Online tool wear classification in turning with time-delay neural networks and process-specific pre-processing

The online determination of a tool's wear in order to exchange it just in time makes high demands on a sophisticated tool monitoring systems. Research has shown that it is possible to use neural networks for the estimation or classification of wear. The article demonstrates that remarkable improvements of the classification can be obtained using available secondary information about physical models of the cutting process and neural networks considering the position of a single input pattern in a pattern sequence. Process models describing the influence of process parameters are used for dedicated pre-processing of the sensor signals. The essential behaviour of these aligned signals in a certain short time window is described by means of polynomial coefficients. The coefficients are used as inputs for feedforward networks considering their temporal development (time-delay neural networks). With a combination of the proposed measures the rate of correct classifications can be increased significantly.