文章信息/Info

Title:

Optimal Design of Microwave Absorbing Material of Carbonyl Iron/Ferroferric Oxide Composite via Machine Learning

文章編號:

1674-3962(2024)07-0652-06

作者:

仲陸祎; 權斌; 車仁超; 陸文聰

1. 上海大學理學院化學系，上海 200444 2. 復旦大學 先進材料實驗室，上海 200433

Author(s):

ZHONG Luyi;  QUAN Bin;  CHE Renchao;  LU Wencong

1. Department of Chemistry, College of Sciences, Shanghai University, Shanghai 200444, China 2. Advanced Materials Laboratory, Fudan University, Shanghai 200433, China

關鍵詞:

吸波材料; 磁導率實部積分值; 磁導率虛部積分值; 機器學習; 高通量篩選

Keywords:

microwave absorbing material;  the integral value of the real part of permeability;  the integral value of the imaginary part of permeability;  machine learning;  high-throughput screening

分類號:

TP181;TB34

DOI:

10.7502/j.issn.1674-3962.202209040

文獻標志碼:

A

摘要:

吸波材料多為復合材料，在抵御電磁干擾和電磁輻射等方面發揮著重要作用，其磁導率與材料對磁能的儲存和消耗以及材料的阻抗匹配有關。以羰基鐵/四氧化三鐵復合吸波材料為研究對象、6個球磨工藝參數為特征變量，分別運用隨機森林回歸（random forest regression，RFR）算法和支持向量回歸（support vector regression，SVR）算法，構建了磁導率實部積分值和虛部積分值的機器學習模型。通過兩步高通量篩選，設計了3個兼顧磁導率實部積分值和虛部積分值的虛擬樣本，并對其中一個樣本進行了實驗驗證。結果表明，磁導率實部積分值和虛部積分值的相對預測誤差分別為3.14%和-6.56%。該研究方法能夠挖掘工藝參數和材料吸波性能之間的關系，加快新材料的研發，為運用機器學習優化設計吸波材料提供了思路。

Abstract:

Microwave absorbing materials are mostly composite materials, which play an important role in resisting electromagnetic interference and electromagnetic radiation. Their permeability is related to the storage and consumption of magnetic energy, as well as the impedance matching of the materials. In this work, microwave absorbing material of carbonyl iron/ferroferric oxide composite is taken as the research object, the machine learning models of the integral value of the real part of permeability (∫μ′) and the integral value of the imaginary part of permeability (∫μ″) are constructed by using random forest regression (RFR) algorithm and support vector regression (SVR) algorithm respectively, with six ball milling process parameters as feature variables. Three virtual samples are designed by two-step high-throughput screening, considering both ∫μ′ and ∫μ″, and the experimental verification is carried out. The results show that the relative prediction error of ∫μ′ is 3.14%, and the relative prediction error of ∫μ″ is -6.56%. Therefore, this research method can be used to explore the relationship between process parameters and absorbing properties of microwave absorbing materials, providing new ideas for the optimal design of microwave absorbing materials by using machine learning.