Sample Partitioning for Uncertainty Reduction in FEA-Based Milling Stability Prediction

Milling stability prediction requires tool and workpiece dynamics and cutting force coefficients, and uncertainty in these inputs propagates to the predicted stability maps. This study experimentally evaluates sample partitioning for uncertainty reduction in milling stability when the workpiece dynamics are predicted using finite element analysis (FEA). The FEA-predicted workpiece natural frequency and modal stiffness differed from the tap-tested reference values by 4.9% and 14.3%, respectively, leading to different predicted stability maps. Initial candidate stability maps were generated by Monte Carlo simulation using the uncertain FEA-predicted workpiece modal parameters. The experimentally identified cutting force coefficients were also treated as uncertain inputs. Physical cutting tests on a constrained motion dynamometer system were then used to retain or reject candidate maps based on the observed stable/unstable outcomes. The candidate maps were reduced from 10,000 to 6 after five partitioning steps, and the retained maps moved toward the tap-tested reference stability map. These results demonstrate that sample partitioning can reduce uncertainty in FEA-based stability maps using a limited number of cutting tests.