Cutting tool monitoring by acoustic emission based upon wavelet-neural networks

摘要

The features of cutting tool states (normal, worn, breakage) were extracted using Acoustic Emission (AE) signals. AE signals were measured by a built-in piezoelectric transducer, which was inserted in the tool holder of an NC lathe. The 8 wavelet packets were taken using wavelet packet analysis for 3-leve. The powers calculating from the 8 wavelet packets were as 8 nodes of the input layer in BP neural networks, which identified three states of cutting tool. The corrected rate of classification in the experiments were normal 100%, worn 95%, breakage 95%. The results obtained show that this method is reliable and efficient.