Rapid uncertainty quantification for structural full-field dynamic responses with extremely high dimension

Conducting a full-field dynamic analysis under structural uncertainties is of great significance for a better understanding of structural mechanics behavior and obtaining structural reliability. However, the full-field dynamic response of structures is of extremely high dimensionality, and traditional uncertainty quantification (UQ) methods may face challenges such as modeling difficulties and low analysis efficiency. To address these issues, this paper proposes a rapid UQ analysis method for the structural full-field responses with extremely high dimension. This method first decouples the ultra-high dimensional full-field response based on modal analysis and further extracts features of the responses based on manifold learning techniques, effectively reducing the dimensionality of the response to be analyzed. Subsequently, by introducing the optimal sparse polynomial chaos expansion technique, an efficient UQ analysis model from structural uncertainty parameters to response is constructed. Three numerical examples are provided to demonstrate the accuracy of the proposed method. Throughout the entire UQ analysis process, only a small amount of low-dimensional features need to be analyzed, and the final UQ accuracy of the full-field dynamic response can be effectively guaranteed. Therefore, the proposed method provides an effective tool for rapid UQ analysis of structural full-field dynamic responses.