加筋薄壳结构多层级自适应代理优化算法研究  

Research on adaptive surrogate-based optimization method with multi levels for grid-stiffened shell structures

作　　者：邵泽山 关成启 吴东涛 Shao Zeshan;Guan Chengqi;Wu Dongtao(Beijing Aerospace Technology Institute,Beijing 100074,China;National Key Laboratory of Aerospace Flight Technology,Xiongan 071700,China)

机构地区：[1]北京空天技术研究所,北京100074 [2]空天飞行技术全国重点实验室,雄安071700

出　　处：《空天技术》2025年第1期68-81,共14页Aerospace Technology

摘　　要：随着代理模型方法的发展及代理模型技术在优化问题中的应用,代理优化(SurrogateBased Optimization,SBO)算法已取代各类传统优化算法(如遗传算法)成为现今航空航天领域求解加筋薄壳结构优化问题的主流算法。但是随着加筋薄壳结构优化问题中设计变量数目的不断增加,传统的SBO算法面临计算效率降低的风险。针对工程中采用传统SBO算法求解设计变量数目较多的加筋薄壳结构优化问题时所面临的困难(计算效率低),将多层级优化策略与传统SBO算法相结合,且补偿考虑了多层级优化策略中各层级子优化问题间的耦合关系,提出了一种多层级自适应代理优化算法,并构建了所提出优化算法的框架。以航空航天工程中典型的加筋薄壳结构为算例,针对其轻量化设计中的优化问题,采用所提出的多层级自适应代理优化算法对其进行优化,对比分析了其与采用传统SBO算法所得结果。结果表明:在求解工程中典型设计变量数目较多的加筋薄壳结构优化问题时,相较于传统SBO算法,所提出的多层级自适应代理优化算法计算效率更高。With the development of surrogate modeling method and the application of surrogate modeling in optimization problems,surrogate-based optimization(SBO)method has displaced various traditional optimization methods(such as genetic algorithm)and become the mainstream method for solving optimization problems of grid-stiffened shell structures in the aerospace field today.However,as the number of design variables in optimization problems of grid-stiffened shell structures increase,traditional SBO method faces the risk of reduced computational efficiency.In response to the difficulties faced by using traditional SBO method to solve the optimization problems of grid-stiffened shell structures with a large number of design variables in engineering(low computational efficiency),an adaptive surrogate-based optimization method with multi levels is proposed by combing the multi-level optimization strategy with the traditional SBO method and compensating for the coupling relationship between sub optimization problems at different levels in the multilevel optimization strategy,additionally the framework for the optimization method proposed is established.Taking a typical grid-stiffened shell structure in aerospace engineering as the example,the optimization problem in its lightweight problem is addressed by using the multi-level adaptive surrogate-based optimization method.Additionally the optimization results are compared with the results obtained by the traditional SBO method,and the results are discussed and analyzed.The results show that when solving typical optimization problems of grid-stiffened shell structures in engineering with a large number of design variables,the multilevel adaptive surrogate-based optimization method has higher computational efficiency than the traditional SBO method.

关 键 词：加筋薄壳结构 代理模型 优化算法 多层级优化策略 自适应抽样 

分 类 号：V414.4[航空宇航科学与技术—航空宇航推进理论与工程]