A visual force control type double-arm robot rail pre-welding derusting and polishing system and method

Abstract

The application relates to a visual force control type rail pre-welding derusting and polishing system and method of a double-arm robot. The system comprises a first robot and a second robot arranged on the two sides of a rail conveying line respectively; a visual system comprising a first industrial camera installed on a flange plate at the tail end of the first robot and a second industrial camera installed on a flange plate at the tail end of the second robot; and an upper computer in communication connection with the visual system, which is used for generating a three-dimensional space model of a rail through three-dimensional point cloud reconstruction according to multi-view image data, and identifying the actual space position of the rail, the end section inclination angle, and the presence and position coordinates of the convex character at the rail waist based on the three-dimensional space model. The application realizes the technical effects of precise removal of the convex character at the rail waist in the rail pre-welding treatment and efficient operation of the double arms in parallel.

Description

Technical Field

[0001] This application relates to the field of railway construction and maintenance technology, and in particular to a vision-force-controlled dual-arm robot system and method for pre-welding rust removal and grinding of rails. Background Technology

[0002] In the construction and maintenance of seamless railway lines, flash butt welding of rails is a key process for connecting long rails. Before welding, the end sections, top surface, bottom surface, and web area of ​​the rails to be welded must be pre-treated by rust removal, grinding, and roughening (i.e., chamfering and deburring) to ensure the quality and service life of the welded joint.

[0003] Currently, pre-welding treatment of rails mainly relies on traditional rust removal machines or single-arm rust removal robots. Traditional rust removal machines are non-intelligent devices, lacking spatial perception capabilities and unable to identify the actual positional deviations of the rails after long-distance transport or the tilt angles of the cross-sections introduced by sawing at the ends. The grinding process is a blind grinding mode, prone to defects such as over-cutting, under-grinding, or damage to the rail base material. While single-arm rust removal robots improve automation to some extent, they only have a single actuator and must sequentially complete processes such as rail top rust removal, rail bottom rust removal, and end face grinding, resulting in long work paths and slow processing cycles. A more prominent problem is that the rail web surface often has raised characters formed by rolling or marking. When these raised characters happen to be in the clamping position of the welding machine's clamping device during flash welding, it severely weakens the contact uniformity and clamping strength of the clamping surface, causing the rail to slip or misalign during welding. Existing traditional equipment and single-arm robots cannot automatically identify and handle this physical interference problem, still requiring manual grinding, which is inefficient and inconsistent.

[0004] Based on the above, how to achieve intelligent recognition and automated removal of raised characters on the rail web during rail welding pretreatment, as well as efficient parallel operation of multiple processes, is a problem to be solved in this technical field. Summary of the Invention

[0005] To address the aforementioned deficiencies or improvement needs of existing technologies, this application provides a vision-force-controlled dual-arm robot system and method for pre-welding rust removal and grinding of rails. This system employs a technique where dual-arm robots are arranged side-by-side on both sides of the rail conveyor line. It uses dual-camera partitioned joint scanning to reconstruct three-dimensional point clouds to identify the rail's spatial pose, end angle, and the position of the raised characters on the rail web. Combined with interference quantization judgment based on clamping area mapping and a dynamic task scheduling strategy for both arms, it achieves the technical effects of precise removal of the raised characters on the rail web and efficient parallel operation of both arms during pre-welding processing.

[0006] The embodiments of this application adopt the following technical solutions: In a first aspect, this application provides a vision-force-controlled dual-arm robot system for pre-welding rust removal and grinding of steel rails, comprising: The first robot and the second robot are respectively positioned on both sides of the rail conveyor line; The vision system includes a first industrial camera mounted on the end flange of a first robot and a second industrial camera mounted on the end flange of a second robot. The first industrial camera and the second industrial camera are used to perform a joint scan of the rail in sections before the grinding operation to obtain multi-view image data. The host computer is connected to the vision system. The host computer is used to reconstruct a three-dimensional point cloud based on the multi-view image data to generate a three-dimensional spatial model of the rail. Based on the three-dimensional spatial model, the host computer identifies the actual spatial position of the rail, the inclination angle of the end section, and the presence and position coordinates of the raised characters at the rail web. According to the pre-stored clamping area parameters of the welding clamping device of the welding machine, the host computer maps the clamping area to the rail web surface area on the three-dimensional spatial model. When the geometric envelope of the detected raised characters overlaps with the mapping range of the clamping area, it is determined to be clamping interference, and the raised character grinding depth parameters are generated. The host computer is also used to perform dynamic task scheduling for the two arms based on the recognition results of the convex characters: when clamping interference is detected, the second robot prioritizes the grinding task of the convex character grinding area based on the convex character grinding depth parameter, and after completion, it switches to the auxiliary rust removal mode to work with the first robot to grind the upper and lower rust removal areas in parallel; when clamping interference is detected, the second robot directly enters the auxiliary rust removal mode; the first robot dynamically corrects the grinding path based on the actual spatial position of the rail and the tilt angle of the end section, and performs grinding tasks of the end rust removal area and the upper and lower rust removal areas.

[0007] Through the above technical solution, the dual-arm robot, in conjunction with the vision system, performs zoned joint scanning and 3D point cloud reconstruction. The host computer, based on the 3D spatial model, achieves accurate perception of the rail spatial pose and intelligent detection of the raised characters on the rail web. It also performs quantitative judgment of the interference of the raised characters by mapping the clamping area. Combined with the dynamic task scheduling of the dual arms, it solves the technical problems of the inability to automatically identify and remove the raised characters on the rail web, the lack of high-precision spatial perception leading to blind grinding, and the low efficiency of single-arm serial operation in the pre-welding processing of rails in one stop.

[0008] In some embodiments, the end rust removal area, the upper and lower rust removal areas, and the raised lettering grinding area are spatially connected; the first robot dynamically updates the grinding path based on the actual spatial position of the rail and the tilt angle of the end section identified by the vision system, and is responsible for roughening and fitting grinding of the end rust removal areas on the rear end face of the front rail and the front end face of the rear rail to ensure the physical fit of the ends of the two rails; the second robot dynamically plans the cutting path based on the position of the raised lettering on the rail web and the grinding depth parameters identified by the vision system, and is responsible for grinding the raised lettering on the surface of the rail web in the raised lettering grinding area to eliminate interference with the welding clamping device.

[0009] Through the above technical solution, the end rust removal area, the upper and lower rust removal areas and the raised lettering grinding area are spatially connected. The two robots dynamically correct their respective grinding paths based on visual recognition data, realizing a precise division of labor for end face roughening and fitting grinding and rail waist raised lettering grinding. This avoids over-cutting or under-grinding caused by blind grinding, ensuring end fit and welding clamping strength.

[0010] In some implementations, when the first robot and the second robot are working on the upper and lower rust removal areas, the two robots adopt time-staggered dynamic interference anti-collision control. The host computer calculates the interference prediction area based on the real-time pose information of the two robots and their respective planned paths, and adjusts the entry sequence of one robot or pauses and waits so that the two robots alternately enter and exit in the shared work space to avoid physical collisions.

[0011] Through the above technical solution, the two robots adopt time-staggered dynamic interference anti-collision control in the shared upper and lower rust removal areas. Based on real-time pose information, interference prediction and timing adjustment are performed. While ensuring the parallel and efficient operation of the two arms, physical collisions between the robots are effectively avoided, ensuring the safety and reliability of the system operation.

[0012] In some embodiments, both the first robot and the second robot are equipped with automatic quick-change grinding heads on their end flanges. The system also includes a first force control electrical box and a second force control electrical box that are electrically connected to the first robot and the second robot, respectively. Each of the first and second force control electrical boxes integrates a constant force feedback control unit. The constant force feedback control unit collects grinding contact force signals in real time and, in conjunction with the rail surface normal vector data pre-acquired by the vision system, corrects the end pose and feed of the first robot or the second robot online, so that the automatic quick-change grinding head always maintains normal constant force contact with the rail surface.

[0013] Through the above technical solution, the constant force feedback control unit combines the real-time force signal with the surface normal vector data obtained by the vision system to correct the robot end pose online. This achieves normal constant force contact control of the grinding head at the irregular curved surfaces of the rail web and rail bottom, effectively avoiding over-cutting or under-grinding problems caused by uneven contact force and protecting the rail base material.

[0014] In some embodiments, a tool magazine is provided within the working envelope range reachable by the first robot and the second robot, and the tool magazine has multiple limiting slots for storing grinding heads of different specifications; The tool magazine includes a first tool magazine and a second tool magazine that are set up independently. The first tool magazine is arranged within the working envelope range that the first robot can reach, and the second tool magazine is arranged within the working envelope range that the second robot can reach, so that the two robots can perform tool changing operations independently.

[0015] With the above technical solution, two tool magazines are independently set up around their respective robots, so that when one robot is performing a grinding operation, the other robot can simultaneously perform a tool changing operation at its own tool magazine, or both can change tools together, realizing parallel tool changing, eliminating tool changing waiting time, and further shortening the overall operation cycle.

[0016] In some implementations, the host computer is also used to control the corresponding robot to drive the industrial camera on it to perform a back-inspection scan of the polished area after the polishing operation in each polishing area is completed, and compare the back-inspection image with the preset polishing quality standard. If residual rust or embossed characters are detected, a re-polishing path instruction is generated and the corresponding robot is controlled to perform the re-polishing operation.

[0017] Through the above technical solution, visual inspection and quality comparison are carried out after each grinding area is completed. For unqualified areas, a secondary grinding path is automatically generated for secondary grinding, realizing closed-loop control of grinding quality, ensuring the process quality and consistency of pre-welding treatment, and preventing unqualified products from flowing into subsequent welding processes.

[0018] In some implementations, when the host computer determines that there is clamping interference, it further calculates the protrusion height of the embossed character protruding from the rail reference surface. When the protrusion height exceeds a preset threshold allowable amount, the embossed character grinding depth parameter is equal to the protrusion height minus the threshold allowable amount, so as to grind the surface of the embossed character to a degree that does not affect the clamping device's fit.

[0019] The above technical solution quantifies and determines the grinding depth based on the difference between the protrusion height of the raised letter and the threshold allowable amount, thereby achieving precise and controllable removal of the raised letter. This ensures the fit requirements of the clamping surface and avoids damage to the rail web base material due to excessive grinding, thus improving the precision of the grinding process.

[0020] In some embodiments, when the first industrial camera and the second industrial camera perform a joint scan of different zones, they move and take pictures of the rail end area, rail top and bottom area and rail waist area from different angles. The host computer then fuses the image data collected by the two industrial cameras from multiple perspectives to generate the three-dimensional spatial model.

[0021] Through the above technical solution, two industrial cameras simultaneously perform zone scanning, making full use of the spatial advantages of the dual arms to achieve full coverage imaging of the rail end and rail web area. Multi-view fusion improves the accuracy and completeness of 3D point cloud reconstruction, providing a more accurate data foundation for subsequent path planning.

[0022] In some embodiments, the automatic quick-change grinding head includes an end face grinding head, a flat rust removal grinding head, and a curved embossed grinding head, which are respectively used for roughening and fitting grinding of the end rust removal area, rust removal grinding of the upper and lower rust removal areas, and smoothing grinding of the embossed grinding area; the first robot and the second robot automatically change grinding heads of different specifications in the tool magazine through the automatic quick-change interface on the end flange.

[0023] Through the above technical solution, special grinding heads are configured for different grinding requirements of end faces, flat surfaces and curved surfaces, and quick tool changing without manual intervention is achieved through automatic quick-change interface. This allows the system to adapt to a variety of different grinding process requirements in the same workstation, improving the flexibility and efficiency of the operation.

[0024] Secondly, this application provides a method for pre-welding rust removal and grinding of rails using a vision-force-controlled dual-arm robot, applied to the vision-force-controlled dual-arm robot pre-welding rust removal and grinding system described in the first aspect, comprising: The first and second robots are controlled to drive the first and second industrial cameras to perform a joint scan of the rail in sections. The host computer then performs 3D point cloud reconstruction based on the multi-view image data to identify the actual spatial position of the rail, the tilt angle of the end section, and the presence and position coordinates of the raised characters at the rail web. The host computer maps the clamping area according to the pre-stored clamping area parameters of the welding clamping device of the welding machine, determines whether the embossed character interferes with the clamping area, and generates the embossed character grinding depth parameters when interference exists. The host computer performs dynamic task scheduling for the two arms based on the judgment results: when there is clamping interference, the second robot prioritizes grinding the convex grinding area, and switches to auxiliary rust removal mode after completion; when there is no clamping interference, the second robot directly enters the auxiliary rust removal mode; the first robot dynamically corrects the grinding path based on the actual spatial position of the rail and the tilt angle of the end section, and performs grinding of the end rust removal area and the upper and lower rust removal areas. During the grinding process, the grinding contact force signal is collected in real time. Combined with the rail surface normal vector data pre-acquired by the vision system, the robot end pose is corrected online to maintain constant normal force contact. After each polishing area is completed, the corresponding robot is controlled to drive the industrial camera to perform a back-inspection scan, generate a re-polishing path instruction for the unqualified area and execute the re-polishing.

[0025] Through the above technical solutions, the entire process from visual scanning, interference judgment, dynamic task scheduling, constant force grinding to re-inspection and re-grinding is automated, achieving the technical effects of intelligent removal of rail web protrusions, multi-area adaptive constant force grinding, and efficient parallel operation of dual arms in the pre-welding treatment of rails.

[0026] In summary, this application includes at least the following beneficial technical effects: 1. By using an industrial camera at the end of a dual-arm robot to perform zoned joint scanning and 3D point cloud reconstruction, and based on the clamping area mapping, intelligent recognition and interference quantification of the raised characters on the rail web are achieved. For the first time, automated detection and precise removal of raised characters are realized in the pre-welding treatment of rails. This eliminates the uneven contact and rail slippage hazards caused by the raised characters in the clamping device of the welding machine, and fills the gap in the existing technology where the raised characters on the rail web cannot be automatically recognized and processed. 2. By coupling the constant force feedback control unit in the force control electrical box with the surface normal vector data obtained by the vision system, the robot end pose is corrected in real time during the grinding process to achieve normal constant force contact. Combined with the closed-loop quality control mechanism of visual back inspection and re-grinding after grinding, the problems of over-cutting, under-grinding and damage to the base material caused by blind grinding are effectively avoided, and the accuracy and consistency of rail pre-welding grinding are significantly improved. 3. Through the dual-arm dynamic task scheduling strategy, time-staggered dynamic interference anti-collision control, and parallel tool changing mechanism supported by independent tool magazine, the two robots achieve parallel and efficient operation of multiple processes such as end grinding, upper and lower rust removal and embossed lettering removal in the same workstation. Compared with single-arm serial operation, the overall processing cycle is significantly shortened, meeting the efficiency requirements of high-speed rail welding production line. Attached Figure Description

[0027] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments of this application will be briefly described below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 A top view of the overall structure of a vision-force-controlled dual-arm robot rail pre-welding rust removal and grinding system provided in this application embodiment; Figure 2This is a schematic block diagram of the robot end effector structure provided in the embodiments of this application; Figure 3 A schematic diagram of the system workflow provided in the embodiments of this application; Figure 4 This is a schematic diagram of the interference determination logic for convex characters provided in an embodiment of this application; Figure 5 This is a schematic diagram of dynamic scheduling of dual-arm tasks provided in an embodiment of this application; Figure 6 A flowchart illustrating a vision-force-controlled dual-arm robot method for pre-welding rust removal and grinding of rails, provided as an embodiment of this application.

[0029] In the diagram: 1. Excavation pit; 2. Outer protective cover; 3. Control cabinet; 4. Electrical cabinet; 5. Shore clamping device; 6. First robot; 7. Second robot; 8. First force control electrical box; 9. Second force control electrical box; 10. Tool magazine; 11. Dust collector; 12. Liquid cooler; 13. Refrigerated dryer; 14. Upper and lower rust removal areas; 15. End rust removal area; 16. Raised lettering grinding area. Detailed Implementation

[0030] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application. Furthermore, the technical features involved in the various embodiments described below can be combined with each other as long as they do not conflict with each other.

[0031] The present application will now be described in detail with reference to the accompanying drawings and embodiments. Example 1

[0032] It should be noted that most existing rail pre-welding equipment consists of traditional rust removal machines or single-arm rust removal robots, which lack intelligent visual perception capabilities and cannot automatically identify the actual spatial position deviation of the rail and the raised characters on the rail web. Furthermore, the efficiency of single-arm serial operation is low, making it difficult to meet the dual requirements of high-speed rail welding production lines for both the quality and efficiency of pre-welding treatment.

[0033] Based on the above problems, this embodiment 1 provides a vision-force-controlled dual-arm robot rail pre-welding rust removal and grinding system, referencing... Figure 1 As shown, the system includes a foundation pit 1, an outer protective cover 2, an operating cabinet 3, an electrical cabinet 4, a shore clamping device 5, a first robot 6, a second robot 7, a first force control electrical box 8, a second force control electrical box 9, a tool magazine 10, a dust collector 11, a liquid cooler 12, and a refrigerated dryer 13.

[0034] Specifically, pit 1 is located within the working area of ​​the rail conveyor line, providing a foundation and working space for the two robots. An outer protective cover 2 surrounds the entire working area, enclosing pit 1 to isolate dust and debris generated during grinding and protect operator safety. An operating cabinet 3, located inside or outside the outer protective cover 2 within operator reach, is used for system start / stop control, parameter setting, and status monitoring. An electrical cabinet 4 houses the system's electrical components and control circuits. A dust collector 11 connects to the working space inside the outer protective cover 2 to collect and filter metal dust particles generated during grinding in real time. A liquid cooler 12 dissipates heat generated during equipment operation; a refrigerated dryer 13 provides dry, clean compressed air to ensure stable equipment operation.

[0035] The bank-side clamping devices 5 are installed on both sides of the rail conveyor line, including clamping devices on both sides of the front rail and clamping devices on both sides of the rear rail. During rust removal and grinding, only one rail needs to be processed; that is, the front and rear rails are fed into the rail conveyor line separately, clamped by their respective clamping devices, and then the rust removal and grinding operation is performed. The processing of the front and rear rails needs to be sequential. Either the front rail enters the station first for processing, or the rear rail enters first. After one rail is processed, it is moved to the next station (i.e., the station where the next welding step will take place; the welding operation is not related to this plan and will not be elaborated here), and then the other rail is processed at this station. After both rails are processed, they are aligned at the welding station, and the welding operation is performed.

[0036] The first robot 6 and the second robot 7 are respectively positioned on both sides of the rail conveyor line. Both robots are multi-axis industrial robots with sufficient working envelope to cover the entire area of ​​the rail to be processed. (Reference) Figure 2 As shown, both the first robot 6 and the second robot 7 are equipped with automatic quick-change interfaces on their end flanges, allowing for the rapid installation and removal of automatic quick-change grinding heads. The automatic quick-change grinding heads include three specifications: end-face grinding heads, flat rust-removing grinding heads, and curved-surface embossed grinding heads. The end-face grinding heads are used for roughening and grinding the end sections of rails, the flat rust-removing grinding heads are used for removing oxide layers and rust from the top and bottom surfaces of the rails, and the curved-surface embossed grinding heads are used for precisely grinding the embossed characters on the rail web.

[0037] The vision system includes a first industrial camera mounted on the end flange of the first robot 6 and a second industrial camera mounted on the end flange of the second robot 7. Figure 2(These are uniformly labeled as industrial cameras). Both industrial cameras are high-resolution industrial-grade cameras that can move and take pictures from multiple angles with the help of a robotic arm, achieving omnidirectional imaging of the rail surface. It should be noted that the installation position of the industrial cameras is not particularly limited, as long as they can be fixed on the robotic arm and take pictures.

[0038] The host computer communicates with the vision system and runs vision recognition software to receive image data from two industrial cameras and perform subsequent calculations such as 3D reconstruction, feature recognition, path planning, and task scheduling.

[0039] The first force control electrical box 8 is electrically connected to the first robot 6, and the second force control electrical box 9 is electrically connected to the second robot 7. Each of the two force control electrical boxes integrates a constant force feedback control unit, which is used to collect the grinding contact force signal between the corresponding grinding head and the rail surface in real time, and control the position and feed of the robot end effector through closed-loop feedback to achieve active constant force control grinding.

[0040] Tool magazine 10 includes independently configured first and second tool magazines. (Reference) Figure 1 As shown, the first tool magazine is arranged around the first robot 6, and the second tool magazine is arranged around the second robot 7. Both tool magazines are located within the working envelope of their respective robots. Each tool magazine 10 has multiple limiting slots for storing grinding heads of different specifications, such as end face grinding heads, flat rust removal grinding heads, and curved surface embossing grinding heads. The independent design of the two tool magazines 10 allows one robot to perform grinding operations while the other robot performs tool changing operations at its own tool magazine, or both robots can perform tool changing operations simultaneously, achieving parallel tool changing and reducing waiting time.

[0041] refer to Figure 1 As shown, the rail grinding areas that the system needs to process include the end rust removal area 15, the upper and lower rust removal areas 14, and the raised lettering grinding area 16. The end rust removal area 15, the upper and lower rust removal areas 14, and the raised lettering grinding area 16 are spatially connected, meaning that the robot can continuously perform grinding operations on different areas without leaving the workspace, achieving seamless processing.

[0042] The above solution, through a system structure of side-by-side dual-arm robots, end vision integrated with automatic quick-change grinding heads, force control electrical box and independent tool magazine, achieves integrated processing capabilities for multiple processes such as end grinding, top and bottom rust removal and lettering removal in the same workstation, providing a hardware foundation for subsequent intelligent vision recognition, adaptive path planning and dual-arm parallel operation. Example 2

[0043] It should be noted that rails may experience actual positional deviations during long-distance transport. The cutting at the ends may introduce a tilt angle to the cross-section, and there may be raised characters on the surface of the rail web. Traditional equipment cannot detect these geometric features, leading to blind grinding operations.

[0044] Based on the above problems, this embodiment 2 obtains the geometric features of the rail through visual scanning. (Reference) Figure 3 When the system is working, after the rail is delivered to the position in step S1 and the shore clamping device is aligned and locked in step S2, the system enters the visual scanning stage as shown in steps S3 and S4.

[0045] During the joint scanning process, the first and second industrial cameras, driven by their respective robots, moved and photographed the rail from different angles. The two industrial cameras worked together, fully utilizing the advantages of their different spatial positions to achieve full-coverage imaging of the area of ​​the rail to be processed.

[0046] After receiving multi-view image data from two industrial cameras, the host computer processes it using built-in vision recognition software. First, the host computer fuses the image data from different perspectives, generating a 3D spatial model of the rail end and web region based on a 3D point cloud reconstruction algorithm. This 3D spatial model accurately reflects the actual geometry and spatial position of the rail.

[0047] Based on the three-dimensional spatial model, the host computer further identifies the following key information: (1) the deviation between the actual spatial position and the theoretical position of the rail, including lateral deviation, longitudinal deviation and height deviation; (2) the tilt angle of the end section: the tilt angle of the rear end face of the front rail and the front end face of the rear rail relative to the vertical plane of the rail axis, which is introduced by the sawing process; (3) the presence and position coordinates of the raised characters at the rail web: identify whether there are raised characters formed by rolling or marking on the surface of the rail web, and if so, obtain its geometric center coordinates, outer contour range and protrusion height and other geometric parameters.

[0048] The above solution uses dual-camera partitioned joint scanning and 3D point cloud reconstruction to accurately acquire the actual spatial pose and surface geometric feature information of the rail, providing an accurate data foundation for subsequent adaptive path planning and convex interference determination, and solving the problem of blind grinding caused by the lack of spatial perception capability of traditional equipment. Example 3

[0049] It should be noted that the raised characters on the rail web surface do not all affect welding quality. Uneven contact and insufficient clamping strength only occur when the raised characters are precisely within the clamping area of ​​the welding machine's clamping device. Current technology cannot intelligently distinguish between raised characters that require grinding and those that do not, resulting in either entirely manual grinding (low efficiency) or complete neglect (potential quality risks).

[0050] To solve the above problems, refer to Figure 3 Middle step S5 and Figure 4 As shown, after the establishment of the three-dimensional spatial model and the recognition of the convex characters are completed, the host computer performs convex character interference quantization judgment based on the clamping region mapping.

[0051] Specifically, the host computer pre-stores the clamping area parameters of the welding clamping device (i.e., the clamping device of the butt welding machine). These parameters define the position and range of the actual clamping contact area of ​​the clamping device on the rail web. Based on these clamping area parameters, the host computer maps the clamping area to the rail web surface area on the three-dimensional space model and marks the boundary of the mapped area of ​​the clamping device in the three-dimensional space model.

[0052] like Figure 4 As shown on the left, the host computer performs spatial overlap analysis on the geometric envelope of the identified convex character (i.e., the area enclosed by the outer contour boundary of the convex character) and the mapping range of the clamping area. When the geometric envelope of the convex character overlaps with the mapping range of the clamping area, the host computer determines that the convex character will interfere with the welding clamping, i.e., it is determined to be a clamping interference state. When the geometric envelope of the convex character is completely outside the mapping range of the clamping area, the convex character will not affect the clamping effect, and it is determined to be a non-interference state, so no convex character grinding is required.

[0053] like Figure 4 As shown below, when clamping interference is detected, the host computer further calculates the protrusion height h of the embossed character protruding from the rail web reference surface. The system has a preset threshold tolerance. When the protrusion height h exceeds this threshold tolerance, the host computer calculates the embossed character grinding depth parameter, which is equal to the protrusion height h minus the threshold tolerance. The goal of setting this grinding depth parameter is to grind the embossed character surface flat to a degree that does not affect the clamping device's fit, ensuring both the flatness requirements of the clamping surface and avoiding excessive grinding that could damage the rail web base material.

[0054] By mapping the clamping area and quantifying the interference of the raised characters, the system can intelligently distinguish between interference raised characters that need to be ground and non-interference raised characters that do not need to be ground, and accurately calculate the required grinding depth. This achieves refined and quantitative control of raised character removal, ensuring welding clamping strength while minimizing unnecessary grinding operations and damage to the base material. Example 4

[0055] It should be noted that the pre-welding treatment of rails involves multiple processes such as end grinding, top and bottom rust removal, and removal of raised lettering. The distribution of raised lettering varies among different rails, so the tasks of the two robots need to be dynamically allocated based on the actual inspection results to achieve optimal work efficiency.

[0056] To achieve the above objectives, refer to Figure 3 Steps S5 to S7 and Figure 5 As shown, the host computer performs dual-arm dynamic task scheduling for the first and second robots based on the result of the convex character interference determination.

[0057] like Figure 5 As shown in scenario one, when the host computer determines that clamping interference exists: the second robot is scheduled to prioritize the embossed lettering grinding task. Based on the embossed lettering grinding depth parameters and embossed lettering position coordinate data generated by the host computer, it dynamically plans the cutting path, installs the curved embossed lettering grinding head to precisely grind the embossed lettering grinding area, and grinds the embossed characters at the rail web clamping position flat. During this period, the first robot simultaneously performs the grinding task of the end rust removal area (end face roughening and fitting grinding). Based on the actual spatial position of the rail and the tilt angle of the end section identified by the host computer, it dynamically corrects the grinding path to ensure the physical fit of the two rail end faces. After the second robot completes the embossed lettering grinding task, it automatically switches to a flat rust removal grinding head at the second tool magazine, switches to auxiliary rust removal mode, and works with the first robot to perform parallel grinding of the upper and lower rust removal areas (upper rust removal area A1 and lower rust removal area A2). It should be noted that upper rust removal is also rail top rust removal, and lower rust removal is rail bottom rust removal.

[0058] like Figure 5 As shown in scenario two, when the host computer determines that there is no clamping interference: the second robot does not need to perform the embossed grinding task and directly enters the auxiliary rust removal mode. From the start of the grinding operation, it works in parallel with the first robot to grind the upper and lower rust removal areas. At this time, because the two robots operate in parallel throughout the entire process, the overall operation cycle is shorter than in scenario one.

[0059] In both cases, the first robot always undertakes the main grinding tasks of the end rust removal area (roughing and bonding grinding of the rear end face of the front rail and the front end face of the rear rail) and the upper and lower rust removal areas, ensuring that the end face bonding and welding conductivity meet the process requirements.

[0060] The above solution employs a dual-arm dynamic task scheduling strategy, enabling the system to intelligently allocate tasks between the two robots based on the actual convex distribution of each rail. When there is interference from convex characters, the critical convex character grinding task is prioritized, while when there is no interference, the second robot is fully utilized to improve rust removal efficiency. This achieves flexibility and optimization in task allocation, minimizing the overall operation cycle time. Example 5

[0061] It should be noted that when grinding the upper and lower rust removal areas in parallel, there is an overlapping area in the workspace of the two robots. If proper anti-collision control is not implemented, the arms or grinding heads of the two robots may collide physically, causing equipment damage and safety accidents.

[0062] To avoid the above situation, refer to Figure 3 The annotation of step S7 and Figure 5 As shown in the time-staggered anti-collision label, when the first robot and the second robot need to work simultaneously in the upper and lower rust removal areas, the two robots adopt a time-staggered dynamic interference anti-collision control strategy.

[0063] Specifically, the host computer acquires the current pose information (including joint angles and end-effector spatial coordinates) of the first and second robots in real time, and predicts the motion trajectories of the two robots over a future period based on their respective planned grinding paths. The host computer performs interference analysis on the predicted trajectories and calculates the interference prediction region, that is, the spatial region and time interval where the motion envelopes of the two robots may intersect.

[0064] When an impending interference is detected, the host computer adjusts the entry sequence of one of the robots to avoid it. For example, when the first robot is grinding the upper rust removal area A1 on the top surface of the rail, and the second robot is about to enter the same area, the host computer issues a pause and wait command, causing the second robot to wait in a safe position until the first robot exits the area, at which point the second robot can then enter and perform its grinding operation. Through this staggered approach, the two robots alternately enter and exit the shared upper and lower rust removal areas, ensuring the efficiency of parallel operations while effectively avoiding physical collisions.

[0065] The above solution achieves safe parallel operation of the two robots in a shared workspace through time-staggered dynamic interference anti-collision control. It ensures the safety of system operation without sacrificing work efficiency and avoids equipment damage and safety hazards caused by collisions between the two arms. Example 6

[0066] It should be noted that the web and bottom of the rail have complex irregular curved surface geometry. When grinding these curved areas, if the contact force between the grinding head and the rail surface is uneven or unstable, it can easily lead to over-cutting (excessive contact force can cause excessive cutting and damage to the base material) or under-grinding (insufficient contact force can cause inadequate grinding), which will affect the quality of pre-welding treatment.

[0067] To avoid the above situation, refer to Figure 2As shown, a force sensor is installed between the robot's end flange and the automatic quick-change grinding head. This force sensor is connected to the corresponding force control electrical box (either the first or second force control electrical box). The constant force feedback control unit integrated inside the force control electrical box collects the grinding contact force signal between the grinding head and the rail surface in real time through the force sensor, including normal and tangential force components.

[0068] During the grinding process, the constant force feedback control unit compares the real-time grinding contact force signal collected with the preset constant force target value to calculate the force deviation. Simultaneously, the constant force feedback control unit obtains the rail surface normal vector data pre-calculated by the vision system during the scanning phase from the host computer. This normal vector data describes the normal direction at various locations on the rail surface. The constant force feedback control unit integrates the force deviation information and the surface normal vector data to online correct the robot end effector's pose (including position and attitude angle) and feed rate, ensuring that the working surface of the grinding head always applies a constant contact force along the normal direction of the rail surface.

[0069] Especially when grinding the raised areas of the rail waist and the bottom surface, the constant force feedback control unit needs to adjust the robot's end effector pose more frequently to track the changes in the surface normal vector, thus maintaining constant normal force contact, due to the significant changes in surface curvature in these areas. The surface normal vector data pre-acquired by the vision system plays a feedforward compensation role in this process, enabling the force control system to perceive the changing trend of the surface shape in advance and reduce contact force fluctuations caused by force control delay.

[0070] The above solution achieves constant force contact of the grinding head on irregular curved surfaces such as the rail web and rail bottom through constant force grinding control with force-vision coupling. The contact force is stable and uniform during the grinding process, effectively avoiding over-cutting and under-grinding, protecting the rail base material from damage, and improving the process quality and consistency of pre-welding grinding. Example 7

[0071] It should be noted that even though the system uses visual guidance and force-controlled grinding, there may still be individual locations where the grinding effect does not meet the standard due to reasons such as wear of the grinding head or uneven hardness of the rail surface. A quality assurance mechanism is needed to ensure that the pre-welding treatment quality of each rail meets the standard requirements.

[0072] To achieve the above objectives, refer to Figure 3 As shown in steps S8 and S9, after the grinding operation in each grinding area is completed, the system automatically enters the post-grinding visual inspection stage.

[0073] Specifically, the host computer controls the corresponding robot to move the industrial camera on its end flange to above the polished area and perform a re-inspection scan and photograph of the polished area. After receiving the re-inspection image data, the host computer compares and analyzes it with preset polishing quality standards. The polishing quality standards include, but are not limited to: whether the surface roughness and roughening angle of the end rust removal area meet the standards, whether the rust removal rate of the upper and lower rust removal areas meets the conductivity requirements, and whether the residual height of the raised lettering in the raised lettering polishing area is below the threshold allowable amount.

[0074] If the re-inspection results show that all ground areas are qualified, the pre-welding treatment of the rail is complete. The system instructs the shore clamping device to release the rail, and the rail continues to be transported to the subsequent welding station via the conveyor line. If the re-inspection results show unqualified areas such as residual rust or embossed lettering, the host computer automatically generates a re-grinding path instruction for the unqualified areas and controls the corresponding robot to perform the re-grinding operation. After the re-grinding is completed, the system performs another re-inspection scan until all areas meet the grinding quality standards.

[0075] By implementing a visual inspection and re-grinding mechanism after grinding, closed-loop control of pre-welding grinding quality is achieved, ensuring that the pre-welding treatment quality of each rail meets the standard requirements, preventing unqualified products from flowing into subsequent welding processes, and improving the first-pass yield rate of welding and the service safety of the rail. Example 8

[0076] Based on the vision-controlled dual-arm robot rail pre-welding rust removal and grinding system provided in the foregoing embodiments, Embodiment 8 of this application provides a vision-controlled dual-arm robot rail pre-welding rust removal and grinding method. (See reference...) Figure 6 As shown, the method includes the following steps: Step 100: Use the shore clamping device to align and rigidly lock the rail to be processed.

[0077] Step 200: Control the first robot and the second robot to drive the first industrial camera and the second industrial camera to perform a joint scan of the rail in sections. The host computer performs three-dimensional point cloud reconstruction based on the multi-view image data to identify the actual spatial position of the rail, the tilt angle of the end section, and the presence and position coordinates of the raised characters at the rail web.

[0078] Step 300: The host computer maps the clamping area according to the clamping area parameters of the welding clamping device, determines whether the embossed character interferes with the clamping area, and generates the embossed character grinding depth parameters when interference exists.

[0079] Step 400: The host computer performs dynamic task scheduling for the two arms based on the judgment result: When there is clamping interference, the second robot prioritizes grinding the convex grinding area, and switches to auxiliary rust removal mode after completion; when there is no clamping interference, the second robot directly enters the auxiliary rust removal mode; the first robot dynamically corrects the grinding path based on the actual spatial position of the rail and the tilt angle of the end section, and performs grinding of the end rust removal area and the upper and lower rust removal areas.

[0080] Step 500: During the grinding process, the first force control electrical box and the second force control electrical box collect the grinding contact force signal in real time. Combined with the rail surface normal vector data pre-acquired by the vision system, the robot end pose is corrected online to maintain constant normal force contact.

[0081] Step 600: After each grinding area is completed, control the corresponding robot to drive the industrial camera to perform a back-inspection scan, generate a re-grinding path instruction for the unqualified area and execute the re-grinding.

[0082] The detailed implementation process of the above method has been described in the foregoing embodiments, and will not be repeated here.

[0083] The above description is merely a preferred embodiment of this application and is not intended to limit this application. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the protection scope of this application.

Claims

1. A vision-force-controlled dual-arm robot rail pre-welding rust removal and grinding system, characterized in that, include: The first robot and the second robot are respectively positioned on both sides of the rail conveyor line; The vision system includes a first industrial camera mounted on the end flange of a first robot and a second industrial camera mounted on the end flange of a second robot. The first industrial camera and the second industrial camera are used to perform a joint scan of the rail in sections before the grinding operation to obtain multi-view image data. The host computer is connected to the vision system. The host computer is used to generate a three-dimensional spatial model of the rail by reconstructing three-dimensional point cloud based on the multi-view image data, and to identify the actual spatial position of the rail, the tilt angle of the end section, and the presence and position coordinates of the raised characters at the rail web based on the three-dimensional spatial model. And based on the pre-stored clamping area parameters of the welding clamping device of the welding machine, the clamping area is mapped to the rail web surface area on the three-dimensional space model. When the geometric envelope of the detected convex character overlaps with the mapping range of the clamping area, it is determined to be clamping interference, and the convex character grinding depth parameters are generated. The host computer is also used to perform dynamic task scheduling for the two arms based on the recognition results of the convex characters: when clamping interference is detected, the second robot prioritizes the grinding task of the convex character grinding area based on the convex character grinding depth parameter, and after completion, it switches to the auxiliary rust removal mode to work with the first robot to grind the upper and lower rust removal areas in parallel; when clamping interference is detected, the second robot directly enters the auxiliary rust removal mode; the first robot dynamically corrects the grinding path based on the actual spatial position of the rail and the tilt angle of the end section, and performs grinding tasks of the end rust removal area and the upper and lower rust removal areas.

2. The vision-force-controlled dual-arm robot rail pre-welding rust removal and grinding system according to claim 1, characterized in that, The end rust removal area, the upper and lower rust removal areas, and the raised letter grinding area are spatially connected. The first robot dynamically updates the grinding path based on the actual spatial position of the rail and the tilt angle of the end section identified by the vision system, and is responsible for roughening and fitting grinding of the end rust removal areas on the rear end face of the front rail and the front end face of the rear rail to ensure the physical fit of the ends of the two rails. The second robot dynamically plans the cutting path based on the position of the raised letter on the rail web and the grinding depth parameters identified by the vision system, and is responsible for grinding the raised letter on the surface of the rail web in the raised letter grinding area to eliminate interference with the welding clamping device.

3. The vision-force-controlled dual-arm robot rail pre-welding rust removal and grinding system according to claim 1, characterized in that, When the first robot and the second robot are working on the upper and lower rust removal areas, the two robots adopt dynamic interference anti-collision control with time staggering. The host computer calculates the interference prediction area based on the real-time pose information of the two robots and their respective planned paths, and adjusts the entry sequence of one of the robots or pauses and waits so that the two robots alternately enter and exit in the shared work space to avoid physical collisions.

4. The vision-force-controlled dual-arm robot rail pre-welding rust removal and grinding system according to claim 1, characterized in that, Both the first robot and the second robot are equipped with automatic quick-change grinding heads on their end flanges. The system also includes a first force control electrical box and a second force control electrical box that are electrically connected to the first robot and the second robot, respectively. Each of the first force control electrical box and the second force control electrical box integrates a constant force feedback control unit. The constant force feedback control unit collects grinding contact force signals in real time and, in conjunction with the rail surface normal vector data pre-acquired by the vision system, corrects the end pose and feed of the first robot or the second robot online, so that the automatic quick-change grinding head always maintains normal constant force contact with the rail surface.

5. The vision-force-controlled dual-arm robot rail pre-welding rust removal and grinding system according to claim 4, characterized in that, A tool magazine is provided within the working envelope range reachable by the first robot and the second robot, and the tool magazine has multiple limiting slots for storing grinding heads of different specifications; The tool magazine includes a first tool magazine and a second tool magazine that are set up independently. The first tool magazine is arranged within the working envelope range that the first robot can reach, and the second tool magazine is arranged within the working envelope range that the second robot can reach, so that the two robots can perform tool changing operations independently.

6. The vision-force-controlled dual-arm robot rail pre-welding rust removal and grinding system according to claim 1, characterized in that, The host computer is also used to control the corresponding robot to drive the industrial camera on it to perform a back-inspection scan of the polished area after the polishing operation in each polishing area is completed, and compare the back-inspection image with the preset polishing quality standard. If residual rust or embossed characters are detected, a re-polishing path instruction is generated and the corresponding robot is controlled to perform the re-polishing operation.

7. The vision-force-controlled dual-arm robot rail pre-welding rust removal and grinding system according to claim 1, characterized in that, When the host computer determines that there is clamping interference, it further calculates the protrusion height of the embossed character protruding from the rail waist reference surface. When the protrusion height exceeds the preset threshold allowable amount, the embossed character grinding depth parameter is equal to the protrusion height minus the threshold allowable amount, so as to grind the surface of the embossed character to a degree that does not affect the clamping device's fit.

8. The vision-force-controlled dual-arm robot rail pre-welding rust removal and grinding system according to claim 1, characterized in that, When the first industrial camera and the second industrial camera perform a joint scan of the rail end area, the rail top and bottom area and the rail waist area from different angles, the host computer moves and takes pictures of the rail end area, the rail top and bottom area and the rail waist area respectively. The host computer then fuses the image data collected by the two industrial cameras from multiple perspectives to generate the three-dimensional spatial model.

9. The vision-force-controlled dual-arm robot rail pre-welding rust removal and grinding system according to claim 5, characterized in that, The automatic quick-change grinding head includes an end face grinding head, a flat rust removal grinding head, and a curved embossed grinding head, which are respectively used for roughening and fitting grinding of the end rust removal area, rust removal grinding of the upper and lower rust removal areas, and smoothing grinding of the embossed grinding area; the first robot and the second robot automatically change grinding heads of different specifications in the tool magazine through the automatic quick-change interface on the end flange.

10. A method for pre-welding rust removal and grinding of rails using a vision-force-controlled dual-arm robot, applied to the pre-welding rust removal and grinding system for rails using a vision-force-controlled dual-arm robot as described in any one of claims 1-9, characterized in that... include: The first and second robots are controlled to drive the first and second industrial cameras to perform a joint scan of the rail in sections. The host computer then performs 3D point cloud reconstruction based on the multi-view image data to identify the actual spatial position of the rail, the tilt angle of the end section, and the presence and position coordinates of the raised characters at the rail web. The host computer maps the clamping area according to the pre-stored clamping area parameters of the welding clamping device of the welding machine, determines whether the embossed character interferes with the clamping area, and generates the embossed character grinding depth parameters when interference exists. The host computer performs dynamic task scheduling for the two arms based on the judgment results: when there is clamping interference, the second robot prioritizes grinding the convex grinding area, and switches to auxiliary rust removal mode after completion; when there is no clamping interference, the second robot directly enters the auxiliary rust removal mode; the first robot dynamically corrects the grinding path based on the actual spatial position of the rail and the tilt angle of the end section, and performs grinding of the end rust removal area and the upper and lower rust removal areas. During the grinding process, the grinding contact force signal is collected in real time. Combined with the rail surface normal vector data pre-acquired by the vision system, the robot end pose is corrected online to maintain constant normal force contact. After each polishing area is completed, the corresponding robot is controlled to drive the industrial camera to perform a back-inspection scan, generate a re-polishing path instruction for the unqualified area and execute the re-polishing.

Images