摘要
电介质复合材料是重要的储能材料,其电容是电容器关键参数之一。传统方法中,实际的电容值是由仪器直接测试得到,给定微观结构的复合材料其电容值可以由泊松方程计算得到。在计算机模拟生成的数据集基础上,采用CNN-SVR模型进行了预测研究,研究结果表明,机器学习方法能够快速有效地预测给定复合材料微观结构的电容值,不仅克服了静态物理特征表达能力的不足,而且摆脱了繁琐的特征设计过程,也避免了复杂的有限元仿真运算。该方法还具有推广和应用于复合材料其他物理特性研究的潜力。
Dielectric composite is an important energy storage material,and its capacitance is one of the key parameters of capacitor.In the traditional method,the actual capacitance value is directly measured by the instrument,and the capacitance value of composite material with given microstructure can be calculated by Poisson equation.Based on the data set generated by computer simulation,the cnn-svr model is used to predict the capacitance value of a given composite microstructure.The results show that the machine learning method can quickly and effectively predict the capacitance value of a given composite microstructure,which not only overcomes the lack of static physical feature expression ability,but also gets rid of the cumbersome feature design process and avoids the complicated finite element simulation.This method also has the potential to be applied to other physical properties of composites.
作者
朱珍
卢天奇
朱文博
陈建文
王金海
黄穗龙
许仁俊
ZHU Zhen;LU Tian-qi;ZHU Wen-bo;CHEN Jian-wen;WANG Jin-hai;HUANG Sui-long;XU Ren-jun(School of Mechatronics Engineering and Automation,Foshan University,Foshan 528225,China)
出处
《佛山科学技术学院学报（自然科学版）》
CAS
2022年第2期36-43,共8页
Journal of Foshan University(Natural Science Edition)
基金
广东省自然科学粤佛联合基金青年项目(2020A1515110601)。
关键词
复合材料
电容
计算机模拟
机器学习
composite materials
capacitance
computer simulation
machine learning
