Dynamic analysis of multiple response patterns in a flexible multibody system with hybrid uncertainties

Multibody systems are often subject to multiple sources of uncertainty, including both random and interval types. The dynamic response of such a multibody system may exhibit multiple patterns that traditional surrogate modeling methods fail to capture effectively. To address this issue, this paper presents a hierarchical hybrid Kriging modeling approach for accurately predicting the multiple dynamic response patterns in a flexible multibody system with hybrid uncertainties. The approach employs adaptive K -means clustering to automatically identify the dynamic response patterns, assigning the inner-layer random samples and their dynamic responses to respective clusters under fixed outer-layer intervals. Following this assignment, variance-based sequential sampling augments the sample sets. The paper details the hierarchical hybrid Kriging model for each pattern, where the inner-layer model captures the influence of random uncertainties on the dynamic response, and the outer-layer model characterizes the propagation of interval uncertainties into the response mean and variance. This integrated process enables high-fidelity modeling and reliable prediction of multiple dynamic response patterns. Two classic examples demonstrate the high accuracy and computational efficiency of the method, which needs only 0.28 % of the computation time of the Monte Carlo-Scanning method.