Enhanced Optical Tracking of Weld Beads in Autonomous Inspection of Separator Vessels

Inspection robots have been developed to support the maintenance of separator vessels. One challenge for such robots is to identify and navigate along the weld bead. This paper proposes a new solution to the weld bead recognition problem, including its tracking, which aims to automate weld bead inspection. A reliable weld bead detection method was developed through five steps that process the line profile sensor data. The weld bead was identified based on the estimation of its center. A non-linear controller paired with a module that compensates for uncertainties generated by gravity was designed to ensure the tracking of weld bead. Such a controller can reduce the impact of the variation in the robot's orientation on the weld bead tracking. An algorithm redirects the robot at the end of the weld bead. The experiments were conducted in a laboratory-scale setup, allowing movement along horizontal and diagonal surfaces. The results show that the robot tracked the weld line in all simulated conditions, with a maximum error of 2.86 mm.