Application of Kohonen neural network for tool condition monitoring

This paper presents a Kohonen neural network application for on-line tool condition monitoring in a turning operation. A wavelet technique is used to decompose dynamic cutting force signal into different frequency bands in time domain. Two features are extracted from the decomposed signal for each frequency band. The two extracted features are mean values and variances of the local maxima of the absolute value of the composed signal. In addition, coherence coefficient in low frequency band is also selected as a signal feature. After scaling, these features are fed to a Kohonen neural network for diagnostic purposes. A framework of a tool condition monitoring system using the neural network is presented. The tool wear criterion and the parameters of the neural network are discussed. The experimental results show that the unsupervised neural network has great potential in an on-line tool condition monitoring system, especially when the number of input features is larger.