Riveting distributed control method and system for intensive bus duct

Abstract

The invention provides a riveting distributed control method and system for an intensive bus duct, and relates to the technical field of riveting control, and the method comprises the steps: enabling a bus duct segment to move to a riveting section on a production line, and after the bus duct segment is fixed, triggering a multispectral industrial camera array, and carrying out the multi-angle real-time image collection; riveting feature points are recognized, and coordinates of distributed riveting reference points are positioned; the cloud end executes a multi-station collaborative scheduling decision and generates a dynamic riveting task distribution instruction; distributed riveting parameters are issued, and multi-thread closed-loop force control operation based on jump riveting is executed; and after the segmented riveting of the bus duct is finished and the bus duct is unlocked and moved out, the riveting quality on-line detection is triggered. The technical problem that in riveting control in the prior art, the positioning and force applying process of riveting points mostly depends on artificial experience or a simple mechanical mode, accurate positioning and real-time force control are lacked, and consequently the riveting quality is unstable is solved.

Description

Technical Field

[0001] This invention relates to the field of riveting control technology, and more specifically to a distributed riveting control method and system for dense busbar trunking. Background Technology

[0002] Busbar trunking is a crucial component of power transmission and distribution systems. It boasts high transmission capacity and space utilization, and is commonly used to connect electrical equipment and power systems. The riveting process is a critical assembly step in the production of busbar trunking, as its quality directly impacts its stability and safety. In traditional riveting processes, the positioning and force application at the riveting points are largely accomplished through manual experience or simple mechanical methods, lacking precise positioning and real-time force control. This imprecise control method introduces variability in the quality of each riveting process, leading to inconsistencies and unreliability over time, making it difficult to meet high production standards. Summary of the Invention

[0003] This application provides a distributed control method and system for riveting of dense bus trunking, which aims to solve the technical problem that in the riveting control of the prior art, the positioning and force application process of the riveting point mostly rely on manual experience or simple mechanical methods, lacking precise positioning and real-time force control, resulting in unstable riveting quality.

[0004] The first aspect disclosed in this application provides a distributed control method for riveting of dense busbar trunking. The method includes: moving busbar trunking segments of a preset length to a riveting section on a production line; fixing them using a hydraulic locking mechanism; triggering a multispectral industrial camera array installed on the side of the riveting section to perform multi-angle real-time image acquisition; identifying riveting feature points based on the multi-angle real-time images; and locating the coordinates of distributed riveting reference points. The cloud platform then performs multi-station collaborative scheduling decisions based on the real-time equipment status of the distributed riveting stations and the coordinates of the distributed riveting reference points, generating... The dynamic riveting task allocation instruction includes a distributed riveting station comprising a fixed riveting array and multiple flexible supplementary riveting units. Based on the dynamic riveting task allocation instruction, the cloud platform guides the distributed riveting station to the distributed riveting reference point of the busbar segment, then sends distributed riveting parameters to the distributed riveting station, controlling the fixed riveting array and multiple flexible supplementary riveting units to perform multi-threaded closed-loop force control operations based on skip riveting. After the busbar segment riveting is completed and the segment is unlocked and removed from the riveting station, online riveting quality detection is triggered.

[0005] The second aspect of this application discloses a distributed riveting control system for dense busbar trunking. The system is used in the aforementioned distributed riveting control method for dense busbar trunking. The system includes: an image acquisition module, used to move busbar trunking segments of a preset length to the riveting section on the production line, and after being fixed by a hydraulic locking mechanism, trigger a multispectral industrial camera array installed on the side of the riveting section to perform multi-angle real-time image acquisition; a feature point recognition module, used to identify riveting feature points based on the multi-angle real-time images and locate the coordinates of distributed riveting reference points; and a scheduling decision module, used to determine the real-time equipment status of the distributed riveting station and the coordinates of the distributed riveting reference points based on the cloud-based system. The system executes multi-station collaborative scheduling decisions and generates dynamic riveting task allocation instructions. The distributed riveting stations include a fixed riveting array and multiple flexible supplementary riveting units. A force control operation execution module guides the distributed riveting stations to the distributed riveting reference points of the busbar segment based on the dynamic riveting task allocation instructions, then sends distributed riveting parameters to the distributed riveting stations, controlling the fixed riveting array and multiple flexible supplementary riveting units to execute multi-threaded closed-loop force control operations based on jump riveting. An online quality detection module triggers online riveting quality detection after the busbar segment riveting is completed and the segment is unlocked and removed.

[0006] One or more technical solutions provided in this application have at least the following beneficial effects: The busbar trunking is segmented and fixed using a hydraulic locking mechanism. A multispectral industrial camera array is triggered at the riveting section to acquire multi-angle images, enabling precise and rapid acquisition of riveting features on the busbar trunking surface, providing accurate data support for subsequent riveting processes. By identifying riveting feature points in the acquired multi-angle real-time images, the coordinates of distributed riveting reference points are accurately located, ensuring that the riveting station and equipment are riveted in the correct positions. Based on the distributed riveting reference point coordinates, the cloud-based system makes multi-station collaborative scheduling decisions based on real-time equipment status. By generating dynamic riveting task allocation instructions, intelligent scheduling of different riveting units can be achieved, enabling… The fixed riveting array and multiple flexible supplementary riveting units work in coordination to improve overall production efficiency. The cloud sends riveting parameters to the fixed riveting array and multiple flexible supplementary riveting units according to dynamic task instructions, controlling them to perform multi-threaded closed-loop force control operations based on jump riveting. Through force control closed-loop technology, it is ensured that the riveting force during the riveting process is always within an optimized control range, thereby guaranteeing riveting quality. When the busbar trough is riveted in sections and removed from the section, online riveting quality detection is automatically triggered. The riveting quality is detected in an automated way, and defects that occur during the riveting process are quickly identified, improving the overall quality controllability of the production process.

[0007] The above description is only an overview of the technical solution of this application. In order to better understand the technical means of this application and to implement it in accordance with the contents of the specification, and to make the above and other objects, features and advantages of this application more obvious and understandable, the following are specific embodiments of this application. Attached Figure Description

[0008] Figure 1 This is a schematic flowchart of a riveting distributed control method for dense busbar trunking provided in an embodiment of this application.

[0009] Figure 2 This is a schematic diagram of a riveting distributed control system for dense busbar trunking provided in an embodiment of this application.

[0010] Figure labeling: Image acquisition module 10, feature point recognition module 20, scheduling decision module 30, force control operation execution module 40, online quality detection module 50. Detailed Implementation

[0011] To further illustrate the technical means and effects of the present invention in achieving its intended purpose, the following detailed description of the specific implementation methods, structures, features, and effects of the present invention, in conjunction with the accompanying drawings and preferred embodiments, is provided below.

[0012] Example 1, as Figure 1 As shown in the figure, this application provides a riveting distributed control method for dense busbar trunking, the method comprising: A100: When the busbar trunking segments of a preset length move to the riveting section on the production line and are fixed by the hydraulic locking mechanism, the multispectral industrial camera array installed on the side of the riveting section is triggered to perform multi-angle real-time image acquisition.

[0013] The busbar trunking consists of multiple pre-set length segments. These segments are transported from the production line to the riveting section via conveyor belts or automated transport equipment. These segments have predetermined dimensions and specifications to ensure precise riveting operations. Upon arrival at the riveting section, a hydraulic locking mechanism secures the segments to the riveting workbench or riveting frame. This mechanism provides stable force and precise positioning, ensuring the segments do not move or deform during riveting, thus guaranteeing the accuracy of the riveting points.

[0014] A multispectral industrial camera array consists of multiple cameras installed at the side of the riveting section or other suitable locations. The camera array begins to synchronously acquire image data based on hardware trigger signals, such as signals indicating the completion of hydraulic fixing or timing signals during the production process. Multi-angle, multispectral image acquisition helps to obtain more detailed riveting feature data from different directions and different spectral ranges.

[0015] A200: Based on the multi-angle real-time images, identify riveting feature points and locate the coordinates of distributed riveting reference points.

[0016] For multispectral images acquired from multiple angles, computer vision and image processing algorithms are used to process the images and identify riveting feature points. These feature points indicate the position, shape, and size of the rivet holes, including the edges, center, and other geometric features related to riveting. Based on the camera array's viewing angle and shooting parameters, the coordinates of distributed riveting reference points are calculated and located. Using 3D reconstruction techniques, such as multi-view geometry or stereo matching methods, the spatial position of each feature point is determined, thus obtaining the accurate coordinates of the riveting reference points. These coordinates serve as a reference for subsequent riveting operations, ensuring that the riveting tools are precisely aligned with the target position.

[0017] A300: The cloud-based system executes multi-station collaborative scheduling decisions based on the real-time equipment status of the distributed riveting stations and the coordinates of the distributed riveting reference points, generating dynamic riveting task allocation instructions. The distributed riveting stations include a fixed riveting array and multiple flexible supplementary riveting units.

[0018] The cloud-based system manages and schedules the entire riveting process. It monitors and collects real-time equipment status data from distributed riveting stations, including current workload, available tools, and equipment health. Based on the real-time equipment status and the coordinates of distributed riveting reference points, it executes multi-station collaborative scheduling decisions. This decision-making process includes calculating the load on each riveting station (e.g., whether it is idle or under maintenance) to ensure reasonable task allocation. Based on the riveting reference point coordinates, tasks are dynamically assigned to different riveting stations. These stations include: fixed riveting arrays for performing high-precision, highly repeatable riveting operations; and multiple flexible, supplementary riveting units for handling riveting points where positional accuracy is difficult to achieve or requires supplementation. Real-time calculations generate riveting task allocation instructions, guiding each riveting unit on how to work collaboratively.

[0019] A400: The cloud platform guides the distributed riveting station to the distributed riveting reference point of the busbar segment according to the dynamic riveting task allocation instruction, and then sends distributed riveting parameters to the distributed riveting station to control the fixed riveting array and multiple flexible supplementary riveting units to perform multi-threaded closed-loop force control operation based on jump riveting.

[0020] Based on the generated dynamic riveting task allocation instructions, the cloud guides the distributed riveting stations to the distributed riveting reference points of the busbar segment. This process ensures that the riveting stations can accurately reach the positions required for the riveting operation. Distributed riveting parameters are sent to the distributed riveting stations, including riveting force, riveting sequence and time, riveting tools and settings. Skip riveting is a riveting method in which the riveting tool jumps between multiple points during the riveting process to improve work efficiency. Employing multi-threaded closed-loop force control means that the cloud can monitor the working status of multiple riveting units (including fixed riveting arrays and flexible supplementary units) in real time and dynamically adjust them through closed-loop force control. If the force exceeds the preset range or a deviation occurs, the operation will be adjusted immediately to ensure riveting accuracy and quality.

[0021] A500: After the busbar trunking is segmented and riveted, and then unlocked and removed from the riveting section, online riveting quality detection is triggered.

[0022] Once the riveting section has completed all scheduled riveting tasks and secured all riveting points, the busbar trunking segment will be unlocked and removed from the riveting station. At this point, the segment is released via mechanical devices or an automated system to ensure its smooth transition to the next stage of processing or transportation. Online riveting quality inspection refers to the automatic initiation of a series of quality inspection procedures after the riveting operation is completed. This process uses high-precision inspection equipment, such as 3D laser scanners and vision inspection systems, to inspect the riveted busbar trunking segments. Based on the scanning results, multimodal fusion analysis is performed, and quality evaluation indicators are output by combining parameters such as riveting head height and head concentricity. If defects are detected, the defect location is marked, and a defect report is generated, providing a basis for subsequent quality improvement.

[0023] Furthermore, the method for identifying riveting feature points and locating distributed riveting reference point coordinates based on the multi-angle real-time images includes: A210: After jointly calibrating the internal and external parameters of the multispectral industrial camera array using a checkerboard calibration board, synchronous exposure control is executed based on a hardware trigger signal to perform multispectral synchronous acquisition of the multi-angle real-time images; A220: Multimodal image pixel-level weighted fusion is performed on the multi-angle real-time images to output a multispectral fusion feature map; A230: Dense stereo matching is performed on the multispectral fusion feature map to generate a three-dimensional dense point cloud model; A240: Subpixel instance segmentation of the riveting hole is performed on the multispectral fusion feature map to locate the subpixel coordinates of the riveting hole center; A250: The subpixel coordinates of the riveting hole center are mapped to the three-dimensional dense point cloud model through a perspective projection matrix, spatial coordinate calculation is performed, and the coordinates of the distributed riveting reference point are output.

[0024] To ensure the accurate acquisition of image data by the multispectral industrial camera array, a joint calibration of the camera's intrinsic and extrinsic parameters is first performed. The calibration uses a checkerboard calibration board, which has known dimensions and a symmetrical structure, making it suitable for calculating the camera's intrinsic and extrinsic parameters. Intrinsic parameter calibration includes calculating parameters such as the camera's focal length, focal point position, and distortion coefficients; extrinsic parameter calibration involves determining the camera's spatial position and orientation relative to the calibration board. Through joint calibration, a unified coordinate system can be established among multiple cameras, ensuring that images from the multispectral cameras can be aligned and matched within the same coordinate system.

[0025] After calibration, the camera exposure process is synchronized via a hardware trigger signal. This hardware trigger ensures that each camera array can acquire image data at the same time, avoiding image errors caused by exposure delays. Multispectral synchronous acquisition refers to acquiring images in different spectral ranges using multiple cameras at the same time. This synchronous acquisition can obtain richer image information, providing higher-precision data support for subsequent tasks such as feature point recognition, 3D reconstruction, and defect detection.

[0026] By using a multispectral industrial camera array, multiple images from different angles are acquired. These images not only have different perspectives in space, but also contain different spectral information, such as data in the visible light, infrared, and ultraviolet bands. The details and features of each image will be different.

[0027] To integrate image information from different perspectives and spectral bands, the pixel values ​​of each image are weighted and fused, assigning different weights to information in different images. The final fused result is then determined based on the reliability and sharpness of each image. In this stage, image pixels are fused one by one; that is, for each pixel, a comprehensive fused pixel value is calculated by combining pixel values ​​from multiple angles or spectral images. The fusion method can be weighted averaging, maximum value selection, minimum value selection, etc., depending on the application scenario and data characteristics.

[0028] After pixel-level weighted fusion is completed, the resulting image is called a multispectral fusion feature map. This image integrates features from multiple angles and multispectral data, has a higher ability to preserve details, and can provide richer image information for subsequent image processing.

[0029] Based on multispectral fusion feature maps, a dense stereo matching algorithm is used to compare images from multiple different angles to find corresponding feature points. The positional differences (i.e., parallax) of these feature points in the image can be used to infer their spatial positions. Dense matching means that depth calculation is performed on each pixel to obtain more accurate 3D information, rather than just local or sparse matching. The dense stereo matching result generates a point cloud dataset containing the spatial coordinates of each point. Each point cloud corresponds to a 3D spatial position, and each point has associated depth information. The 3D dense point cloud model is a 3D model formed by combining all point cloud data in space. It can realistically reflect the 3D structure and shape of the photographed object or workpiece.

[0030] Instance segmentation refers to separating a target (in this case, a rivet hole) from the background in an image. This involves not only locating the target edges at the pixel level but also accurately labeling each target instance. Subpixel segmentation further improves localization accuracy, enabling the identification of targets even in the most subtle differences between pixels. Modern image processing algorithms, such as convolutional neural networks, are used to achieve accurate instance segmentation, allowing for the independent identification of each rivet hole and precise segmentation of its boundaries.

[0031] After the rivet holes are segmented, the center coordinates of the rivet holes are precisely located. Subpixel precision means not only determining the pixel position of the rivet holes, but also being accurate to the finer positions between pixels, for example, through interpolation. Subpixel positioning finds the precise center of the boundary by calculating the grayscale changes between pixels and mapping it to finer coordinate values.

[0032] A perspective projection matrix is ​​a mathematical tool for converting two-dimensional image coordinates into three-dimensional spatial coordinates. In this stage, the perspective projection matrix maps the two-dimensional sub-pixel coordinates obtained from the center of the riveting hole to its actual position in three-dimensional space. Through this mapping process, combined with the camera's calibration parameters and the pixel positions in the image, the spatial coordinates of each riveting hole in three-dimensional space are obtained. This means that the actual position of the riveting hole can be accurately located in three-dimensional space, further providing a precise reference for the riveting operation. The final result is a set of three-dimensional coordinates containing all the riveting holes; these coordinates are called distributed riveting reference point coordinates.

[0033] Furthermore, the method involves segmenting the riveting hole sub-pixel instances into the multispectral fused feature map and locating the sub-pixel coordinates of the riveting hole center. A241: Perform surface anti-reflection preprocessing on the multispectral fusion feature map to generate an illumination equalization feature map; A242: Perform multi-instance segmentation on the illumination equalization feature map using the YOLOv7-R industrial-grade target detection model to output distributed candidate bounding boxes; A243: Expand the pixel region using the distributed candidate bounding boxes as the spatial topology reference to construct a distributed multimodal spatial anchoring domain; A244: Perform sub-pixel level edge refinement on the distributed multimodal spatial anchoring domain to locate the sub-pixel coordinates of the riveting hole center.

[0034] In multispectral images, surface reflections, such as gloss and specular reflection, can lead to the loss or distortion of image information in certain areas. To address this issue, surface anti-reflection preprocessing techniques are used to reduce the influence of illumination. This preprocessing method includes: identifying and removing highlights generated by the light source in the image while preserving the reflection characteristics of the true surface; and processing existing reflective areas in the image to prevent them from interfering with subsequent analysis. The preprocessed image appears more uniform and clearer, reducing the impact of illumination and providing more reliable image data for subsequent analysis. The processed image is called an illumination equalization feature map, which is more suitable for further target detection and image processing than the original image.

[0035] YOLO (You Only Look Once) is a real-time object detection system. YOLOv7 is its latest version, offering higher accuracy and speed. YOLOv7-R refers to a specific variant of YOLOv7, optimized for industrial applications, enabling efficient object detection and the ability to detect multiple objects in an image. The YOLOv7-R industrial-grade object detection model identifies rivet holes in an illumination-equalized feature map and distinguishes them from the background. Through object detection algorithms, multiple targets, i.e., rivet holes, can be located in the image, and a bounding box is generated for each target. These bounding boxes are predicted boxes representing the regions in the image where rivet holes exist. Distributed candidate bounding boxes mean that multiple detected rivet holes are distributed across different regions of the image, and the location of each rivet hole is individually labeled.

[0036] Distributed candidate bounding boxes serve as a spatial topological baseline, providing a reference framework for subsequent operations. Pixel region expansion is performed, meaning that a certain pixel range is extended outward from the boundary of each candidate bounding box to ensure that all rivet holes are completely contained within the processing area. The spatial anchoring domain refers to a specific region in the image. In this step, a distributed multimodal spatial anchoring domain is constructed by combining the candidate bounding boxes and the expanded pixel region, covering multiple rivet hole regions. Multimodality means that this spatial anchoring domain not only considers the visual information of the image but also incorporates other types of information, such as depth and reflection. By combining this multimodal information, the localization of rivet holes can be processed more comprehensively.

[0037] In target detection, the edges of rivet holes may not be perfectly precise, especially in pixel-level images. Subpixel-level edge refinement aims to refine the edges using high-precision algorithms to obtain more accurate target boundaries. Techniques such as interpolation, edge detection, and image sharpening are used to further optimize the edges of the rivet holes, making their positional information more accurate. Through this refined edge refinement, the center position of the rivet hole can be determined more accurately. Subpixel precision means being able to pinpoint minute differences between pixels, achieving higher accuracy than traditional pixel-level positioning.

[0038] Furthermore, the cloud-based system executes multi-station collaborative scheduling decisions based on the real-time equipment status of the distributed riveting stations and the coordinates of the distributed riveting reference points, generating dynamic riveting task allocation instructions. The method includes: A310: Analyze the normal vector and spatial coordinates of the distributed riveting reference point coordinates to construct a riveting point topology matrix; A320: Analyze the real-time device status to obtain multiple working envelope space ranges and multiple multi-axis real-time attitude angles of multiple fixed riveting units in the fixed riveting array, and multiple global positioning coordinates and multiple joint degrees of freedom states of multiple flexible filler riveting units; A330: Based on the multiple working envelope space ranges and multiple multi-axis real-time attitude angles, traverse the riveting point topology matrix, calculate the riveting reachability of the multiple fixed riveting units, and generate a riveting point weight allocation matrix; A340: Perform riveting competition based on the riveting point weight allocation matrix to generate a fixed workstation scheduling decision; A350: Using the fixed workstation scheduling decision as a space occupancy taboo, perform filler riveting optimization of the riveting point topology matrix based on the multiple global positioning coordinates and multiple joint degrees of freedom states to generate multiple flexible filler displacement decisions; A360: Spatially coordinate the multiple flexible filler displacement decisions and fixed workstation scheduling decisions to generate the dynamic riveting task allocation instruction.

[0039] Using positional data in three-dimensional space, the normal vector of each riveting point is calculated. For example, a geometric algorithm is used to calculate the normal direction using data from neighboring points, ensuring that the riveting head is parallel or perpendicular to the target surface, depending on design requirements. Based on the coordinates of distributed riveting reference points, the three-dimensional spatial coordinates of each riveting point are resolved. These coordinates can be based on the workpiece's global coordinate system or local coordinates relative to certain reference positions. Using the resolved normal vectors and spatial coordinate data, a riveting point topology matrix is ​​constructed. This matrix represents the spatial relationships between different riveting points and their relative positions during the riveting process.

[0040] The fixed riveting array comprises multiple fixed riveting units, each performing a riveting task at a preset position. The working envelope of each fixed riveting unit determines its operable spatial area, which is a three-dimensional space defining the area the unit can access. The attitude of each fixed riveting unit consists of angles along multiple axes (such as rotation angles). By analyzing these real-time attitude angles, the current orientation and position of the riveting unit can be accurately determined, used to adjust the position and angle of the rivet joint. Multiple flexible supplementary riveting units offer greater flexibility, enabling precise riveting at different positions. Analyzing their global positioning coordinates reveals their current positions, and the joint degrees of freedom provide the range of motion that the flexible supplementary riveting units can perform and whether any obstacles are present.

[0041] Riveting accessibility refers to whether each fixed riveting unit can reach and successfully perform the task at the riveting point. In this step, by combining the working envelope space range of each fixed riveting unit with its multi-axis real-time attitude angles, it is determined whether each riveting unit can contact and complete the operation at a specific riveting point. Using the constructed riveting point topology matrix, each riveting point is traversed to analyze which fixed riveting units can reach these points, and the accessibility of each riveting point is calculated. Based on the calculated accessibility, a weight allocation matrix is ​​generated. Each element in this matrix represents the degree of matching between a specific riveting unit and the riveting point. For example, a fixed riveting unit has a high accessibility to a riveting point, so the weight of that position is higher.

[0042] In scenarios with multiple fixed riveting units and multiple riveting points, there may be situations where multiple fixed riveting units can reach the same riveting point. To resolve this conflict, a riveting competition is implemented. This involves determining which riveting unit will ultimately execute the task based on the weight of each riveting point. During this process, a riveting point weight allocation matrix is ​​used to calculate the priority, reachability, and task allocation strategy between each riveting unit and riveting point, selecting the most suitable fixed riveting unit to perform the riveting task. Based on the results of the riveting competition, it is determined which fixed riveting unit will execute the task at each riveting point, generating a fixed workstation scheduling decision.

[0043] The generated fixed-position scheduling decisions are used as space occupancy taboos to prevent flexible replacement riveting units from attempting to occupy the same position or path at the same time, ensuring that each riveting unit has an independent working space. Replacement riveting refers to supplementing or correcting hard-to-reach riveting points using flexible replacement riveting units. Due to their high flexibility and adjustability, flexible replacement riveting units can perform replacement operations as needed, ensuring that all riveting points are correctly riveted. The riveting point topology matrix is ​​optimized based on multiple global positioning coordinates and multiple joint degrees of freedom states, calculating the optimal displacement path for the flexible replacement riveting unit to efficiently reach the riveting point and complete the riveting task. Through the above analysis, multiple flexible replacement displacement decisions are generated, instructing each flexible replacement riveting unit how to move within its operable range, ensuring that it can accurately reach the replacement riveting point and complete the riveting task.

[0044] The obtained multiple flexible offset displacement decisions and fixed workstation scheduling decisions are spatially coordinated to generate dynamic riveting task allocation instructions. The instructions include specific work positions, work sequences, task priorities and other parameters to ensure that all riveting units can execute tasks according to the optimal path and plan.

[0045] Furthermore, based on the riveting point weight allocation matrix, riveting competition is performed to generate fixed workstation scheduling decisions. The method includes: A341: After performing multi-constraint normalization on the riveting point weight allocation matrix, perform normal-distance joint priority weighting to generate a standardized weight matrix; A342: Construct a device spatial capability constraint domain based on the multiple multi-axis real-time attitude angles and multiple global positioning coordinates; A343: Construct a bipartite graph model based on the device spatial capability constraint domain and distributed riveting reference point coordinates; A345: After modeling the topological connection relationship of the bipartite graph model based on the standardized weight matrix, solve for maximum weight matching and output the device-riveting point pairing set; A346: Perform motion path planning based on the device-riveting point pairing set and output the fixed workstation scheduling decision.

[0046] Multi-constraint normalization integrates multiple different constraints into a unified standardized scale. This step normalizes the original anchor point weight assignment matrix to eliminate scale differences caused by different constraints. The normalization process converts each constraint into a uniform weight range, such as 0 to 1, so that the weights of all constraints can be calculated on the same scale.

[0047] In riveting tasks, normal vectors and distances significantly influence task priority. For example, closer riveting points are prioritized as task targets, or the normal vectors of certain riveting points determine their riveting order or method. Based on the normal vector of each riveting point and its distance to the riveting unit, the priority of each riveting point is jointly calculated. This priority incorporates the effects of spatial and geometric constraints through a weighted method, and higher-priority riveting points are assigned to more suitable riveting units for processing.

[0048] By using multi-constraint normalization and priority weighting, a standardized weight matrix is ​​finally generated. Each element in this matrix represents the degree of matching between the riveting point and the riveting unit, as well as the priority of the task, which serves as the basis for subsequent task scheduling and optimization.

[0049] During the riveting process, riveting equipment possesses a certain spatial capability, meaning the tasks it can complete within its operational range. This spatial capability is limited by the equipment's attitude angles and its positioning in space. Specifically, each riveting unit has multiple degrees of freedom to adjust its angles or postures when performing a task. These attitude angles affect the equipment's accessibility to the riveting point and its operational range. Global positioning coordinates refer to the equipment's precise position in the workspace, which, together with the equipment's attitude angles, determines the working range that the equipment can cover, i.e., the equipment's workspace. Based on this information, a spatial capability constraint domain is constructed, representing the spatial area that the equipment can cover and operate within.

[0050] A bipartite graph is a graph structure where nodes can be divided into two disjoint subsets, and each edge connects a node in one of these subsets. The two subsets are: the device set, which includes all riveting units (fixed riveting arrays and flexible patching riveting units); and the riveting point set, which includes all reference points that need to be riveted (riveting holes). If a device has a riveting point within its working range, the bipartite graph model will connect the device node and the riveting point node, with the edge representing the device's ability to perform operations on that riveting point. The bipartite graph model is constructed by establishing edge relationships between devices and riveting points, creating a one-to-one correspondence between each device and the riveting points it can handle.

[0051] Based on the standardized weight matrix, each edge in the bipartite graph model is assigned a weight. The goal of topological connectivity modeling is to determine which device should perform the task at which riveting point using these weights, thus forming a reasonable device-riveting point allocation. Maximum weight matching is a classic algorithm in graph theory. Its goal is to find a set of device-riveting point matches in a bipartite graph that maximizes the total weight of the match. The maximum weight matching algorithm optimizes the pairing relationships between devices and riveting points to ensure that each device performs its task with the highest efficiency or lowest cost. The output device-riveting point pairing set indicates which device is responsible for which riveting point's task under the maximum weight matching.

[0052] During the riveting process, the equipment needs to move along a specific trajectory to reach the target riveting point and complete the task. Motion path planning calculates the optimal trajectory for each piece of equipment based on its paired riveting points. Motion path planning not only finds the shortest path from the current position to the target position but also considers obstacles in the workspace, dynamic constraints of the equipment, and the timing requirements of the task. Path planning algorithms, such as A* and Dijkstra's algorithm, are used to calculate the optimal path from the equipment's current position to the riveting point based on the equipment's kinematic model, ensuring that collisions or interference do not occur during the movement. After the motion path planning is completed, a fixed-station scheduling decision is generated based on the pairing set and the planning results, instructing each fixed riveting unit how to move along the predetermined trajectory and complete the riveting task.

[0053] Furthermore, taking the fixed workstation scheduling decision as a space occupancy taboo, and based on the multiple global positioning coordinates and multiple joint degree of freedom states, the riveting optimization of the riveting point topology matrix is ​​performed to generate multiple flexible displacement decisions. The method includes: A351: Calculate the riveting motion trajectory based on the fixed workstation scheduling decision and generate a spatiotemporal occupancy cube; A352: Based on the equipment-riveting point pairing set, locate multiple riveting coordinates to be filled in the opposite direction at the distributed riveting reference point coordinates; A353: Using the spatiotemporal occupancy cube as a dynamic obstacle avoidance constraint field, and using the multiple global positioning coordinates and multiple joint degrees of freedom states as initial values ​​for kinematic solution, optimize the filling riveting trajectory based on the multiple riveting coordinates to be filled in, and generate the multiple flexible filling displacement decisions.

[0054] The fixed-station scheduling decision has clearly defined which equipment needs to perform which riveting tasks. Based on these decisions, the riveting motion trajectory of each equipment during task execution is calculated. When calculating the trajectory, kinematic constraints of the equipment, such as the range of joint angle changes, equipment length, and dynamic limitations, are considered to ensure the feasibility of the motion trajectory. A spatiotemporal occupancy cube is used to represent the equipment's workspace. This not only describes the area occupied by the equipment in space but also incorporates the time dimension, representing the equipment's occupancy status over time during movement. Through this cube, the spatial area occupied by the equipment at different points in time can be clearly determined.

[0055] By using a device-riveting point pairing set, it is determined which riveting points require replacement operations. These replacement operations are handled by a flexible replacement riveting unit, especially when the fixed riveting array cannot directly reach certain riveting points. Reverse positioning involves determining the specific riveting coordinates that need replacement based on the known coordinates of distributed riveting reference points. These replacement coordinates are located in irregular positions or, due to workspace limitations, cannot be directly completed by the main riveting array.

[0056] A spatiotemporal occupancy cube is used as a dynamic obstacle avoidance constraint field, providing information on the spatial area and temporal changes occupied by the device. This constraint field ensures effective collision avoidance with other devices during the riveting and patching operation. Initial kinematic values ​​refer to the device's initial attitude and position during the riveting and patching operation. Based on the device's global positioning coordinates and joint degrees of freedom, a preliminary estimate of the device's trajectory from its current state to the target position is calculated. After obtaining the riveting and patching coordinates, trajectory optimization is performed, calculating the optimal path from the device's current position to the target riveting point. This process considers multiple factors, including kinematic constraints, obstacle avoidance constraints, and task timing requirements, optimizing the path planning of the riveting and patching unit to ensure the shortest and safest path while avoiding collisions with other devices. After the riveting and patching trajectory optimization is complete, specific flexible riveting displacement decisions are generated to guide the flexible riveting and patching unit in accurately executing the riveting task.

[0057] Furthermore, the cloud-based system, based on the dynamic riveting task allocation instruction, guides the distributed riveting station to the distributed riveting reference point of the busbar segment, and then sends distributed riveting parameters to the distributed riveting station. This controls the fixed riveting array and multiple flexible supplementary riveting units to perform multi-threaded closed-loop force control operations based on skip riveting. The method includes: A410: Divide the distributed riveting reference point coordinates into an odd-numbered riveting reference point coordinate group and an even-numbered riveting reference point coordinate group; A420: Set the phase difference trigger delay for the odd-numbered riveting reference point coordinate group and the even-numbered riveting reference point coordinate group; A430: Based on the odd-numbered riveting reference point coordinate group and the even-numbered riveting reference point coordinate group, divide the fixed riveting array and multiple flexible supplementary riveting units into a first group of riveting control units and a second group of riveting control units; A440: Using the phase difference trigger delay as a synchronization control reference, use the distributed riveting parameters to intermittently control the first group of riveting control units and the second group of riveting control units, and perform a three-stage closed-loop force control operation in the odd-numbered riveting reference point coordinate group and the even-numbered riveting reference point coordinate group.

[0058] The distributed riveting reference point coordinates are divided into two groups: an odd group and an even group. This division is to assign different tasks to different control units during riveting, thereby enhancing the parallelism and flexibility of the operation. The odd-numbered riveting reference point coordinate group contains all riveting reference points located at odd positions; the even-numbered riveting reference point coordinate group contains all riveting reference points located at even positions.

[0059] Phase difference trigger delay is a time delay value used to indicate the start-up time difference between two sets of tasks. This value is derived by analyzing various factors in the production process, such as equipment capacity, riveting accuracy requirements, and workspace constraints. By setting a reasonable phase difference trigger delay, parallel operation between tasks can be effectively achieved, while avoiding equipment overload or performance degradation due to task conflicts.

[0060] Fixed riveting arrays are used to handle riveting tasks with relatively fixed positions and high repetitiveness; these are assigned to the first group of riveting control units. Flexible, complementary riveting units are used to handle riveting tasks that are more flexible and require complementary positioning; these tasks are assigned to the second group of riveting control units. Based on different task requirements, the control units can execute corresponding riveting tasks according to the assigned riveting point groups. This grouping method allows for more efficient management of riveting tasks and reduces the complexity of task execution.

[0061] Using the phase difference trigger delay as the synchronization control benchmark for the execution of two sets of tasks, the riveting tasks of the first and second sets can be executed alternately based on the phase difference trigger delay, avoiding conflicts between equipment and workspace. Cross-intermittent control is a method of allocating tasks by staggering execution times, ensuring that control units of different groups do not execute tasks simultaneously, thereby effectively allocating equipment resources and avoiding bottlenecks caused by excessive task concentration. Based on the phase difference trigger delay, the start-up timing of the first and second sets of riveting control units is alternately controlled, achieving efficient scheduling of the two sets of tasks. Three-stage closed-loop force control refers to riveting during the process, using three different control strategies based on real-time force feedback: Initial control: when the equipment quickly approaches the target riveting point, a higher control speed is used, but precise control of pressure and force is still required; Middle control: as the equipment approaches the riveting point, the control unit adjusts the riveting force to ensure riveting accuracy; Final control: when the riveting operation is completed, the speed is slowed down, and fine force feedback adjustment is performed to ensure riveting quality and accuracy. This control method effectively avoids excessive or insufficient riveting force, ensuring riveting quality while maximizing production efficiency.

[0062] Furthermore, the method also includes: A441: Retrieve the riveting point material and riveting thickness based on the unique identifier of the busbar segment; A442: Using the riveting point material as the primary key and the riveting thickness as the secondary index, match and obtain the initial riveting parameters in the riveting parameter library; A443: Collect the real-time temperature field of the busbar segment; A444: Use the real-time temperature field to correct the initial riveting parameters and output the distributed riveting parameters.

[0063] Each busbar trunking segment has a unique identifier, a system-generated code or serial number, used to identify and track specific information about the segment. This unique identifier is used to retrieve the material of the riveting points associated with that busbar trunking segment, such as steel or aluminum, and the riveting thickness, from the riveting point information database. This information is used to determine the basic parameters of the riveting operation, as different materials and thicknesses require different riveting forces, speeds, and tool settings.

[0064] The riveting parameter library stores detailed parameters for different types of riveting tasks, such as the required riveting force, riveting speed, and tool type. These parameters are based on years of riveting experience, experimental results, and standard specifications, and can help the riveting system be appropriately adjusted under different conditions. The riveting parameter library is searched using the riveting point material as the primary key, and then further indexed by the riveting thickness to accurately match the initial riveting parameters suitable for the specific busbar segment riveting point, including the force to be applied during riveting, the appropriate riveting speed, and the tools used.

[0065] During the riveting process, temperature has a significant impact on the riveting quality. Too high or too low a temperature may lead to riveting failure or non-compliance. Therefore, it is necessary to collect the temperature information of the busbar trunking segments in real time. The real-time temperature field refers to the temperature distribution at different locations in the entire busbar trunking segment, which is monitored in real time through temperature sensor arrays or infrared imaging technology.

[0066] Based on the acquired real-time temperature field, the initial riveting parameters are adjusted according to the influence of temperature on the riveting parameters. For example, if the temperature is too high, the material will soften, requiring a reduction in riveting force or an adjustment in riveting speed; if the temperature is too low, the material will harden, requiring an increase in riveting force or an adjustment in speed. Based on materials science, mechanics, and thermodynamics, the real-time temperature field is applied to the initial riveting parameters to generate temperature-corrected distributed riveting parameters, which guide the actual riveting operation, thereby ensuring high precision and high quality in riveting.

[0067] Furthermore, after the busbar trunking segment riveting is completed and the section is unlocked and removed from the riveting section, an online riveting quality detection is triggered. The method includes: A510: After the busbar segment is unlocked and removed from the riveting section, a workpiece removal signal is triggered; A520: The 3D laser scanner array responds to the workpiece removal signal and performs a full-domain topographic scan of the busbar segment to obtain a distributed riveting point cloud dataset; A530: The distributed riveting point cloud dataset is parsed to obtain the distributed riveting uphead height and the distributed riveting uphead concentricity; A540: The distributed riveting uphead height and the distributed riveting uphead concentricity are used to perform a multimodal fusion quality decision evaluation and output the coordinates of riveting quality defects.

[0068] After the riveting section completes the riveting task, the busbar section is unlocked from the riveting station and removed, triggering a workpiece removal signal. This signal is an event notification that the riveting process has been completed, and quality inspection is required next.

[0069] When the busbar trunking segment is moved out, a 3D laser scanner is triggered to perform a comprehensive scan of the busbar trunking segment. The 3D laser scanner scans the surface of the entire busbar trunking segment with a laser beam, generating high-precision 3D point cloud data, which becomes a distributed riveting point cloud dataset. This dataset contains detailed spatial information of the entire busbar trunking segment.

[0070] By analyzing a distributed riveting point cloud dataset, the distributed riveting uphead height and distributed riveting pier concentricity were extracted. The uphead, the protruding part generated during riveting, is a crucial indicator of riveting quality; an uphead that is too high or too low indicates problems with riveting quality, such as excessive or insufficient riveting force. The pier is a circular structure formed during riveting; its concentricity refers to the symmetry of its center. Excessive deviation can affect the connection quality, leading to weak connections or potential structural problems. The uphead height and pier concentricity were calculated based on data from the riveting area in the point cloud. This involved geometric analysis and fitting techniques, such as using least squares fitting to determine the pier's center and calculating its deviation.

[0071] By integrating two quality indicators—distributed riveting uphead height and distributed riveting head concentricity—and combining two different types of defect indicators, the riveting quality is comprehensively evaluated: if the uphead height exceeds the preset tolerance range, it indicates excessive force or improper operation during riveting; if the concentricity of the uphead deviates too much, it indicates a deviation in the riveting, affecting the stability of the connection. Finally, based on the above analysis, specific riveting quality defect coordinates are output. These defect coordinates indicate the location of the problematic riveting point during the riveting process, specifically including the exact locations of abnormal uphead height and uphead concentricity deviations.

[0072] Example 2, based on the same inventive concept as the riveting distributed control method for dense busbar trunking in the foregoing examples, such as... Figure 2 As shown in the figure, this application provides a riveting distributed control system for dense busbar trunking, the system comprising: Image acquisition module 10 is used to move busbar segments of a preset length to the riveting section on the production line. After being fixed by a hydraulic locking mechanism, it triggers a multispectral industrial camera array installed on the side of the riveting section to perform multi-angle real-time image acquisition. Feature point recognition module 20 is used to identify riveting feature points based on the multi-angle real-time images and locate the coordinates of distributed riveting reference points. Scheduling decision module 30 is used to execute multi-station collaborative scheduling decisions in the cloud based on the real-time equipment status of the distributed riveting stations and the coordinates of the distributed riveting reference points, generating dynamic riveting task allocation instructions. The distributed riveting station includes a fixed riveting array and multiple flexible supplementary riveting units; the force control operation execution module 40 is used to guide the distributed riveting station to the distributed riveting reference point of the busbar segment according to the dynamic riveting task allocation instruction, and then send distributed riveting parameters to the distributed riveting station to control the fixed riveting array and multiple flexible supplementary riveting units to perform multi-threaded closed-loop force control operation based on jump riveting; the online quality detection module 50 is used to trigger online riveting quality detection after the busbar segment riveting is completed and the segment is unlocked and removed.

[0073] Furthermore, the feature point recognition module 20 is used to perform the following operation steps: After joint calibration of the intrinsic and extrinsic parameters of the multispectral industrial camera array using a checkerboard calibration board, synchronous exposure control is executed based on hardware trigger signals to perform multispectral synchronous acquisition of real-time images from multiple angles. Multimodal image pixel-level weighted fusion is performed on the real-time images from multiple angles to output a multispectral fusion feature map. Dense stereo matching is performed on the multispectral fusion feature map to generate a three-dimensional dense point cloud model. Subpixel instance segmentation of the riveting holes is performed on the multispectral fusion feature map to locate the subpixel coordinates of the riveting hole center. The subpixel coordinates of the riveting hole center are mapped to the three-dimensional dense point cloud model through a perspective projection matrix, spatial coordinate calculation is performed, and the coordinates of the distributed riveting reference points are output.

[0074] Furthermore, the feature point recognition module 20 is used to perform the following operation steps: The multispectral fusion feature map is preprocessed with surface anti-reflection to generate an illumination equalization feature map; the illumination equalization feature map is segmented into multiple instances using the YOLOv7-R industrial-grade target detection model to output distributed candidate bounding boxes; the pixel region is expanded using the distributed candidate bounding boxes as a spatial topology reference to construct a distributed multimodal spatial anchoring domain; sub-pixel level edge refinement is performed on the distributed multimodal spatial anchoring domain to locate the sub-pixel coordinates of the riveting hole center.

[0075] Furthermore, the scheduling decision module 30 is used to perform the following operation steps: The distributed riveting reference point coordinates are analyzed using normal vectors and spatial coordinates to construct a riveting point topology matrix. The real-time device state is analyzed to obtain multiple working envelope space ranges and multiple multi-axis real-time attitude angles for multiple fixed riveting units in the fixed riveting array, as well as multiple global positioning coordinates and multiple joint degrees of freedom states for multiple flexible filler riveting units. Based on the multiple working envelope space ranges and multiple multi-axis real-time attitude angles, the riveting point topology matrix is ​​traversed to calculate the riveting reachability of the multiple fixed riveting units, generating a riveting point weight allocation matrix. Riveting competition is performed based on the riveting point weight allocation matrix to generate a fixed workstation scheduling decision. Using the fixed workstation scheduling decision as a space occupancy taboo, filler riveting optimization of the riveting point topology matrix is ​​performed based on the multiple global positioning coordinates and multiple joint degrees of freedom states, generating multiple flexible filler displacement decisions. The multiple flexible filler displacement decisions and fixed workstation scheduling decisions are spatially coordinated to generate the dynamic riveting task allocation instruction.

[0076] Furthermore, the scheduling decision module 30 is used to perform the following operation steps: After performing multi-constraint normalization on the riveting point weight allocation matrix, a normal-distance joint priority weighting is performed to generate a standardized weight matrix. A device spatial capability constraint domain is constructed based on the multiple multi-axis real-time attitude angles and multiple global positioning coordinates. A bipartite graph model is constructed based on the device spatial capability constraint domain and the coordinates of distributed riveting reference points. After modeling the topological connectivity of the bipartite graph model using the standardized weight matrix, the maximum weight matching is solved, and a device-riveting point pairing set is output. Motion path planning is performed based on the device-riveting point pairing set, and the fixed workstation scheduling decision is output.

[0077] Furthermore, the scheduling decision module 30 is used to perform the following operation steps: Based on the fixed workstation scheduling decision, the riveting motion trajectory is calculated, and a spatiotemporal occupancy cube is generated. Based on the equipment-riveting point pairing set, multiple riveting coordinates to be filled are located in reverse at the coordinates of the distributed riveting reference points. Using the spatiotemporal occupancy cube as a dynamic obstacle avoidance constraint field, and using the multiple global positioning coordinates and multiple joint degrees of freedom as initial values ​​for kinematic solution, the riveting trajectory is optimized based on the multiple riveting coordinates to be filled, and the multiple flexible filling displacement decisions are generated.

[0078] Furthermore, the force control operation execution module 40 is used to perform the following operation steps: The distributed riveting reference point coordinates are divided into odd-numbered riveting reference point coordinate groups and even-numbered riveting reference point coordinate groups; a phase difference trigger delay is set for the odd-numbered and even-numbered riveting reference point coordinate groups; based on the odd-numbered and even-numbered riveting reference point coordinate groups, the fixed riveting array and multiple flexible supplementary riveting units are mixed and divided into a first group of riveting control units and a second group of riveting control units; using the phase difference trigger delay as a synchronization control reference, the first group of riveting control units and the second group of riveting control units are cross-intermittently controlled using the distributed riveting parameters, and a three-stage closed-loop force control operation is performed in the odd-numbered and even-numbered riveting reference point coordinate groups.

[0079] Furthermore, the force control operation execution module 40 is used to perform the following operation steps: The material and thickness of the riveting point are retrieved based on the unique identifier of the busbar segment; the initial riveting parameters are obtained by matching the riveting parameter library with the material of the riveting point as the primary key and the riveting thickness as the secondary index; the real-time temperature field of the busbar segment is collected; the initial riveting parameters are corrected using the real-time temperature field, and the distributed riveting parameters are output.

[0080] Furthermore, the online quality detection module 50 is used to perform the following operation steps: After the busbar segment is unlocked and removed from the riveting section, a workpiece removal signal is triggered. The 3D laser scanner array responds to the workpiece removal signal and performs a full-domain topographic scan of the busbar segment to obtain a distributed riveting point cloud dataset. The distributed riveting point cloud dataset is analyzed to obtain the distributed riveting uphead height and the distributed riveting uphead concentricity. The distributed riveting uphead height and the distributed riveting uphead concentricity are used to perform a multimodal fusion quality decision evaluation and output the coordinates of riveting quality defects.

[0081] The above description is merely a preferred embodiment of the present invention and is not intended to limit the present invention in any way. Although the present invention has been disclosed above with reference to preferred embodiments, it is not intended to limit the present invention. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the present invention. Any modifications, equivalent changes, and alterations made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the scope of the present invention.

Claims

1. A riveting distributed control method for dense busbar trunking, characterized in that, The method includes: The busbar trunking segments of a preset length are moved to the riveting section on the production line and fixed by the hydraulic locking mechanism. This triggers the multispectral industrial camera array installed on the side of the riveting section to perform multi-angle real-time image acquisition. Based on the multi-angle real-time images, riveting feature points are identified, and the coordinates of distributed riveting reference points are located. The cloud platform executes multi-station collaborative scheduling decisions based on the real-time equipment status of the distributed riveting station and the coordinates of the distributed riveting reference point, generating dynamic riveting task allocation instructions. The distributed riveting station includes a fixed riveting array and multiple flexible supplementary riveting units. The cloud platform guides the distributed riveting station to the distributed riveting reference point of the busbar segment according to the dynamic riveting task allocation instruction, and then sends distributed riveting parameters to the distributed riveting station to control the fixed riveting array and multiple flexible supplementary riveting units to perform multi-threaded closed-loop force control operation based on jump riveting. After the busbar trunking is segmented and riveted, and then unlocked and removed from the riveting section, an online inspection of the riveting quality is triggered.

2. The riveting distributed control method for dense busbar trunking as described in claim 1, characterized in that, Based on the multi-angle real-time images, riveting feature points are identified, and the coordinates of distributed riveting reference points are located. The method includes: After the internal and external parameters of the multispectral industrial camera array are jointly calibrated using a checkerboard calibration board, synchronous exposure control is executed based on a hardware trigger signal to perform multispectral synchronous acquisition of the multi-angle real-time images. Multimodal image pixel-level weighted fusion is performed on the multi-angle real-time images to output a multispectral fusion feature map; Dense stereo matching is performed on the multispectral fusion feature map to generate a three-dimensional dense point cloud model; The multispectral fusion feature map is segmented into sub-pixel instances of the riveting hole, and the sub-pixel coordinates of the center of the riveting hole are located. The sub-pixel coordinates of the center of the rivet hole are mapped to the three-dimensional dense point cloud model through a perspective projection matrix, and spatial coordinates are calculated to output the coordinates of the distributed rivet reference point.

3. The riveting distributed control method for dense busbar trunking as described in claim 2, characterized in that, The method involves segmenting the riveting hole sub-pixel instances into the multispectral fused feature map and locating the center sub-pixel coordinates of the riveting hole. The multispectral fusion feature map is preprocessed with surface anti-reflection to generate an illumination equalization feature map; The illumination equalization feature map is segmented into multiple instances using the YOLOv7-R industrial-grade object detection model, and distributed candidate bounding boxes are output. Using the distributed candidate bounding boxes as a spatial topology reference, pixel regions are expanded to construct a distributed multimodal spatial anchoring domain; Subpixel-level edge refinement is performed on the distributed multimodal spatial anchoring domain to locate the subpixel coordinates of the center of the rivet hole.

4. The riveting distributed control method for dense busbar trunking as described in claim 1, characterized in that, The cloud-based system executes multi-station collaborative scheduling decisions and generates dynamic riveting task allocation instructions based on the real-time equipment status of the distributed riveting stations and the coordinates of the distributed riveting reference points. The method includes: The normal vector and spatial coordinates of the distributed riveting reference points are analyzed to construct the riveting point topology matrix; The real-time device status is analyzed to obtain multiple working envelope space ranges and multiple multi-axis real-time attitude angles of multiple fixed riveting units in the fixed riveting array, as well as multiple global positioning coordinates and multiple joint degrees of freedom of multiple flexible replacement riveting units. Based on the multiple working envelope space ranges and multiple multi-axis real-time attitude angles, the riveting point topology matrix is ​​traversed, the riveting reachability of the multiple fixed riveting units is calculated, and a riveting point weight allocation matrix is ​​generated. Based on the riveting point weight allocation matrix, riveting competition is conducted to generate fixed workstation scheduling decisions; Taking the fixed workstation scheduling decision as a space occupancy taboo, the riveting optimization of the riveting point topology matrix is ​​performed based on the multiple global positioning coordinates and multiple joint degree of freedom states, generating multiple flexible replacement displacement decisions; The spatial coordination of multiple flexible displacement decisions and fixed workstation scheduling decisions generates the dynamic riveting task allocation instruction.

5. The riveting distributed control method for dense busbar trunking as described in claim 4, characterized in that, Based on the riveting point weight allocation matrix, riveting competition is performed to generate fixed workstation scheduling decisions. The method includes: After performing multi-constraint normalization on the riveting point weight allocation matrix, a normal-distance joint priority weighting is performed to generate a standardized weight matrix. The device spatial capability constraint domain is constructed based on the multiple multi-axis real-time attitude angles and multiple global positioning coordinates; A bipartite graph model is constructed based on the equipment space capability constraint domain and the coordinates of the distributed riveting reference points. After modeling the topological connectivity of the bipartite graph model based on the standardized weight matrix, the maximum weight matching is solved, and the device-riveting point pairing set is output. Based on the set of equipment-riveting point pairs, motion path planning is performed, and the fixed workstation scheduling decision is output.

6. The riveting distributed control method for dense busbar trunking as described in claim 5, characterized in that, Taking the fixed workstation scheduling decision as a space occupancy taboo, and based on the multiple global positioning coordinates and multiple joint degree of freedom states, the riveting optimization of the riveting point topology matrix is ​​performed to generate multiple flexible displacement decisions. The method includes: Based on the fixed workstation scheduling decision, the riveting motion trajectory is calculated, and a spatiotemporal occupancy cube is generated; Based on the device-riveting point pairing set, multiple riveting coordinates to be filled are located in reverse at the coordinates of the distributed riveting reference points; Using the spatiotemporal occupancy cube as the dynamic obstacle avoidance constraint field, and the multiple global positioning coordinates and multiple joint degrees of freedom states as the initial values ​​for kinematic solution, the replacement riveting trajectory is optimized based on the multiple replacement riveting coordinates, and the multiple flexible replacement displacement decisions are generated.

7. The riveting distributed control method for dense busbar trunking as described in claim 1, characterized in that, The cloud platform, based on the dynamic riveting task allocation instruction, guides the distributed riveting station to the distributed riveting reference point of the busbar segment, and then sends distributed riveting parameters to the distributed riveting station. This controls the fixed riveting array and multiple flexible supplementary riveting units to perform multi-threaded closed-loop force control operations based on skip riveting. The method includes: The coordinates of the distributed riveting reference points are divided into odd-numbered riveting reference point coordinate groups and even-numbered riveting reference point coordinate groups. Set the phase difference trigger delay between the odd-number riveting reference point coordinate group and the even-number riveting reference point coordinate group; Based on the odd-number riveting reference point coordinate group and the even-number riveting reference point coordinate group, the fixed riveting array and multiple flexible filler riveting units are mixed and divided into a first group of riveting control units and a second group of riveting control units. Using the phase difference trigger delay as the synchronization control reference, the first group of riveting control units and the second group of riveting control units are controlled intermittently using the distributed riveting parameters. Three-stage closed-loop force control operations are performed in the odd-numbered riveting reference point coordinate group and the even-numbered riveting reference point coordinate group.

8. The riveting distributed control method for dense busbar trunking as described in claim 1, characterized in that, The method further includes: The material and thickness of the riveting points are retrieved based on the unique identifier of the busbar segment. Using the material of the rivet point as the primary key and the rivet thickness as the secondary index, the initial rivet parameters are obtained by matching in the rivet parameter library. The real-time temperature field of the busbar trunking segment is collected; The initial riveting parameters are corrected using the real-time temperature field, and the distributed riveting parameters are output.

9. The riveting distributed control method for dense busbar trunking as described in claim 1, characterized in that, After the busbar trunking is segmented and riveted, and then unlocked and removed from the riveting section, an online inspection of the riveting quality is triggered. The method includes: After the busbar section is unlocked and removed from the riveting section, a workpiece removal signal is triggered; In response to the workpiece removal signal, the 3D laser scanner array performs a full-domain topography scan of the busbar groove segments to obtain a distributed riveting point cloud dataset. The distributed riveting point cloud dataset is analyzed to obtain the height of the distributed riveting uphead and the concentricity of the distributed riveting pier. The distributed riveting uphead height and the distributed riveting pier concentricity are used to perform multimodal fusion quality decision evaluation, and the coordinates of riveting quality defects are output.

10. A distributed riveting control system for dense busbar trunking, characterized in that, For implementing the riveting distributed control method for dense busbar trunking according to any one of claims 1-9, the system comprises: The image acquisition module is used to move busbar sections of a preset length to the riveting section on the production line. After being fixed by the hydraulic locking mechanism, it triggers the multispectral industrial camera array installed on the side of the riveting section to perform multi-angle real-time image acquisition. The feature point recognition module is used to identify riveting feature points based on the multi-angle real-time images and locate the coordinates of distributed riveting reference points. The scheduling decision module is used to execute multi-station collaborative scheduling decisions in the cloud based on the real-time equipment status of the distributed riveting station and the coordinates of the distributed riveting reference point, and generate dynamic riveting task allocation instructions. The distributed riveting station includes a fixed riveting array and multiple flexible supplementary riveting units. The force control operation execution module is used to guide the distributed riveting station to the distributed riveting reference point of the busbar segment according to the dynamic riveting task allocation instruction in the cloud, and then send distributed riveting parameters to the distributed riveting station to control the fixed riveting array and multiple flexible supplementary riveting units to perform multi-threaded closed-loop force control operation based on jump riveting. The online quality inspection module is used to trigger online inspection of the riveting quality after the segmented riveting of the busbar trunking is completed and after the segment is unlocked and removed from the riveting section.

Images