Abstract
In industrial assembly lines like for automobile manufacturing, seam sealing is an important process where sealant is painted on seams locate at the joints by thin plate parts for watertight or a rust prevention. Manual tasks in sealing is so exhausted even for skillful human operators because the curve to be sealed is too complex to trace in high-speed. Therefore, most of the sealing processes have already been automated by teaching-playback robots. However, it requires much redundant amount of sealant to cope with the positioning error or shape deformation of the workpiece. This is one of the problems degrading the manufacturing cost. In this paper, we propose a highly reliable algorithm for seam position computation from sensed profile range data adjacent to the seam. The data is assumed to be sensed by a high-speed scanning laser range sensor that is equipped with a robot arm with the sealing gun. Our algorithm employs statistical jump detection from the range data because the jump in the data is a robust feature and the computation for the detection is fairly small. Based on the computed seam position, the sealing gun can be controlled as tracking the seam at such high speed as 300 [mm/sec] . It is proved experimentally that the sealing system with the developed algorithm is very effective, especially in the point of reduction of the redundant sealant.
