Abnormal State Detection in Cutting Process by Using Model Parameters

Abstract

This paper deals with features that suggest states of cuttings and a method for detection of abnormal states of cuttings. Damping ratios, natural frequencies and adaptation-errors of the time series models used are estimated by fitting the time series models to acceleration signals detected at the tools in normal cuttings, in cuttings with worn tools and with broken tools, and in cuttings when chatter vibration occurs. An adaptive filter is used to identify the time series model. A screening of the estimated parameters reveals that the damping ratios and the adaptation-errors are acceptable as the features that suggest the states of cuttings. Salient increases and decreases in values of the features are observed especially for the tool breakage and the chatter vibration. A method applying the fuzzy inference on the damping ratios and the adaptation-errors is provided for judgement of states in cuttings and proves effective through a simulation with experimental data.