基于自适应采样和变复杂度近似的多学科稳健性设计优化方法研究

项目名称： 基于自适应采样和变复杂度近似的多学科稳健性设计优化方法研究

项目编号： No.51505163

项目类型： 青年科学基金项目

立项/批准年度： 2016

项目学科： 机械、仪表工业

项目作者： 蒋平

作者单位： 华中科技大学

项目金额： 21万元

中文摘要： 本项目围绕复杂工程系统多学科稳健性设计优化（MRDO）问题，从其计算复杂性、组织复杂性和建模复杂性特点出发，拟重点开展：研究同时考虑输入空间分布信息和输出空间函数特征信息的自适应试验设计方法，提出基于设计空间划分与缩减的自适应采样方法；探明高/低复杂度模型融合机理，研究以高/低复杂度模型的差异信息为导向的变复杂度近似模型构建方法，提出融合定量先验信息与定性先验知识的变复杂度近似模型构建技术和多输入多输出变复杂度建模技术；从“逆向设计思维”角度出发，建立基于性能空间到设计变量不确定性空间逆向映射的多学科多目标目标敏感性与可行性稳健性综合评价体系。在以上理论研究的基础上，以船舶与海洋工程装备为研究对象，开展关键技术应用验证。.本项目将为复杂工程系统多学科稳健性设计优化提供支撑性的新方法和技术，推进多学科设计优化学科的发展，具有重要的理论研究与工程应用价值。

中文关键词： 多学科稳健性设计优化；自适应采样；变复杂度近似建模；不确定性

英文摘要： This project focuses on the Multidisciplinary Robust Design Optimization (MRDO) of complex engineering systems. In order to solve the computing complexity, organizing complexity and modeling complexity owned by the complex engineering systems MRDO problems, the planning work mainly contains: firstly, the hybrid information of space-filling, projective properties in input design space and highly nonlinear, multi-modal characteristics in output response space will be considered in sampling strategies, and then adaptive sampling strategies combining design space partition with design space reduction are presented to make trade-off between global exploration and local exploitation; secondly, the variable-complexity(VC) approximation method is studied to find the mechanism of fusing between high and low complexity models, and active learning VC metamodeling approach that makes full use of the already-collected experimental data for modeling correction will be proposed. Furthermore, to extend the applicability of VC metamodeling approach, the ways of incorporating quantitative and qualitative prior-knowledge into the VC surrogate modeling process and VC metamodeling approach for multiple input and output parameters scenarios will be developed; thirdly, deterministic robust optimization approach considering multi-objective and feasibility robustness will be presented from the “Reverse Design’ view, that executes robust design by calculating the variance of variables for a given acceptable objective variations region and acceptable constraint variations region; At last, the naval architecture and ocean engineering equipment will be taken as the research subject to verify the crucial theories and methods proposed in this project..This project will provide new ways to dealing with MRDO of complex engineering systems, and facilitate the development of MDO. In conclusion, the research work of this project is very important and valuable.

英文关键词： Multidisciplinary robust design optimization;Adaptive sampling;Variable-complexity approximation modeling;Uncertainty