基于免疫遗传算法的BP神经网络在纱线条干预测上的应用  

Application of BP neural network in yarn evenness prediction based on immune Genetic Algorithm

作　　者：查刘根 谢春萍[1] ZHA Liugen;XIE Chunping(Key Laboratory of Eco-Textiles,Ministry of Education,Jiangnan University,Wuxi 214122,China)

机构地区：[1]江南大学生态纺织教育部重点实验室,江苏无锡214122

出　　处：《丝绸》2019年第2期19-26,共8页Journal of Silk

基　　金：国家重点研发计划项目(2017YFB0309200);江苏省产学研项目(BY2016022-16);江苏省自然科学基金项目(BK20170169);纺织服装产业河南省协同创新项目(hnfx14002);中央高校基本科研业务费专项资金资助项目(JUSRP51731B)

摘　　要：为了避免因随机生成BP神经网络初始权值和阈值而带来的不确定性,以及得到更好的预测纱线条干CV值的精度和速度,借助免疫遗传算法对传统单一的BP神经网络进行权值和阈值的优化。免疫算法中特有的浓度调节机制有效地解决了遗传算法后期未成熟收敛的问题。利用Matlab构建单一的BP神经网络模型、遗传BP神经网络模型和免疫遗传BP神经网络模型进行纱线条干CV值的预测实验,通过仿真训练结果的对比分析可得出,免疫遗传算法优化的BP神经网络能够更准确、更快速、更稳定地完成纱线条干CV值的预测。In order to avoid uncertainty caused by randomly generating initial weights and threshold values of BP neural network,and obtain higher prediction accuracy and speed of yarn evenness CV value,weights and threshold values of traditional single BP neural network were optimized with the aid of immune genetic algorithm.The unique concentration adjustment mechanism of immune algorithm effectively works in solving the problem of premature convergence in later period of genetic algorithm.Matlab was applied to construct a single BP neural network model,genetic BP neural network model and immune genetic BP neural network model to predict yarn evenness CV value.Comparative analysis of simulation training results indicate that the BP neural network developed via optimization with immune genetic algorithm performs more accurately,quickly and stably in predicting yarn evenness CV value.

关 键 词：纱线条干 纱线质量预测 BP神经网络 免疫算法 遗传算法 

分 类 号：TS111.914[轻工技术与工程—纺织材料与纺织品设计]