BAYESIAN AND REGRESSION APPROACHES TO ONLINE PREDICTION OF RESIDUAL TOOL LIFE

In this paper, two statistical approaches to on-line prediction of cut ting tool life are presented and discussed. A Bayesian approach utiliz es in-process information about the cutting tool state and constitutes a valuable basis for improved prediction. A second approach is based on the cutting forces and facilitates a prediction of the tool life wi th an uncertainty of 15% after 1.5-2.0 cutting minutes. Traditional to ol condition monitoring can be improved by increased reliability of to ol life predictions, increased utiltzation of the cutting tools togeth er with reduced need for pre-process data and calibrating procedures. (C) 1998 John Wiley & Sons, Ltd.