关键词: 焊缝, 机器人磨削, 点云, 特征提取

Abstract: Service life and dynamic performance of structural components were affected by the accuracy and surface consistency of weld grinding. Currently, weld grinding mainly relied on manual or trajectory teaching methods, which had problems such as poor uniformity, low efficiency and high cost. So, a point cloud-based automated grinding system for robotic weld seams was proposed. A quadratic streamlining method for weld seam point cloud was proposed to obtain the global information of the weld seam surfaces accurately, and the weld seam width, height, and centerline features were extracted based on the distances between sampling points and neighborhood center of gravity points. The errors of height and width of the extracted weld seams are as 0.09 mm and 0.2 mm respectively, and the minimum residual height of the weld seams after grinding is as 0.17 mm, and the vlaue of surface roughness Ra is up to 0.498 μm.

Key words: weld seam, robotic grinding, point cloud, feature extraction