摘要
军事和民用领域对吸波材料的需求与日俱增,利用机器学习(machine learning,ML)有助于加快设计具有多目标特性的吸波材料。开发了吸波材料设计方法(absorbing material design method,AMDM)。研究对象为羰基铁/铁粉吸波材料,利用5个球磨工艺参数作为特征变量,采用灰狼(grey wolf optimizer,GWO)算法和支持向量回归(support vector regression,SVR)算法,建立了磁导率实部和虚部积分的预测模型。经过模型预测,得到4个性能优于原始数据库的虚拟样本,并制备了4个样本进行实验验证。结果显示,磁导率实部积分预测误差绝对值最小为3.735%,最大为8.328%,磁导率虚部积分预测误差绝对值最小为1.124%,最大为4.423%。该研究为高效开发和设计复合吸波材料提供了有效途径。
The demand for microwave-absorbing materials is rapidly increasing in both military and civilian applications.Machine learning(ML)offers an effective approach for accelerating the design of absorbing materials with multi-objective performance requirements.In this study,an absorbing material design method(AMDM)is developed and applied to carbonyl iron/iron-powder composite absorbers.Five ball-milling process parameters are used as feature variables,and predictive models for the real and imaginary parts of the magnetic permeability integrals are constructed using the grey wolf optimizer(GWO)and support vector regression(SVR)algorithms.Model predictions yielded four virtual samples with performance superior to those in the original database.These samples were then fabricated and experimentally validated.The results show that the absolute prediction error of the real-part magnetic-permeability integral ranges from 3.735%to 8.328%,while that of the imaginary-part integral ranges from 1.124%to 4.423%.This study provides an effective pathway for the efficient development and design of composite microwave-absorbing materials.
作者
寇龚权治
陈将伟
KOUGONG Quanzhi;CHEN Jiangwei(College of Electronic and Optical Engineering,College of Flexible Electronics(Future Technology),Nanjing University of Posts&Telecommunications,Nanjing 210023,P.R.China)
出处
《重庆邮电大学学报(自然科学版)》
北大核心
2025年第6期911-920,共10页
Journal of Chongqing University of Posts and Telecommunications(Natural Science Edition)
关键词
各向异性
粒径
非线性回归
羰基铁
反射损耗
anisotropy
particle size
nonlinear regression
carbonyl iron
reflection loss
