基于Kriging模型和两目标约束应对策略的代理优化算法  

Surrogate-based optimization algorithm based on Kriging models and bi-objective constraint-handling strategy

作　　者：张建侠[1] 马义中[1] 张延静[1] 欧阳林寒 ZHANG Jianxia;MA Yizhong;ZHANG Yanjing;OUYANG Linhan(School of Economics and Management, Nanjing University of Science and Technology, Nanjing 210094, China;School of Economics and Managements Nanjing University of Aeronautics and Astronautics, Nanjing 210016. China)

机构地区：[1]南京理工大学经济管理学院,江苏南京210094 [2]南京航空航天大学经济与管理学院,江苏南京210016

出　　处：《计算机集成制造系统》2018年第12期3001-3007,共7页Computer Integrated Manufacturing Systems

基　　金：国家自然科学基金资助项目(71471088;71371099;71702072);中央高校基本科研业务专项资金资助项目(3091511102);江苏省自然科学基金资助项目(BK20170810);江苏省研究生科研创新计划资助项目(KYCX17_0405)~~

摘　　要：为了解决包含黑箱约束的复杂工程系统的优化设计问题,提出一种基于Kriging模型和两目标约束应对策略的代理优化算法。该算法将改进目标函数的期望改进准则和刻画可行域边界的可行性概率准则同时作为优化目标,再从得到的Pareto集中选取新试验点,不仅提高了新试验点选取的目的性也使新试验点兼具探索最优解和开发可行域边界的能力。最后,通过两个数学算例和一个工程算例将所提算法与已有算法进行比较,计算结果表明基于两目标约束应对策略的代理优化算法具有更高的优化精度、效率和稳健性。To solve the design optimization problems of complex engineering systems with black-box constraints,a surrogate-based optimization algorithm was proposed based on Kriging models and bi-objective constraint-handling strategy.By taking the expected improvement criterion of improving the objective function and the feasibility probability criterion of characterizing the feasible region boundary as objectives simultaneously,the new trials from Pareto sets purposefully was selected so as to make the new added trials have the ability to balance optimal solution exploration and feasible region boundary exploitation.The proposed algorithm was tested on two numerical and one engineering benchmarks and was compared with the existed algorithms.The numerical results indicated that the proposed algorithm based on the bi-objective constraint-handling strategy was more accurate,efficient and robust.

关 键 词：黑箱约束 KRIGING模型 代理优化算法 两目标约束应对策略 期望改进 可行性概率 

分 类 号：N945.15[自然科学总论—系统科学] 0212.6