Force-based approach to on-line flute breakage detection in a peripheral end milling process

This paper presents a force-based algorithm that is capable of rapidly detecting flute breakage and that is also robust to the effects of transients (changing cutting conditions), cutter runout, and noise in the force signal. The algorithm is based on the periodicity of the steady state cutting force signal and employs efficient algorithms providing potential real-time application. The algorithm utilizes a two-step signal processing approach to filter out the dynamic components contained within the raw force signal. The step employs a differencing technique, corresponding to the number of points per revolution, to quickly reduce the dynamic structure in the cutting force signal. The second step further reduces the dynamic structure in the 'revolution differenced' cutting force signal through an autoregressive model that is recursively updated. Flute breakage detection limits with a desired level of confidence can be imposed about the generated forecast error series by estimating its variance. The algorithm is demonstrated to be effective in detecting flute breakage through a series of down milling experiments including three actual flute breakages: a 'natural' flute breakage occurring during steady state cutting, a breakage on entry, and a breakage on exit. The algorithm is also demonstrated to be insensitive to the effects of cutter runout, entry/exit cuts, and noise in the force signal. Results indicate that the algorithm shows excellent potential for practical, on-line application.