摘要
为了优化岸桥轨道夹结构,针对复杂黑箱约束问题,提出一种自适应约束代理优化算法。先使用改善的约束并行期望改进准则(Constrained Parallel Expected Improvement,CPEI)和重要边界采样准则(Importance Boundary Sampling,IBS)探索全局最优解区域,再使用信任域法进行局部开发,并引入客观的约束函数Kriging模型来更新策略,减少不必要的仿真分析,节约了计算成本。通过一个数学算例将所提算法与约束期望改进(Constrained Expected Improvement,CEI)和两目标加点优化算法(Expected Improvement-Probability of Feasibility,EI-PoF)进行比较,结果表明:所提算法收敛更快,精确度更高。
In order to improve the rail clamp structure of container crane and solve the optimization problem of black box complex engineering system under constraint conditions,an adaptive constraint surrogate-based optimization algorithm is proposed.The improved constrained parallel expectation improvement(CPEI)criterion and important boundary sampling(IBS)criterion are used to explore the global optimal solution region,and then the trust region method is used for local exploration.For saving the calculation cost,the objective constraint function Kriging model updating strategy is introduced to reduce unnecessary computer simulation analysis.The proposed algorithm is tested on one numerical benchmark and is compared with CEI and EI-POF.The results show that the proposed algorithm has faster convergence and higher accuracy.
作者
余镇
樊志华
李志华
YU Zhen;FAN Zhihua;LI Zhihua(School of Mechanical Engineering,Hangzhou Dianzi University,Hangzhou Zhejiang 310018,China)
出处
《杭州电子科技大学学报（自然科学版）》
2022年第2期71-77,共7页
Journal of Hangzhou Dianzi University：Natural Sciences
基金
浙江省自然科学基金资助项目(LY19E050013,LY18E050008)。
关键词
黑箱约束
KRIGING模型
代理优化算法
并行加点
black box constraint
Kriging model
surrogate-based optimization algorithm
parallel filling
