An adaptive orbital system based on laser vision sensor for pipeline GMAW welding

This study focused on the improvement of an adaptive orbital welding system, applying a laser-based vision sensor, with the objective of increasing its reliability and robustness by proposing a calibration for the manipulator's rail deformations, and also developing novel methods and algorithms based on neural network techniques for the adaptive parameters setup. The main problem with mobile-based robotic welding systems that use computer vision for trajectory control arises from the offset between the sensing point and the welding torch, in conjunction with the geometric deformations of the mounted rail. This configuration in moving base manipulators generates an error on the online trajectory correction. In this context, methods of verification and calibration of this systematic error intrinsic to the geometric composition are necessary. Welding tests were conducted in order to validate the proposed methods, for both rail form calibration and adaptive welding parameters setup. The enhanced system was capable of overcoming significant geometrical variations of the gap and pipe mismatch. A previous scan was used to calibrate the rail deformation, allowing for single laser line projection sensors to be utilized for this purpose. The proposed methods are restricted to not only welding operations of large-diameter pipes, but also for general welding applications and possible inspection operations as well.

Automated summary

This study aimed to improve an adaptive orbital welding system by proposing novel methods and algorithms based on neural network techniques, as well as a calibration method for manipulator's rail deformations. The proposed methods were validated through welding tests and demonstrated the system's capability to overcome significant geometrical variations.