焊接
弹道
计算机科学
机器人
机器人焊接
计算机视觉
运动规划
切线
背景(考古学)
校准
人工智能
点云
算法
粒子群优化
工作区
点(几何)
光学(聚焦)
特征(语言学)
电弧焊
几何学
工程类
迭代最近点
数学
机械工程
熔池
移动机器人
迭代法
钥匙(锁)
作者
Hao Wang,Xiujun Wang,Chunguo Liu,C Chen,Xiaohui Zhao
标识
DOI:10.1016/j.rcim.2026.103315
摘要
• Online synchronous planning of welding paths and torch postures for structures with unknown spatial geometries is realized. • RMSEs for welding path and torch posture planning are ≤ 0.1334 mm and ≤ 0.8064°, respectively. • The trajectory planning scheme demonstrates exceptional adaptability to complex spatial geometries, multi-layer multi-pass welding, and diverse joint configurations. To address the limitations of existing welding trajectory planning schemes, which predominantly focus on welding paths while neglecting collaboration of torch postures and overly rely on prior models or idealized assumptions, an innovative solution for online synchronous planning of welding paths and torch postures for structures with unknown spatial geometries was developed. The GMAW system assisted by line laser vision was independently designed and implemented. The camera calibration leveraging the circular tangential gradient orthogonal method (CTGOM) and adaptive particle swarm optimization (APSO) was formulated. Optical plane calibration and hand-eye calibration were then sequentially performed, achieving an average absolute error of 0.1111 mm in the robot’s perception of spatial targets. The cascaded 3D inverse modeling framework for spatial joints was established, successively employing YOLOv11, context information-based iterative centerline extraction algorithm (CIICEA), Pauta criterion and K-means-based quasi-uniform cubic B-spline fitting algorithm (K-QCBFA). The approach for online planning of robot welding trajectories based on point cloud geometric differentiation was proposed. By constructing local differential geometries from groove cross-sectional planes and base metal tangent planes, the full six-degree-of-freedom welding trajectories ( X B , Y B , Z B , R X , R Y , R Z ) for unknown spatial geometries were planned synchronously online. The root mean square error (RMSE) for online planning of welding paths in butt and fillet joints was 0.0829 mm and 0.1334 mm, respectively, while that for torch postures was 0.7782° and 0.8064°, respectively. The proposed welding trajectory planning scheme demonstrates exceptional adaptability to complex spatial geometries, multi-layer multi-pass welding, and diverse joint configurations.
科研通智能强力驱动
Strongly Powered by AbleSci AI