Panoramic scanning based teach-less robot welding system and welding control method

Abstract

The application is based on a panoramic scanning teaching-free robot welding system and a welding control method, which comprises remote process preprocessing, on-site workpiece loading, global visual panoramic scanning positioning, fine positioning visual weld seam correction, intelligent welding path planning and parameter calling, robot automatic welding, and workpiece unloading process; the welding system comprises an automatic welding mechanism, including an inverted robot, a ground rail walking mechanism, and a welding system; a visual identification system, including a global visual unit and a fine positioning visual unit; a software system, including welding intelligent planning software, welding intelligent execution software, and an integrated control system; and three parts are cooperated to complete the work. In the panoramic scanning stage, workpiece model identification and pose detection are strengthened to ensure positioning accuracy; in the weld processing stage, through double visual cooperation, the influence of workpiece error on weld quality is reduced; in the welding process, the path and the called process parameters are strictly executed according to the software self-planning, and the influence of human factors on the welding quality is reduced.

Description

Technical Field

[0001] This invention belongs to the field of automatic welding technology, specifically relating to a teachless robotic welding system based on panoramic scanning and its welding control method. Background Technology

[0002] In the manufacturing process of railway freight cars, the welding production of small components, such as brake cylinder mounting brackets and 120 valve mounting brackets, traditionally relies mainly on manual operation, which presents the following technical problems: Manual welding is highly dependent on the skills of operators, and the difference in skill levels among different operators leads to unstable weld quality, easily resulting in defects such as undercut, weld beads, porosity, and weld deviation, which do not meet the welding quality standards for railway freight car parts; The welding equipment for small components in the material preparation and production area is old, and its condition is still not good after major repairs, requiring frequent maintenance to barely maintain production, resulting in low equipment reliability and affecting production continuity; Manual welding is inefficient, with long working hours for welding a single batch of small components and requiring a large number of welding operators, resulting in high annual labor costs and making it difficult to meet the needs of mass production of railway freight cars; The diverse and small-batch characteristics of small components mean that traditional robot welding requires manual teaching programming for different workpieces, which is time-consuming and lacks flexibility when switching workpiece models, making it unsuitable for discrete manufacturing scenarios.

[0003] While robotic welding has been attempted for welding small components of railway freight cars, traditional robotic welding, though improving efficiency to some extent, relies on professional personnel for teaching and programming. Changing workpiece models involves lengthy teaching cycles, making it unsuitable for the diverse, small-batch production characteristics of these components. Furthermore, existing welding equipment lacks effective panoramic scanning and intelligent path planning capabilities, making it difficult to address issues such as large overall workpiece errors, complex weld seam shapes leading to inaccurate positioning, and difficulties in weld seam correction. Consequently, it fails to meet the high-quality, high-efficiency welding requirements for small components of railway freight cars.

[0004] Therefore, we are conducting research on the application of teach-free robot programming technology based on panoramic scanning in the welding of small parts of railway freight cars, in order to improve welding quality, increase production efficiency, and reduce labor costs. Summary of the Invention

[0005] This invention discloses a teach-free robot welding system and its welding control method based on panoramic scanning, addressing the shortcomings of existing technologies. The purpose of this invention is to provide a welding system and its welding control application method based on panoramic scanning teach-free robot programming technology, which addresses the characteristics of railway freight car small parts being diverse, produced in small batches, and requiring high welding quality. This solves the problems of unstable quality, low efficiency, and high cost of traditional manual welding, as well as the need for manual teaching of traditional robots, thereby achieving automated and high-quality welding of small parts.

[0006] This invention is achieved through the following technical solution:

[0007] A teach-free robotic welding system based on panoramic scanning is characterized by comprising an automatic welding mechanism, a vision recognition system, and a software control system.

[0008] The automated welding mechanism includes an inverted robot, a ground-rail walking mechanism, and a welding system;

[0009] The visual recognition system includes a global visual unit and a fine-positioning visual unit;

[0010] The software control system includes intelligent welding planning software, intelligent welding execution software, and integrated control system.

[0011] Furthermore, the automatic welding mechanism includes: a ground rail fixed to the ground of the work area and a mobile platform running on the ground rail. The mobile platform is provided with a vertical support column and a horizontal cantilever beam. The welding robot is inverted and fixed under the horizontal cantilever beam. The welding gun of the welding system is provided at the end of the welding robot's mechanical arm.

[0012] Furthermore, the welding system includes a welding power source, a cooling water tank, a wire feeder, a torch cleaning station, and an anti-collision welding torch.

[0013] Furthermore, the global vision unit of the vision recognition system is fixed at the end of the transverse cantilever beam and consists of an industrial camera and a laser line structured light for panoramic scanning of the workpiece; the precision positioning vision unit of the vision recognition system is integrated at the front end of the robot arm and uses a high frame rate camera with more than 5 million pixels for acquisition of the micro-contour of the weld and real-time correction.

[0014] Furthermore, the software control system consists of welding intelligent planning software, welding intelligent execution software, and an integrated control system. The planning software supports the import of 3D models and automatic extraction of weld information, the execution software controls the robot's actions, the integrated system realizes the coordinated scheduling of hardware and vision, and reserves a communication interface for the MES manufacturing execution system to support equipment status monitoring and data traceability.

[0015] This invention also discloses a welding control method using the above-mentioned panoramic scanning-based teachless robot welding system, comprising: performing a panoramic scan of the workpiece to be welded using a global vision unit to obtain a three-dimensional model of the workpiece and weld information; generating a welding path planning scheme using intelligent planning software based on the weld information and calling pre-stored welding process parameters; controlling the inverted robot to move along the ground track walking mechanism to the predetermined welding position, and correcting the weld position using a precision positioning vision unit; adjusting the robot welding torch posture according to the correction result and performing automatic welding operations.

[0016] Specifically, the following methods are included:

[0017] (1) Remote process preprocessing: Import the three-dimensional model of the workpiece through intelligent planning software. The software automatically extracts the number, type and size parameters of the weld. Based on the process database, it recommends welding current, voltage and welding speed. After manual confirmation, it is saved to the model library.

[0018] (2) On-site workpiece loading: The small parts to be welded are hoisted to the work area by a crane, and the workpiece loading position is adaptive within a range of 5.5m×2.5m;

[0019] (3) Global vision panoramic scanning and positioning: The automatic welding mechanism moves to drive the global vision unit to move laterally, collect the three-dimensional point cloud data of the workpiece, automatically identify the workpiece model and detect the pose, including the X / Y / Z axis coordinates and attitude angle; the position error of the same specification workpiece is required to be ≤20mm, and the global position matrix is ​​called.

[0020] (4) Precision positioning vision correction of weld seam: The robot on the automatic welding mechanism carries a precision positioning vision unit, scans the weld seam according to the path of global vision planning, obtains micro three-dimensional contour information, and autonomously corrects the welding path deviation, adapting to the scenario where the workpiece assembly error is ≤10mm.

[0021] (5) Intelligent path planning and parameter calling: The integrated control system analyzes the precise positioning data, calls the corresponding workpiece process parameters and welding torch trajectory in the model library, and automatically generates the welding program;

[0022] (6) Automatic welding by robot: Start the program, the robot welds according to the planned path, and the cleaning station cleans foreign objects from the welding gun, applies silicone oil and cuts wire in real time;

[0023] (7) Workpiece unloading: After welding is completed, the system sends a signal and the crane transfers the workpiece to the storage area to complete one welding cycle.

[0024] Furthermore, the global visual panoramic scanning and positioning includes:

[0025] Scanning path optimization: The "partition scanning + stitching algorithm" is adopted to divide the working area into 500mm×500mm sub-areas. The ground rail drives the camera to move in an "S-shaped" path. The scanning time of a single area is ≤8s. After stitching the entire area, the point cloud resolution reaches 0.1mm and the positioning error is ≤±1mm.

[0026] Workpiece recognition algorithm: Based on the improved SIFT algorithm, extract feature points on the workpiece surface and match them with feature templates in the model library;

[0027] Pose calculation model: Establish a "point cloud data - world coordinate system" transformation model, fit the deviation between the actual pose of the workpiece and the theoretical model by the least squares method, and output the translation amount of the X / Y / Z axes and the rotation angle around the three axes to provide the initial path for precise positioning scanning.

[0028] Furthermore, the precise positioning and visual weld seam correction includes:

[0029] High Dynamic Range Imaging: HDR high dynamic range technology is used to suppress welding arc interference. The camera exposure time is adaptively adjusted within 10 to 100 μs to ensure that the weld area (dark area) and the workpiece surface (bright area) are clearly imaged at the same time.

[0030] Weld contour extraction: Based on the Canny edge detection algorithm and morphological filtering, the weld groove and molten pool edge features are extracted from the image, and the three-dimensional coordinates of each point on the weld centerline are calculated. The contour extraction accuracy is ±0.05mm.

[0031] Real-time correction control: Adopting a "position feedback + feedforward compensation" control strategy, the precision positioning vision unit collects one frame of image for every 1mm movement of the robot, analyzes the path deviation and outputs correction instructions. The correction response time is ≤50ms, ensuring that the welding torch is always aligned with the center of the weld, adapting to scenarios where the workpiece assembly error is ≤10mm.

[0032] Furthermore, the intelligent path planning and parameter calling includes: using the model library matching function of the intelligent planning software to automatically load the pre-stored welding program of the same specification workpiece, adopting a dynamic path planning algorithm to adjust the welding trajectory in real time to adapt to the workpiece pose change, and automatically compensating for weld geometric deviation through welding current feedback closed-loop control.

[0033] The present invention provides a teaching-free robotic welding system and welding control method based on panoramic scanning, which is particularly suitable for welding scenarios of small parts of railway freight cars, such as brake cylinder hangers, 120 valve hangers, and medium-sized parts of export cars. It provides a process means and measures to solve the problems of low efficiency, unstable quality, and high dependence on operators in manual welding.

[0034] The welding system and its control welding method of the present invention utilize a global vision unit to achieve panoramic scanning, recognition and positioning of the workpiece, and combine it with a precision positioning vision unit to complete the accurate correction of the weld seam. Combined with an intelligent software system, the welding path is autonomously planned and process parameters are called, eliminating the need for manual teaching and programming, and realizing automated and high-quality welding of small parts of railway freight cars.

[0035] Specifically addressing the characteristics of discrete manufacturing of railway freight car parts—characterized by diverse types, small batches, significant overall workpiece errors, and complex weld seam patterns—as well as the challenges of traditional manual welding relying on skilled operators, aging equipment requiring frequent maintenance, and welding quality being easily affected by human factors, this study optimizes three aspects: welding equipment architecture, visual recognition system, and software control process. Corresponding countermeasures and specific application methods are formulated: The hardware architecture of an inverted robot + ground rail expands the operating range; dual recognition using global vision + precise positioning vision ensures positioning and weld seam accuracy; and intelligent planning and execution software enables teach-free programming, controlling positioning errors and weld seam defects during the welding process. This achieves the goal of meeting the welding quality requirements of railway freight car parts without manual teaching, thereby improving production efficiency and reducing labor costs.

[0036] The welding system and welding method of this invention can identify whether competitors are using this technology by detecting the weld quality of the welded workpiece (such as appearance and dimensional accuracy), tracing the equipment operation log (welding parameters and path planning records), comparing the application effects of traditional manual welding or other robotic welding technologies; at the same time, the core algorithm and process parameter model library of the software system are unique and can be used as a basis for technology identification.

[0037] The welding system and welding control method of this invention employ a global vision unit for panoramic scanning of the workpiece, achieving automatic workpiece identification, position and orientation matching, and intelligent planning of a precise positioning scanning path, eliminating the need for manual workpiece positioning. The robot's front end is equipped with a precision positioning vision unit, enabling accurate scanning, positioning, and tracking of the weld seam, with real-time deviation correction to ensure weld quality, without requiring manual adjustment of the welding torch position.

[0038] This invention's welding system and welding control method support the automatic extraction of weld information, matching of process parameters, and calculation of welding torch trajectory from imported 3D models, achieving manual teaching and programming-free operation and adapting to the welding of various small parts. The software system has a reserved communication interface for connection to enterprise information networks, facilitating equipment status monitoring and data management. The modification and integration costs are controllable, and the application effect is significant. It enables equipment operation status monitoring, fault alarms, and process parameter management, facilitating production management and data traceability. The system adopts an inverted robot + ground rail hardware architecture, with an effective ground rail travel of 5.5 meters and a robot arm span ≥2000mm, expanding the working space and meeting the welding needs of small parts of different specifications.

[0039] This invention's welding system and welding control method overcome the difficulties of unstable quality in manual welding of small parts for railway freight cars and the need for manual teaching of traditional robots. It achieves automated welding without teaching, significantly improving welding quality and avoiding defects such as undercut, weld beads, and porosity. The welding method is suitable for welding various types of small-batch, small-part components. Remote process preprocessing allows for rapid switching of workpiece models without repeated teaching, improving production flexibility and efficiency. The welding system's hardware architecture is stable and reliable. The ground rail and inverted robot are guaranteed for rigidity and accuracy through finite element analysis. The vision system has an IP54 protection rating, adapting to complex environments such as electromagnetic interference, arc light, and spatter at the welding site. Attached Figure Description

[0040] Figure 1 This is a schematic diagram of the overall architecture of the welding system of the present invention.

[0041] In the diagram, 1 is the welding torch, 2 is the precision scanning unit, 3 is the global scanning unit, 4 is the robotic arm, 5 is the track, and 6 is the centralized control unit. Detailed Implementation

[0042] The present invention will be further described below with reference to specific embodiments. These specific embodiments are further explanations of the principles of the present invention and are not intended to limit the present invention in any way. Any technology that is the same as or similar to the present invention does not exceed the scope of protection of the present invention.

[0043] The present invention relates to a teachless robotic welding system based on panoramic scanning, which consists of an automatic welding mechanism, a vision recognition system, and a software control system. The automatic welding mechanism includes an inverted robot, a ground track walking mechanism, and a welding system. The vision recognition system includes a global vision unit and a precision positioning vision unit. The software control system includes intelligent welding planning software, intelligent welding execution software, and an integrated control system.

[0044] The welding system consists of three parts: an automatic welding mechanism, a vision recognition system, and a software control system. During the panoramic scanning stage, it enhances workpiece model recognition and posture detection to ensure positioning accuracy. During the weld treatment stage, it reduces the impact of workpiece errors on weld quality through dual vision coordination. During the welding process, it strictly follows the path planned by the software and the process parameters called, reducing the impact of human factors on welding quality.

[0045] The welding method of the teachless robot welding system based on panoramic scanning of the present invention includes remote process preprocessing, on-site workpiece loading, global visual panoramic scanning positioning, precise positioning visual weld seam correction, intelligent welding path planning and parameter calling, automatic robot welding, and workpiece unloading.

[0046] The following describes in detail the teach-free robot welding system and welding method based on panoramic scanning of the present invention.

[0047] I. System Overall Design.

[0048] 1. System Architecture. Based on the collaborative concept of "hardware support - visual perception - software control", a teach-free robotic welding system is designed. The overall architecture is divided into three parts, such as... Figure 1 As shown.

[0049] (1) Automatic welding mechanism: The welding system is configured with an inverted robot and a ground rail walking mechanism, including a welding power source, a cooling water tank, a wire feeder, a torch cleaning station, and an anti-collision welding torch. The ground rail has an effective travel of 5.5m, which can realize a large range of lateral movement; the inverted robot has an arm span of ≥2000mm and adopts an inverted installation method to expand the vertical working space, which is suitable for workpiece working areas of 5.5m×2.5m×2.5m;

[0050] (2) Visual recognition system: including a global vision unit and a fine positioning vision unit. The global vision unit is installed on the inverted L-shaped column beam and uses a 2-megapixel industrial camera + laser line structured light to realize panoramic scanning of the workpiece; the fine positioning vision unit is integrated at the front end of the robot and uses a 5-megapixel high frame rate camera to complete the acquisition of the micro-contour of the weld and real-time correction.

[0051] (3) Software control system: It consists of welding intelligent planning software, welding intelligent execution software and integrated control system. The planning software supports the import of three-dimensional models and automatic extraction of weld information, the execution software controls the robot's actions, the integrated system realizes the coordinated scheduling of hardware and vision, and reserves a communication interface for MES (Manufacturing Execution System) to support equipment status monitoring and data traceability.

[0052] 2. Core process flow method. A seven-step automated process flow method is adopted, and the entire welding process of welded components is free from manual intervention.

[0053] (1) Remote process preprocessing: Import the workpiece 3D model through intelligent planning software (supports Tekla, STEP and other formats), the software automatically extracts the number, type (fillet weld / butt weld) and size parameters of welds, and recommends welding current (180-220A), voltage (22-26V) and welding speed (300-500mm / min) based on the process database. After manual confirmation, it is saved to the model library. Only one preprocessing is required for workpieces of the same specification.

[0054] (2) On-site workpiece loading: The small parts to be welded are hoisted to the work area by a crane. Precise positioning is not required. The workpiece loading position is adaptive within a range of 5.5m×2.5m.

[0055] (3) Global vision panoramic scanning and positioning: The ground rail drives the global vision unit to move laterally, collect the three-dimensional point cloud data of the workpiece, automatically identify the workpiece model and detect the pose (X / Y / Z axis coordinates and attitude angle). If the position error of the same specification workpiece is ≤20mm, only one scan is required, and the global position matrix can be directly called afterward.

[0056] (4) Precision positioning vision correction of weld seam: The robot carries a precision positioning vision unit, scans the weld seam according to the path of global vision planning, obtains micro three-dimensional contour information, and autonomously corrects the welding path deviation, adapting to the scenario where the workpiece assembly error is ≤10mm.

[0057] (5) Intelligent path planning and parameter calling: The integrated control system analyzes the precise positioning data, calls the corresponding workpiece process parameters and welding torch trajectory in the model library, and automatically generates the welding program;

[0058] (6) Automatic welding by robot: Start the program and the robot welds according to the planned path. The cleaning station cleans foreign objects from the welding gun, applies silicone oil and cuts wire in real time. The anti-collision sensor will stop the machine immediately when it is triggered. It supports sway welding (sway amplitude 0-5mm) and multi-layer multi-pass welding functions, and the welding accuracy reaches ±0.5mm.

[0059] (7) Workpiece unloading: After welding is completed, the system sends a signal and the crane transfers the workpiece to the storage area to complete one welding cycle.

[0060] The global visual panoramic scanning positioning technology of this invention can achieve "large-scale, high-speed, and high-precision" workpiece positioning; the specific method adopts:

[0061] Scanning path optimization: The "partition scanning + stitching algorithm" is adopted to divide the 5.5m×2.5m working area into 11 500mm×500mm sub-areas. The ground rail drives the camera to move in an "S" shaped path. The scanning time of a single area is ≤8s. After stitching the entire area, the point cloud resolution reaches 0.1mm and the positioning error is ≤±1mm.

[0062] Workpiece recognition algorithm: Based on the improved SIFT (Scale Invariant Feature Transform) algorithm, it extracts feature points on the workpiece surface, such as bolt holes and edge contours, and matches them with feature templates in the model library. The recognition accuracy is ≥99.5%, and it is adaptable to interference such as oxide scale and oil stains on the workpiece surface.

[0063] Pose calculation model: Establish a "point cloud data - world coordinate system" transformation model, fit the deviation between the actual pose of the workpiece and the theoretical model by the least squares method, and output the translation amount of the X / Y / Z axes and the rotation angle around the three axes to provide the initial path for precise positioning scanning.

[0064] The precision positioning visual weld seam correction technology of this invention can solve the problems of "weld seam micro-contour extraction" and "real-time path correction"; the specific implementation method is as follows:

[0065] High Dynamic Range Imaging: HDR (High Dynamic Range) technology is used to suppress welding arc interference. The camera exposure time can be adaptively adjusted within 10 to 100 μs to ensure that the weld area (dark area) and the workpiece surface (bright area) are clearly imaged simultaneously.

[0066] Weld contour extraction: Based on the Canny edge detection algorithm and morphological filtering, features such as weld groove and molten pool edge are extracted from the image, and the three-dimensional coordinates of each point on the weld centerline are calculated. The contour extraction accuracy reaches ±0.05mm.

[0067] Real-time correction control: Adopting the "position feedback + feedforward compensation" control strategy, the precision positioning vision unit collects one frame of image for every 1mm movement of the robot, analyzes the path deviation and outputs correction instructions. The correction response time is ≤50ms, ensuring that the welding torch is always aligned with the center of the weld, adapting to scenarios where the workpiece assembly error is ≤10mm.

[0068] The teach-free programming and process parameter matching technology of this invention can be adapted to a variety of small parts; the specific method adopts:

[0069] Automatic extraction of weld information: When the intelligent planning software analyzes the 3D model, it identifies welds through the "face-edge-volume" topological relationship, distinguishes between types such as fillet welds and butt welds, and automatically calculates parameters such as weld length and weld leg size, improving extraction efficiency by 80% compared to manual annotation;

[0070] Intelligent matching of process parameters: Establish a database of "workpiece material - weld type - process parameters" which includes welding parameter templates for commonly used railway freight car materials such as Q345 steel and Q450. It supports adaptive adjustment of parameters based on workpiece thickness (3-10mm). For example, the current increases by 5-8A for every 1mm increase in thickness.

[0071] Automatic welding torch trajectory generation: Based on the three-dimensional coordinates of the weld seam, the software uses the "Cartesian space interpolation" algorithm to generate the welding torch trajectory, including the transition path planning of the arc starting point and arc ending point, to avoid collision between the welding torch and the workpiece. The trajectory generation time is ≤10s / piece, and no manual teaching is required.

[0072] II. Analysis of Application Effects

[0073] 1. Test Conditions and Indicators. A test platform was set up in the welding workshop for small parts of railway freight cars. The test workpieces were selected as brake cylinder hangers made of Q345 steel with 6 welds and 120 valve hangers made of Q345 steel with 4 welds. The application effects of traditional manual welding, traditional robot welding and this technology were compared. The test indicators included welding quality, production efficiency and cost-effectiveness.

[0074]

[0075] As can be seen from the table above, manual welding is fast but labor-intensive, with unstable weld quality, worker fatigue, and safety risks. Dedicated machine welding requires manual grinding, is inconvenient for production changes, has poor flexibility, requires manual programming, has fast welding speed, requires tooling and fixtures, has high requirements for material accuracy and assembly gap, has two workstations, can only weld two at a time, cannot be left unattended, requires frequent loading and unloading, and its productivity is not effectively developed.

[0076] Intelligent welding systems offer several advantages: they eliminate the need for programming and teaching, and require no tooling or fixtures for positioning, thus reducing production costs; they enable remote model processing of intelligent welding, allowing software to autonomously extract weld seams and generate job tasks; they eliminate the need for long-term worker supervision at the welding site, fully leveraging employee productivity; and they offer multi-functionality, enabling rapid production changes and enhancing flexible intelligent welding capabilities.

[0077] 2. Improved Welding Quality. The welding quality of the welding system and its control method of this invention, as tested according to the "Technical Conditions for Welding of Railway Freight Cars" (TB / T 1580), is as follows:

[0078] Appearance quality: The weld surface is smooth and free of defects such as undercut (depth ≤ 0.5mm), weld beads, and porosity. The appearance qualification rate is 99.2%, which is 14.2 percentage points higher than manual welding (85%) and 7.2 percentage points higher than traditional robot welding (92%).

[0079] Dimensional accuracy: Weld width deviation ≤ ±0.3mm, weld leg size deviation ≤ ±0.2mm, overall welding accuracy reaches ±0.5mm, meeting the first-class accuracy requirements for welding small parts of railway freight cars.

[0080] Mechanical properties: Tensile test shows that the weld tensile strength is ≥490MPa and the impact energy (-40℃) is ≥34J, both of which are better than the performance requirements of Q345 steel base material.

[0081] 3. Production efficiency, cost-effectiveness, and environmental adaptability.

[0082] Production efficiency: The welding time for a single brake cylinder hanger has been reduced from 25 minutes by manual welding to 8 minutes, increasing efficiency by 68%; the workpiece model switching time has been reduced from 2 to 4 hours by traditional robots to ≤10 minutes, significantly improving flexible production capabilities.

[0083] Cost-effectiveness: A single workstation can reduce the number of welding operators by 2, increase the effective operating rate of the equipment from 70% of the old equipment to 92%, and reduce downtime for maintenance by about 500 hours per year.

[0084] Environmental adaptability: The vision system has an IP54 protection rating and uses a "lens dust cover + airflow purging" design to prevent splash contamination. In the arc light and electromagnetic interference environment of the welding site, the failure rate during continuous operation is ≤2%, and the equipment stability meets the needs of 24-hour continuous production.

[0085] In summary, the environmental system and welding control method of this invention, through an integrated design of "hardware architecture optimization - dual vision collaboration - software intelligent planning," solves the core problems of unstable quality, low efficiency, and poor flexibility in the welding of small components for railway freight cars. The dual vision system achieves seamless integration of "global positioning - precise positioning," achieving a welding accuracy of ±0.5mm and increasing the appearance qualification rate to 99.2%. The no-teach programming technology eliminates reliance on manual teaching, reducing workpiece changeover time by more than 95%, and is suitable for multi-variety, small-batch discrete manufacturing scenarios. It saves annual costs per workstation, has a short investment payback period, and combines technological advancement with economic feasibility.

Claims

1. A welding method for a teach-free robotic welding system based on panoramic scanning, characterized in that: The welding was completed using a teach-free robotic welding system, which consists of an automatic welding mechanism, a vision recognition system, and a software control system. The automatic welding mechanism includes an inverted robot, a ground rail walking mechanism, and a welding system. The vision recognition system includes a global vision unit and a precision positioning vision unit. The software control system includes intelligent welding planning software, intelligent welding execution software, and an integrated control system. Welding methods include: (1) Remote process preprocessing: Import the three-dimensional model of the workpiece through intelligent planning software. The software automatically extracts the number, type and size parameters of the weld. Based on the process database, it recommends welding current, voltage and welding speed. After manual confirmation, it is saved to the model library. (2) On-site workpiece loading: The small parts to be welded are hoisted to the work area by a crane, and the workpiece loading position is adaptive within a range of 5.5m×2.5m; (3) Global Vision Panoramic Scanning and Positioning: The automatic welding mechanism moves, driving the global vision unit to move laterally, collecting the workpiece's three-dimensional point cloud data, automatically identifying the workpiece model and detecting its pose, including X / Y / Z axis coordinates and attitude angles; requiring the position error of workpieces of the same specification to be ≤20mm, calling the global position matrix; specifically including: Scanning path optimization: The "partition scanning plus stitching algorithm" is adopted to divide the working area into 500mm×500mm sub-areas. The ground rail drives the camera to move in an "S"-shaped path. The scanning time of a single area is ≤8s. After stitching the entire area, the point cloud resolution reaches 0.1mm and the positioning error is ≤±1mm. Workpiece recognition algorithm: Based on the improved SIFT algorithm, extract feature points on the workpiece surface and match them with feature templates in the model library; Pose calculation model: Establish a "point cloud data and world coordinate system" transformation model, fit the deviation between the actual pose of the workpiece and the theoretical model by the least squares method, and output the X / Y / Z axis translation and rotation angle around the three axes to provide the initial path for fine positioning scanning; (4) Precision positioning vision correction of weld seams: The robot on the automatic welding mechanism carries a precision positioning vision unit, scans the weld seam according to the path planned by the global vision, obtains microscopic three-dimensional contour information, and autonomously corrects the welding path deviation, adapting to scenarios where the workpiece assembly error is ≤10mm; specifically including: High Dynamic Range Imaging: HDR high dynamic range technology is used to suppress welding arc interference, and the camera exposure time is adaptively adjusted within 10 to 100 μs to ensure clear imaging of the weld area and the workpiece surface at the same time. Weld contour extraction: Based on the Canny edge detection algorithm and morphological filtering, the weld groove and molten pool edge features are extracted from the image, and the three-dimensional coordinates of each point on the weld centerline are calculated. The contour extraction accuracy is ±0.05mm. Real-time correction control: The "position feedback plus feedforward compensation" control strategy is adopted. For every 1mm the robot moves, the precision positioning vision unit collects one frame of image, analyzes the path deviation and outputs a correction command. The correction response time is ≤50ms, which ensures that the welding torch is always aligned with the center of the weld and is suitable for scenarios with workpiece assembly error ≤10mm. (5) Intelligent path planning and parameter calling: The integrated control system analyzes the precise positioning data, calls the corresponding workpiece process parameters and welding torch trajectory in the model library, and automatically generates the welding program; (6) Automatic welding by robot: Start the program, the robot welds according to the planned path, and the cleaning station cleans foreign objects from the welding torch, applies silicone oil and cuts wire in real time; (7) Workpiece unloading: After welding is completed, the system sends a signal and the crane transfers the workpiece to the storage area to complete one welding cycle.

2. The welding method of the teachless robot welding system based on panoramic scanning according to claim 1, characterized in that: The intelligent path planning and parameter calling includes: using the model library matching function of the intelligent planning software to automatically load the pre-stored welding program of the same specification workpiece, adopting the dynamic path planning algorithm to adjust the welding trajectory in real time to adapt to the workpiece posture change, and automatically compensating for weld geometric deviation through welding current feedback closed-loop control.

3. The welding method of the teachless robot welding system based on panoramic scanning according to claim 2, characterized in that: The automatic welding mechanism includes: a ground rail fixed to the ground of the work area and a mobile platform running on the ground rail. The mobile platform is equipped with a vertical support column and a horizontal cantilever beam. The welding robot is inverted and fixed under the horizontal cantilever beam. The welding gun of the welding system is installed at the end of the welding robot's robotic arm.

4. The welding method of the teachless robot welding system based on panoramic scanning according to claim 2, characterized in that: The welding system includes a welding power source, a cooling water tank, a wire feeder, a torch cleaning station, and an anti-collision welding torch.

5. The welding method of the teachless robot welding system based on panoramic scanning according to claim 2, characterized in that: The global vision unit of the vision recognition system is fixed at the end of the transverse cantilever beam and consists of an industrial camera and laser line structured light for panoramic scanning of the workpiece. The precision positioning vision unit of the vision recognition system is integrated at the front end of the robot arm and uses a high frame rate camera with more than 5 million pixels for acquisition of weld micro-contours and real-time correction.

6. The welding method of the teachless robot welding system based on panoramic scanning according to claim 2, characterized in that: The software control system consists of welding intelligent planning software, welding intelligent execution software, and an integrated control system. The planning software supports the import of 3D models and automatic extraction of weld information, the execution software controls the robot's movements, the integrated system realizes the coordinated scheduling of hardware and vision, and reserves a communication interface for the MES manufacturing execution system to support equipment status monitoring and data traceability.

Images