Surface defect detection of aluminum alloy welds with 3D depth image and 2D gray image

  • PDF / 2,872,433 Bytes
  • 12 Pages / 595.276 x 790.866 pts Page_size
  • 19 Downloads / 273 Views

DOWNLOAD

REPORT


ORIGINAL ARTICLE

Surface defect detection of aluminum alloy welds with 3D depth image and 2D gray image Zhihong Yan 1,2 & Bowei Shi 1,3 & Luping Sun 1,2 & Jun Xiao 1,2 Received: 28 January 2020 / Accepted: 5 August 2020 # Springer-Verlag London Ltd., part of Springer Nature 2020

Abstract Towards the trend and demand of automatic welding inspection in industry, a composite vision system enabling simultaneous 3D-depth and 2D-gray imaging of the bead surface is constructed to detect typical surface defects of aluminum alloy weld beads. In this vision system, the structured laser light is responsible for obtaining 3D-depth image of the bead surface; meanwhile, the multi-angle illuminations are used to capture gray images. Then, four methods are proposed to extract the weld bead boundaries according to its different characters shown in the 3D depth images and 2D gray images. In the 3D depth image, the extraction algorithms of defects such as collapse, undercut, burning-through, excessive reinforcement, surface porosity, spatter, and poor forming are studied. In multi-angle gray images, the extraction algorithm of defects such as cracks and surface blackening is also proposed and studied. Keywords Aluminum alloy weld . Surface defect extraction . Composite vision . Multi-angle illumination . Depth image and gray image

1 Introduction The detection of welding shape and welding defects is the main task of welding quality control, and there are many related researches, but most of them focus on the nondestructive testing (NDT) of the weld defects, especially the NDT of internal defects [1–3]. In recent years, with the wide application of automatic welding, there is an urgent demand for the automatic detection of surface shape and defects of weld beads. For detecting of the bead surface shape and defects, there are two common methods: 3D morphology detection and 2D image detection. Among the 3D morphology detection methods, the most commonly used one is structured laser * Zhihong Yan [email protected] 1

College of Mechanical Engineering and Applied Electronics Technology, Beijing University of Technology, Beijing 100124, China

2

Engineering Research Center of Advanced Manufacturing Technology for Automotive Components, Ministry of Education, Beijing University of Technology, Beijing 100124, China

3

Beijing Aerospace Institute of Microsystems, Beijing 100094, China

scanning for its high reliability and real-time response [4]. Chu et al. [5] defined weld bead geometry such as reinforcement and width and welding defects such as undercut and plate displacement through key points of weld cross-section surface profile, and then obtained these key points through processing of laser stripe image, and obtained the parameters above. Li et al. [6, 7] have done similar works; the difference is that they divide the welding bead into two types: root-pass and cap. In root-pass welding bead, the groove width, filling depth, and misalignment parameters are added. Furthermore, Han et al. [8] added an automatic recognition algorithm o