Pyroelectric sensor signals processing for early-warning generation in CO2 laser welding

Three different techniques were chosen for processing pyroelectric sensor signals that would provide early warnings of malfunctioning in the CO2 laser welding. The corresponding algorithms were implemented using the mean, the root mean square and variance of a suitable number of digitized signal samples. Each pair of these algorithms was found complementary in reaching 100% success in the early individuation of the defects. The one based on the variance was more efficient than the other two in identifying the causes of malfunctioning. The processing times ranged from 20 μs to 200 ms, and the spatial resolutions varied from 18 μm to 0.5 mm.