一种基于Blind Kriging模型和差分进化的电磁结构优化算法  

Hybrid algorithm for electromagnetic structure optimization based on Blind Kriging and differential evolution

作　　者：陈晓辉[1] 郭欣欣[1] 裴进明[1] 

机构地区：[1]安徽工程大学电气工程学院,芜湖241000

出　　处：《电子测量与仪器学报》2016年第11期1694-1700,共7页Journal of Electronic Measurement and Instrumentation

基　　金：安徽省高等教育提升计划项目(TSKJ2014B05)资助

摘　　要：各类电磁结构日趋复杂,设计自由度不断提升.传统优化算法需要对大量的参数组合进行全波仿真试探,设计效率普遍较低。针对这一问题,提出BK(blind Kriging)模型和差分进化相结合的电磁结构优化算法.相比普通Kriging模型,BK模型通过贝叶斯参数选择算法将影响性能的主要因子加入回归模型,提高对响应的预测精度;依据BK模型的预测结果从每代差分进化种群中选择最优个体执行电磁仿真。由于优化过程中大量的电磁计算转移到快速的BK模型,优化效率得到显著提升。通过一个圆波导多螺钉极化转换器的优化设计,表明该方法的求解质量和收敛速度优于现有算法。Various electromagnetic（ EM） structures become more complex and often have increasing degrees of design freedom. Classical optimization methods require numerous simulation trials of different parameter combinations,which lead to a low design efficiency. To address this problem,a hybrid EM structure optimization algorithm combined blind Kriging（ BK） model with differential evolution（ DE） is proposed in this paper.Comparing with ordinary Kriging model,the prediction accuracy can be improved by adding main factors that are identified by Bayesian variable selection technique to the regression model. According to the predicted responses by BK models,the optimal individual is selected from every DE generation and reevaluated by EM simulation. Since most EM computation burden is shifted to efficient BK models,the design efficiency can be improved significantly.The proposed algorithm is validated by the optimization of a multi screw polarization converter,and it outperforms other existing algorithms in the quality of the solution and the convergence rate.

关 键 词：电磁结构 优化算法 BLIND KRIGING 差分进化 

分 类 号：TN80[电子电信—信息与通信工程]