Prediction of small drill bit breakage by wavelet transforms and linear discriminant functions

Though the discrete wavelet transform (DWT) has gained attention in the fie ld of machining monitoring, the potential uses of DWT-based classification techniques have not yet been fully explored. In this paper, linear discrimi nant analysis is used to post-process DWT output for on-line prediction of the breakage of small drill bits. Bit failure is characterized by two types of transient ('sawtooth' and 'screeching') in the cutting force signal. To detect these transients, instead of traditional Fourier-based methods the DWT is used, which is better suited to analysis of time-localized phenomena . Three index functions ('energy', 'waviness' and 'irregularity') are adopt ed to test for the presence of transients in the DWT expansion. The indices are used to perform linear discriminant analysis, thereby classifying the input signals by state (normal or prefailure). Experiments showed that the DWT-based linear discriminant analysis method can accurately identify impen ding breakage about one to three cycles prior to failure, even when the cut ting conditions change. (C) 1999 Elsevier Science Ltd. All rights reserved.