A robotic passive vision system for texture analysis in weld beads

Robots have been increasingly used in applications involving welding of large metal structures, such as the naval industry, ensuring higher efficiency and repeatability at lower costs. However, inadequate communication between the robotics and the welding system can lead to internal and surface defects in the final product. Problems that occur during welding can be detected with the help of visual inspection. In the present work a passive monocular camera was used to quantify the texture found in weld beads as part of a fully-computerised vision system. The textures identified were associated with the presence of welding discontinuities. An algorithm based on Principal Component Analysis was developed to analyse weld beads, where part of the beads was produced using conditions that purposefully resulted in welding discontinuities, identifying the most important features that characterized one group with healthy beads and another group containing discontinuities. After this stage, a machine learning method was used in new weld beads, in order to classify them as healthy or defective. The accuracy of the proposed method for texture identification was 96.4%.