Real-time roughness estimation in laser oxidation cutting via coaxial process vision

Laser cutting is an established technology for the processing of metal sheets and tubes given its elevated productivity and high part quality. However, external influences or variations in the process conditions may affect the quality of the final product. In oxidation cutting, cuts are typically evaluated by means of profile roughness whilst critical defect formation consists in loss of cut. Real-time estimation of the cut quality via process monitoring is of great interest since it enables inline evaluation of the manufactured components and identification of defected parts. Such capabilities were investigated during the cutting of high thickness mild steel, acquiring process emission images with a coaxial monitoring system and correlating them to the profile roughness via Machine Learning algorithms. Results indicate roughness predictions with good fitting (R2>80%) and with a Mean Absolute Error below 10 μm (RRzz parameter).