基于Kriging代理模型的多点加点序列优化方法.pdfVIP

29 卷 4 期 Vol.29 No.4 工 程 力 学 2012 年 4 月 Apr. 20 12 ENGINEERING MECHANICS 90 文章编号：1000-4750(2012)04-0090-06 基于 代理模型的多点加点序列优化方法 1 2 高月华 ，王希诚 (1. 大连交通大学交通运输工程学院，辽宁，大连 116028 ；2. 大连理工大学工业装备结构分析国家重点实验室，辽宁，大连 116024) 摘 要：基于 Kriging 代理模型提出了一种同时考虑预测响应值及其不确定性的多点加点准则，并基于该准则发 展了一套序列近似优化方法。多点加点准则基于初始样本信息和所预测的对象函数特征增加新样本集，以在寻优 迭代过程中自适应地提高代理模 的精度。该文方法依据多点加点准则在一次迭代中增加多个空间无关的新样本 点，适用于多机同时计算或并行计算，从而提高计算效率。以两个经典的数学函数为例，将该优化方法与期望提 高准则方法进行了比较，结果表明该文提出的优化方法能够有效地提高最优解的全局性。将方法用于一盒式注塑 件的成 工艺优化设计，优化结果也表明了该方法的有效性。 关键词：Kriging ；代理模 ；序列优化；加点准则；注塑成 中图分类号：TP18 文献标志码：A A SEQUENTIAL OPTIMIZATION METHOD WITH MULTI-POINT SAMPLING CRITERION BASED ON KRIGING SURROGATE MODEL 1 2 GAO Yue-hua , WANG Xi-cheng (1. Traffic Transportation School, Dalian Jiaotong University, Dalian, Liaoning 116028, China; 2. State Key Laboratory of Structural Analysis for Industrial Equipment, Dalian University of Technology, Dalian, Liaoning 116024, China) Abstract: A multi-point sampling criterion considering the predictor and its uncertainty simultaneously is proposed based on Kriging surrogate model, and a kind of sequential approximation optimization method is developed. Multi-point sampling criterion is used to add the new samples by considering the distribution of the initial samples and the characteristics of the predicted objective function. The proposed method selects more than one new sample point for each optimization iteration, thus it can be performed by parallel computation or multi-computer runs which improves the computational efficienc