IN-PROCESS TOOL-FAILURE DETECTION BY MEANS OF AR MODELS

The present paper proposes a cutting tool breaking and chipping detect ion system for continuous and interrupted cutting, based on the analys is of the cutting force components F-x and F-y. A multifactorial exper imental design has been carried out, to take account of the variabilit y of the force signal. An adaptive signal processing algorithm is prop osed, which detects catastrophic failure when at least one component d eviates outside an estimated oscillation band for a time duration long er than a prefixed interval. The algorithm has been implemented on a f our-microprocessor transputer board Several tests confirmed the validi ty of the approach for detecting breaking and chipping phenomena in a few milliseconds, both in turning and in milling operations.