Monitoring of Cutting-Tool Wear Based on Pattern Recognition

The study of on-line tool wear monitoring is transformed into a problem of 2-class classifier design in statistical pattern recognition. Feature vector is extracted from vibration signal in the cutting process. The optimizing feature plane is formed according to projection theory. On the basis of this a discrimination function the G(D), having self-learning characteristics, has been proposed. It classifies situations of tool wear as the G(D). The results indicate that the recognition rate is 95%, the leak away rate less than 0.6%, discrimination time less than 15s on a microcomputer PC 286. The proposed method can provide an on-line monitoring of tool wear in the cutting process.