Abstract

Characteristic spark cone images are produced beneath the workpiece during reactive gas-assisted laser-cutting processes. Laser-cutting quality is controlled using robust optical pattern recognition and classification of the spark cone images. High irradiance fully formed spark cones are characteristic of a high quality dross-free cut; whereas, a sparse cone indicates a poor quality cut with dross attachment to the lower surface of the workpiece. It was found that by incorporating both the out-of-class image, sparse spark cone, and the in-class image, fully formed spark cone, into a Wiener filter construction, the discriminant ability of the imaging control system was significantly enhanced, the incorporation of three out-of-class images gave optimum performance. The Wiener filters formed from spark cone images, recorded when cutting different thickness materials, were integrated into a synthetic discriminant function (SDF) filter. When spark cone images from the laser cutting process were cross-correlated with the SDF, it gave a high output peak for a good quality cut, and a low output noise signal for a poor quality cut.^1-5

© 1994 IEEE