Uncertainty quantification and dynamic characteristics identification for predicting milling stability lobe based on surrogate model

Abstract

The prediction of chatter-free machining parameters suffers from inaccuracies in dynamic milling model inputs and simplification in milling process modeling, which may lead to a significant mismatch between the predicted stability boundary of the mathematical model and actual physical experiments. This study proposes a novel stability analysis method for milling operations based on a surrogate model that considers the effects of both uncertainties and variations in model inputs. The uncertainties of inputs are quantified by considering the statistical distribution of both cutting force coefficients and modal parameters, and the variations of modal parameters are identified through operational modal analysis (OMA). Furthermore, the proposed method introduces the statistical Kriging surrogate model of the spectral radius in the model parameter domain to propagate uncertainties to the stability lobe diagram (SLD). The confidence interval of the predicted stability boundary is obtained using the estimated prediction variance of the generated Kriging surrogate model. Finally, a mathematical measurement of SLD quality is presented, based on the similarities both in shape and position between the predicted and experimental stability boundaries. The cutting experimental verification and numerical analysis indicated that the robustness and accuracy of the SLD are considerably improved compared to the state-of-the-art methods. Thus, the proposed method holds significant promise for practical engineering applications in controlling milling stability on machining equipment such as CNC tools and industrial robots.