文章信息/Info

Title:

Research on Optimization and Prediction Mechanism of Material Properties Based on Gradient and Feature Analysis in Convolution Neural Network

文章編號(hào):

1674-3962(2020)05-0385-06

作者:

曹卓1; 但雅波1; 李想1; 牛程程2; 董容智1; 錢(qián)松榮1; 胡建軍1; 3

（1. 貴州大學(xué)機(jī)械工程學(xué)院，貴州 貴陽(yáng) 550025） (2. 貴州大學(xué) 現(xiàn)代制造技術(shù)教育部重點(diǎn)實(shí)驗(yàn)室，貴州 貴陽(yáng) 550025) (3. 美國(guó)南卡羅來(lái)納大學(xué)計(jì)算機(jī)科學(xué)與工程系，南卡羅萊納 哥倫比亞 29208）

Author(s):

CAO Zhuo1;  DAN Yabo1;  LI Xiang1;  NIU Chengcheng2;  DONG Rongzhi1;  QIAN Songrong1;  HU Jianjun1; 3

(1. School of Mechanical Engineering, Guizhou University, Guiyang 550025, China) (2. Key Laboratory of Advanced Manufacturing Technology, Ministry of Education, Guizhou University, Guiyang 550025, China) (3. Department of Computer Science and Engineering, University of South Carolina, Columbia SC 29208，USA)

關(guān)鍵詞:

材料信息學(xué); 卷積神經(jīng)網(wǎng)絡(luò); 形成能; 梯度分析; 特征抽取

Keywords:

materials informatics; convolutional neural network; formation energy; gradient analysis; feature extraction

分類(lèi)號(hào):

TB3

DOI:

10.7502/j.issn.1674-3962.201905008

文獻(xiàn)標(biāo)志碼:

A

摘要:

材料信息學(xué)作為材料領(lǐng)域一種新的研究方法，引起了國(guó)內(nèi)外廣泛的關(guān)注。隨著材料數(shù)據(jù)的快速增加，機(jī)器學(xué)習(xí)方法也越來(lái)越多地被應(yīng)用在材料數(shù)據(jù)的分析中，并有望從大量的材料數(shù)據(jù)中獲取具有指導(dǎo)性的材料學(xué)規(guī)律。采用卷積神經(jīng)網(wǎng)絡(luò)模型，使用從材料數(shù)據(jù)庫(kù)中收集得到的4000多種材料的數(shù)據(jù)，對(duì)材料的形成能進(jìn)行預(yù)測(cè)并得到了較為準(zhǔn)確的預(yù)測(cè)結(jié)果。隨后對(duì)材料特征矩陣的梯度進(jìn)行分析，發(fā)現(xiàn)了梯度與材料性能間有一定的相關(guān)性，并可在梯度矩陣的指導(dǎo)下找到具有目標(biāo)性能的材料特征矩陣分布。最后對(duì)卷積神經(jīng)網(wǎng)絡(luò)中識(shí)別出的特征模式進(jìn)行了分析，進(jìn)一步驗(yàn)證了卷積神經(jīng)網(wǎng)絡(luò)具有較好的材料性能預(yù)測(cè)能力。

Abstract:

As a new research mode in material science, material informatics has attracted wide attention. With the rapid increase of material data, machine learning methods are more and more used in the analysis of material data to obtain instructive physical and chemical laws from a large number of material data. This paper focuses on the convolutional neural network, using data from more than 4000 materials collected from the Material Project database to predict formation energy of materials, and the prediction results are accurate. Then, the gradient of feature map is analyzed, we observe that there are some certain correlations between gradient and material properties, and under the guidance of gradient matrix, the possible distribution of feature map with target properties can be found. Finally, the patterns recognized by the convolutional neural network are analyzed, which further verifies that the convolutional neural network can achieve excellent prediction results of material property.