Sample partitioning for milling stability and surface location error model uncertainty reduction

Abstract

This paper describes uncertainty reduction in milling dynamics process modeling, including both stability and surface location error, using sample partitioning. In this approach, the model inputs are defined by distributions and samples are retained or rejected using stability test results. Stability maps are calculated using uncertain inputs, including modal parameters that represent the tool tip frequency response functions and cutting force model coefficients that relate the cutting force to the commanded chip area. After the uncertain inputs are partitioned by limited stability tests, the combined stability/surface location error map is predicted for final milling parameter selection. A numerical case study is provided to evaluate the approach. It is shown that 10000 initial samples were reduced to 224 samples after five tests. The maximum error between the input and output parameters was 3.8% and the minimum was -0.03%, providing initial validation of the approach.