A renewable power material recovery system and method

Abstract

This invention belongs to the field of waste electrical equipment recycling control, and provides a renewable electrical equipment recycling system and method. The renewable electrical equipment recycling system includes a dismantling dispatch center, waste material storage equipment, recycling storage and logistics equipment, and several dismantling production lines. The dismantling dispatch center generates matching waste electrical equipment outbound loading instructions and issues them to the waste material storage equipment based on the waste electrical equipment dismantling plan, combined with waste electrical equipment storage information and the location of each dismantling production line. After the dismantling production lines complete the dismantling process, they generate recycling storage and logistics equipment dispatch instructions based on the current status information of the recycling storage and logistics equipment, the type of dismantled material, and its warehouse location. The recycling storage and logistics equipment receives and responds to the recycling storage and logistics equipment dispatch instructions, transporting the corresponding recycled material to the corresponding warehouse.

Description

Technical Field

[0001] This invention belongs to the field of waste electrical materials recycling control, and particularly relates to a renewable electrical materials recycling system and method. Background Technology

[0002] The statements in this section are merely background information related to the present invention and do not necessarily constitute prior art.

[0003] Waste electrical equipment is diverse, including but not limited to waste cables, transformers, and electricity meters. To meet the requirements for the reuse of waste electrical equipment, it is currently dismantled, and usable materials are recycled. However, traditional waste electrical equipment dismantling and recycling technologies have the following technical problems:

[0004] (1) The dismantling control center can only dismantle and recycle one type of waste electrical materials, and is not applicable to the dismantling and recycling control of multiple types of waste electrical materials.

[0005] (2) The traditional dismantling process of waste electrical materials is complicated, and the dismantling and storage of the various parts after dismantling are relatively chaotic, which is not conducive to the later recycling and other treatments.

[0006] (3) In the process of transporting recycled materials, there are problems such as information asymmetry and mismatch between vehicles and vehicles during loading and transportation, which increases logistics costs and reduces recycling efficiency. Summary of the Invention

[0007] To address the technical problems mentioned above, this invention provides a renewable energy material recycling system and method, which is applicable to the dismantling and recycling control of various waste energy materials. It also enables the orderly storage of dismantled parts, and ensures information symmetry and route matching between vehicles transporting recycled materials, thereby reducing logistics costs and improving recycling efficiency.

[0008] To achieve the above objectives, the present invention adopts the following technical solution:

[0009] The first aspect of the present invention provides a renewable energy resource recycling system.

[0010] A renewable energy material recycling system includes: a dismantling dispatch center, waste material storage equipment, recycling storage and logistics equipment, and several dismantling production lines;

[0011] The dismantling and dispatching center is used to: generate matching waste power material loading and unloading instructions based on the waste power material dismantling plan, combined with waste power material storage information and the location of each dismantling production line, and issue them to the waste material storage equipment.

[0012] The waste material storage equipment is used to: retrieve the corresponding type of waste electrical materials from the designated warehouse and transport them to the corresponding dismantling production line according to the waste electrical materials loading and unloading instruction;

[0013] The dismantling production line is used to: automatically trigger the dismantling process when the corresponding waste electrical materials are detected to be in place, generate the dismantled materials, and feed back its dismantling progress to the dismantling scheduling center;

[0014] The dismantling scheduling center is also used to: after the dismantling production line completes the dismantling process, generate a scheduling instruction for the recycling storage and logistics equipment based on the current status information of the recycling storage and logistics equipment, the type of dismantled material and its warehouse location;

[0015] The recycling warehousing and logistics equipment is used to: receive and respond to the scheduling instructions of the recycling warehousing and logistics equipment, and transport the corresponding recycled materials to the corresponding warehouse.

[0016] As one implementation method, the dismantling and scheduling center is also used to: predict the amount and type of waste electrical materials generated in the future period based on historical waste electrical material dismantling data and pre-trained machine learning algorithms, and then optimize the waste electrical material dismantling plan.

[0017] As one implementation method, the process by which the dismantling and dispatching center generates dispatching instructions for recycling and warehousing logistics equipment includes:

[0018] After the dismantling production line completes the dismantling process, it identifies the box volume of the dismantled material, retrieves the warehouse location matching the dismantled material, and the current status information of the recycling warehousing and logistics equipment.

[0019] The number of recycling storage and logistics equipment is determined based on the packing volume of the disassembled materials and the conveying capacity of the recycling storage and logistics equipment.

[0020] Prioritize selecting recycling and storage logistics equipment that is currently idle, and then filter all recycling and storage logistics equipment for the current delivery task based on the time closest to the end of the previous delivery task.

[0021] With the goal of minimizing the transport path, and combining the starting point of the disassembled materials' packing and the warehouse location that matches the disassembled materials, a scheduling instruction for recycling and warehousing logistics equipment is generated.

[0022] As one implementation method, in each of the dismantling production lines, automatic loading and unloading target detection is also performed on the corresponding waste electrical materials. The process is as follows:

[0023] Obtain the current process number of the scrap electrical materials and its corresponding actual process image set;

[0024] According to the current process number, retrieve the corresponding target detection model from the set power material and equipment model library, and use the target detection model to identify the actual image set of the process, and identify whether the actual image set of the current process matches the standard image set of the process.

[0025] In this system, each process in the power material and equipment model library is associated with the target detection model through process number;

[0026] The target detection models for each process are trained using standard image sets of each process corresponding to waste electrical materials.

[0027] As one implementation method, the target detection model is the CrossKD target detection model.

[0028] As one implementation method, the standard image set for each process includes standard initial state images of waste electrical materials in each process and standard component images that need to be assembled or disassembled in each process.

[0029] As one implementation, both the standard initial state image and the standard component image include a first sample image with configured labels and a second sample image without configured labels; the CrossKD target detection model for each process is trained using the standard image set of each process, including:

[0030] Arbitrarily select a process standard image set, and based on the first sample image, use the teacher model to perform knowledge distillation training on the student model, and determine the trained student model as the initial target detection model for each process;

[0031] Based on the second sample image, the teacher model is used to perform knowledge distillation training on the initial target detection model of each process, and the trained initial target detection model is determined as the CrossKD target detection model of each process.

[0032] The teacher model is a model pre-trained based on the first sample image, and the knowledge distillation training process includes feature layer distillation training and logic layer distillation training.

[0033] In one implementation, the teacher model includes a first feature extraction module and a first segmentation module;

[0034] The student model includes a second feature extraction module, which, based on the first sample image, performs knowledge distillation training on the student model using the teacher model, including:

[0035] The first sample image is input into the first feature extraction module and the second feature extraction module respectively to obtain the first image features output by the teacher model and the second image features output by the student model;

[0036] The first image features are input into the first segmentation module to obtain the first target detection result output by the teacher model;

[0037] The second image features are input into the first segmentation module to obtain the second target detection result of the teacher model on the student model.

[0038] Calculate the first loss between the first image features and the second image features with respect to feature layer distillation training, and calculate the second loss between the first target detection result and the second target detection result with respect to logic layer distillation training;

[0039] The model parameters of the student model are updated in reverse based on the first loss and the second loss.

[0040] In one implementation, the first sample image is input to the first feature extraction module and the second feature extraction module respectively to obtain the first image features output by the teacher model and the second image features output by the student model, including:

[0041] The first sample image is input to the first feature extraction module and the second feature extraction module respectively to obtain the first feature map output by the first feature extraction module and the second feature map output by the second feature extraction module.

[0042] The pixel values ​​in the first feature map and the second feature map are normalized respectively to obtain the first image feature and the second image feature.

[0043] As one implementation method, the process of updating the model parameters of the student model in reverse based on the first loss and the second loss further includes:

[0044] Determine whether the first loss and the second loss have converged;

[0045] If so, the student model is determined to be trained successfully, and the trained student model is determined as the initial object detection model.

[0046] As one implementation method, the actual image set of the current process corresponds to the standard image set of the process as follows:

[0047] The actual parts that need to be assembled or disassembled in the current process are consistent with the images of the standard parts that need to be assembled or disassembled in the current process, and the actual initial state image of the waste electrical materials in the current process is consistent with the standard initial state of the current process.

[0048] The discrepancy between the actual image set of the current process and the standard image set of the process is as follows:

[0049] The actual parts that need to be assembled or disassembled in the current process are inconsistent with the standard parts images that need to be assembled or disassembled in the current process, or / and the actual initial state image of the waste electrical materials in the current process is inconsistent with the standard initial state of the current process.

[0050] As one implementation method, in the dismantling production line, when the scrap electrical equipment is a scrap transformer, the corresponding dismantling process is as follows:

[0051] Obtain the RGB and depth images of the transformer;

[0052] The upper surface of the transformer is segmented using RGB and depth images to obtain the segmented image;

[0053] Using the segmented image, candidate locations for bolts are determined from a preset list; the information in the preset list includes the transformer type and the bolt location corresponding to the transformer type.

[0054] In an RGB image, a preset bolt position detection model is used to detect bolt positions. The detected bolt positions are compared with candidate positions, and the bolt positions are corrected. The bolt position detection model is a weighted bounding box fusion model.

[0055] Using the corrected bolt position, the three-dimensional coordinates of the bolt are determined in a preset three-dimensional model;

[0056] The three-dimensional coordinates of the bolt are converted into the coordinates of a preset robot end effector to disassemble the bolt, thereby disassembling the transformer cover.

[0057] As one implementation method, in the dismantling production line, when the scrap electrical materials are scrap transformers, the corresponding dismantling process utilizes a long short-term memory artificial neural network to segment the upper surface of the transformer. Specifically, firstly, the depth information is converted into three channels: disparity, surface normal, and height. Then, a recurrent event network is used to extract contextual information in different directions on the upper surface of the transformer and propagates it bidirectionally in both directions. Simultaneously, for the RGB channel information in the image of the upper surface of the transformer, features are extracted using the convolutional structure in the long short-term memory artificial neural network, and interpolation is used to restore the features at each level to the same resolution, and then concatenate them. Finally, a recurrent event network is used to obtain the contextual information.

[0058] As one implementation method, in the dismantling production line, when the waste electrical materials are waste transformers, in the corresponding dismantling process, the point cloud of only the upper surface of the transformer is retained by the connection component marking algorithm and noise removal to obtain the upper surface model of the transformer; by comparing it with the pre-formed upper surface model of the transformer, the 6D pose of the transformer is obtained; using the obtained 6D pose of the transformer, the segmentation result of the upper surface of the transformer is corrected in the two-dimensional RGB image.

[0059] As one implementation method, in the dismantling production line, when the scrap electrical materials are scrap transformers, the corresponding dismantling process involves converting RGB information and depth information into point cloud information to segment the upper surface of the transformer; preprocessing is used to retain only the point cloud of the upper surface of the transformer; the retained point cloud of the upper surface of the transformer is matched with a pre-formed upper surface model of the transformer to obtain the 6D pose of the transformer, wherein the upper surface model of the transformer is a point-pair feature model; and the obtained 6D pose of the transformer is used to correct the segmentation result of the upper surface of the transformer in a two-dimensional RGB image.

[0060] As one implementation method, in the dismantling production line, when the scrap electrical materials are scrap transformers, in the corresponding dismantling process, when determining the candidate position of the bolt, the bolt position of the corresponding model transformer is directly retrieved from the preset list, and the detected bolt position is compared with the candidate position. Specifically, firstly, the candidate position of the bolt of the corresponding model transformer is retrieved from the preset list and matched with the previously detected bolt position. For bolts that were not detected due to external interference, the candidate position is taken as the bolt position. For positions where the distance between the candidate position and the previously detected position is greater than a set threshold, they are considered false detections. For bolts where the distance between the candidate position and the previously detected position is less than the threshold, the detected position is taken as the bolt position.

[0061] As one implementation method, in the dismantling production line, when the waste electrical materials are waste transformers, in the corresponding dismantling process, after the transformer top cover is dismantled, the magnetic attraction point cloud is matched in the three-dimensional transformer top surface point cloud using the transformer 6D pose and the transformer top surface model to obtain the three-dimensional coordinates of the midpoint of the magnetic attraction position; the three-dimensional coordinates of the magnetic attraction position are converted into the tool coordinates of the electromagnetic tooling, guiding the electromagnetic tooling to the magnetic attraction position, automatically attracting the transformer top cover, and lifting the transformer top cover;

[0062] The preset oil pumping robot automatically inserts the end oil pumping tool into the transformer housing to start oil pumping, and automatically stops when the oil pumping is completed.

[0063] The electromagnetic tooling transports the transformer cover to the coil disassembly station, where the robotic arms at both ends automatically use hydraulic shears to separate the connection between the transformer cover and the windings.

[0064] The RGB image of the transformer is segmented. Using the RGB image segmentation results, the three-dimensional coordinates of the yoke and coil positions on the winding are determined from the pre-constructed transformer type and winding model. The three-dimensional coordinates are converted into the coordinates of the preset robot end-effector tool, which guides the robot end-effector clamping fixture to clamp the yoke for disassembly.

[0065] As one implementation method, when the waste electrical materials are waste transformers, the corresponding dismantling production line includes a conveyor line, as well as a weighing mechanism and a barcode scanning mechanism installed on the conveyor line.

[0066] Near the conveyor line are a screw removal robotic arm, an oil extraction robotic arm, a assisted cantilever crane, a coil dismantling robotic arm, a sorting robotic arm, and an automated warehouse; the screw removal robotic arm, the oil extraction robotic arm, the coil dismantling robotic arm, and the sorting robotic arm are all equipped with rotating parts through adjustment mechanisms and connecting rods, and the rotating parts are equipped with connectors for connecting various tools.

[0067] As one implementation method, when the waste electrical materials are waste cables, the corresponding dismantling production line includes: an automatic feeding device, an automatic picking and grabbing device, an automatic dismantling device, and a crushing and screening device arranged sequentially along the main transmission line.

[0068] The automatic feeding device is used to automatically transport waste meters to the inlet of the automatic picking and grabbing device;

[0069] The automatic picking and grabbing device is used to automatically pick up individual waste meters and transport them to the workstation where the automatic dismantling device is located.

[0070] The automatic dismantling device is used to automatically dismantle the waste electricity meters at its workstation into designated components;

[0071] The crushing and screening device is used to crush the disassembled components, separate and screen them according to the properties of recyclable materials, and store them in the corresponding receiving bins.

[0072] As one implementation method, when the waste electrical materials are waste cables, the corresponding dismantling production line includes: an automatic cable feeding device, a primary stripping device, a core separation and feeding device, and an automatic crushing and sorting device arranged sequentially along the main transmission line; a secondary stripping device and an automatic packaging device are arranged sequentially on the branch transmission line connected to the rear end of the core separation and feeding device.

[0073] The automatic cable feeding device is used to divide a pile of cables into individual cables through a stepped transmission method, and then sequentially transmit each individual cable to a primary stripping device.

[0074] The primary stripping device is used to adaptively clamp the cable according to the cable diameter and perform primary stripping to obtain a primary stripping product; the primary stripping product includes inner and outer sheaths, filler, steel armor and wire core.

[0075] The core separation and feeding device is used to transfer and feed the product generated from the first stripping process in layers.

[0076] The secondary stripping device is used to strip the wire core a second time to obtain the metal wire core, which is then conveyed to the automatic packaging device for packaging.

[0077] The automatic crushing and sorting device is used to automatically crush and sort the inner and outer sheaths, fillers, and steel armor.

[0078] As one implementation method, when the waste electrical materials are composite insulators, the corresponding dismantling production line includes: a feeding conveyor line for transporting the insulators to be recycled;

[0079] The insulator fixing mechanism receives insulators from the feeding conveyor line and, in cooperation with the feeding and handling mechanism, fixes the insulators to be recycled.

[0080] The cutting robot, using a cutting tool carried at its end, cuts and separates the shed part of the insulator along the axial direction of the insulator to be recycled, with the cooperation of the insulator fixing mechanism, and then cuts and separates the fiberglass rod and metal head in the insulator after the shed is separated.

[0081] The sorting and conveying line uses sorting robots to sort the separated umbrella skirts, fiberglass rods, and metal heads into the corresponding recycling containers.

[0082] A second aspect of the present invention provides a method for recycling renewable energy resources.

[0083] A method for recycling renewable energy resources includes:

[0084] Based on the scrap electrical materials dismantling plan, combined with the scrap electrical materials storage information and the location of each dismantling production line, the dismantling dispatch center generates matching scrap electrical materials outbound loading instructions and issues them to the scrap materials storage equipment.

[0085] According to the instructions for loading and unloading waste electrical materials, the waste material storage equipment will take out the corresponding type of waste electrical materials from the designated warehouse and transport them to the corresponding dismantling production line.

[0086] When the corresponding waste electrical materials are detected to be in place, the dismantling process of the dismantling production line is automatically triggered, the dismantled materials are generated, and the dismantling progress is fed back to the dismantling dispatch center.

[0087] After the dismantling production line completes the dismantling process, the dismantling dispatch center also generates a dispatch instruction for the recycling storage and logistics equipment based on the current status information of the recycling storage and logistics equipment, the type of dismantled materials and their warehouse location.

[0088] The recycling warehousing and logistics equipment receives and responds to the scheduling instructions of the recycling warehousing and logistics equipment, and transports the corresponding recycled materials to the corresponding warehouse.

[0089] The beneficial effects of this invention are:

[0090] (1) This invention innovatively provides a control technology for the dismantling and recycling of waste electrical materials. Based on the dismantling plan of waste electrical materials, the dismantling production line and the status information of recycling storage and logistics equipment, the dismantling scheduling center generates a corresponding scheduling dismantling strategy, which realizes the orderly storage of some dismantled parts. Through the dynamic and static combined scheduling planning of recycling storage and logistics equipment, logistics costs are reduced and recycling efficiency is improved.

[0091] (2) This invention innovatively provides an automatic loading and unloading target detection technology for waste power materials. By recognizing the actual image set of the process, it can identify whether the actual image set of the current process matches the standard image set of the process. This solves the shortcomings of the existing technology in terms of recognition accuracy, computing resource requirements and environmental adaptability. It realizes the target detection of the complex and diverse structure and changing operating environment of waste power materials, and improves the accuracy and stability of waste power material detection.

[0092] (3) This invention innovatively proposes an automatic transformer disassembly strategy based on a pre-stored list to determine bolt positions. It utilizes images segmented by YOLACT neural network instances and matched with pre-formed 6D pose matching of targets on the transformer to determine candidate bolt positions from a pre-stored list including transformer types. In the RGB image, a pre-stored bolt position detection model is used to detect bolt positions. The detected bolt positions are compared with the candidate positions to correct the bolt positions. Using the corrected bolt positions, the three-dimensional coordinates of the bolts are determined in a pre-stored three-dimensional model. The bolt positions of the corresponding transformer models are directly retrieved from the pre-stored list, and the bolt positions are corrected by comparing the detected bolt positions with the candidate positions. While ensuring the accuracy of bolt positions, this approach avoids excessive data acquisition and updates during the judgment and correction process, reducing the impact on control speed and efficiency.

[0093] (4) This invention creatively proposes a human-machine collaborative method for dismantling batch waste transformers. During the transportation of waste transformers on the conveyor line, the weighing mechanism and the scanning mechanism are used to weigh and scan information respectively. The waste transformers are dismantled by the screw dismantling robot arm, the oil extraction robot arm, the coil dismantling robot arm and the sorting robot arm, which solves the problem of dismantling large batches of waste transformers and improves the dismantling efficiency of batches of waste transformers.

[0094] Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention. Attached Figure Description

[0095] The accompanying drawings, which form part of this invention, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an improper limitation of the invention.

[0096] Figure 1 This is a schematic diagram of a renewable energy material recycling system according to an embodiment of the present invention;

[0097] Figure 2 This is a teacher-student guidance framework diagram of the CrossKD object detection model according to an embodiment of the present invention;

[0098] Figure 3 This is a flowchart illustrating the training process of the CrossKD object detection model according to an embodiment of the present invention.

[0099] Figure 4 This is a schematic diagram of a waste transformer dismantling production line according to an embodiment of the present invention;

[0100] Figure 5 This is a schematic diagram of a waste electricity meter dismantling production line according to an embodiment of the present invention;

[0101] Figure 6 This is a schematic diagram of the structure of an automatic dismantling device for a waste electricity meter dismantling production line according to an embodiment of the present invention;

[0102] Figure 7 This is a schematic diagram of the crushing and screening device structure of a waste electricity meter dismantling production line according to an embodiment of the present invention.

[0103] Among them, 1-1, conveyor line; 1-2, weighing mechanism; 1-3, barcode scanning mechanism; 1-4, screw removal robotic arm; 1-5, oil extraction robotic arm; 1-6, assisted cantilever crane; 1-7, coil disassembly robotic arm;

[0104] 2-1 Main conveyor line; 2-2 Automatic picking and gripping device; 2-3 Automatic dismantling device; 2-4 Material sorting operating table; 2-5 Double-layer conveyor line; 2-6 Crushing and screening device; 2-7 Receiving box; 2-8 First spindle milling cutter; 2-9 Coordinate robot arm; 2-10 Vision system; 2-11 Second spindle milling cutter; 2-12 Shredder; 2-13 Magnetic separator; 2-14 Air separator; 2-15 Eddy current separator; 2-16 Conveyor belt. Detailed Implementation

[0105] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0106] It should be noted that the following detailed description is illustrative and intended to provide further explanation of the invention. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention pertains.

[0107] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0108] Figure 1 This is a schematic diagram of a renewable energy resource recycling system according to an embodiment of the present invention. (Combined with...) Figure 1 An embodiment of the present invention provides a renewable energy material recycling system, comprising: a dismantling dispatch center, waste material storage equipment, recycling storage and logistics equipment, and several dismantling production lines;

[0109] The dismantling and dispatching center is used to: generate matching waste power material loading and unloading instructions based on the waste power material dismantling plan, combined with waste power material storage information and the location of each dismantling production line, and issue them to the waste material storage equipment.

[0110] The waste material storage equipment is used to: retrieve the corresponding type of waste electrical materials from the designated warehouse and transport them to the corresponding dismantling production line according to the waste electrical materials loading and unloading instruction;

[0111] The dismantling production line is used to: automatically trigger the dismantling process when the corresponding waste electrical materials are detected to be in place, generate the dismantled materials, and feed back its dismantling progress to the dismantling scheduling center;

[0112] The dismantling scheduling center is also used to: after the dismantling production line completes the dismantling process, generate a scheduling instruction for the recycling storage and logistics equipment based on the current status information of the recycling storage and logistics equipment, the type of dismantled material and its warehouse location;

[0113] The recycling warehousing and logistics equipment is used to: receive and respond to the scheduling instructions of the recycling warehousing and logistics equipment, and transport the corresponding recycled materials to the corresponding warehouse.

[0114] In the specific implementation process, the dismantling and scheduling center is also used to: predict the amount and type of waste electrical materials generated in the future period based on historical waste electrical material dismantling data and pre-trained machine learning algorithms, and then optimize the waste electrical material dismantling plan.

[0115] The machine learning algorithm here can be implemented using existing neural network architecture models, such as CNN, which will not be described in detail here.

[0116] In one or more embodiments, the process by which the dismantling and scheduling center generates scheduling instructions for the recycling and storage logistics equipment includes:

[0117] After the dismantling production line completes the dismantling process, it identifies the box volume of the dismantled material, retrieves the warehouse location matching the dismantled material, and the current status information of the recycling warehousing and logistics equipment.

[0118] The number of recycling storage and logistics equipment is determined based on the packing volume of the disassembled materials and the conveying capacity of the recycling storage and logistics equipment.

[0119] Prioritize selecting recycling and storage logistics equipment that is currently idle, and then filter all recycling and storage logistics equipment for the current delivery task based on the time closest to the end of the previous delivery task.

[0120] With the goal of minimizing the transport path, and combining the starting point of the disassembled materials' packing and the warehouse location that matches the disassembled materials, a scheduling instruction for recycling and warehousing logistics equipment is generated.

[0121] By designing a packaging scheme for recycling warehousing and logistics equipment, as well as a static and dynamic scheduling plan for such equipment, problems such as information symmetry and route matching between the equipment and the warehousing and logistics facilities can be solved, thereby reducing logistics costs and improving recycling efficiency.

[0122] As one implementation method, in each of the dismantling production lines, automatic loading and unloading target detection is also performed on the corresponding waste electrical materials. The process is as follows:

[0123] S101, Obtain the current process number of the waste electrical materials and its corresponding actual process image set.

[0124] In this embodiment, the actual process image set of the current process of the waste electrical materials includes: the actual initial state image of the waste electrical materials in the current process and the images of the actual parts that need to be assembled or disassembled in the current process.

[0125] S102, retrieve the corresponding target detection model from the model library according to the number of the current process, and use the target detection model to identify the actual image set of the process, and identify whether the actual image set of the current process matches the standard image set of the process.

[0126] Each process in the model library is associated with a target detection model through a process number;

[0127] The target detection models for each process are trained using standard image sets of each process corresponding to waste electrical materials.

[0128] In the specific implementation process, the actual image set of the current process matches the standard image set of the process as follows:

[0129] The actual parts that need to be assembled or disassembled in the current process are consistent with the images of the standard parts that need to be assembled or disassembled in the current process, and the actual initial state image of the waste electrical materials in the current process is consistent with the standard initial state of the current process.

[0130] The discrepancy between the actual image set of the current process and the standard image set of the process is as follows:

[0131] The actual parts that need to be assembled or disassembled in the current process are inconsistent with the standard parts images that need to be assembled or disassembled in the current process, or / and the actual initial state image of the waste electrical materials in the current process is inconsistent with the standard initial state of the current process.

[0132] In one or more embodiments, a prompt signal is also output based on the judgment result regarding the actual initial state of the waste electrical materials in the current process and whether the actual components that need to be assembled or disassembled are detected as normal.

[0133] If the actual image set of the current process matches the standard image set of the process, output a prompt signal indicating that the actual initial state of the waste electrical materials in the current process and the actual parts that need to be assembled or disassembled are normal.

[0134] If the actual image set of the current process does not match the standard image set of the process, output alarm signals regarding the actual initial state of the waste electrical materials in the current process and the abnormal detection of actual parts that need to be assembled or disassembled.

[0135] In this embodiment, the target detection model is the CrossKD target detection model, such as... Figure 2 As shown.

[0136] The standard image set for each process includes standard initial state images of waste electrical materials in each process and standard component images that need to be assembled or disassembled in each process.

[0137] Specifically, the standard initial state image and standard component image both include a first sample image with configured labels and a second sample image without configured labels; the CrossKD target detection model for each process is trained using the standard image set of each process, such as... Figure 3 As shown, it includes:

[0138] Arbitrarily select a process standard image set, and based on the first sample image, use the teacher model to perform knowledge distillation training on the student model, and determine the trained student model as the initial target detection model for each process;

[0139] Based on the second sample image, the teacher model is used to perform knowledge distillation training on the initial target detection model of each process, and the trained initial target detection model is determined as the CrossKD target detection model of each process.

[0140] The teacher model is a model pre-trained based on the first sample image, and the knowledge distillation training process includes feature layer distillation training and logic layer distillation training.

[0141] It should be noted that the acquisition module of this invention can simultaneously acquire a first sample image with configured labels and a second sample image without configured labels. Based on the first sample image, the teacher model is used to perform knowledge distillation training on the student model, and the trained student model is determined as the initial object detection model. Based on the second sample image, the teacher model is used to perform knowledge distillation training on the initial object detection model to obtain the object detection model. By using knowledge distillation to train the object detection model, the feature extraction capability of the object detection model can be improved, ensuring the accuracy of object detection while ensuring successful application on in-vehicle platforms. Furthermore, during the training process, a semi-supervised training method is adopted to accurately utilize unlabeled image data to improve model performance, eliminating the need for extensive image data annotation and reducing the annotation cost of training samples.

[0142] Specifically, the standard image set for each process includes multiple standard initial state images and multiple standard component images. The number of standard initial state images and standard component images can be selected according to actual needs. In this invention, the number of standard initial state images and standard component images is preferably 200. In the 200 standard initial state images, the shooting angle of each standard initial state image is different.

[0143] In this sample image set to a preset ratio, labeled images are designated as the first sample image, while the remaining unlabeled images are designated as the second sample image. Preferably, the preset ratio is 50%. After acquiring the sample images, a small number of sample images can be pre-labeled, specifically by labeling each pixel in the sample image with a classification label, such as the number of each process step. In the 200 standard component images, the shooting angle of each standard component image varies. The images with the preset ratio are labeled images and designated as the first sample image, while the remaining unlabeled images are designated as the second sample image.

[0144] Preferably, the preset ratio is 50%. After acquiring sample images, a small number of sample images can be pre-labeled. Specifically, each pixel in the sample image can be labeled with a classification label, such as the name of each component. The labeled sample images are then designated as the first sample images, and the remaining unlabeled sample images are designated as the second sample images. In the subsequent training of the object detection model, semi-supervised training can be performed based on the first and second sample images. This accurately utilizes unlabeled image data to improve model performance without requiring extensive image data labeling, thus reducing the labeling cost of training samples.

[0145] Specifically, the teacher model includes a first feature extraction module and a first segmentation module;

[0146] The student model includes a second feature extraction module, which, based on the first sample image, performs knowledge distillation training on the student model using the teacher model, including:

[0147] The first sample image is input into the first feature extraction module and the second feature extraction module respectively to obtain the first image features output by the teacher model and the second image features output by the student model;

[0148] The first image features are input into the first segmentation module to obtain the first target detection result output by the teacher model;

[0149] The second image features are input into the first segmentation module to obtain the second target detection result of the teacher model on the student model.

[0150] Calculate the first loss between the first image features and the second image features with respect to feature layer distillation training, and calculate the second loss between the first target detection result and the second target detection result with respect to logic layer distillation training;

[0151] The model parameters of the student model are updated in reverse based on the first loss and the second loss.

[0152] In some optional embodiments, the CWD method can be used to input the first sample image into the first feature extraction module in the teacher model and the second feature extraction module in the student model, respectively. The first feature extraction module outputs multiple first feature maps at a preset number of feature layers. Then, softmax normalization is performed on each pixel value in each of the multiple first feature maps to distribute each pixel value between 0 and 1, resulting in the first image feature T1. Simultaneously, the second feature extraction module outputs multiple second feature maps at a preset number of feature layers. Then, softmax normalization is performed on each pixel value in each of the multiple second feature maps to distribute each pixel value between 0 and 1, resulting in the second image feature S1. The preset number of feature layers can be set according to the actual application scenario, such as 256 layers, 512 layers, etc.

[0153] The CrossKD method can be used. The teacher model propagates forward, inputting the first image features into the first segmentation module to generate a class probability map for each pixel, which is the first object detection result. The second image features output by the student model are extracted and input into the segmentation head (i.e., the first segmentation module) of the teacher model. The first segmentation module processes the second image features to the desired number of classes, for example, 19 classes. The shape of the feature vector at this time is [2,256,224,224]. After processing by the segmentation head, the vector shape is [2,19,224,224]. 2 is the batch size, 19 is the number of channels, and 224 is the width and height of the image. Finally, the class probability map of each pixel output by the first segmentation module is obtained, which is the second object detection result of the teacher model on the student model.

[0154] Specifically, the first sample image is input into the first feature extraction module and the second feature extraction module respectively to obtain the first image features output by the teacher model and the second image features output by the student model, including:

[0155] The first sample image is input to the first feature extraction module and the second feature extraction module respectively to obtain the first feature map output by the first feature extraction module and the second feature map output by the second feature extraction module.

[0156] The pixel values ​​in the first feature map and the second feature map are normalized respectively to obtain the first image feature and the second image feature.

[0157] In some alternative embodiments, the DIST method can be used. After obtaining the first and second image features, the Pearson correlation coefficient can be used as the loss function to calculate the first loss between the second and first image features. The student model parameters are then updated based on backpropagation using the first loss. Here, the first loss represents the completion degree of feature layer distillation training under labeled data training; the closer the first loss is to 0, the higher the completion degree of feature layer distillation training. After obtaining the first and second object detection results, the Pearson correlation coefficient can be used as the loss function to calculate the second loss between the second and first object detection results. The student model parameters are then updated based on backpropagation using the second loss. Here, the second loss represents the completion degree of logic layer distillation training under labeled data training; the closer the second loss is to 0, the higher the completion degree of logic layer distillation training.

[0158] The process of updating the model parameters of the student model in reverse based on the first loss and the second loss also includes:

[0159] Determine whether the first loss and the second loss have converged;

[0160] If so, the student model is determined to be trained successfully, and the trained student model is determined as the initial object detection model.

[0161] In some alternative embodiments, the teacher model includes a first feature extraction module and a first segmentation module, the initial object detection model includes a third feature extraction module and a second segmentation module, and the knowledge distillation training process performed on the initial object detection model using the teacher model based on the second sample image includes:

[0162] The second sample image is input into the first feature extraction module and the third feature extraction module respectively to obtain the third image feature output by the teacher model and the fourth image feature output by the initial target detection model;

[0163] The third image features are input into the first segmentation module to obtain the third target detection result output by the teacher model;

[0164] The fourth image feature is input into the second segmentation module to obtain the fourth target detection result output by the initial target detection model;

[0165] Calculate the third loss between the third image features and the fourth image features with respect to feature layer distillation training, and calculate the fourth loss between the third target detection result and the fourth target detection result with respect to logic layer distillation training;

[0166] The model parameters of the initial target detection model are updated in reverse based on the third loss and the fourth loss.

[0167] Specifically, the method further includes:

[0168] The DIST method can be used. After obtaining the third and fourth image features, the Pearson correlation coefficient can be used as the loss function to calculate the third loss between the fourth and third image features. The initial object detection model parameters are then updated based on backpropagation using this third loss. The third loss represents the completion degree of feature layer distillation training under unlabeled data training; the closer the third loss is to 0, the higher the completion degree of feature layer distillation training. Similarly, after obtaining the third and fourth object detection results, the Pearson correlation coefficient can be used as the loss function to calculate the fourth loss between the fourth and third object detection results. The initial object detection model parameters are then updated based on backpropagation using this fourth loss. The fourth loss represents the completion degree of logistic layer distillation training under unlabeled data training; the closer the fourth loss is to 0, the higher the completion degree of logistic layer distillation training.

[0169] Accordingly, when updating the student model parameters through backpropagation, the model gradient corresponding to the third or fourth loss can be calculated based on the backpropagation algorithm, and the model parameters of the initial object detection model can be updated according to the model gradient. In specific application scenarios, the number of training epochs of the initial object detection model can be set according to the data size of the second sample image. After determining that the initial object detection model has reached the required number of training epochs, it can be determined whether the third and fourth losses have converged.

[0170] In summary, the technical solution in this application, by utilizing knowledge distillation to train the target detection model, can improve the feature extraction capability of the target detection model. This enables the target detection model to achieve performance comparable to larger models, ensuring the accuracy of target detection, while also allowing for successful application in target monitoring schemes for automated loading and unloading of waste electrical materials. Furthermore, during training, a semi-supervised training method is employed, precisely utilizing unlabeled image data to improve model performance, eliminating the need for extensive image data annotation and reducing the cost of labeling training samples.

[0171] When the third and fourth losses reach convergence, the initial target detection model can be considered to have completed training, and this trained initial target detection model is determined as the final target detection model. The preset thresholds for feature layer distillation training and logic layer distillation training can be the same or different, specifically values ​​between 0 and 1. The closer the preset threshold is to 0, the higher the required training accuracy for the corresponding feature layer distillation training or logic layer distillation training. Specific values ​​can be set according to the actual application scenario and are not specifically limited here.

[0172] Specifically, the step of inputting the second sample image into the first feature extraction module and the third feature extraction module respectively to obtain the third image features output by the teacher model and the fourth image features output by the initial target detection model includes:

[0173] The second sample image is input into the first feature extraction module and the third feature extraction module respectively to obtain the third feature map output by the first feature extraction module and the fourth feature map output by the third feature extraction module.

[0174] The pixel values ​​in the third feature map and the fourth feature map are normalized respectively to obtain the third image feature and the fourth image feature.

[0175] Specifically, the method further includes:

[0176] Determine whether the third loss and the fourth loss have converged; if so, determine that the initial target detection model has been trained and determine the trained initial target detection model as the target detection model.

[0177] To address the problem of target conflict, this invention proposes a cross-head knowledge distillation method (CrossKD), aiming to optimize the knowledge transfer process and enhance the performance of the student model in the automated loading and unloading task of waste electrical materials. The core idea of ​​CrossKD is to generate cross-head predictions by passing the intermediate feature maps of the student model to the detection head of the teacher model. This method increases the consistency between the teacher and student models, effectively reducing the impact of target conflict.

[0178] Specifically, for dense detectors (such as RetinaNet), their detector head consists of a series of convolutional layers. Assuming the detector head has a total of n convolutional layers, the first n-1 layers generate intermediate feature maps f. i The final layer generates the prediction result p. CrossKD operates by passing the feature maps from the intermediate layers of the student model to the corresponding layers of the teacher's detection head; for example, it passes the i-th layer feature map of the student model. Passed to the teacher at the (i+1)th convolutional layer This generates cross-head prediction. This process can be represented as:

[0179]

[0180] Among them, S(·) and These are the region selection principle and the normalization factor. This loss calculation method avoids the trap of direct teacher-student prediction discrepancies, allowing some of the student's detection heads to better optimize for the real target. This distillation method is particularly suitable for complex task scenarios, such as the automated loading, unloading, and dismantling of waste electrical materials, because it can effectively handle multi-target and multi-scale detection requirements.

[0181] By employing this cross-head prediction approach, CrossKD effectively reduces the inconsistency between teacher and student models, thereby mitigating the negative impact of goal conflict on distillation results. Compared to traditional prediction imitation methods, CrossKD's uniqueness lies in the fact that it does not simply minimize the direct prediction discrepancies between teacher and student models. Instead, it enhances the collaborative work between teacher and student models through the sharing of cross-head features, enabling the student model to better learn the deeper features and prediction strategies of the teacher model.

[0182] CrossKD is particularly well-suited for complex task scenarios, such as the automated loading, unloading, and dismantling of waste electrical materials. These tasks typically involve multi-target and multi-scale detection requirements, coupled with challenges related to diverse equipment and complex environments. By sharing feature maps, CrossKD can better handle multi-target and multi-scale detection needs, enhancing the model's generalization ability and robustness.

[0183] Furthermore, since the automated handling of waste electrical materials requires high-precision target positioning and classification, CrossKD demonstrates significant performance improvements in such scenarios. It effectively reduces prediction errors caused by target conflicts, thereby improving the accuracy and efficiency of the system in practical applications.

[0184] The overall training loss function consists of a weighted sum of the detection loss and the distillation loss:

[0185]

[0186] in, and Let represent the classification loss and regression loss, respectively, calculated between the student model's prediction and the true target. The additional CrossKD loss... and Used for distillation between cross-head prediction and teacher prediction.

[0187] In the automated loading and unloading of scrap electrical materials, classification loss can be handled using mass focus loss for soft label prediction, while regression loss can be achieved using GIoU loss or KL divergence, depending on the detector's regression form. Through CrossKD optimization, the model can more accurately and efficiently detect, locate, and classify transformers, thereby improving the accuracy and robustness of the automated loading and unloading system.

[0188] In automated loading and unloading tasks for waste electrical equipment, the classification loss can utilize Quality Focal Loss to handle soft-label prediction. This loss function better balances the ratio between positive and negative samples, making it particularly suitable for dense detection tasks. Regression losses, depending on the detector's regression pattern, can employ GIoU loss or KL divergence, among others. These loss functions provide more accurate bounding box adjustments during regression, improving the model's prediction accuracy for target locations.

[0189] CrossKD's distillation loss aligns with the teacher model's predictions through cross-head predictions, thus avoiding the pitfall of direct teacher-student prediction discrepancies in traditional distillation methods. This design better preserves the teacher model's prediction strategy while allowing the student model to more accurately capture the true target through partial head optimization.

[0190] This optimization approach is particularly suitable for multi-target, multi-scale detection in complex scenarios. For example, automated operating systems for waste electrical equipment need to accurately detect, locate, and classify targets under different environments and equipment types. CrossKD improves the design of distillation losses, enabling the model to better cope with these challenges, thereby enhancing the accuracy and robustness of automated systems.

[0191] In one or more embodiments, in the dismantling production line, when the scrap electrical equipment is a scrap transformer, the corresponding dismantling process is as follows:

[0192] Obtain the RGB and depth images of the transformer;

[0193] The upper surface of the transformer is segmented using RGB and depth images to obtain the segmented image;

[0194] Using the segmented image, candidate locations for bolts are determined from a preset list; the information in the preset list includes the transformer type and the bolt location corresponding to the transformer type.

[0195] In an RGB image, a preset bolt position detection model is used to detect bolt positions. The detected bolt positions are compared with candidate positions, and the bolt positions are corrected. The bolt position detection model is a weighted bounding box fusion model.

[0196] Using the corrected bolt position, the three-dimensional coordinates of the bolt are determined in a preset three-dimensional model;

[0197] The three-dimensional coordinates of the bolt are converted into the coordinates of a preset robot end effector to disassemble the bolt, thereby disassembling the transformer cover.

[0198] In the dismantling production line, when the scrap electrical materials are scrap transformers, the corresponding dismantling process utilizes a Long Short-Term Memory (LSTM) artificial neural network to segment the upper surface of the transformer. Specifically, firstly, depth information is converted into three channels: disparity, surface normal, and height. Then, a recurrent event network is used to extract contextual information in different directions on the upper surface of the transformer and propagates it bidirectionally in both directions. Simultaneously, for the RGB channel information in the image of the upper surface of the transformer, features are extracted using the convolutional structure in the LTM artificial neural network, and interpolation is used to restore the features at each level to the same resolution and concatenate them. Finally, a recurrent event network is used to obtain contextual information.

[0199] In the dismantling production line, when the waste electrical materials are waste transformers, the corresponding dismantling process uses a connection component marking algorithm and noise removal to retain only the point cloud of the upper surface of the transformer, thus obtaining the upper surface model of the transformer. By comparing it with the pre-formed upper surface model of the transformer, the 6D pose of the transformer is obtained. Using the obtained 6D pose of the transformer, the segmentation result of the upper surface of the transformer is corrected in a two-dimensional RGB image.

[0200] In the dismantling production line, when the scrap electrical materials are scrap transformers, the corresponding dismantling process involves converting RGB and depth information into point cloud information to segment the upper surface of the transformer; preprocessing is used to retain only the point cloud of the upper surface of the transformer; the retained point cloud of the upper surface of the transformer is matched with a pre-formed upper surface model of the transformer to obtain the 6D pose of the transformer, wherein the upper surface model of the transformer is a point-pair feature model; the obtained 6D pose of the transformer is used to correct the segmentation result of the upper surface of the transformer in a two-dimensional RGB image.

[0201] In the dismantling production line, when the scrap electrical materials are scrap transformers, the corresponding dismantling process involves determining the candidate positions of bolts. The bolt positions of the corresponding transformer model are directly retrieved from a preset list, and the detected bolt positions are compared with the candidate positions. Specifically, the candidate bolt positions of the corresponding transformer model are first retrieved from the preset list and matched with the previously detected bolt positions. For bolts that were not detected due to external interference, the candidate positions are used as their positions. Positions where the distance between the candidate position and the previously detected position is greater than a set threshold are considered false detections. For bolts where the distance between the candidate position and the previously detected position is less than the threshold, the detected position is used as the bolt position.

[0202] In the dismantling production line, when the waste electrical materials are waste transformers, in the corresponding dismantling process, after the transformer top cover is dismantled, the magnetic attraction point cloud is matched in the three-dimensional transformer top surface point cloud using the transformer 6D pose and the transformer top surface model to obtain the three-dimensional coordinates of the midpoint of the magnetic attraction position; the three-dimensional coordinates of the magnetic attraction position are converted into the tool coordinates of the electromagnetic tooling, guiding the electromagnetic tooling to the magnetic attraction position, automatically attracting the transformer top cover, and lifting the transformer top cover;

[0203] The preset oil pumping robot automatically inserts the end oil pumping tool into the transformer housing to start oil pumping, and automatically stops when the oil pumping is completed.

[0204] The electromagnetic tooling transports the transformer cover to the coil disassembly station, where the robotic arms at both ends automatically use hydraulic shears to separate the connection between the transformer cover and the windings.

[0205] The RGB image of the transformer is segmented. Using the RGB image segmentation results, the three-dimensional coordinates of the yoke and coil positions on the winding are determined from the pre-constructed transformer type and winding model. The three-dimensional coordinates are converted into the coordinates of the preset robot end-effector tool, which guides the robot end-effector clamping fixture to clamp the yoke for disassembly.

[0206] In another embodiment, when the scrap electrical equipment is a scrap transformer, such as Figure 4 As shown, the corresponding dismantling production line includes a conveyor line 1-1, and a weighing mechanism 1-2 and a barcode scanning mechanism 1-3 installed on the conveyor line;

[0207] Near the conveyor line 1-1, there are screw removal robotic arms 1-4, oil extraction robotic arms 1-5, assisted cantilever cranes 1-6, coil dismantling robotic arms 1-7, sorting robotic arms, and an automated warehouse. Each of the robotic arms 1-4, 1-5, 1-6, 1-7, and sorting robotic arms is equipped with rotating parts through adjusting mechanisms and connecting rods. The rotating parts are equipped with connectors for connecting various tools.

[0208] Specifically, near the conveyor line, there are screw removal robotic arms 1-4, oil extraction robotic arms 1-5, assisted cantilever cranes 1-6, coil dismantling robotic arms 1-7, sorting robotic arms, and an automated warehouse. This allows for separate operations of casing dismantling, oil extraction, and coil dismantling on a single conveyor line. It eliminates the need to complete the dismantling of one transformer and sort the dismantled parts before proceeding to the next, ensuring that the dismantling of transformers is unaffected. This system is suitable for dismantling large quantities of used transformers. Simultaneously, the weighing mechanism 1-2 and the barcode scanning mechanism 1-3 on conveyor line 1-1, combined with the sorting robotic arms, enable tracking and categorized management of the entire dismantling process for different transformer models. This solves the problem of chaotic management of dismantled parts and facilitates subsequent recycling and other processing.

[0209] Optional, such as Figure 1 As shown, the conveyor line 1-1 includes a track and AGV trolleys mounted on the track. The AGV trolleys are used to transport used transformers to various workstations such as weighing, barcode scanning, screw removal, and oil extraction. Each used transformer has a QR code label on both sides. When a work task is received, the QR code on the transformer is scanned using a mobile terminal. After confirming that it matches the task list, the AGV trolley transports the transformer to the dismantling assembly line.

[0210] Optionally, the track includes a loading conveyor line and a unloading conveyor line; wherein, the loading conveyor line is used for loading the scrap transformers; and the unloading conveyor line is used for unloading the scrap transformers. Optionally, the weighing mechanism 1-2 is used to weigh the dismantled materials; optionally, the weighing mechanism 1-2 can be a weighbridge or a weight scale, etc. The barcode scanning mechanism 1-3 is used to scan the label codes on the scrap transformers and record the information of the dismantled materials.

[0211] like Figures 5-7 As shown, when the waste electrical materials are waste cables, the corresponding dismantling production line includes: an automatic feeding device, an automatic picking and grabbing device 2-2, an automatic dismantling device 2-3, and a crushing and screening device 2-6 arranged sequentially along the main transmission line 2-1.

[0212] The automatic feeding device is used to automatically transport waste meters to the inlet of the automatic picking and grabbing device;

[0213] The automatic picking and grabbing device 2-2 is used to automatically pick up a single waste electricity meter and transport it to the workstation where the automatic dismantling device is located.

[0214] The automatic dismantling device 2-3 is used to automatically dismantle the waste electricity meters at its workstation into set components;

[0215] The crushing and screening devices 2-6 are used to crush the disassembled components, separate and screen them according to the properties of recyclable materials, and classify and store them into the corresponding receiving bins.

[0216] In one or more embodiments, a material sorting station 2-4 and a double-layer conveyor line 2-5 are sequentially arranged after the automatic disassembly device. The material sorting station 2-4 is a manual assistance position where workers remove the circuit board and terminal block from the lower shell, cut the wire between the circuit board and the terminal block with scissors, and place the circuit board into the material basket. Workers then place the upper and lower shells into the lower layer of the double-layer conveyor line 2-5, and place the remaining materials into the upper layer.

[0217] In this embodiment, the main transmission line is a double-sided transmission line.

[0218] In practice, the automatic feeding device can be a forklift.

[0219] Used electricity meters are placed in beam racks. When a production task is assigned, the used electricity meters are transported by forklift to the double-sided conveyor line next to the robot, unpacked, and then transported on the double-sided conveyor line.

[0220] In practice, the automatic picking and gripping device is a loading robotic arm.

[0221] like Figure 6 As shown, the automatic disassembly device includes a main controller, a first spindle milling cutter 2-8, a coordinate robot 2-9, a vision system 2-10, and a second spindle milling cutter 2-11;

[0222] The first spindle end mill 2-8, the coordinate robot 2-9, the vision system 2-10, and the second spindle end mill 2-11 are all connected to the main controller;

[0223] The first spindle milling cutter 2-8 is used to mill the locking screws and tail cover screws of the energy meter under the action of the main controller;

[0224] The main controller is also used to send a signal that the screws have been removed to the coordinate robot 2-9, so as to control the coordinate robot to take out the upper shell and place it on the belt conveyor of the main transmission line;

[0225] The vision system 2-10 is used to determine the material category and model by calling backend database data based on material characteristics, and then visually identify the position of the circuit board card fixing screws and transmit it to the main controller.

[0226] The main controller is used to control the second spindle milling cutter 2-11 to perform screw milling action based on the position of the circuit board card fixing screws identified by vision.

[0227] This allows for the accurate milling away of clips or screws connecting the circuit board to the lower housing, facilitating manual removal of the circuit board from the lower housing.

[0228] like Figure 7 As shown, the crushing and screening device includes a shredder 2-12, a magnetic separator 2-13, an air separator 2-14, and an eddy current separator 2-15 arranged in series; one end of the conveyor belt 2-16 is connected to the output end of the automatic dismantling device, and the other end is connected to the input end of the shredder 2-12.

[0229] The shredder 2-12 is used to shred the receiving housing and terminal block;

[0230] The magnetic separator 2-13 is used to separate iron components from the output material of the shredder;

[0231] The air separator 2-14 is used to separate the light impurities in the material after the iron components have been separated by the action of wind, and the light impurities are put into the receiving box 7, while the remaining material components continue to be conveyed.

[0232] The eddy current separator 2-15 is used to separate copper and plastic from the debris output from the air separator, and discharge them through different discharge ports into the receiving box.

[0233] In one or more embodiments, the shredder is also connected to a start controller, which is connected to a material arrival sensor. The material arrival sensor is used to detect whether there is material at the input port of the shredder and transmit the corresponding signal to the start controller.

[0234] The automatic feeding device is used to automatically transport waste meters to the inlet of the automatic picking and grabbing device;

[0235] The automatic picking and grabbing device is used to automatically pick up individual waste meters and transport them to the workstation where the automatic dismantling device is located.

[0236] The automatic dismantling device is used to automatically dismantle the waste electricity meters at its workstation into designated components;

[0237] The crushing and screening device is used to crush the disassembled components, separate and screen them according to the properties of recyclable materials, and store them in the corresponding receiving bins.

[0238] In one or more embodiments, when the waste electrical materials are waste cables, the corresponding dismantling production line includes: an automatic cable feeding device, a primary stripping device, a core separation and feeding device, and an automatic crushing and sorting device arranged sequentially along the main transmission line; a secondary stripping device and an automatic packaging device are arranged sequentially on the branch transmission line connected to the rear end of the core separation and feeding device.

[0239] The automatic cable feeding device is used to divide a pile of cables into individual cables through a stepped transmission method, and then sequentially transmit each individual cable to a primary stripping device.

[0240] The primary stripping device is used to adaptively clamp the cable according to the cable diameter and perform primary stripping to obtain a primary stripping product; the primary stripping product includes inner and outer sheaths, filler, steel armor and wire core.

[0241] The core separation and feeding device is used to transfer and feed the product generated from the first stripping process in layers.

[0242] The secondary stripping device is used to strip the wire core a second time to obtain the metal wire core, which is then conveyed to the automatic packaging device for packaging.

[0243] The automatic crushing and sorting device is used to automatically crush and sort the inner and outer sheaths, fillers, and steel armor.

[0244] In other embodiments, when the waste electrical materials are composite insulators, the corresponding dismantling production line includes: a feeding conveyor line for transporting the insulators to be recycled;

[0245] The insulator fixing mechanism receives insulators from the feeding conveyor line and, in cooperation with the feeding and handling mechanism, fixes the insulators to be recycled.

[0246] The cutting robot, using a cutting tool carried at its end, cuts and separates the shed part of the insulator along the axial direction of the insulator to be recycled, with the cooperation of the insulator fixing mechanism, and then cuts and separates the fiberglass rod and metal head in the insulator after the shed is separated.

[0247] The sorting and conveying line uses sorting robots to sort the separated umbrella skirts, fiberglass rods, and metal heads into the corresponding recycling containers.

[0248] Based on the above-mentioned renewable energy resource recycling system, the renewable energy resource recycling method includes:

[0249] Step 1: The dismantling dispatch center generates matching waste power material loading instructions based on the waste power material dismantling plan, combined with the waste power material storage information and the location of each dismantling production line, and issues them to the waste material storage equipment.

[0250] Step 2: The waste material storage equipment retrieves the corresponding type of waste electrical materials from the designated warehouse and transports them to the corresponding dismantling production line according to the waste electrical materials outbound and loading instructions.

[0251] Step 3: When the corresponding waste electrical materials are detected to be in place, the dismantling process of the dismantling production line is automatically triggered, the dismantled materials are generated, and the dismantling progress is fed back to the dismantling dispatch center.

[0252] Step 4: After the dismantling production line completes the dismantling process, the dismantling dispatch center will generate a dispatch instruction for the recycling and storage logistics equipment based on the current status information of the recycling and storage logistics equipment, the type of dismantled materials and their warehouse location.

[0253] Step 5: The recycling warehousing and logistics equipment receives and responds to the scheduling instructions of the recycling warehousing and logistics equipment, and transports the corresponding recycled materials to the corresponding warehouse.

[0254] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention. Various modifications and variations can be made to the present invention by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention.

Claims

1. A renewable energy resource recycling system, characterized in that, include: The facility includes a dismantling and dispatching center, waste material storage equipment, recycling and logistics equipment, and several dismantling production lines. The dismantling and dispatching center is used to: generate matching waste power material loading and unloading instructions based on the waste power material dismantling plan, combined with waste power material storage information and the location of each dismantling production line, and issue them to the waste material storage equipment. The waste material storage equipment is used to: retrieve the corresponding type of waste electrical materials from the designated warehouse and transport them to the corresponding dismantling production line according to the waste electrical materials loading and unloading instruction; The dismantling production line is used to: automatically trigger the dismantling process when the corresponding waste electrical materials are detected to be in place, generate the dismantled materials, and feed back its dismantling progress to the dismantling scheduling center; The dismantling scheduling center is also used to: after the dismantling production line completes the dismantling process, generate a scheduling instruction for the recycling storage and logistics equipment based on the current status information of the recycling storage and logistics equipment, the type of dismantled material, and its warehouse location. This includes: after the dismantling production line completes the dismantling process, identifying the box volume of the dismantled material, retrieving the warehouse location matching the dismantled material, and the current status information of the recycling storage and logistics equipment; determining the number of recycling storage and logistics equipment based on the box volume of the dismantled material and the conveying capacity of the recycling storage and logistics equipment; prioritizing the selection of idle recycling storage and logistics equipment, and then filtering all recycling storage and logistics equipment for the current conveying task based on the closest time to the end time of the previous conveying task; and generating a scheduling instruction for the recycling storage and logistics equipment with the shortest conveying path as the objective, combined with the box starting point of the dismantled material and the warehouse location matching the dismantled material. The recycling warehousing and logistics equipment is used to: receive and respond to the scheduling instructions of the recycling warehousing and logistics equipment, and transport the corresponding recycled materials to the corresponding warehouse; In each of the dismantling production lines, automatic loading and unloading target detection is also performed on the corresponding waste electrical materials. According to the number of the current process, the target detection model corresponding to it is retrieved from the set electrical material equipment model library. The target detection model is used to identify the actual image set of the process and to identify whether the actual image set of the current process matches the standard image set of the process.

2. The renewable energy material recycling system as described in claim 1, characterized in that, The dismantling and scheduling center is also used to: predict the amount and type of waste electrical materials generated in the future based on historical waste electrical material dismantling data and pre-trained machine learning algorithms, and then optimize the waste electrical material dismantling plan.

3. The renewable energy material recycling system as described in claim 1, characterized in that, In each of the aforementioned dismantling production lines, the corresponding waste electrical materials undergo automatic loading and unloading target detection. The process is as follows: Obtain the current process number of the scrap electrical materials and its corresponding actual process image set; In this system, each process in the power material and equipment model library is associated with the target detection model through process number; The target detection models for each process are trained using standard image sets of each process corresponding to waste electrical materials.

4. The renewable energy material recycling system as described in claim 3, characterized in that, The target detection model is the CrossKD target detection model.

5. The renewable energy resource recycling system as described in claim 4, characterized in that, The standard image set for each process includes standard initial state images of waste electrical materials in each process and standard component images that need to be assembled or disassembled in each process.

6. The renewable energy resource recycling system as described in claim 5, characterized in that, The standard initial state image and standard component image both include a first sample image with configured labels and a second sample image without configured labels; the CrossKD object detection model for each process is trained using the standard image set of each process, including: Arbitrarily select a process standard image set, and based on the first sample image, use the teacher model to perform knowledge distillation training on the student model, and determine the trained student model as the initial target detection model for each process; Based on the second sample image, the teacher model is used to perform knowledge distillation training on the initial target detection model of each process, and the trained initial target detection model is determined as the CrossKD target detection model of each process. The teacher model is a model pre-trained based on the first sample image, and the knowledge distillation training process includes feature layer distillation training and logic layer distillation training.

7. The renewable energy material recycling system as described in claim 6, characterized in that, The teacher model includes a first feature extraction module and a first segmentation module; The student model includes a second feature extraction module, which, based on the first sample image, performs knowledge distillation training on the student model using the teacher model, including: The first sample image is input into the first feature extraction module and the second feature extraction module respectively to obtain the first image features output by the teacher model and the second image features output by the student model; The first image features are input into the first segmentation module to obtain the first target detection result output by the teacher model; The second image features are input into the first segmentation module to obtain the second target detection result of the teacher model on the student model. Calculate the first loss between the first image features and the second image features with respect to feature layer distillation training, and calculate the second loss between the first target detection result and the second target detection result with respect to logic layer distillation training; The model parameters of the student model are updated in reverse based on the first loss and the second loss.

8. The renewable energy resource recycling system as described in claim 7, characterized in that, The first sample image is input into the first feature extraction module and the second feature extraction module respectively to obtain the first image features output by the teacher model and the second image features output by the student model, including: The first sample image is input to the first feature extraction module and the second feature extraction module respectively to obtain the first feature map output by the first feature extraction module and the second feature map output by the second feature extraction module. The pixel values ​​in the first feature map and the second feature map are normalized respectively to obtain the first image feature and the second image feature.

9. The renewable energy material recycling system as described in claim 7, characterized in that, The process of updating the model parameters of the student model in reverse based on the first loss and the second loss also includes: Determine whether the first loss and the second loss have converged; If so, the student model is determined to be trained successfully, and the trained student model is determined as the initial object detection model.

10. The renewable energy resource recycling system as described in claim 3, characterized in that, The actual image set of the current process matches the standard image set of the process as follows: The actual parts that need to be assembled or disassembled in the current process are consistent with the images of the standard parts that need to be assembled or disassembled in the current process, and the actual initial state image of the waste electrical materials in the current process is consistent with the standard initial state of the current process. The discrepancy between the actual image set of the current process and the standard image set of the process is as follows: The actual parts that need to be assembled or disassembled in the current process are inconsistent with the standard parts images that need to be assembled or disassembled in the current process, or / and the actual initial state image of the waste electrical materials in the current process is inconsistent with the standard initial state of the current process.

11. The renewable energy material recycling system as described in claim 1, characterized in that, In the dismantling production line, when the scrap electrical equipment is a scrap transformer, the corresponding dismantling process is as follows: Obtain the RGB and depth images of the transformer; The upper surface of the transformer is segmented using RGB and depth images to obtain the segmented image; Using the segmented image, candidate locations for bolts are determined from a preset list; the information in the preset list includes the transformer type and the bolt location corresponding to the transformer type. In an RGB image, a preset bolt position detection model is used to detect bolt positions. The detected bolt positions are compared with candidate positions, and the bolt positions are corrected. The bolt position detection model is a weighted bounding box fusion model. Using the corrected bolt position, the three-dimensional coordinates of the bolt are determined in a preset three-dimensional model; The three-dimensional coordinates of the bolt are converted into the coordinates of a preset robot end effector to disassemble the bolt, thereby disassembling the transformer cover.

12. The renewable energy material recycling system as described in claim 11, characterized in that, In the dismantling production line, when the waste electrical materials are waste transformers, the corresponding dismantling process uses a long short-term memory artificial neural network to segment the upper surface of the transformer. Specifically, firstly, the depth information is converted into three channels: parallax, surface normal, and height. Then, a recurrent event network is used to extract contextual information in different directions on the upper surface of the transformer and propagates it bidirectionally in both directions. At the same time, for the RGB channel information in the image of the upper surface of the transformer, the convolutional structure in the long short-term memory artificial neural network is used to extract features, and interpolation is used to restore the features at each level to the same resolution and cascade them. Finally, the context information is obtained using a recurring event network.

13. The renewable energy material recycling system as described in claim 11, characterized in that, In the dismantling production line, when the waste electrical materials are waste transformers, the corresponding dismantling process uses a connection component marking algorithm and noise removal to retain only the point cloud of the upper surface of the transformer, thus obtaining the upper surface model of the transformer. By comparing it with the pre-formed upper surface model of the transformer, the 6D pose of the transformer is obtained. Using the obtained 6D pose of the transformer, the segmentation result of the upper surface of the transformer is corrected in a two-dimensional RGB image.

14. The renewable energy resource recycling system as described in claim 11, characterized in that, In the dismantling production line, when the waste electrical materials are waste transformers, the corresponding dismantling process converts RGB information and depth information into point cloud information to segment the upper surface of the transformer. Preprocessing retains only the point cloud of the upper surface portion of the transformer; The transformer's 6D pose is obtained by matching the point cloud of the preserved upper surface portion of the transformer with a pre-formed upper surface model of the transformer, wherein the upper surface model of the transformer is a point-pair feature model; the obtained 6D pose of the transformer is used to correct the segmentation result of the upper surface of the transformer in a two-dimensional RGB image.

15. The renewable energy resource recycling system as described in claim 11, characterized in that, In the dismantling production line, when the scrap electrical materials are scrap transformers, the corresponding dismantling process involves determining the candidate positions of bolts. The bolt positions of the corresponding transformer model are directly retrieved from a preset list, and the detected bolt positions are compared with the candidate positions. Specifically, the candidate bolt positions of the corresponding transformer model are first retrieved from the preset list and matched with the previously detected bolt positions. For bolts that were not detected due to external interference, the candidate positions are used as their positions. Positions where the distance between the candidate position and the previously detected position is greater than a set threshold are considered false detections. For bolts where the distance between the candidate position and the previously detected position is less than the threshold, the detected position is used as the bolt position.

16. The renewable energy material recycling system as described in claim 11, characterized in that, In the dismantling production line, when the waste electrical materials are waste transformers, in the corresponding dismantling process, after the transformer top cover is dismantled, the magnetic attraction point cloud is matched in the three-dimensional transformer top surface point cloud using the transformer 6D pose and the transformer top surface model to obtain the three-dimensional coordinates of the midpoint of the magnetic attraction position; the three-dimensional coordinates of the magnetic attraction position are converted into the tool coordinates of the electromagnetic tooling, guiding the electromagnetic tooling to the magnetic attraction position, automatically attracting the transformer top cover, and lifting the transformer top cover; The preset oil pumping robot automatically inserts the end oil pumping tool into the transformer housing to start oil pumping, and automatically stops when the oil pumping is completed. The electromagnetic tooling transports the transformer cover to the coil disassembly station, where the robotic arms at both ends automatically use hydraulic shears to separate the connection between the transformer cover and the windings. The RGB image of the transformer is segmented. Using the RGB image segmentation results, the three-dimensional coordinates of the yoke and coil positions on the winding are determined from the pre-constructed transformer type and winding model. The three-dimensional coordinates are converted into the coordinates of the preset robot end-effector tool, which guides the robot end-effector clamping fixture to clamp the yoke for disassembly.

17. The renewable energy material recycling system as described in claim 1, characterized in that, When the waste electrical materials are waste transformers, the corresponding dismantling production line includes a conveyor line, as well as a weighing mechanism and a barcode scanning mechanism installed on the conveyor line. Near the conveyor line are a screw removal robotic arm, an oil extraction robotic arm, a assisted cantilever crane, a coil dismantling robotic arm, a sorting robotic arm, and an automated warehouse; the screw removal robotic arm, the oil extraction robotic arm, the coil dismantling robotic arm, and the sorting robotic arm are all equipped with rotating parts through adjustment mechanisms and connecting rods, and the rotating parts are equipped with connectors for connecting various tools.

18. The renewable energy resource recycling system as described in claim 1, characterized in that, When the waste electrical materials are waste cables, the corresponding dismantling production line includes: an automatic feeding device, an automatic picking and grabbing device, an automatic dismantling device, and a crushing and screening device arranged sequentially along the main transmission line. The automatic feeding device is used to automatically transport waste meters to the inlet of the automatic picking and grabbing device; The automatic picking and grabbing device is used to automatically pick up individual waste meters and transport them to the workstation where the automatic dismantling device is located. The automatic dismantling device is used to automatically dismantle the waste electricity meters at its workstation into designated components; The crushing and screening device is used to crush the disassembled components, separate and screen them according to the properties of recyclable materials, and store them in the corresponding receiving bins.

19. The renewable energy material recycling system as described in claim 1, characterized in that, When the waste electrical materials are waste electricity meters, the corresponding dismantling production line includes: an automatic cable feeding device, a primary stripping device, a wire core separation and feeding device, and an automatic crushing and sorting device arranged sequentially along the main transmission line; a secondary stripping device and an automatic packaging device are arranged sequentially on the branch transmission line connected to the rear end of the wire core separation and feeding device. The automatic cable feeding device is used to divide a pile of cables into individual cables through a stepped transmission method, and then sequentially transmit each individual cable to a primary stripping device. The primary stripping device is used to adaptively clamp the cable according to the cable diameter and perform primary stripping to obtain a primary stripping product; the primary stripping product includes inner and outer sheaths, filler, steel armor and wire core. The core separation and feeding device is used to transfer and feed the product generated from the first stripping process in layers. The secondary stripping device is used to strip the wire core a second time to obtain the metal wire core, which is then conveyed to the automatic packaging device for packaging. The automatic crushing and sorting device is used to automatically crush and sort the inner and outer sheaths, fillers, and steel armor.

20. The renewable energy material recycling system as described in claim 1, characterized in that, When the waste electrical materials are composite insulators, the corresponding dismantling production line includes: a feeding conveyor line for transporting the insulators to be recycled; The insulator fixing mechanism receives insulators from the feeding conveyor line and, in cooperation with the feeding and handling mechanism, fixes the insulators to be recycled. The cutting robot, using a cutting tool carried at its end, cuts and separates the shed part of the insulator along the axial direction of the insulator to be recycled, with the cooperation of the insulator fixing mechanism, and then cuts and separates the fiberglass rod and metal head in the insulator after the shed is separated. The sorting and conveying line uses sorting robots to sort the separated umbrella skirts, fiberglass rods, and metal heads into the corresponding recycling containers.

21. A control method for a renewable energy resource recycling system based on any one of claims 1-20, characterized in that, include: Based on the scrap electrical materials dismantling plan, combined with the scrap electrical materials storage information and the location of each dismantling production line, the dismantling dispatch center generates matching scrap electrical materials outbound loading instructions and issues them to the scrap materials storage equipment. According to the instructions for loading and unloading waste electrical materials, the waste material storage equipment will take out the corresponding type of waste electrical materials from the designated warehouse and transport them to the corresponding dismantling production line. When the corresponding waste electrical materials are detected to be in place, the dismantling process of the dismantling production line is automatically triggered, the dismantled materials are generated, and the dismantling progress is fed back to the dismantling dispatch center. After the dismantling production line completes the dismantling process, the dismantling dispatch center also generates a dispatch instruction for the recycling storage and logistics equipment based on the current status information of the recycling storage and logistics equipment, the type of dismantled materials and their warehouse location. The recycling warehousing and logistics equipment receives and responds to the scheduling instructions of the recycling warehousing and logistics equipment, and transports the corresponding recycled materials to the corresponding warehouse.

Images