Comparing Diffuse Approximation and Kriging Method for Predicting the Tool Life when Milling Ultrahigh Strength Steel

Abstract

Tool wear can affect surface quality of the work-piece and frequent tool changing leads to low machining efficiency. Tool life of milling ultrahigh strength steel is thus an important indicator in manufacturing. In this study, we compare Diffuse Approximation (DA) and Kriging method for predicting tool life of milling ultrahigh strength steel. Latin hypercube method is employed as design of experiments after comparing three different designs of experiments (full factorial design, Latin Hypercube Sampling (LHS) and random design). An example with LHS is taken to investigate accuracy of approximation of DA and Kriging method. The results show that the DA provides significant improvement in accuracy of prediction compared with Kriging method. Tool life of milling ultrahigh strength steel predicted by DA is almost close to ‘exact’ values, especial for big mutation of adjacent measured values. The most errors with DA are less than ones with Kriging method.