The capability of OCT technology for real-time seam tracking and weld bead quality evaluation was examined in robot-guided automated GMAW and tactile laser welding and brazing of automotive parts made of bare steel, coated steel and aluminum with different joint types at different welding speeds. The permissible gap size and gap compensation were investigated for both automated conventional and tactile laser welding. During tactile laser welding and brazing, the filler wire speed and laser power were varied. Satisfactory results were achieved: The industry-proven, standardized OCT functions and fieldbus communication enabled accurate and consistent seam tracking, gap bridging and online quality monitoring of the seam topography of the tested materials during processing with variable welding or brazing parameters. Thanks to the flexibility, insensitivity to environmental conditions and very high accuracy of the OCT system, the joining processes could be precisely controlled.

This study aims to facilitate adaptive beam shaping using a deformable mirror for real-time control in laser butt joint welding by estimating gap widths. Two methods capable of real-time gap estimation were investigated: adaptive Kalman filtering and deep learning. The Kalman filter was evaluated experimentally, enabling adaptive beam shaping for gap bridging. Results demonstrated significant reductions in residual distortion and compliance with ISO 13919-1 Level D standards for joint gap widths up to 0.55 mm. Separately, deep learning methods, particularly Convolutional Neural Networks, were evaluated for gap width classification and tack weld detection, achieving over 99% accuracy under controlled conditions and 96% in noisy conditions. These findings highlight the potential of both adaptive Kalman filtering and deep learning for advanced gap characterization for real-time control, underscoring their roles in improving welding adaptability and weld quality.

During laser deep penetration welding, involuntary disturbances of the nominal process can lead to seam imperfections. Due to their possible impairing characteristics on the mechanical properties of the seam, it is of high interest to reliably detect these defects. This can be done by means of sensor based inline process monitoring. While the sensor data can be used to identify the defect location and perform post process repair work, it can also build the foundation to reweld the detected seam segments inline. This study demonstrates an approach to detect process anomalies during laser deep penetration welding of hidden T-joints inline and use a LabVIEW based control algorithm to utilize the scanning capabilities of the welding optic to reweld the defected seam area inline to enhance the seam quality and possibly eliminate necessary post process work.

The requirement of industrial development is comprehensively controlled production to ensure component quality despite short development times or due to batch fluctuations during production. Process evaluation through multi-modal sensor technology, real-time data processing and the use of artificial intelligence (AI) plays a decisive role in processing and control in order to ensure optimal process parameters. The paper presents possibilities for recording process emissions using sensor technology (e.g. acoustics, image-evaluation) for highly dynamic laser welding using 2D-scanner exemplary for copper. Furthermore, a real-time capable platform is presented, which is used for data acquisition and provides an interface for AI-based data processing for subsequent process control. The experimentally determined sensor data is used to establish a correlation between process quality of cop-per weldings and process parameters. The first AI-models developed in this way have been trained with an initial data set of approximately 160 trials and are currently in an evaluation phase.

To protect operators from laser radiation, laser machines must have a protective enclosure. The requirements for laser protective shielding are defined in the current standard EN 60825-4. According to this, extensive case-related laser stability tests must be carried out on the protective material. These case-related tests make it difficult to compare and select suitable materials and thus considerably increase the effort for manufacturers. Therefore, the SALSA project is taking a different approach: The influence of the beam diameter and the irradiance on the laser resistance of common protection materials is systematically investigated and the relationships found are expressed in simple mathematical formulas. This should enable manufacturers to use exemplary laser stability measurements to extrapolate service lifetimes for any operating conditions. The study describes the relationships found using green laser radiation, since laser radiation in the visible wavelength range is currently of particular importance for applications in e-mobility.

The increasing adoption of robotic laser welding highlights the need for robust seam-tracking systems to ensure precision and adaptability. Current methods struggle with fixturing errors, part-to-part variations, thermal deformation and the need for precise alignment of the laser beam, especially in three-dimensional welding scenarios. Laser welding is particularly sensitive to variations in joint morphology and material reflectivity, requiring advanced solutions to maintain high welding quality. Spatial beam shaping techniques, based on beam oscillation or “wobbling”, partially mitigate these issues by accommodating variable joint gaps and high-reflectivity materials. However, solutions, capable of real-time trajectory planning and adaptive process control, are essential to meet the requirements of this technology. This work presents a comprehensive analysis of the technological and computational requirements for seam-tracking operations, providing a foundation for developing systems capable of operating at high welding velocities, while ensuring quality and minimizing rework, paving the way for a more-effective implementation in smart-manufacturing.