Weld Joint Reconstruction and Classification Algorithm for Trajectory Generation in Robotic Welding

Abstract

Automation of welding with robotic arms has become an inevitable trend in modern manufacturing technologies. This process can be automated by using a "click and go" in which the robot will weld a line where the spot is described or by using an in-line tracking algorithm in which the robot will choose the spot where to weld the line in each layer. This paper presents a simple methodology for the reconstruction of the weld joint and the classification of the joint geometry to serve as a first step to the automatic determination of the robot trajectory. The weld joint has been reconstructed using a laser profilometer placed as a tool on the robot. Spurious data has been removed by signal processing. The joint has been reconstructed three-dimensionally. The classification of the joint profiles was generated using an algorithm based on signal processing and artificial intelligence. This algorithm has been tested for the classification of V-joints (bevel-bevel) and single bevel joints.