基于证据理论的结构高维全场响应不确定性分析OA

The propagation analysis of high-dimensional structural responses under complex uncertain variables is prone to problems such as low modeling efficiency and poor analytical accuracy.To address this issue,a rapid analysis method for high-dimensional structural response uncertainty based on evidence theory was proposed.Firstly,the basic probability assignment of evidence variables was utilized to generate sample sets efficiently via optimal Latin hypercube sampling.Secondly,principal component analysis was applied to reduce the dimensionality of the high-dimensional structural responses,extracting low-dimensional features and eigenvectors to decrease modeling complexity.Finally,an extreme learning machine was employed to construct the mapping relationship between the uncertainty parameters and the low-dimensional features,enabling the prediction of uncertainty propagation at any position of the high-dimensional responses with respect to the input variables.Validation through the examples demonstrated that the proposed method could effectively quantify the uncertainty distribution at arbitrary positions of both high-dimensional time-domain and spatial responses,achieving high accuracy with relatively high modeling efficiency.The proposed method can significantly reduce the complexity of uncertainty analysis for high-dimensional full-field responses and serve as an effective tool for uncertainty propagation analysis of complex engineering structures.