Gas tungsten arc welding (GTAW)(Ref. 1) is the primary process used for precise joining of metals. The GTAW process is illustrated in Fig. 1. An arc is established between the nonconsumable tungsten electrode and the base metal. The base metal is melted by the arc forming a liquid weld pool that joins the two pieces of base metal together after solidification. Manual
GTAW is commonly used in industry,especially for applications where feedback from the process may help to maintain weld quality and overcome possible process variations. In this process, skilled welders can often make adjustments based on their observation of the liquid weld pool
surface.Those real-time adjustments often lead to desired weld bead geometry characterized by backside bead width wb and positive reinforcement hb—Fig. 2. Welding robots are preferred in
many applications since they outperform human welders whose performance degrades because of their physical limitations (inconsistent concentration,fatigue, stress, and long-term
health issues). Unfortunately, current industrial welding robots are basically articulated arms with a preprogrammed set of movements, and they lack the intelligence skilled human welders possess. They require precision prepared workpieces with little variation in geometry and material properties. Therefore, their applications are mostly limited to assembly lines for mass-produced products,such as automobiles, where workpiece preparation is controllable at reasonable costs.However, as outlined in the national robotic report (Ref. 2), the trend in
manufacturing is to produce customized products in small batches where ideal automated production lines are not cost effective. As such,welding robots that possess intelligence
comparable to skilled welders but with fewer physical restrictions and that can work in harsh environments will be one of the keys to maintaining a competitive manufacturing industry despite relatively high labor costs/wages. The resultant intelligent welding robots may also help resolve the skilled welder shortage issue the manufacturing industry is currently facing.
In this research, a novel human-machine cooperative welding paradigm,virtualized welding (Ref. 3), is utilized to transfer human intelligence to welding robots. In this framework,
a welding robot working in the actual welding environment was augmented with sensors to observe the workpiece and reconstruct the 3D weld pool surface.The obtained data from the sensors
as feedback from an actual welding process is viewed by a human welder in a virtualized welding environment,and the welder adjusts the welding parameters accordingly. The data and adjustments would also be recorded/analyzed to model how the welder responded to the 3D weld pool
surface, which is believed to be the major source of feedback information a welder may acquire during the welding process, to enable transformation of human intelligence to the welding
robot to form autonomous intelligent welding robots. This research serves as the first study in modeling and analyzing human adjustment using the proposed virtualized welding platform.Major welding parameters in manual GTAW include welding current,welding speed, torch orientation, and
arc length. In a particular automated control system, however, only a few selected parameters should be adjusted to compensate against the effects from possible variations in the process. Among all the major welding parameters, an increase in the welding current and a decrease in the welding speed will significantly increase the heat input into the welding process,thus considerably influencing the weld pool surface geometry.In the authors’ previousstudies (Refs.
4–6), welding current was utilized to control the welding process. However,in many pipe welding applications,the pipe is often fixed and cannot be rotated during welding (e.g., 5G fixed position,that is, the axis of the pipes is horizontal;the pipe stays stationary during welding; and the welding torch will be moving along the weld joint (Ref. 7).Normally, welders choose a predefined welding current and move the torch along the pipe since the movement of the torch can be conveniently adjusted by a human welder to overcome the effects from variations. In this study, a welder’s movement along the welding direction was studied. Although other welding parameters (such as torch orientation and arc length) can certainly have an impact on the welding process, for the top part of the pipe, controlling the welding speed, as confirmed by experiments,is sufficient to generate satisfactory welds.The learned correlation between
the welding current and welder’s corresponding movement (i.e., the welding speed) can be used for humanmachine cooperative controlled pipe welding applications where an unskilled human welder operates a virtualized welding torch determining the welding speed while the welding machine could compensate for his/her incorrect movement by adjusting the welding current. For automated welding machines that need to simultaneously control the frontside weld pool characteristic parameters and backside weld penetration by adjusting welding current and speed, the proposed correlation could also provide an interval/constraint for welding process
input parameters, which will then be utilized to calculate the optimized welding current and speed.The remainder of the paper is organized as follows: In the next section,related work is detailed. In the third section, an overview of the virtual welding system is provided. In the
fourth section, experimentation is detailed and data from nine teleoperated welding experiments are presented/analyzed. A linear correlation was found between welding current and speed. Automated welding experiments were conducted under different welding currents, in which the proposed correlation was utilized to calculate the welding speed needed for each welding current. Experimental results are presented in the fifth section, followed by conclusions.