An approach based on singular spectrum analysis and the Mahalanobis distance for tool breakage detection

Abstract

The failure of cutting tools significantly decreases machining productivity and product quality; thus, tool condition monitoring is significant in modern manufacturing processes. A new method that is based on singular spectrum analysis and Mahalanobis distance are combined to extract the crucial characteristics from spindle motor current to monitor the tool's condition. The singular spectrum analysis is a novel nonparametric technique for extracting the properties of nonlinear and nonstationary signals. However, because the components are not completely independent, the original singular spectrum analysis eventually leads to misinterpretation of the final results. The proposed method is used to overcome the weakness of the original singular spectrum analysis. The singular spectrum analysis algorithm is adopted to decompose the original signal and the useful singular values that correspond to the tool condition can be extracted. The Mahalanobis distance of the singular values is proposed as a feature that can effectively express the tool condition. The experiments on a CNC Vertical Machining Center demonstrate that this method is effective and can accurately detect the tool breakage in mill process.