Comparison of methods for data analysis in the remote monitoring of remote laser welding

Abstract A proven solution for the on-line monitoring of the gap in remote fibre laser welding of overlapped zinc-coated steel is based on the analysis of the visible optical emission that the welding process generates. Traditionally, different monitoring approaches are used, such as monitoring the overall emission or parts of the emission with filtered photodiodes or spectroscopic analysis of the wavelength domain. In the monitoring of welding defects, these approaches can lead to different performance results. In the present paper, different methods to monitor and analyse the visible emission are compared. The monitoring strategy uses an hardware known as Through Optical Combiner Monitoring (TOCM) that allows the signal emitted by the welding to be acquired directly at the laser source. The paper aims to evaluate the ability of monitoring methods to identify the effects of variable factors, such as the gap between the plates and the location inside the weld seam at which the variation of the gap occurs, on the monitored emission. The optical emission from 400 nm to 800 nm is monitored during remote laser welding from the optical combiner of the fibre laser source. First, the entire optical emission is examined with a spectroscope with a wavelength resolution of 0.57 nm. Second, multivariate data analysis is used to evaluate different indicators, such as the overall emission in the considered range, the emissions in separated wavelength ranges according to physical evaluations of the welding process or the entire spectrum. For each of the obtained indicators, variance analysis is performed, and the statistical significance of the gap value and its location in the weld seam are used to compare the performance of the tested methods.