In-Line Monitoring of Laser Welding Using a Smart Vision System

This paper presents a vision system for the in-line monitoring of laser welding. The system is based on a coaxial optical setup purposely chosen to guarantee robust detection of the joints and optimal acquisition of the melt pool region. Two procedures have been developed: The former focuses on keeping the laser head locked to the joint during the welding; the latter monitors the appearance of the keyhole region. The system feedbacks the joint position to the robot used to move the welding laser and monitors the penetration state of the laser. The goal is to achieve a continuous adaptation of the laser parameters (power., speed and focusing) to guarantee the weld quality. The developed algorithms have been designed to optimize the system performance in terms of the elaboration time and of accuracy and robustness of the detection. The overall architecture follows the Industrial Internet of Things approach, where vision is embedded, edge-based analysis is carried out, actuators are directly driven by the vision system, a latency-free transmission architecture allows interconnection as well as the possibility to remotely control multiple delocalized units.