计算机集成制造系统 ›› 2018, Vol. 24 ›› Issue (第12): 3107-3119.DOI: 10.13196/j.cims.2018.12.018

• 当期目次 • 上一篇    下一篇

基于继承拉丁超立方采样与局部Kriging近似的可靠性设计优化

李晓科1,2,马军1,2,陈振中3,文笑雨1,2,邱浩波4+   

  1. 1.郑州轻工业学院河南省机械装备智能制造重点实验室
    2.郑州轻工业学院机电工程学院
    3.东华大学机械工程学院
    4.华中科技大学机械科学与工程学院
  • 出版日期:2018-12-31 发布日期:2018-12-31
  • 基金资助:
    国家自然科学基金资助项目(51675198,U1404621,51405302);国家973计划资助项目(2014CB046705);河南省高校科技创新团队资助项目(18IRTSTHN015);河南省高等学校青年骨干教师资助计划资助项目(2015GGJS-183);郑州轻工业学院2016年度博士基金资助项目(2016BSJJ012)。

Reliability-based design optimization using inherited Latin hypercube sampling and local Kriging approximation

  • Online:2018-12-31 Published:2018-12-31
  • Supported by:
    Project supported by the National Natural Science Foundation,China(No.51675198,U1404621,51405302),the National Basic Research Program,China(No.2014CB046705),the S&T Innovation Team Foundation of Universities in Henan Province,China(No.18IRTSTHN015),the Supporting Program for Young Backbone Teachers of Universities in Henan Province,China(No.2015GGJS-183),and the 2016 Doctoral Foundation of Zhengzhou University of Light Industry,China(No.2016BSJJ012).

摘要: 针对基于Kriging近似的可靠性设计优化样本最优配置问题,提出基于继承拉丁超立方采样与局部Kriging近似的可靠性设计优化方法。采用继承拉丁超立方样本构建Kriging近似,并求解最大可能失效点作为局部序列采样中心。针对有效概率约束构建基于Kriging近似误差、功能函数非线性度量及目标可靠度的局部序列采样区域。以重要抽样策略计算失效概率及灵敏度,并采用序列近似规划求解最优设计点。通过3个算例及机床横梁设计优化应用验证了所提方法的有效性。

关键词: 可靠性设计优化, 继承拉丁超立方采样, Kriging近似, 失效概率评估

Abstract: To improve the sample configuration of Kriging-based Reliability-based Design Optimization(RBDO)methods,a novel method using Inherited Latin Hypercube Sampling(ILHS)and local Kriging approximation method was proposed.ILHS was used to ensure the accuracy of Kriging at the initial optimization phase,and the Most Probable Point(MPP)was calculated to be the center of a local sampling region for active probabilistic constraints.The local sampling region was determined according to the Kriging approximation error,performance function nonlinearity and target reliability.The important sampling combined with sequential approximation programming was used to calculate the optimal design.Two numerical examples and the application of box girder were used to demonstrate the effectiveness of proposed method.

Key words: reliability-based design optimization, inherited Latin hypercube sampling, Kriging model, failure probability calculation

中图分类号: