Pattern-recognition analysis of sound radiation in metal cutting

Abstract

Pattern-recognition analysis of sound radiation was developed as a basis for monitoring the metal-cutting process. The sound-pressure signals radiated during cutting under fixed conditions were recorded using a sharp tool, a worn tool and a broken tool. A round bar of heat-treated AISI 4340 steel was machined in cylindrical turning, using a titanium-carbide-coated cutting tool. The spectral components in the 0–10 kHz range were used as features. Using resubstitution, signals coming from sharp and worn tools were easily distinguished, resulting in a 100% classification with just two features, being the average of the low (0–5 kHz) and high (5–10 kHz) frequency ranges. With tool breakage, two acceptable classifications were obtained using independent testing, the first with a performance of 100%, 75% and 100% for sharp, worn and breakage signals respectively, using nine features, and the second with 80%, 88% and 80%, using eight features.