Research evolution on intelligentized technologies for robotic welding at SJTU

Abstract

This paper addresses the new evolution of the study on intelligentized technologies for robotic welding based on our works in SJTU from 2006, which contains multi-information acquirement of arc welding process, such as extraction of weld pool image, voltage, current, and sound features; multi-information fusion algorithms for prediction of weld penetration by neural network and DS evidence theory; intelligentized modeling of welding process and robot system, such as welding dynamic knowledge models by SVM and RS methods, the MLD modeling of welding robot system; and intelligent control strategies for welding pool and penetration process, such as the adaptive inverse control scheme based on SVM-FRDS, control scheme based on the knowledge model by the RS theory, and the Model-Free adaptive control of pulsed GTAW; the 3-D seam tracking during robotic welding by combining arc sensing and visual sensing; and a new welding robot system for autonomously avoiding obstacles. © 2011 Springer-Verlag Berlin Heidelberg.