苑昭阔,吴俐俊,王骏,张萍,韦增志.基于神经网络遗传算法的超疏水涂层优化[J].表面技术,2022,51(1):240-246, 271.
YUAN Zhao-kuo,WU Li-jun,WANG Jun,ZHANG Ping,WEI Zeng-zhi.Optimization of Superhydrophobic Coatings Based on Neural Network and Genetic Algorithm[J].Surface Technology,2022,51(1):240-246, 271
基于神经网络遗传算法的超疏水涂层优化
Optimization of Superhydrophobic Coatings Based on Neural Network and Genetic Algorithm
投稿时间:2021-07-29  修订日期:2021-11-21
DOI:10.16490/j.cnki.issn.1001-3660.2022.01.025
中文关键词:  超疏水涂层  BP神经网络  遗传算法  水接触角  导热系数
英文关键词:superhydrophobic coatings  BP neural network  genetic algorithm  water contact angle  thermal conductivity
基金项目:国家重点研发计划(2020YFC1910100)
作者单位
苑昭阔 同济大学 机械与能源工程学院,上海 201804 
吴俐俊 同济大学 机械与能源工程学院,上海 201804 
王骏 南京同诚节能环保装备研究院,南京 211100 
张萍 同济大学 机械与能源工程学院,上海 201804 
韦增志 同济大学 机械与能源工程学院,上海 201804 
AuthorInstitution
YUAN Zhao-kuo School of Mechanical Engineering, Tongji University, Shanghai 201804, China 
WU Li-jun School of Mechanical Engineering, Tongji University, Shanghai 201804, China 
WANG Jun Nanjing Tongcheng Energy Saving and Environmental Protection Equipment Research Institute, Nanjing 211100, China 
ZHANG Ping School of Mechanical Engineering, Tongji University, Shanghai 201804, China 
WEI Zeng-zhi School of Mechanical Engineering, Tongji University, Shanghai 201804, China 
摘要点击次数:
全文下载次数:
中文摘要:
      目的 探究超疏水涂层各成分的含量对涂层水接触角和导热系数的影响,找到最优成分组合,使涂层水接触角和导热系数同时获得最大值。方法 根据设计的L25(55)正交试验,制作和测试涂层试样,借助Matlab软件建立结构为5-8-2的BP神经网络,通过正交试验结果训练和测试神经网络,得到涂层水接触角和导热系数的预测模型。调用训练好的预测模型,采用遗传算法对涂层各成分含量进行全局寻优。使用寻优得到的参数和调整后的参数进行试验,检验寻优计算结果。结果 BP神经网络预测模型水接触角的最大误差为0.061 98,导热系数的最大误差为0.065 77。基于遗传算法的优化结果,涂层成分(质量分数)为纳米SiO2 10.1%+TiO2 6.4%+碳粉5%+纳米石墨烯0.6%+MTES 1.8%时,涂层的水接触角达到164.24°,导热系数达到14.19 W/(m.K),其误差分别为3.80%和2.31%。采用调整后的参数进行试验,测试得到涂层的水接触角为155.02°,导热系数为13.25 W/(m.K),其误差分别5.64%和5.58%。结论 通过BP神经网络预测模型和遗传算法寻优,可以使涂层的水接触角和导热系数都获得较大的提高。
英文摘要:
      The work aims to explore the influence of the content of each coatings component on the coatings water contact angle (WCA) and thermal conductivity, and find the optimal composition so as to maximize the WCA thermal conductivity simultaneously. The coatings samples were made and tested according to the L25(55) orthogonal experimental design. The BP neural network with the structure of 5-8-2 was established by Matlab software. The prediction model of WCA and thermal conductivity of coatings was obtained by training and testing the neural network with the results of orthogonal test. The genetic algorithm was used to optimize the content of each component by calling the trained prediction model. The optimized and adjusted parameters were used to test and verify the optimization results. After the BP neural network model was trained, the prediction results showed that the maximum error was 0.061 98 and WCA 0.065 77 for thermal conductivity. Based on the optimization results of genetic algorithm, the coatings would have 164.24° for WCA and 14.19 W/(m.K) for thermal conductivity, with 10.1wt% Nano-SiO2, 6.4wt% TiO2, 5wt% carbon powder, 0.6wt% nano graphene and 1.8wt% MTES. In the meanwhile, the error of the WCA and thermal conductivity was 3.80% and 2.31%, respectively. The coatings made with adjusted parameters had 155.02° for WCA and 13.25 W/(m.K) for thermal conductivity, with errors of 5.64% and 5.58%, respectively. Through BP neural network prediction model and genetic algorithm optimization, the water contact angle and thermal conductivity of coatings both got greatly improved.
查看全文  查看/发表评论  下载PDF阅读器
关闭

关于我们 | 联系我们 | 投诉建议 | 隐私保护 | 用户协议

您是第19498161位访问者    渝ICP备15012534号-3

版权所有:《表面技术》编辑部 2014 surface-techj.com, All Rights Reserved

邮编:400039 电话:023-68792193传真:023-68792396 Email: bmjs@surface-techj.com

渝公网安备 50010702501715号