Application of singular spectrum analysis to tool wear detection using sound signals

Abstract

The aim of the present work is to study the applicability of singular spectrum analysis (SSA) to the processing of the sound signal from the cutting zone during a turning process, in order to extract information correlated with the state of the tool. SSA is a novel non-parametric technique of time series analysis that decomposes a given time series into an additive set of independent time series. The correspondence between the singular spectrum obtained using SSA and the frequency spectrum of the signal is the basis of this processing technique. Finally, some of the features extracted from the SSA-processed sound signal were presented to a feedforward back-propagation (FFBP) neural network to determine the tool flank wear. The results showed that the proposed processing technique is well suited to the task of signal processing in the area of tool condition monitoring (TCM).