Quantitative identification of laser cutting quality relying on visual information

In the industrial practice of the laser cutting, the cut quality is defined in a qualitative manner by skilled technicians. Specific features lying on the cut edges in fact compose the overall quality, i.e. inclination of striations, presence of burr and different process zones along the edge. These attributes are evaluated by experts which at the end assess the cut quality on the base of their personal judgment. On the other hand, measurements of roughness and burr height in accordance to standards or internal procedures are also carried out. However, measuring is time consuming and more important is not always in agreement with the qualitative evaluation given by skilled technicians. In this scenario, the paper presents a method relying on visual information able to measure quantitatively a new class of quality attributes, which opportunely combined provide an index of performance consistent to the qualitative one based on experience. In this study, images of the cut edge of 5mm thick stainless steel AISI 304 cut with nitrogen as assisting gas are analyzed. The image analysis algorithm utilizes both standard gradient techniques, wavelets decomposition and analyses in the frequency domain for measuring periodic and not periodic quantities. Burr profile is isolated, the typical process zones are successfully identified and striation’s angle is computed for each zone. The gray analysis method combining multiple outputs from the image analysis algorithm is applied in order to compute the overall quality. The weights are set to express correctly the judgement of technicians. The method proved reliable, relatively fast and promising for further extension to different thicknesses and materials.