Figure 13 depicts the welding speeds in nine experiments versus the torch orientation angle θ (0 deg corresponds to 12 o’clock) in relation to the gravitational direction. It is noticed
that the average welding speed corresponding to a different degree is fluctuating for all the three welding currents.However, no noticeable correlation is found between the fluctuations
of the welding speed to the position of the torch. It will be further proved in the welding experiments that a constant welding speed is sufficient to produce satisfactory welds in the top
part of the pipe, i.e., –25 deg ≤ θ ≤ 25 deg (approximately 11 to 1 o’clock).For full position pipe welding, however,different speeds should be applied for different positions and will be
studied in the future using an advanced system whose projector and Leap sensor must move with the torch movement.The data points to be used to correlate the welding current and welding
speed are plotted in Fig. 14. A linear model is fitted using the least squares algorithm (Ref. 43):s = 0.0462l – 1.3 (2) or equivalently,l = 10.13s + 39.77 (3) where s is the welding speed in mm/s and l is the welding current in A.The above correlation between the welding current and welding speed can be used for human-machine cooperative teleoperated pipe welding applications
where an unskilled human welder operates the torch (determining the actual welding speed) while the welding machine could compensate his/her inaccurate movement (inaccurate welding speed) by adjusting the welding current.It could also be used to provide a parameter interval for simultaneously controlling the frontside weld pool surface characteristic parameters and
backside joint penetration in automated welding. This learned correlation will then be applied in welding experiments to demonstrate the effectiveness of the learned welding speed
from the human welder.