Development of seam tracking sensor for cartesianautomation robots

In today's industrial manufacturing processes, automation is rapidly expanding in parallel with technological advancements, and many tasks that were previously performed manually are now carried out through robotic systems. One of the most prominent examples of this transformation is the welding process. Robotic welding systems are widely adopted in production lines due to their ability to operate continuously, 24 hours a day and 7 days a week, without being exposed to harmful environmental conditions such as toxic fumes, ultraviolet radiation, and elevated temperatures. In addition to improving productivity and efficiency, such systems significantly reduce health risks for human operators. Nevertheless, not every welding application can be directly integrated into robotic automation. Conventional robotic systems generally operate along pre-programmed paths, which makes them vulnerable to deviations arising from thermal deformation of the workpiece, shrinkage of molten metal during rapid solidification, or insufficient production tolerances. These positional deviations directly affect weld quality and therefore necessitate the use of seam tracking sensors that enable the welding torch to dynamically follow the actual joint path. Within the scope of this thesis, a laser-based seam tracking sensor is developed for integration into Cartesian automation robots. The proposed sensor is capable of positioning the welding torch along the seam with a precision of ±0.1 mm and operates reliably on both matte and highly reflective metallic surfaces. The mechanical design of the sensor is configured to remain minimally affected by adverse conditions during welding, such as elevated temperature, arc light, spatter, and fumes. Distortions in visual data caused by these interferences are addressed through a hybrid approach that combines traditional image processing techniques with artificial intelligence–based filtering methods, with customized algorithms tailored for different surface properties. The sensor extracts the geometrical profile of the weld seam by analyzing the deformations of a projected laser line on the workpiece surface. This profile data is processed using the laser triangulation method, and by fusing it with position feedback from the Cartesian robotic system, the three-dimensional coordinates of the weld path are determined. Based on this information, a novel position control algorithm is implemented to guide the robotic torch in real time, enabling accurate seam tracking and consistent weld quality.