Welding AI
Brief description
The aim of the project is to develop an AI-supported robotic arm system that uses optical sensors to recognize different weld seam geometries and performs the welding process automatically with high precision. Weld seam shapes and defects are examined in order to define quality characteristics on the basis of which an evaluation score is developed that can be used to classify a weld seam using suitable sensor technology. 3D profile sensors with an accuracy of +- 0.5 mm are integrated and optimized under various reflection and spark conditions. A 3D volume model of the weld object is created, which automatically compensates for deviations of +- 0.25 mm and continuously adjusts the welding parameters. In addition, a calibration system is developed that ensures the quality of the data required for the project's AI and calibrates the sensors to reduce rejects to less than 1%. Finally, a prototype with plug-and-play capability is created that can be integrated into existing systems within 6 hours, while sensor technologies and safety measures are optimized to prevent errors.
