A Neural Network Architecture for Tool Wear Detection through Digital Camera Observations

Abstract

In flexible manufacturing systems with unmanned machining operations, one of the most important issues is to control tool wear growth in order to identify when the tool needs to be replaced. Tool monitoring systems can be divided into on-line and off-line methods. The authors have already conducted both on-line and off-line analyses. The simplest way to check the tool status is to measure either the flank and the crater wear levels or the presence of a cutter breakage. This task, which can be easily performed off-line by the operator, gives a lot of problems when it is conducted automatically on-line. In the present paper a neural network for image recognition is applied for the wear level detection. The network receives as input an image of the tool, acquired by a digital camera mounted near the machine tool storage, and provides a binary output which indicates whether the tool can continue to work.