Use of contour signatures and classification methods to optimize the tool life in metal machining

Abstract

Tool replacement operations have a great influence over the cost of machined parts. At present, the common criteria used to determine the tool life do not optimize the use of tools and lead to significant economic losses. The main objective of this work is to define a new procedure to improve the decision about the time for tool replacement. The approach followed is based on digital images of the cutting edge that has already been explored in the literature. The novelty of the present work relies on the use of contour signatures of the wear region as the input to two techniques of classification, k-nearest neighbour and a neural network. The influence of the signature size vector is also analysed. A total of 1383 flank wear images were acquired and the error rate estimation was about 5%.