Correlation between roughness of turned workpieces and vibration signature of the lathe bed

The identification of non-desired working conditions during machining is a necessary step to reduce faulty workpieces. To complicate the matter are the wide variety of possible variables which can change the resulting quality (machine parameters, geometry and material of the raw bar, …). The identification of a key feature from the vibratory signals of the machine can provide a Data-Driven solution. An experimental campaign was carried out and the vibratory signals were analyzed using traditional frequency approaches without success. A more advanced method, involving Spectral Kurtosis, was thus applied. It was found that the PSDs of the faulty signals, processed with the mentioned method, show a much higher power order than the healthy ones. Conclusively, it was found a possible connection between the faulty conditions and the critical frequencies of the bar free-length.