Interval prediction of responses for uncertain multidisciplinary system
Interval prediction of responses for uncertain multidisciplinary system
复制标题
DOI：
10.1007/s00158-016-1601-4
复制
发表时间：
2017-06
影响因子：
3.9
通讯作者：
Xiaojun Wang;Ruixing Wang;Xianjia Chen;Lei Wang;Xinyu Geng;Weichao Fan
中科院分区：
文献类型：
--
作者：
Xiaojun Wang;Ruixing Wang;Xianjia Chen;Lei Wang;Xinyu Geng;Weichao Fan
考虑到概率方法需要大量的样本数据点，当信息不足时，非概率区间分析方法可以作为替代。本文提出了一种新的求解多学科系统响应界的策略，即基于迭代算法的区间不确定性分析方法（IA-IUAMs）。新方法分别采用Jacobi迭代和Seidel迭代两种迭代过程。基于Jacobi迭代的区间不确定性分析方法（JI-IUAM）利用并行子系统分析策略提高计算效率，而基于Seidel迭代的区间不确定性分析方法（SI-IUAM）利用最新信息加速收敛。两种IA-IUAM都能够准确、快速地评估响应的边界。并与基于区间不确定性分析的一般灵敏度分析方法（SIUAM）和传统的蒙特卡罗模拟方法（MCS）进行了比较。两个数值算例和两个工程算例表明了新方法的有效性。结果表明，IA-IUAM方法避免了数百次系统分析，比MCS方法效率更高; IA-IUAM方法避免了线性近似和全局灵敏度计算，比SIUAM方法精度更高，适用范围更广。
Considering that numerous sample data points are required in the probabilistic method, a non-probabilistic interval analysis method can be an alternative when the information is insufficient. In the paper, new strategies, which are iterative algorithm based interval uncertainty analysis methods (IA-IUAMs), are developed to acquire the bounds of the responses in multidisciplinary system. Two iterative processes, Jacobi iteration and Seidel iteration, are applied in the new methods respectively. The Jacobi iteration based interval uncertainty analysis method (JI-IUAM) utilizes the strategy of concurrent subsystem analysis to improve computational efficiency while the Seidel iteration based interval uncertainty analysis method (SI-IUAM) can accelerate convergence by utilizing the newest information. Both IA-IUAMs are able to evaluate the bounds of responses accurately and quickly. The presented methods are compared with general sensitivity analysis based interval uncertainty analysis method (SIUAM) and conventional Monte Carlo simulation approach (MCS). The validity and efficiency of the new methods are demonstrated by two numerical examples and two engineering examples. Results show that, on the one hand, IA-IUAMs are more efficient than MCS by avoiding hundreds of system analyses, on the other hand, IA-IUAMs are more accurate and have a wider range of application than SIUAM by avoiding linear approximation and global sensitivity calculation.