Active Learning Polynomial-Chaos Kriging Based Uncertainty Quantification and Reliability Analysis of a Spline-Shafting System

The flexible rotor system connected by a floating involute spline may experience self-excited vibrations due to internal friction excitation. The self-excited vibration induced by inevitable manufacturing tolerances, installation errors, component deformations, and other influencing factors exhibits significant uncertainty. In this study, we employ the active learning surrogate model to analyze the reliability of the nonlinear self-excited vibrations of the spline-shafting system. Firstly, considering the tooth surface friction, a nonlinear dynamic model of the rotor system is established for deterministic analysis of self-excited vibration response. Subsequently, considering the randomness of input parameters, Polynomial-Chaos Kriging (PCK) surrogate model is constructed to evaluate the failure probability of nonlinear self-excited vibrations in the rotor system under supercritical speed. The surrogate models are treated as black-box models, facilitating their combination with time-domain numerical integration and Monte Carlo Simulation (MCS), enabling rapid estimation of the relevant statistical information of self-excited vibrations. This approach contributes to the extension of applied research on uncertainty quantification and probabilistic reliability assessment for complex rotating machinery.