Tool wear monitoring in turning using a pattern recognition technique

The automation and optimization of the manufacturing process play an important role in improving productivity. For this, monitoring and diagnostic systems are becoming increasingly necessary in manufacturing. In this paper, pattern recognition analysis by the linear discriminent functions approach is attempted for in-process detection of tool wear in turning. Different methods of representation of forces are put forward and their relative merits are analyzed. Log-linear models have been constructed to estimate force components for sharp cutting edges, as influenced by feed, speed and depth of cut. The data obtained are separated into training and checking sets, a criterion function being proposed for determining the data that should be placed in the training set. Various heuristics involved in pattern recognition analysis, such as the amount of data used for training and the value of the learning rate, have been studied. © 1993.