Abstract: To make full use of computing resources and reduce the number of iterations,a surrogate-based optimization algorithm which could add batch points was proposed.To explore the optimum solution and to exploit its area,the expected improvement and the WB2 criterion were used correspondingly.The constraint boundary was characterized by using the probability of feasibility and the multi-objective optimization framework.Two corresponding multi-points infilling algorithms were designed in the exploration and exploitation phases and an adaptive switching strategy for this two phases was designed according to the distance between new sample points and known sample points.The performance of the algorithm was verified by three different types of numerical and one engineering benchmarks.The results showed that the proposed algorithm was more efficient in convergence and the solution was more precise and robust.

Key words: Kriging model, surrogate-based optimization, infill sampling criteria, probability of feasibility, multi-points infill

摘要： 为了充分利用计算资源,减少迭代次数,提出一种可以批量加点的代理优化算法。该算法分别采用期望改进准则和WB2(Watson and Barnes)准则探索存在的最优解并开发已存在最优解的区域,利用可行性概率和多目标优化框架刻画约束边界。在探索和开发阶段,设计了两种对应的多点填充算法,并根据新样本点和已知样本点的距离关系,设计了两个阶段的自适应切换策略。通过3个不同类型算例和一个工程实例验证算法性能,结果表明,该算法收敛更快,其结果具有较好的精确性和稳健性。

关键词: Kriging模型, 代理优化, 加点准则, 可行性概率, 多点填充