摘要
对于大系统的优化,子系统由于分析模型计算复杂而大量采用近似模型代替与优化。常用的近似建模技术有响应面模型(RS)、计算机试验设计与分析模型(DACE)和神经网络中的误差后向传播模型(BP)。拟对以上三种近似建模技术进行了比较研究,简要阐述了三种近似建模技术的理论基础,并应用实例分别进行了考核。在各近似模型的近似精度和计算费用等方面进行了初步探讨与研究。
Approximation is often utilized in complex system optimization because of the expensive computation of an analytical model. Response surface (RS), design and analysis of computer experiments (DACE) and BP neural network are three common approximations usually used to surrogate the original simulation model in MDO. The aforementioned methods are mathematically reviewed and three problems are tested. Three models are compared in terms of accuracy and computational cost.
出处
《系统工程与电子技术》
EI
CSCD
北大核心
2005年第5期834-836,917,共4页
Systems Engineering and Electronics
基金
高等学校博士学科专项科研基金资助课题
关键词
近似建模
响应面
BP神经网络
approximate modeling
response surface
BP neural network
