Welding robot welding seam automatic extraction and path planning system

Abstract

The invention relates to an automatic welding seam extraction and path planning system for a welding robot. The automatic welding seam extraction and path planning system comprises a three-dimensional model analysis module, an automatic welding seam extraction and feature calculation module, a welding key point and welding gun posture generation module, a path planning and reachability verification module and a data output and state management module. The welding seam automatic extraction and feature calculation module automatically identifies a welding seam based on a geometric topological relation and extracts geometric and topological features of the welding seam; the welding key point and welding gun posture generation module generates a welding path point, an arrival point, a departure point and a touch locating point according to the welding seam characteristics, and calculates a corresponding welding gun space posture; the path planning and reachability verification module performs multi-constraint verification on the executable performance of the welding path; and the data output and state management module is used for uniformly managing and outputting the welding path and the multi-state judgment result thereof. The system is high in welding path performability, high in adaptability to complex structures, high in system expandability, high in engineering efficiency and high in practicability.

Description

Technical Field

[0001] This invention belongs to the field of industrial robots and intelligent manufacturing technology, specifically relating to an automatic weld seam extraction and path planning system for welding robots. Background Technology

[0002] As global manufacturing rapidly advances towards automation and intelligence, the application of industrial robots, especially welding robots, continues to grow in the manufacturing sector. According to the International Federation of Robotics (IFR) 2025 report, the global installed base of industrial robots reached approximately 542,000 units in 2024, exceeding 500,000 units for the fourth consecutive year. The Chinese market accounted for approximately 54% of the global new deployments, while Asia as a whole accounted for 74%. By the end of 2024, the cumulative number of industrial robots in operation globally reached approximately 4,664,000 units, an increase of approximately 9% over the previous year.

[0003] Among industrial robot equipment, welding robots are a typical application, and their market size and application penetration rate continue to expand. According to data from Fortune Business Insights, the global welding robot market size was approximately US$736 million in 2024 and is projected to grow to approximately US$810 million by 2025, maintaining a high compound annual growth rate in the coming years. Meanwhile, other market research indicates that the global welding robot market size was approximately US$1.084 billion in 2024 and is projected to reach US$3.114 billion by 2034, with a compound annual growth rate of approximately 10.2% from 2025 to 2034.

[0004] From an industry application perspective, the penetration rate of welding robots in key manufacturing sectors is continuously increasing. Research data shows that over 30% of global welding robot deployments are used in the automotive industry, where the industry's demands for consistent welding quality and high production speeds are driving a significant acceleration in the automation of manual welding. Meanwhile, according to a Chinese industry monitoring report, the overall penetration rate of welding robots in China's manufacturing sector is projected to increase from approximately 42% in 2024 to approximately 58% in 2025.

[0005] Despite the rapid growth in the scale and penetration rate of welding robot equipment, significant technical limitations remain in practical engineering applications. Traditional welding robot programming primarily relies on manual teaching, where operators guide the welding torch along the weld seam point by point using a teach pendant, recording each critical pose point. This method is not only time-consuming and inefficient but also highly dependent on operator experience; skill differences between operators can easily lead to inconsistent welding quality. Furthermore, when workpiece structures are complex, the number of weld seams is large, or product models change frequently, manual teaching is almost inadequate for meeting the demands of flexible production.

[0006] To improve the programming efficiency of welding robots, several offline programming and automated path generation methods have emerged in recent years. These methods are typically based on 3D CAD models, generating welding paths by analyzing the workpiece geometry. However, existing technologies still face many problems and shortcomings.

[0007] First, in terms of weld identification, some existing methods rely on manual annotation of welds in CAD models, which has a low degree of automation and still has a high cost of human intervention. Other methods are based on simple geometric boundaries or feature lines for identification, which makes it difficult to accurately distinguish between real welds and non-welded structural boundaries. In complex assembly models, they are prone to generating a large number of misidentification results and lack a systematic weld screening and organization mechanism.

[0008] Secondly, in terms of calculating weld geometry and welding posture, existing technologies often only focus on extracting the weld centerline, while neglecting to consider the normal relationship between adjacent weld surfaces, the included angle, the welding direction, and the process posture. This leads to problems such as unreasonable posture, welding torch interference, or unstable welding quality in the actual execution of the generated welding path. Especially in spatially tortuous welds or multi-faceted intersecting structures, the continuity and stability of the welding torch posture are difficult to guarantee.

[0009] Furthermore, regarding welding robot path planning and reachability verification, some existing systems only generate geometric paths without in-depth verification of robot kinematics, lacking systematic analysis of inverse kinematics solvability, joint constraints, singularities, and collision risks. This leads to frequent issues such as "path unreachable" or "safety hazards" in the generated welding paths during simulation or field execution, severely impacting system usability. In addition, existing methods provide relatively coarse feedback on path planning results, making it difficult to provide engineers with effective decision-making support.

[0010] Furthermore, in the touch positioning and error calibration operations commonly used in the welding preparation stage, existing technologies mostly rely on manual setting of touch points, resulting in low automation and difficulty in ensuring calibration accuracy and repeatability, which restricts the application of welding robots in high-precision welding scenarios.

[0011] Therefore, there is an urgent need for an automatic weld seam extraction and path planning system for welding robots that can automatically complete weld seam extraction, key point and welding torch pose calculation based on a 3D model, and perform path planning and accessibility verification in combination with robot kinematics, so as to realize the efficient, intelligent and engineering application of the welding programming process. Summary of the Invention

[0012] The purpose of this invention is to address the common problems in existing welding robot programming processes, such as reliance on manual weld seam identification, low teaching efficiency, insufficient reliability of automatic welding path generation, and incomplete path reachability verification. This invention proposes an automatic weld seam extraction and path planning system for welding robots. This system, without requiring manual teaching or manual weld seam annotation, can automatically complete weld seam extraction, weld seam geometric feature analysis, and calculation of key welding points and welding torch pose based on a 3D workpiece model. It also combines the kinematic constraints of the welding robot to verify the reachability of the welding path, thereby significantly improving the efficiency of welding robot programming, reducing manual intervention costs, and enhancing the engineering feasibility of welding paths and the stability of welding quality.

[0013] This invention provides an automatic weld seam extraction and path planning system for a welding robot, comprising: a 3D model input and analysis module, a geometric topology construction module, an automatic weld seam extraction and feature calculation module, a welding key point and welding torch posture generation module, a path planning and reachability verification module, and a data output and status management module; the output of the 3D model input and analysis module is connected to the input of the geometric topology construction module, the output of the geometric topology construction module is connected to the input of the automatic weld seam extraction and feature calculation module, the output of the automatic weld seam extraction and feature calculation module is connected to the input of the welding key point and welding torch posture generation module, the output of the welding key point and welding torch posture generation module is connected to the input of the path planning and reachability verification module, and the output of the path planning and reachability verification module is connected to the input of the data output and status management module; wherein: The 3D model input and parsing module is used to read the 3D model data of the workpiece and construct the geometric topology structure, which is then transmitted to the geometric topology construction module. The automatic weld extraction and feature calculation module includes an automatic weld extraction module and a weld geometric feature calculation module. The automatic weld extraction module can read and parse 3D model data (STEP format), automatically identify the contact boundaries between workpieces, and extract geometric intersections that may constitute welds. It filters and selects welds through multi-level rules, including constraints on the number of faces, face angles, weld length thresholds, and work surface classification, to achieve accurate identification of valid welds. Simultaneously, the automatic weld extraction and feature calculation module can automatically generate a unique identifier for each weld and organize and manage welds according to work surfaces. The weld geometric feature calculation module generates welding path points, arrival points, departure points, and touch-and-find points based on the weld geometric features, calculates key geometric features such as the start point, end point, direction vector, and length of the weld, and solves for the normal vectors and angle relationships between adjacent weld surfaces. Then, it automatically generates transition points in the weld path according to process requirements, making the welding process more stable. Through precise geometric model analysis, this module can reliably identify welding locations in complex structures, laying the foundation for subsequent path calculations; The welding key point and welding torch posture generation module supports the automatic generation of multiple key points required for the welding process, including the weld arrival point, departure point, touch-finding point, and welding path point. The touch-finding point is constructed based on the normal vector of adjacent surfaces, the direction of the angle bisector, and the process distance requirements, and includes paired touch points in the X, Y, and Z directions for positioning error calibration before welding. The arrival point and departure point are calculated using a trapezoidal symmetry strategy to make the welding torch movement more stable. At the same time, this module uses the rotation matrix and ZYX Euler angle method to calculate the welding torch posture, handle special cases such as gimbal lock, and ensure posture stability. The path planning and reachability verification module converts the poses of key points on the weld into the point format required for robot motion planning, and performs inverse kinematics solution, joint limit check, singular point detection, and collision detection on each of these points, as well as reachability detection, to determine whether the path is executable. This module outputs the planning results in the form of status codes, including reachability, no inverse solution, collision risk, singular point, axis limit, etc., to help users evaluate the feasibility of welding. The planning process also includes a strategy for loading and closing robx files to ensure proper management of system resources. The data output and status management module is used for unified management and structured output of weld feature data, welding key point poses, path planning results, and multi-state judgment information. It can export all weld and its key point information in JSON format, including weld geometry, key point coordinates, pose angles, and path data. This module provides a complete log recording and error feedback mechanism, supporting users to analyze and verify welding tasks. Through unified configuration management functions, users can adjust process-related settings such as weld extraction parameters, path calculation parameters, and contact point distances.

[0014] The specific operation steps of the automatic weld seam extraction and path planning system for welding robots are as follows: (1) The 3D model input and parsing module identifies weld features from the project file (STEP format); (2) The automatic weld extraction and feature calculation module performs geometric feature calculation on the identified welds; (3) The welding key point and welding gun posture generation module generates the key points required for the welding process and the corresponding welding gun posture; (4) The path planning and reachability verification module plans and verifies the feasibility of the welding path; (5) The data output and status management module manages and outputs the welding results in a unified manner.

[0015] In this invention, step (2) is specifically performed as follows: (2.1) The automatic weld extraction module reads the STEP 3D model of the workpiece and converts it into a geometric shape representation. It traverses the face, edge and volume structure in the model, constructs a face-edge adjacency relationship table, and forms a complete geometric topology. During the construction process, it synchronously records the normal information, surface type, area and geometric accuracy tolerance of each topological face, and records the geometric type, arc length, endpoint coordinates and the information of the part to which each edge belongs. At the same time, it marks whether the edge is shared by different parts to establish the index relationship between face-edge-part. (2.2) Based on the above geometric topology, candidate welds are screened step by step using a rule chain method: First, only edges located at the contact boundaries of different parts are retained as candidate welds; then, face number constraints are applied to retain only edges shared by two topological faces; if a weld intersects with other faces geometrically, the weld is truncated according to the intersection line to obtain non-interference weld segments; then, excessively short welds are filtered according to the minimum weld length threshold related to the welding process; finally, welds are grouped according to the working face orientation or part posture for subsequent posture planning and path sorting. (2.3) Calculate the precise start point, end point and center line direction vector of the selected weld and obtain the weld arc length; at the same time, extract the normal vector, surface type and their included angle relationship of the two adjacent surfaces of the weld to form a geometric and topological feature description of the weld, and cache the working surface identifier, adjacent part identifier and weld type information together as input for subsequent welding gun attitude calculation and path planning.

[0016] In this invention, step (3) is specifically performed as follows: (3.1) Uniform sampling is performed along the centerline of the weld according to the arc length, and the sampling interval is controlled by the welding process parameters; at the start and end points of the weld, the arrival point and departure point are generated outward along the weld direction to ensure the continuity of the welding torch movement during the arc initiation and arc termination stages. When the weld length is insufficient to meet the preset outward distance, the sampling interval is adaptively adjusted to ensure that at least the basic path configuration including the start point, midpoint and end point is generated; (3.2) A local coordinate system for the welding torch is constructed based on the weld direction vector and the normal vector of the adjacent surface. The weld normal is used as the Z-axis of the welding torch, and the weld direction is used as the Y-axis. The X-axis of the welding torch is obtained by cross product. The welding torch posture is optimized in combination with the orientation of the working surface to reduce lateral interference between the welding torch and the workpiece. In the presence of baffles or trihedral structures, the welding torch posture is tilted and adjusted according to the vector pointing from the touch base point to the weld to enable the welding torch to bypass the obstacle structure. The posture of the intermediate path point of the weld is calculated by interpolation to ensure the continuity and smoothness of the welding torch posture changes during the welding process. (3.3) To achieve pre-welding positioning calibration, contact base points are constructed on the angle bisectors of the normals of adjacent weld surfaces, and the positions of the base points are determined based on the contact distance parameters and plate thickness information. Then, pairs of contact finding points are generated along the X, Y, and Z directions to correct the clamping and positioning errors of the welding robot in the three axes. When there is a lack of effective adjacent surfaces in a certain direction, contact points are supplemented by other directions to ensure that calibration can be completed in all three axes. (3.4) The welding path points, arrival points, departure points and touch-finding points are uniformly encapsulated into a set of welding key points. The key points include spatial position coordinates and welding torch posture information. The posture is represented in a unified Euler angle format. The posture of the touch-finding point inherits the posture of the weld start point, so that the welding torch posture in the positioning stage is consistent with that in the formal welding stage. Weld identification and working surface grouping information are retained in the key point data.

[0017] In this invention, step (4) is specifically performed as follows: (4.1) The path planning module converts the set of welding key points into a sequence of target poses of the welding robot end effector, and performs inverse kinematics solution for each target pose to generate multiple sets of joint solutions; during the solution process, a small perturbation is applied to the position of the robot base to increase the diversity of feasible solutions, and the joint solution sequence with the minimum cost is selected in the solution space of each path point through the bundle search method. The cost function comprehensively considers the joint motion amplitude, the continuity of posture change and the reachability constraint. (4.2) Perform singularity detection and joint limit detection on the obtained joint solutions in sequence to determine whether the robot is close to a kinematic singular state or whether there are joint angles or base travels that exceed the allowable range in the corresponding pose; when the dynamic limit information is unavailable, the preset static limit range is used as a fallback constraint. (4.3) Further collision detection is performed on the joint solutions that have passed the above detection. The detection objects include the robot body, welding torch, workpiece, positioner and other environmental components. When necessary, the fine collision detection mode is activated to detect different component combinations one by one. Once interference is found, the corresponding solution is eliminated to avoid potential collision risks. (4.4) Based on the results of inverse kinematics solution, singularity detection, joint limit detection and collision detection, the reachability determination state of each key point in the welding path is generated, and the specific reasons for unreachability or risk are recorded.

[0018] In this invention, step (5) is specifically performed as follows: (5.1) The feasibility determination results of the welding path are divided into multiple state types, including at least: path reachable and collision-free state, inverse kinematics unsolvable state, joint or base stroke exceeding limit state, singular point risk state, and collision risk state. (5.2) During the path planning process, the status identifier of each path point is output in real time; when an unreachable state occurs, the planning can be terminated in advance and the corresponding reason is returned; after the entire weld path planning is completed, the overall status of the path is summarized to provide a basis for welding process adjustment and manual review. (5.3) The weld geometry information, welding key point pose information, path planning results and multi-state judgment results are uniformly stored as structured data and exported in standard data format for offline simulation, on-site execution and subsequent quality traceability and process optimization.

[0019] The automatic weld seam extraction and path planning system for welding robots provided by this invention significantly improves the reliability and engineering feasibility of automatic programming for welding robots by combining three-dimensional geometric topology analysis, welding torch posture calculation, and multi-constraint path planning and reachability verification strategies. It performs particularly well in handling complex assembly structures, spatially tortuous weld seams, and multi-interference scenarios. Compared with existing technologies, this invention has the following advantages: (1) High executability of welding path: By comprehensively constraining and verifying the geometric features of weld, welding gun posture, inverse kinematics solution, joint limit, singular point and collision risk, this system can effectively avoid unreachable or unsafe welding paths during the path generation stage, and significantly improve the success rate of welding path execution in actual robot systems.

[0020] (2) Strong adaptability to complex structures: Through geometric topology analysis based on three-dimensional models and multi-level rule chain screening mechanism, the system can accurately identify effective welds in complex assembly models and multi-part scenarios. In complex working conditions such as baffles and trihedral structures, the system ensures the accessibility of the welding torch through posture optimization and tilting processing, demonstrating good adaptability to complex welding scenarios.

[0021] (3) The system has strong scalability: The system adopts a modular design. Functional modules such as weld seam extraction, key point and attitude calculation, path planning and accessibility verification interact through standardized interfaces, which facilitates functional expansion and system integration according to different robot models, welding process parameters or application scenarios.

[0022] (4) High engineering efficiency and strong practicality: By introducing multiple inverse kinematics solution, bundle search strategy and early stopping mechanism in the path planning process, the overall calculation time is effectively controlled while ensuring the safety and reliability of the welding path. It is suitable for quickly generating welding paths in offline programming and engineering application environments, thereby improving the programming efficiency of welding robots. Attached Figure Description

[0023] Figure 1 This is a schematic diagram of the overall process of the automatic weld seam extraction and path planning system for welding robots of the present invention.

[0024] Figure 2 This is a schematic diagram of the overall architecture of the welding robot automatic weld seam extraction and path planning system of the present invention. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to specific embodiments and accompanying drawings. However, the scope of protection of this invention is not limited to the following embodiments. Example 1:

[0026] Example 1: 1. System Overall Process Description The overall process of the automatic weld seam extraction and path planning system for welding robots of the present invention is as follows: Figure 1 As shown, the main steps include STEP 3D model input, geometric topology construction, automatic weld seam extraction, weld seam feature calculation, welding key point generation, welding torch attitude calculation, inverse kinematics multi-solution solution, bundle search path optimization, multi-state reachability determination, welding path and data output, etc.

[0027] The system is an offline automated process that requires no manual teaching. All weld positions, welding postures, and robot motion paths are automatically generated by the system based on the 3D model and process parameters, and are output after being verified by multiple constraints.

[0028] 2. System Overall Architecture Description The overall system architecture is as follows Figure 2 As shown, the automatic weld seam extraction and path planning system for welding robots provided by this invention includes a 3D model input and analysis module, a geometric topology construction module, an automatic weld seam extraction and feature calculation module, a welding key point and welding torch posture generation module, a path planning and reachability verification module, and a data export and status management module. The output of the 3D model input and analysis module is connected to the input of the geometric topology construction module; the output of the geometric topology construction module is connected to the input of the automatic weld seam extraction module; the output of the automatic weld seam extraction module is connected to the input of the welding key point and welding torch posture generation module; the output of the welding key point and welding torch posture generation module is connected to the input of the path planning and reachability verification module; and the output of the path planning and reachability verification module is connected to the input of the data output and status management module.

[0029] The overall architecture design of the system follows the engineering process from geometric modeling to path execution. It starts with the input of the STEP 3D model and ends with the output of the welding path and its multi-state executability results. The modules interact with each other through standardized data structures to form a complete and closed-loop offline programming solution for welding robots.

[0030] 3. Weld geometry and topology feature extraction module The weld geometry and topology feature extraction module is the basic module of the system, responsible for automatically identifying welds from the 3D model and constructing their geometric and topological feature representations.

[0031] 3.1 Formal Representation and Topology Construction of 3D Models In this embodiment, the system first reads the STEP format 3D model file and converts it into a unified geometric representation. The model consists of several parts, each composed of topological elements such as faces, edges, and volumes. The system constructs a face-edge adjacency table and a face-volume attribution table by traversing the topological elements in the model, which are used to describe the spatial connection relationships between the parts.

[0032] During topology construction, the system records the normal information, surface type, area, and geometric accuracy tolerance for each topological surface; and records the geometric type, arc length, endpoint coordinates, and the identifier of the part to which each edge belongs, and marks whether the edge is a shared edge across parts. The above topology information constitutes the basic data for subsequent weld identification and attitude calculation.

[0033] 3.2 Automatic identification and rule chain screening of welds Based on the above topological representation, the system first identifies shared edges located at the contact boundaries of different parts as candidate welds. Subsequently, the system uses a multi-level rule chain to filter the candidate welds.

[0034] Specifically, the system first applies a face number constraint, retaining only welds shared by two topological faces; when a candidate weld intersects geometrically with a third face, the system truncates the weld according to the intersection line to obtain non-interference weld segments; subsequently, the system filters out excessively short welds based on threshold parameters related to plate thickness and minimum weld length; finally, the system groups welds according to the working face orientation or part posture, providing a basis for subsequent welding torch posture optimization and path sorting.

[0035] Through the above rule chain processing, the system can stably extract effective welds in complex assembly models and ensure that the weld results have clear geometric and topological semantics.

[0036] 3.3 Calculation of weld geometry For the selected welds, the system calculates their precise start point, end point, centerline direction vector, and arc length, and extracts the normal vectors of the two adjacent surfaces of the weld and their included angle. Simultaneously, the system records information such as the working surface identifier of the weld, the identifiers of adjacent parts, the weld type, and whether a baffle structure exists, forming a weld feature package, which serves as direct input for subsequent welding torch attitude calculation and path planning.

[0037] 4. Welding Key Points and Welding Torch Posture Calculation Module The welding key point and welding torch posture calculation module is responsible for converting the geometric features of the weld into a spatial pose sequence that can be executed by the welding robot.

[0038] 4.1 Generation of Welding Path Points and Arrival / Departure Points The system samples uniformly along the weld centerline according to the arc length, with the sampling interval controlled by welding process parameters. At the start and end points of the weld, the system extends outward along the weld direction to generate arrival and departure points, ensuring the continuity of the welding torch's movement during the arc initiation and termination phases. When the weld length is insufficient to meet the preset parameters, the system automatically adjusts the sampling strategy to ensure the generation of the minimum executable path.

[0039] 4.2 Construction and Continuity Guarantee of Welding Torch Posture The system constructs a local coordinate system for the welding torch based on the weld direction vector and the normals of adjacent surfaces. The weld normal is used as the Z-axis of the welding torch, the weld direction is used as the Y-axis, and the X-axis is obtained through cross product. The welding torch posture is optimized by combining the orientation of the working surface to reduce the risk of lateral interference between the welding torch and the workpiece.

[0040] When a baffle or trihedral structure is detected, the system adjusts the welding torch posture based on the directional relationship between the contact point and the weld, causing the torch to tilt to avoid the obstacle. The posture of the intermediate path point of the weld is calculated through interpolation, thus ensuring the continuity and smoothness of the welding torch posture changes during the welding process.

[0041] 4.3 Generation of touch-based search points To achieve pre-welding positioning calibration, the system constructs contact points on the angle bisectors of the normals of adjacent weld surfaces. Based on the contact distance and plate thickness parameters, it generates paired contact finding points in the X, Y, and Z directions to correct clamping and positioning errors. When a certain direction lacks effective adjacent surfaces, the system uses a compensation strategy to ensure calibration can be completed in all three axes.

[0042] 5. Welding path planning and reachability verification module The welding path planning and accessibility verification module is responsible for converting welding key points into robot motion paths and verifying their executability.

[0043] 5.1 Inverse Kinematics Solution and Path Combination The system converts welding key points into a sequence of target poses for the robot's end effector and performs inverse kinematics solving for each pose, generating multiple sets of joint solutions. During the solution process, a small perturbation is applied to the robot's base position to increase the diversity of feasible solutions. Subsequently, the system employs a bundle search strategy to combine the entire path solution chain in the solution space of each path point.

[0044] 5.2 Multi-constraint verification and state determination The system sequentially performs joint limit detection, singularity detection, and collision detection on the path solution, checking the robot body, welding torch, workpiece, and related equipment. If an unreachable or risky situation is detected, the system immediately marks the corresponding state and removes the solution. Finally, the system generates a multi-state reachability determination result for the entire weld path.

[0045] 6. Results Output and Data Management Module The system stores weld features, key welding point poses, path planning results, and their reachability status as structured data, and supports exporting in JSON format for offline simulation, on-site execution, and quality traceability.

Claims

1. A robotic weld seam auto-extraction and path planning system for a welding robot, characterized by include: The system comprises a 3D model input and parsing module, a geometric topology construction module, an automatic weld seam extraction and feature calculation module, a welding key point and welding torch posture generation module, a path planning and reachability verification module, and a data output and status management module. The output of the 3D model input and parsing module is connected to the input of the geometric topology construction module; the output of the geometric topology construction module is connected to the input of the automatic weld seam extraction and feature calculation module; the output of the automatic weld seam extraction and feature calculation module is connected to the input of the welding key point and welding torch posture generation module; the output of the welding key point and welding torch posture generation module is connected to the input of the path planning and reachability verification module; and the output of the path planning and reachability verification module is connected to the input of the data output and status management module. The 3D model input and parsing module is used to read the 3D model data of the workpiece and construct the geometric topology structure, which is then transmitted to the geometric topology construction module. The automatic weld extraction and feature calculation module includes an automatic weld extraction module and a weld geometric feature calculation module. The automatic weld extraction module can read and parse 3D model data (STEP format), automatically identify the contact boundaries between workpieces, and extract geometric intersections that may constitute welds. It filters and selects welds through multi-level rules, including constraints on the number of faces, face angles, weld length thresholds, and work surface classification, to achieve accurate identification of valid welds. Simultaneously, the automatic weld extraction and feature calculation module can automatically generate a unique identifier for each weld and organize and manage welds according to work surfaces. The weld geometric feature calculation module generates welding path points, arrival points, departure points, and touch-and-find points based on the weld geometric features, calculates key geometric features such as the start point, end point, direction vector, and length of the weld, and solves for the normal vectors and angle relationships between adjacent weld surfaces. Then, it automatically generates transition points in the weld path according to process requirements, making the welding process more stable. Through precise geometric model analysis, this module can reliably identify welding locations in complex structures, laying the foundation for subsequent path calculations; The welding key point and welding torch posture generation module supports the automatic generation of multiple key points required for the welding process, including the weld arrival point, departure point, touch-finding point, and welding path point. The touch-finding point is constructed based on the normal vector of adjacent surfaces, the direction of the angle bisector, and the process distance requirements, and includes paired touch points in the X, Y, and Z directions for positioning error calibration before welding. The arrival point and departure point are calculated using a trapezoidal symmetry strategy to make the welding torch movement more stable. At the same time, this module uses the rotation matrix and ZYX Euler angle method to calculate the welding torch posture, handle special cases such as gimbal lock, and ensure posture stability. The path planning and reachability verification module converts the poses of key points on the weld into the point format required for robot motion planning, and performs inverse kinematics solution, joint limit check, singular point detection, and collision detection on each of these points, as well as reachability detection, to determine whether the path is executable. This module outputs the planning results in the form of status codes, including reachability, no inverse solution, collision risk, singular point, axis limit, etc., to help users evaluate the feasibility of welding. The planning process also includes a strategy for loading and closing robx files to ensure proper management of system resources. The data output and status management module is used for unified management and structured output of weld feature data, welding key point poses, path planning results, and multi-state judgment information; it can export all weld and its key point information in JSON format, including weld geometry, key point coordinates, pose angles, and path data; this module provides a complete log recording and error feedback mechanism, supporting users to analyze and verify welding tasks; through unified configuration management functions, users can adjust weld extraction parameters, path calculation parameters, and touch point distance process-related settings; The specific operation steps of the automatic weld seam extraction and path planning system for welding robots are as follows: (1) The 3D model input and parsing module identifies weld features from the project file (STEP format); (2) The automatic weld extraction and feature calculation module performs geometric feature calculation on the identified welds; (3) The welding key point and welding gun posture generation module generates the key points required for the welding process and the corresponding welding gun posture; (4) The path planning and reachability verification module plans and verifies the feasibility of the welding path; (5) The data output and status management module manages and outputs the welding results in a unified manner.

2. A welding robot seam auto-extraction and path planning system according to claim 1, characterized in that Step (2) is performed as follows: (2.1) The automatic weld extraction module reads the STEP 3D model of the workpiece and converts it into a geometric shape representation. It traverses the face, edge and volume structure in the model, constructs a face-edge adjacency relationship table, and forms a complete geometric topology. During the construction process, it synchronously records the normal information, surface type, area and geometric accuracy tolerance of each topological face, and records the geometric type, arc length, endpoint coordinates and the information of the part to which each edge belongs. At the same time, it marks whether the edge is shared by different parts to establish the index relationship between face-edge-part. (2.2) Based on the above geometric topology, candidate welds are screened step by step using a rule chain method: First, only edges located at the contact boundaries of different parts are retained as candidate welds; then, face number constraints are applied to retain only edges shared by two topological faces; if a weld intersects with other faces geometrically, the weld is truncated according to the intersection line to obtain non-interference weld segments; then, excessively short welds are filtered according to the minimum weld length threshold related to the welding process; finally, welds are grouped according to the working face orientation or part posture for subsequent posture planning and path sorting. (2.3) Calculate the precise start point, end point and center line direction vector of the selected weld and obtain the weld arc length; at the same time, extract the normal vector, surface type and their included angle relationship of the two adjacent surfaces of the weld to form a geometric and topological feature description of the weld, and cache the working surface identifier, adjacent part identifier and weld type information together as input for subsequent welding gun attitude calculation and path planning.

3. A system for automatic extraction of weld seams and path planning for a welding robot according to claim 1, characterized in that Step (3) is performed as follows: (3.1) Uniform sampling is performed along the centerline of the weld according to the arc length, and the sampling interval is controlled by the welding process parameters; at the start and end points of the weld, the arrival point and departure point are generated outward along the weld direction to ensure the continuity of the welding torch movement during the arc initiation and arc termination stages. When the weld length is insufficient to meet the preset outward distance, the sampling interval is adaptively adjusted to ensure that at least the basic path configuration including the start point, midpoint and end point is generated; (3.2) A local coordinate system for the welding torch is constructed based on the weld direction vector and the normal vector of the adjacent surface. The weld normal is used as the Z-axis of the welding torch, and the weld direction is used as the Y-axis. The X-axis of the welding torch is obtained by cross product. The welding torch posture is optimized in combination with the orientation of the working surface to reduce lateral interference between the welding torch and the workpiece. In the presence of baffles or trihedral structures, the welding torch posture is tilted and adjusted according to the vector pointing from the touch base point to the weld to enable the welding torch to bypass the obstacle structure. The posture of the intermediate path point of the weld is calculated by interpolation to ensure the continuity and smoothness of the welding torch posture changes during the welding process. (3.3) To achieve pre-welding positioning calibration, contact base points are constructed on the angle bisectors of the normals of adjacent weld surfaces, and the positions of the base points are determined based on the contact distance parameters and plate thickness information. Then, pairs of contact finding points are generated along the X, Y, and Z directions to correct the clamping and positioning errors of the welding robot in the three axes. When there is a lack of effective adjacent surfaces in a certain direction, contact points are supplemented by other directions to ensure that calibration can be completed in all three axes. (3.4) The welding path points, arrival points, departure points and touch-finding points are uniformly encapsulated into a set of welding key points. The key points include spatial position coordinates and welding gun posture information. The posture is represented by a unified Euler angle format. The orientation of the touch point inherits the orientation of the weld start point, ensuring that the orientation of the welding torch is consistent with that of the formal welding stage, and retains the weld identification and working face grouping information in the key point data.

4. The welding robot seam auto-extraction and path planning system of claim 1, wherein Step (4) is operated as follows: (4.1) The path planning module converts the set of welding key points into a sequence of target poses of the welding robot end effector, and performs inverse kinematics solution for each target pose to generate multiple sets of joint solutions; During the solution process, a small perturbation is applied to the position of the robot base to increase the diversity of feasible solutions. The joint solution sequence with the minimum cost is selected from the solution space of each path point by the bundle search method. The cost function comprehensively considers the joint motion amplitude, the continuity of posture change and the reachability constraint. (4.2) Perform singularity detection and joint limit detection on the obtained joint solutions in sequence to determine whether the robot is close to a kinematic singular state or whether there are joint angles or base travels that exceed the allowable range in the corresponding pose; when the dynamic limit information is unavailable, the preset static limit range is used as a fallback constraint. (4.3) Further collision detection is performed on the joint solutions that have passed the above detection. The detection objects include the robot body, welding torch, workpiece, positioner and other environmental components. When necessary, the fine collision detection mode is activated to detect different component combinations one by one. Once interference is found, the corresponding solution is eliminated to avoid potential collision risks. (4.4) Based on the results of inverse kinematics solution, singularity detection, joint limit detection and collision detection, the reachability determination state of each key point in the welding path is generated, and the specific reasons for unreachability or risk are recorded.

5. The welding robot seam auto-extraction and path planning system of claim 1, wherein Step (5) is performed as follows: (5.1) The feasibility determination results of the welding path are divided into multiple state types, including at least: path reachable and collision-free state, inverse kinematics unsolvable state, joint or base stroke exceeding limit state, singular point risk state, and collision risk state. (5.2) During the path planning process, the status identifier of each path point is output in real time; when an unreachable state occurs, the planning can be terminated in advance and the corresponding reason is returned; after the entire weld path planning is completed, the overall status of the path is summarized to provide a basis for welding process adjustment and manual review. (5.3) The weld geometry information, welding key point pose information, path planning results and multi-state judgment results are uniformly stored as structured data and exported in standard data format for offline simulation, on-site execution and subsequent quality traceability and process optimization.

Images