Remote laser welding systems can usually focus a laser beam to diameters of 0.1 mm. Therefore, high quality clamping and precise teaching is needed in order to achieve appropriate process tolerance for a sound weld. This brings reservation in the field of small series and user-customized manufacturing, where product individualization requires flexible and adaptive systems as workpiece geometry is not exact due to manufacturing tolerances and thermal deformations during welding. The preparation for welding is therefore often time-consuming. To solve this, we have developed an innovative system, which enables in-line adaptive 3D seam tracking. The system consists of an industrial robot (Yaskawa MC2000), a scanning head (HighYag RLSK; working area 200 mm × 300 mm × 200 mm) with optical triangulation feedback and a fiber laser (IPG, YRL-400-AC; 400 W). A feed-forward loop was used to achieve positioning accuracy of under 0.05 mm during on-the-fly welding. Experimental results show that between welding speeds of 25 and 150 cm/min, average tracking deviations are 0.043 mm and 0.276 mm in y and z directions, respectively. Moreover, teaching times for a specified seam can be shortened for more than 10 times due to the fact that only rough seam teaching is required. The proposed system configuration could be adapted to other classical welding processes.

No abstract

The increase in complex workpieces with changing geometries demands advanced control algorithms in order to achieve stable welding regimes. Usually, many experiments are required to identify and confirm the correct welding parameters. We present a method for controlling laser power in a remote laser welding system with a convolutional neural network (CNN) via a PID controller, based on optical triangulation feedback. AISI 304 metal sheets with a cumulative thickness of 1.5 mm were used. A total accuracy of 94% was achieved for CNN models on the test datasets. The rise time of the controller to achieve full penetration was less than 1.0 s from the start of welding. The Gradient-weighted Class Activation Mapping (Grad-CAM) method was used to further understand the decision making of the model. It was determined that the CNN focuses mainly on the area of the interaction zone and can act accordingly if this interaction zone changes in size. Based on additional testing, we proposed improvements to increase overall controller performance and response time by implementing a feed-forward approach at the beginning of welding.

Original scientific paper One of the key challenges in robotic remote laser 3D processing (RL3DP) is to achieve high accuracy for the laser's working trajectory relative to the features of the workpiece. This paper presents a novel RL3DP system with an automatic 3D teaching functionality for a precise and rapid determination of the working trajectory which comprises a robot manipulator, 3D scanning head, fibre laser and an off-axis positioned camera. The 3D measurement is based on laser triangulation with laser-stripe illumination using the laser's pilot beam and scanning head. The experimental results show that the system has a precision better than 70 µm and 120 µm along lateral and vertical direction respectively inside the measuring range of 100 × 100 mm. The teaching time is 30-times shorter compared to a visual teaching procedure. Therefore, such a system can lead to large cost reductions for modern production lines that have constant changes to the products' geometries and functionalities. Keywords: edge detection; laser triangulation; remote laser processing; robot teaching; three-dimensional measurement Automatsko učenje robotskih laserskih daljinskih 3D obradnih sustava na osnovu laserske triangulacijeIzvorni znanstveni članak Jedan od bitnih izazova u daljinskoj laserskoj 3D obradi (RL3DP) je postizanje visoke točnosti jer laser radi po rubovima predmeta koji obrađuje. Ovaj rad prikazuje novi RL3DP sustav s automatskom 3D nastavnom funkcionalnošću radi točnog i brzog određivanja prijenosa detektiranih rubova koje obrađuje robot sa 3D skenerom, fiber laserom i izvan aksijalno pozicioniranom kamerom. 3D skeniranje se bazira na laserskoj triangulaciji sa pilotskom laserskom trakom. Eksperimentalni rezultati pokazuju da sustav ima preciznost bolju od 70 µm i 120 µm u bočnom i vertikalnom smjeru u mjernom području od 100 × 100 mm. Vrijeme učenja je 30-puta kraće u odnosu na vizualni postupak. Stoga, takav sustav može značajno smanjiti troškove obrade sa modernim proizvodnim sistemima koji se moraju prilagođavati stalnim promjenama geometrije i funkcionalnosti proizvoda.

No abstract