基于Kriging模型和条件风险值的多响应优化设计  

Multiple responses optimization design based on Kriging and conditional value at risk

作　　者：朱连燕[1,2] 马义中[1] 吴锋[1] 张建侠[1] 欧阳林寒[1] 

机构地区：[1]南京理工大学经济管理学院,江苏南京210094 [2]南京科技职业学院基础科学部,江苏南京210048

出　　处：《计算机集成制造系统》2016年第6期1581-1589,共9页Computer Integrated Manufacturing Systems

基　　金：国家自然科学基金资助项目(71471088,71371099,71401080);教育部高等学校博士学科点专项科研基金资助项目(20123219120032);江苏省研究生科研创新计划资助项目(KYZZ15_0126)~~

摘　　要：针对具有风险规避特性的多响应随机仿真优化问题,结合稳健参数设计思想和条件风险值准则,提出基于Kriging模型的均值—条件风险值优化策略。利用元建模技术,分别建立了均值响应和条件风险值响应的Kriging模型,在此基础上构建了具有风险参数描述的均值—条件风险值决策模型;采用bootstrap方法度量环境变量的不确定性对多响应优化Pareto前沿的影响,同时给出不同风险参数对Pareto前沿的影响。仿真实验结果表明所提方法能够有效处理具有风险规避特性的多响应随机仿真优化问题,验证了所提方法的有效性和合理性。Aiming at the risk-averse stochastic simulation optimization problems with multiple responses,by combining robust design approach with conditional value at risk criterion,an optimal strategy of mean-conditional value at risk was proposed based on Kriging model was proposed.Kriging models for mean response and conditional value at risk response were constructed respectively with meta modeling technology,and the decision model of mean-conditional value at risk with risk parameter description was built on this basis.The influence of environment variable's uncertainty on Pareto frontier of multiple responses was measured by bootstrap method,and the influence of different risk parameters on Pareto frontier was also given.The simulation experiment results showed that the proposed approach could dispose the risk-averse stochastic simulation optimization problems with multiple responses effectively.

关 键 词：条件风险值 稳健参数设计 风险规避Kriging模型 

分 类 号：F270.5[经济管理—企业管理] C934[经济管理—国民经济]