Study on Design of Experiments based Kriging Model for Tool Life Prediction

For the tool life prediction, it is hard to obtain the exact model to do prediction. The establishment of surrogate model for the prediction or optimization of complex issues is a useful way. However, the accuracy of surrogate model mostly depends on the arrangement of samples (experiment data). Therefore, in this paper, three types of design of experiments (DOE) are analyzed and compared based the Kriging model. The six-hump camel back function is used to test the accuracies and amounts of computation of three DOEs. The results show that Latin hypercube sampling design with the Kriging model method is the best combination for the accuracy of surrogate model and the amounts of computation of DOE. Finally, this combination is applied to develop the surrogate model of tool life prediction and the accuracy is accepted in practical engineering.