Sensing of Arc length and wire extension using neural network in robotic welding

Abstract

It is important to design intelligent welding robots to obtain a good quality of the welding. It is required to detect bead height and deviation from center of gap and to control arc length and back bead by adjusting welding conditions such as welding speed, power source voltage and wire feed rate. Authors propose arc sensor using neural networks to detect arc length and wire extension. By using them and by means of geometric method, the bead height is detected. Moreover, authors propose switch back welding method to get stable back bead. That is, welding torch is not only woven in the groove, but also moved backward and forward.