value score 110a can represent the value of the entire WPCB when recycling the WPCB 200, and the material value score 110b can represent the value of the portion 210 of the WPCB when recycling the WPCB 200. The material value score 110 can indicate whether it is cost-effective to recover materials from the WPCB 200 or the portion 210.

[0030] According to one embodiment, the apparatus 100 can be configured to divide the WPCB 200 into a plurality of portions 210 and assign an individual material value score 110b to one or more of these portions for each portion. Optionally, the apparatus 100 can be configured to assign an individual material value score 110b to only the portion 210 with the highest material value score 110b. In this case, the portion 210 of the WPCB 200 that is most cost-effective for material recovery can be efficiently indicated.

[0031] According to one embodiment, the apparatus 100 can be configured to show both the material value score 110a for the entire WPCB 200 and the material value score 110b for one or more portions 210 in the WPCB 200. The overall material value score 110a makes it possible to determine whether the WPCB 200 is waste or has recycling value, and the individual material value scores 110b for each part make it possible to distinguish the most valuable parts in the recovery of valuable materials.

[0032] According to one embodiment, the material value score 110 depends on the content of one or more predetermined materials to be recovered from the printed circuit board waste 200. In other words, the material value score 110 may depend on the amount of one or more predetermined materials in the WPCB 200, for example, the amount of one or more predetermined materials in the components 220 of the WPCB 200.

[0033] Optionally, the apparatus 100 may comprise any of the features described with respect to the apparatus of FIG. 2.

[0034] FIG. 2 shows an apparatus 100 comprising a dual-energy or spectral X-ray image acquisition unit 120 and a material value score determination unit 130.

[0035] The dual-energy or spectral X-ray image acquisition unit 120 can be configured to acquire a dual-energy or spectral X-ray image 122 of the WPCB 200 or a portion 210 thereof. Optionally, instead of the apparatus 100 comprising the dual-energy or spectral X-ray image acquisition unit 120, the apparatus 100 may be configured to receive the dual-energy or spectral X-ray image 122 from an external dual-energy or spectral X-ray image acquisition unit 120.

[0036] The device 100 is configured to obtain a material value score 110 by simultaneously analyzing the two sides 230, 240 of the printed circuit board waste 200. In contrast to each of the other methods based on visual images or NIR images, the X-ray image 122 can show the components 220 on the WPCB 200 regardless of the orientation of the WPCB 200, that is, the components 220 on the front surface 230 and the back surface 240 can be detected simultaneously. Further, in applications in an environment with a lot of sand, dust or powder, the XRT 120 is a more robust imaging technology.

[0037] The material value score determination unit 130 is configured to determine a material value score 110 for the WPCB 200 or one or more of its parts 210. This determination is performed based on the dual energy or spectral X-ray image 122 (XRT).

[0038] In any form, the material value score determination unit 130 is, for example, a machine learning module.

[0039] In any form, the material value score determination unit 130 can include an image analysis unit 131 for analyzing the dual energy or spectral X-ray image 122 with respect to the individual components 220 of the WPCB 200. For example, the image analysis unit 131 may be configured to obtain information about the component types of the components 220 arranged on the WPCB 200 and information about the number of components 220 per component type. In any form, in addition to this, the image analysis unit 131 may also obtain information about the size and / or weight of the components 220.

[0040] The material value score determination unit 130 is configured to obtain a plurality of component material value scores 112 by determining, for example, for each component type, a component material value score using, for example, the component material value determination unit 136. The component material value score can be determined by multiplying the number of components 220 of each component type by the average component material value score associated with each component type. Each average component material value score can be a predefined value score assigned to each component type. The component material value determination unit 136 may include a database that indicates a plurality of average component material value scores and, for each average component material value score, indicates the associated component type. Alternatively, the component material value determination unit 136 may be configured to update the corresponding average component material value score for each component type based on the variation in the market value of the valuable materials included in each component type. The apparatus 100 is configured to execute the determination of the material value score 110 based on the plurality of component material value scores 112, for example, by obtaining the sum of the plurality of component material value scores 112.

[0041] The image analysis unit 131 is, for example, a machine learning module.

[0042] Optionally, the apparatus 100 may be configured to detect the components 220 of the printed circuit board waste 200 from the dual energy or spectral X-ray image 122 using, for example, the component detection unit 132. In addition, the apparatus 100 may be configured to estimate characteristics such as the type, weight, and / or size of the components 220 from the dual energy or spectral X-ray image 122 using, for example, the component characteristic determination unit 133. The apparatus 100 may be configured to determine the component material value for each component 220 based on one or more characteristics associated with each component 220.

[0043] The device 100 may be configured to obtain information regarding the average mass of a predefined material per size or weight of a certain component type. In any form, the average mass may be obtainable for different materials and / or different component types respectively. Based on the information regarding the average mass and the information regarding the size and / or weight of the individual components 220 on the WPCB 200, the device 100 is configured to determine the amount of a predefined material included in the WPCB 200. In addition to this, the device 100 may be configured to obtain the rate - price of a predefined material included in the WPCB 200 and determine a material value score 110 based on the rate - price of the predefined material and the amount of the predefined material. This can be performed for one or more predefined materials, and the material value score may indicate the value of an individual material or the combined value of all the predefined materials.

[0044] Similarly, the device 100 may be configured to obtain information regarding the average mass of a predefined material per component of a certain component type. In any form, the average mass can be made obtainable for different materials and / or different component types respectively. Based on the information regarding the average mass and the information regarding the number of the individual components 220 on the WPCB 200, the device 100 is configured to determine the amount of a predefined material included in the WPCB 200. In addition to this, the device 100 may be configured to obtain the rate - price of a predefined material included in the WPCB 200 and determine a material value score 110 based on the rate - price of the predefined material and the amount of the predefined material. This can be performed for one or more predefined materials, and the material value score may indicate the value of an individual material or the combined value of all the predefined materials.

[0045] FIG. 3 shows a block diagram of a method 300 for the assignment 310 of a material value score 110 to a WPCB 200 or a portion 210 thereof. In any form, the method 300 comprises a determination 320 of the material value score 110. The determination 320 can be performed based on a dual energy or spectral X-ray image 122 of the WPCB 220 or a portion 210 thereof. The determination 320 is advantageously considered to be performed by subjecting the dual energy or spectral X-ray image 122 to a machine learning process.

[0046] The method 300 can include the features and / or functions described with respect to the apparatus 100 of FIGS. 1 and / or 2.

[0047] FIG. 4 shows a system 400 for sorting printed circuit board waste 200, e.g., 200 1 ~200 5 The system 400 comprises an apparatus 100 for assigning a material value score 110, e.g., 110 1 ~110 5 to the WPCB 200 or a portion 210 thereof. In addition to this, the system 400 comprises a sorting apparatus 410 for classifying a plurality of printed circuit board wastes 200 into two or more grades 420, e.g., 420 1 ~420 2 for the printed circuit board waste 200 according to the material value scores 110 assigned to each of the plurality of printed circuit board wastes 200 by the apparatus 100.

[0048] For example, the first grade 420 1 represents waste, i.e., a WPCB 200 that is not worth recycling, and the second grade 420 2 can represent a valuable WPCB 200, i.e., a WPCB 200 that is worth recycling.

[0049] The apparatus 100 can comprise the features and / or functions described with respect to the apparatus 100 of FIGS. 1 and / or 2.

[0050] FIG. 5 shows a block diagram of a method 500 for sorting printed circuit board waste 200. The method 500 includes a method 300 for assigning a material value score 110 to the WPCB 200 or a portion 210 thereof, and classifying 510 a plurality of printed circuit board wastes 200 into two or more grades 420 for the printed circuit board waste 200 according to the material value scores 110 assigned to each of the plurality of printed circuit board wastes 200 by the method 300.

[0051] The method 300 can include the features and / or functions described with respect to the method 300 of FIG. 3.

[0052] The method 500 can include the features and / or functions described with respect to the method 400 of FIG. 4.

[0053] FIGS. 4 and 5 illustrate a sorting system 400 and a method 500 for automatically sorting WPCBs 200 based on the predicted value, i.e., the material value score 110, of each individual WPCB 200. The system 400 can be configured to acquire dual energy or spectral X-ray images 122 (XRT) from the WPCBs 200 on a conveyor belt or chute. After preprocessing, these images 122 can be input into, for example, a deep neural network (or other suitable machine learning method, such as a method for detecting objects in an image), such that the deep neural network detects components 220, such as ICs, BGAs / PGAs, tantalum capacitors, connectors, etc. The system 400 calculates the value of each component 220 on the WPCB 200, i.e., the component material value, using a model based on features such as the type, size, and / or weight of the component 220. It is also possible to estimate both the size and weight of the detected component 220 from the X-ray image 122.

[0054] The monetary value of the WPCB200 obtained as a result from this model, i.e., the plurality of component material values 112, may be adapted to the individual metallurgical processes used for the processing of these WPCB200s at a later stage. Further, the model for predicting the value of the WPCB200, i.e., the material value score 110, may be adapted to the actual processes of the user of the system 400, i.e., for example, the yield in the actual metallurgical process, the current rate of the material value, or the components that impede the process and reduce the value. The user can set a threshold for the predicted value or value content 110 of the WPCB200, in which case, if it falls below this threshold, the system 400 rejects and separates out the WPCB, for example, classifies the WPCB into grade 420 1 for separation. The value content here means the value or monetary value per unit mass or per unit area, for example, euros / kg or euros / cm 2 .

[0055] FIG. 6 shows a process diagram of a method 600 for determining a model that can be used by the apparatus 100 described in the present specification for determining the material value score 110.

[0056] FIG. 6 shows, for example, the training of a model 132 for object detection, for example, the training of a deep learning model (DL model). In machine learning, a representative database is essential for the training and evaluation of a model. Therefore, the first step may be to generate such a machine-readable database and label base from samples of the real-world WPCB200. In one experiment, the inventors, for example, from individual WPCB200s (where the number of WPCBs can be 104), for example, performed an annotation 610 on components 220 (for example, representative components) to obtain annotated components 612 (where the number of annotated components can be 1514). For example, for training, an annotation 610 is performed on the components 220 of the WPCB200 to obtain an annotation dataset or database, that is, annotated components 612. As a further step, there may be something related to the acquisition 120 of the dual-energy or spectral X-ray image 122 of the WPCB. Based on the X-ray image 122 of the WPCB200 and based on the annotated components 612 of the same WPCB200, the model 132 for object detection can be trained. The trained model 132 and / or the model 132 during training are configured to identify, recognize, or detect the components of the WPCB200 to obtain detected components 614.

[0057] According to one embodiment, the object detection model 132 may be configured to estimate characteristics such as the type, weight, and / or size of the component 220 from the dual-energy or spectral X-ray image 122.

[0058] In addition to this, method 600 may include steps of preparing and excluding representative samples of different component classes from the WPCB 200. Next, for these samples, i.e., the WPCB 200, an analysis 620, for example, by ICP - OES (Inductively Coupled Plasma Optical Emission Spectroscopy. Any method capable of obtaining the content of other elements can be used), is performed to obtain the content 622 of one or more predetermined materials, such as valuable metals. Based on this information, i.e., the content 622 of one or more predetermined materials in the representative component 220 and the detected component 614, a true value that serves as the basis for calibrating the model 130 is obtained, thereby predicting the actual value of all corresponding components, such as the detected component 614, and ultimately the value 110 of the WPCB 200, for example, it as the sum or weighted sum of the values of all corresponding components. When obtaining the elemental content 622 of one or more predetermined materials in the representative components of the WPCB 200, the elemental analysis 620 of the representative components may be performed only once per applicable case. Using the obtained elemental content 622, the model 130 can be calibrated to perform value estimation.

[0059] The representative components may include integrated circuits (ICs), tantalum capacitors, ball grid arrays (BGAs), pin grid arrays (PGAs), and / or connectors. The selection of the representative components may depend on one or more predetermined materials to be recovered from the WPCB. The representative components may also include components that interfere with subsequent process steps in recycling and thus result in a negative value, such as large aluminum heat sinks or components containing toxic substances.

[0060] According to one embodiment, the apparatus 100 described in the present specification for determining the material value score 110 may be configured to obtain a dual - energy or spectral X - ray image 122 of the WPCB 200 and determine the detected component 614 using a model 132 trained for object detection. In addition to this, the apparatus 100 may be configured to determine the material value score 110 based on the detected component 614 using the model 130 for value estimation.

[0061] From here, the machine learning model can be trained to detect the corresponding component 220. From ICP-OES, the content of valuable materials in all component classes is known. Using the weight and material appearance rate estimated from the X-ray image 122, the value of each material in each component can be estimated. Therefore, the value 110 of the entire WPCB 200 can be calculated by obtaining the sum of the values of all components, that is, the sum of the multiple component material values 112. Depending on subsequent metallurgical processes and other user-specific parameters, the weighted sum of material values and the weighted sum of total values can be used to predict the material value (in the sense of achievable profit). The actual weights need to be adjusted to reflect each process. Also, these weights may reflect the recovery rate of a given material in the process. For some components, they may be assigned a negative value that hinders subsequent process steps.

[0062] In the method described in this application for detecting the component 220 on the WPCB 200 by a deep learning method, the inventor has found a suitable method for integrating prior art object detection networks such as YOLO [3] and EfficientDet [4]. In the training and evaluation of the database, the inventor followed an established evaluation scheme in machine learning. The data is divided into three separate sets, namely, for training, verification, and testing. The verification set is used to adjust the training process of the model to verify its performance. After this final model is found, the test data is used as an independent final evaluation. When the inventor obtained the scores based on the harmonic mean of the reproducibility and accuracy for four component classes, namely, IC, tantalum capacitor, connector, and BGA·PGA, they were 87.83%, 82.54%, 79.26%, and 88.89%, respectively. According to this component detection performance and the statistical findings obtained from chemical analysis, the model shows good results in predicting the material value score 110 from individual WPCBs 200 of WEEE.

[0063] Advantages include the ability to detect component 220 regardless of its visibility. That is, component 220 can appear in the X-ray image 122 whether the PCB 200 is placed above (230) or below (240) component 220, or mounted on both. Thus, since the X-ray based approach is independent of the position and orientation of the PCB during the imaging process, it does not need to be explicitly positioned and yet can still detect all components 220 on the PCB 200.

[0064] The evaluation of value 110 based on detected component 614 is advantageous because it can flexibly incorporate prioritization of important types of materials etc. for the user and factors such as the recovery rate of components of the material.

[0065] Although several aspects have been described in the context of an apparatus, these aspects also represent a description of corresponding methods, and it is clear that a block or apparatus corresponds to a method step or a feature of a method step. Similarly, aspects described in the context of a method step also represent a description of corresponding blocks or items or features of a corresponding apparatus.

[0066] Each embodiment of the invention may be implemented in hardware or in software, depending on the requirements of a particular implementation. This implementation can be carried out using a digital storage medium, such as a floppy disk, DVD, CD, ROM, PROM, EPROM, EEPROM or flash memory, which stores electronically readable control signals and which, when cooperating (or capable of cooperating) with a programmable computer system, causes the corresponding method to be executed.

[0067] Some embodiments according to the invention include a data carrier having electronically readable control signals and capable of cooperating with a programmable computer system so as to cause one of the methods described herein to be executed.

[0068] Generally, each embodiment of the present invention can be realized as a computer program product having program code, and the program code operates to execute one of the above-described methods when the computer program product is executed on a computer. The program code can be stored, for example, in a machine-readable carrier.

[0069] Other embodiments include a computer program for executing one of the methods described in the present specification, which is stored in a machine-readable carrier.

[0070] Therefore, in other words, one embodiment of the method of the present invention is a computer program having program code for executing one of the methods described in the present specification when the computer program is executed on a computer.

[0071] Therefore, a further embodiment of the method of the present invention is a data carrier (or digital storage medium, or computer-readable medium) including a recorded computer program for executing one of the methods described in the present specification.

[0072] A further embodiment includes a processing means, such as a computer or a programmable logic device, configured or adapted to execute one of the methods described in the present specification.

[0073] A further embodiment includes a computer installed with a computer program for executing one of the methods described in the present specification.

[0074] In some embodiments, a programmable logic device (e.g., a field programmable gate array) may be used to execute some or all of the functions in the methods described in the present specification. In some embodiments, the field programmable gate array may cooperate with a microprocessor to execute one of the methods described in the present specification. Generally, the method may be executed by any hardware device.

[0075] Each of the above embodiments merely illustrates the principles of the present invention. It is understood that those skilled in the art will be obvious to make changes and modifications to the configurations and details described in the specification of the present application. Therefore, it is intended to be limited only by the appended claims, rather than by the specific details presented as the description and explanation of each embodiment in the specification of the present application.

[0076] [References] [1] Mallaiyan Sathiaseelan, M. A., Paradis, O. P., Taheri, S., & Asadizanjani, N. (2021). Why Is Deep Learning Challenging for Printed Circuit Board (PCB) Component Recognition and How Can We Address It?. Cryptography, 5(1), 9. [2] Silva, L. H. D. S., Junior, A. A., Azevedo, G. O., Oliveira, S. C., & Fernandes, B. J. (2021). Estimating Recycling Return of Integrated Circuits Using Computer Vision on Printed Circuit Boards. Applied Sciences, 11(6), 2808. [3] Redmon, J., Divvala, S., Girshick, R., & Farhadi, A. (2016). You only look once: Unified, real-time object detection. In Proceedings of the IEEE conference on computer vision and pattern recognition (pp. 779-788). [4] Tan, M., Pang, R., & Le, Q. V. (2020). Efficientdet: Scalable and efficient object detection. In Proceedings of the IEEE / CVF conference on computer vision and pattern recognition (pp. 10781-10790).

Claims

1. An apparatus (100) for assigning a material value score (110) to a printed circuit board waste (200) or a portion (210) thereof, configured to determine the material value score (110) and to perform the determination based on a dual energy or spectral X-ray image (122) of the printed circuit board waste (200) or a portion (210) thereof, wherein the performance of the determination comprises applying the dual energy or spectral X-ray image (122) to a machine learning module (132) to obtain the type and size of components (220) of the printed circuit board waste (200) or a portion (210) thereof, and determining the material value score (110) based on the type and size of the components (220).

2. The apparatus according to claim 1, wherein the machine learning module (132) is trained based on a dataset of annotated components (612), and the annotated components (612) represent annotations of representative components of a WPCB (200).

3. The apparatus (100) according to claim 2, wherein the representative components of the WPCB (200) are selected according to the value of the components of the WPCB (200) that are the subject of a recycling process.

4. The apparatus (100) according to any one of claims 1 to 3, further configured to perform the determination of the material value score based on one or more metallurgical processes for processing the printed circuit board waste (200) or a portion (210) thereof.

5. The apparatus (100) according to any one of claims 1 to 3, further configured to perform the determination of the material value score based on one or more metallurgical processes for processing the printed circuit board waste (200) or a portion (210) thereof, and to reduce the material value score if one or more of the components (220) impede the one or more metallurgical processes.

6. The apparatus (100) according to any one of claims 1 to 3, configured to obtain the material value score (110) by simultaneously analyzing both sides (230, 240) of the printed circuit board waste (200).

7. The apparatus (100) according to any one of claims 1 to 3, wherein the material value score (110) depends on the content of one or more predetermined materials to be recovered from the printed circuit board waste (200). **Claim 8** A method (300) for assigning a material value score (110) to a printed circuit board waste (200) or a part (210) thereof, the method comprising determining (320) the material value score (110), the determination (320) being performed based on a dual-energy or spectral X-ray image (122) of the printed circuit board waste (200) or a part (210) thereof, wherein the performing comprises applying the dual-energy or spectral X-ray image (122) to a machine learning module (132) to obtain the type and size of components (220) of the printed circuit board waste (200) or a part (210) thereof, and determining the material value score (110) based on the type and size of the components (220), the method (300) being executed using a hardware device, a computer, or a combination of a hardware device and a computer. **Claim 9** A system (400) for sorting printed circuit board waste (200), comprising the apparatus (100) for assignment according to any one of claims 1 to 3, and a sorting device (410) for classifying a plurality of printed circuit board wastes (200) into two or more grades (420) for the printed circuit board waste (200) according to the material value scores (110) assigned to each of the plurality of printed circuit board wastes (200) by the apparatus (100) for assignment. **Claim 10** A method (500) for sorting printed circuit board waste (200), comprising the method (300) for assignment according to claim 8, and classifying (510) a plurality of printed circuit board wastes (200) into two or more grades (420) for the printed circuit board waste (200) according to the material value scores (110) assigned to each of the plurality of printed circuit board wastes (200) by the method (300) for assignment.

Images