In laser beam welding of hidden T-joints the web sheet is completely covered by the face sheet, thus impeding the determination of the relative position between the laser beam and the web sheet. This circumstance usually raises high demands on the clamping accuracy as an in-process correction of the beam path by means of optical reference measurements is difficult. Previous research has shown that optical coherence tomography is capable of distinguishing between alignment or misalignment between the beam position and the web sheet. While this distinction has already been employed for controlling the weld path, it was not yet possible to determine from the measurement data the information in which direction the laser deviates from the web sheet, resulting in a random initial guess for the beam path correction. In this research the asymmetry of the process zone when deviating from the web sheet position is exploited to derive information about the direction in which the weld path deviation occurs. For this purpose, the OCT measuring beam is shifted laterally to the laser beam to capture asymmetric keyhole features which are specific for the respective direction of the weld path deviation. The OCT signals are evaluated by means of analytical approaches as well as neural networks. It is shown that the direction of the weld path deviation can often be determined from the OCT measurement data, thus allowing for a loopless setting of initial beam path correction.
Keywords
- Hidden T-Joints
- Laser Beam Welding
- Neural Networks
- Process Controlling