In robot applications for automatic welding of ship hull assembly process, sensing of the exact location and orientation of a workpiece and its detailed surface geometry are needed for the robot guidance. Among various sensors of the contact and non-contact type, the laser vision sensor which is based on optical triangulation of the scanned point beam and the through-arc sensor using the electric signals obtained from the rotating welding arc are effective to track the fillet joint of ship hull assembly process.
In the hull assembly process, the skin plates are welded together to become a block which is approximately 20㎡, and longitudinal and transverse webs are fitted and welded, so stationary-base welding robots cannot be used. Mobile manipulators, which have unlimited size of operational space and a high degree of kinematic redundancy, are suitable for that process.
The active vision sensor based on optical triangulation is classified into two types according to the beam characteristics: projected sheet of beam or scanned point beam. The vision sensor with scanned point beam is rarely influence by arc noise and the preprocessing time of the image is relatively short.
In this paper, a conceptual mobile platform-manipulator welding system to which a laser vision sensor with scanned point beam is attached was designed for automatic welding of the 3-D seam in the hull block. An image processing algorithm using neural network was developed for proper feature recognition. Experimental results from using the laser vision sensor which is based on optical triangulation of the scanned point beam are also presented.
Through-arc sensing using the electric signals obtained from the welding arc is cost effective way to track the welding joint. During GMA welding with a constant voltage power source, the welding current and the voltage generally vary with the CTWD(contact tip-to-work distance). The torch weaving method is widely used to intentionally stimulate the difference of the CTWD. But the conventional torch weaving methods have a relatively poor resolution and limit the oscillation frequency due to the mechanical restraints. The rotating arc sensor can improve the sensor characteristics such as the sensitivity and the responsibility due to the high-speed rotation frequency. During the high-speed rotating arc welding, the self-regulation of welding arc is not fully performed because the period of rotation is shorter than the time constant of the self-regulation process. The rapid change of CTWD generates the variation of the arc length rather than that of the electrode extension length. Therefore, the rotating arc sensor operates in a dynamic state enhancing its sensitivity. In addition to the joint tracking, the through-arc sensing can be applied to detecting an end-point of the weld joint because the shape of the joint is reflected on the electrical signals such as welding current and voltage, which are related to the CTWD.
In this paper, the following topics are investigated: (a) A mathematical modeling which investigates the GMA welding system. (b) Geometrical sensing model of the rotating arc sensor in fillet welding which has the open and closed end. (c) Current simulations and experiments in open and closed fillet welding.
