关键词: 人工神经网络, 多学科优化, 响应面, 协同策略

Abstract: There are two challenges facing the Complex Multidisciplinary Optimization Design, which include information exchange among coupled subsystems and complexity of system analysis. Based on Simultaneous Analysis and Design (SAND) algorithm and Concurrent Subspace Optimization (CSSO) algorithm, the use of Artificial Neural Network (ANN)-based Response Surface (RS) approximations in Multidisciplinary Design Optimization (MDO) was proposed. It was a two-level optimization architecture, that is to say, the discipline level only satisfied the local constraints and the system level offered some collaborative mechanism to guarantee all of the discipline designs agreements on linking/coupled variables so as to facilitate improvements on design. A winged missile system was adopted as an example to verify the effectiveness of this algorithm.

Key words: artificial neural network, multidisciplinary optimization, response surface, collaborative strategy