This research utilized a new human machine welding paradigm, virtualized welding, to learn a human welder’s speed adjustment under different welding currents. Learning experiments
through teleoperation were conducted by a human welder to generate satisfactory welds under different welding currents. A correlation between the welding current and welding
speed was proposed for GTAW pipe welding with specified welding conditions.Consistent penetration and satisfactory welds were generated in automated welding experiments. It is
also observed that for the top part of the pipe welding, instead of manipulating
a full set of welding parameters (including torch orientation and arc length, etc.), adjusting the welding speed was sufficient to generate satisfactory welds.Future work of the authors is to
further study skilled human welders’intelligence in dynamic welding conditions and adaptively control the welding process. Full position pipe welding will also be studied where welding
speed and torch orientation could be simultaneously controlled to generate satisfactory welds.