Cutting tool wear and work hardening monitoring through cutting sound classification and machine learning in 304 stainless steel

Abstract

Currently, there is a very marked trend toward digitization and robotic control of manufacturing processes; thus the target is to optimize the economic resources. Therefore, the monitoring resources usage such as tools monitoring in the machining process become vital into enterprise optimization strategy. The above, since important savings in manufacturing processes can occur. The 304 stainless steels are widely used in the industry due to its mechanical and corrosion protection features. However, its metallurgical characteristics are difficult to machine. In that sense, this work is focused to analyze surface finishing during milling by varying different operational parameters. Then, a Deep Learning algorithm was applied in order to predict the work hardening through its surface finishing by means of the differences in sound recorded during machining. The results suggests that most of the energy is concentrated below 12 kHz. In addition, test specimens machined without coolant fluid yielded shallow surface strains in comparison to the machined ones with coolant fluid. Finally, the mean absolute percentage error for roughness prediction was about 28% suggesting a good correlation. Thus, the proposed methodology can be incorporated to an appropriate alarm and roughness measurements can be employed to detect the increase in work hardening. In both cases to prevent the tool failure.

Graphical abstract