Audio-visual modelling and control of GMA welding

点击翻译按钮获取中文摘要

This research program is in the area of experimental gas metal arc (GMA) welding process identification and control. The objective is to investigate the relationships between visual and acoustic signals that expert human welders can sense when welding, and the control actions they employ in response to these.The motivation for pursuing this research is that the current state-of-the-art in robotic GMA welding systems still rely on what is essentially open-loop control: parameter settings are preset to values that should result in a good quality weld for a given joint geometry under constant conditions. However, expert human welders continually make adjustments to the welding torch position and speed, as well as to the voltage and wire feed-rate, in response to what they see and hear while they are welding. This human feedback control system is learned through experience, and is very effective at producing high quality welds. However, arc welding is a tiring and hazardous job involving exposure to noxious fumes, heat and intense ultra violet radiation, and it would therefore be of great benefit to develop fully automated / robotic welding systems capable of providing comparable feedback control that human welders excel at.The expected significance of this research is a better understanding of how professional welders react to the acoustic and visual cues they receive while welding, and how this can be applied to the automatic control of the GMA welding process. The benefits to society include the potential for reduced human exposure to the hazardous arc welding environment, while retaining the need for skilled welding professionals to oversee and program the robotic welding operations.