Modelling of multivariate time series for tool wear estimation in finish-turning

Abstract

In finish-turning, the traditional flank or crater wear estimation alone is no longer adequate, whilst the wear at the minor cutting edge is of primary concern since it directly affects the surface quality and dimensional accuracy of a finished product. This paper describes the development of an overall tool wear monitoring strategy in finish-turning. To estimate various types of wear developing at the different faces of a tool insert, multidimensional force and vibration signals are used, and a stochastic technique, based on multivariate autoregressive moving average vector models (ARMAV), is used to quantify the process dynamics embedded in the signals. Based on that, a dispersion analysis is carried out to discriminate among signal ingredients each of which is sensitive to a particular type of wear. Experimental results show that the groove wear at the minor cutting edge, once formed, dominates the finishing-tool life, while in the case of no groove formation at the minor cutting edge, the minor flank wear reaches its critical point earlier than the major flank or crater wear does. The results also show that the method derived from the dispersion analysis is a feasible approach to on-line tool wear estimation in finish-turning.