基于GRA与正则化RBF的稳健参数优化  

Robust Parameter Optimization Based on GRA and Regularized RBF

作　　者：黄鸿琦 HUANG Hongqi(School of Management,Henan Institute of Technology,Xinxiang 453003,China)

机构地区：[1]河南工学院管理学院,河南新乡453003

出　　处：《新乡学院学报》2020年第9期33-38,共6页Journal of Xinxiang University

基　　金：河南省教育厅人文社会科学研究一般项目(2019-ZDJH-086);新乡市社科重点调研课题(2020-139)。

摘　　要：为优化金属聚丙烯薄膜电容器热聚合工艺中多响应参数的设计,利用灰色关联度分析方法对实验参数进行综合模糊评价,依据预期改善效果,利用层次分析法给出各响应指标满足一致性检验的优先优化权重,使综合评价指标优先考虑优先优化权重大的响应指标。利用正则化径向基神经网络建立参数与综合评价指标的网络模型,并通过真值检验判断模型的拟真效果,在全局范围内找出最优参数组合。结果表明:在优先考虑响应稳健性及损耗角正切的基础上,正则化径向基神经网络能够很好地反映真实映射关系,可使热聚合工艺中多响应参数达到整体最优的效果。In order to optimize the design of multiple response parameters in the thermal polymerization process of metal polypropylene film capacitor,a comprehensive fuzzy evaluation of the experimental parameters was carried out using the grey relational analysis method.According to the expected improvement effect,the priority optimization weight of each response index meeting the consistency test was given by using the analytic hierarchy process,so that the comprehensive evaluation index gave priority to the response index with high priority optimization weight value.The regularized RBF neural network was used to establish the network model of parameters and comprehensive evaluation indexes,and the simulation effect of the model was judged by truth value inspection,and the optimal parameter combination was found in the global scope.The results show that the regularized RBF neural network can well reflect the real mapping relationship on the basis of giving priority to response robustness and loss angle tangent.The multiple response parameters in the thermal polymerization process can achieve the overall optimal effect.

关 键 词：多响应参数稳健优化 金属化聚丙烯薄膜电容器 正则化径向基神经网络 灰色关联度分析 层次分析法 

分 类 号：TP114.2[自动化与计算机技术—控制理论与控制工程] TP391.9[自动化与计算机技术—控制科学与工程]