A statistical approach to acoustic monitoring of laser welding

Acoustic emission at frequencies between 20 Hz and 20 kHz during CO2 laser welding of steel contains important diagnostic information related to weld morphology, depth of penetration and the heat-affected zone. We have recorded acoustic spectra under a variety of welding conditions with mild and galvanized steel sheet and find that a well-defined spectrum is obtained under conditions that lead to an optimization of these weld characteristics. A statistical approach is developed which yields a predictive measure of weld quality, based on a comparison between real-time acoustic spectra, and that obtained under optimized conditions. This technique is applied to an analysis of bead on plate, lap and butt laser welds in mild and galvanized steel.