Reliability Analysis of Multidisciplinary System with Spatio-temporal Response using Adaptive Surrogate Modeling

摘要

This paper presents an adaptive surrogate modeling framework for the reliability analysis of a multidisciplinary system with spatio-temporal response. The Kriging surrogate method in conjunction with singular value decomposition is first employed to replace the original computer simulation models in the multidisciplinary system analysis. Due to the limited computational resources, the initial surrogate models may not accurately represent the original computer simulation models, which results in errors in the multidisciplinary reliability analysis (MDRA). To tackle this problem, this paper then develops a method to analyze the effects of surrogate model uncertainty on the results of MDRA. Variability of the system responses over time and space caused by the surrogate model uncertainty is considered during the analysis. Based on uncertainty effect analysis, a new approach is proposed to adaptively allocate the computational resources to train individual disciplinary surrogate models and thus to improve the accuracy of MDRA. A four-step refinement procedure is developed based on the variance-based global sensitivity analysis (GSA) method to determine at what input setting, which discipline, when, and which surrogate model to allocate the computational resources. A hypersonic vehicle panel subjected to hypersonic flow conditions is used to demonstrate the proposed method. The results demonstrate that the proposed adaptive surrogate modeling method is able to efficiently and accurately perform the reliability analysis of a multidisciplinary system with spatio-temporal output.