On-line Tool Condition Monitoring Based on PCA and Integrated Neural Networks for Cold Blast Machining Operation

Abstract

In order to realize on-line monitoring of tool condition for cold blast turning operation, a multi-sensor intelligent system based probability neural network was investigated to build the relation between signal features and tool condition. A KISTLER 9257B dynamometer and three KISTLER 8702B50M1 accelerometers were used to monitor the change of tool condition. Cutting force signals and vibration signals were processed by different methods and features sensitive to tool condition were selected by principal component analysis. The structure of probability neural networks (PNN) was optimized by PCA and probability theory. After training with 428 samples, the PCA-PNN could accurately predict tool condition and maximal classifying precision is 98.2% for condition D. Experimental results show that the optimized PNN has a simpler structure than the traditional PNN to get a similar result.